MIT LIBRARIES
661^
3 9080 01917
Digitized by the Internet Archive
in
2011 with funding from
Boston Library Consortium IVIember Libraries
http://www.archive.org/details/countryportfolioOOkraa
DE\A/BY
Massachusetts Institute of Technology
Department of Economics
Working Paper Series
COUNTRY PORTFOLIOS^
Bank
Aart Kraay, The World
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
http://papers.ssm.com/paper.taf?abstract
id=XXXXXX
iNSTlTUTE
OF TECHNOLOGY
'M^^'CHUSETTS
OCT 2
5
LIBRARIES
Massachusetts Institute of Technology
Department of Economics
Working Paper Series
COUNTRY PORTFOLIOS^
Bank
Aart Kraay, The World
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
http://papers.ssm.conVpaper.taf/abstract
id=XXXXXX
at
Abstract
How do countries hold their financial
on
capital located at
countries from 1966
home and
to 1997.
wealth?
We construct a new database of countries' claims
abroad, and international borrowing and lending, covering 68
We find that a small amount of capital
poor countries. Countries' foreign asset positions are remarkably
form of foreign loans rather than foreign equity.
To
persistent,
and mostly take the
model of intemafional capital flo9ws that highlights
between diminishing returns, production risk and sovereign risk. We show that in
interpret these facts,
the interplay
we build
flows from rich countries to
a simple
the presence of reasonable diminishing returns and production risk, the probability that
international crises occur twice a century
is
enough
to generate a set
of counfry portfolios
that are
roughly consistent with the data.
*We thank Mark Aguiar, Daron Acemoglu,
Lucas and Mark Wright for
insightful
Giancarlo Corsetti, Raquel Fernandez, Robert E.
comments on
Ferretti for his advice in the construction
a previous draft and Gian Maria Milesi-
of our dataset.
We are
also grateful to Cesar Calderon,
Alex Karaivanov, George Monokroussos, and Rashmi Shankar for excellent research assistance.
The opinions expressed here are the authors' and do not necessarily reflect those of the World
Bank, its executive directors, or the countries they represent. Financial support from the
Regional Studies Program of the Latm American Region of the World Bank is gratefully
acknowledged.
Introduction
How do
description of country portfolios
main features. By country
how
it
is
Our
countries hold their financial wealth?
objective
we
portfolio,
we
the net foreign assets of a country,
seem
to
net foreign asset positions?
refer to the country's holdings of foreign equity
is
equity? With the answers at hand,
To determine
we go
of
to the theory
and
1997 and includes 68 countries
account
trade. Constructing this
them and
draw from
this
regularities
1
.
we
fill
in
for
database forces us
make
the gaps.^ Despite
data are robust.
highlight here.
In fact,
The
How
persistent are
and ask:
we
Why?
construct a
there
this,
is
over 90 percent of world production
to
choose among fragmentary and
(heroic?) assumptions on
we
20 percent
in
to
we
nothing subtle about the empirical
following are
all
very striking features of the data:
in
absolute value and
negative for most countries. Roughly 80 percent of the observations
whose
how
feel confident the inferences
Net foreign asset positions as a share of wealth are small
consist of countries
new
Our sample covers the period 1966-
loans.
imperfect sources of information and then
reconcile
country characteristics
domestic equity and loans, and domestic
residents' holdings of foreign equity
that
We first turn to the
the relative importance of foreign loans and
the main features of country portfolios,
database on foreigners' holdings
and
What
positive net foreign asset positions?
What
and
domestic capital and various foreign assets. By
large are net foreign asset positions?
be associated with
of their
refer to the financial wealth of the country
and loans minus foreigners' holdings of domestic equity and loans.
How
to provide a
and advance a parsimonious explanation
distributed across holdings of
data and ask:
is
in
our sample
net foreign assets as a share of wealth are less than
absolute value.
We also find that the net foreign
negative for about 80 percent of the observations
in
asset position
is
our sample.
See Sinn [1990], Rider [1994] and Lane and Milesi-Ferretti [1999] for alternative sources of
data on foreign asset positions.
^
Introduction
How do
description of country portfolios
main features. By country
how
it
is
seem
to
we
portfolio,
the net foreign assets of a country,
and loans minus
net foreign asset positions?
we
What
is
of their
and
refer to the country's holdings of foreign equity
and loans.
What
We first turn to the
country characteristics
positive net foreign asset positions?
equity? With the answers at hand,
To determine
to provide a
domestic capital and various foreign assets. By
large are net foreign asset positions?
be associated with
is
refer to the financial wealth of the country
foreigners' holdings of domestic equity
How
objective
and advance a parsimonious explanation
distributed across holdings of
data and ask:
Our
countries hold their financial wealth?
How
persistent are
the relative importance of foreign loans and
we go
to the theory
and ask: Why?
the main features of country portfolios,
we
construct a
new
database on foreigners' holdings of domestic equity and loans, and domestic
residents' holdings of foreign equity
and
1997 and includes 68 countries
account
and
trade. Constructing this
that
them and
draw from
this
regularities
1
.
we
fill
in
for
database forces us
imperfect sources of information and then
reconcile
Our sample covers the period 1966-
loans.
make
the gaps.^ Despite
data are robust.
highlight here.
In fact,
The
there
this,
is
over 90 percent of world production
to
choose among fragmentary and
(heroic?) assumptions
we
20 percent
in
we
nothing subtle about the empirical
following are
all
very striking features of the data:
in
absolute value and
negative for most countries. Roughly 80 percent of the observations
whose
to
feel confident the inferences
Net foreign asset positions as a share of wealth are small
consist of countries
on how
in
our sample
net foreign assets as a share of wealth are less than
absolute value.
We also find that the
negative for about 80 percent of the observations
net foreign asset position
in
is
our sample.
See Sinn [1990], Rider [1994] and Lane and Milesi-Ferretti [1999] for alternative sources of
data on foreign asset positions.
^
2.
There
is
a strong positive relationship between financial wealth and the net
foreign asset position
when
in
a cross-section of countries.
financial wealth doubles, the
by three to
seems
six
even
The share
of net foreign assets
a R^ of 93 percent. Since
variation in
wealth,
it
changes
wealth increases
in
net foreign asset
a set of variables that are designed to capture
wealth
is
very persistent over time.
seems reasonable
A simple
lagged value delivers a slope coefficient of 0.98 and
its
we
in
find that
to
between 40 percent and 70 percent of the
conclude that persistence
important source of persistence
4.
find that
financial wealth
net foreign assets can be attributed to
in
in
we
aggregate production functions.
in
regression of this share on
in
sense) most of the variation
after controlling for
cross-country differences
3.
share of net foreign assets
percentage points. Cross-country variation
to explain (in a statistical
positions,
In particular,
in
changes
in relative
in
relative
wealth
is
an
foreign assets.
Gross foreign asset positions are small and consist mostly of foreign loans rather
than foreign equity.
In
developing countries, foreign equity assets and
liabilities
are roughly 0.3 and 2.8 percent of wealth, while foreign loan assets and
account
for 4.5 percent
equity assets and
liabilities
foreign loan assets
To sum
up,
and 8.8 percent
and
of wealth. In industrial countries, foreign
are roughly 3.3 and 3.9 percent of wealth, while
liabilities
we observe
liabilities
account
for
1 1
percent of wealth each.
that net foreign asset positions are small relative to
wealth and tend to be negative, except for a few rich countries. These net foreign
asset positions are remarkably persistent as a fraction of wealth, and mostly consist
of foreign loans rather than foreign equity holdings. This picture that
the data
is
so clear that
we
think
it
emerges from
should constitute the main target of any successful
theory of international capital flows.
It
seems safe
to
argue that such a theory requires at least two ingredients. To
explain the strong positive association between wealth
positions, the theory
must contain
at least
one force
and net foreign asset
that creates incentives for capital
to
move from
poor countries. Natural candidates for
rich to
and country-specific production
returns at the country level
this role
risk.
If
are diminishing
either of these
forces are present, the risk-adjusted rate of return to capital declines as
is
invested
a country, creating an incentive to invest
in
capital. In the
eliminated
if
absence
keep most
risk.
we
in rich
While
second ingredient
variables
popular view
of their capital
this
why
net foreign asset
needs
to recognize that
ingredient to explain
that the theory just
is
countries even
is likely
to
this is
why
investors
the presence of diminishing returns
in
be true to some extent, we look elsewhere
of the theory, for three reasons.
First,
we
find that
standard
(human
think are associated with better aggregate production functions
capital, quality of institutions,
and others) seem
to
be either unrelated to the net
foreign asset position of a country or, alternatively, explain very
(See Table 3 and the discussion
production functions
in rich
small, they cannot explain
of
it
countries can explain
why gross
predict that they
why
of
its
variation
net foreign asset positions are
foreign equity positions are also small.
risk,
choose large gross foreign equity positions
balanced. Finally, better aggregate production functions
why most
little
below). Second, while better aggregate
extent that investors have a desire to diversify production
explain
little
would only be
have better aggregate production functions, and
and production
for the
A
capital
were equalized across countries.
Hence the theory needs a second
rich countries
countries that have
in
of a counten/ailing force, this incentive
capital stocks
positions are so small. ^
more
two
To
the
the theory would
that are roughly
in rich
countries cannot
international trade consists of loans rather than equity. While both
assets are useful to transfer capital across countries, equity has the additional benefit
of allowing countries to share production risk
preferred over loans.
Why
and should therefore always be
are observed foreign equity positions so small?
foreign loans rather than foreign equity the asset that
To answer these questions, we would
still
need
to
add
theory anyway.
See Lucas
is
[1990] for a review of alternative possibilities.
Why
are
most traded internationally?
additional elements to the
In this
paper
we
explore the alternative hypothesis that sovereign
be the second ingredient that the theory needs. ^
In
hedge against the
demand
Should
for capital that
we
positions?
in
risk of foreign default.
might explain
why
might
the presence of this sort of
domestic capital offers domestic investors not only the value of
also a
risk
This creates a
its
risk,
production flow, but
home
bias
in
the
net foreign asset positions are small.
also expect that sovereign risk leads to small gross foreign equity
Is
it
even possible
that sovereign risk explains
why most
international trade
assets consists of loans rather than equity? The answers to these questions
depend
crucially
Assume
on the consequences
first
that
if
for investors of
a country defaults on
its
a foreign
foreign obligations, foreign
owns abroad and then using
countries respond by seizing the assets that this country
these assets to
(partially)
compensate
Assume
creditors.
which assets are seized and transferred
default.
to creditors
also that the process by
does not give
loss of value. In this case, the loss suffered by creditors
in
rise to
any costs or
the event of default
is
determined by their net foreign asset position vis-a-vis the defaulting country. To
minimize exposure to sovereign
positions. But they
do not have
risk,
choose small net foreign asset
investors then
to hold small gross foreign equity positions. In fact, to
the extent that investors have a desire to diversify production risk they would again
choose large foreign equity positions
assumptions, sovereign
are small, but
can
it
explain
If
it
risk
that are roughly balanced.
provides a rationale for
why
Under these
net foreign asset positions
cannot explain why gross foreign equity positions are small. Neither
why most
we want
international trade consists of loans rather than equity.
to hold sovereign risk responsible for the small gross foreign equity
positions observed
in
the data,
we need
to
remove
at least
one
of the
assumptions
of
the previous paragraph. Perhaps legal systems do not allow creditors to seize the
foreign assets of defaulting countries. Or, even
^
if
they do, the transfer of ownership
Obstfeld and Rogoff [1996, p. 349] define sovereign risk as referring to "... any situation in
which a government defaults on loan contracts with foreigners, seizes foreign assets located
within its borders, or prevents domestic residents from fully meeting obligations to foreign
creditors." This is a good description of what we have in mind.
might involve large costs and a substantial loss of value.
exposure
to sovereign risk
is
no longer
case, investors'
In this
their net foreign asset position vis-a-vis the
defaulting country, but instead sonne fraction of their gross foreign asset position.
Now the
existence of sovereign
risk
can
potentially explain
why
investors
choose
small gross foreign equity positions.
With the help of the additional (and
ownership of equity
transferring
sovereign
risk
is
we
think realistic)
costlier than transferring
assumption that
ownership of loans,
can also explain why most international trade
assets consists of
in
foreign loans rather than foreign equity. Investors issuing foreign loans
willing to sell
they receive
these assets at a discount that reflects the value for them of the gain
in
the event of
What
default.''
borrowed abroad receives the
full
claims to foreigners receives the
so on). This gain
is
to better
gain?
A
domestic investor who has
full
is
exactly equal to the
A domestic investor who
has sold equity
value of the equity, but loses any valuable
advantage that foreign investors brought
know-how, access
is this
value of the loans. This gain
loss experienced by the foreign investor.
in
and equity are
to the firm (experience,
managerial
skills,
technology or relationships, firm-specific knowledge, and
less than the loss experienced by the foreign investor. Therefore
the event of default, foreign loans give rise only to pure transfers while foreign
equity creates losses. While the latter allows investors to
risk,
it
is
hedge against production
a worse hedge against sovereign risk than the former.
