Country portfolios

MIT LIBRARIES

Digitized

by

the

Internet

Archive

in

2011

with

funding

from

Boston

Library

Consortium

IVIember

Libraries

DE\A/BY

Massachusetts

Institute

of

Technology

Department

of

Economics

Working

Paper

Series

COUNTRY

PORTFOLIOS^

Aart

Kraay,

The World

Bank

Norman

Loayza,

Banco

Central

de Chile

and

The World

Bank

Luis S

erven,

The

World

Bank

Jaume

Ventura,

MIT

Working

Paper

00-16

July

2000

Room

E52-251

50

Memorial

Drive

Cambridge,

MA

02142

This

paper

can

be

downloaded

without

charge

from

the

Social

Science

Research Network

Paper

Collection

at

'M^^'CHUSETTSiNSTlTUTE

OFTECHNOLOGY

OCT

2 5

Massachusetts

Institute

of

Technology

Department

of

Economics

Working Paper

Series

COUNTRY

PORTFOLIOS^

Aart Kraay,

The

World

Bank

Norman

Loayza,

Banco

Central

de

Chile

and

The

World Bank

Luis

Serven,

The

World

Bank

Jaume

Ventura,

MIT

Working Paper

00-16

July

2000

Room

E52-251

50

Memorial

Drive

Cambridge,

MA

02142

This

paper

can

be downloaded

without

charge from

the

Social

Science

Research Network Paper

Collection

at

Abstract

How

do

countries holdtheirfinancial wealth?

We

constructa

new

database

of

countries' claims

on

capital locatedat

home

and

abroad,

and

internationalborrowing

and

lending,covering 68

countries

from

1966to 1997.

We

find that asmall

amount

ofcapital flows from richcountries to

poor

countries. Countries' foreign assetpositions areremarkablypersistent, and mostlytakethe

form

offoreign loansratherthan foreign equity.

To

interpretthese facts,

we

build asimple

model

ofintemafional capital flo9wsthathighlights the interplay

between

diminishingreturns, production risk

and

sovereign risk.

We

show

that in the presence ofreasonable diminishingreturns and productionrisk, the probabilitythat

international crises occurtwice a centuryis

enough

to generate a setof counfryportfolios that are

roughlyconsistentwiththe data.

*We

thank

Mark

Aguiar,

Daron Acemoglu,

Giancarlo Corsetti, Raquel Fernandez, RobertE.

Lucas

and

Mark

Wright for insightful

comments

on

aprevious draft and

Gian Maria

Milesi-Ferretti for his advice inthe construction of ourdataset.

We

are also grateful toCesarCalderon,

Alex

Karaivanov,

George Monokroussos,

and

Rashmi

Shankarforexcellentresearchassistance.

The

opinionsexpressedhere are the authors'and

do

not necessarilyreflectthoseofthe

World

Bank, itsexecutive directors, or the countriestheyrepresent. Financial support

from

the

Regional Studies

Program

ofthe

Latm

American Region

ofthe

World Bank

is gratefully

Introduction

How

do

countries hold theirfinancial wealth?

Our

objective is to provide a

description ofcountry portfolios

and

advance

a parsimonious explanation oftheir

main

features.

By

country portfolio,

we

refer tothefinancial wealth ofthe country

and

how

it is distributed across holdings ofdomesticcapital

and

variousforeign assets.

By

the net foreign assets ofa country,

we

refer tothe country's holdings offoreign equity

and

loans

minus

foreigners' holdings ofdomesticequity

and

loans.

We

first turn tothe

data

and

ask:

How

large are netforeign asset positions?

What

countrycharacteristics

seem

to

be

associatedwith positive netforeign asset positions?

How

persistent are netforeign asset positions?

What

is the relativeimportance of foreign loans

and

equity?Withthe

answers

athand,

we

go

to the theory

and

ask:

Why?

To

determinethe

main

features ofcountry portfolios,

we

constructa

new

database

on

foreigners' holdings ofdomestic equity

and

loans,

and

domestic

residents' holdings offoreign equity

and

loans.

Our sample

covers the period

1966-1997 and

includes

68

countriesthat accountfor over90 percent ofworld production

and

trade. Constructing this

database

forces us to

choose

among

fragmentary

and

imperfect sources of information

and

then

make

(heroic?) assumptions

on

how

to

reconcile

them and

fill in the gaps.^ Despitethis,

we

feel confidentthe inferences

we

draw

fromthis data are robust. In fact, there is nothing subtle aboutthe empirical

regularities

we

highlight here.

The

following are all verystriking features ofthe data:

1. Net foreignasset positions as a share ofwealth are small in absolute value

and

negative for

most

countries.

Roughly 80

percent ofthe observations in our

sample

consist ofcountries

whose

net foreign assets as ashare ofwealth are less than

20

percentin absolute value.

We

alsofindthat the net foreign asset position is

negative forabout

80

percent oftheobservations in our sample.

^ _{See Sinn}

[1990], Rider [1994] andLane and Milesi-Ferretti[1999] foralternativesourcesof data onforeignassetpositions.

Introduction

How

do

countries hold theirfinancial wealth?

Our

objective is to provide a

description of country portfolios

and advance

aparsimonious explanation oftheir

main

features.

By

countryportfolio,

we

refertothefinancial wealth ofthe country

and

how

it is distributed across holdings ofdomesticcapital

and

variousforeign assets.

By

the net foreign assetsofa country,

we

referto thecountry's holdings of foreign equity

and

loans

minus

foreigners' holdingsofdomestic equity

and

loans.

We

firstturn tothe

data

and

ask:

How

large are netforeign asset positions?

What

country characteristics

seem

to

be

associated with positive netforeign asset positions?

How

persistentare

netforeign asset positions?

What

is therelative importance offoreign loans

and

equity?Withthe

answers

at hand,

we

go

tothe theory

and

ask:

Why?

To

determinethe

main

featuresofcountry portfolios,

we

constructa

new

database on

foreigners' holdings ofdomestic equity

and

loans,

and

domestic

residents' holdings offoreign equity

and

loans.

Our sample

covers the period

1966-1997

and

includes

68

countries that accountforover

90

percent ofworld production

and

trade. Constructing this

database

forcesus to

choose

among

fragmentary

and

imperfectsources of information

and

then

make

(heroic?) assumptions

on

how

to

reconcile

them and

fill in the gaps.^ Despite this,

we

feel confident the inferences

we

draw

from this data are robust. Infact, there is nothing subtle aboutthe empirical

regularities

we

highlighthere.

The

following areall verystriking features ofthe data:

1. Net foreign asset positions as a shareofwealth are small in absolute value

and

negativefor

most

countries.

Roughly 80

percentofthe observations in our

sample

consistofcountries

whose

netforeign assets as a shareofwealth are lessthan

20

percent in absolutevalue.

We

alsofindthat the net foreign asset position is

negative for about

80

percentofthe observations in our sample.

^

_See

_Sinn

[1990], Rider [1994] and Lane and Milesi-Ferretti [1999] foralternative sources of data onforeign assetpositions.

