journal 03

/20

10/#

28

the journal of financial transformation

Recipient of the APEX Awards for Publication Excellence 2002-2009

Cass-Capco Institute Paper Series on Risk

MSc in Insurance and Risk Management

Cass is one of the world’s leading academic centresin the insurance field. What's more, graduatesfrom the MSc in Insurance and Risk Managementgain exemption from approximately 70% of theexaminations required to achieve the AdvancedDiploma of the Chartered Insurance Institute (ACII).

For applicants to the Insurance and RiskManagement MSc who already hold a CII Advanced Diploma, there is a fast-track January start, giving exemption from the first term of the degree.

To find out more about our regular informationsessions, the next is 10 April 2008, visitwww.cass.city.ac.uk/masters and click on'sessions at Cass' or 'International & UK'.

Alternatively call admissions on:

+44 (0)20 7040 8611

With a Masters degree from Cass Business School,you will gain the knowledge and skills to stand out in the real world.

Minimise risk,optimise success

Editor

Shahin Shojai, Global Head of Strategic Research, Capco

Advisory Editors

Cornel Bender, Partner, Capco

Christopher Hamilton, Partner, Capco

Nick Jackson, Partner, Capco

Editorial Board

Franklin Allen, Nippon Life Professor of Finance, The Wharton School,

University of PennsylvaniaJoe Anastasio, Partner, Capco

Philippe d’Arvisenet, Group Chief Economist, BNP Paribas

Rudi Bogni, former Chief Executive Officer, UBS Private Banking

Bruno Bonati, Strategic Consultant, Bruno Bonati Consulting

David Clark, NED on the board of financial institutions and a former senior

advisor to the FSA

Géry Daeninck, former CEO, Robeco

Stephen C. Daffron, Global Head, Operations, Institutional Trading & Investment

Banking, Morgan Stanley

Douglas W. Diamond, Merton H. Miller Distinguished Service Professor of Finance,

Graduate School of Business, University of Chicago

Elroy Dimson, BGI Professor of Investment Management, London Business School

Nicholas Economides, Professor of Economics, Leonard N. Stern School of

Business, New York University

José Luis Escrivá, Group Chief Economist, Grupo BBVA

George Feiger, Executive Vice President and Head of Wealth Management,

Zions Bancorporation

Gregorio de Felice, Group Chief Economist, Banca IntesaHans Geiger, Professor of Banking, Swiss Banking Institute, University of Zurich

Wilfried Hauck, Chief Executive Officer, Allianz Dresdner Asset Management

International GmbH

Pierre Hillion, de Picciotto Chaired Professor of Alternative Investments and

Shell Professor of Finance, INSEAD

Thomas Kloet, Senior Executive Vice-President & Chief Operating Officer,

Fimat USA, Inc.

Mitchel Lenson, former Group Head of IT and Operations, Deutsche Bank Group

David Lester, Chief Information Officer, The London Stock Exchange

Donald A. Marchand, Professor of Strategy and Information Management,

IMD and Chairman and President of enterpriseIQ®

Colin Mayer, Peter Moores Dean, Saïd Business School, Oxford University

Robert J. McGrail, Executive Managing Director, Domestic and International Core

Services, and CEO & President, Fixed Income Clearing Corporation

John Owen, CEO, Library House

Steve Perry, Executive Vice President, Visa Europe

Derek Sach, Managing Director, Specialized Lending Services, The Royal Bank

of Scotland

John Taysom, Founder & Joint CEO, The Reuters Greenhouse Fund

Graham Vickery, Head of Information Economy Unit, OECD

Norbert Walter, Group Chief Economist, Deutsche Bank Group

TABlE of conTEnTs

opinion

8 A partial defense of the giant squidSanjiv Jaggia, Satish Thosar

12 Thou shalt buy ‘simple’ structured products onlySteven Vanduffel

14 Enterprise friction — the mandate for risk managementSandeep Vishnu

19 Enterprise Risk Management — a clarification of some common concernsMadhusudan Acharyya

22 capital at risk — a more consistent and intuitive measure of riskDavid J. Cowen, David Abuaf

ARTiclEs

27 Economists’ hubris — the case of risk managementShahin Shojai, George Feiger

37 Best practices for investment risk managementJennifer Bender, Frank Nielsen

45 lessons from the global financial meltdown of 2008Hershey H. Friedman, Linda W. Friedman

55 What can leaders learn from cybernetics? Toward a strategic framework for managing complexity and risk in socio-technical systemsAlberto Gandolfi

61 financial stability, fair value accounting, and procyclicalityAlicia Novoa, Jodi Scarlata, Juan Solé

77 can ARMs’ mortgage servicing portfolios be delta-hedged under gamma constraints?Carlos E. Ortiz, Charles A. Stone, Anne Zissu

87 The time-varying risk of listed private equityChristoph Kaserer, Henry Lahr, Valentin Liebhart, Alfred Mettler

95 The developing legal risk management environmentMarijn M.A. van Daelen

103 interest rate risk hedging demand under a Gaussian frameworkSami Attaoui, Pierre Six

109 non-parametric liquidity-adjusted VaR model: a stochastic programming approachEmmanuel Fragnière, Jacek Gondzio, Nils S. Tuchschmid, Qun Zhang

117 optimization in financial engineering — an essay on ‘good’ solutions and misplaced exactitudeManfred Gilli, Enrico Schumann

123 A VaR too far? The pricing of operational risk Rodney Coleman

131 Risk management after the Great crashHans J. Blommestein

As the variety and volume of shocks facing the financial system increase the demand for those who have a

thorough understanding of risk, its measurement and management will continue to rise. And this has certainly

been the case over the past couple of years. The financial system came very close to collapsing and had it

not been for the aggressive and immediate responses from the global monetary authorities many of today’s

surviving institutions might have also disappeared, without commenting on the moral hazard debate that this

has engendered.

In order to avoid similar events taking place in the future, the world’s leading executives and policy makers

have come together to develop means of predicting and perhaps preventing similar crises from engulfing the

financial system in the future, sadly with each sitting on either side of the debate. While these efforts are to

be commended it is not clear whether they are necessarily targeting the real reasons behind the current crisis

and whether such preventions would be effective for future crises that will have their own specificities and

implications.

From our perspective, these efforts will be in vain unless executives, policymakers, and developers of scientific

models come together to have an open debate about the true causes and implications of the current crisis and

what actions need to be taken in order to rectify them, since many of the main culprits are still present within

the system. This is the reason that we established the Cass-Capco Institute Paper Series on Risk in collaboration

with Cass Business School. To focus on the real issues that impact our industry and not just those that get press

attention.

These issues are covered in a number of the articles in this edition of the paper series, which we are confident

will be of interest, and certainly of useful benefit to our executive readers.

At Capco we strive to bring together leading academic thinking along with business application, as either in

isolation will be of limited benefit to anyone. We hope that this issue of the paper series has also achieved that

objective.

Good reading and good luck.

Yours,

Rob Heyvaert,

Founder and CEO, Capco

Dear Reader,

The past year or so has been very bruising for executives sitting at the helms of the world’s major financial insti-

tutions. Many came very close to failing, and quite a few in fact did. Many of the most respected global financial

brands have now disappeared.

There have been numerous reasons put forward for the current crisis, but most have focused on the most obvious

causes and not necessarily on the underlying issues that are most pertinent. The aim of this edition of the Cass-

Capco Institute Paper Series on Risk is to give adequate attention to those issues that have fallen under the radar,

mainly through neglect by headline-grabbing journalists or because press attention has been focused elsewhere.

The papers address the damage that securitization has caused through the disappearance of due diligence at the

issuers of mortgage lending. They also examine why it is that most, if not all, of the risk measurement and manage-

ment models look great on paper but are of very little value to the institutions that need them most.

Risk management had in recent years reached the status of a science. Yet if anything, the current crisis proved

that it is anything but. There are a myriad of models and formulas that have been found wanting during the recent

crisis. In fact, none of the quantitative models worked. Those institutions that survived did so simply because of

the actions of the monetary authorities of the countries in which they were based.

We still have a long way to go before the models developed at academic institutes that view economics and finance

as science rather than art can actually be put to effective practice within financial institutions. The gap between

theory and practice remains as large today as it has ever been.

Similarly, in the field of risk management the gap between hoped-for and actual effectiveness has also remained

quite large. Financial institutions will have a very tough time instituting effective risk management policies so long

as they are dealing with spaghetti-like IT systems that simply don’t communicate with one another. This situation

prevents management from getting a holistic picture of the different pockets containing highly correlated risks.

This edition of the paper series highlights these issues so that actions can be taken in time for the next inevitable

crisis. It insists that financial executives wake up to the fact that investing in new and more advanced models is of

little benefit if they are not aware of all the risks they face. Those at the helm need to get their IT in order before

any of the models can be applied. We cannot continue putting the cart before the horse.

We hope that you enjoy this issue of the Journal and that you continue to support us in our efforts to mitigate the

gap between academic thinking and business application by submitting your ideas to us.

On behalf of the board of editors

The dream of effective ERM

opinion

A partial defense of the giant squid

Thou shalt buy ‘simple’ structured products only

Enterprise friction — the mandate for risk management

Enterprise Risk Management — a clarification of some common concerns

Capital at risk — a more consistent and intuitive measure of risk

Taibbi, M., 2009, “The great American bubble machine,” Rolling Stone magazine, 1

issue 1082-83, July 2.

See: Don’t dismiss Taibbi: what the mainstream press can learn from a Goldman take-2

down, The Audit, posted on the CJR website on August 08, 2009.

Based on the ranking system developed in: Carter, R. B., F. H. Dark, and A. K. Singh, 3

1998, “Underwriter reputation, initial returns and the long-run performance of IPO

stocks,” Journal of Finance, 53:1, 285-311.

Spinning involves the underwriter allocating underpriced IPOs to favored executives 4

– the quid pro quo being a promise of future business. Laddering involves allocations

conditioned upon buyers agreeing to purchase additional shares of the IPO in the after-

market. The SEC sanctioned various underwriting firms including Goldman Sachs, which

paid a fine of U.S.$40 million without admitting wrongdoing. The firm also reportedly

paid U.S.$110 million to settle an investigation by New York state regulators.

Taibbi quotes Professor Jay Ritter, a leading IPO researcher at the University of 5

Florida: “In the early eighties, the major underwriters insisted on three years of

profitability. Then it was one year, then it was a quarter. By the time of the Internet

bubble, they were not even requiring profitability in the foreseeable future.”

See: Jaggia, S., and S. Thosar, 2004, “The medium-term aftermarket in high-tech 6

IPOs: patterns and implications,” Journal of Banking and Finance, 28, 931-950.

For those who have been meditating at a Buddhist monastery over

the last year, the giant squid in the title refers to Goldman Sachs

Inc. Matt Taibbi writing in Rolling Stone magazine1 characterizes

the investment bank as: “a great vampire squid wrapped around

the face of humanity, relentlessly jamming its blood funnel into

anything that smells like money.” The article is (to put it mildly)

a colorful polemic hurled at Goldman Sachs accusing it of essen-

tially creating and profiting from various financial bubbles since the

onset of the Great Depression.

Taibbi’s rhetoric was not well received in the mainstream business

press. Reactions were dismissive (along the lines of: simplistic analy-

sis; he’s not a real business reporter!), indignant (basically objecting

to the article’s over-the-top language), defensive (all of them do it,

why pick on Goldman?) but seemed not to engage with the substance

of Taibbi’s accusations. In fact, an ‘audit’ done by the Columbia

Journalism Review’s Dean Sparkman largely validates Taibbi’s sub-

stantive claims2.

One of these claims relates to the tech sector bubble of the late

1990s. With the benefit of hindsight, it is clear that many of the

high-tech IPOs launched in this period were based on dubious

valuations. Goldman was certainly active in IPO underwriting and

had the highest ranking in terms of underwriter reputation [Carter

et al. (1998)]3. The firm also had its share of high-profile misfires

(for example: Webvan, Etoys) when the dot.com mania peaked and

crashed in the spring of 2000. Goldman was also arguably involved

in activities such as spinning and laddering4; the latter has the

effect of artificially pumping up the stock prices of IPO firms in the

aftermarket. But was Goldman a particularly egregious offender in

a climate in which underwriting best practices had slipped precipi-

tously?5 And how should this be evaluated?

As it turns out, we were involved in researching high-tech firms

that had an IPO in the late 1990s. We found significant positive

momentum and sharp reversals within a six-month aftermarket

window6. When we were doing the study, underwriter reputation

was not a central concern — it was only one of several control vari-

ables we employed. However, in the wake of the Taibbi article and

the considerable controversy it has generated, we thought it would

be interesting to revisit our sample to see if we could uncover any

interesting facts related to underwriter identity.

The set upOur primary sample was drawn from ipo.com, which lists the

universe of U.S. IPOs with dates, offer prices etc. classified in a

number of categories. We chose all IPOs from January 1, 1998

through October 30, 1999 in the following sectors: biotechnol-

ogy, computer hardware, computer software, electronics, Internet

services, Internet software, and telecommunications. This resulted

in a sample of 301 high-tech IPO firms. We stopped at October 30,

1999 because we wanted to study medium-term aftermarket price

behavior beyond the IPO date while excluding the market correc-

tion that commenced in 2000 [Jaggia and Thosar (2004)].

In Figure 1, we provide selected descriptive statistics relating to our

sample broken down by three lead underwriter reputation tiers:

top, medium, and bottom. The top-tier underwriters are those that

received the highest score of 9 in the Carter et al. (1998) rank-

ing system. These are: Goldman Sachs, Credit Suisse First Boston

(renamed Credit Suisse), Hambrecht & Quist, and Salomon Smith

Barney. The medium-tier underwriters are those with a score

between 8.75 and 8.99, while the bottom tier includes all firms with

a score below 8.75.

There do not appear to be any obvious differences between IPO

firms represented by top-tier underwriters and the others in terms

of objective quality criteria. If anything, metrics such as: the level

of initial underpricing, percentage of profitable firms, and firm age

A partial defense of the giant squid Sanjiv Jaggia, Professor of Economics and Finance, California Polytechnic State University

Satish Thosar, Professor of Finance, University of Redlands

8

Underwriter reputation

Variables Top-tier Medium-tier Bottom-tier

Cumulative market-adjusted return (CMAR) at the end of six months

44.71(140.91)

19.95(101.88)

-16.55(57.23)

Percentage change from offer to market open price

65.74(72.26)

68.38(105.15)

43.66(73.18)

Percentage of firms with positive net income in pre-IPO year

21.11(41.04)

16.90(37.61)

24.64(43.41)

Revenue in pre-IPO year ($ millions) 83.71(285.64)

50.71(200.94)

75.84(397.87)

Offer size ($ millions) 134.48(176.86)

126.97(490.20)

48.60(47.48)

Percentage of firms with green-shoe (over-allotment) option

64.44(48.14)

54.23(50.00)

55.07(50.11)

Percentage of firms belonging to the Internet services or software categories

61.11(49.02)

60.56(49.04)

59.42(49.46)

Firm age at IPO date (years) 4.45(3.40)

5.61(5.88)

5.90(6.03)

Number 90 142 69

Notes:Standard deviations are in parentheses below the sample means.1. Top-tier underwriter firms are those assigned the highest point score of 9 in the 2. Carter et al. (1998) system. This category includes Goldman Sachs. Medium-tier firms are those with a score of 8.75 – 8.99. Bottom-tier are all those below 8.75.A green-shoe provision gives the underwriter the option to purchase additional 3. shares at the offer price to cover over allotments. Presence of the provision indirectly increases underwriter compensation.

Figure 1 – Selected descriptive statistics for IPO firms classified by underwriter

reputation

Let P7 i1 represent the day 1 open price of the ith firm and let Pm1 be the corresponding

level of the market (Nasdaq) index. Similarly, Pit and Pmt represent the open price at

day t of the ith firm and the market respectively. The CMAR of the firm at time t is

calculated as: CMARit = [Pit/Pi1] ÷ [Pmt/Pm1] -1. The time in question does not refer to

calendar time, but to the time from the IPO date.

seem to favor the bottom-tier group. The average IPO offer size is

considerably larger for the top-tier compared to the bottom-tier

group, which is not surprising. One expects that superior reputation

carries with it the ability to tap more extensive investor networks

to raise larger chunks of capital at a given time. Also worth noting

is that the top-tier group has the highest proportion of contracts

with a green-shoe option. A green-shoe provision gives the under-

writer the option to purchase additional shares at the offer price to

cover over-allotments and thereby indirectly increases underwriter

compensation.

A striking and somewhat surprising difference is in the cumula-

tive market-adjusted returns (CMAR) registered by each group.

To study this in greater detail, we graph (Figure 2) the CMAR for

each group using the post-IPO day 1 open price as the base through

trading-day 125 or approximately six months after the IPO date7.

There are visual and arguably economically significant differences

across groups. The bottom-tier group (green) immediately slips into

negative territory and stays there for a six-month CMAR of -16.5

percent. The medium (red) and top (blue) groups display strong

positive momentum and reach a CMAR peak of 40.5 percent (at

day 114) and 61.5 percent (at day 112) respectively. The CMAR then

tapers off possibly due to the onset of lock-up expiration pressures

and at the end of six months ends up at 20.0 percent and 44.7 per-

cent for the medium and top groups respectively.

This can be viewed in a number of ways. If the market is behav-

ing rationally and recognizing ‘true value’ as time elapses, the

45 percent CMAR displayed by the top group represents serious

underestimation of the initial IPO offer prices. It represents in

effect a wealth transfer from the founders and seed financiers of

the firm to outside investors and this is over and above the initial

underpricing of 66 percent for this group (Figure 1). Under normal

circumstances, the underwriters could be justly accused either of

incompetence in terms of valuation or extorting their IPO clients to

enrich themselves and their favored customers.

On the other hand, if informed investors recognize that tech sector

stock prices are inflated, unsustainable, and are in the market to

exploit the ‘greater fool,’ the CMAR patterns may reflect the ability

of certain underwriters through their analyst coverage, laddering

arrangements, etc., to not only stabilize but pump up prices in the

aftermarket until the wealth transfer from uninformed to informed

investors is duly complete.

We decided that a closer disaggregated look at the top-tier group

might be useful.

The defenseIn Figures 3 and 4, we report descriptive statistics and CMAR pat-

terns for the IPOs underwritten by top-tier firms. Hambrecht &

Quist and Salomon Smith Barney are combined so as to represent

a reasonable sample size; Credit Suisse and Goldman Sachs are

reported separately.

A few metrics are worth noting. More firms represented by Goldman

(27 percent) were profitable in their pre-IPO year than Credit Suisse

9

Note: Top-tier underwriter firms are those assigned the highest point score of 9 in

the Carter et al. (1998) system. In our sample, they represent 90 firms. Medium-tier

are those with a score of 8.75 – 8.99 representing 142 firms.

Bottom-tier are all those below 8.75 representing 69 firms.

Figure 2 – Cumulative market-adjusted returns (CMAR) for IPO firms grouped by lead

underwriter reputation

-20

0

20

40

60

80

100

120

0 25 50 75 100 125

Trading days

CMAR CS GS Rest

-30

-20

-10

0

10

20

30

40

50

60

70

0 25 50 75 100 125

Trading days

Top-tier Medium-tier Bottom-tierCMAR

Variables cs Gs Rest

Cumulative market-adjusted return (CMAR) at the end of six months

80.09(165.39)

27.51(131.84)

30.55(121.06)

Percentage change from offer to market open price

74.34(75.17)

85.67(81.23)

26.63(28.61)

Percentage of firms with positive net income in pre-IPO year

14.29(35.64)

27.03(45.02)

20.00(40.83)

Revenue in pre-IPO year ($ millions) 41.34(147.31)

37.32(50.56)

199.81(504.84)

Offer size ($ millions) 101.97(103.99)

155.59(222.89)

139.63(165.43)

Percentage of firms with green-shoe (over-allotment) option

67.86(47.56)

86.47(34.66)

28.00(45.83)

Percentage of firms belonging to the Internet services or software categories

71.43(46.00)

59.46(49.77)

52.00(50.99)

Firm age at IPO date (years) 4.01(2.13)

5.03(3.98)

4.07(3.63)

Number 28 37 25

Notes:Standard deviations are in parentheses below the sample means.1. Top-tier underwriter firms are those assigned the highest point score of 9 in 2. the Carter et al. (1998) system; CS = Credit Suisse, GS = Goldman Sachs, Rest = Hambrecht & Quist and Salomon Smith Barney A green-shoe provision gives the underwriter the option to purchase additional 3. shares at the offer price to cover over allotments. Presence of the provision indirectly increases underwriter compensation.

Figure 3 – Selected descriptive statistics for IPO firms classified by top-tier

underwriters

10 – The journal of financial transformationThis may reflect Goldman’s greater clout even within the top-tier group.8

(14 percent) and the rest (20 percent). Goldman firms were also

marginally longer in business before the IPO date. On the other

hand, Goldman firms were subject to greater initial underpricing on

average. They also had a higher average offer size and were more

likely to be subject to a green-shoe provision8. But the most sug-

gestive statistic in our view is the six-month CMAR. The Goldman

group’s CMAR at 28 percent is significantly lower than that of the

Credit Suisse group which racked up 80 percent. Thus aftermarket

momentum (or manipulation if one were to take the cynical view) is

lowest for firms represented by Goldman.

This is borne out by the CMAR patterns in Figure 4. The red line

representing Goldman firms is virtually flat in the immediate after-

market, when most purported price pumping takes place. The blue

(Credit Suisse) and green (Hambrecht & Quist and Salomon Smith

Barney) lines suggest higher levels of momentum and reversal

within a six-month period — more of a bubble within a bubble pat-

tern with the benefit of hindsight.

After all is said and done, the tech bubble is only one instance of a

series of such events in recorded history. And, while these events

result in a lot of wealth destruction, the firms left standing in the

end usually signify technological and productivity gains to society,

which may in the long-run exceed the Schumpeterian costs.

We do not profess to know how to execute such a cost-benefit analy-

sis. Instead, we decided to undertake an outlier analysis within our

sample. We carried out a case study-type analysis of the 20 firms

that registered a six-month CMAR of more than 100 percent and

were represented by top-tier lead underwriters. We were essentially

projecting ourselves back in time before the crash and picking a small

subset of the likeliest candidates for success. How did they perform

over the long-term? We traced the fortunes of these 20 firms from

their IPO date up until the present (August 2009). We examined

available financials, stock price performance, mergers, consolida-

tions etc. Several firms were targets of class-action lawsuits filed by

aggrieved stockholders claiming misstatements in the IPO prospec-

tus and the like. Our findings are summarized in Figure 5, which is

essentially a status report on each firm. We assigned each firm into

one of following categories, or letter grades if you will.

A These are all firms that have survived and thrived. In our judg-

ment, they all have successful business models and good pros-

pects going forward. An investor who bought shares soon after

the IPO date and held on to them till August 2009 would have

realized significant positive returns. Only four of the 20 firms

receive an A grade — three of these (Ebay, Juniper Network,

Allscripts) were lead underwritten by Goldman Sachs. The fourth

(F5 Networks) was underwritten by Hambrecht & Quist.

B The three firms in this category are viewed as viable ongoing

enterprises. There is a fair amount of within-group variation. For

-20

0

20

40

60

80

100

120

0 25 50 75 100 125

Trading days

CMAR CS GS Rest

-30

-20

-10

0

10

20

30

40

50

60

70

0 25 50 75 100 125

Trading days

Top-tier Medium-tier Bottom-tierCMAR

Note: CS = Credit Suisse representing 28 firms; GS = Goldman Sachs representing 37

firms; Rest = Hambrecht & Quist and Salomon Smith Barney together representing

25 firms.

Figure 4 – Cumulative market-adjusted returns (CMAR) for IPO firms grouped by top-

tier underwriters

ipo firm name/current or former ticker symbol

cMAR % lead underwriter status

Infospace Inc (INSP) 199.2 Hambrecht & Quist B

Art Technology Group (ARTG) 255.4 Hambrecht & Quist B

F5 Networks Inc (FFIV) 469.9 Hambrecht & Quist A

Inktomi Corp (INKT) 142.7 Goldman Sachs C

Ebay Inc (EBAY) 623.4 Goldman Sachs A

Viant Corp (VIAN) 183.6 Goldman Sachs D

Active Software Inc (ASWX) 248 Goldman Sachs C

Allscripts Inc (MDRX) 103.9 Goldman Sachs A

Tibco Software (TIBX) 182.8 Goldman Sachs B

Inet Technologies (INTI) 119 Goldman Sachs C

Juniper Network Inc (JNPR) 116.5 Goldman Sachs A

NetIQ Corp (NTIQ) 141.9 Credit Suisse C

Appnet Systems Inc (APNT) 158.4 Credit Suisse C

Commerce One Inc (CMRC) 609.8 Credit Suisse C

E.Piphany Inc (EPNY) 138.3 Credit Suisse C

Phone.com Inc (PHCM) 141.4 Credit Suisse C

Software.com Inc (SWCM) 237.3 Credit Suisse C

Tumbleweed Software Corp (TMWD) 201.7 Credit Suisse C

Liberate Technologies (LBRT) 487.2 Credit Suisse C

Vitria Technology (VITR) 250 Credit Suisse C

Notes:The above firms were represented by top-tier lead underwriters and experienced 1. post-IPO six-month cumulative market-adjusted returns (CMAR) greater than 100 percent.Status (August 2009) definitions are given below:2. Successful ongoing enterprises; significant positive returns realized by early long-3. term investors.Viable ongoing enterprises.4. Merged, restructured or otherwise consolidated; significant impairment to early 5. valuations.Defunct.6.

Figure 5 – Current status of selected IPO firms launched during the dotcom bubble era

11Jonathan A. Knee, senior managing director at Evercore Partners, in the New York 9

Times, DealBook Dialogue, October 6, 2009.

Critics may point out that Goldman would likely have gone under (or at least taken 10

large losses) if the U.S. taxpayer had not bailed out AIG and thereby its counterpar-

ties.

instance, Infospace (Hambrecht & Quist) has negative income

in its latest financial year but still has a market capitalization of

U.S.$293 million. Early post-IPO investors who held on to their

position would see a negative return. In contrast, Tibco Software

(Goldman) is profitable, has a current market capitalization of

U.S.$1.61 billion, and a P/E multiple of 27. The only reason Tibco

did not get an A grade is that early buy-and-hold investors would

register a negative stock return.

C The twelve firms in this group were severely impacted in the

tech sector crash of 2000. While a small number survive with

their original stock ticker symbol, none of these are profitable or

actively traded. Most have merged, restructured, or otherwise

consolidated. The common element is that early investors who

had not divested before the crash would have suffered signifi-

cant (if not quite total) losses. Goldman represented three firms

in this group.

D The one firm in this category (Viant; Goldman) declared bank-

ruptcy in 2003 and is essentially defunct.

Hambrecht & Quist represented only three firms (1 A; 2 Bs), all of

which survive and in aggregate delivered considerable value to

early investors. Goldman’s record is mixed with three As and a B

balanced out with three Cs and a D. Credit Suisse has the poorest

record in terms of our sample (9 Cs). None of the firms they rep-

resented were successful in weathering the tech sector shakeout.

Thus, even among the small subset of IPO firms represented by

top-tier underwriters and greeted with sustained enthusiasm by

investors, ex-post analysis reveals considerable variation in the

staying power of their business models.

conclusionA respected market observer recently commented: “When faced

with market euphoria, whatever its source, financial institutions

will always be confronted with the same stark choice: lower your

standards or lower your market share.”9

Goldman was certainly part of the general deterioration of under-

writing standards but our analysis reveals that they did represent

some very good firms and in terms of our CMAR analysis were a

reasonably responsible player in the IPO aftermarket. Perhaps their

quality control mechanisms were not quite so compromised. More

recently, they seem to have recognized the risks stemming from

subprime lending well ahead of their competitors, hedged with

some success, and have emerged from the financial crisis more or

less intact10. We doubt that Taibbi would set much store by this but

there it is.

12 – The journal of financial transformation

Thou shalt buy ‘simple’ structured products onlySteven Vanduffel, Faculty of Economics and Political Science, Vrije Universiteit Brussel (VUB)

Structured products are a special type of investment product; they

appear to offer good value in two situations. The first is when you

can sell them with an attractive margin, such that the payoff pro-

vided at the end of the investment horizon T>0 is hedged, and not

of your concern. This method of value creation is possible for banks

and financial planners but is not really in reach of retail customers.

The second possibility consists of purchasing these instruments

and making sufficiently high investment returns with them. In this

note we claim that one must be extremely careful when one is

on the buyer’s side; the odds may go against you, transforming a

potential cash cow into a loss generator. We will elaborate on this

point further.

An obvious question people ask themselves when investing is how

to do it in an optimal way. Of course, everyone who prefers more

to less, which is akin to having an increasing utility curve [von

Neumann and Morgenstern (1947)], will want to invest the available

funds in a product that will ultimately provide the highest return.

Unfortunately an investor cannot predict with certainty which

investment product will outperform all others. However, ex-ante

one may be able to depict for each investment vehicle all possible

returns together with the odds to achieve these, hence determining

the distribution function for the wealth that will be available at time

T, and this will be called the wealth distribution.

Hence, the quest for the best investment vehicle amounts to

determining the most attractive wealth distribution first. There

is, however, no such thing as a wealth distribution function that is

universally optimal; simply because optimality is intimately linked

to the individual preference structure, the utility curve, of the

investor at hand. Some people may prefer certainty above all, and

in such instance putting the available money in a state deposit is

the best one can do. Others may prefer a higher expected return

at the cost of more uncertainty and may find it appropriate to

invest in a fund that tracks the FTSE-100. For some people neither

of these options is satisfactory and they may wish to seek capital

protection while taking advantage of increasing markets as well.

For example, they may want to purchase a product that combines

the protection of capital with a bonus of 50% in case the FTSE-100

has increased in value during the entire investment horizon under

consideration. Others may find this too risky and may prefer a

product which protects their initial capital augmented with 50%

of the average increase (if any) of the FTSE-100 during the invest-

ment horizon.

The different examples hint that building the optimal wealth dis-

tribution and corresponding investment strategy often consists

of combining several underlying products, such as fixed income

instruments, equities, and derivatives. In the context of this note,

such a cocktail of products will be further called a structured prod-

uct. Hence, using structured products one is able to generate any

desired wealth distribution, and in this sense there is no doubt that

structured products can offer good value.

Is this the end of the story? Not really, because once a convenient

wealth distribution has been put forward and a structured product

has been designed to achieve this, one may still raise the following

question: how can we obtain this wealth distribution at the lowest

possible cost? Of course, such a cost efficient strategy, if it exists,

would be preferred by all investors. But does it exist? Intuitively

one may think that two products with the same wealth distribution

at maturity should bear the same cost. Surprisingly this common

belief is absolutely untrue and careless design of a structured prod-

uct can cause a lot of harm.

cost efficiency in a complete single-asset market Dybvig (1988) showed that in a so-called complete single-asset

market the most efficient way to achieve or build a wealth distribu-

tion is by purchasing ‘simple’ structured products. Here ‘complete-

ness’ refers to a financial market where all payoffs linked to the

single underlying risky asset are hedgeable or replicable [Harrison

and Kreps (1979)], and ‘simple’ refers to the feature that the payoff

can be generated using a (risk-free) zero-coupon bond and plain

vanilla derivatives (calls and puts).

Indeed, Dybvig showed that optimal payoffs only depend on the

value of the underlying asset at maturity not at intermediate times.

In other words, they should be path-independent. He also showed

that they should be non-decreasing in the terminal value of the

underlying risky asset. For example, if S(t) (0≤t≤T) is reflecting the

price process of the risky asset, then the exotic payoff S2(T)-3 is a

path-independent and non-decreasing payoff, whereas the payoff

S(T/2).S(T) is path-dependent.

Moreover, since these path-independent payoffs can be approxi-

mated to any degree of precision by a series of zero-coupon

bonds and calls and puts, the logical conclusion of Dybvig’s work

would indeed be that proper investments only consist of simply

purchasing the appropriate proportions of bonds and plain vanilla

derivatives (calls and puts) written on the underlying risky asset.

Path-dependent, thus complex, structured products are then to

be avoided. We remark that Cox and Leland (1982, 2000) already

showed (using other techniques) that optimal payoffs are necessar-

ily path-independent but this result is more limited because it only

holds for investors that are risk averse, whereas Dybvig’s result

holds for all investors (assuming they all prefer more to less).

More general marketsOne can argue that the above mentioned results are limited in the

sense that they only hold for a single-asset market, which is quite

13

restrictive, and also that they are contingent on a major assump-

tion, i.e., the claim that markets are complete.

Let us start with the first assumption and analyze to which extent

it can be relaxed to multi-asset markets. Assuming that the sub-

sequent periodical asset returns of the different risky assets are

(multivariate) normally and independently distributed (nowadays

also referred to as a multidimensional Black-Scholes market),

Merton (1971) already proved that risk averse investors exhibiting

a particular, so-called CRRA utility function would always allocate

their funds to a riskless account and a mutual fund representing

the so-called market portfolio; implying that their payoff at matu-

rity can be understood as a particular path-independent deriva-

tive written on an underlying market portfolio. Recently Maj and

Vanduffel (2010) have generalized this to include all risk averse

decision-makers. They have shown that optimal pay-offs are neces-

sarily path-independent in the underlying market portfolio. Hence,

in a multidimensional Black-Scholes market, the only valuable struc-

tured products are those which can be expressed as a combination

of a fixed income instrument and plain vanilla derivatives that are

written on the underlying market-portfolio.

Regarding the assumption of normally distributed returns, it is well-

known that this is a false assumption in general. See, for instance,

Madan and Seneta (1990) for evidence, but also look at Vanduffel

(2005), where it is argued that this assumption is also dependent

on the length of the investment horizon involved. Nevertheless, a

widely accepted paradigm to describe asset returns involves the

use of Lévy processes. Hence, let us assume that this is actually

being used by market participants, and also that they agree to use

the so-called Esscher tranform (exponential tilting) to price deriva-

tives. Vanduffel et al. (2009a) have shown that optimal structured

products are necessarily path-independent; providing further evi-

dence against the use of complex (path-dependent) structures.

final remarksIn this note we have given some theoretical evidence that from

a buyer’s point of view the only (financially) valuable structured

products are path-independent and thus ‘simple.’ While compli-

cated structured products may provide some emotional value or

happiness to the investor they do not seem to be the right vehicles

for generating financial value. For illustrations of the different

theoretical results we refer to Dybvig (1988), Ahcan et al. (2009),

Vanduffel et al. (2009a,b), Maj and Vanduffel (2010), and Bernard

and Boyle (2010).

Let us also remark that besides theoretical inefficiency, path-

dependent payoffs also have a practical cost disadvantage. Indeed,

as compared to plain vanilla products they also suffer from a lack of

transparency and liquidity, creating the potential for their promot-

ers to make them more expensive than they fairly should be.

We also remark that at this point we do not claim that in all theoreti-

cal frameworks path-independent structures are value-destroying,

and we leave it for future research to analyze to what extent the

different findings can be generalized and interpreted further. Our

conjecture is that, at least, structured products should be designed

‘in some particular way’ in order to be potentially optimal.

All in all, from a buyer’s perspective the only good structured prod-

ucts to invest in seem to be the simple ones and one can argue that

even these may be unnecessary. Indeed, using basic products such

as equities, property, and bonds one may be able to design a well-

balanced wealth distribution as well; hence avoiding all other bells

and whistles that are associated with derivatives.

Finally, note that the optimal design of structured products is a

research topic that has also been picked up by some other authors

including Boyle and Tian (2008) and Bernard and Boyle (2010).

ReferencesBernard, C., and P. Boyle, 2010, “Explicit representation of cost efficient strategies,” •

working paper, University of Waterloo

Boyle, P., and W. Tian, 2008, “The design of equity indexed annuities,” Insurance, •

Mathematics and Economics, 43:3, 303-315

Cox, J. C., and H. E. Leland, 1982, “On dynamic investment strategies,” Proceedings of •

the seminar on the Analysis of Security Prices, 262, Center for Research in Security

Prices, University of Chicago

Cox, J. C., and H. E. Leland, 2000, “On dynamic investment strategies,” Journal of •

Economic Dynamics and Control, 24, 1859-1880

Dybvig, P, H., 1988, “Inefficient dynamic portfolio strategies or how to throw away a •

million dollars in the stock market,” The Review of Financial Studies, 1:1, 67-88

Harrison, J. M., and D. M. Kreps, 1979, “Martingales and arbitrage in multiperiod securi-•

ties markets,” Journal of Economic Theory, 20, 381-408

Madan, D. B., and E. Seneta, 1990, “The variance gamma (VG) model for share market •

returns,” Journal of Business, 63, 511-524

Maj, M., and S. Vanduffel, 2010, “Improving the design of financial products,” working •

paper, Vrije Universiteit Brussel

Merton, R., 1971, “Optimum consumption and portfolio rules in a continuous-time •

model,” Journal of Economic Theory, 3, 373-413

Vanduffel, S., A. Ahcan, B. Aver, L. Henrard, and M. Maj, 2009b, “An explicit option-•

based strategy that outperforms dollar cost averaging,” working paper

Vanduffel, S., A. Chernih, M. Maj, and W. Schoutens, 2009a, “A note on the suboptimal-•

ity of path-dependent pay-offs for Lévy markets,” Applied Mathematical Finance, 16:4

http://www.informaworld.com/smpp/title~db=all~content=t713694021~tab=issueslist~

branches=16 – v16, 315-330

Vanduffel, S., 2005, “Comonotonicity: from risk measurement to risk management,” •

PhD thesis, University of Amsterdam

von Neumann, J., and O. Morgenstern, 1947, “Theory of games and economic •

behavior,” Princeton University Press, Princeton

14 – The journal of financial transformationwww.opriskandcompliance.com1

Enterprise friction — the mandate for risk managementSandeep Vishnu, Partner, Capco

In today’s battered economy, few are willing to put in place any-

thing that might meddle with earnings potential. Even fewer are

willing to spend money on something that may offer only a theo-

retical return on investment. Behind the polite but forced smiles

and handshakes from executives, there is a silent accusation: risk

management dampens revenue and puts brakes on innovation. This

is a challenge faced by risk managers as they try to put in place

structures to guard against losses.

But risk management is not about playing it safe; it is about play-

ing it smart. It is about minimizing, monitoring, and controlling the

likelihood and/or fallout of unfavorable events caused by unpredict-

able financial markets, legal liabilities, project failures, accidents,

security snafus — even terrorist attacks and natural disasters. There

is always risk in business, and risk management should be designed

to help companies navigate the terrain.

Sure, risk management may at times call on companies to pull back

on the reins, and it certainly is not free. However, risk management

provides a counterpoint to enterprise opportunity — friction, if you

will — that not only avoids unnecessary losses, but enhances the

ability of organizations to respond effectively to the threats and

vulnerabilities to which they are exposed in the course of busi-

ness.

Friction is a much-maligned term. The connotations are more often

negative than positive: a retarding effect, in-fighting, etc. Even in

physics, it is characterized as a necessary evil. The fact that friction

allows us to walk properly is often overlooked. One can understand

the importance of friction simply by trying to walk on ice, where the

absence of friction causes one to slip and slide. By contrast, high

friction makes walking on wet sand inordinately difficult.

Imagine Michael Jordan without his Nikes. One might argue that a

stockinged Jordan might be less encumbered with the extra weight

and trappings of footwear. But few could argue that he would have

been as effective on the polished court that was his field of opera-

tions.

Achieving a proper balance of risk in strategic planning and opera-

tions for the enterprise is critical — especially in today’s environ-

ment, when resilience and agility in the face of uncertainty are as

important as the effective use of identified variables.

Now, even more than before, the goal of enterprise risk manage-

ment is to define and deliver the right level of friction, because

getting this calibration exactly right can help improve enterprise

agility. Too little friction, and a company could slip into dangerous

scenarios; too much friction, and a company could just get stuck.

Getting this right can not only drive corrective measures, but can

also serve as an effective counterbalance.

To create that necessary, well-balanced friction, companies need

to take some fundamental steps. In this report, we will examine the

three cornerstones of robust risk management — cornerstones that

will help win over the naysayers and more importantly, ensure that

companies have the most-efficient risk management programs in

place. But for this to work, we believe that:

Companies will need to change their view of risk management ■■

as a necessary evil, and elevate the role of this critical function

within their organizations. Risk should have a strong voice at the

management table, and risk mitigation should be given as much

importance as risk taking.

Executives will need to design a new blueprint for strategic ■■

management that integrates risk management into every criti-

cal element of the enterprise and thereby drives a risk-sensitive

culture.

Organizations will have to buttress the risk management infra-■■

structures they already have in place, including, data, analytics,

and reporting.

History lessons: extending the time horizon of risk managementOne of the toughest issues associated with implementing an effec-

tive risk management program revolves around effectively funding

the efforts. The challenge often boils down to cost versus benefit. In

the months and years that led to the housing market and mortgage

meltdown, credit crisis and subsequent recession, companies paid

less attention to risk management mainly because times were so

good for so long. Typically, managing risks was not an embedded

element in critical business processes; it was a bolt-on activity.

Consider a 2007 global risk information management survey

conducted by OpRisk & Compliance magazine1. Although a major-

ity of respondents said they were at least somewhat effective in

providing the right information to the right people at the right

time to meet the organization’s business requirements, the survey

indicated that many felt the information they had was not being

used effectively to create a risk management culture within their

organizations. That same survey also illuminated the ever-present

ROI hurdle. Only 9 percent of respondents said their firms were able

to trace the ROI of information management initiatives designed to

capture and manage vital corporate data.

The biggest problem with risk management is in the establishment

of its ROI. Risk management is primarily about loss avoidance, and

it is difficult to measure what has been avoided and thus has not

occurred. Efficiency and effectiveness metrics are often dwarfed by

the magnitude of loss avoidance; however, while the former can be

measured, the latter are estimated.

Executives are often heard saying, “We haven’t faced a serious

15

problem; why are we spending this much money?” It is a classic

Catch-22.

The ROI problem is compounded by the fact that risk management

can sometimes act as a retardant to growth. Anything — especially

the cost to fund an initiative with fuzzy ROI metrics or threaten a

company’s profit margin — is definitely going to be suspect.

We suggest that firms refrain from looking at risk management

costs within quarterly financial reporting time frames. Analysis of

risk management’s return needs to be considered over much longer

time horizons.

Sometimes retardants are necessary for growth. With 20/20 hind-

sight, many contend that Wall Street should have invested more

strategically in risk management practices and infrastructure. If

risk profiles had been based on a 30-year historical record rather

than the standard 100-year window, a fair number of the subprime

and adjustable rate mortgage (ARM) loans processed during the

housing bubble and in the years leading up to the recession would

have been seen more clearly for what they became — toxic assets

that damaged the health of the financial industry and ultimately the

entire U.S. economy.

In all likelihood, many on Wall Street probably knew intuitively the

risks they were taking; they just chose to make a management

decision that the risk was acceptable because the top line return

was so attractive. We would not call it willful negligence, but we do

believe it was a measured decision recognizing that the benefit of

the upside potentially outweighed the likelihood of the downside. Of

course, we doubt anybody suspected that the downside would turn

out to be as severe as it has been. This phenomenon is exacerbated

when dealing with new products and structures [i.e., collaterized

debt obligations (CDO) and mortgage backed securities (MBS)],

which do not have the same experiential basis as traditional prod-

ucts such as fixed-rate mortgages.

While longer-term views are desirable for assessments, ROI calcu-

lations, etc., they are sometimes hard to achieve. Consequently,

enterprises should try to embed risk management in all ‘risk-taking’

activities and ultimately drive a risk-sensitive culture. This can be

facilitated by the increased use of risk-sensitive measures such

as risk-adjusted return on capital (RAROC), economic value added

(EVA), shareholder value added (SVA), etc. For example, within

investment banks this would mean that the desk heads are not just

responsible for P&L, but also for the amount of capital used to gen-

erate that P&L. Extending this sentiment further, incentive compen-

sation can also be based on RAROC or similar metrics. These met-

rics allow for a degree of normalization across the enterprise and

allow the board and senior management to determine investment

strategy and consistently evaluate returns. If such metrics and

management approaches gain broad support, then markets would

begin assessing enterprise performance through risk-adjusted

measures in addition to traditional top-line and bottom-line metrics.

Regulatory standardization and public availability of such informa-

tion has greater potential to drive a shift toward sustainable growth

versus short-term results.

Fast forward to today, and risk management is still receiving short

shrift because so many companies are scrambling to make ends

meet. Few feel they have the financial resources or time to finance

and bolster existing risk management practices. It is once again a

Catch-22 situation:

When times are good, people do not want to pay attention to risk ■■

management because they are too busy making money.

When times are bad, people do not want to pay too much atten-■■

tion to risk management because they are already incurring

losses, and do not want to spend more money.

In the end, risk management gets only lip service. That needs to

change.

striking a balanceEnlightened enterprises promote creative tension between strategy

and risk management, and put in place a set of checks and balances

to guard against the exploitation of short-term opportunities at the

expense of long-term viability. Failure to strike this balance can

have devastating consequences, as evidenced by Countrywide’s

demise in 2009. In 2005, it was the largest mortgage originator;

however, 19 percent of its loans were option ARMs, and of those, 91

percent had low documentation.

More specifically, strategic considerations and risk assessments

need to be made in tandem. There must be a dynamic — even sym-

biotic — interaction between these two perspectives. They should

be seen as two sides of the same coin — like the classic ying-yang

balance principle.

Enterprise strategy

Enterprise risk

management

To effectively integrate risk considerations into the critical strate-

gic decision-making processes, organizations should incorporate

the following principles into every aspect of their management

philosophy.

16 – The journal of financial transformation

promote a culture of resilienceExecutives may well consider revisiting many of the major pillars of

their organization and refine critical processes by integrating risk

considerations into their enterprise architecture. Resilience and

agility should be primary goals of such efforts and should address

foundational elements such as data, as well as derived capabilities,

including analytics, and feedback loops driven through reporting.

Often organizations conduct risk assessments as a bolt-on activity.

But organizations that integrate resilience (and risk management

in general) into their culture in a granular manner stand a better

chance of not only mitigating risks more effectively, but also more

cost efficiently.

The agile software development process adopted by high-tech

organizations has demonstrated that integrating quality assurance

into the development process results in both higher-quality and

less-expensive final products. Checking for mistakes ‘after the fact’

is almost always more expensive.

Robust enterprise risk management (ERM) needs to leverage formal

structures — data, processes, and technology used for creating, stor-

ing, sharing, and analyzing information — as well as informal networks

represented by the communication and relationships both within and

without the risk management organization. Informal networks have

repeatedly shown their usefulness in identifying and mitigating fraud,

and often provide early warnings of potential tail events.

informal networks

informal networks

formal structures

info

rmal

net

wor

ks

info

rmal

net

wor

ks

The interplay between formal structures and informal networks

is important because this allows risk managers to compensate

for shortcomings in one by using the other. But this requires the

right culture to be in place: one that encourages staff to ask tough

questions without fear of being seen as inhibitors to growth. Risk

identification should not have punitive consequences. A culture

of appropriately calibrated enterprise friction should be fostered.

Doing this would allow critical elements of the organization to

accelerate their pursuit of opportunities while knowing that they

have the perspective, and operational ability, to slow down, acceler-

ate, or change course because of an appropriate sensitivity to key

risk parameters.

Compensation is, and always has been, a key lever in determining

the nature of a corporate culture. Culture depends heavily on incen-

tives, which today are often skewed toward rewarding upside ben-

efits and not necessarily avoiding downside losses. Compensation

practices need to become more risk-sensitive, so that they reward

long-term value creation and not just short-term gains. Risk mitiga-

tion should be as important as risk-taking to drive the appropriate

culture.

Governance

Reporting

Analytics

Data

Resilience

Data as a foundation for risk managementThere is a growing consensus among risk managers across indus-

tries — from government, to financial services, to manufacturing

and health care — that the data upon which key organizational

decisions are made represent the foundational layer for enterprise

risk management (ERM). Bad data can have an immediate and nega-

tive impact at any point of the organization, but the downstream

impacts of bad data can snowball out of control.

Some data challenges, such as completeness and timeliness, are

harder to overcome than others. However, incorporating a risk

management perspective on the design of a robust data model

can help reduce inconsistency and inaccuracy, and drive overall

efficiency. This can help address the challenges that result from

the fact that data often exists in silos, making it difficult to get an

accurate view of a related set of information across these silos.

Wachovia’s write-down of the Golden West financial portfolio, which

stemmed largely from overreliance on poor data, offers an example

of disproportionate emphasis being placed on valuations rather

than on borrower income and assets.

Another challenge relates to inconsistent labels, which make it hard

to match customers to key metrics over a common information

life cycle. Different technological platforms (which help create the

silos in the first place) make aggregation and synthesis challenging.

Most organizations lack a clear enterprise-wide owner in charge of

addressing such data quality issues, which complicates their identi-

fication and remediation.

17

Analytical risksAnalytical frameworks help translate data into actionable informa-

tion. However, analytics should not just be simple characterizations

of data. They should be timely and insightful so that analysis can

enable appropriate actions. In the financial services industry, the

credit crisis demonstrated how neglected — or inappropriate —

analytical frameworks prevented organizations from identifying

knowable risks (i.e., flawed model assumptions) and illustrated why

key decision-makers were unable to break through the opacity of

others (i.e., lack of transparency into the risk of underlying assets

being traded in secondary markets, especially when it related to

second-order derivatives).

All too often, analytical frameworks emerge as simplistic char-

acterizations of the ‘real world’ that may not be able to convey

a complete risk profile. This is evidenced by the overreliance on

value-at-risk as a key risk metric in the recent financial crisis. The

dissolution of Lehman Brothers and the near collapse of AIG offer

good examples of the shortcomings of traditional analytics, which

were unable to adequately account for dramatic increases in lever-

age, counterparty risk, and capital impacts as markets and ratings

deteriorated.

Reporting deficienciesReporting is a multidimensional concept that does not necessarily

capture the dynamic nature of information presentation. Typically,

reporting has at least four major stakeholders, two external (regu-

lators and investors) and two internal (senior management, includ-

ing the board of directors, and line management).

A strong risk-information architecture is crucial to delivering the

right information to the right audience in a timely manner. It should

present salient information as a snapshot, as well as provide the

ability to drill down into the detail. Well-defined business usage will

help drive overall requirements, while integrated technology plat-

forms can help deliver the processing efficiency needed to manage

the volumes and timeliness of information presentation.

Reporting has often been segmented into regulatory reporting

and management reporting, directed toward specific compliance

requirements for the former and financial statements for the lat-

ter. The financial crisis highlighted the need for organizations in

many industries to develop both ad-hoc and dynamic reporting,

which not only meet compliance requirements, but also, and more

importantly, improve the decision-making process. Many organiza-

tions are coming to the conclusion that current architectures and

infrastructures might not necessarily facilitate easy achievement

of these requirements. For example, a March 2007 statement to

investors by Bear Stearns represented that only 6 percent of one

of its hedge funds was invested in subprime mortgages. However,

subsequent examination revealed that it was closer to 60 percent.

Governance imperativesGovernance has many definitions and flavors, which span the

strategic as well as the tactical. It is probably simplest to think of

governance as the way that an enterprise steers itself. This involves

using key conceptual principles to define objectives as well as moni-

toring the performance of processes to ensure that objectives are

being met.

Reporting, or information presentation, is the mechanism that

enables governance. Governance relies on this function to provide

timely and insightful information that allows executives to take

preventative and corrective action so that they can avoid imbal-

ance and tail events. For example, executives from Bear Stearns

and the SEC, which was providing regulatory oversight, failed to

recognize that risk managers at Bear Stearns had little experience

with mortgage-backed securities, where the greatest risk was con-

centrated. Reporting mechanisms were oriented toward capturing

and characterizing transactions and did not appropriately address

competencies and capabilities, thereby creating a knowledge gap.

Defining and facilitating the integrated management of different

risk types should become a primary activity for enterprise gover-

nance.

conclusion: where to go from hereRisk management is not new; most companies already have infra-

structure in place to help execute their risk management strategy.

The good news is that companies do not need to scrap what they

have got. Instead, firms need to enhance and buttress current risk

management infrastructures to drive the right level of enterprise

friction.

The first order of business an organization needs to put in place

to begin reversing negative attitudes about risk management is to

elevate its role. The CEO needs to be involved, along with boards of

directors and senior executives across the lines of business. Senior

management needs to define and then promulgate a shared set of

risk management values across the company. One positive outcome

of the recession is that risk management has greater executive

mindshare, and risk managers need to capitalize on that.

Specific goals for creating a risk management culture include:

Institutionalize risk management. Develop and articulate an ■■

explicit risk management strategy. Establish roles that reflect

the organization’s risk management model.

Determine and define who has ownership over risk management ■■

issues and actions, and who will take on the roles established in

the risk management model.

Nurture a culture of risk awareness and action. Include risk-■■

based metrics in performance scorecards and operate a reward

system that balances risk taking with risk mitigation.

18 – The journal of financial transformation

Be pragmatic. Focus on business needs, such as compliance and ■■

shareholder value. Then attack those needs in bite-size portions

to demonstrate success early and often.

In the end, creative tension between strategy and risk management

should be seen as a positive development in organizations. This

helps to ensure that short-term opportunities are not exploited at

the expense of long-term viability. However, these strategic consid-

erations and risk assessments should be made at the same time,

ensuring a symbiotic interaction between these two perspectives.

In summary, three components are critical to delivering enterprise

friction:

Enterprise risk management should have a strong voice at the ■■

management table and should work in tandem with enterprise

strategy across all enterprise activities.

Formal risk management structures must be buttressed across ■■

data, analytics, reporting, and governance to help the enterprise

achieve the appropriate level of resilience.

Informal networks should be encouraged. These networks can ■■

evolve to fill the white space left uncovered by formal struc-

tures.

19The author is grateful to John Fraser, Chief Risk Officer of Hydro One, for revision of 1

the manuscript before submission.

Enterprise Risk Management — a clarification of some common concernsMadhusudan Acharyya, The Business School, Bournemouth University1

There have been several discussions on the topic of Enterprise

Risk Management (ERM) both in practitioner and academic com-

munities. However, there still remains considerable confusion and

misunderstanding in the ongoing debate on this topic. This may be

because we have focused more on the application of the subject

rather than the causes of the confusion. This article articulates a

different understanding of ERM and provides clarifications of sev-

eral misunderstood issues.

It is understood that risk is universal and it has a holistic effect

on the organization. All organizational functions are exposed to

risk to various extents. In addition, there exist differences in risk

attitude both at individual and organizational levels. One of the

controversial areas of risk management is the classification of

risk. Risk is traditionally classified by professionals depending on

the phenomena they observe in their functions at various levels

of the management hierarchy (i.e., lower-medium-top). From

financial and economic perspectives, risks are often classified as

market (price, interest rate, and exchange risks), liquidity risks

(which involve interaction between the firm and the external

market) and credit risk. From the strategic perspective, risk can

be classified as systemic risk and reputational risk (which involves

policy and decision-making issues at the top management level).

From the operational perspective risk can be classified as fraud

risk and model risk (which involves human and process at the

organizational level). From the legal perspective risk can be clas-

sified as litigation risk, sovereignty risk, regulatory/compliance

risk, and so on. The list of risk categories is so long that it never

stops, thus creating complexities to understand them. There exists

another understanding that attempts to classify the risks of a firm

from the various sources, i.e., both internal and external sources.

Organizations take risks knowingly and unknowingly, and risks are

also produced during the operations of the business. Some of the

risks cause massive failure and some do not. Moreover, some risks

are core to the business and some are ancillary. As such, the list of

significant risks differs extensively from one industry to another.

For a bank the source of core risks are from lending activities (i.e.,

credit risk), for an insurance company the core risks are providing

the cover of insurable risks at the right price and offloading them

through appropriate pooling and diversification. In addition, the

professional expertise of the leader also influences the risk man-

agement priorities of the organization. To some organizations risk

management may be thought of as a matter of common sense but

it might be the subject of particular specialization to some others.

On one hand, risk management is close to the general management

function, with particular attention to risk built on a management

process (i.e., identification, measurement, mitigation, and monitor-

ing). It is a matter of understanding the culture, growing aware-

ness, selecting priorities, and effective communication. On the

other hand, it is a tool to maintain the survival of the organization

from unexpected losses with the features of capturing potential

opportunities. Alternatively, in the latter view, risk is concerned

with the extreme events which are rare but have massive power

of destruction. Statisticians call them lower tail risks with lower

probability and higher severity which provide volatility in value

creation. In sum, some view risk as a danger and to others taking

risks is an opportunity.

The meaning of ERMSo, what does ERM mean? It actually means different things to dif-

ferent people depending on their professional training, nature of

job, type of business, and the objective to be achieved. There exist

two types of definitions of ERM. From the strategic perspective,

ERM is to manage the firm’s activities in order to prevent the failure

of achieving its corporate objectives. From another perspective,

ERM is related to the successful operationalization of the corporate

strategy of the firm in the dynamic market environment, which

requires management of all significant risks in a holistic framework

rather than in silos. It is argued that ERM is not to deal with the risks

that are related to the day-to-day business functions of the firm

and are routinely checked upon by lower or line level management.

Here is the key difference between business risk management and

ERM. It is evident that the practitioners, at least in the financial

industry, are developing ERM to deal with the unusual risks, which

the statisticians called outliers or extreme events. Enterprise risk is

a collection of extraordinary risks that are full of surprising potenti-

ality and can threaten the survival of business. Indeed, overlooking

the less vulnerable areas, which currently might seem less risky,

is dangerous as they may produce severe unexpected risks over a

period of time. Consequently, a comparison and differentiation of

the key significant risks with the less significant risks is an inherent

issue in ERM. A continuous process of selecting key significant risks

and the relevant time frame are essential elements of identification

of enterprise risks. Notwithstanding, whether ERM is a specialized

branch of [corporate] finance or an area of general business man-

agement is still an issue of debate in both practitioner and academic

communities.

The evolution of ERMHow did ERM evolve? In some cases it was the internal initiatives

of the organizations and in some others the motivation was exter-

nal (i.e., regulations). Indeed, the regulators and rating agencies

play an observational (or monitoring) role in the business journey

of financial firms. They are there to maintain the interest of the

20 – The journal of financial transformation

customers and investors. Theoretically, the innovation remains

within the expertise of the firm and most [large] corporations, in

essence, want to get ahead of the regulatory curve to maintain

their superior role in the competitive market. It would be frighten-

ing for the market if the goals of regulators and rating agencies

in terms of risk management are not aligned to the business goals

of the corporations. If not, at least two things could happen —

first, the creation of systemic risk and second, an increase in the

cost of the product that is often charged to the customers. It is

interesting to see if the ERM initiatives of regulators and rating

agencies influence the firm’s internal (or actual) risk manage-

ment functions. Nevertheless, organizations, in particular those

that have large scale global operations, should be given enough

flexibility to promote innovations. Otherwise, the current move-

ment of ERM of financial firms will be limited to the boundary of

regulators and rating agencies’ requirements. On the other hand,

it is true that strict (or new) regulations trigger innovations mainly

in the areas of governance. Interestingly, there appears a positive

indication through the introduction of a principle-based regula-

tory approach.

The uniqueness of ERMUniqueness is an important aspect of ERM. Since risks of each firm

(and industry) are different from another it is important to view

ERM from a specific firm or industry perspective. Take the example

of the insurance business. The key sources of risk of a typical

insurance company are underwriting, investment, and finance/

treasury functions. Similar to other industries, insurers also face

several types of risks, such as financial, operational, strategic, haz-

ard, reputational, etc. However, the risks are traditionally seen in

silos and the view of an underwriter and an actuary on risk is very

different from the investment and finance people. The opposite is

also true. In the banking world the views of a trader on risk is very

different from that of the lenders. The risk profile of an investment

bank and commercial bank will be different. Consequently, some

common questions are emerging in both practitioner and academic

communities in recognizing ERM. Does ERM seek a common view of

risks of all these professionals? Is it necessary? Is it possible? The

short answer is ‘indeed not.’ So, what does a common language of

risk mean? In my view, it means that everybody should have the

capability of assessing the downside risk and upside opportunity of

their functions and making decisions within their position in terms

of the corporate objectives of the firm. This understanding places

the achievement of corporate objectives and strategic decision-

making at the heart of ERM. Alternatively, employees should have

the ability to judge the risk and return of their actions in a proac-

tive fashion, judging the implication of their functions on the entire

organization (i.e., another department/division), and communicat-

ing their concerns across the firm. In practice, a ‘group risk policy,’

which is the best example of such a common risk language, provides

valuable guidance.

Value of ERMThe value proposition of ERM is another disputed area. What is the

objective of ERM? Why should a firm pursue ERM? In reality, the

goal of a firm is to create value for its owners. Consequently, maxi-

mization of profit is the overriding objective of an enterprise, which,

in modern terms is called creation of [long-term] shareholder

value. This is fully aligned with the expectation of the sharehold-

ers of a firm. Within this justification, senior executives are paid

commensurate to their risk taking and managing capabilities (an

issue of agency theory). There remains a lot of speculation as to

the benefits of ERM, such as value creation for the firm, securing

competitive advantage, reducing the cost of financial distress, low-

ering taxes, etc. In addition, there remains analysis of the benefits

of risk management from ex-ante (pre-loss) and ex-post (post-loss)

situations. However, the recent 2007 financial crisis demonstrated

the failure of several large organizations that were believed to have

ERM in place.

Risk ownershipA further unresolved issue of ERM is the risk ownership structure

within the organization. Take the example of a CEO. How does he/

she view risks of the firm? How much risk of the firm does he/she

personally hold? In fact, the CEO is (at the executive level) the

ultimate owner of the risk of the firm of going burst (i.e., survival).

A relevant question for risk ownership is who else, along with the

CEO, owns and constantly monitors the total risk of the firm?

Although, at the upper level it is the board of directors, it might

be too late to get them involved to deal with the risks of the firm

that have already caused irreparable damage to the organization.

In essence, the CEO is the only person who takes the holistic view

of the entire organization. Consequently, it is important to support

ERM by creating risk ownership across the various levels of man-

agement hierarchy within the organization.

Risk appetite and toleranceHow much risk should an organization take? This is often referred

to as risk appetite or level of risk tolerance. This needs a bottom-

up assessment of risk with the integration of several key risks that

exist within the firm. Moreover, it is directly linked to the corporate

objectives of the firm which, in essence, means where the firm

wants to be in a certain period of time in future. Certainly, it is not

limited to tangible risks of the firm, those associated with the capi-

tal market variables, such as asset and liability risks. In essence, it

includes a lot of intangible issues, such as the culture of the firm,

the expertise of the people who drive the business process, the

market where the firm operates, and the future cash flows that the

firm wants to produce. In fact, the appetite for risk is an essential

element of formulating a firm’s corporate strategy. Indeed, it is dan-

gerous to rely solely on statistical models that generate numbers

where the scope for including managerial judgment is limited.

21

challenges of ERMHow can the ERM objectives be achieved? Should we prefer the

mathematical approach? Indeed, a mathematical treatment (or

extrapolation) of risk (i.e., quantitative modeling) is necessary to

transform ideas into actions; but the limitation of mathematics,

as a language, is that it cannot transform all ideas into numerical

equations. This is equally true for the progress of ERM, as it is still

going through the transition period of getting to maturity. Indeed,

risk involves a range of subjective elements (i.e., individual experi-

ence, judgment, emotion, trust, confidence, etc.) and ignoring these

behavioral attributes will likely lead to failure in adoption of ERM.

It is important to remember that beyond mathematical attempts to

theorize risk effects/impacts, risk management is about processes

and systems that involve human understanding and actions. Ideally,

neither approach is singly sufficient to handle the enterprise risk of

the firm. An effective ERM must balance both in a common frame-

work. ERM is truly a multidisciplinary subject.

The role of cRoWho shares the responsibility of the CEO in relation to risk? The

common practice is to have a ‘risk leadership team’ equivalent

to a ‘group risk committee’ comprising of the head of each major

function. Theoretically, this team is supported by both technical

and non-technical persons; hence there could be communication

problems since they might speak with different languages of risk.

