Journal of Development Economics

Vol. 63 Ž2000. 45–58

www.elsevier.comrlocatereconbase

Debt management and crisis in developing

countries
Michael P. Dooley
Social Sciences I, Department of Economics, UniÕersity of California, Santa Cruz, CA 95064, USA

Abstract

Debt management policy for governments of developing countries must balance conflict-
ing objectives. The structure of explicit and implicit government debt influences the amount
of lending private creditors are willing to extend, contractual debt service costs, the
probability of default and the costs of default. Because default is not relevant for
governments of industrial countries, their debt management policies are not a useful guide
for developing countries. The model developed suggests that minimizing debt service costs
is likely to be a very inefficient policy for governments of developing countries because
such a policy increases the cost of default. q 2000 Elsevier Science B.V. All rights
reserved.

JEL classification: F21; F30

Keywords: Debt; Crisis; Default

1. Introduction

Recent financial crises in Asia, Latin America and Russia have demonstrated
the vulnerability of the real sectors of developing countries to changes in financial
market conditions. In this paper we argue that better management of governments’
balance sheets might considerably reduce the frequency of crises and their
associated costs.
Our approach to debt and asset management policy or, more generally, govern-
ments’ financial intermediation, is influenced by characteristics of developing

E-mail address: mpd@cats.ucsd.edu ŽM.P. Dooley..

0304-3878r00r$ - see front matter q 2000 Elsevier Science B.V. All rights reserved.
PII: S 0 3 0 4 - 3 8 7 8 Ž 0 0 . 0 0 0 9 9 - 7
46 M.P. Dooleyr Journal of DeÕelopment Economics 63 (2000) 45–58

countries not shared by industrial countries. Experience suggests that a wide range
of developing counties might choose or be forced to default on their domestic
andror their international debt. Moreover, renegotiations of contracts seem to
generate losses in output for some interval following default. These characteristics
imply that governments must manage the flow of income and expenditures as well
as the nature of the financial contracts they enter into, in order to avoid crises. In
particular, the benefits associated with governments’ financial intermediation must
be balanced against costs of associated financial crises.
The existing literature on debt and asset management assumes that, as long as
the government remains solvent, there are welfare improving roles for government
financial intermediation as a way to overcome a variety of market failures. Several
of these arguments seem compelling when applied to developing countries.
Nevertheless, the implied participation in financial markets by governments of
developing countries carries with it the risk of default and the associated costs. A
preliminary assessment of costs and benefits suggests to us that governments of
developing countries should substantially reduce their role in financial intermedia-
tion and treat avoidance of crises as their primary objective in managing remaining
positions.
We offer some unorthodox rules of thumb for debt and asset management for
developing countries. The basic conclusion is that governments that might be
forced into default should not borrow in order to lend. This would rule out, or
severely limit, apparently unrelated activities such as unfunded pension plans and
sterilized foreign exchange market intervention. This, in turn, rules out all but
polar exchange rate regimes. Free floating or currency board arrangements are
consistent with prudent debt management policies but intermediate regimes are
not. Finally, governments should issue only one type of debt and this one type of
debt should be long term, not indexed and denominated in domestic currency.

2. Debt management policy

A good place to start a discussion of debt management is a AMiller –Mod-

iglianiB theorem for the scale and structure of a government’s gross assets and
liabilities. The conditions under which changes in the government’s balance sheet
are neutral have been carefully set out in the literature ŽStiglitz, 1983; Chan,
1983.. If there is no uncertainty, and if representative, perpetual households
consider the government’s portfolio their own, it seems clear that any financial
position the government takes would be fully offset by the private sector.
With uncertainty, the private sector must also offset the expected and condi-
tional value of their net tax liabilities generated by changes in the market value of
the government’s portfolio. For example, if the government takes a highly
leveraged position by issuing domestic currency debt in order to acquire foreign-
 M.P. Dooleyr Journal of DeÕelopment Economics 63 (2000) 45–58 47

currency assets, neutrality would require that the private sector offset the govern-
ment’s position in the same or equivalent markets. This is the basic reason why
sterilized intervention, or other changes in the economic exposure of the govern-
ment’s portfolio, have no economic effects in a neoclassical model in which the
private sector and the government are assumed to have the same access to credit
markets. One of the many strong assumptions necessary for this result is that
households must be able to forecast their share of taxes associated with the
government’s capital gains and losses, as well as the gains and losses on their
share of government liabilities. Clearly, the household cannot manipulate its
portfolio to offset its contingent tax liability if it does not know what that tax
liability will be ŽFischer, 1983..

