The Geography of International Portfolio Flows ...

The Geography of International PortfolioFlows, International CAPM, and the Role of

Monetary Policy Frameworks∗

Roberto A. De SantisEuropean Central Bank

Using bilateral data on international equity and bond flows,we find that (i) the prediction of the International CapitalAsset Pricing Model is partially met and that global equitymarkets might be more integrated than global bond markets,and (ii) asset allocators engage in trend-chasing activities, par-ticularly in bond markets. Moreover, over the turbulent 1998–2001 period characterized by an equity bubble and the subse-quent burst, we find evidence that investors preferred portfolioassets of countries where the central bank gave relative impor-tance to money. As for EMU, once controlling for diversifica-tion benefits and the elimination of the exchange rate risk, weshow that cross-border portfolio flows among euro-area coun-tries have increased due to the catalyst effect of EMU. Coun-tries’ shares in the world market portfolio, home bias, initialdegree of misallocation across countries, past returns, diversi-fication benefits, and EMU can explain 35–40 percent of thetotal variation in equity and bond asset flows.

JEL Codes: C13, C21, F37, G11.

1. Introduction

The International Capital Asset Pricing Model (IntCAPM) suggeststhat international investors should hold assets of each country in

∗The views expressed in this paper are those of the author and do not nec-essarily reflect those of the European Central Bank or the Eurosystem. AuthorContact: European Central Bank, Kaiserstrasse 29, 60311 Frankfurt am Main,Germany. E-mail: roberto.de [email protected]; Tel: +49-69 1344 6611.

147

148 International Journal of Central Banking June 2010

proportion to the country’s share in the world market portfolio.1

This implies that all countries, in a world without transaction andinformation costs, would hold the same portfolio and would diver-sify their investment in other countries in proportion to the size oftheir financial markets. In this respect, global indices such as thepopular Morgan Stanley Capital International (MCSI) All CountryWorld Index (ACWI), Datastream Global index, and Standard andPoor’s (S&P’s) Global index are widely used by investors as theirperformance benchmarks for the global asset portion of their equityportfolio. The first aim of this paper is to examine whether countriesallocate money according to the simple prediction of the IntCAPM.We also assess the empirical relevance of the IntCAPM for the bondmarket. In order to carry out this test, we employ the Lehman Broth-ers Multiverse index released in 2001 as the benchmark for the globalbond market.

It is useful to point out that recent studies have looked at the con-temporaneous link between the actual weight of country j in fundi’s equity portfolio and the optimal weight suggested by the Int-CAPM for emerging markets (Gelos and Wei 2005). We instead aimat investigating whether portfolio flows (divided by the size of theinvestors) and, therefore, the global reallocation of portfolio capitalacross twenty-three developed countries and seven emerging-marketeconomies are functions of the optimal weights at the beginning ofthe period as suggested by the IntCAPM.

In order to carry out such a study, we construct a database onbilateral cross-border equity and bond flows for thirty countries cov-ering in 2001 80 percent of world GDP, 84 percent of world inter-national investment in equity portfolios, and 71 percent of worldinternational investment in bonds and notes portfolios. Therefore,the thirty countries of the sample and the cross-section of 870 bilat-eral observations encompass all important international investmentdecisions (excluding official investments by monetary authorities).

Specifically, we construct measures of bilateral net asset flows ofequities and bonds mainly on the basis of the International Monetary

1The International Capital Asset Pricing Model and the Intertemporal AssetPricing Model are often referred to using the same label: I-CAPM. We useIntCAPM to avoid confusion.

Vol. 6 No. 2 The Geography of International Portfolio Flows 149

Fund (IMF) Coordinated Portfolio Investment Survey (CPIS) data-base, which reports bilateral holdings at end-1997 and end-2001. Wedefine net asset flows of equities and bonds as the purchases minusthe sales of country k’s equities and bonds by citizens of country c.The considered period was characterized by the sharp rise in cross-border capital flows globally, the increased percentage of householdsavings invested in capital markets, the boom and bust of the equitybubble, and the establishment of European Economic and MonetaryUnion (EMU) in January 1999. Looking at cumulated portfolio netasset flows over the four-year period 1998–2001 has the advantageof abstracting from short-run variations in international portfolioflows, which could be due to unexpected economic news and cyclicaldevelopments, as well as phenomena which are difficult to pin down.

The second aim of the paper is to assess whether, during thisturbulent period for asset markets, central banks’ monetary pol-icy frameworks across countries influenced the geography and sizeof international capital flows.2 As a consequence of the asset-priceshocks, international investors might have preferred to reallocatetheir portfolio assets toward countries that gave importance to spe-cific characteristics of monetary policy frameworks. Therefore, weemploy the results of a survey commissioned and coordinated by theBank of England in 1998 aiming at measuring several key charac-teristics of the monetary frameworks consistently across ninety-fourcentral banks (Mahadeva and Sterne 2000). Moreover, given the newmonetary policy framework, which came to light in Europe in Janu-ary 1999, we also look and control for the potential impact of EMUon global portfolio flows as found by Lane (2006) for fixed-incomeinvestments and by De Santis and Gerard (2009) for both equityand fixed-income investments.

International portfolio flows have skyrocketed in the last fifteenyears, and a large number of studies have tried to explain their

2A large body of the literature pointed to the importance of institutionalcharacteristics of the monetary policy framework—such as central banks’ inde-pendence, accountability, and transparency—to control inflationary expectationsand safeguard financial stability (i.e., Rogoff 1985; Alesina and Summers 1993).Several measures of central banking independence were also constructed andused to study the impact on inflation (Grilli, Masciandaro, and Tabellini 1991;Cukierman 1992; Jacome and Vazquez 2005).


determinants. However, data limitations have meant that these con-tributions focused on country aggregates of net equity and debt flows(De Santis and Luhrmann 2009), on country aggregates of inflowsof equity capital—foreign direct investment plus portfolio equitysecurities—(Alfaro, Kalemli-Ozcan, and Volosovych 2007), or haveonly considered a single source country, most often the United Statesbeing the recipient or the source of the investment (Bohn and Tesar1996; Brennan and Cao 1997; Coval and Moskowitz 1999; Froot,O’Connell, and Seasholes 2001; Huberman 2001; Ahearne, Griever,and Warnock 2004). Portes and Rey (2005) is the only study thatlooks at bilateral cross-border equity flows between fourteen coun-tries pointing to the role played by information costs. De Santis andGerard (2009) is the only study looking at the determinants of bilat-eral changes in portfolio country weights in both equity securitiesand fixed income for a panel of thirty countries pointing to the roleplayed by the non-linear fully hedged diversification benefits, the ini-tial degree of misallocation, and the establishment of EMU. Anotherbranch of the literature has studied the determinants of bilateralbank and portfolio holdings using empirical methods borrowed fromthe traditional gravity models of international goods trade (Faruqee,Li, and Yan 2004; Lane 2006; Aviat and Coeurdacier 2007; Lane andMilesi-Ferretti 2008; Papaioannou 2009).

Our main contribution to the literature is that we find clear evi-dence that portfolio asset flows are influenced positively by the rela-tive size of the recipient countries’ financial markets. The predictionsof the IntCAPM are only partially met, as the estimated coefficienton the benchmarks is less than unity: 0.4 for the equity portfolioand 0.2 for the bond portfolio. Moreover, we find that asset alloca-tors engage in trend-chasing activities, particularly in bond markets.These results do not change when controlling for home bias and theinitial degree of underweight, which enters non-linearly, thereforepotentially proxying for initial direct and indirect costs.

We also find that, during the turbulent 1998–2001 period, inter-national investors in both equity and fixed-income markets had atendency to purchase assets issued by countries whose monetaryauthorities gave importance to money.

We estimate on a 95 percent confidence interval the potential cat-alyst effect of the euro within the euro area to amount to USD 22–47billion in equity securities and USD 32–76 billion in bonds and notes,


which implies that EMU might have enhanced risk sharing amongeuro-area member states. The catalyst effect of the euro is esti-mated after controlling for the elimination of the exchange rate riskamong euro-area member states and the effect of being a memberof the European Union (EU). EMU boosted the cross-border invest-ment activity among euro-area member states due to the removal ofintra-area currency-matching rules, the sharing of common tradingplatforms, and the cross-border merger of the Amsterdam, Brussels,Lisbon, and Paris exchanges (Euronext).

The proportion of the total variation explained by our empiricalmodels amounts to 35–40 percent for both equity and bond flows.This is a valuable result given that we look at the geography ofinternational portfolio flows during a very difficult period for assetmanagement.

The remainder of the paper is organized as follows. Section 2discusses the key ideas. Section 3 describes the databases used forthe analysis. Section 4 outlines the empirical approach and teststhe IntCAPM. Section 5 assesses the role of monetary policy frame-works and estimates the potential impact of EMU on portfolio flows.Section 6 reports how robust the model specifications are when con-trolling for other variables, which could potentially affect portfolioflows. Section 7 concludes.

2. A Simple Theory

2.1 The IntCAPM

Consider a representative investor from a generic country c and letγc denote the investor’s degree of risk aversion. Also, indicate withμc the (N × 1)-vector of expected returns in excess of the risk-freerate on the N risky assets, and with Σc the expected (N × N)-covariance matrix for the risky assets, where the subscript c indi-cates that returns are measured in the currency of country c. If theinvestor faces no constraints on foreign holdings and financial mar-kets are perfectly integrated, mean-variance optimization implies thefollowing portfolio allocation:

w∗c,t =

1γc

Σ−1c,t μc,t,


where w∗c,t is the (N × 1)-vector of optimal weights for the N risky

assets.The optimal weights are not observable. However, in a fully

integrated world where purchasing power parity (PPP) holds, Sol-nik (1974) and Sercu (1980) show that the international version ofthe simple CAPM of Sharpe (1964) and Lintner (1965) holds: themarket equilibrium is achieved when all investors hold the worldmarket portfolio, where each country portfolio is weighted by itsmarket capitalization. Therefore, an observable estimate of the opti-mal weight invested in each country k is equal to that country k’smarket capitalization weight in the world index portfolio, wBench

k,97 :

w∗c,t = wBench

k,t .

There exist several standard benchmarks for the equity portfolio,such as the popular MCSI ACWI, Datastream Global index, andS&P’s Global index, as they provide consistent data, have suffi-ciently long price history, and are widely used by global investors.We use the Datastream Global index and compare the results withS&P’s Global index, as both include the country coverage used inthis study.

To our knowledge, a similar benchmark for the bond portfoliofor such a number of countries was never used due to difficulty incompiling countries’ bond market capitalization at market value. InJanuary 2001, Lehman Brothers launched a new index (i.e., Multi-verse index), which provides a broad-based measure of the interna-tional fixed-income bond market, with index history dating back toJanuary 1999. Multiverse index provides information on the overallstatus of the global debt asset class and offers a means to comparethe entire global debt asset class across countries.

Table 1 provides the estimated market shares across countries inboth equity and bond markets, which will then be used to test theIntCAPM. The first and second columns report the equity portfolioweights computed using the market value of the Thomson Data-stream and S&P’s Global indices, respectively. The last two columnsreport, respectively, the bond portfolio weights computed using themarket value of the Lehman Brothers Multiverse index and the out-standing amount as reported by the Bank for International Settle-ments (BIS). The equity portfolio weights of Datastream and S&P


Table 1. Portfolio Weights (percentage share)

Equity Portfolio Bond Portfolio

Datastream S&P Lehman Bros.Global Global MultiverseIndex Index Index BISEnd-97 End-97 End-98 End-97

Argentina 0.31 0.26 0.39 0.16Australia 1.35 1.28 0.55 0.56Austria 0.17 0.15 0.72 0.64Belgium 0.64 0.59 1.96 1.43Bermuda — 0.01 0.02 0.01Canada 2.47 2.46 2.56 2.07Chile 0.29 0.31 0.03 0.19Denmark 0.40 0.41 0.88 1.26Finland 0.35 0.32 0.83 0.26France 3.32 2.92 6.12 4.10Germany 4.18 3.57 6.06 7.76Iceland — 0.01 0.00 0.02Indonesia 0.12 0.13 0.04 0.29Ireland 0.24 0.21 0.25 0.12Israel 0.13 0.20 0.01 0.43Italy 1.80 1.49 5.65 5.46Japan 12.51 9.59 7.18 16.37Korea 0.14 0.20 0.17 0.37Malaysia 0.36 0.40 0.01 0.18Netherlands 2.75 2.03 2.14 1.45New Zealand 0.17 0.13 0.13 0.07Norway 0.30 0.29 0.18 0.26Portugal 0.30 0.17 0.32 0.24Singapore 0.53 0.46 0.11 0.07Spain 1.25 1.26 2.48 1.31Sweden 1.10 1.18 1.09 1.03Thailand 0.09 0.10 0.01 0.05United Kingdom 10.50 8.64 4.74 2.78United States 44.90 48.92 53.57 48.02Venezuela 0.04 0.06 0.11 0.11Subtotal 90.69 87.73 98.31 97.06Total (USD billion) 17,634 23,116 10,355 19,054

Note: This table presents the sample descriptive statistics for portfolio weights usedto test the validity of the IntCAPM.


are very similar. Some small differences can be identified for the bondportfolio weights mainly due to the fact that the Lehman BrothersMultiverse index is evaluated at market value, while the BIS bondoutstanding is at face value.

The computations reported in table 1 indicate that accordingto the IntCAPM, almost 50 percent of world portfolio should beinvested in U.S. securities. As for the euro area (e.g., Luxembourgand Greece), approximately 14 percent of world portfolio should beinvested in euro-area equity securities and 25 percent in euro-areabonds.

We will report the empirical results using Datastream weights forequity portfolio and Lehman Brothers weights for bond portfolio.3

It is useful to point out that all the results remain invariant whenusing, respectively, S&P weights and BIS weights.

2.2 The Role of Central Banks

There is no doubt that capital flows are affected by the institutionsof a country (Alfaro, Kalemli-Ozcan, and Volosovych 2007; Coeur-dacier, De Santis, and Aviat 2009; De Santis and Luhrmann 2009).However, there is no empirical study looking at the role of centralbanking institutional setting. Surely, monetary policy influences thedecisions of international investors, but less obvious is understand-ing why specific central bank monetary policy characteristics shouldaffect international investment decisions. Moreover, at the turn ofthe century, as a consequence of the asset-price shocks, internationalinvestors might have preferred to reallocate their portfolio assetstoward countries that gave importance to specific characteristics ofmonetary policy frameworks.