If
the desire to avoid
diminishing returns induces investors to transfer capital from rich to poor countries,
foreign loans
production
will
risk.
be a more
attractive asset
Thus, sovereign
risk
if
sovereign
has also the
risk is
high relative to
potential to explain
why
loans are
the preferred asset to finance capital flows.
The
notion that foreign equity
novel or controversial
interesting issue
"*
It
Is
among observers
whether reasonable
to sovereign risk
of international financial markets.
probabilities of default
is
hardly
The
can quantitatively
in loan contracts shows up In the Interest rate. The evidence Is overwhelming
developing countries usually command a higher Interest rate than domestic
much less clear whether we could find any comparable discount on equity.
This discount
that loans to
loans.
is
and loans are subject
explain the main features of country portfolios.
To determine
this,
we
construct a
simple North-South model of international capital flows. The production technology
exhibits diminishing returns at the country level
model, the world
assets, which
economy experiences
end up
in
a
crisis
period
and country-specific
risk. In this
periods with substantial international trade
in
which South defaults on
its
in
foreign
obligations. North investors seize South's foreign assets, but this transfer of assets
costly.
The
default initiates a crisis period in which international financial markets shut
down. Eventually, international trade
above.
is
It
seems
If
sovereign
clear that this
in
assets resumes and the cycle starts again.
model can,
in
risk is sufficiently high,
principle, explain the facts
net capital flows
will
either diminishing returns or country-specific production risk
observe a tendency
for capital to flow
from
is
be small (Fact
important,
poor countries (Fact
rich to
extent that transferring ownership of foreign equity
is
discussed
2).
we
1).
If
should
To
the
costlier than transferring
ownership of loans, gross foreign asset positions should be small and consist
primarily of loans rather than equity (Fact 4).
If
the model generates persistence
in
the
world distribution of wealth, this can naturally explain the persistence of foreign asset
positions (Fact 3).
It
is
much
less clear
however whether
a reasonable quantitative description of the data.
even
in
A
this
model
able to provide
is
perhaps surprising finding
the presence of reasonable diminishing returns and production
probability that international crises occur twice a century
is
enough
to
risk,
is that,
the
generate a set
of country portfolios that are roughly consistent with the data.
The paper
is
organized as follows: Section
I
provides a brief description of our
database and extensively documents the four facts mentioned above. Section
presents the model and discusses
Appendix
A
its
main
qualitative
provides a detailed discussion of
Appendix B contains some
proofs.
how we
and quantitative
II
implications.
constructed our database, while
A
I.
Description of Country Portfolios
In this section,
we
describe the main characteristics of country portfolios using
overview of the sources and
We first provide an
methodology used to construct the data. We then
summarize
in
a new database covering 68 countries from 1966
its
main features
the data and describes
A New
1.1
to 1997.
the form of four facts. Appendix
how we account for changes
in
A
provides details on
the value of stocks of assets.
Database on Country Portfolios
Our database contains estimates
assets of countries.
^
In particular,
of
domestic capital stocks and foreign
we have measures
for the following quantities:
k = Domestic capital stock
e = Domestic equity owned by foreign residents.
e* = Foreign equity
owned by domestic
residents.
= Loans issued by domestic residents and owned by foreign residents.
I
I*
= Loans issued by foreign residents and owned by domestic residents.
We measure these quantities
per domestic resident
in
constant 1990
US
dollars.
We
define a=k+e*-e+r-l and f=e*-e+r-l as the financial wealth or portfolio of the country
and
its
its
net foreign assets, respectively.
First,
we use
as a creditor (debtor)
if
initial
value. Second,
We abstract from
capital.
component
of financial wealth
in
two steps.
the limited available information on stocks of these assets to determine
we use
these assets to extend the
^
refer to a country
net foreign assets are positive (negative).
We construct estimates of each
an
We
other
initial
flow data
and estimates of changes
in
the value of
stocks forward and backward over time.
components
of wealth
such as
land, natural resources,
We rely on
and human
this
method
available
in
of cumulating flows
even
for
those countries where more stock data
is
order to avoid a potential bias. Since long time series of stock data are
few
available only for a
essentially result
rich countries,
using these as the primary source would
methods being used
in different
and poor
to construct stocks for rich
These differences would then contaminate our inferences regarding how
countries.
portfolios vary with wealth.
We rely on
of
data from a number of standard sources.
We obtain
initial
stocks
domestic capital from the Penn World Tables, and use flow data on gross domestic
investment to build up stocks of capital valued
in
US
PPP.
dollars at
In
order to
determine foreign holdings of domestic equity and domestic holdings of foreign
equity,
we
rely primarily
Investment reported
Yearbook.
Statistics
reported
in this
in
on data on stocks and flows of
direct
and
portfolio equity
the International Monetary Fund's Balance of
We again
and a variety
use the
limited available information
of other sources to determine the
Payments
on stocks
initial
level of
each
asset for each country, and then use corresponding flow data from the balance of
payments
to construct stocks for remaining years.
the debts of developing countries reported
in
Finally,
we combine
stock data on
the World Bank's Global Development
Finance with data on stocks and flows on debt from the Balance of Payments
Statistics
Yearbook
to build
up stocks of borrowing and lending
for
all
countries
in
our
sample.^
Our sample
with a
sample
of
of countries
is
determined primarily by data
98 countries with population greater than one
GDP greater than
1
000 US
dollars at
PPP
in 1
990. Of these
with missing, incomplete, or inconsistent balance of
with an unbalanced panel of
between 1966 and 1997.
In
availability.
million
we
We
begin
and per capita
discard 25 countries
payments data. This leaves us
73 countries spanning on average 25
the empirical work that follows
we
of the years
restrict attention to
a
We assume throughout that stocks of debt reported in these sources constitute solely the
assets and liabilities of domestic residents. To the extent that these reflect debt issued by or
owed to foreign-owned firms operating in the country, we are overestimating the loan assets
and liabilities of domestic residents. Given that foreign holdings of domestic equity are small
relative to wealth this mismeasurement is unlikely to be empirically important.
®
set of
68 countries excluding
classified
all
five transition
economies. Table
1
lists
these countries
by geographical region. As the table shows, our sample includes basically
and a substantial number
industrial countries
regions of the world.
The
countries
in
It
is
reasonable to think that these countries
also account for a similar fraction of world trade
coverage,
we do
sample.
the case of claims on equity,
not find that net foreign assets
claims on foreign equity
is
all
our sample account for over 90 percent of
world production and commodity trade.
In
of developing countries from
we
in
assets.
sum
to
find that the
''
Despite this wide
zero across countries
sum
of
all
in
our
countries reported
on average about 3 percent less than reported foreign
claims on domestic equity.
In
the case of lending the discrepancy
is
world
larger, with
reported borrowing exceeding lending by about 12 percent. While these
discrepancies are not
trivial,
known discrepancies
in
1.2
they are of a comparable order of magnitude to well-
flows on foreign assets.
Main Features of Country Portfolios
In this
subsection
we examine
database described above.
the main features of country portfolios using the
We organize the discussion
around four main facts or
findings.
Fact
1
Net foreign assets as a share of wealth are small and negative
for
most
countries.
Figure t
shows the
distribution of the
share of foreign assets
in
wealth,
—
a
pooling the available 1717 observations for
all
countries and years. Overwhelmingly,
net foreign assets represent a small fraction of the wealth of domestic residents.
Our procedure results in estimates of wealth that are very small (and occasionally negative)
few country-year observations, corresponding to countries with very large external debt.
We exclude these observations by limiting the sample to those where the ratio of wealth to
''
for a
GDP
is
greater than 0.5.
Roughly 80 percent of the country/year observations are less than 20 percent
in
absolute value. Also, about 80 percent of the country-year observations are negative.
Perhaps Figure
is
1
concealing important variation across regions and over time
the size and sign of the net foreign asset position.
reports the share of net foreign assets
regions and time periods.
different
Clearly, the finding that
rule out this possibility,
(median) country
stock.
in
each group
most countries have a small but negative net
and over
time.
Since net foreign asset positions are small relative
is
in
reports wealth-weighted averages,
for the typical
foreign asset position holds across regions
there
Table 2
wealth, aggregating across countries
The top panel
and the lower panel reports the share
and period.
in
To
in
to wealth,
it
follows that
a strong relationship between countries' financial wealth and their capital
To see
this,
Figure 2 plots the capital stock,
k,
against financial wealth,
averaging the available years over the period 1966-1997 for each country.
a,
A simple
regression of the capital stock on domestic wealth delivers and R^ of 98 percent and
a slope coefficient of 0.93. Clearly, the world distribution of capital stocks
to the world distribution of wealth. This
between these two
appears
coefficient
one
is
and
countries
in
to
be very close
random
is
is
zero. That
is,
that differences
to one, the null hypothesis that this coefficient
on average the
We also reject the
capital stock
null
is
hypothesis
exceeds wealth
in
poor
less than wealth in rich countries. This leads us to the next fact:
The share of net foreign assets
2:
very close
or uninteresting. Although the slope
rejected at conventional significance levels.
that the intercept
Fact
distributions are
does not mean however
is
in the
country portfolio increases with wealth
a cross-section of countries.
The
wealth
apparent
is
wealth,
strong positive association between the share of net foreign assets and
—
,
in
Figure
3,
which plots the average share of net foreign assets
against the logarithm of wealth,
ln(a), for
each country over 1966-1997.
a
Virtually
all
(44 out of 47) developing countries are debtors, as are two-thirds of
10
in
our sample are mostly
industrial countries (14 out of 21). Tiie ten creditor countries in
rich industrial countries
(Belgium/Luxembourg, Switzerland, Germany, France, United
Kingdom, Japan, and the Netherlands), and three developing countries (Saudi
Arabia, Singapore, and Lesotho).^
The
subperiods.
(shown
in
between wealth and net foreign asset positions holds across
relationship
The
the
row of Table 3 confirms that the simple results
first
first
in
Figure 3
column) hold across the different subperiods (shown
the
in
remaining columns). The estimated coefficients range from 0.04 to 0.08 and are
from zero
significantly different
in
each subperiod. The magnitudes
when wealth
coefficients indicate that
doubles, the share of foreign assets
increases by three to six percentage points.
regional
of these
dummies. With the exception
In
the next row of Table 3
we
row of Table 3 we check whether the changes
an
artifact of
in
wealth
introduce
on the
of the first subperiod, the coefficient
logarithm of wealth increases slightly, and remains significantly positive.
is
in
the third
In
the wealth elasticity across periods
the changing composition of the sample by restricting attention to a
balanced panel of 8-year average observations. The results do not change
substantially.
There are reasons
foreign asset position
easy
is
to think that the relationship
between wealth and the net
even stronger than what a simple regression would
to think of factors that are positively correlated with wealth
and are
negatively correlated with the net foreign asset position of a country.
strongly correlated with
human
capital,
technology and
likely to
First,
institutional quality,
which raise the returns to capital and make foreign assets less attractive
countries. Second, the variability of returns
making foreign assets less
effects
whereby
sample.
Lesotho
is
in
some
of the larger, richer
arguments can be summarized by saying
All
of these
is
somewhat
of
11
be
wealth
all
is
is
of
for rich
may
also be scale
economies
that
it
in
is likely
an anomaly, as it runs large current account surpluses
South Africa.
primarily workers' remittances from
It
also negatively correlated with wealth,
attractive for rich countries. Third, there
returns are higher
find.
our
that rich
reflecting
countries have better aggregate production functions and therefore find foreign
assets less
To
attractive.
the extent that this
is
the case, the simple regressions of the
net foreign asset position on wealth contain omitted variables that bias
downwards
We therefore introduce a number of additional control variables
We proxy for human capital with the number of years of
the slope coefficient.
into the regression.
secondary education
the workforce, and control for scale effects using the
in
logarithm of population.
We include openness to international trade measured
as
M2
to
trade volumes as a share of
GDP,
financial depth
GDP, government consumption as a share
of
measured as the
GDP, and an
ratio of
index of
civil liberties,
as
proxies for the quality of the institutional environment.
We measure the variability of
returns using the standard deviation of real per capita
GDP
period. Finally,
we
dummies
include a set of regional
growth over the indicated
to control for other
unobserved
region-specific heterogeneity.
