PUBLIC DEBT IN EMERGING MARKETS: CHAPTER III IS · PDF filethe suggestion that—despite the currently benign ... budgetary funds, ... cies between the debt and the fiscal flows data

The potential risks associated with highpublic debt have long been a concernof economic policymakers around theglobe. In the industrial countries, the

need to strengthen fiscal positions and reducepublic debt to accommodate the pressures thatpopulation aging will put on government budg-ets in the future has received considerable atten-tion in recent years (see, for example, the May2001 World Economic Outlook; Economic PolicyCommittee, 2001; and Turner and others, 1998).For emerging market economies, high publicdebt has often had more immediate conse-quences for economic performance, with debtcrises—and the resulting painful periods of eco-nomic adjustment—having been a recurring fea-ture of the histories of many of these countries.

Following a period of relative calm in the firsthalf of the 1990s, during which public debt inmany countries declined, recent developmentshave once again brought to the fore the issue ofpublic debt in emerging market economies.Public debt has increased quite sharply in recentyears across a broad range of emerging marketeconomies; there have been high profile andcostly debt defaults or distressed debt restructur-ings in Argentina, Ecuador, Pakistan, Russia,Ukraine, and Uruguay; and other countries—Turkey, for example—have experienced severe fis-cal difficulties. These developments have led tothe suggestion that—despite the currently benignenvironment in global financial markets—emerg-ing market economies may once again be on theverge of serious public debt problems.

Discussions of the economic impact of publicdebt go back at least as far as the eighteenth

century, when debt problems in France andGreat Britain began to mount. More recently,the political economy aspects of public debthave also received increasing attention.1 Thereare of course valid reasons why a governmentmay choose to borrow and accumulate debt. Thedebt may be used to fund spending that con-tributes to broader economic and social objec-tives. Financing public investment—for example,by improving physical infrastructure—mightraise the rate of return on private capital or pro-vide something that the private sector would notprovide because of externalities, while higherspending on education or health care mayenhance a nation’s human capital. Further, ifgovernment spending has to be temporarily hightoday because of, say, a war or a natural disaster,debt could be used as a buffer to limit the needto immediately raise taxes (see Barro, 1979).Financing countercyclical fiscal policy also hasan important role in helping stabilize economiesand smooth business cycles.

High public debt can, however, have a signifi-cant negative effect on economic activity. Itrequires high taxes to finance and puts upwardpressure on real interest rates, “crowding out”private investment. When a government is nolonger able to finance its deficits, it is forced tocontract spending or raise revenues, often at atime when fiscal policy is needed to help stabi-lize the economy (fiscal policy becomes procycli-cal rather than countercyclical). When it cannottake these actions, a debt crisis ensues and thegovernment is forced to default or inflate thedebt away (an implicit default), both of whichentail large economic and welfare costs.

49

CHAPTER IIIPUBLIC DEBT IN EMERGING MARKETS:IS IT TOO HIGH?

Note: The main authors of this chapter are Tim Callen (lead), Marco Terrones, Xavier Debrun, James Daniel, andCeline Allard, with consultancy support from Enrique Mendoza. Nathalie Carcenac, Carolina Gutierrez, and BennettSutton provided able research assistance.

1In this literature, debt is seen in a strategic context where the government can use it to finance higher expenditures ortax cuts to boost its reelection prospects, or to try to constrain the actions of successor regimes (see Rogoff, 1990, andPersson and Svenson, 1989).

CHAPTER III PUBLIC DEBT IN EMERGING MARKETS: IS IT TOO HIGH?

50

Obtaining reliable and comparable cross-country time-series data on public debt and itskey components for emerging marketeconomies is not an easy task. Indeed, the con-siderable difficulties indicate the need for con-certed efforts to improve the quality of the datain this important area.

Some of the major issues that arise in puttingtogether a data set on public debt in emergingmarket economies are as follows.• Data availability. Many countries have only

recently started producing reasonably com-prehensive measures of public debt, and theyhave only a limited time series that typicallydoes not go back beyond the early 1990s on aconsistent basis. Sometimes a longer timeseries is available for a narrower definition ofdebt—usually central government—althoughthis is not always the case. Information onexternal public debt is typically more readilyavailable than that for domestic public debt.Data on other key aspects of public debt—including its foreign/local currency denomi-nation and average maturity—are rare(although improving).

• Coverage of the data. For an analysis of fiscal sus-tainability, it is important to have as broad acoverage as possible of public sector liabilities.Preferably, in addition to the liabilities of thecentral government, the liabilities of subna-tional governments and public sector enter-prises should be included, as well as thecontingent liabilities of the government(which may include loan guarantees, publicsector pension liabilities, and even the poten-tial costs of bank recapitalization). In reality,however, it is not possible to put together dataon contingent liabilities for a large sample ofcountries. Even obtaining a comprehensivemeasure of explicit debt is difficult. For exam-ple, some countries only have data availableon a central or general government basis,while even when public sector data is avail-able, its coverage varies between countries.

Some countries include public sector banksand the central bank in the definition of thepublic sector; others do not. Coverage ofextrabudgetary institutions also varies. Forexample, South Africa’s data exclude extra-budgetary funds, while for Korea andThailand the data include the debt of bankrestructuring agencies. In general, data forLatin American countries tend to have thewidest coverage of the public sector, and datafor Middle Eastern countries the narrowest.

• Other definitional issues. Definitions, even for agiven coverage, also vary greatly betweencountries. A major difference is in the use ofgross or net data, and which assets are nettedout. Brazil, because of its experience withhigh inflation, also uses a nonstandard(“valorized”) definition of GDP, where adjust-ment for the effect of inflation on GDP ismade. These adjustments can be substantial.For example, Brazil’s general governmentdebt at end-2002 was 55 percent of valorizedGDP on a net basis, but 86 percent of stan-dard GDP on a gross basis (with two-thirds ofthe increase due to the different GDP defini-tion). Another related statistical issue is thecomparability of public debt and fiscal data.While this is generally less of a problem whencentral or even general government debt dataare used, when public sector debt is consid-ered it is sometimes not possible to get fiscalrevenues, expenditures, and balance data on acomparable basis. In such cases, inconsisten-cies between the debt and the fiscal flows dataare inevitable.Because of the limitations with currently avail-

able public debt data for emerging marketeconomies, the approach that has been taken inthis chapter is to construct two separate data-bases. Data on contingent liabilities, implicitdebt, and arrears are not included for eitherindustrialized or emerging market economies.

The first database focuses on obtaining themost comprehensive measure of public sectordebt that is available, and contains data for 34emerging market and 20 industrial countries cov-ering the period 1990–2002 (although for some

Box 3.1. Data on Public Debt in Emerging Market Economies

Note: The main author of this box is James Daniel.

Given the recent rise in public debt in emerg-ing market economies, two increasingly impor-tant questions are at what point does public debtbecome too high?2 and what policy actions doesa government need to take to ensure that itsdebt is sustainable? A recent paper by Reinhart,Rogoff, and Savastano (2003) has investigatedthe “intolerance” of some emerging marketeconomies to external debt, and has examinedepisodes of large external debt reductions inthese economies. To date, however, few studieshave empirically examined public debt sustain-ability or large public debt reductions in emerg-ing market economies, partly because of thedifficulties in constructing a data set on publicdebt in these countries. This chapter seeks toaddress this gap, and build on the work ofReinhart, Rogoff, and Savastano. In particular, itcompiles a comprehensive cross-country data-base on public debt in emerging marketeconomies, and then applies a number of differ-ent approaches to assess sustainability and ana-lyze past instances in which countries haveundertaken large public debt reductions.

Innovative aspects of the analysis include aninvestigation of how fiscal policy in emergingmarket economies responds to public debt, andthe implications of the greater inherent volatilityof emerging market economies for the sustain-ability of their public debt.

As already discussed, compiling a data set is amajor challenge for any study of public debt inemerging market economies. The availabilityand coverage of public debt data vary consider-ably between countries, and there is no singlesource from which the data can be obtained. Forthe purposes of this chapter, two new data setswere constructed. They both focus on gross pub-lic sector debt, rather than net debt (i.e., wherepublic sector assets are netted out) or the netpresent value of the debt, because of data limita-tions. The first data set contains a broad meas-ure of public debt for the period 1990–2002,and the second a narrower definition of publicdebt, but over a longer time period (1970–2002).The reasons for creating two separate data sets,and the strengths and weaknesses of each, arediscussed in Box 3.1.

PUBLIC DEBT IN EMERGING MARKETS: IS IT TOO HIGH?

51

emerging market economies, data are not avail-able for the whole period). The data for emerg-ing market economies were collected from IMFstaff reports and country economists. Of these 34countries, 19 had data for the public sector, 10for the general government, and 5 for the cen-tral government. For most countries, data are ona gross basis (i.e., financial assets are not nettedout), with Brazil, Egypt, Jordan, Pakistan, andTurkey being exceptions. For the industrialcountries, data are on a general governmentbasis and are taken from the World EconomicOutlook and OECD Analytical databases.

The second data set focuses on constructing alonger time series of data—which is essential forsome of the econometric exercises and event

analyses conducted in the chapter—for abroader sample of countries. This has data from1970 to 2002 for 79 countries (20 industrial, 32emerging market, and 27 other developing) andwas constructed from the World Bank’s GlobalDevelopment Finance database, the IMF’sGovernment Finance Statistics database, theOECD Analytical database, and country-specificsources. For industrial countries, data are againon a general government basis. For emergingmarket economies, total public debt is con-structed as the sum of separately constructedseries for external and domestic public debt.The external debt data are a comprehensivepublic sector debt measure, but the domesticdebt data are on a central government basis.

2Economic theory provides little practical guidance on the optimum level of public debt as it is dependent on the speci-fication of the model (see Aiyagari and McGrattan, 1998).

Public Debt and Fiscal Policy inEmerging Market Economies

Public debt in emerging market economieshas risen quite sharply since the mid-1990s, andcurrently averages about 70 percent of GDP(Figure 3.1).3 This increase in debt has morethan reversed the decline that took place in thefirst half of the 1990s, so that despite the Bradydebt restructuring initiative and large-scale priva-tization programs in many countries, public debtin emerging markets is higher than it was at thebeginning of the 1990s. This is not to say therehave not been success stories—Bulgaria, forexample, has reduced its public debt ratio fromclose to 160 percent of GDP in the early 1990s toless than 60 percent of GDP in 2002—but manyother countries have experienced very largeincreases in their debt ratios. In Argentina, pub-lic debt has risen from 30 percent of GDP in theearly 1990s to 150 percent of GDP at end-2002,while in Lebanon it has increased from 50 per-cent of GDP to close to 180 percent of GDP overthe same period.

The increase in public debt in emerging mar-ket economies in recent years has been concen-trated in Latin America and Asia, with the latterseeing the most notable rise owing to the impactof the financial crisis in the region in the late1990s. In contrast, debt ratios in the transitioncountries in Europe have fallen sharply as anumber of these economies have implementedsignificant economic and fiscal reforms whilethey move toward accession to the EuropeanUnion. In the Middle East and Africa, debt hasremained broadly unchanged, but at uncomfort-ably high levels. The rise in public debt has beenaccounted for by increased issuance of domesticdebt, spurred by domestic financial liberaliza-tion, the decline in inflation (particularly in


52

1992 94 96 98 2000 020

20

40

60

80Latin America

1992 93 94 95 96 97 98 99 2000 01 0210

20

30

40

50

60

70

80

1992 94 96 98 2000 020

20

40

60

80 Asia

All Emerging Markets

Total debt External debt Domestic debt

Source: IMF staff estimates. Unweighted averages. Only countries for which continuous data are available are included. For some countries, continuous data are available for total public debt, but not for the external and domestic subcomponents. Hence, external and domestic public debt do not always sum to total public debt.

Figure 3.1. Public Debt in Emerging Market Economies (Percent of GDP)

Public debt has risen across a broad range of emerging market economies since the mid-1990s. This rise has been due to domestic debt, which now accounts for nearly one-half of total debt.

1

1992 94 96 98 2000 020

20

40

60

80

100

120Transition Countries

1992 94 96 98 2000 020

20

40

60

80

100

120 Middle East and Africa

1

3Emerging market economies are here defined as thosethat were in the EMBI global index at the beginning of2002 plus Costa Rica, Indonesia, India, Israel, and Jordan.Data are for nonfinancial public sector debt (externaland domestic) where available, or the broadest definitionof public sector that is otherwise available. Average figuresare unweighted and only include countries for which con-tinuous data are available for the sample period.

Latin America), and bank restructuring debt.4 Incontrast, the share of external public debt hasdeclined, and now accounts for about one-halfof the total, compared to about two-thirds at thebeginning of the 1990s.

The increase in public debt in emerging mar-ket economies stands in contrast to develop-ments among the industrial countries, wheredebt ratios have generally declined in recentyears (with the notable exception of Japan)(Figure 3.2). Strikingly, after being well belowindustrial country levels during the 1990s, theaverage public debt ratio in emerging marketeconomies is now higher than the average ratioin industrial countries (and much higher as apercent of government revenues).5 It is alsonoticeable that despite the decline in the shareof external debt in total public debt to about 50percent in emerging market economies, it stillremains well above the 25 percent share inindustrial countries. The difference in debtdenominated in, or indexed to, foreign currencyis even larger. Based on a limited sample ofemerging market economies, the foreign cur-rency component is about 60 percent of totaldebt because some domestic government debt islinked to foreign currencies.

What have been the main factors behind theincrease in public debt in emerging marketssince the mid-1990s? The rise appears to belargely accounted for by interest and exchangerate movements and the recognition of off-balance-sheet and contingent liabilities (all cap-tured in the “other” item in Figure 3.3). In anumber of countries, the costs of recapitalizingbanking systems have been particularly high.6

Growth, on the other hand, has acted to reducethe public debt ratio. The primary fiscal balance

PUBLIC DEBT AND FISCAL POLICY IN EMERGING MARKET ECONOMIES

53

All emerging markets

Asia

Latin America

Middle East & Africa

Transition countries

0 20 40 60 80

Industrial countries

All emerging markets

Asia

Latin America

Middle East & Africa

Transition countries

0 100 200 300 400

1992 93 94 95 96 97 98 99 2000 01 0255

60

65

70

75

80Public Debt (percent of GDP)

Source: IMF staff estimates. Unweighted averages. G-7 only.

Figure 3.2. Comparison of Public Debt Levels in Emerging Market and Industrial Economies

Public debt in emerging market economies is now higher than in industrial countries when compared to GDP, and is significantly higher in relation to government revenues. External debt also accounts for a higher proportion of public debt in emerging markets.

Public Debt (ratio to revenue; average, 1992–2002)

External Public Debt (percent of total debt; average, 1992–2002)


Emerging markets

1

12

Industrial countries2

4Reinhart, Rogoff, and Savastano (2003) similarly notethese trends, but across a much smaller subset ofcountries.