Therefore, we employ the results of a comprehensive survey com-missioned by the Bank of England in 1998, because it aimed atmeasuring consistently the diversity in monetary frameworks acrossninety-four central banks with a coverage of characteristics thatstretched beyond previous studies (Mahadeva and Sterne 2000).Specifically, we look at the following characteristics: (i) short- and

3The employed portfolio weights for Bermuda and Iceland amount to 0.01percent as indicated by S&P’s Global index.


medium-term policy focus—inflation, money, exchange rate, and dis-cretionary policy; (ii) institutional characteristics—independence,accountability, and transparency of policy explanations; and (iii)structural characteristics—importance given to financial stability inthe setting of monetary policy instruments.

Central bank independency is defined by a range of character-istics covering legal objectives, goal, instruments, finance of thegovernment deficit, and term of office of the governor.

The measure of accountability was constructed by assessing howfar the central bank has a legal or informal responsibility to explainand defend its policies to government and parliament and to involveparliament in monetary policy decisions. Therefore, the measurerelates to accountability to a specific target as well as to govern-mental and parliamentary monitoring of the central bank.

The measure of policy explanations is defined by the effortmade by the central bank in explaining policy decisions, assess-ment of the economy, and forecasts and forward-looking analy-sis. It can be interpreted as one aspect of transparency in thatdeeper explanations of policy, which allow us to understand itsgoal and the means by which policymakers react to changes in eco-nomic conditions, is one important manifestation of higher degree oftransparency.

As for the measures of monetary policy objectives, they aredefined and classified by the exchange rate, money growth, andinflation dimensions, rather than just one dimension, and give thedegree to which a country’s policy focused on a particular objective.Therefore, the survey is constructed to avoid a 100 percent commit-ment to a single objective, as in most cases definitions that focus onthe explicit variable targeted may not fully capture policy prefer-ences. The general measure of discretion is a non-linear combinationof the scores for exchange rate focus, money focus, and inflationfocus.

Finally, the importance of financial stability in the monetaryframework is defined by various financial stability issues, such asthe volatility of asset prices, domestic and overseas financial sectorinsolvency, and credit rationing.

The eight indices range between 0 and 100, where a high scoreimplies more independence, more accountability of central bank togovernment, more policy explanations to those outside the central


bank, and a higher degree of importance given to policy objectivesand financial stability issues (see table 2).4

Some of these indices are also strongly correlated (see table 3).The focus on the exchange rate is strongly correlated with the focuson inflation and money, with the transparency index, and with theimportance given to financial stability; in turn, the focus on moneyis strongly correlated with the importance of financial stability.

Interestingly, independence and accountability are negativelycorrelated (−20 percent), which implies that explaining and defend-ing monetary policies to government and parliament and the involve-ment of parliament in monetary policy decisions might be partly seenby central banks as an infringement on their independence.

2.3 The Role of EMU

The establishment of EMU in January 1999 was a fundamentalinstitutional change in the world economy that has affected thedirection and the magnitude of global portfolio flows. Lane (2006)and De Santis and Gerard (2009) uncover evidence of euro-areainvestors having assigned a higher weight to portfolio investmentin euro-area countries, which implies that EMU has facilitatedportfolio market access, enhancing risk sharing and regional finan-cial integration. EMU boosted the cross-border investment activityamong euro-area member states due to the removal of intra-areacurrency-matching rules, the sharing of common trading platforms(such as the MTS model composed of several national Europeanmarkets for trading fixed-income securities),5 and the cross-bordermerger of the Amsterdam, Brussels, Lisbon, and Paris exchanges(Euronext).

In most countries, portfolio funds are often subject to someform of restriction on the level of their non-domestic investment.

4For a comprehensive analysis on the construction of all the indices measuringthe monetary policy frameworks across countries, refer to chapter 4 of Mahadevaand Sterne (2000). Bermuda and Venezuela are not included in the survey. There-fore, we use the U.S. characteristics for Bermuda, as the Bermuda dollar is atpar with the U.S. dollar, and we use the Uruguay characteristics for Venezuelagiven the similarities of the monetary policy frameworks of these two countries(see Jacome and Vazquez 2005).

5Detailed information can be found at http://www.mtsgroup.org/.


Tab

le2.

Sam

ple

Des

crip

tive

Sta

tist

ics

for

the

Char

acte

rist

ics

ofth

eM

onet

ary

Pol

icy

Fra

mew

ork

Str

uct

ura

lPol

icy

Focu

sIn

stit

uti

onal

Fac

tors

Fac

tor

Exch

ange

Fin

anci

alIn

flat

ion

Mon

eyR

ate

Dis

cret

ion

Indep

enden

ceA

ccou

nta

bility

Tra

nsp

aren

cySta

bility

Arg

enti

na0

010

00

7910

053

58A

ustr

alia

940

06

7383

788

Aus

tria

00

8813

6867

2717

Bel

gium

00

946

7733

688

Ber

mud

a19

250

8492

8395

33C

anad

a88

06

1691

100

7933

Chi

le88

031

2893

1783

17D

enm

ark

00

946

8875

700

Fin

land

560

6366

9192

748

Fran

ce40

4658

8490

8353

50G

erm

any

1988

1329

9617

7033

Icel

and

190

7534

5992

658

Indo

nesi

a50

636

6656

8383

83Ir

elan

d19

075

3487

8378

8Is

rael

880

1319

6610

068

33It

aly

4444

5094

8858

8133

Japa

n50

00

5093

089

50K

orea

6375

659

7383

8858

(con

tinu

ed)


Tab

le2.

(Con

tinued

)

Str

uct

ura

lPol

icy

Focu

sIn

stit

uti

onal

Fac

tors

Fac

tor

Exch

ange

Fin

anci

alIn

flat

ion

Mon

eyR

ate

Dis

cret

ion

Indep

enden

ceA

ccou

nta

bility

Tra

nsp

aren

cySta

bility

Mal

aysi

a44

038

7585

8371

67N

ethe

rlan

ds0

088

1391

8379

8N

ewZea

land

940

06

8910

092

8N

orw

ay0

069

3157

5089

17Por

tuga

l6

075

2885

8378

33Si

ngap

ore

190

5653

9025

4917

Spai

n56

063

6680

8359

33Sw

eden

100

013

697

8395

42T

haila

nd31

66

7582

5067

83U

nite

dK

ingd

om10

00

00

7710

094

16U

nite

dSt

ates

1925

084

9283

9533

Ven

ezue

la25

088

2570

834

17

Note

s:T

his

table

pre

sent

sth

esa

mple

des

crip

tive

stat

isti

csfo

rth

ech

arac

teri

stic

sof

the

mon

etar

ypol

icy

fram

ewor

k(s

eeM

ahad

eva

and

Ste

rne

2000

,pp.14

1–62

).T

he

eigh

tin

dic

esra

nge

bet

wee

n0

and

100,

wher

ea

hig

hsc

ore

implies

mor

ein

dep

enden

ce,m

ore

acco

unt

ability

togo

vern

men

ts,m

ore

pol

icy

expla

nat

ions

toth

ose

outs

ide

the

cent

ralban

k,an

da

hig

her

deg

ree

ofim

por

tance

give

nto

pol

icy

obje

ctiv

esan

dfinan

cial

stab

ility

issu

es.


Tab

le3.

Cor

rela

tion

Mat

rix

ofth

eSta

teof

Dev

elop

men

t,Q

ual

ity

ofth

eIn

stitution

s,an

dM

onet

ary

Fra

mew

orks

Var

iable

s

Dev

elopm

ent

Qual

ity

ofth

eIn

stit

uti

ons

Monet

ary

Policy

Fra

mew

ork

s

GP

BC

RD

CL

LP

ICP

CIN

FM

ON

FX

DIS

IND

AC

CT

RA

FS

GP

1.00

BC

0.29

1.00

RD

0.66

0.32

1.00

CL

−0.

65−

0.09

−0.

451.

00LP

0.87

0.31

0.67

−0.

651.

00IC

0.78

0.08

0.42

−0.

540.

791.

00P

C0.

710.

020.

53−

0.58

0.85

0.77

1.00

INF

−0.

17−

0.08

0.18

−0.

01−

0.01

−0.

180.

151.

00M

ON

−0.

100.

030.

100.

17−

0.17

−0.

27−

0.35

−0.

051.

00FX

0.08

−0.

44−

0.24

−0.

070.

010.

300.

01−

0.68

−0.

321.

00D

IS−

0.15

0.32

−0.

090.

30−

0.20

−0.

22−

0.37

−0.

120.

44−

0.21

1.00

IND

0.21

0.39

0.29

−0.

240.

370.

400.

370.

120.

06−

0.21

0.16

1.00

AC

C−

0.10

−0.

37−

0.02

−0.

19−

0.04

−0.

230.

080.

26−

0.19

0.03

−0.

12−

0.20

1.00

TR

A0.

280.

370.

34−

0.31

0.31

0.12

0.27

0.37

0.12

−0.

600.

080.

280.

021.

00FS

−0.

580.

23−

0.26

0.56

−0.

61−

0.70

−0.

720.

090.

42−

0.38

0.49

−0.

02−

0.05

0.07

1.00

Note

s:T

heva

riab

lesof

this

tabl

ear

ela

bel

edas

follo

ws:

GP

=C

ount

ries

’GD

Pper

capi

tam

inus

the

GD

Pper

capi

taof

Nor

way

;BC

=B

ank

cred

itto

the

priv

ate

sect

or/G

DP

min

usw

orld

shar

e;R

D=

R&

Dex

pen

ditu

re/G

DP

min

usw

orld

shar

e;C

L=

Civ

illib

erti

esin

dex;

LP

=La

Por

tain

dex;

IC=

ICR

Gin

dex;

PC

=Per

ceiv

edco

rrup

tion

inde

x;IN

F=

Infla

tion

focu

sin

dex;

MO

N=

Mon

eyfo

cus

inde

x;FX

=E

xcha

nge

rate

focu

sin

dex;

DIS

=D

iscr

etio

nfo

cus

inde

x;IN

D=

Inde

pen

denc

ein

dex;

AC

C=

Acc

ount

abili

tyin

dex;

TR

A=

Tra

nspa

renc

yin

dex;

FS

=R

ole

offin

anci

alst

abili

tyin

dex.


Currency-matching rules for portfolio funds, for example, are set toensure that foreign currency risk is reduced. Since the introductionof the euro in January 1999, the intra-euro-area currency-matchingrule has shifted from national currencies to the euro. The resultinggreater flexibility allowed individual euro-area country portfolios tosecure better diversification of investment risk by purchasing morenon-domestic euro-area assets.6

To control for the effect of EMU on global capital flows as wellas to measure its average impact, we include two sets of binary vari-ables. First we include a dummy which takes the value of 1 if thecountry receiving the investment belongs to the EMU. The coeffi-cient of this dummy measures the average portfolio asset flow intoindividual EMU countries for all investors. However, the effect of thesingle currency may be more pronounced on the investment decisionsof investor-residents in the euro area. To control for this differentialeffect, we include a dummy variable which takes the value of 1 whenboth investing and receiving countries belong to the EMU. The coef-ficient of this dummy measures the average portfolio asset flow intoindividual euro-area countries for all euro-area investors that comesin addition to the average flow observed for all investors. Accord-ingly, it quantifies the average financial integration effect of EMUfor the individual euro-area member state.7

6For example, Europe’s life insurance companies could not hold more than 20percent of their assets in foreign currencies unless they were matched by liabili-ties denominated in the same currency. As the vast majority of those liabilitieswere denominated in national currency, so were most of the assets. Quantitativerestrictions are also typical for pension funds. For a description of the restric-tions in the EU before EMU, see International Monetary Fund (1997, table 63,p. 213).

7A further complication comes from the role of the London market as a majorintermediary of foreign investments from and to the rest of the world. Due tothe large size and higher sophistication of the London markets, many investorschoose to make their foreign investments via the United Kingdom. For exam-ple, a Japanese investor may choose to select a British investment managerto invest in euro-area equities and bonds. The IMF data on portfolio holdingsreport an accurate country breakdown of bilateral investment, which tries to iden-tify the residence of the issuer. Nevertheless, since the city of London is a keyEuropean player, we control for that by including two additional dummies. Thefirst dummy takes a value of 1 if the receiving country is the United Kingdom.A second dummy takes the value of 1 if the investing country belongs to theEMU and the receiving country is the United Kingdom.


3. Data: The Geography of International Portfolio Flows

The IMF Coordinated Portfolio Investment Survey (CPIS) databasereports the portfolio positions of international investors, excludingthe official holdings of monetary authorities disaggregated by regionsand instruments. More specifically, the CPIS data set provides ageographical breakdown of international portfolio holdings disag-gregated by three instruments—equity securities, long-term debtsecurities, and short-term debt securities—and includes virtually allmajor international investment, excluding foreign direct investment.An additional advantage of this data set is the consistency of thecompilation criteria:

• Participants undertake a benchmark portfolio asset survey atthe same time.

• Participants follow definitions and classifications that aremutually consistent by following the methodology set out inthe fifth edition of the IMF Balance of Payments Manual.

• All participants provide a breakdown of their stock of port-folio investment assets by the country of residency of thenon-resident issuer.

There are also several issues related to the CPIS database: (i)data on the rather poor countries is missing; (ii) coverage is incom-plete due to lack of important portfolio holders, such as Middle Eastcountries and China; (iii) several transactions are reported as con-fidential for offshore financial centers and emerging markets; (iv)liabilities are bigger than assets, pointing toward underreporting;and (v) collection methods differ from country to country, which isan issue particularly for identifying the nationality and/or institu-tional type of the investor. However, we focus the analysis on themost important countries, which have a long experience in com-piling balance of payments data and international investment posi-tions, while at the same time encompassing all important interna-tional investment decisions. The exclusion of Middle East countries,China, and monetary authorities is desired, as they have been pur-chasing securities to control developments in the exchange rate orstrategic assets. Both objectives are not related to the mean-varianceoptimization.