Consequently, there should be somebody responsible for coordina-

tion (or facilitation) in order to maintain the proper communication

of risk issues across the organization in terms of the require-

ments of the corporate objectives. This person, at least theoreti-

cally, should be technically familiar with all the subject areas (which

practically appears impossible) but should have the capability to

understand the sources of risks and their potential impact either

solely or with the help of relevant experts. Currently such a role

is emerging and is often called the Chief Risk Officer (CRO). In the

meantime, the presence of CROs has appeared both in the financial

(i.e., banking and insurance) and the non-financial sectors (i.e., com-

modity). Typically, a CRO is responsible for developing ERM with

adequate policy and process for managing risks at all levels of the

firm. In addition to an adequate knowledge of the quantitative side

of risk management, a CRO should have a fair understanding of the

behavioral aspects of risk as he/she has to deal with the human

perceptions and systems (i.e., communication and culture). One of

the challenging jobs of CROs is to report to the board of directors

(most of whose members often do not have a technical knowledge

of the risks) through the CFO or the CEO depending on the specific

structure of the firm. Ideally, the board of directors is, by regula-

tions (e.g., Combined Code in the U.K., Sarbanes Oxley in the U.S.,

and similar regulations in some other countries), responsible for all

risks of the firm. Another big challenge for a CRO is to promote a

risk awareness and ownership culture throughout the firm, includ-

ing the units at the corporate centre and the subsidiaries/divisions

at various geographical locations. The recent Sir Walker’s review in

the U.K. has highlighted the significance of the role of the CRO and

recommended the establishment of a board level risk committee for

banking and other bank-like financial firms (i.e., life insurers).

future of ERMWhere is ERM going? Certainly, globalization influences businesses

to change their business patterns and operational strategies. As

stated earlier, ERM is currently maturing and more significant

developments are likely in future. It is assumed that ERM will move

from its narrow focus of core business risks to take on a more

general perspective. Risk management will gradually be embedded

within firms’ strategic issues. Certainly, risk is an integral part of all

businesses and their success depends on the level of each firm’s

capability for managing risks. However, integrating the two views

of managing risks (i.e., fluctuation of performance in the area of

corporate governance with the volatility in the shareholder value

creation) in a common framework is challenging. Importantly, the

robustness of an ERM program depends on the commitment of the

top management in promoting a strong risk management culture

across the organization. Moreover, an innovative team of people

within the organization with a structured risk-taking ability and

approach is significantly important for the success of ERM.

ReferencesBrian W. N. and R. M. Stulz, 2006, “Enterprise risk management: theory and practice,” •

Journal of Applied Corporate Finance 18, 8-20

Calandro, J., W. Fuessler, and R. Sansone, 2008, “Enterprise risk management – an •

insurance perspective and overview.” Journal of Financial Transformation, 22, 117-122

Dickinson, G., 2001, “Enterprise risk management: its origins and conceptual founda-•

tion,” The Geneva Papers on Risk and Insurance: Issues and Practice, 26, 360-366

Fraser, J. R. S., and B. J. Simkins, 2007, “Ten common misconceptions about enter-•

prise risk management,” Journal of Applied Corporate Finance, 19, 75-81

Mehr, R. I. and B. A. Hedges, 1963, Risk management in the business enterprise, Richard •

D. Irwin, Inc. Homewood, IL

Walker, D., 2009, “A review of corporate governance in U.K. banks and other financial •

industry entities – Final recommendations,” H M Treasury

A reference to Nassim Taleb’s popular book Black Swan. David Cowen and Nassim 1

worked at the same firm in London, Bankers Trust, 16 years ago for a brief period.

Nassim has led the charge against VaR as a risk model. 22

Wall Street Journal, August 2007. http://online.wsj.com/article/2

SB118679281379194803.html. For a more in-depth look at the limitations of VaR see

this recent article: http://www.nytimes.com/2009/01/04/magazine/04riskt.html?_

r=2&ref=magazine&pagewanted=all

capital at risk — a more consistent and intuitive measure of riskDavid J. Cowen, Quasar Capital, and President and CEO, Museum of American Finance David Abuaf

This paper will explain a risk methodology for traders and hedge

funds that trade in the most liquid of markets like G10 futures and

foreign exchange. The methodology is called ‘capital at risk’ (CaR)

and is a replacement for ‘value at risk’ (VaR). CaR obviates the need

to worry about fat tails or outliers called Black Swans as it virtually

eliminates downside surprises1. It is a conservative measure of risk

and focuses on assessing the maximum downside to the portfolio.

The traditional profile of a risk manager who should use CaR is a

short-term trader investing in the most liquid of markets — where

slippage is almost entirely avoidable; however, CaR is by no means

exclusive to short-term traders. In the volatility of 3rd and 4th quar-

ters of 2008 this tool would have been very useful to those with a

medium to longer term trading horizon as well.

problems with traditional risk metricsIn traditional risk management, VaR is used by traders to assess the

probability of a deviation to the portfolio’s return in excess of a spe-

cific value. This measurement, like many others, has flaws. The most

obvious is its basis on past performance — wherein historical volatil-

ity is indicative of future volatility. This flaw leads to two discreet

problems. The first is that it cannot take into account severe market

dislocations that are not reflected in historical data. The second is

that from a practitioner’s standpoint VAR can be completely differ-

ent from one trader to the next due to subjective limitations, i.e., the

time limit utilized or confidence threshold.

In the more liquid markets which short-term traders frequent, VaR is

a risk model that has the potential for catastrophic drawdowns. When

using VaR, only historical returns are factored into future volatility

expectations, and as a result infrequent occurrences are not reflect-

ed (stock market crashes, high commodity demand, terrorist attacks,

etc.). Additionally, when one sees a return indicated at 3σ VaR (99%

confidence level) we can expect the event to occur 2.5x per year,

yet amongst traders it is not uncommon to observe moves of this

magnitude or greater more than 5x per year. And with VaR there is

the required subjective nature of the timeframe used. For instance, a

VaR model that has twenty years of look-back data might seem suf-

ficient; however it would not include the October 1987 market crash.

A five-year model would not have the technology bubble of 2000.

Therefore, there are inherent caution flags when using VaR.

To see the problems of using VaR, one need only to look at the

performance of hedge funds and statistical arbitrage traders dur-

ing the summer of 2007. Many funds lost 20% of their capital in

those months alone. Matthew Rothman, head of Quantitative Equity

Strategies for now defunct Lehman Brothers was quoted in the

Wall Street Journal as saying “[Today] is the type of day people will

remember in quant-land for a very long time. Events that models

predicted would happen once in 10,000 years happened every day

for three days”2. That quote is testimony enough to find a different

measure of risk.

What is caR?When should it be used?

CaR is a measure of risk originally designed to value the maximum

downside to the portfolio without using any assumptions. There are

specific conditions which must exist in order to properly use CaR:

1 Each trade must have a predetermined stop-loss.

Stop-loss levels are continually readjusted for profitable a.

trades to lock-in profits. Consequently, CaR is not a static

number. Additionally, even if the stop-loss level does not

move, because the market is moving, CaR will by definition be

a dynamic number.

Even if two trades have a high degree of correlation they must b.

be treated as separate trades to their stop loss. For instance,

if one was short equivalent Australian dollar against U.S.

dollar with 25 basis points of risk to the portfolio and long

equivalent New Zealand dollar with 25 basis points of risk to

the portfolio, CaR will report 50 basis points. VAR risk models

would look at this as cross trade, long New Zealand dollar

versus short Australian dollar and say that those two currency

pairs are highly correlated, which they are, and then report a

significantly lower risk to the portfolio, say something on the

order of 10 basis points.

2 Trades must be in liquid futures or spot foreign exchange so that

slippage is mitigated.

Emerging markets trades should take caution using CaR a.

methodology for these currencies have substantial gap risk

negating the usefulness of CaR.

3 If options are used then it is calculated at the full premium of the

option no matter what maturity or delta. It is the full cost of the

option.

Therefore CaR can only be used in a long only option based a.

strategy. It has limitations if the risk manager is naked short-

ing options.

These conditions are not just optimal but essential.

How to calculate caR?One of the more fortunate effects of transacting in extremely liq-

uid markets and not relying on historical performance or outside

assumptions is the ease of calculation of CaR:

1 Revalue all cash and futures positions to their stop loss levels, so

if the risk manager was stopped out what would you have lost.

Add up the total cost of all your options based on if revaluation 2

went to zero.

23Washington Post, December 30, 2008. http://www.washingtonpost.com/wp-dyn/con-3

tent/article/2008/12/30/AR2008123003431_5.html?sid=ST2008123003491&s_pos

Add 1 and 2.3

Divide the above amount by the total capital of the portfolio.4

The end result is the maximum amount of loss to the portfolio, 5

CaR.

CaR’s ease of use in calculation can be applied to all trades and is

easily aggregated to specific market or portfolio levels.

CaR is not the most difficult of measures to use or calculate. The

benefit of CaR is that a trader can rest assured that they know their

maximum portfolio loss in the event of a catastrophe. It is always a

worst case scenario. In that manner there are no portfolio surprises

as one is always cognizant of the full risk to the portfolio.

Benefits of caREase of use.■■

Ease of calculation.■■

Easily understandable.■■

Not easily manipulated.■■

Not subject to historical data or assumptions.■■

More intuitive manner in which to assess risks of a trade.■■

Eliminates downside surprise.■■

Example usage of caRThe following illustrates a portfolio of $10,000,000 invested in a

few securities.

AUM 10,000,000

contract Quantity current price stop cAR

Day 1 Feb COMEX gold call 20 13.1 26.2bp

DEC S&P Mini s.1 20 840 823 17bp

DEC S&P Mini s.2 10 840 830 5bp

Total portfolio cAR 48.2bp

Day 2 Feb COMEX gold call 20 14 28bp

DEC S&P Mini s.1 20 866 850 16bp

DEC S&P Mini s.2 10 866 850 8bp

DEC S&P Mini s.3 10 866 845 10.5bp

Total portfolio cAR 62.5bp

The notations s.1, s.2, and s.3 indicate distinct sets of contracts

Here we see that as of the close of day 1 the firm has 20 outstand-

ing Feb gold calls presently priced at 13.1 and 30 outstanding

December S&P mini contracts with different stops. The CaR was

calculated assuming the call would fall from present prices to zero

and that the mini-contracts could fall from 840 to their respective

stops. Each gold contract is worth $100 per point. Consequently,

we multiplied 13.1 x $100 x 20 to reach $26,200. Then we divided

$26,200 by $10,000,000. The mini example is (840-823) = 17. 17 is

multiplied by $50 per mini contract and then by 20, or 17 x $50 x

20 = $17,000. $17,000 is then divided by $10,000,000 to achieve

the 17 basis points of risk.

On day 2, we see the portfolio has more risk associated with the

gold call — this is because the price of the call has risen, so the pos-

sible value lost has increased. In the first two mini contracts, the

stops were rolled upwards, locking in profit but still exposing the

portfolio to a loss from day 2’s NAV. We also see the addition of a

third S&P mini contract, further adding to the portfolio’s CaR.

Real world examples of not knowing your riskIn 2008, we witnessed catastrophic losses at Bear Stearns, Lehman

Brothers, AIG, and a score of other high profile firms. The cost to

the economy and taxpayers has been enormous. To use AIG as but

one example, the U.S. Government has pumped U.S.$152 billion

into AIG in the form of direct loans, investment in preferred stock,

and acquisition of troubled assets. AIG’s exposure was through its

Financial Products Division, which became a major player in the

derivatives market, both in credit default swaps and collateralized

debt obligations. In the case of the credit default swap, in exchange

for fees AIGFP would guarantee, or insure, a firm’s corporate debt

in case of default. It was a growing market and the firm was book-

ing hefty profits. With respect to CDOs, the firm had a portfolio of

structured debt securities that held either secured or unsecured

bonds or loans. By the end of 2005 AIG had almost U.S.$80 billion

of CDOs.

The firm was comfortable with its derivatives portfolio. In a public

forum in August 2007 AIG Financial Products President Joseph

Cassano boasted on a conference call to investors that the deriva-

tives portfolio was secure: “It is hard for us, without being flippant,

to even see a scenario within any kind of realm of reason that would

see us losing $1 in any of those transactions”3. What prompted this

confidence? According to interviews with top officials at AIG they

were relying on a computer model to assess their credit default

swap portfolio. According to their proprietary model, there was only

a 0.15% chance of paying out, or a 99.85% of reaping reward. AIG

believed that there was only a 1.5 chance in 1,000 of disaster. Once

again a model which predicted only a slim chance of an event occur-

ring, the so-called fat tail, sunk a once prestigious firm.

While AIG had credit risk they never faced the reality of the magni-

tude of their risk. Had CaR been used, it would not have presented

a probability, but rather a finite amount of risk. In particular what

hurt AIG was that it had to post collateral against these swaps

and when their AAA rating became imperiled the calls for margin

occurred. Perhaps if AIG had thought in terms of full collateral capi-

tal at risk they would have thought differently about their risk. The

24 – The journal of financial transformation

reality of life is that risk in the financial markets is never 0.15% no

matter what the model states.

And we know that VaR in practice is simply unable to handle the

stress. Bear Stearns reported on May 31, 2007 an interest rate

risk to their portfolio of U.S.$30.5 million using a VaR confidence

level of 95%. Moreover, they claimed diversification benefits from

elsewhere offset the entire firm-wide exposure to only U.S.$28.7

million. When Bear Stearns failed we know that they had levered

their money so high, over thirty times, that a run developed that

could not be met. How does a firm state a risk of less than U.S.$30

million when its true risk is significantly higher? We know that VaR

is part of the problem and not part of the answer.

conclusionCaR has been used by David Cowen to measure risk in his hedge

funds. Over his two decade trading career he was never satisfied

with VaR. David set out to find a simplistic method to value the

maximum downside to the portfolio. The result is CaR. This is an

easy to calculate measure which appropriately factors in all risks to

the portfolio without looking at historical and often erroneous data.

While David originated this idea he does not take sole authorship

for the concept. Others could have easily simultaneously developed

this method as well.

Articles

Economists’ hubris — the case of risk management

Best practices for investment risk management

Lessons from the global financial meltdown of 2008

What can leaders learn from cybernetics? Toward a strategic framework for managing complexity and risk in socio-technical systems

Financial stability, fair value accounting, and procyclicality

Can ARMs’ mortgage servicing portfolios be delta-hedged under gamma constraints?

The time-varying risk of listed private equity

The developing legal risk management environment

Interest rate risk hedging demand under a Gaussian framework

Non-parametric liquidity-adjusted VaR model: a stochastic programming approach

Optimization in financial engineering — an essay on ‘good’ solutions and misplaced exactitude

A VaR too far? The pricing of operational risk

Risk management after the Great Crash

26 – The journal of financial transformation

27The views expressed in this paper reflect only those of the authors and in no way 1

are representative of the views of Capco, Contango Capital Advisors, or any of their

partners

Articles

Economists’ hubris — the case of risk management1

AbstractIn this, the third paper in the Economists’ Hubris series, we high-

light the shortcomings of academic thought in developing models

that can be used by financial institutions to institute effective

enterprise-wide risk management systems and policies. We find

that pretty much all of the models fail when put under intense sci-

entific examinations and that we still have a long way to go before

we can develop models that can indeed be effective. However, we

find that irrespective of the models used, the simple fact that the

current IT and operational infrastructures of banking institutions

does not allow the management to obtain a holistic view of risk

and the silos they sit within means that instituting an effective

enterprise-wide risk management system is as of today nothing

more than a panacea. The main worry is that it is not only academ-

ics who fail to realize this fact, practitioners also believe that these

models work even without having a holistic view of the risks within

their organizations. In fact, we can state that this is the first paper

in which we highlight not only the hubris exhibited by economists

but also the hubris of practitioners who still believe that they are

able to accurately measure and manage the risk of the institutions

they manage, monitor, or regulate.

shahin shojaiGlobal Head of Strategic Research, Capco

George feigerCEO, Contango Capital Advisors

28 – The journal of financial transformation

Economists’ hubris — the case of risk management

In this, our third article in the economists’ hubris series, we look

at the shortcomings of academic thinking in financial risk man-

agement, a very topical subject. In the previous two articles, we

examined whether contributions from the academic community

in the fields of mergers and acquisitions [Shojai (2009)] and asset

pricing [Shojai and Feiger (2009)] were of much practical use to the

practitioners and demonstrated that economists have drifted into

realms of sterile, quasi-mathematical and a priori theorizing instead

of coming to grips with the empirical realities of their subjects. In

this sense, they have stood conventional scientific methodology,

which develops theories to explain facts and tests them by their

ability to predict, on its head. Not surprisingly this behavior has

carried also into the field of risk management, with an added twist.

Rather like the joke about the man who looks for his dropped keys

under the street light because that is where the light is rather than

where he dropped the keys, financial economists have focused on

things that they can ‘quantify’ rather than on things that actually

matter. The latter include both the structure of the financial system

and the behavior of its participants. Consequently, the gap between

academic thinking and business application remains as large today

as it has ever been.

Irrespective of one’s views regarding academic finance, or even the

practioners who are expected to apply the models devised, few can

deny that there were serious failures in risk management at major

global financial institutions, perpetrated in all probability by the

belief that the models developed work and that they can withstand

environments such as the recent financial crisis. However, it is not

enough to simply make such generalized statements without taking

the time to get a better understanding of why such failures took

place and whether we will be able to correct them in the future.

Our opinion is that the tools that are currently at the disposal of

the world’s major global financial institutions are not adequate to

help them prevent such crises in the future and that the current

structure of these institutions makes it literally impossible to avoid

the kind of failures that we have witnessed.

Our objective with this article is not to exonerate the risk manage-

ment divisions of these institutions, nor is it to suggest that the

entire enterprises should be forgiven for the incalculable dam-

age that they have caused. Our aim is to demonstrate that even

if the risk management divisions of these institutions had acted

with the best intentions of their organizations, and even the other

stakeholders, in mind they would have still had a difficult task in

effectively managing the risks within their enterprises given the

tools that were at their disposal and the structures of the firms

they operated in. Furthermore, and in our opinion given the focus of

this paper, more importantly, had the risk management divisions of

these institutions effectively instituted the tools that were provided

to them by academic finance the situation would be no better than

it is today.

The more one delves into the intricacies of academic finance the

more one realizes how little the understanding is among a majority

of academics about what actually takes place within these institu-

tions and how difficult it really is to implement the theories that are

devised in so-called scientific finance institutions at major business

schools in the West.

In this paper, we will focus on a number of these issues. We will high-

light why it is that the current structures within the major financial

institutions make it almost impossible to have a holistic view of the

enterprise’s risk, despite the many different suggestions as to its

viability in academic literature; which we must add is not that many.

We will discuss why it is that the current compensation structures

make it very hard for the management to control risks at the indi-

vidual, divisional, and group level. And, finally, we will explain why it

would still be impossible to prevent such crisis in the future, even if

the two former issues were somehow miraculously solved, given the

tools that are available to risk managers and their management.

However, before we get too deep into the intricacies of financial

institutions and their risk management operations, it is important

to cast a critical eye on what has been suggested to be the main

cause of the recent crisis and discuss why it is that one of the main

causes has been overlooked, namely the disappearance of due

diligence by banks.

What are causes of the recent market crisis?If one were to choose the one area which has borne the most criti-

cism for the current crisis it would be the CDO market, and those that

rate them [Jacobs (2009)]. Of course, the regulators and central

bankers have also not got away unscathed, but most studies seem

to suggest that had we got a better understanding of CDOs and their

risks we might have been able to prevent the current crisis.

The first problem with this point of view is the expectation that com-

plex financial assets, such as CDOs or other securitized assets, can

be accurately valued using scientific methods. Even if we were able

to calculate the risk of simple, vanilla structure, financial assets to

correctly price them, and by that we mean that the pricing models

arrive at the same price that the asset is trading at in the markets,

we would still have a very difficult time pricing securitized assets

with any degree of accuracy. The reason is that the prepayment

rights of these instruments, which are related, with some friction,

to movements in interest rates, would necessitate an absolutely

perfect prediction of interest rate movements into the future, and

a similarly accurate assessment of the proportion of borrowers who

choose to repay [Boudoukh et al. (1997)]. As one can appreciate,

that is quite an unrealistic expectation.

The sad fact is that academic finance has failed in its efforts to

even provide valuation models that can price simple assets, such

29

Economists’ hubris — the case of risk management

as equities, with any degree of accuracy. Expecting these models to

perform any better for highly complex instruments is nothing more

than wishful thinking [Shojai and Feiger (2009)].

If we accept that these assets cannot be priced with any degree of

accuracy then we must accept that neither the financial institutions

that created or traded these assets nor the rating agencies would

have been able to help prevent the crisis that was brought about

by the subprime mortgage market even if they did know what they

were doing.

But, in our opinion, focusing on the pricing of these assets, espe-

cially since all of the people involved with these instruments are

familiar with their intricacies, ensures that we ignore the main

reason why this market became so big and finally imploded. In our

opinion, even if interest rates had not been kept so low for as long

as they were we would still have not been able to prevent the crisis.

And that is because securitization, by its mere existence, creates an

environment that would lead to such crises. The fact that it had not

happened before is the real surprise.

Now, our job is not to provide a primer on securitization or how

the U.S. financial system works for those of our readers who are

significantly more qualified to talk on the subject than we are,

but it would not hurt to just examine how the mere existence of

securitized mortgages can lead to potential crises. If we look at

how mortgages were issued in the past we would find that prior

to the development of the so-called individual risk rating models

[Hand and Blunt (2009), Hand and Yu (2009)] banks used to use

the information that they had about their client of many years to

decide on whether and how much mortgage to issue them with. The

decision and the amount were, therefore, related to the history of

that client’s relationship with the bank.

When banks started lending to nonclients more aggressively they

had to resort to using the information provided by personal risk

ranking organizations, such as Experian. Now, the issue is not

whether these rating agencies provide information that is genuinely

accurate or of much use, though what we do find is that it is not as

remotely scientific in its accuracy as many are lead to believe. The

fact is that the loan was still issued by the bank or its brokers after

doing some sort of due diligence on the client, and more importantly

the loan remained on the books of the bank. The fact that the loan

remained on the bank’s own books ensured that they monitored the

activities of their brokers more closely, since any defaults would

directly cost the bank itself. What securitization does is to allow

the bank to package these mortgages and sell them to third-party

investors. In other words, the loss would no longer hit the bank

but the investors, who in the case of credit enhanced MBSs would

have no idea who the original issuing bank(s) was (were). When this

happens, the bank is no longer a mortgage company but simply

a financial engineer of mortgages. Its income will come not from

the difference between its borrowing rate and the rate at which it

lends, less any loss experienced as a result of defaults, but from

the fees it can charge from manufacturing mortgage pools that are

then sold to investors. As a result it does not spend as much time

on due diligence as it used to do; hence the accusations that banks

were pushing mortgages to those they knew would not be able to

repay them. The fact was that once the mortgages were packaged

non-payment was somebody else’s problem. However, since house

prices were continually going up and interest rates were going in

the opposite direction few would default since they could simply

extract more value from their homes by remortgaging; yet another

source of fees for the banks.

With the banks relinquishing their role as the conductors of due dili-

gence, it was left to the rating agencies to act as sort of auditors for

these issues. But, they never had access to the underlying borrow-

ers to have any idea of their true state of health. That information

was with the issuing bank, and remember that they had stopped

caring about collecting that kind of information when they started

selling the mortgages onto other investors [Keys et al. (2010)]. So,

the rating agencies had to use aggregate data to rate these instru-

ments, or the credit quality of the insurer who had provided credit

enhancement to the security [Fabozzi and Kothari (2007)]. In either

case, neither the credit enhancer nor the rating agency had any

idea about the underlying quality of the borrowers. Consequently,

all that was needed to bring the house of cards down was a correc-

tion in house prices, which is exactly what happened.

Consequently, no matter how we change the regulations govern-

ing rating agencies, unless they admit that they have no idea how

to rate securitized assets, which they really cannot for the afore-

mentioned reasons, such crises are likely to occur again and again.

Given that the number of securitized asset issues dwarf other more

traditional issues that rating agencies used to live off it is highly

unlikely that they will admit to having no idea about how to rate

these instruments, no matter how complex the mathematical model

they use is.

The moral of this overview is that when risk is passed on and com-

pensation is based on the number of sales, due diligence goes out

of the window. It is this lack of due diligence that brought about the

current crisis, and not only in terms of CDOs and MBSs but across

the entire financial ecosystem. From traders who can only win to

asset managers that share in client winnings but do not have to

recoup their losses, or even share in them, when they lose.

Because of the aforementioned issues it is essential that when we

do examine institutional risk management policies we recognize

the importance that human aspects play in their successful imple-

mentation.

30 – The journal of financial transformation

Economists’ hubris — the case of risk management

Having shared with you our perspective on what actually caused

the recent crisis, we will, in the next section, highlight the short-

comings of academic finance in developing models that financial

institutions can use to institute effective enterprise-wide risk man-

agement systems and policies.

A framework for thinking about financial risksIn order to evaluate the relevance of academic thinking on risk

management it is helpful to use a 3-level schema:

Risk at the level of the individual financial instrument.■■

Risk at the level of a financial institution holding diverse instru-■■

ments.

Risk at the level of the system of financial institutions.■■

A financial instrument might be a credit card or a residential mort-

gage or a small business loan. In the U.S., risks in these instruments

have been intensively studied and are moderately predictable in

large pools. One example that has become common parlance is to

equate the default rate on national pools of seasoned credit card

balances to the national unemployment rate. A financial institution

holding a diverse portfolio of such instruments might be a bank

which originates them and retains all or some or an investing insti-

tution like a pension fund or a hedge fund or an insurance company

(or, indeed, an individual investor with a personal portfolio).

The system of financial institutions is the total of these individual

players, in particular, embracing the diverse obligations of each to

the others. A bank may have loaned unneeded overnight cash to

another bank or it may have borrowed to fund its portfolio of trad-

able securities by overnight repurchase agreements; a hedge fund

may have borrowed to leverage a pool of securities; an insurance

company may have guaranteed another institution’s debt, backing

that guarantee by pledging part of its own holding of securities;

and an individual may have guaranteed a bank loan to his small

business by pledging a real estate investment, itself leveraged by

a mortgage.

The fallacy in academic approaches to risk management (enthu-

siastically adopted by the financial institutions themselves) is to

assume that the techniques shown to be reasonably useful for the

analysis of large samples of individual instruments can be of signifi-

cant value in assessing the other levels of risk.

Academic approach to risk management within financial institutionsThe academic approach to handling the risk of a financial institu-

tion holding a diverse pool of instruments is to look at some set of

historic correlations among the instruments and model the institu-

tion as the portfolio of the instruments that it holds. Technically it

assumes that the outcomes of all the instruments are drawn from

a stationary joint probability distribution, from which all sorts of

enticing estimates are possible, such as Enterprise Value at Risk

and others.

Anyone who has attempted to estimate the variance/covariance

matrices of the instruments knows that the distributions are not

stationary. Indeed, they vary markedly according to the sample

period chosen. Hence the statistically trained have used various

weighting schemes to create ‘more relevant’ data. For example,

weighting recent data more heavily than older data using some

distributed lag scheme. This is typical of the economists’ methods:

create some theoretical structure and impose it on the data. A more

productive approach would be to inquire why the distributions are

not stationary.

Without getting into too much detail of the underlying models at

this stage we would like to suggest that at least 3 factors play a role

in causing the nonstationarity of risk distributions which lead to the

practical downfall of the ‘portfolio approach’ to risk management

within an institution:

Failure to attempt to understand ‘causation.’■■

Neglect of the ‘convergence of trading behavior’ in the financial ■■

system we have created.

Adherence to false notions of ‘market efficiency’ which causes ■■

neglect of credit-fueled valuation bubbles.

Let us start with ‘causation.’ Unlike the predictably random move-

ments of electrons in an atom, economic events sometimes have

causes which bear rather sharply on the correlation among instru-

ments. It is easy to give a practical illustration. Some time ago one

of the authors had a client with what was, supposedly, a well-diver-

sified portfolio — equities, bonds, real estate, even some direct busi-

ness interests. Unfortunately, all the investments were located in

or were claims on businesses in Houston, Texas. When the price of

oil collapsed, so did the value of the portfolio. There was a common

causal factor in all the returns. It is not so easy, ex-ante, to calculate

the variance/covariance matrix of returns in Houston conditional

on the price of oil or of returns in the Midwest conditional on the

health of General Motors. Let alone to then create the probability

distribution of the fate of these causal factors.

Consider now ‘convergence of trading behavior.’ Correlations of

returns of assets like European and U.S. and emerging stock indices,

various commodities, and the like have been highly variable but

clearly rising over the past 15 years [Ned Davis research]. A combina-

tion of deregulation of global capital flows, development of sophisti-

cated capital market intermediaries operating globally, and data and

trading technology have enabled more and more money to be placed

on the same bets at the same time. And it is. Once, Harvard and Yale

were unique among institutions in investing heavily in timberland and

private equity and hedge strategies. Now everyone is doing it and the

returns are falling while the volatility is rising.

31

Economists’ hubris — the case of risk management

And finally ‘credit-fueled valuation bubbles.’ Asset markets do not

price efficiently, no matter what your professor said in business

school. Reinhardt and Rogoff (2009) have demonstrated that for

centuries, valuation bubbles funded by excess credit creation have

occurred in all economies for which we have decent records. Why

do investors not simply recognize these and “trade them away,”

as the efficient markets hypothesis would imply? It is not so easy,

because while the bubbles are ‘obvious,’ when they will burst is not.

Smithers (2009) explains this in eloquent terms. However, if you

think about it, had you started shorting the tech bubble aggres-

sively in 1998 you would have been bankrupt well before the end.

As Keynes said, the markets can remain irrational longer than you

can remain solvent.

systemic connections and liquidity riskOnce we acknowledge that valuations are not perfect but can

undershoot and overshoot we can explain systemic risk that is,

essentially, broad loss of liquidity in most assets. Why pay more for

something today than it is likely to be worth tomorrow?

Consider the example of the CLO market in 2008. Several hundred

billion of corporate loans were held in CLOs, on the largely invisible

books of offshore hedge funds and underwriting institutions (via

off-balance-sheet vehicles), in structures leveraged 20 times and

more. The leverage came from major banks and insurance compa-

nies which, we should remember, devoted the bulk of their other

business activities to making loans to entities like businesses and

real estate developers and to each other. They in turn raised their

liabilities by issuing bonds and commercial paper.

Some of the loans in the CLOs started falling in price in secondary

trading (the market making in which, by the way, was provided by

the same banks which were providing the leverage to the CLO hold-

ers). This precipitated margin calls on the holders that they could

not all meet. With leverage of 20 times the fund equity could quickly

disappear so the only recourse was to dump loans and delever-

age as quickly as possible. So we sat at our Bloomberg screens in

amazement as blocks of hundreds of millions or billions of loans

were thrown out for whatever bid they could get. Well, would you

buy then, even if you thought that almost all of the loans were

ultimately money-good, as we indeed did? Of course not, because

in that panic it was certain that the prices would fall further, which

they did.

Normally the market makers would buy bargains but they were

providing the leverage and they were holding trading inventories

that were tumbling in price. So they withdrew the leverage and

reduced the inventories and so forced prices down further. This

killed the hedge funds but also inflicted losses on other holders,

including their own balance sheets, and created credit losses on

the loans extended to the leveraged CLO holders. Now the banks

were in trouble themselves. So they dried up the interbank lend-

ing market, essential for liquidity in the world trading system, and

their commercial paper appeared risky and fell in price, damaging

the money market fund industry which held a large part of liquid

assets in the US.

Similar things happened with mortgage-backed CDOs and other

instruments. We need not elaborate on the history but you can

see why the variance/covariance matrix did not work out to be

relevant. And, consequently, why the core of the academic work

on risk management did not turn out to be relevant. And indeed,

how insidious the concept of market efficiency has been in blinding

market participants to the nature of real risk by implying that it has

already been priced in to all assets as well as it can be.

The role of collateral and incentivesWe have not quite finished with criticism of academic treatment of

risk management. We need to turn now to the two academic nos-

trums for the kind of risk we have described: ‘adequate collateral’

and ‘appropriate incentives.’

Collateral, whether in the hands of a central counterparty or put up

over the counter against individual transactions (the equity cush-

ion in all those leveraged CLOs) is claimed to allow the system to

absorb unanticipated shocks. That may be, but the question is, how

much collateral is enough? Here we come back to the stationary

joint probability distribution of asset prices, which defines the likely

magnitude of these unanticipated shocks, and we can start over.

On reflection, systemic risk renders collateral least helpful when

you need it most. Think of something as simple as a mortgage on

a house. The collateral is the down payment, which is the lender’s

cushion against default. If your house goes on the market in ‘normal

times,’ say because of a divorce, all will be fine and the lender is

likely to recoup his loan. If a real estate bubble has collapsed and

every house on your block is for sale, the cushion is non-existent.

This is not an easy problem to solve. Think dynamic collateral

requirements driven by rules on market value versus ‘true value.’

Let us now turn to incentives, particularly the notion that if incen-

tives are paid in deferred common stock of the originators, there

will be a much lower likelihood of people making ‘bad trades.’ We

believe that such attempts will fail the test of practical reality, for

at least the three following reasons.

They are incompatible with a free market for talent. Institutions ■■

that attempt long deferral of incentive payments into restricted

vehicles will experience loss of their highest performers to insti-

tutions that do not do this. You can see this happening right now

on Wall Street.

The way in which we choose to measure corporate performance ■■

encourages the opposite, namely short-term risk taking for near-

32 – The journal of financial transformation

Economists’ hubris — the case of risk management

term results. We measure results quarterly and annually even

though the economic cycle takes place over years, not quarters.

Who doubts that any financial institution CEO who dropped out

of playing in late 2006 would have long lost his job and his key

staff, well before the markets collapsed? Is this not the meaning

of the infamous remark by Chuck Prince, former CEO of Citi that

“as long as the music is playing we have to dance.”

The proposed holding periods for the restricted incentive pay-■■

ments are always some calendar interval. The economy does not

oblige by revealing the quality of systemic risk provoking actions

like credit bubbles over short periods.

Having highlighted what we believe to have been the main causes

of the recent financial crisis and some of the failures of academic

finance in the area of risk management for financial institutions,

in the rest of the paper we will highlight the shortcomings of VaR

and its offsprings and highlight why it is that irrespective of the

effectiveness of the models, enterprise-wide risk management will

remain a panacea until internal infrastructures of financial institu-

tions are modified to eliminate many of the problems that exist

today.

VaR and its shortcomingsUnlike what many who study the subject of risk might believe there

are very few studies dedicated to the institution of effective risk

management systems and policies within financial institutions. In

fact, had it not been for Value-at-Risk (VaR) [Jorion (2006)], and

its numerous offsprings [Kuester et al. (2006)], we would not have

had much to refer to in this article. The reality is that as of today we

are still forced to refer back to the contributions of Harry Markowitz

in the early 1950s [Markowitz (1952)] towards diversification ben-

efits of less than perfect assets within a portfolio when trying to

determine both risk tolerance and appetite of financial firms (as we

discussed in the previous section).

VaR at its most basic tries to measure of the maximum loss that a

firm is willing to accept within a prespecified parameter, which in

most cases is the simple confidence interval over a set time hori-

zon. It is used to determine what the firm’s appetite is for risk. A

simple example that most use is that if we use a 95% confidence

interval and the firm’s portfolio of investments has a one-day 5%

VaR of $100 million, there is a 5% probability that the portfolio

will fall in value by more than $100 million over a one-day period.

Another way of stating this is that a loss of $100 million or more

on this portfolio is expected on 1 day in 20. However, VaR assumes

that markets are normal and there is no trading, i.e., no change in

the portfolio composition.

In reality the composition of the firm’s portfolio is changing every

second, with every trade that the traders make or every new instru-

ment that is engineered. In fact, for complex instruments the expo-

sure could change during the day without any additional trades or

changes in portfolio composition.

The assumption that the bank’s entire portfolio of assets follows

a normal distribution is also not very valid. Different institutions

have different portfolio compositions and based on that the shape

of their risk profile could be very different from their peers. Using

the incorrect distribution could result in arriving at risk appetite

levels that are completely inappropriate for the firm in question.

In fact, even if one starts off with the right distribution it does not

mean that they will end with the correct distribution, as distribu-

tions could change over time, such as the subprime mortgages that

became more risky with different distribution profiles over time.

The fact that different assets have different distributions means

that their risks cannot simply be added to each other to arrive

at a group risk figure. Given the enormous difficulties that fund

management companies, and their consultants, face when trying to

compute the overall risk of a portfolio of shares that follow similar

distributions it should come at no surprise just how difficult it would

be if one were to try to calculate the overall risk of a financial

institution.

More importantly, the relationships between the different assets

within the portfolio could be, and most probably are, miscalculated,

even if we assume that historical patterns will hold into the future.

How would the model account for liquidity risk when recalibrating

the overall VaR [Fragnière et al. (2010)]? What happens to the cor-

relations during a crisis? If we have learned nothing else from the

recent crisis we have learned that correlations among assets that

were previously unrelated become significantly stronger during cri-

ses. If the correlations merge we would lose some of the benefits of

diversification that we were relying on, pushing VaR figures against

the wall. And sadly, the fact is that historical patterns rarely hold in

the future. They are of little use in correctly determining VaR, espe-

cially since no future crisis will exhibit the patterns of behavior that

its predecessors did. Hence, it is literally impossible to determine

the true number of bad events that fall outside our predetermined

confidence interval.

Even if you were able to accurately determine the number of events

that will fall beyond your accepted confidence level, VaR will not

help you determine their magnitude. The problem is that VaR looks

at the number of events that might fall beyond the confidence

interval we have established and not their magnitude. What is the

point of knowing that there is a 1 in 20 chance of losing more than

$100 million when because of that single event the bank could go

bankrupt? The recent events have proved that a one-off outlier

could bring the entire system down.

VaR also fails to account for the interrelationships between finan-

33

Economists’ hubris — the case of risk management

cial institutions, which in recent years, thanks to the CDO and MBS

markets, has increased dramatically. The fact is that many banks

had used MBSs as collateral with the Fed and when the crisis hit

their values plummeted, preventing the rebalancing needed in light

of higher home loan borrower default rates. As a result, banks were

left with just one alternative when default rates shot up, decreasing

credit availability by lending less and calling-in outstanding loans

[Hunter (2009)]. The implications were that the interbank market

dried up and lending to businesses disappeared, which had second-

ary and tertiary impacts on different assets of other institutions.

The reality of the situation is that VaR, and all other risk manage-

ment methodologies, are relying on data that is backward looking,

is incomplete, and fails to account for the important events that

financial institutions really need to look out for. These risk manage-

ment tools behave like the ratings that ratings agencies issue. It

really does not matter whether a firm has a BBB rating rather than

AAA rating if neither experience default. All that happens is that

the investors in the BBB issue earn a higher return for the same

risk. The differences in ratings matter when the BBB firm defaults

while the AAA firm does not, and we have seen that the granularity

that investors expect simply does not exist. All borrowers become

equally highly risky when the situation turns sour. Citigroup is a

perfect example of an institution that went from being a superstar

to a destroyed entity.

Most importantly, all risk measurements are far more subjective than

many want to accept. They are based on individual decisions about

what should be incorporated into the model and what should not.

The problem sadly is that even if the methodologies that were

provided were reliable, which they are far from being, the current

structures of most financial institutions make instituting effective

risk management policies literally impossible. And it is these institu-

tional intricacies that we will turn to in the next section.

Why risk management is a panacea in today’s financial environmentEven if we choose to ignore the fact that different instruments have

different risk profiles and follow disparate distributions, we cannot

ignore the fact that they are based within different silos of the

institution that is trying to manage its risk.

These divisions have different management styles, compensation

structures, and risk appetites. More importantly, in order to deter-

mine the firm’s overall risk appetite we need to be able to firstly

place each of the risks within their correct buckets, such as credit,

market, operational, etc. The reality of the situation is that most

institutions have major difficulties in quantifying the risks of their

credit and market instruments, let alone the firm’s operational or

liquidity risks.

Most institutions have only recently started to come to grips with

the fact that operational risk is a major risk and that it must be

managed. Learning how to quantify it will take years, if not decades.

Once that is done, we need to be able to compute the operational

risks of each division. Many institutions still segregate their differ-

ent business, hence it is literally impossible to quantify operational

risk for the group and determine what diversification benefits could

be derived. For example, many institutions combine credit and FX

instruments within the same divisions, others keep them separate.

Some combine complex instruments with the fixed income division,

others with equities. In some firms FX, credit, and equities each

have their own quants teams, whose risks no one can understand.

As a result, the firm will have a view on the risk of each silo, but

will not be able to aggregate them in any useful way for the group.

Similarly, since they are sitting in different parts of the business it

is difficult to accurately determine the correlation benefits that the

firm is experiencing in a meaningful way.

The other problem is that there is just too much data to deal with,

even if we are able to aggregate them. The risk management teams

are facing a hosepipe of data which they have to decipher, and its

contents change with every trade and even by a simple change in

the hour of the day.

Even if so much data could be effectively analyzed the next tough

task is to present them in a useful way to the management, since

it is they who determine the firm’s overall risk appetite and not the

risk management division. Once the management sets the param-

eters then it is the risk management team’s job to ensure that

everyone operates within the set boundaries.

These kinds of complex reports are very hard for the manage-

ment to understand. And the simpler they are made the greater

sacrifices have to be made about specificities of risk. The result is

that either the firm does not allocate adequate capital to the risks

it faces or it allocates too much, hence limiting some of the firm’s

profit potential.

Even if the firm was fully intent on getting all the available informa-

tion and it was able to correctly label all of the risks it faces and

place them in the correct buckets, IT and human aspirations will still

make their task impossible.

implications of iT and human greedWhile academic literature does account for personal greed, and

a number of papers have been written on the subject by Michael

Jensen and his many colleagues and students of the subject

[Jensen and Meckling (1976)], it is usually overlooked when it

comes to subjects that deal with anything other than corporate

finance. For some reason, there is an assumption that just like

34 – The journal of financial transformation

Economists’ hubris — the case of risk management

Shojai, S., 2009, “Why the rush to repay TARP?” Riskcenter, May 121

investors all behave rationally individuals are also mostly honest

and as a result models should work.

The fact is that any risk model is as reliable as the data that is

inputted into it, and financial executives are inclined to reduce the

allocation of risk to their activities insofar as possible so that they

can take on greater risks. After all, their compensation is based

purely on the returns they generate. They have an option on the

bank. They share in the gains and not the losses. Consequently, the

higher the risk, the higher the return, with a floor attached.

Given the complexity of many of the instruments that financial

institutions are dealing with it is only natural that they will be col-

lecting underestimated risk profiles from the traders and engineers.

Furthermore, because of the silo structures many firms will find

that they have assets that are highly correlated with each other but

sitting in different silos. As a result, during the downturn the firm

might find itself significantly more exposed than it thought and not

as capitalized as necessary.

The desire to underestimate risk is not limited to the traders or

engineers sitting on the quants desks, the firm’s own management

are compensated on their annual performance, if not quarterly.

Hence they would also like to be able to take more risks than they

should be permitted to. And none of these people are stupid. The

traders know that they will be bailed out by the bank if the situation

goes sour and at worst lose their jobs, but if the bet pays off they

will be able to retire with savings that even their grandchildren can-

not spend. And, the management knows that if the bank goes close

to failure it will be bailed out by the government if it is important

enough to the local or global economy. The moral hazard problem,

as some had predicted1, has become severely worse since banks are

now even more confident of taking big risks.

Sadly, however, there is nothing that can be done about that. And

the different Basle concordats will in no way help, since they are

based on parameters that have no connection with the real world.

They are arbitrary numbers plucked from the air. BIS is today as

effective in preventing financial risk as the U.N. is in preventing

global conflicts.

In addition to the human aspects of risk management, most firms

are unable to get a holistic view of their risk due to the fact that

their IT systems are simply unable to provide that kind of data. Most

financial institutions are unable to determine the actual cost of

the instruments they develop let alone their risk, and how it might

change over time.

In today’s world very few financial institutions have not undergone

some sort of a merger or acquisition, with the pace increasing in the

past couple of years. The result is a spaghetti like infrastructure of

systems that are simply unable to communicate with one another

[Dizdarevic and Shojai (2004)]. Incompatible systems make the

development of a viable enterprise-wide risk management close to

impossible, since without accurate and timely data the firm will be ill

prepared to respond to changes in asset compositions and prices.

And, sadly that situation will remain for many years to come.

Despite the huge annual investments made by major financial insti-

tutions, different systems supporting different instruments are sim-

ply unable to share information in any meaningful way. This means

that no matter how remarkable the models used are, the data that

they will be dealing with are incomplete.

Consequently, despite the best of ambitions, the dream of having

a reliable and effective enterprise-wide risk management shall

remain just that, a dream. And as long as firms remain incapable

of determining their own exposures, irrespective of which of the

reasons mentioned above is the main culprit, they will continue to

face enormous risks at times when markets do correct themselves

rapidly and sadly the precedent set in the recent crisis does noth-

ing but put flame to the fire. Financial institutions are now even

more confident of taking risks than they were two years ago, which

means that the next crisis can only be greater than the one we just

lived through.

conclusionIn this, the third article in the Economists’ Hubris series, we turn our

attention to not only the hubris of economists but also the hubris

of practitioners, be they bankers or regulators. We find that while

a number of models have been developed to help financial institu-

tions manage their risk, none is really that reliable when placed

under strict scientific examinations. Similar to other economic

disciplines, risk management is also prone to being effective solely

within the confines of the academic institutions in which they are

developed.

However, the models are not our only problem. In order to institute

effective risk management systems and policies at financial institu-

tions we first need to be able to collect reliable data, which given

today’s operational and IT infrastructures is literally impossible to

do. Risk data are sitting in a number of silos across the globe with-

out any viable way of aggregating them in any useful way. It is this

issue that needs to be dealt with first before we work on developing

new models that can effectively manage them.

The sad fact is that just like the academics who develop these mod-

els, the practitioners who use them also assume that they work.

Unlike asset pricing where there is a clear gap between academic

thinking and business application, when it comes to enterprise-wide

risk management, both are equally wrong. Consequently, practitio-

ners, both bankers and their regulators, were living under a false

35

Economists’ hubris — the case of risk management

sense of security that was shattered with the recent financial crisis.

The current governmental and regulatory responses are focusing

on the periphery. The real issue is that banks should be forced

to improve their operational and IT infrastructure in order to be

able to get a holistic view of the risks that are sitting within the

many pockets of their organizations. Our aim with this paper is to

highlight the main issues that financial institutions face in institut-

ing effective risk management systems and policies so that public

policy debates get diverted to focus on those issues that truly mat-

ter and not those that simply get press attention.

ReferencesBoudoukh, J., M. Richardson, and R. Stanton, 1997, “Pricing mortgage-backed securities •

in a mulitfactor interest rate environment: a multivariate density estimation approach,”

Review of Financial Studies, 10, 405-446

Dizdarevic, P., and S. Shojai, 2004, “Integrated data architecture – the end game,” •

Journal of Financial Transformation, 11, 62-65

Fabozzi, F. J., and V. Kothari, 2007, Securitization the tool of financial transformation,” •

Journal of Financial Transformation, 20, 33-45

Fragnière, E., J. Gondzio, N. S. Tuchschmid, and Q. Zhang, forthcoming 2010, “Non-•

parametric liquidity-adjusted VaR model: a stochastic programming approach”, Journal

of Financial Transformation

Hand, D. J., and G. Blunt, 2009, “Estimating the iceberg: how much fraud is there in the •

U.K.,” Journal of Financial Transformation, 25, 19-29

Hand, D. J., and K. Yu, 2009, “Justifying adverse actions with new scorecard technolo-•

gies,” Journal of Financial Transformation, 26, 13-17

Hunter, G. W., 2009, “Anatomy of the 2008 financial crisis: an economic analysis post-•

mortem,” Journal of Financial Transformation, 27, 45-48

Jacobs, B. I., 2009, “Tumbling tower of Babel: subprime securitization and the credit •

crisis,” Financial Analysts Journal, 66:2, 17 – 31

Jensen. M. C., and W. H. Meckling, 1976, “Theory of the firm: managerial behavior, •

agency costs and ownership structure,” Journal of Financial Economics, 3:4, 305-360

Jorion, P., 2006, Value at Risk: the new benchmark for managing financial risk, 3rd ed. •

McGraw-Hill

Keys, B. J., T. Mukherjee, A. Seru, and V. Vig, 2010, “Did securitization lead to lax •

screening? Evidence from subprime loans,” Quarterly Journal of Economics, forthcom-

ing

Kuester, K., S. Mittnik, and M. S. Paolella, 2006, “Value-at-Risk prediction: a comparison •

of alternative strategies,” Journal of Financial Econometrics, 4:1, 53-89

Markowitz, H., 1952, “Portfolio selection,” Journal of Finance, 7:1, 77-91•

Reinhart, C. M., and K. Rogoff, 2009, This time is different: eight centuries of financial •

folly, Princeton University Press

Shojai, S., 2009, “Economists’ hubris – the case of mergers and acquisitions,” Journal •

of Financial Transformation, 26, 4-12

Shojai, S., and G. Feiger, 2009, “Economists’ hubris: the case of asset pricing,” Journal •

of Financial Transformation, 27, 9-13

Smithers, A., 2009, Wall Street revalued: imperfect markets and inept central bankers, •

John Wiley & Sons

37

Articles

Best practices for investment risk management

Jennifer BenderVice President, MSCI Barra

frank nielsenExecutive Director, MSCI Barra

AbstractA successful investment process requires a risk management

structure that addresses multiple aspects of risk. In this paper, we

lay out a best practices framework that rests on three pillars: risk

measurement, risk monitoring, and risk-adjusted investment man-

agement. All three are critical. Risk measurement means using the

right tools accurately to quantify risk from various perspectives.

Risk monitoring means tracking the output from the tools and flag-

ging anomalies on a regular and timely basis. Risk-adjusted invest-

ment management (RAIM) uses the information from measurement

and monitoring to align the portfolio with expectations and risk

tolerance.

38 – The journal of financial transformation

Best practices for investment risk management

The last 18 months have brought risk management to the forefront

and highlighted the need for guidance on best practices for inves-

tors. Many institutional investors were surprised by the violent mar-

ket moves during the current crisis. Some have argued that current

risk management practices failed when they were needed most, and

with multi-sigma events extending across formerly uncorrelated

asset classes, investors have questioned the very meaning of the

term ‘well diversified portfolio.’ What does sound risk management

mean for plans, foundations, endowments, and other institutional

investors? How should these institutions think about best practices

in risk management? We start with three guiding principles:

Risk management is not limited to the risk manager; any-1

one involved in the investment process, from the cio to

the portfolio managers, should be thinking about risk — risk

management should not be limited to an after-the-fact reporting

function but must be woven into the investor’s decision-making

process, whether it is the asset allocation decision or the process

for hiring managers. Those responsible for asset allocation and

management should be risk managers at heart and consider risk

and return tradeoffs before making investment decisions.

if you cannot assess the risk of an asset, maybe you should 2

not invest in it — for those institutions invested in alternative

asset classes, such as private equity and hedge funds, or who

have exposure to complex instruments, such as derivatives and

structured products, the risk management requirements have

greatly increased. These investors need a framework for manag-

ing risk that far exceeds what was needed for the plain vanilla

stock and bond investing that prevailed only ten years ago. We

argue that one should assess one’s risk management capabilities

before making the decision to invest in certain asset types.

proactive risk management is better than reactive risk 3

management — being prepared for unlikely events is perhaps

the most important lesson learned from the recent crisis. This

applies to both market and nonmarket risks such as counter-

party, operational, leverage, and liquidity. Addressing this issue

transcends the simple use of the output of models and tools. It

requires an institutional mindset that analyzes the global eco-

nomic outlook, understands the aggregate portfolio exposures

across asset classes, and is willing to use the model output intel-

ligently to align the portfolio structure with the plan sponsor’s

assessment of the risks that may impact the portfolio.

In this paper, we propose a risk management framework that:

Is aligned with the investment objectives and investment hori-■■

zon.

Tackles multiple aspects of risk and is not limited to a single ■■

measure like tracking error or Value at Risk (VaR).

Measures, monitors, and manages exposures to economic and ■■

fundamental drivers of risk and return across asset classes to

avoid overexposures to any one risk factor.

Manages risk for normal times but is cognizant of and aims to be ■■

prepared for extreme events.

We developed this framework with institutional investors and

their risk management challenges in mind, but this framework

can be adapted easily to the requirements of asset managers. In

the first section of the paper, we describe a framework that takes

into account the three guiding principles. In the second section of

the paper, we illustrate this framework in more detail and provide

examples for its implementation.

Three pillars for risk managementRisk management has evolved rapidly over the last few decades,

marked by key developments like the adoption of the 1988 Basel

Accord and significant episodes like the U.S. Savings and Loan crisis,

the collapse of LTCM, the ‘Dot.com’ bust, and the recent financial

crisis. However, the degree to which various risk methodologies

and practices have been implemented by institutional investors has

varied. In particular, there remains a wide range in how market par-

ticipants (pension plans, endowments, asset managers, hedge funds,

investment banks, etc.) have integrated the risk management func-

tion. Best and Reeves (2008), for example, highlight the divergence

in risk management practices between buy-side and sell-side institu-

tions, the latter being subject to greater regulatory pressure.

Our goal is to establish a framework for sound market risk man-

agement for institutional investors. We rely on three pillars:

risk measurement, monitoring, and management (or risk-adjusted

investment management, RAIM). Risk measurement refers to the

tools institutional investors use to measure risk. Risk monitoring

focuses on the process of evaluating changes in portfolio risk over

time. RAIM refers to how investors may adjust their portfolios in

response to expected changes in risk. Robust risk management

integrates all three areas.

The risk manager’s toolkit may include a variety of measures captur-

ing different views of risk. Figure 1 illustrates one way of categorizing

the suite of tools needed. We distinguish between risk measures for

Alpha (active risk) Beta (total risk)

normal Extreme normal Extreme

Tracking error Stress testing active bets

Asset class volatility/beta

Stress testing asset classes

Contribution to tracking error

Active exposures

Active contribution to tail risk

Contributions to total risk

Total contribution to tail risk

Benchmark misfits Maximum active drawdown

Sources of return/exposures

Maximum drawdown, contagion effect

Figure 1 – Structure for risk measurement and monitoring

39

Best practices for investment risk management

assets, both equity and fixed income, together with commodities, hedge funds, and

currencies, are then combined into a single risk model. This makes it suitable for a

wide range of investment purposes, from conducting an in-depth analysis of a single-

country portfolio to understanding the risk profile of a broad set of international

investments across several asset classes.

VaR captures the expected loss at some threshold, while Expected Shortfall captures 3

the expected loss once that threshold has been exceeded. Maximum drawdown is

defined as the largest drop from a peak to a bottom in a certain period.

Performance attribution, the attribution of realized returns to a set of exposures 1

times factor returns, can provide valuable insight to risk managers seeking to iden-

tify where their investments or managers added value. In addition, it can highlight

similarities between asset groups or managers’ strategies in a way that is far more

informative than looking at historical inter-manager correlations alone.

The Barra Integrated Model (BIM) is such a multi-asset class model for forecasting 2

the asset- and portfolio-level risk of global multi-asset class portfolios or plans. The

model begins with a detailed analysis of individual assets from 56 equity markets and

46 fixed income markets to uncover the factors that drive their risk and return. The

normal and extreme times as well as risk measures that relate to

absolute losses or losses relative to a benchmark. On one hand, insti-

tutional investors need to manage the total risk of their investments,

which means protecting themselves from asset-liability deficits,

declines in broad asset classes, and more generally, any losses large

enough to make it difficult to meet the investor’s obligations. On the

other hand, institutions need to manage the risk of managers under-

performing their benchmarks, which involves monitoring the tracking

error and performance relative to the assigned benchmark.

To assess future risks, it is essential to measure and monitor risk

both at the aggregate level and at the factor level. For risk mea-

surement, most institutional investors measure aggregate portfolio

risk with volatility or tracking error, which rely on individual volatili-

ties and correlations of asset classes and managers. However, while

volatility, tracking error, and correlations capture the overall risk of

the portfolio, they do not distinguish between the sources of risk,

which may include market risk, sector risk, credit risk, and interest

rate risk, to name a few. For instance, energy stocks are likely to

be sensitive to oil prices, and BBB corporate bonds are likely to

be sensitive to credit spreads. Sources of risk, or factors, reflect

the systematic risks investors are actually rewarded for bearing

and are often obscured at the asset class level [Kneafsey (2009)].

Institutional investors can decompose portfolio risk using a factor

model to understand how much return and risk from different asset

classes or managers resulted from prescribed factor exposures in

the past1, or how much risk to expect going forward.2

Risk monitoring enables institutions to monitor changes in the

sources of risk on a regular and timely basis. For instance, many

well diversified U.S. plans saw a growing exposure to financial sec-

tor, housing, and credit risk from 2005-2006. While risk managers

may not have foreseen a looming correction, the ability to monitor

these exposures would have at least alerted them to the risks in the

event of a correction.

Portfolio decomposition plays an important role in stress testing.

Here, the sources of risk are stressed by the risk manager to assess

the impact on the portfolio. Stress testing is flexible in enabling risk

managers to gauge the impact of an event on the portfolio. The

stress scenario might be real or hypothetical, commonplace or rare,

but stress tests are used typically to assess the impact of large and

rare events. Scenarios can come in different flavors, such as macro

shocks, market shocks, or factor shocks. The intuition behind stress

testing for market risk can be applied to nonmarket or systemic

risks, such as leverage and liquidity risk. When leverage and liquid-

ity shocks occur, as in 2008, it may result in unexpected increases

in investment commitments for which no immediate funding source

is available. While largely unpredictable, the impact of such shocks

can be analyzed using stress tests. Below we list a number of stress

test categories that investors might employ on a regular basis to

assess the immediate impact on the portfolio as well as the change

in impact over time.

systemic shock

Liquidity shock■■

Leverage shock■■

Macro shock

Interest rate shock■■

Oil price shock■■

Market-wide shock

Market-wide decline in equity prices■■

Targeted shock

U.S. value stocks hit■■

Japan growth stocks hit■■

Our discussion of stress testing segues naturally into the problem

of managing tail risk, or the risk of some rare event occurring.

Whereas stress tests do not address the likelihood of extreme

shocks occurring, other methods for analyzing tail risk do. This

recent period of turmoil has acutely highlighted both the impor-

tance of managing tail risk and the inadequacy of generic tail risk

measures, such as parametric VaR.

While the simplest measure of parametric VaR assumes that

returns are normally distributed, more sophisticated methods

for calculating VaR do not. These span a wide range of modeling

choices that may rely on parametrically or non-parametrically

specified non-normal distributions, or Extreme Value Theory. For

a more detailed discussion on the latter, we refer to Barbieri et al.

(2009) and Goldberg et al. (2009). Other measures of tail risk, such

as Expected Shortfall (conditional VaR) and Maximum Drawdown,3

seek to capture a different and potentially more relevant facet of

tail risk. In general, turbulent times highlight the need for moni-

toring appropriate tail risk measures. Such times also call for the

frequent reporting of exceptional developments, i.e., reporting that

highlights unusual changes in risk measures or increases in expo-

sure to certain factors.

Before we move on to the third pillar, RAIM, it is important to point

out that risk monitoring requires the necessary IT and infrastruc-

ture resources for support. First, accurate data is essential, as is

sufficient coverage of the assets held in the portfolio. Delays in a

40 – The journal of financial transformation

Best practices for investment risk management

risk manager’s ability to view changes in holdings, prices, or char-

acteristics are often caused by infrastructure limitations. In some

cases, data may not be readily available, or the resources required

to collect data from custodians or individual managers may be

prohibitively expensive. In addition, hard-to-model assets, such as

complex derivatives, hedge funds, and private equity, can pose a

challenge for even the most advanced systems. In sum, institutions

should consider the costs of implementing the necessary risk man-

agement systems when they decide in which assets to invest.

One consequence of the current crisis may be that investors

become more cautious when they choose their investments.

Warren Buffett, for example, commented at his recent shareholder

meeting on complex calculations used to value purchases: “If you

need to use a computer or a calculator to make the calculation, you

shouldn’t buy it.” Even though that statement may be extreme, the

point is well taken. The damage that exotic, illiquid, and hard-to-val-

ue instruments have triggered over the last 18 months highlighted

the need to be able to assess the risks of such investments before

money is allocated to them.

The third pillar in our framework is RAIM, which puts risk measure-

ment and monitoring outputs into action. While risk measurement

provides the measures, and risk monitoring ensures that the

measures are timely and relevant, without the ability to make

adjustments to the portfolio, this information is of limited value

for protecting the investor against losses. RAIM aligns the invest-

ment decision-making process with the risk management function.

For instance, RAIM might be used to make portfolio adjustments

as either the correlations between assets or managers rise or the

probability of certain tail risk or disaster scenarios increases. RAIM

could also facilitate the management of risks coming from certain

sources of return, or it could aid in better diversifying the portfolio.

Specifically, RAIM could be used in the development of overlay

strategies that would facilitate certain hedges, such as currency

hedges, or tail risk insurance.

As an example, the declines in the broad equity market last year

caused many pension plans to become underfunded. Decision-

makers may decide that their tolerance for losses should be limited

to a specific percentage. They should then decide whether that limit

should be maintained through a passive hedge or through a trigger

mechanism defined by the breach of clearly defined parameters of

a risk measure. Some pension plans started hedging their equity

exposure to limit downside risk, though for many it was too late.

One reason why pension plans may not have hedged their market

exposure more frequently is the cost of hedging. Hedging reduces

the performance of the portfolio in up markets, but in periods when

the market declines, hedging limits the downside. Figure 2 illus-

trates a successful market hedge that includes a stop-loss plan at a

point when assets drop below a specified level.

All three pillars — risk measurement, risk monitoring, and RAIM — are

indispensable to a complete risk management structure. Figure 3

summarizes the three pillars, illustrated with specific examples.

The Figure uses the same idea we presented before, namely, that

risk measures can be categorized by normal and extreme times and

relative versus absolute investment objectives.

implementing a market risk management frameworkIn practice, the needs of institutional investors can be wide ranging,

and their ideal measurement, monitoring, and managing capabili-

ties will differ. In this section, we illustrate the case of a hypotheti-

cal but typical U.S. plan sponsor. Although there may be additional

criteria, the three critical drivers of risk management requirements

are as follows:

1 Return requirements — the plan’s liabilities or expected pay-

outs will influence not only the assets in which it invests but

also which benchmarks are used and how much it can lose over

60%

40%

20%

0%

-20%

-40%

-60%

-80%

-100%

Cumulative return

Plan bears cost of insurance during

normal markets but benefits from

large unexpected drops due to

systemic blow-ups

Portfolio

insurance

takes effect

Dec-0

8

Feb-0

9

Oct-0

8

Aug-08

Jun-08

Apr-08

Feb-0

8

Dec-0

7

Oct-0

7

Aug-07

Jun-07

Apr-07

Feb-0

7

Dec-0

6

Uninsured portfolio Insured portfolio

Figure 2 – Risk-adjusted investment management to protect against downside risk

Risk measurement Risk monitoring RAiM

nor

mal

Total Volatility Monitor sources of volatility

Limit exposure to biggest sources of volatility

Active Tracking error Monitor sources of tracking error

Limit exposure to biggest sources of tracking error

Extr

eme

Total Stress tests/tail risk measures

Monitor expected shortfall of the total plan

Implement portfolio insurance plan

Active Stress tests/tail risk measures

Monitor changes in potential active losses if market declines by X%

Ask managers to limit exposures to certain sources of risk

Figure 3 – Three pillars of risk management

41

Best practices for investment risk management

certain periods. The latter, in turn, may drive how much risk it is

willing to take and with how much exposure to certain sources of

return/risk it is comfortable.

2 investment horizon — the plan’s investment horizon, or willing-

ness to sustain shorter-term shocks, will influence which risk

measures are appropriate and how frequently they need to be

monitored.

3 complexity of investments — plans that invest in difficult-to-

value assets with potentially non-normal return distributions

or unusually high exposure to tail events require additional risk

measures, higher monitoring frequencies, and advanced RAIM

capabilities.

These criteria are naturally linked, although the degree of impor-

tance might vary from plan to plan. For instance, return require-

ments may play the primary role in driving the choice of instru-

ments and asset classes for some plans, while they may play a

less important role for other plans. For some plans, the investment

horizon is tied directly to their return requirements, while for oth-

ers, it is more a function of how much they are willing to lose over a

given period. Regardless, these three criteria determine the guiding

principles for any plan’s risk management function.

For example, a plan sponsor with reasonable and relatively

infrequent payout obligations, a large surplus, and with limited

exposure to alternatives and complex instruments does not need

short-term measures or frequent monitoring. This plan would

benefit from focusing on longer-term risk measures. Instead of

setting up a system to calculate 10-day VaR measures, the plan

could focus on how multiyear regime shifts in different risk fac-

tors, such as interest rate cycles, may affect the portfolio’s value.

In contrast, a plan with frequent and significant expected payouts,

a limited ability to sustain short-term losses, and with substantial

exposure to alternatives and complex instruments would require

a wide variety of risk measures, frequent risk monitoring, and a

well developed RAIM process. Most plans are likely to fall between

these two extremes.

Another example may help to shed light on these ideas. Below, we

group investors into one of three categories using the third criteria —

complexity of instruments and asset classes.

A Type-1 plan invests in a straightforward allocation to equities and

fixed income. Equity and fixed income allocations may be limited

to the domestic market, and fixed income investments are mostly

concentrated in government bonds and AAA-rated corporate. A

Type-2 plan may invest in equities globally, including emerging

markets. Fixed income investments may include high yield bonds

and mortgage-backed securities, and the plan may also invest in

alternatives and complex derivatives. However, as a percentage of

the plan’s total value, the investments in alternatives and deriva-

tives would be small. A Type-3 plan would invest in a variety of

alternative asset classes as well as complex instruments but to a

larger extent than Type-2 plans.

Currently, the vast majority of pension plans fall into the second

group. We, therefore, consider a hypothetical Type-2 plan for our

illustration of a risk management framework. Its asset allocation

is as follows: equity (60%) [U.S. (36%), international (24%)], fixed

income (U.S.) (25%), alternatives (15%) [real estate (5%), private

equity (5%), hedge funds (5%)].