5The median public debt ratio is also higher in emerg-ing market economies—66 percent in 2002 comparedwith 62 percent of GDP in industrial countries—and hasshown the same upward trend since the mid-1990s.

6Burnside, Eichenbaum, and Rebelo (2001) model theimpact of contingent financial sector liabilities in the con-text of the Asian financial crisis.

(revenues less noninterest expenditures) has notitself added to the debt stock during this period,but it has not acted in any significant way to off-set the increase in debt that has been caused byother factors. Indeed, primary fiscal balanceshave weakened somewhat since the mid-1990s inall regions except the Middle East and Africa ata time when a strong fiscal effort was needed(Figure 3.4).

The increase in public debt to high levels inmany emerging market economies in recentyears has once again raised concerns about debtsustainability and whether there could be arepeat of the 1980s debt crisis. The long historyof debt crises in many emerging marketeconomies suggests that such concerns are notunfounded. Indeed, the fact that some emergingmarket economies have a long history of default-ing on their sovereign debt raises the questionof why international investors continue to lendto these countries. Evidence, however, suggeststhat investors may not have lost by investing inthese economies, although the ex post risk pre-mia earned on their investment has been small.For example, Klingen, Weder, and Zettelmeyer(2003) find that during 1970–2002 the rate ofreturn on lending to emerging markets was thesame as the return on U.S. government bonds.Over a more recent sample, the ex post risk pre-mium was found to be small, but positive.

Casual observation of sovereign debt defaultepisodes in emerging markets over the past 30years indicates that while the level of publicdebt at the time of a default has varied substan-tially, in many cases it has been quite low. In 55percent of the defaults recorded, public debtwas below 60 percent of GDP—the benchmarkestablished for European Union members inthe Maastricht treaty—in the year before thedefault, and in 35 percent of the cases thedefault actually occurred at a debt ratio of lessthan 40 percent of GDP (Figure 3.5).7 Indeed,


54

-20

-10

0

10

20

30

40

Figure 3.3. Emerging Market Economies: Contributions to the Change in the Public Debt Stock Since 1997 (Percent of GDP)

Primary balance contribution

The recognition of contingent liabilities and interest and exchange rate developments—captured in the "other" category—have largely been responsible for the rise in public debt.

Total increase in public debtGrowth contribution Other contribution

Source: IMF staff estimates. Unweighted averages.

All emerging Asia Latin America Middle East & Transition markets Africa countries

1

1

7Looking at external debt at the time of sovereign debtdefault over the same period, Reinhart, Rogoff, andSavastano (2003) find that external debt was less than60 percent of GNP in 47 percent of cases, but less than

the median public debt-to-GDP ratio in the yearbefore a default was about 50 percent of GDP.Governments in emerging markets have alsodefaulted on their domestic debt through highinflation, particularly in the 1980s and early1990s when several of these economies hadtriple-digit annual inflation rates (and a fewexperienced hyperinflation).8

Not all emerging market economies, however,have experienced debt crises or very high infla-tion rates, indicating that it is difficult to makegeneralizations about these economies as agroup. Indeed, a number of emerging marketeconomies—such as India and Malaysia—havemanaged to maintain relatively high public debtfor a long period without a default. A compari-son between emerging market country defaulters(since 1998) and nondefaulters points to a num-ber of noticeable differences between the twogroups.9 The countries that have defaulted have,on average, a higher ratio of public debt to GDP,a higher debt-to-revenue ratio, a higher propor-tion of external debt in total public debt, and alower ratio of broad money to GDP than thosethat did not default (Figure 3.5).10 Indeed, in anumber of cases it bears noting that debt ratiosprior to the crisis were held down by overvaluedexchange rates, given the importance of foreigncurrency–denominated debt in such cases.

The default experience of many emergingmarket economies stands in stark contrast to thatof industrial countries, where there has been no


55

1992 94 96 98 2000 02-15

-10

-5

0

5

1992 94 96 98 2000 02-15

-10

-5

0

5

1992 94 96 98 2000 02-6

-4

-2

0

2

4Latin America

1992 93 94 95 96 97 98 99 2000 01 02-6

-4

-2

0

2

4

Figure 3.4. Fiscal Balances in Emerging Market Economies (Percent of GDP)

Source: IMF staff estimates. Unweighted averages. Only countries for which continuous data are available are included.

1992 94 96 98 2000 02-6

-4

-2

0

2

4 Asia

Primary balances have weakened slightly since the mid-1990s, except in the Middle East and Africa.

All Emerging Markets

Transition CountriesMiddle East and Africa

Primary balance Overall balance

1

1

40 percent of GNP in only 17 percent of cases. For thecalculations reported here, the default data are takenfrom Standard & Poor’s (2002b) and refer to defaultevents on both external and domestic government debt.Default episodes were matched with available data ontotal public debt to generate the 38 defaults that underliethe chart. Periods of severe fiscal stress that do not resultin default are not captured.

8See the May 2001 World Economic Outlook.9Hemming, Kell, and Schimmelpfennig (2003) provide

a detailed analysis of the role of fiscal policy in 11 recentcrisis episodes in emerging market economies.

10There may of course be other differences between thedefaulters and nondefaulters. In particular, differences inthe maturity structure of the debt may also have played arole. Data limitations, however, precluded examining thisissue in this chapter.

explicit public debt default since World War II(although inflation in many industrial countrieshas eroded the real value of debt, particularlyduring the 1970s). These differences in defaulthistory have led to the view that because of thecharacteristics of emerging market economies—including their inherent volatility, weaker institu-tions, and poor credit history—the level of publicdebt that they can sustain is much lower than forindustrial countries (see Reinhart, Rogoff, andSavastano, 2003, and IMF, 2002).

Certainly, there are a number of features ofthe fiscal structure in emerging marketeconomies that have an important bearing onthe level of public debt that they can sustain.These include the following.• Revenue ratios in emerging market economies are

low. On average, the revenue-to-GDP ratio isabout 27 percent of GDP, compared with 44percent of GDP in industrial countries(Figure 3.6). There are, however, consider-able differences among emerging marketeconomies, with, for example, many of thetransition economies and Israel having ratioson par with industrial countries. Effective taxrates in emerging market economies are gen-erally much lower than in industrial coun-tries.11 The difference is particularly strikingfor direct tax rates, where industrial countriesgenerally have effective direct tax rates of30 percent or more and emerging marketsoutside eastern Europe, often only about 10percent. This low effective tax rate is theresult of inefficient tax systems, significant tax


56

Comparison of Key Indicators of Public Debt(percent)

Public debt ratios are often quite low at the time of a default. There are, however, noticeable differences between defaulters and nondefaulters.

0–20 20–40 40–60 60–80 80–100 100+ Public debt/GDP

0

100

200

300

400

500

0

20

40

60

80

100

120

Figure 3.5. Debt Default and Public Debt Ratios

Defaulters

Debt/revenues Debt/GDP Share of external Broad debt/total debt money/GDP

0

10

20

30

40

Source: IMF staff estimates. Data are an average of 1998–2002. Defaulters refer to countries that have defaulted since 1998.

Distribution of Public Debt Ratios in the Year Before a Default

Shar

e of

all

defa

ults

Nondefaulters

1

1

Left scale

Right scale Right scale Right scale

0

11Estimates of effective direct and indirect tax rateswere computed for a subset of industrial and emergingmarket economies for which data were available. Datawere taken from the United Nations National AccountsStatistics and the IMF’s Government Finance Statistics, andthe calculations use a simplified version of the methodol-ogy proposed by Mendoza, Razin, and Tesar (1994). Thelength of the sample varied across countries dependingon data availability. The effective direct tax rate was calcu-lated as the ratio of total tax and nontax revenue net ofdomestic taxes on goods and services divided by the sumof compensation to employees and total operating sur-plus. The effective indirect tax rate was calculated as theratio of all domestic taxes on goods and services dividedby private consumption.


57

Source: IMF staff estimates. Calculations for the ratios of public revenue to GDP are generally for 1990–2002. Data for emerging market economies are on a nonfinancial public sector basis where available; otherwise, data are on the broadest basis available. For industrial countries, data are for the general government. Effective tax rate calculations are for country-specific periods for which detailed tax and national account data are available. To generate a larger sample of countries, Czech Republic, Sri Lanka, Mauritius, and Tunisia—which have detailed tax and national accounts data available—were included in the calculations.

1

Sweden

Denmark

Canada

Norway

France

Finland

United States

Australia

Ireland

Poland

Czech Rep.

Mauritius

Tunisia

South Africa

Chile

Costa Rica

Philippines

Korea

Mexico

Thailand

Sri Lanka

Peru

0 10 20 30 40 50 60

Emerging market economies generally have lower revenue ratios and effective tax rates than industrial countries.

Figure 3.6. Revenue Ratios and Effective Tax Rates in Emerging Market and Industrial Economies(Percent)

Ratios of Public Revenue to GDP

1

Denmark

Norway

Finland

Ireland

Sweden

France

Canada

Australia

United States

Czech Rep.

Poland

Chile

Thailand

South Africa

Mauritius

Sri Lanka

Tunisia

Mexico

Korea

Peru

Costa Rica

Philippines

0 10 20 30 40 50

Effective Indirect Tax Rates Effective Direct Tax Rates 2

2

2

SwedenDenmark

AustriaNorwayFinlandFrance

BelgiumNetherlands

GermanyItaly

CanadaPortugal

SpainUnited Kingdom

IrelandGreece

New ZealandAustralia

JapanUnited States

IsraelCroatia

HungaryPoland

MalaysiaUkraineBulgaria

BrazilRussia

ChileJordan

UruguayNigeria

PanamaVenezuelaColombia

South AfricaMoroccoEcuador

PeruMexico

ArgentinaTurkey

ThailandKorea

PhilippinesCôte d'Ivoire

IndiaIndonesia

PakistanLebanon

Costa RicaChina

0 10 20 30 40 50 60

exemptions, and a large informal sector. Thedifference in effective indirect tax ratesbetween industrial and emerging marketeconomies is also noticeable.

• Revenues are volatile in emerging market economies.The volatility of revenues—measured by thecoefficient of variation—in emerging marketeconomies is generally much higher than inindustrial countries, although there are excep-tions (Figure 3.7). This is partly due to thegreater underlying volatility of the economy;income, consumption, and the terms of trade(which are often driven by the prices of a fewcommodities) are more volatile in emergingmarkets (see Kose, Prasad, and Terrones,2003).12 There is also a considerable differ-ence in the volatility of effective tax rates(measured by the coefficient of variation).

• Interest costs account for a high proportion of gov-ernment expenditure in emerging market economiesand are volatile. At 5 percent of GDP, interestexpenditures are almost twice as high inemerging market economies as in industrialcountries, and account for an average of about17 percent of expenditures (compared with10 percent in industrial countries). Interestexpenditures are also more volatile in emerg-ing markets because of the structure of publicdebt. With a large proportion of debt eitherexternal or denominated in foreign currency,and revenues in domestic currency, highexchange rate volatility can result in largespikes in interest (and principal) payments rel-ative to government income. Further, domes-tic debt is often of a short maturity, so interestcosts are more sensitive to changes in thedomestic interest rate environment.These differences in the budget and public

debt structures between emerging and industrialcountries are striking and, as will be discussed in

the next section, they have important implica-tions for debt sustainability.

Assessing the Sustainability of PublicDebt in Emerging Market Economies

Before proceeding, it is first necessary todefine the related concepts of government sol-vency and public debt sustainability. A govern-ment is said to be solvent if it is expected to beable to generate sufficient future primary budgetsurpluses to be able to repay its outstanding debt(in more technical terms, the present dis-counted value of future primary fiscal surplusesmust be at least equal to the value of the existingstock of public debt).13 This criterion, however,is not very practical or demanding because, forexample, it would permit a government to runlarge primary deficits for a period of time if itcould commit to running primary surpluses of asufficient size thereafter and so satisfy the sol-vency condition. In reality, a government cannotcommit to such action—running large primarysurpluses for a long period of time would becostly and politically very difficult.

So solvency needs to be viewed in relation to afiscal adjustment path that is both economicallyand politically feasible, and a given debt level isusually thought of as being sustainable if itimplies that the government’s budget constraint(in present value terms) is satisfied without anunrealistically large future correction in the pri-mary balance (see IMF, 2002). Liquidity condi-tions are also important. Even if a governmentsatisfies its present value budget constraint, itmay not have sufficient assets and financing avail-able to meet or roll over its maturing liabilities.Unfortunately, there is no simple rule for deter-mining whether, in practice, a government’s debtis sustainable or not.14 This section therefore


58

12The impact of commodity prices and commodity exports on government revenues is important even for those emerg-ing market economies that have diversified their exports away from primary commodities. In Mexico, for example, oilexports are less than 15 percent of total exports, but oil-related revenues still account for about one-third of public sectorrevenue. Regression results reported in Appendix 3.1 confirm the importance of commodity price developments for theprimary budget balance in emerging market economies.

13Appendix 3.1 shows why the government’s primary fiscal balance, rather than the overall fiscal balance, is the key forthe analysis of public debt sustainability.

14See Chalk and Hemming (2000) for a survey of methods for assessing fiscal sustainability.

ASSESSING THE SUSTAINABILITY OF PUBLIC DEBT IN EMERGING MARKET ECONOMIES

59

Canada

United States

France

Australia

Denmark

Ireland

Norway

Sweden

Finland

Poland

Chile

Czech Rep.

South Africa

Sri Lanka

Korea

Philippines

Thailand

Peru

Mexico

Costa Rica

Mauritius

Tunisia

0 5 10 15 20 25 30 35

Canada

Denmark

Norway

Sweden

France

Australia

United States

Finland

Ireland

Chile

Costa Rica

Poland

South Africa

Czech Rep.

Mexico

Thailand

Peru

Philippines

Korea

Tunisia

Sri Lanka

Mauritius

0 5 10 15 20

Emerging market economies generally have more volatile revenue ratios and effective tax rates than industrial countries.

Figure 3.7. Volatility of Revenues and Effective Tax Rates in Emerging Market and Industrial Economies(Coefficient of variation)

Source: IMF staff estimates. Calculations for the ratios of public revenue to GDP are generally for 1990–2002. Data for emerging market economies are on a nonfinancial public sector basis where available; otherwise, data are on the broadest basis available. For industrial countries, data are for the general government. Effective tax rate calculations are for country-specific periods for which detailed tax and national account data are available. To generate a larger sample of countries, Czech Republic, Sri Lanka, Mauritius, and Tunisia—which have detailed tax and national accounts data available—were included in the calculations.