The remainder
of
Table 3 summarizes the results of
regression. Averaging over
remains very
significant.
make domestic
capital
all
this
augmented
years, the coefficient on wealth rises to 0.07 and
Consistent with the view that high levels of
more
we
attractive,
find that
level.
surprisingly, population size enters positively, suggesting that there
diseconomies of scale or congestion effects that make domestic
large countries. Public consumption as a share of
approaches significance
demand
for foreign
at the
10 percent
financial
which
may
coefficients
may be
capital less attractive
enters negatively and
may
reflect
an increased
reflect the fact that countries with well-
markets have less need
for
recourse to international financial
markets. The remaining control variables, openness,
growth do not enter
GDP
Somewhat
loans to finance public consumption. Finally, financial depth
enters positively and significantly. This
developed
level,
capital
years of secondary education
enters negatively although not quite significantly at the 10 percent
in
human
significantly.
on these variables
qualitatively similar to
civil liberties,
and the
volatility of
Although the magnitude and significance of the
differs
those obtained
somewhat across subperiods,
in
the
first
12
column.
the results are
From Table 3 we can also
obtain a sense of the magnitude and relative
importance of wealth relative to other factors as a determinant of foreign assets
Consider the regression
positions.
in
the
A one
data over the period 1966-1997.
first
column of Table 3 based on average
standard deviation increase
in
the logarithm of
wealth (which corresponds to roughly a three-fold increase) leads to roughly twothirds of a standard deviation increase in the net foreign asset position, or
percentage
points. In contrast,
a one standard deviation increase
significant control variables leads to
an increase
(in
the the net foreign asset position. Another
in
importance of wealth
explaining the cross-country variation
perform a variance decomposition of the
fitted
can be attributed to wealth.
that almost
60 percent
^
in
all
but the
first
in the
(in
pooling
figures
it
vertical axis, against
its
is
is
due
in
is
(in
to
we
find
(in
we
is
due
similarly find that
to variation in wealth.
wealth
value lagged one year
the second panel) and 10 years
all
foreign assets
predicted foreign assets
country portfolio
Figure 4 plots the share of net foreign assets
years
in
subperiod
predicted foreign assets
The share of net foreign assets
year on the
see the
of the cross-country variation in predicted foreign assets
the majority of the variation
3:
to
Averaging over the entire sample period,
to cross-country variation in wealth. In
Fact
in
way
values from these regressions. The
bottom row of Table 3 reports the share of the variance
that
the remaining
in
absolute value) of only one-third
of a standard deviation
in
about 10
is
persistent over time.
in
a given country and
the
first
panel), five
the third panel) on the horizontal axis,
country-year observations over the period 1966-1997. From these three
clear that the share of foreign assets
in
wealth
is
very persistent.^" The
We use a decomposition of the variance of the sum of two correlated random variables
suggested by Klenow and Rodriguez-Clare [1997]. In particular, if Z=X+Y, they define the
share of the variance of Z attributable to X as COV(X,Z)A/AR(Z). This number has the
following natural interpretation: If we observe that Z is one percent above its mean but we did
not observe X and Y separately, then we would infer that X is COV(X,Z)A/AR(Z) percent
above its mean value. We apply this defining Z as fitted foreign assets, and X as the estimated
coefficient on wealth multiplied by wealth.
^°
This fact was also noted by Kraay and Ventura [Forthcoming] in a smaller sample of 13
OECD economies, who argue that this observation can explain the long-standing puzzle
raised by Feldstein and Horioka [1980].
®
13
simple correlation between the share of foreign assets
year and the
same share
simple correlation
foreign assets
in
is
lagged one year
country portfolios
holds roughly -10 percent of
is
around
2.
country portfolios
in
a given
Even over a 10-year horizon the
a respectable 0.55. This very strong persistence of the share of
is all
its
wealth
more
the
small foreign assets are relative to wealth.
GDP
0.96.
is
in
surprising
our sample the typical (median) country
In
foreign assets,
in
when one considers how
and
ratio of
its
wealth to
GDP
This implies that a current account surplus of 5 percent of
sustained for only four years would be sufficient to entirely erase a country's net
foreign asset position. Yet
we see
the data
in
that net foreign assets as a share of
wealth on average barely change over this horizon, as indicated by the estimated
slope coefficients which are very close to one.
The pooled data
over time
in
Figure 4 nevertheless
in
the persistence of foreign assets
in
year indicated on the horizontal axis,
we
its
the vertical axis as a
one-year
dashed
lag,
line.
average of these estimated slopes.
the foreign assets of
to the
all
country portfolios.
all
Figure 5
we
solid line
When
in
wealth
plot the resulting slope coefficients
this
on
shows a three-year centered moving
slope
greater than (less than) one,
is
countries on average expand (contract).
mid-1980s, the slopes are
In
regress the share of foreign assets
and then
The
interesting variation
the top panel of Figure 4 by year. For each
disaggregate the annual persistence
on a constant and
in
mask some
From the mid-1970s
greater than one. This reflects primarily the rapid
buildup of debt of developing countries financed by borrowing from rich countries and
oil
producers, followed by an even greater increase
of these countries
debt
crisis
was
eventually resolved
fall
in
share of net foreign assets
immediate
is
in
recorded debts as
crisis of
the 1980s.
the late 1980s, the portfolios of
below one. Somewhat
near one during the 1990s, providing
Why
their
suspended payment during the debt
contracted and the slopes
is
in
little
all
many
As the
countries
surprisingly, the estimated slope
evidence of systematic increases
in
country portfolios during this period.
the net foreign asset position as a share of wealth so persistent?
possibility is that the persistence of net foreign assets reflects the
persistence of wealth, which
the
we have seen
to
14
be an important determinant of the
One
cross-country variation
empirically by regressing the
change
logarithm of wealth and the changes
3,
We investigate this
net foreign assets.
in
net foreign assets on the
in
in
change
in
the
the other control variables considered
in
Table
using the three eight-year changes implied by the four 8-year averaged periods
using
Table
in
3.
We then
perform the
determine what share of the variation
same
in
variance decomposition as before to
changes
We find that between 40
changes
in
wealth.
changes
in
net foreign assets can be attributed to
to
iiypothesis
us that persistence
in
wealth
is
in
net foreign assets are
due
to
percent and 70 percent of the variation
changes
in
in
wealth. This suggests
an important source of persistence
in
foreign
assets.
Fact
4:
Gross foreign asset positions are small and consist mostly of foreign loans
rather than foreign equity.
Net foreign assets consist of net claims on foreign equity (e*-e) and net
lending abroad,
(l-l*).
We
illustrate
the cross-country variation
in
the relative importance of the latter
country portfolios
in
explaining
two ways. Table 4 shows the
in
composition of foreign assets as they vary across regions, income groups, and time.
Averaging over
all
years, claims on foreign equity consist of only 0.3 percent of the
wealth of developing countries, while foreign claims on domestic equity account for
2.8 percent of wealth.
Among
industrial countries
claims on domestic equity are only
and 3.9 percent
somewhat
claims on foreign equity and foreign
larger
and consist of only 3.3 percent
of wealth." In contrast gross borrowing
and lending account
percent and 4.5 percent of wealth of developing countries, and
each
for industrial countries. Finally,
borrowing and lending
in
it
is
1 1
for 8.8
percent of wealth
interesting to note that the share of gross
the wealth of industrial countries
is
much
larger than that of
developing countries, especially during the 1990s.
"
A
documented that the holdings of foreign equity of investors in a few
are very small (French and Poterba [1991], Tesar and Werner [1992] and
others). Lewis [1999] provides an excellent survey of alternative explanations for this empirical
large literature has
rich countries
regularity.
Our data confirms
that
it
applies to a
much broader
15
set of countries
and assets.
In
contrast with Table 4 which focuses on gross positions, Figure 6 provides
evidence on the importance of net lending
net foreign assets.
We plot net
the share of foreign assets
in
in
explaining the cross-country variation
lending as a share of wealth on the vertical axis, and
wealth on the horizontal axis, averaging over the period
1966-97. The top panel shows the results for
shows the same
relationship
is
in
all
countries, while the bottom panel
information for industrial countries only. Since the slope of this
more than the covariance between net
nothing
foreign assets
and net
lending divided by the variance of net foreign assets, this slope can be interpreted as
the fraction of the variance
sample
In
our
in
our sample
full
in
net foreign assets attributable to variation
of countries this fraction
this is not surprising at
all,
is
in
net lending.
82 percent. For the developing countries
since
we have seen
that gross equity
same number
positions are small.
What
industrial countries
where cross-border claims on equity are much more prevalent
than
in
is
more
surprising
developing countries. From
country variation
rather than
in
in
up,
portfolios is small
and
evidence
we conclude
that
most
that the share of net foreign assets
typically negative (Fact 1), exhibits
association with wealth
in
in
of the cross-
net lending
and consists
in
country
a strong positive
a cross-section of countries (Fact
persistent over time (Fact 3),
applies to
portfolio equity investment.
we have seen
and
foreign equity (Fact
that the
net foreign assets can be attributed to differences
foreign direct
To sum
this
is
2), is
remarkably
primarily of foreign loans rather than
4).
16
II.
Towards an Explanation of Country
We
next develop a model that emphasizes the interplay of diminishing returns,
production risk and sovereign risk
in
a world populated by
variance investors. ^^ Diminishing returns and production
these were the only forces
at
work,
and the world
portfolios,
we would observe
homogeneous mean-
risk
adjusted rate of return to capital declines as more capital
same
Portfolios
all
is
imply that the risk-
invested
in
a country.
If
countries choosing the
distribution of capital stocks
would be determined by
the equalization of risk-adjusted rates of return. Sovereign risk however generates a
home
bias
in
the
demand
would hold only domestic
mimic the world
for capital.
capital
A Model
is
were the only force
this
and the world
distribution of wealth.
distribution of capital stocks
11.1
If
The
at
work, countries
distribution of capital stocks
set of country portfolios
shaped by the
and the world
interaction of these forces.
of International Capital Flows
The world contains two
countries, North
and South; one factor
capital;
and a single good
good
used as the numeraire. Each country contains a continuum
is
would
that
can be used
for
of production,
consumption and investment. This
of identical
consumer/investors that evaluate consumption sequences as follows:
(1)
where c
EJlnc(t)e"^'
is
dt
(5>0)
consumption. The time index
confusing. Throughout,
North has higher
initial
we use an
will
be omitted whenever
this is not
asterisk to denote South variables.
wealth than South,
i.e.
We assume
a(0)>a*(0).
^^
The key property of these investors is that the share of each asset in their portfolio does not
depend on their wealth, but only on the menu of available assets. By homogeneous, we mean
that all Investors have identical (homothetic) preferences, although possibly different wealth
and menu
of available assets.
17
The production technology
North and South.
To produce one
required. Since capital
is
is
is
quite simple. Let k
processes with independent increments. That
and E[do>do/]=0. Then the flow
7t=9k"^
in
and
(0<y<1; 9>0)
measures the strength
is,
unit is
and
co*
in
its
return
be two standard Wiener
E[do)]= E[dco*]=0, E[dco^]=E[dcL)*^]=dt
is
given by
South; where n and n* are short-hand for
and a
is
a positive constant. The parameter y
effect. ^^
The parameter a measures the importance
country-specific production risk. Therefore, this formulation
same technology and embodies
adjusted rate of return to capital decreasing
for capital to
always one and
is
of diminishing returns which, for simplicity, are treated here as
an externality or congestion
countries have the
capital stocks of
consumption good
of production net of depreciation
North and 7T;*dt+adco*
7i:*=0k*"''
co
be the
k*
unit of the
each
reversible, the price of
the flow of production net of depreciation. Let
Ttdt+adca
one
unit of capital,
and
move from
rich to
in
assumes
of
that both
the two forces that
make
the risk-
the capital stock and create incentives
poor countries.
Domestic investors own the domestic stock of
and equity contracts with foreign
capital
and can enter
investors. International loans promise to
into loan
pay an
instantaneous interest rate rdt. At the beginning of the period, the lender gives the
principal to the borrower. At the
The borrower might honor
plus interest.
its
end
of the period there are
promise,
The borrower might
in
two possible outcomes.
which case the lender receives the principal
also default on
its
promise,
in
which case the
borrower keeps the principal and interest and the lender receives nothing.
North (South) equity has price v
generated by one
unit of
(v*)
and promises
North (South)
capital. At the
buyer gives the value of the equity to the
unit of capital
at the
end
promise,
and allows him/her
of the period there are
in
to
to
seller.
The
A
share of
pay the net flow of production
beginning of the period the
seller provides the
operate the production technology.
two possible outcomes. The
seller
buyer with a
Once
again,
might honor
its
which case the buyer receives the value of the equity and keeps the net
flow of production.
The
seller
might also default on
its
promise,
in
which case the
^^
At the cost of further notation, we could generate this dependence by assuming there is a
production factor that is not priced. Since this is well known, we dispense with the formalities.
18
seller
keeps the value
unit of capital
amount
of the equity
and the net flow
and then pays a cost X (0<X<1)
The buyer receives
of production.
South and e
of lending from North to
(e*)
to
repossess the
nothing. Let
I
be the
be the number of North (South)
shares owned by South (North).
It
is
equilibrium
evident that international loan and equity contracts
if
and only
if
the probability that they are honored
will
be used
in
high enough. This
is
observation raises a familiar time-inconsistency problem. Since governments cannot
punish foreign citizens, international trade
willingness to punish their
foreigners. 'Ex-ante' both
own
citizens
if
in
assets relies on governments'
they default on their obligations towards
governments would
commit
like to
to
do
this
and allow
investors to exploit beneficial trade opportunities. However, 'ex-post' governments
not have an incentive to punish default
of default are clear. But,
if
the benefits exceed the costs.
what are the costs
of default?