3. Portfolio management in an imperfect world

If the private sector cannot offset the exposure and leverage of the government’s
portfolio, portfolio management policy can contribute to or detract from residents’
welfare for a variety of Asecond bestB reasons. One of the more interesting
distortions emphasized in the literature is associated with incomplete credit
markets. By issuing various AnewB types of debt, the government can provide
vehicles that allow the private sector to increase welfare. In some cases, the new
asset allows trade between generations.1 In others the new asset has a desirable
covariance with risks that cannot be hedged using existing markets.2 The idea is
that, once established, a liquid market for a new asset such as a fixed interest
nominal bond of varying maturities provides free information about market
expectations to the private sector. Since the information is free to all there may be
under investment without government intervention. If such debt really is useful to
the private sector, then investors will hold it at a lower yield. This suggests that
minimizing debt service cost is a good indicator that the government is providing
welfare-improving debt instruments.
For emerging markets, it seems likely that this argument is relevant for optimal
portfolio management policy. Emerging markets lack many of the financial
markets that are potentially useful to the private sector. For example, the develop-
ment of a liquid market for long-term, fixed-interest government debt may fill an
important gap in the ability of residents to finance long-term investment.
Another interesting class of models points to debt management as a way for the
government to commit to utilizing conventional taxes rather than surprise inflation

1
Gale Ž1990..
2
See Bohn Ž1988, 1990a,b. for models that evaluate different types of government debt, including
foreign-currency denominated debt.
48 M.P. Dooleyr Journal of DeÕelopment Economics 63 (2000) 45–58

to satisfy the government’s budget constraint.3 The government can commit to

price stability by issuing assets and liabilities that would not benefit from surprise
inflation. The Atime consistency problemB implies that the government can reduce
real debt-service costs by shortening the maturity of domestic currency debt,
indexing returns to inflation or replacing domestic-currency-denominated debt
with foreign-currency debt. This is clearly at odds with the idea that the govern-
ment should develop a long-term domestic bond market in order to AcompleteB the
market.

4. Multiple equlibria

The lesson from the credibility literature is that private expectations concerning
future inflation determine this period’s debt-service costs if the government issues
debt that is subject to the inflation tax. This raises the possibility that changes in
private expectations can quickly generate a crisis. Under some circumstances,
shifts in private expectations can lead to multiple equlibria and self-fulfilling
expectations of high rates of inflation. If there are real costs associated with
financial crises, the government should avoid short-term domestic debt and spread
refinancing evenly over time. In this context, lengthening maturity minimizes the
chance that the debt will have to be rolled over just when expectations are at a bad
point.4
Clearly, multiple equilibrium models provide an alternative reason why govern-
ments must do more than live within their intertemporal budget constraints. These
models also highlight an obvious conflict between policy objectives. The inability
to commit to low inflation means that the private sector will not give up its option
to alter the inflation risk premium by buying long-term bonds unless they are
induced to do so by a very steep yield curve. But the possibility of shifts in private
expectations means that governments should avoid relatively cheap short-term
debt if they want to avoid a crisis. In the following pages we develop the idea that
crises can occur without shifts in private expectations and that this suggests
additional constraints on the government’s participation in financial markets.

5. Default risk

While inflation is clearly a default option for developing countries, we have

considerable evidence that outright default on some types of financial obligations

3
Lucas and Stokey Ž1983. argue that the inability to commit to an inflation policy is a fundamental
distortion in a monetary economy. Blanchard and Missale Ž1994. present a model in which the
maximum maturity of government debt is a decreasing function of the size of the debt.
4
See Calvo Ž1988., Giavazzi and Pagano Ž1990., Alesina and Drazen, Ž1991..
 M.P. Dooleyr Journal of DeÕelopment Economics 63 (2000) 45–58 49

is also an option. Even countries that have never defaulted are exposed to
expectations that such an option is available. This additional option makes
portfolio management for emerging markets more difficult and more important.
Any debt management policy must be evaluated for its contribution to probability
that a default will occur and for the costs of the default if it does occur. The
assumption that inflation is the relevant default technology for industrial countries
is consistent with the idea that alternative types of default are too costly to deserve
serious consideration.5
A very general way to express the problem is that the government must
compare the benefits of acquiring any asset with the costs associated with adding a
liability to the portfolio. This includes some estimate of the contribution of the
liability to the probability that some creditor group will offer terms for rolling over
the liability that the debtor will reject, thereby forcing a renegotiation. The debtor
must also estimate the contribution of a liability to the costs of bargaining
following default.