1

1

2

Effective Direct Tax Rates Effective Indirect Tax Rates 2Ratios of Public Revenue to GDP 2

CanadaAustria

SpainDenmark

FinlandFrance

BelgiumGermany

ItalyUnited States

NetherlandsJapan

United KingdomSwedenPortugalNorway

New ZealandIreland

AustraliaGreece

South AfricaHungary

IsraelUruguay

IndiaCroatia

PanamaChile

RussiaJordan

MoroccoPakistanMalaysia

PolandColombiaThailandEcuadorMexico

PeruBulgaria

PhilippinesUkraine

Côte d'IvoireBrazil

ArgentinaIndonesiaVenezuela

KoreaTurkeyChina

LebanonCosta Rica

Nigeria

0 5 10 15 20 25 30 35

applies a number of different approaches thathave been developed in the economics literatureto look at the issue of public debt sustainability inemerging market economies, and how the situa-tion compares with industrial countries. The aimof the analysis is to look at trends across a broadrange of countries, rather than to focus on thesituation in any one country.

It should be noted up front that the followinganalysis does not take account of the risks thatgovernments face from contingent and otheroff-balance-sheet liabilities. This is because of thedifficulties in compiling cross-country data onsuch liabilities. The recent experience in manycountries, however, has shown that the recogni-tion of contingent or implicit liabilities—particu-larly those associated with the recapitalization offinancial sectors—can add significantly to publicdebt, and in some cases push a situation thathad previously appeared to be sustainable intoone that is clearly not. Box 3.2 provides a discus-sion of the main contingent and other off-budget liabilities that are faced by governmentsin emerging and industrial countries, and therisks that these may present to the fiscal outlook.IMF (2003) also discusses contingent liabilitiesand public debt sustainability.

A Simple Approach to Public Debt Sustainability

Methods for assessing public debt sustainabil-ity usually start from the basic accounting iden-tity that links public sector revenues andexpenditures to the change in the debt stock.One commonly used approach is to view fiscalpolicy as sustainable if it delivers a ratio of publicdebt to GDP that is stable, and then to calculatethe primary budget balance that would achieve

that (known as the “debt stabilizing primary bal-ance”).15 If the actual primary balance is lessthan the debt stabilizing balance, current fiscalpolicy implies an increasing ratio of public debtto GDP, and is therefore viewed as unsustain-able. The difference between the actual anddebt stabilizing primary balance indicates thedegree of fiscal adjustment that is needed toachieve a constant debt-to-GDP ratio. A judg-ment can then be made as to whether such anadjustment is attainable in the political and eco-nomic environment of the country concerned.

Over the past few years, only a small numberof emerging market economies (mainly in Asia)appear to have been running primary budgetsurpluses consistent with what is required to sta-bilize or reduce the ratio of public debt to GDP(Figure 3.8).16 For others—particularly countriesin Latin America—there has been a significantdifference between the actual and debt stabiliz-ing primary balance. Of course, a number ofemerging market economies have recently madeconsiderable efforts to increase their primary fis-cal surpluses, and such actions, if sustained,could address such sustainability concerns.Further, were growth to be stronger or real inter-est rates lower than in the past, a smaller pri-mary surplus would be needed to stabilize thedebt ratio. Among the industrial countries, onlyJapan has had a large gap between its actual anddebt stabilizing primary balance in recent years.

While these types of indicators of debt sustain-ability are useful because they are quite simpleto construct and have a straightforward interpre-tation, their drawback is that they are based onan arbitrary definition of sustainability (i.e., sta-bilize the debt-to-GDP ratio). Incurring tem-porarily high deficits and debt levels, however,


60

15See Buiter (1985), Blanchard (1990), and Blanchard and others (1990). This method is based on long-run, perfectforesight considerations that transform the government’s budget constraint into an equation that maps the long-run pri-mary fiscal balance as a share of GDP into a “sustainable” debt-to-GDP ratio that remains constant over time. The debt sta-bilizing primary balance depends on the debt-to-GDP ratio, the real growth rate, and the real interest rate on governmentdebt. The real interest rate on debt is in practice difficult to measure accurately, and requires, among other factors, abreakdown of debt and interest payments into local and foreign currency that is not always available. Here, an emergingmarket country’s real interest rate is taken as the U.S. long-term real interest rate plus its average EMBI spread. For indus-trial countries, the real 10-year bond yield is used.

16The figure is based on the average primary balance and ratio of public debt to GDP for 2000–02, the average real inter-est rate for 1998–2002, and the average real growth rate for 1990–2002 (1997–2002 for transition economies).

may be appropriate in some circumstances, andit is certainly unlikely that a country should tryand maintain a stable debt-to-GDP ratio at alltimes. Further, it may be of little practical policyuse to know what is needed to stabilize the debtratio when it is already at a high level and leavesa country vulnerable to shocks, such as a suddenstop in capital flows.

How Does Fiscal Policy Respond to PublicDebt Accumulation?

A more flexible approach to assessing debtsustainability is to look at it within the context ofthe broader objectives and constraints of the fis-cal policy decision-making process. One way todo this is to look at the relationship between fis-cal policy instruments (the variables deemed toreflect the actions of policymakers) and theobjectives of fiscal policy (such as stabilizing out-put fluctuations and maintaining debt sustain-ability). Such “reaction functions” or “policyrules” are well established in the analysis of mon-etary policy, but they are much less developed instudies of fiscal policy, and to date have not beenapplied to emerging market economies.17

Fiscal policy reaction functions were sepa-rately estimated for both industrial and emerg-ing market economies, with the primary fiscalbalance being considered the key operating tar-get of the fiscal authorities. The primary fiscalbalance is assumed to respond to public debt,but it is also affected by temporary factors suchas the level of economic activity.18 Within this


61

-10 -8 -6 -4 -2 0 2 4 6 8 10

-4

-2

0

2

4

6

8

10

Emerging markets

Fallingdebt

Figure 3.8. Emerging Market and Industrial Economies: Actual and Debt Stabilizing Primary Balances (Percent of GDP)

Actual primary balance

Source: IMF staff estimates. Calculated using the average primary surplus and public debt during 2000–02, the average real interest rate during 1998–2002, and the average real growth rate during 1990–2002 (1997–2002 for transition countries).

Debt

-sta

biliz

ing

prim

ary

bala

nce

Primary balances in many emerging market economies have fallen short of what has been needed to stabilize the public debt ratio in recent years. This stands in contrast to most industrial countries.

Increasingdebt

1

1


17Such fiscal policy studies for industrial countriesinclude Bohn (1998) for the United States; Mélitz (1997)for OECD countries; Debrun and Wyplosz (1999) foreuro area countries; and Gali and Perotti (2003) forEuropean countries. Favero (2002) makes joint estimatesof monetary and fiscal policy rules.

18For emerging market economies, four temporary fac-tors that affect the primary balance were considered (all ofwhich were found to significantly affect the primary surplusin the estimated fiscal policy reaction function): the busi-ness cycle, inflation, commodity prices, and debt restruc-turing or default. For industrial economies, the temporaryfactors considered were limited to the business cycle andinflation. Appendix 3.1 contains details of the sample selec-tion and econometric methodology used in this section.


62

Conventional approaches to fiscal sustainabil-ity, such as those discussed in the main text,focus on obligations explicitly recognized as lia-bilities in the budgetary system—that is, in prac-tice, the total public debt and the financing ofpresent expenditures. These, however, make uponly a fraction of a government’s potential obli-gations. As has been clearly illustrated in recentyears, the recognition of off-budget obligationscan significantly alter a government’s debt posi-tion. Therefore, sustainability analysis needs topay due attention to the government’s off-budget obligations, which consist of two maincategories.• First, governments face a moral or implicit

commitment to provide public goods andservices in the future.1 These depend on aseries of interrelated factors such as futurepotential growth, demography, and specificpressures on health expenditures, such asthose related to HIV-AIDS.

• Second, governments face obligations that willonly come due if a specific event occurs.2

These contingent liabilities may be the result ofcontractual or legal commitments such asloan guarantees and state insurance schemes,or stem from implicit understandings that, forexample, the government should providerelief in the event of natural disasters, that it

should honor the financial commitments ofinstitutions involved in quasi-fiscal activities,and that it should intervene beyond itsexplicit obligations under deposit insuranceor other guarantees if the stability of thefinancial system is at risk.Estimating off-budget items is subject to con-

siderable uncertainty both because of the noto-rious unreliability of long-run projections andbecause of the very nature of contingent liabili-ties. As a consequence, implicit and contingentliabilities are generally interpreted as a source ofrisk affecting the “core” fiscal outlook resultingfrom conventional debt sustainability analysis.Of course, that fiscal risk comes in addition tothe impact of macroeconomic risk—also dis-cussed extensively in the main text. In recentyears, events in a number of countries indicatethat these two sources of risk can combine toproduce full-blown financial and fiscal crises.

Among the many potential sources of fiscalrisk, obligations related to adverse demographictrends—mainly population aging, but also theimpact of HIV-AIDS on life expectancy in anumber of developing countries—and implicitcontingent liabilities—mainly associated withthe preservation of financial system stability—stand out as the greatest threats to fiscal sustain-ability in both industrial and emerging marketeconomies.

Impact of Aging

The consequences of population aging for fis-cal sustainability are particularly significant inindustrial economies, where demographic pres-sures combine with extensive social security sys-tems. The problem is especially acute whenpay-as-you-go (unfunded) pension systems andpublic health insurance prevail. A recent studyby the European Commission (forthcoming)concluded that by 2050, age-related increases inpension and health care outlays would rangebetween 2.6 percent of GDP in the UnitedKingdom and 11.8 percent of GDP in Greece,bearing “clear risks” to fiscal sustainability in atleast six member states of the European Union.Other studies attempt to estimate the broad

Box 3.2. Fiscal Risk: Contingent Liabilities and Demographics

Note: The main author of this box is Xavier Debrun.A recent and comprehensive coverage of fiscal riskcan be found in Polackova Brixi and Schick (2002).

1If these implicit obligations exceed the future rev-enues implied by current tax policy (in net presentvalue terms), the current policy amounts to shiftingthe payment of the government’s bills to future gener-ations. In a context of population aging, such inter-generational transfers raise particularly difficultissues—not discussed here—and have recentlyattracted a lot of attention (see Auerbach, Kotlikoff,and Leibfritz, 1999).

2Guidance on accounting treatment and disclosureof various kinds of contingent liabilities is provided inthe International Public Sector Accounting Standard(IPSAS), paragraph 19, as released by the Interna-tional Federation of Accountants in October 2002,and in paragraphs 62–66 of the IMF’s Manual on FiscalTransparency.

framework, the connection between policyactions and long-run debt sustainability—the keyissue of interest here—lies in the fact that a posi-tive response of the primary balance to publicdebt generally implies the consistency of fiscal

policy with long-run solvency (see Bohn, 1998,for a formal demonstration, and Appendix 3.1).As discussed earlier, however, long-run solvency(satisfying the present-value budget constraint)is a relatively undemanding criterion as it only


63

stock of implicit liabilities that would be impliedby a continuation of current tax and expendi-ture policies, including unfunded pension liabil-ities and age-related increases in expenditures.For example, in a comparative analysis of 19OECD countries, Frederiksen (2001) found thatthe stock of net implicit government liabilitiesvaried between 84 percent of GDP in the UnitedKingdom and almost 400 percent of GDP inSpain, well above the respective public debtstocks in those countries. Such estimates areinevitably imprecise, and much higher numberscan be found elsewhere in the literature.

Implicit Contingent Liabilities

Although much harder to measure, contin-gent liabilities also constitute a significant risk tothe fiscal outlook, with the implicit guaranteesextended to the financial system and large nonfi-nancial enterprises being the most importantsources of such liabilities. In recent years, astring of banking crises have dramatically illus-trated the vulnerability of the fiscal position tocontingent liabilities. Emerging market econo-mies were particularly exposed, with an averageestimated fiscal cost of almost 20 percent of GDPbeing incurred; in some cases this cost exceeded50 percent of GDP (for example, Indonesia atthe end of the 1990s). The realization of obliga-tions vis-à-vis the financial sector has also affectedindustrial economies, although the average fiscalimpact of banking crises has generally beenmuch smaller than in emerging markets.

Today, a number of countries remain highlyexposed to contingent financial sector obliga-tions. Despite their inevitable imprecision, esti-mated ranges of financial sector liabilitiessuggest potentially very large fiscal costs in somecountries in the event of future banking crises.For example, across a sample of 80 industrial

and emerging market economies, Standard &Poor’s (2002a) most conservative estimates ofthese liabilities range from 3 percent of GDP(Mexico) to 64 percent of GDP (China).

How to Deal With Fiscal Risk?

Contingent liabilities and the growing pres-sures on expenditures from population agingpresent significant fiscal risks in many countries.Consequently, it is essential for fiscal decisionmakers—and for those to whom they areaccountable—to be fully aware of these risksand the alternative fiscal strategies that may beneeded to deal with them. Awareness impliesthe need for realistic long-run projections—anaspect especially important for countries facinglarge age-related obligations—and the completedisclosure of explicit contingent liabilities.3

Budget documents should also provide adetailed discussion of the risks such liabilitiesimply for fiscal sustainability. A better grasp offiscal risk, as well as greater public awarenessabout it, would encourage governments toadopt more prudent fiscal policies, includinglower medium-term public debt objectives(which require structurally stronger fiscal posi-tions in the short to medium run), provisionsfor impending expenditure shocks, and, last butnot least, reforms of pension systems, socialsecurity, and other entitlements. Along withmeasures to make labor and product marketsmore competitive, those reforms would alsohelp boost productivity (see the April 2003World Economic Outlook), with direct feedbackeffects on the speed and the credibility of thefiscal consolidation process.

3It is understandably difficult to disclose implicit lia-bilities, if only because of the moral hazard problemsuch disclosure might create.

requires a commitment to adjust policy in the(possibly distant) future.

Two conclusions follow from examining thelink between the adjusted primary balance (i.e.,after the impact of temporary factors has beenaccounted for) and public debt.19 First, emerg-ing market economies as a group exhibit a loweraverage adjusted primary balance than industrialcountries at any level of public debt (Figure 3.9).Second, the response of the primary surplusweakens as the debt-to-GDP ratio rises in emerg-ing market economies, and this response stopsaltogether when debt exceeds 50 percent ofGDP. This suggests that—on average—the con-duct of fiscal policy in emerging marketeconomies is not consistent with ensuring sus-tainability once public debt exceeds a thresholdof 50 percent of GDP. In contrast, industrialcountries respond strongly to rising debt whendebt is at a high level. Indeed, when debt isabove 80 percent of GDP, the estimated adjust-ment in the primary surplus is almost threetimes as large as that at lower debt levels. Theseestimates of course are for a large sample ofemerging and industrial countries, and thereported results are an average for each sample.Therefore, this behavior is not true for everycountry in either the emerging market or indus-trial country group; some emerging marketeconomies have acted quite strongly to maintaina sustainable debt position.