2.

There

is a strong positive relationship

between

financial wealth

and

the net foreign asset position in a cross-section ofcountries. In particular,

we

find that

when

financial wealth doubles, the share of netforeign assets in wealth increases

bythree to six percentage points. Cross-country variation in financial wealth

seems

to explain (in a statistical sense)

most

ofthevariation in net foreign asset

positions,

even

after controlling for aset ofvariablesthat aredesigned to capture

cross-country differences in aggregate production functions.

3.

The

share ofnetforeign assets in wealth is very persistent overtime.

A

simple

regression ofthis share

on

its lagged value delivers a slope coefficient of0.98

and

a R^ of93 percent. Since

we

find that

between

40

percent

and 70

percentofthe

variation in

changes

in netforeign assets can

be

attributed to

changes

in relative

wealth, it

seems

reasonable to conclude that persistence in relative wealth is

an

importantsource ofpersistence in foreign assets.

4.

Gross

foreign asset positions are small

and

consist mostly offoreign loans rather

than foreign equity. In developing countries, foreign equity assets

and

liabilities

are roughly 0.3

and

2.8 percent ofwealth, while foreign loan assets

and

liabilities

accountfor4.5 percent

and

8.8 percent ofwealth. In industrial countries, foreign

equity assets

and

liabilities are roughly 3.3

and

3.9 percent ofwealth, while

foreign loan assets

and

liabilitiesaccountfor 1 1 percent ofwealth each.

To

sum

up,

we

observe that net foreign asset positions are small relative to

wealth

and

tend to be negative, exceptfor a

few

rich countries.

These

net foreign

asset positions are remarkably persistent as afraction ofwealth,

and

mostly consist

offoreign loans rather thanforeign equity holdings. This picture that

emerges

from

the data is so clear that

we

think itshould constitute the

main

target of

any

successful

theoryof international capital flows.

It

seems

safeto argue that such a theory requires at leasttwo ingredients.

To

explain the strong positive association

between

wealth

and

netforeign asset

to

move

from richto poor countries. Natural candidatesfor this role are diminishing

returns atthecountry level

and

country-specific production risk. If either ofthese two

forces are present, the risk-adjusted rate of return to capital declines as

more

capital

is invested in a country, creating

an

incentive to investin countries that

have

little

capital. Inthe

absence

ofa counten/ailing force, this incentivewould only

be

eliminated if capital stocks

were

equalized across countries.

Hence

the theory

needs

a

second

ingredient to explain

why

net foreign asset

positionsare so small.^

A

popularview isthatthetheoryjust

needs

to recognizethat

rich countries

have

betteraggregate productionfunctions,

and

this is

why

investors

keep most

oftheircapital in rich countries

even

in the presence ofdiminishing returns

and

production risk. While this is likelyto betrue to

some

extent,

we

look elsewhere

forthe

second

ingredient ofthetheory, forthree reasons. First,

we

findthat standard

variables

we

thinkare associated with betteraggregate production functions

(human

capital, quality ofinstitutions,

and

others)

seem

to beeither unrelated to the net foreign asset position ofa country or, alternatively, explain very little ofitsvariation

(See Table 3

and

the discussion of itbelow). Second, while better aggregate

production functions in rich countries

can

explain

why

netforeign asset positions are

small, theycannotexplain

why

grossforeignequity positions are alsosmall.

To

the

extent that investors

have

a desireto diversify production risk, the theory

would

predict thatthey

choose

largegrossforeign equitypositions that are roughly

balanced. Finally, betteraggregate production functions in rich countriescannot

explain

why

most

international trade consists ofloans ratherthan equity. While both

assets are useful to transfer capital across countries, equityhas the additional benefit ofallowing countries to share production risk

and

should therefore always be

preferred over loans.

Why

are observedforeign equity positions so small?

Why

are

foreign loans ratherthan foreign equity the assetthat is

most

traded internationally?

To

answer

thesequestions,

we

would still

need

to

add

additional elements tothe

theory anyway.

Inthis

paper

we

explore the alternative hypothesis thatsovereign risk might

be

the

second

ingredient thatthetheory needs.^ Inthe

presence

ofthissort of risk,

domesticcapital offers domestic investors notonly the value ofits productionflow, but

alsoa

hedge

against the risk offoreign default. This creates a

home

bias inthe

demand

for capital that mightexplain

why

net foreign asset positions are small.

Should

we

also expectthat sovereign risk leadsto small grossforeign equity

positions? Is it

even

possible that sovereign risk explains

why

most

international trade

in assets consists ofloans ratherthan equity?

The

answers

tothesequestions

depend

crucially

on

the

consequences

forinvestors ofaforeign default.

Assume

first thatifa country defaults

on

itsforeign obligations, foreign

countries respond by seizing the assetsthat this country

owns

abroad

and

then using

these assetsto (partially)

compensate

creditors.

Assume

alsothatthe process by which assets are seized

and

transferred to creditors

does

not give rise to

any

costs or loss of value. In this case, the loss suffered by creditors in the event ofdefault is

determined bytheir netforeign asset position vis-a-visthe defaulting country.

To

minimize

exposure

to sovereign risk, investorsthen

choose

small net foreign asset

positions. But they

do

not

have

to hold small grossforeign equitypositions. Infact, to

the extentthat investors

have

adesireto diversify production riskthey

would

again

choose

largeforeign equity positions thatare roughly balanced.

Under

these assumptions, sovereign risk provides a rationalefor

why

netforeign asset positions

are small, but it cannot explain

why

grossforeign equity positions aresmall. Neither

can it explain

why

most

international trade consistsofloans ratherthan equity.

If

we

want

to hold sovereign risk responsibleforthe small gross foreign equity positions observed in the data,

we

need

to

remove

at least

one

oftheassumptions of

the previous paragraph.

Perhaps

legal

systems

do

not allowcreditors to seize the

foreign assets ofdefaulting countries. Or,

even

iftheydo, thetransfer ofownership

^_{Obstfeld and Rogoff}

[1996, p.349] define sovereign riskas referringto"...anysituation in

which a government defaults on loan contractswith foreigners, seizes foreign assets located

within itsborders, or prevents domesticresidentsfromfullymeetingobligationstoforeign

might involve large costs

and

asubstantial loss of value. Inthis case, investors'

exposureto sovereign risk is

no

longertheir net foreign asset position vis-a-visthe

defaulting country, but instead sonnefraction oftheirgrossforeign asset position.

Now

the existence ofsovereign risk

can

potentiallyexplain

why

investors

choose

small grossforeign equitypositions.

With the help ofthe additional (and

we

think realistic) assumption that

transferring ownership ofequity is costlierthantransferring ownershipofloans,

sovereign risk can also explain

why

most

international trade in assets consistsof foreign loans ratherthanforeign equity. Investors issuing foreign loans

and

equityare

willing to sell these assets ata discountthat reflectsthe value for

them

ofthe gain

they receive in the eventofdefault.''

What

is this gain?