A first critical step is to adopt tools that enable the plan to mea-

sure and monitor risk at the source, or factor level, and not just at

the aggregate level. The plan should then monitor its exposure to

different risk sources on a regular basis — monthly or quarterly at

the very least. This would occur at the total portfolio level, look-

ing across asset classes. In addition, the plan can require from its

managers more detailed information on risk exposures. Many plans

receive only high-level performance summaries focusing on real-

ized returns and tracking error. Requiring estimates of exposures

to various sources of risk is a crucial extension. For illustration, we

show how this would fit in the framework we have used so far:

Total risk:

normal periods —■■ the plan can evaluate sources of return and

risk across its overall portfolio using a multi-asset class factor

model. Sources of risk can include macroeconomic and market

factors. An example of the type of analysis that can be done is

to look at the performance of the plan’s portfolio in different

macroeconomic regimes. The plan could then adjust its asset

allocation during the next review period.

Extreme events —■■ using the sources of risk for the overall port-

folio, the plan can shock certain factors or sources of risk, i.e.,

those likely to suffer in the event of a market dislocation, includ-

ing a systemic meltdown or series of external shocks. These

stress tests would enable the plan to evaluate how individual or

multiple simultaneous shocks impact the overall portfolio (i.e.,

tail risk and tail correlations). The plan could then establish an

action plan if asset values drop by some absolute or relative (i.e.,

to liabilities) amount.

Active risk:

normal periods —■■ the plan can ask for reports on their sources

of risk from all equity, fixed income, and alternatives managers,

or the plan can estimate them internally. Sources of active risk

should be detailed and focused on the specific risk and return

drivers of the manager’s investment strategy. For instance,

analyzing a value, small cap equity manager’s tracking error will

focus on the active bets relative to the agreed upon benchmark,

ideally a small cap value benchmark like the MSCI Small Cap

Value Index. Measuring and monitoring will focus on questions

42 – The journal of financial transformation

Best practices for investment risk management

of active bets relative to this benchmark, for example, does the

manager’s portfolio have a consistent small cap value bias or did

the portfolio move towards growth-oriented companies over the

last few years when value stocks underperformed? The active

risk analysis should ensure that the hired manager is following

his or her mandate and is not deviating from the agreed upon

guidelines.

Extreme events —■■ the plan may want to stress test the impact

of the joint underperformance of a number of active strategies

that historically have been uncorrelated. For example, during

the quant meltdown in August 2007, a number of return factors

became suddenly highly correlated, leading to severe negative

portfolio performance relative to their respective benchmarks.

Such stress tests enable the plan to evaluate how shocks impact

an entire group of managers. Other useful measures for rare

events are tail risk and tail risk correlations of active bets across

managers. Certain factors or strategies may become highly cor-

related across asset classes or markets during crises (i.e., value

or momentum across equity markets) and could lead to vastly

higher losses relative to their respective benchmarks than esti-

mated by the tracking error.

The above examples illustrate how a model that decomposes risk

along its sources can help institutions evaluate different types of

risk across different dimensions. It can be applied similarly to real-

ized returns in order to attribute past performance. For our hypo-

thetical Type-2 plan, a suggested set of minimum components for

risk management may include:

Aggregate measures of volatility and tracking error across man-■■

agers and asset classes.

An accurate decomposition of return and risk across asset ■■

classes, utilizing an integrated (across asset classes) multi-factor

risk model.

A stress testing framework and/or extreme risk measures for ■■

understanding tail risk and tail risk correlations in the portfolio.

An appropriate set of benchmarks.■■

The exact measures, monitoring frequencies, and RAIM processes

the plan adopts will depend on its return requirements and expect-

ed payouts, and its investment horizon and willingness to tolerate

shorter-term losses. For instance, a plan with limited ability to

withstand short-term losses may want to build out its ability to

assess tail risk over different horizons using risk measures such as

Expected Shortfall based on Extreme Value Theory [Goldberg et al.

(2009)], which is more conservative than parametric VaR. These

plans may also want to implement extensive stress tests across

asset classes and within certain subcategories of investments.

Meanwhile, plans with greater ability to withstand short-term losses

may opt for more basic tail risk measures and stress tests.

Our example focused on a Type-2 plan. For a Type-3 plan, this illus-

tration would also be relevant, but the requirements for measuring,

monitoring, and managing different types and sources of risk would

be more extensive. For a Type-1 plan, the extent to which it invests

in risk management should depend on its liabilities structure and its

short-term risk tolerance.

Most plans, regardless of their specific characteristics, can take

some basic actions on an organizational or administrative level to

manage risk. Our hypothetical plan may establish a risk commit-

tee consisting of the CIO, risk managers, senior portfolio manag-

ers, and legal and compliance officers that meets at least once a

quarter to discuss the overall economic and financial environment.

Participants can discuss their concerns regarding systemic risk

issues such as liquidity and leverage, review the results of stress

tests, and debate whether hedging strategies should be activated to

address undesired exposures or potential tail risk events. The plan

could also develop a reporting framework where the risk committee

would receive at least monthly reports on unusual developments

identified by the risk manager. Then, if the investment committee

is sufficiently concerned about exposure to a certain segment, it

could ask those managers with large exposures to hedge them or

to eliminate the undesired exposures.

Figure 4 illustrates this type of setup, where the risk manager pre-

pares risk reports and recommendations for the risk committee and

deliverers risk management services and advice to the different

asset class managers.

Finally, the plan may establish minimum risk management require-

ments for external managers. For instance, the external managers

could be required to demonstrate their ability to calculate tracking

error, VaR, or other measures, as well as how risk management

impacts their portfolio construction.

senior investment /risk committee

Risk manager

Equity Alternatives fixed income

Initial asset/manager allocation•

Standard risk reports•

Red flags/exceptions reporting•

Figure 4 – Organizational structure for risk management

43

Best practices for investment risk management

conclusionRecent events have put into stark relief the inadequacy of the cur-

rent state of risk management. Much has been said about the need

for better risk management and a greater degree of risk aware-

ness in the broader investment community. Risk management is

a dynamic area, and any set of best practices are bound to evolve

over time. Here we set out to clarify some of the principles and

tools that we believe are required for a sound risk management

framework.

Specifically, we lay out a framework that rests on three pillars — risk

measurement, monitoring, and RAIM (or risk-adjusted investment

management). Each of the three domains is critical for risk manage-

ment. Risk measurement means having the right tools to measure

risk accurately from various perspectives. Risk monitoring means

observing the risk measures on a regular and timely basis. RAIM

means using the information from the measurement and monitor-

ing layers intelligently to ensure that the portfolio management

process is aligned with expectations of risk and risk tolerance.

While each pillar encompasses a different aspect of risk manage-

ment, each is indispensable to a strong risk management process.

Moreover, they are interdependent and should be aligned with the

investor’s objectives. Their interconnectedness drives the key con-

ceptual theme — that risk management and the investment process

should be fully integrated.

ReferencesBarbieri, A., V. Dubikovsky, A. Gladkevich, L. Goldberg, and M. Hayes, 2009, “Central •

limits and financial risk,” MSCI Barra Research Insights

Best, P., and M. Reeves, 2008, “Risk management in the evolving investment manage-•

ment industry,” Journal of Financial Transformation, 25, 88-90

Goldberg, L., M. Hayes, J. Menchero, and I. Mitra, 2009, “Extreme risk analysis,” MSCI •

Barra Research Insights

Kneafsey, K., 2009, “Four demons,” Journal of Financial Transformation, 26, 18-23•

Articles

45

lessons from the global financial meltdown of 2008

Hershey H. friedmanProfessor of Business and Marketing, Department of Economics, Brooklyn College, City University of

New York

linda W. friedmanProfessor, Department of Statistics and Computer

Information Systems, Baruch College and the Graduate Center of the City University of New York

AbstractThe current financial crisis that threatens the entire world has

created an ideal opportunity for educators. A number of impor-

tant lessons can be learned from this financial meltdown. Some

are technical and deal with the value of mathematical models and

measuring risk. The most important lesson, however, is that unethi-

cal behavior has many consequences. This debacle could not have

occurred if the parties involved had been socially responsible and

not motivated by greed. Conflicts of interest and the way CEOs are

compensated are at the heart of this financial catastrophe that has

wiped out trillions of dollars in assets and millions of jobs. We pres-

ent a set of lessons as teaching opportunities for today’s students

and tomorrow’s decision makers.

46 – The journal of financial transformation

lessons from the global financial meltdown of 2008

The current financial crisis is the worst debacle we have experienced

since the Great Depression. Millions of jobs have been lost and tril-

lions of dollars in market value have evaporated. The entire world is

suffering. The financial crisis is far from over and is adversely affect-

ing people all over the world. Are there lessons that can be learned

from this financial debacle? In fact, it is the perfect teaching tool

since it makes it so easy to demonstrate what can go wrong when

firms do not behave in an ethically, socially responsible manner.

It is interesting to note that the financial meltdown of 2008 did not

suddenly appear out of nowhere. The corporate world was heading

down a dangerous path for more than 20 years. We feel that an

early warning was the Savings and Loans disaster in which 1,043

banks failed with a cost to U.S. taxpayers of about U.S.$124 billion

[Curry and Shibut (2000)]. That happened between 1986 and 1995.

The financial world ignored this warning. A few years later came

several colossal corporate scandals including Enron, which filed

for bankruptcy in late 2001, Tyco International, Adelphia, Global

Crossing, WorldCom, and many other firms. These companies were

found to use dubious accounting practices or engage in outright

accounting fraud to deceive the public and enrich executives. In

fact, the Sarbanes-Oxley Act of 2002 was enacted in order to

prevent future financial disasters such as Enron. Suskind (2008)

reports that Alan Greenspan, Chairman of the Federal Reserve,

was at a meeting on February 22, 2002 after the Enron debacle

and was upset with what was happening in the corporate world.

Mr. Greenspan noted how easy it was for CEOs to “craft” financial

statements in ways that could deceive the public. He slapped the

table and exclaimed: “There’s been too much gaming of the system.

Capitalism is not working! There’s been a corrupting of the system

of capitalism.” This was another warning sign that it was easy for

unrestrained greed to harm the entire economy.

The dot.com bubble which took place between 1995 and 2001

(NASDAQ peaked at over 5100 in March 2000), was a different

kind of crisis. It was fueled by irrational spending on Internet stocks

without considering traditional business models. Investors did not

seem to care about earnings per share or other more traditional

measures. Moreover, there were too many companies trying to

create online businesses. The price of many dot.com stocks came

tumbling down, but the bubble was not based on fraud as much as

overvaluation of stocks (especially the IPOs) and excessive specu-

lation. The housing bubble, on the other hand, that helped cause

the current financial meltdown was fueled to a large degree by the

ready availability of deceitful mortgages. What was apparent from

the dot.com bubble was that prices cannot go up forever — a lesson

not heeded by those in the mortgage business.

Long Term Capital Management (LTCM), a hedge fund founded

in 1994, which went out of business in 2000, showed how risky a

highly-leveraged hedge fund could be. In 1998, LTCM had borrowed

over U.S.$125 billion, but only had equity of U.S.$5 billion. The

financial crisis started by LTCM at that time also demonstrated how

the entire financial system could be at risk because of the actions

of one fund. The Federal Reserve Bank was involved in a U.S.$3.5

billion rescue package in 1998 to protect the financial markets

from a total collapse because of the actions of LTCM. One lesson

that should have been learned from this financial debacle was that

we should not rely so much on sophisticated mathematical models.

LTCM’s models were developed by two Nobel laureates — Myron

Scholes and Robert C. Merton — who were both members of the

board of the hedge fund.

Lowenstein (2008) observes that “regulators, too, have seemed to

replay the past without gaining from the experience. What of the

warning that obscure derivatives needed to be better regulated

and understood? What of the evident risk that intervention from

Washington would foster yet more speculative behavior — and pos-

sibly lead to a string of bailouts?” Lowenstein (2008) states that

only six months after the LTCM fiasco, Alan Greenspan “called for

less burdensome derivatives regulation, arguing that banks could

police themselves.” Needless to say, Greenspan was proven quite

wrong in this assertion.

The scandal involving Bernard Madoff which has been called the larg-

est Ponzi scheme ever has also served to cast serious doubts on how

well our financial system is being monitored. Gandel (2008) notes

that KPMG, PricewaterhouseCoopers, BDO Seidman, and McGladrey

& Pullen all signed off that all was well with the many feeder funds

that had invested with Madoff. What has shocked everyone is that

the auditors did not recognize that billions of dollars of assets were

just not there. Auditors are supposed to check that the stated assets

actually exist. Interestingly, Madoff himself was not a client of any of

the large auditing firms; he used a tiny accounting firm in New City,

NY that had only three employees. It is now apparent that this alone

should have been an indication that something was very wrong with

the way Madoff conducted business. One lawyer remarked: “All they

really had to substantiate the gains of these funds was Madoff’s own

statements. They were supposed to be the watchdogs. Why did they

sign off on these funds’ books?” [Gandel (2008)].

What can be learned from the above crises, especially the financial

meltdown of 2008? Firstly, we should recognize that what we have

experienced is not the breakdown of an economic system. This is

a values meltdown more than anything else. In fact, a recent poll

showed that bankers are near the bottom of the list when it comes

to respect felt by the public and barely beat prostitutes and con-

victed felons [Norris (2009)].

is the pursuit of self-interest always good?One pillar of mainstream economics taught in most economics

courses is based on the famous saying of Adam Smith in his classic

47Available at: http://www.americanrhetoric.com/MovieSpeeches/moviespeechwall-1

street.html

lessons from the global financial meltdown of 2008

work, “Wealth of Nations,” that: “it is not from the benevolence of

the butcher, the brewer or the baker that we expect our dinner,

but from their regard to their own interest.” Smith demonstrated

how self-interest and the “invisible hand” of the marketplace

allocates scarce resources efficiently. Students are taught that

“economic man” or homo economicus acts with perfect rationality

and is interested in maximizing his/her self-interest. This results in

an economic system (capitalism) that is efficient and productive.

For the corporation, self-interest is synonymous with maximiza-

tion of profits and/or maximization of shareholder wealth. Indeed,

students are being taught that self interest plus free markets and

deregulation results in prosperity for everyone. The famous speech

by Gordon Gekko in the movie Wall Street is based on the idea that

the pursuit of self-interest is good for all of us1: “Greed, for lack of

a better word, is good. Greed is right. Greed works. Greed clarifies,

cuts through, and captures the essence of evolutionary spirit. Greed

in all of its forms, greed for life, for money, for love, knowledge has

marked the upward surge of mankind. And greed, you mark my

words, will not only save Teldar Paper, but that other malfunction-

ing corporation called the USA.”

Change the word “greed” to pursuit of self-interest and you have in

effect what has been taught to millions of business and economics

students. Even the so-called watchmen and gatekeepers — cor-

porate directors, investment bankers, regulators, mutual funds,

accountants, auditors, etc. — have fallen into the self-interest trap

and disregarded the needs of the public [Lorsch et al. (2005)].

It is ironic that the discipline of economics started as part of the

discipline of moral philosophy, and even became a moral science

[Alvey (1999)]. Adam Smith, in his first book, “The Theory of Moral

Sentiments,” made it clear that he believed that economic growth

depended on morality. To Smith, benevolence — not pursuit of self-

interest — was the highest virtue [Alvey (1999)]. The following quo-

tation demonstrates what Smith actually believed: “Man ought to

regard himself, not as something separated and detached, but as a

citizen of the world, a member of the vast commonwealth of nature

and to the interest of this great community, he ought at all times to

be willing that his own little interest should be sacrificed.”

Robinson made the point more than 30 years ago that the pursuit

of self-interest has caused much harm to society and that Adam

Smith should not be associated with this doctrine. In actuality,

Smith believed that “society, however, cannot subsist among those

who are at all times ready to hurt and injure one another.” Raw self-

interest without a foundation of morality is not what Adam Smith

is all about. Robinson ended a commencement address with the

following warning: “I hope … that you will find that the doctrines of

Adam Smith are not to be taken in the form in which your profes-

sors are explaining them to you” [Robinson (1977)].

Howard (1997) uses the expression “tragedy of maximization” to

describe the devastation that the philosophy of maximizing self-

interest has wrought. Unrestrained capitalism that is obsessed with

self-interest and is unconcerned about the long run, can lead to

monopoly, inequitable distribution of income, unemployment, and

environmental disaster [Pitelis (2002)].

In 1937, at his second inaugural address, President Franklin D.

Roosevelt stated: “We have always known that heedless self-

interest was bad morals; we know now that it is bad economics.”

[Roosevelt (1937)]. Lawrence H. Summers, in a 2003 speech to the

Chicago Economic Club made the following remark: “For it is the

irony of the market system that while its very success depends

on harnessing the power of self-interest, its very sustainability

depends upon people’s willingness to engage in acts that are not

self-interested.”

The financial meltdown of 2008 shows quite clearly what happens

when everyone is solely concerned with self-interest. Closely tied to

the concept of pursuit of self-interest is the belief that free markets

do not need regulation. Many economists and CEOs promoted the

belief that capitalism could only work well with very little regulation.

free markets and the role of regulationThere has been a small movement in economics that questions

the neoclassical model in economics and its belief that free mar-

kets and laissez-faire economics will solve all problems [Cohen

(2007)]. According to one economist, only 5% to 10% of America’s

15,000 economists are “heterodox,” i.e., do not follow the neoclas-

sical model promoted by free market enthusiasts such as Milton

Friedman. Some heterodox economists feel that neoclassical eco-

nomics has become “sycophantic to capitalism;” the discipline is

concerned with mathematical solutions that do not resemble the

real world. The discipline is more concerned about models than

solving social problems [Monaghan (2003)].

Lichtblau (2008) believes that the federal government did not do

a good job monitoring Wall Street. He cites what Arthur Levitt,

former chairman of the SEC, said regarding his former agency: “As

an overheated market needed a strong referee to rein in danger-

ously risky behavior, the commission too often remained on the

sidelines.” Sean Coffey, who used to be a fraud prosecutor, also

was not happy with the performance of the SEC: the SEC “neutered

the ability of the enforcement staff to be as proactive as they could

be. It’s hard to square the motto of investor advocate with the

way they’ve performed the last eight years.” Not only was there a

relaxation of enforcement, there was also a reduction in SEC staff.

Coffey asserts that the administration used the argument that

loosening up regulations was necessary in order to make it possible

for American companies to compete globally [Lichtblau (2008)].

Senator Charles E. Schumer also believed that the rules had to be

48 – The journal of financial transformation

lessons from the global financial meltdown of 2008

changed to encourage free markets and deregulation if the United

States was to remain competitive [Lipton and Hernandez (2008)].

The problems began at a meeting on April 28, 2004 between five

major investment banks and the five members of the SEC The

investment banks wanted an exemption from a regulation that lim-

ited the amount of debt — known as the net capital rule — they could

have on their balance sheets. By increasing the amount of debt they

would be able to invest in the “opaque world of mortgage-backed

securities” and credit default swaps (CDSs). The SEC agreed to

loosen the capital rules and also decided to allow the investment

banks to monitor their own riskiness by using computer models to

analyze the riskiness of various securities, i.e., switch to a voluntary

regulatory program [Labaton (2008)]. The firms did act on the new

requirements and took on huge amounts of debt. The leverage ratio

at Bear Stearns rose to 33:1. This made the firm very risky since it

held only $1 of equity for $33 of debt. Regarding what transpired,

James D. Cox said [Labaton (2008)]: “We foolishly believed that

the firms had a strong culture of self-preservation and responsibil-

ity and would have the discipline not to be excessively borrowing.

Letting the firms police themselves made sense to me because I

didn’t think the SEC had the staff and wherewithal to impose its own

standards and I foolishly thought the market would impose its own

self-discipline. We’ve all learned a terrible lesson.”

Alan Greenspan finally admitted at a congressional hearing in

October, 2008 that he had relied too much on the “self-correcting

power of free markets” [Andrews (2008)]. He also acknowledged

that he did not anticipate the “self-destructive power of wanton

mortgage lending” [Andrews (2008)]. Greenspan was asked wheth-

er his ideology — i.e., belief in deregulation and that government

regulators did not do a better job than free markets in correcting

abuses — had contributed to the financial crisis. He admitted that he

had made a mistake and actually took partial responsibility. Some

of the mistakes Greenspan made had to do with risky mortgages

and out-of-control derivatives.

Risky mortgagesGreenspan allowed the growth of highly risky and fraudulent mort-

gages without recognizing the “self-destructive power” of this type

of mortgage lending with virtually no regulation [Skidelsky (2008)].

The Fed could have put a stop to it by using its power under a 1994

law (Home Owner Equity Protection Act) to prevent fraudulent

lending practices. It was obvious that the mortgage industry was

out of control and was allowing individuals with very little money

to borrow huge sums of money. Greenspan could also have used

the monetary powers of the Fed to raise interest rates which would

have ended the housing bubble.

Here are just a few examples of how deregulation affected the sub-

prime mortgage market:

WaMu, for example, lent money to nearly everyone who asked for

it. Loan officers were encouraged to approve mortgages with virtu-

ally no checking of income [Goodman and Morgenson (2008)]. To

encourage people with little income to borrow money, WaMu used

option ARMs (Adjustable Rate Mortgages). The very low initial rates

enticed people to take out mortgages. Of course, many borrowers

thought the low payments would continue indefinitely and would

never balloon [Goodman and Morgenson (2008)]. The number of

ARMs at WaMu increased from 25% (2003) to 70% (2006). It did

not take long for word to spread that WaMu would give mortgages

to nearly anyone. WaMu even ran an advertising campaign telling

the world that they would give mortgages to anyone, “The power

of yes.” WaMu did exactly that: it approved almost every mortgage.

Revenues at WaMu’s home lending unit increased from $707 million

to approximately $2 billion in one year. As one person noted: “If

you were alive, they (WaMu) would give you a loan. Actually, I think

if you were dead, they still would give you a loan” [Goodman and

Morgenson (2008)].

In the mortgage business, NINJA loans are mortgage loans made

to people with “No income, no job or assets.” These mortgage

loans were made on the basis of unsubstantiated income claims

by either the applicant or the mortgage broker or both. There are

stories of individuals with income of U.S.$14,000 a year purchasing

U.S.$750,000 homes with no money down and no mortgage pay-

ments for two years [Friedman (2008)]. Other types of mortgages

that became popular include the “balloon mortgage” (borrower only

makes the interest payment for 10 years but then has to pay a huge

amount — the balloon payment); the “liar loan” (borrower claims an

annual income, no one checks and there is no documentation); the

“piggyback loan” (this combines a first and second mortgage so a

down payment is not necessary); the “teaser loan” (mortgage at very

low interest rate for first two years but when it is readjusted after the

two years, the borrower does not have the income to make the pay-

ments); the “option ARM loan” (discussed above); and the “stretch

loan” (borrower has to use more than 50% of his monthly income to

make the mortgage payments) [Pearlstein (2007)].

President Bush wanted to encourage homeownership, especially

among minorities. Unfortunately, his approach, which encouraged

easy lending with little regulation, helped contribute to the financial

crisis [Becker et al. (2008)]. The increase in homeownership was

accomplished through the use of toxic mortgages. President Bush

also encouraged mortgage brokers and corporate America to come

up with innovative solutions to enable low-income people to own

homes. L. William Seidman, an advisor to Republican presidents,

stated: “This administration made decisions that allowed the free

market to operate as a bar room brawl instead of a prize fight.”

Bush’s banking regulators made it clear to the industry that they

would do everything possible to eliminate regulations. They even

used a chainsaw to symbolically show what they would do to the

49

lessons from the global financial meltdown of 2008

banking regulations. Various states, at one point, did try to do

something about predatory lending but were blocked by the federal

government. The attorney general of North Carolina said: “They

took 50 sheriffs off the beat at a time when lending was becoming

the Wild West” [Becker et al. (2008)].

The policy of encouraging homeownership affected how Fannie

Mae and Freddie Mac conducted business. Fannie and Freddie are

government-sponsored, shareholder-owned companies, whose job

is to purchase mortgages. They purchase mortgages from banks

and mortgage lenders, keep some and sell the rest to investors.

This enables banks to make additional loans, thus allowing more

people to own homes. Fannie and Freddie were encouraged by

President Bush to do everything possible to help low-income people

buy homes. One way to accomplish this was to reduce the down

payments. There was a drive to allow new homeowners to obtain

a federally-insured mortgage with no money down. There was little

incentive to do something about the super-easy lending practices,

since the housing industry was helping to pump up the entire econ-

omy. The increasing home values helped push consumer spending.

More and more people were becoming homeowners in line with

what President Bush wanted.

Back in the year 2000, Fannie Mae, under CEO Franklin D. Raines

and CFO J. Timothy Howard decided to expand into riskier mort-

gages. They would use sophisticated computer models and rank

borrowers according to how risky the loan was. The riskier the

loan, the higher the fees that would be charged to guarantee the

mortgage. The big risk was if a huge number of borrowers would be

unable to make the payments on their mortgages and walk away

from their obligations. That danger was not seen as likely in 2000.

Moreover, the computer models would ensure that the higher fees

for the riskier mortgages would offset any losses from mortgage

defaults. The company announced that it would purchase U.S.$2

trillion in loans from low-income (and risky) borrowers by 2010

[Duhigg (2008)].

What this accomplished — besides enriching the executives at Fannie

Mae — was that it made subprime mortgages that in the past would

have been avoided by lenders more acceptable to banks all over

the country. These banks did not have the sophistication or experi-

ence to understand the kind of risk they were taking on. Between

2001 and 2004, the subprime market grew from U.S.$160 billion to

U.S.$540 billion [Duhigg (2008)]. In 2004, there were allegations

of accounting fraud at Freddie and Fannie and both had to restate

earnings. Daniel H. Mudd became CEO of Fannie Mae in 2005 after

Raines and Howard resigned from Fannie Mae under a cloud. Under

Mudd’s watch, Fannie purchased even riskier mortgages, those that

were so new that the computer models could not analyze them

properly. Mr. Mudd was warned by Enrico Dallavecchia, his chief risk

officer, that the company was taking on too much risk and should

charge more. According to Dallavecchia, Mudd’s response was that

“the market, shareholders, and Congress all thought the companies

should be taking more risks, not fewer. Who am I supposed to fight

with first?” [Duhigg (2008)].

As early as February 2003, Armando Falcon, Jr., who ran the Office

of Federal Housing Enterprise Oversight (OFHEO), wrote a report

that warned that Fannie and Freddie could default because they

were taking on far too many mortgages and did not have the capi-

tal to protect themselves against losses. Falcon almost got fired

for this report. After some accounting scandals at Freddie, the

President decided to keep Falcon on. A bill was written by Michael

Oxley, a Republican and Chairman of the House Financial Services

Committee that would have “given an aggressive regulator enough

power to keep the companies from failing” [Becker et al. (2008)].

Since the bill was not as strong as what President Bush wanted,

he opposed it and it died in the Senate. The bottom line was that

in Bush’s desire to get a tougher bill, he ended up with no bill at

all. Eventually, James B. Lockhart III became the Director of the

OFHEO. Under his watch, Freddie and Fannie purchased U.S.$400

billion of the most risky subprime mortgages. In September, 2008,

the federal government had to take over Fannie Mae and Freddie

Mac.

Derivatives out of controlGreenspan also admitted that he allowed the market for derivatives

to go out of control. Greenspan was opposed to tighter regulation

of derivatives going back to 1994 [Andrews (2008)]. George Soros,

the renowned financier, avoided derivatives “because we don’t

really understand how they work”; Felix G. Rohatyn, the prominent

investment banker, called derivatives “potential hydrogen bombs”;

and Warren E. Buffett remarked that derivatives were “financial

weapons of mass destruction.” Greenspan, on the other hand, felt

that “derivatives have been an extraordinarily useful vehicle to

transfer risk from those who shouldn’t be taking it to those who are

willing and capable of doing so” [Goodman (2008)].

Greenspan also admitted that the market for credit default swaps

(CDSs), which became a multi-trillion dollar business, went out of

control [Andews (2008)]. The CDSs were originally developed to

insure bond investors against default risk but they have taken on

a life of their own and have been used for speculative purposes.

A CDS is a credit derivative and resembles insurance since the

buyer makes regular payments and collects if the underlying finan-

cial instrument defaults. In a CDS, there is the protection buyer who

can use this instrument for credit protection; the protection seller

who gives the credit protection; and the “reference entity,” which

is the specific bond or loan that could go bankrupt or into default.

It could be compared to buying fire insurance on someone else’s

house. Imagine how the insurance industry would work if 1000

people were able to buy fire insurance on Jane Doe’s house. With

50 – The journal of financial transformation

lessons from the global financial meltdown of 2008

traditional insurance, if Jane Doe’s house burns down, there is only

one payment. If, on the other hand, 1000 people were allowed to

each own a fire insurance policy on Jane Doe’s home, one thousand

payouts must be made. The insurance company would be collecting

fees from 1,000 customers and have a nice revenue stream, but the

risk would be very great. In fact, we would have a situation where

one thousand people would do everything possible to make sure the

Jane Doe home burns down.

Another difference between a CDS and traditional insurance is that

there is a market for the CDS, which makes it attractive for specula-

tors, since the owner of the CDS does not actually have to own the

underlying security. One problem with this is that a hedge fund can

buy a CDS on a bond and then do everything possible (i.e., sell the

stock short to force the price of the stock down) to make sure that

the reference entity does indeed default [Satow (2008)].

It should be noted that the market for derivatives and CDSs became

unregulated thanks to the Commodity Futures Modernization Act of

2000. This law was pushed by the financial industry in the name of

free markets and deregulation. It also made it virtually impossible

for states to use their own laws to prevent Wall Street from doing

anything about these financial instruments. This bill was passed “at

the height of Wall Street and Washington’s love affair with deregu-

lation, an infatuation that was endorsed by President Clinton at the

White House and encouraged by Federal Reserve Chairman Alan

Greenspan” [Kroft (2008)].

According to Eric Dinallo, the insurance superintendent of New

York State [Kroft (2008)]: “As the market began to seize up and as

the market for the underlying obligations began to perform poorly,

everybody wanted to get paid, had a right to get paid on those

credit default swaps. And there was no ‘there’ there. There was no

money behind the commitments. And people came up short. And

so that’s to a large extent what happened to Bear Sterns, Lehman

Brothers, and the holding company of AIG. It’s legalized gambling.

It was illegal gambling. And we made it legal gambling…with abso-

lutely no regulatory controls. Zero, as far as I can tell.” In response

to the question as to whether the CDS market was like a “bookie

operation,” Dinallo said: “Yes, and it used to be illegal. It was very

illegal 100 years ago.”

SEC Chairman, Christopher Cox, stated that it is of utmost impor-

tance to bring transparency to the unregulated market in CDSs.

They played a major role in the financial meltdown and were also

the cause of the near bankruptcy of AIG which necessitated a gov-

ernment bailout [Landy (2008), Cox (2008)]. According to Landy

(2008), no one even knows the exact size of the CDS market;

estimates range from U.S.$35 trillion to U.S.$55 trillion. When AIG

was bailed out by the federal government it held U.S.$440 billion of

CDSs [Philips (2008)].

One banker from India made the point that his colleagues were sur-

prised at the “lack of adequate supervision and regulation.” What

made it even more amazing was that all this occurred after the

Enron debacle and the passing of the Sarbanes-Oxley bill [Nocera

(2008)]. In fact, India avoided a subprime crisis because a bank

regulator by the name of V. Y. Reddy, who was the opposite of

Greenspan, believed that if “bankers were given the opportunity to

sin, they would sin.” Unlike Greenspan, he felt that his job was to

make sure that the banking system would not be part of a huge real

estate bubble. He used his regulatory powers to deflate potential

bubbles by not allowing bank loans for the purchase of undeveloped

land. Bankers were not happy that Reddy was preventing them

from making huge amounts of money, like their American peers

were making. Now everyone knows that Reddy was right. As one

Indian banker explained, regarding what happened in the United

States: “It was perpetuated by greedy bankers, whether investment

bankers or commercial bankers. The greed to make money is the

impression it has made here.” [Nocera (2008)].

credit rating agenciesThere are three major credit-rating agencies: Moody’s, Standard

& Poors, and Fitch Ratings. All have been accused of being overly

generous in how they rated the securities that consisted of bundled,

low-quality mortgages. The big question that has arisen is whether

these firms assigned very good ratings (AAA) because of sheer

incompetence or to make more money. By ingratiating themselves

with clients, they were able to steer more business to themselves.

There is no question that the firms were able to charge consider-

ably more for providing ratings for complex financial securities than

for simple bonds. Rating a complex U.S.$350 million mortgage pool

would generate approximately U.S.$250,000 in fees, as compared

to U.S.$50,000 for an equally sized municipal bond. Morgenson

(2008a) quotes a managing director at Moody’s, a firm that rates the

quality of bonds, as saying that: “These errors make us look either

incompetent at credit analysis or like we sold our soul to the devil for

revenue, or a little of both.”

Moody’s graded the securities that consisted of Countrywide Financial’s

mortgages — Countrywide is the largest mortgage lender in the United

States. Apparently, the ratings were not high enough and Countrywide

complained. One day later, Moody’s raised the rating. This happened

several times with securities issued by Countrywide.

Morgenson (2008a) provides an interesting example of how unreli-

able the ratings had become. A pool of residential subprime mort-

gages was bundled together by Goldman Sachs during the summer

of 2006 (it was called GSAMP 2006-S5). The safest part of this,

consisting of U.S.$165 million in debt received a AAA rating from

Moody’s and S & P on August 17, 2006. Eight months later, the rat-

ing of this security was lowered to Baa; and on December 4, 2007,

it was downgraded to junk bond status.

51

lessons from the global financial meltdown of 2008

Method of compensation may encourage risk takingWarren Buffet once said: “in judging whether corporate America

is serious about reforming itself, CEO pay remains the acid test”

[Kristof (2008)]. It is obvious to all that corporate America has not

performed well on this test. The fact that as many as 29% of firms

have been backdating options makes it appear that the system of

compensation of executives needs an overhaul [Burrows (2007)].

According to a Watson Wyatt survey, approximately 90% of insti-

tutional investors believe that top executives are dramatically

overpaid [Kirkland (2006)].

Richard Fuld, CEO of Lehman Brothers, earned approximately half-

a-billion dollars between 1993 and 2007. Kristof (2008) observes

that Fuld earned about U.S.$17,000 an hour to destroy a solid,

158-year old company. AIG Financial Products, a 377-person office

based in London, nearly destroyed the mother company, a trillion-

dollar firm with approximately 116,000 employees. This small office

found a way to make money selling credit default swaps to financial

institutions holding very risky collateralized debt obligations. AIG

Financial Products made sure that its employees did very well

financially: they earned U.S.$3.56 billion in the last seven years

[Morgenson (2008b)].

One of the big problems at many Wall Street firms was how

compensation was determined. Bonuses made up a huge part of

how people were compensated. One individual at Merrill Lynch

received U.S.$350,000 as salary but U.S.$35,000,000 as bonus

pay. Bonuses were based on short-term profits, which distorted

the way incentives work. Employees were encouraged to take

huge risks since they were interested in the bonus. Bonuses were

based on the earnings for that year. Thus, billions in bonuses were

handed out by Merrill Lynch in 2006 when profits hit U.S.$7.5 bil-

lion. Of course, those profits were only an illusion as they were

based on toxic mortgages. After that, the company lost triple that

amount, yet the bonuses were not rescinded [Story (2008)]. Lucian

Bebchuk, a compensation expert, asserted that “the whole organi-

zation was responding to distorted incentives” [Story (2008)].

It is clear that bonuses based on the profits of a particular year

played a role in the financial meltdown. Firms are now changing

the way bonuses work so that employees have to return them if

the profits turn out to be illusory. The money will be kept in escrow

accounts and not distributed unless it is clear that the profits are

real. E. Stanley O’Neal, former CEO of Merrill Lynch, not only col-

lected millions of dollars in bonuses but was given a package worth

about U.S.$161 million when he left Merrill Lynch. He did quite well

for someone whose claim to fame is that he nearly destroyed Merrill

Lynch; it was finally sold to Bank of America.

According to WaMu employees, CEO Kerry K. Killinger put a huge

amount of pressure on employees to lend money to borrowers with

little in the way of income or assets. Real estate agents were given

fees of thousands of dollars for bringing borrowers to WaMu. WaMu

also gave mortgage brokers generous commissions for the riskiest

loans since they produced higher fees and resulted in increased com-

pensation for executives. Between 2001 and 2007, Killinger earned

approximately U.S.$88 million [Goodman and Morgenson (2008)].

Cohan (2008) also sees “Wall Street’s bloated and ineffective com-

pensation system” as a key cause of the financial meltdown. Several

politicians are examining the way bonuses work in order to make

sure that they will not be given to executives of firms that are receiv-

ing government assistance. Cohan (2008) feels very strongly that

compensation reform in Wall Street is badly needed. He feels that

the change of the old system, where the big firms (i.e., Donaldson,

Lufkin, and Jenrette; Merrill Lynch; Morgan Stanley; Goldman

Sachs; Lazard; etc.) switched from being a partnership to a public

company, contributed to the financial debacle. When these firms

were partnerships, there was collective liability so the firms were

much more cautious. Once they became corporations with common

stock, “bankers and traders were encouraged to take short-term

risks with shareholder’s money.” These bankers and traders did not

mind taking on more and more risk since their bonuses depended

on the annual profits. Put these ingredients together — encourage

risk taking, no accountability, and use of shareholder’s money — and

you get a financial meltdown. Siegel (2009) feels that the CEOs

deserve most of the blame for the financial crisis. When the major

investment banks such as Lehman Brothers and Bear Stearns were

partnerships, they were much more conservative since they were

risking their own money. Once they became public companies, they

did not mind taking on huge amounts of risk since they were no

longer risking their own wealth. In effect, they were using other

people’s money to become super wealthy.

Cohan (2008) feels that compensation, which consumes approxi-

mately 50% of generated revenues, is far too high. It is a myth

that these high salaries and bonuses are needed to keep good

executives from leaving. There was a time that CEOs earned about

30 to 40 times more than the ordinary worker. Recently, that ratio

at large companies has been 344 to 1 [Kristof (2008)]. For those

who believe that large bonuses lead to improved performance,

research by Dan Ariely indicates the opposite. Bonuses cost the

company more and lead to worse performance [Ariely (2008)].

The reason given by Ariely is that the stress caused by trying to

win the big bonus overwhelms any ability of the bonus to motivate

performance.

It appears that the method of compensation used by Wall Street

firms was a distorted incentive that did not improve performance.

Rather, it encouraged firms to take on huge amount of risk — risk

that eventually destroyed many of these firms. Amy Borrus, dep-

uty director at the Council of Institutional Investors, asserts that

52 – The journal of financial transformation

lessons from the global financial meltdown of 2008

“poorly structured pay packages encouraged the get-rich-quick

mentality and overly risky behavior that helped bring financial mar-

kets to their knees and wiped out profits at so many companies”

[Morgenson (2009)].

To make matters even worse, the public has learned that Merrill

Lynch rushed U.S.$4 billion in end-of-year bonuses to top manage-

ment a few days before the company was taken over by Bank of

America. These bonuses were paid out while the company was

reporting a fourth quarter loss of more than U.S.$15 billion. Clearly,

the bonuses are not rewards for a job well done. John Thain,

former CEO of Merrill Lynch, argued that “if you don’t pay your

best people, you will destroy your franchise.” The massive losses

suffered by Merrill Lynch forced Bank of America to ask for addi-

tional billions in bailout money from the government on top of the

U.S.$25 billion they had already been promised. Thain also agreed

to reimburse Bank of America for the U.S.$1.2 million renovation of

his office a year earlier [LA Times (2009)]. The renovation included

area rugs for U.S.$131,000, chairs for U.S.$87,000, a U.S.$13,000

chandelier, and an antique commode for U.S.$35,000 [Erman

(2009)]. Citigroup was pressured by the Obama administration to

cancel a deal to purchase a U.S.$50 million new French corporate

jet. Citigroup was castigated by the media and called Citiboob by

the New York Post for this planned purchase. Citigroup has lost bil-

lions of dollars, fired 52,000 employees, and received U.S.$45 bil-

lion in government bailout money, yet they were ready to purchase

this luxury for its executives [Keil and Bennett (2009)].

The year 2008 was a time in which Wall Street firms lost billions.

The country was surprised to learn that this did not stop these

firms from giving out bonuses totaling U.S.$18.4 billion — sixth larg-

est amount ever. Apparently, bonuses are also doled out for a job

poorly done. President Obama declared these bonuses “shameful”

and “the height of irresponsibility.” Vice President Biden stated: “I’d

like to throw these guys in the brig. They’re thinking the same old

thing that got us here, greed. They’re thinking, ‘Take care of me’”

[Stolberg and Labaton (2009)].

What has become clear to the public is that Wall Street just does

not get it. The public and the media see the Wall Street executives

who are indifferent to what they caused. They still feel that they

deserve enormous amounts of money even if their firms have shed

thousands of jobs and have nearly destroyed the financial system.

Ira Kay, an executive consultant with Watson Wyatt, feels that you

can make a strong case that the Wall Street culture (some refer

it to the “eat what you kill” mentality) as well as the bonuses that

resulted from this kind of selfish culture have contributed greatly

to the current financial meltdown [Nocera (2009b)].

There is no question that executive compensation is going to be

scrutinized very carefully in the future, even by boards that were

far too chummy with CEOs. There is talk of passing legislation that

will force CEOs to return past pay — this is known as a “clawback” —

in cases where compensation was based on profits that turned out

to be illusory. Frederick E. Rowe, a founder of Investors for Director

Accountability, feels that “there is a fine line that separates fair

compensation from stealing from shareholders. When manage-

ments ignore that line or can’t see it, then hell, yes, they should

be required to give the money back” [Morgenson (2009)]. There is

even talk of “clawing back” executives’ pensions in cases where the

profit earned by the firm turned out to be imaginary.

Models have only limited valueModels are representations of reality. They will not work if conditions

have changed dramatically. They certainly will not work if compensa-

tion is tied to risk taking and executives insist that employees take

risks to enrich themselves. As noted above, executives were doing

everything possible to show stellar performance — even if it meant

taking on huge amounts of risk — since they wanted a fat yearly

bonus. Whenever there are conflicts of interest, poor decisions are

likely to be made. Models rely on interpretation by people. If the

employees interpreting the model are afraid of losing their jobs, they

will construe the models in a way that pleases top management.

As noted above, Fannie Mae decided to expand into riskier mort-

gages in the year 2000. They believed that their sophisticated mod-

els would allow them to purchase riskier mortgages and that these

computer models would protect them from the increased risk. After

all, Fannie Mae would charge higher fees for risky mortgages. The

mortgages, though, became super risky because of very low down

payments and unverified income on the part of the borrower. It is

doubtful that any model could anticipate that millions of homeown-

ers would find it easy to walk away from their mortgage. In the past,

relatively few people defaulted on a mortgage because they had to

make a substantial down payment [Duhigg (2008)].

Lewis and Einhorn (2009) feel that the rating agencies did not

measure risk properly. American financial institutions took on a

great deal more risk without having to face a downgrading of their

securities. This was probably due to the fact that the credit rating

agencies were making money on the risky securities being created

by these very same financial institutions. Companies such as AIG,

Fannie Mae, Freddie Mac, GE, and MBIA which guarantee municipal

bonds kept their AAA rating for a very long time. Lewis and Einhorn

(2009) make the point that MBIA deserved its AAA rating in 1990

when it insured municipal bonds and had U.S.$931 million in equity

and U.S.$200 million of debt. By the year 2006, MBIA was insuring

CDOs (collateralized debt obligations) which are extremely risky;

the company had U.S.$26.2 billion in debt and only U.S.$7.2 billion

in equity. It kept its AAA rating for quite a while until it was obvious

to everyone that MBIA was no longer a secure firm. The models

that were being used by the credit rating agencies may not have

53

lessons from the global financial meltdown of 2008

worked quite well. However, no one wanted to kill the goose that

was producing so much profit for the credit rating agencies. Indeed,

the agencies, as noted above, were quite pleased with all the risky

securities they were paid to evaluate.

Back in 2004, as noted above, the investment banks wanted the

SEC to give them an exemption from a regulation that limited the

amount of debt — known as the net capital rule — they could have on

their balance sheets. By increasing the amount of debt, they would

be able to invest in the “opaque world of mortgage-backed securi-

ties” and credit default swaps (CDSs). The SEC agreed and decided

to allow the investment banks to monitor their own riskiness by

using computer models to analyze the riskiness of various securi-

ties [Labaton (2008)]. One individual was opposed to the changes

and said the computer models would not work in periods of “severe

market turbulence.” He pointed out that computer models did not

protect Long-Term Capital Management from collapse [Labaton

(2008)]. Of course, no one listened to him.

One measure of risk that was very popular with the financial engi-

neers was VaR (Value-at-Risk). This is a short-term measure of risk

(daily, weekly, or for a few weeks) expressed in dollars and is one

number and is based on the normal distribution. A weekly VaR of

U.S.$100 million indicates that, for that week, there is a 99% prob-

ability that the portfolio will not lose more than U.S.$100 million.

Of course, this is based on what has happened in the past. Alan

Greenspan noted the following when trying to explain what went

wrong during the financial meltdown: “The whole intellectual edifice,

however, collapsed in the summer of last year because the data input

into the risk-management models generally covered only the past

two decades, a period of euphoria. Had instead the models been fit-

ted more appropriately to historic periods of stress, capital require-

ments would have been much higher and the financial world would be

in far better shape today, in my judgment [Nocera (2009a)].

The problem with VaR, and similar measures, is that it ignores what

can happen 1% of the time. Taleb (2007) considers VaR a dishonest

measure since it does not measure risks that come out of nowhere,

the so called “black swan.” It provides a false sense of security

because over the long run, things that have low probabilities of

occurrence do happen. Indeed, Long Term Capital Management col-

lapsed because of a black swan, unexpected financial crises in Asia

and Russia [Nocera (2009a)].

conclusionPhelps (2009) observes: “Whether in Enron’s creative accounting,

the packaging of high-risk subprime mortgages as top-grade col-

laterized debt obligations, or Bernard Madoff’s $50 billion scam

operation, the recent riot of capitalist irresponsibility has shattered

the fantasy that the free market, left to its own devices, will pro-

duce rationality and prosperity.”

The key lessons one can learn from the global financial crisis of

2008 are the following:

some self-interest is good —■■ capitalism requires it and human

beings are programmed that way. However, voracious self-

interest with total disregard for everyone else is not good for

society.

Too much regulation may be a bad thing — ■■ it stifles innovation

and is not good for business or society. On the other hand, too

little regulation is even more dangerous for society and busi-

ness. China is discovering this with their recent problem with

tainted milk which killed several children. Apparently, executives

at several dairy companies sold dairy products adulterated with

melamine, a toxic chemical, in order to make the protein count

appear higher than it actually was. One executive, who has been

sentenced to death, was convicted of selling 600 tons of “protein

powder” contaminated with melamine to dairy firms [Barboza

(2009)]. The poor conditions of the Peanut Corporation of

America plant in Blakely, Georgia which resulted in a salmonella

outbreak also demonstrated what can happen when there is too

little regulation. ConAgra, manufacturer of Peter Pan peanut

butter, had similar problems and upgraded its safety proce-

dures after a salmonella outbreak in 2007. Government officials

admitted that there were not enough agents (60) to monitor

thousands of food businesses. It is clear that the plant inspec-

tors missed many serious and chronic problems such as a leaky

roof in the Blakely plant [Moss (2009)]. Coming up with the right

amount of regulation is not impossible. What is needed is suf-

ficient regulation to discourage huge risk taking but enough to

encourage innovation [Knowledge@Wharton (2009)].

incentives are an effective way of motivating people — ■■

However, bonuses may not be the appropriate kind of incentive

for executives. In fact, bonuses can help destroy firms if they

encourage executives to focus on short-term profits rather than

the long-term health of a company.

Mathematical models —■■ might provide some insights but ulti-

mately it is people who have to interpret them. They can be used

to justify almost anything and can cause as much harm as good.

The global financial crisis would not have occurred if executives

were truly ethical. There is no question that lack of ethics played

a significant role in the meltdown. A large number of people knew

that the mortgages they were dealing with were toxic. It does not

take a great financial expert to know that mortgages with no down

payments given to people with no income is extremely foolhardy.

The excuse that they believed that housing prices would continue to

keep going up (at one point houses were doubling in value every six

or seven years) is not credible and indeed it is not even a legitimate

justification for this behavior. The truth is that as long as there

was money to be made virtually no one said anything. To hide how

risky these toxic mortgages were, they turned them into securities.

To make matters worse, mortgage brokers were encouraged to do

54 – The journal of financial transformation

lessons from the global financial meltdown of 2008

everything possible to get people to take out mortgages. The rating

agencies were making money on rating these securities so they did

not do their job. All the parties figured someone else would end up

holding the bag.

One thing is clear: free markets do not work well unless there is

accountability, responsibility, ethics, and transparency. Predatory

capitalism that disregards the concern for others and is based pure-

ly on self-interest may even be more dangerous than communism.

ReferencesAlvey, J. E., 1999, “Why spirituality deserves a central place in liberal education,” •

Journal of Markets and Morality, 2:1, 53-73

Andrews, E. L., 2008, “Greenspan concedes error on regulation,” New York Times, •

October 24, B1, B6

Ariely, D., 2008, “What’s the value of a big bonus?” New York Times, OP-ED, November •

20, A43

Barboza, D., 2009, “China plans to execute 2 in scandal over milk,” New York Times, •

January 23, A5

Becker, J., S. G. Stolberg, and S. Labaton, 2008, “White House philosophy stoked mort-•

gage bonfire,” New York Times, December 21, 1, 36-37

Burrows, P., 2007, “He’s making hay as CEOs squirm,” Business Week, January 15, •

64-65

Cohan, W. D., 2008, “Our risk, Wall Street’s reward,” New York Times, Week in Review, •

November 16, 13

Cohen, P., 2007, “In economics departments, a growing will to debate fundamental •

assumptions,” New York Times, July 11, B6

Cox, C., 2008, “Swapping secrecy for transparency,” New York Times, Week in Review, •

October 19, 12

Curry, T., and L. Shibut, 2000, “The cost of the savings and loan crisis: Truth and •

consequences,” FDIC Banking Review, 13(2), retrieved December 29, 2008 from http://

www.fdic.gov/bank/analytical/banking/2000dec/brv13n2_2.pdf

Duhigg, C., 2008, “Pressured to take more risk, Fannie reached a tipping point,” New •

York Times, October 5, 1, 34

Erman, B., 2009, “Big bonuses, costly commodes: Thain out at B of A,” Globeandmail.•

com, January 23. Retrieved January 28, 2009 from http://www.theglobeandmail.com/

servlet/story/LAC.20090123.RTHAIN23/TPStory/Business

Friedman, T. L., 2008, “The great unraveling.” New York Times, OP-ED, December 17, •

A39

Gandel, S., 2008, “The Madoff fraud: How culpable were the auditors?” Time, •

December 17, retrieved December 29, 2008 from http://www.time.com/time/print-

out/0,8816,1867092,00.html#

Goodman, P. S., 2008, “Taking a hard look at a Greenspan legacy,” New York Times, •

October 9, A1, A29

Goodman, P. S., and G. Morgenson, 2008, “By saying yes, WaMu built empire on shaky •

loans,” New York Times, December 28, 1, 22

Howard, G. S., 1997, “The tragedy of maximization,” Ecopsychology Online, October, •

retrieved June 25, 2007 from http://ecopsychology.athabascau.ca/1097/index.

htm#politics

Keil, J.G. and C. Bennett, 2009, “Citigroup cancels plans to buy $50m jet,” New •

York Post, January 27, retrieved January 28, 2009 from http://www.nypost.

com/seven/01272009/news/nationalnews/citigroup_cancels_plans_to_buy_50m_

jet_152270.htm

Kirkland, R., 2006, “The real CEO pay problem,” Fortune, June 30. Retrieved •

July 29, 2008 from http://money.cnn.com/magazines/fortune/fortune_

archive/2006/07/10/8380799/index.htm

Knowledge@Wharton, 2009, “Getting it right: making the most of an opportunity to •

update market regulation,” January 21, retrieved January 22, 2009 from http://knowl-

edge.wharton.upenn.edu/article.cfm?articleid=2145

Kristof, N., 2008, “Need a job? $17,000 an hour. No success required,” New York •

Times, OP-ED, September 18, A29

Kroft, S., 2008, “The bet that blew up Wall Street,” 60 Minutes, October 26, retrieved •

January 4, 2008 from http://www.cbsnews.com/stories/2008/10/26/60minutes/

main4546199.shtml

Labaton, S., 2008, “Agency’s ’04 rule let banks pile up new debt and risk,” New York •

Times, October 3, A1, A23

Landy, H., 2008, “Credit default swaps oversight nears,” Washington Post, November •

15, D03

Lewis, M., and D. Einhorn, 2009, “The end of the financial world as we know it,” New •

York Times, Week in Review, January 4, 9-10

Lichtblau, E., 2008, “Federal cases of stock fraud drop sharply” New York Times, •

December 25, A1

Lipton, E., and R. Hernandez, 2008, “A champion of Wall Street reaps benefits,” New •

York Times, December 14, 1, 36

Lorsch, J. W., L. Berlowitz, and A. Zelleke, 2005, Restoring trust in American business, •

MIT Press, Cambridge, MA

Los Angeles Times, 2009, “John Thain defends Merrill Lynch bonuses amid Bank of •

America Takeover,” January 27, latimes.com, retrieved January 28, 2009 from http://

www.latimes.com/business/la-fi-thain27-2009jan27,0,476991.story

Lowenstein, R., 2008, “Long-Term Capital Management: It’s a short term memory,” •

International Herald Tribune, September 7. Retrieved December 29, 2008 from http://

www.iht.com/articles/2008/09/07/business/07ltcm.php

Monaghan, P., 2003, “Taking on ‘rational man’,” Chronicle of Higher Education, •

January 24, A12-A15

Morgenson, G., 2008a, “Debt watchdogs: Tamed or caught napping,” New York Times, •

December 7, 1, 40

Morgenson, G., 2008b, “Behind insurer’s crisis, blind eye to a web of risk,” New York •

Times, September 28, 1, 28

Morgenson, G., 2009, “After huge losses, a move to reclaim executives’ pay,” Sunday •

Business, New York Times, February 22, 1, 7

Moss, M., 2009, “Peanut case shows holes in food safety net,” New York Times, •

February 9, A1, A12

Nocera, J., 2008, “How India avoided a crisis,” New York Times, December 20, B1, B8•

Nocera, J., 2009a, “Risk mismanagement,” New York Times Magazine, January 4, •

23-50

Nocera, J., 2009b, “Getting theirs: It’s not the bonus money. It’s the principle,” New •

York Times, Business Day, January 31, 1, 4

Norris, F., 2009, “Failing upward at the Fed,” New York Times, Business Day, February •

27, B1, B4

Pearlstein, S., 2007, “No money down falls flat,” Washington Post, March 14, D01•

Phelps, C., 2009, “Beyond economic revival: which New Deal approaches are worth •

emulating?” Chronicle Review, February 13, B10-B12

Philips, M., 2008, “The monster that ate Wall Street,” Newsweek, October 6, retrieved •

on January 4, 2009 from http://www.newsweek.com/id/161199

Pitelis, C., 2002, “On economics and business ethics,” Business Ethics: A European •

Review, 11(2), 111-118

Robinson, J., 1977, “Morality and economics” Commencement Address, retrieved July •

9, 2007 from Economist’s View http://economistsview.typepad.com/economists-

view/2007/07/morality-and-ec.html

Roosevelt, F. D., 1937, “Second Inaugural Address,” retrieved October 25, 2008 from •

Bartleby.com http://www.bartleby.com/124/pres50.html

Satow, J. 2008, “Default swaps may be next in credit crisis,” New York Sun, September •

22, http://www.nysun.com/business/default-swaps-may-be-next-in-credit-crisis/86312/

Siegel, J., 2009, “Lesson one: what really lies behind the financial crisis,” Knowledge@•

Wharton, January 21, retrieved January 22, 2009 from: http://knowledge.wharton.

upenn.edu/article/2148.cfm

Skidelsky, R., 2008, “The remedist,” New York Times Magazine, December 14, 21-22•

Stolberg, S. G., and S. Labaton, 2009, “Banker bonuses are ‘shameful,’ Obama •

declares,” New York Times, January 30, A1, A20

Story, L., 2008, “On Wall Street, bonuses, not profits, were real,” New York Times, •

December 18, A1, A35

Suskind, R., 2008, “The crisis last time,” New York Times, OP-ED, September 25, A29•

Taleb, N. N., 2007, The black swan: the impact of the highly improbable, Random •

House, New York

Articles

55

What can leaders learn from cybernetics? Toward a strategic framework for managing complexity and risk in socio-technical systemsAlberto Gandolfi

Professor, Department of Management and Social Sciences, SUPSI University of Applied Science of

Southern Switzerland

AbstractHow can leaders cope with the seemingly inexorable increase in

complexity of their world? Based on the cybernetic “law of requi-

site variety,” formulated almost fifty years ago by W. R. Ashby, we

propose a pragmatic framework that might help strategic decision

makers to face and manage complexity and risk in organizational,

political, and social systems. The law of requisite variety is a funda-

mental hypothesis in the general theory of regulation and defines

an upper limit to the controlling ability of a system, based on its

“variety” (complexity) level. The strategic framework discussed

here assumes that complexity breeds systemic risks and suggests

three mutually complementary approaches to managing complexity

at a strategic level: 1) where possible, reduce complexity in the sys-

tem to be controlled; 2) effectively manage residual complexity; and

3) increase complexity of the controlling system. Each approach is

then discussed, and practical examples provided.

56 – The journal of financial transformation

What can leaders learn from cybernetics? Toward a strategic framework for managing complexity and risk in socio-technical systems

Decision makers in literally every area are facing an increasing level

of complexity. This inexorable trend in turn has lead to increased

difficulty in managing organizational, technological, political, and

socio-economical systems and their related risks. Everything seems

to become more complex, and therefore more difficult to manage

and control: markets and business opportunities, corporate gover-

nance, projects, fiscal systems, technology, demands from all kinds

of stakeholders, management systems, and supply chains.

Partly as a response to this important trend, in the last three

decades the interdisciplinary study of the structure and the behav-

ior of complex systems — to which we will refer here as complexity

science — has grown considerably and has brought relevant inputs

to many scientific fields, such as biochemistry, ecology, physics,

chemistry, genetics, and biology [see for example Kauffman (1993)

for genetics and evolutionary biology or Pimm (1991) for ecology].

Many scholars have then tried to apply the instruments and analyti-

cal concepts developed by complexity science in order to improve

the control of socio-technical systems, including organizational,

economical, technological, social, and political systems [Iansiti

and Levien (2004), Kelly (1994), Senge (1992), Malik (1984)]. Even

though this transfer from “hard” to “soft” sciences has brought

some interesting insight for the latter, a general model addressing

complexity management in socio-technical systems is still lacking.

What we still need is a more pragmatic approach, enabling manag-

ers, politicians, and other decision makers in the real world to better

understand and actively influence the behavior and control risks of

the systems they are supposed to control. Our research has led to

a conceptual model for complexity management in social and socio-

technical systems, which will be outlined in this paper. The current

model should not be considered as complete, but as work in prog-

ress which will be further enhanced and modified over the coming

years, focusing primarily not on theoretical elegance, but rather at

usefulness in a real decision making environment.

complexity and riskBefore we can proceed to explain our framework, it is necessary

to clarify what we mean by the term “complexity”. In this paper

we will adopt a relatively simple, management-oriented definition,

viewing complexity from the perspective of the decision makers

who are supposed to plan, manage, and control real socio-technical

systems, surrounded by real environments. Although it is not in the

scope of this paper to discuss the different meanings of “complex-

ity,” for our purposes two complementary dimensions of complexity

become particularly relevant [Casti (1994), Malik (1984)] [see also

Sommer and Loch (2004) for applications in project management,

and Perrow (1999) for applications in risk management]:

The number of states of the systems (system variety) —■■ what

does cybernetics mean by the term “state” of a system? In short,

it is a particular configuration of the different elements of a

system, exhibiting a particular behavior [Ashby (1956)]. System

variety depends, therefore, on the total number of different ele-

ments in the systems and on the number of different internal

(behavior relevant) configurations of each element. For example,

in a productive system, a status could be a particular combina-

tion of all relevant system elements: machines, raw material,

components, people, schedule, data, procedures, and orders.

The interactions among elements (system dynamics) —■■ lead-

ing to discontinuous and nonlinear changes in system behavior —

from a decision-making point of view, the perceived complexity

increases by increasing the number of relationships between ele-

ments, and by increasing nonlinearity and opacity (i.e., nontrans-

parency) of these relationships. Opacity occurs when an observer

(in this case, the decision-maker) is no longer able to determine

clear and straightforward cause-effect relationships in the sys-

tem dynamics. Nonlinearity occurs when the system changes in

an erratic, discontinuous way or when the size of change is not

correlated with the size of the input that caused it.

We can note that those two aspects represent, in a way, a concep-

tual extension of the NK-model for evolutionary fitness landscapes

[Kauffman (1993)], where N is the number of elements of the land-

scape and K is the total number of relationships between elements.

Following these preliminary considerations, in the remainder of this

paper we will adopt this twofold definition of complexity.

A very important point is the relationship between complexity and

risk. Results from several research areas have consistently dem-

onstrated that an increase in complexity breeds more, and more

diverse, potential risks for a socio-technical system, as also sug-

gested by our daily experiences [Johnson (2006)]. The dynamics of

complex systems show a set of well-known features, such as feed-

back loops [Johnson (2006), Sterman (2000), Forrester (1971)],

nonlinear behavior (phase transitions) [Perrow (1999), Sterman

(2000), Ball (2005), Kauffman (1993)], network effects (network

dynamics) [Barabasi (2002)], emerging properties (system effects,

self-organization) [Johnson (2006), Holland (2000), Kauffman

(1993), Ashby (1956)], delayed effects [Jervis (1997), Sterman

(2000)], and indirect, unintended consequences (side effects)

[Jervis (1997), Sterman (2000), Tenner (1997)].

As Jervis (1997, p.29) summarized it: “many crucial effects are

delayed and indirect; the relations between two actors often are

determined by each one’s relation with others; interactions are

central and cannot be understood by additive operations; many

outcomes are unintended; regulation is difficult.”

Casti (1994) suggests that complexity generates risks for manage-

ment and produces system ‘surprises.’ Such surprises make the

system’s behavior opaque, unstable, counterintuitive, sometimes

57

What can leaders learn from cybernetics? Toward a strategic framework for managing complexity and risk in socio-technical systems

chaotic, and eventually unpredictable [Forrester (1971)]. Under such

circumstances, control, regulation, and, consequently, effective risk

management are extremely difficult or even impossible to achieve.

Errors and failures, and in the worst cases catastrophes, are the

often unavoidable consequences [Choo (2008), Perrow (1999)].

Ashby’s seminal hypothesis: “only variety can destroy variety”In the 1960s, Ashby proposed a law of ‘requisite variety,’ which

states that “only variety can destroy variety” [Ashby (1956)], where

variety means the number of different states (or behaviors) the

system can display. To fully understand this metaphor, one has

to consider the cybernetic definition of control, as suggested by

Ashby: to control a system means to reduce the variety of its poten-

tial behaviors. One controls a system if he/she is able to limit the

number of different states the system may exhibit. In the extreme

case, the system is entirely under control, such that it may exhibit

only one particular behavior (or status), the one the controlling

system has chosen [Hare (1967)]. An alternative way of understand-

ing the law of requisite variety, sometimes called Cybernetics I [De

Leeuw (1982)], is to view control as the ability to reduce or to block

the flow of variety from environmental disturbances, thus enabling

an internal “physiological” stability and the survival of the system

[Ashby (1956)].

In this paper, we will assume that the cybernetic concept of variety

might be substituted by the more broad (and bold) term of ‘com-

plexity’ [Jackson (2000), Malik (1984), Van Court (1967)]. In fact,

we believe the term complexity can better capture the rich nature

of real socio-technical systems, where variety is only one of the

aspects of complexity [Johnson (2006)].

The three leversOn the basis of the conceptual schema provided by the law of

requisite variety, our research led to the elaboration of a strategic

framework, with a pragmatic orientation. Its objective, and the

criterion which we will adopt to measure its effectiveness, is to

help decision-makers out in the field to manage their complex work

environment. The target audience for the strategic framework is

those people faced with problems, environments, and situations

of increasing complexity, such as board members, executives of

private or publicly owned companies, consultants, politicians, plan-

ners, and project managers.

Van Court, in one of the rare comments and conceptual develop-

ments of Ashby’s law of requisite variety, noted that “there are only

two major ways to adjust the ability of the analyst (or controller)

to the requirements of the system to be controlled: (1) increase the

variety of the controller, or (2) reduce the variety of the system to

be controlled” [Van Court (1967)]. This issue was initially proposed

by Stafford Beer with his seminal concept of ‘variety engineering’

[Beer (1979, 1981)]. Beer argued that management can balance its

‘variety (complexity) equation’ by using both ‘variety reducers’

for filtering out environmental variety and ‘variety amplifiers’ for

increasing its own variety vis-à-vis its environment.