The analysis also indicates clear differencesbetween emerging market and industrial coun-tries in terms of the cyclicality of fiscal policy(Figure 3.10). While a 1 percentage pointimprovement in the output gap is estimated toresult in an average improvement in the pri-mary balance of only 0.04 percentage point ofGDP in Latin America and 0.23 percentagepoint of GDP in non–Latin American emergingmarkets, it leads to a 0.87 percentage point of


64

Primary Surplus at Different Debt Levels

0

1

2

3

4

5

6

7

0–25 25–50 50–100 100+

Figure 3.9 Relationship Between Public Debt and the Primary Balance(Percent of GDP; 1990–2002)

20 40 60 80 100-2

0

2

4

6

8

10

20 40 60 80 100-6

-4

-2

0

2

4

6

8

10

1

Source: IMF staff estimates. Primary balance adjusted for the impact of transitory shocks.1

Primary surpluses respond much more strongly to debt in industrial countries than in emerging market economies. Indeed, fiscal policy in emerging market economies stops responding to an increase in public debt when debt is above 50 percent of GDP. This stands in contrast to industrial countries, where fiscal policy responds more aggressively when debt is above 80 percent of GDP.

Prim

ary

surp

lus

Emerging Market Economies

Total public debt20 40 60 80 100 120

-6

-4

-2

0

2

4

6

8

10Pr

imar

y su

rplu

sIndustrial Countries

Total public debt

Prim

ary

surp

lus


Emerging markets

Emerging marketsIndustrial countries

Prim

ary

surp

lus

Total public debt

Relationship Between Public Debt and Primary Surplus

Total public debt

19The figures and econometric results discussed in thissection refer to the association between the primary sur-plus adjusted for the influence of temporary factors (as apercent of GDP) and the ratio of public debt to GDPobserved at the end of the preceding year.

GDP improvement in industrial countries.20

These differences are primarily driven by thebehavior of expenditures, which, as a percent ofGDP, are unreactive to cyclical fluctuations inemerging markets (in Latin American countries,expenditures actually appear to be slightly pro-cyclical). In cyclical upswings, outlays expand atthe same pace as economic activity (or faster inLatin America), but when economic growthweakens, revenues decline and lending condi-tions tighten, and outlays fall.21 This behaviorcontrasts to that in industrial countries, whereexpenditures increase by less than economicgrowth in an upturn and fall by less than activityin a downturn, thus exerting a stabilizing influ-ence on the economy. This behavior likelyreflects the significant automatic stabilizers atwork through the extensive social security sys-tems in industrial countries, giving to govern-ment expenditure an insurance role againstmacroeconomic volatility (see Rodrik, 1998, andFatàs and Mihov, forthcoming). Interestingly,better institutional quality is found to be associ-ated with a more countercyclical policy inemerging market economies, suggesting thatthe ability to control expenditures (and raiserevenues) is less of a problem in countries withbetter institutions (see Appendix 3.1).

These results are suggestive of a link betweendebt sustainability and the short-term conduct offiscal policy. Because their behavior indicates astrong commitment to debt sustainability, indus-trial countries can run countercyclical fiscal poli-cies without lenders becoming concerned aboutsustainability issues. In many emerging marketeconomies, however, the ability to adjust fiscal


65

0.0

0.2

0.4

0.6

0.8

1.0

0.030

0.035

0.040

0.045

0.050

0.055

0.060

0.065

Figure 3.10. Emerging Market and Industrial Economies: Sensitivity of Fiscal Policy to the Business Cycle (Percent of GDP; 1990–2002)

-1.0

-0.8

-0.6

-0.4

-0.2

0.0

0.2

0.4Cyclical Sensitivity of Primary Expenditure

Source: IMF staff estimates. Response, in percent of GDP, to a percentage point improvement in the output gap.

Fiscal policy is much more countercyclical in industrial countries than in emerging market economies. Most of the cyclical sensitivity of the primary balance is due to the cyclical response of primary expenditure.

Primary Balance: Cyclical Sensitivity and Response to Debt

Industrial Emerging markets All emerging Latin America countries excluding markets

Latin America

Cyclical sensitivity (left scale)

Response of primary balance to an increase in debt

(linear; right scale)

1

1

Industrial Emerging markets All emerging Latin America countries excluding markets

Latin America

20A number of other studies have found evidence ofprocyclical fiscal policies. For example, Talvi and Végh(2000) argue that fiscal policy is procyclical in most coun-tries outside the G-7, while the April 2002 World EconomicOutlook found that fiscal policy was procyclical in a num-ber of Latin American countries.

21Procyclical fiscal policy in Latin America has implica-tions for social spending and the poor. Braun and DiGrescia (2003) find that social spending in the region isprocylical (although less so than total government spend-ing), and that in crisis situations governments oftenreduce social spending, which adversely affects the poor.

policy to maintain debt sustainability is often indoubt. Lenders therefore quickly become con-cerned when deficits widen, and the tightresource constraint forces governments to cutexpenditures during a downturn, further addingto the economic weakness.

Do Governments in Emerging MarketEconomies Overborrow?

A third approach to assessing public debt sus-tainability is to see if a government is “overbor-rowing” in the sense of whether its debt stockexceeds the present discounted value of itsexpected future primary surpluses. To opera-tionalize such a calculation, expected futureprimary balances are here approximated bythe average primary balance achieved duringthe sample period, on the assumption that agovernment’s fiscal policy track record is thebest guide to what it can be expected toachieve in the future. A benchmark level ofpublic debt (as a percent of GDP) is thencalculated and compared with actual debt. Theextent of over- or underborrowing is measuredby the ratio of actual public debt to the bench-mark level of debt, with a ratio greater than1 suggesting that a government is overborrow-ing relative to what is justified by its fiscal policytrack record.22 The discount rate—the differ-ence between the real interest rate and real out-put growth—is proxied by the differencebetween the real LIBOR interest rate plus acountry-specific spread and the average realGDP growth.23

The benchmark debt-to-GDP ratio was calcu-lated for 50 countries (14 industrial, 21 emerg-ing market, and 15 developing) using data forthe 1985–2002 period.24 The median value ofthe ratio for industrial countries is estimated at75 percent of GDP, three times higher than the25 percent of GDP estimate for emerging mar-ket economies (Figure 3.11). Comparing theactual and benchmark public debt levels sug-gests that many emerging market economieshave indeed been overborrowing as the typical(median) emerging market economy has a ratioof public debt to GDP that is 2!/2 times largerthan its fiscal policy track record would suggestis warranted.25 While this is lower than for the“other developing countries” group, it comparesunfavorably with the typical industrial country,where the ratio is less than 1. There are differ-ences, however, among emerging marketregions. Asian countries have a similar ratio toindustrial countries, while countries in LatinAmerica and other regions have a ratio of 2!/2

and 6, respectively, suggesting significant over-borrowing. Further, the typical emerging marketeconomy with a default history has an overbor-rowing ratio of 3!/2, compared with a ratio of lessthan 1 for a nondefaulter. These results conveythe same message as before: many emergingmarket economies need to generate larger pri-mary surpluses than they have done in the pastto be able to sustain their public debt levels.

The fact that many countries overborrowraises the question of whether there are anycommon features that help to explain thisbehavior. An econometric analysis suggests that


66

22This overborrowing ratio is closely related to the public debt sustainability measure discussed earlier, but it does notprovide a quantitative estimate of the primary balance adjustment needed to stabilize the debt-to-GDP ratio. For a countrythat has undertaken significant fiscal reforms in recent years and is now achieving a higher sustained primary surplus thanit has historically, the assumption that its past track record provides a good guide to future primary surpluses may of coursenot be valid.

23If future growth rates are expected to be higher, or real interest rates lower, than their historic average, this will affectthe estimated overborrowing ratio. Because data on spreads are not available for the whole sample period or for allcountries, the Institutional Investor rating—which is highly correlated with spreads—is used to derive a proxy (seeAppendix 3.1).

24The calculation was not made for those countries where the average primary balance was negative or the discount fac-tor was negative in the sample period.

25Because of a number of outliers, the mean overborrowing ratio for emerging market economies at 16 is much higherthan the median.

the following policy variables are importantdeterminants of overborrowing.26

• Government revenues. Governments with low rev-enues will often have difficulty meeting theirdesired expenditures from revenues, increas-ing the pressure on them to borrow. Theeconometric results suggest that an increase inemerging market economies’ revenue ratio tothe industrial country average would, otherthings remaining unchanged, reduce the over-borrowing ratio by about 35 percent.

• Trade openness. Openness has a positive effecton economic growth, which helps mitigate theexisting debt burden. Further, more openeconomies are able to generate the largertrade surpluses needed to service foreign debtafter an exchange rate depreciation, and aretherefore less likely to experience difficultieswith external public debt.27 The estimates sug-gest that reducing foreign exchange raterestrictions for current transactions—theproxy used here for trade openness—to indus-trial country levels would, other things remain-ing unchanged, reduce the overborrowingratio in emerging markets by 60 percent.28

• The quality of domestic institutions and the natureof the political system. A number of studies havefound a relationship between the quality offiscal institutions—the rules and regulationsby which budgets are constructed andimplemented—and fiscal outcomes.29 Further,


67

0

1

2

3

4

5

6

0

20

40

60

80

100

Figure 3.11. Do Governments in Emerging Market Economies Overborrow?(Median values)

Governments in emerging market countries have a tendency to overborrow; however, there are important regional differences.

"Benchmark" Public Debt Ratio(percent of GDP)

0

1

2

3

4

5

6

Latin American countries Asian countries Other emerging market economies

Source: IMF staff estimates.

Overborrowing Ratio

Industrial countries Emerging market Other developing economies countries

Overborrowing Ratio(emerging market economies)

Industrial countries Emerging market Other developing economies countries

26Other factors not directly under the control of policy-makers—macroeconomic volatility and relative (to theU.S.) per capita income—were also included in theregressions, as was an industrial country dummy variable(see Appendix 3.1 for details).

27On openness and economic growth, see the survey byBerg and Krueger (2003), and on openness and externaldebt difficulties, see Sachs (1985).

28The index of exchange rate restrictions for currenttransactions is used here because it is available for thecountries during the full sample period of the analysis.The reported results, however, remain broadly unchangedwhen alternative measures of trade openness—such asthat developed by Sachs and Warner (1995)—are used.

29See, for example, von Hagen (1992) and von Hagenand Harden (1995). Alesina and others (1998) find thenature of the budget process strongly influences fiscaloutcomes in Latin America.

good institutions are associated with strongergrowth, which boosts revenues and eases thedebt servicing burden.30 On the other hand,political systems that deliver weak (minority orcoalition) governments often delay fiscaladjustment and accumulate public debt basedon short-term needs.31 Simple correlationssuggest that good institutions are associatedwith less overborrowing. In the econometricanalysis, however, only the protection of prop-erty rights was found to be a significantexplanatory variable, with the estimated coeffi-cient suggesting that were the protection ofproperty rights in emerging market economiesto be raised to the level of industrial countries,the overborrowing ratio would be reduced byabout 50 percent.

Uncertainty and Public Debt Sustainability

One of the problems with the threeapproaches to debt sustainability that have beendiscussed so far in this chapter is that they donot take account of the uncertainties that facegovernments in emerging market economies.32

As outlined earlier, government revenues inemerging market economies are more variablethan in industrial economies, and a governmentcould find itself in a situation where it is facedwith low revenues for an extended period oftime because of, say, a collapse in the price ofthe country’s primary commodity export.Further, emerging market governments also faceconsiderable uncertainty from interest andexchange rate movements. There have recentlybeen a number of attempts to incorporate suchuncertainties into the analysis of public debt sus-tainability. One approach has been to apply theValue-at-Risk (VaR) methodology that is com-

monly used in the assessment of financialinstitution risk to look at the risks faced by thegovernment—see Box 3.3 for a discussion of thismethodology. A different approach has been touse economic models that incorporate uncer-tainty to derive estimates of sustainable publicdebt ratios (see Mendoza and Oviedo, 2003).

One way to look at the impact of uncertaintyon public debt sustainability is to consider thecase of a government that is credibly committedto servicing its debts in all circumstances. Such agovernment would need to take into account thefact that its future revenues—and consequentlyprimary balance outcomes—are uncertain, andthat it could be faced with the possibility of along period of low revenues in the future. To becredibly committed to servicing its debt in all cir-cumstances, the government cannot borrowmore than the debt that it would be able to sus-tain with the primary balances that would occurwith these low revenue outcomes.33 This is not tosay that the government could not borrow at all:if actual debt were below the maximum sustain-able debt level, the government would be able toborrow until the threshold was reached, at whichpoint it would need to reduce expenditures tomaintain the credibility of its commitment.

The requirement that a government shouldonly borrow up to the debt level that it could sus-tain in the face of a long period of low revenuesmay seem a stringent one. Emerging markets,however, have faced long periods of low revenuerealizations in the past when the price of theirmain commodity export has fallen. For example,governments in oil-exporting countries faced thissituation after the collapse of oil prices in the1980s.34 In such circumstances, the governmentis suddenly confronted with a debt stock that ithad believed was sustainable when revenues


68

30See the April 2003 World Economic Outlook for an analysis of the relationship between growth and institutions.31Alesina, Perotti, and Tavares (1998) find that coalition governments often have a harder time consolidating fiscal pol-

icy than do single party governments.32See Gavin and others (1996) for an extensive discussion of the effects of volatility on fiscal policies in Latin America.33Revenue volatility can also create liquidity problems even if long-run public debt is sustainable.34Indeed, slumps in commodity prices—particularly oil—are generally quite long lasting. For example, Cashin,

McDermott, and Scott (2002) find that slumps in commodity prices typically last for about three and a half years, withslumps in oil prices on average lasting over four years.

related to commodity exports were high, butwhich is not sustainable with the new reality oflower revenues from commodity exports.

To implement these ideas, it is first necessaryto determine what constitutes a low revenueoutcome, and in such circumstances, what fiscaladjustment the government could make. Here,a low revenue outcome is characterized by arevenue-to-GDP ratio that is two standard devia-tions below the average level, and the range ofprimary expenditure reductions that emergingmarkets have made in the past is taken as anindication of the fiscal adjustment that a govern-ment could potentially achieve. Using theseassumptions, Figure 3.12 shows the maximumsustainable public debt ratios for two “typical”emerging market economies and an industrialeconomy for different assumptions about thepossible variability of their future revenues(measured by the coefficient of variation) andtheir commitment to adjust expenditures if alow revenue outcome occurs. In the calcula-tions, both emerging market economies areassumed to have revenue and primary expendi-ture ratios of 20 percent of GDP on average—broadly the averages seen in non-Europeanemerging markets—while one (a “low-risk”country—Case A) has a real interest rate onpublic debt that is 5 percentage points higherthan its growth rate, and the other (a “high-risk” country—Case B) has a real interest ratethat exceeds its growth rate by 10 percentagepoints.35 The industrial country (Case C) hasrevenue and primary expenditure ratios of40 percent of GDP on average, and a real inter-est rate that is 2.5 percentage points higherthan its growth rate.