A

domestic investor

who

has

borrowed abroad

receives the full value ofthe loans. This gain is exactly equal tothe

loss experienced by theforeign investor.

A

domestic investor

who

has sold equity

claims toforeigners receives thefull value oftheequity, but loses

any

valuable

advantage

that foreign investors broughtto thefirm (experience, managerial skills,

know-how,

access to bettertechnologyor relationships, firm-specific knowledge,

and

so on). This gain is less than the loss experienced bytheforeign investor. Therefore

inthe event ofdefault,foreign loans give rise onlyto puretransferswhile foreign

equity creates losses. Whilethe latterallows investors to

hedge

against production

risk, it is a

worse

hedge

against sovereign risk than the former. Ifthe desireto avoid

diminishing returns induces investorsto transfercapital from richto poorcountries,

foreign loanswill

be

a

more

attractive asset ifsovereign risk is high relative to

production risk. Thus, sovereign risk has also the potential to explain

why

loans are

the preferred asset tofinance capital flows.

The

notion that foreign equity

and

loans are subject to sovereign risk is hardly

novel or controversial

among

observers of international financial markets.

The

interesting issue is

whether

reasonable probabilities of default can quantitatively

"*

This discount in loancontracts shows up Inthe Interest rate. Theevidence Isoverwhelming thatloanstodeveloping countries usually

command

a higherInterestratethan domestic

explain the

main

features ofcountry portfolios.

To

determine this,

we

constructa

simple North-South

model

of international capital flows.

The

production technology

exhibits diminishing returns at the country level

and

country-specific risk. In this

model, the world

economy

experiences periods with substantial international trade in

assets, which

end up

in a crisis period in which

South

defaults

on

itsforeign

obligations. North investors seize South'sforeign assets, butthis transfer ofassets is

costly.

The

defaultinitiates a crisis period in which international financial marketsshut

down.

Eventually, international trade in assets

resumes and

the cycle starts again.

It

seems

clear thatthis

model

can, in principle, explain thefacts discussed above. Ifsovereign risk is sufficiently high, net capital flows will

be

small (Fact 1). If eitherdiminishing returns or country-specific production risk is important,

we

should

observe a tendencyfor capital toflowfrom rich to poor countries (Fact 2).

To

the

extentthat transferring ownership of foreign equity is costlierthantransferring

ownership of loans, grossforeign assetpositions should

be

small

and

consist

primarily of loans ratherthan equity (Fact4). Ifthe

model

generates persistence inthe

world distribution ofwealth, this can naturallyexplain the persistence offoreign asset

positions (Fact3). It is

much

less clear

however whether

this

model

is able to provide

a reasonable quantitative description ofthe data.

A

perhapssurprising finding is that,

even

in the presence ofreasonable diminishing returns

and

production risk, the

probability that international crises occur twice a century is

enough

to generate a set

ofcountryportfolios that are roughly consistent with the data.

The

paper is organized asfollows: Section I provides a briefdescription ofour

database

and

extensively

documents

the four facts

mentioned

above. Section II

presents the

model and

discusses its

main

qualitative

and

quantitative implications.

Appendix

A

provides a detailed discussion of

how

we

constructed our database, while

I.

A

Description

of

Country

Portfolios

Inthis section,

we

describe the

main

characteristics of countryportfolios using

a

new

database

covering

68

countries from

1966

to 1997.

We

first provide

an

overview ofthe sources

and methodology used

to constructthe data.

We

then

summarize

its

main

features in theform offourfacts.

Appendix

A

providesdetails

on

the data

and

describes

how

we

accountfor

changes

in thevalue ofstocks ofassets.

1.1

A

New

Database

on Country

Portfolios

Our

databasecontains estimates ofdomestic capital stocks

and

foreign

assets of countries.^ _{In particular,}

_we

_have

_measures

_{forthefollowing quantities:}

k =

Domestic

capital stock

e =

Domestic

equity

owned

byforeign residents.

e* = Foreign equity

owned

by domestic residents.

I =

Loans

issued bydomestic residents

and

owned

byforeign residents.

I* =

Loans

issued by foreign residents

and

owned

by domestic residents.

We

measure

these quantities perdomestic resident in constant

1990

US

dollars.

We

define a=k+e*-e+r-l

and

f=e*-e+r-l asthe financialwealth or portfolioofthe country

and

its net foreign assets, respectively.

We

referto a country

as

a creditor (debtor) if

its netforeign assets are positive (negative).

We

construct estimates of

each

component

of financial wealth in two steps.

First,

we

use the limited available information on stocksofthese assetsto determine

an

initial value. Second,

we

use flow data

and

estimates of

changes

in the value of

these assets to extend the initial stocks forward

and backward

overtime.

We

rely

on

^

We

_{abstractfrom othercomponents ofwealth suchas land, natural resources, and}

_human

this

method

ofcumulating flows

even

forthose countries

where

more

stock data is

available in orderto avoid a potential bias. Since long time series ofstock data are

available only fora

few

rich countries,using these as the primary source

would

essentially result in different

methods

being

used

to construct stocks for rich

and

poor

countries.

These

differences

would

then contaminate ourinferences regarding

how

portfolios varywith wealth.

We

rely on data from a

number

of standard sources.

We

obtain initial stocks

ofdomesticcapital from the

Penn

World

Tables,

and

use flow data

on

gross domestic investment to build

up

stocks of capital valued in

US

dollars at

PPP.

In orderto

determineforeign holdings ofdomesticequity

and

domestic holdings offoreign equity,

we

rely primarily

on

data

on

stocks

and

flows of direct

and

portfolio equity

Investment reported in the International

Monetary

Fund's Balance of

Payments

Statistics Yearbook.

We

again

use

the limited availableinformation

on

stocks

reported in this

and

a variety ofothersourcesto determine the initial level of

each

assetfor

each

country,

and

then use corresponding flowdata from the balanceof

payments

to construct stocksfor remainingyears. Finally,

we

combine

stock data on

the debts ofdeveloping countries reported in the

World

Bank's Global

Development

Financewith data

on

stocks

and

flows

on

debt from the Balance of

Payments

Statistics

Yearbook

to build

up

stocks ofborrowing

and

lending forall countries in our

sample.^

Our

sample

ofcountries is determined primarily by data availability.

We

begin

with a

sample

of

98

countries with population greaterthan

one

million

and

percapita

GDP

greaterthan 1

000

US

dollars at

PPP

in 1990.

Of

these

we

discard

25

countries

with missing, incomplete, or inconsistentbalance of

payments

data. This leaves us

with an unbalanced panel of

73

countries spanning

on

average

25

ofthe years

between 1966 and

1997. Inthe empirical

work

thatfollows

we

restrict attention toa

®

We

_assume

_{throughoutthat stocksofdebt reported inthesesources constitute solelythe}

assets and liabilities ofdomestic residents.