The CPIS database for the year 1997 covers twenty-nine of thelargest economies in the world, nine of which belong to the euroarea—Austria, Belgium, Finland, France, Ireland, Italy, Nether-lands, Portugal, and Spain; the three old EU member states butnot members of the euro area—Denmark, Sweden, and the UnitedKingdom; another ten developed countries—Australia, Bermuda,Canada, Iceland, Israel, Japan, New Zealand, Norway, Singapore,and the United States; four Asian emerging markets—Indonesia,Korea, Malaysia, and Thailand; and three Latin American emergingmarkets—Argentina, Chile, and Venezuela. Germany did not reportdata in 1997, but did so in 2001. Since Germany is a key euro-areamember and its international portfolio holdings are substantial, weused an annual database on international investment positions fromthe Bundesbank to derive the geographical allocation of equities andbonds and notes positions abroad held by German residents at end-1997. Specifically, we use the Bundesbank 1997 and 2001 recordsand adjust all the 1997 positions consistently (including exchangerate movements) to make them comparable to the 2001 holdingsrecorded in the CPIS. In all, we employ a matrix formed by thirtycountries (that is 870 observations). However, we excluded from thedatabase the investing countries that allocated explicitly to specificreceiving countries less than 75 percent of their international port-folio either in 1997 or in 2001, or those countries that held less than100 million in U.S. dollars in their international portfolio in 1997or in 2001. In other words, we opted for excluding from the cross-section analysis economies that reported undetermined investmentpositions vis-a-vis the countries in the sample.8

The change in foreign holdings from end-1997 to end-2001 couldbe due to capital gains, exchange rate changes, portfolio trans-actions, and other adjustments (i.e., reclassification). Under thehypothesis that cross-border other adjustments are relatively neg-ligible, the actual portfolio flows from the investing country c tothe receiving country k over the period 1998–2001, Tck,t, can becomputed using the IMF data model widely employed in the field of

8We excluded the investment of Argentina, Indonesia, Israel, Thailand, andVenezuela from the equity holdings database and the investment of Iceland, Israel,and New Zealand from the long-term debt instrument holdings database. How-ever, in doing so, we simply lose, respectively, 0.05 percent and 0.20 percent ofallocated global equity and bond holdings.


balance of payments, international investment positions, and exter-nal debt statistics:9

Tck,t =(

Invc,k,01

ek,01pk,01− Invc,k,97

ek,97pk,97

)ek,tpk,t

=[

Invc,k,01

(1 + xk,t) (1 + rk,t)− Invc,k,97

]ek,tpk,t

ek,97pk,97,

where Invc,k is the amount invested by country c in country k financialassets and held in country c currency; ek and pk are, respectively, theexchange rate (i.e., country c currency per unit of country k currency)and the asset price in country k financial assets at the end of 1997and at the end of 2001; xk,t and rk,t are, respectively, the change inthe exchange rate and the total asset return over the four-year period;and ek,t and pk,t are, respectively, the average exchange rate and theaverage asset price over the same period. This approach implies thattransactions are assumed to occur uniformly over the period 1998–2001. However, it facilitates the computation of the bilateral cross-border portfolio flows, as they do not depend on the choice of the priceindices’ base year.

As for the exchange rate adjustments, one should note that mostof the global portfolio allocation is in U.S. dollars and euros. Accord-ing to a subtotal of thirteen countries used in this study, 75 percentof equity portfolio and 80 percent of long-term debt portfolio are heldin these two currencies. International investors also held less than10 percent of their portfolios in British pounds and Japanese yen.Therefore, it would be a mistake to use local currencies in estimat-ing cross-border portfolio flows. Moreover, London is a key Europeanfinancial center generally issuing European assets in euros. Since theJapanese yen depreciated by only 1 percent against the U.S. dollarover the 1998–2001 period, and given the lack of a disaggregated cur-rency and geographical breakdown of portfolio holdings, a potentialalternative is to assume that all assets issued by European countriesare issued in euros, while all assets issued by non-European coun-tries are issued in U.S. dollars. This assumption finds its support inthe European Central Bank’s studies on the international role of theeuro, which point out that (i) the use of the euro in international

9See Committeri (2000) for a comprehensive analysis.


markets has a strong regional dimension, as it is focused on countriesand financial centers geographically close to the euro area, and (ii)the city of London plays a pivotal role regarding the use of the eurooutside the euro area (European Central Bank 2003).

Table 4 reports the estimated cross-border portfolio flows aggre-gated for the thirty countries in the sample and the aggregate cumu-lated figures reported by the International Financial Statistics (IFS)of the IMF over the same period. The estimated flows and the IFSfigures are not directly comparable, as important countries such asoffshore centers and several Asian and Latin American countries arenot included in the CPIS database. Moreover, the cross-border flowsof debt instruments reported by the IFS include official flows frommonetary authorities and are the sum of bond and money marketinstrument flows. For example, the estimated foreign capital flows inU.S. long-term debt securities amount to USD 173 billion, while theU.S. debt liabilities reported by the IFS, which include the flows offoreign monetary authorities and offshore centers, amount to USD869 billion. However, it is useful to point out that the reserve assetsof Japan, China, and Southeast Asian countries increased by approx-imately USD 450 billion over the 1998–2001 period, and it is gener-ally agreed that the Asian monetary authorities purchased mostlyU.S. Treasury securities. It is also generally believed that monetaryauthorities might have a different profit-maximizing behavior thanprivate investors. Hence, the exclusion of their investment decisions,which is implicit when using the CPIS database, is cardinal to testthe IntCAPM.

Although differences for individual transactions clearly exist,the cross-section is acceptable. In fact, the correlation coefficientsbetween the estimated and the IFS figures amount to 95 percentand 88 percent on the asset side of equity and bond securities, andto 90 percent and 78 percent on their respective liability side.

The aggregate results for four country groupings show that port-folio flows of euro-area member states were substantially largercompared with other regions, particularly in the bond market (seetable 5). A deeper inspection of the data reveals, first, that allregions of the world purchased euro-area assets over the period 1998–2001 (see figure 1). Second, intra-euro-area allocation was extremelyhigh. The portfolio transactions among euro-area member statesamounted to USD 315 billion in equity securities and USD 754


Tab

le4.

Est

imat

edan

dIF

SPor

tfol

ioFlo

ws

by

Cou

ntr

y,19

98–2

001

(USD

Million

s)

Est

imat

edU

sing

CP

IS(T

hir

tyC

ountr

ies)

IFS

(All

Countr

ies)

Equity

Bonds

Equity

Deb

tIn

stru

men

ts

Ass

etLia

bility

Ass

etLia

bility

Ass

etLia

bility

Ass

etLia

bility

Countr

ies

(1a)

(2a)

(3a)

(4a)

(1b)

(2b)

(3b)

(4b)

Arg

enti

na72

18−

1155

0−

156

−35

509

3057

−14

178

1894

9046

Aus

tral

ia26

894

−44

9742

10−

5550

2133

528

269

1391

934

153

Aus

tria

1694

6−

1668

2993

147

256

2591

220

3452

613

8933

0B

elgi

um32

027

1144

467

878

4369

4N

AN

AN

AN

AB

erm

uda

−28

958

5757

420

7724

9N

AN

AN

AN

AC

anad

a66

765

−11

080

−28

79−

2338

887

687

4631

810

347

7349

Chi

le17

39−

2394

1403

3372

6646

024

6043

58D

enm

ark

2310

5−

605

1212

6−

2865

2446

431

8423

611

2012

2Fin

land

1642

346

761

2649

3−

3564

1994

332

284

3017

682

33Fr

ance

9688

695

292

2045

0415

9606

9693

113

0205

3181

0128

6335

Ger

man

y16

8440

8782

316

6482

1963

1931

2046

1344

1131

7931

3735

99Ic

elan

d12

0212

461

1197

1340

6914

228

36In

done

sia

215

2845

6−

4435

NA

−57

320

−92

Irel

and

8789

444

127

1071

9926

771

1166

3724

8967

2273

3640

136

Isra

el11

1933

9729

9914

4814

3069

6635

86−

94It

aly

1031

3217

1214

2121

7951

1768

7172

1517

8246

2957

27Ja

pan

3070

310

8697

8991

8−

7570

7740

315

7815

3623

9413

3065

Sout

hK

orea

111

2226

0−

2066

−34

398

1201

3928

855

58−

5681

Mal

aysi

a−

401

−77

73−

9−

6245

NA

NA

028

3

(con

tinu

ed)


Tab

le4.

(Con

tinued

)

Est

imat

edU

sing

CP

IS(T

hir

tyC

ountr

ies)

IFS

(All

Countr

ies)

Equity

Bonds

Equity

Deb

tIn

stru

men

ts

Ass

etLia

bility

Ass

etLia

bility

Ass

etLia

bility

Ass

etLia

bility

Countr

ies

(1a)

(2a)

(3a)

(4a)

(1b)

(2b)

(3b)

(4b)

Net

herl

ands

8077

180

545

1123

5518

3460

1255

7064

234

1648

4119

5377

New

Zea

land

2071

−33

2319

44−

1362

3482

−33

211

8715

34N

orw

ay31

222

2658

2908

393

1534

820

3296

3657

420

665

Por

tuga

l30

4137

4411

857

2483

550

8856

4220

095

2221

9Si

ngap

ore

9342

1591

630

799

8844

3446

127

3912

749

801

Spai

n40

517

2949

065

833

7951

863

973

4697

213

2118

1007

06Sw

eden

4350

3−

1246

319

732

−75

6259

076

1143

931

101

1182

1T

haila

nd−

7930

3320

8−

6708

NA

2486

505

−33

29U

nite

dK

ingd

om−

1075

334

4075

5939

873

361

1209

2239

1493

1885

4715

2808

Uni

ted

Stat

es28

4130

2331

16−

6828

017

3220

4315

0646

9311

3948

686

8766

Ven

ezue

la1

−12

50−

930

−13

873

1761

−17

863

2Tot

al11

3501

511

3501

510

9404

810

9404

818

6125

818

1491

421

7533

726

7070

5

Cor

r.C

oeffi

cien

t(1

):0.

95(2

):0.

91(3

):0.

88(4

):0.

78

Note

s:T

he

firs

tfo

ur

colu

mns

ofth

eta

ble

repor

tth

ere

sult

sof

the

esti

mat

ednet

asse

ttr

ansa

ctio

ns

over

the

cum

ula

ted

per

iod

1998

–200

1ag

greg

ated

for

the

thir

tyco

unt

ries

ofth

esa

mple

.T

he

last

four

colu

mns

repor

tth

eco

unt

ries

’eq

uity

and

deb

tflow

sov

erth

esa

me

per

iod

repor

ted

byth

eIF

Sdat

abas

eof

the

IMF.T

he

esti

mat

edan

dth

eIF

Sfigu

res

are

not

dir

ectl

yco

mpar

able

,as

offsh

ore

cent

ers,

seve

ralA

sian

and

Lat

inA

mer

ican

count

ries

,an

dth

eoffi

cial

flow

sof

mon

etar

yau

thor

itie

sar

enot

incl

uded

inth

eC

PIS

dat

abas

e.M

oreo

ver,

the

deb

tin

stru

men

tflow

sre

por

ted

byth

eIF

Sar

eth

esu

mof

bon

ds

and

mon

eym

arke

tin

stru

men

tflow

s.


Tab

le5.

Est

imat

edan

dIF

SPor

tfol

ioFlo

ws

by

Reg

ions,

1998

–200

1(U

SD

Billion

s)

Est

imat

edfr

om

CP

IS(T

hir

tyC

ountr

ies)

IFS

(All

Countr

ies)

Equity

Bonds

Equity

Deb

tIn

stru

men

ts

Ass

etLia

bility

Ass

etLia

bility

Ass

etLia

bility

Ass

etLia

bility

Countr

ies

(1a)

(2a)

(3a)

(4a)

(1b)

(2b)

(3b)

(4b)

Ten

EM

U64

6.1

397.

691

4.0

975.

894

3.0

672.

014

41.5

1411

.7T

hree

Non

-EM

UE

U55

.933

1.0

91.3

62.9

204.

540

6.1

243.

318

4.8

Ten

Non

-EU

424.

540

2.6

89.9

156.

469

3.5

714.

548

0.4

1069

.1D

evel

oped

Seve

nE

mer

ging

8.6

3.9

−1.

1−

100.

111

.522

.410

.25.

2M

arke

tsTot

al11

35.0

1135

.010

94.0

1094

.018

61.3

1814

.918

05.1

2337

.6

Cor

r.C

oeffi

cien

t(1

):0.

99(2

):0.

96(3

):0.

97(4

):0.

84

Note

s:T

he

figu

res

repor

tth

epor

tfol

ioflow

sof

equity

and

fixe

d-inco

me

aggr

egat

edfo

rfo

ur

count

rygr

oupin

gs.T

he

ten

EM

Uco

unt

ries

are

Aust

ria,

Bel

gium

,Fin

land,

Fra

nce

,G

erm

any,

Irel

and,

Ital

y,N

ether

lands,

Por

tuga

l,an

dSpai

n.

The

thre

enon

-EM

UEU

count

ries

are

Den

-m

ark,

Sw

eden

,an

dth

eU

nit

edK

ingd

om.T

he

ten

non

-EU

dev

elop

edco

unt

ries

incl

ude

Aust

ralia,

Ber

muda,

Can

ada,

Icel

and,Is

rael

,Ja

pan

,N

ewZea

land,

Nor

way

,th

eU

nit

edSta

tes,

and

Sin

gapor

e.T

he

seve

nem

ergi

ng

mar

kets

are

Arg

enti

na,

Chile,

Indon

esia

,K

orea

,M

alay

sia,

Thai

land,

and

Ven

ezuel

a.T

he

firs

tfo

ur

colu

mns

ofth

eta

ble

repor

tth

ere

sult

sof

the

esti

mat

ednet

asse

ttr

ansa

ctio

ns

over

the

cum

ula

ted

per

iod

1998

–200

1ag

greg

ated

for

the

thir

tyco

unt

ries

ofth

esa

mple

.T

he

last

four

colu

mns

repor

tth

eco

unt

ries

’eq

uity

and

deb

tflow

sov

erth

esa

me

per

iod

repor

ted

byth

eIF

Sdat

abas

eof

the

IMF.T

he

esti

mat

edan

dth

eIF

Sfigu

res

are

not

dir

ectl

yco

mpar

able

,as

offsh

ore

cent

ers,

seve

ralA

sian

and

Lat

inA

mer

ican

count

ries

,an

dth

eoffi

cial

flow

sof

mon

etar

yau

thor

itie

sar

enot

incl

uded

inth

eC

PIS

dat

abas

e.M

oreo

ver,

the

deb

tin

stru

men

tflow

sre

por

ted

byth

eIF

Sar

eth

esu

mof

bon

dan

dm

oney

mar

ket

inst

rum

ent

flow

s.


