The Growth of Finance, Financial Innovation, and

Systemic Risk

Lecture 3

BGSE Summer School in Macroeconomics, July 2013

Nicola Gennaioli, Universita’ Bocconi, IGIER and CREI

Risk Taking, Leverage and Financial Innovations

We saw previously how macroeconomic developments were linked to the growth in risk taking by certain financial sector players

We now consider the corresponding growth in leverage.

Two key points: It was inextricably lined to the creation of new

financial instruments These new instruments were critical in the

recent crisis

2

Shadow banking and the 2007-2008 crisis

3

Shadow (/securitized) banking: Provision of short-term safe debt to

intermediaries Debt is collateralized through securitization:

Intermediaries: originate, acquire, and pool loans Tranching of loan pools to create safe pieces

In 2007-8, as loan pools lost value, the system unraveled: External financing stopped, intermediaries lost from

retained risks

The Shadow Banking Sector

Origination: finance companies, financed through commercial paper

Loan warehousing, pooling, and tranching into ABS and their intermediation: broker dealers, structured investment vehicles (SIVs), etc.

Funding of above activities: money market funds, securities lenders

4

Shadow Banking and Leverage

5

Shadow Banking and Innovation

Traditional banking: banks raise deposits, originate loans, and keep these loans in their balance sheets.

Originate and distribute banking: banks raise deposits, originate loans, but sell these loans to the markets. Loans are pooled by shadow banks, and used

to crated ABS to raise collateralized financing.

The originate and distribute model was taught to bring stability: it would allow banks to reduce the risk in their balance sheets. Not quite what happened.

6

Build a “neglected risks” model of Shadow Banking and Securitization7

Demand for safety: Outside investors only want riskless debt.

Securitization: Intermediaries use funds to originate safe and risky loans. Risky loans are subject to institution-specific idiosyncratic risk (and to aggregate risk). Trading and pooling of risky loans eliminates idiosyncratic risk.

Neglected Risks: investors and intermediaries neglect low probability aggregate risks (GS 2010, GSV 2011).

Neglected Risks8

This assumption captures:

Uncertainty of model-economy: overweighting of historical trends

Difficulty in measuring risk in complex, interconnected financial institutions

These problems exacerbated by financial innovations. New securities are difficult to understand for the price and to price for intermediaries Wrong models (Coval Jurek and Stafford, 2010)

How robust is the financial system to these problems?

Main results I9

Investors’ wealth drives securitization/shadow banking: As investors’ wealth becomes large, intermediaries

make marginal, risky loans. To create safe collateral, they securitize and pool them.

Intermediaries’ assets (loan portfolios) and liabilities (riskless debt) grow together. Aggregate risk yields a carry trade.

Pooling of risks endogenously renders intermediaries interconnected. Under RE, stability and welfare go up.

Main results II

With neglected risks, securitization creates a “diversification myth”: each single intermediary now looks safer.

Yet, pooling of idiosyncratic risks also raises the exposure of all intermediaries to any neglected aggregate tail risks. Ex-ante pooling creates ex-post fragility and

illiquidity

Securitization: expand ex-ante financing but creates fragility ex-post by capitalizing on investors’ misperception of risks

10

Some Related literature11

Link shadow banking to growing investor wealth (Caballero et al. 2008). Account for link between risk-taking and low interest rates (Maddaloni and Peydro 2011, Jimenez et al. 2011). Show that with, neglected risks, insurance creates catastrophe bonds (Coval, Jurek, and Stafford 2009b).

Explain comovement of assets and leverage (Adrian and Shin, 2010) and intermediaries’ risk retention (Acharya, Schnabl, and Suarez 2010).

Endogenize bank interconnectedness and systematic risk. Shin (2009a) and Allen and Gale (2000). We focus on neglect of aggregate tail risks.

Explain how banks lose a fortune holding other banks’ risks in a crisis. See Benmelech and Dlugosz on CDO’s.

Ex-post illiquidity. Geanakoplos (2009), SV (2010), Gorton and Metrick (2010), etc. We focus on ex-ante insurance, not on short term debt.

Model pooling and tranching, but not as a result of asymmetric information (De Marzo and Duffie 1999) or ring-fencing.

Organization of Presentation

12

The model with rational expectations The model with local thinking and results Extensions (briefly)

13

Two dates t = 0, 1

There is a measure 1 of infinitely risk averse investors

Eω[C0 + minωC1,ω]

they receive wealth w at t = 0

There is a measure 1 of risk neutral intermediaries

Eω[C0 + C1,ω ]

they receive wealth wint < 1 at t = 0

Intermediaries invest by using their own wealth wint and by issuing riskless debt to investors: Issue debt D at t = 0, promise to repay rD at t =

1.

14

Intermediaries have access to projects that pay in t=1:

Safe (H): invest IH,j and obtain RIH,j at t = 1

Limited aggregate supply of 1.