To

theextentthatthese reflectdebt issued byor

owed

toforeign-ownedfirmsoperating in the country,

we

are overestimating the loan assets

and liabilities ofdomestic residents. Given that foreign holdingsofdomesticequityare small

set of

68

countries excluding fivetransition economies. Table 1 liststhese countries

classified by geographical region.

As

the table shows, our

sample

includes basically

all industrial countries

and

a substantial

number

ofdeveloping countriesfrom all

regions ofthe world.

The

countries in our

sample

accountforover

90

percent of

world production

and commodity

trade. It is reasonabletothink that these countries

alsoaccountfora similar fraction ofworld trade in assets. ''

Despitethis wide

coverage,

we

do

not findthat netforeign assets

sum

tozero across countries in our

sample. Inthe

case

ofclaims

on

equity,

we

find thatthe

sum

ofall countries reported

claims

on

foreign equityis

on

average about3 percent less than reported foreign

claims

on

domestic equity. In the case of lending the discrepancy is larger, with world

reported borrowing exceeding lending by about 12 percent. While these

discrepancies are nottrivial, they are ofa

comparable

order ofmagnitudeto

well-known

discrepancies in flows

on

foreign assets.

1.2

Main

Features

of

Country

Portfolios

Inthis subsection

we

examine

the

main

features ofcountryportfolios using the

database described above.

We

organize the discussion around four

main

facts or

findings.

Fact 1

:

Netforeign assets

as

a

share of wealth are small

and

negative for

most

countries.

Figure t

shows

the distribution ofthe shareofforeign assets in wealth,

—

,

a

pooling the available

1717

observations forall countries

and

years. Overwhelmingly,

net foreignassets representa small fraction ofthe wealth ofdomestic residents.

''

Our procedureresults inestimates ofwealth thatare very small (and occasionally negative)

forafewcountry-year observations, correspondingtocountries with very large external debt.

We

excludethese observations by limitingthesample tothosewherethe ratioofwealth to

Roughly

80

percentofthe country/year observations are lessthan

20

percent in

absolute value. Also, about

80

percent ofthecountry-yearobservations are negative.

Perhaps

Figure 1 is concealing important variation across regions

and

overtime in the size

and

sign ofthe net foreign asset position.

To

rule outthis possibility, Table 2

reports theshareofnetforeign assets in wealth, aggregating across countries in

different regions

and

time periods.

The

toppanel reportswealth-weighted averages,

and

the lower panel reports the shareforthe typical (median) countryin

each group

and

period. Clearly, thefinding that

most

countries

have

a small butnegative net

foreign asset position holds across regions

and

overtime.

Since net foreignasset positions are small relative to wealth, it follows that

there is a strong relationship

between

countries' financialwealth

and

theircapital

stock.

To

see

this, Figure 2 plots the capital stock, k, against financial wealth, a,

averaging the available years over the period

1966-1997

for

each

country.

A

simple

regression ofthe capital stock

on

domesticwealth delivers

and

R^of

98

percent

and

a slope coefficient of 0.93. Clearly, the world distribution of capitalstocks is very close

to the world distribution ofwealth. This

does

not

mean

however

that differences

between

these two distributions are

random

or uninteresting. Although the slope

coefficient

appears

to

be

very close to one, the null hypothesis thatthis coefficient is

one

isrejected atconventional significance levels.

We

also rejectthe null hypothesis

thatthe intercept is zero. That is,

on average

the capital stock

exceeds

wealth in poor

countries

and

is less than wealth in richcountries. This leads us tothe nextfact:

Fact 2:

The

share of netforeign assetsin the countryportfolio increases with wealth

in across-section of countries.

The

strong positive association

between

the share of netforeign assets

and

wealth is apparent in Figure 3, which plotsthe

average

shareof net foreign assets in

wealth,

—

, againstthe logarithmofwealth, ln(a), for

each

country over 1966-1997.

a

Virtually all (44 out of47) developing countries are debtors, as aretwo-thirds of

industrial countries (14 outof21). Tiie ten creditor countries in our

sample

are mostly

rich industrial countries (Belgium/Luxembourg, Switzerland,

Germany,

France, United

Kingdom,

Japan,

and

the Netherlands),

and

three developing countries (Saudi

Arabia, Singapore,

and

Lesotho).^

The

relationship

between

wealth

and

netforeignasset positions holds across

subperiods.

The

first row ofTable 3 confirmsthat the simple results in Figure 3

(shown

in the firstcolumn) hold acrossthe differentsubperiods

(shown

in the

remaining columns).

The

estimated coefficients range from 0.04to 0.08

and

are

significantly differentfrom zero in

each

subperiod.

The magnitudes

ofthese

coefficients indicate that

when

wealth doubles, the share offoreign assets in wealth

increases bythree to six percentage points. In the next

row

ofTable 3

we

introduce

regional

dummies.

With the exception ofthe first subperiod, thecoefficient

on

the

logarithmofwealth increases slightly,

and

remains significantlypositive. Inthe third

row

ofTable 3

we

check

whetherthe

changes

in the wealth elasticity across periods is

an

artifactofthechanging composition ofthe

sample

by restricting attention to a

balanced panel of8-yearaverage observations.

The

results

do

not

change

substantially.

There

are reasonsto think thatthe relationship

between

wealth

and

the net foreign asset position is

even

stronger than

what

a simple regression

would

find. It is

easy

to think offactors thatare positively correlated with wealth

and

are likelyto be

negatively correlated withthe net foreign asset position ofa country. First, wealth is

stronglycorrelated with

human

capital, technology

and

institutional quality, all of

which raisethe returns to capital

and

make

foreign assets less attractivefor rich countries. Second, the variability ofreturns is also negatively correlated with wealth,

making

foreign assets less attractive for rich countries. Third, there

may

also bescale

effects

whereby

returnsare higher in

some

ofthe larger, richer

economies

in our

sample. All ofthese

arguments can be

summarized

bysaying that it is likelythat rich

Lesotho is

somewhat

ofan anomaly, as itruns large current accountsurplusesreflecting

primarilyworkers' remittancesfrom SouthAfrica.

countries

have

better aggregate production functions

and

therefore findforeign

assets less attractive.

To

the extentthat this is the case, the simple regressionsofthe net foreign asset position

on

wealth contain omitted variables that bias

downwards

the slopecoefficient.

We

therefore introduce a

number

ofadditional control variables

into the regression.

We

proxyfor

human

capital with the

number

ofyears of

secondary

education in the workforce,

and

controlfor scaleeffects using the

logarithmof population.

We

include

openness

to international trade

measured

as

trade

volumes

as a share of

GDP,

financial depth

measured

as the ratio of

M2

to

GDP,

government consumption

as a share of

GDP,

and

an indexofcivil liberties, as

proxiesfor thequality ofthe institutional environment.

We

measure

thevariability of

returns using the standard deviation ofreal per capita

GDP

growth overthe indicated

period. Finally,

we

include a set of regional

dummies

to controlforother

unobserved

region-specific heterogeneity.