Looking at the first emerging market countryexample (Case A), the more stable its revenues—i.e., the smaller the coefficient of variation of therevenue ratio—the higher is the maximum ratio


69

85

31 0

20

40

60

80

0.01 0.03 0.05 0.07 0.10

Source: IMF staff estimates. Based on the Mendoza-Oviedo Model. Zeros indicate that the government can not sustain a positive debt-to-GDP ratio under these conditions.

85

31 0

50

100

150

200

250

300

350

0.01 0.02 0.03 0.04 0.05

85

31 0

25

50

75

100

125

150

0.01 0.03 0.05 0.07 0.10

Countries with more variable tax revenues, less ability to adjust expenditures, and a larger difference between the real interest rate and the real growth rate are able to sustain lower public debt ratios.

Figure 3.12. Maximum Ratios of Sustainable Public Debt to GDP

Low-Risk Emerging Market Economy (Case A)

Coefficient of variation of revenue-to-GDP ratio

Max

imum

sus

tain

able

deb

t-to-

GDP

High-Risk Emerging Market Economy (Case B)

Industrial Economy (Case C)

1

1


Max

imum

sus

tain

able

deb

t-to-

GDP


Max

imum

sus

tain

able

deb

t-to-

GDP

Adjustment in ratio of primary

expenditures to GDP


expenditures to GDP


expenditures to GDP

35While these assumed differences between the realinterest rate and the real growth rate may seem high, theyare intended to capture a situation where a country hasbeen hit by a shock and spreads have increased sharplyand growth weakened.


70

Conventional approaches to assessing fiscalsustainability and vulnerability—such as scenarioanalysis or summary indicators—are subject to anumber of weaknesses. Besides limitations incoverage (they usually focus only on debt, andtherefore exclude a range of contingent assetsand liabilities and other components of the pub-lic sector balance sheet), they do not adequatelyaddress the downside risk of adverse future eco-nomic and financial outcomes. Attempts to cor-rect this deficiency—particularly critical foremerging market economies that typically face avolatile economic environment—have usuallyfocused on sensitivity (or stress) tests withrespect to arbitrary changes in selectedvariables.

A recent extension of the Value-at-Risk (VaR)methodology—which is commonly used in theassessment of financial institution risk—to thepublic sector balance sheet provides a morecomprehensive assessment of a country’s fiscalsustainability. In this approach, the fiscal risksthat are faced by the government and the poten-tial impact that these may have on its net worthposition (assets minus liabilities, which aredefined to include the present value of contin-gent assets and liabilities) can be explicitly mod-eled.1 Among other things, such fiscal risks mayinclude interest and exchange rate movements,commodity price changes, and output fluctua-tions. The VaR analysis assesses the effect thatmovements in such variables, and their co-movements—for example, between oil pricesand exchange rates—have on a government’snet worth position, and it summarizes the worstpossible position that the government could bein after a given time period for a given level ofconfidence. Put a different way, the VaR pres-ents a numerical estimate of the potential loss in

net worth the government could face over agiven period of time if a “worst-case” scenariowere to develop.

The VaR analysis proceeds broadly as follows.The first step is to calculate the government’scurrent net worth from its balance sheet. Butthis net worth position in the future is uncer-tain, and will be strongly affected by the out-comes of key variables that affect the value ofthe government’s assets and liabilities. For exam-ple, in a country that is a large oil exporter, thevalue of the government’s assets will be influ-enced by the future price of oil, which is uncer-tain. Likewise, the value of both assets andliabilities may be affected by future exchangerate movements. To capture these uncertainties,the possible future movements of the main riskvariables that affect the balance sheet, and anyco-movements between these variables, are esti-mated. Based on these estimates, an overallprobability distribution of the government’s networth is calculated, and the overall risks to thegovernment’s balance sheet can then beassessed. For example, while the estimated networth may currently be, say, 100 percent of GDP,the calculations may suggest that because of therisks the government faces there is a 5 percentchance that in one year its net worth will only be60 percent of GDP. In this case, the govern-ment’s “value-at-risk” is said to be 40 percent ofGDP.

From the VaR analysis, the government canidentify the factors that present the most signifi-cant risks to its net worth position, and thepotential size of these risks. To the extent possi-ble, it can then act to mitigate or limit theserisks, while also pursuing other fiscal reformsthat could strengthen its overall balance sheetposition and make it more resilient to suchshocks if they occur.

While in principle the VaR approach could beapplied to any country, incorporating the spe-cific behavioral and institutional features of thatcountry, at present the absence of data on thepublic sector balance sheet in most emergingmarket countries precludes the widespreadapplication of the technique. A stylized applica-

Box 3.3. Assessing Fiscal Sustainability Under Uncertainty

Note: The main authors of this box are GeorgeKopits and Tim Callen.

1See Barnhill and Kopits (2003). In a similar spirit,IMF (2003) reports the results of a stochastic simula-tion exercise that looked at the probability distribu-tion of future public debt outcomes for a sample of 41emerging market countries.

of sustainable public debt to GDP for any givenlevel of expenditure adjustment that it can com-mit to. The rationale for this is that when thegovernment is faced with a low revenue out-come, the actual revenue-to-GDP ratio will behigher, and consequently the primary surpluslarger, than if the variability of revenues isgreater. For example, if this country has a coeffi-cient of variation on its revenue ratio of 5 per-cent and can commit to adjust primaryexpenditures by 5 percent of GDP, then its maxi-mum sustainable public debt ratio is 60 percentof GDP. For the “high-risk” emerging marketcountry (Case B) with similar revenue andexpenditure characteristics, the maximum sus-tainable debt ratio is just 30 percent of GDP.But, if the coefficient of variation for this coun-try is 7 percent, then the maximum debt level isonly 22 percent of GDP. For the industrial coun-try (Case C), the combination of a higher aver-age revenue ratio, low revenue volatility, and asmaller difference between the real interest rateand the real growth rate means its maximum

sustainable debt ratio is higher than for theemerging market economies even if it can onlycommit to a modest cut in expenditures. Forexample, with a commitment to cut primaryexpenditures by 3 percent of GDP and revenuevolatility of 3 percent, the maximum sustainabledebt ratio for the industrial country is about85 percent of GDP.

These calculations illustrate the link betweenrevenue generation capacity, revenue variability,and primary expenditure adjustment—all ofwhich affect the primary balance—and debt sus-tainability. If a country has low and variable gov-ernment revenues, it will be able to sustain alower public debt level than a country with ahigher and more stable revenue base. Thismeans that the sustainable debt level may vary—potentially by a considerable amount—betweencountries (it will also depend on real interestrates and growth). The implication is that differ-ences in sustainable debt levels can be expectednot only between industrial and emerging mar-ket economies, but also among emerging market


71

tion of the VaR approach has, however, recentlybeen conducted for Ecuador, a country wherethe necessary data are available (see Barnhilland Kopits, 2003). The results from this exer-cise, which are briefly discussed below, arebroadly illustrative of some of the fiscal risksthat many emerging market countries face, andshould not be taken to imply that these risks aregreater in Ecuador than elsewhere.

The VaR calculations—which were carried outon the balance sheet for the year 2000—suggestthat the public sector in Ecuador faces a num-ber of important fiscal risks. The government’snet worth position was found to be vulnerable toshocks that could affect the present value of itsfuture income from petroleum reserves, theprofits of its state-owned enterprises, the net lia-bilities of the public pension system, and itsexternal liabilities. The VaR analysis was alsoused to look at the relative importance of spe-

cific fiscal risks in Ecuador by separately assess-ing the potential impact of interest andexchange rate movements and oil prices on thegovernment’s net worth position. Interest ratevolatility was found to be the single most impor-tant source of risk facing the government inEcuador. If, hypothetically, the government wasable to eliminate the interest rate risk it faced, itwould significantly reduce the potential erosionin net worth that it could experience in a worst-case scenario. Exchange rate movements werealso an important source of risk at the time, andEcuador’s move to dollarization—which wasactually introduced in early 2000—consequentlyreduced the risks to the government’s balancesheet. Finally, if oil prices—the only truly exoge-nous source of risk—had been hypotheticallystabilized, this would also have limited thedownside to net worth, but by less than eliminat-ing interest or exchange rate risk.

economies themselves. For example, India—which has relatively stable government rev-enues—could be expected to sustain a higherdebt level than Venezuela, where revenues aremuch more variable. (Of course, there may alsobe other reasons why India could sustain ahigher public debt ratio, including the maturityprofile and interest costs of the debt, the size ofthe domestic bond market, and its relativelystrong growth rate.) Indeed, countries withhigher average revenue ratios and lower revenuevariability do in general have higher public debtratios (Figure 3.13). Because revenue variabilityhas important implications for debt sustainabil-ity, proposals have been made to create debtinstruments that could help cushion emergingmarkets from changing economic conditions(Box 3.4).

Can Governments in Emerging MarketsEconomies Sustain Their Current Debt Levels?

A common theme running through theresults presented in this section is that histori-cally many emerging market economies havenot generated large enough primary budget sur-pluses to ensure the sustainability of their pub-lic debt. This stands in contrast to industrialcountries. This inability to generate adequateprimary surpluses is both a function of weak rev-enue bases (which generally have low yields andare volatile) and an inability to control expendi-tures during economic upswings (this appearsto be particularly important in Latin America).These factors suggest that emerging marketeconomies can generally sustain lower publicdebt ratios than industrial countries. Althoughthis sustainable debt level will certainly vary—and potentially by a considerable amount—thecalculations based on past fiscal performancesuggest that for the typical emerging marketeconomy it is quite low. Of course, industrialcountries face considerable pressures from pop-ulation aging going forward, so this analysisshould not be taken as suggesting that publicdebt levels in these countries are currently at acomfortable level.


72

10 20 30 40 50 600

20

40

60

80

100

120

140

Figure 3.13. Ratios of Revenue and Public Debt to GDP

Revenue-to-GDP ratio

0 5 10 15 200

20

40

60

80

100

120

Coefficient of variation of public revenue

Source: IMF staff estimates.

Publ

ic d

ebt-t

o-GD

P ra

tio

Countries with higher and less volatile revenue ratios tend to have higher public debt levels.

Publ

ic d

ebt-t

o-GD

P ra

tio

How Can High Public Debt LevelsBe Reduced?

If governments face high public debt levels,what can they do to reduce them? Governmentshave a number of potential policy options avail-able to them to reduce their debt: (1) they canadjust fiscal policy and run primary budget sur-pluses sufficient to reduce the debt; (2) they canseek to grow or inflate their way out of theirdebt difficulties; (3) they can sell assets to retiredebt; or (4) they can explicitly default on thedebt.

While reducing the public debt ratio throughstrong economic growth would generally be agovernment’s preferred option, growth isbeyond the direct control of the government. Ofcourse, the government can play an importantrole by creating an environment conducive togrowth through the implementation of soundmacroeconomic and structural policies (includ-ing by not accumulating excess debt that couldadversely affect private sector activity).36 Theother options each have advantages and disad-vantages. Reducing public debt by running pri-mary budget surpluses, for example, maintainsthe fiscal credibility of the government, but isoften difficult politically—particularly if high pri-mary surpluses need to be maintained for anylength of time—and may involve decisions that,

at least in the short run, have a detrimentaleffect on activity.37 An explicit default or highinflation provide ways of reducing debt withouthaving to run larger primary surpluses, but theyboth entail costs. If it defaults, a government islikely to suffer a loss of reputation that couldprevent or limit its future borrowing, and henceconstrain its future fiscal policy options, whilehigh inflation has significant negative effects oneconomic activity and welfare.38 Finally, a policyof selling government assets is only likely to besuccessful in reducing debt if accompanied byresponsible fiscal policy (so the proceeds are notsimply spent), and the policy does not changethe underlying net worth position of the govern-ment although it reduces debt.

To examine how large public sector debtreductions have occurred in practice, data for 79industrial, emerging market, and other develop-ing countries for the period 1970–2002 wereused, and a sample of large public debt reduc-tions was constructed as follows. Cases were iden-tified where public debt was reduced over athree-year period, and then the top 15 percentof these episodes (in terms of the size of thedebt reduction, which in the sample corre-sponded to a drop in public debt of at least 18percent of GDP) were chosen. Lastly, cases inwhich the debt stock at the end of the three-year

HOW CAN HIGH PUBLIC DEBT LEVELS BE REDUCED?

73

36A simple correlation between public debt and growth in emerging market economies since 1990 shows a clear negativerelationship. More formally, Pattillo, Poirson, and Ricci (2002) find that external debt begins to have a negative effect ongrowth once it exceeds 35–40 percent of GDP.

37Assessing the impact of fiscal consolidation on economic activity is not straightforward. While most evidence points tothe conclusion that fiscal multipliers are positive—i.e., that a fiscal consolidation will have a negative impact on growth inthe short run—this appears not always to be the case (see Hemming, Kell, and Mahfouz, 2002). Recent studies in advancedcountries have shown that if fiscal consolidation is mainly achieved through a reduction in current spending it may beexpansionary (see Alesina, Perotti, and Tavares, 1998). For emerging market economies where there is a public debtsustainability problem and the risk premia on interest rates are high, a credible fiscal consolidation could result in a largefall in interest rates, spurring private activity and more than offsetting the withdrawal of fiscal stimulus. Hemming andTer-Minassian (2003) discuss the impact of fiscal tightening during crisis episodes.

38The costs of an explicit default and/or high inflation are difficult to measure. For an extensive discussion of reputationand sovereign debt, see Obstfeld and Rogoff (1996) and the references therein. A default affects a country’s access to capi-tal markets, its borrowing costs, and its trade relations with its debtors. Empirical evidence on the size of the costs ofdefault, however, is mixed. For example, Lindert and Morton (1989) argue that investors pay little attention to the pastrepayment record of a borrowing government. Özler (1993), however, finds that countries with default histories facedhigher commercial bank interest rates in the 1970s. In terms of costs through the trade channel, Rose (2002) finds that asovereign debt default is associated with a decline in bilateral trade between a debtor and its creditors of about 8 percent ayear and this persists for about fifteen years. With regard to the costs of high inflation, Lucas (2003) estimates that thegains from eliminating an inflation rate of 200 percent—a level observed in many Latin American countries during the1980s—are in excess of 5 percent of income in the long run.


74

Highly leveraged emerging market economiesare heavily exposed to volatility in economicconditions, resulting in increased risk of finan-cial distress and even debt crises. Debt instru-ments with repayments linked to keymacroeconomic variables could help cushionemerging markets from unexpected changes ineconomic conditions. In particular, the idea ofcreating bonds indexed to GDP—or, equiva-lently, GDP growth—has recently regainedattention (see Borensztein and Mauro, 2002).