Risky (L). Invest IL,j and obtain:

There are three aggregate states ω = g, d, r, with

πg > πd > πr and Pr(πω) = φω

- there is both idiosyncratic and aggregate risk

- a pool of projects yields AIL πω in aggregate state ω

otherwise

ororyprobabilitwithAI rdgjL

0

),(,

Intermediaries’ return to investment

15

Technology features decreasing returns + increasing risk

Marginal Return

R

1

E(πω)A

Total Investment

A

0

Timing

16

At t = 0 each intermediary borrows Dj , invests IH,j and IL,j, sells SL,j units of risky investment, buys TL,j units of diversified pools of other intermediaries’ investments. Interest rate r, price of loans pL: competitively

set at t=0.

At t = 1 state ω and intermediaries’ returns are revealed. Investment pays off and debt is repaid.

Investors lend all of their wealth w if r > 1, they are indifferent between lending or not at r = 1.

Intermediaries’ expected profits I17

At t = 0 each intermediary j has expected profits:

R∙IH,j +

+ [Eω(πω)∙A∙(IL,j–SL,j) + Eω(πω)∙A∙TL,j + pL(SL,j–TL,j)] +

+ Dj – IH,j – IL,j + wint – rDj.

Return from idiosyncratic risk kept (IL,j–SL,j) is 0 or A.

Return from securitized pool TL,j is πω∙A in ω. The pool is only subject to aggregate risk

about πω.

Intermediaries’ expected profits II18

Intermediary j holds two types of risky investments. Risky investments originated and kept:

(IL,j–SL,j) →

Risky securitized pools purchased in the

market:

TL,j →

A∙(IL,j–SL,j) πg φg+ πd φd+ πr φr

0 otherwise

A∙πg ∙TL,j φg

A∙πd ∙TL,j φd

A∙πr∙TL,j φr

19

The constraints faced by the intermediary are:

Feasibility: at t = 0 cannot invest more than resources raised

Riskless debt constraint:

rDj ≤ R∙IH,j + πr∙A∙TL,j.

Pledge safe return and securitized pool in worst state πr.

Intermediaries’ “carry trade” is [ Eω(πω)A – r ]∙TL,j

Feasibility of Securitization:

SL,j ≤ IL,j

Cannot sell more investments than those undertaken

Preliminaries

20

Risky asset is securitized only if debt constraint is binding Pooling of risks relaxes investors’ demand for

safe collateral

Securitization pools are bought by intermediaries: they are the high value buyers. Thus, TL,j = SL,j. Use pools to back debt. Securitization supports growth in leverage and… Allows intermediaries to earn a return above

safe debt

21

low w : no risky investment, no debt, no securitizationInterest rate

Total wealth

Risky investment

and securitization

R

wint

22

higher w: no risky investment, some debt, no securitizationInterest rate

Total wealth

Risky investment

and securitization

R

wint1

R > R(1- wint)

23

Intermediate w: some risky investment, debt, no securitizationInterest rate

Total wealth

Risky investment

and securitization

R

wint1

R > E(πω)Aw

E(πω)A

IL = w + wint – 1

R / E(πω)A

24

high w : risky investment, debt, securitizationInterest rate

Risky investment

and securitization

R

wint1

R + πrASL > E(πω)Aw

E(πω)A

IL = w + wint – 1

SL = [E(πω)/πr]w – R/Aπr

R/E(πω)A wint + w*

25

Very high w : risky investment, debt, maximal securitizationInterest rate

Risky investment

and securitization

R

wint1

R + πrAIL > r(w)wE(πω)A

R/E(πω)A wint + w* wint + w**

1

Securitization under RE I26

Securitization endogenously arises to meet the demand “w” for riskless debt. Driven by marginal, risky, projects. By lowering idiosyncratic risk, pooling boosts

safe collateral and debt capacity. Growth of assets and leverage

Pools allow intermediaries to earn a yield (“carry trade”)

When at t = 1 returns are revealed, not much happens Some intermediaries do better than others (if

securitization is partial), but all debt is truly safe.

Securitization is welfare improving.

Securitization under RE II27

In worst case scenario πr :

A fraction 1 – πr of intermediaries get 0 on their

projects:

[R + πr∙A∙SL + 0∙(IL–SL)] – [R + πr∙A∙SL] = 0

A fraction πr of intermediaries get A on their

projects:

[R + πr∙A∙SL + A∙(IL–SL)] – [R + πr∙A∙SL] = A∙(IL–SL) > 0

Securitization and Local Thinking

28

Local thinking (GS 2010, GSV 2011): neglect unlikely (tail) risk. Here is “recession”, because φr = min φω.

At t = 0, agents think only of “growth” and “downturn”

Two things change with respect to RE at t = 0 Assess higher average return ELT(πω)A > E(πω)A

Relax debt constraint:

rDj ≤ R + πd∙A∙TL,j.