The

remainderofTable 3

summarizes

theresults ofthis

augmented

regression. Averaging overall years, thecoefficient

on

wealth rises to 0.07

and

remains very significant. Consistentwith the view thathigh levels of

human

capital

make

domestic capital

more

attractive,

we

findthatyears ofsecondary education

enters negativelyalthough not quite significantly atthe 10 percent level.

Somewhat

surprisingly, population size enters positively, suggesting thatthere

may

be

diseconomies

ofscale orcongestion effects that

make

domestic capital less attractive

in large countries. Public

consumption

as a share of

GDP

enters negatively

and

approaches

significance atthe 10 percent level, which

may

reflect

an

increased

demand

forforeign loans tofinance public consumption. Finally, financial depth

enters positively

and

significantly. This

may

reflect thefact that countrieswith

well-developed

financial markets

have

less

need

for recourse to international financial

markets.

The

remaining control variables, openness, civil liberties,

and

the volatility of

growth

do

notenter significantly. Although the

magnitude and

significance ofthe

coefficients

on

these variables differs

somewhat

across subperiods, the results are

qualitatively similar tothose obtained in thefirstcolumn.

From

Table 3

we

can also obtain a

sense

ofthe

magnitude

and

relative

importance ofwealth relativeto otherfactors as adeterminant offoreign assets

positions. Consider the regression in thefirst

column

ofTable 3

based on average

data over the period 1966-1997.

A

one

standard deviation increase inthe logarithm of

wealth (which correspondsto roughly athree-fold increase) leads to roughly

two-thirds ofa standard deviation increase inthe net foreign asset position, or about 10 percentage points. In contrast, a

one

standard deviation increase in the remaining

significant control variables leads to

an

increase (inabsolute value) ofonlyone-third ofa standard deviation inthe the net foreign asset position. Another

way

to see the

importance ofwealth in explaining the cross-country variation in foreign assets is to

perform a variance decomposition ofthefitted values from these regressions.

The

bottom

row

of Table 3reports the share ofthe variance in predicted foreign assets

that can

be

attributed towealth.^ _{Averaging overthe entire}

_sample

_period,

_we

_find

that almost

60

percent ofthe cross-countryvariation in predicted foreignassets is

due

to cross-countryvariation inwealth. In all butthe first subperiod

we

similarly findthat

the majority ofthe variation in predictedforeign assets is

due

to variation in wealth.

Fact 3:

The

share ofnetforeign assetsin the countryportfolio ispersistent overtime.

Figure4 plotstheshare ofnetforeign assets in wealth in a given country

and

year

on

thevertical axis, against its value lagged

one

year (inthefirst panel), five

years (in the

second

panel)

and

10 years (in thethird panel)

on

the horizontal axis,

pooling all country-year observationsoverthe period 1966-1997.

From

thesethree

figures it is clear that the shareofforeign assets inwealth is very persistent.^"

The

®

We

_{useadecomposition ofthevariance}

ofthe

sum

oftwocorrelated randomvariables

suggested byKlenow and Rodriguez-Clare[1997]. In particular, ifZ=X+Y, they define the

shareofthevarianceofZ attributableto

X

as COV(X,Z)A/AR(Z). This numberhas the

following natural interpretation: If

we

observethat

Z

is onepercent above its

mean

but

we

did

notobserve

X

and

Y

separately, then

we

would inferthat

X

is COV(X,Z)A/AR(Z) percent

above its

mean

value.

We

applythis defining Z asfitted foreignassets, and

X

as the estimated coefficienton wealth multiplied bywealth.

^°

Thisfact

was

also noted by KraayandVentura [Forthcoming] in a smaller sampleof 13

OECD

economies,

who

argue thatthisobservationcan explain thelong-standing puzzle

raised by Feldstein and Horioka [1980].

simple correlation

between

the shareofforeign assets in country portfolios in agiven

year

and

the

same

share lagged

one

year is 0.96.

Even

over a 10-year horizon the

simplecorrelation is a respectable 0.55. Thisvery strong persistence ofthe share of

foreign assets in country portfolios is all the

more

surprising

when

one

considers

how

small foreign assets are relative to wealth. In our

sample

the typical (median) country

holds roughly

-10

percentofits wealth in foreign assets,

and

its ratio ofwealth to

GDP

is around 2. This impliesthata current account surplusof5 percentof

GDP

sustainedfor only fouryears

would be

sufficient to entirely erase a country's net

foreign asset position. Yet in thedata

we

see

that netforeign assets as a share of

wealth

on average

barely

change

overthis horizon, as indicated by the estimated

slope coefficients which are very closeto one.

The

pooled data in Figure 4 nevertheless

mask

some

interesting variation

overtimein the persistenceofforeign assets in country portfolios. In Figure 5

we

disaggregate theannual persistence in thetop panel ofFigure4 by year. For

each

year indicated

on

the horizontal axis,

we

regress the share offoreign assetsin wealth

on

a constant

and

its one-yearlag,

and

then plotthe resulting slope coefficients

on

the vertical axis as a

dashed

line.

The

solid line

shows

a three-yearcentered

moving

averageofthese estimated slopes.

When

this slope is greaterthan (less than) one, the foreign assets ofall countries

on average

expand

(contract).

From

the

mid-1970s

to the mid-1980s, the slopes are all greaterthanone. This reflects primarilythe rapid

buildupofdebtofdeveloping countriesfinanced by borrowing from rich countries

and

oil producers, followed

by an

even

greater increase intheir recorded debts as

many

ofthese countries

suspended

payment

during the debtcrisis ofthe 1980s.

As

the

debt crisis

was

eventually resolved in the late 1980s, the portfolios ofall countries

contracted

and

theslopes fall

below

one.

Somewhat

surprisingly, the estimated slope

is near

one

during the 1990s, providing little evidence ofsystematic increases inthe

share ofnetforeign assets in country portfolios duringthis period.

Why

isthe netforeign asset position as ashare ofwealth so persistent?

One

immediate possibility is thatthe persistenceofnet foreign assets reflectsthe

persistence ofwealth, which

we

have

seen

to be

an

important determinant ofthe

cross-countryvariation in netforeign assets.

We

investigatethis iiypothesis empirically by regressing the

change

in netforeign assets

on

the

change

in the

logarithm ofwealth

and

the

changes

in the othercontrolvariables considered in Table

3, using the three eight-year

changes

implied bythefour 8-year

averaged

periods

using in Table 3.

We

then perform the

same

variance decomposition as beforeto

determine

what

shareofthe variation in

changes

in netforeign assets are

due

to

changes

in wealth.

We

find that

between 40

percent

and 70

percentofthe variation in

changes

in net foreign assets can

be

attributed to

changes

in wealth. This suggests

to us that persistence in wealth is an important source ofpersistence inforeign

assets.

Fact4:

Gross

foreign assetpositions aresmall

and

consistmostlyofforeign loans

ratherthan foreign equity.

Net foreign assets consistofnet claims

on

foreignequity (e*-e)

and

net

lending abroad, (l-l*).

We

illustrate the relative importance ofthe latter in explaining

the cross-country variation in country portfolios in two ways. Table 4

shows

the

composition offoreign assets as theyvaryacross regions,

income

groups,

and

time.