Growth-indexed bonds could work as follows.Consider a country whose real GDP has beengrowing for many years at 3 percent, and isexpected to continue doing so. Assume that thiscountry can issue regular bonds at 7 percentinterest. That country could contemplate issuinggrowth-indexed bonds whose yearly coupon pay-ments will be increased by, for example, 1 per-centage point for every percentage point bywhich GDP growth exceeds its 3 percent trend,with a symmetric reduction in the coupon ratewhen growth falls short of the reference growthrate (the contract could specify that coupon pay-ments cannot fall below zero). In years whengrowth turns out to be 1 percent, the couponwill be 5 percent, and in years when growthturns out to be 5 percent, the coupon will be9 percent.

Such growth indexation results in a numberof advantages. First, growth-indexed bondswould help to reduce the volatility of debt-to-GDP ratios, thereby reducing the likelihood ofcrises. When GDP growth turns out lower thanusual, interest payments due will also be lowerthan in the absence of indexation, and viceversa. Second, growth-indexed bonds provide an“automatic-stabilizer”-type mechanism, thusreducing the need for procyclical policies.These bonds would help avoid politically diffi-cult adjustments in the primary balance at timesof weak economic performance and, conversely,they would help avoid excessive fiscal expan-

sions in times of strong growth. This is espe-cially important for emerging market econo-mies, where economic downturns often lead towaning confidence in international financialmarkets, forcing them into untimely fiscaltightening to defend credibility. But interestsavings, and the corresponding room for asomewhat lower primary fiscal surplus, in timesof weak economic growth might also proveappealing for advanced countries—particularlythose with limits on their overall fiscal deficits,such as the EMU countries. Finally, growth-indexed bonds would improve risk sharing atthe international level. Indeed, individual GDPrisk has relatively low correlation with globalrisk and can be largely diversified in a financialportfolio; global investors holding growth-indexed bonds issued by a variety of countriesmight thus be willing to accept a relatively lowrisk premium (see Borensztein and Mauro,2002).

While focusing on GDP risk has intrinsicappeal, there are also other sources of riskaffecting the debt-service capacity of emergingmarkets. Terms of trade risk has been stressed inthis regard, supporting the idea of debt instru-ments with repayments adjusted to the worldprice of some key commodity, for example.However, the economic structure of manyemerging market countries is becoming increas-ingly more diversified. Indexing bond paymentsto the prices of one or two key commodities mayprovide significant insurance only to a handfulof countries. Indexing to the growth of GDP—the broadest measure of how the economy isdoing—would provide far greater insurancebenefits.

Similar proposals have been on the table forsome time. In particular, the idea of linkingdebt payments to growth, exports, or exportprices generated considerable interest in theaftermath of the debt crisis in the 1980s. In theevent, a few of the Brady bonds that were issuedto restructure syndicated bank loans included“value recovery rights” (VRRs) that occasionallyprovided for a higher payoff to bondholdersin the event that GDP (or GDP per capita) of

Box 3.4. The Case for Growth-Indexed Bonds

Note: The main authors of this box are EduardoBorensztein and Paolo Mauro.

HOW CAN HIGH PUBLIC DEBT LEVELS BE REDUCED?

75

the debtor country rose above a certain level.Precedents of countries that have used elementsof GDP indexation in debt restructuringsinclude Bosnia and Herzegovina, Bulgaria, andCosta Rica. For example, a portion of Bulgaria’sBrady bonds provided for a GDP “kicker” suchthat, once real GDP exceeded 125 percent of its1993 level, creditors would be entitled to anadditional 0.5 percent in interest for every 1 per-cent of real GDP growth in the year prior tointerest payment. A more recent precedent is abond issued by the City of Buenos Aires in 2003as part of a debt restructuring operation, whichincludes indexation of principal repayments tothe city’s tax revenues.

Still, creating markets for new financial instru-ments is by no means easy, and there are manypractical and conceptual difficulties with theimplementation of a growth-indexed bond mar-ket. Would investors find growth-indexed bondstoo complicated or risky? Could countries misre-port their growth rates or lose incentive to growrapidly?

International investors already invest heavilyin stocks of emerging market countries, whoseprices are much more volatile than the GDPgrowth rates of the same countries. Moreover,international investors are already highlyexposed to GDP risk under standard debt con-tracts, though implicitly, as growth slowdownscause drops in bond prices, and may evenprompt defaults. While some investors may beturned away by instruments that are difficultto understand and price, the indexationmechanism is not alien to financial markets, asinflation-indexed bonds are well established inseveral sovereign debt markets in both advancedand developing countries. In addition, as notedabove, a few emerging market countries havealready issued bonds that include payment con-ditions that are contingent on GDP.

If growth-indexed bonds were to constitute alarge fraction of a country’s external debt, thatcountry’s authorities might be tempted tounderstate its growth rate or even reduce thegrowth orientation of its policies. How strongthat temptation would be, and whether it could

be resisted, are open questions, but they mightreasonably make some investors reluctant tohold growth-indexed bonds. Nevertheless, oneshould note that it is high growth rather thanlow growth that typically gets politiciansreelected. Data revisions could present anotherobstacle for investors. The bond contractswould therefore need to establish a clearmethod for dealing with revisions: for example,coupon payments for each date x could bebased on GDP as estimated on date y, ignoringany subsequent data revision. Of course, thecontract should specify that, for the purposesof the bond payment, the methodology usedto estimate GDP data cannot be changed inmidcourse.

Beneficial financial innovation is often hin-dered by the need to coordinate the actions ofmany potential market participants before anew instrument can be launched. As a result,historically, innovations in sovereign borrowinghave been limited and have often seemed toemerge from a combination of historical acci-dent, special circumstances, and strong officialintervention. For example, a forceful case hasbeen made for the introduction of inflation-indexed bonds by distinguished economistssince the nineteenth century. Yet, inflation-indexed bonds represent a large share of debtin only a few countries. Moreover, the timingof introduction and the popularity of thesebonds present no obvious regularities in termsof economic circumstances such as inflationhistory or the level of development. Similarly,financial flows to emerging markets switchedfrom the bond format—the historical norm—to syndicated bank loans in the 1970s, and backto bonds in the 1990s largely on account ofofficial encouragement and guarantees, reg-ulations in financial markets, and the interna-tional financing needs that originated fromthe oil price shock in 1973. The episodes offinancial turbulence that prompted the effortsunder way to rethink the international financialarchitecture might thus also help bring about anew era of innovation in sovereign financeinstruments.

period was still above the level three years priorto the event were eliminated. This selectionprocess highlighted 26 debt reduction episodesin the emerging market economies in thesample.39

A large majority (19 out of 26) of theseepisodes were associated with a debt default.While it is not possible to identify the exactimpact that the restructuring had on the out-standing debt, it appears to have generally beenan important factor behind the decline in thedebt ratio. The seven remaining episodes (whichtook place in five different countries) were thenexamined to understand the principal factorsbehind the debt reductions that have notinvolved a restructuring.40 In these seven cases,the median decline in the public sector debtratio was 34 percent of GDP over the three-yearperiod (Figure 3.14). Strong growth appears tohave been a significant contributing factor to thedecline in the debt ratio, with real GDP growthaveraging 8.5 percent a year. Fiscal consolidationplayed an important role as well, with a signifi-cant improvement in the primary balance begin-ning immediately before the debt began to fall.The fiscal consolidation was largely the result ofexpenditure restraint—with current expenditurebeing reduced and capital spending remainingconstant—although the revenue ratio alsoincreased somewhat. Moderate inflation ofabout 5 percent also helped, while exchangerate appreciation acted to reduce outstandingexternal public debt.

This analysis suggests that while large debtreductions have often occurred in conjunctionwith debt defaults, there are cases where theyhave been brought about by a combination ofstrong economic growth and fiscal consolida-tion. Interestingly, in all five of the countrieswhere debt was reduced without a restructuring,the public debt ratio is still below the level at the


76

4

6

8

10

12

14Public Investment

Figure 3.14. How Do Emerging Market Countries Reduce their Public Debt?(Percent of GDP)

10

20

30

40

50

60

70

20

40

60

80

100

120

140 Total Public Debt

Lower quartile

External Public Debt

Median

Upper quartile

Upper quartile

Median

10

20

30

40

50

60

70

-4

-2

0

2

4

6

8 Primary Balance

Lower quartile

Median

Upper quartileUpper quartile

Median

Lower quartile

10

20

30

40

50 Current Expenditure

Lower quartile

Median

Upper quartileUpper quartile

Median

Lower quartile

Strong growth and fiscal consolidation were key factors behind the decline in the debt-to-GDP ratio. Fiscal consolidation was largely achieved through a reduction in the ratio of current expenditures to GDP.

Lower quartile

Total Expenditure

Source: IMF staff estimates. Only includes debt reduction episodes that do not involve debt defaulter countries.

0

5

10

15

20

25

30

0

2

4

6

8

10

12 Output Growth

Lower quartile

MedianUpper quartile Upper quartile

MedianLower quartile

Inflation

1

1

t-3 t-2 t-1 t t+1 t+2 t+3 t-3 t-2 t-1 t t+1 t+2 t+3

t-3 t-2 t-1 t t+1 t+2 t+3 t-3 t-2 t-1 t t+1 t+2 t+3

t-3 t-2 t-1 t t+1 t+2 t+3 t-3 t-2 t-1 t t+1 t+2 t+3

t-3 t-2 t-1 t t+1 t+2 t+3 t-3 t-2 t-1 t t+1 t+2 t+339This exercise is roughly parallel to the analysis of

major reductions in external debt in Reinhart, Rogoff,and Savastano (2003).

40These occurred in Hungary, Israel (twice), Korea,Malaysia (twice), and Thailand.

beginning of the identified debt reductionepisode (although in the Asian countries, theratio has again risen in recent years followingthe financial crisis in the region). The outcomeis more mixed in the cases where debt reductionwas associated with a default. While in 10 ofthese countries debt has remained below thelevel prevailing at the beginning of the debtreduction episode, in 5 cases the country haseither defaulted again and/or debt is currentlyabove the level at the beginning of the debtreduction episode. This suggests that defaultdoes not always provide a long-term solution topublic debt problems, and that, unless it isaccompanied by complementary changes in fis-cal and other economic policies, it will not besuccessful in fostering sustainably lower debtlevels.

Whether it is achieved with or without a debtrestructuring, a substantial and sustained reduc-tion in public sector debt requires the imple-mentation of sound economic and fiscal policiesover a number of years. For example, Chile hasimplemented strong and sustained fiscal (andother economic) reforms since it defaulted onits external public debt in the 1980s, and thegovernment has reduced its debt from 54 per-cent of GDP in 1990 to 21 percent of GDP in2002. Several elements have contributed to thissuccessful adjustment, including expenditurerestraint, improved revenue collection, and stateenterprise reform that transformed losses intosignificant profit transfers to the government.Privatization proceeds have also been used toreduce debt, and real exchange rate apprecia-tion has reduced external debt in relation toGDP. Chile did not impose specific rules for thefiscal balance, but other institutional factorsplayed useful roles in maintaining fiscal disci-pline, including giving more power to thefinance ministry than to other ministries or thelegislature; prohibiting the central bank fromextending credit to the government; and pre-venting lower levels of government from borrow-ing. Since 2001, the government has committedto an annual target—a surplus of 1 percent ofGDP—for the central government structural bal-

ance (adjusted for cyclical effects and copperprice movements), thus allowing automatic stabi-lizers to work.

The benefits of these sustained policy actionsare clear. The financial markets have confidencein Chile’s fiscal policies, and spreads on govern-ment debt are well below those of other govern-ments in the region. Further, uninterruptedaccess to the capital markets has enabled theChilean government to avoid the forced pro-cyclical fiscal policies seen in other countries inthe region, reinforcing confidence in its eco-nomic management.

A number of other countries have also madeprogress in reducing high levels of public debt.In Hungary, public debt has fallen from about85 percent of GDP in the mid-1990s to less than60 percent now as a result of strong growth, aperiod of sustained primary budget surpluses(which, however, ended in 2002), and the pro-ceeds from the sale of government assets.Bulgaria has reduced its public debt from about160 percent of GDP in the early 1990s to lessthan 60 percent of GDP in 2002 as a result ofdebt restructuring, a fiscal consolidation pro-gram that has seen primary budget surpluses sus-tained since 1994, and high inflation (up to1997). Lastly, in Mexico, public debt wasreduced in the early 1990s as the countryemerged from its Brady debt restructuring.Despite the Tequila crisis in 1995, which entaileda costly restructuring of the banking system,debt is currently about 50 percent of GDP, andthe last of Mexico’s Brady debt has recently beenrepaid.

ConclusionsHigh public debt is a cause for concern in

many emerging market economies. At about 70percent of GDP, the average public debt ratio inemerging market economies now exceeds that inindustrial countries. Not only does this highlevel of public debt raise the risk of a fiscal crisisin some countries, but it also imposes costs onthe economy by keeping borrowing costs high,discouraging private investment, and constrain-

CONCLUSIONS

77

ing the flexibility of fiscal policy. Lower publicdebt levels would likely enable governments inemerging markets to run a more countercyclicalfiscal policy, with benefits for economic stability.

The analysis in this chapter suggests that, his-torically, many emerging market economies havenot generated large enough primary budget sur-pluses to ensure the sustainability of their publicdebt. This stands in contrast to industrial coun-tries. The inability to generate adequate primarysurpluses appears to stem from the characteris-tics of the fiscal systems: governments in emerg-ing market countries generally have weakrevenue bases (with lower yields and highervolatility) and are less effective at controllingexpenditures during economic upswings (this isparticularly the case in Latin America).

While the sustainable level of public debtvaries between countries—depending on thecharacteristics of each country—for the typicalemerging market economy it is often quite low.For example, the analysis of overborrowing sug-gested that, based on past fiscal performance,the sustainable public debt level for a typicalemerging market economy may only be about 25percent of GDP, while the estimates of the fiscalpolicy reaction functions indicated that emerg-ing market economies as a group have failed inthe past to respond in a manner consistent withensuring fiscal sustainability once public debtexceeds 50 percent of GDP.41 There are, how-ever, regional differences, with Asian countriesgenerally doing more to ensure debt sustainabil-ity than countries in other regions.

What can policymakers do to reduce publicdebt and cushion themselves against the risksthat high debt presents? It is important to recog-nize that the past does not necessarily conditionthe future—policies and institutions do change.The example of Chile, in particular, shows thatstrong fiscal and structural policy reforms—sometimes in combination with an initial debt

restructuring—can be effective in putting publicdebt on a firm and lasting downward path. To besuccessful, however, a broad and sustained pack-age of reforms is needed that encompasses thefollowing.• Tax and expenditure reforms. Reforms to

strengthen and broaden the tax base areneeded so that governments have access tohigher and less variable revenues. Effective taxrates in emerging market economies are gen-erally low, suggesting that tax avoidance—through either legal or illegal means—andweak tax administration are serious issues thatneed to be addressed. The continued relianceon taxes and transfers related to commodityexports is a weakness of many current tax sys-tems, and efforts are needed to broaden thetax base to reduce its variability. Better controlof expenditures during economic upswings isalso essential to ensure that periods of strongrevenue growth result in higher primary sur-pluses rather than increased spending.