Figure 1. Changes in Non-Domestic Euro-AreaAssets by Region

Notes: Panel A reports the estimated portfolio net flows of EMU assets transactedby residents of region i over the period 1998–2001 aggregated for four country group-ings. Panel B reports the estimated flows relative to foreign assets held in region iover the average period 1998–2001. The ten EMU countries are Austria, Belgium,Finland, France, Germany, Ireland, Italy, Netherlands, Portugal, and Spain. Thethree non-EMU EU countries are Denmark, Sweden, and the United Kingdom.The ten non-EU developed countries include Australia, Bermuda, Canada, Iceland,Israel, Japan, New Zealand, Norway, Singapore, and the United States. The sevenemerging markets are Argentina, Chile, Indonesia, Korea, Malaysia, Thailand, andVenezuela.


billion in bonds and notes (see figure 1A), which represent, respec-tively, 31 percent and 51 percent of the non-domestic equity andbond assets held on average by euro-area member states over theperiod 1998–2001 (see figure 1B).

Several transactions are not reported in the CPIS database. Thisimplies either a rational portfolio allocation decision of not investingor the desire of not reporting confidential transactions. Specifically,the zeros reported by countries with more than 75 percent of theirinternational portfolio allocated account for 6.7 percent and 16.6 per-cent in equity and bond portfolios, respectively. It is important tostress that the zero flows can still occur if there were no transactionsor if transactions just offset each other during the 1997–2001 period.Also, negative transactions can occur if countries disinvested in 2001relative to 1997. Such zero and negative flows are not truncated; theyremain part of the database. However, given the zero reporting forsome holding positions, we estimate the same model, excluding andincluding such transactions. As shown in various regressions, resultsare barely affected, partly suggesting that most of the zero holdingsare based on actual portfolio investment decisions.

4. The Empirical Approach: Testing the IntCAPM

In a fully integrated world where purchasing power parity (PPP)holds, Solnik (1974) and Sercu (1980) show that the internationalversion of the simple CAPM of Sharpe (1964) and Lintner (1965)holds. Moreover, the equilibrium is achieved when all investorshold the world market portfolio, where each country’s portfolio isweighted by its market capitalization. In this model, the optimalshare invested in each country k is equal to that country k’s marketcapitalization weight in the world index portfolio, wBench

k,97 .The IntCAPM predicts a coefficient on wBench

k,97 equal to unity.However, due to home bias, marginal and fixed transaction costs,asymmetric information, heterogeneous belief about market perfor-mance, and trend-chasing behavior, the prediction of the IntCAPMmight not hold. Therefore, first we regress bilateral portfolio flowsin equity and bond securities upon their respective benchmarks,wBench

k,97 . Next, we control for the degree of home bias, the non-lineardegree of misallocation at the beginning of the period, and assetperformance in the previous period. Specifically, we estimate thefollowing model:


tck,t = α0 + α1wBenchk,97 + α2HBc,97 + α3ΔHBc,01

+ α4DW 3ck,97 + α5rk,t−1 + εt,

where tck,t denotes country c’s international transactions invested incountry k divided by country c’s average international holdings overthe 1998–2001 period; HBc,97 and ΔHBc,01 are country c’s degreeof home bias at the beginning of the period and its first differenceover the sample period; DWck,97 denotes the degree of underweightof country k assets in investor c’s portfolio at the beginning of theperiod; rk,t−1 is the total returns on country k’s market portfolioin the previous period; and εt is a well-behaved term for all otherdeterminants of portfolio asset flows.

Home bias on the part of an investor, HBc,t, is broadly definedas the tendency to invest more in domestic assets even though therisk is shared more effectively if foreign assets are held. We expectthat the higher the degree of home bias, the larger the benefits offurther cross-border investments and the stronger the incentives forinternational diversification. Hence, net portfolio flows should bepositively related to the degree of home bias at the beginning of theperiod (α2 > 0) and negatively related to its increase over the period(α3 < 0). An index that is generally used to measure home bias is 1minus the Foreign Asset Acceptance Ratio (FAAR).10 FAAR meas-ures the extent to which the share of foreign assets in an investor’sportfolio diverges from the share of foreign assets that would be heldin a “borderless” global portfolio. By this metric, home bias is higherthe lower FAAR is from unity. Specifically, FAAR is computed as theactual share of foreign assets in total country holdings, wf

c,t, dividedby the optimal share of foreign assets in the total country portfolio,1 − wBench

c,t . This implies that

HBc,t = 1 −wf

c,t

1 − wBenchc,t

.

Since this measure is investor specific, it also plays the role of countryc’s fixed effect.

10See, for example, International Monetary Fund (2005) and De Santis andGerard (2009).


The degree of underweight, DWck, is defined as the differencebetween the optimal and actual share of country k assets in investorc’s portfolio. It is generally agreed that the higher the costs in a par-ticular foreign market, the more severely underweighted that countrywill be in the investor’s portfolio (Ahearne, Griever, and Warnock2004). Therefore, we use the initial degree of bilateral misallocationpartly to instrument the role played by direct and indirect costs andasymmetric information on bilateral cross-border asset transactions.The larger the initial difference between optimal and actual share,the stronger the incentive to learn about the country and to reducethe associated asset allocation costs in order to trade back to opti-mal weights, reducing the position when the actual weight exceedsthe optimal weight and increasing the investment in an asset whenit is underweighted. Since our data focus exclusively on the foreignholdings of each country, the optimal weight to be invested in coun-try k by country c is equal to country k’s market capitalization inthe world market index, excluding the investing country c. Then

DWck,97 = wBenchck,97 − wck,97,

where wBenchck,97 = wBench

k,97

1−wBenchc,97

and wck,97 = Invc,k,97∑k Invc,k,97

.Since rebalancing a portfolio entails both direct and indirect fixed

transaction costs, it is unlikely to take place when bilateral actualportfolio weights differ only slightly from bilateral optimal portfolioweights. Therefore, we introduce some non-linearities by taking thecube of this measure. We expect that the degree of underweight atthe beginning of the period affects non-linearly and positively thegeography of portfolio flows (α4 > 0).11

11Ahearne, Griever, and Warnock (2004) and Portes and Rey (2005) pointedout that direct and indirect costs, such as information costs, to trade assets ina particular foreign market are key reasons why that country is more severelyunderweighted in the investor’s portfolio. While these factors undoubtedly influ-ence firms’ decisions about where to invest, a comprehensive evaluation of thismotivation is well beyond the scope of this paper, because we do not have meas-ures of bilateral cross-border fixed costs on equity and bond allocation, whichoften take the form of legal barriers and restrictive regulations. However, oncecontrolling for home bias, the initial non-linear degree of bilateral misalloca-tion could proxy at least partly the initial fixed costs on cross-border portfolioallocation.


If portfolio decisions are based partly on past returns, theninvestors might tend to underweight countries whose stock marketshave performed poorly. Bohn and Tesar (1996) found that inter-national portfolio flows co-move with lagged measures of expectedreturns. This suggests that international investors engage in positive-feedback trading, also called “trend chasing.” To capture this type of“returns-chasing” behavior a la Bohn and Tesar, we use past returnsand expect α5 to be positive.

The results reported in table 6 indicate that wBenchk,97 is statisti-

cally significant, but its coefficient is less than unity: it is equal to0.4 for equities and 0.2 for bonds (see specification 1). A countrythat sees its market size increase by 1 percentage point relative tothe world market capitalization would attract international equity(bond) flows amounting to 0.4 percent (0.2 percent) of the equity(bond) assets held abroad by foreigners. The finding that the coeffi-cient on the bond benchmark is half that on the equity benchmarkmight imply that global equity markets are more integrated thanglobal bond markets.

The results do not change when we control for home bias (seespecification 2). On average, the decline in home bias in country cincreases international investment toward all destination countriesk in both models.

The results also do not change when we control for the non-lineardegree of underweight and past performance. Interestingly, the coef-ficient on DWck,97 is not statistically significant when taking thelinear measure (not reported). Conversely, it is strongly statisticallysignificant for the bond flow model when taking the non-linear meas-ure. One potential interpretation is the initial fixed-cost argumentdiscussed above. The willingness to close the initial gap betweenthe share of foreign assets that would be held in a “borderless”global portfolio and actual foreign investment weights is an impor-tant determinant of bond flows, as it raises the adjusted R2 by 7 per-centage points, from 14.6 percent in specification 2 to 21.9 percentin specification 3.

Finally, past performance in the destination country is statisti-cally significant for the equity and bond flow models (see specifi-cation 4a). While no prudent investor assumes future returns willmirror past returns, trend-chasing behaviors still characterize thegeography of international equity and bond flows in the long term.

Tab

le6.

Inte

rnat

ional

Por

tfol

ioA

sset

Flo

ws

and

IntC

AP

M

t ck

,t=

α0

+α

1w

Ben

chck

,97

+α

2H

Bc,9

7+

α3Δ

HB

c,0

1+

α4D

W3 ck

,97

+α

5r k

,t−

1+

ε t

Equity

Net

Ass

etFlo

ws

Bond

Net

Ass

etFlo

ws

Spec

if.1

Spec

if.2

Spec

if.3

Spec

if.4a

Spec

if.4b

Spec

if.1

Spec

if.2

Spec

if.3

Spec

if.4a

Spec

if.4b

Cst

0.00

5∗∗∗

−0.

036

−0.

036∗

−0.

039∗

∗−

0.03

8∗∗

0.01

2∗∗∗

−0.

011∗

∗−

0.01

1∗∗

−0.

023∗

∗∗−

0.01

7∗∗∗

(0.0

02)

(0.0

19)

(0.0

02)

(0.0

19)

(0.0

19)

(0.0

02)

(0.0

06)

(0.0

06)

(0.0

06)

(0.0

05)

wBen

chk

,97

0.42

7∗∗∗

0.42

5∗∗∗

0.44

9∗∗∗

0.44

1∗∗∗

0.44

7∗∗∗

0.19

0∗∗∗

0.19

3∗∗∗

0.23

3∗∗∗

0.23

2∗∗∗

0.23

4∗∗∗

(0.0

95)

(0.0

92)

(0.1

28)

(0.1

27)

(0.1

27)

(0.0

67)

(0.0

67)

(0.0

61)

(0.0

61)

(0.0

61)

HB

c,9

70.

038

0.03

80.

038

0.03

70.

015∗

∗0.

016∗

∗0.

017∗

∗0.

012∗

(0.0

24)

(0.0

24)

(0.0

24)

(0.0

23)

(0.0

08)

(0.0

08)

(0.0

08)

(0.0

08)

ΔH

Bc,0

1−

0.11

1∗∗∗

−0.

111∗

∗∗−

0.11

2∗∗∗

−0.

105∗

∗∗−

0.07

4∗∗∗

−0.

068∗

∗∗−

0.06

9∗∗∗

−0.

059∗

∗∗

(0.0

27)

(0.0

27)

(0.0

27)

(0.0

26)

(0.0

13)

(0.0

12)

(0.0

13)

(0.0

13)

DW

3 ck

,97

−0.

460

−0.

491

−0.

509

1.38

4∗∗∗

1.38

6∗∗∗

1.39

1∗∗∗

(1.1

49)

(1.1

47)

(1.1

46)

(0.2

85)

(0.2

82)

(0.2

84)

r k,t

−1

0.00

6∗∗∗

0.00

5∗∗∗

0.03

9∗∗∗

0.03

0∗∗∗

(0.0

02)

(0.0

02)

(0.0

07)

(0.0

06)

Adj

ust

edR

20.

283

0.31

50.

315

0.31

70.

319

0.10

40.

146

0.21

90.

235

0.23

7F

−S

tat

263.

5510

2.92

77.6

363

.06

68.9

274

.85

37.3

745

.65

40.1

549

.53

Note

s:T

his

table

repor

tsth

ere

sult

sof

the

cros

s-se

ctio

nal

regr

essi

onof

the

equity

and

bon

dnet

asse

tflow

s.T

he

expla

nat

ory

vari

able

sar

edes

crib

edin

appen

dix

1.T

he

sam

ple

size

for

spec

ifica

tion

s1–

4ais

n=

667

for

equity

flow

san

dn

=63

9fo

rbon

dflow

san

dex

cludes

all

zero

hol

din

gs.T

he

sam

ple

size

for

spec

ifica

tion

4bis

n=

725

for

equity

flow

san

dn

=78

3fo

rbon

dflow

san

din

cludes

allze

rohol

din

gs.O

LS,W

hit

ehet

eros

kedas

tici

ty-c

onsi

sten

tst

andar

der

rors

,an

dco

vari

ance

.Sta

ndar

der

rors

for

the

coeffi

cien

tsar

ere

por

ted

inpar

enth

eses

.**

*,**

,an

d*

den

ote

stat

isti

calsi

gnifi

cance

at1,

5,an

d10

per

cent

,re

spec

tive

ly.


All in all, the proportion of the total variation explained by themodel is quite significant: 31.7 percent in equity flows and 23.5percent in bond flows.

The results are barely affected when treating the zero report-ing for some holding positions in the sample as actual portfolioinvestment decisions (see specification 4b).

5. Portfolio Flows, Monetary Policy Frameworks,and EMU

The second aim of the paper is to assess whether central banks’monetary policy frameworks across countries influenced the geog-raphy and size of international capital flows during the boom andbust of asset prices at the turn of the century. In this context wealso investigate the role of EMU on global portfolio reallocation.