Averaging overall years, claims

on

foreign equity consist ofonly0.3 percent ofthe

wealthofdeveloping countries, while foreign claims

on

domestic equityaccountfor

2.8 percent ofwealth.

Among

industrial countriesclaims

on

foreign equity

and

foreign

claims

on

domesticequityare only

somewhat

larger

and

consist ofonly 3.3percent

and

3.9 percentofwealth." In contrastgross borrowing

and

lending accountfor 8.8

percent

and

4.5 percent ofwealth ofdeveloping countries,

and

1 1 percent ofwealth

each

for industrial countries. Finally, it is interesting to notethatthe shareofgross borrowing

and

lending in thewealth of industrial countries is

much

larger than that of

developing countries, especiallyduring the 1990s.

"

_A

_{largeliterature hasdocumentedthatthe holdingsof foreign equity of investors in afew}

rich countries are very small (Frenchand Poterba [1991],Tesar andWerner [1992] and others). Lewis[1999] provides an excellent surveyofalternativeexplanationsfor this empirical

regularity. Ourdata confirmsthat itapplies toa

much

broaderset ofcountries and assets.

In contrastwith Table

4

which focuses

on

gross positions, Figure 6 provides

evidence

on

the importanceofnet lending in explaining the cross-country variation in

net foreign assets.

We

plot net lending as a share ofwealth

on

the vertical axis,

and

the share of foreign assets in wealth

on

the horizontalaxis, averaging overthe period

1966-97.

The

top panel

shows

the results for all countries, while the bottom panel

shows

the

same

informationfor industrial countries only. Sincethe slope ofthis relationship is nothing

more

than the covariance

between

netforeign assets

and

net

lending divided by the variance ofnetforeign assets, this slope can

be

interpreted as

thefraction ofthe variancein net foreign assets attributable to variation in net lending.

In ourfull

sample

ofcountriesthisfraction is

82

percent. Forthe developing countries in our

sample

this is not surprising at all, since

we

have seen

thatgross equity

positions are small.

What

is

more

surprising isthatthe

same

number

applies to industrial countries

where

cross-border claims

on

equityare

much

more

prevalent

than in developing countries.

From

this evidence

we

conclude that

most

ofthe

cross-country variation in net foreign assets

can be

attributed todifferences in netlending

ratherthan inforeign direct

and

portfolio equity investment.

To

sum

up,

we

have

seen

thatthe share ofnet foreign assets in country

portfolios is small

and

typically negative (Fact 1), exhibits a strong positive

association with wealth in a cross-section ofcountries (Fact 2), is remarkably

persistent overtime (Fact3),

and

consists primarily offoreign loans ratherthan

foreign equity (Fact 4).

II.

Towards

an

Explanation

of

Country

Portfolios

We

next develop a

model

that

emphasizes

the interplay ofdiminishing returns,

production risk

and

sovereign risk in aworld populated by

homogeneous

mean-variance investors.^^ _{Diminishing returns}

_and

_{production risk implythat the}

risk-adjusted rate ofreturn to capital declines as

more

capital is invested in a country. If

these

were

the only forces atwork,

we

would observe all countries choosing the

same

portfolios,

and

the world distribution of capital stocks

would

be

determined by

the equalization ofrisk-adjusted rates ofreturn. Sovereign risk

however

generates a

home

bias in the

demand

for capital. Ifthis

were

theonly force atwork, countries

would

hold onlydomestic capital

and

the world distribution of capital stocks

would

mimic

the world distribution ofwealth.

The

set ofcountry portfolios

and

the world

distribution of capital stocks is

shaped

bythe interaction oftheseforces.

11.1

A

Model

of

International

Capital

Flows

The

world containstwo countries, North

and

South;

one

factor ofproduction,

capital;

and

a single

good

that can

be

used

for consumption

and

investment. This

good

is

used

as the numeraire.

Each

country contains a continuum of identical

consumer/investorsthat evaluate

consumption sequences

as follows:

(1) EJlnc(t)e"^' dt (5>0)

where

c is consumption.

The

time index will be omitted

whenever

this is not

confusing. Throughout,

we

use

an

asterisk todenote South variables.

We

assume

North has higher initialwealth than South, i.e. a(0)>a*(0).

^^_The

keypropertyofthese investors is thatthe shareofeach assetintheir portfolio does not

depend ontheirwealth, butonlyonthe

menu

ofavailable assets. By homogeneous,

we mean

thatall Investors haveidentical (homothetic) preferences, although possiblydifferentwealth

and

menu

ofavailable assets.

The

productiontechnology is quite simple. Let k

and

k*

be

thecapital stocksof

North

and

South.

To

produce

one

unitofcapital,

one

unitofthe

consumption

good

is

required. Since capital is reversible, the price of

each

unitis always

one and

its return

is the flowofproduction net of depreciation. Letco

and

co*

be

two standard

Wiener

processes with independent increments. That is, E[do)]= E[dco*]=0, E[dco^]=E[dcL)*^]=dt

and

E[do>do/]=0.

Then

the flow ofproduction netofdepreciation is given by

Ttdt+adcain North

and

7T;*dt+adco* in South;

where n

and

n* are short-hand for 7t=9k"^

and

7i:*=0k*"'' (0<y<1; 9>0)

and a

is a positive constant.

The

parametery

measures

the strength ofdiminishing returns which, for simplicity, are treated here as

an

externality or congestion effect.^^

_The

_parameter

_a

_measures

_{the importance of}

country-specific production risk. Therefore, thisformulation

assumes

thatboth

countries

have

the

same

technology

and

embodies

the

two

forcesthat

make

the

risk-adjusted rate of return to capital decreasing in the capital stock

and

create incentives

forcapital to

move

from rich to poorcountries.

Domestic

investors

own

the domesticstock of capital

and

can enter into loan

and

equity contracts with foreign investors. International loans promiseto

pay an

instantaneous interest rate rdt. At the beginning ofthe period, the lender gives the

principal tothe borrower. Atthe

end

ofthe periodthere are two possible outcomes.

The

borrower might honor its promise, in which

case

the lender receives the principal plus interest.

The

borrower mightalsodefault

on

its promise, in which

case

the

borrower

keeps

the principal

and

interest

and

the lender receives nothing.

A

shareof

North (South) equityhas pricev (v*)

and

promises to

pay

the net flow ofproduction

generated by

one

unitofNorth (South) capital. At the beginning ofthe period the

buyergives the value ofthe equityto the seller.

The

sellerprovides the buyerwith a

unitofcapital

and

allows him/herto operate the production technology.

Once

again,

at the

end

ofthe period there aretwo possible outcomes.

The

sellermight honor its

promise, inwhich

case

the buyerreceives the value ofthe equity

and keeps

the net

flowofproduction.