Equilibrium under Local Thinking at t = 029

Securitization and leverage expand. At low w this raises interest rates, at higher w this also boosts investment

Interest rate

R

E(πω)A

wint+w* wint + w**

1

ELT(πω)A

wint + w**,LT

higher r higher r, D

Securitization and Local Thinking at t=1

30

If at t = 1 the state is g or d, debt is sustainable, as with RE

In worst case scenario πr :

Share 1 – πr of unsuccessful intermediaries fail:

[R + πr∙A∙SL + 0∙(IL–SL)] – [R + πd∙A∙SL] = – (πd – πr)∙A∙SL < 0

Share πr of successful intermediaries also fail

iff:

[R + πr∙A∙SL + A∙(IL–SL)] – [R + πd∙A∙SL] =

= A∙(IL–SL) – (πd – πr)∙A∙SL < 0

31

Default and repayment in recession

All intermediaries fail if:

< 1 + (πd – πr)

A “successful” intermediary is more likely to fail if more investment is securitized! Pooling creates correlation in

intermediaries’ assets Small mistakes create massive fragility

when w is large

L

L

S

I

Securitization and Market Liquidity at t = 132

At t = 1, state ω is learned only partially. Observe s in {l, h} Here h is informative of {g, d}, while l is

informative of {d, r} In s, a share qs of risky projects pays off A at t =

1. qh > ql

Two implications from imperfect learning and “early” projects: We can study retrading and market liquidity at t

= 1 Early intermediaries may have liquidity to buy

claims at t = 1.

Due to “early” projects, some debt repayment occurs at t = 1 Still focus on long term debt, but promising two

coupons For simplicity, return R is ring fenced by most

senior debt class

Event tree nesting the previous setup

33

Event tree under local thinking

34

As ql is revealed, the agent realizes to be in the lower branch, which did not come to mind at t = 0.

Basic results with partial information

35

Not much changes at t = 0. If at t = 1 neglected state ql realizes, investors learn that debt may default at t = 2 if state is πr.

In ql investors value each securitized asset πrA, intermediaries value the same asset at E(πω|ql)A > πrA. Can a trade arise?

The total liquidity of “early” intermediaries is equal to:

ql∙[A∙(IL,j–SL,j) – (πd – πr)A∙SL,j]

Which increases in the unsecured portion of projects and decreases in the unexpected drop in collateral (πd – πr)A

Market Fragility at t = 1

36

In neglected state ql the price of securitized projects drops to investors’ reservation value πrA when:

IL,j/SL,j < 1 + (πd – πr) + πd/ql

High securitization reduces the liquidity of successful intermediaries

Securitization creates fragility also by draining out market liquidity after neglected risks realize. Limited securitization leaves “spare liquidity” ex-post. Correlation in balance sheet costly when neglect

risk occursThis goes beyond idea that intermediaries

commit all of their wealth at t = 0 (see SV 2010 and GSV 2011)

37

Ex-ante and ex-post liquidity

Due to market liquidity, investors might be willing to lend against risky collateral (and debt) based on resale value. Or, equivalently, to directly buy securitized

assets. Markets do the “pooling,” regardless of who holds the risky assets

If early intermediaries buy back risky assets at t = 1 at p1 > πdA, investors may lend more than reservation value at t = 0

Market trading at price p1

38

Failure of market insurance at t = 1

Liquidity in good times is consistent with illiquidity in bad times if

“Normal times” market liquidity at t = 1 can coexist with illiquidity when neglected risks materialize.

Investors lend more than reservation value πdA because they expect the t = 1 market to be liquid. If securitization is large, as ql realizes the market becomes illiquid and price drops to πrA

Securitization creates both liquidity in normal times and illiquidity when neglected risks materialize. This boosts fragility and creates spikes in risk premia

drop price unexpecteddropliquidity unexpected

)(E

r

h

l

hq

q

q

Some Facts (I)39

Mortgage Origination and Subprime Securitization

Some Facts (II)

Securitization and decline in lending standards

40

Some Facts (III)

Centrality of the collapse of AAA securities

41

Some Facts (IV)

Collapse of commercial paper market

42

Some Facts (V)

Securitization, subprimes and the collapse of ABS

43

Some Facts (VI)

Unusual securities

44

Some Facts (VII)

Collapse in the issuance of ABS

45

Conclusions I

We offered the following theory for the positive correlation between assets and leverage, and the resulting financial fragility: Large wealth of risk averse investors creates

enormous pressure/opportunities for banks to manufacture safe assets

This induces banks to securitize and expand their balance sheets by holding “safer” pooled risks

Banks make enormous profits out of the resulting carry trade

The system becomes highly interconnected. As some regional housing markets cool off and delinquencies rise, the system collapses

46

Conclusions I

Neglected risks are subtle and changing Cannot expect regulators to stay ahead Capital requirements are a crude but

appropriate instrument for reducing bets Problems with market-based risk weighting

Extreme concentration of exposures to a given asset class should raise a red flag

Deeper skepticism about innovations that capitalize on neglect of risk, such as prime MMF.

47