Basically, our model extends this approach, suggesting three meth-

ods, or strategic levers, for the decision maker to better manage, and

survive, complexity. Assuming that a controlling system C wants to

control and manage a system s, the three strategic levers are:

first lever —■■ reduce complexity in the controlled system.

second lever —■■ manage the remaining complexity.

Third lever —■■ increase complexity of the controlling system C,

and understand and accept complexity.

first lever — reduce complexity in the controlled system cThe first lever suggested by our strategic framework aims to

reduce the complexity of the controlled system — where and when

such a reduction is possible, realistic, and adequate. One can adopt

different approaches in order to reduce the overall complexity of a

system, such as:

complexity transfer — in this first approach a certain ‘amount

of complexity’ is transferred elsewhere, outside the system to be

controlled.

Example 1 —■■ outsourcing decisions usually allow for transferring

a considerable amount of organizational and technical complex-

ity of a process to a specialized partner, outside the company.

Here, it is interesting to note that the overall net management

burden and costs are often reduced, due to the specialization

of the outsourcing partner in a few specific processes, such as

transportation of goods, IT management, or market research.

The result is that the complexity decrease in the outsourcing

company is greater than the complexity increase in the special-

ized partner.

Example 2 —■■ the introduction of traffic circles leads to a sub-

stantial reduction in the complexity of local traffic management

by public authorities, usually achieved by traffic lights. In this

very successful case, complexity of coordination has been effec-

tively distributed to thousands of single car drivers, each of them

managing his/her own space- and time-limited complexity when

approaching the crossing.

Example 3 —■■ from an industrial point of view, Alberti has sug-

gested a general concept for the reduction of complexity in pro-

duction planning and production management [Alberti (1996)],

which includes the transfer of complexity from department and

plant managers to other elements of the manufacturing system,

such as product structure, production infrastructure (i.e., factory

layout), workers, and suppliers.

58 – The journal of financial transformation

What can leaders learn from cybernetics? Toward a strategic framework for managing complexity and risk in socio-technical systems

simplification of system elements (reduction of both number and

variety of elements) — in this approach, complexity is reduced by

standardizing and rationalizing the individual elements of a socio-

technical system. The result is a reduction in the overall number

and variety of elements of the system, and hence of system variety

[Gottfredson and Aspinall (2005), Schnedl and Schweizer (2000)].

Example 1 —■■ some airlines have dramatically reduced the number

of different planes in their fleet. For example, Southwest Airlines

ended up having only one kind of plane [Freiberg and Freiberg

(1996)]. This leads to a reduced complexity in terms of training

pilots, crew and technical personnel, maintenance and safety

management, and purchasing of new planes and spare parts.

Example 2 —■■ standardizing the documents which are created

and circulated in an organization is also a complexity reducing

measure. For example, a company could decide that all sales reps

use only one standard format to register new sales orders. For

them and for their colleagues from all other departments this

would mean a less demanding task, reduced error probability,

and a reduction in training time. By the same token, imposing a

standardization of the different softwares used by an organiza-

tion will lead to a significant complexity reduction, particularly in

the IT department and in the help desk.

simplification of the interaction between elements (reduction

of the dynamical complexity) — in complex systems the relation-

ships between elements are usually numerous, nonlinear, and non-

transparent. This approach focuses on moving the system in the

opposite direction, i.e., reducing the number of interactions, making

them more linear or more transparent for the decision maker. The

objective is to obtain a system with lower overall dynamic complex-

ity [Gottfredson and Aspinall (2005)].

Example —■■ a growing number of project leaders have adopted

web-based communication platforms (often with a “pull-philos-

ophy” for getting the data) to better manage communication

and collaboration in complex projects, with many stakeholders

involved. Such a tool may help to organize and simplify commu-

nication flow between project members, usually characterized

by infinite and disturbing numbers of emails, the coexistence of

different versions of the same document, and frequent mistakes.

Here complexity reduction is obtained by reducing the number

of interactions in the communication network between project

members, that is, moving from a n:n-communication topology

(everybody interacts with everybody) to a n:1 topology (every

project member interacts within the communication platform.)

second lever — managing residual complexityThe second approach suggested by our framework concentrates

on managing the remaining complexity in the controlled system.

Obviously, the potential for reducing complexity in the controlled

system has its limits, due to either physical-structural boundaries or

limitations in the actual power of the controlling system (in this case

the decision maker managing a complex socio-technical system).

In the last decades several scientific disciplines have developed

supporting tools and concepts, which can be extremely helpful in

enabling decision makers to better manage existing socio-technical

complexities. We will focus here on two of such tools and concepts:

using decision support software and segmenting complexity levels

Decision support software — we refer here to software tools which

enable the effective simulation and optimization of the behavior

of complex systems. One should distinguish between (a) general

purpose, generic applications and (b) task- or system-specific appli-

cations (i.e., for logistics, transportation, production scheduling).

Basically, even many functionalities of the traditional Enterprise

Resource Planning (ERP) applications aim to reduce complexity

for decision-makers in organizations, enabling a centralized, non-

redundant, fully-integrated data management [Davenport and

Harris (2005)].

segmenting complexity levels — another approach tries to seg-

ment (or segregate) complexity, that is, to set apart different levels

of complexity in different processes, with specific rules, procedures,

resources, and strategies. The segmentation of process complexity

is one of the basic tenets of Business Process Reengineering, BPR

[Hammer and Champy (1993)]. In the case of many insurance com-

panies, which have subdivided the process of evaluation of requests

for compensation according to the complexity of the individual

cases, simple cases follow a more simplified procedure, with less

informed collaborators — while more complex cases are managed

by a team of specialists through a more elaborate process. This per-

mits allocation of complex tasks to the best (and most expensive)

specialized resources, while avoiding the overburdening of these

resources with numerous simple and non-critical cases.

Third lever — increase the complexity of the decision makerAfter having reduced the complexity of the controlled system,

and having better managed the remaining complexity of system,

a third approach suggests increasing the variety (or complexity)

of the controlling system, the decision making body. What does

“increasing complexity of the decision maker” mean? According to

our working definition of complexity, to increase complexity means

here to increase the number of potential perspectives, opinions,

decisions, and the variety of approaches for selecting among them

the best, most effective one. This translates into two complementa-

ry options: both (a) increase the number or the (internal) complex-

ity of elements of the decision making system, and (b) improve the

number and variety of the interactions among these elements. In

real life — say, a manager who has to lead a complex organization —

we can envisage the following approaches to reach this goal.

59

What can leaders learn from cybernetics? Toward a strategic framework for managing complexity and risk in socio-technical systems

increase the “decision complexity” of human resources —

through different training and development measures, one has

to enable single employees and managers to think and act in a

more complex way: considering more potential frames and solu-

tion ideas when solving a problem; understanding the relationship

between activities and objectives of different processes, teams

and departments; communicating and collaborating with others in

a more effective way; acting as an interconnected element for the

long term interests of the whole system; and deliberately delaying

judgment and remaining open to alternative systems [McKenzie et

al. (2009)]. For example, decision makers should understand and

accept complexity, the logic behind the structure and behavior of

complex socio-technical systems, and its profound consequences

for the management of those systems. One essential conceptual

frame for understanding complexity is System Thinking [Sterman

(2000), Vester (1999), Senge (1990)]. Basically, System Thinking

is a holistic way of seeing and interpreting the world, in terms of

systems, made up of networks of interconnected elements. Often

the emergent behavior of a system is determined by the relation-

ship between elements, rather than by the behavior of the single

elements. System Thinking has been implicitly adopted and inte-

grated in strategic planning tools such as the well-known Balanced

Scorecard, BSC [Kaplan and Norton (2001)], developed in order to

escape from the limits of traditional strategic planning which was

too focused on financial indicators — and hence not systemic. BSC,

on the contrary, recognizes the role of many other dimensions and

objectives (market, internal, learning and growth objectives) for the

long-term success of an organization.

increase variety of the whole decision system [schwaninger

(2001)] — a first, simple step is to go from individual to collective

decision-making [Van Buuren and Edelenbos (2005)]. Further,

one can increase complexity by diversifying the decision-making

group from different perspectives: diversity of age, professional

or cultural background, thinking style, race, sex, knowledge of

the problem, role in the organization, experiences, and interests.

A more diverse team is potentially able to generate more diverse

mental states, solutions, or decisions [Page (2007)]. For example,

some innovative companies have made their research and develop-

ment teams more diverse by integrating “outsiders,” that is, people

with little or no knowledge (or a very naïve knowledge) of a specific

technological product, like children, older people, housewives, and

artists. At the opposite end, there are clues which show that some

big decision mistakes might be traced back to an insufficient vari-

ety of decision bodies (i.e., the board of directors). Schütz (1999)

describes the case of Union Bank of Switzerland (UBS) in the 1990s.

He shows that the board of directors of UBS was in those years very

homogeneous: members were all men, Swiss, more than 50 years

old, army officers, with an economic or legal background, having

followed a traditional banking career. This complexity (variety) level

was apparently inadequate to cope with an increasing complexity in

the financial market environment. Schütz argues that such a board

could not effectively control or manage some critical and complex

processes of the bank. Eventually, UBS ended up in serious finan-

cial distress, and had to accept a merger with a smaller and less

powerful competitor, SBS (the new bank maintained the name UBS,

although most of the new management came from SBS).

improve knowledge management in the organization — more

knowledge enables access to more diverse mental states and

more dynamic interactions, and thus, potentially, to more complex

decisions. In this way, effective knowledge management can be

a powerful tool for increasing the complexity of decision-making

throughout the organization.

create networks and decide and operate in networks — there is

some evidence that a network can generate a higher decision and

behavior complexity than other topological configurations, such

as hierarchies or chains, or a single decision maker [Surowiecki

(2004), Vester (1999), Kelly (1994), Peters (1992)]. One of the most

effective ways of dealing with very complex problems is to create

a network of different and complementary competencies, as in

the case of scientific research, where the trend is to build large

research networks (competence centers) on specific complex top-

ics. At the Civico City Hospital of Lugano (Switzerland), a medical

competence centre for diagnosis and therapy of breast cancer was

established in 2005, in order to integrate into a virtual network all

the involved specialists (oncologists, radiologists, surgeons, phar-

macologists, nurses, psychologists, and researchers). One of the

most important features of such networks is a massive improve-

ment of communication between the nodes, which leads to an

increase in relationship complexity of the system [Evans and Wolf

(2005)]. The effectiveness of networked decision making can be

further improved by implementing methodologies and approaches

designed to foster cohesion, communication, and synergy in larger

groups of individuals, as defined for example by Stafford Beer with

his Team Syntegrity Model [Schwaninger (2001)].

conclusion Complexity and risks of markets, legal systems, technological land-

scapes, environmental, and social issues, to mention only a few,

will further increase in the next decades. Drawing from the insight-

ful intuition of R.W. Ashby, this paper tries to outline a pragmatic

strategic framework to managing complexity in socio-technical

systems, as summarized in Figure 1. The model should be further

improved in the next few years; very important progress is to be

expected if researchers eventually become able to measure com-

plexity levels or at least to quantify changes in the complexity levels

of socio-technical systems.

We believe that the application of concepts and insights of complex-

ity science to management still has a huge development potential.

60 – The journal of financial transformation

What can leaders learn from cybernetics? Toward a strategic framework for managing complexity and risk in socio-technical systems

We must narrow down the gap between the two worlds: complexity

science and system thinking on the one hand and “down-to-earth,”

empirical management reality on the other. We also believe that the

model discussed in this paper can be a humble contribution to this

fascinating challenge on the edge of these two worlds.

ReferencesAlberti, G., 1996, Erarbeitung von optimalen Produktionsinfrastrukturen und Planungs- •

und steuerungssystemen, Dissertation Nr. 11841, ETH, Zürich

Ashby, W. R., 1956, An introduction to cybernetics, Chapman & Hall, London•

Beer, S., 1985, Diagnosing the system, Wiley, New York•

Beer, S., 1981, Brain of the firm, second edition, Wiley, Chichester•

Beer, S. 1979, The heart of enterprise, Wiley, Chichester•

Casti, J., 1994, Complexification, Harper Collins, New York•

Choo, C.W., 2008, “Organizational disasters: why they happen and how they may be •

prevented,” Management Decision, 46:1, 32-45

Davenport, T. H., and J. G. Harris, 2005, “Automated decision making comes of age,” •

Sloan Management Review, Summer, 83-89

De Leeuw A., 1982, “The control paradigm: an integrating systems concept in organi-•

zation theory,” in Ansoff, H. I., A. Bosman, and P. M. Storm (eds.), understanding and

managing strategic change, North Holland, New York

De Leeuw A., and H. Volberda, 1996, “On the concept of flexibility: a dual control per-•

spective,” Omega, International Journal of Management Sciences, 24:2, 121-139

Evans, P., and B. Wolf, 2005, “Collaboration rules,” Harvard Business Review, July-•

August, 96-104

Forrester, J. W., 1971, “Counterintuitive behavior of social systems,” Theory and •

Decision, 2:2, 109-140

Freiberg, K. L., and J. A. Freiberg, 1996, Nuts! Bard Press, New York•

Gandolfi, A., 2004, La Foresta delle decisioni, Casagrande, Bellinzona•

Gottfredson, M., and K. Aspinall, 2005, “Innovation versus complexity: what is too •

much of good thing?” Harvard Business Review, November, 62-71

Hammer, M., and J. Champy, 1993, Reengineering the corporation, Harper Business, •

New York

Holland, J. H. 2000, Emergence – from chaos to order, Oxford University Press, Oxford•

Iansiti, M., and R. Levien, 2004, The keystone advantage, Harvard Business School •

Press, Boston

Jackson, M. C., 2000, Systems approaches to management, Kluwer Academic, New •

York

Jervis, R., 1997, Systems effects, Princeton University Press, Princeton•

Johnson, J., 2006, “Can complexity help us better understanding risk?” Risk •

Management, 8, 227-267

Kauffman, S. A., 1993, The origin of order, Oxford University Press, Oxford•

Kaplan, R. S., and D. P. Norton, 2001, The strategy focused organization, Harvard •

Business Press, Boston

Kelly, K., 1994, Out of control: the new biology of machines, social systems and the •

economic world, Addison-Wesley Reading

Küppers, B. O., 1987, “Die Komplexität des Lebendigen,” in Küppers, B. O., Ordnung aus •

del chaos, Piper, München

Malik, F., 1984, Strategie des managements komplexer systeme, Haupt, Bern•

McKenzie, J., N. Woolf, and C. Van Winkelen, 2009, Cognition in strategic decision mak-•

ing, Management Decision, 47:2, 209-232

Mintzberg, H., 1994, The rise and fall of strategic planning, The Free Press, New York•

Ormerod, P., 1998, Butterfly economics, Faber and Faber, London•

Page, S., 2007, The difference, Princeton University Press, Princeton•

Perrow, C., 1999, Normal accidents, Princeton University Press, Princeton•

Peters, T., 1992, Liberation management, Fawcett Columbine, New York•

Pimm, S. L., 1991, The balance of nature? University of Chicago Press, Chicago•

Schnedl, W., and M. Schweizer, 2000, Dynamic IT management in the 21st century, •

PricewaterhouseCoopers

Schütz, D., 1999, Der fall der UBS, Weltwoche-ABC-Verlag, Zürich•

Schwaninger, M., 2001, “Intelligent organizations: an integrative approach”. System •

Research and Behavioral Science, 18, 137-158

Senge, P. M., 1990, The fifth discipline, Doubleday, New York•

Sommer S. C., and C. H. Loch, 2004, “Selectionism and learning in projects with com-•

plexity and unforeseeable uncertainty,” Management Science, 50:10, 1334-1347

Sterman, J., 2000, Business dynamics, Irwin McGraw-Hill, Boston•

Surowiecki, J., 2004, The wisdom of crowds, Little Brown, London•

Tenner, E., 1997, Why things bite back, Vintage Books, New York•

Van Buuren A., and J. Edelenbos, 2005, “Innovations in the Dutch polder,” Science and •

Society Conference, Liverpool, 11-15 September

Van Court, H., 1967, Systems analysis: a diagnostic approach, Harcourt, Brace & World, •

New York

Vester, F., 1999, Die Kunst vernetzt zu denken, DVA, Stuttgart•

strategic lever some approaches

first lever — reduce complexity in the controlled system C

Complexity transfer

Simplification of system elements (reduction of variety)

Simplification of the interaction between elements (reduction of the dynamical complexity)

second lever — manage remaining complexity

Decision support applications

Segmenting complexity levels

……

Third lever — increase complexity of the controlling system C

Increasing decision complexity of human resources

Increasing variety of the whole decision system

Improving knowledge management

Deciding and operating in networks

Figure 1 – Overview of the strategic framework to better manage complexity in socio-

technical systems – the four strategic levers and the corresponding approaches (the

list is not exhaustive).

Articles

61The authors are thankful to Kenneth Sullivan, and participants at the 101 th Accountants

Forum of the IMF, as well as a number of staff of the IMF, for their comments and

suggestions. We are also indebted to Yoon Sook Kim and Xiaobo Shao for their

research and technical support.

AbstractIn light of the uncertainties about valuation highlighted by the

2007–2008 market turbulence, this paper provides an empirical

examination of the potential procyclicality that fair value account-

ing (FVA) could introduce in bank balance sheets. The paper finds

that, while weaknesses in the FVA methodology may introduce

unintended procyclicality, it is still the preferred framework for

financial institutions. It concludes that capital buffers, forward-

looking provisioning, and more refined disclosures can mitigate the

procyclicality of FVA. Going forward, the valuation approaches for

accounting, prudential measures, and risk management need to be

reconciled and will require adjustments on the part of all parties.

Alicia novoaEconomist, Financial Oversight Division, IMF

Jodi scarlataDeputy Division Chief, Financial Analysis Division,

IMF

Juan soléEconomist, Global Financial Stability Division, IMF

financial stability, fair value accounting, and procyclicality1

62 – The journal of financial transformation

financial stability, fair value accounting, and procyclicality

Non-derivative financial assets with fixed or determinable payments and fixed matu-2

rity that an entity has the intention and ability to hold to maturity.

Namely, when they are risk-managed on a FV basis, though differences remain 3

between FAS159 and IAS39.

The paper utilizes that major classification of financial assets as set out by the IASB 4

and FASB, recognizing that there are additional, specific categories of instruments

subject to fair valuation.

Since the 2007 market turmoil surrounding complex structured

credit products, fair value accounting and its application through

the business cycle has been a topic of considerable debate. As

the illiquidity of certain products became more severe, financial

institutions turned increasingly to model-based valuations that,

despite increased disclosure requirements, were nevertheless

accompanied by growing opacity in the classification of products

across the fair value spectrum. Moreover, under stressed liquidity

conditions, financial institutions made wider use of unobservable

inputs in their valuations, increasing uncertainty among financial

institutions, supervisors, and investors regarding the valuation of

financial products under such conditions.

It has been during this period that the procyclical impact of fair

value accounting on bank balance sheets and, more specifically,

the valuation of complex financial instruments in illiquid markets,

came to the fore, raising questions on the use of market prices

below “theoretical valuation” and the validity of “distressed sales.”

Financial products were fair valued despite concerns that the cur-

rent market prices were not an accurate reflection of the product’s

underlying cash flows or of the price at which the instrument

might eventually be sold. Sales decisions based on fair value pric-

ing in a weak market with already falling prices resulted in further

declines in market prices, reflecting a market illiquidity premium.

Additionally, falling prices can, and did, activate margin calls and

sale triggers that are components of risk management criteria,

contributing further to the downward trend. As bank net worth is

positively correlated with the business cycle, and as fair market

values for collateral values fall, losses have been passed through

to banks’ capital [Kashyap (2005)]. The weakening of bank balance

sheets and regulatory requirements for prudential capital replen-

ishment has served to heighten concerns as to the future course of

some markets, the health of banks and, more broadly, the financial

system.

This paper reviews the principles and application of fair value

accounting (FVA), the implications of its features and how these

impact bank balance sheets. Using a simple model, it provides

empirical support for the public discussions regarding the procycli-

cality of FVA on bank balance sheets. Utilizing representative bank

balance sheets from a sample of actual institutions, it examines the

application of FVA to banks’ balance sheets during the course of a

normal business cycle, as well as during extreme shocks, such as

those that have recently occurred, to distill in what manner FVA

may contribute to procyclicality. The paper examines the results

obtained, discusses actual and proposed alternatives to FVA, and

elaborates on policy implications going forward.

The paper finds that, while the application of FVA methodology

introduces unwanted volatility and measurement difficulties, FVA

nevertheless is the correct direction for proceeding as it provides

as faithful a picture as possible of a bank’s current financial condi-

tion — alternative techniques have their own shortcomings. Yet

despite its advantages, difficulties exist not only in determining

FV prices in downturns and illiquid markets, but also during boom

times in active markets when prices can overshoot and incorporate

risk premia that inflate profits. Under such circumstances, market

prices may not accurately reflect risks and can result in exagger-

ated profits that distort incentives (i.e., management compensa-

tion) and amplify the cyclical upturn. In rapidly evolving financial

markets, inaccurate valuations may quickly alter the implications

for solvency and more broadly, financial stability.

The paper emphasizes that FVA should be structured so that it

contributes to good risk management and ensures that financial

statements include adequate disclosure of valuations, method-

ologies, and volatilities such that inherent uncertainties are well

understood. While the volatility of estimation errors in valuation

techniques should be reduced as much as possible, genuine eco-

nomic volatility should be faithfully reflected in financial statements

and preserved by regulators and supervisors [Barth (2004), Borio

and Tsatsaronis (2005)]. The paper concludes by providing some

quantitative insight for regulators and supervisors to better assess

the implications of FVA on bank balance sheets and capital and puts

forward proposals for dealing with issues of the volatility of FVA

and FV classification. Importantly, it stresses the need for resolving

the tensions between valuation approaches across risk managers,

accountants, and prudential supervisors and regulators, so as to

ensure that accounting frameworks do not unduly contribute to

potential financial instability.

fair value accounting through the business cyclefair value accounting and its application

The current accounting framework

Both U.S. Generally Accepted Accounting Principles (U.S. GAAP)

and International Financial Reporting Standards (IFRS) use a mixed

attributes model, where different valuation criteria are applied to

different types of assets and liabilities, depending on their char-

acteristics and on management’s intentions in holding them to

maturity or not. In essence, both frameworks require FV valuation

for financial assets and liabilities held for trading purposes and

available-for-sale (AFS) assets, and all derivatives. Held-to-maturity

(HTM) investments2, loans, and liabilities not fair valued are valued

at amortized cost. Both frameworks provide a carefully specified

option to fair value (FVO) certain financial assets and liabilities3 that

would normally be valued at amortized cost4. The mixed attributes

model is intended to be as neutral as possible — without emphasiz-

ing one accounting principle over another. But, its uneven applica-

tion to balance sheets produces accounting volatility and may not

fully capture the effects of economic events in all instruments

included in the banks’ financial statements.

63

financial stability, fair value accounting, and procyclicality

September 30, 2008, the U.S. SEC jointly with the U.S. FASB issued new guidance

clarifying the use of FVA under the current environment and, on October 10, 2008,

the U.S. FASB staff issued Staff Position No. 157-3 providing guidance on how to

determine the FV of a financial asset when the market for that asset is not active.

IFRS 7, “Financial instruments: disclosures,” became effective on January 1, 2007.9

For those financial assets measured at amortized cost, the entity must also disclose 10

the FV in the notes to the statements.

Including audit-related programs.11

The FSF recommends disclosures about price verification processes to enhance gov-12

ernance and controls over valuations and related disclosures. Disclosures regarding

risk management governance structures and controls would also be welcome.

Examples are the U.S. SEC letters of March 2007 and March 2008 to major financial 13

institutions outlining the nature of recommended disclosures and the most current

letter of September 16, 2008.

“Leading-practice disclosures for selected exposures,” April 11, 2008. Twenty large, 14

internationally oriented financial firms were surveyed (15 banks and five securities

firms) as of end-2007.

Nevertheless, differences are disappearing given the international convergence to 5

IFRS currently underway, led jointly by the International Accounting Standards Board

(IASB) and the U.S. Financial Accounting Standards Board (FASB), which is aimed to

achieve a single set of high-quality international accounting standards.

This language is U.S. GAAP-specific and not IFRS, but it is used extensively in the 6

banking industry and in financial statements of IFRS users as well.

IFRS do not explicitly mention some risk factors (i.e., counterparty credit risk, liquid-7

ity risk), which may have added confusion to financial statement preparers during

the 2007–08 turmoil. An International Accounting Standards Board Expert Advisory

Group is currently working on this and other FV issues. The U.S. Financial Accounting

Standards Board is reevaluating some disclosure requirements (i.e., credit deriva-

tives) and has issued new standards (i.e., FAS 161 on derivatives and hedging). Both

Boards are working jointly on FV issues and examining requirements for off-balance

sheet entities as well.

White papers prepared by the six largest international audit firms and other audit 8

firms summarize guidance on what constitutes an active market, FV measurement

in illiquid markets, and forced sales, CAQ (2007) and GPPC (2007). Further, on

What is fair value?

IFRS and U.S. GAAP similarly define FV as the amount for which an

asset could be exchanged, and a liability settled, between knowl-

edgeable, willing parties, in an arm’s length, orderly transaction.

U.S. GAAP (FAS 157) are more prescriptive than IFRS because they

consider that FV is an “exit” or “selling” price5. Both accounting

frameworks prescribe a hierarchy of FV methodologies that start

with observable prices in active markets (Level 1), using prices

for similar instruments in active or inactive markets or valuation

models using observable inputs (Level 2), and moving to a mark-to-

model methodology with unobservable inputs and model assump-

tions (Level 3)6. The absence of market prices, trading activity, or

comparable instruments’ prices and inputs is a prominent feature

of complex structured credit products, many of which are held off

balance sheet. Consequently, both frameworks require extensive

disclosures of information on the FV methodologies used, specific

assumptions, risk exposures, sensitivities, etc.

Thus defined, FV does not require the presence of deep and liquid

markets to be applied. FV can be estimated when a market does not

exist, as FV valuation models comprise the expected, risk-discount-

ed cash flows that market participants could obtain from a finan-

cial instrument at a certain point in time. While FV incorporates

forward-looking assessments, it must also reflect current market

conditions, measures of risk-return factors7, and incorporate all fac-

tors that market participants consider relevant, with firm-specific

risk preferences or inputs kept to a minimum. Under this definition,

two key issues underlying the FV methodology present a challenge:

what constitutes an active market and what can be considered an

observable price or input.

Forced or “fire” sales would not be valid determinants of market

prices, because the accounting frameworks presume that a report-

ing entity is a going concern that does not need or intend to liqui-

date its assets, or materially curtail the scale of its operations. Yet,

accounting standard setters have decided to leave to judgment

(namely, of management, supervisors, and auditors) how to deter-

mine “regularly occurring” or “distressed” sales, and when sales in

thin markets, at heavy discounts, could be used for balance-sheets’

FVA8. Consequently, market participants and supervisors would

expect to see banks’ external auditors use a very cautious approach

to examining the prices and inputs used to FV financial instruments,

in order to minimize late write-downs or write-offs and opportuni-

ties for management to “cherry-pick” the accounting treatment of

financial instruments.

Disclosures of fVA

Both IFRS and U.S. GAAP mandate various disclosures, particularly

when information other than market inputs is used to estimate FV.

For example, IFRS 7 requires disclosure (i) if the transaction price

of a financial instrument differs from its FV when it is first recorded

in the balance sheet; and (ii) of the implications of using “reason-

ably possible alternative assumptions” to reflect the sensitivities of

FV measurement9. IFRS 7 also contain reporting requirements that

include the publication of sensitivity tests for individual items of

the financial statements. Similarly, FAS 157 requires banks’ balance

sheets to be sufficiently clear and transparent as to fully explain to

market participants, through quantitative and qualitative notes to

the financial statements, the nature of the changes and the meth-

odologies used, to name a few items10.

Although some U.S. and European Union (E.U.) financial institutions

voluntarily provide such disclosures, neither IFRS nor U.S. GAAP

require disclosure on the governance and management control

processes11 surrounding FV valuation12. Enhancement of disclosures

in this direction could increase confidence in the banks’ balance-

sheets and lower investors’ aversion to transact in instruments

whose valuations may not be well understood13. This would not

necessarily indicate a need for more disclosures, but for a more

appropriate composition, medium (i.e., websites), and frequency of

disclosures.

Along this line, at the request of the Financial Stability Forum (FSF)

a Senior Supervisors Group conducted a survey of disclosure prac-

tices for selected financial exposures, such as special-purpose enti-

ties and collateralized debt obligations, among others, and issued

a report concluding that disclosure practices currently observed

could be enhanced without amending existing accounting disclo-

sure requirements14. The FSF is encouraging financial institutions

to use these disclosure practices for their 2008 financial reports

and urging supervisors to improve risk disclosure requirements in

Pillar 3 of Basel II.

64

Canada has postponed adoption of the full International Financial Reporting 15

Standards until 2011.

Barth (2004) argues that mixed-attributes models impair the relevance and reliability 16

of financial statements and that this constitutes one of the primary reasons behind

hedge accounting. IAS 39 was aimed to alleviate mismatches in assets and liabilities

valuations due to the mixed-attributes model and the complexities of hedge accounting.

It should be noted that procyclicality of accounting and reporting standards existed 17

prior to the recent attention to FVA. It has long been recognized that as the business

cycle and market sentiment change, so too will valuations of assets and liabilities.

IFRS and U.S. GAAP accounting standards – and FVA is no exception – are applicable 18

to reporting entities irrespective of their size or systemic importance.

financial stability, fair value accounting, and procyclicality

One intention of the FVO in both accounting frameworks is to enable entities to 19

reduce accounting mismatches by applying FV on matching assets and liabilities.

Bank supervisors use prudential filters as a tool to adjust changes in the (accounting) 20

equity of a bank due to the application of the accounting framework, so that the qual-

ity of regulatory capital may be properly assessed. For example, when the gains that

result from a deterioration in a bank’s own creditworthiness (fair valued liability) are

included in a bank’s prudential own funds, they must be “filtered out” by the supervi-

sor in order to determine the true amount of regulatory own funds.

In principle, valuations are thus better aligned with the prevailing mark-to-model 21

techniques used in risk management.

A preliminary reading of financial reports prepared for mid-2008 by

some U.S., European Union, and Canadian banks would show that

U.S. banks are including more quantitative notes in their financial

statements, as compared with their end-2007 reporting15, typically

providing information on financial assets securitized, cash flows

received on Special Purpose Entities (SPE) retained interests,

assets in non-consolidated variable-interest entities (VIE), and

maximum exposures to loss in consolidated and non-consolidated

VIEs, with details broken down by instrument.

Volatility and procyclicality of fVA

Barth (1994 and 2004) argues that there are three potential

channels through which FV may introduce volatility into financial

statements. The first is the volatility associated with changes in

the underlying economic parameters. The second is the volatility

produced by measurement errors and/or changing views regarding

economic prospects throughout the business cycle. As to the third,

volatility may be introduced by relying on the “mixed attributes”

model that applies FVA to certain instruments and amortized

cost to others, reducing the netting effect that full fair valuation

of assets and liabilities would produce16. Each of these sources of

volatility is either explicitly or implicitly present in the simulation

exercises examined later in the paper.

The mixed attributes model adopted by IFRS and U.S. GAAP has

embedded volatility and procyclicality aspects17. On the one hand,

historical cost accounting, applicable to HTM investments and loans,

is less volatile and also backward looking. When such an investment

or loan is correctly priced at origination, its FV equals its face value.

Over the life of the asset and until maturity, its reported stream of

profits is stable and its carrying value is based on its value at origi-

nation. But if market conditions negatively affect these portfolios

and there is evidence of a credit loss event and asset impairment,

then the reporting values must be reassessed and provisions for

losses must be accrued or write-offs recorded. The latter is often a

late recognition of excess risk taken earlier, in good times. In this

sense, historical costs are subject to a backward-looking assess-

ment of value. Thus, amortization of loans, when combined with

procyclical provisioning, often coincides with a downturn of an

economic cycle, adding to stresses.

On the other hand, FVA introduces more volatility in earnings and

capital during the life of an asset or liability than historical cost

accounting and incorporates forward-looking assessments18. Gains

and losses in fair-valued instruments can generally affect the

income statement and this increased volatility of FVA and resulting

procyclical effects may create incentives for banks to restructure

their balance sheets (i.e., lower loan originations, higher/lower

securitization, introduce hedging, etc.)19. Nevertheless, higher FV

volatility, per se, would not necessarily be a problem if market

participants are well informed and could correctly interpret the

information provided in the financial statements. In this sense,

increased volatility may be thought of as part of the process of fair

valuing financial instruments, and a reflection of genuine economic

volatility, not as a cause itself of procyclicality.

However, in some cases, the symmetrical treatment within FVA

produces seemingly misleading results. For example, the use of

FVA on a bank’s own debt, where the price of the bank’s bonds and

notes falls due to a decline in its own creditworthiness, will result in

a gain that must be recognized in the bank’s financial statements,

equal to the difference between the original value of the debt and

its market price. As counter-intuitive as this situation may be, it is

still a faithful representation of FV and is a signal to supervisors or

other users of financial statements to have appropriate tools (i.e.,

prudential filters)20 for understanding the implications of FVA and

the impact on regulatory capital.

As valuation moves from market prices to mark-to-model valua-

tion, FVA poses reliability challenges to which markets, particularly

under distress, are sensitive21. These “subjective” aspects of FVA

may compound market illiquidity or price spirals if they increase

uncertainty around valuations. Both in the United States and the

European Union, financial institutions’ balance sheets are heav-

ily represented in Level 2 instruments, a possible indication that

financial institutions are biased towards using Level 2 methods

2228 25

7267

69

6 5 6

0

10

20

30

40

50

60

70

80

90

100

U.S. Financial Institutions European Financial Institutions Total

Level 1 valuations Level 2 valuations Level 3 valuations

Figure 1 – Aggregate fair value hierarchy, end-2007 (in percent)

Source: Fitch Ratings

65

financial stability, fair value accounting, and procyclicality

IASB’s November 2009 exposure draft, Financial instruments: amortized cost and 22

impairment, proposes an expected loss model for provisioning.

due to their flexibility, as well as a desire to avoid “obscure” Level

3 assets and liabilities (Figure 1). Falling valuations can activate

certain management decision rules that trigger the liquidation of

certain assets or portfolios, adding additional stress. Hence, there

is a need for good risk management practices to be consistent with

FV mark-to-model valuations. Clear and transparent quantitative

and qualitative notes to the financial statements regarding the

nature of the changes and methodologies could enhance reliability

of mark-to-model valuations.

Although more volatile, FVA could play a role by partially mitigating

the crisis if warning signals are heeded, thereby helping markets

to recover before damaging self-fulfilling downturns worsen. FVA

that captures and reflects current market conditions on a timely

basis could lead to a better identification of a bank’s risk profile, if

better information is provided. An earlier warning that can prompt

corrective action by shareholders, management, and supervisors

allows for a timelier assessment of the impact of banks’ risky

actions on regulatory capital and financial stability. Moreover, since

FVA should lead to earlier recognition of bank losses, it could have

a less protracted impact on the economy than, for example, loan

portfolios whose provisions for losses are usually made when the

economy is already weak22. Raising new capital at an earlier stage

might enable banks to retain written-down assets or other assets

originally not for sale on their balance sheets and, thus, to avoid

asset price spirals.

On the prudential front, the negative impact of vastly lower valu-

ations stemming from recent market conditions raises questions

as to whether increases in regulatory capital may be needed for

complex structured products, off-balance sheet entities (OBSEs),

or other risks. Basel II, Pillar 2 guidance could encourage banks to

put greater attention into FV during periods of falling or rising asset

prices, so that they may better control for procyclical aspects of

FVA. Pillar 3 disclosures could improve transparency of valuations,

methodologies, and uncertainties. Nevertheless, FVA can serve as

an early warning system for supervisors to pursue closer scrutiny

of a bank’s risk profile, risk-bearing capacity, and risk management

practices.

off-balance-sheet entities and procyclicality

Recent market turmoil has heightened public awareness of the

extensive use of off-balance-sheet entities (OBSEs) by financial

institutions. With variations, both IFRS and U.S. GAAP have spe-

cific criteria to determine when instruments transferred to OBSEs

should be consolidated on-balance-sheet. Any retained interest

in securitized financial assets should be on-balance-sheet and

accounted for at FV, usually in the trading book.

Mandatory disclosures on OBSEs are not prevalent. Their absence

may have added to market confusion and contributed to procyclical

behavior by helping to create a market perception that the banks

were standing behind their OBSEs. Both the IASB and the U.S. FASB

have different projects under way to improve OBSE disclosures and

enhance the criteria for derecognition and consolidation of OBSEs.

Examples are the IASB’s consolidation and derecognition projects,

and the FASB’s changes to FAS 140 and Interpretation 46(R). The

FASB’s recently revised standard, FAS 140, will go into effect at the

end of 2009.

Regardless, OBSEs require financial supervisors to revisit pruden-

tial reporting so that the integrity of banks’ risk exposures can be

better captured and explained, as well as adequately buffered (i.e.,

capital) to the satisfaction of supervisors.

procyclicality in the Basel ii framework

A key improvement in the Basel II framework is its enhanced risk

sensitivity. Yet this very feature is associated with the unintended

effect of heightening its procyclical propensity. Basel II recognizes

possible business cycle effects and how they should be addressed

in both Pillar 1 (minimum capital requirements) and Pillar 2 (super-

visory review process) of the framework. If Basel II is properly

implemented, then greater risk sensitivity can lead banks to restore

capital earlier in a cyclical downturn, thus preventing a build-up of

required capital when it could amplify the cycle.

Under Basel II’s Standardized Approach, risk weights are based on

external ratings constructed to see through the cycle, so that cyclical

effects are muted. It is in the internal- ratings-based (IRB) approach-

es that deterioration in credit risk feeds more directly into the capital

requirements. The three main risk components in the IRB approaches

(i.e., probability of default, loss given default, and exposure at

default) are themselves influenced by cyclical movements and may

give rise to a cyclical impact on banks’ capital requirements.

Basel II includes mitigating measures to address these concerns.

Although Pillar 1 does not mandate the use of through-the-cycle mod-

els, it promotes estimates of risk components based on observations

that “ideally cover at least one economic cycle,” and whose valida-

tion must be based on data histories covering one or more complete

business cycles. Sound stress testing processes must be in place

that involve scenarios based on economic or industry downturns and

include specific credit risk stress tests that take into account a mild

recession to assess the effects on the bank’s risk parameters.

Pillar 2 places the onus on both banks and supervisors to assess

business cycle risk and take appropriate measures to deal with it.

Banks are required to be “mindful of the stage of the business cycle

in which they are operating” in their internal assessment of capital

adequacy, perform forward-looking stress tests, address capital

volatility in their capital allocation, and define strategic plans for

raising capital. In turn, encouraging forward-looking credit risk

The U.S. Financial Accounting Standards Board has a project under way to address 23

provisioning and related credit risk disclosures.66

financial stability, fair value accounting, and procyclicality

In mid-October 2008, the IASB amended IAS39 to allow some reclassifications of 24

financial instruments held for trading or AFS to the HTM category, meeting certain

criteria, with the desire to reduce differences between IFRSs and US GAAP. As of

November 2009, IFRS 9, Financial statements, requires reclassifications between

amortized cost and fair value classification when the entity’s business model changes.

assessments or higher provisioning for loan losses (that consider

losses over the loans’ whole life) is left to national supervisors23.

Thus, where Pillar 1 does not adequately capture business cycle

effects, supervisors should take remedial action under Pillar 2,

including through additional capital buffers.

The capital disclosures required by Pillar 3 may assist markets and

stakeholders in exercising pressure on the banks to maintain their

capital levels throughout the full business cycle. In its recent report,

“Enhancing market and institutional resilience,” the Financial

Stability Forum called for the Basel Committee to develop Pillar 2

guidance on stress testing practices and their use in assessing capi-

tal adequacy through the cycle, to examine the balance between

risk sensitivity and cyclicality, and update the risk parameters and

the calibration of the framework, if needed [Financial Stability

Forum (2008)]. In response, the committee is establishing a data

collection framework to monitor Basel II’s impact on the level and

cyclicality of prudential capital requirements over time across

member countries. The committee is expected to use these results

to further calibrate the capital adequacy framework.

options for the application of fair value accounting to mitigate procyclicalityThe procyclicality of FVA has prompted the search for options that

allow financial institutions to cope with situations of market turmoil.

Alternatives range from considering a wider selection of “observ-

able” prices or inputs to a change in the accounting treatment of

financial instruments, as follows:

consensus pricing services

Consensus pricing services, often independent brokers and agen-

cies, can provide price quotes for complex or illiquid financial

instruments, often using prices based on their own sales of relevant

instruments that allow them to observe price behavior and market-

test their estimates. Through this approach, illiquid products could

obtain a Level 2 price, potentially limiting valuation uncertainty

and underpricing in downturns. However, difficulties may remain if

there is a wide dispersion of values that do not reflect the features

of the specific financial product or if banks contend that values do

not reflect market conditions, thereby obliging banks to use internal

valuation methodologies.

Valuation adjustments

Banks could estimate the “uncertainty” surrounding the price of cer-

tain assets and make a valuation adjustment to the carrying value of

an instrument disclosed in the financial statements. Valuation adjust-

ments would allow banks to work with less perfect prices that are

corrected to reflect current market conditions. These estimates of

“uncertainty” might incorporate the liquidity of inputs, counterparty

risk, or any market reaction likely to occur when the bank’s position

is realized. Valuation adjustments could improve fair value measure-

ments and discipline in reporting, yet they need close monitoring to

ensure that this practice does not evolve into management “cherry

picking,” providing a means to evade a certain accounting fair value

level classification, or improving the balance sheet.

Reclassifications

The transfer of assets from available-for-sale or trading to the

held-to-maturity (HTM) category could avoid the volatility resulting

from valuation changes amid a downward spiral. However, from an

accounting perspective, reclassifications could be penalized by not

allowing banks to revert to the trading book when markets rebound.

Further, assets transferred from the trading category to HTM would

be subject to impairment assessment (as they should were they

moved into the AFS category). From a prudential standpoint, dete-

riorated HTM assets would require higher regulatory capital, while

changes in AFS assets would be considered additional but not core

capital. Allowing reclassifications, particularly if not fully disclosed,

may postpone the weaknesses of the balance sheets, and promote

cherry-picking elements of the accounting framework24.

full fair value accounting

Recognizing the significant challenges that FVA poses, a longer-

term alternative would be to adopt a full-fair-value (FFV) model for

all financial assets and liabilities in a balance sheet, irrespective of

an entity’s intention in holding them. One single FV principle, with

some limited exceptions, would reduce the complexity of financial

instruments reporting, balance sheet window dressing, and cherry

picking, and allow for more transparent representations of the

financial condition of an entity. It could improve the comparability

of financial information across balance sheets and enhance market

discipline, but it would pose significant challenges for implementa-

tion, modeling capabilities, and auditing estimates.

internal decision rules

Without searching for a FVA alternative, regulators could require

banks to have internal decision rules based on FV that require a

careful review of all the implications of changing FV and the specific

occasions when such changes could trigger management decisions,

so that these decisions do not adversely affect regulatory capital or

accentuate downward price spirals.

smoothing techniques and circuit breakers

Smoothing asset prices and circuit breakers could be used as price

adjusters to FVA to reduce excessive price volatility in the bal-

ance sheet. Smoothing techniques involve the averaging of asset

prices over a given period. A circuit breaker imposes rules to stem

the recognition of a fall in asset prices. However, both reduce the

information content of financial statements by suspending equity

at an artificially higher-than-fair-value calculated level. The simula-

tion exercises examine the following alternatives: FFV accounting,

smoothing techniques, circuit breakers, and reclassifications.

67

financial stability, fair value accounting, and procyclicality

See Enria et al. (2004), who examine the impact of several one-off shocks on the 25

balance-sheet of a representative European bank under alternative accounting

frameworks.

Modeling fVA through the business cycle using simulationsUsing model simulations, this section assesses the effects that

changes in financial instruments’ fair value have on the balance

sheet of three types of large, internationally active financial institu-

tions — U.S. commercial banks, U.S. investment banks, and European

banks — as well as more retail-oriented U.S. and E.U. banks (Table 1).

The balance sheets of a sample of representative institutions were

taken as of end-2006 to construct prototypical institutions. The

simulations illustrate the impact of changes in valuations and, ulti-

mately, on these representative banks’ equity capital. The section

also explores possible alternatives related to FVA and its current

application — full fair value, smoothing techniques, circuit breakers,

and reclassifications — that aim to reduce its volatility on balance

sheets (Box 3.4).

The first simulation serves as the baseline for subsequent scenarios

and consists of tracking the evolution of the banks’ balance sheets

throughout a normal business cycle. Four scenarios are applied to

the normal cycle with the goal of gauging the degree to which fair

valuations amplify fluctuations in balance sheet components, and

more notably, on accounting capital25. The sources of increased

cyclicality are (i) a bust-boom cycle in equity valuations; (ii) a bust-

boom cycle in the housing market; (iii) a widening and then contrac-

tion of banks’ funding spreads; and (iv) a bust-boom cycle in debt

securities’ valuations, all of which are calibrated using the most

current cyclical movements (Table 2). As noted by Fitch (2008a

and 2008b) among others, the sensitivities of FV measurements

to changes in significant assumptions are particularly important

when valuations are model-based and/or markets become highly

illiquid. Specifically, the method by which an institution chooses

to value components of its balance sheet constitutes one of the

three main transmission channels through which FVA introduces

volatility into the balance sheet [Barth (2004)]. The simulations

help underscore this point and provide a sense of the magnitude of

these effects. In addition, the simulations illustrate how a sudden

tightening in banks’ funding conditions, or changes in the liquidity

conditions in securities markets, exacerbate cyclical fluctuations in

balance sheets.

It is worth noting that from a cash flow perspective, the changes

in assumptions underlying valuations (such as those made in the

simulations below) may not necessarily be of future consequence

to the reporting institution, as those gains and losses have not been

realized and may never be. In this sense, the ensuing changes in

regulatory capital produced by the updated valuations are some-

what artificial. With these considerations in mind, the simulation

results should be interpreted as a simple exercise to gauge how

changes in the underlying valuation parameters in the presence of

FVA may lead to substantial fluctuations in banks’ equity.

Data and modeling assumptionsThis section presents the construction of the simulation exercises

and reviews the assumptions underlying the various scenarios.

Banks’ balance sheets

To accurately reflect the balance sheets of a representative

large U.S. commercial bank, a large U.S. investment bank, a large

European bank, and retail-oriented U.S. and European banks, the

financial statements at end-2006 for these five banking groups

U.s. commercial

bank

U.s. investment

bank

European bank

U.s. retail-

oriented bank

European retail-

oriented bank

financial assets

Securities

Debt securities 21.82 27.85 15.71 14.96 17.72

Trading book FV1 21.82 27.85 14.98 5.09 16.59

Banking book2 — — 0.73 9.87 1.13

Shares 6.73 7.50 6.55 0.64 2.96

Trading book FV1 6.73 7.50 6.32 0.47 2.96

Banking book2 — — 0.23 0.17 —

Derivatives (trading) 2.67 5.28 14.71 1.19 4.44

Interest rate swaps 1.48 1.87 7.76 ... ...

Other derivatives 1.20 3.41 6.96 ... ...

Loans

Corporate/Consumer 10.11 5.63 23.77 23.00 25.84

Short-term (fixed rate) < 1 year FV1 4.72 2.82 11.88 6.84 12.92

Medium-term ( > 1 year < 5 year) 3.66 2.82 3.57 10.97 3.88

Fixed rate FV1 0.72 1.41 1.78 1.71 1.94

Variable rate FV1 2.94 1.41 1.78 9.26 1.94

Long-term (> 5year) 1.73 n.a. 8.32 5.19 9.04

Fixed rate FV1 0.46 n.a. 4.16 2.03 4.52

Variable rate FV1 1.27 n.a. 4.16 3.16 4.52

Mortgages 16.51 n.a. 6.54 37.44 26.43

Fixed rate FV1 12.83 n.a. 1.40 29.09 10.78

Variable rate FV1 3.68 n.a. 5.14 8.35 15.65

other assets 28.60 43.27 20.93 17.34 5.41

financial liabilities

Debt securities/equity (trading) FV1 4.68 8.68 12.77 0.01 12.71

Derivatives (trading) 3.20 5.49 15.34 0.96 3.47

Interest rate swaps 2.09 1.73 7.84 ... ...

Other derivatives 1.10 3.76 7.49 ... ...

Short-term and long-term financial liabilities/Bonds

FV1 18.25 27.21 10.35 19.56 18.97

other liabilities 65.26 51.52 56.23 69.72 61.16

of which: deposits and interbank borrowing

42.44 3.72 24.88 60.12 56.72

net equity3 7.65 3.71 2.86 9.75 4.36

Table 1 – Balance sheet of representative U.S. and European financial institutions (in

percent of total assets, December 31, 2006)

Sources: Annual Reports; and SEC’s 10-K filings.

Note: Columns may not add to 100 percent as some balance sheet items are not

displayed in the table.

1 Valued at fair value.

2 Annual statements showed negligible or zero holdings for the sampled U.S. banks.

3 Net equity in percent of total (non-risk weighted) assets.

The filing period was chosen to be December 2006 in order to obtain balance sheets 26

that are relatively recent, while at the same time do not reflect too closely banks’ bal-

ance sheet structures in the run-up or fall-out of the 2007-08 U.S. sub-prime meltdown.

For simulation purposes, all banks were assumed to be newly established, so that all 27

balance sheet items are at FV at the start of the simulations. Thus, the shocks applied

to the baseline reflect only the pure impact of the shocks, and not a combination of the

imposed shock plus any initial deviations from fair value.

IAS 39 prevents the valuation of demand deposits at a value less than face value, even 28

if a significant portion of these display economic characteristics of a term deposit.

Consequently, deposits remain at face value in the exercise.

68

financial stability, fair value accounting, and procyclicality

Despite being a central element in the 2007–08 turmoil, an explicit breakdown of credit 29

derivative exposures was unavailable in the 2006 reports. Some mortgage-backed

securities were included in the debt securities category.

Strictly speaking, PDt is the conditional probability of default at time t. That is, the 30

probability that, conditional on not having defaulted before, a loan defaults on period t.

It should be noted that the Quantitative impact study 5 (QIS-5) estimated the PD for a 31

group of G-10 (ex-U.S.) banks’ retail mortgage portfolio at 1.17 percent, very close to the

estimate of 1.18 percent for the trend period used here.

Although this may be a less realistic assumption than allowing LGDs to evolve through 32

the cycle, the qualitative results of the simulations would not be altered.

were compiled from the institutions’ annual reports and the U.S.

Securities and Exchange Commission’s 10-K filings26. Individual

bank balance sheets were then used to construct a weighted aver-

age for each type of institution, and the resulting representative

balance sheets (Table 1). Table 1 indicates the line items that were

fair valued in the simulations27, 28. Not all the items in the balance

sheet were fair valued in the simulations: items that are typically

not available for sale (i.e., securities in the banking book) and items

that fall under the “other” categories were held constant.29

Valuation of assets and liabilities under fair value

Loans and debt securities are valued at their expected net present

value (NPV), which takes into account the probability of default and

the loss given default of each instrument. In other words, the value

of a given security (or loan) with a maturity of T years is given by

the expression

NPV =E CFt( )1 +δt( )t

t = 1

T

∑

where δt is the discount rate for year t, and E(CFt) is the expected

cash-flow for year t factoring in the possibility that the security (or

loan) defaults:

E(CFt) = [PDt · (1 + rt) · N · (1 – LGDt)] + [(1 – PDt) · rt · N] for all t<T,

and

E(CFT) = [PDT · (1 + rT) · N · (1 – LGDT)] + [(1 – PDt) · (1 + rT) · N]

where PDt stands for probability of default30, rt is the interest rate

on the loan, N is the notional amount of the loan, and LGDt is the

loss-given-default.

Under FV, traded shares are valued at their market price. Since

the detailed composition of the shares portfolio of banks was

not available, it was assumed that banks hold a generic type of

share which represents the Standard & Poor’s 500 stock market

index. Therefore, the number of shares for each type of bank was

obtained by dividing the value of their shares portfolio at end-2006

by the value of the S&P 500 index at the same date.

characterization of the business cycles

To simplify the analysis, the paper considers a stylized business

cycle consisting of four periods representing different points in a

typical business cycle: trend, trough, peak, and back to trend. Each

point in the business cycle is characterized by a different prob-

ability of default (PD) on securities and loans. To construct the

normal business cycle, the PDs on loans and debt securities were

assumed to change with the pulse of the cycle, increasing during

economic downturns and decreasing during upswings. To isolate

the effect of the evolving PDs on valuations, the baseline simula-

tion abstracts from changes in interest rates during the cycle and

initially assumes a flat yield curve.

In principle, different classes of securities and loans may have dif-

ferent PDs and evolve differently throughout the cycle. For simplic-

ity, however, this paper assumes that all securities and loans have

the same PD and display the same cyclical behavior, except for the

scenario of the bust-boom cycle in real estate, where a different

PD for mortgages is assumed. In addition, loans are assumed to be

bullet instruments, whose principal is repaid in full upon maturity.

The specific values for these PDs were derived from Nickell et

al. (2000), who investigate the dependence of securities rating

transition probabilities on the state of the economy [Pederzoli and

Torricelli (2005), Bangia et al. (2002), and Altman et al. (2005)].

The probabilities of default at different stages of the business cycle

were computed using their estimated transition matrices at differ-

ent points in the cycle (Table 2)31.

To compute the net present value (NPV) of loans and securities,

it is also necessary to have a measure of losses in the event of

default. Thus, loss-given-default (LGD) rates were taken from the

BIS’s Fifth quantitative impact study QIS-5 [BIS (2006a)], and equal

20.3 percent for mortgage loans and 46.2 percent for corporate

loans. To isolate the effect of the evolving PDs, the LGD rates were

held constant through the cycle (except in the bust-boom cycle

in the housing market and in the downward price spiral for debt

securities)32.

Business cycle trend

points

Business cycle trough

points

Business cycle peak

points

Normal cycle PD for all loans and securities 1.18 1.40 0.73

LGD for mortgages 20.30 20.30 20.30

LGD for loans1 and securities 46.20 46.20 46.20

Stock market index 100.00 100.00 100.00

Stock market cycle

PD for all loans and securities 1.18 1.40 0.73

LGD for mortgages 20.30 20.30 20.30

LGD for loans1 and securities 46.20 46.20 46.20

Stock market index 100.00 80.00 120.00

Real estate market cycle

PD for mortgages 1.18 5.29 0.73

PD for loans1 and securities 1.18 1.40 0.73

LGD for mortgages 20.30 30.50 20.30

LGD for loans1 and securities 46.20 46.20 46.20

Stock market index 100.00 100.00 100.00

Note: PD = probability of default; LGD = loss given default.1 Loans excluding mortgages.

Table 2 – Parameter values for each simulation (in percent)

Sources: IMF staff estimates; and Nickell et al. (2000).

69

financial stability, fair value accounting, and procyclicality

The results are presented in terms of the evolution of banks’ normalized equity 37

through the cycle — that is, at each point in the cycle, banks’ equity is divided by their

initial level of equity (i.e., at end-2006). All figures for this section are presented at

the end.

Note however that this result reflects only one element of countercyclical forces, as 38

“other liabilities” represents about 50 percent of the balance sheet and can poten-

tially introduce additional countercyclicality.

Chapter 4 of IMF (2008a) examines procyclicality of leverage ratios of U.S. invest-39

ment banks, finding their extreme variation across the cycle. Note this is consistent

with the scenario conducted later in this paper where funding spreads vary through

the cycle, producing the same procyclicality found in IMF (2008a).

See Guerrera and White (2008). Additionally, Barth et al. (2008) suggest that these 40

counterintuitive effects are attributable primarily to incomplete recognition of con-

temporaneous changes in asset values.

The initial price of the representative stock held by banks was normalized to the value 33

of the S&P 500 index at end-2006, which closed at 1418 on December 29th, 2006.

To estimate the PDs during the 2007–08 U.S. housing crisis, it was assumed that 100 34

percent of foreclosures and 70 percent of delinquencies beyond 90 days end up in

default. These percentages are then combined with the respective PDs to yield an over-

all estimated PD of 5.29 percent for all mortgages. See UBS (2007); data source: Merrill

Lynch, April 2008.

The rationale behind this characterization of distressed markets follows Altman et. 35

al (2005) in that during times of distress, the demand for securities declines hence

reducing both the market price and the recovery rate (i.e., the inverse of LGD) of

securities. See Acharya et al. (2007), Altman et al. (2005), and Bruche and González-

Aguado (2008) for papers discussing the link between distressed markets and

increases in LGD rates.

Derived from Bruche and González-Aguado (2008).36

characterization of the economic shocks

The first scenario considered is a bust-boom cycle in stock market

valuations where, concurrent with a normal cycle, share prices ini-

tially plummet by 20 percent during the downturn of the economic

cycle and then surge to a level that is 20 percent above the original

level, to ultimately return to their trend value (Table 3)33.

The second scenario is a bust-boom cycle in the housing market, in

which mortgage default rates and LGD rates dramatically increase

during the downturn, and then rebound during the recovery. In

this scenario, PDs of mortgage loans increase to 5.29 percent in

the trough of the cycle — a magnitude which is commensurate with

the recent meltdown in the U.S. housing market34. Additionally, the

reduction in house values — and thus the expected decline in recov-

eries — was factored in through a 50 percent increase in the LGD

rate over the average values reported in the QIS-5 (i.e., from 20.3

percent to 30.5 percent).

To simulate the cycle in funding conditions, the paper assumes that

during the business cycle trough, banks’ cost of funding increases by

58.7 basis points. This increase in spreads was obtained by comput-

ing the average rise in Libor-OIS spreads for U.S. and European banks

during the summer of 2007. Conversely, to analyze the effects of

ample liquidity conditions, the simulation assumes that banks’ fund-

ing costs decrease by the same amount during the cycle peak.

To construct the scenario of distressed securities markets and then

recovery, it was assumed that the LGD rates for debt securities

sharply increase during troughs and decrease by the same amount

during peaks35. During the cycle trough, the LGD rate for debt

securities increases to 67.3 percent36 from its initial base of 46.2

percent. Subsequently, the simulation applies the same shock mag-

nitude (but reversed sign) to the LGD during the cycle peak; that is,

LGD decreases to 25.1 percent.

simulation results

The simulations highlight three key points regarding FVA and its

potential regulatory and financial stability implications: (i) strong

capital buffers are crucial to withstand business cycle fluctuations

in balance sheet components, especially when FV is applied more

extensively to assets than liabilities; (ii) fair valuing an expanded set

of liabilities acts to dampen the overall procyclicality of the balance

sheet; and (iii) when combined with additional liquidity shortages in

financial markets, the FVA framework magnifies the cyclical volatil-

ity of capital.

The effects of economic shocks under full fair value

In the normal cycle, fair valuing both sides of the balance sheet

produces fluctuations that are mild compared to the bust-boom

scenarios below (Figure 2), an intuitive result37. However, it is worth

noting that, in the case of the representative U.S. investment bank,

equity behaves in a countercyclical manner due to the strong effect

of fair valuing the liabilities. Under full FV (FFV), the value of the

bank’s liabilities declines as economic activity weakens and prob-

abilities of default (PDs) rise, mitigating the decline in equity. This

effect arises because of the asset/liability structure of the invest-

ment banks’ balance sheet, which consists of a large proportion of

financial liabilities that are fair valued. Liabilities at FFV, as is done

by some U.S. investment banks, can introduce an element of coun-

tercyclicality by serving as an implicit counter-balancing hedge to

the fair valuation of assets38, 39. This phenomenon has raised related

concerns by some market observers who regard with unease a

bank’s ability to record revaluation gains as its own creditworthi-

ness weakens and the price of its own debt declines40. The presence

of gains that are a construct of the particular technique chosen

for valuation, signals the need for clear disclosure of underlying

assumptions to avoid misrepresentation of financial statements.

In both the bust-boom cycles in equity valuations and in the housing

market, the European banks exhibit the largest deviations from trend.

For the equity price shock, despite roughly comparable magnitudes

U.s. commercial banks Baseline period 1 period 2 period 3 period 4

Business cycle trend

Business cycle

trough

Business cycle trend

Business cycle peak

Business cycle trend

Normal cycle 7.6 7.5 7.6 7.9 7.6

Bust-boom cycle in share prices 7.6 6.3 7.3 9.1 7.6

Bust-boom cycle in real estate 7.6 5.4 7.6 7.9 7.6

U.s. investment banks Baseline period 1 period 2 period 3 period 4

Business cycle trend

Business cycle

trough

Business cycle trend

Business cycle peak

Business cycle trend

Normal cycle 3.7 3.8 3.7 3.6 3.7

Bust-boom cycle in share prices 3.7 2.3 3.4 5.0 3.7

Bust-boom cycle in real estate 3.7 3.8 3.7 3.6 3.7

European banks Baseline period 1 period 2 period 3 period 4

Business cycle trend

Business cycle

trough

Business cycle trend

Business cycle peak

Business cycle trend

Normal cycle 2.9 2.8 2.9 3.0 2.9

Bust-boom cycle in share prices 2.9 1.6 2.6 4.2 2.9

Bust-boom cycle in real estate 2.9 1.9 2.9 3.0 2.9

Table 3 – Equity to assets ratio through the business cycle (in percent)

Source: IMF staff estimates.

Some portion of the lower equity position in European banks may stem from differ-41

ences in IFRS versus U.S. GAAP accounting treatments [Citigroup (2008), Financial

Times (2008)].

Note, however, that retail-oriented European banks also have a larger fraction of debt 42

securities and financial liabilities than the larger European banks.

70

financial stability, fair value accounting, and procyclicality

In effect, valuing these instruments at amortized cost would produce comparable 43

results to being classified as HTM.

of equity shares across the three banks’ portfolios, a combination

of two effects are at work. First, there is the countercyclical effect

of the relatively greater proportion of FV liabilities for investment

banks. Second, the European bank has a lower capital base and thus

the relative size of valuation changes to normalized equity capital is

larger. In the housing market scenario, the European bank exhibits

wider fluctuations, despite the fact that the U.S. commercial bank

holds a much larger fraction — about two-and-half times greater — of

its loan portfolio in mortgages. In both scenarios, the lower capital

base of the European bank vis-à-vis the U.S. commercial bank is a

key element. Similar results in terms of capital-to-assets ratios are

presented in Table 3, but reflect a less dramatic impact on European

banks41. More generally, a bank’s balance sheet would evolve through

the cycle — contracting in downturns and expanding in upturns — such

that it would restore a bank’s capital adequacy ratio, a result that is

not testable in this simple framework.

The recent events have raised two interesting scenarios regarding

increased funding costs and a downward spiral in the valuation of

debt securities. Sudden changes in bank’s ability to obtain funding

largely exacerbate the fluctuations in balance sheets (Figure 3).

This exercise underscores the significance of general liquidity

conditions in driving balance sheet fluctuations, and how the FVA

framework recognizes these changes promptly. Interestingly, the

countercyclical behavior observed in the U.S. investment banks’

equity disappears. In fact, the U.S. investment bank is hardest hit by

both the tightening of funding conditions and the distress in secu-

rities markets. This should not be surprising given that, contrary

to the U.S. commercial and European banks, the U.S. investment

bank does not rely on deposits — which are not fair valued — to

fund its activities. Note, too, that these simulations do not account

for structured credit products or the OBSEs that were so central

to much of the 2007–08 turmoil and would likely increase the

procyclicality of the balance sheets. Such a deterioration of banks’

balance sheets could affect market confidence and overall share

prices, which in turn could generate additional volatility in banks’

balance sheets.

The results presented thus far have focused on the balance sheets

of large internationally active institutions. Comparatively, the

more retail-oriented banks tend to have larger loan and mortgage

portfolios and rely more extensively on deposits for their funding42.

To illustrate the effects of these two structural characteristics,

simulations comprising the cycle in funding spreads and the bust-

boom cycle in real estate were conducted for all banks, excluding

the representative U.S. investment bank. The results corroborate

the supposition that the more retail-oriented institutions are less

vulnerable to changes in funding conditions than their internation-

ally active counterparts (Figure 4). Conversely, the retail-oriented

banks are harder hit by a bust in the housing market than the inter-

nationally active banks.

The effects of mixed-attributes models

Using two versions of the mixed-attributes model, this exercise

shows how the degree to which financial institutions apply FV to

their assets and liabilities affects the extent to which there can be

offsetting volatility effects. Table 4 shows that financial institutions

apply FV differentially. But what is not shown in the table is the

extent to which the vast majority of banks continue to use amor-

tized cost to value their loan portfolio. Thus, for the purposes of the

simulations, two variations of the model are considered: (i) “finan-

cial liabilities and bonds” are valued at amortized cost throughout

the cycle; and then (ii) in addition, “loans” and “mortgages” are also

valued at amortized cost43.