• Steps to improve the credibility of fiscal policy.Governments need to be able to demonstratethat their overall debt burden is manageable,and that it is likely to remain so under mostcircumstances. Building this credibilityrequires not only the implementation of effec-tive fiscal reforms, but also a record of adher-ing to these reforms through upturns anddownturns. The strengthening of fiscal institu-tions has a very important role to play in thisregard. Fiscal rules—broadly defined as a per-manent constraint on fiscal performance—insome cases may play a useful role in strength-ening fiscal policy credibility if appropriatelydesigned and obeyed. For example, the FiscalResponsibility Law introduced in Brazil in2000—which established policy rules consist-ing of limits and targets for selected fiscal indi-cators for all levels of government, includingdebt ceilings and transparency requirements—


78

41These thresholds are not dissimilar from those found in recent studies on external debt crises in emerging markets.For example, IMF (2002) estimates a threshold of 40 percent of GDP, Manasse, Roubini, and Schimmelpfennig (2003) esti-mate a threshold of 50 percent of GDP, and Reinhart, Rogoff, and Savastano (2003) derive country-specific thresholds inthe range of 15–20 percent of GDP for countries that have repeatedly defaulted on their sovereign debt.

appears to have helped strengthen the govern-ment’s credibility in financial markets.42

Poland has also introduced a constitutionallimit on public debt of 60 percent of GDP(including the risk-weighted stock of outstand-ing government guarantees) and correctiveprocedures that kick in when public debtexceeds 50 percent of GDP.

• Steps to reduce exposure to exchange rate and inter-est rate movements. Given the structure of theirpublic debt, many emerging marketeconomies are exposed to considerable inter-est rate and foreign exchange risk. Steps areneeded to reduce the reliance on domesticallyissued foreign currency and short-term debt.Policies to promote more open economieswould help reduce the risks from externaldebt as exchange rate depreciations wouldthen provide more of a boost to exports andgovernment revenues to mitigate the impacton the budget of higher debt servicing costs.Recent proposals to create GDP-linked bondscould also provide some cushion during timesof economic stress.

• Structural reforms to boost growth prospects.Historic experience suggests that it is difficultto bring public debt ratios down withoutrobust economic growth. In this context, theimplementation of a broad-based agenda ofstructural reforms is a crucial complement tofiscal consolidation efforts. As emphasized inthe April 2003 World Economic Outlook, thestrengthening of institutions could beexpected to provide a significant boost togrowth over the medium term. Addressingcorporate and financial sector weaknesses willalso be a key, while further steps to liberalizetrade and promote long-term foreign invest-ment will have lasting growth benefits.

• Addressing the risks from contingent and implicitliabilities. It is also important that governmentsact to minimize the risks they face from con-tingent and implicit liabilities. This applies not

only to countries trying to reduce high debtlevels, but also to those that currently have rel-atively low debt. The experience of manycountries in recent years has shown that therecognition of such liabilities can significantlyadd to public debt and quickly raise questionsabout sustainability. The recapitalization ofbanking systems, in particular, has provedcostly, while government guarantees on privatesector projects are a further source of risk.Governments need to be fully aware of thecontingent and implicit liabilities they face—in this regard, improving fiscal transparencywould help—and act to reduce them to theextent possible. Improving financial sectorsupervision is an essential step toward thisgoal.More generally, the mechanisms for the

restructuring of sovereign debt also need to bestrengthened. Defaults on external public debthave been common among emerging marketeconomies, and certainly cannot be ruled out inthe future. It is therefore important that mecha-nisms are in place to deal with such events in anorderly manner to minimize, to the extent possi-ble, the costs and disruptions to all the involvedparties. To this end, current efforts to promotethe inclusion of collective action clauses in debtcontracts and, more generally, to find ways toimprove arrangements for sovereign debtrestructuring within the existing legal frameworkare important.

Appendix 3.1. Assessing FiscalSustainability: Data andEconometric Methods

The main authors of this appendix are MarcoTerrones and Xavier Debrun.

This appendix provides further details on thedata and the econometric methodology andresults discussed in the main text.

APPENDIX 3.1. ASSESSING FISCAL SUSTAINABILITY: DATA AND ECONOMETRIC METHODS

79

42Kopits (2001) contains a detailed discussion of fiscal policy rules. For more detail on fiscal policy rules in Brazil, seeGoldfajn and Guardia (forthcoming).

Data Issues

The main issues related to the two data setsused in the chapter have been discussed in Box3.1. The emerging market and other developingcountries in the data sets are as follows.

Data set 1: 1990–2002: Argentina, Brazil,Bulgaria, Chile, China, Colombia, Costa Rica,Côte d’Ivoire, Croatia, Ecuador, Egypt, Hungary,India, Indonesia, Israel, Jordan, Korea, Lebanon,Malaysia, Mexico, Morocco, Nigeria, Pakistan,Panama, Peru, the Philippines, Poland, Russia,South Africa, Thailand, Turkey, Ukraine,Uruguay, and Venezuela.

Data set 2: 1970–2002: Algeria, Argentina,Bangladesh, Benin, Bolivia, Botswana, Brazil,Burkina Faso, Burundi, Cameroon, Chile,China, Colombia, Costa Rica, Côte d’Ivoire,Czech Republic, Dominican Republic, Ecuador,Egypt, El Salvador, Gabon, Ghana, Guatemala,Haiti, Honduras, Hungary, India, Indonesia,Israel, Jamaica, Jordan, Korea, Malaysia,Mauritius, Mexico, Morocco, Niger, Nigeria,Pakistan, Papua New Guinea, Panama,Paraguay, Peru, the Philippines, Poland, Russia,Singapore, South Africa, Sri Lanka, Syrian ArabRepublic, Tanzania, Togo, Thailand, Tunisia,Turkey, Ukraine, Uruguay, Venezuela, andZimbabwe.

The industrial economies common to bothdata sets are Australia, Austria, Belgium, Canada,Denmark, Finland, France, Germany, Greece,Ireland, Italy, Japan, the Netherlands, NewZealand, Norway, Portugal, Spain, Sweden, theUnited Kingdom, and the United States.

Estimating Fiscal Policy Reaction Functions

A major issue in the specification of fiscalreaction functions is the choice of the policyinstrument or operational target. Given thischapter’s focus, it is important to choose a policyvariable directly related to debt dynamics. Thebudget identity (1) indicates that the stock ofpublic debt at the beginning of period t + 1(Bt+1) results from the inherited debt, Bt , towhich the period t financing requirements (theoverall balance) Ft is added:

Bt+1 = Bt + Ft . (1)

Since Ft depends on interest payments (inprinciple not a choice variable of the fiscalauthorities), and thereby on Bt itself (whichreflects past policies), interest payments can beseparated from other expenditures, and theidentity rewritten as:

Bt+1 = (1 + r)Bt + St – Rt = (1 + rt)Bt – Pt . (2)

where St is noninterest (primary) governmentspending, Rt is total government revenues, and ris the interest rate paid on existing debt. Thevariable Pt ≡ Rt – St is the primary balance (sur-plus). To account for the effect of growth onborrowing capacity, (2) can be rewritten in termsof ratios to GDP (denoted by lowercase letters):

Bt+1Yt+1 Bt Pt–––––– = (1 + r)–– – –– Yt+1 Yt Yt Yt

(1 + g)bt+1 = (1 + r)bt – pt, (3)

where Yt is the level of GDP and g is the nominalgrowth rate. From equation (3), the primary bal-ance that stabilizes the debt ratio (that is, bt+1) isgiven by p–t = bt (r – g), where r and g can also bemeasured in real terms as the effect of inflationdisappears with the use of ratios. Since the realinterest rate is generally higher than real growth,the primary surplus consistent with a constantdebt-to-GDP ratio increases with the initial debtstock and the difference between the real inter-est rate and the real growth rate.

With these identities in mind, and in line withBohn (1998) and other studies, the primary bal-ance is used as the operational target in the fis-cal reaction function:

pi,t = αi + ∑J

j=1βjXj,i,t + ρbi,t–1 + εi,t , (4)

where pi,t is the primary balance in country i attime t; αi is a country-specific intercept (fixedeffect) accounting for heterogeneity in thegroup of countries under consideration; bi,t–1 isthe debt level at the end of the previous period;εi,t is an error term; and Xj is a vector of macro-economic variables explaining changes in theprimary balance unrelated to the long-run sol-


80

vency requirement. In the spirit of Barro’s(1979) “tax smoothing” theory, these variablesreflect transitory shocks to expenditure andrevenues, such as business cycle fluctuationsand exceptional events such as wars or naturaldisasters. That conjecture conveniently limitsthe number of potential explanatory variableswhile remaining consistent with a well-specifiedtheory of fiscal policy.43 Finally, as discussed inthe main text, the connection between currentpolicy actions and long-run solvency lies in theassumption that the primary balance systemati-cally responds to past changes in the publicdebt, an aspect captured by the coefficientρ in equation (4). While Bohn (1998) demon-strates that a positive value for ρ is sufficient toensure long-run solvency under very weaktechnical assumptions, it is interesting to seehow the dynamics of the debt-to-GDP ratio areaffected by assuming (as in equation (4)) thatpt = ρbt + xt, where xt summarizes the determi-nants of the primary surplus unrelated to debtsustainability concerns. Equation (3) can thenbe rewritten as

1 + r – ρ xtbt+1 – bt = –[1 – (––––––– )]bt – –––––. (5)1 + g 1 + g

Assuming that xt is “stationary” (which in prac-tice excludes a downward trend in the non-debt-related surplus), the sign of the term insquare brackets determines whether the debtratio is mean reverting, in the sense of converg-ing toward some finite level pinned down by theaverage of xt. A positive sign implies mean rever-sion and will be observed if r – ρ < g. Hence, ρcan be interpreted as the largest differencebetween the real interest rate and real growththat remains consistent with a mean-revertingdebt ratio.

Equation (4) was separately estimated for pan-els of emerging market and industrial economiesfor the period 1990–2002. Four transitory deter-minants of fiscal policy were incorporated, with

the latter two only being included for emergingmarkets: the output gap, defined as the relativedeviation of real GDP from its Hodrick-Prescott(HP)-filtered trend (to capture the impact of thebusiness cycle on the primary balance); the CPIinflation rate (to account for shocks to seignior-age revenues); an indicator to capture the yearsin which a country experienced a debt default orrestructuring (to account for the lack of financ-ing that generally accompanies such situations);and the deviation of oil and non-oil commodityprice cycles from their respective HP-filteredtrends (to capture the direct effect of commodityprice swings on government revenues in com-modity exporting countries).

Commodity price data (oil price index and anindex of food and metal commodities) are fromthe Commodity Price System database. All otherdata, including budgetary series in industrialcountries, come from the World EconomicOutlook database. For the econometric exercise,countries in a state of war or an extended periodof default or restructuring during the sampleperiod were eliminated on the grounds thatunusual events were affecting the primary surplus.Countries in which there were clear trends in thefiscal series were also excluded.44 The resultsreported in the main text refer to this restrictedsample. As confirmed by the last column in Table3.1, the exact country composition of the sampledoes not qualitatively affect the results for emerg-ing market countries, although there is now astatistically significant positive, but very weak, reac-tion of the primary surplus to debt. For industrialcountries, however, the inclusion of Japan in thesample does mean that the 80 percent of GDPthreshold reported in Table 3.2 (and discussedbelow) could no longer be found.

Tables 3.1 and 3.2 present different variants ofequation (4) for emerging market and industrialeconomies respectively. A key dimension of theempirical investigation was to capture statisticallythe nonlinear relationship between debt and


81

43See Favero (2002), Galí and Perotti (2003), and Fatàs and Mihov (forthcoming) for detailed discussions of the issuesrelated to the specification of fiscal reaction functions.

44Of course, the short time-series dimension of the panel prevents formal stationarity tests.

the primary balance suggested by Figure 3.9.Although various powers (quadratic and cubic)of the debt variable were added to the basicspecification (4) to test the statistical significanceof a nonlinear relationship, spline regressions—that is, models allowing for a kink in the regres-sion line—were found to fit the data very well,suggesting that the reaction of the primary sur-plus to the debt was indeed contingent on thedebt level itself. A variety of debt thresholds weretested (from 30 percent of GDP to 90 percent ofGDP by increments of 5 percentage points). Foremerging market economies, the 50 percent ofGDP threshold provided the best fit to the data(highest adjusted R2), and this was also the low-est debt-to-GDP ratio beyond which the positivedebt feedback effect disappeared (in the senseof not being statistically different from zero).

For industrial countries, spline regressions con-firmed the visual impression from Figure 3.9,and indicated that the debt feedback effect wasstatistically larger when the debt-to-GDP ratiowas above 80 percent.

In addition to the discussion in the main text,it is worth noting that all transitory determinantsof fiscal policy have the expected signs and aregenerally highly significant in the regressions.Interestingly, whereas higher inflation is gener-ally associated with a larger primary surplus inemerging market economies (in line with theeffect on seigniorage revenues), inflationappears to have a strongly negative effect on thesurplus in industrial economies, perhaps reflect-ing contemporaneous efforts to reduce inflationand to adjust the fiscal balance in a number ofcountries in the 1990s.


82

Table 3.1. Emerging Market Economies: Fiscal Policy Reaction Functions, 1990–2002(Dependent variable: primary surplus, percent of GDP)

FullExplanatory Variables No Control Controls Regional Spline Openness Institutions Sample

Output gap (YG) . . . 0.141*** . . . 0.134*** 0.057 –0.098 0.128***Total public debt (TD)-lagged 0.039*** 0.047*** . . . 0.123*** 0.134*** 0.132*** 0.091***

ControlsInflation . . . 0.001*** 0.001*** 0.001*** 0.001*** 0.001*** 0.001***Oil price cycles (if oil producer) . . . 0.054** 0.050*** 0.071*** 0.071*** 0.071*** 0.088***Non-oil commodity price cyles

(if commodity producer) . . . 0.069 0.067 0.078* 0.081* 0.066 0.105**Default/restructuring . . . 0.715*** 0.541*** 0.654*** 0.493*** 0.686*** 0.822***

Nonlinearities and interactionsInteraction of YG with:

Trade openness . . . . . . . . . . . . 0.131** . . . . . .Institutional quality . . . . . . . . . . . . . . . 0.078*** . . .Regional dummies . . .

Latin American . . . . . . 0.044* . . . . . . . . . . . .Non–Latin American . . . . . . 0.235*** . . . . . . . . . . . .