6. Incentives and information

To understand these issues, we must make some assumptions about the

incentives and information structures relevant for sovereign debtors and their
creditors. Following a relatively well-developed literature for corporate finance,
the key assumption is that creditors can observe that a debtor country pays or not
but not the reason for nonpayment. There are two possibilities. Bad luck may
erode the government’s tax base so that payment is impossible or the government
may choose not to pay and suffer whatever penalties creditors can impose.
Countries with little or no risk of bad luck, and the associated unavoidable
default, will favor debt structures that insure maximum punishment following a
default that could be avoided, that is, a strategic default. The problem for both the
corporate and sovereign debtors is to determine how different numbers or types of
creditors affect the optimal debt structure. The corporate case has been modeled as
follows. All the managers, potential managers, creditors, and potential creditors
look forward to the bids that each will make following a seizure of the firm’s
assets. The composition of debt AmattersB under any set of assumptions where the
value of the asset to individual participants in this game depends on the behavior
of the other participants. Bolton and Scharfstein Ž1996., for example, assume that

5
In the literature, default is generally referred to as a capital levy. A onetime surprise tax on holders
of debt is equivalent to a default on that debt. Capital levies were extensively discussed as an option for
reducing the large debts accumulated during World War II. See Eichengreen Ž1990. for an excellent
discussion of the economics and political economy of this era.
50 M.P. Dooleyr Journal of DeÕelopment Economics 63 (2000) 45–58

new managers earn lower returns on the asset and incur costs of evaluating the
assets. In order to distinguish between one and many creditors, they also assume
that each creditor has the right to seize a well-defined asset and that these assets
are worth more when used together than separately. The ability of a creditor to
block coalitions that can best utilize the seized asset provides the result that the
number of original creditors AmattersB for the expected value of the investment to
the initial manager.
The structure of the game that determines the equilibrium for different debt
structures is quite specific to the assumptions appropriate for corporate finance.
Some of these assumptions do not travel well to the case of sovereign debt. But
the intuition we want to pursue for the case of sovereign debt does emerge in the
AcorporateB models. In general, the government and the corporate AmanagerB must
consider both the contribution of a given debt structure to the probability that a
default will occur and the costs of that default should it occur.
The assumption that makes a corporate or sovereign model interesting is that
which insures that conflict resolution generates dead weight losses for all the
participants. In the corporate finance model outlined above this is a static problem
in which the salvage value of the company’s assets depends on the structure of
debt. For sovereign debt it seems natural to model the dead weight losses that
result from the passage of time during which the creditors and debtors restructure
the existing debt. In international finance this has been a well-known argument
over the effects of a debt overhang. During this interval we will assume that the
ability of residents of debtor countries to utilize the domestic capital stock is
impaired. Since this also reduces the creditors’ potential income there is a
coordination problem that is related to the structure of the debt. This might be
modeled as a war of attrition ŽDrazen and Grilli, 1993. where creditors’ uncer-
tainty about the attributes of the debtor and other creditors generates delay in
settlement.
Clearly if there are one creditor and one debtor, they might assess each other’s
bargaining power and, rather than watch the asset melt, will divide the spoils.6
However, if the two sides are uncertain about the other’s preferences or have very
different expectations about a AfairB division, some waiting and dead weight losses
are likely. The equilibrium condition is that each creditor, and if not liquidity
constrained the debtor, compares the marginal cost of waiting with the marginal
benefit of winning. The benefit is the probability that the other side will concede
in the next instant times the difference between the winner’s and loser’s share of
what remains of the asset. If there are a AfewB creditors, they will have to assess
the others’ expected returns for delaying a settlement. If there are very many
creditors, it might be difficult to arrange any agreement for a very long time.