We shall
The
do
benefits
assume they vary
overtime. Let s={0,1} be the state of the wortd. During 'normal times' (s=0) both
countries can credibly commit to retaliate with penalties that are large
ensure that default never occurs. During
credibly
commit
(North) defaults
Ae*
to penalties
if
its
beyond
'crisis
(v* e*+l-ve<-A,e).^'* Let
a
and
dt
|3dt
to
periods' (s=1) countries cannot
retaliation in kind.
net foreign asset position
enough
is
be the
As a
result,
negative enough,
if
s=1 South
i.e.
ve-l-v*e*<-
probabilities that the wortd transitions
from 8=0 to s=1 and vice versa; and assume these transitions are independent of
production shocks, E[dcods]=E[dca*ds]=0.
The value
of
ds
is
revealed after the
beginning-of-penod payments of loan and equity contracts have already been made.
What
is
the equilibrium probability of default?
probability of default
is
zero.
If
investors believe the
a*>a(1-A), the country portfolios chosen by investors
are consistent with this belief.^^
^'*
Assume
In this
The seminal papers on sovereign
risk
case, the equilibnum probability of default
and the
ability of
is
various types of penalties to sustain
trade are Eaton and Gersovitz [1981] and Bulow and Rogoff [1989]. Eaton and Fernandez
[1995] and Obstfeld and Rogoff [1996, chapter 6] are two excellent surveys of this topic.
^ If
investors believe that the probability of default
v=v*=1, k=k*=(a+a*)/2, e=a*/2, e*=a/2 and
if
s=1
,
is
zero, the equilibrium of the
no country defaults
a*>a (1-X).
1=0. Naturally,
North does not default and neither does South
19
if
if
model implies
s=0. But even
zero and sovereign
risk is
on that a*(0)<a(0)-(1-X).
dependent.
Assume
the
simply not an issue.
case, the equilibrium probability of default
In this
state
initial
We shall therefore assume from
is
is
now
state-
s=1 and investors believe that default occurs
if
the state does not change. Then, investors do not purchase foreign loans and equity
since they expect default to occur with probability close to one,
both countries are indifferent on whether to default or not.
equilibrium exists
we assume
i.e.
1-|3dt.
To ensure
believe that default occurs
If
is
restrict
in
To sum
up, the world
in
s=0 and investors
economy
The parameters a and
it
will
A.
in
the cycle starts again. Although
is
probability,
small. This implies that
exhibits periods of trade
s=0
to s=1) in
measure the
occurs, a period ensues
Eventually, international trade
i.e.
strategies followed by investors.
is
We shall
the
in
in
assets that
which the debtor country
probability
which there
assets resumes
in
adt
is
and the destructiveness
no trade
(s transitions
in
from s=1 to s=0) and
normal times there might be substantial trade
We describe these strategies
We
in
on the
next.
can construct a mixed-strategy Nash equilibrium as follows: Let p be the probability of
choose country portfolios such that ve-l-v*e*=-Ae*. South is
indifferent between defaulting or not. There is an equilibrium in which South defaults with
default that leads investors to
probability (p/a)dt.
20
of
assets.
assets, the (small) probability that a crisis occurs has an important effect
^^
purchase
'normal times'.
crises (s transitions from
this crisis. After
In
the chosen country portfolios are consistent
which a
in
equilibrium probability of default
defaults.
is
which the country follows a pure strategy.^®
in
the analysis to the case
in
state
no trade
the unlikely event the state of the world changes. Otherwise,
no Nash equilibrium
culminate
initial
is
the state changes. In this case, investors
this probability is not too large,
with South defaulting
there
if
next that the
and equity since default occurs only with a very small
foreign loans
adt.
Assume
'crisis periods'.
an
that
they default (on their non-existent foreign obligations)
so that the beliefs of investors are consistent. This implies that there
assets during
But then
11.2
Investment Strategies
In crisis periods,
and save. This
the only decision that investors face
limited choice
embedded
is
in this
is
how much
to
consume
budget constraint which applies
only during crisis periods:
(2)
da = (7ia-c)dt +
In
aadco
normal times there
trade
is
in
constraint of the representative investor
(3)
in
and we can
write the
North as follows:
restriction applies:
k-ve + v*e*+l
a =
The
first
two terms of the budget constraint
expected return and
volatility of
state of the world not changing.
(3)
are standard and
schemes and impose
last restriction is
The
third
consumer
in
short-sale constraints
on the holdings
North maximizes
(1)
rule out Ponzi
of foreign equity. This
risk.
subject to (2)-(4) and the dynamics of asset phces
i.e.
Since the representative consumer
the laws of motion of
is
infinitesimal,
r,
v
,
v*, n, n*,
a and
the South solves a similar problem. Appendix
The representative
B shows
that the first-order
conditions associated with the investor's problem can be written as follows:
21
a*.
he/she understands that his/her
actions have no influence on these prices and their evolution.
in
we
the optimal consumption and portfolio rules, the representative
their return characteristics,
consumer
the
term describes the wealth shock that the
a logical implication of sovereign
To determine
show how
wealth depends on asset choice, conditional on the
investor experiences at the onset of a crisis period. Throughout,
and
budget
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, of course, the following
(4)
assets,
6a
(5)
c =
(6)
7T:-p
=a
k-e
2
a
p =
a
a-_
(7)
r
(8)
p-v-n = -a
k-^e
^
k-e
2
,,
a
,
+ a(A-v)
k-Xe
a
(9)
7i*-r-v*<0, e*=0.
If
where p
is
Otherwise,
re
e*
-p v* = a^
*
•
+a v
k-Xe
the multiplier associated with constraint (3) divided by the marginal
utility
of
wealth. This quantity can be interpreted as the risk-free rate that applies on loans
between North
residents.
Equation
(5) is
familiar result that
of time preference
A similar set
of first-order conditions apply to South.
the first-order condition associated with
c;
and shows the
consumption equals the annualized value of wealth. Since the rate
is
used as the discount
factor, the
consumption
rule
is
independent
of the state of the world.
Equations
describe
(6)
how
(6)
and
(7)
are the
first
investors value production
says that the premium
order conditions associated with k and
and sovereign
risk,
I;
and
respectively. Equation
for holding production risk, n-p, is the
covariance between
the return to one unit of capital and one unit of the investor's portfolio, a^(k-e)/a.
Equation
(7)
says that the premium
proportional to the covariance
portfolio, i.e.
a
.
for holding sovereign risk, r-a-p,
between one
But since
unit of
is
also
loans and one unit of the investor's
this time the effect of the
shock
is 'large'
or
a
non-local, the factor of proportionality
wealth before and after a
crisis
is
occurs,
not
i.e.
one but the
.
k-Ae
22
ratio of
the marginal value of
Equations
(8)
and
and describe the supply
(9)
of North equity
to sell equity.
is
at
a discount,
i.e.
in this
the event of a
v<1. Using Equations (6) and
equal to the gain obtained
income
1-A. in
in
crisis.
(7),
e*;
South equity,
for
at
which North
Each share sold by the North reduces income by one
output, but also provides a gain of
be sold
and the demand
can be interpreted as determining the price
respectively. Equation (8)
is willing
are the first-order conditions associated with e and
This
we
is
why
unit of
equity
will
find that this discount
the event of a crisis times the price of one unit of
state of the world,
-
1-v =(1-^)
i.e.
Equation
.
(9)
defines the
r
demand
for
South equity. This asset contains both production
and the required premium
risk
and sovereign
risk
reflects just this.
This completes the description of the model. Equations (4)-(9) and their
counterparts for South jointly determine asset prices
distribution of capital stocks (k,k*)
trade
(e,e*,i)
(n,7i*, a,a*).
(p,p*,r,v,v*),
and consumptions
(c,c*);
as a function of the distribution of wealth
We
use
this set of
(a, a*)
the world
and the pattern
of asset
and technology
equations to derive the cross-sectional implications of
the theory. Then, the budget constraints (2)-(3) determine the law of motion of the
world
economy as
a function of the
the realizations of the shocks.
initial
distribution of wealth
We use this additional
and technology and
set of equations to derive the
time-series implications of the theory.
11.3
Three Examples
Before
we embark
examples or special cases
that the theory
in
in
a quantitative analysis of the model,
that help build intuition
emphasizes.
In
these examples,
normal times and then ask: What
does
it
is
the
initial
evolve over time?
23
on the
we
discuss three
role of the different forces
we assume
the world
pattern of trade
in
economy
assets and
starts
how
EXAMPLE
no discount on
#1
Let
:
a^O
so that there
international assets,
i.e.
choose the same
As
portfolios.
risk. In this
case, there
effectively, the
a result, there
is
same menu
of assets, they
no borrowing or lending. Since
technologies are identical, both countries invest 50 percent of their wealth
capital
and the
rest in foreign equity.
each country and the world
is
p=p*=r and v=v*=1. Since countries have
and homothetic preferences and,
identical
no sovereign
is
Thus, half of the world capital stock
in
domestic
located
is
in
interest rate equals the marginal product of capital:
^^
(10)
k = k*
(11)
r
=
-^
^a+a*v^
=e
^
\
If
one
countries
in
+
2
J
interprets North
Table
1
,
we have
and South as the
= 0.8
that
set of industrial
;
and developing
and net foreign asset positions (as a
a + a*
percentage of wealth) of North and South are 37.5 and -150 percent, respectively.
The
distribution of wealth
portfolios
^
=
An
a*
2
implication of this
The
is
predictions of this
exhibits
rate:
^
^
that net foreign asset positions are time invariant.
example are both well-known and dead wrong. The
two features that are present
in
the data: net foreign asset positions
are positively associated with wealth (Fact 2) and very persistent (Fact
model also predicts net foreign asset positions
data (Fact
same
da
^^
= (r-p).dt + -(dco + dco*)
a
model
constant over time, since both countries choose the
and therefore have the same growth
da
(12)
is
1).
It
that are
much
shows
that countries hold
little
24
in
the
and no borrowing and
foreign equity
of their net foreign asset positions with foreign loans (Fact 4).
But the
larger than those
also predicts very large foreign equity positions
lending, while the data
3).
and finance most
EXAMPLE
poor countries
the
same
same
return
assets
asset.
is
in
:
a^O
so that the only reason
to exploit higher rates of return.
is
return
#2 Let
in
normal times.
If
why
capital flows
Now foreign
equity
from
rich to
and loans
deliver
X=0, foreign equity and loans also deliver the
the event of default and, consequently, the composition of foreign
indeterminate.
If
X
is strictly
positive, foreign equity
becomes a dominated
case, countries do not hold foreign equity and finance their net foreign
In this
asset positions with foreign loans. The world distribution of capital stocks and the
world interest rate are implicitly determined by:
(13)
r
= ek-^+a-
(14)
r
= ek*"^+a-
(15)
k
k
k^
\3'
+ k* = a + a*
Equations (13) and (14) describe the
Equation (15)
is
demand
the market-clearing condition.
home
demand
risk
creates a
rate
exceeds the marginal product
risk
premium required
bias
to
in
the
for North
and South
These equations show
for capital. Also,
capital,
that sovereign
note that the world interest
of capital in both countries. This difference
compensate lenders
for
and
sovereign
risk.
is
The top panel
the
of
Table 5 computes the foreign asset position of North under alternative assumptions
for Y
and a (Remember the net foreign asset position
North). Net foreign asset positions for North
37.5 and -150 percent (as a^O).
instance, a=0.02
is
What
consistent with the
episodes of large-scale defaults
in
value of aggregate production
9
is
if
South
is
-A times
are reasonable values for
20'*^
that of
and South range from zero (as a^'>=)
to
a and y? For
century experience which features two
the 1930s and the 1980s. Since the expected
k^'^,
values of y around 0.6 correspond to the
neoclassical growth model, while values of y near
model. Table 5 shows that
of
one are close
to the linear
growth
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 half of
those
we found
in
the previous example.
If
y=0.1, the net foreign asset positions are
25
These nunnbers are now seven times smaller than
5.4 and -21 .6 percent.
previous example and start to resemble those
A
nice feature of the
first
and the world
be the case also
^
(16)
=
3
distribution of wealth
in this
^
same
was
second example,
portfolios,
of constant net foreign
they had the
albeit for different reasons:
= (r-a-p).dt
3
North capital and lends the rest to South. South invests more than
and borrows the difference from North. Since there
obtains a higher return on
rate
same growth
constant. Interestingly, this turns out to
Countries choose quite different portfolios. North invests part of
capital
the
find in the data.
example was the prediction
asset positions. Since countries chose the
rate
we
in
its
its
less
wealth
wealth
South
in
in
South
capital,
South
domestic capital than North does. But since the interest
on foreign loans exceeds the marginal product
an even higher return on
is
its
its
of capital of South, North obtains
foreign loans during normal times.
and both countries share the same growth
rate.
An
implication
These
is
effects balance
that net foreign
asset positions are constant over time.