Figure 5 underscores the idea that the asymmetric application of

a mixed-attributes model, where FV is applied more extensively to

value assets than liabilities, has the effect of increasing the procy-

clical behavior of the balance sheet. In other words, the fluctuations

in equity — for all types of institutions and for all the scenarios

considered — are larger when a smaller fraction of liabilities are

fair valued (compare with Figure 2, the results under FFV). Thus,

the benefits intended by the introduction of the FVO, which were

to reduce the accounting volatility of the mixed attributes methods

and the need for FV hedge accounting techniques, are lessened.

This supports an expanded application of FV, rather than a reduced

application, as some would like to propose. Bear in mind, however,

that the application of FV to banks’ own debt may produce revalu-

financial institutions Assets at fV on a recurring

basis

liabilities at fV on a recurring

basis

Return on equity

JPMorgan Chase & Co. 41 16 12.86

Citigroup 39 22 3.08

Bank of America 27 6 10.77

Goldman Sachs 64 43 31.52

Lehman Brothers 42 22 20.89

Merrill Lynch 44 33 -25.37

Morgan Stanley 44 27 9.75

Credit Suisse 64 39 17.88

Societe Generale 46 32 3.36

Royal Bank of Scotland 45 31 15.13

BNP Paribas 65 55 16.98

Deutsche Bank 75 48 18.55

UBS 54 35 -10.28

HSBC 40 25 16.18

Barclays 52 39 20.50

Credit Agricole 44 24 10.67

Table 4 – Application of fair value by U.S. and European banks, 2007

(in percent of total balance sheet)

Sources: Fitch; and Bloomberg L.P.

71

financial stability, fair value accounting, and procyclicality

This simulation abstracts from the effect of revaluing interest rate swaps. 47

Unfortunately, it was not possible to obtain a sufficiently complete and consistent

dataset on these instruments to include them in the simulation. Nevertheless, pre-

liminary results using available data on interest rate swaps showed similar qualitative

results.

Moving to an expected loss model of provisioning could decrease volatility.44

Although this simulation is subject to the Lucas critique in that bank behavior is 45

assumed not to change in response to policy adjustments, it provides some insights

into the interaction between FVA and interest rates.

Interestingly, the addition of changes in the yield curve counteracts the effect of the 46

evolution of PDs. The drop in the yield curve in the downturn results in higher valu-

ations and thus counterbalances the downward effect of the PDs, while the positive

effect on valuations stemming from lower PDs is counterbalanced by a higher yield

curve in the upturn.

ation gains as the value of liabilities declines on their balance sheet

and that this should be properly disclosed.

This simulation highlights that the greater the imbalance of the

mixed attributes application to assets and liabilities, the greater is

the accounting volatility. When financial instruments are valued at

a historical cost that does not represent the current market condi-

tions, an accurate picture of a bank’s equity becomes blurred and

the informational content of the accounting statements weakens.

Historical costs have low information content for investors who rely

on current financial figures as a basis for investment decisions. For

a regulator, making an accurate assessment of the health of a bank,

and formulating the appropriate regulatory response, becomes

increasingly difficult44.

The second simulation (not shown), where financial liabilities plus

loans and mortgages are all valued at amortized cost, showed that

the range of fluctuations diminished further than in the above simu-

lation. Thus, although the wider application of the mixed attributes

model can reduce fluctuations in the balance sheet, the cost comes

in the form of a further reduction in up-to-date information.

smoothing techniques and circuit breakers on reporting pricesSimulations using proposed alternatives to smooth balance sheet

volatility show that a smoothing/averaging technique for falling

asset prices blurs the bank’s capital position, in magnitudes varying

by the amount and period over which the averages are calculated.

Smoothing techniques and other impediments to allowing valua-

tions to adjust, so called “circuit breakers,” make it harder for regu-

lators and investors to accurately assess the financial position of a

bank as it hides the economic volatility that should be accounted

for in the balance sheet.

To illustrate, two simulations were conducted, each averaging

share prices over different lengths. The first simulation uses a two-

period average, whereas the second simulation is extended to three

periods. As shown in Figure 6, the longer the averaging length, not

surprisingly, the smoother is the path of the balance sheet. Notably,

the application of a smoothing technique might reduce the occa-

sion for forced sales, as it could avoid sale triggers in some cases.

Accordingly, this could lessen a downward price spiral in the market

for a financial product by avoiding forced sales, but comes at the

expense of a reduction in the informational content of financial

statements and potentially lengthening the resolution period.

Similarly, concepts such as a circuit breaker, whereby rules stem the

recognition of a fall in asset prices, mask the underlying equity posi-

tion by suspending equity at an artificially higher level than under

FV and, more generally, may hamper price discovery. However,

in this case, the cycle may be extenuated even longer than with a

smoothing technique because the circuit breaker can maintain the

same value for a given period, while the smoothing is a rolling aver-

age that is updated during each period of the cycle. Additionally, this

measure is asymmetrically applied, as the circuit breaker has gener-

ally been proposed for when valuations are falling. Even though not a

preferred technique, for symmetry, one could apply circuit breakers

during bubble periods to stop the artificial inflation of equity. If not,

asymmetric treatment of valuations may create perverse risk-taking

incentives for managers as long as financial institutions are able to

benefit from the upside in valuation while the downside would remain

capped.

The effects of a changing yield curve Yield curve effects are introduced to the baseline scenario to evalu-

ate how the change in interest rates over the cycle affect the bal-

ance sheet45. The paper follows Keen (1989) and assumes the fol-

lowing stylized facts regarding the cyclical behavior of yield curves

[Piazzesi and Schneider (2006), Keen (1989)]: (i) both short- and

long-term rates tend to decline during business cycle downturns

and rise during expansions; and (ii) short rates tend to rise more

relative to long rates during expansions (i.e., the yield curve flat-

tens) and fall more relative to long rates during recessions (i.e., the

yield curve steepens) (Figure 7)46.

The influence of interest rates tends to dominate the effect of

the change in PDs, such that the interest rate effect dampens

the magnitude of procyclical equity fluctuations for the European

bank, and even becomes countercyclical for the U.S. commer-

cial bank (Figure 8). For the U.S. investment bank, the change in

interest rates renders the evolution of equity procyclical, rather

than countercyclical, as in the baseline simulation. This reversal

in behavior is due to the fact that the U.S. investment bank has

a slightly larger share of FV liabilities than assets being revalued

when interest rates change47. But this also highlights the European

banks as an intermediate structure between the investments banks

and retail bank characteristics. Regardless of the balance sheet

structure, changes to interest rates and other monetary policy

tools can dampen procyclical influences, suggesting countercyclical

monetary policy could have the beneficial outcome of also helping

to counteract the effects of the asset valuation cycles on banks’

equity. Note, however, these simulations do not allow the financial

institutions to respond to policy changes, and thus these results,

while informative, should be taken with caution.

72 – The journal of financial transformation

financial stability, fair value accounting, and procyclicality

Although the weaknesses are related more to issues of OBSEs, consolidation, and 48

derecognition, than to FV.

conclusions and policy recommendationsThe financial turmoil that started in July 2007 unveiled weaknesses

in the application of some accounting standards48 and with the

valuation and reporting of certain structured products. While these

weaknesses may have contributed to the current crisis, they also

provide an opportunity to better understand them.

The paper finds that, despite concerns about volatility and mea-

surement difficulties, FVA is the appropriate direction forward and

can provide a measure that best reflects a financial institution’s

current financial condition, though various enhancements are

needed to allow FVA to reinforce good risk management techniques

and improved prudential rules. Nevertheless, the application of

FVA makes more transparent the effects of economic volatility on

balance sheets that, under certain risk management frameworks,

could exacerbate cyclical movements in asset and liability values.

Exaggerated profits in good times create the wrong incentives.

Conversely, more uncertainty surrounding valuation in downturns

may translate into overly tight credit conditions, and negatively

affect growth at a time when credit expansion is most needed.

This is not to say that alternative accounting frameworks, such

as historical cost accounting, avoid such fluctuations, but rather

that FVA recognizes them as they develop. Regardless, accounting

frameworks are not meant to address the market-wide or systemic

outcomes of their application, as they are applied only to individual

institutions. Nevertheless, much of the controversy surrounding

FV stems more from the risk management and investment deci-

sion rules using FV outcomes, rather than the framework itself.

Delinking the interaction of FV estimates from specific covenants,

such as sales triggers, margin calls, or additional collateral require-

ments during downturns, or compensation tied to short-term prof-

its during upturns, are options that could mitigate the procyclical

impact of FVA.

Overall, the simulations confirmed a number of issues in the ongo-

ing FVA debate and underscored three key points regarding FVA

and its potential regulatory and financial stability implications: (i)

strong capital buffers and provisions make an important contribu-

tion to withstanding business cycle fluctuations in balance sheets,

especially when FVA is applied more extensively to assets than

liabilities; (ii) when combined with additional liquidity shortages in

financial markets, the FVA framework magnifies the cyclical volatil-

ity of capital; and (iii) fair valuing an expanded set of liabilities acts

to dampen the overall procyclicality of the balance sheet. However,

the latter may also give rise to the counterintuitive outcome of

producing gains when the valuation of liabilities worsens. This is of

particular concern when a deterioration in a bank’s own credit wor-

thiness, and the subsequent decline in value of own debt, results

in profits and a false sense of improvement in the bank’s equity

position.

Proposals for alternative accounting methods, such as historical

cost or simplistic mechanisms to smooth valuation effects on bank

balance sheets, reduce the transparency of a financial institution’s

health by blurring the underlying capital position. While these tech-

niques may avoid sale triggers incorporated in risk management

covenants and limit downward price spirals, the measurement

variance introduced by such techniques can increase uncertainties

regarding valuations. The loss of transparency makes it more dif-

ficult for all users of financial statements, for example, for supervi-

sors to conduct adequate oversight of financial institutions and rec-

ommend appropriate regulatory measures to deal with prudential

concerns, and for investors who will demand increased risk premia

in the face of uncertainty.

policy proposalsMost proposals should aim to deal with the use of FV estimates to

lessen the volatility that FVA can introduce to the balance sheet.

Assessments of provisioning and capital adequacy should take

better account of the business cycle. Improved transparency can

be achieved not necessarily by more disclosures, but better dis-

closures. Financial, accounting, and regulatory bodies are already

providing guidance and recommendations in this direction.

The simulations support the relevance of establishing a capital ■■

buffer that looks through the cycle, augmenting the capital posi-

tion during boom cycles to withstand the burden on capital that

stems from economic downturns. Although a partial analysis, the

simulations show that FVA can introduce financial statement vola-

tility and provide a first indication that buffers of around 24 per-

cent of additional capital would help banks weather normal cycli-

cal downturns, whereas higher buffers — on the order of 30–40

percent extra capital — would be needed to offset more severe

shocks. Recognizing that these estimates do not reflect concur-

rent changes in risk-weighted assets, they nevertheless provide an

initial estimate of the magnitude of the needed capital buffer, as

well as the direction for further analysis. Note that these are not

adjustments to FV calculations, per se, but are adjustments meant

to help mitigate the impact on bank balance sheets. Consideration

to making other changes to the accounting framework so that

the FV calculations themselves obviate the need for these other

adjustments would be useful at this juncture.

Broadening the current narrow concept of provisions to incor-■■

porate additional methods of retaining income in upswings could

provide a way of better offsetting balance sheets’ procyclical

effects, for not-fair-valued assets. It is generally agreed that

provisions protect against expected losses and capital protects

against unexpected losses. A build-up of provisions better linked

to the expected volatility, higher risks, and potentially larger

losses of an asset, could better anticipate the potential negative

effects on the balance sheet that would be reflected through the

73

financial stability, fair value accounting, and procyclicality

Basel Committee on Banking Supervision (2006b) and IAS 2001.50

FASB’s XRBL project for financial institutions would provide data online in about 51

three years, as discussed in the IMF April 2008 edition of the Global financial stability

report [IMF (2008b)].

This would be separate from U.S. SEC 10-Q filings.52

Forward-looking provisioning denotes provisions based on the likelihood of default 49

over the lifetime of the loan, reflecting any changes in the probability of default

(after taking into account recovery rates). Dynamic (or statistical) provisioning can be

considered an extension of forward-looking provisions with reliance on historical data

on losses for provisioning calculations. Conceptually, dynamic provisioning would

entail that during the upside of the cycle, specific provisions are low and the statisti-

cal provision builds up generating a fund; during the downturn, the growth in specific

provisions can be met using the statistical fund instead of the profit and loss account

[Enria al (2004) and Bank of Spain (www.bde.es)].

cycle, as long as the build-up does not provide a way for manipu-

lating earnings. Coordination between accounting standard set-

ters and supervisors would be needed to effect such changes.

Similarly, the use of forward-looking provisioning■■ 49, combined with

a supervisor’s experienced credit judgment in assessing the prob-

ability of default, loss given default, and loan loss provisioning50,

could mitigate the procyclical forces on the balance sheet. The

recognition of credit losses in the loan portfolio earlier in a down-

ward cycle would lessen an accompanying decline in bank profits

and the potential for a squeeze in credit extension that could

contribute to a further downward economic trend. Similarly, on

the upside, dividend distributions should only come from realized

earnings that are not biased by upward cyclical moves.

From an oversight perspective, the simulations underscore the ■■

importance of understanding the cyclical implications of FVA.

An enhanced role for prudential supervisors will be needed to

ensure close inspection of a bank’s risk profile and risk manage-

ment practices, and make appropriate recommendations for

augmented capital buffers and provisions, as needed. A compre-

hensive bank supervisory framework should include stress tests

of FV positions through the business cycle. Similarly, auditors

will have a critical role to play in ensuring credibility, consistency,

and neutrality in the application of FVA, and overall in support-

ing market confidence rather than appearing to augment pro-

cyclicality by encouraging lower valuations during a downturn.

A closer collaborative framework among audit and accounting

standard setters and supervisors would be highly beneficial for

markets and financial stability to ensure greater consistency in

assessing and interpreting financial statements.

In light of the different dynamics through the financial cycle and ■■

the doubts that can surround valuations, FV estimates should be

supplemented by information on a financial instrument’s price

history, the variance around the FV calculations, and manage-

ment’s forward-looking view of asset price progression and how

it will impact the institution’s balance sheet. Reporting a range

within which the FV price could fall would help users of financial

statement to better understand and utilize the volatilities with

which they are dealing. FV estimates should be supplemented

with detailed notes on the assumptions underlying the valuations

and sensitivity analyses, so that investors can conduct their own

scenario analyses and determine whether the FV price is repre-

sentative of market conditions.

More refined disclosures could meet the expanding needs of ■■

various users, including investors, supervisors, and deposi-

tors, in a common framework of disclosure. For example, a

series of shorter reports that would be available on websites51

and issued more frequently (i.e., quarterly)52 and cater to a

narrower group of user’s needs could highlight the most rel-

evant information, with a particular emphasis on risk develop-

ments. Further, the volatility associated with a FV balance

sheet may mean that the balance sheet is no longer the pri-

mary medium for evaluating bank capital. Market participants

and supervisors may increasingly turn to cash flow state-

ments, income and equity statements, and risk measures —

all of which provide distinct financial information and that must

evolve in response to users’ needs.

Albeit of a simple structure and subject to isolated shock sce-■■

narios, the simulations point to the fact that the application of

FV to both sides of the balance sheet would introduce a coun-

tercyclical component that may cushion some of the financial

shocks that can result in large swings in bank equity. This result,

however, arises in the shock scenarios, in part, from a deteriora-

tion in the own-debt values as risk premia rise on the liability

side of the balance sheet. This logically compensates for the

deterioration of the asset side during a downturn. From the view

point of assessing the riskiness of the financial institution or its

future prospects, the result can be viewed as paradoxical, as it

can hardly be regarded as a positive factor for the financial insti-

tution to have its own debt values deteriorate. The simulations

also illustrate how a bank’s response to a particular shock varies

substantially depending on the specific balance sheet structure

and thus there is a need to discern the source of the cyclicality

through additional disclosures.

A key challenge going forward will be to enrich the FVA framework

so that market participants and supervisors are better informed,

in order to promote market discipline and financial stability. The

fragmented solution that currently exists between the accounting,

prudential, and risk management approaches to valuation is insuf-

ficient and must be reconciled. Importantly, this will require adjust-

ments on the part of all three disciplines to resolve these tensions.

74 – The journal of financial transformation

financial stability, fair value accounting, and procyclicality

Normal Cycle

100.098.3103.8100.0

101.5

96.997.6

105.1

949698

100102104106

U.S. commercial banks U.S. investment banks European banks

Bust-Boom Cycle in Share Prices

100.0100.0

80.7

121.4

61.0

137.3

53.5

149.3

40

60

80

100

120

140

160

Bust-Boom Cycle in Real Estate

100.0

69.1

103.8100.0101.5

96.9

66.7

105.1

40

60

80

100

120

140

160

Business cycletrend

Business cycletrough

Business cycletrend

Business cycletrend

Business cycletrend

Business cycletrough

Business cycletrend

Business cycletrend

Business cycletrend

Business cycletrough

Business cycletrend

Business cycle Business cycletrendpeak

Business cyclepeak

Business cyclepeak

Normal Cycle

100.0100.0100.0 98.3 103.8

97.6

105.1

97.8

104.9

96.5

107.8

94

99

104

109

U.S. commercial banks European banksRetail-oriented U.S. banks Retail-oriented European banks

Cycle in Funding Spreads

100.0100.0

85.0

117.4

62.0

141.9

88.9114.0

64.4

140.9

40

60

80

100

120

140

160

Bust-Boom Cycle in Real Estate

100.0

14.4

100.0

69.1 103.8

66.7

105.1

45.9

104.9

107.8

10

40

70

100

130

160

Business cycletrend

Business cycletrough

Business cycletrend

Business cycletrend

Business cycletrend

Business cycletrough

Business cycletrend

Business cycletrend

Business cycletrend

Business cycletrough

Business cycletrend

Business cycle Business cycletrendpeak

Business cyclepeak

Business cyclepeak

U.S. Commercial Bank

100.0100.0100.0 98.3

103.8

85.0

117.4

91.6107.3

55

75

95

115

135

Normal cycle Cycle in funding spreads

Cycle in debt securities' LGDs

U.S. Investment Bank

100.0100.0 101.5

96.9

58.8

140.9

83.9

106.3

55

65

75

85

95

105

115

125

135

145

European Bank

100.0100.0100.0 97.6

105.1

62.0

141.9

85.3

111.7

55

65

75

85

95

105

115

125

135

145

Business cycletrend

Business cycletrough

Business cycletrend

Business cycletrend

Business cycletrend

Business cycletrough

Business cycletrend

Business cycletrend

Business cycletrend

Business cycletrough

Business cycletrend

Business cycle Business cycletrendpeak

Business cyclepeak

Business cyclepeak

Normal Cycle

95.5

100.0100.0

96.3

107.9

109.4

94.7

111.3

90

95

100

105

110

115

U.S. commercial banks U.S. investment banks European banks

Bust-Boom Cycle in Share Prices

100.0100.0

78.7

125.4

55.1

149.9

50.5

155.5

40

60

80

100

120

140

160

Bust-Boom Cycle in Real Estate

100.0100.0100.0

67.1

107.9100.0 95.5

109.4

63.8

111.3

40

60

80

100

120

140

Business cycletrend

Business cycletrough

Business cycletrend

Business cycletrend

Business cycletrend

Business cycletrough

Business cycletrend

Business cycletrend

Business cycletrend

Business cycletrough

Business cycletrend

Business cycle Business cycletrendpeak

Business cyclepeak

Business cyclepeak

Figure 2 – Simulation of full fair value Figure 4 – Simulation of full fair value: international versus retail-oriented banks

Figure 3 – Simulation of full fair value: changes in fund conditions and financial

market distress

Figure 5 – Simulation of partial fair value (includes short-term and long-term financial

liabilities valued at amortized cost)

75

financial stability, fair value accounting, and procyclicality

U.S. Commercial Banks

100.0

80.7

121.4

89.5

110.4108.8

100.0102.3

104.4

70

80

90

100

110

120

130

Bust-boom cycle in share prices Two period average

Three period average Circuit breaker

U.S. Investment Banks

120.2

61.0

137.3

100.0

112.1100.0

81.2 93.6

40

60

80

100

120

140

160

European Banks

100.0

122.1

111.0100.0

53.5

149.3

75.5

121.7

101.5

40

60

80

100

120

140

160

Business cycletrend

Business cycletrough

Business cycletrend

Business cycletrend

Business cycletrend

Business cycletrough

Business cycletrend

Business cycletrend

Business cycletrend

Business cycletrough

Business cycletrend

Business cycle Business cycletrendpeak

Business cyclepeak

Business cyclepeak

Normal Cycle

100.0100.0

103.0

98.4

95.8

102.798.4

103.9

90

95

100

105

110

U.S. commercial banks U.S. investment banks European banks

Bust-Boom Cycle in Share Prices

100.0

85.4

116.0

100.0

55.4

143.1

54.2

148.1

40

60

80

100

120

140

160

Bust-Boom Cycle in Real Estate

100.0100.0

79.8 98.4

95.8 102.7

70.3

103.9

40

60

80

100

120

140

160

Business cycletrend

Business cycletrough

Business cycletrend

Business cycletrend

Business cycletrend

Business cycletrough

Business cycletrend

Business cycletrend

Business cycletrend

Business cycletrough

Business cycletrend

Business cycle Business cycletrendpeak

Business cyclepeak

Business cyclepeak

Figure 6 – Simulation of smoothing techniques Figure 8 – Simulation of full fair value with upward sloping yield curve

3,25

3,75

4,25

4,75

5,25

5,75

6,25

6,75

1 2 3 4 5 7 10 20 30

Maturities (In years)

Trend Trough Peak

Figure 7 – Yield curves and business cycles (in percent)

76 – The journal of financial transformation

financial stability, fair value accounting, and procyclicality

ReferencesAcharya, V. V., S. T. Bharath, and A. Srinivasan, 2007, “Does industry-wide distress •

affect defaulted firms? Evidence from creditor recoveries,” Journal of Financial

Economics, 85:3, 787–821

Altman, E. I., B. Brady, A. Resti, and A. Sironi, 2005, “The link between default and •

recovery rates: theory, empirical evidence, and implications,” Journal of Business,

78:6, 2203–2227

Bangia, A., F. Diebold, A. Kronimus, C. Schagen, and T. Schuermann, 2002, “Ratings •

migration and the business cycle, with application to credit portfolio stress testing,”

Journal of Banking and Finance, 26

Bank of Spain, “Dynamic provisioning in Spain: impacts of the new Spanish statistical •

provision in the second half of 2000,” Available at (www.bde.es)

Barth, M., 2004, “Fair values and financial statement volatility,” in Borio, C., W. C. •

Hunter, G. G Kaufman, and K. Tsatsaronis (eds), The market discipline across countries

and industries, MIT Press, Cambridge, MA

Barth, M., L. D. Hodder, and S. R. Stubben, 2008, “Fair value accounting for liabilities •

and own credit risk,” The Accounting Review, 83:3.

Basel Committee on Banking Supervision, 2006a, Results of the fifth quantitative •

impact study (QIS 5), Bank for International Settlements, Basel

Basel Committee on Banking Supervision, 2006b, Sound credit risk assessment and •

valuation for ), Bank for International Settlements, Basel

Borio, C., and K. Tsatsaronis, 2005, “Accounting, prudential regulations and financial •

stability: elements of a synthesis,” BIS Working Papers No. 180, ), Bank for International

Settlements, Basel

Bruche, M., and C. González-Aguado, 2008, “Recovery rates, default probabilities, and •

the credit cycle,” CEMFI Working Paper, Centro de Estudios Monetarios y Financieros,

Madrid

Calza, A., T. Monacelli, and L. Stracca, 2006, “Mortgage markets, collateral constraints, •

and monetary policy: do institutional factors matter?” Center for Financial Studies,

University of Frankfurt, Working Paper No. 2007/10

Center for Audit Quality, 2007, “Measurements of fair value in illiquid (or less liquid) •

markets,” White Papers, October

Citigroup, 2008, “There’s a hole in my bucket: further deterioration in European banks’ •

capital ratios,” Industry focus, Citigroup Global Markets: Equity Research

Enria, A., L. Cappiello, F. Dierick, S. Grittini, A. Haralambous, A. Maddaloni, P. A.M. •

Molitor, F. Pires, and P. Poloni, 2004, “Fair value accounting and financial stability,”

ECB Occasional Paper Series No. 13, European Central Bank

Financial Stability Forum, 2008, “Report of the Financial Stability Forum on enhancing •

market and institutional resilience,” April 7

Financial Times, 2008, “Banks according to GAAP.” Available at www.ft.com•

Fitch Ratings, 2008a, “Fair value accounting: is it helpful In illiquid markets?” Credit •

Policy Special Report, 28 April

Fitch Ratings, 2008b, “Fair value disclosures: a reality check,” Credit Policy Special •

Report, June 26

Guerrera, F., and B. White, 2008, “Banks find way to cushion losses,” Financial Times, •

July 8

Global Public Policy Committee, 2007, “Determining fair value of financial instruments •

under IFRS in current market conditions,” Available at www.pwc.com

International Accounting Standards Board, 2001, “Financial instruments: recognition •

and measurement,” IAS 39, Available at www.iasb.org

International Accounting Standards Board, 2007, “Disclosure of financial instruments,” •

IFRS 7, available at www.iasb.org

International Monetary Fund (IMF), 2008a, “Financial stress and economic downturns,” •

Chapter 4, World Economic Outlook, World Economic and Financial Surveys, October

International Monetary Fund (IMF), 2008b, Global financial stability report, World •

Economic and Financial Surveys, April

Kashyap, A., 2005, “Financial system procyclicality,” Joint Conference of the IDB and •

Federal Reserve Bank of Atlanta’s Americas Center: Toward better banking in Latin

America, InterAmerican Development Bank, Washington, January 10

Keen, H., 1989, “The yield curve as a predictor of business cycle turning points,” •

Business Economics, October, Available at: http://findarticles.com/p/articles/mi_

m1094/is_n4_v24/ai_7987167.

Nickell, P., W. Perraudin, and S. Varotto, 2000, “Stability of rating transitions,” Journal •

of Banking and Finance, 24:1–2, 203–227

Office of Federal Housing Enterprise Oversight, 2008, Data statistics•

Pederzoli, C., and C. Torricelli, 2005, “Capital requirements and business cycle regimes: •

Forward-looking modeling of default probabilities,” Journal of Banking and Finance,

29:12, 3121-140

Piazzesi, M., and M. Schneider, 2005, “Equilibrium yield curves,” NBER Working Paper •

12609

Union Bank of Switzerland (UBS), 2007, “Subprime loss protection – a do-it-yourself •

kit,” Mortgage Strategist, June 26

Articles

77

can ARMs’ mortgage servicing portfolios be delta-hedged under gamma constraints?

carlos E. ortizAssociate Professor, Department of Mathematics

and Computer Science, Arcadia University

charles A. stoneAssociate Professor, Department of Economics,

Brooklyn College, City University of New York

Anne ZissuAssociate Professor and Chair, Department of

Business, CityTech, City University of New York, and Research Fellow, Department of Financial

Engineering, New York University, The Polytechnic Institute

AbstractOrtiz et al. [2008, 2009] develop models for portfolios of mortgage

servicing rights (MSR) to be delta-hedged against interest rate risk.

Their models rely on this fundamental relationship between pre-

payment rates (cpr) and interest rates, represented as a sigmoid

function (S-shape). Defaults that lead to foreclosures or loan modi-

fications on mortgages will either truncate or extend the stream of

servicing income greeted by pools of adjustable rate mortgages.

Ortiz et al.’s previous research focuses on mortgage services rights

for fixed rate mortgages. In this paper we will extend their research

to MSR for adjustable rate mortgages (ARMs).

78 – The journal of financial transformation

can ARMs’ mortgage servicing portfolios be delta-hedged under gamma constraints?

The market for ARMs and the current financial crisis are closely

related. The story goes like this: home values were increasing at

increasing rates between 1994 and 2007, making the public goal

of widespread home ownership more difficult to achieve. Not only

were house prices increasing, the rate of increase was going up for

most of this period as well (Figure 1).

As home prices increased home ownership became too costly for

large segments of the population, particularly people with sub-

prime credit scores. Subprime lending filled the gap and subprime

lenders bent over backwards to accommodate more and more bor-

rowers who wanted to get involved with the home buying frenzy.

Underwriting standards were lowered and speculative/ponzi loans

cloaked in the fabric of thirty-year amortizing loans were originated.

Rising home values increased the value of leverage. Leverage was

more affordable when the borrower assumed the interest rate risk

(adjustable rate mortgages). Competition among mortgage lenders

for origination and servicing fees created a supply of unstable mort-

gage loans that were effectively short term loans with an option

to extend or refinance. Rising home prices were expected to act

as cover for weak underwriting. Ponzi finance always depends on

rising asset values and underpriced credit. Both of these elements

became increasingly and extremely scarce at the end of 2006 and

the beginning of 2007. As refinancing for subprime borrowers

became less available, default became the only exit for a large per-

cent of the subprime market place.

Adjustable rate mortgages encompass a wide range of contract

designs. The significant differences across ARM contracts are the

frequency of interest rate adjustment, the index used to determine

the interest rate, the margin over the index, the caps and floors

in the contract, and the required minimum interest payment. The

basic ARM product will use a money market index such as LIBOR or

the constant maturity treasury (CMT) as the benchmark. The inter-

est rate the mortgagor is required to pay will be set at a margin

above this index and will adjust periodically. The adjustment period

may be a year or longer. The amount of rate adjustment that can

take place within an adjustment period and over the length of the

loan may be capped and floored. Some ARMS known as hybrid

mortgages have a fixed rate period that may range from three

to ten years. After the fixed rate period is over the adjustable

rate phase begins. The mortgage rate begins to float at a margin

above an index. Another product that became common during the

extreme acceleration of home prices in 2005 is the option ARM.

The distinguishing feature of this ARM product is the choice it

gives the borrower to defer interest payments. Deferred interest

payments create negative amortization of the loan as the amounts

deferred are added onto the outstanding mortgage loan principal.

Negative amortization increases the leverage the borrower is using

to finance the property. Negative amortization was a short-term

solution for borrowers who were trying to conserve cash in the

short run. Negative amortization creates an unstable situation

when house prices begin to decline and the thin equity cushion the

owner has is quickly eroded. This is what happened at the end of

the housing boom.

The underwriters underpriced the credit risk because they overes-

timated the future value of the mortgaged property. It appears that

the premise that was prevalent in the mortgage market between

2005 and 2007 was that rising home prices would erase all traces

of weak or questionable underwriting. Risky underwriting includes

basing loan amounts on the discounted rates as opposed to fully

indexed rates. This technique was used extensively to tease bor-

rowers into using more leverage than was prudent. The unsurpris-

ing result of offering below market rates is payment shock when

the discounted rate is adjusted to a fully indexed rate. Optimists

believed that payment shock could be dealt with by either refinanc-

ing an outstanding loan with a loan that was more affordable or

by sale of the house for more than the amount of the mortgage

principal. When both the mortgage and housing markets collapsed

in 2007, payment shock led to accelerating default rates.

Servicing income has become an increasingly important source

of income for financial institutions as securitization became the

preferred method of financing mortgages. Securitization enables

financial institutions to originate mortgages in greater volumes

than they are able to finance mortgages. This is true at the firm

level and the industry level. Financial institutions (FI) have trans-

formed their balance sheets from spaces where long-term financing

takes place to refineries where the raw material of MBS and ABS

are distilled into primary components and then refinanced in the

broader capital markets all over the world. What stays behind on

the balance sheets of the FI is the servicing contract. The benefits

that accrue to FI from operating a financial refinery are the fees

associated with originating and servicing financial assets. In this

paper we are examining how servicers can construct delta hedges

to offset losses in value to ARM servicing portfolios from changes in

interest rates. Interest rate changes affect the rates of mortgagor

prepayment and default.

Figure 1 – Percentage change in Case-Shiller national home price index

(Q1-1988 to Q1-2009)

-25.00%

-20.00%

-15.00%

-10.00%

-5.00%

0.00%

5.00%

10.00%

15.00%

20.00%

1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008

Time

% Change

0

2

4

6

8

1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009

6-month LIBOR

1-Year CMT

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

2005 2006 2007 2008 2009

79

can ARMs’ mortgage servicing portfolios be delta-hedged under gamma constraints?

Servicing income is derived from the outstanding mortgage princi-

pal that is being serviced. Servicing fees range from 25 to 38 basis

points of serviced principal. Servicing costs are not constant. The

servicer is responsible for the costs associated with delinquencies,

workouts, and foreclosures on top of the normal costs of operating

servicing centers. The value of servicing is higher when mortgagors

make payments on time. Delinquencies, defaults, and foreclosures

are costly. Workouts, while costly, are also a source of revenue

because while defaults truncate the stream of servicing income,

workouts extend the servicing income. The fact that mortgage

defaults were too many and too fast, “too many to fail” has pro-

moted the government, investors, and lenders to promote the work-

out loans rather than move to foreclose on properties. This makes

sense in a deflating property market. Not only does the relationship

of the fixed costs to variable costs makes the value of servicing

uncertain, the principal amount that is serviced at any point in time

is not known in advance because mortgagors can prepay or default.

While both prepayments and defaults reduce the future stream

of servicing income, the impact of defaults and prepayments are

not equivalent. As already mentioned, a default is more costly to

service than a prepayment. If the prepayment is financed with a

mortgage that is serviced by the same servicer as the original mort-

gage, then servicing income may boost income from the profits

associated with the origination of a new loan. Full defaults simply

end the servicing income because the principal of the loan is settled

with the investors from the sale of the foreclosed property.

The common elements that drive prepayment of ARMs are the rela-

tive cost of available fixed rate financing at a point in time to the

expected cost of floating rate financing going forward. If an ARM is

designed to adjust upward or downward every six months and mort-

gage rates continue to increase after the mortgage is originated,

the cause for refinancing will not be to secure an immediate savings

but rather to lower the expected present value of future funding

costs. Fixed rates will always be above adjustable rates except for

the rare case of a very steep downward sloping yield curve. As

long as fixed rates are above adjustable rates, the exercise of the

ARM prepayment option would be prompted by a mortgagor trying

to protect resources from future increases in the ARM index. The

incentive to refinance an ARM is not one sided as it is for a fixed

rate mortgage. When fixed rates fall by enough from their levels

that existed when the ARM was originated, the mortgagor may

have the incentive to switch into the less risky fixed rate mortgage.

Hybrid ARMs that combine features of fixed rate and adjustable

rate instruments have become popular. Mortgages such as the 5/1

ARM offer fixed rates for a five-year period after which the rates

begin to adjust. There is evidence that at the cusp between the

adjustable rate and fixed rate periods, mortgagors act to avoid a

sharp upward revision in interest payments by attempting to refi-

nance out of ARM into a more affordable loan if one is available

[Pennington-Cross and Ho (2006)].

The flow of subprime risk into the mortgage market increased dra-

matically in the years leading up to the crash in the housing market

and the collapse of the capital and money markets in 2008. For

example, as a percent of total mortgage originations in terms of

principal in 2006, 2007, and 2008, the percentages of MBSs backed

by subprime mortgages that were issued were 16.8%, 9.01%, and

0.13%, respectively. These numbers are important because servic-

ing income became more unstable as it was backed by mortgages

that were speculative in nature. If borrowers were unable to call

their mortgage by refinancing with a more affordable loan then

they often exercised the other embedded option, the right to put

the mortgaged property to the borrower. The current wave upon

wave of foreclosures was triggered by the collapse of the subprime

market. Going forward, what will have an important impact on ser-

vicing income derived from ARMs are the public policy and private

initiatives that are being executed to modify outstanding mortgage

loans, many of which are and will be subprime ARMs. “Currently,

about 21 percent of subprime ARMs are ninety days or more delin-

quent, and foreclosure rates are rising sharply.” (Chairman Ben S.

Bernanke at the Women in Housing and Finance and Exchequer

Club Joint Luncheon, Washington, D.C. January 10, 2008, Financial

Markets, the Economic Outlook, and Monetary Policy)

To arrive at a rough idea of how significant the market for mort-

gage servicing rights is, we assume there is a 25 bps servicing fee

charged against all outstanding first lien mortgage loans financing

1 to 4 family residences, whether securitized or not. The approxi-

mate servicing income generated in 2005, 2006, 2007, and Q1

2008 is provided in Figure 2. Hedging this income correctly at a

reasonable cost can protect the capital of banks and make cash

flows less volatile. Figure 4 illustrates the breakdown between fixed

rate mortgages and adjustable rate mortgages. As of May 2007,

subprime ARMs accounted for two-thirds of first lien subprime

mortgage market and 9% of all first lien mortgages. “After rising

at an annual rate of nearly 9 percent from 2000 through 2005,

house prices have decelerated, even falling in some markets. At the

same time, interest rates on both fixed- and adjustable-rate mort-

gage loans moved upward, reaching multi-year highs in mid-2006.

Some subprime borrowers with ARMs, who may have counted on

refinancing before their payments rose, may not have had enough

home equity to qualify for a new loan given the sluggishness in

2005 2006 2007 Q4-2008 Q2-2009

Outstanding mortgage principal (first lien)

U.S.$ 9.385 trillion

U.S.$ 10.451 trillion

U.S.$ 11.140 trillion

U.S.$ 11.042 trillion

U.S.$ 10.912 trillion

Servicing income = 25 bp x outstanding mortgage principal

U.S.$ 23.462 billion

U.S.$ 26.128 billion

U.S.$ 27.852 billion

U.S.$ 27.606 billion

U.S.$ 27.281 billion

Figure 2 – Estimate of annual mortgage servicing income Source for outstanding mortgage principal is Freddie Mac.

80 – The journal of financial transformation

can ARMs’ mortgage servicing portfolios be delta-hedged under gamma constraints?

house prices.” (Chairman Ben S. Bernanke at the Federal Reserve

Bank of Chicago’s 43rd Annual Conference on Bank Structure and

Competition, Chicago, Illinois May 17, 2007, The Subprime Mortgage

Market)

If we look at the CMT (constant maturity Treasury) rate or the

six month LIBOR we see that rates rise incrementally, placing

increasing pressure on the borrowers’ ability to repay as ARM

rates increase. At some point the mortgagor will have to decide if

further increases in rates are sustainable. Rising ARM indices will

prompt the mortgagor to search for a way out of the increasingly

costly contract. The three paths out are prepayment, modification,

or default. It must be noted that while conventional mortgages are

typically issued without prepayment penalties, subprime mortgages

often do have prepayment penalties attached. These penalties

make prepayment more costly.

Institutions such as Citibank, Wells Fargo, Bank of America, or

Countrywide rely on servicing income and take measures to hedge

this income. We are offering a technique for hedging the nega-

tive fallout to servicing income derived from ARM loans that face

increasing risks of prepayment and default as interest rates rise.

The subprime mortgage loan cohorts that had the worst perfor-

mance are 2005, 2006, and 2007. Subprime mortgages originated in

2007 experienced the most rapid foreclosure rates. The 2007 cohort

loans were originated as home processes began their steep decline

and refinancing became next to impossible (Testimony before

the Joint Economic Committee U.S. Congress HOME MORTGAGES

Recent Performance of Nonprime Loans Highlights the Potential

for Additional Foreclosures Statement of William B. Shear, Director

Financial Markets and Community Investment, For Release on

Delivery Expected at 10:00 a.m. EDT Tuesday, July 28, 2009).

It is estimated that 21.9% of subprime and Alt-A loans set to reset in

the third quarter of 2009 are already 30+ days overdue. This makes

these loans likely candidates for default or prime candidates for

loan workouts. (Data report No. 1, February 2008, State Foreclosure

Prevention Working Group)

To stem the tide of foreclosures, the U.S. Government has initiated

the “Making Home Affordable Program” (HAMP). This program offers

mortgagors the chance to refinance with a more affordable mortgag-

es or to modify a mortgage that is delinquent or at risk of becoming

delinquent. Servicers are offered compensation to facilitate the mod-

ification of loans that qualify for the program. These fees along with

the savings that servicers will gain from avoiding foreclosure and the

increased longevity of the loan make HAMP interesting. The program

is being ramped up so that the rate of foreclosures can be stemmed,

which would help to stabilize the housing market. HAMP supplements

the efforts of lenders who are taking actions to save value by working

with mortgagors. Loan modifications extend the life of mortgages,

enhancing the value of servicing contracts. Rather than lose the

mortgage principal that is being serviced via a foreclosure, the idea

is to modify loans that allow borrowers to stay current with the new

terms. (OCC and OTS Mortgage Metrics Report Disclosure of National

Bank and Federal Thrift Mortgage Loan Data Second Quarter 2009,

Office of the Comptroller of the Currency Office of Thrift Supervision

Washington, D.C. September 2009)

prepayment function for variable rate mortgagesOur model of ARM prepayments is based on a number of general

assumptions about prepayment behavior. First of all, the diversity

of ARM contracts means that it is not possible to make accurate

general statements about all ARMs. The ARM contracts that our

analysis is most applicable to are subprime ARMs that were issued

with initial index discounts, Option ARMs and Hybrid ARMs. All

of these mortgage contracts set up the potential for an extreme

increase in required payments once the teaser period is over or the

fixed rate period is over and the interest is fully indexed to current

money market rates. A glance at Figure 3 shows how rapidly money

market rates to which many ARMs are indexed (the six-month

LIBOR rate and the 1 year CMT) increased between 2004 and 2006.

This rapid increase in interest rates resulted in serious stress on

the ability of many households that had issued ARMs to continue

to make monthly payments. Since the rate increases happened dis-

cretely and index resets are not continuous, mortgagors have time

to make decisions that lower the value of their payment burdens.

These options include prepayment and default [Pennington-Cross

and Ho (2006)]. This increases the likelihood of a delinquency lead-

ing to default or delinquency leading to prepayment.

The following assumptions about mortgagor prepayment and ser-

vicing are integral to our model of delta-hedging servicing income.

The prepayment rate, cpr for ARMs (adjustable rate mortgages), ■■

is a function of the market interest rate y and default rates d. In

our model, the cpr of ARMs increases when interest rates go up,

or are expected to increase because mortgagors have an incen-

tive to lock-in current rates using fixed rate mortgages or search

for alternative ARMs with lower rates. Mortgage defaults impact

the cpr in the same way as prepayments. As a result of this

assumption the cpr function is no longer the S-Shape found with

fixed-rate mortgages [Stone and Zissu (2005)]. The cpr-function

for ARMs that prompt prepayments or defaults in rising interest

rate environments becomes an inverted S-shape (mirror of the

prepayment function for fixed rate mortgages). As interest rates

decline relative to the contract rate, borrowers will still have

an incentive to refinance when the obtainable fixed rate is less

costly than the expected adjustable rates in the future.

Default rates on ARMs are positively correlated with rising inter-■■

est rates and diminishing home values.

81

can ARMs’ mortgage servicing portfolios be delta-hedged under gamma constraints?

The magnitude and value of income derived from mortgage servic-■■

ing rights (MSR) depend on default rates and prepayment rates.

We delta-hedge a portfolio of mortgage servicing rights for ARMs

with other fixed income securities such that the value of the servic-

ing portfolio is not affected by increases or decreases in market

rates. In order to obtain the portfolio that requires the lowest cost

delta hedge, we compare hedge ratios dynamically. This paper

applies Ortiz et al.’s model of a delta-hedge-ratio function applied to

a portfolio of fixed rate mortgages to a portfolio of servicing rights

derived from a pool of ARMs. We develop the gamma-hedge-ratio

function using three types of fixed-income securities, a coupon pay-

ing bond with n years to maturity (case 1), a zero coupon bond with

n years to maturity (case 2), and a bond that pays coupons only

twice in the life of the bond, at n/2 and then at n (case 3).

“The fair value of the MSRs is primarily affected by changes in

prepayments that result from shifts in mortgage interest rates. In

managing this risk, Citigroup hedges a significant portion of the value

of its MSRs through the use of interest rate derivative contracts,

forward purchase commitments of mortgage-backed securities, and

purchased securities classified as trading (primarily mortgage-backed

securities including principal-only strips).” (Citigroup’s 2008 annual

report on Form 10-K). We have selected cash purchases of bonds to

affect our delta hedge. Our research can be extended to incorporate

the instruments that banks like Citigroup employ to hedge MSR.

The option to prepay an ARM is not as straight-forward as it is

for a fixed rate mortgage. Mortgagors have financial incentives to

prepay ARMs when interest rates are either rising or falling. This is

because ARM mortgages shift the interest rate risk to the borrower.

Our contribution is to set up a delta hedge for MSR that diminishes

as mortgagors try to avoid further rate increases.

“The adjustable-rate mortgage loans in the trust generally adjust

after a one month, six month, one year, two year, three year, five

year, or seven year initial fixed-rate period. We are not aware of

any publicly available statistics that set forth principal prepay-

ment experience or prepayment forecasts of mortgage loans of

the type included in the trust over an extended period of time, and

the experience with respect to the mortgage loans included in the

trust is insufficient to draw any conclusions with respect to the

expected prepayment rates on such mortgage loans.” (Prospectus

supplement to prospectus dated June 28, 2006, DLJ Mortgage

Capital, Inc. Sponsor and seller adjustable rate mortgage-backed

pass-through certificates, series 2006-3).

Model and applicationsWe express the relationship between prepayment rate and the

change in basis points in equation (1). The prepayment rate, cpr, is

primarily a function of the difference between the new mortgage

rate y in the market and the contracted mortgage rate r.

cpr = a/(1 + eb(y-r)) (1)

Figure 5 shows the relationship between the spread between mar-

ket rates y and contractual rates on an adjustable-rate mortgage r

and the cpr of the mortgage. The greater the spread is, the higher

the incentive mortgagors will have to switch from an ARM to a

fixed rate mortgage or default on their loan ARM if new affordable

financing is not available. The default option is typically exercised

when home equity has become negative. Declining home equity can

be the result of either an increase in the value of the mortgage lia-

bility or a decline in the value of the mortgaged property or both.

Of course, declining home values are the fundamental cause of

negative home equity. Default rates have been stronger than refi-

nancing rates during the subprime crisis. Default rates by subprime

borrowers have swamped refinancing rates by these borrowers

because of rapidly declining values of home equity and increasing

joblessness, even as rates have come down to historical lows.

Kalotay and Fu (2009) illustrate that the option value offered to

mortgagors differs across mortgage contracts. He shows that the

5/1 ARM has a lower option value than the fixed rate mortgage and a

higher option value than the 1 year ARM. The option value increases

Figure 3 – ARM indicies, 6-month LIBOR rates, and 1-year CAT Figure 4 – Conventional 30-year mortgage rate conventional Treasuy indexed 5-1

hybrid adjustable rate mortgages

-25.00%

-20.00%

-15.00%

-10.00%

-5.00%

0.00%

5.00%

10.00%

15.00%

20.00%

1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008

Time

% Change

0

2

4

6

8

1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009

6-month LIBOR

1-Year CMT

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

2005 2006 2007 2008 2009

-25.00%

-20.00%

-15.00%

-10.00%

-5.00%

0.00%

5.00%

10.00%

15.00%

20.00%

1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008

Time

% Change

0

2

4

6

8

1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009

6-month LIBOR

1-Year CMT

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

2005 2006 2007 2008 2009

82 – The journal of financial transformation

can ARMs’ mortgage servicing portfolios be delta-hedged under gamma constraints?

as the mortgage incorporates more elements of the 30-year fixed

rate mortgage. A lower option value has generally translated into

a lower initial interest rate for the borrower. (Consumer Mortgage

Decisions by Andrew J. Kalotay and Qi Fu, Research Institute for

Housing America June 2009)

Valuation of MsR The cash flow of a MSR portfolio at time t is equal to the servicing

rate s times the outstanding pool in the previous period:

MSRt = (s)m0(1 – cpr)t-1Bt-1 (2)

where:

m0 = number of mortgages in the initial pool at time zero

B0 = original balance of individual mortgage at time zero

r = mortgage coupon rate

cpr = prepayment rate

m0(1-cpr)t = number of mortgages left in pool at time (t)

Bt = outstanding balance of individual mortgage at time (t)

s = servicing rate

We express the value of mortgage servicing rights as:

V(MSR) = (s)m0 [Σ(1 – cpr)t-1Bt-1] ÷ (1 + y)t (3)

with t = 1,…..n (through the entire paper).

Equation (3) values a MSR portfolio by adding each discounted cash

flow generated by the portfolio to the present, where n is the time

at which the mortgages mature, and y is the yield to maturity.

After replacing equation (1) in equation (3) we obtain the MSR

function as:

V(MSR) = (s)m0 [Σ(1 – (a/(1 + expb(y-r)))t-1Bt-1] ÷ (1 + y)t (3a)

We use the following values for Equation (3a) to generate the MSR

over a range of market rates y and we present them in Figure 6:

m0 = 100

B0 = $100,000

r = 6%

a = .4

b = 100

cpr = prepayment rate

s = .25%

n = 30

Before commenting on Figure 6, it is important to understand that

a mortgage servicing right is equivalent to an interest only secu-

rity (IO). The ultimate cash flow that an IO generates over time is

directly tied to the amount of underlying principal outstanding at

any point in time. An interest only security is derived by separating

interest from principal cash flows when the mortgagor’s payment

is made, before distributing it to investors. The interest component

is distributed to the IO investors and the principal component is

distributed to the PO (principal only) investors. The typical graph of

an IO security over market rates shows an increase in the value of

MSR as market rates increase because prepayment rate decreases

(prepayment effect is stronger than discount effect) until its value

reaches a maximum (prepayment effect is equal to discount effect)

and then it starts to decrease (discount effect is greater than pre-

payment effect).

-0.06 -0.04 -0.02 0(y-r)

0.02

0.02

0.04

0.06

0.08

0.10cpr(y)

0.12

0.14

0.16

0.18

0.04 0.06

Plot of cpr vs (y-r)

Figure 6 – Value of mortgage servicing rights

Figure 5 – Prepayment function

0 0.05 0.10 0.15y

$

5.5 x 106

5 x 106

4.5 x 106

4 x 106

Plot of function V(IO) vs y

83

can ARMs’ mortgage servicing portfolios be delta-hedged under gamma constraints?

The value of MSR derived from a pool of ARMs with respect to chang-

es in market interest rates differs from the value of MSR for fixed

rate mortgages because the prepayment function is inverted. This

is illustrated in Figure 6. When market rates start to increase, mort-

gagors who anticipate that further rate increases are on the horizon

will attempt to refinance their mortgages with fixed rate mortgages.

Those who cannot secure a fixed rate mortgage at acceptable terms

may have to endure further rate increases which will lead to higher

default rates. The effect of rising interest rates for certain classes of

subprime ARMs will be a diminution of outstanding principal. Since

this principal is the source of servicing fees and the discount rate is

now higher the value of MSR must fall. We are discounting a smaller

stream of cash flows at a higher rate.

ARMs do not all reset at the same margin above an index. Rates

that were initially set at a discount from market value at some

point will reset to the fully indexed margin. An extreme change

in rates when the teaser period is over leads to what is known as

prepayment shock. Anticipation of payment shock increases the

incentive to prepay. In 2004, the Fed began raising the Fed funds

rate. In December 2003, the rate was 1%; by June 2007 it had risen

to 5.25%. This increase in money market rates led to increases in

ARM indices, a lower demand for mortgage finance, higher default

rates, and the beginning of the reduction in the supply of mortgage

credit. A negative feedback loop was set in motion that led to falling

house prices that increased default rates that lowered the supply of

mortgage credit even further that placed further downward pres-

sure on house prices.

The LIBOR index rose by more than the CMT index, especially in

2008 when interbank lending became severely curtailed. Rising

interest rates can lead to payment shock for initially discounted

ARMs. If mortgagors who have issued ARMs expect rates to con-

tinue rising they can see that their household finances will become

stressed. While prepayment into a fixed rate mortgage will not

necessarily lower payments, it may lower the value of the mortgage

liability relative to the initial ARM by shielding the borrower from

further rate increases.

Rapidly increasing interest rates will increase the incentive to

refinance and the likelihood to default. The incentive to get out of

ARMs that are becoming too costly will depress the value of MSR.

Declining interest rates and lower defaults rates will enhance the

value of MSR in our model over a range of rates. Notice that the plot

of MSR in Figure 6 is far from linear. At first, rising rates diminish

the value of the MSR, the IO strip. Once MSR reaches a minimum,

further rate increases lead to an increase in the value of MSR. This

is explained by the absence of refinancing opportunities that offer

savings and perhaps coupled with rising home values. Again at

some point the discount rate effect takes over and the discounted

value of MSR begin to decline.

The delta hedge ratioWe use the same model (OSZ) [Ortiz et al. (2008)] developed

to obtain a delta-hedged portfolio of fixed rate bonds and MSRs

derived from fixed rate mortgages. In this paper we are hedging

MSRs that are derived from ARMs with fixed rate bonds with the

difference that the MSR are backed by adjustable rate mort-

gages.

αV(MSR) + βV(B) = K (4)

Where K is the constant value of the portfolio of MSRs and bonds;

and α and β are the shares of MSRs and bonds respectively that are

consistent with a zero-delta portfolio that satisfies the constraint K.

OSZ obtain the hedge ratio α as a function of the MSR’s and bond’s

deltas and their respective values:

α = [-K(dV(B)/dy)] ÷ [(dV(MSR)/dy)V(B) – (dV(B)/dy)V(MSR)] (5)

osZ simulations and analysisIn the following three cases we use the OSZ model to create the

delta hedge of ARM mortgage serving rights.

case 1

A regular bond with yearly coupons and face value paid at maturity:

V(B) = c Σ 1/(1+y)t + Face/(1+y)n

c = $350,000

Face = $5,000,000

y = 5%

n = 10

case 2

A zero-coupon bond

V(B) = Face/(1+y)n

We keep the value of the variables the same as before:

Face = $5,000,000

y = 5%

n = 10

case 3

A bond that pays coupons only twice in the life of the bond: at n/2

and then at n:

V(B) = c/(1+y)n/2 + c/(1+y)n + Face/(1+y)n

c = $350,000

n = 10

84 – The journal of financial transformation

can ARMs’ mortgage servicing portfolios be delta-hedged under gamma constraints?

Face = $5,000,000

y = 5%

α corresponds to the share of MSR in the portfolio of MSR and

bonds that delta-hedges the portfolio of MSR for small changes

in interest rates. A high α means that less bond principal must

be positioned to delta-hedge the MSR portfolio, making for a less

costly hedge. The optimal hedge is the one that requires the highest

α bond portfolio.

We now examine Figure 7, which plots on the vertical axis the value

of α and on the horizontal axis the market interest rate available

to mortgagors who have issued ARMs but are now considering

prepaying their mortgage. The interest rate on the underlying ARM

mortgage pool is 5%. Again α is the share of the MSR in the hedged

portfolio.

We have divided the horizontal axis into three zones.

Zone 1 — starts at a market rate of 0% and ends where the three α

functions cross the horizontal axis for the first time. This crossover

point corresponds to the minimum value of the MSR in Figure 6. In

zone 1 the maximum α (most efficient hedge) is achieved by pur-

chasing zero coupon bonds. As market rates begin to rise, the value

of MSR derived from certain pools of ARMs will begin to decline as

defensive prepayments begin to increase.

Zone 2 — corresponds to the range of market rates where the three

α functions are negative. In this range, the optimal delta hedge is

obtained by shorting n year coupon bonds.

Zone 3 — starts where the three α functions cross for the second

time the horizontal axis; that point corresponds to the maximum

value of MSR in Figure 6. In that zone, the highest level of α is

achieved by adding to the MSR zero-coupon bonds (case 2).

conclusionAdjustable rate mortgages shift, to varying degrees, interest rate

risk from lenders/investors to borrowers. The hybrid mortgage

combines elements of the fixed-rate mortgage and the adjustable

rate mortgage by offering borrowers a fixed period prior to the

adjustable period. Adjustable-rate mortgages were a significant

component of the subprime mortgage market. The value of mort-

gage servicing rights is extremely sensitive to changes in market

rates and of course to default rates which are directly impacted by

market rates. Holders of MSR can delta-hedge their MSR portfolios

against interest rate risk. This paper simulates three different port-

folios of MSR and bonds. While the servicer will sometimes choose

to take a long position in zero-coupon bonds and sometimes to

short the n year coupon paying bond, the intermediate case is never

selected. The intermediate bond has a shorter duration than the

zero-coupon bond and a longer duration than the n year coupon

bond.

We have illustrated that the financial instrument that servicers

must use to effectively delta-hedge a portfolio of servicing rights

that is losing value as interest rates rise changes as the spread

between the market mortgage rate and contract rate changes.

Hedging servicing rights derived from ARMs that expose the mort-

gagor to payment shock as fixed rates becomes adjustable and the

contract rate is fully marked to a margin above the ARM index. In

addition to the changing incentive to refinance into a more afford-

able mortgage there is the increasing risk of default if the opportu-

nity to refinance is not available. The hedging of AMRs in our model

also incorporates the discount affect on future servicing income.

A very import dimension of the market for mortgage servicing

rights and particularly servicing income derived from subprime

mortgages, and even more specifically from subprime ARMs, is the

effort the government and the private sector are making to modify

loans before they default. These efforts will boost the value of MSRs

because the servicers will continue to service principal that would

have otherwise been lost to foreclosures. In addition, servicers that

participate in the government’s HAMP program will retain servicing

income that could have been lost to other lenders/servicers. 0.05 0.10 0.15

Plots of alpha for the three cases

-0,2

0

0.2

0.4

0.6

Case 1: red line (n-year annual coupon bonds)Case 2: blue line (n-year zero coupon bonds) Case 3: green line (bond with duration less than case 1 and greater than case 2)

Figure 7 – The share of MSR (α) in a delta-hedged portfolio

85

can ARMs’ mortgage servicing portfolios be delta-hedged under gamma constraints?

ReferencesBierwag, G. O., 1987, Duration analysis: managing interest rate risk, Ballinger Publishing •

Company, Cambridge, MA

Boudoukh, J., M. Richardson, R. Stanton, and R. F. Whitelaw, 1995, “A new strategy •

for dynamically hedging mortgage-backed securities,” The Journal of Derivatives, 2,

60-77, 1995

Ederington, L. H., and W. Guan, 2007, “Higher order Greeks,” Journal of Derivatives, •

14:3, 7-34

Eisenbeis, R. A., W. Frame Scott, and L. D. Wall, 2006, “An analysis of the systemic •

risks posed by Fannie Mae and Freddie Mac and an evaluation of the policy options for

reducing those risks,” Federal Reserve Bank of Atlanta, Working Paper 2006-2

Goodman, L. S., and J. Ho, 1997, “Mortgage hedge ratios: which one works best?”, The •

Journal of Fixed Income, 7:3, 23-34

Kalotay, A. J., and Q. Fu, 2009, “Consumer mortgage decisions,” Research Institute for •

Housing America, June

Office of Thrift Supervision, 2007, “Hedging – mortgage servicing rights,” examination •

book, 750.46-750.52, July

Ortiz, C., C. A. Stone, and A. Zissu, 2008, “Delta hedging of mortgage servicing portfo-•

lios under gamma constraints” Journal of Risk Finance, 9:4, 379-390

Ortiz, C., C. A. Stone, and A. Zissu, 2009, “Delta hedging a multi-fixed-income-securities •

portfolio under gamma and vega constraints,” Journal of Risk Finance, 10:2, 1691-72

Pennington-Cross, A., and G. Ho, 2006, “The termination of subprime hybrid and fixed •

rate mortgages,” in Research Division Federal Reserve Bank of St. Louis Working Paper

Series, July

Posner, K., and D. Brown, 2005, “Fannie Mae, Freddie Mac, and the road to redemp-•

tion”, Morgan Stanley, Mortgage Finance, July 6

Stone, C. A., and A. Zissu, 2005, The securitization markets handbook: structure and •

dynamics of mortgage- and asset-backed securities, Bloomberg Press

Articles

87

AbstractStructure and stability of private equity market risk are still nearly

unknown, since market prices are mostly unobservable for this

asset class. This paper aims to fill this gap by analyzing market

risks of listed private equity vehicles. We show that aggregate

market risk varies strongly over time and is positively correlated

with the market return variance. Cross-sections of market risk are

highly unstable, whereas ranks of individual vehicles within yearly

subsamples change slightly less over time. Individual CAPM betas

are predictable only up to two to three years into the future and

quickly converge to a stationary distribution when measured in

risk classes in an empirical Markov transition matrix. We suspect

that market risk of private equity is affected by factors unique to

this sector: acquisitions and divestments that constantly rebalance

portfolios, scarcity of information about portfolio companies, and

rapid changes within portfolio companies. Unstable market risk

seems to be a fundamental characteristic of private equity assets,

which must be incorporated into valuation and portfolio allocation

processes by long-term private equity investors. Large increases in

systematic risk in recent years cast doubt on diversification ben-

efits of private equity in times of crisis.

christoph KasererFull Professor, Department of Financial

Management and Capital Markets, Center for Entrepreneurial and Financial Studies (CEFS),

Technische Universität München

Henry lahrCenter for Entrepreneurial and Financial Studies

(CEFS), Technische Universität München

Valentin liebhartCenter for Entrepreneurial and Financial Studies

(CEFS), Technische Universität München

Alfred MettlerClinical Associate Professor, Department of

Finance, J. Mack Robinson College of Business, Georgia State University

The time-varying risk of listed private equity

88 – The journal of financial transformation

The time-varying risk of listed private equity

Performance measurement and portfolio allocation are notoriously

difficult when dealing with private equity assets due to the non-

transparent nature of this asset class. To evaluate the success of an

investment, its risk premium as derived from factor pricing models

can serve as a benchmark for required returns in excess of the risk-

free rate. Private equity’s market risk, albeit hard to measure in

traditional private equity funds, can be obtained from listed private

equity (LPE) vehicles. In this paper, we measure aggregate and indi-

vidual market risk and its variability over time. Private equity inves-

tors can benefit from a quantification of private equity market risks

and their variability, since this asset class represents a substantial

share of international investment opportunities. The private equity

sector had more than U.S.$2.5 trillion of capital under management

in 2008 according to the International Financial Services London

[IFSL (2009)]. This large volume demands for a time variation anal-

ysis of systematic risks. In addition, industry-specific characteristics

caused by private equity business models shape the evolution of

risk within this asset class: acquisitions and divestments of portfolio

companies constantly rebalance fund portfolios, which should lead

to highly unstable market risk.

Several authors have focused on non-constant risk premia in equi-

ties, bonds, and REITs. Early papers discussed the impact of risk

variability on portfolio decisions [Levy (1971), Blume (1971)]. Later

work developed the conditional capital asset pricing model and sim-

ilar models using mostly public market data [Lettau and Ludvigson

(2001), Ghysels (1998), De Santis and Gérard (1997), Jagannathan

and Wang (1996), Bollerslev et al.(1988), Ferson et al. (1987)]. Time-

varying risk properties of private equity, however, have not been

examined by empirical research.

The difficulty with risk measurement in traditional (unlisted) private

equity vehicles lies in the opacity of their price formation. Time

series of returns are hardly observable, which renders estimation of

market risk nearly impossible. Many attempts at measuring system-

atic risk are based on voluntarily reported returns of private equity

funds, internal rate of return (IRR) distributions [Kaplan and Schoar

(2005)], cash flows [Driessen et al. (2009)], transaction values of

portfolio companies [Cochrane (2005), Cao and Lerner (2009)], or

the matching of portfolio companies to public listed companies with

similar risk determinants [Ljungqvist and Richardson (2003), Groh

and Gottschalg (2009)].

Private equity vehicles that are listed on international stock

exchanges provide a natural alternative to unlisted ones when esti-

mating their risk. Return data are readily available and can be used

to answer risk-related questions: what are the market risk patterns

of listed private equity throughout the life of the security? How

stable are the market risks of the listed private equity companies?

We first analyze the market risk structure of listed private equity.

For this purpose, we measure market risk in an international capital

asset pricing model (CAPM) using Dimson (1979) betas. While Lahr

and Herschke (2009) measure constant betas over the lifetime of

listed private equity vehicles, we take a step further in considering

their time series properties. In our model, market risk is measured

over a rolling window to generate a continuous set of beta obser-

vations, which describes the aggregate asset class risk over time.

Second, we examine the stability of individual betas. Correlations of

cross-sections for consecutive years can offer insights into relative

changes of risk within the asset class. We find that market risk of

listed private equity is quite unstable if time periods longer than

two years are considered. In a second step, we compute transition

probabilities between risk classes. Our results reflect the instability

of risk in general, but highlight a moderate persistence of excep-

tionally high and low risks.

stability of market riskA broad picture of how market risk evolves in listed private equity

can be seen from yearly cross-sections. In this section, we show the

main properties of market risk measurement for our listed private

equity sample for different time horizons: rolling windows of one

and two years and total lifetime. Our sample of listed private equity

vehicles is based on the data from Lahr and Herschke (2009). They

generate a comprehensive list of listed private equity companies,

which we extend to account for recent listings. Our final sample

includes 278 vehicles, the largest proportions being investment

companies and funds that are managed externally. The time hori-

zon chosen for our analysis is January 1st, 1989 to March 25, 2009,

although not all companies have returned data during the entire

time period due to new listings and delistings.

To measure the market risk of our sample vehicles, we use individual

return indices from Thomson Datastream in U.S. dollars, the 3-month

Treasury bill rate as a proxy for the risk-free rate, and MSCI World

index returns in U.S. dollars as a proxy for market returns. All return

data are converted to logarithmic returns. During the time period

studied in this paper, 33 companies were delisted. All companies

enter our analysis when return data becomes available and drop out

if they delist or if trading volume is zero after some date.