6
This type of bargaining is assumed, for example, in Bulow and Rogoff Ž1989.. In that model the
adversaries evaluate one another and immediately agree to a payment.
 M.P. Dooleyr Journal of DeÕelopment Economics 63 (2000) 45–58 51

For sovereign debt, the lack of collateral Žor the means to seize it. means that
some alternative threat is necessary to provide an incentive for repayment. The
typical threats that have been modeled involve trade sanctions or prohibition of
future borrowing. The trouble with these enforcement mechanisms is that the
former has never been observed and the latter seems very weak relative to the
amount of debt observed. Moreover, if countries had the ability to impose such
sanctions they should do so regardless of the debt. An alternative is to assume that
creditors can block the use of assets acquired by the debtor government. The threat
evaporates as the assets depreciate. This provides an end to the recontracting game
that seems more consistent with experience. Sovereign debt reschedulings are
eventually resolved.

7. A sovereign debt model

Assume a world which lasts for three periods. In the first period, a foreign
creditor lends the government K to buy assets, where K is a dollar amount. The
risk-free interest rate is assumed equal to zero.
The government uses K to import an asset that in the second period yields x
with probability u and zero with probability 1 y u . The asset lasts for one more
period but depreciates uniformly during the period and yields a certain return y if
utilized for the entire third period. The government agrees to pay z in the second
period. If the government pays less than z, the creditor has the right to impair the
government’s use of the asset until a new agreement is reached for a share of the
residual value of the asset, y.
During the third period the asset is not productive if a negotiation for sharing y
is in progress. The value of the asset declines during the third period from y to
zero. This specification of the punishment technology is appealing because it
means that the creditors are only able to interfere with the government’s ability to
utilize the assets acquired with the foreign funds and for only as long as the assets
last. The alternative interpretation that the creditors can punish the debtor forever
and without regard to the seriousness of the offence is less appealing. One might
think of a subsistence economy lifted temporarily to a higher level of output by an
infusion of foreign capital, but once the capital depreciates the creditor has no way
to push output below the initial level.
If the government can pay, which occurs with probability Q , it will consider a
strategic default. The temptation to keep z, the contractual payment in period two,
is compared to the value of y that the government expects to capture following a
negotiation with the creditorŽs.. The incentive constraint for the government to pay
z if x occurs is

2 z - y y gyts Ž 1.
52 M.P. Dooleyr Journal of DeÕelopment Economics 63 (2000) 45–58

where gyts is the expected share of y that goes to the government following a
strategic default and a negotiation lasting 0 - ts - 1. The value of strategic default
depends on the expected duration of the negotiation in period three. If x is less
than z, the government is AsolventB but illiquid in period two and we assume for
now that the difference is Arolled overB into a payment due at the end of period
three. This simple expression highlights what we believe is a fundamental feature
of international debt contracts. By entering into a contract that is difficult to
renegotiate the debtor can credibly commit to repayment when it is able to repay.
If the government cannot pay, which happens with probability Ž1 y Q ., there is
a similar negotiation. The problem is that the contracts have been designed to
make the necessary renegotiation just as costly. One difference between the two
states of nature is that following a strategic default the government has AsecretB
resources that can be used to overcome the coordination problem and speed a
settlement. This probably accounts for clauses in international debt contracts that
blocks buy-backs of debt at market prices.
After taking all this into account, the creditor must expect to make a fair rate of
return

u Ž z . q Ž 1 y u . cyt b y K s 0 Ž 2.
where cyt b is the share of y that goes to the creditor following an unavoidable
default.
Note that gy, cy, ts and t b are a complicated function of the structure of debt
and that ts will generally not be equal to t b but they are probably related. Thus, a
country with a very small chance of bad luck would choose a debt structure that
generated costly renegotiation following a strategic default. But that debt structure
might also generate costly renegotiations following an unavoidable default.
Following a strategic default, the debtor has resources that might help speed a
settlement. For example, secret buy-backs of sovereign debt may have helped
some debtor countries reduce their debt following the 1982 debt crisis. But in the
case of sovereign debt, it is difficult for the government to conceal assets from its
creditors and knowing a debtor is withholding payment might make creditors even
more unwilling to settle. International organizations may have some power to
punish strategic default because they can withhold additional credits following
default.
The problem for the government is to design a debt structure that maximizes its
net revenue from investment. The general form of the government’s net revenue
function is

u Ž x q y y z . q Ž 1 y u . gyt b s R Ž 3.
Substituting Eq. Ž2. into Eq. Ž3., we arrive at:

R s u x q y y K y Ž 1 y u . yt b Ž 4.
 M.P. Dooleyr Journal of DeÕelopment Economics 63 (2000) 45–58 53