This example shares with the
first
one the
positions are positively associated with wealth
they have
in
common. Although
large relative to those
in
substantially. Moreover,
arbitrarily small)
we
and very
persistent. But this
sovereign
example
risk
is
a
To counteract
still
a
bit
are willing to
assume
that X
is
positive (although
the model also predicts that foreign equity holdings are zero and net
is
It
seems
therefore that this
a major improvement over the previous one, and a small dose of
moves us
bit
that
gap between theory and data has narrowed
foreign asset positions are financed through foreign loans.
second example
is all
the predicted net foreign asset positions are
the data, the
if
predictions that net foreign asset
a long
way towards
misleading since
this impression,
we
it
reconciling theory
and data. But
suggests that X plays a small
explain what
theory.
26
this
role in the analysis.
happens when we leave X out
of the
EXAMPLE
#3 Let X-^0 so
equity. In this case,
production
:
no costs of transferring ownership of
that there are
follows from Equations (6)-(9) that countries share
it
all
the
risk:
e
k-e
k
e*
k*-e*
k*
a*
a
a+a*
a
a*
a+a*
(17)
This equations state that countries receive a fraction of each of the outputs
that
is
equal to their share of world wealth. The
the exposure to sovereign risk
against this type of
risk
by holding small net foreign asset positions. This does not
preclude them from hedging against production
equity positions that are roughly balanced.
no borrowing and lending
of the
first
hedge
the net foreign asset position, countries
is
Since
intuition for this result is simple:
example? Not
if
Does
A,=0.
quite.
A
this
risk
In fact,
mean
it
by holding large gross foreign
is
that
possible to
we
show
that there
is
are back to the predictions
major departure from the
first
example
is
that
countries hold smaller net foreign asset positions. The distribution of capital stocks
implicitly
determined by:
(18)
=
r
k
ek"^+a-
a + a^
vay
kM
/i.*^-''
(19)
r
=
ek*^^+a
-a
k
2
va
k
(20)
again. Equations (18) and (19) describe the
capital,
interest rate
capital also
and Equation
still
command
commands
panel of Table 5
this
(20)
is
demand
the market-clearing condition.
need not exceed the expected marginal product
foreign loans
choose
a+a
+ k* = a + a*
Once
South
is
show
for
Now the
a premium to compensate for production
the predicted net foreign asset positions
it
is
world
of capital. Although
a premium to compensate for sovereign
number because
North and
risk,
risk.
domestic
The middle
when o
is
0.05.
We
the average standard deviation of the growth rate
27
in
a sample of 88 countries from 1960 to 1994. Since production
creates an
risk
additional incentive for capital to flow from North to South, the net foreign asset
positions are larger here than
A second
is
now
the second example, but only slightly so.
in
difference with the
consistent with a
example
first
measured home bias
creates a discount on foreign equity, that
is,
in
is
that
full
sharing of production
country portfolios. Sovereign
v<1 and v*<1. This
means
that
risk
risk
full
sharing of production risk does not require countries to invest a fraction of their wealth
in
foreign equity that
stock. For instance,
is
proportional to the share of foreign capital
assume
discounts were zero, v=v*=1
25 percent
that
,
sharing of production
full
invest 25 percent of their wealth
in
South
discounts are 50 percent, v=v*=0.50,
invests 52.5 percent of
percent. This
is
its
enough
of the world capital
wealth
in
for North to
full
capital
is
in
the world capital
located
risk implies that
and the
in
South.
If
both countries
rest in North capital.
If
sharing of production risk implies that South
South
capital, while
buy the
rights to
North invests only 12.5
75 percent of South's output.
If
the discounts are large enough, the model might be consistent with the observation
that foreign equity holdings are small. Table 6
The discount on South
equity (not
not large
shown
enough
in
shows these discounts when a=0.05.
equity ranges from 20 to 30 percent.
Table
6) is
a
little bit
larger.
The discount on North
Nevertheless, these discounts are
to reconcile theory with data.
This example shares with the other two the result that both countries have the
same growth
(21)
^
/^
=
a
.^^
9
k + k*
a
This
production
rates:
result,
risk,
p
dt
+a
V
k
.
dco +
k+k*
which follows from the fact that there
implies that net foreign asset positions
28
is
do
\
Ir*
k
L-
dco*
k+k
complete sharing of
not
change over
time.
As
the previous two examples
in
we have
the prediction that net foreign asset
and very
positions are positively associated with wealth
persistent.
As
the second
in
example, predicted net foreign asset positions are small enough to resemble those
we observe
in its
in
the data. But this
example departs dramatically from the previous one
predictions regarding the composition of net foreign assets.
example,
we have now
large foreign equity positions
determined
is
in
losses.
If
risk.
a-X
is
But
in
the event of default, equity
large relative to o^, investors hold
finance net foreign asset positions. This
where o^O.
If
a-X
is
11.4
is
is
it
is
first
how
lending.
the composition
is
a better
allows countries to share
a worse asset since
equity
little
what happens
in
it
generates
and instead use loans
to
the second example
small relative to (/, investors hold large equity positions and do
not use foreign loans. This
This
the
the theory. During normal times, equity
asset to transfer capital from North to South since
production
in
and no borrowing and
This difference between the second and third example illustrates
of foreign assets
As
as far as
is
what happens
the third
in
we go with examples.
example where X^O.
We next turn to the full-fledged
model.
Quantitative Implications
Can
the model developed here replicate the main features of the data with a
reasonable set of parameter values? The key issue
'reasonable' set of parameters values.
Y=0.2, a=0.05, a=0.02
somewhere
because
in
and
A,=0.20.
A
choice of y=0.2
average standard deviation
countries from 1960 to 1994.^''
to
agree on what constitutes a
We propose the following
between the neoclassical and the
this is the
is
means
linear
of the
that
benchmark
we
settle
growth models.
growth rate
We also choose a=0.02
in
because the
We set o=0.05
a sample of 88
20'^ century
experienced two episodes of large-scale defaults by developing countries
1930s and the 1980s. Admittedly, we have
little
29
intuition
values:
in
the
as to what a reasonable
has
value for X
good
We therefore chose A,=0.20, which
is.
results.
parameters,
we assume
Since there
we
also
is
is
a value that delivers relatively
substantial uncertainty regarding the values of
show how the
results vary with
that the share of world wealth in North
each
is
0.8,
which
capital
is
located
in
in
is
consistent with the
Table
1.
predicts net foreign asset positions for the North
10.5 percent and -42 percent, respectively. This
North and 28.5 percent
in
means
the
of the them. Throughout,
interpretation of North as the set of industrialized countries
Our base case
all
and South
of
that 71.5 percent of the world
South. Therefore, only 8.5 percent of
The North
the world capital stock flows from North to South.
portfolio contains
89.5
percent of North capital, and 6.25 and 4.25 percent of foreign loans and equity,
respectively. North
does not
percent of South capital. This
and 17 percent
much
trade
in
is
The South
equity to South.
sell
142
financed by selling foreign loans and equity worth 25
of South's wealth, respectively. Overall, this
assets relative to the data (See Figure
numbers show
portfolio contains
that substantial progress
dose
of sovereign risk in the theory. This
case
to the
has been
1
and Table
made by
becomes apparent
standard model without sovereign
predicts net foreign asset positions of 37.5
risk
in
4).
But these
including a very
if
we compare
(See Example
and 150 percent
implies that half of the capital would be located
base case predicts too
in
1).
The
modest
this
base
latter
North and South. This
each country and 30 percent
of the
world capital stock would flow from North to South. Moreover, the country portfolios
are predicted to be identical and contain 50 percent of domestic capital and 50
percent of foreign equity.
The
next step, of course,
to the particular choice of
is
to
determine
how
sensitive are these predictions
parameter values. Figure 7 shows how the predictions of
the model vary with the two parameters that describe sovereign
the Figure
we change one
benchmark values. Not
"
of the
risk. In
parameters holding constant the rest
each panel of
at their
surprisingly, increasing the probability of crises (a) and/or the
same sample, the average growth rate was 0.02. We set the rate of time preference
6=0.02 and choose a value for G=0.04 so as to match this average growth rate given a
choice of units such that a+a*=1
In this
at
30
destruction these generate {X) leads to a reduction
in
net foreign asset positions.
Halving the probability of crises from twice to once a century almost doubles the
predicted net foreign asset positions (from 10.5 and
-A4
to 18.75
and -75 percent).
Doubling the probability of crises from twice century to four times a century almost
halves the predicted net foreign asset positions (from 10.5 and -A4 to 5.75 and -23
percent).
It
seems
changes
sensitive to
example
therefore, that predicted net foreign asset positions are quite
to the
in
a. This
second and
is
what we found also when we moved from the
ones. Figure 7 shows that
third
asset positions are not very sensitive to changes
case when we moved from the second
Figure 7 also shows
with
a and X
both
a and X make
to the third
We
this predicted net foreign
also found this to be the
example.
the base case predictions for the South portfolio vary
(the implications for the North portfolio follow immediately). Increases
enough
to close the
enough
to close the
North foreign equity market, but not
South foreign equity market. Moderate increases
in
a and X
are sufficient to close the South foreign equity market. While moderate reductions
a and X would open
the North equity market,
it
would take large reductions
parameters to raise foreign equity holdings substantially.
for the
composition of foreign assets
changes
a and
in
the value of
X.
If
in
equity less attractive relative to loans. In the base case, these
variables are already high
large
how
in X.
first
On
A,.
this
in
the other hand,
It
seems
in
in
these
that the predictions
our base case are not very sensitive to small
we have
also a
lot
of uncertainty regarding
value were close to zero, the predictions regarding the
composition of foreign assets would change substantially, as shown
in
the second
example above.
Figure 8
shows how the
predictions of the
model vary with the parameters
that
describe technology. Simple inspection of this Figure reveals that predicted net
foreign asset positions are quite sensitive to our assumptions about diminishing
returns
(y),
example,
if
but not very sensitive to our assumptions on production risk (a). For
we
(from 10.5 and
raise y from 0.2 to 0.6, the net foreign asset positions basically double
-44
to
20.5 and -82 percent). As y goes to zero, net foreign asset
31
positions collapse to zero as well. This reflects the fact that there are no longer
any
return differentials to provide incentives for capital flows and, for our choice of
parameters, the
risk of default
diversifying production risk.
values),
it
is
necessary
changes
that
in
to
a have
is
sufficiently high to
aX>a^ (which
If
assume
that
y>0
is
outweigh any benefits of
a restriction satisfied by the benchmark
to generate capital flows.
small effects on net foreign asset positions
The observation
was already made
when we moved from
the second to the third example. Not surprisingly, the relative
importance of y and a
for the
Increases
in
production
and induce investors
to
composition of country portfolios
risk raise
is
just the opposite.
the diversification benefits of holding foreign equity
use more of
it
and
less loans.
We see that as a increases first
the North begins to purchase South equity and eventually the South also holds North
equity. In the process, holdings of foreign loans decline.
An
important feature of the data
is
the persistence of net foreign asset
positions. This requires that relative wealth positions
Figure 9 confirms this
is likely
to
do not change much over
be the case, since the expected values
growth rate of wealth are quite similar across countries.
average growth rate
the
same
year).
is
is slightly
each
our base case, South's
true for the standard deviation of this growth rate (6 vs. 4.7 percent per
Over the 30-year horizon covered by our dataset
6%. This
sensitive to
for the
higher than North's (2.7 vs. 2.5 percent per year);and
growth rates would lead to a cumulative increase
only
In
time.
finding of
changes
of the three
To sum
in
an almost constant world
the parameters.
We
in
average
this difference in
the relative wealth of the South of
distribution of wealth
had already encountered
is
not very
this result in
examples discussed above.
up,
if
diminishing returns at the country level are weak, a small doses
of sovereign risk
can move us a long way towards reconciling the theory's predicted
capital flows with
those
we observe
positions are small (Fact
persistent (Fact 3).
If
1),
in
the data. Both indicate that net foreign asset
positively associated with wealth (Fact 2),
and very
diminishing returns at the country level are strong, the theory
predicts net foreign asset positions that are too large relative to those
32
in
the data.
A
natural
In
way
to
reduce them would be to assume that North has a better technology.
any case, the key parameters regulating the size
of crises
and the strength
portfolios,
have
in
we
of diminishing returns. With respect to the composition of
find that the theory
can
replicate the data only
the event of crises and/or production risk
does the theory
of capital flows are the probability
is
if
is
financed with loans (Fact
33
is
a bad asset to
not very important. Only then
predict that foreign equity holdings are small
foreign asset position
equity
3).
and most
of the net
Appendix A: Data Sources
In this
appendix
we
describe
estimates of country portfolios. To do
stock, foreign claims
data and methodology used to construct
tine
on the domestic
we
this,
require data on the domestic capital
capital stock, the stock of
domestic residents'
holdings of capital located abroad, and domestic residents lending to and borrowing
from abroad.