Market risk estimation

To obtain market risks we regress excess LPE stock returns on

excess market returns (MSCI World). We employ a Dimson regres-

sion to account for autocorrelation in asset returns caused by

illiquidity in LPE vehicles. Early studies showed that in similar set-

tings autocorrelation on the portfolio level can be a problem [Fisher

(1966), French et al. (1987)]. We use the results of Dimson (1979)

and incorporate 7 lagged market returns in the estimation model to

adjust for autocorrelation. In a second step we aggregate the lags

as proposed by Dimson to obtain a measure of market risk. Since

our sample consists of vehicles traded on international exchanges,

89

The time-varying risk of listed private equity

we account for currency risk by introducing risk factors for the

euro, yen, and pound sterling, which represent excess returns of

currency portfolios in U.S. dollar. The international capital asset

pricing model is thus given by

ri,t = αi + βi,krm,t−k +γ i,1GBPt +k=0

7

∑ γ i,2EURt +γ i,3YPYt +εi,t

where ri,t and rm,t—k are the respective observed logarithmic

(excess) individual vehicle and market returns at time t and t-k,

whereas k corresponds to the respective lag. The intercept α rep-

resents a vehicle-specific constant excess return, β and g are the

slope coefficients, and e is an error term. We use this regression

equation to calculate the market risks for different time periods:

one year, two years, and lifetime market risks.

Yearly cross-sections and aggregate market risk

We first illustrate the behavior of aggregate market risk in a time

series context before taking a closer look at individual risk stability.

Time series of aggregate risk can be constructed from measures

of market risk in rolling windows. We define two such rolling win-

dows: the first spans 52 weekly observations and the second has

104 observations. Similar windows are used by Bilo (2002), who

measures historical return variances but does not examine the time

series properties of systematic risk.

Table 1 summarizes the main market risk statistics for different

observation windows. Mean one-year betas range from a minimum

of 0.22 in 1993 to a maximum of 1.36 in 2000. Two-year betas

are highest for the periods 1999-2000 and 2007-2008. One-year

betas as well as two-year betas vary around an average that is

almost equal to unity. All periods and estimation windows exhibit a

large cross-sectional variation in market risk. This might be caused

by the huge diversity within the listed private equity asset class,

which includes many small vehicles with strongly differing business

models. Interestingly, mean betas are positively correlated to their

standard deviation. This suggests that mean betas are driven by

vehicles with huge betas as indicated by a skewed distribution of

betas.

Figure 1 shows a more detailed picture over time. Listed private

equity betas are first estimated for each vehicle in rolling windows.

These individual betas are then weighted equally over all vehicles

that we were able to calculate a beta for at a given point in time.

Figure 1 reveals the volatile nature of private equity market risk.

Even mean betas vary widely over time. Beta variability is smaller

lifetime beta Deciles

Mean sD no. > 0 no. < 0 Min 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 Max

1.31 0.93 270 8 -2.89 0.32 0.62 0.85 1.03 1.18 1.41 1.62 1.88 2.44 5.28

Yearly cross-sections

one-year betas Two-year betas

period ending Mean sD Max-Min no. < 0 n period ending Mean sD Max-Min no. < 0 n

1989 0.51 1.58 6.74 13 35

1990 0.84 0.86 4.01 6 41 1990 0.93 0.71 2.69 5 41

1991 0.84 1.40 8.45 9 48

1992 0.26 1.74 8.72 22 53 1992 0.56 1.06 6.75 12 53

1993 0.22 1.52 8.25 27 57

1994 0.52 1.68 8.77 20 61 1994 0.65 1.32 8.14 16 61

1995 0.64 2.47 16.61 25 69

1996 1.15 4.32 38.72 27 74 1996 0.86 2.06 15,00 23 74

1997 0.97 2.23 16.94 32 88

1998 1.16 1.33 7.34 14 103 1998 1.26 1.26 6.45 7 103

1999 1.31 2.53 20.75 32 120

2000 1.36 2.31 17.89 32 135 2000 1.33 1.30 8.49 17 135

2001 1.19 1.36 9.96 21 160

2002 1.16 1.45 11.49 29 177 2002 1.19 1.16 7.12 11 177

2003 1.09 2.09 15.21 42 182

2004 1.35 2.95 30.14 47 183 2004 1.30 2.07 16.94 42 182

2005 1.05 2.63 20.78 58 190

2006 0.95 1.51 13.83 40 203 2006 1.16 1.29 11,00 23 202

2007 1.04 2.12 23.33 52 226

2008 1.20 1.33 12.02 30 244 2008 1.37 1.29 8.15 18 243

Mean 0.94 1.97 15.00 Mean 1.06 1.35 9.07

Table 1 – Beta cross-sections from 1989 to 2008

90 – The journal of financial transformation

The time-varying risk of listed private equity

for two-year betas due to smoothing but higher at the beginning

of our observation period. This higher variability could be caused

by the smaller number of vehicles compared to later years, which

makes the sample mean a less efficient estimator for the true mean.

The one-year beta time series exhibits a mean-reverting behavior,

oscillating between low values around zero during the early 1990s

and peaking in 1996 and 2000. Its long term average, however, is

a moderate 0.98. Two-year betas behave similarly around a time

series mean of 1.07. They are lower than the market average during

the 1990s and again from 2006 through 2008 with a large hike dur-

ing the financial crisis. Both charts have more or less pronounced

peaks during the Asian Financial Crisis (1997-1998), the dot.com

bubble (1999-2001), the year 2004, and the recent financial crisis

(2007-2009).

Exogenous shocks and extreme market movements are possible

causes for beta variability. The green lines in Figure 1 show market

return variances for the corresponding rolling windows (one-year

and two-year) for weekly MSCI World return data. Betas and the

market return variance are significantly correlated with a coeffi-

cient of 0.28 (p<0.05), which is surprising, since betas are inversely

related to the market’s variance in a CAPM context by definition.

This result suggests a large increase in covariance between listed

private equity vehicles and the market in times of uncertainty.

If systematic risk of private equity is about the same as the mar-

ket’s risk, and even worse rises in times when investors seek portfo-

lio insurance, the often purported benefits of this asset class might

turn out to be hard to achieve.

Do risks move together?Although aggregate market risk seems to be rather unstable over

medium to long time periods, individual risks might still move

parallel to the general mean. There could thus be considerable

relative stability within the listed private equity asset class despite

its apparent irregular behavior. We measure risk stability relative

to the listed private equity asset class by estimating Pearson cor-

relation coefficients. Betas can be huge in magnitude, which can

strongly influence linear estimators such as Pearson correlation

coefficients. Spearman’s rank correlation coefficient can provide a

robust measure of relative beta stability.

We first calculate the Pearson correlation to capture beta move-

ments between two points in time. Correlation coefficients are

calculated from all vehicles with a risk estimate available for two

consecutive years. Figure 2 shows that one-year betas are correlated

especially at lags one and two, but only for observation periods after

1993. Betas prior to 1994 seem to behave randomly. Correlations of

one-year betas after 1995 for lag one vary between 0.16 and 0.3.

Interestingly, betas are not always significantly correlated even for

recent years. There is, for example, almost no correlation between

2006 and 2008. An explanation for weak relations in general could

be mean reversion. LPE vehicles tend to have a beta around an indi-

vidual mean. If market risk deviates from this mean in a given year

due to an exogenous shock affecting an individual vehicle, it tends to

drift back to the individual mean after some time. This would reduce

the correlation between individual betas.

Another explanation is real changes in the vehicles’ underlying

businesses, which cause beta variability. Private equity funds buy

portfolio companies to realize capital gains and management fees

over some holding period, which is usually less than ten years,

-0,5

0

0,5

1

1,5

2

2,5

3

0

0,00035

0,0007

0,00105

0,0014

0,00175

0,0021

0,00245

0

0,2

0,4

0,6

0,8

1

1,2

1,4

1,6

1,8

2

1990 1992 1994 1996 1998 2000 2002 2004 2006 2008

Market risk

0

0,00015

0,0003

0,00045

0,0006

0,00075

0,0009

0,00105

0,0012

0,00135

0,0015

Variance of weekly returns of the MSCI World

Total

Mean

Variance MSCI

Market risk Variance of weekly returns of the MSCI World

1990 1992 1994 1996 1998 2000 2002 2004 2006 2008

Figure 1 – Mean one-year beta (upper panel) and two-year beta (lower panel) from

1989 to 2009.

8990919293949596979899000102030405060708

90

92

94

96

98

00

02

04

06

08

89

90 91

92

93

94

95

96

97

98

99

00 01

02

03

04

05

06

07

08

90 92

94

96

98

00

02

04

06

08

One-year betas

-0.4 0.05 0.5

Two-year betas

Figure 2 – Correlation of cross-sectional betas between observation periods.

Pearson correlation coefficients are below the main diagonal, Spearman rank

correlation coefficients are above.

91

The time-varying risk of listed private equity

depending on the funds’ focus. Portfolio turnover is thus higher

than in other companies that grow organically or merge with newly

acquired subsidiaries. If the portfolio companies’ market risk is

diverse across portfolio acquisition, private equity betas experience

jumps according to the market risk of the often substantial amounts

of assets acquired or sold.

The portfolio companies’ risk itself might not be constant either.

Private equity funds and firms that specialize in turnaround man-

agement and venture capital in particular can experience large

swings in market risk. Restructuring can affect operational risk as

well as market risk, while rapid growth of successful companies

causes changes in portfolio risk even if individual portfolio com-

pany risk remains constant. This effect also depends on portfolio

size. Acquisitions and divestments must be large relative to port-

folio size to cause substantial risk changes on the portfolio level.

It seems reasonable to expect such a rebalancing effect over the

medium to long term.

Short-term beta variability is likely caused by estimation errors,

which in turn arise for primarily two reasons. First, listed private

equity vehicles are often quite small by market capitalization and

illiquid. The low informational quality of prices in thinly traded

stocks — although mitigated by the Dimson regression — can carry

over to insignificant or unstable beta coefficients. Second, the

nontransparent structure and characteristics of portfolio com-

panies can reduce informational efficiency. Since most portfolio

companies are not listed and investors in private equity must rely

on the information provided by the fund manager, this information

can sometimes be as scarce as unreliable. If company betas are

seen as a moving average of partially unreliable information about

true market risk, betas become increasingly unstable over short

horizons.

Rank correlations

As a robustness check for the Pearson correlation matrices, we

calculate the Spearman rank correlation to capture the rela-

tive rank movements. Instead of correlating betas directly, the

Spearman correlation computes a coefficient based on the ranks

of individual betas in two consecutive time periods. This calculation

yields a matrix with yearly entries from 1990 to 2008 in the one-

year case and with two-year intervals where betas are estimated

over 104 weeks. Estimation of correlations is based on all vehicles

with observable betas in two consecutive periods, which leads to a

changing number of degrees of freedom.

The entries above the diagonal in both panels in Figure 2 show rank

correlation matrices for one-year and two-year betas. Similar to the

one-year Pearson correlation matrix, betas are correlated at the

first few lags in the one-year case. The highest correlation for one-

year betas is 0.319 at the first lag. Betas begin to be significantly

correlated from 1996 on, which is partly due to their higher magni-

tude and partly due to increasing degrees of freedom.

Results are different in the two-year case but similar to Pearson

correlations. Coefficients are about 0.2 higher than in the one-

year beta case. This is likely the result of better estimates due

to the increasing degrees of freedom when estimating beta over

104 weeks, which makes estimates less sensitive to outliers in the

return distribution. If betas are measured more accurately, correla-

tions increase as well.

Our results suggest that opacity in portfolio companies and illi-

quidity lead to estimation error on the short run, while portfolio

rebalancing changes individual betas over the medium to long term.

Estimated market risk seems to be most stable over horizons span-

ning two to three years. The two-year beta correlations are slightly

higher than in the one-year case, while significant correlations can

be observed over the last decade only.

Evolution of risk over timeThe time series perspective can be combined with an assessment

of cross-sectional stability if one assumes that individual betas are

generated by a Markov process common to all vehicles. We estimate

an empirical Markov transition matrix under the assumption that

future betas depend only on their current value. Transition prob-

abilities from one risk class to another within the empirical Markov

Transition Matrix (MTM) are calculated as the relative frequency

of moving from one risk class to another in the next observation

period. For this purpose, we construct risk classes in two different

ways: betas deciles and fixed beta classes.

persistency in beta deciles

The deciles-oriented MTM is based on quantiles of the distribution

of individual company market risk. All companies in a decile are

assigned to one risk class. As a result of changes in aggregate beta

over time, decile boundaries may change as well. Since betas are

measured with error, boundaries of upper and lower deciles fluctu-

ate most. Quantiles around the median are more stable.

Because a company’s risk class is assigned by the decile function, its

risk class depends on the risk exposure of the entire listed private

equity market. An increase in beta, however, does not change the

risk class if all companies are affected by this increase to the same

extent. This property has an important influence on the interpreta-

tion of our results. The probability of a transition from one risk class

into another does not reflect changes of absolute risk exposure but

gives insight into how the risk of one company behaves compared

to the risk of all other companies. If, for example, a flat transition

matrix is found, betas are equally likely to move from one risk class

to any other. In other words, the relative risk structure would be

completely random over time. This could hold for companies in the

92 – The journal of financial transformation

The time-varying risk of listed private equity

venture capital market in particular, whose risks are driven by real

options and less by fundamental data. If, to the contrary, only posi-

tive entries along the main diagonal exist, relative risks within the

industry do not change over time.

When interpreting results, we have to take care of the fact that

new companies get listed and some become delisted. If we allow

companies to move in and out of our sample, decile boundaries can

change even without any variation in risks, which may force a com-

pany to change its risk class. This can be a problem if the sample

is not random.

Transitions probabilities shown in Figure 3 confirm our results

for cross-sectional rank correlations. Betas are moderately stable

relative to the listed private equity sector over short time periods.

Interestingly, high-risk companies remain in the highest risk class

with a 24% (one-year) and a 25% (two-year) probability. This

suggests that there might be companies that have a persistently

higher risk than other companies. The same result can be seen for

low risk classes. For both observation horizons, risk is comparably

stable (17.8% for the one-year betas and even 18.3% for the two-

year betas). Note that in both cases a relatively high proportion of

companies switch from the highest risk class into the lowest risk

class from one period to the next and vice versa. These companies

are most likely outliers, whose beta cannot be estimated reliably

and therefore is unstable compared to the industry as a whole.

Considering the almost random correlation structure prior to 1994

in Figure 2, we calculated transition matrices excluding these early

betas, but find similar results.

The flat probability structure in deciles 5 and 6 can be observed

for several reasons: first, it can be driven by listings and delistings.

These changes in the underlying sample can influence decile bound-

aries, if newly added companies do not follow the same risk distri-

bution as existing ones. A second explanation could be that the risk

of private equity changes fast. Companies having extremely high or

low risk remain in their respective risk classes if their risk does not

change too much, whereas medium-risk companies switch between

deciles more often for similar beta changes. Incomplete and noisy

information about portfolio companies might not allow the market

to generate stable betas over short time periods.

Moderately stable ranks are good news for private equity investors.

If private equity vehicles remain in their risk deciles over periods

of two years and even longer (results for three and four years are

not shown here), investors can base their portfolio allocation deci-

sions on betas relative to the private equity sector. Although betas

exhibit large absolute swings, which will be shown below, long-term

investors can still target specific high or low risk vehicles for inclu-

sion in their portfolio.

persistency in fixed beta classes

Absolute persistency of market risks can be examined by using

fixed risk classes. In this case, we do not measure the behavior of

company risk relative to other companies but absolute risk change.

We define the following ten risk classes for individual betas βi,t for

vehicle i at time t: qk-1<βi,t<qk with {q0,...,q10} = {-∞,-3,-2,-1,0,1,2,3,4,

5,∞}, where each class is denoted by its upper boundary. If risks

were stable in this sense, we would expect matrix entries along the

main diagonal. If risks change their size randomly, each row should

reflect the unconditional distribution of betas.

The MTM with fixed risk classes in Figure 4 yields a different impres-

sion of market risk. When using fixed class boundaries, transition

matrices cannot be used to reach conclusions about risk move-

ments within the listed private equity sector anymore. Instead,

transition probabilities reflect the behavior of individual risks, which

include changes in market risk as well as idiosyncratic exogenous

factors. As expected from our correlation analysis, betas show

a highly mean-reverting behavior. This effect can be seen from

classes five and six (representing the industry general mean mar-

ket risk), which have the highest transition probability from almost

every other risk class. Transition probabilities seem to converge to

the stationary distribution after a short time, which again indicates

that betas become unstable due to portfolio rebalancing and time-

varying risk within portfolio companies.

Our suspicion that extreme betas are due to estimation error is

confirmed by the fact that one-year betas in risk classes 1 and

10 behave quite randomly between two observation periods. An

economic reason for unstable negative betas could be that private

equity has no short selling strategies. Although results are simi-

lar for one-year and two-year betas, there are a few differences.

Except for one outlier in element p(-3,5) (not shown), the lowest risk

class for two-year betas does not contain any entries. High risks are

10

9

8

7

6

5

4

3

2

1

10

9

8

7

6

5

4

3

2

1

1 2 3 4 5 6 7 8 9 10

1 2 3 4 5 6 7 8 9 10

One-year betas

0 0.25 0.5

Two-year betas

Figure 3 – Markov transition matrices with risk classes constructed from deciles

Left panel – one year betas from 1990 to 2008, N = 2158; right panel: two year betas

from 1991 to 2007, N = 914

t

t+1

t

t+1

93

The time-varying risk of listed private equity

more persistent than medium risks, as can be seen from risk classes

9 and 10, whose transition probabilities are shifted to the right. This

effect is strongest in one-year betas in the high risk classes, which

are more stable than two-year betas.

conclusionAggregate market risk of listed private equity vehicles varies

strongly over time and is positively correlated with the market

return variance. Individual CAPM betas are highly unstable, where-

as ranks of individual vehicles within a cross-section change slightly

less over time. Individual CAPM betas are predictable only up to 2 to

3 years into the future and quickly converge to a stationary distri-

bution when measured in risk classes in an empirical Markov transi-

tion matrix. High- and low-risk companies, however, are more likely

to remain within their risk classes than medium-risk companies. The

probability that a company can be found in the same decile in the

next observation period is about 25% for high-risk companies and

18% for low-risk ones.

We suspect that market risk of private equity is affected by factors

unique to this sector: acquisitions and divestments that constantly

rebalance portfolios, scarcity of information about portfolio com-

panies, and rapid changes within portfolio companies. Unstable

market risk seems to be a fundamental characteristic of private

equity assets, which must be incorporated in the valuation process

and which casts doubt on diversification benefits of private equity

in times of crisis. Particularly important, investors usually hold

traditional private equity shares to maturity, which can be up to 10

years. Unpredictable changes in market risk pose a challenge for

portfolio allocation, since investors would be buying assets that

behave entirely different from what they were supposed to when

first included in the investor’s portfolio. However, targeting vehicles

with specific risks relative to the asset class might be a feasible

strategy for long-term private equity investors.

ReferencesBilo S., 2002, “Alternative asset class: publicly traded private equity, performance, •

liquidity, diversification potential and pricing characteristics,” Doctoral dissertation,

University of St. Gallen

Blume, M. E., 1971, “On the assessment of risk,” Journal of Finance, 26:1, 1-10•

Bollerslev, T., R. F. Engle, and J. M. Wooldridge, 1988, “A capital asset pricing model •

with time-varying covariances,” Journal of Political Economy, 96:1, 116-131

Cao, J., and J. Lerner, 2009, “The performance of reverse leveraged buyouts,” Journal •

of Financial Economics, 91:2, 139-157

Cochrane, J. H., 2005, “The risk and return of venture capital,” Journal of Financial •

Economics, 75:1, 3-52

De Santis, G., and B. Gérard, 1997, “International asset pricing and portfolio diversifica-•

tion with time-varying risk,” Journal of Finance, 52:5, 1881-1912

Dimson, E., 1979, “Risk measurement when shares are subject of infrequent trading,” •

Journal of Financial Economics, 7, 197-226

Driessen, J., T. Lin, and L. Phalippou, 2009, “A new method to estimate risk and return •

of non-traded assets from cash flows: the case of private equity funds,” NBER Working

Paper No. W14144

Ferson, W. E., S. Kandel, and R. F. Stambaugh, 1987, “Tests of asset pricing with time-•

varying expected risk premiums and market betas,” Journal of Finance, 42:2, 201-220

Fisher, L., 1966, “Some new stock market indices,” Journal of Business, 39:1, 191-225•

French, K., W. Schwert, and R. Stambaugh, 1987, “Expected returns and volatility,” •

Journal of Financial Economics, 19:1, 3-29

Ghysels, E., 1998, “On stable factor structures in the pricing of risk: do time-varying •

betas help or hurt?” Journal of Finance, 53:2, 549-573

Groh, A., and O. Gottschalg, 2009, “The opportunity cost of capital of US buyouts,” •

Working Paper

International Financial Services London, 2009, “Private equity 2009,”•

Jagannathan, R., and Z. Wang, 1996, “The conditional CAPM and the cross-section of •

expected returns,” Journal of Finance, 51:1, 3-53

Kaplan S. N., and A. Schoar, 2005, “Private equity performance: returns, persistence, •

and capital flows,” Journal of Finance, 60, 1791-1823

Lahr, H. and F. T. Herschke, 2009, “Organizational forms and risk of listed private •

equity,” The Journal of Private Equity, 13:1, 89-99

Lettau, M., and S. Ludvigson, 2001, “Resurrecting the (C)CAPM: a cross-sectional test •

when risk premia are time-varying,” Journal of Political Economy, 109:6, 1238-1287

Levy, R. A., 1971, “On the short-term stationarity of beta coefficients,” Financial •

Analysts Journal, 27:6, 55-62

Ljungqvist, A., and M. Richardson, 2003, “The cash flow, return and risk characteristics •

of private equity,” NBER Working Paper No. 9454

max

5

4

3

2

1

0

-1

-2

-3

max

5

4

3

2

1

0

-1

-2

-3

-3 -2 -1 0 1 2 3 4 5

max -3 -2 -1 0 1 2 3 4 5

max

One-year betas

0 0.25 0.5

Two-year betas

Figure 4 – Markov transition matrices with fixed risk classes.

Left panel – one year betas from 1990 to 2008, N = 2158; right panel: two year betas

from 1991 to 2007, N = 914.

t

t+1

t

t+1

Articles

95

The developing legal risk management environment

Marijn M.A. van DaelenLaw and Management lecturer / researcher,

Department of Business Law / Center for Company Law (CCL), Tilburg University

AbstractFinancial institutions are facing an increasing number of risk

management regulations with different national and international

approaches. The legal framework, including its risk management

provisions that existed prior to the current financial crisis, has been

severely tested in ‘real life’ and did not hold up to its expectations

(whether reasonable or not). As a reaction to the financial crisis

lawmakers and policymakers have been focusing on, inter alia, risk

management regulations to restore public confidence in companies

and the overall market. The underlying question here is whether

new regulations can indeed prevent the next crisis.

The developing legal risk management environment

96

van Daelen, M. M. A. and C. F. Van der Elst, 2009, “Corporate regulatory frameworks 4

for risk management in the US and EU,” Corporate Finance and Capital Markets Law

Review, 1:2, 83-94, p. 84

van Daelen, M. M. A., forthcoming 2010, “Risk management from a business law 5

perspective,” in van Daelen, M. M. A. and C. F. Van der Elst, eds., Risk management

and corporate governance: interconnections in law, accounting and tax, Edward Elgar

Publishing, Cheltenham

Basel Committee on Banking Supervision, 2005, “Compliance and the compliance 1

function in banks,” Bank for International Settlements, p. 1

Enriques, L., 2009, “Regulators’ response to the current crisis and the upcoming 2

reregulation of financial markets: one reluctant regulator’s view,” University of

Pennsylvania Journal of International Economic Law, 30:4, 1147-1155

Section 8.01(b) of the US MBCA 2005 and Regulation 70 of the U.K. Table A as 3

amended on 1 October 2007 as well as Article 3 of the Model Articles for Public

Companies of the Companies Act 2006.

As a reaction to the financial crisis lawmakers and policymakers

have been focusing on, inter alia, risk management regulations to

restore public confidence in companies and the overall market. The

underlying question here is whether new regulations can indeed

prevent the next crisis. Of course, new regulations can improve

the legal environment of financial institutions, thereby reducing

the imperfections shown by this crisis. However, there seems to be

nothing to gain with extensive additional regulation that can only

prevent a crisis with attributes that are similar to the one the mar-

ket is facing now. First of all, previous crises have had their specific-

ities and the next crisis will most likely have its own features. Hence

it is more than doubtful whether a lot of this kind of regulation is

needed to prevent the next crisis. Secondly, introducing new rules

without reducing existing rules that should have (but might not

have) tackled the same or relating problems, will lead to a pile-up

of regulations. Compliance with all applicable rules and regulations

becomes prohibitively costly for companies. At the same time, the

compliance risk will increase, which in turn can also increase costs.

Compliance risk can be defined as “the risk of legal or regulatory

sanctions, material financial loss, or loss to reputation a [financial

institution] may suffer as a result of its failure to comply with laws,

regulations, rules, related self-regulatory organization standards,

and codes of conduct applicable to its […] activities”1. This is why, in

the words of Enriques (2009), “excessive reregulation today is the

best guarantee of effective pressure towards deregulation tomor-

row” and “regulators should make a lot of noise and show a lot

of activism, all the while producing very little change”2. Excessive

regulation will have a lot of negative side effects and may not even

ensure justified public faith in the reliability of companies and the

overall market in the future. The upshot is that it makes sense first

to have a closer look at existing risk management regulation and

determine the goals in the long run before deciding on the need for

more financial regulation.

Two types of rules provide the basis for the legal risk management

environment. Firstly, there are regulations that demand compa-

nies to have an appropriate risk management system present.

This includes not only ensuring that there is a system to identify

and analyze risks, but also ensuring that the system is adequately

maintained and monitored. Secondly, there are regulations that

require the company to disclose information on (a) the company’s

risk management system or (b) the risks that the company faces

(thus indirectly require a system). This firm-specific information set

is important for minimizing the information asymmetry between

managers and shareholders. After all, managers are involved with

the day-to-day business and have better access to all company

information whereas the shareholders, the providers of capital, only

receive publicly available information. The information needed to

reduce this asymmetry must be disclosed in prospectuses (one time

disclosure document) and in half year or annual reports (ongoing

disclosure documents).

In specific, financial institutions are facing an increasing number of

risk management regulations with different national and interna-

tional approaches. The legal framework, including its risk manage-

ment provisions that existed prior to the current financial crisis, has

been severely tested in ‘real life’ and did not hold up to its expecta-

tions (whether reasonable or not). These risk management provi-

sions can be divided into general and sector-specific provisions. As

the sector-specific rules supplement the general requirements, the

latter will be discussed in the next section. The following section

will address the sector-specific risk management regulation and the

final section provides for some concluding remarks.

General risk management regulationTo start off, in general the board of directors is responsible for man-

aging the company. For example, the U.S. statutory model states that

the board of directors has to manage and oversee the business and

exercise corporate powers and the U.K. model articles of association

state that the board has to manage the company’s business3. In both

the U.S. and U.K. practices, management directs operations since del-

egation of board authority is recognized, but policymaking remains

a task of the board of directors4. Throughout the years, the duty of

directors has been further developed. In the 20th century the duties

of the directors included maintaining a system of internal controls

and disclosing the company’s risks5. For years, the U.S. was forerun-

ner with requirements that focused solely on the financial reporting

aspect of the internal control and risk management framework. The

foundations of American federal securities law are laid down in the

1933 Securities Act, demanding issuers to publicly disclose significant

information about the company and the securities, and the 1994

Securities Exchange Act, requiring mandatory corporate reporting

and independent audits. Congress enacted these two securities

Acts in response to the American stock market crash of 1929 and

the Great Depression in order to reduce the information asymmetry

between managers and shareholders. The Securities and Exchange

Commission (SEC) was established pursuant to the 1934 Act. In the

1933 Act, regulatory recognition was given to the importance of

internal control (which had a more limited scope than risk manage-

ment as we know it today). Regulation S-X Rule 2-02(b) of the 1933

Act required external auditors to give appropriate consideration

to the adequacy of the system of internal control implemented by

the corporation. In its final report of 1940, the SEC recommended a

more thorough assessment of the internal control system. After a

number of corporate scandals, which were related to the bribery of

foreign officials in the mid-1970s, the American regulatory approach

to internal control changed. The 1977 Foreign Corrupt Practices Act

(FCPA) required reporting companies to keep books, records, and

97

The developing legal risk management environment

Committee on Corporate Governance, 1997, Corporate governance in the Netherlands 13

– forty recommendations, Paragraphs 4.2, 4.3, 3.2 and 6.4

Heier, J. R., M. T. Dugan and D. L. Sayers, 2004, “Sarbanes-Oxley and the culmination 14

of internal control development: a study of reactive evolution,” American Accounting

Association 2004 Mid-Atlantic region meeting paper, p. 14

Romano, R., 2005, “The Sarbanes-Oxley Act and the making of quack corporate gov-15

ernance,” Yale ICF Working Paper 05 (23), p. 1

Section 8.01(c), subsections (2) and (6) of the Model Business Corporation Act 16

2005. This Act has been adopted in whole or in part by more than 30 U.S. states.

Amendments to the act were adopted December 2009 regarding proxy voting.

Committee of Sponsoring Organizations of the Treadway Commission, 2004, 17

Enterprise risk management – integrated framework, executive summary, New York:

AICPA Inc., p. 3

15 U.S.C. section 78m (b) (2) (B)6

17 CFR 229.303.a-3-ii and “Instructions to paragraph 303(a)”, under no. 37

See, for instance, the Treadway Commission, 1987, Report of the National 8

Commission on Fraudulent Financial Reporting, New York: AICPA Inc., p. 12.

Establishing an audit committee was already recommended by the SEC in 1972 and

demanded by the NYSE in 1978.

The Cadbury Committee was set up by the Financial Reporting Council, the London 9

Stock Exchange, and the accountancy profession. Its recommendations are focused

on the control and reporting functions of boards, and on the role of auditors.

Cadbury Committee, 1992, Report on the financial aspect of corporate governance, 10

London: Gee, Recommendation 4.31 and 4.32

Cadbury Report 1992, Recommendation 4.35, section (e), under (v)11

Hampel Committee, 1998, Committee on Corporate Governance – final report, 12

London: Gee, Section D (Accountability and Audit) under II and subsection 2.20, p. 21

accounts as well as maintaining a system of internal accounting con-

trols in order to control management activities6. A few years later,

companies needed to assess their risks as item 303 of the MD&A

was added to Regulation S-K. It required managements’ discussion

and analysis report to include “material events and uncertainties

known to management that would cause reported financial informa-

tion not to be necessarily indicative of future operating results or of

future financial condition”7. Around that time, U.S. recommendation

reports and guidelines — such as the 1978 Cohen Report and the

1979 Minahan Report and later the 1987 Treadway Report and 1992

COSO I Report — were starting to stress a broader internal control

framework. Moreover, recommendations towards the oversight duty

of audit committees of the board of directors regarding the financial

reporting process and internal controls started to develop8.

Years after the U.S. 1933 Act, the FCPA, and the MD&A, the U.K.

followed with self-regulatory but more detailed provisions to man-

age companies’ risks. The main rules on mandatory disclosure were

given by the Companies Act of 1985 and the Listing Rules. Section

221 of the Companies Act required companies to keep account-

ing records in order to show and explain their transactions and to

disclose their financial position. The Listing Rules required listed

companies to include a statement of compliance with the provisions

of the 1992 Cadbury Report9 in their annual report and accounts on

a comply-or-explain basis. This self-regulatory report provided that

the board of directors had to maintain a system of internal control

over the financial management of the company — including proce-

dures to mitigate corporate governance risks and failures — and that

the directors had to make a statement in the annual report on the

effectiveness of their internal control system10. The Cadbury Report

also recommended that all listed companies should establish an audit

committee, comprising at least three non-executives. The report

gave the audit committee’s duties, which include reviewing the

company’s statement on internal control systems11. The 1998 Hampel

Report broadened the U.K. internal control perspective by arguing

that the system did not only have to cover financial controls but also

operational and compliance controls, as well as risk management12.

As the Hampel Committee suggested, the London Stock Exchange

issued the Combined Code on Corporate Governance, which included

the provisions of, inter alia, the Cadbury Report and Hampel Report.

Later, other European member states followed the U.K. with inter-

nal control and risk management regulations. For instance, the

Netherlands issued a self-regulatory code (the 1997 Peters Report)

that stressed the board of directors’ responsibility for effective sys-

tems of internal control and recommends the supervisory board to

consider whether to appoint an audit committee. This committee was

recommended specific duties such as supervising external financial

reports, compliance, and the control of company risks13.

Obviously, the legal internal control and risk management envi-

ronment significantly changed when the U.S. Congress passed

the Sarbanes Oxley Act (SOX) after the corporate failures and

fraud cases between 2000 and 2003. It has been said to be the

culmination of a century of internal control developments14. This

2002 federal law was intended to restore public faith and trust

by, inter alia, improving the accuracy and reliability of corporate

disclosures. It contains not only disclosure requirements, but also

substantive corporate governance mandates15. The legal duties of

corporate constituents regarding a system of internal controls are

further developed by this Act and other legislative measures. The

well-known SOX Section 404 demands an annual internal control

report in which management’s responsibility for “establishing and

maintaining an adequate internal control structure and procedures

for financial reporting” is stressed. The report also has to include an

assessment of the effectiveness of these structures and procedures.

Section 302 requires the CEO and CFO — thus not management as

Section 404 does — to certify the fairness of the financial state-

ments and information as well as their responsibility for establish-

ing and maintaining internal controls. The CEO and CFO also have

to present their conclusions — not the total evaluation of Section

404 — about the effectiveness of the internal controls based on their

evaluation. The duties of audit committees also continued to evolve

as Section 301 requires that audit committees establish procedures

for, “the receipt, retention, and treatment of complaints […] regard-

ing accounting, internal accounting controls, or auditing matters.”

Section 205(a) stresses that the purpose of the audit committee is

to oversee the company’s accounting and financial reporting pro-

cesses and audits of the financial statements. Other U.S. legislative

measures as well as guidelines have a wider internal control and risk

management perspective, such as the MBCA and the COSO II Report.

The MBCA provides the scope of the board’s oversight responsi-

bilities relating to the company’s major risks and the effectiveness

of the company’s internal financial, operational, and compliance

controls16. The COSO II Report broadens reporting to encompass

non-financial information and internal reporting and it adds a fourth

category, the strategic objectives, to the existing financial reporting,

operational, and compliance objectives17. Eversince the corporate

failures manifested themselves, it has been argued that new regula-

The developing legal risk management environment

98

Corporate Governance Code Monitoring Committee, 2008, The Dutch corporate gov-24

ernance code – principles of good corporate governance and best practice provisions

(DCGC 2008), Principle II.1

DCGC 2008, Best practice provision III.5.425

The 1984 Eighth Company Law Directive (84/253/EEC, OJ L 126, 12 May 1984, p. 26

20–26) harmonized the approval of persons responsible for carrying out the statu-

tory audits of accounting documents. Articles 3 and 24 demanded such persons to be

independent and of good repute.

Article 5 and IV of Annex I of Directive 2003/71/EC of the European Parliament and 27

of the Council of 4 November 2003 on the prospectus to be published when securi-

ties are offered to the public or admitted to trading and amending Directive 2001/34/

EC, OJ L 345, 31 December 2003, p. 64–89

Ribstein, L. E., 2002, “Market vs. regulatory responses to corporate fraud: a critique 18

of the Sarbanes-Oxley Act of 2002”, Journal of Corporation Law, 28:1, p. 5

Cunningham, L. A., 2002, “Sharing accounting’s burden: business lawyers in Enron’s 19

dark shadows”, Boston College Working Paper, pp. 16-17

Bratton, W. W., 2002, “Enron and the dark side of shareholder value,” Available at 20

SSRN: <http://ssrn.com/abstract=301475>, p. 13

Committee on Corporate Governance, 2000, The Combined Code – principles of 21

good governance and code of best practice (Combined Code 2000), Principle D.2 and

Provision D.2.1 (Principle C.2 and Provision C.2.1 of the 2008 Combined Code)

Committee on Corporate Governance, 2003, The Combined Code – principles of good 22

governance and code of best practice (Combined Code 2003), Provision C.3.2

Financial Reporting Council, 2009, Review of the Combined Code: Final Report, p. 2723

tions such as SOX might not succeed in regulating frauds or that

their effectiveness would be limited as the frauds that preceded this

legal response occurred despite several levels of monitoring in place

at the time18. For example, Cunningham notes that “[h]istory offers

no reason to expect that new rules will prevent a repeat of account-

ing scandals even of this large size or frequency”19 and Bratton,

that “[t]he costs of any significant new regulation can outweigh the

compliance yield, particularly in a system committed to open a wide

field for entrepreneurial risk taking”20.

Within Europe, the legal internal control and risk management envi-

ronment also changed after the failures and frauds around the new

millennium, but with a more principle-based and self-regulatory

approach. Following the 2003 European Commission’s Plan to Move

Forward, E.U. member states have drawn up or updated their nation-

al corporate governance codes for listed companies. In the U.K., due

to the Hampel Report, the 2000 Combined Code already underlined

the board’s duty to maintain a sound system of internal controls,

on a comply-or-explain basis. The Code added that the board has to

report annually to the shareholders that it has reviewed the effec-

tiveness of the group’s internal control system, covering financial,

operational, and compliance controls and risk management21. A few

years later, the provisions dealing with the audit committee’s duties

were updated due to the 2003 Higgs and Smith Reports requiring

the committee to review the company’s internal control and risk

management systems22. The 2009 Review of the Combined Code

announced amendments to the internal control principle in order

to stress “the board’s responsibility for defining the company’s risk

appetite and tolerance and maintaining a sound risk management

system” and to the provisions in order to include that the board has

to “satisfy itself that appropriate systems are in place to enable it

to identify, assess and manage key risks”23. Other E.U. member

states also issued corporate governance codes that emphasized

the board’s and audit committee’s duties. For instance, the Dutch

code provides that the board is responsible for complying with all

relevant primary and secondary legislation and managing the risks

associated with the company’s activities. In addition, the board

has to report related developments to and discuss the internal risk

management and control systems with the supervisory board and

the audit committee24. The audit committee has to monitor the

activities of the board with respect to the operation of the internal

risk management and control systems25.

Traditionally, E.U. lawmakers focused mainly on corporate disclosure

rules and not so much on requiring management systems to endorse

the reliability of the reporting and an internal control framework26.

Responding to the corporate failures and fraud around the new mil-

lennium, the E.U. became more active in areas such as company law,

accounting, and auditing law, although parts of these areas remain

controlled by the national legislators. The E.U. legislative movement

brought forward several general as well as sector-specific direc-

tives and recommendations. One of these general directives is the

Prospectus Directive with the purpose of harmonizing, inter alia,

the information contained in the prospectus in order to provide

equivalent investor protection. It requires the prospectus to include

key information on the company’s risk factors and a summary in

which “the essential characteristics and risks associated with the

issuer, any guarantor and the securities” are disclosed27. In addition,

the Transparency Directive 2004/109/EC harmonizes transparency

requirements to ensure a higher level of investor protection and mar-

ket efficiency within the E.U. The directive acknowledges the impor-

tance of disclosure of the companies’ main risks as Articles 4 and 5

of the directive require the annual and half-yearly financial reports

to include a management report in which a description is given of the

“principal risks and uncertainties” that the company faces. Next to

disclosing general information on the companies’ risks, it is required

to disclose information on the risk management systems regarding

the financial reporting process. That is because Articles 1 and 2 of the

2006/46/EC amendment to the Accounting Directives — the Fourth

and Seventh Company Law Directives — provide that the annual

corporate governance statement must include a description of the

main features of the company’s (or group’s) internal control and risk

management systems for the financial reporting process. Moreover,

specific features of the audit committee’s duties are given at E.U.

level. In a Commission Recommendation (2005/162/EC), the audit

committee is recommended to assist the board to review the internal

control and risk management systems. The committee should do so

in order to ensure that the main risks the company faces are properly

identified, managed, and disclosed. This monitoring role of the audit

committee is further developed in the 2006/43/EC Audit Directive

for monitoring the financial reporting process and the effectiveness

of the company’s internal control and risk management systems.

Thus, the legislative measures at the E.U. level require reporting on

the main features of the systems for the financial reporting process

and monitoring the effectiveness of the systems. Before implement-

ing these directives in national laws and regulations, most member

states had already issued corporate governance codes focusing on a

broader concept of internal control and risk management, emphasiz-

ing the financial reporting, operational, and compliance aspects.

99

The developing legal risk management environment

Articles 11 and 22 and Annex v of Directive 2006/48/EC of the European Parliament 31

and of the Council of 14 June 2006 relating to the taking up and pursuit of the

business of credit institutions, OJ L 177, 30 June 2006, p. 1–200, and Article 34 of

Directive 2006/49/EC of the European Parliament and of the Council of 14 June

2006 on the capital adequacy of investment firms and credit institutions, OJ L 177,

30 June 2006, p. 201–255

Directive 2009/138/EC of the European Parliament and of the Council of 25 32

November 2009 on the taking-up and pursuit of the business of Insurance and

Reinsurance (Solvency II), OJ L 335, 17 December 2009, p. 1–155

See for a more thorough analysis of risk management within financial law: Van der 28

Elst, C. F., forthcoming 2010, Risk management in financial law, in van Daelen, M. M.

A., and C. F. Van der Elst, eds., Risk management and corporate governance: intercon-

nections in law, accounting and tax, Edward Elgar Publishing, Cheltenham

Belgium, Canada, France, Germany, Italy, Japan, the Netherlands, Sweden, the U.K. 29

and the U.S. and later also Switzerland.

Basel Committee on Banking Supervision, 2004, International convergence of capital 30

measurement and capital standards – a revised framework

The upshot is that, despite some inconsistencies, both in the E.U.

and the U.S. most parts of the reporting and monitoring level

are covered, as shown in Table 1. In the U.S. this is accompanied

by legislative measures focusing on establishing and maintaining

an internal control system for financial reporting, whereas E.U.

member states further the framework by provisions related to the

overall system for establishment and maintenance. In addition, in

the U.S. the establishing, maintaining, reporting, and monitoring

provisions regarding the financial reporting systems are provided

by law. On the contrary, in the E.U. the establishing, maintaining,

and reporting provisions regarding the overall systems are pro-

vided by self-regulation, although this regulation has a legal basis

in several member states.

Establish/identify

Maintain/manage

Report on Monitor

General risks

(E.U. / U.S. state)

(E.U. / U.S. state)

E.U./U.S. E.U. MS / U.S. state

General systems

E.U. MS E.U. MS E.U. MS E.U./E.U. MS / U.S. state

Financial reporting systems

U.S. U.S. E.U./U.S. U.S.

Table 1 – Aspects of the main general E.U. and U.S. internal control and risk

management provisions

sector-specific risk management regulationNext to these general provisions, the legal risk management environ-

ment — especially with regards to the financial industry — is shaped by

sector-specific legal measurements. The financial industry-specific

provisions cover mainly the banking system, insurance, and securi-

ties. The previous section shows that the foundation of the legal

risk management environment is given by mainly two types of rules.

Firstly, having — including maintaining and monitoring — internal

control and risk management systems and secondly, disclosing infor-

mation about those systems and the risks the company faces. As the

current section will show, the financial industry regulations and guide-

lines supplement these two levels — though much more detailed —

but also expand the first level. Indeed, the general rules provide that

a system must be in place and stress the board’s duty to maintain

internal control and risk management systems and the audit com-

mittee’s monitoring role therein. The financial industry-specific

provisions, however, regulate certain internal functions within the

organization and emphasize the external monitoring role of the

supervisory authorities. Several financial industry-specific provisions

are described below in order to further explain these expansions28.

At E.U. level, financial industry-specific directives that refer to risk

management are, inter alia, the Capital Requirements Directives,

the Solvency Directive, and the MiFID. The Basel Committee on

Banking Supervision was established in 1974 by the central bank

governors of the Group of Ten countries29. Without formal supra-

national supervisory authority, the committee issues supervisory

standards and guidelines which national authorities can implement.

The 1988 Basel Capital Accord introduced a capital measurement

system which provided for the implementation of a credit risk mea-

surement framework. In 2004 a revised framework was issued. This

2004 Basel II Accord provides for requirements relating to minimum

capital, supervisory review, and market discipline and disclosure30.

It stresses that risk management is fundamental for an effective

assessment of the adequacy of a bank’s capital position. The Basel

II framework is introduced into European legislation through the

Capital Requirements Directives, comprising Directive 2006/48/EC

and Directive 2006/49/EC. It affects credit institutions and certain

types of investment firms. In line with the above described gen-

eral legal provisions, Article 138 of Directive 2006/48/EC requires

credit institutions to have adequate risk management processes

and internal control mechanisms in place, including reporting and

accounting procedures. Article 135 of that Directive reads that

E.U. member states have to demand that “persons who effectively

direct the business of a financial holding company be of sufficiently

good repute and have sufficient experience to perform those

duties.” Consequently, where general legal provisions develop what

the duty of the board and managers includes, this sector-specific

provision regulates what a proper person for performing certain

duties would be like. In addition, credit institutions and certain

types of investment firms must have effective processes to iden-

tify, manage, monitor, and report their risks as well as adequate

internal control mechanisms. Their management body — consisting

of at least two persons with sufficiently good repute and sufficient

experience to perform such duties — should approve and review the

strategies and policies for indentifying, managing, monitoring, and

mitigating the risks, taking into account specific criteria regarding

the credit and counterparty risk, residual risk, concentration risk,

securitization risk, market risk, interest rate risk arising from non-

trading activities, operational risk, and liquidity risk31. Obviously,

these requirements, especially the specific criteria, are much more

detailed than the general ones described in the previous section.

Another sector-specific European Directive that introduces a

comprehensive framework for risk management and regulates

certain internal functions within the organization is the Solvency

II Directive32. It has a much wider scope than the Solvency I

Directive and contains thorough revision and realignment of the

E.U. Directives relevant to (re)insurers. The Solvency II Directive

has similarities to the Basel II banking regulation. This set of regu-

latory measurements for insurance companies includes provisions

regarding capital, governance, and risk management, effective

The developing legal risk management environment

100

Section 4.2 of the Dutch Banking Code: Nederlandse Vereniging van Banken (NVB), 39

Code Banken, 9 September 2009, p. 10. This self-regulatory code will most likely

receive a legal basis in 2010.

Recommendations 23-27 and Annex 10 (Elements in a board risk committee report) of 40

the Walker Review, 2009, A review of corporate governance in U.K. banks and other

financial industry entities – Final recommendations. In addition, see the U.K. Turner

Review, 2009, A regulatory response to the global banking crisis, p. 93.

Financial Reporting Council, 2009, Review of the Combined Code: Final Report, p. 25 41

Articles 41, 44, 46 and 101 of Directive 2009/138/EC33

Articles 42, 46 and 47 of Directive 2009/138/EC34

Commission Directive 2006/73/EC of 10 August 2006 implementing Directive 35

2004/39/EC of the European Parliament and of the Council as regards organisational

requirements and operating conditions for investment firms and defined terms for

the purposes of that Directive, OJ L 241, 2 September 2006, p. 26–58 (MiFID level 2

Directive)

Articles 6 and 7 of Commission Directive 2006/73/EC36

Section 303A of the NYSE’s Listed Company Manual 37

COSO, Effective enterprise risk oversight: the role of the board of directors, 2009; 38

Section 5 of the Shareholder Bill of Rights Act of 2009 (S. 1074) of 19 May 2009. The

bill proposes to amend the Securities Exchange Act of 1934 (15 U.S.C. 78a et seq.) by

inserting after section 14 section 14A and by adding at the end subsection (e) ‘corpo-

rate governance standards,’ (5) ‘risk committee’

supervision, and disclosure. It requires written and implemented

policies for the company’s risk management, internal control, and

internal audit. In addition, the (re)insurance companies must have

an effective and well integrated risk-management system in order

to identify, measure, monitor, manage, and report risks, covering,

inter alia, market, credit, liquidity, concentration, and operational

risks. The risk management system must include risk mitigation

techniques and the companies have to conduct risk and solvency

assessments. Next to the risk management system, an effective

internal control system with administrative and accounting pro-

cedures, an internal control framework, and appropriate reporting

arrangements is required33. To sum up, compared to the general

provisions described above, this directive introduces a more com-

prehensive set of risk management and internal control require-

ments. The directive also prescribes internal functions and the

duties of as well as the personal qualifications for those functions.

For instance, for the evaluation of the adequacy and effective-

ness of the internal control system, an internal audit function is

required. Furthermore, the (re)insurance companies are instructed

to have a compliance function within their organization. This com-

pliance function has the duty to advise the management or super-

visory body on compliance with laws and regulations, to identify

and assess compliance risks, and to assess the possible impact of

any changes in the legal environment. Moreover, it demands that

the “persons who effectively run the undertaking or have other key

functions” are fit and proper, that is, have adequate professional

qualifications, knowledge, and experience and are of good repute

and integrity respectively34.

A third financial industry-specific piece of E.U. legislation is the

MiFID, the markets in financial instruments directive, which pro-

vides organizational requirements and operating conditions for

investment firms35. Like the previous directives it requires com-

panies to establish, implement, and maintain risk management

policies and procedures in order to detect risks, set the level of risk

tolerance, and includes risk minimizing procedures. The directive

further demands investment companies to monitor the adequacy

and effectiveness of these risk management policies and proce-

dures. With regard to the internal functions of the organization, it

provides that investment companies have to establish and maintain

a compliance function, for which a compliance officer must be

appointed, with the duty to monitor and assess the adequacy and

effectiveness of the company’s measures and procedures. It goes

on to describe that this compliance function must have “the nec-

essary authority, resources, expertise, and access to all relevant

information” in order to create an environment in which it can

discharge its responsibilities properly and independently. In addi-

tion, likewise the insurance companies, the investment companies

need to have an internal audit function for the evaluation of the

adequacy and effectiveness of the internal control system36.

As the regulatory reform is making its entrance at E.U. level, E.U.

member states are introducing their own financial industry-specific

guidelines and the U.S. is developing general regulations regarding

certain internal functions within the organization. In the U.S., the New

York Stock Exchange corporate governance rules require audit com-

mittees to discuss the guidelines and policies to govern the process

of risk assessment and risk management37. In addition, in May 2009

legislation entitled ‘Shareholder Bill of Rights Act of 2009’ has been

introduced in the U.S. Senate by Senator Charles Schumer that would,

if passed, mandate risk committees for publicly traded companies in

general. The role of these risk committees — that are to be composed

of independent directors — is to be responsible for the establishment

and evaluation of the risk management practices of the issuer38. Like

in the U.S., in E.U. member states the idea of requiring companies to

have a risk committee is also starting to be considered. For instance,

the Dutch self-regulatory Banking Code requires banks — not listed

companies in general — to have a risk committee39. Furthermore,

the U.K. Walker Review recommends certain listed banks and insur-

ance companies to establish a risk committee in order to, inter alia,

oversee and advise the board on current risk exposures and the

future risk strategy40. Similar to the Netherlands, but contrary to

the U.S., this U.K. recommendation is not extended to non-financial

listed companies41. In general, from a legal perspective reform as a

reaction to the financial crisis includes the (further) development of

the duty of the board, senior management, the supervisory body or

non-executives, the audit committee, the internal audit, the compli-

ance function, and the risk committee.

Next to regulating the duty of and personal qualifications for cer-

tain internal functions within the organization, the more recent

financial industry-specific provisions emphasize the external moni-

toring role of the supervisory authorities. Financial markets are

more global than they used to be and since the financial crisis can

be defined as a global systemic crisis, lawmakers are searching for

ways to address regulatory repair internationally. The reaction at

101

The developing legal risk management environment

Proposal for a directive of the European Parliament and of the Council Amending 45

Directives 1998/26/EC, 2002/87/EC, 2003/6/EC, 2003/41/EC, 2003/71/EC, 2004/39/

EC, 2004/109/EC, 2005/60/EC, 2006/48/EC, 2006/49/EC, and 2009/65/EC in

respect of the powers of the European Banking Authority, the European Insurance

and Occupational Pensions Authority, and the European Securities and Markets

Authority, 26 October 2009, COM(2009) 576. Legislative measures to implement the

European Systemic Risk Board are not included in this proposal.

See for this discussion Morley, J. D., and R. Romano, eds., 2009, “The future of finan-46

cial regulation”, John M. Olin Center for Studies in Law, Economics, and Public Policy,

Research Paper No. 386, 108-116

Morley, J. D., and R. Romano, R. eds., 2009, “The future of financial regulation”, John 47

M. Olin Center for Studies in Law, Economics, and Public Policy, Research Paper No.

386, p. 88

See for the U.S. reaction: H.R.4173 – Wall Street Reform and Consumer Protection 42

Act of 2009. U.S. House passed this financial services reform bill on 11 December

2009, which is said to be ‘the biggest change in financial regulation since the Great

Depression.’ When the bill becomes law, it would provide for, inter alia, protection of

consumers and investors, enhanced Federal understanding of insurance issues, and

regulated over-the-counter derivatives markets. This legislation would also establish

a Consumer Financial Protection Agency and give the Treasury Department new

authority. See in addition, Department of The Treasury, financial regulatory reform

– a new foundation: rebuilding financial supervision and regulation, 17 June 2009

(the White Paper on Financial Regulatory Reform of June 2009 from the Obama

Administration).

The de Larosière Group, 2009, Report of the high-level group on financial supervision 43

in the EU, Brussels, p. 4

Communication from the Commission, European financial supervision, 27 May 2009, 44

COM(2009) 252

the E.U. level42 includes introducing European supervisory authori-

ties to supplement the national supervision in order to repair the

lack of cohesiveness and form a supervisory front. The European

Commission mandated the de Larosière Group, a high-level group

on financial supervision in the E.U., to propose recommendations

on the future of European financial regulation and supervision.

The report of the de Larosière Group provides for a framework

that points out three main items to strive for: a new regulatory

agenda (to reduce risk and improve risk management), stronger

coordinated supervision (macro- and micro-prudential), and effec-

tive crisis management procedures (to build confidence among

supervisors)43. In reaction to these recommendations, the European

Commission proposes reforms relating to the way financial markets

are regulated and supervised. It recommends a European financial

supervisory framework composed of two new pillars. The first pil-

lar is a European Systemic Risk Board (ESRB) “which will monitor

and assess potential threats to financial stability that arise from

macro-economic developments and from developments within the

financial system as a whole (‘macro-prudential supervision’)”44. The

second pillar is a European System of Financial Supervisors (ESFS)

that should consist of a network of national financial supervisors

as well as new European Supervisory Authorities. At the moment,

three financial sector-specific committees are already in place

at the EU level: the Committee of European Banking Supervisors

(CEBS), the Committee of European Insurance and Occupational

Pensions Committee (CEIOPS), and the Committee of European

Securities Regulators (CESR). In order to establish European

Supervisory Authorities, a directive is proposed which transforms

these committees into a European Banking Authority (EBA), a

European Insurance and Occupational Pensions Authority (EIOPA),

and a European Securities and Markets Authority (ESMA)45.

concluding remarks Since the financial crisis started, the legal risk management envi-

ronment is under construction. Especially for the financial services

industry, regulations relating to the internal organization of the

company and the external monitoring role of the supervisory

authorities are piling up. As argued above, the underlying question

here is whether new regulations can indeed prevent the next crisis.

For new regulations to improve the legal environment of financial

institutions — thereby reducing the imperfections shown by this

crisis — firstly the current legal environment must be clear and sec-

ondly, the primary problem has to be understood. This seems only

logical, but gauging the precise problem is far from easy. Where

Fein argues that it might not be bank regulation that is broken, but

rather bank supervision, Kashyap argues that regulation is broken

at the most basic level46. Even so, if new regulations can prevent

a crisis such as this one, there is no guarantee that this type of

regulation is needed to prevent the next one, as the next crisis will

most likely have its own features. Besides that, at a roundtable on

the future of financial regulation Harring argued that there might

almost never be a perfect time for reform. “When profits are high

and markets are buoyant, it’s only we ivory tower types who think

about it. And when there’s a crash, risk aversion rises to such an

extent that tightening regulations is unnecessary because institu-

tions and markets are already too risk-averse to rekindle economic

growth”47. To conclude, it might not be the right time for regulatory

reform, but when it is, it might be best to focus on what we want

to achieve in the long run and how that can be achieved keeping

in mind an adequate balance between the interests of corporate

constituents, consumers, and investors on the one hand, and the

overall costs, on the other. After all, the financial services industry

is one of the most heavily regulated industries already.

Articles

103

interest rate risk hedging demand under a Gaussian framework

sami AttaouiEconomics and Finance Department,

Rouen Business School

pierre sixEconomics and Finance department,

Rouen Business School

AbstractThis article analyzes the state variables Merton-Breeden hedg-

ing demand for an investor endowed with a utility function over

both intermediate consumption and terminal wealth. Based on the

three-factor model of Babbs and Nowman (1999), we show that this

demand can be simply expressed as weighted average zero-coupon

bonds sensitivities to these factors. The weighting parameter is

actually the proportion of wealth our investor sets aside for future

consumption rather than for terminal wealth.

interest rate risk hedging demand under a Gaussian framework

104The first component is the speculative mean-variance demand that depends on asset 4

direction that can be much more easily studied than the Merton-Breeden hedging

demand.

The study of the variance covariance matrix as well as the covariance of assets with 5

state variables can be easily carried out for various assets in our affine Gaussian

framework and is thus omitted in the sequel.

Our analysis can also be carried out in the Dai and Singleton (2000) framework.1

We consider the case 2 g>1. This is standard in the theoretical literature and has been

supported by empirical evidence.

See Cox and Huang (1991) for the equivalence between the martingale and dynamic 3

programming approach.

Portfolio management of fixed-income securities has recently

received a lot of attention in the literature. However, to the best of

our knowledge, none of these studies focus on the Merton-Breeden

hedging demand [Merton (1973), Breeden (1979)] for an arbitrary

fixed income security. Considering an investor with a utility func-

tion over intermediate consumption and terminal wealth, we focus

in this study on a stochastic opportunity set composed of stochas-

tic interest rates modeled by the three-factor model of Babbs and

Nowman (1999). Indeed, Litterman and Scheinkman (1991) have

identified three common factors that account for more than 82%

of innovations in bond returns.

Relying on the martingale approach for portfolio management

[Karatzas et al. (1987), Cox and Huang (1989, 1991)], we provide

useful insights on the factors related to interest rate risk hedging

demand. Our results show that this demand is intimately linked to

the sensitivities of zero-coupon bonds to the various factors as

well as to the proportion of wealth an investor sets aside for future

consumption rather than for terminal wealth. In addition, these

quantities can be straightforwardly computed.

settingWe consider the three-factor interest rate model of Babbs and

Nowman (1999)1. The three state variables Xi(t), i = 1, 2, 3 are

assumed to be governed by the following dynamics

dX(t) = -diag(κ)Xdt + σdz(t);

X(0) ≡ X, (1)

where dz(t) ≡ [dz1(t) dz2(t) dz3(t)]’ is a vector of independent

Brownian motion defined on a filtered probability space (Ω, F0≤t≤T,

F, P), T designates the end of the economy, and P the historical

probability. The market prices of risk, θ ≡ [θ1 θ2 θ3]’, linked to these

Brownian motions are assumed to be constant. Moreover, κ ≡ [κ1 κ2

κ3]’ is a vector of positive constant mean reversion speed of adjust-

ments and diag(κ) is the diagonal matrix whose ith element is κi.

σ ≡ (σij), 1≤i, j≤3, is a lower triangular matrix of constants that repre-

sents the sensitivity of the factors to the various shocks.

The instantaneous risk-free interest rate is specified as follows:

r(t) ≡ μ – X1(t) – X2(t) –X3(t) (2)

where μ is a positive constant.

Under these assumptions, Babbs and Nowman (1999) give a closed-

form formula for the time-0 price of a zero-coupon bond maturing

at time TB:

B TB, X( ) = exp −TB. r∞ − w TB( )( ) − D K TB( )′ X

, (3)

where D K TB( ) ≡ Dκ1TB( ) Dκ2

TB( ) Dκ3TB( )[ ]'

,

with Dx y( ) ≡ 1

x1 − exp −x.y( )( ).

These three functions represent the sensitivity of the bond price to

the three factors. The long term interest rate r∞ and the determin-

istic function w(TB) are given in the appendix. Applying Ito’s lemma

to Equation (3), we obtain the zero coupon volatility vector:

σ B(TB ) = ′ σ D K TB( )

(4)

We consider an individual that invests in a riskless asset and in a

portfolio composed of three non-redundant fixed-income securities.

Let π(t) denote the vector of risky proportions, and W(t) and C(t)

the investor’s wealth and consumption, respectively. Furthermore,

we assume that our investor has a constant relative risk aversion2,

g, and a finite investment horizon, TI.

The investor maximizes his/her expected utility over consumption

and terminal wealth subject to a budget constraint. Formally,

J ≡ supπ (s)0≤ s ≤ TI

ECs

1 −γ

1 −γds

0

TI∫ +WTI

1 −γ

1 −γ

(5)

dW(t)

W (t)+ C(t)

W (t)dt= r(t)+ ϖ (t ′ ) π(t)( )θ[ ]dt+ ϖ (t ′ ) π(t)( )′dz(t)

W (0) ≡ W

(6)

and C(0) ≡ C is the investor optimal initial consumption.

Using the dynamic programming approach3, Merton (1973) has

shown that:

π(0) = JW

−WJWW

Σ−1 (0)e(0)+ JWi

−WJWW

Σ−1 (0)i

3

∑ ωi (0),

(7)

where JW and JWW denote the first and second partial derivative

of the value function J with respect to wealth, respectively. JWi is

the second cross partial derivative of J with respect to wealth and

state variable Xi. Σ(t) ≡ ϖ(t) ϖ(t)’ is the variance covariance matrix,

with ϖ(t) denoting the matrix of volatility of assets. e(t) represents

the vector of excepted return in excess of the risk free rate of these

assets. ωi(t) stands for the vector of covariance between state vari-

able Xi and the three risky assets.

The second component4 of the right hand side of Equation (7) is the

so-called Merton-Breeden hedging demand for risky assets whose

number is equal to that of state variables.

As stated above, we focus on the part of the hedging demand that

does not depend on the type of asset selected in the portfolio:

JWi(t) ÷ -W(t)JWW(t), i=1,2,35. Merton (1973) has shown that JWi(t) ÷

-JWW(t) could be couched in terms of two sensitivities:

105

interest rate risk hedging demand under a Gaussian framework

Full demonstration is available from the authors upon request8

The proof is available from the authors upon request.9

We show in the appendix that T10 i(g, TI, X) <TI, which implies that Hi(g, Ti(g, TI, X) < Hi(g, TI).

If 6 ∂iC<0 (>0) then the investor will demand more of the risky asset whose returns are

positively (negatively) correlated with changes in the state variable i. Therefore, an

unfavorable shift in the opportunity set is offset by higher level of wealth (see Merton

(1973) for a more detailed discussion of this point).

|| stands for the usual euclidean norm.7

JWi ÷ -JWW = -∂iC ÷ ∂WC (8)

where ∂iC denotes the partial derivative of consumption with

respect to state variable Xi and ∂WC is the partial derivative of

consumption with respect to wealth. Since -∂iC ÷ ∂WC can be posi-

tive or negative, we provide below a comprehensive investigation

of this term6.

In order to study this hedging component, we use the martingale

approach [Karatzas et al. (1987), Cox and Huang (1989, 1991)],

where the budget constraint (6) is cast into a martingale:

ECs

Gs

ds0

TI∫ +WTI

GTI

= W

(9)

The numeraire Gt is the growth portfolio whose dynamics obeys the

following equation:7

dG(t)/G(t) ≡ [r(t) + |θ|2]dt + θ’ dz(t)

G(0) = 1 (10)

Using the results of Munk and Sorensen (2007), the value function

can be computed as follows:

J(γ,TI , X,W ) = W 1 −γ

1 −γW

Cγ,TI , X( )

γ

.

(11)

The optimal ratio of wealth to consumption is given only in terms

of the growth portfolio:

W

Cγ,TI , X( ) = Q γ,TI , X( ) + q γ,TI , X( ),

(12)

with Q γ,TI , X( ) ≡ q γ,u, X( )du0

TI

∫

and q γ,u, X( ) ≡ E G(u)1

γ−1

.

Equation (12) stresses the importance of the function q(g, u, X),

which describes how state variables affect the value function and

then the hedging component under scrutiny. The closed-form

expression of q(g, u, X) can be obtained by a change of probability

measures where a particular numeraire8, whose volatility vector

is 1/g · θ + [1 – 1/g]σB(TI), is introduced. We obtain the following

proposition:

proposition 1 — The functions q(g, TI, X) are given by:

q γ,TI , X( ) = B TI , X( )1 − 1γ exp − γ −1

2γ 2σ B(v)−θ 2

dv0

TI

∫

(13)

where the explicit expression of σ B(v)−θ 2dv

0

TI

∫ is given in the

appendix.

proof — available from the authors upon request.