The first three terms of Eq. Ž4. are the first best expected return on the asset if
there is no default. The fourth term is the dead weight loss associated with
rescheduling. A full description of how a portfolio of debt might be designed to
maximize the expected value of Eq. Ž4. is very difficult. A result that stands out is
that the optimal debt structure is related to the probability that the investment will
fail for reasons beyond the control of the debtor government. Analytic and
simulation solutions for this problem are topics for future research. In principle,
the key parameters could be estimated from data on how long it takes to
renegotiate different debt structures and how costly is the renegotiation.
In fact, the difficulty of the problem when viewed in this context makes us very
skeptical about the effects of recent suggestions for altering the architecture of the
international monetary system. Any change in the rules of the game that reduce the
costs of renegotiating international contracts also reduces their value as a commit-
ment mechanism. It is not surprising, for example, that industrial countries have
not been willing to incorporate clauses in bond contracts that would reduce the
voting majority necessary to renegotiate bond contracts. Since industrial countries
are very unlikely to experience an unavoidable default, they have no reason to
favor debt structures that can be renegotiated at lower cost.

8. Liquidity crises

In the above framework, we assumed that conditions one and two always hold
ex ante and that creditors do not force a solvent debtor to default. A more
complete model would take into account the possibility that some creditors might
find it in their narrow interest to force a default if the debtor is illiquid. This would
look like a solvent default in that following a good outcome in period two, the
debtor is unable to make the payment z unless some creditor agrees to provide
additional credit. Suppose there are initially two AgentleB creditors. These two see
it in their mutual interest to roll over the debt of the solvent creditor as assumed
above. But now suppose that one of the gentle creditors is bought out by a AtoughB
creditor. Also assume the tough creditor expects to get all of the residual value of
the asset following a strategic default. In this case the tough creditor might do
better by forcing default. The remaining gentle creditor could still save the day by
buying out the tough creditor but this means rolling over the entire gap between x
and z rather than its share. If there are many gentle creditors this could be a
difficult coordination problem.

9. Adverse selection and debt service costs

Other things being equal, tough creditors will offer relatively good terms as the
probability of default rises because they expect to recover a greater share of their
54 M.P. Dooleyr Journal of DeÕelopment Economics 63 (2000) 45–58

contractual payments at the expense of other creditors. But the tough bargainer
does not internalize the dead weight loss associated with the loss in output
generated by her actions. It follows that market prices are distorted and are a
misleading guide for management of the debt.7 In particular, as default becomes
more likely, debtors may be tempted to choose debt that minimizes contractual
payments. This would be a mistake since the cost of this structure of debt in
default states might be very high.
To avoid misleading market signals, it might be optimal to create as few
instruments and classes of creditors as possible. If there are no AnichesB in which
particularly hostile creditors can gather, they might self-select out of the credit
supply. The industrial countries have adopted just such a rule. In general,
industrial country governments issue one class of liabilities with a range of
maturities but with no implicit or explicit seniority among credits or creditors. An
alternative might be to borrow from a very large number of creditors that cannot
block the use of the asset during renegotiation. The point is that at present we
know very little about the cost of alternative debt structures following default.
Recent proposals to AreformB the structure of international financial arrangements
ŽEichengreen,1999; G-22, 1998. are unlikely to make headway until these issues
are better understood and subjected to empirical testing.
If only one type of debt is issued how should it be indexed? In our view the
dominant consideration is the government’s limited ability to generate changes in
the foreign-currency value of their principal asset — that is, conventional tax
revenue. It follows that governments cannot afford to issue foreign debt, or more
precisely foreign-currency-denominated debt, even if that debt helps commit the
government to price stability. For governments with revenues from oil or a single
commodity this means that financial debt should be partially indexed to these
commodities. But for most countries potential changes in the real exchange value
of their domestic currency implies that domestic-currency-denominated debt is
preferred to foreign-currency debt. Real exchange rate changes have been large,
unexpected and uncorrelated with other economic variables. There is no apparent
way to hedge a mismatch between government receipts and foreign-currency debt.
Indexing the debt to the domestic price level might be optimal but it should be
remembered that giving up a tax base for inflation increases, other things equal,
the probability that the government will have to default on indexed debt.