Data on stocks of
for
some years
some
coverage
Second, estimates of stocks
time.
set of countries, years,
stock
is
is
limited to rich countries
of assets are constructed using
on flows
of investment in
and assets. The flow of investment
flows on international assets
in
countries and
is
available
in
methodologies which are
and
countries, assets,
in
much
larger
the domestic capital
and a consistent treatment
(B0PSY4 and B0PSY5).
Payments
In
Statistics
order to expand country
coverage and ensure a consistent methodology across assets and countries, we
up estimates
of stocks of assets
existing estimates of stocks,
(A1)
where
in
S,,
Set is
based
primarily
on
constant 1990
build
some
identity that:
=V,tS,,,_i+F„
US
US
this available flow data,
and the standard accounting
the value of the stock of a given asset
constant 1990
of
the balance of payments statistics
the International Monetary Fund's Balance of
Yearbook, Revisions 4 and 5
to recent years.
these assets exist for a
readily available from national accounts data,
reported
and
documented and may vary considerably across
contrast, data
In
some
a variety of sources. This existing stock data suffers from two
in
deficiencies. First, the
often poorly
of these assets are available for
dollars,
dollars,
F^
and
is
the flow of
Vet is the
the value of the stock of that asset
in
in
country c at the end of period
new investment
in
that asset, also
gross change between periods
period
34
t-1
t-1
and
in
t
in
t
In
its
stock
order to implement this approach for each asset,
some
in
initial
period, data
information on changes
describes
how we do
in
require an estimate of
on the corresponding flow of investment, and
the value of the asset.
each asset
this for
we
The remainder
of this
appendix
of interest.
Domestic Capital
Flows:
PPP
We use data
on gross domestic investment
in
constant 1990 dollars at
Summers and Heston Penn World Table Version
from the
(IxRGDPCHxPOP).^^ This covers
and equipment, and changes
in
non-residential
and
5.6
residential building,
We depart from the
inventories.
machinery
usual practice of
cumulating only gross domestic fixed investment (which excludes inventory
accumulation) since inventories themselves form a (small) component of wealth.
Initial
Stocks:
countries starting
start
GDP
initial
delivers
Since
We then
capital stock for
GDP
capital output ratios
initial
many
We
1965 (KAPWx(1+KRES/100)xRGDPCHxPOP/RGDPW).
per capita.
data on per capita
and
all
use the
fitted
of the countries
and investment are
GDP
per capita
is
in
of the countries in our final
on our estimated stocks
measurement
after
ratio
on the
log-level
values of this regression to estimate
our sample,
available.
in
the
first
year
for
which
Since the correlation between
very large, this procedure by construction
capital stocks that are very similar to the
the 1950s, unavoidable
^®
report estimates of capital stocks for 61
by estimating a cross-country regression of the capital/GDP
of real
the
in
Summers and Heston
Summers-Heston estimates.
sample have data on investment beginning
errors
in initial
stocks
will
in
have minimal impact
1966 when our data on foreign assets begins.
We work with
an expanded version of the Penn World Tables which extends the coverage
GDP, investment, and population by using growth rates of local-currency
constant price GDP and investment, as well as population, as reported in the World Bank
World Tables,
of real per capita
35
Valuation.
be measured
at
we would
In principle
like
the capital stock, and
market value. An obvious choice would be
value of capital by changes
a share price index. This
in
to
changes
is
may be
developing countries where markets are
and proxy changes
Finally,
we need
in this
thin.
by the change
in
We
is
in
the
inappropriate for
in
not representative of the stock of capital as a whole.
share prices and the underlying value of firms
in
other assets, to
proxy changes
several reasons. Capital listed on the stock market, especially
countries,
all
developing
The
link
between
tenuous, especially
in
instead consider replacement cost,
the local currency investment deflator.
account physical depreciation. This of course varies
to take into
across assets, and probably also across countries (consider the adverse effects of
poor infrastructure).
the
In
absence
of better information
we
are forced to rely on the
average value of 6 percent used by Summers and Heston.
summary, we measure the gross change
In
stock during period
t
Pt
where 5=0.06
is
value of the domestic capital
as:
Pf-I
v„=(l-5)-!-i-
(A2)
in
^ct
"^c,!
©ct-i
Pi,,_i
the depreciation rate, Pt
is
the
US
price level, ectis the
in local
currency units per
US
time
A
with the timing of these prices.
t.
small concern
end-of-period stocks
exchange
is
dollar,
we would
like to
rates. Unfortunately gross
mid-year averages
for
and
P^^ is
the investment deflator
Since
some
in
country c at
are measuring
domestic investment deflators are reported as
in
countries with very high inflation,
timing
we
find
is
extremely
large revaluations of the domestic capital stock that probably only reflect errors
timing of
we
exchange
leave them
in
rate
and
but discard
rate
use corresponding end-of-period prices and
most countries, so that some mismatch
unavoidable. Finally, for
we
exchange
price data.
in
the
Rather than censor these from our dataset,
them when appropriate
36
in
empirical work.
Cross-Border Claims on Equity
order to isolate the portion of the domestic capital stock
In
we need
residents,
measure domestic
to subtract foreign claims
residents' claims
on
on
owned by domestic
this asset. Similarly,
capital located abroad.
we need
to
These claims can take
the form of direct investment or portfolio equity investment, although the dividing line
between the two
is
vague. For example,
ownership stake to constitue
this paper,
distinction
we do
is
However, we construct data on the two
other applications.
We rely on flows
Flows:
investment as reported
in
In
between majority and minority ownership as the
irrelevant for our purposes.
in
countries consider a 10 percent
direct investment, but this threshold varies for others.
not distinguish
stocks separately for use
many
the
of inward
B0PSY5
and outward
(lines
direct
and
portfolio equity
4505, 4555, 4610 and 4660). Data on
these items are also available under the Revision 4 presentation of the Balance of
Payments
Statistics
definitions
between the
between the
countries,
B0PSY4
We therefore
almost perfect
extend the coverage of the
Stocks:
The B0PSY5
and equity investment
data from the
first
is
practice, the
in
for
in
correspondence
these items for most
B0PSY5 backwards
using
data wherever possible.
and 8660), and
the
B0PSY4 and B0PSY5,
B0PSY5 and B0PSY4
Initial
of direct
Yearbook (B0PSY4). Although there are minor changes
for
some
for
reports data on the stocks corresponding to flows
most
industrial
economies
(lines
8505, 8555, 8610
countries these stocks can be extended backwards using
B0PSY4. For most
available stock reported
in
where these stocks are
countries
the
available,
BOPSY4 and B0PSY5, and then
we use
use Equation
(A1) and the data on valuation changes discussed below to construct stocks for
years for which flow data are available before and after
stocks of direct investments are provided
in
the
sources, they are quite similar to those reported
OECD
and UNCTAD, which
rely
BOPS
in
this date.
Note that since
as reported by country
other publications such as the
on the same national sources.
37
all
For most developing countries, estimates of the stock of inward direct
investment originating
use
this to
aware
in
measure
OECD
in
initial
economies
or originating
in
available
stocks of direct investment
any comprehensive source
of
is
of data
in
(1970) for 1967.
We are
these countries.
in
on stocks
OECD
We
not
of portfolio equity investment
developing countries, other than the
B0PSY5 where
coverage of
this variable is very poor.
For countries
we
sources,
relative to
for
infer initial
which no stock information
an estimate of a zero
Valuation.
We
capital.
method as
Applying this
more
observed
stock, which
initial
is
same
difficult
for the
this
use the
since for each country
we need
type of information starting
in
direct
to
we
and
know
distribution of the stock of direct investment
in
therefore use the
in
6c',t-1
38
same
Equation (A2).
portfolio equity
investment
the destination by recipient
OECD
economies,
value of capital located
V,,=(l-5)-Zw,,,..%l.^^.^
rt
of this
OECD.
across recipient countries to
recipient country:
cVc
most
the value of
the mid-1980s as reported by the
construct a weighted average of the changes
(A3)
in
domestic capital stock, as summarized
outward
the
probably correct.
do not vary systematically with the ownership
principle for valuing
In
in
flows are zero, and so this
country of their investment abroad. For direct investment by
have
that asset
the average investment ratio
For inward direct and portfolio equity investment,
valuation
in
of these
available to implement this approach.
is
initial
we use
proceed from the assumption that changes
capital located in a country
is
portfolios,
for portfolio equity investment, the
in
any
order to smooth out year-to-year deviations from the equality
In
three years for which flow data
results
in
stocks as the ratio of the flow of investment
between marginal and average
cases
available
gross domestic investment, multiplied by the domestic capital stock
obtained above.
first
is
Pc',t-i
in
each
we
We
where w^ ^.j
the share of foreign direct investment by country c located
is
We use this weighted
c'.
investment
average to measure changes
our sample,
FDI across destination countries
similar to that of
is
For
portfolio equity investment,
is
distribution
that the distribution of
of the VctS for the 13 countries
the data required to implement Equation (A3).
distribution across recipient countries
its
in
OECD economies. We therefore
measure Vd for these countries as a simple average
we have
the value of outward direct
we assume
in
which
country
13 countries for which data on FDI by destination are available
for the
1990, For the remaining countries
for
in
in
across countries
possible to use the
same
is
we do
not have comparable data on
We therefore assume that
by source country.
sufficiently similar to that of direct
valuation adjustment as
its
we do
investment that
for direct investment.
it
^^
Cross-Border Borrowing and Lending
Flows:
B0PSY5
credits
We rely on
(lines
4619+4703, and 4669+4753). This includes
and other
loans.
with equity, there are
B0PSYY4, which
sample, and so
inward and outward non-equity flows as reported
in
we
all
Data on these items are also available
some changes
in definition
between the
practice turn out to be minor for
extend the coverage of the
most
in
the
debt securities, trade
in
the
BOPSY4. As
B0PSYY5
and
of the countries in our
B0PSY5 backwards
using
B0PSY4
data wherever possible.
Initial
Stocks: Stock data corresponding to these flows are available from a
variety of sources.
For developing countries, the most comprehensive available data
^^
It is worth noting a small caveat here.
Data on flows of direct investment include reinvested
earnings of foreign-owned firms, while data on flows of portfolio equity investment do not. In
principle changes in the stock market valuation of equities will reflect these reinvested
earnings, while changes in the replacement cost of capital do not. To the extent these
reinvested earnings are important, our procedure will understate the stock of portfolio equity
claims. However, the alternative of using stock market data to value equity Is even less
attractive, given the weaknesses of stock market data noted above.
39
on stocks
of debt
is
the World Bank's Global Development Finance (GDF) report,
which provides data since 1970
(line
DT.DOD.DECT.CD). For most
countries, data on stocks of borrowing
and lending are reported
B0PSY4. The
only source of data that
countries
B0PSY5 and B0PSY4, where
is
the
those countries where
lending,
we
infer
Changes
for three
in
reasons.
denominated may change
Finally, the
value of debt
repayment.
In principle
coverage
information at
all
is
very weak. Finally, for
on stocks of borrowing and
the dollar value of outstanding debt net of amortization
The value
of the currencies
relative to the
US
dollar.
may change
each
of
these
with
will
changes
in
which the debt
difficult.
country debts,
we
in
will
We therefore
rely
on the
include adjustments for
full
changes
change over
in
changes
in
time.
and
makes comprehensive
the secondary
recent,
valuation
adopt a more limited strategy. For developing
time series of stocks reported by the
in
of
the perceived probability of
Unfortunately, secondary markets are thin
of debt.
is
Depending on the structure
be reflected
especially for developing country debt, which
adjustments
B0PSY5 and
are aware of for lending by developing
debt and the term structure of interest rates, the value of debt
market price
the
stocks from flows as described above for equity.
Valuation:
can occur
we have no
we
in
industrial
GDF, as these
the value of denominating currencies. For
developing country lending, and for the borrowing and lending of industrial countries,
we do
not have data
maturity structure.
of debt
is
We therefore can do
constant, so that the
V„ =
(A4)
where
even on the currency composition
Pt is
the
change
of these assets,
let
alone their
no more than assume that the nominal value
in
the real value of debt outstanding
is
simply:
^
t-1
US consumer
price index.
It
is
interesting to note that for industrial
countries with high-quality stock data on borrowing and lending, this simple method
yields very similar estimates.
40
other Items
The stocks
in
of cross-border assets described
above cover the majority
we
For completeness,
countries' international investment positions.
of items
measure
also
the reserves of the monetary authority, distinguishing between gold and non-gold
reserves. Stocks of gold are a
component
on foreigners. For the purposes
domestic
capital stock.
as reported
in
of wealth that
of this paper,
we
does not
constitute a claim
therefore include
We measure the stock of gold
in
it
with the
ounces
millions of fine troy
the International Monetary Fund's International Financial Statistics
(IFS) (line 1ad) multiplied by the dollar price of gold.
For most countries, short-term
dollar-denominated debt instruments are an important component of non-gold
reserves. For the purposes of this paper
part of lending abroad.
reported
in
the IFS
We
we
therefore count non-gold reserves as a
measure the stock
of non-gold reserves directly as
(line 1ld).