The first term of Equation (13) is a positive decreasing function of

the investment horizon and the second term is obviously also a

positive decreasing function of the investment horizon. Hence, q(g,

TI, X) and Q(g, TI, X) are decreasing and increasing functions of the

investment horizon, respectively. As a consequence, the behavior

of the wealth to consumption ratio [Equation (12)] as a function of

the investment horizon has to be examined numerically. Moreover,

Equation (12) shows that the wealth to consumption ratio does not

depend on either wealth or consumption. Thus, using Equation (11),

our hedging term can be couched solely in terms of consumption:

JWi

−WJWW

γ,TI , X( ) = − ∂iC

Cγ,TI , X( )

(14)

The hedging demand component is the opposite of the elasticity

of consumption with respect to state variables. To the best of our

knowledge, this feature has not been much emphasized in the lit-

erature.

To further study this elasticity we need to define another variable

which measures, in proportion, how much wealth an investor sets

aside to satisfy future consumption rather than terminal wealth.

Building on Karatzas et al. (1987)9, we show that this proportion,

πC(g, TI, X), is given by:

πC γ,TI , X( ) = 1 +q γ,TI , X( )Q γ,TI , X( )

−1

.

(15)

πC(g, TI, X) is an increasing function of the investment horizon, is

positive, and lies between 0 and 1. When πC(g, TI, X) ≡ 1, the investor

focuses only on consumption, and in the case πC(g, TI, X) ≡ 0, he/she

is solely concerned by terminal wealth. We are now able to state the

main result of our article:

proposition 2 — the elasticity of consumption with respect to state

variables is given by:

∂iC

Cγ,TI , X( ) = πC γ,TI , X( )H i γ,Ti γ,TI , X( )( )+ 1 − πC γ,TI , X( )[ ]H i γ,TI( )

(16)

with H i γ,T ()( ) ≡ 1 − 1

γ

Dκ i

T ()( ) and Ti(g,TI,X) is given in the appendix.

proof — Available from the authors upon request.

Equation (16) states that the elasticity of consumption can be

decomposed in weighted average of bond sensitivities. The weight-

ing parameter is the proportion of wealth for future consumption.

Furthermore, Hi(g, TI) and Hi(g, Ti(g, TI, X)) are the risk-aversion

adjusted sensitivities of a zero-coupon bond price with respect to

the state variable. They are related to a bond having TI as terminal

maturity and to a bond maturing at an intermediate horizon Ti(g, TI, X),

respectively10. Moreover, in the case of the first bond the investor is

solely concerned about terminal wealth, whereas, in the case of the

second bond he/she is concerned about intermediate consumption.

106 – The journal of financial transformation

interest rate risk hedging demand under a Gaussian framework

Finally, we point out that the elasticity of consumption is always

positive.

numerical illustrationIn this section, we provide various numerical analysis for πC(g, TI, X),

Hi(g, TI), Hi(g, Ti(g, TI, X)), Ti(g, TI, X), and ∂iC/C(g, TI, X). The base case

parameters (Table 1) are taken from the empirical study in Babbs

and Nowman (1999). The initial values for the state variables are

set equal to their long-term means, i.e. zero.

We consider three levels of risk aversion, g = 3, 6, 9, and three

investment horizons, TI = 3, 10, 30.

Table 2 provides results for πC(g, TI, X). It is an increasing function of

both the risk aversion and the investment horizon. It reaches 100%

for a very long investment horizon. The investor, in this case, solely

considers the hedging component linked to consumption.

Table 3 reports the bond sensitivities in the case of terminal

maturity only (Panel A) and in the case of intermediate maturity

(Panel B). Both components are increasing in risk aversion and

investment horizon. However, they differ in size impact. It is higher

for the sensitivity linked to terminal wealth than for that linked to

consumption. Moreover, the magnitude across the three factors

varies significantly, especially for the component linked to terminal

wealth. For example, for g = 6 and TI = 10, we obtain, in the case of

terminal wealth, H1(g,TI) =1.27, H2(g,TI) = 5.98, and H3(g,TI) = 6.48,

and, in the case of intermediate consumption, H1(T1(g,TI,X)) = 0.89,

H2(T2(g,TI,X)) = 2.08, and H3(T3(g,TI,X)) = 2.16.

The size and pattern of the sensitivity-linked consumption is

reflected in the behavior of the intermediate horizon (Table 4). This

consumption horizon is decreasing in risk aversion and increasing

in investor’s terminal horizon.

We finally turn to the ultimate variable of the paper, that is, the

elasticity of consumption with respect to the state variables

(Table 5). Despite the conflicting behavior of the various compo-

nents of demand, the pattern of this component is monotonic. For

all factors, this term is increasing in risk aversion and investment

horizon. Moreover, the elasticity of consumption is substantially

higher for the third factor than for the first one.

κ1 κ2 κ3 θ1 θ2 θ3 μ

65.53% 7.05% 5.25% 15.82% 9.61% 1.73% 7.01%

σ11 σ21 σ21 σ31 σ32 σ31

2.14% -1.78% 0.65% 1.43% -0.46% 0.64%

Table 1 – Base case parameters

πc(g,TI,X)

TI =3 TI =10 TI =30

g=3 79.21% 98.39% 100%

g=6 80.15% 99.02% 100%

g=9 80.46% 99.17% 100%

Table 2 – πC(g, TI, X) as a function of g and TI

panel A — linked to terminal wealth

H1(g,TI) H2(g,TI) H3(g,TI)

TI=3 TI=10 TI=30 TI=3 TI=10 TI=30 TI=3 TI=10 TI=30

g=3 0.875 1.02 1.02 1.8 4.78 8.32 1.85 5.19 10.1

g=6 1.09 1.27 1.27 2.25 5.98 10.4 2.31 6.48 12.6

g=9 1.17 1.35 1.36 2.4 6.38 11.1 2.47 6.92 13.4

panel B — linked to intermediate consumption

H1(T1(g,TI,X)) H2(T2(g,TI,X)) H3(T3(g,TI,X))

TI=3 TI=10 TI=30 TI=3 TI=10 TI=30 TI=3 TI=10 TI=30

g=3 0.546 0.738 0.748 0.875 1.81 1.94 0.89 1.88 2.04

g=6 0.675 0.89 0.897 1.08 2.08 2.16 1.1 2.16 2.25

g=9 0.718 0.938 0.945 1.14 2.16 2.23 1.16 2.24 2.32

Table 3 – Bond sensitivities

T1(g,TI,X) T2(g,TI,X) T3(g,TI,X)

TI=3 TI=10 TI=30 TI=3 TI=10 TI=30 TI=3 TI=10 TI=30

g=3 0.504 0.847 0.872 0.00684 0.015 0.0162 0.00381 0.00843 0.00918

g=6 0.496 0.788 0.801 0.00673 0.0136 0.0142 0.00375 0.00767 0.00804

g=9 0.493 0.771 0.781 0.0067 0.0132 0.0137 0.00373 0.00744 0.00774

Table 4 – Intermediate horizon

∂1C/C(g,TI,X) ∂2C/C(g,TI,X) ∂3C/C(g,TI,X)

TI=3 TI=10 TI=30 TI=3 TI=10 TI=30 TI=3 TI=10 TI=30

g=3 61.46% 74.24% 74.84% 106.77% 185.66% 194.45% 108.94% 193.72% 203.78%

g=6 75.83% 89.36% 89.7% 131.04% 211.49% 216.26% 133.68% 219.93% 225.38%

g=9 80.54% 94.18% 94.46% 138.93% 219.17% 223.02% 141.72% 227.68% 232.08%

Table 5 – Elasticity of consumption as a function of g and TI

107

interest rate risk hedging demand under a Gaussian framework

concluding remarksThis article provides an in-depth analysis of the state variables

Merton-Breeden hedging demand when the opportunity set is

affected by interest rates only. Relying on the three-factor model

of Babbs and Nowman (1999), we show that the Merton-Breeden

hedging demand boils down to a portfolio of risk-adjusted bond

sensitivities with respect to the state variable. The portfolio weight-

ing parameter is the wealth proportion that will be used for future

consumption. Our analysis could be extended to take account of a

stochastic behavior of the market price of risk.

AppendixA. Zero coupon parameters

Following Babbs and Nowman (1999), the bond parameters are as

follow:

r∞ = µ + θ j

σ ij

κ ii=1

3

∑j=1

3

∑ − 1

2

σ ij

κ ii=1

3

∑

2

j=1

3

∑

(A1)

w u( ) = Dκiu( ) θ j

σ i, j

κ i

−σ kjσij

κ kκ ik=1

3

∑j=1

3

∑j=1

3

∑

i=1

3

∑ + 1

2Dκ i +κ j

u( ) σ ikσ jk

κ iκ jk=1

3

∑i, j

3

∑

(A2)

B. Closed-form expression of σ B(v)−θ 2dv

0

TI

∫

Direct computation leads to:

σ B(v)−θ 2dv

0

TI

∫ = θ 2TI + σ ijθ j

1

κ i

TI − Dκi(TI)( )

i, j

3

∑ + mij

1

κ iκ j

TI − Dκi(TI)− Dκ j

(TI)+ Dκi +κ j(TI)( )

i, j

3

∑

σ B(v)−θ 2dv

0

TI

∫ = θ 2TI + σ ijθ j

1

κ i

TI − Dκi(TI)( )

i, j

3

∑ + mij

1

κ iκ j

TI − Dκi(TI)− Dκ j

(TI)+ Dκi +κ j(TI)( )

i, j

3

∑

(A3)

with mij ≡ (σσ’)ij.

c.

It can be shown from the authors that T1(g,TI,X), is given by:

Ti γ,TI , X( ) ≡ κ i log 1 −κ i ψ γ,u, X( )Dκiu( )

0

TI

∫ du

−1

(C1)

with ψ γ,u, X) ≡q γ,u, X( )

Q γ,TI , X( )( . Since κ i log 1 −κ i x( )−1( ) is the inverse

function of Dκi(x), Equation (C1) clearly demonstrates a weighted

average formulae which implies that Ti(g,TI,X)<TI.

ReferencesBabbs, S., and K. Nowman, 1999, “Kalman filtering of generalized Vasicek term struc-•

ture models,” Journal of Financial and Quantitative Analysis, 34:1, 115-130

Breeden, D., 1979, “An intertemporal asset pricing model with stochastic consumption •

and investment opportunities,” Journal of Financial Economics, 7, 263-296

Cox, J., and C. Huang, 1989, “Optimum consumption and portfolio policies when asset •

prices follow a diffusion process,” Journal of Economic Theory, 49, 33-83

Cox, J., and C. Huang, 1991, “A variational problem arising in financial economics,” •

Journal of Mathematical Economics, 21, 465-488

Dai, Q., and K. Singleton, 2000, “Specification analysis of affine term structure mod-•

els,” The Journal of Finance, 55, 1943-1978

Karatzas, I., J. Lehoczky, and S. Shreve, 1987, “Optimal portfolio and consumption •

decisions for a small investor on a finite horizon,” Siam Journal on Control and

Optimization 25, 1557-1586

Litterman, R., and J. Scheinkman, 1991, “Common Factors Affecting Bond Returns,” •

Journal of Fixed Income, 1, 62-74

Merton, R., 1973, “An intertemporal capital asset pricing model,” Econometrica, 41, •

867-887

Munk, C., and C. Sørensen, 2007, “Optimal real consumption and investment strategies •

in dynamic stochastic economies”, in Jensen, B. S., and T. Palokangas, (eds), Stochastic

economic dynamics, CBS Press, 271-316

Articles

109

Emmanuel fragnièreProfessor, Haute École de Gestion de Genève,

and Lecturer, University of Bath

Jacek GondzioProfessor, School of Mathematics,

University of Edinburgh

nils s. TuchschmidProfessor, Haute École de Gestion de Genève

Qun ZhangSchool of Mathematics, University of Edinburgh

non-parametric liquidity-adjusted VaR model: a stochastic programming approach

AbstractThis paper proposes a Stochastic Programming (SP) approach for the

calculation of the liquidity-adjusted Value-at-Risk (LVaR). The model

presented in this paper offers an alternative to Almgren and Chriss’s

mean-variance approach (1999 and 2000). In this research, a two-

stage stochastic programming model is developed with the intention

of deriving the optimal trading strategies that respond dynamically

to a given market situation. The sample paths approach is adopted

for scenario generation. The scenarios are thus represented by a

collection of simulated sample paths rather than the tree structure

usually employed in stochastic programming. Consequently, the SP

LVaR presented in this paper can be considered as a non-parametric

approach, which is in contrast to Almgren and Chriss’s parametric

solution. Initially, a set of numerical experiments indicates that the

LVaR figures are quite similar for both approaches when all the

underlying financial assumptions are identical. Following this sanity

check, a second set of numerical experiments shows how the ran-

domness of the different types (i.e., bid and ask spread) can be easily

incorporated into the problem due to the stochastic programming

formulation and how optimal and adaptive trading strategies can

be derived through a two-stage structure (i.e., a recourse problem).

Hence, the results presented in this paper allow for the introduction

of new dimensionalities into the computation of LVaR by incorporat-

ing different market conditions.

110 – The journal of financial transformation

non-parametric liquidity-adjusted VaR model: a stochastic programming approach

Gondzio and Grothey (2006) showed in their research that they could solve a qua-1

dratic financial planning problem exceeding 109 decision variables by applying a

structure-exploiting parallel primal-dual interior-point solver.

Developed over the last couple of decades, Value-at-Risk (VaR)

models have been widely used as the main market risk manage-

ment tool in the financial world [Jorion (2006)]. VaR estimates

the likelihood of a portfolio loss caused by normal market move-

ments over a given period of time. However, VaR fails to take into

consideration the market liquidity impact. Its estimate is quite

often based on mid-prices and the assumption that transactions do

not affect market prices. Nevertheless, large trading blocks might

impact prices, and trading activity is always costly. To overcome

these problems, some researchers have proposed the calculation

of liquidity adjusted VaR (LVaR) [Dowd (1998)]. Differing from the

conventional VaR, LVaR takes both the size of the initial holding

position and liquidity impact into account. The liquidity impact is

commonly subcategorized into exogenous and endogenous illiquid-

ity factors. The former is normally measured by the bid-ask spread,

and the latter is expressed as the price movement caused by mar-

ket transactions [Bangia et al. (1999)]. From this perspective, LVaR

can be seen as a complementary tool for risk managers who need

to estimate market risk exposure and are unwilling to disregard the

liquidity impact.

Bangia et al. (1999) proposed a simple but practical solution that

is directly derived from the conventional VaR model in which an

illiquidity factor is expressed as the bid-ask spread. Although this

approach avoids many complicated calculations, it fails to take

into consideration endogenous illiquidity factors. Hence, liquidity

risk and LVaR are underestimated. A more promising solution for

LVaR estimation stems from the derivation of optimal trading

strategies as suggested by Almgren and Chriss (1999 and 2000).

In their model, Almgren and Chriss adopted the permanent and

temporary market impact mechanisms from Holthausen et al.’s

work (1987) and assumed linear functions for both of them. By

externally setting a sales completion period, they derived an opti-

mal trading strategy defined as the strategy with the minimum

variance of transaction cost, or of shortfall, for a given level of

expected transaction cost. Or inversely, a strategy that has the

lowest level of expected transaction cost for a given level of vari-

ance. With the normal distribution and the mean and variance of

transaction cost, LVaR can also be determined and minimized

to derive optimal trading strategies. In this setting, LVaR can be

understood as the pth percentile possible loss that a trading posi-

tion can encounter when liquidity effects are incorporated into

the risk measure computation. Later on, Almgren (2003) extended

this model by using a continuous-time approximation, and also

introduced a non-linear and stochastic temporary market impact

function. Another alternative is the liquidity discount approach

presented by Jarrow and Subramanian (1997 and 2001). Similar

to Almgren and Chriss’s approach (1999 and 2000), the liquidity

discount approach requires that the sales completion period be

given as an exogenous factor. The optimal trading strategy is then

derived by maximizing an investor’s expected utility of consump-

tion. Note that both approaches require externally setting a fixed

horizon for liquidation. Aiming to overcome this problem, Hisata

and Yamai (2000) extended Almgren and Chriss’s approach by

assuming a constant speed of sales and by using continuous

approximation. They could derive a closed-form analytical solution

for the optimal holding period. In this setting, the sales comple-

tion time thus becomes an endogenous variable. Yet, Hisata and

Yamai’s model relies on the strong assumption of a constant

speed of sales.

Krokhmal and Uryasev (2006) argued that the solution offered

by Almgren and Chriss and that of Jarrow and Subramanian were

unable to dynamically respond to changes in market conditions.

Therefore, they suggested a stochastic dynamic programming

method and derived an optimal trading strategy by maximizing the

expected stream of cash flows. Under their framework, the optimal

trading strategy becomes highly dynamic as it can respond to mar-

ket conditions at each time step. Another methodology that incor-

porates these dynamics into an optimal trading strategy is that of

Bertsimas and Lo (1998). They applied a dynamic programming

approach to the optimal liquidation problems. Analytical expres-

sions of the dynamic optimal execution strategies are derived by

minimizing the expected trading cost over a fixed time horizon.

In this paper, we present a new framework for the calculation

of non-parametric LVaR by using stochastic programming (SP)

techniques. Over the past few years, stochastic programming has

grown into a mature methodology used to approach decision mak-

ing problems in uncertain contexts. The main advantage of SP is

its ability to better tackle optimization problems under conditions

of uncertainty over time. Due to the fast development of com-

puting power, it has been used to solve large scale optimization

problems1. Therefore, we believe it is a promising methodology for

LVaR modeling.

The SP approach presented in this paper is extended from

Almgren and Chriss’s framework (1999 and 2000). The sample

path approach is adopted for scenario generation, rather than the

scenario tree structure usually employed in SP. The scenario set

is represented by a collection of simulated sample paths. Differing

from Almgren and Chriss’s parametric formulation of LVaR, we

present a non-parametric formulation for LVaR. Both exogenous

and endogenous illiquidity factors are taken into account. The for-

mer is measured by the bid-ask spread, and the latter is expressed

by linear market impact functions, which are related to the quantity

of sales. The model in this paper is built in a discrete-time manner,

and the holding period is required to be determined externally. The

permanent and temporary market impact mechanism proposed by

Holthausen et al. (1987) is adopted to formulate the market impact,

and both permanent and temporary market impacts are assumed

as linear functions.

111

non-parametric liquidity-adjusted VaR model: a stochastic programming approach

For the estimation of temporary and permanent market impact coefficients, Almgren 2

and Chriss did not propose a specific method. They assumed that: for the temporary

market impact, trading each 1% of the daily volume incurs price depression of one bid-

ask spread, and for the permanent market impact, trading 10% of the daily volume will

have a significant impact on price, and incur price depression of one bid-ask spread.

Since this paper focuses on the LVaR modeling, not the estimation of market impact

stochastic programming lVaR modelThis paper proposes a SP approach to estimate the non-parametric

LVaR, which is based on Almgren and Chriss’s mean-variance

approach (1999 and 2000). While their model has been shown to

be an interesting methodology for the calculation of LVaR and has

a huge potential in practice, the optimal trading strategies derived

by their model fail to respond dynamically to the market situation,

as they rely on a ‘closed-form’ or ‘static’ framework. For instance,

if an increasing trend is observed in the market price, investors

may decide to slow their liquidation process. If, on the other hand,

unexpected market shocks occur, investors may decide to adjust

their trading strategy and to speed up the completion of their sale.

These market situations can be simulated and incorporated into

scenarios. Clearly, any ‘closed form’ solution cannot deal with this

type of uncertainty in such a dynamic manner. The LVaR formula-

tion in Almgren and Chriss’s model is based on the mean-variance

framework; thus, it can be considered as a parametric approach.

In contrast, the LVaR formulation presented in this paper is non-

parametric and allows for the incorporation of various dynamics

in the liquidation process. Thus, we propose a new framework for

LVaR modeling.

Almgren and chriss’s mean-variance model

According to Almgren and Chriss’s framework, a holding period T

is required to be set externally. Then, this holding period is divided

into N intervals of equal length (t = T/N). The trading strategy is

defined as the quantity of shares sold in each time interval, which

is denoted by a list of n1,…, nk,…, nN, where nk is the number of

shares that the trader plans to sell in the kth interval. Accordingly,

the quantity of shares that the trader plans to hold at time tk = kt

is denoted by xk. Suppose a trader has a position X that needs to be

liquidated before time T, then we have:

1k k kn x , 1

N

kk

X n=

=1 1

, 0,..., .k N

k j jj j k

x X n n k N= = +

= = = = and 1k k kn x , 1

N

kk

X n=

=1 1

, 0,..., .k N

k j jj j k

x X n n k N= = +

= = = =

Price dynamics in Almgren and Chriss’s model are formulated as an

arithmetic random walk as follows:

1k

k k k

nS S gµ= + + (1)

where Sk is the equilibrium price after a sale, μ and σ are the drift

and volatility of the asset price, respectively, and xk is a random

number that follows a standard normal distribution N (0, 1). The last

term, g(nk/t), describes the permanent market impact from a sale.

The actual sale price is calculated by subtracting the temporary

impact, h(nk/t), from the equilibrium price:

kk k

nS S h=� (2)

According to Almgren and Chriss’s framework (1999 and 2000),

both g(nk/t) and h(nk/t) are assumed to be linear functions:

k kn ng = 1

2k kn n

h = + (3) and k kn ng = 1

2k kn n

h = + (4)

where g and h are the permanent and temporary market impact

coefficients2, respectively, and e denotes the bid-ask spread. They

are all assumed to be constant.

Based on the previously presented equations, the formula for the

actual sale price is derived as:

01 1

1

2

k kk

k j kj j

nS S k nµ

= =

= + +�

I II III

(5)

As you can see from this formula, the actual sale price can be

decomposed into three parts. Part I is the price random walk, which

describes the price dynamics without any market impacts. Parts II

and III are the price decline caused by the permanent and tempo-

rary market impact, respectively.

Then the total proceeds can be calculated by summing the sale

values over the entire holding period:

( ) 20 1 1

1 1 1 1 1

2 20 1 1

1 1 1

1

2

1 1 1

2 2 2

N N N N N

k k k k k k k kk k k k k

N N N

k k k kk k k

n S XS x x n X x X n

XS x x X X n

µ

µ

= = = = =

= = =

= = + +

= + +

�total proceed

(6)

Consequently, ‘liquidation cost’ (LC)3 can be derived by subtract-

ing the total actual sale proceeds from the trader’s initial holding

value, that is:

2 20 1 1

1 1 1 1

1 1 1

2 2 2

N N N N

k k k k k kk k k k

LC XS n S x x X X nµ= = = =

= = + + +� (7).

Almgren and Chriss derive the formulae for the mean and variance

of the liquidation cost as:

[ ] 2 21

1 1

1 1 1

2 2 2

N N

k kk k

E LC X x X nµ= =

= + +

[ ] 2 21

1

N

kk

V LC x=

=

(8)

and

[ ] 2 21

1 1

1 1 1

2 2 2

N N

k kk k

E LC X x X nµ= =

= + +

[ ] 2 21

1

N

kk

V LC x=

= (9)

Finally, they formulate the LVaR by using the parametric approach

with the mean and variance of the LC as:

[ ] [ ]clLVaR E LC V LC= + (10)

where cl denotes the confidence level for the LVaR estimation, and

αcl is the corresponding percentile of the standard normal distribu-

tion. As expressed, LVaR measures a possible loss with a given

position while taking into consideration both market risk conditions

and liquidity effects.

coefficients, Almgren and Chriss’s simple assumption is adopted for all the numerical

experiments in this paper.

In Almgren and Chriss’s paper, this ‘cost’ is referred as the transaction cost. However, 3

the transaction cost is commonly known as the fees involved for participating in the

market, such as the commission to the brokers. Therefore, in order to avoid any con-

fusion, it is named ‘liquidation cost’ in this paper.

112 – The journal of financial transformation

non-parametric liquidity-adjusted VaR model: a stochastic programming approach

An extension is presented below.4

The optimal trading strategy could be derived by minimizing the

LVaR. The mathematical programming formulation of this optimiza-

tion problem is thus written as:

[ ] [ ]

1

1

1

min

. 0,..., 1

,..., 0.

kcl

n

N

k jj k

N

kk

N

E LC V LC

st x n k N

X n

n n

= +

=

+

= =

=

Based on this brief introduction to Almgren and Chriss’s mean-vari-

ance approach, we can thereon proceed to present the SP approach

to LVaR modeling.

stochastic programming transformationIn stochastic programming, uncertainty is modeled with scenarios

that are generated by using available information to approximate

future conditions. Before conducting the SP transformation, we

need to briefly introduce the scenario generation technique used

in this paper.

The liquidation process of investors’ positions is a multi-period

problem. The most commonly used technique is to model the evo-

lution of stochastic parameters with multinomial scenario trees, as

shown in Figure 1(a).

However, the use of scenario tree structures often leads to con-

siderable computational difficulty, especially when dealing with

large scale practical problems. In the scenario tree structure, the

uncertainties are represented by the branches that are gener-

ated from each node. Increasing the number of branches per node

can improve the quality of the approximation of the uncertainty.

However, it causes an exponential growth in the number of nodes.

Indeed, in order to approximate the future value of the uncertain

parameters with a sufficient degree of accuracy, the resulting

scenario tree could be of a huge size. It is commonly known as

the “curse of dimensionality” [Bellman (1957)]. It is a significant

obstacle for dynamic or stochastic optimization problems. An

alternative method to overcome this problem is to simulate a col-

lection of sample paths to reveal the future uncertainty as shown

in Figure 1(b). Each simulated path represents a scenario. These

sample paths can be generated by using Monte Carlo simulation,

historical simulation, or bootstrapping. There have been several

interesting papers regarding the application of the sample paths

method in stochastic programming [Hibiki (2000), Krokhmal and

Uryasev (2006)]. Using sample paths is advantageous because

increasing the number of paths to achieve a better approximation

causes the number of nodes to increase linearly rather than expo-

nentially. This advantage is also present when the time period is

increased. The number of nodes increases linearly with the sample

paths method and exponentially with scenario tree structure.

Let ( ){ }0 1 , 2, , ,, , , , , , | 1 , ,s s k s N sC C C C C s Sc= =… … … , be a collection of sample paths

( ){ }0 1 , 2, , ,, , , , , , | 1 , ,s s k s N sC C C C C s Sc= =… … … ,

where Ck,s represents the information about relevant parameters.

In Almgren and Chriss’s model (1999 and 2000), we should recall

that the only randomness considered is market price. Hence, to

set a point of comparison between their results and the results

from the SP approach, we first assume the only randomness that

is considered in the sample paths will come from the market price

component, k. Yet, this restrictive assumption can clearly be easily

relaxed, and randomness can be added to other parameters, such

as the bid-ask spread or the temporary and permanent market

impact coefficients4.

Under the SP framework, the trading strategy is no longer a vector

but a two dimensional matrix

1 ,1 ,1 ,1

1 , , ,

1 , , ,

k N

s k s N s

Sc k Sc N Sc

n n n

n n n

n n n

� �� � �

� � �� �

strategy=

first stage second stage

where nk,s is the quantity of shares sold in kth interval on path s, s is

the index of scenarios, and Sc is the number of scenarios. This is a

two stage SP problem. n1,s (s = 1,...,Sc) are the first stage variables,

and nk,s (k = 2,…,N and s = 1,...,Sc) are the second stage variables.

Due to the nonanticipativity in the first stage, the first stage vari-

ables must be locked:

1 , 1 , 1 2,...s sn n s Sc= = .

For the actual sale price formulation, recall Equation (5). Taking

into account the scenarios and replacing part I with k,s (i.e., the Figure 1 – Scenario generation

(a) scenario tree

Time period Time period

(b) sample paths

113

non-parametric liquidity-adjusted VaR model: a stochastic programming approach

asset price without market impacts in kth interval for each scenario),

the actual sale price is reformulated as:

,, , ,

1

1ˆ2

kk s

k s k s j sj

nS S n

=

=� (11)

As we now have the sale price formulation, the total sale proceeds

corresponding to each scenario is naturally obtained by summing

up the sale proceeds over the entire set of N intervals:

2,

, , , , , , ,1 1 1

2 2, , ,

1 1

1ˆ2

1 1 1ˆ2 2 2

N N kk s

k s k s k s k s k s k s j sk k j

N N

k s k s k sk k

nn S S n n n n

S n X X n

= = =

= =

= =

=

�.

total proceed

(12)

Consequently, the liquidation cost under scenario s is derived by

subtracting the corresponding total sale proceeds from the trader’s

initial holding value, that is:

2 20 , , 0 , , ,

1 1 1

1 1 1ˆ2 2 2

N N N

s k s k s k s k s k sk k k

LC XS n S XS X X S n n= = =

= = + + +�

(13)

The deterministic equivalent formulation of this SP problem with

nonanticipativity constraints is:

,1

,1

1 , ,

1 , 1 , 1

min

. 1 ,...,

, ..., 0 1 ,...,

2,...

k s

Sc

s sn

s

N

k sk

s N s

s s

p LC

st X n s Sc

n n s Sc

n n s Sc

=

=

= =

== =

where ps is the probability of scenario s. Since the scenarios are

obtained by the Monte Carlo simulation, they are thus equally prob-

able with ps = 1 / Sc.

The resulting problem is a quadratic optimization one. The objec-

tive, the expected value of the liquidation cost, is a quadratic func-

tion, and all constraints are linear.

non-parametric lVaR formulation

Depending on the set of assumptions, the calculation methodology,

and their uses, two different types of VaRs usually exist, i.e., the

parametric VaR and the non-parametric VaR. The same categoriza-

tion obviously applies to LVaR. The LVaR estimated by Almgren

and Chriss (1999 and 2000) is parametric as shown in Equation

[10]. In this paper, we have to rely on a non-parametric formulation

because it stems from the SP solution that we have adopted. More

precisely, the calculation procedure is as follows:

Solve the stochastic optimization problem stated above and 1

obtain the optimal trading strategy matrix.

Apply the optimal trading strategies to the corresponding sce-2

nario and calculate the liquidation cost for that specific scenario.

That is to say, we substitute the optimal trading strategy matrix

into the liquidation cost formula (Equation [13]), and calculate

LC, which is a vector indexed by s.

Sort the vector LC, and find the value of the 3 αth percentile LC, i.e.,

the α%-LVaR. The most commonly used α value is 95 and 99.

numerical experiments iAs previously mentioned, we first conducted a sanity check. This

section details the numerical experiments for both the SP model

and the Almgren and Chriss’s mean-variance model with the restric-

tion of randomness on one component only, i.e., the ‘pure market

price.’

JP Morgan’s stock data was collected for the numerical experi-

ments. The holding period, T, was set to be 5 days, and we selected

the time interval to be 0.5 day. Thus, the total number of sales, N,

was 10. The selection of the holding period and time interval was

arbitrary.

For the price sample paths generation, the Monte Carlo simulation

was applied. The stochastic evolution of the price was assumed to

follow a geometric Brownian motion:

21

1ˆ ˆ exp2k k kS S µ= +

(14)

Under Almgren and Chriss’s mean-variance framework, market

price was assumed to follow an arithmetic random walk because it

is ultimately rather difficult to derive a closed-form solution based

on an assumption of geometric Brownian motion. Yet, with Monte

Carlo simulations, formulating the price evolution under different

assumptions creates no issues related to the underlying distribu-

tions that could generate prices and returns. Since the geometric

random walk is the most commonly used assumption for price sto-

chastic processes, it is used in this paper even though differences

between these two random walks are almost negligible over a short

period of time.

10,000 sample paths were generated by using the Monte Carlo

simulation. The simulated prices form a 10-by-10,000 matrix. The

initial price is 37.72. These simulated sample paths are displayed

in Figure 2.

Five different initial holdings were chosen for the numerical experi-

ments with the aim of observing how the initial position affected

the LVaR estimation. The LVaRs were calculated with the most

commonly seen confidence levels of 95% and 99%. The results are

summarized in Table 1.

The numerical results show that the LVaR ratios computed by the

SP model are slightly lower than those computed by Almgren and

114 – The journal of financial transformation

non-parametric liquidity-adjusted VaR model: a stochastic programming approach

Chriss’s model at both the 95% and 99% confidence levels. Also,

as expected, the numerical results show that the LVaR estimates

increase when the initial holdings increase. This is true for both

Almgren and Chriss’s model and the SP model. As previously stated,

as the initial holding becomes larger, the market impact becomes

stronger when a trader liquidates his position. Consequently, the

LVaR will increase as well. This is clearly a characteristic that dis-

tinguishes the LVaR from the traditional VaR.

SP LVaRs are lower than Almgren and Chriss’s LVaRs because the

SP model’s optimal trading strategies can dynamically adapt to the

market situation. This fits investors’ actual trading behaviors in

the market, as they will adjust their trading plans according to the

market environment. Therefore, the SP model can provide more

precise LVaR estimates due to the characteristic of the SP model’s

adaptive trading strategies.

Generalization of the stochastic programming lVaR modelA simple stochastic programming LVaR model that was transformed

from Almgren and Chriss’s mean-variance model was presented

above in order to compare with other models discussed (i.e., both

two LVaR approaches were used under the same setting). Let us now

extend the analysis and show some of the advantages provided by the

SP approach. Contrary to Almgren and Chriss’s model that assumes

that both the bid-ask spread and market impact coefficients are con-

stants, we generalize the SP LVaR model by relaxing this assumption

and treating these two components as random variables.

By incorporating randomness in the bid-ask spread and both the

permanent and temporary market impact coefficients, the formula

of the actual sale price needs to be rewritten as:

, ,, , , , ,

1

1ˆ2

kk s k s

k s k s j s j s k sj

nS S n

=

=�

(15)

For the formulation of ek,s, we employ a standardization process.

Since bid-ask spreads tend to be proportional to asset prices, past

observations may not accurately reflect the current variations.

Bangia et al. (1999) suggested calculating a relative bid-ask spread

that is equal to the bid-ask spread divided by the mid-price. By

employing this calculation, the bid-ask spread is expressed as a pro-

portion of the asset price; thus, the current bid-ask spread variation

is sensitive to the current asset price rather than past observations.

The relative bid-ask spread, as a normalizing device, can improve

the accuracy of the bid-ask spread variation estimation. The bid-ask

spread is thus formulated as:

, , ,ˆ

k s k s k sS= (16)

where , , ,ˆ

k s k s k sS= is the relative bid-ask spread at time tk on path s. Recall

the sample path set

0 0.5 1 1. 5 2 2.5 3 3.5 4 4.5 532

34

36

38

40

42

44

Figure 2 – Simulated price scenarios

Price

Time (days)

Price scenario (simulated random walk

initial holding (shares) 1000000 500000 100000 50000 10000

Almgren and chriss’s Mean-Variance Model

Parametric LVaR 95%

1.425E+06 6.186E+05 1.010E+05 4.845E+04 9.311E+03

Parametric LVaR per share

1.425 1.237 1.010 0.969 0.931

Parametric LVaR ratio

3.78% 3.28% 2.68% 2.57% 2.47%

Parametric LVaR 99%

1.863E+06 8.216E+05 1.388E+05 6.718E+04 1.304E+04

Parametric LVaR per share

1.863 1.643 1.388 1.344 1.304

Parametric LVaR ratio

4.94% 4.36% 3.68% 3.56% 3.46%

stochastic programming Model

Non-parametric LVaR 95%

1.290E+06 5.399E+05 8.144E+04 3.832E+04 7.218E+03

Non-parametric LVaR per share

1.290 1.080 0.814 0.766 0.722

Non-parametric LVaR ratio

3.42% 2.86% 2.16% 2.03% 1.91%

Non-parametric LVaR 99%

1.811E+06 7.944E+05 1.342E+05 6.442E+04 1.249E+04

Non-parametric LVaR per share

1.811 1.589 1.342 1.288 1.249

Non-parametric LVaR ratio

4.80% 4.21% 3.56% 3.42% 3.31%

Table 1 – Numerical results summary

*LVaR per share = LVaR/Initial holding; LVaR ratio = LVaR per share/Initial price

115

non-parametric liquidity-adjusted VaR model: a stochastic programming approach

( ){ }0 1 , 2, , ,, , , , , , | 1 , ,s s k s N sC C C C C s Sc= =… … … .

By incorporating the randomness into the relative bid-ask spread

and market impact coefficients, Ck,s is extended to

( ), , , , ,ˆ , , ,k s k s k s k s k sC S= .

In other words, in the generalized SP model, each node on the simu-

lated sample paths contains information for the asset price, the

relative bid-ask spread, and the permanent and temporary market

impact coefficients.

As we now have the new sale price formulation, the formula for

liquidation cost under scenario s (as shown in Equation [13]) is

rewritten as:

2, ,

0 , , 0 , , , , , , , ,1 1 1

1ˆ ˆ2

N N kk s k s

s k s k s k s k s k s k s k s k s k s j sk k j

nLC XS n S XS S n S n n n

= = =

= =� (17)

The deterministic equivalent formulation and the LVaR calculation

procedure are the same as shown above. By generalizing the SP

model, more parameters are incorporated in the sample path set,

which leads to a more accurate approximation of future uncertain-

ties.

numerical experiments iiThis section reports the numerical experiments for the generalized

SP LVaR model presented above. We use the same dataset that

was used for aforementioned numerical experiments. The holding

period and time interval remain identical, i.e., 5 days and half a day,

respectively.

For the sample paths’ generation of the relative bid-ask spread, and

the permanent and temporary market impacts, we assumed that

they followed independent lognormal distributions and simulated

each of them simply as a white noise:

(18)

(19)

(20)

2, ,

1exp

2k s k sµ= +

2, ,

1exp

2k s k sµ= +

2, ,

1exp

2k s k sµ= +

where μ and σ are the means and standard deviations, respectively,

of the three random variables (i.e., e, g and h).

Once again 10,000 sample paths were generated by using the

Monte Carlo simulation for each parameter. The LVaRs at the 95%

and 99% confidence levels were computed for the same five initial

holding scenarios employed above. The results are summarized in

Table 2.

In Figure 3, we can see that the LVaR ratios computed by the SP

model with the incorporation of randomness into the bid-ask spread

and the market impact coefficients are slightly lower than those

computed by the SP model with the constant bid-ask spread and

market impact coefficients. When the initial holding is small, incor-

porating these new random variables does not cause a significant

change to the LVaR estimate. However, when the initial holding is

large, the differences are substantial. For instance, when the initial

holding is 1,000,000, incorporating randomness reduces the 95%

LVaR ratio from 3.42% to 2.87% and the 99% LVaR ratio from

4.80% to 4.30%.

The main reason for these differences must lie in the way the opti-

mal trading strategies that are derived by the SP model respond to

the variation of the bid-ask spread and market impact coefficients.

For example, if we assume that the bid-ask spread is constant, the

initial holding (shares) 1000000 500000 100000 50000 10000

Non-parametric LVaR 95%

1.084E+06 4.740E+05 7.800E+04 3.739E+04 7.119E+03

Non-parametric LVaR per share

1.084 0.948 0.780 0.748 0.712

Non-parametric LVaR ratio

2.87% 2.51% 2.07% 1.98% 1.89%

Non-parametric LVaR 99%

1.621E+06 7.389E+05 1.312E+05 6.403E+04 1.243E+04

Non-parametric LVaR per share

1.621 1.478 1.312 1.281 1.243

Non-parametric LVaR ratio

4.30% 3.92% 3.48% 3.40% 3.30%

Table 2 – Numerical results

Figure 3: LVaR ratio comparison

1,5%

2,0%

2,5%

3,0%

3,5%

4,0%

4,5%

5,0%

1.000.000 500.000 100.000 50.000 10.000

Initial holding (shares)

LVaR ratio 95% with incorporation of the variation of spread and market impact coefficients

LVaR ratio 95% with constant spread and market impact coefficients

LVaR ratio 99% with incorporation of the variation of spread and market impact coefficients

LVaR ratio 99% with constant spread and market impact coefficients

LVaR ratio

116 – The journal of financial transformation

non-parametric liquidity-adjusted VaR model: a stochastic programming approach

loss caused by the spread in the whole liquidation process is e X/2

for each scenario as shown in Equation (13) (e is the mean value

of the bid-ask spread). With the incorporation of randomness into

the bid-ask spread, the optimal trading strategies are adjusted in

accordance with its variation. When the spread is high, the optimal

trading strategy may suggest selling less. On the contrary, when

it is low, the optimal trading strategy may suggest selling more.

Therefore, the average loss caused by the spread can be expected

to be lower than e X/2. Stated otherwise, the SP model’s optimal

trading strategies can take advantage of changes by acting in a

flexible and timely manner. Also note that introducing the calcula-

tion of the relative bid-ask spread and the Monte Carlo simulation

itself can cause certain differences. However, the effects are pre-

sumably small.

Finally, it is worthwhile mentioning that incorporating randomness

into the bid-ask spread and market impact coefficients within the

Almgren and Chriss’s model will definitely enlarge the resulting

LVaR estimates. Indeed, it would add new variance terms to the

variance of the liquidation cost, since the variations of parameters

are represented by their variances. This would lead to the increase

of the LVaR estimates. The SP solution and its numerical experi-

ments indicate that if uncertainty is handled well, it does not nec-

essarily cause an increase in the LVaR estimates. It highlights the

strength of the SP approach, which provides adaptive strategies

(or ‘recourse strategies’). Moreover, adding new random variables

in the model does not increase the difficulty of the problem due to

the non-parametric nature of the SP LVaR.

conclusionThis paper presents a stochastic programming approach for LVaR

modeling, which is extended from Almgren and Chriss’s mean-

variance approach. In contrast to their approach, the optimal trad-

ing strategies are derived by minimizing the expected liquidation

cost. Thus, the SP strategies dynamically adapt to new market

situations. This is the strength of SP in the context of decision

making under uncertainty. Another contribution from this paper

is the non-parametric formulation of the SP LVaR. It contrasts

with the LVaR modeling methodologies that quite often rely on

parametric approaches. Overall, the numerical results indicate that

the two approaches are not identical. Indeed, the LVaRs computed

using the SP model in this paper are lower than those computed by

Almgren and Chriss’s model. Yet, LVaR modeling still remains in its

infancy, especially when using SP in this context.

ReferencesAlmgren, R. and N. Chriss, 1999, “Value under liquidation,” Risk, 12, 61-63•

Almgren, R. and N. Chriss, 2000, “Optimal execution of portfolio transactions,” Journal •

of Risk, 3, 5-39

Almgren, R., 2003, “Optimal execution with nonlinear impact functions and trading-•

enhanced risk,” Applied Mathematical Finance, 10, 1-50

Bangia, A., F. Diebold, T. Schuermann and J. Stroughair, 1999, “Liquidity on the out-•

side,” Risk, 12, 68-73

Bertsimas, D., and A. Lo, 1998, “Optimal control of execution costs,” Journal of •

Financial Markets, 1, 1-50

Bellman, R.E., 1957, Dynamic programming, Princeton University Press, Princeton, New •

Jersey

Dowd K., 1998, Beyond Value at Risk: the new science of risk management, Wiley and •

Sons, New York

Gondzio, J. and A. Grothey, 2006, “Solving nonlinear financial planning problems with •

109 decision variables on massively parallel architectures,” in Costantino, M., and C.A.

Brebbia (eds.), Computational finance and its applications II, WIT transactions on mod-

elling and simulation, 43, WIT Press, Southampton

Hibiki, N., 2000, “Multi-period stochastic programming models for dynamic asset allo-•

cation,” Proceedings of the 31st ISCIE international symposium on stochastic systems

theory and its applications, 37-42

Hisata, Y. and Y. Yamai, 2000, “Research toward the practical application of liquidity •

risk evaluation methods,” Monetary and Economic Studies, 83-128

Holthausen, R. W., R. W. Leftwich and D. Mayers, 1987, “The effect of large block trans-•

actions on security prices: a cross-sectional analysis,” Journal of Financial Economics,

19, 1987, 237-268

Holthausen, R. W., R. W. Leftwich and D. Mayers, 1990, “Large-block transactions, the •

speed of response, and temporary and permanent stock-price effects,” Journal of

Financial Economics, 26, 1990, 71-95

Jarrow, R. A., and A. Subramanian, 1997, “Mopping up liquidity,” Risk, 10, 170-173•

Jarrow, R. A., and A. Subramanian, 2001, “The liquidity discount,” Mathematical •

Finance, 11:4, 447-474

Jorion P., 2006, Value at Risk: the new benchmark for managing financial risk, 3rd Ed., •

McGraw-Hill, New York

Krokhmal, P., and S. Uryasev, 2006, “A sample-path approach to optimal position liqui-•

dation,” Annals of Operations Research, Published Online, November, 1-33

Articles

117This paper is based on Manfred Gilli’s leçon d’adieu, given at the Conférence Luigi 1

Solari 2009 in Geneva. Both authors gratefully acknowledge financial support from

the E.U. Commission through MRTN-CT-2006-034270 COMISEF.

Manfred Gilli1University of Geneva and Swiss Finance Institute

Enrico schumannUniversity of Geneva

optimization in financial engineering — an essay on ‘good’ solutions and misplaced exactitude

AbstractWe discuss the precision with which financial models are handled,

in particular optimization models. We argue that precision is only

required to a level that is justified by the overall accuracy of the

model, and that this required precision should be specifically ana-

lyzed in order to better appreciate the usefulness and limitations of

a model. In financial optimization, such analyses are often neglect-

ed; operators and researchers rather show an a priori preference

for numerically-precise methods. We argue that given the (low)

empirical accuracy of many financial models, such exact solutions

are not needed; ‘good’ solutions suffice. Our discussion may appear

trivial: everyone knows that financial markets are noisy, and that

models are not perfect. Yet the question of the appropriate preci-

sion of models with regard to their empirical application is rarely

discussed explicitly; specifically, it is rarely discussed in university

courses on financial economics and financial engineering. Some

may argue that the models’ errors are understood implicitly, or

that in any case more precision does no harm. Yet there are costs.

We seem to have a built-in incapacity to intuitively appreciate ran-

domness and chance. All too easily then, precision is confused with

actual accuracy, with potentially painful consequences.

118 – The journal of financial transformation

optimization in financial engineering — an essay on ‘good’ solutions and misplaced exactitude

Der Mangel an mathematischer Bildung gibt sich durch nichts

so auffallend zu erkennen wie durch maßlose Schärfe im

Zahlenrechnen. (Carl Friedrich Gauß)

Imagine you take a trip to the Swiss city of Geneva, known for

international organizations, expensive watches and, not least, its

lake. After leaving the train station, you ask a passerby how far it is

to the lakefront. You are told, ‘Oh, just this direction, along the Rue

du Mont-Blanc. It’s 512.9934 meters.’

This is not a made-up number. Google Earth allows you to track

your path with such a precision. We measured this route around

10 times and found that it ranged between roughly 500 and 520

meters. (Unfortunately, if you had actually tried to take this route,

during most of 2009 you would have found that the street was

blocked by several construction sites.) When we are told ‘it’s 500

meters to the lake,’ we know that this should mean about, say,

between 400 and 600 meters. We intuitively translate the point

estimate into a range of reasonable outcomes.

In other fields, we sometimes seem to lack such an understanding. In

this short article, we shall look at such a field, financial engineering.

We will argue that a misplaced precision can sometimes be found

here, and we will discuss it in the context of financial optimization.

financial modelingIn setting up and solving an optimization model, we necessarily

commit a number of approximation errors. [A classic reference on

the analysis of such errors is von Neumann and Goldstine (1947).

See also the discussion in chapter 6 of Morgenstern (1963).] ‘Errors’

does not mean that something went wrong; these errors will occur

even if all procedures work as intended. The first approximation is

from the real problem to the model. For instance, we may move

from actual prices in actual time to a mathematical description of

the world, where both prices and time are continuous (i.e., infinitely-

small steps are possible). Such a model, if it is to be empirically

meaningful, needs a link to the real world, which comes in the form

of data, or parameters that have to be forecast, estimated, simu-

lated, or approximated in some way. Again, we have a likely source

of error, for the available data may or may not well reflect the true,

unobservable process.

When we solve such models on a computer, we approximate a solu-

tion; such approximations are the essence of numerical analysis. At

the lowest level, errors come with the mere representation of num-

bers. A computer can only represent a finite set of numbers exactly;

any other number has to be rounded to the closest representable

number, hence we have what is called roundoff error. Then, many

functions (i.e., the logarithm) cannot be computed exactly on a

computer, but need to be approximated. Operations like differ-

entiation or integration, in mathematical formulation, require a

‘going to the limit,’ i.e., we let numbers tend to zero or infinity. But

that is not possible on a computer, any quantity must stay finite.

Consequently, we have so-called truncation error. For optimization

models, we may incur a similar error. Some algorithms, in particular

the methods that we describe below, are stochastic. As a result, we

do not (in finite time) obtain the model’s ‘exact’ solution, but only an

approximation (notwithstanding other numerical errors).

In sum, we can roughly divide our modeling into two steps: from

reality to the model, and then from the model to its numerical solu-

tion. Unfortunately, large parts of the quantitative finance literature

seem only concerned with assessing the quality of the second step,

from model to implementation, and attempt to improve here. In the

past, a certain division of labor has been necessary: the economist

created his model, and the computer engineer put it into numerical

form. But today, there is little distinction left between the research-

er who creates the model, and the numerical analyst who imple-

ments it. Modern computing power allows us to solve incredibly

complex models on our desktops. (John von Neumann and Herman

Goldstine, in the above-cited paper, describe the inversion of ‘large’

matrices where large meant n>10. In a footnote, they ‘anticipate

that n~100 will become manageable’ (fn. 12). Today, Matlab inverts

a 100x100 matrix on a normal desktop PC in a millisecond. (But

please, you will not solve equations by matrix inversion.) But then of

course, the responsibility to check the reasonableness of the model

and its solution lies — at all approximation steps — with the financial

engineer. And then only evaluating problems with respect to their

numerical implementation falls short of what is required: any error

in this step must be set into context, we need to compare it with

the error introduced in the first step. But this is, even conceptually,

much more difficult.

Even if we accepted a model as ‘true,’ the quality of the model’s

solution would be limited by the attainable quality of the model’s

inputs. Appreciating these limits helps us decide how ‘exact’ a solu-

tion we actually need. This decision is relevant for many problems

in financial engineering since we generally face a trade-off between

the precision of a solution and the effort required (most apparently,

computing time). Surely, the numerical precision with which we

solve a model matters; we need reliable methods. Yet, empirically,

there must be a required-precision threshold for any given problem.

Any improvement beyond this level cannot translate into gains

regarding the actual problem anymore; only in costs (increased

computing time or development costs). For many finance problems,

we guess, this required precision is not high.

Example 1 — in numerical analysis, the sensitivity of the problem

is defined as follows: if we perturb an input to a model, the change

in the model’s solution should be proportional. If the impact is far

larger, the problem is called sensitive. Sensitivity is often not a

numerical problem; it rather arises from the model or the data. In

119

optimization in financial engineering — an essay on ‘good’ solutions and misplaced exactitude

finance, many models are sensitive. Figure 1 shows the S&P 500

from 31 December 2008 to 31 December 2009, i.e., 253 daily prices.

The index level rose by 23%. 23.45%, to be more precise (from

903.25 to 1115.10). But does it make sense to report this number to

such a precision?

Suppose we randomly pick two observations (less than one percent

of the daily returns), delete them, and recompute the yearly return.

Repeating this jackknifing 5000 times, we end up with a distribu-

tion of returns (right panel of Figure 1). The median return is about

23%, but the 10th quantile is 20%, the 90th quantile is 27% (the

minimum is only about 11%, the maximum is 34%!). Apparently,

tiny differences like adding or deleting a couple of days cause very

meaningful changes.

This sensitivity has been documented in the literature [for instance

in Acker and Duck (2007); or Dimitrov and Govindaraj (2007)],

but it is rarely heeded. The precision with which point estimates

are sometimes reported must not be confused with accuracy. We

may still be able to give qualitative findings (like ‘this strategy

performed better than another’), but we should not make single

numbers overly precise; we need robustness checks. Returns are

the empirical buildings blocks of many models. If these simple

calculations are already that sensitive, we should not expect more

complex computations to be more accurate.

Example 2 — the theoretical pricing of options, following the papers

of Black, Scholes, and Merton in the 1970s, is motivated by an arbi-

trage argument according to which we can replicate an option by

trading in the underlier and a riskless bond. A replication strategy

prescribes to hold a certain quantity of the underlier, the delta. The

delta is changing with time and with moves in the underlier’s price,

hence the options trader needs to rebalance his positions. Suppose

you live in a Black-Scholes-Merton world. You just sold a one-month

call (strike and spot price are 100, no dividends, riskfree rate is at

2%, volatility is constant at 30%), and you wish to hedge the posi-

tion. There is one deviation from Black-Scholes-Merton, though: you

cannot hedge continuously, but only at fixed points in time [Kamal

and Derman (1999)].

We simulate 100,000 paths of the stock price, and delta-hedge

along each path. We compute two types of delta: one is the delta

as precise as Matlab can get and the other is rounded to two digits

(i.e., 0.23 or 0.67). Table 1 shows the volatility of the hedging-error

(i.e., difference between the achieved payoff and the contractual

payoff) as a percentage of the initial option price. (It is often help-

ful to scale option prices, i.e., price to underlier, or price to strike.)

Figure 2 shows replicated option payoffs.

The volatility of the profit-and-loss is practically the same, so even

in the model world nothing is lost by not computing delta to a high

precision. Yet in research papers on option pricing, we often find

prices and Greeks to 4 or even 6 decimals.

Here is a typical counterargument: ‘True, for one option we don’t

need much precision. But what if we are talking about one million

options? Then small differences matter.’ We agree; but the question

is not whether differences matter, but whether we can meaningfully

compute them. (Your accountant may disagree. Here is a simple

rule: whenever you sell an option, round up and when you buy,

round down.) Between buying one share of IBM stock or buying one

million shares, there is an important difference: you take more risk.

We can rephrase our initial example: you arrive at the train station

in Geneva, and ask for the distance to Lake Zurich.

optimization in financial engineeringHeuristics

The obsession with precision is also found in financial optimization;

researchers are striving for exact solutions, better even if in closed-

form. Finding these exact solutions is not at all straightforward, for

most problems it is not possible. Importantly, optimization methods

like linear or quadratic programming place — in exchange for exact

solutions — considerable constraints on the problem formulation.

We are often required to shape the problem such that it can be

solved by such methods. Thus, we get a precise solution, but at the

Jan 09 Apr 09 Jul 09 Oct 09 Jan 10

700

800

900

1000

1100

10% 15% 20% 25% 30% 35%

Figure 1 – Left: The S&P 500 in 2009. Right: Annual returns after jackknifing 2

observations. The vertical line gives the realized return.

frequency of rebalancing with exact delta with delta to two digits

once per day 18.2% 18.2%

five times per day 8.3% 8.4%

Table 1 – Volatility of profit-and-loss under different hedging schemes.

-5

0

5

10

15

20

25

80 85 90 95 100 105 110 115 120-5

0

5

10

15

20

25

80 85 90 95 100 105 110 115 120

Figure 2 – Payoff of replicating portfolios with delta to double precision (left), and

delta to 2 digits (right).

120 – The journal of financial transformation

optimization in financial engineering — an essay on ‘good’ solutions and misplaced exactitude

price of possibly incurring more approximation error at an earlier

stage. An example from portfolio optimization can illustrate this

point. Markowitz (1959, chapter 9) compares two risk measures,

variance and semi-variance, along the dimensions cost, conve-

nience, familiarity, and desirability, and concludes that variance

is superior in terms of cost, convenience, and familiarity. For vari-

ance, we can compute the exact solution to the portfolio selection

problem; for semi-variance, we can only approximate the solution.

But with today’s computing power (the computing power we have

on our desktops), we can test whether even with an inexact solution

for semi-variance the gains in desirability are worth the effort.

To solve such a problem, we can use optimization heuristics. The

term heuristic is used in different fields with different, though relat-

ed, meanings. In mathematics, it is used for derivations that are not

provable (sometimes even incorrect), but lead to correct conclu-

sions nonetheless. [The term was made famous by George Pólya

(1957).] Psychologists use the word for simple ‘rules of thumb’ for

decision making. The term acquired a negative connotation through

the works of Kahnemann and Tversky in the 1970s, since their ‘heu-

ristics and biases’ program involved a number of experiments that

showed the apparent suboptimalitiy of such simple decision rules.

More recently, however, an alternative interpretation of these

results has been advanced [Gigerenzer (2004, 2008)]. Studies indi-

cate that while simple rules underperform in stylized settings, they

yield (often surprisingly) good results in more realistic situations,

in particular in the presence of uncertainty. The term heuristic is

also used in computer science; Pearl (1984) describes heuristics as

methods or rules for decision making that are (i) simple, and (ii) give

good results sufficiently often.

In numerical optimization, heuristics are methods that aim to provide

good and fast approximations to optimal solutions [Michalewicz and

Fogel (2004)]. Conceptually, they are often very simple; implement-

ing them rarely requires high levels of mathematical sophistication

or programming skills. Heuristics are flexible, we can easily add,

remove, or change constraints, or modify the objective function.

Well-known examples for such techniques are Simulated Annealing

and Genetic Algorithms. Heuristics employ strategies that differ

from classical optimization approaches, but exploit the processing

power of modern computers; in particular, they include elements

of randomness. Consequently, the solution obtained from such

a method is only a stochastic approximation of the optimum; we

trade-off approximation error at the solution step against approxi-

mation error when formulating the model. Thus, heuristics are not

‘better’ methods than classical techniques. The question is rather

when to use which approach [Zanakis and Evans (1981)]. In finance,

heuristics are appropriate [Maringer (2005) gives an introduction

and presents several case studies].

Minimizing downside riskIn this section, we will consider a concrete example: portfolio opti-

mization. Our first aim is to evaluate the precision provided by a

heuristic technique. To do that, we need to compare the in-sample

quality of a solution with its out-of-sample quality. Then, we will

compare several selection criteria for portfolio optimization, and

discuss the robustness of the results.

Required precision — we use a database of several hundred

European stocks to run a backtest for a simple portfolio strategy:

minimize semi-variance, subject to (i) the number of assets in the

portfolio being between 20 and 50, (ii) any weight of an included

asset being between 1% and 5%. We construct a portfolio using

data from the last year, hold the portfolio for three months and

record its performance; then we rebalance. In this manner, we

‘walk forward’ through the data which spans the period from

January 1999 to March 2008 [Details can be found in Gilli and

Schumann (2009)].

The solution to this optimization problem cannot be computed

exactly. We use a heuristic method called Threshold Accepting.

This method, however, only returns stochastic solutions: running

the method twice for the same dataset will lead to different optimal

portfolios. With this method, we face an explicit trade-off between

computing time and precision. So when we let the algorithm search

7

8

9

10

11

12

13

14

0 100 200 300 400 500 600-0,2

0

0,2

0,4

0,6

0,8

0 100 200 300 400 500 600

Figure 3 – Risk and risk-adjusted return. The grey dots give the actual portfolios, the

dark line is a local average.

Ranks Average risk Average risk-adjusted return

all 9.55 (1.67) 0.48 (0.19)

1-50 8.03 (0.04) 0.66 (0.05)

51-100 8.07 (0.05) 0.66 (0.05)

101-150 8.09 (0.04) 0.64 (0.05)

151-200 8.17 (0.08) 0.63 (0.07)

201-300 8.51 (0.13) 0.59 (0.08)

301-400 9.16 (0.24) 0.49 (0.11)

401-500 10.94 (0.40) 0.35 (0.09)

501-600 12.49 (0.55) 0.19 (0.11)

Table 2 – Risk and risk-adjusted return (out-of-sample).

121

optimization in financial engineering — an essay on ‘good’ solutions and misplaced exactitude

for longer, then on average we get better solutions. We execute, for

the same dataset, 600 optimization runs with differing numbers of

iterations, and obtain 600 solutions that differ in their in-sample

precision. Higher in-sample precision is associated with more com-

puting time. We rank these portfolios by their in-sample objective

function (i.e., in-sample risk) so that rank 1 is the best portfolio and

rank 600 is the worst.

The left-hand panel of Figure 3 shows the resulting out-of-sample

risk of the portfolios, sorted by in-sample rank. We observe an

encouraging picture: as the in-sample risk goes down, so does the

out-of-sample risk. In other words, increasing the precision of the

in-sample solutions does improve the out-of-sample quality of the

model. At least up to a point: for the best 200 portfolios or so, the

out-of-sample risk is practically the same. So once we have a ‘good’

solution, further improvements are only marginal. We also compute

risk-adjusted returns (Sortino ratios with a required return of zero),

shown in the right-hand panel. It shows a similar pattern, even

though it is much noisier. Table 2 gives details (all annualized). The

numbers in parentheses are the standard deviations of the out-of-

sample results.

This randomness, it must be stressed, follows from the optimization

procedure: in each of our 600 runs we obtained slightly different

portfolios, each portfolio maps into a different out-of-sample per-

formance. For the best portfolios, the improvements are minuscule.

For example, the average risk per year of the best 50 portfolios is 4

basis points lower than the risk of the next-best 50 portfolios.

To judge the relevance of this randomness introduced by our

numerical technique, we need to compare it with the uncertainty

coming from the data. To build intuition, we jackknife from the

out-of-sample paths just as we did in Example 1 above. An example

is illustrated in Figure 4. In the upper panel, we picture the out-

of-sample performance of one euro that was invested in the best

portfolio (rank-1; this path corresponds to the left-most grey dot in

Figure 3). In the lower panel, we see several paths computed after

having randomly-selected and deleted one percent of the data

points. The average risk of the best 50 portfolios in our tests was

8.03% per year, with a standard deviation of 4 basis points. With

jackknifing one percent of the data, we obtain risk between 7.75%

and 8.17%; with jackknifing five percent, we get a risk between

7.58 and 8.29, far greater than the randomness introduced by our

method. For risk-adjusted return – in which we are naturally more

interested – things are even worse. The best 50 portfolios had an

average risk-return ratio of 0.66. Just jackknifing the paths of these

portfolios by one percent, we already get a range of between 0.42

and 0.89.

In sum, heuristics may introduce approximation error into our

analysis, but it is swamped by the sensitivity of the problem with

respect to even slight data changes. Hence, objecting to heuristics

because they do not provide exact solutions is not a valid argument

in finance.

Robustness checks — we run backtests for a large number of

alternative selection criteria, among them partial moments (i.e.,

semi-variance), conditional moments (i.e., Expected Shortfall), or

quantiles (i.e., Value-at-Risk). In the study described above, we only

investigated the approximation errors of the optimization method,

compared with the errors coming from the data. But now we wish

to compare different models.

We implement a backtest like the one described above, but we also

add a robustness check, again based on a jackknifing of the data:

suppose a small number of in-sample observations were randomly

selected and deleted (we delete 10% of the data). The data has

changed, and hence the composition of the computed portfolio

will change. If the portfolio selection strategy is robust, we should

expect the resulting portfolio to be similar to the original one, as

the change in the historical data is only small, and we would also

expect the new portfolio to exhibit a similar out-of-sample perfor-

mance. Repeating this procedure many times, we obtain a sampling

distribution of portfolio weights, and consequently also a sampling

distribution of out-of-sample performance. We do not compare

the differences in the portfolio weights since it is difficult to judge

what a given norm of the difference between two weight-vectors

practically means. Rather, we look at the changes in out-of-sample

results. This means that for any computed quantity that we are

interested in we have a distribution of outcomes. Figure 5 gives

some examples for different strategies.

0.8

1

1.2

1.4

1.6

1.8

2

Jan00 May01 Oct02 Feb04 Jul05 Nov06 Apr08

Jan00 May01 Oct02 Feb04 Jul05 Nov06 Apr08

0.8

1

1.2

1.4

1.6

1.8

2

Figure 4 – Upper panel: out-of-sample performance of one euro invested in the rank-1

portfolio. Lower panel: out-of-sample performance after jackknifing.

122 – The journal of financial transformation

optimization in financial engineering — an essay on ‘good’ solutions and misplaced exactitude

(More details can be found in Gilli and Schumann, forthcoming.) The

figure shows the out-of-sample returns of three strategies: mini-

mum variance, the upside-potential ratio [Sortino et al. (1999)], and

Value-at-Risk; we also plot Sharpe ratios. Portfolios constructed

with the upside-potential ratio, for instance, have a median return

that is more than a percentage point higher than the return of the

minimum variance portfolio; Sharpe ratios are also higher. Even

VaR seems better than its reputation. But most remarkable is the

range of outcomes: given a 10% perturbation of in-sample data,

returns differ by more than 5 percentage points per year.

conclusionIn this article, we have discussed the precision with which financial

models are handled, in particular optimization models. We have

argued that precision is only required to a level that is justified by

the overall accuracy of the model. Hence, the required precision

should be specifically analyzed, so that the usefulness and limita-

tions of a model can be better appreciated. Our discussion may

appear trivial; everyone knows that financial markets are noisy,

and that models are not perfect. Yet the question of the appropri-

ate precision of models with regard to their empirical application is

rarely discussed explicitly. In particular, it is rarely discussed in uni-

versity courses on financial economics and financial engineering.