5.1 The Impact of Monetary Policy Frameworks

We use the IntCAPM as our benchmark and add explanatory vari-ables describing key characteristics of monetary policy frameworksacross countries. Therefore, we estimate the following model:

tck,t = α0 + α1wBenchk,97 + α2HBc,97 + α3ΔHBc,01 + α4DW 3

ck,97

+ α5rk,t−1 + λ′Monk,t−1 + εt,

where the vector Monk,t−1 includes variables measuring the follow-ing characteristics of the monetary policy framework in country k:(i) the central bank’s objectives, such as the focus on inflation, onmoney, on the exchange rate, and on discretionary policy, (ii) thecentral bank’s institutional factors, such as independence, account-ability, and transparency, and (iii) the importance of financial stabil-ity in the setting of monetary policy instruments. Results reportedin table 7 are based on the sample that excludes the zero entrypositions. The results are the same when zero holding positions areincluded in the analysis (not reported; see also table 8).

Among the characteristics related to monetary policy, the focuson money played a significant and positive role in influencing thegeography of the portfolio flows. Given the link between asset pricesand credit growth (Borio and Lowe 2004; Detken and Smets 2004),

Vol. 6 No. 2 The Geography of International Portfolio Flows 175Tab

le7.

Inte

rnat

ional

Por

tfol

ioFlo

ws

and

Char

acte

rist

ics

ofM

onet

ary

Pol

icy

Fra

mew

orks

t ck,t

=α

0+

α1w

Ben

chk,9

7+

α2H

Bc,9

7+

α3Δ

HB

c,0

1+

α4D

W3 ck,9

7+

α5r k

,t−

1+

λ′ M

onk,t

−1

+ε t

Spec

if.

Expla

nat

ory

Equity

Flo

ws

Bon

dFlo

ws

Num

ber

Var

iable

Coeff

.s.

e.R

2C

oeff

.s.

e.R

2

Pol

icy

Focu

s

5In

flation

k,9

80.

004

(0.0

06)

0.31

7−

0.01

1∗∗

(0.0

04)

0.23

76

Mon

eyk,9

80.

021∗

∗∗(0

.007

)0.

322

0.06

2∗∗∗

(0.0

12)

0.28

67

Exc

hang

era

tek,9

8−

0.00

6(0

.006

)0.

309

0.00

5(0

.005

)0.

234

8D

iscr

etio

n k,9

80.

003

(0.0

07)

0.31

70.

010

(0.0

08)

0.23

6

Inst

itut

iona

lFa

ctor

s

9In

depe

nden

cek,9

80.

003

(0.0

14)

0.31

70.

046∗

∗∗(0

.016

)0.

240

10Acc

ount

abili

tyk,9

80.

006

(0.0

08)

0.31

7−

0.02

5∗∗∗

(0.0

09)

0.24

511

Tra

nspa

renc

y k,9

80.

006

(0.0

08)

0.31

7−

0.00

7(0

.006

)0.

234

Stru

ctur

alFa

ctor

sk

12Fin

anci

alSt

abili

tyk,9

80.

014

(0.0

09)

0.31

8−

0.00

6(0

.008

)0.

233

Pol

icy

Focu

sw

ith

Ger

man

yB

eing

the

Ben

chm

ark

for

Oth

erE

uro-

Are

aC

ount

ries

5aIn

flation

k,9

80.

003

(0.0

07)

0.31

7−

0.01

4∗∗∗

(0.0

04)

0.23

96a

Mon

eyk,9

80.

009∗

∗(0

.004

)0.

319

0.03

0∗∗∗

(0.0

04)

0.27

57a

Exc

hang

era

tek,9

8−

0.01

1∗∗∗

(0.0

04)

0.31

9−

0.01

8∗∗∗

(0.0

04)

0.24

08a

Dis

cret

ion k

,98

−0.

004

(0.0

14)

0.31

3−

0.01

6∗∗

(0.0

07)

0.23

6(c

ontinu

ed)


Tab

le7.

(Con

tinued

)

Spec

if.

Expla

nat

ory

Equity

Flo

ws

Bon

dFlo

ws

Num

ber

Var

iable

Coeff

.s.

e.R

2C

oeff

.s.

e.R

2

Inst

itut

iona

lFa

ctor

sw

ith

Ger

man

yB

eing

the

Ben

chm

ark

for

Oth

erE

uro-

Are

aC

ount

ries

9aIn

depe

nden

cek,9

80.

003

(0.0

12)

0.31

70.

063∗

∗∗(0

.014

)0.

247

10a

Acc

ount

abili

tyk,9

8−

0.00

0(0

.006

)0.

317

−0.

029∗

∗∗(0

.005

)0.

261

11a

Tra

nspa

renc

y k,9

80.

012

(0.0

09)

0.31

8−

0.00

6(0

.007

)0.

234

Stru

ctur

alFa

ctor

sw

ith

Ger

man

yB

eing

the

Ben

chm

ark

for

Oth

erE

uro-

Are

aC

ount

ries

12a

Fin

anci

alSt

abili

tyk,9

80.

010

(0.0

09)

0.31

7−

0.00

6(0

.007

)0.

234

Note

s:T

his

table

repor

tsth

ere

sult

sof

the

cros

s-se

ctio

nal

regr

essi

onof

por

tfol

ionet

asse

tflow

s.Eac

hex

pla

nat

ory

vari

able

ofth

ista

ble

isad

ded

asan

addit

ional

regr

esso

rto

the

IntC

APM

repor

ted

insp

ecifi

cati

on4

ofta

ble

6.A

sfo

rth

ech

arac

teri

stic

sof

the

mon

etar

ypol

icy

fram

ewor

k,th

ein

dic

esra

nge

bet

wee

n0

and

1,an

da

hig

her

scor

eis

asso

ciat

edw

ith

the

hig

her

deg

ree

ofim

por

tance

give

nto

pol

icy

obje

ctiv

es,

toin

stit

uti

onal

char

acte

rist

ics,

and

tofinan

cial

stab

ility

issu

es.

The

expla

nat

ory

vari

able

sar

edes

crib

edin

appen

dix

1.T

he

sam

ple

size

isn

=66

7fo

req

uity

flow

san

dn

=63

9fo

rbon

dflow

san

dex

cludes

allze

rohol

din

gs.O

LS,W

hit

ehet

eros

kedas

tici

ty-c

onsi

sten

tst

andar

der

rors

,an

dco

vari

ance

.Sta

ndar

der

rors

for

the

coeffi

cien

tsar

ere

por

ted

inpar

enth

eses

.**

*,**

,an

d*

den

ote

stat

isti

calsi

gnifi

cance

at1,

5,an

d10

per

cent

,re

spec

tive

ly.

Tab

le8.

Inte

rnat

ional

Por

tfol

ioFlo

ws,

IntC

AP

M,an

dM

onet

ary

Pol

icy

Fra

mew

orks

t ck

,t=

α0

+α

1w

Ben

chck

,97

+α

2H

Bc,9

7+

α3Δ

HB

c,0

1+

α4D

W3 ck

,97

+α

5r k

,t−

1+

λM

onk

,t−

1+

ε t

Equity

Ass

etFlo

ws

Bond

Ass

etFlo

ws

Spec

if.13

Spec

if.14a

Spec

if.14b

Spec

if.13

Spec

if.14a

Spec

if.14b

Cst

−0.

043∗

∗(0

.019

)−

0.04

3∗∗

(0.0

19)

−0.

042∗

∗(0

.019

)−

0.02

8∗∗∗

(0.0

08)

−0.

027∗

∗∗(0

.008

)−

0.02

0∗∗

(0.0

07)

wBen

chk

,97

0.43

3∗∗∗

(0.1

27)

0.43

6∗∗∗

(0.1

28)

0.44

3∗∗∗

(0.1

27)

0.23

7∗∗∗

(0.0

60)

0.21

4∗∗∗

(0.0

62)

0.21

3∗∗∗

(0.0

62)

HB

c,9

70.

037

(0.0

24)

0.03

7(0

.024

)0.

036

(0.0

23)

0.01

6∗∗

(0.0

08)

0.01

5∗(0

.008

)0.

012∗

(0.0

07)

ΔH

Bc,0

1−

0.11

2∗∗∗

(0.0

27)

−0.

112∗

∗∗(0

.027

)−

0.10

5∗∗∗

(0.0

26)

−0.

071∗

∗∗(0

.011

)−

0.07

2∗∗∗

(0.0

11)

−0.

060∗

∗∗(0

.008

)D

W3 ck

,97

−0.

516

(1.1

43)

−0.

519

(1.1

45)

−0.

536

(1.1

43)

1.36

3∗∗∗

(0.2

87)

1.36

6∗∗∗

(0.2

75)

1.37

7∗∗∗

(0.2

75)

r k,t

−1

0.00

8∗∗∗

(0.0

02)

0.00

8∗∗

(0.0

03)

0.00

6∗∗∗

(0.0

02)

0.03

3∗∗∗

(0.0

07)

0.06

4∗∗∗

(0.0

12)

0.05

4∗∗∗

(0.0

10)

Mon

eyk

,t98

0.02

1∗∗∗

(0.0

07)

0.02

1∗∗∗

(0.0

07)

0.02

1∗∗∗

(0.0

06)

0.02

6∗∗

(0.0

05)

0.05

9∗∗∗

(0.0

12)

0.05

4∗∗∗

(0.0

11)

Inflat

ion k

,98

——

——

——

0.00

7∗(0

.004

)0.

003

(0.0

04)

−0.

000

(0.0

03)

Acc

ount

abili

tyk

,98

——

——

——

−0.

012∗

∗(0

.005

)−

0.02

3∗∗∗

(0.0

07)

−0.

018∗

∗∗(0

.006

)dM

oney

k,9

80.

002

(0.0

05)

0.00

4(0

.004

)−

0.00

8(0

.017

)−

0.00

7(0

.015

)dI

nflat

ion k

,98

0.03

3∗∗

(0.0

16)

0.02

3(0

.014

)dA

ccou

ntab

ility

k,9

8−

0.03

9(0

.024

)−

0.03

2(0

.022

)Adj

uste

dR

20.

322

0.32

10.

323

0.27

60.

301

0.30

0F

−S

tat

53.7

045

.99

50.3

431

.37

25.9

430

.88

Note

s:T

his

table

repor

tsth

ere

sult

sof

the

cros

s-se

ctio

nal

regr

essi

onof

por

tfol

ionet

asse

tflow

s,su

mm

ariz

ing

findin

gsin

table

s6

and

7.T

his

table

also

repor

tsth

ere

sult

sw

hen

cont

rollin

gfo

rth

epot

enti

alch

ange

inth

ech

arac

teri

stic

sof

the

mon

etar

ypol

icy

fram

ewor

ks,as

aco

nse

quen

ceof

the

esta

blish

men

tof

EM

U,usi

ng

the

Ger

man

char

acte

rist

ics

asa

targ

etben

chm

ark

for

other

euro

-are

aco

unt

ries

.O

ther

expla

nat

ory

vari

able

sar

edes

crib

edin

appen

dix

1.T

he

sam

ple

size

for

spec

ifica

tion

s13

–14a

isn

=66

7fo

req

uity

flow

san

dn

=63

9fo

rbon

dflow

san

dex

cludes

all

zero

hol

din

gs.T

he

sam

ple

size

for

spec

ifica

tion

14b

isn

=72

5fo

req

uity

flow

san

dn

=78

3fo

rbon

dflow

san

din

cludes

allze

rohol

din

gs.O

LS,

Whit

ehet

eros

kedas

tici

ty-c

onsi

sten

tst

andar

der

rors

,an

dco

vari

ance

.Sta

ndar

der

rors

for

the

coeffi

cien

tsar

ere

por

ted

inpar

enth

eses

.**

*,**

,an

d*

den

ote

stat

isti

calsi

gnifi

cance

at1,

5,an

d10

per

cent

,re

spec

tive

ly.


global portfolio investors might have purchased assets of countrieswhere monetary analysis played a more prominent role.

If the focus on monetary aggregates, credit growth, and financialflows helps central banks identify inflated asset prices and financialimbalances, then portfolio investors would indeed give importance tothis characteristic. The empirical results on both equity and bondflows support this hypothesis. In particular, the R

2of the model

explaining the geography of bond flows increases from 23.5 percentin the benchmark model (see specification 4a of table 6) to 28.6percent when adding the focus on money by the recipient country’scentral bank (see specification 6 of table 7).

Among the characteristics related to the institutional factors,independence (specifications 9 and 9a of table 7) is found to be sig-nificant and with the expected sign only for bond flows, while beingaccountable is a push factor of bond flows (specifications 10 and10a). As already pointed out, independence and accountability arenegatively correlated, which implies that explaining and defendingmonetary policies to government and parliament and the involve-ment of parliament in monetary policy decisions might be seen byinternational investors as an infringement on central banks’ inde-pendence, putting at risk the price stability mandate, which is nowa typical mandate by monetary authorities across the globe.

Qualitatively similar results are obtained if we assume that thecharacteristics of the monetary policy frameworks of the euro-areamember states converged with those of Germany with the estab-lishment of EMU (see specifications 5a–12a of table 7). Under thishypothesis, the focus on the exchange rate becomes strongly neg-atively significant in both equity and bond flow specifications (seespecification 7a). This might imply that global asset allocators overthe period 1998–2001 reduced their investment vis-a-vis countrieswhose monetary authorities gave relative importance to exchangerate targets.

When pooling the indices, which are statistically significant inthe same model, independence is no longer significant in the bondflow specification, while the focus on inflation has the positivesign (see specification 13 of table 8). Given the establishment ofEMU in January 1999, we also control for the potential change inthe characteristics of the monetary policy frameworks of the euro-area member states using the German monetary framework as the


benchmark. Results, which are reported in specifications 14a and14b of table 8, indicate that the focus on money and accountabil-ity continue to be robust. These will be the only variables that areconsistently significant when we study and control for the effect ofEMU.12

The fact that the focus on inflation or independence has not sys-tematically influenced international capital flows may be simply dueto the fact that the core objective of most of the central banks ofthe countries in the sample is to enforce price stability, with inde-pendence being key to control inflationary expectations. Most likely,foreign investors care about inflation only at relatively high levels.

5.2 The Impact of EMU

To control for the effect of EMU on global capital flows as wellas to measure its average impact, we include two sets of binaryvariables. However, the binary variables per se are not sufficientto study the catalyst effect of EMU, because portfolio capital couldhave been reallocated globally to better exploit the expected diversi-fication benefits and the elimination of the exchange rate risk amongeuro-area member states.