Data and Results
We implement these
to the
GDP
sample
in
of
ideas using data described above.
98 countries with population greater than
1990 greater than 1000 1990
US
dollars at
with fewer than five years of data on balance of
there are substantial anomalies
were unable
in
PPP.
payments
is
the
million
We then
items,
and per capita
discard countries
and those
for
which
the reported balance of payments statistics that
to resolve with reference to country sources.
problematic source of data
1
We restrict attention
B0PSY5
By
far the
most
and B0PSY4. The electronic versions
these sources often report zero stocks when there are positive flows of assets.
B0PSY4 the
In
we
of
the
data occasionally contain outright errors which needed to be eliminated
by inspecting the data for each variable, country and year. Since data coverage
weak and
we
is
the prevalence of these types of difficulties increases sharply prior to 1966,
begin our sample
in this
year. This leaves us with
an unbalanced panel of 73
countries covering an average of 25 years per country over the period 1966-1997.
41
Finally, occasionally
our procedures described above result
stocks of assets that are negative
in
some
years, most
extrapolating flows backwards from a stock estimate
estimates of
In
commonly as a
in later
years.
In
result of
the vast
majority of cases these negative stocks are very small, less than 0.5 percent of
in
absolute value.
If
the
minimum observation
no smaller than -0.5 percent of GDP, we
shift
in
a given stock series
up the series
is
The data are
available at www.worldbank.org/research/growth.
42
negative but
to eliminate these.
handful of remaining negative stocks are discarded.
GDP
The
Appendix B: Solution Details
we
appendix,
In this
that Equations (5)-(9) describe his/her optimal
consumption and
investment opportunity set that the consumer faces
We shall
X=(v,v*,r,Tc,Ti*,a,a*).
dynamics
(A1)
dt
dX| =|i,
In
denote the
element be given as
of this
+ xi/, dco + xi/*
equilibrium,
the representative
i^i,
\\iu
\^*,
consumer
is
is fully
element of
portfolio rules.
The
described by the vector
this vector
as
x,.
Let the
follows:^"
•dco*+E,i
and
i'"
^i
ds
might be functions of aggregate variables, but
infinitesimal
and does
his/her choices affect these aggregates. Let V°
representative
show
solve the representative consumer's problem and
consumer when s=0 and s=1,
not take into consideration
and V^ be the value functions
respectively. Then,
we have
how
of the
that:
.y^.i.!iii.yi!yl.a.a.^,.
(A2)
ZZ^(M',-v^+^:w;).pk-v^]
j'^x^ax,
,
and the
(A3)
first-order condition
=
1
c"'
Also,
20
is:
- dy'
da
we have
During crises,
state
associated with this Bellman equation
v, v*
that:
and
r
are to be interpreted as the asset prices that would apply
changed.
43
if
the
5 V° = maxl
In
•
<c>
c+
[jt
•
(k
- e) + n *
-e *
+r
•
I
-
+
c]
da
[
I
^ aa'
a-[v^-V°]
and the
^V
2
'
subjectto a =
I
first-order conditions
=
(A5)
k- v-e + v *e*-l
associated with
+
I
2
'
9xf
[i-.
9Xi
"^
.2
2
(A4)
V
^
;
k
>e >
;
e* > 0.
equation are:
this Bellnnan
^^
c-^
aa
(A6)
0=—
— —^-cf
^+
9a
—— r--
=
(A7)
da
^
dV°
=
'
where
Tt *
av°
is
+
—
^-^—-^
+Xv3
^dadx,
9a
P+
ili
-
da
•
k-e
da
da
p
k-e
——
— 7^-—^c^
da
(A9)
a-^
da
-—
=
(A8)
•
3a
d^V°
—
da^
,
•
a*
e
*
-Xv^
^dady.^
-e^a^V"
+>
^dadx;
a-(t)-(X-v) +
—
da
.
a
-^
f^V,
dV''
W;
^'
aa
a
pv-Tii+r|2
da
•
(b
•
v
dV°
*
^
the multiplier of the constraint a=k-ve+v*e*+l; and
*
+ri,
"
rjs
are
P v
^
aa
and
rii ri2
aa
the multipliers associated with the constraints that k>e,
The usual Kuhn-Tucker complementary-slack
ri3e*=0.
It
is
straightforward to verify that
e>0 and e*>0,
respectively.
conditions apply: rir(k-e)=0, •r|2e=0 and
1
1
V° = — •lna + f°(X|) and V^ = —
5
Ina
+ f\X|)
6
solve the Bellman equations (A2) and (A4). Using these value functions and the
order conditions,
it
follows that (A5)-(A9) correspond to Equations (5)-(9)
44
in
first-
the text.
References
Bulow, Jeremy and Kenneth Rogoff, "Sovereign Debt: Is to Forgive to Forget?,"
American Economic Review LXXIX (1989), 43-50.
.
Eaton, Jonatiian and Mark Gersovitz, "Debt with Potential Repudiation: Theoretical
and Empirical Analysis," The Review of Economic Studies XLVIII (1981),
289-309.
,
Eaton, Jonathan and Raquel Fernandez, "Sovereign Debt," in Gene Grossman and
Kenneth Rogoff, eds. Handbook of International Economics Amsterdam,
.
Elsevier, 1995.
and Charles Horioka, "Domestic Savings and
Economic Journal XC (1980), 314-329.
Feldstein, Martin,
Flows,"
International Capital
,
French, Kenneth and James Poterba "Investor Diversification and International Equity
Markets," American Economic Review LXXXI (1991), 222-226.
,
International
Monetary Fund, "Balance
of
Payments
Statistics
Yearbook," various
issues.
International
Monetary Fund, "International Financial
Statistics," various issues.
Jaume Ventura, "Current Accounts in Debtor and Creditor Countries"
Quarterly Journal of Economics forthcoming.
Kraay, Aart and
,
Klenow, Peter J. and Andres Rodriguez-Clare, "The Neoclassical Revival in Growth
Economics; Has It Gone Too far?" in Ben S. Bernanke and Julio Rotemberg
eds.
NBER
Macroeconomics Annual 1997
.
Lane, Philip and Gian Maria Milesi-Ferretti, "The External Wealth of Nations:
Measures of Foreign Assets and Liabilities for Industrial and Developing
Countries," International Monetary Fund Working Paper No. 99/115, 1999.
Lewis, Karen, "Trying to Explain
Economic
Literature
,
Home
Bias
in
Equities
and Consumption," Journal
XXXVII (1999), 571-608.
"Why
doesn't capital flow from rich to poor countries?," American
Economic Review LXXX (1990), 92-96.
Lucas, Robert
E.,
,
and Kenneth Rogoff, Foundations of International
Macroeconomics Cambridge, MIT Press, 1996.
Obstfeld, Maurice
,
45
of
Organization for Economic Cooperation and Development, "Stocl<s of Private Direct
Investments by DAC Countries in Developing Countries End 1967, Paris,
OECD.
Organization for Economic Cooperation and Development, "International Direct
Investment Statistics Yearbook", various issues.
Rider, Mark, "External debt
and
Liabilities of Industrial
Countries," Research
Discussion Paper 9405, Reserve Bank of Australia, 1994.
Sinn, Stefan, "Net External Asset Positions of 145 Countries: Estimation
Interpretation," Institut
fiJr
and
Weltwirtschaft, University of Kiel, 1990.
Summers, Robert and Alan Heston, "Penn World Table Mark 5: An Expanded Set of
International Comparisons", Quarterly Journal of Economics CVI (1991), 327,
68.
Tesar, Linda, and Ingrid Werner, "Home Bias and the Globalization of Securities,"
National Bureau of Economic Research, Working Paper No. 4218, 1992.
United Nations Center on Transnational Corporations, "World Investment Directory",
various issues.
46
>
_
w
>2O
OOO
CO
_l
2W<
2
N
til
J3
o
z
d. °-
n
<2
E
.— *j
—
IS
u
Q.
ce a:
.2
c
ra
c
.Q
CO
o
^ o
ro
w
<
<
CO
CD
•a
n <^
m ^ _i CL
"^
.-Oi6^o$i<>='3
QL1J£5^t2OOTC0I-H
™ bi 5
< LU i
E
CO
>,
OWW
"
:
15
3
-5
c'
t:
CO
3
"
c
CO
CD
uj
'
j»:
2 —
"2 C
O
(D
10
CD
O
OOO
3
CD
<
3
m ^
CD
d)
2
o
CO OT
CO
(A
0)
c
ra
4)
C
3
O
n
z _ ce CO _j Q. <
I o>IOX
y ^ Q_ CO H
O^
to
O
>sr
o
e)-i<_j-i-2z3^Q3X,, q: -^
>; z
QiOccxoarOoH-z^iuyiu > o
^1 a:
<mcQOOOQliJOl-S^ZQ.CO H 3 m
>
JO
o
0)
O)
o
CD
Q.
c
3
j:
E
O
ac
ni
w
XI
•D
n
O)
c
OL
o
<u
>
<1>
(A
<
CO
CD
ro
CD
_-
5?
f-
ro
>^
S 2
c o -p
Q. CL ^
o ro ro
O "o ^m =
z
^
^
D.
-c "a
CJ
OJ
5; q:
!=
ro
I-
CO
ro
c
tu
E
^"
<
.=
.E
-- lCO
"c
ro
cu
^o
2^
<
m
.9
E
ro
1-
CD CD
-Ij=
O
CD
X 2 ^ => ^ D- ^ < K O
33_!<XlUZcoltt:COD^d<
<<CDOCJQQUJU-Li-OOtt
CO
'-'
'-'
OQ
=>
c F
o m
?
OX
_i
ON
CC
S
CO
ro
c
CJ)
0)
CO
-Q
3
5
3
E
E
CD
3
0
Ul
T.
_J
c
gium-
Stralia
3
DC
<<
itzeria
nada
stria
3 3
CD
CD
ro
^
CO
ro
t=
!=
CD
"D
c
E E
1CD
n
cz
CO
CO
c
Cl)
Q
>.
CD
CO LL u_
CD
CD
en
C K
3
i^
to
CO
>
"O
_
,„
CD
ro
3
CO
c
§>
c=
CD
ro
O)
V)
0)
—
o
CO
o
^ o o o
CJ
o C
CD
CD
y o E
CO I-
U)
*
X3
ro
re
h-
H
0)
re
(0
to
O
S
M
CO
> ^ 3 ^
^- Q.=; E ? t: S
o O)
Z Z
o 5
Q. CO
0)
=
3
c;
ro
;
Z
E
CL LU
H3>
Table
2:
Net Foreign Assets Are Small
(Share of net foreign assets
66-73
in
country portfolios
74-81
f/a)
82-89
90-97
66-97
Weighted Average
Industrial Countries
0.013
0.007
0.000
-0.006
0.004
Developing Countries
-0.099
-0.037
-0,081
-0.065
-0.068
EAP
LAC
-0.090
-0.036
-0.051
-0.036
-0.037
-0.095
-0.102
-0.150
-0.101
-0.112
MENA
-0.169
-0.034
-0.121
-0.068
SA
SSA
-0.054
0.133
-0.043
-0.054
-0.063
-0.054
-0.195
-0.138
-0.118
-0.069
-0.137
-0.011
-0.028
-0.025
-0.037
-0.016
-0.120
-0.116
-0.167
-0.139
-0.137
EAP
LAC
-0.098
-0.113
-0.117
-0.080
-0.093
-0.116
-0.122
-0.215
-0.150
-0.153
MENA
-0.108
-0.072
-0.164
-0.165
-0.171
SA
SSA
-0.086
-0.102
-0.115
-0.095
-0.085
-0.202
-0.185
-0.112
-0.035
-0.206
Median
Industrial Countries
Developing Countries
•
Notes: Weighted averages are computed Dver an un balanced Danel 8- year
averages for 68 countries. As a result, changes across periods to a small extent
reflect changes in the composition of the sample. Results using a smaller
balanced panel are similar.
48
t
h-
in
m
^
CO
CD
CNJ
d
o
d
CM
CD
o
d
r--
x—
CM
CD
ci
CD
CO
cz>
O
CM
q
d
CD
^
O o O O o
CO
o oo
CD
CM q o q
d o d d d d d d
CD CM CM CD
CM CM CO
CO
'S-
,-^
CT)
6
a>
o
O
i-
LU
6
CO
^
o
d
q
d
C7>
CD
CD
^
CD CD
in
o
d
*
"^_
1
CD in CO in CO in
CD 'Sh- ^
CO
CM CO
00
T—
CD
CD
^
o o m o O ^
d dd9 d ^
00
00
00
oo
in t^
d
d
,_
CM in
CD in
CD
CO
CM
en CM •* CD ^ o
o ^
CO CO •* in CM
CO
CO o o o X— CO O O oo
d d d d d dd d d
CO
T
cT)
o
d
T
en
op
CNl
CD
o
o
M—
a>
CXI
o
O
C3
in
o
d
o
o
d
o
CD s 00 o
o
CM CO
r^
o o 00
CM
^ d CD CD d
X
in CD
CD
r-^
^
•* CO .,_ 00
00 ^
00 CO
in
CD
o
d q
d q
d
Vt^
d
1
d
sz
1^
ro
(V
^
^
+J
>
a>
(A
(0
a>
in
CO oo
o CO
T—
*
o o O O
T— CM o o
O o
d d d d d d d d d
i_*
u.