Again, some may argue, the errors are understood implicitly (just

like ‘500 meters’ means ‘between 400 and 600 meters’), or that in

any case more precision does no harm; but here we disagree. We

seem to have a built-in incapacity to intuitively appreciate random-

ness and chance, hence we strive for ever more precise answers. All

too easily then, precision is confused with accuracy; acting on the

former instead of the latter may lead to painful consequences.

ReferencesAcker, D., and N. W. Duck, 2007, “Reference-day risk and the use of monthly returns •

data,” Journal of Accounting, Auditing and Finance, 22, 527-557

Dimitrov, V., and S. Govindaraj, 2007, “Reference-day risk: observations and exten-•

sions,” Journal of Accounting, Auditing and Finance, 22, 559-572

Gigerenzer, G., 2004, “Fast and frugal heuristics: the tools of bounded rationality,” in •

Koehler, D. J. and N. Harvey (eds), Blackwell handbook of judgment and decision mak-

ing, Blackwell Publishing

Gigerenzer, G., 2008, “Why heuristics work,” Perspectives on Psychological Science, •

3, 20-29

Gilli, M., and E. Schumann, 2009, “Optimal enough?” COMISEF Working Paper Series, •

No. 10

Gilli, M., and E. Schumann, forthcoming, “Risk-reward optimisation for long-run inves-•

tors: an empirical analysis,” European Actuarial Journal

Kamal, M., and E. Derman, 1999, “When you cannot hedge continuously: the corrections •

of Black-Scholes,” Risk, 12, 82-85

Maringer, D., 2005, Portfolio management with heuristic optimization, Springer•

Markowitz, H. M., 1959, Portfolio selection, Wiley•

Michalewicz, Z., and D. B. Fogel, 2004, How to solve it: modern heuristics, Springer, •

2nd edition

Morgenstern, O., 1963, On the accuracy of economic observations, Princeton University •

Press, 2nd edition

Pearl, J., 1984, Heuristics, Addison-Wesley•

Pólya, G., 1957, How to solve it, Princeton University Press, 2nd edition (expanded •

reprint, 2004)

Sortino, F., R. van der Meer, and A. Plantinga, 1999, “The Dutch triangle,” Journal of •

Portfolio Management, 26, 50-58

von Neumann, J., and H. H. Goldstine, 1947, “Numerical inverting of matrices of high •

order,” Bulletin of the American Mathematical Society, 53, 1021-1099

Zanakis, S. H., and J. R. Evans, 1981, “Heuristic ‘optimization’: why, when, and how to •

use it,” Interfaces, 11, 84-91

10 11 12 13 14 15 16 17 18 19 200

0.5

1

Annualised return in %

Minimum variance

Upside potential ratio

Value-at-Risk

0.5 0.55 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 10

0.5

1

Annualised Sharpe ratios

Minimum variance

Upside potential ratio

Value-at-Risk

Figure 5 – Annualized returns and Sharpe ratios for different portfolio selection

strategies.

Articles

123

Rodney colemanDepartment of Mathematics,

Imperial College London

A VaR too far? The pricing of operational risk

AbstractThis paper is a commentary on current and emerging statistical

practices for analyzing operational risk losses according to the

Advanced Measurement Approaches of Basel II, the New Basel

Accord. In particular, the limitations of the ability to model opera-

tional risk loss data to obtain high severity quantiles when the

sample sizes are small and exposed. The viewpoint is that of a

mathematical statistician.

124 – The journal of financial transformation

A VaR too far? The pricing of operational risk

Value-at-Risk (VaR) entered the financial lexicon as a measure of

volatility matched to riskiness. Basel I saw that it could be a risk-

sensitive pricing mechanism for computing regulatory capital for

market and credit risks. Basel II extended its scope to operational

risk, the business risk of loss resulting from inadequate or failed

internal processes, people, or systems, or from external events. The

European Parliament agreed and has passed legislation ensuring

that internationally active banks and insurers throughout the E.U.

will have adopted its provisions from 2012.

However, to put the problem of applying the Basel II risk-sensitive

advanced measurement approaches [BCBS (2006)] in perspective,

consider the role of quantitative risk modeling in the recent global

financial crisis. This began with the credit crunch in August 2007

following problems created by subprime mortgage lending in the

U.S. A bubble of cheap borrowing was allowed to develop, with debt

securitized (i.e. off-loaded) and insured against default, creating a

vicious spiral. In their underwriting, the insurers failed to adjust for

the risk that property values might fall, with consequent defaults on

mortgage repayment. This bad business decision caused havoc at

banks who lent the money, and with investors who bought the debt,

and insurers without the means to pay out on the defaults. Where

was VaR when it was needed? That is to say, what part did financial

modeling play in preventing all this? The answer has to be nothing

much. Further, it can only marginally be put down to operational

risk, since operational risk excludes risks which result in losses from

poor business decisions.

Oprisk losses will often stem from weak management and from

external events. Basel II, the New Basel Accord, set out a regulatory

framework for operational risk at internationally active banks. In so

doing, it gave a generally accepted definition for operational risk,

previously categorized as ‘other risks.’ More importantly, it is Basel

II’s role in driving developments in operational risk management

practices, the search for robust risk measurement, and the pros-

pect of being subject to regulatory supervision and transparency

through disclosure (Pillars 2 and 3 of the Accord) that have led in a

short time to operational risk becoming a significant topic in busi-

ness and management studies.

Basel ii The Accord sets out a risk sensitive way of calculating reserve capi-

tal to cover possible defaults. Institutions are required to categorize

operational risk losses by event type, promoting identification of

risk drivers. There is no mandated methodology.

Pillar 1 of Basel II gives three ways of calculating the operational risk

capital charge, with increasing complexity, but benefiting from a

reducing charge. We shall be considering its requirements in respect

of its highest level, the Advanced Measurement Approaches (AMA),

which requires that the banks model loss distributions of cells over

a business line/loss event type grid using operational risk loss data

that they themselves have collected, supplemented as required by

data from external sources.

Pillar 2 of the Accord requires banks to demonstrate that their man-

agement and supervisory systems are satisfactory. Pillar 3 relates

to transparency, requiring them to report on their operational risk

management. It is these two latter pillars that are probably going

to have a greater impact in protecting global finance than loss

modeling.

Solvency II, the European Union’s regulatory directive for insurers,

has adopted the same three pillars. This directive will come into

force throughout the E.U. in 2012.

In November 2007, the U.S. banking agencies approved the U.S.

Final Rule for Basel II.

Banks will be grouped into the large or internationally active banks

that will be required to adopt AMA, those that voluntarily opt-in to

it, and the rest who will adopt an extended version of the earlier

Basel I.

We note that in the rule book, summarized in BCBS (2006), opera-

tional risk sits in paragraphs 644 to 679, occupying the final 12

pages, pages 140 to 151.

Advanced measurement approaches Attention is directed to the following passages taken from BCBS

(2006) that cover the modeling requirements for the advanced

measurement approaches.

(665) — A bank’s internal measurement system must reasonably

estimate unexpected losses based on the combined use of internal

and relevant external loss data, scenario analysis, and bank-specific

business environment and internal control factors (BEICF).

(667) — The Committee is not specifying the approach or distribu-

tional assumptions used to generate the operational risk measure

for regulatory capital purposes. However, a bank must be able to

demonstrate that its approach captures potentially severe ‘tail’

loss events. Whatever approach is used, a bank must demonstrate

that its operational risk measure meets a soundness standard [...]

comparable to a one-year holding period and a 99.9th percentile

confidence interval.

(669b) — Supervisors will require the bank to calculate its regula-

tory capital requirement as the sum of expected loss (EL) and

unexpected loss (UL), unless the bank can demonstrate that it is

adequately capturing EL in its internal business practices. That is,

to base the minimum regulatory capital requirement on UL alone,

125

A VaR too far? The pricing of operational risk

the bank must be able to demonstrate [...] that it has measured and

accounted for its EL exposure.

(669c) — A bank’s risk measurement system must be sufficiently

`granular’ to capture the major drivers of operational risk affecting

the shape of the tail of the loss estimates.

(669d) — The bank may be permitted to use internally determined

correlations in operational risk losses across individual operational

risk estimates [...]. The bank must validate its correlation assump-

tions.

Event types

Internal fraud

External fraud

Employment practices and workplace safety

Clients, products, and business practice

Damage to physical assets

Business disruption and system failures

Execution, delivery, and process management

Table 1 – Event types for banking and insurance under Basel II and Solvency II

(699f) — There may be cases where estimates of the 99.9th percen-

tile confidence interval based primarily on internal and external loss

event data would be unreliable for business lines with a heavy-tailed

loss distribution and a small number of observed losses. In such

cases, scenario analysis, and business environment and control

factors, may play a more dominant role in the risk measurement

system. Conversely, operational loss event data may play a more

prominent role in the risk measurement system for business lines

where estimates of the 99.9th percentile confidence interval based

primarily on such data are deemed reliable.

(672) — Internally generated operational risk measures used for

regulatory capital purposes must be based on a minimum five-year

observation period of internal loss data.

Business unit Business line

investment banking Corporate finance

Trading and sales

Banking Retail banking

Commercial banking

Payment and settlement

Agency services

others Asset management

Retail brokerage

Table 2 – Business units and business lines for international banking under Basel II

(673) — A bank must have an appropriate de minimus gross loss

threshold for internal loss data collection, for example, €10,000.

(675) — A bank must use scenario analysis of expert opinion in

conjunction with external data to evaluate its exposure to high-

severity events. [...] These expert assessments could be expressed

as parameters of an assumed statistical loss distribution.

Event types and business lines Table 1 shows the seven designated loss event types (ETs) given in

Basel II, also adopted by Solvency II. Table 2 shows the eight broad

business lines (BLs) within the banking sector given in Basel II.

Together they create an 8 by 7 grid of 56 BL/ET cells.

The Operational Risk Consortium Ltd (ORIC) database of opera-

tional risk events, established in 2005 by the Association of British

Insurers (ABI) has nearly 2000 events showing losses exceeding

£10,000 in the years 2000 to 2008. Table 3, based on a report for

ABI [Selvaggi (2009)], gives percentages of loss amounts for those

BL/ET cells having at least 4% of the total loss amount.

Event type

Business line cp&Bs ED&pM BD&sf others Total

Sales and distribution 18.9 6.8 0.7 26.4

Customer service/policy 13.2 2.3 15.5

Accounting/finance 23.4 0.1 23.5

IT 6.0 6.6 12.6

Claims 4.0 1.4 5.4

Underwriting 6.3 0.3 6.6

Others 5.1 11.8 1.4 10.0

Total 24.0 65.5 7.4 11.4 100.0

Table 3 – Business line/event type grid showing percentages of loss amounts (min

4%) in the ORIC database (2000-08)

Source: Selvaggi (2009)

This information tells little about the actual events. For this we need

Level 2 and Level 3 categories, the seven event types being Level

1. To illustrate this, again from Selvaggi (2009), Table 4 shows the

most significant Level 2 and Level 3 event types in terms of both

severity and frequency (values over 4%) from ORIC. This table

excludes losses arising from the widespread mis-selling of endow-

ment policies in the U.K. in the 1990s.The victims were misled with

over-optimistic forecasts that their policies would pay off their

mortgages on maturity. Among the insurers, Abbey Life was fined

a record £1 million in December 2002, and paid £160 million in com-

pensation to 45,000 policy holders. The following March, Royal &

Sun Alliance received a fine of £950,000. Later that year, Friends

Provident was fined £675,000 for mis-handling complaints, also an

operational risk loss.

126 – The journal of financial transformation

A VaR too far? The pricing of operational risk

A widely accepted approach in risk management is to identify the

major risks faced by an organization, measured by their impact in

terms of their frequency and severity. In many cases, not much

more than a handful of operational risks will give rise to the most

serious of the loss events and loss amounts.

However, the day-to-day management requires bottom up as well

as top down risk control. There needs to be an understanding of

the risk in all activities. Aggregating losses over business lines and

activities would tend to conceal low impact risks behind those hav-

ing more dramatic effect. The bottom-up approach is thus a neces-

sary part of seeing the complete risk picture. Aggregation at each

higher level will inform management throughout the business.

Expected loss and unexpected loss An internal operational loss event register would typically show

high impact events at low frequency among events of high fre-

quency but low impact. A financial institution might therefore sort

its losses into: ‘expected loss’ (EL) to be absorbed by net profit,

‘unexpected loss’ (UL) to be covered by risk reserves (so not totally

unexpected), and ‘stress loss’ (SL) requiring core capital or hedging

for cover. The expected loss per transaction can easily be embed-

ded in the transaction pricing. It is the rare but extreme stress loss-

es that the institution must be most concerned with. This structure

is the basis of the loss data analysis approach to operational risk.

Hard decisions need to be made in choosing the EL/UL and UL/SL

boundaries. As well as these, a threshold ‘petty cash’ limit is needed

to set a minimum loss for recording it as an operational loss. Loss

events with recovery and other near miss events will also need to

be entered into the record.

We see that, for regulatory charges, Basel specifies EL to be the

expectation of the fitted loss distribution and UL to be its 99.9th

percentile. For insurers, Solvency II sets UL at the 99.5th percentile.

Basel further permits the use of UL as a single metric for charging.

Let us compare this with the classical VaR, the difference between

UL and EL on a profit-and-loss plot. Expectation measures location,

and VaR measures volatility. Oprisk losses are non-negative, so Basel

appears to be using zero as its location measure, making it possible

to rely on just UL as its opVaR metric. Basel’s referring to ‘confidence

interval’ for the confidence limit UL adds further weight to thinking

it refers to the interval (0, UL). Now UL gives no information about

potential losses larger than itself. In fact, the very high once-in-a-

thousand-years return value is deemed by Basel to capture pretty

much all potential losses. This will not always be the case for heavy-

tailed loss distribution models. In survival analysis and extreme value

theory we would also estimate the expected shortfall, the expecta-

tion of the loss distribution conditioned on the values in excess of UL.

This has been called CVaR (conditional VaR).

There being a minimum threshold to registered losses, we really

need to reconstruct the loss distribution for under-the-threshold

losses to include them in the expectation and 99.9th percentile.

I have always felt that the use of the difference between a moment

measure (expectation) and a quantile measure (percentile) in the

VaR calculation unnecessarily complicates any investigation of its

statistical properties. Why not have median and high percentile?

External data During a recent supervisory visit to an insurance company, the

FSA was told that they had only two operational risk loss events

to show, with another six possibles. This extreme example of the

paucity of internally collected oprisk data, particularly the large

losses that would have a major influence in estimating reserve

funding, means that data publicly available or from commercial or

consortia databases needs to be explored to supplement internal

loss events.

Fitch’s OpVar is a database of publicly reported oprisk events show-

ing nearly 500 losses of more than ten million dollars between

1978 and 2005 in the U.S. The 2004 Loss Data Collection Exercise

(LDCE) collected more than a hundred loss events in the U.S. of 100

million dollars or more in the ten years to 2003.

The Operational Riskdata eXchange Association (ORX) is well estab-

lished as a database of oprisk events in banking. It is a consortium

level 2 level 3 size% freq%

Advisory activities Mis-selling (non-endowment) 13 9

Transaction capture, execution, maintenance

Accounting error 12

Inadequate process documentation

8

Transaction system error 8 6

Management information error 7

Data entry errors 7 5

Management failure 5

Customer service failure 4 16

Suitability, disclosure, fiduciary

Customer complaints 6 4

Systems Software 6

Customer or client account management

Incorrect payment to client/customer

9

Payment to incorrect client/customer

4

Theft and fraud Fraudulent claims 4

Total 76 57

Table 4 – Levels 2 and 3 event categories from insurance losses in ORIC (2000-08)

showing loss amount or loss frequency of 4% or more

Source: Selvaggi (2009)

127

A VaR too far? The pricing of operational risk

collecting data from thirty member banks from twelve countries. It

has more than 44,000 losses, each over €20,000 in value. Apart

from ORIC, individual insurance companies will have their own

claims records containing accurate settlement values.

combining internal and external data When data from an external database is combined with an in-house

dataset, the former are relocated and scaled to match. The exter-

nal data are standardized by subtracting its estimated location

parameter value (i.e., its sample mean m) and dividing the result

by its estimated scale value (i.e., its sample standard deviation s)

from each datum. Then we adopt the location m’ and scale s’ of

the internal data. So, if y is a standardized external datum, the new

value of the external datum is z = m’ + s’y. Finally we check the

threshold used for the internal database against the transformed

threshold for the relocated and rescaled external data, and set the

larger of the two as the new threshold, eliminating all data points

that fall below.

This can lead to strange statistical results. Table 5 shows that sam-

ple statistics for pooled data can lie outside the range given by the

internal and external data sets [Giacometti et al. (2007, 2008)].

We note that the same statistics using logscale data show this phe-

nomenon for commercial banking skewness and kurtosis. Perhaps

the location and scaling metrics are inappropriate, and we need

model specific metrics for these. We are told that the sample size of

the external data is about 6 times that of the internal set.

The small sample problem An extreme loss in a small sample is overrepresentative of its 1 in a

1000 or 1 in a 10,000 chance, yet under-represented if not observed.

Indeed we find the largest losses are overly influential in the fitted

model. So, we must conclude that fitting a small dataset cannot

truly represent the loss process, whatever the model used.

As a statistician I am uncomfortable with external data. An alterna-

tive device which I feel to be more respectable is to fit a heavy-

tailed distribution to the data and then simulate a large number of

values from this fitted distribution, large enough to catch sufficient

high values for statistical analysis. A basic principle in statistics

is that inferences about models in regions far outside the range

of the available data must be treated as suspect. Yet here we are

expected to do just that when estimating high quantiles.

stress and scenario testing Scenario analysis has as its object to foresee and consider respons-

es to severe oprisk events. However, merging scenario data with

actual loss data will corrupt that data and distort the loss distribu-

tion. These scenario losses would be those contributing to a higher

capital charge than otherwise would be the case.

Stress testing is about the contingency planning for these adverse

events based on the knowledge of business experts.

A potential loss event could arise under three types of scenario:

Expected loss (optimistic scenario) ■■

Unexpected serious case loss (pessimistic scenario) ■■

Unexpected worst case loss (catastrophic scenario) ■■

This mimics the expected loss, unexpected loss, and stress loss of

the actuarial approach.

probability modeling of loss data Loss data is not gaussian. The normal distribution model that backs

so much of statistical inference will not do. Loss data by its nature

has no negative values. There are no profits to be read as negative

losses. Operational losses are always a cost. Further a truncated

Business line sample statistic

internal data

External data

pooled data

Retail banking mean 15,888 37,917 11,021

std deviation 97,668 184,787 59,968

skewness 20.17 26.53 26.60

kurtosis 516 910 953

min 500 5,000 1,230

max 2,965,535 8,547,484 2,965,535

Commercial banking mean 28,682 40,808 19,675

median 2,236 12,080 1,347

std deviation 133,874 653,409 83,253

min 500 5,000 521.54

max 1,206,330 20,000,000 2,086,142

Table 5 – Some descriptive statistics for pooled data [abstracted from Giacometti et

al. (2007)]

Figure 1

0.0 200 400 600 800 1000

0.005

0.004

0.003

0.002

0.001

0.0

128 – The journal of fi nancial transformation

A VaR too far? The pricing of operational risk

normal distribution taking only positive values gives too little prob-

ability to its tail, making insuffi cient allowance for large and very

large losses. The lognormal distribution has been used instead

historically in econometrics theory, and the Weibull in reliability

modeling. In practice, even the lognormal will fail to pick up on the

extremely large losses.

Two models that can allow large observations come from Extreme

Value Theory. These are the Generalized Extreme Value distribu-

tion (GEV) and the Generalized Pareto Distribution (GPD). They are

limit distributions as sample sizes increase to infi nity. These distri-

butions are used in environmental studies (hydrology, pollution, sea

defenses, etc.) as well as in fi nance. The GEV and GPD each have

three parameters, μ giving location, σ scale, and x shape, which we

vary to obtain a good fi t. The location μ and shape x are not to be

identifi ed with the population mean and population variance. For

the GPD μ is the lower bound of the range. Figure 1 shows the form

of their respective probability density functions. The lognormal and

Weibull have just two parameters, and so lack fl exibility. Four and

more parameter models, such as Tukey’s g-and-h class of distribu-

tions, are also gaining users, but require more data than is usually

available, though they have been seen to capture the loss distribu-

tion of aggregated fi rm-wide losses. An extensive list with plots and

properties can be found in Young and Coleman (2009).

fitting severity models We fi t the GEV and GPD to the 75 losses given in Cruz (2002,

p.83). For each model we use two fi tting processes: maximum

likelihood for all three parameters and maximum likelihood for the

location and scale, but the Hill estimator for the shape parameter

[Hill (1975)]. Figure 2 shows the sample cumulative distribution

function (the observed proportion of values less than x) shown as

steps, together with four fi tted cumulative distribution functions

(the height y is the probability of obtaining a future value less than

x). The range of observation is (143, 3822). From Figure 2 we can

see a good fi t in each case. In Table 6 what we also see is that the

estimated losses at large quantiles (reading x-values from fi tted

y-values) differ greatly between the fi tted models, that is to say, at

values way beyond the largest observation.

Estimation far outside a dataset is always fraught and can lead

to signifi cant errors in high quantile estimation. Basel II asks for

the 99.9 percentile, Q(0.999), Solvency II for the 99.5 percentile,

Q(0.995). These quantiles are the opVaR.

A simulation study of GPD (0.70, 150, 125) gave an estimated 95%

confi dence interval for Q(0.999) of (5200, 9990), very wide indeed.

The computations were made using Academic Xtremes [Reiss and

Thomas (2007)]. The statistical operations were carried out without

seeking any great precision. The data in thousands of dollars were

rounded to the nearest thousand dollars, the parameters of the

fi tted models are rounded to two signifi cant digits, the fi ts were

Figure 2

0.0 1000 2000 3000 4000

1

0.8

0.6

0.4

0.2

0.0

fitted model GEV GEV GpD GpD

parameter estimates

μ 230 230 135 150

σ 130 100 165 125

x 0.53 0.70 0.44 0.70

Quantiles

Q(0.9) 777 793 866 793

Q(0.95) 1230 1169 1425 1161

Q(0.975) 1960 1706 2333 1661

Q(0.99) 3663 2794 4457 2605

Q(0.995) 5906 4046 7258 3619

Q(0.999) 18066 9525 22452 7595

Data Model values

917 1089 1057 1251 1053

1299 1288 1214 1497 1204

1416 1614 1459 1902 1435

2568 2280 1925 2733 1857

3822 4842 3470 5929 3160

Table 6 – The parameters, quantiles, and fi tted values of GEV and GPD models when

fi tted to loss data

Source: Young and Coleman (2009, pp. 400-403)

μ σ xGEV 230 130 0.53

Simulation 1 234 135 0.56

Simulation 2 224 120 0.49

Simulation 3 221 119 0.45

Simulation 4 230 131 0.51

Average 227 126 0.50

Table 7 – Parameter estimates of GEV (x, μ, σ) from four simulations of 1000 values

from GEV (0.53, 230, 130)

129

A VaR too far? The pricing of operational risk

judged by eye, the simulation for the estimated confidence interval

was based on a simulation of only 4000 values. My first big surprise

was that a good fit could be achieved, secondly that it could be

achieved so easily, my third that we could have four close fits, and

with such a variety of parameter values. Table 7 shows parameter

estimation variability through four simulations of 1000 values from

GEV (0.53, 230, 130). We see that 1000 values can be a small sample

in that it may not be enough to provide precise estimation.

Why such lack of concern for precision? Highly sophisticated meth-

ods can do no better when we have such variability in the resulting

output at high quantiles far beyond the data. We see this variability

right away in the estimated parameter values. The fit was judged

by eye. Why were goodness-of-fit tests not used? They are by their

nature conservative, requiring strong evidence for rejecting a fit

(evidence not available here).

Modeling body and tail separately With sufficient data we may see that the tail data needs to be mod-

eled separately from the main body. We might consider fitting a

GEV to the body and a GPD to the tail, reflecting the Extreme Value

Theory properties. Three parameters each and one for the loca-

tion of the join between the two parts makes seven, but having the

two probability densities meeting smoothly provides two relations

between them, and using the same shape parameter brings the

problem down to four unknowns [Giacometti et al. (2007), Young

and Coleman (2009, p. 397)].

Modeling frequency Standard statistical methods can be used to fit Poisson or negative

binomial probability distributions to frequency counts. Experience

shows that the daily, weekly, or monthly frequencies of loss events

tend to occur in a more irregular pattern than can be fitted by

either of these models.

Basel asks that we combine the fitted frequency model with the fit-

ted severity model to obtain a joint model. This loses the correlation

structure from the loss events.

Some topics left out:

Basel asks for correlation analysis: a big problem with small data-■■

sets. Mathematical finance has provided us with a correlation

theory based on copulas, but not useful here.

Validation techniques, such as the resampling methods of the ■■

jackknife and bootstrap [Efron and Tibshirani (1993)] can be

used to obtain sampling properties of the estimates such as

confidence intervals.

Bayes hierarchical modeling [Medova (2000), Kyriacou and ■■

Medova (2000), Coles and Powell (1996)] treats non-stationarity

by letting the GPD parameters be themselves random from dis-

tributions with parameters (hyper-parameters).

Dynamic financial analysis refers to enterprise-wide integrated ■■

financial risk management. It involves (mathematical) modeling

of every business line, with dynamic updating in real time.

Bayes belief networks (BBNs) are acyclic graphs of nodes con-■■

nected by directed links of cause and effect. The nodes are

events, and the states are represented by random variables.

Each node event is conditioned on every path to it. For the link A

to B, the random variable X associated with event A is given by

the multi-on-multi-variate history leading to A. The BBN requires

starting probabilities for each node, and the calculation for P(B |

A) where A incorporates all links to it. This is a formidable task.

The introspection forces risk assessment for every activity, and,

once the network is up and running, it can be used for stress

testing.

To sum up The point being emphasized is that no methodology on its own can

provide an answer. Multiple approaches should be used, both quali-

tative and quantitative, to aid management in acquiring a sensitivity

to data and its interpretation and its use in decision making.

References Basel Committee on Banking Supervision, 2006, “International convergence of capital •

measurement and capital standards,” Bank for International Settlements

Coles S., and E. Powell, 1996, “Bayesian methods in extreme value modelling: A review •

and new developments, International Statistical Review, 64, 119-136

Cruz, M.G., 2002, Modeling, measuring and hedging operational risk, Wiley •

Efron, B., and R. Tibshirani, 1993, An introduction to the bootstrap, Chapman & Hall •

Embrechts P., (editor), 2000, Extremes and integrated risk management, Risk Books •

Giacometti, R., S. Rachev, A. Chernobai, and M. Bertocchi, 2008, “Aggregation issues in •

operational risk,” The Journal of Operational Risk, 3:3, 3-23

Giacometti, R., S. Rachev, A. Chernobai, M. Bertocchi, and G. Consigli, 2007, “Heavy-•

tailed distributional model for operational risk,” The Journal of Operational Risk, 3:3,

3-23

Hill, B.M., 1975, “A simple general approach to inference about the tail of a distribu-•

tion,” Annals of Statistics, 3, 1163-1174

Kyriacou, M.N., and E.A. Medova, 2000, “Extreme values and the measurement of •

operational risk II,” Operational Risk, 1:8, 12-15

Lloyd’s of London, 2009, “ICA: 2009 Minimum Standards and Guidance,”, Lloyd’s •

Medova, E.A., 2000, “Extreme values and the measurement of operational risk I”, •

Operational Risk, 1:7, 13-17

Reason, J., 1997, Managing the risk of organisational accidents, Ashgate •

Reiss, R.-D., and M. Thomas, 2007, Statistical analysis of extreme values (3rd edition), •

Birkhauser

Selvaggi, M., 2009, “Analysing operational risk in insurance,” ABI Research Paper 16. •

Young, B., and R. Coleman, 2009, Operational risk assessment, Wiley •

Articles

131The views expressed are personal ones and do not represent the organisations with 1

which the author is affiliated. All errors are mine.

Hans J. Blommestein1

PwC Professor of Finance, Tilburg University and Head of Bond Markets and Public Debt, OECD

Risk management after the Great crash

AbstractThis study takes a closer look at the role of risk (mis)management

by financial institutions in the emergence of the Great Crash. It is

explained that prior to the crisis too much reliance was placed on

the quantitative side of risk management, while not enough atten-

tion was paid to qualitative risk management. In this context it is

argued that there is an urgent need for dealing more effectively

with inherent weaknesses related to institutional- and organiza-

tional aspects, governance issues, and incentives. More sophistica-

tion and a further refinement of existing quantitative risk manage-

ment models and techniques is not the most important or effective

response to the uncertainties and risks associated with a fast-mov-

ing financial landscape. In fact, too much faith in a new generation

of complex risk models might lead to even more spectacular risk

management problems than the one we experienced during the

Great Crash. Against this backdrop, the most promising approach

for improving risk management systems is by providing a coherent

framework for addressing systematical weaknesses and problems

that are of a qualitative nature. However, given the inadequate and

very imperfect academic knowledge and tools that are available,

risk management as a scientific discipline is not capable of dealing

adequately with fundamental uncertainties in the financial system,

even if a coherent picture associated with the aforementioned

qualitative issues and problems is being provided. This perspective

is providing an additional motivation to those authorities that are

contemplating to constrain or restructure parts of the architecture

of the new financial landscape.

There are various channels through which an unsound incentive structure can have 2

an adverse structural influence on the pricing of financial assets. For example, Fuller

and Jensen (2002) illustrate (via the experiences of Enron and Nortel) the dangers

of conforming to market pressures for growth that are essentially impossible, leading

to an overvalued stock. Other channels through which perverse incentives are being

transmitted originate from situations where traders and CEOs pursue aggressive risk

strategies, while they are facing largely an upside in their rewards structures (and

hardly a downside). For example, successful traders can make fortunes, while those

that fail simply lose their jobs (in most cases they move on to a trading job in other

financial institutions). CEOs of institutions that suffer massive losses walk away with

132

Risk management after the Great crash

very generous severance and retirement packages. There are also fundamental flaws

in the bonus culture. Costs and benefits associated with risk-taking are not equally

shared and the annual merry-go-round means financial institutions can end up paying

bonuses on trades and other transactions that subsequently prove extremely costly

[Thal Larsen (2008)]. Rajan (2008) points out that bankers’ pay is deeply flawed. He

explains that employees at banks (CEOs, investment managers, traders) generate

jumbo rewards by creating bogus ‘alpha’ by hiding long-tail risks. In a similar spirit,

Moody’s arrives at a very strong conclusion that the financial system suffers from

flawed incentives that encourage excessive risk-taking [Barley (2008)].

The literature on the causes of the global credit/liquidity crisis

(Great Crash for short) is growing exponentially [see, for example,

Senior Supervisors Group (2008), FSA (2009), BIS (2009), among

others]. The many studies and policy reports reveal serious failures

at both the micro- (financial institutions) and macro-levels (financial

system as a whole) and mistakes made by different actors (bankers,

rating agencies, supervisors, monetary authorities, etc.). This paper

(i) takes a closer look at the role of risk (mis)management by finan-

cial institutions in the creation of the Great Crash; and (ii) outlines

the most promising ways for improving risk management, in par-

ticular by paying much more attention to qualitative risk manage-

ment issues. However, it will also be explained why it would still be

impossible to prevent major crises in the future, even if a coherent

picture involving qualitative issues or aspects would be provided.

However, in doing so, we do not exclude the importance of the role

of other (important) factors in the origin of the Great Crash. On the

contrary, mistaken macroeconomic policies as well as deficiencies

in the official oversight of the financial system were also significant

factors in the global financial crisis, especially in light of global imbal-

ances and the emergence of systemic risks and network externali-

ties during the evolution of the crisis. However, most commentators

would agree that serious failures in risk management at the level of

major (and even some medium-sized) financial institutions played an

important role in the origin and dynamics of the Great Crash.

Why did risk management fail at the major financial institutions? The conventional storyline before the Great Crash was that risk

management as a science has made considerable progress in the

past two decades or so and that financial innovations such as risk

transfer techniques had actually made the balance sheet of finan-

cial institutions stronger. However, my analysis of risk management

failures associated with the global financial crisis (and supple-

mented by insights gained from studying earlier crisis episodes

such as the crash of LTCM in 1998) [Blommestein (2000)] uncovers

deep flaws in the effectiveness of risk management tools used by

financial institutions.

The crisis has not only shown that many academic theories were

(are) not well-equipped to properly price the risks of complex

instruments such as CDOs [Blommestein (2008b, 2009)], espe-

cially during market down-turns [Rajan (2009)], but also that

risk management methodologies and strategies based on basic

academic finance insights were not effective or even misleading.

A core reason is that academic risk-management methodologies

are usually developed for a system with ‘well-behaved or governed’

financial institutions and markets that operate within a behavioral

framework with ‘well-behaved’ incentives (that is, risk management

is not hampered by dysfunctional institutions, markets, or financial

instruments and/or undermined by perverse incentives).

In this paper, I will, therefore, not only focus on the quantitative

dimension of risk management systems such as the mispricing of

risks (of complex products) and measurement problems but also on

the qualitative dimension covering such issues like those involving

the above mentioned ‘practical’ obstacles of an institutional, orga-

nizational and incentive nature. In addition to the well-documented

role of the perverse incentives associated with misconstrued com-

pensation schemes, it will be argued that the institutional, or orga-

nizational, embedding of risk management is of crucial importance

as well. It will be suggested that the qualitative dimension of risk

management is equally (or perhaps even more) important to the

measurement or quantitative side of the process of risk manage-

ment.

This view implies that even if the risk management divisions of

these financial institutions had acted with the longer-term inter-

ests of all stakeholders in mind (which many of them did not), they

would still have had an uphill battle in effectively managing the risks

within their firms because of (a) the inadequate tools that were at

their disposal (based, crucially, on insights from academic finance);

(b) the complex organizational architecture, and sometimes dys-

functional institutional environment, in which many risk managers

had (have) to operate; and (c) excessive risk-taking associated

with business strategies that incorporated perverse compensa-

tion schemes. In other words, had the risk management divisions

of these institutions effectively implemented the state-of-the-art

tools that were provided to them by academic finance (includ-

ing, crucially, reliable information about the riskiness of complex

financial instruments such as structured products), they would still

have to struggle with the complex institutional or organizational

embedding of risk management situations as well as the various

channels through which an unsound incentive structure (operating

between financial institutions and the market) can have an adverse

structural impact on the pricing of financial assets2.

This perspective can then also be used to explain why it is very

difficult or even impossible to effectively manage the totality of

complex risks faced by international banks and other financial insti-

tutions. More specifically, it is the reason why effective enterprise-

wide risk management is an extremely hard objective, especially in

133

Risk management after the Great crash

a rapidly-changing environment. This echoes a conclusion from a

2005 paper on this topic: “But successful implementation of ERM

is not easy. For example, a recent survey by the Conference Board,

a business research organization, shows that only 11 per cent of

companies have completed the implementation of ERM processes,

while more than 90 per cent are building or want to build such a

framework” [Blommestein (2005b)].

For all these reasons we have to have a realistic attitude towards

the practical capacity of risk management systems, even advanced

ones based on the latest quantitative risk measures developed in

the academic literature. An additional reason to take a very modest

view on real abilities of available risk control technologies is the fact

that academically developed risk measures predominantly deal with

market-and-credit risk. Academic finance has much less to say about

the analytical basis for liquidity risk, operational risk, and systemic

risk. Unfortunately, the latter types of risks played major roles in the

origin of the Great Crash [Blommestein (2008a)], showing that they

were not adequately diagnosed, managed, and/or supervised.

Against this backdrop, we will, first, cast a critical eye at what has

been suggested to be the principal cause of the recent financial cri-

sis: the systemic mispricing of risks and the related degeneration of

the risk management discipline into a pseudo quantitative science.

After that we will show that the disappearance or weakening of due

diligence by banks in the securitization process was an important

crisis-factor that was not detected by conventional, quantitative

risk management systems. In fact, it is another important example

why individual financial institutions can fail, and complete systems

collapse, when not enough attention is being paid to the qualitative

dimension of risk management [Blommestein et al. (2009)].

How important was the mispricing of risks? An increasingly interconnected and complex financial system made

it harder to price risks correctly. Both market participants and

supervisors underestimated the increase in systemic risk. Early on

I concluded in this context: “The sheer complexity of derivatives

instruments, coupled with consolidation in the financial industry,

has made it increasingly hard for regulators and bankers to assess

levels of risk. In the credit derivatives market, risks that have been

noted include a significant decline in corporate credit quality, little

information on counterparties, operational weaknesses that may

result from the novelty of these instruments, and a disincentive to

manage actively portfolio credit risk. As a result, systemic risk in

this complex, often opaque financial landscape is likely to be higher

than before” [Blommestein (2005b)].

Although risk managers had more rigorous risk management tools

at their disposal than in the past, the rapidly changing financial

landscape (characterized by more complex products and markets,

a higher level of systemic risk, and increased financial fragility)

weakened the applicability and conditions under which these quan-

titative tools and techniques can be used effectively. Many market

participants (including sophisticated ones) had difficulties in under-

standing the nature and pricing of new products and markets, due

to the sheer complexity of many new financial instruments and the

underlying links in the new financial landscape. In a 2005 study

I noted: “Even sophisticated market participants might, at times,

have difficulties understanding the nature of these new products

and markets. Consequently, risks may be seriously mispriced. In

the market for collateralized debt obligations (CDOs), the high pace

of product development requires the rapid adaptation of pricing

machines and investment strategies. Although the ability to value

risky assets has generally increased, concerns have been raised

about the complex risks in this fast-growing, market segment and,

more to the point, whether investors really understand what they

are buying” [Blommestein (2005b)].

Moreover, as explained above, securitization was adversely affected

by problems with incentives and information as well as the pricing

of tail events [Cechetti (2009)]. More generally, a number of widely

held assumptions proved to be wrong and costly, in particular that

the originate-and-distribute (securitize) model would decrease

residual risk on the balance sheet of banks, that the growing use

of credit risk transfer instruments would result in better allocated

risks and a more stable banking sector, and that ample market

liquidity would always be available.

The outbreak of the financial crisis proved these assumptions

wrong, whereby the dark side of the risk paradox became visible

[Blommestein (2008c)]. In effect, it became increasingly clear that

there had been a significant and widespread underestimation of

risks across financial markets, financial institutions, and countries

[Trichet (2009)]. For a variety of reasons, market participants did

not accurately measure the risk inherent in financial innovations

and/or understand the impact of financial innovations on the over-

all liquidity and stability of the financial system. Indeed, there is

growing evidence that some categories or types of risks associated

with financial innovations were not internalized by markets; for

example, tail risks were underpriced and systematic risk (as exter-

nality) was not priced, or was priced inadequately. This widespread

and systemic underestimation of risks turned out to be at the core

of the financial crisis. At the same time, the availability of sophisti-

cated quantitative risk tools created a false sense of security and

induced people to take greater risks. “Professional enthusiasm

about new risk control technology may give rise to overconfidence

and even hubris” [Helwig (2009)].

The underestimation of risks reflected to an important degree mis-

takes in both the strategic use of risk management systems as well

as the technically inadequate risk management tools. During the

unfolding of the crisis, many financial institutions revealed a huge

Let us focus on operational risk (op risk) in a large bank as an example. Shojai and 3

Feiger(2010) note that many institutions have only recently started to come to grips

with the fact that operational risk is a major risk fraught with obstacles. First, quantifi-

cation is a very challenging task. Second, once op risk has been measured properly, we

need to be able to compute the operational risks of each division. Third, how do firms

aggregate op risk across organisations as a whole? Many larger institutions still segre-

gate their different businesses (perhaps for good reasons). Hence it is nearly impossible

(a) to quantify operational risk for the group and (b) to determine what diversification

benefits could be derived. Moreover, in some banks, FX, credit, and equities each have

their own quants teams, whose aggregate risks for the bank no one can really under-

stand [or even for the financial system as a whole [Patterson (2010)].

See Patterson (2010) for a non-technical account of the role of Process Driven 4

Trading (PDT) during the crisis. The formulas and complicated models of quants

traded huge quantities of securities and as the housing market began to crash, the

models collapsed. In the words of Patterson (2010): “The result was a catastrophic

domino effect. The rapid selling scrambled the models that quants used to buy and

sell stocks, forcing them to unload their own holdings. By early August, the selling

134

Risk management after the Great crash

had taken on a life of its own, leading to billions in losses. The meltdown also revealed

dangerous links in the financial system few had previously realized — that losses in

the U.S. housing market could trigger losses in huge stock portfolios that had nothing

to do with housing. It was utter chaos driven by pure fear. Nothing like it had ever

been seen before. This wasn’t supposed to happen!”

Some academic economists were certainly aware of the limitations and weaknesses 5

of these models for use in the financial sector. For example, Merton (1994) gave the

following general warning: “The mathematics of hedging models are precise, but the

models are not, being only approximations to the complex, real world. Their accuracy

as a useful approximation to that world varies considerably across time and place.

The practitioner should therefore apply the models only tentatively, assessing their

limitations carefully in each application”. However, other academics and many users

of academic models in the financial industry were often ill-informed and ignorant

about the deeper weaknesses of using these kinds of models across time and differ-

ent market places [Blommestein (2009)].

Ironically enough, the October 1987 crash marked the birth of VAR as a key risk man-6

agement tool. For a very brief history of the birth of VAR, see Haldane (2009).

concentration of risks, suggesting that risk management systems

failed (a) to identify key sources of risks, (b) to assess how much

risk was accumulated, and (c) to price financial risks properly (or to

use reliable market prices, in particular for structured products). The

underlying problem was that risk management did not keep pace with

the risks and uncertainty inherent in financial innovations and the

fast-changing financial landscape. Risk managers placed too much

trust into the existing risk models and techniques (see below), while

underlying assumptions were not critically evaluated. Unfortunately,

the use of these models proved to be inadequate, both from a techni-

cal and a conceptual point of view. On top of this, risk management

fell short from a qualitative perspective; that is, too little attention

was paid to corporate governance processes, the architecture and

culture of organizations, business ethics, incentives, and people.

fatal flaws in the origination and securitization process: failures in quantitative — and qualitative risk managementThe (impact of the) securitization of mortgages and financial

innovations such as CDO and CDS markets came under heavy

criticism as being an important cause of the global financial crisis

[Blommestein (2008a); Tucker (2010); Jacobs (2009)]. Risks were

significantly underpriced [Blommestein (2008b)] (in particular by

rating agencies) while risk management systems failed.

Naturally, also regulators and central bankers made mistakes. Most

studies, however, seem to suggest that had we gotten a better

understanding of the correct pricing of complex structured products

such as CDOs, CLOs and CDSs over the cycle, then we might have

been able to prevent the seriousness of the global financial crisis. For

example, popular CDO pricing models such as the Gaussian copula

function are based on the dubious key assumption that correlations

are constant over the cycle. The above reasoning implies that had

we been able to employ a far superior method (in terms of accuracy

and/or robustness) than the Gaussian copula function, then we would

have valued more accurately structured products over the cycle. As

a result, the Great Crash would not have occurred.

The limits of pricing models and quantitative risk management

However, the conclusion that better quantifications would have

prevented a major crisis can be challenged on the following three

key grounds. First, as noted above, pricing in the fast-moving,

complex financial landscape is a huge challenge. Pricing models

are therefore subjected to significant model risk. For example, the

foundation of the pricing of risk in structured products such as

CDOs and CDSs is based on the key theoretical notion of perfect

replication. Naturally, perfect replication does not exist in reality

and has to be approximated by historical data which in many cases

is very incomplete and of poor quality. Instead, researches and

practitioners had to rely on simulation-based pricing machines. The

input for these simulations was very shaky as they were based on

“relatively arbitrary assumptions on correlations between risks and

default probabilities” [Colander et al. (2009)].

Second, many institutions have major difficulties in quantifying

the ‘regular’ risks associated with credit and market instruments.

However, the Great Crash demonstrated that this was even more

so the case for firms’ operational and liquidity risks. These compli-

cations multiply when one tries to aggregate the various risks of

divisions or departments within larger financial institutions3.

Third, from a more conceptual perspective, the financial crisis

brought to light that the risk management discipline had developed

too much into a pseudo quantitative science with pretensions beyond

its real risk management capabilities4. The over-reliance on sophis-

ticated though inadequate risk management models and techniques

contributed to a false sense of security [Honohan (2008)]. Indeed,

many professionals were too confident in the ability of quantitative

models to reliably measure correlations and default probabilities

[Helwig and Staub (1996)]. It was assumed that quantitative risk

management models represented stable and reliable stochastic

descriptions of reality. Ironically, by relying to an increasing degree

on sophisticated mathematical models and techniques, the risk man-

agement discipline lost its ability to deal with the fundamental role

of uncertainty in the financial system5. In addition to this fundamen-

tal methodological problem, the financial crisis revealed technical

failures in risk management in the sense that even sophisticated

methods and techniques turned out not to be refined enough. At

the core of many risk management systems was (is) the concept of

Value-At-Risk (VAR), which became a key tool in the quantification of

risk, the evaluation of risk/return tradeoffs, and in the disclosure of

risk appetite to regulators and shareholders6. This concept is effec-

tively based on the idea that the analysis of past price movement

patterns could deliver statistically robust inferences relating to the

135

Risk management after the Great crash

A clear example is the trading strategy used by a number of Citigroup employees. 9

On 2 August 2004, Citigroup pushed through €11 billion in paper sales in two minutes

over the automated MTS platform, throwing the market into confusion. As the value

of futures contracts fell and traders moved to cover their positions, Citigroup reen-

tered the market and bought back about €4 billion of the paper at cheaper prices.

The strategy was dubbed Dr Evil, in an internal e-mail circulated by the traders. In

2007, an Italian Court indicted seven (by that time former) Citigroup traders on

charges of market manipulation in the sale and repurchase of government bonds on

the MTS electronic fixed income network. (Citi bond traders indicted over ‘Dr Evil’

trade, http://www.finextra.com/fullstory.asp?id=17210, 19 July 2007.)

A similar problem would occur when one would focus only on ‘macro’ factors such 7

as global imbalances and low interest policies. Even if interest rates had not been

kept so low for as long as they were or global imbalances would have been smaller,

we would still have not been able to prevent the erosion of financial stability and the

structural weaknesses in the financial sector (in both the banking sector and security

markets). In retrospect, the global crisis of the new financial landscape was an acci-

dent waiting to happen due to (semi-) hidden unsound structural features (see below).

Rajan (2009) notes in this context that ‘….originators could not completely ignore 8

the true quality of borrowers because they were held responsible for initial defaults.”

However, he concludes that even this weak source of discipline was undermined by

steadily rising housing prices.

probability of price movements in the future [FSA (2009)]. However,

the financial crisis revealed severe problems with applying the VAR

concept to the world of complex longer-term social and economic

relationships [Danielsson (2002)].

complex risks and uncertainties and the importance of qualitative risk managementFocusing too much on the technical intricacies of models for the

‘correct’ pricing of these assets, although important, ensures that

we ignore another, often neglected, crucial reason why markets

for securitized assets became so big and fragile and finally col-

lapsed7. In fact, as noted before, the quantitative approach to risk

management does not fully cover the range of important risks and

uncertainties.

First, the insight that the ‘originate- to- securitize’ process (and its

embedded risks) was capable of generating not very well under-

stood negative spillovers via the shadow banking sector to com-

mercial banks [Tucker (2010)].

Second, the originate-to-securitize business model or process had

as (unintended) consequence the fatally weakening of due diligence

undertaken by originators8. With the originators relinquishing their

role as the conductors of due diligence it was left to the credit

rating agencies (CRAs) to fill this information gap. But, on top of

their inadequate pricing methodologies, CRAs never had sufficient

access to the required information about underlying borrowers to

have any idea of their true state of health. That crucial information

was in principle in the hands of the issuing bank, but, as noted, they

had stopped caring about collecting that kind of information when

they started selling the mortgages onto other investors [Keys et al.

(2010)]. So, the rating agencies had to use aggregate data to rate

these instruments, or to rely on the credit quality of the insurer

who had provided credit enhancement to the security [Fabozzi and

Kothari (2007)]. In either case, neither the credit enhancer nor the

rating agency had any idea about the underlying quality of the bor-

rowers [Shojai and Feiger (2010)].

Third, perverse incentives, associated with flawed compensation

structures, are keeping valuations from their ‘true’ (equilibrium)

prices [Blommestein (2008b)]. As a result, excessive risk-taking

(was) is manifesting itself in part through asset bubbles with sig-

nificantly underpriced risks. Moreover, experience and tests show

that humans have an ingrained tendency to underestimate outliers

[Taleb (2007)] and that asset markets have a tendency to gener-

ate a pattern of bubbles (with prices much higher than the intrinsic

value of the asset), followed by crashes (rapid drops in prices)

[Haruvy et al. (2007)].

However, prior to the crisis, these underlying structural problems

did not dampen the demand from institutional investors for AAA

paper. Institutional investors believed that it was possible to

squeeze out a large quantity of paper rated AAA via the slicing and

dicing through repeated securitization of the original package of

assets (mortgages, other loans). Consequently, all that was needed

to expose the underlying weaknesses was a correction in house

prices, which is exactly what happened. Moreover, it became indeed

painfully clear that the complex securities issued by CDOs are very

hard to value, especially when housing prices started to drop and

defaults began to increase.

The key insight of this overview is that modern risk transfer

schemes (that reallocate risk, or, in many cases more accurately

stated, uncertainty) may undermine due diligence (and prudence

more generally), especially in combination with compensation

schemes that encourage excessive risk-taking. This structural lack

of prudential behavior infected not only the structured finance seg-

ments but the entire financial system. All types of players (bankers,

brokers, rating agencies, lawyers, analysts, etc.) were operating

under business plans with a (implicit) short-term horizon that put

institutions and systems at risk. Deeply flawed incentive schemes

encouraged dangerous short-cuts, excessive risk-taking, but also

unethical practices9. The culture of excessive risk-taking and dubi-

ous ethics [Blommestein (2005a)] in the banking industry spread

like a virus during the past decades and became firmly entrenched

[Blommestein (2003)]. Even if the top management of banks aim to

maximize long-term bank value, it may then be extremely hard to

impose incentives and control systems that are consistent with this

objective. In fact, prior to the crisis, there was a move away from

this long-term objective, with an increasing number of bank CEOs

encouraging business strategies based on aggressive risk-taking.

This in turn engendered excess risk-taking and non-ethical prac-

tices within firms and at all levels (traders, managers, CEOs).

Hence, a relatively small and local crisis could transform itself into

the Great Crash.

Clearly, the many complex and new risks and uncertainties in the

fast-moving financial landscape could not be effectively diagnosed

and managed via a purely quantitative approach. In fact, it encour-

aged additional risk-taking induced by a false sense of confidence in

sophisticated risk-control technologies. We, therefore, need a para-

136This point is also emphasized by the CFO of the Dutch KAS Bank. He observes that 10

many risk management models work fine when there is no crisis. However, they can

fail spectacularly during a crisis because of left-out factors. It is important (1) to be

aware which risk factors have been (deliberately) omitted and why and (2) which

actions to take when a crisis erupts [Kooijman (2010)].

Risk management after the Great crash

Of interest is that the conclusions from a 2008 report on risk management practices 11

during the crisis, drafted by a group of 8 financial supervisors from 5 countries,

focused predominantly on organizational and institutional issues [Senior Supervisors

Group (2008)].

digm shift in risk management that also includes an assessment

of uncertainties through the lens of qualitative risk management.

Only in this way would we be able to tackle the adverse influences

of organizational issues, human behavior, and incentives schemes

[Blommestein (2009)]. This would also allow us to account for the

fact that all risk measurements systems are far more subjective

than many experts want to accept or admit. Empirical risk control

systems are the result of subjective decisions about what should be

incorporated into the risk model and what should not10.

conclusionsThe first key conclusion is that prior to the Great Crash too much

reliance was placed on the quantitative side of risk management

and too little on the qualitative dimension. In this context it was

argued that there was an urgent need for dealing effectively with

inherent weaknesses related to institutional, organizational, gover-

nance, and incentives’ aspects11. More sophistication and a further

refinement of existing quantitative risk management models and

techniques is not the most important or effective response to the

uncertainties and risks associated with a fast-moving financial

landscape. In fact, too much faith in a new generation of complex

risk models might even lead to more spectacular risk management

problems as the one witnessed during the last decade. Instead, as

noted by Blommestein et al. (2009), a more holistic and broader

approach to risk management is needed as part of a paradigm shift

where more attention is given to the qualitative dimension of risk

management.

A final key finding is that it would still be impossible to prevent

major crises in the future, even if a coherent picture associated with

the aforementioned qualitative issues is being provided. The under-

lying epistemological reason is the (by definition) imperfect state

of academic knowledge about new uncertainties and risks associ-

ated with a fast-moving society, on the one hand, and the inher-

ently inadequate risk-management responses that are available as

tools to risk managers, their top management and, indeed, also to

their supervisors, on the other. Indeed, we have shown in a related

analysis that the Great Crash is another illustration of the fact that

risk management as a scientific discipline is not capable of dealing

adequately with fundamental uncertainties in the financial system

[Blommestein et al. (2009)]. From this perspective it is therefore

no surprise that some authorities are considering to constrain or

restructure parts of the architecture of the new financial landscape

[see Annex below and Group of Thirty (2009)].

Annex: structural weaknesses waiting to erupt in the new financial landscapeTucker (2010) recently analyzed the question of whether the struc-

ture or fundamental architecture of the new financial landscape

needs to be constrained or restructured by the authorities. In doing

so he focused on a key weakness in the new financial landscape: the

shadow banking sector.

In Tucker’s analysis, the dangerous side of ‘shadow banking’ refers

to “those instruments, structures, firms or markets which, alone or

in combination, and to a greater or lesser extent, replicate the core

features of commercial banks: liquidity services, maturity mismatch

and leverage.” The un(der)regulated shadow banking activities can

then create an unstable and fragile banking sector. For example,

the money fund industry is a major supplier of short-term funding

to banks, while its own maturity mismatch served to mask the true

liquidity position of the banking sector. This in turn fatally injected

additional fragility into the financial system as a whole. Warnings

were published quite a few years ago [Edwards (1996)], while Paul

Volcker, former chairman of the U.S. Federal Reserve, is reported to

have expressed serious concerns at internal Federal Reserve meet-

ings around thirty years ago [Tucker (2010)].

So, like in the case of the ‘originate- to- securitize’ process, this

was an example of a structural weakness waiting to erupt, although

the wait was longer. But during the global financial crisis they both

became a reality. “When the Reserve Fund “broke the buck” after

Lehman’s failure, there was a run by institutional investors” …

“Echoing Paul Volcker’s concerns, the Bank of England believes

that Constant-NAV money funds should not exist in their current

form” [Group of 30 (2009)].

137

Risk management after the Great crash

ReferencesBarley, R., 2008, “Ability to track risk has shrunk ‘forever’” Moody’s, Reuters.com, 6 •

January

Blommestein, H. J., 2009, The financial crisis as a symbol of the failure of academic •

finance?(A methodological digression), Journal of Financial Transformation, Fall

Blommestein, H. J.,2008a, “Grappling with uncertainty”, The Financial Regulator, March•

Blommestein, H. J., 2008b, “Difficulties in the pricing of risks in a fast-moving financial •

landscape (A methodological perspective), Journal of Financial Transformation, 22

Blommestein, H. J., 2008c, “No going back (derivatives in the long run)”, Risk, August•

Blommestein, H. J. 2006, The future of banking, SUERF Studies.•

Blommestein, H. J., 2005a, How to restore trust in financial markets?” in Dembinski, P. •

H., (ed.), Enron and world finance – a case study in ethics, Palgrave

Blommestein, H. J., 2005b, “Paying the price of innovative thinking,” Financial Times •

(Mastering Risk), 23 September

Blommestein, H. J., 2003, “Business morality audits are needed,” Finance & Common •

Good, No. 15, Summer

Blommestein H. J. 2000, “Challenges to sound risk management in the global financial •

landscape,” Financial Market Trends No 75, April, OECD

Blommestein, H. J., L. H. Hoogduin, and J. J. W. Peeters, 2009, “Uncertainty and risk •

management after the Great Moderation, The role of risk (mis)management by finan-

cial institutions,” Paper presented at the 28th SUERF Colloquium on “The Quest for

Stability”3-4 September 2009, Utrecht, The Netherlands and Forthcoming in the SUERF

Studies Series

Boudoukh, J., M. Richardson, and R. Stanton, 1997, “Pricing mortgage-backed securities •

in a mulitfactor interest rate environment: a multivariate density estimation approach,”

Review of Financial Studies, 10, 405-446

Cechetti, S. G., 2009, “Financial system and macroeconomic resilience,” Opening •

remarks at the 8th BIS Annual Conference, 25-26 June 2009

Colander, D., H. Föllmer, A. Haas, M. Goldberg, K. Juselius, A. Kirman, and T. Lux, 2009, •

“The financial crisis and the systematic failure of academic economics,” Kiel Working

Papers No. 1489, Kiel Institute for the World Economy

Danielsson, J., 2002, “The emperor has no cloths: limits to risk modelling,” Journal of •

Banking and Finance, 26:7, 1273-1296

Edwards, F., 1996, “The new finance: regulation and financial stability,” American •

Enterprise Institute

Fabozzi, F. J., and V. Kothari, 2007, Securitization the tool of financial transformation,” •

Journal of Financial Transformation, 20, 33-45

Financial Services Authority, 2009, “The Turner review: a regulatory response to the •

global banking crisis,” March

Fuller, J., and M. C. Jensen, 2002, “Just say no to Wall Street,” Journal of Applied •

Corporate Finance, 14:4, 41-46

Group of Thirty, 2009, “Financial reform: a framework for financial stability,” January 15•

Haldane, A. G., 2009, “Why banks failed the stress test,” Speech, 13 February•

Haruvy, E., Y. Lahav, and C. N. Noussair, 2007, “Traders’ expectations in asset markets: •

experimental evidence,” American Economic Review, 97:5, 1901-1920

Honohan, P., 2008, “Risk management and the cost of the banking crisis”, IIIS •

Discussion Paper, No. 262

Jacobs, B. I., 2009, “Tumbling tower of Babel: subprime securitization and the credit •

crisis,” Financial Analysts Journal, 66:2, 17 – 31

Jensen. M. C., and W. H. Meckling, 1976, “Theory of the firm: managerial behavior, agen-•

cy costs and ownership structure,” Journal of Financial Economics, 3:4, 305-360

Jorion, P., 2006, Value at Risk: the new benchmark for managing financial risk, 3rd ed. •

McGraw-Hill

Keys, B. J., T. Mukherjee, A. Seru, and V. Vig, 2010, “Did securitization lead to lax •

screening? Evidence from subprime loans,” Quarterly Journal of Economics, forthcom-

ing

Kooijman, R., 2010, “Risk management: a managerial perspective,” Presentation at the •

IIR Seminar on Risk Management for Banks 2010, Amsterdam, 26-27 January

Kuester, K., S. Mittnik, and M. S. Paolella, 2006, “Value-at-Risk prediction: a comparison •

of alternative strategies,” Journal of Financial Econometrics, 4:1, 53-89

Markowitz, H., 1952, “Portfolio selection,” Journal of Finance, 7:1, 77-91•

Merton, R.C., 1994, “Influence of mathematical models in finance on practice: Past, pres-•

ent and future”, Philosophical Transactions (Royal Society), 347 (1684), p. 451-462

Patterson, S., 2010, The quants: how a new breed of math whizzes conquered Wall •

Street and nearly destroyed it, Random House, Inc

Rajan, R. G., 2009, The credit crisis and cycle-proof regulation, Federal Reserve Bank of •

St. Louis Review, September/October

Rajan, R., 2008, “Bankers’ pay is deeply flawed,” Financial Times, 8 January•

Shojai, S., and G. Feiger, 2010, “Economists’ hubris: the case of risk management,” •

Journal of Financial Transformation, Forthcoming

Senior Supervisors Group, 2008, “Observations on risk management practices during •

the recent market turbulence”, March 6

Taleb, N. N., 2007, The black swan – the impact of the highly improbable, Random •

House, New York

Thal Larsen, P., 2008, “The flaw in the logic of investment banks’ largesse,” Financial •

Times, 6 January

Trichet, J-C. , 2009, “(Under-)pricing of risks in the financial sector”, speech 19 January•

Tucker, P., 2010, Shadow banking, capital markets and financial stability, Remarks, •

Bernie Gerald Cantor (BGC) Partners Seminar, Bank of England, London, 21 January

138 – The journal of financial transformation

Manuscript guidelines

All manuscript submissions must be in English.

Manuscripts should not be longer than 7,000 words each. The maximum number of A4 pages allowed is 14, including all footnotes, references, charts and tables.

All manuscripts should be submitted by e-mail directly to the editor@capco.com in the PC version of Microsoft Word. They should all use Times New Roman font, and font size 10.

Where tables or graphs are used in the manuscript, the respective data should also be provided within a Microsoft excel spreadsheet format.

The first page must provide the full name(s), title(s), organizational affiliation of the author(s), and contact details of the author(s). Contact details should include address, phone number, fax number, and e-mail address.

Footnotes should be double-spaced and be kept to a minimum. They should be numbered consecutively throughout the text with superscript Arabic numerals.

for monographsAggarwal, R., and S. Dahiya, 2006, “Demutualization and cross-country merger of exchanges,” Journal of Financial Transformation, Vol. 18, 143-150

for booksCopeland, T., T. Koller, and J. Murrin, 1994, Valuation: Measuring and Manag-ing the Value of Companies. John Wiley & Sons, New York, New York

for contributions to collective worksRitter, J. R., 1997, Initial Public Offerings, in Logue, D. and J. Seward, eds., Warren Gorham & Lamont Handbook of Modern Finance, South-Western Col-lege Publishing, Ohio

for periodicalsGriffiths, W. and G. Judge, 1992, “Testing and estimating location vectors when the error covariance matrix is unknown,” Journal of Econometrics 54, 121-138

for unpublished materialGillan, S. and L. Starks, 1995, Relationship Investing and Shareholder Activ-ism by Institutional Investors. Working Paper, University of Texas

Guidelines for manuscript submissions

Guidelines for authors

In order to aid our readership, we have established some guidelines to ensure that published papers meet the highest standards of thought leader-ship and practicality. The articles should, therefore, meet thefollowing criteria:

1. Does this article make a significant contribution to this field of research? 2. Can the ideas presented in the article be applied to current business mod-

els? If not, is there a road map on how to get there.3. Can your assertions be supported by empirical data?4. Is my article purely abstract? If so, does it picture a world that can exist in

the future?5. Can your propositions be backed by a source of authority, preferably

yours?6. Would senior executives find this paper interesting?

subjects of interestAll articles must be relevant and interesting to senior executives of the lead-ing financial services organizations. They should assist in strategy formula-tions. The topics that are of interest to our readership include:

• Impact of e-finance on financial markets & institutions• Marketing & branding• Organizational behavior & structure• Competitive landscape• Operational & strategic issues• Capital acquisition & allocation• Structural readjustment• Innovation & new sources of liquidity • Leadership • Financial regulations• Financial technology

Manuscript submissions should be sent toProf. Shahin Shojai, Ph.D.The EditorEditor@capco.com

CapcoBroadgate West9 Appold StreetLondon EC2A 2APTel: +44 207 426 1500Fax: +44 207 426 1501

139

The world of finance has undergone tremendous change in recent years. Physical barriers have come down and organizations are finding it harder to maintain competitive advantage within today’s truly global market place. This paradigm shift has forced managers to identify new ways to manage their operations and finances. The managers of tomorrow will, therefore, need completely different skill sets to succeed.

It is in response to this growing need that Capco is pleased to publish the ‘Journal of financial transformation.’ A journal dedicated to the advancement of leading thinking in the field of applied finance.

The Journal, which provides a unique linkage between scholarly research and business experience, aims to be the main source of thought leadership in this discipline for senior executives, management consultants, academics, researchers, and students. This objective can only be achieved through relentless pursuit of scholarly integrity and advancement. It is for this reason that we have invited some of the world’s most renowned experts from academia and business to join our editorial board. It is their responsibility to ensure that we succeed in establishing a truly independent forum for leading thinking in this new discipline.

You can also contribute to the advancement of this field by submitting your thought leadership to the Journal.

We hope that you will join us on our journey of discovery and help shape the future of finance.

Prof. Shahin ShojaiEditor@capco.com

Request for papers — Deadline 8 July, 2010

For more info, see opposite page

2010 The Capital Markets Company. VU: Prof. Shahin Shojai,

Prins Boudewijnlaan 43, B-2650 Antwerp

All rights reserved. All product names, company names and registered trademarks in

this document remain the property of their respective owners.

Design, production, and coordination: Cypres — Daniel Brandt and Pieter Vereertbrugghen

© 2010 The Capital Markets Company, N.V.

All rights reserved. This journal may not be duplicated in any way without the express

written consent of the publisher except in the form of brief excerpts or quotations for review

purposes. Making copies of this journal or any portion there of for any purpose other than

your own is a violation of copyright law.

www.capco.com

Capco offices

Amsterdam

Antwerp

Bangalore

Chicago

Frankfurt

Geneva

London

Luxembourg

Mumbai

New York

Paris

Pune

San Francisco

Toronto

Washington DC

T +32 3 740 10 00