The

sellermight alsodefault

on

its promise, in which

case

the

^^_At

thecostoffurthernotation,

we

couldgeneratethisdependence byassumingthere isa

productionfactor thatis not priced. Sincethis is wellknown,

we

dispense withtheformalities.

seller

keeps

the value ofthe equity

and

then pays a costX (0<X<1)to repossess the

unitof capital

and

the net flowof production.

The

buyer receives nothing. Let I bethe

amount

of lending from North to South

and

e (e*)

be

the

number

ofNorth (South)

shares

owned

by South (North).

It isevidentthat international loan

and

equity contractswill be

used

in

equilibrium if

and

onlyifthe probability thatthey are

honored

ishigh enough. This

observation raises a familiartime-inconsistency problem. Since

governments

cannot

punishforeign citizens, internationaltrade in assets relies

on

governments'

willingnessto punishtheir

own

citizens iftheydefault

on

theirobligations towards

foreigners. 'Ex-ante' both

governments

would like to

commit

to

do

this

and

allow

investors toexploit beneficial trade opportunities.

However,

'ex-post'

governments do

not

have an

incentive to punish default ifthe benefits

exceed

the costs.

The

benefits of default are clear. But,

what

are the costsofdefault?

We

shall

assume

they vary

overtime. Let s={0,1}

be

the state ofthewortd. During 'normal times' (s=0) both

countries can credibly

commit

to retaliatewith penalties that are large

enough

to

ensure that default never occurs. During 'crisis periods' (s=1) countries cannot

credibly

commit

to penalties

beyond

retaliation in kind.

As

a result, if s=1 South (North) defaults ifits netforeign asset position is negative enough, i.e.

ve-l-v*e*<-Ae*

(v*e*+l-ve<-A,e).^'* Let

a

dt

and

|3dt

be

the probabilitiesthat the wortd transitions

from

8=0

tos=1

and

vice versa;

and

assume

thesetransitions are independentof

production shocks, E[dcods]=E[dca*ds]=0.

The

value ofds is revealed afterthe

beginning-of-penod

payments

of loan

and

equity contracts

have

already

been

made.

What

is the equilibrium probabilityofdefault?

Assume

investors believe the

probability of default is zero. Ifa*>a(1-A), the country portfolios

chosen

by investors

are consistentwith this belief.^^ In this case, the equilibnum probability ofdefaultis

^'*

The seminal papers on sovereign risk andthe abilityofvarioustypes of penaltiestosustain trade areEaton and Gersovitz [1981] and Bulowand Rogoff [1989]. Eaton and Fernandez

[1995] and Obstfeldand Rogoff[1996, chapter6] aretwoexcellent surveysofthis topic.

^ _{If investors believe thatthe probability of defaultiszero, the equilibrium of the model implies}

v=v*=1, k=k*=(a+a*)/2, e=a*/2, e*=a/2and 1=0. Naturally, no countrydefaults ifs=0. Buteven ifs=1, Northdoesnot default and neither does Southifa*>a(1-X).

zero

and

sovereign riskis simply not

an

issue.

We

shall therefore

assume

from

now

onthat a*(0)<a(0)-(1-X). In this case, the equilibrium probability of default is

state-dependent.

Assume

the initial state is s=1

and

investors believe that defaultoccurs if

the state

does

not change. Then, investors

do

not purchase foreign loans

and

equity

since they expectdefault to occurwith probability close toone, i.e. 1-|3dt. But then

both countries are indifferent

on whether

to default or not.

To

ensurethat

an

equilibrium exists

we

assume

theydefault(on their non-existentforeign obligations)

so thatthe beliefs ofinvestors are consistent. This implies thatthere is

no

trade In

assets during 'crisis periods'.

Assume

nextthatthe initial state is

s=0 and

investors

believe that default occurs ifthe state changes. In this case, investorswill

purchase

foreign loans

and

equitysince default occurs onlywith avery small probability, i.e. adt. Ifthis probability is not too large, the

chosen

country portfolios are consistent

with South defaulting in the unlikelyeventthe state ofthe world changes. Otherwise,

there is

no

Nash

equilibrium in which the countryfollows a purestrategy.^®

We

shall

restrictthe analysis to the case in which

a

is small. This implies that

adt

is the

equilibrium probability of default in 'normal times'.

To

sum

up, theworld

economy

exhibits periods oftrade in assets that

culminate in crises (s transitions from

s=0

to s=1) inwhich the debtor country

defaults.

The

parameters

a

and

A.

measure

the probability

and

the destructiveness of

this crisis. After itoccurs, a period

ensues

in which there is

no

trade in assets.

Eventually, international trade in assets

resumes

(stransitionsfrom s=1 tos=0)

and

the cycle starts again. Although in normal times there might

be

substantial trade in

assets, the (small) probability that acrisis occurs

has an

important effect

on

the

strategies followed by investors.

We

describethese strategies next.

^^

_We

canconstruct a mixed-strategyNash equilibrium asfollows: Let p bethe probabilityof defaultthat leads investorstochoosecountryportfolios suchthatve-l-v*e*=-Ae*. South is

indifferent between defaulting ornot. There is an equilibrium in which Southdefaults with

probability (p/a)dt.

11.2

Investment

Strategies

In crisis periods, the only decision that investorsface is

how much

to

consume

and

save. This limited choice is

embedded

inthis budget constraintwhich applies

onlyduring crisis periods:

(2)

da

=

(7ia-c)dt

+

aadco

In normal times there istrade in assets,

and

we

can writethe budget

constraint ofthe representative investorin North asfollows:

(3)

da

= [7r(k-e) +

7i*e*+rl-c]dt

+(7-[(k-e)d(o+

e*d(o*]+[(v-A)e-v*e*-l]ds

where, ofcourse, the following restriction applies:

(4) a=

k-ve

+

v*e*+l

The

firsttwo terms ofthe budget constraint (3) arestandard

and

show

how

the

expected return

and

volatilityofwealth

depends on

asset choice, conditional

on

the

state ofthe world not changing.

The

thirdterm describes the wealth

shock

that the

investorexperiences atthe onset ofa crisis period. Throughout,

we

rule out Ponzi

schemes

and impose

short-sale constraints

on

the holdings offoreign equity. This

last restriction is a logical implication ofsovereign risk.

To

determine the optimal consumption

and

portfolio rules, the representative

consumer

in North

maximizes

(1) subject to (2)-(4)

and

the

dynamics

ofasset

phces

and

their return characteristics, i.e. the laws ofmotion ofr, v , v*, n, n*,

a and

a*.

Since the representative

consumer

is infinitesimal, he/she understands that his/her

actions

have no

influence

on

these prices

and

their evolution.

The

representative

consumer

in the South solves a similar problem.

Appendix

B

shows

that thefirst-order

conditions associated with the investor's problem can

be

written as follows:

(5) c

=

6a

(6) 2

k-e

7T:-p

=

a

a (7) a r p

=

a-_

^

k-^e

2

k-e

,, , a (8)

p-v-n

=

-a

+

a(A-v)

a

k-Xe

e*

(9) If7i*-r-v*<0, e*=0. Otherwise, re*

-p

•v* =a^ +

a

v

k-Xe

where

pis the multiplierassociated with constraint(3) divided bythe marginal utility of

wealth. This quantity can

be

interpreted as the risk-free rate that applies

on

loans

between

North residents.