CO
UJ
O
S
CO
in
CD
CD h- CO
CM CM
"*
h-
CZ)
o
d
o
d
,_
in
CD
o
d
CM
CO
CT)
CJ>
00
4
t--
H—
O
d
§
u
O
CO
in CD
CD
o
d
CO
"Y
o
CO
q
d
^
in CD t- CO *
o 00
o o CO
O
o CM
q
d d d o CD d d
CT>
CT>
r--
CM
CD
s
d
d
to
»->
0)
(A
o
o
1-
<
UJ
C
^
D)
CD
o
d
o
CN
q
d
CO
in
o
d
o
d
c:^
CO
00 CD in
o
O
in
CM CO
o CD CD CO CM
o ^
o ^.
q
d d d d d d d d ^
CD r^
CM
in
.^—
"51-
^
CO
O
U.
CD
CD
CO
rc^
2
o
0)
z
o
o
CD
r
o
o
o q o o o O o 5
d d d d d q d d
1^ CD
Oi CD
00 in
lO T
CO CM CM 00 CM T
CO
in
CD
CD CD
m^
i^
CO in
UO CO
ro
O
CD
CO
d
d
Q.
, ,
CO
£
a>
ro
^
<i>
i-i
.^
f—
LU
CO
r~-
CJ)
S
CO
(0
q
CD
o
,—
t--
CM
^
q
d
5
o
h- in CO ^ o ^
in
CO CO LO o
CO
CM o o o T— CM o o r^
d d d d d d d dd
^
t~--
c
T
T
q
d
^
^
V
"D
(1>
a.
s
TO
(A
O
**
.^
n
0)
0)
o
O
g
c
c
o
c»
ID
o
d
00
in
o
d
00
CO
CD CD
CD
o
d
^
E
E
3
a
c
c
CD
0>
CJ
O)
c
m
2
c
o
•-•
o
o
z
o
o
o o
d d
^— a>
CD
CO
CM CT)
CO
CD
CD CD
CL
CO X CJ>
CO 00 LU CO
CD in en in
d
<D
Z3
>
"ra
o
en
^
^ o o o
d
o d
ro
-a
1
O
O
in en 00 r- CD
CD CD
CM CO
CO
X—
CD
>,
c
O
c
ro
«
c
o
O
CO
0)
CO
Q)
E
E
D
E
Q
ro
c
o
en
CD
a:
E
D
Q
ro
CO
O)
c
"o
<1>
o
a:
Si
D
CO
CJ
c
ro
O
Dc
a oo
n
ni
CD
ro
Q.
"D
cyi
c
en
.9 .9
ro
ro
£ £
ro
CD
O)
ro
CD
ro
^ JD
O O
IX CD
tt **
c O
o c
CD
en
<X>
en
ro
V)
O
*-•
c
o
o
CD
Q.
ro
O
-2
'n.
CD
CO
en
5 CD
£ >
ro
"a
a
Q
O
CL
Q
a
en
o
E
E
+
Q-
13
in
c
(D
t Q O
o
Q. CD
X CM
LU
!Z
en
en
CD
c
o
o
-C
Q.
Q.
en
0}
g
t:
> >
o
2 O O
o
ro
o
O
zs
>
a.
CD
c
(0
o
a q
X!
c
.5
ro
>
o
ro
>
en
en
O
ro
=!
o
o
CO ro
x:
tt a: CO
d
Table
4:
Foreign Assets Consist Primarily of Borrowing and Lending
fifi-73
74-81
82-89
90-97
fifi-97
Net Foreign Assets
Owned by Foreigners
0.013
0.025
0.007
0.000
-0.006
0.004
Equity
0.024
0.042
0.033
Equity Held Abroad
0.028
0.029
Gross Lending
Gross Borrowing
0.041
0.061
0.031
0.059
0.029
0.032
0.124
0.127
-0.099
-0.037
0,039
0.022
Equity Held Abroad
0.004
0.002
Gross Lending
Gross Borrowing
0.024
0.088
Weighted Average
Industrial Countries
0.053
0.039
0.158
0.112
0.175
0.114
-0.081
-0.065
-0.068
0.029
0.028
0.005
0.003
0.060
0.023
0.002
0.048
0.041
0.045
0.077
0.108
0.082
0.088
-0.011
-0.028
-0.025
-0.037
-0.016
0.021
0.021
0.030
0.061
0.035
0.007
0.007
0.023
0.053
0.028
0.046
0.050
0.119
0.140
0.105
0.044
0.083
0.161
0.199
0.145
-0.120
-0.116
-0.167
-0.139
-0.137
0.039
0.000
0.027
0.031
0.033
0.035
0.000
0.000
0.001
0.001
0.028
0.043
0.128
0.045
0.174
0.056
0.051
0.155
0.160
Developing Countries
Net Foreign Assets
Equity
Owned
by Foreigners
Median
Industrial Countries
Net Foreign Assets
Equity
Owned
by Foreigners
Equity Held Abroad
Gross Lending
Gross Borrowing
Developing Countries
Net Foreign Assets
Equity
Owned
by Foreigners
Equity Held Abroad
Gross Lending
Gross Borrowing
0.105
Notes: Weighted averages are computed over an unbalanced panel of 8-year
averages for 68 countries. As a result, chang es across periods to a small extent
reflect changes in the composition of the sam pie. Results using a smaller balanced
panel are similar.
50
Table
5:
Examples 2 and 3
Foreign Assets
/
Wealth in the North
Sigma=0.00
Alpha
Gamma
0.00
0.01
0.02
0.03
0.04
0.05
0.00
0.375
0.000
0.000
0.000
0.000
0.000
0.10
0.375
0.098
0.054
0.037
0.028
0.023
0.20
0.375
0.161
0.099
0.071
0.055
0.045
0.40
0.375
0.229
0.163
0.126
0.102
0.086
0.60
0.375
0.264
0.203
0.165
0.138
0.119
Sigma:=0.05
Alpha
Gamma
0.00
0.01
0.02
0.03
0.04
0.05
0.00
0.375
0.030
0.015
0.010
0.008
0.006
0.10
0.375
0.119
0.067
0.046
0.035
0.028
0.050
0.20
0.375
0.175
0.110
0.079
0.062
0.40
0.375
0.236
0.170
0.132
0.107
0.090
0.60
0.375
0.268
0.208
0.169
0.142
0.122
In this calculation,
9=0.04
we
(k^"^+k*^"^)"\
set the
average return
We also set world wealth
51
to capital constant at
equal to one, that
is,
4 percent, that
a+a*=1.
is,
we
set
Table
6:
Example
3
Discount on South Equity
(v*)
Alpha
Gamma
0.00
0.01
0.02
0.03
0.04
0.05
0.00
1.000
0.815
0.677
0.578
0.504
0.447
0.10
1.000
0.860
0.730
0.630
0.553
0.492
0.20
1.000
0.879
0.763
0.667
0.591
0.529
0.40
1.000
0.894
0.797
0.713
0.643
0.585
0.60
1.000
0.901
0.814
0.739
0.675
0.621
Foreign Investment
Gamma
0.00
0.01
0.02
0.03
0.04
0.05
0.00
1.500
0.728
0.570
0.476
0.409
0.359
0.10
1.500
1.014
0.737
0.592
0.498
0.431
0.20
1.500
1.193
0.875
0.698
0.583
0.502
0.40
1.500
1.389
1.067
0.867
0.729
0.629
0.60
1.500
1.492
1.189
0.987
0.842
0.734
In this calculation,
e=0.04(k^"^+k*''
South (e*v*/a*)
Alpha
in
we
')"'
set the average return to capital constant at 4 percent, that
We also
set world wealth equal to one, that
52
is,
a+a*=1.
is,
we
set
Figure
1:
Net Foreign Assets Are Small
0.4
0.35
0.3
>
u
c
3
oo
0.25
(1)
0.2
0.15
0.1
0.05
< -0.5 < -0.4 < -0.3 < -0.2 <
-0.1
<0
< 0.1
< 0.2
< 0.3
< 0.4
< 0.5
>
Net Foreign Assets/Wealtli
Notes: This figure plots the frequency distribution of the share of net foreign assets
all country-year observations in an unbalanced panel spanning 68
wealth, pooling
countries over the period 1966-1997.
53
in
.5
Figure
2:
Domestic Capital Stocl<s and Wealth Are Highly Correlated
CHE
NOR
60000
y = 0.9345x+ 1624.1
.-
50000
R^ = 0.9776
Q.
re
u
^ 40000
NZLcAN
75
Q.
30000
re
O
o
«)
(U
20000
£
o
10000
10000
20000
30000
40000
50000
Wealth Per Capita
Notes: Switzerland (a=$1 10000, k = $90000)
is
not
54
shown
to scale.
60000
Figure
3:
The Share of Net Foreign Assets
In
Country Portfolios Increases with Wealth
0.6
0.058X
-
0.6427
CHE
SVEffillW
AUS
CAN
CMR
JAM
CIV
-0.6
8
9
10
11
12
ln(Wealth Per Capita)
Notes: Congo (ln(a)=7.22,f/a -1.55) and Egypt (ln(a)=7.2, f/a-0.98) are influential and are
excluded. Including these two observations gives a slope of 0.09 with a standard error of
0.01.
55
Figure 4: The Share of Net Foreign Assets In Country Portfolios
Is
Persistent
1
y=0.98x-0.00
0.8
R' = 0.93
0,6
<
-1
-0.6
-0.8
0.2
-0.4
0.4
0.6
O.f
Foreign Assets/Wealth(t-1)
1
0.8
y = 0.84X
-
0.03
R^ = 0.58
<
-1
-0.8
-0.6
6
0.2
-p^'
0.4
0.6
0.8
1
Foreign Assets/Wealth(t-5)
1
0.8
0.6
y=0.64x-0.06
R' = 0.30
0.4
y»0.2
0.4
0.6
0.8
1
Foreign Assets/Wealth(t-10)
Notes: This graph plots the share of foreign assets in wealth against its one-year (five-year)
(10-year) lag in the first (second) (third) panels, pooling all country-year observations in an
unbalanced panel spanning 68 countries over 1966-97.
56
Figure
5:
The Persistence
1970
1975
of the Net Foreign Asset Share
1980
1985
Changes Over Time
1990
1995
Notes: This graph plots the slope coefficient of a cross-sectional regression of the share of
in wealth on itself lagged one year, for the year indicated on the horizontal
foreign assets
axis.
The heavy
line is
a three-year centered moving average of these slope coefficients.
57
Figure
6:
Foreign Assets Consist Primarily of Borrowing and Lending
(All
Countries)
0.5
1
0.4y =
0.8243x+ 0.0176
SAU
"
R" = 0.8857
°-2
1
0.20.1
ChJE^
SGP
c
BLX^-'^''^^
-
TTO
nding
Le
-0.4
-0.3
-0.2_^|M§^)
0.2
0.1
0.3
0.4
0.5
Net
of
1
p
>
Share
C_
V
C^
-0.4
-
-0.5
-
Share of Foreign Assets
in
Wealth
(Industrial Countries)
15
0.5
1
0.4
-
y = 0.81 34x+ 0.0035
^'^
R^ = 0.8994
0.2
0.1
c
-
BLX^/^
-
#^
c
yf^T*«c,uy
Le
en
Net
CHE
-0.4
-0.3
-0.2
i^^^S'
)
,
0.1
0.2
of
Share
-0.3
-
-0.4
-
-0.5
-
Share of Foreign Assets
58
in
Wealth
^
,
,
,
'
0.3
0.4
0.5
IT)
CO
CM
o
O
o
o
o
c
o
V
c
o
o
3
DO
E
D)
(0
D>
*
>
o
-c
3
O
>
ro
*2
1
O
1
Si
<
c
o
'^
inoOT-aJt^LDcor-T-roLn
E
3
w
inrOT-CT)h~.incOT-T-coio
T^t-'i-^ooocbcJooc)
(A
<
o
O
CD
0)
c
o
00
o
d
D
0)
Q
si
-
T3
/
O
CO
o
i
O
>
CD
+^
W
c
O
d
1
/
0)
flj
en
O
00
1
/
CM
0>
o
3
/
CNJ
o
1
IT
1
LO
CO
T-
/—
'
T—
CO
LD
I^
O)
o o o o o c o O
1
1
1
r
1
>T1
00
ir>
T1
1
tnco-^-T-coioh-co
cziddddddd
ns
E
Figure
9:
Implications of Theory for Persistence of Foreign Asset Positions
Average Growth
Volatility of
5236 CkO
61
Growth
Date Due
Lib-26-67
MIT LIBRARIES
3 9080 01917 6616