In 1998, the Maastricht process was well under way and investorswere keenly aware of the high likelihood that the intra-EMU cur-rency risk would disappear. Therefore, the expected volatility of theexchange rate is assumed to be zero among euro-area countries andequal to past volatility for the other cross-border transactions. Morespecifically, the expected volatility of the exchange rate among euro-area member states and other countries of the world is assumed toequal the past volatility of the rest of the world’s currencies vis-a-visthe deutsche mark.

As a measure of the expected diversification benefit, De San-tis and Gerard (2009) suggest to employ the marginal impact onportfolio risk of increasing or decreasing the investor’s position in aparticular asset. Recall that the foreign investment portfolio variancecan be computed as

σ2Pct = w′

c,tΣc,twc,t.

12We have also controlled for the average inflation rate over the ten-year period1992–2001, and the results do not vary.


wc,t is the vector of weights for the N foreign assets and Σc thecovariance matrix of returns of the foreign assets, where the sub-script c indicates that the covariance and weight are computed fromthe investing country c’s perspective. Then, the decrease in portfoliovariance for a marginal increase in the weight invested in asset k canbe interpreted as a measure of the diversification benefit, DBc,t:

DBc,t = − ∂

∂wc,tσ2

Pct = − ∂

∂wc,t

[w′

c,tΣc,twc,t

]= −2Σc,twc,t.

That is,

DBck,t = − ∂

∂wck,t

[w′

c,tΣc,twc,t

]= −2

K∑l=1

wcl,tσlk,t,

where DBck,t measures the diversification benefit of adding asset k toinvestor c’s position. Therefore, we should expect it to be positivelyrelated to portfolio asset flows.

For an international investor, however, the return on any foreignasset varies not only because of asset-specific risk but also becauseof unpredictable fluctuations in exchange rates. Since the currency-risk exposure of asset portfolios can be hedged through derivativestransactions, it may be of interest to distinguish between the pureasset component and the currency-risk component of the diversifica-tion benefit motive of portfolio reallocation. Therefore, we considerthree measures of diversification benefits: (i) an aggregate measureof diversification benefits based on the investor’s foreign investmentsreturns denominated in his domestic currency, DBAgg

ck = DB(rck); (ii)

a measure of diversification benefits based on the investor’s foreigninvestments fully hedged returns, DBFH

ck = DB(rkk); and (iii) a meas-

ure of diversification benefits based on the currency component ofthe investor’s foreign investments, DBCurr

ck = DB(xck).

The first two measures of the diversification benefit are easy tocompute based on investor’s currency denominated asset returns andlocal currency denominated returns, respectively. Since rc

k = rkk +xc

k,where rc

k is the continuously compounded (or log) return on countryk’s portfolio denominated in currency c, and xc

k is the change in theexchange rate between currency k and currency c, the third measure,the currency component of the investor’s diversification benefits, isthen computed by taking the difference between the first two:


DBCurrck = DBAgg

ck − DBFHck .

In our context currency risk is important also because the intro-duction of the euro eliminated a substantial component of currencyrisk for many international investments in our sample. Therefore,we would also like to disentangle the currency-risk effects of theadoption of the euro from the aggregate currency-risk effects of achange in portfolio allocation. We use the same methodology amplyexplained in De Santis and Gerard (2009) to construct the measureof diversification benefits based on currency components.

Since rebalancing a portfolio entails transaction costs, it isunlikely to take place when estimated marginal diversification ben-efits are of small magnitude. Therefore, to introduce some non-linearities, we take the cube of the estimated values of the diver-sification benefits.

The results point out that the coefficient on the expected volatil-ity of the exchange rate is not statistically significant for equity flowsand is only significant at 10 percent for bond flows, with the cor-rect sign in both cases (see specification 15 of tables 9 and 10). Asfor the diversification benefit, the aggregate marginal diversificationbenefit is not statistically significant (not reported), while the fullyhedged marginal diversification benefit is positive and statisticallysignificant, particularly for bond flows. The results contrast withthe findings of Portes and Rey (2005), who found weak supportfor the diversification motive, possibly because they use bilateralcovariances of returns in a common currency as a measure of riskdiversification. The impact of the volatility of the exchange rate andof the marginal diversification benefits arising from the currencycomponent are generally small. These findings imply that investorsmight have preferred to hedge against exchange rate risks over theperiod 1998–2001.

The econometric results summarized in tables 9 and 10 also sug-gest that, on top of the mere elimination of the exchange rate risk,EMU played a key role in the allocation of portfolio capital amongcountries worldwide as well as among euro-area member states,thereby enhancing regional financial integration and risk sharing.The catalyst effect of EMU—which is on top of the mere eliminationof the exchange rate risk and due to the reduction of legal barriers,such as the removal of intra-area currency-matching rules, and the

182 International Journal of Central Banking June 2010Tab

le9.

Inte

rnat

ional

Equity

Flo

ws,

IntC

AP

M,M

onet

ary

Pol

icy

Fra

mew

orks,

and

EM

U

t ck

,t=

α0

+α

1w

Ben

chck

,97

+α

2H

Bc,9

7+

α3Δ

HB

c,0

1+

α4D

W3 ck

,97

+α

5r k

,t−

1+

λM

onk

,t−

1

+γFX

vole c

k,t

−1

+β

1D

k⊂

EM

U+

β2D

k⊂

EM

U·D

c⊂

EM

U+

β3D

k⊂

UK

+β

4D

k⊂

EM

U·D

c⊂

UK

+β

5(D

k⊂

EU

−D

k⊂

EM

U)+

β6(D

k⊂

EU·D

c⊂

EU

−D

k⊂

EM

U·D

c⊂

EM

U)+

ϕD

B3 ck

,t−

1+

ε t

Spec

if.15

Spec

if.16

Spec

if.17

aSpec

if.17

b

Cst

−0.

046∗∗

(0.0

19)

−0.

046∗∗

(0.0

20)

−0.

046∗∗

(0.0

19)

−0.

046∗∗

(0.0

20)

wBen

chk

,97

0.42

2∗∗∗

(0.1

16)

0.45

6∗∗∗

(0.1

11)

0.42

8∗∗∗

(0.1

16)

0.42

5∗∗∗

(0.1

32)

HB

c,9

70.

045∗

(0.0

24)

0.04

5∗(0

.024

)0.

045∗

(0.0

24)

0.04

5∗(0

.023

)Δ

HB

c,0

1−

0.07

9∗∗∗

(0.0

24)

−0.

077∗∗

∗(0

.025

)−

0.07

6∗∗∗

(0.0

25)

−0.

072∗∗

∗(0

.025

)D

W3 ck

,97

−0.

348

(1.0

96)

−0.

491

(1.0

87)

−0.

343

(1.0

98)

−0.

311

(1.1

59)

r k,t

−1

0.00

5∗(0

.003

)0.

001

(0.0

03)

0.00

3(0

.003

)0.

003

(0.0

02)

Mon

eyk

,t−

10.

022∗∗

∗(0

.006

)0.

018∗∗

∗(0

.006

)0.

021∗∗

∗(0

.006

)0.

019∗∗

∗(0

.005

)FX

vole c

k,t

−1

−0.

068

(0.2

47)

−0.

087

(0.2

48)

−0.

064

(0.2

47)

−0.

011

(0.0

76)

Dc,E

MU

−0.

007

(0.0

05)

−0.

006

(0.0

06)

−0.

006

(0.0

05)

−0.

006

(0.0

04)

DEM

U,E

MU

0.03

3∗∗∗

(0.0

06)

0.03

8∗∗∗

(0.0

07)

0.03

6∗∗∗

(0.0

07)

0.03

5∗∗∗

(0.0

05)

Dc,U

K0.

013

(0.0

32)

——

0.00

9(0

.032

)0.

009

(0.0

34)

DEM

U,U

K0.

083∗∗

(0.0

40)

——

0.07

6∗(0

.039

)0.

077∗

(0.0

40)

Dc,E

U−

Dc,E

MU

——

0.01

4(0

.011

)0.

004

(0.0

06)

0.00

5(0

.006

)D

EU

,EU

−D

EM

U,E

MU

——

0.01

8∗∗(0

.009

)0.

010

(0.0

07)

0.01

0(0

.007

)D

BFX

3

ck

,t−

117

.14∗∗

(8.5

55)

20.1

7∗∗(8

.278

)17

.64∗∗

(8.4

99)

16.8

1∗(9

.011

)D

BEM

U3

ck

,t−

10.

971∗

(0.5

09)

0.58

8(0

.542

)0.

795

(0.5

19)

0.81

7(0

.503

)D

BX

−EM

U3

ck

,t−

10.

063

(0.0

49)

0.06

4(0

.049

)0.

062

(0.0

49)

0.04

5(0

.036

)Adj

uste

dR

20.

366

0.35

20.

366

0.36

8F

−St

at28

.48

26.8

525

.04

27.3

7

Note

s:T

his

table

repor

tsth

ere

sult

sof

the

cros

s-se

ctio

nal

regr

essi

onof

por

tfol

ioeq

uity

net

asse

tflow

s,su

mm

ariz

ing

findin

gsin

table

s6–

8.T

his

table

also

repor

tsth

ere

sult

sw

hen

cont

rollin

gfo

rth

epot

enti

aleff

ect

ofEM

Uan

dof

mar

ginal

div

ersi

fica

tion

ben

efits

.Exp

lanat

ory

vari

able

sar

edes

crib

edin

appen

dix

1.T

he

sam

ple

size

for

spec

ifica

tion

s15

–17a

,w

hic

hex

cludes

all

zero

hol

din

gs,

isn

=66

7.T

he

sam

ple

size

for

spec

ifica

tion

17b,w

hic

hin

cludes

allze

rohol

din

gs,is

n=

725.

OLS,W

hit

ehet

eros

kedas

tici

ty-c

onsi

sten

tst

andar

der

rors

,an

dco

vari

-an

ce.

Sta

ndar

der

rors

for

the

coeffi

cien

tsar

ere

por

ted

inpar

enth

eses

.**

*,**

,an

d*

den

ote

stat

isti

cal

sign

ifica

nce

at1,

5,an

d10

per

cent

,re

spec

tive

ly.

Vol. 6 No. 2 The Geography of International Portfolio Flows 183Tab

le10

.In

tern

atio

nal

Bon

dFlo

ws,

IntC

AP

M,M

onet

ary

Pol

icy

Fra

mew

orks,

and

EM

U

t ck

,t=

α0

+α

1w

Ben

chck

,97

+α

2H

Bc,9

7+

α3Δ

HB

c,0

1+

α4D

W3 ck

,97

+α

5r k

,t−

1+

λM

onk

,t−

1+

β1D

k⊂

EM

U+

γF

Xvo

le ck

,t−

1

+β

1D

k⊂

EM

U+

β2D

k⊂

EM

U·D

c⊂

EM

U+

β3D

k⊂

UK

+β

4D

k⊂

EM

U·D

c⊂

UK

+β

5(D

k⊂

EU

−D

k⊂

EM

U)

+β

6(D

k⊂

EU

·Dc⊂

EU

−D

k⊂

EM

U·D

c⊂

EM

U)+

ϕD

B3 ck

,t−

1+

ε t

Spec

if.15

Spec

if.16

Spec

if.17

aSpec

if.17

b

Cst

−0.

019∗∗

(0.0

07)

−0.

020∗∗

∗(0

.007

)−

0.01

9∗∗(0

.007

)−

0.01

6∗∗(0

.006

)w

Ben

chk

,97

0.21

2∗∗∗

(0.0

58)

0.21

5∗∗∗

(0.0

58)

0.21

3∗∗∗

(0.0

58)

0.21

8∗∗∗

(0.0

60)

HB

c,9

70.

012

(0.0

07)

0.01

2(0

.007

)0.

012

(0.0

07)

0.01

2∗(0

.007

)Δ

HB

c,0

1−

0.05

0∗∗∗

(0.0

10)

−0.

050∗∗

∗(0

.010

)−

0.05

1∗∗∗

(0.0

10)

−0.

036∗∗

∗(0

.008

)D

W3 ck

,97

1.35

4∗∗∗

(0.2

58)

1.36

0∗∗∗

(0.2

56)

1.35

4∗∗∗

(0.2

59)

1.38

3∗∗∗

(0.2

62)

r k,t

−1

0.04

2∗∗∗

(0.0

11)

0.04

0∗∗∗

(0.0

12)

0.04

1∗∗∗

(0.0

12)

0.03

4∗∗∗

(0.0

10)

Mon

eyk

,t−

10.

047∗∗

∗(0

.011

)0.

046∗∗

∗(0

.011

)0.

045∗∗

∗(0

.011

)0.

042∗∗

∗(0

.010

)Acc

ount

abili

tyk

,98

−0.

017∗∗

(0.0

07)

−0.

014∗∗

(0.0

07)

−0.

016∗∗

(0.0

07)

−0.

014∗∗

(0.0

06)

FX

vole c

k,t

−1

−0.

158∗

(0.0

92)

−0.

166∗

(0.0

92)

−0.

155∗

(0.0

93)

−0.

115∗

(0.0

64)

Dc,E

MU

0.00

9∗(0

.004

)0.

010∗∗

(0.0

05)

0.00

9∗∗(0

.005

)0.

006

(0.0

04)

DEM

U,E

MU

0.03

7∗∗∗

(0.0

08)

0.03

8∗∗∗

(0.0

08)

0.03

6∗∗∗

(0.0

08)

0.04

2∗∗∗

(0.0

08)

Dc,U

K0.

039∗∗

∗(0

.013

)—

—0.

037∗∗

∗(0

.013

)0.

038∗∗

∗(0

.012

)D

EM

U,U

K−

0.01

7(0

.019

)—

—−

0.01

6(0

.019

)−

0.01

7(0

.018

)D

c,E

U−

Dc,E

MU

——

0.01

5∗∗(0

.006

)0.

003

(0.0

06)

0.00

2(0

.005

)D

EU

,EU

−D

EM

U,E

MU

——

−0.

003

(0.0

06)

−0.

001

(0.0

05)

0.00

2(0

.005

)D

BF

X3

ck

,t−

137

283∗∗

∗(4

880)

3613

5∗∗∗

(492

5)37

244∗∗

∗(4

836.

8)38

729∗∗

∗(4

872.