A

similar set offirst-orderconditions applyto South.

Equation (5) is thefirst-order condition associatedwith c;

and

shows

the

familiar result that

consumption

equalsthe annualized value ofwealth. Sincethe rate oftime preference is

used

as the discount factor, the

consumption

rule is independent

ofthe state ofthe world.

Equations (6)

and

(7) are the firstorder conditions associated with k

and

I;

and

describe

how

investors value production

and

sovereign risk, respectively. Equation

(6) says thatthe

premium

for holding production risk, n-p, is the covariance

between

the return to

one

unitof capital

and

one

unitofthe investor'sportfolio, a^(k-e)/a.

Equation (7) says thatthe

premium

forholding sovereign risk, r-a-p, is also

proportionalto the covariance

between one

unit ofloans

and

one

unitofthe investor's

portfolio, i.e.

a

. Butsince this time the effectofthe

shock

is 'large' or

a

non-local,thefactor of proportionality is not

one

butthe ratioofthe marginal valueof

wealth before

and

aftera crisis occurs, i.e. .

k-Ae

Equations (8)

and

(9) are thefirst-orderconditions associated with e

and

e*;

and

describe the supplyofNorth equity

and

the

demand

for South equity,

respectively. Equation (8) can

be

interpreted asdetermining the price atwhich North iswilling to sell equity.

Each

share sold bythe North reduces

income

by

one

unitof output, but also provides a gain of 1-A.in the eventofacrisis. This is

why

equitywill

be

sold at adiscount, i.e. v<1. Using Equations (6)

and

(7),

we

find thatthis discount is equal tothe gain obtained in the event of a crisistimes the price of

one

unitof

income

in this state oftheworld, i.e. 1-v=(1-^)

—

-. Equation (9) defines the

r

demand

forSouth equity. Thisasset contains both production risk

and

sovereign risk

and

the required

premium

reflects just this.

Thiscompletes the description ofthe model. Equations (4)-(9)

and

their

counterparts for Southjointly determine asset prices (p,p*,r,v,v*), the world

distribution of capital stocks (k,k*)

and consumptions

(c,c*);

and

the pattern ofasset

trade (e,e*,i) as afunction ofthedistribution ofwealth (a,a*)

and

technology

(n,7i*,a,a*).

We

use this set ofequations toderive the cross-sectional implications of

thetheory. Then,the budgetconstraints (2)-(3)determine the lawofmotion ofthe

world

economy

asa function ofthe initial distribution ofwealth

and

technology

and

the realizations ofthe shocks.

We

usethis additional set ofequationstoderive the

time-series implications ofthe theory.

11.3

Three

Examples

Before

we

embark

ina quantitative analysis ofthe model,

we

discuss three

examples

or special cases that help build intuition

on

the role ofthe differentforces

thatthe theory emphasizes. In these examples,

we

assume

the world

economy

starts

in normal times

and

then ask:

What

is the initial pattern oftrade in assets

and

how

does

it evolve overtime?

EXAMPLE

#1: Let

a^O

sothatthere is

no

sovereign risk. Inthis case, there is

no

discount

on

international assets, i.e. p=p*=r

and

v=v*=1. Since countries

have

identical

and

homothetic preferences and, effectively, the

same menu

of assets,they

choose

the

same

portfolios.

As

a result, there is

no

borrowing or lending. Since

technologies are identical, both countries invest

50

percent oftheirwealth in domestic

capital

and

the restin foreign equity. Thus, halfoftheworld capital stock is located in

each

country

and

theworld interest rate equalsthe marginal product ofcapital:

(10) k =k*

=

-^+ ^^ 2 (11) r

=

e

^a+a*v^

\ ^ J

If

one

interprets North

and

South as the set of industrial

and

developing

countriesin Table 1,

we

have

that =0.8;

and

netforeign assetpositions (as a

a+

a*

percentage ofwealth) of North

and South

are37.5

and

-150 percent, respectively.

The

distribution of wealth is constant overtime, since both countries

choose

the

same

portfolios

and

therefore

have

the

same

growth rate:

da

da

(12)

^

=

^^

=

(r-p).dt +

-(dco

+dco*)

a

a*

2 ^ ^

An

implication ofthis is that netforeign asset positions are time invariant.

The

predictions ofthis

example

are both well-known

and

dead

wrong.

The

model

exhibitstwo featuresthat are present inthe data: net foreign asset positions

are positivelyassociated with wealth (Fact 2)

and

very persistent (Fact 3). But the

model

alsopredicts netforeign asset positions thatare

much

largerthan those in the

data (Fact 1). Italso predicts very large foreign equity positions

and no

borrowing

and

lending, whilethe data

shows

thatcountries hold little foreign equity

and

finance

most

oftheirnetforeign asset positions with foreign loans (Fact4).

EXAMPLE

#2

: Let

a^O

so thatthe only reason

why

capital flowsfrom rich to

poor countries is to exploit higherrates ofreturn.

Now

foreign equity

and

loans deliver

the

same

return in normaltimes. IfX=0, foreign equity

and

loans also deliverthe

same

return inthe event ofdefault and, consequently, the composition offoreign

assets is indeterminate. IfX is strictly positive,foreign equity

becomes

a

dominated

asset. In this case, countries

do

not holdforeign equity

and

finance theirnetforeign

asset positions with foreign loans.

The

world distribution of capital stocks

and

the

world interest rate are implicitlydetermined by:

(13) r

=

ek-^+a-k (14) r=

ek*"^+a-(15) k+k*

=

a +

a*

k^ \3'

Equations (13)

and

(14) describe the

demand

for North

and

South capital,

and

Equation (15) is the market-clearing condition.

These

equations

show

that sovereign

risk creates a

home

bias inthe

demand

forcapital. Also, notethat theworld interest

rate

exceeds

the marginal product ofcapital in both countries. This difference isthe

risk

premium

required to

compensate

lendersforsovereign risk.

The

toppanel of

Table 5

computes

the foreign asset position of North

under

alternative

assumptions

for_Y

and

a (Remember

the net foreign asset position of South is

-A

times that of

North). Netforeign assetpositions for North

and South

range from zero (as a^'>=) to

37.5

and -150

percent (as

a^O).

What

are reasonable valuesfor

a

and

y? For

instance,

a=0.02

is consistent with the 20'*^ _{century experience which features two}

episodes of large-scale defaults in the

1930s and

the 1980s. Since the expected

value ofaggregate production is 9k^'^, values of yaround 0.6 correspond to the

neoclassical growth model, while values of ynear

one

are close tothe lineargrowth

model. Table 5

shows

that if

a=0.02 and

y=0.6, the net foreign asset positions of

North

and

South are 20.3

and

-81.2 percent.

These

numbers

are roughly halfof

those

we

found in the previous example. Ify=0.1, the netforeign asset positions are