10. Third parties

Another important factor that conditions the negotiation among private creditors
is the possibility that third parties, in this case governments and international

7
Other creditors will complain and the market value of their claims will fall but this might not be
perceived as a problem for the debtor.
 M.P. Dooleyr Journal of DeÕelopment Economics 63 (2000) 45–58 55

organizations, might be induced to enter into the negotiations. Dooley Ž1995.

develops the hypothesis that the nearly 10 years of depressed economic activity in
debtor countries following the 1982 debt crisis was caused by the prolonged
negotiation between private and official creditors. In that episode commercial
banks knew that a prolonged recession in debtor countries would eventually force
the official sector to buy out private creditors and this expectation blunted the
incentives for private creditors to renegotiate the debt. Moreover, the gradual
amortization of private debt made possible by new credits from the official sector
after 1982 was fully consistent with those expectations.
More recently, the substitution of dollar-indexed, short-term debt ŽTesobonos.
by the government of Mexico for short-term domestic currency debt ŽCetes. in the
months leading up to the December 1994 crisis may have contributed to a quick
renegotiation of the debt. The existing literature on debt management would
emphasize the fact that by indexing the liability the government altered its
incentives to inflate and devalue the currency. Thus, the probability that default in
the form of inflation and devaluation would be optimal was reduced and this
reduced debt service costs.
Our model emphasizes a different aspect of the same policy. The substitution of
Tesobonos for Cetes changed the nature of the rescheduling exercise that followed
the 1994 crisis. The recontracting in this case was resolved quickly by forceful
intervention by official creditors. In retrospect the issue of Tesobonos was a very
good idea if the decision of official creditors, in this case the U.S. government and
international organizations, to lend quickly and in large amounts was due to their
unwillingness to fight a long costly war with a well organized and a powerful
group of holders of Tesobonos.

11. Implementation of a debt management policy

While many countries have established debt and asset management procedures
in recent years most have focused on managing a very limited set of financial
assets and liabilities. The widespread practice of establishing a AbenchmarkB
portfolio for foreign-currency-denominated assets and an independent benchmark
portfolio for foreign-currency-denominated liabilities violates the fundamental rule
that risks are associated with portfolios. The problem with separating asset and
liability portfolios is that there is a perfectly predictable covariance of minus one
between the value of similar assets and liabilities denominated in the same
currency.
But merging portfolios of financial assets and liabilities is only the tip of the
iceberg. We cannot evaluate an activity’s contribution to risk until we know how
the associated balance sheet items interact with all of the other positions that
comprise the balance sheet. Developing country governments have diverse and
highly leveraged portfolios. Important examples of activities that generate assets
56 M.P. Dooleyr Journal of DeÕelopment Economics 63 (2000) 45–58

and liabilities include foreign exchange market intervention, deposit insurance,

lender of last resort commitments, state owned enterprises, membership in interna-
tional organizations and social insurance commitments. Moreover, governments in
developing countries are probably more exposed to capital gains and losses than
they realize. At a minimum, governments that manage exchange rates must
consider the implicit liability to guarantee the domestic currency value of assets
and liabilities in the domestic banking system. If a government decides to
liberalize its domestic credit markets, prudent debt management policy would
immediately recognize an implicit liability that will arise if important institutions
prove not viable in this new environment. A good portfolio is one that balances the
welfare gains from each activity with the costs generated by that activity’s
contribution to default risk.
An elementary principle of financial management is that risks are associated
with portfolios, not with the behavior of individual assets and liabilities. It follows
that the first and perhaps most difficult step in implementing a coherent debt and
asset management policy is to set out a comprehensive balance sheet for the
government. This is a difficult task both conceptually and empirically. Although
only a subset of the government’s marketable assets and liabilities are actively
managed, sensible debt management policy is concerned with the value of the
entire portfolio. The values of assets such as the stream of conventional tax
receipts are difficult to quantify because they are not marketable assets, but
variations in the values of such assets are an important part of the risk profile of
the government’s portfolio. Moreover, implicit and derivative assets and liabilities
must also be quantified.