7)D

BE

MU

3

ck

,t−

1−

12.3

67(8

2.86

)−

13.8

13(8

2.58

)−

11.9

69(8

2.93

)−

19.7

09(1

5.83

)D

BX

−E

MU

3

ck

,t−

1−

0.15

0(0

.100

)−

0.15

6(0

.100

)−

0.15

0(0

.100

)−

0.03

1(0

.015

)Adj

uste

dR

20.

366

0.35

90.

364

0.36

0F

−St

at25

.59

24.8

522

.52

26.8

3

Note

s:T

his

table

repor

tsth

ere

sult

sof

the

cros

s-se

ctio

nal

regr

essi

onof

bon

dnet

asse

tflow

s,su

mm

ariz

ing

findin

gsin

table

s6–

8.T

his

table

also

repor

tsth

ere

sult

sw

hen

cont

rollin

gfo

rth

epot

enti

aleff

ect

ofEM

Uan

dof

mar

ginal

div

ersi

fica

tion

ben

efits

.Exp

lanat

ory

vari

able

sar

edes

crib

edin

appen

dix

1.T

he

sam

ple

size

for

spec

ifica

tion

s15

–17a

,w

hic

hex

cludes

allze

rohol

din

gs,is

n=

639.

The

sam

ple

size

for

spec

ifi-

cati

on17

b,w

hic

hin

cludes

allze

rohol

din

gs,is

n=

783.

OLS,W

hit

ehet

eros

kedas

tici

ty-c

onsi

sten

tst

andar

der

rors

,an

dco

vari

ance

.Sta

ndar

der

rors

for

the

coeffi

cien

tsar

ere

por

ted

inpar

enth

eses

.**

*,**

,an

d*

den

ote

stat

isti

calsi

gnifi

cance

at1,

5,an

d10

per

cent

,re

spec

tive

ly.


sharing of common platforms, such as Euronext—is the estimatedcoefficient on DEMU,EMU. It is positive and strongly statisticallysignificant for both equity and bond flows.

Unfortunately, the lack of back data on bilateral portfolio flowsdoes not allow us to verify whether the upward trend in intra-euro-area portfolio flows from 1998 onward already started before theestablishment of EMU. Indeed, we could have captured the processof financial integration in the EU, as a result of EU policies aimingat liberalizing cross-border portfolio allocation among EU memberstates in the 1990s. However, we can run a similar exercise includ-ing binary variables aiming at controlling for the EU effect. Giventhe strong correlation between the dummies capturing both the EUand the EMU effects, we subtract the EMU dummies from the EUdummies. The estimated coefficient on DEMU,EMU remains stronglystatistically significant in both equity and bond markets (see spec-ification 16 of tables 9 and 10). Moreover, the results reported inspecifications 16 and 17a of tables 9 and 10 indicate that the Euro-pean dummies are not statistically significant once we control forLondon as a major intermediary of euro-area foreign investmentsfrom and to the rest of the world.

Needless to say, the results remain invariant when we includethe zero holding positions in the sample and treat them as actualportfolio investment decisions (see specification 17b of tables 9and 10).

Therefore, there is evidence of a positive EMU effect on cross-border portfolio flows. On a 95 percent confidence level, the catalysteffect of the euro within the euro area over the cumulated period1998–2001 amounted to USD 22–47 billion in equity securities andUSD 32–76 billion in bonds and notes. On average, the impact onthe fixed-income market is larger possibly because European insti-tutional investors invested massively in domestic government debt.The adoption of the euro currency-matching rule allowed them torebalance into euro-denominated bonds issued by other EMU mem-ber states.

All in all, the link between international portfolio flows, the Int-CAPM, characteristics of monetary policy frameworks, and EMUis quite robust also in terms of adjusted R2, which is above 35percent in both models. This result is very important given thatthe analysis has been carried out in the middle of an equity bubble


and burst, during therefore an extremely difficult period for globalasset management.

6. Adding Control Variables

The literature on portfolio flows has put forward a number of vari-ables potentially affecting the cross-border investment allocation.The aim of this section is to assess how robust the results are whenadding control variables to specification 15 of tables 9 and 10. Theresults reported in table 11 are obtained using the sample, exclud-ing the zero reporting for some holding positions. The results areinvariant when considering such zeros in the sample (not reported).

The “stages of development hypothesis” postulates an inverseU-shaped relationship between capital inflows and relative per capitaincome. Countries in the early stages of development tend to experi-ence capital inflows, which arise from building the infrastructure andexpanding domestic markets. In a subsequent phase, as new ideasare transformed into products and services and the country devel-ops some comparative advantages in specific industries, its per capitaincome rises and capital inflows decline. However, it is often arguedthat portfolio flows are particularly sizable among developed coun-tries, against the prediction of the “stages of development hypothe-sis” (Lucas 1990). Indeed, country k’s GDP per capita polynomial isnot statistically significant for the equity market and is even positivefor the bond market (see specification 18 of table 11).

Investment in research and development (R&D) is generally con-sidered a good policy to enhance the productivity of a country.If capital has a higher return in countries investing in R&D, theallocation of portfolios should also be partly affected. The imper-fection in international credit markets can affect the amount andthe direction of portfolio flows (Portes and Rey 2005).13 A struc-tural determinant of national savings is the demographic profile of acountry. Relatively high youth and old-age dependency ratios wouldbring about net capital inflows, as a relatively large population ofdependent young and old has a relatively lower savings rate (Ando

13R&D and bank credit to the private sectors are strongly correlated withthe GDP per capita (see table 3), because richer countries generally have moreresources to finance the private sector and the expenditure in R&D.


Tab

le11

.In

tern

atio

nal

Por

tfol

ioFlo

ws,

Dev

elop

men

t,In

stitution

s,D

ista

ntn

ess,

and

Net

Issu

ance

Spec

if.

Expla

nat

ory

Equity

Flo

ws

Bon

dFlo

ws

Num

ber

Var

iable

Coeff

.s.

e.R

2C

oeff

.s.

e.R

2

Deg

ree

ofE

cono

mic

Dev

elop

men

t

pcG

DP

k,9

7−

0.00

0(0

.001

)0.

002∗

∗∗(0

.001

)18

pcG

DP

2 k,9

70.

002

(0.0

20)

0.36

40.

055∗

∗(0

.025

)0.

367

19Ban

kcr

edit/G

DP

k,9

70.

002

(0.0

05)

0.36

5−

0.00

3(0

.002

)0.

366

20R

&D

/G

DP

k,9

70.

081

(0.2

24)

0.36

5−

0.10

2(0

.158

)0.

366

You

ngk,9

70.

031

(0.0

21)

0.00

5(0

.032

)21

Old

k,9

70.

028

(0.0

52)

0.36

50.

007

(0.0

60)

0.36

4so

phc,9

80.

001

(0.0

02)

0.00

1(0

.002

)so

phk,9

80.

000

(0.0

02)

0.00

5∗∗∗

(0.0

02)

22ds

oph k

,t−

0.00

3(0

.007

)0.

364

0.00

9∗∗

(0.0

05)

0.36

7

Inst

itut

ion

Qua

lity

23Lib

erties

k,9

70.

006

(0.0

07)

0.36

50.

008

(0.0

05)

0.36

624

LaPor

tak,9

7−

0.00

4(0

.014

)0.

365

0.01

4(0

.019

)0.

366

25IC

RG

k,9

7−

0.03

8(0

.035

)0.

365

0.00

8(0

.021

)0.

366

26C

orru

ptio

n k,9

7−

0.00

3(0

.007

)0.

365

0.00

9(0

.007

)0.

366

(con

tinu

ed)


Tab

le11

.(C

ontinued

)

Spec

if.

Expla

nat

ory

Equity

Flo

ws

Bon

dFlo

ws

Num

ber

Var

iable

Coeff

.s.

e.R

2C

oeff

.s.

e.R

2

Dis

tant

ness

27D

ista

nce c

k,9

7−

0.00

4(0

.003

)0.

367

0.00

1(0

.002

)0.

366

28Tra

deck,9

70.

043∗

(0.0

26)

0.36

80.

002

(0.0

46)

0.36

629

Tel

ck,9

8−

0.00

7(0

.007

)0.

366

0.00

2(0

.008

)0.

366

30La

ngua

geck,9

7−

0.00

0(0

.009

)0.

365

0.01

5∗∗

(0.0

07)

0.37

1

Eco

nom

icG

row

th

31Δ

GD

Pk,t

−1

0.00

1(0

.027

)0.

365

−0.

039∗

∗(0

.017

)0.

368

Net

New

Issu

ance

32Is

sues

k,t

0.03

2∗(0

.017

)0.

366

0.02

2∗∗∗

(0.0

06)

0.37

2

Note

s:T

his

table

repor

tsth

ere

sult

sof

acr

oss-

sect

ional

regr

essi

onof

por

tfol

ionet

asse

tflow

s.Eac

hex

pla

nat

ory

vari

able

ofth

ista

ble

isad

ded

asan

addit

ional

regr

esso

rto

the

mod

elsp

ecifi

cati

ons

15of

table

s9

and

10.T

he

expla

nat

ory

vari

able

sar

edes

crib

edin

appen

dix

1.T

he

sam

ple

size

isn

=66

7fo

req

uity

flow

san

dn

=63

9fo

rbon

dflow

san

dex

cludes

allze

rohol

din

gs.O

LS,W

hit

ehet

eros

kedas

tici

ty-c

onsi

sten

tst

andar

der

rors

,an

dco

vari

ance

.Sta

ndar

der

rors

for

the

coeffi

cien

tsar

ere

por

ted

inpar

enth

eses

.**

*,**

,an

d*

den

ote

stat

isti

calsi

gnifi

cance

at1,

5,an

d10

per

cent

,re

spec

tive

ly.


and Modigliani 1963). All these variables are not statistically signif-icant (see specifications 19–21 of table 11). Let us stress that theseresults are based upon a cross-section analysis, where most of thecountries share similar characteristics.

Portes and Rey (1995) argue that the sophistication of finan-cial markets is an important determinant of equity flows. However,our results do not support this hypothesis. At the same time, thesophistication of financial markets in the recipient country and itsdevelopment over time are statistically significant for bond flows (seespecification 22).

Several studies argue that institutions matter in shaping thenet flow of capital across countries (Alfaro, Kalemli-Ozcan, andVolosovych 2007; Coeurdacier, De Santis, and Aviat 2009; De Santisand Luhrmann 2009). International investment decisions are affectedby risks as well as by the countries’ institutional framework, as tur-moil, violence, instability, rule of law, property rights, and freedominfluence economic market sentiment. We expect that countries withbetter institutions should attract portfolio flows. To assess the roleof the quality of the institutions, we look at standard indices suchas the degree of civil liberties; the degree of political, financial, andeconomic risks; the degree of perceived corruption; and the degree ofcontract repudiation and expropriation risks.14 The results are gen-erally weak (see specifications 23–26), possibly because most of thebilateral transactions in this study occur across developed countries,which generally have an adequate institutional setting.

To assess the role of distantness, we consider geographical dis-tance, bilateral trade intensity, phone-traffic volume, and commonlanguage. The phone-traffic volume can be interpreted as a proxy forinformation flows (Portes and Rey 2005), while common languagecan be interpreted as a proxy for cultural similarities.15 Trade in

14All indices proxying for the quality of the institutions are strongly correlatedamong them as well as vis-a-vis the GDP per capita, in that richer countries havebetter institutions (see table 3).

15Common language is a dummy that takes the value of 1 if receiving andinvesting countries share the same language. The language groupings are as fol-lows: English (Australia, Bermuda, Canada, Ireland, New Zealand, Singapore, theUnited Kingdom, and the United States); Spanish (Argentina, Chile, Spain, andVenezuela); French (Belgium, Canada, and France); German (Austria and Ger-many); Dutch (Belgium and Netherlands); and Scandinavian (Denmark, Norway,and Sweden).


goods and services could facilitate the information flow across tradepartners, increasing the willingness to conduct cross-border portfoliotransactions. However, trade costs can also explain the equity port-folio home bias (Obstfeld and Rogoff 2001). In both interpretations,the deeper the trade relationship between countries, the deeper theirasset trade flow. There is some mild evidence for the trade variableaffecting equity flows (see specification 28) and for common languagehaving an impact on bond flows (see specification 30).

Finally, we also control for economic growth in the destinationcountry. Negative economic growth performance in the destinationcountry over the previous four years affects positively the subse-quent bond flows, possibly because a rebound in GDP growth mightbe expected (see specification 31).

The analysis so far presented focused on changes in demand.However, could important shifts occurring also on the supply sideaffect the results? The Maastricht Accord imposed tight restrictionson government debt. The effects of this were not symmetric acrosscountries and may have had a significant impact on the compositionof fixed-income securities available to investors. To assess whetherthe results remain robust to changes on the supply side, we useBIS data to compute the net new international equity and bondissues (the difference between completed issues and redemptions ina given period) over the period 1998–2001, which permit us to meas-ure the amount of new funds raised on the international markets.The net issuance of each individual country is then scaled by thetotal country portfolio. The results indicate that portfolio flows werealso affected by the new funds raised on the international marketsby the destination country (see specification 32).

All the results on the other regressors presented in tables 9and 10 remain unaltered when controlling for the effects of all thesevariables, as implicitly suggested by reporting the adjusted R2 intable 11.

7. Conclusions

It is generally believed that the predictions of the InternationalCapital Asset Pricing Model (IntCAPM) do not hold because of


home bias, transaction costs, asymmetric information, speculativebehaviors of investors, etc. We have presented a simple modelingframework showing that the geography of international portfolioflows is only partly influenced by the IntCAPM. Using bilateral dataon international equity and bond flows, we find clear evidence thatinvestors do not hold assets of each country in proportion to thecountry’s share in the world market portfolio. However, the pre-dictions of the IntCAPM are partially met because the estimatedcoefficient, rather than being 1, is 0.4 for equity flows and 0.2 forbond flows. A country that sees its market size increase by 1 percent-age point relative to the world market capitalization would attractinternational equity (bond) flows amounting to 0.4 percent (0.2 per-cent) of the equity (bond) assets held abroad by foreigners. Thismight also imply that global equity markets are more integratedthan global bond markets and that there is still room for furtherintegration in both markets. The results remain invariant when con-trolling for home bias, the initial degree of misallocation, and pastreturns.