12. Conclusions

Governments may have a variety of good reasons to engage in financial

intermediation. But while benefits are associated with individual activities, costs
are associated with that activity’s contribution to the risk and return profile of the
whole portfolio. Optimal debt and asset management policy balances the welfare
gains from each activity with the costs generated by the marginal contribution of
that activity to default risk. In evaluating costs we depart from the literature on
debt management for industrial countries in that we focus on the probability that
the government will choose or be forced to default on some or all of its
obligations. This type of analysis is familiar in corporate finance but has not been
widely applied to the problem of sovereign finance.
Some of the conclusions that emerge from the analysis are striking and contrary
to lessons drawn from models that do not focus on default. Governments should
avoid asset and liability structures that are likely to trigger default and it should
avoid debt and asset structures that prolong and intensify the decline in economic
activity that follows default. The usual objective of choosing a riskrreturn trade
 M.P. Dooleyr Journal of DeÕelopment Economics 63 (2000) 45–58 57

off along an efficient frontier implicitly ignores default risk. Such portfolios can be
very costly in that they can both raise the probability of default and the costs that
follow. In fact, we argue that apparently costly portfolios in terms of debt service
costs are likely to be preferred when the cost of default is included in the decision
making process.
Finally, it seems likely to us that welfare gains usually associated with financial
intermediation are probably quite small relative to welfare losses associated with
the larger, and larger variance, portfolios created. If this is the case, governments
of emerging markets should avoid financial intermediation. An important aspect of
this is to limit implicit assets and liabilities associated with exchange rate and
lender of last resort commitments. If such commitments are in place, the govern-
ment should aggressively regulate the behavior of the private sector in order to
control the growth of implicit liabilities.

References

Alesina, A., Drazen, A., 1991. Why are stabilizations delayed? American Economic Review, No. 1 81
Ž5., 1170–1188.
Blanchard, O., Missale, A., 1994. The debt burden and debt maturity. American Economic Review 84
Ž1., 309–329, March.
Bohn, H., 1988. Why do we have nominal government debt? Journal of Monetary Economics 21,
l27–140, January.
Bohn, H., 1990a. Tax smoothing with financial instruments. American Economic Review 80 Ž5.,
1217–1230, December.
Bohn, H., 1990b. A positive theory of foreign currency debt. Journal of International Economics,
273–292.
Bolton, P., Scharfstein, D., 1996. Optimal debt structure. Journal of Political Economy 104 Ž1.,
February.
Bulow, J., Rogoff, K., 1989. A constant recontracting model of sovereign debt. Journal of Political
Economy 97 Ž1., 155–178.
Calvo, G.A., 1988. Servicing the public debt: the role of expectations. American Economic Review 78
Ž4., 647–661, September.
Chan, L.K.C., 1983. Uncertainty and the neutrality of government financing policy. Journal of
Monetary Economics 11, 351–373.
Dooley, Michael, P., 1995. In: Kenen, P. ŽEd.., A Retrospective on the Debt Crisis, Understanding
Interdependence: The Economics of the Open Economy. Princeton Univ. Press.
Drazen, A., Grilli, V., 1993. The benefit of crises for economic reform. American Economic Review
83 Ž3., 598–607, June.
Eichengreen, B., 1990. The capital levy in theory and practice. In: Dornbusch, R., Draghi, M. ŽEds..,
Public Debt Management: Theory and History. Cambridge Univ. Press, Cambridge, England, pp.
191–215.
Eichengreen, B., 1999. Toward a New International Financial Architecture: A Practical Post Asia
Agenda. Institute for International Economics, Washington DC.
Fischer, S., 1983. Welfare aspects of government issue of indexed bonds. In: Dornbusch, R., Simonsen,
M.H. ŽEds.., Inflation Debt and Indexation. MIT Press, Cambridge, MA, pp. 223–246.
G-22, 1998. Report of Working Group on Strengthening Financial Systems, October 5.
58 M.P. Dooleyr Journal of DeÕelopment Economics 63 (2000) 45–58

Gale, D., 1990. The efficient design of public debt. In: Dornbusch, R., Draghi, M. ŽEds.., Public Debt
Management: Theory and History. Cambridge Univ. Press, Cambridge, England.
Giavazzi, F., Pagano, M., 1990. Confidence crises and public debt management. In: Dornbusch, R.,
Draghi, M. ŽEds.., Public Debt Management: Theory and History. Cambridge Univ. Press,
Cambridge, England, pp. 94–124.
Lucas, R.E., Stokey, N.L., 1983. Optimal fiscal and monetary policy in an economy without capital.
Journal of Monetary Economics 12, 55–94.
Stiglitz, J.E., 1983. On the relevance or irrelevance of public financial policy: indexation, price
rigidities, and optimal monetary policies. In: Dornbusch, R., Simonsen, M.H. ŽEds.., Inflation Debt
and Indexation. MIT Press, Cambridge, MA, pp. 183–222.