Additional findings suggest that (i) a decline in home bias gener-ates portfolio outflows vis-a-vis all countries; (ii) the higher the ini-tial non-linear degree of misallocation, which might be due to higherfixed transaction costs and information asymmetries, the greater theincentive to reduce them and, consequently, the larger the subse-quent bond flows; and (iii) asset allocators engage in trend-chasingactivities, particularly in bond markets.

We have also investigated whether characteristics of the mone-tary policy frameworks and the establishment of EMU influencedportfolio asset flows over the turbulent 1998–2001 period for theasset markets. We find clear evidence that investors preferred port-folio assets of countries where the central bank gave relative impor-tance to money. With consumer price inflation well anchored, mon-etary analysis might have provided a framework for monitoringand assessing developments in asset prices and financial imbalances,cardinal to international investors when forming expectations onfuture risk-adjusted asset returns. As for EMU, once controlling fordiversification benefits and the elimination of the exchange rate risk,we show that cross-border portfolio flows among euro-area countrieshave increased due to the catalyst effect of EMU (i.e., reductionof legal barriers, sharing of common platforms, and simplification


of cross-border regulations). Therefore, we can safely say that EMUhas enhanced regional financial integration among euro-area memberstates in both equity and bond markets.

All in all, countries’ shares in the world market portfolio, homebias, initial degree of misallocation across countries, past returns,diversification benefits, monetary policy frameworks, and EMU canexplain 35–40 percent of the total variation in equity and bondportfolios over the turbulent 1998–2001 period, characterized by anequity bubble and the subsequent burst.

Appendix 1. Definitions of Variables

Variables Definitions

tck,t International transactions invested in countryk divided by country c’s average internationalholdings

wBenchk,97 Country’s share in the world market portfolio

HBc,97 Home bias of investing country c in 1997ΔHBc,97 HBc,01 − HBc,97

DWck,97 Difference between optimal and actual weights in1997

Retk,t−1 Total market return of receiving country k

end-1993 to end-1997Inflationk,98 Inflation focus index in country k in 1998: 0

(lowest importance) and 1 (highest importance)Moneyk,98 Money focus index in country k in 1998: 0

(lowest importance) and 1 (highest importance)Exchange ratek,98 Exchange rate focus index in country k in

1998: 0 (lowest importance) and 1 (highestimportance)

Discretionk,98 Discretion focus index in country k in 1998: 0(lowest importance) and 1 (highest importance)

Independencek,98 Independence index in country k in 1998: 0(lowest importance) and 1 (highest importance)

(continued)



Accountabilityk,98 Accountability index in country k in 1998: 0(lowest importance) and 1 (highest importance)

Transparencyk,98 Transparency index in country k in 1998: 0(lowest importance) and 1 (highest importance)

Financial Financial stability role index in country k inStabilityk,98 1998: 0 (lowest importance) and 1 (highest

importance)FXvoleck,t−1 Standard deviation of the bilateral exchange rate

change from 1993 to 1997 with deutsche markbeing the EMU currency

Dc,EMU Dummy is 1 if receiving country belongs to theeuro area

DEMU,EMU Dummy is 1 if receiving and investing countriesbelong to the euro area

Dc,UK Dummy is 1 if receiving country belongs to theUnited Kingdom

DEMU,UK Dummy is 1 if investing country belongs to EMUand receiving country to the United Kingdom

Dc,EU Dummy is 1 if receiving country belongs to theEU

DEU,EU Dummy is 1 if receiving and investing countriesbelong to the EU

DBFHck,t−1 Expected diversification benefits—fully hedged

returnsDBEMU

ck,t−1 Expected diversification benefits—internal EMUcurrency exposure

DBX−EMUck,t−1 Expected diversification benefits—external EMU

currency exposurepcGDPk,97 Country k GDP per capita minus the GDP per

capita of Norway in 1997 (2000 internationalPPP—$1,000)

pcGDP2k,97 The square of country k GDP per capita minus

the GDP per capita of Norway in 1997 dividedby 1,000

(continued)



Bank creditk,97 Bank credit to the private sector as a ratio to GDPof country k in 1997 minus the world average

R&Dk,97 R&D expenditure as a ratio to GDP of country k in1997 minus the world average

Youngk,97 Young dependents to working-age population incountry k relative to the world average in 1997

Oldk,97 Old dependents to working-age population incountry k relative to the world average in 1997

sophc,98 Sophistication of financial markets in country c in1998

sophk,98 Sophistication of financial markets in country k in1998

dsophk,t Change in the sophistication of financial markets incountry k, 1998 to 2001

Libertiesk,97 Civil liberties index in country k in 1997:0 (highest degree of freedom) and 1 (lowest degree offreedom)

La Portak,97 La Porta et al. (1998) index in country k in 1997: 0(highest risk) and 1 (lowest risk)

ICRGk,97 International Country Risk Guide rating in countryk in 1997: 0 (highest risk) and 1 (lowest risk)

Corruptionk,97 Perceived corruption index in country k in 1997: 0(highest risk) and 1 (lowest risk)

Distanceck,97 Logarithm of physical distance between capital citiesTradeck,97 Country k’s export share in country c plus country

c’s export share in country k in 1997Telck,97 Phone traffic (minutes per 1,000 subscribers)

between investing and receiving markets in 1997Languageck,97 Dummy is 1 if receiving and investing countries

share the same languageΔGDPk,t−1 Log-difference of the receiving country’s GDP in

USD at constant prices from 1993 to 1997Issuesk,t Net new issuance (completed issues minus redemp-

tions) from 1998 to 2001 divided by total country k

portfolio in 1997

194 International Journal of Central Banking June 2010A

ppen

dix

2.D

ata

Sou

rces

Dat

aP

rim

ary

Sou

rces

Sec

ondar

ySou

rces

Inte

rnat

iona

lPor

tfol

ioH

oldi

ngs

IMF-C

PIS

1997

,200

1G

erm

any:

Bun

desb

ank,

1997

Rea

l,N

omin

al,an

dPer

Cap

ita

Wor

dB

ank—

WD

IG

DP

Equ

ity

Mar

ket

Cap

ital

izat

ion

Tho

mso

nD

atas

trea

m/S

&P

Ber

mud

aSt

ock

Exc

hang

eB

ond

Mar

ket

Cap

ital

izat

ion

Leh

man

Bro

ther

sD

omes

tic

Bon

dsO

utst

andi

ngB

ISE

quity

Mar

kets

Tot

alR

etur

nsT

hom

son

Dat

astr

eam

Ber

mud

aSt

ock

Exc

hang

eB

ond

Mar

kets

Tot

alR

etur

nsJP

Mor

gan

Exc

hang

eR

ates

Tho

mso

nD

atas

trea

mP

hone

-Tra

ffic

Vol

ume

Wor

ldB

ank—

WD

IA

geD

epen

denc

yR

atio

sU

NW

orld

Pop

ulat

ion

Pro

spec

tsTra

deV

alue

sIM

FB

erm

uda

and

Isra

el:O

EC

DB

ank

Cre

dit

toth

eP

riva

teW

orld

Ban

k—W

DI

Sect

orR

&D

Exp

endi

ture

Wor

ldB

ank—

WD

IC

ivil

Lib

erti

esIn

dex

Free

dom

Hou

seLa

Por

taIn

dex

La

Por

taet

al.(

1998

)IC

RG

Inde

xIn

tern

atio

nalC

ount

ryR

isk

Gui

deC

orru

ptio

nIn

dex

Tra

nspa

renc

yIn

tern

atio

nal

Mon

etar

yPol

icy

Fram

ewor

ksM

ahad

eva

and

Ster

ne(2

000)

Soph

isti

cati

onof

Fin

anci

alW

orld

Eco

nom

icFo

rum

Mar

kets

Net

New

Issu

ance

BIS


References

Ahearne, A. G., W. L. Griever, and F. E. Warnock. 2004. “Infor-mation Costs and the Home Bias: An Analysis of US Holdingsof Foreign Equities.” Journal of International Economics 62 (2):313–36.

Alesina, A., and L. H. Summers. 1993. “Central Bank Indepen-dence and Macroeconomic Performance: Some ComparativeEvidence.” Journal of Money, Credit, and Banking 25 (2):151–62.

Alfaro, L., S. Kalemli-Ozcan, and V. Volosovych. 2007. “Capi-tal Flows in a Globalized World: The Role of Policies andInstitutions.” In Capital Controls and Capital Flows in Emerg-ing Economies: Policies, Practices, and Consequences, ed. S.Edwards. National Bureau of Economic Research ConferenceReport. University of Chicago Press.

Ando, A., and F. Modigliani. 1963. “The ‘Life Cycle’ Hypothesis ofSaving: Aggregate Implications and Tests.” American EconomicReview 53 (1): 55–84.

Aviat, A., and N. Coeurdacier. 2007. “The Geography of Trade inGoods and Asset Holdings.” Journal of International Economics71 (1): 22–51.

Bohn, H., and L. L. Tesar. 1996. “U.S. Equity Investment in ForeignMarkets: Portfolio Rebalancing or Return Chasing?” AmericanEconomic Review 86 (2): 77–81.

Borio, C. E. V., and P. Lowe. 2004. “Securing Sustainable PriceStability: Should Credit Come Back from the Wilderness?” BISWorking Paper No. 157.

Brennan, M. J., and H. H. Cao. 1997. “International Portfolio Invest-ment Flows.” Journal of Finance 52 (5): 1851–80.

Coeurdacier, N., R. A. De Santis, and A. Aviat. 2009. “Cross-BorderMergers and Acquisitions and European Integration.” EconomicPolicy 24 (57): 55–106.

Committeri, M. 2000. “Effects of Volatile Asset Prices on Balanceof Payments and International Investment Position Data.” IMFWorking Paper No. 191.

Coval, J. D., and T. J. Moskowitz. 1999. “Home Bias at Home: LocalEquity Preference in Domestic Portfolios.” Journal of Finance54 (6): 2045–73.


Cukierman, A. 1992. Central Bank Strategy, Credibility, and Inde-pendence: Theory and Evidence. Cambridge, MA: MIT Press.

De Santis, R. A., and B. Gerard. 2009. “International PortfolioReallocation: Diversification Benefits and European MonetaryUnion.” European Economic Review 53 (8): 1010–27.

De Santis, R. A., and M. Luhrmann. 2009. “On the Determinants ofNet International Portfolio Flows: A Global Perspective.” Jour-nal of International Money and Finance 28 (5): 880–901.

Detken, C., and F. Smets. 2004. “Asset Price Booms and MonetaryPolicy.” In Macroeconomic Policies in the World Economy, ed.Horst Siebert. Berlin: Springer.

European Central Bank. 2003. Review of the International Role ofthe Euro. Frankfurt am Main: European Central Bank.

Faruqee, H., S. Li, and I. K. Yan. 2004. “The Determinants of Inter-national Portfolio Holdings and Home Bias.” IMF Working PaperNo. 34.

Froot, K. A., P. G. O’Connell, and M. S. Seasholes. 2001. “ThePortfolio Flows of International Investors.” Journal of FinancialEconomics 59 (2): 151–93.

Gelos, R. G., and S.-J. Wei. 2005. “Transparency and InternationalPortfolio Holdings.” Journal of Finance 60 (6): 2987–3020.

Grilli, V., D. Masciandaro, and G. Tabellini. 1991. “Political andMonetary Institutions and Public Financial Policies in the Indus-trial Countries.” Economic Policy 6: 342–91.

Huberman, G. 2001. “Familiarity Breeds Investment.” Review ofFinancial Studies 14 (3): 659–80.

International Monetary Fund. 1997. International Capital Markets,Developments, Prospects, and Key Policy Issues. Washington,DC: International Monetary Fund.

———. 2005. Global Financial Stability Report. Washington, DC:International Monetary Fund.

Jacome, L. I., and F. F. Vazquez. 2005. “Any Link Between LegalCentral Bank Independence and Inflation? Evidence from LatinAmerica and the Caribbean.” IMF Working Paper No. 75.

Lane, P. R. 2006. “Global Bond Portfolios and EMU.” InternationalJournal of Central Banking 2 (2): 1–23.

Lane, P. R., and G. M. Milesi-Ferretti. 2008. “International Invest-ment Patterns.” Review of Economics and Statistics 90 (3): 538–49.


La Porta, R., F. Lopez-de-Silanes, A. Shleifer, and R. W. Vishny.1998. “Law and Finance.” Journal of Political Economy 106 (6):1113–55.

Lintner, J. 1965. “The Valuation of Risk Assets and the Selectionof Risky Investments in Stock Portfolios and Capital Budgets.”Review of Economics and Statistics 47 (1): 13–37.

Lucas, R. E. 1990. “Why Doesn’t Capital Flow from Rich to PoorCountries?” American Economic Review 80 (2): 92–96.

Mahadeva, L., and G. Sterne. 2000. Monetary Policy Frameworks ina Global Context. London: Routledge.

Obstfeld, M., and K. Rogoff. 2001. “The Six Major Puzzles inInternational Macroeoconomics: Is There a Common Cause?”In NBER Macroeconomics Annual 2000, ed. B. S. Bernanke andK. Rogoff, 339–90. Cambridge, MA: MIT Press.

Papaioannou, E. 2009. “What Drives International Bank Flows? Pol-itics, Institutions and Other Determinants.” Journal of Develop-ment Economics 88 (2): 269–81.

Portes, R., and H. Rey. 2005. “The Determinants of Cross-BorderEquity Flows.” Journal of International Economics 65 (2): 269–96.

Rogoff, K. 1985. “The Optimal Degree of Commitment to an Inter-mediate Monetary Target.” Quarterly Journal of Economics 100(4): 1169–90.

Sercu, P. 1980. “A Generalization of the International Asset PricingModel.” Revue de l’Association Francaise de Finance 1: 91–135.

Sharpe, W. F. 1964. “Capital Asset Prices: A Theory of MarketEquilibrium under Conditions of Risk.” Journal of Finance 19(3): 425–42.

Solnik, B. H. 1974. “An Equilibrium Model of the InternationalCapital Market.” Journal of Economic Theory 8 (4): 500–24.

The Geography of International Portfolio Flows ...

Documents

Transcript of The Geography of International Portfolio Flows ...