Interaction of TD with: Trade openness . . . . . . . . . . . . –0.018** . . . . . .Institutional quality . . . . . . . . . . . . . . . –0.003** . . .Regional dummies

Latin American . . . . . . 0.041*** . . . . . . . . . . . .Non–Latin American . . . . . . 0.051*** . . . . . . . . . . . .

NonlinearitiesSpline regression coefficient

Break at 50 percent . . . . . . . . . –0.115*** –0.113*** –0.111*** –0.072***Test if slope equal to 0 (Wald χ2) . . . . . . . . . 1.641 . . . . . . 8.601***

Adjusted R 2 0.497 0.578 0.562 0.630 0.578 0.623 0.682Number of observations 249 249 249 249 249 249 362

Note: All equations have been estimated with Generalized Least Squares allowing for fixed effects and using a heteroscedasticity-consistentvariance-covariance matrix for statistical tests. The symbols *, **, and *** indicate that the estimated coefficient is significantly different fromzero at the 10, 5, and 1 percent level, respectively. Except for the full sample column, the following countries have been excluded for one of thereasons explained in the text: Bulgaria, China, Côte d’Ivoire, Croatia, Egypt, India, Israel, Jordan, Lebanon, Nigeria, and Pakistan.

The impact of broader economic and institu-tional factors on fiscal policy in emerging mar-ket economies was also investigated. Two pointsemerged from the econometric analysis (seeTable 3.1).• First, trade openness and institutional quality

(defined as a combination of lower corruption,better bureaucracy, greater democraticaccountability and a more effective rule of lawaccording to indicators from InternationalCountry Risk Guide) tend to be associated with amuch more countercyclical response of fiscalpolicy. The effect of trade openness may partlyreflect the generally bigger size of governmentsin more open economies and the consequentlylarger automatic stabilizers (Rodrik, 1998).The effect of institutions is consistent with the

conjecture that good institutions are associatedwith a better ability to raise revenues, more fis-cal policy credibility, and correspondinglylooser resource constraint, allowing the gov-ernment to run more countercyclical policies.

• Second, governments in open emerging mar-ket economies tend to behave more like indus-trial countries as they post higher averageprimary surpluses and react less strongly todebt sustainability concerns at low debt levels.Countries with better institutions have on aver-age lower public debt, which explains whytheir governments need to react less to debtsustainability concerns.Finally, Table 3.3 shows strong evidence of a

procyclical expenditure policy in Latin America,in contrast to other emerging markets and, even


83

Table 3.2. Industrial Economies: Fiscal Policy Reaction Functions, 1990–2002(Dependent variable: primary surplus, percent of GDP)

Explanatory Variables No Control Controls Spline Full Sample

Output gap (YG) . . . 0.971*** 0.960*** 0.961***Total public debt (TD) – lagged 0.057*** 0.060*** 0.045*** 0.039***

Inflation . . . –0.407*** –0.349*** –0.416***

Spline regression coefficient . . . . . . . . . . . .Break at 80 percent . . . . . . 0.086*** –0.042

Adjusted R2 0.367 0.691 0.702 0.621Number of observations 191 191 191 227

Note: All equations have been estimated with Generalized Least Squares allowing for fixed effects and using a heteroscedasticity-consistentvariance-covariance matrix for statistical tests. The symbols *, **, and *** indicate that the estimated coefficient is significantly different fromzero at the 10, 5, and 1 percent level, respectively. Except for the full sample column, the following countries have been excluded for one of thereasons explained in the text: Germany (effect of reunification), Norway (effect of oil revenues), and Japan (increasing primary deficits).

Table 3.3. Expenditure Equations, 1990–2002(Dependent variable: Primary expenditure, percent of GDP)

Emerging Markets Industrial Countries__________________________________ ______________________________________________Explanatory Variables Restricted sample Full sample Restricted sample Full sample

Output gap . . . . . . . . . –0.968*** –0.318*** –0.953*** –0.331***Latin America 0.156*** 0.121*** 0.122*** . . . . . . . . . . . .Emerging markets excluding

Latin America –0.231*** –0.113*** –0.064*** . . . . . . . . . . . .Inflation –0.003*** –0.002** –0.001*** 0.149** 0.023 0.196*** 0.039Total public debt lagged 0.002 –0.012** –0.005 –0.006 –0.056*** 0.015** –0.039***Default/restructuring –0.851*** –0.718*** –0.711*** . . . . . . . . . . . .Lagged dependent variable . . . 0.428*** 0.535*** . . . 0.769*** . . . 0.759***

Adjusted R 2 0.926 0.939 0.940 0.928 0.972 0.925 0.970Number of observations 242 242 350 191 191 227 227

Note: All equations have been estimated with Generalized Least Squares allowing for fixed effects and using a heteroscedasticity-consistentvariance-covariance matrix for statistical tests. The symbols *, **, and *** indicate that the estimated coefficient is significantly different fromzero at the 10, 5, and 1 percent level, respectively.

more so, to industrial economies, in which pri-mary expenditures have a stabilizing influence onthe business cycle. The results also suggest thatindustrial countries react to debt accumulationthrough a contraction in expenditure that appearsstronger than in emerging market economies.

Determinants of Overborrowing

Following the literature on the determinantsof debt default and fiscal crises, the role of acountry’s economic and institutional structureon public sector overborrowing was investi-gated.45 The analysis assumes a linear cross-section regression model of the form

y = α + Xβ + u (6)

where y is a (n × 1) vector of the overborrowingratio; X is a (n × K) matrix of economic and

institutional characteristics; β is a (K × 1) vectorof parameters; and n is the number of countriesin the sample. The model was estimated usingboth ordinary least squares (OLS) and instru-mental variables (IV) techniques.

As discussed in the main text, the overborrow-ing ratio is measured as the ratio of actual publicdebt to the benchmark level of debt (both as per-cent of GDP). The benchmark debt level is calcu-lated as the present discounted value of futureprimary balances (here proxied by the averagehistorical primary balance), with the discount fac-tor being the difference between the average realinterest rate (measured by the real LIBOR rateplus a spread proxied from a country’sInstitutional Investor rating) and the average realGDP growth rate.46 It was possible to calculatethe benchmark debt-to-GDP ratio for 50 coun-tries (14 industrialized countries and 36 develop-


84

Table 3.4. Overborrowing and Institutions: Bivariate Regression Results

Regression_____________________________________________________________________________________(1) (2) (3) (4) (5) (6)

Corruption –0.75(0.18)*** . . . . . . . . . . . . . . .

Property rights –0.69. . . (0.20)*** . . . . . . . . . . . .

Constraint on the executive –0.15. . . . . . (0.06)** . . . . . . . . .

Quality of bureaucracy –0.79. . . . . . . . . (0.18)*** . . . . . .

Democratic accountability –0.55. . . . . . . . . . . . (0.16)*** . . .

Law and order –0.84. . . . . . . . . . . . . . . (0.22)***

Constant 1.1 3.2 1.55 2.91 3.21 1.18(0.26)*** (0.80)*** (0.33)*** (0.60)*** (0.80)*** (0.28)***

R2 0.18 0.18 0.07 0.19 0.13 0.17Number of observations 50 46 50 49 49 50

Notes: Robust standard errors are in brackets. The symbols *, **, and *** indicate statistical significance at 10, 5, and 1 percent levels,respectively. The dependent variable is the logarithm of the ratio of overborrowing. For definitions and sources of the three first measures ofinstitutions, see Appendix 3.1 of the April 2003 World Economic Outlook. For definitions and sources of the three last measures of institutionssee the International Country Risk Guide.

45See, for instance, Manasse, Roubini, and Schimmelpfennig (2003).46The benchmark debt-to-GDP ratio, ν, is calculated using the following present value formula:

pν = ∫

∞

0

psexp – {(r – g)s}ds ≈ –––– ,r – g

where p is the average primary balance as percent of GDP, r is the real interest rate, g is output growth, and s is time. Theaverage primary balance is used to predict future fiscal policies under the premise of no major change in policies. Ofcourse, if there is a structural break in the conduct of fiscal policy, this assumption may not be valid. Data on spreads arenot available for all countries or all years in the sample. Therefore, the spread was proxied by (100 – the Institutional

ing countries, of which 21 were emerging marketeconomies) using data for 1985–2002.47

The following economic factors were consid-ered in the analysis: trade openness, the ratio ofgovernment revenue to GDP, economic volatility,and relative (to the United States) income percapita. Trade openness is measured as the aver-age of the foreign exchange restrictions for cur-rent account transactions as compiled by theAnnual Report on Exchange Arrangements andExchange Restrictions; government revenue is cal-culated as the average ratio of total governmentrevenue to nominal GDP, obtained fromGovernment Finance Statistics and InternationalFinance Statistics, respectively. Output volatility ismeasured as the standard deviation of thegrowth rate of real GDP, and relative income ismeasured as the ratio of the PPP-adjusted realper capita income of each country relative toUnited States using the WEO database.

A number of different measures of institu-tional quality were used in the analysis.48 Simplebivariate regressions suggest that measures ofinstitutional quality are inversely related tooverborrowing (Table 3.4). Two institutionalmeasures were found to be important in theregression analysis: an index of property rightsand a measure of corruption. The property rightsindex, obtained from Heritage Foundation’sIndex of Economic Freedom, measures theextent of protection of private property, whilethe measure of corruption is the freedom fromgraft index—see Kaufmann, Kray, and Zoido-Lobatón (1999)—which measures the extentpublic investiture is used for corruption or pri-vate benefit.

The main results are reported in Table 3.5. Inaddition to the variables reported in the main

text, the regression also included output vola-tility, relative income per capita, and an indus-trial country dummy. The first and the lastregressors were statistically significant. Oneconcern about the OLS results—reported in thefirst column—is that the revenue-to-GDP ratiomay be endogenously determined as the degreeof overborrowing could influence a govern-ment’s policy response, including its tax policy.Although the Wu-Hausman test does not revealstrong evidence of such endogeneity, in view of


85

Table 3.5. Determinants of Overborrowing

Ordinary IndepentLeast Squares Variable

Openness (trade restrictions) 1.64 1.65(0.59)*** (0.55)***

Government revenues –0.05 –0.04(percent of GDP) (0.02)** (0.02)**

Property rights (index) –0.62 –0.64(0.35)* (0.33)**

Volatility of growth rate of output 25.15 25.50(11.16)** (10.40)**

Relative (to the U.S.) –0.57 –0.62income per capita (1.35) (1.21)

Constant 2.21 2.11(1.15)* (1.02)**

R2 0.48 . . .Number of observations 46 46

Wu-Hausman F test . . . 0.29P-value . . . (0.59)

Sargan test . . . 0.77P-value . . . (0.86)

Notes: Robust standard errors are in brackets. The symbols *,**, and *** indicate statistical significance at 10, 5, and 1 percentlevels, respectively. The dependent variable is the logarithm of theratio of overborrowing. The regressions also included an industrialcountry dummy. The following variables were used as instruments:government revenues as percent of GDP for the 1970–85 period,average terms of trade growth, an index of corruption, and anemerging market country dummy. The null hypothesis under theWu-Hausman test is that the ratio of government revenues to GDPis exogenous. The null hypothesis under the Sargan test is that theinstruments are uncorrelated with the error term and correctlyexcluded from the estimated equation.

Investor (II) index); the II index is based on international banks’ risk assessment of individual countries and has a scale of0–100, with 100 representing the least chance of a debt default. Reinhart, Rogoff, and Savastano (2003) find a strong cor-relation between the Institutional Investor index—which is available for a large number of countries since 1979—andexternal debt, and between external debt and spreads. A simple regression on a subsample of countries where spreads dataare available confirms a strong positive relationship between actual spreads and those constructed from the II index.

47The calculations were not made for those countries where the average primary balance or the discount factor was neg-ative during the sample period. The ratios were calculated for 1985–2002 so that data for 1970–84 could be used to instru-ment variables in the regressions.

48Different measures of institutions are discussed in Appendix 3.1 of the April 2003 World Economic Outlook.

earlier results—whereby governments improvetheir primary balances in response to increaseddebt—this remains a concern. To deal with thispossibility, instrumental variables analysis wasemployed, with the following variables used asinstruments: the average ratio of governmentrevenues to GDP during 1970–1984; averageterms of trade for 1985–2002; the corruptionindex; and a Latin American country dummy.These instruments seem adequate, as theSargan test fails to reject the null hypothesisthat they are uncorrelated with the error termand correctly excluded from the estimatedequations. The results from the first-stageregression suggest that the ratio of governmentrevenues to GDP is strongly positively correlatedwith terms of trade growth and the freedomfrom graft index. In other words, high terms oftrade growth and less corruption have a positiveimpact on the ratio of revenues to GDP, whichin turn affects overborrowing.

ReferencesAlesina, Alberto, Roberto Perotti, and José Tavares,

1998, “The Political Economy of Fiscal Adjustment,”Brookings Papers on Economic Activity: 1, BrookingsInstitution, pp. 197–248.

Alesina, Alberto, Ricardo Hausmann, Rudolf Hommes,and Ernesto Stein, 1998, “Budget Institutions andFiscal Performance in Latin America,” IADBWorking Paper No. 394 (Washington: Inter-American Development Bank).

Aiyagari, Rao S., and Ellen R. McGrattan, 1998, “TheOptimum Quantity of Debt,” Journal of MonetaryEconomics, Vol. 42 (December), pp. 447–69.

Auerbach, Alan J., Laurence J. Kotlikoff, and WilliLeibfritz, eds., 1999, Generational AccountingAround the World (Chicago: University of ChicagoPress).

Barnhill Jr., Theodore M., and George Kopits, 2003,“Assessing Fiscal Sustainability Under Uncertainty,”IMF Working Paper 03/79 (Washington: Interna-tional Monetary Fund).

Barro, Robert J., 1979, “On the Determination ofPublic Debt,” Journal of Political Economy, Vol. 87(October), pp. 940–71.

Berg, Andrew, and Anne Krueger, 2003, “Trade,Growth, and Poverty: A Selective Survey,” IMF

Working Paper 03/30 (Washington: InternationalMonetary Fund).

Blanchard, Olivier J., 1990, “Suggestions for a New Setof Fiscal Indicators,” OECD Working Paper No. 79(Paris: Organization for Economic Cooperationand Development).

———, Jean-Claude Chouraqui, Robert P. Hagemann,and Nicola Sartor, 1990, “The Sustainability ofFiscal Policy: New Answers to an Old Question,”OECD Economic Studies, No. 15 (Autumn),pp. 7–36.

Bohn, Henning, 1998, “The Behavior of U.S. PublicDebt and Deficits,” Quarterly Journal of Economics,Vol. 113 (August), pp. 949–63.

Borensztein, Eduardo, and Paolo Mauro, 2002,“Reviving the Case for GDP-Indexed Bonds,” IMFPolicy Discussion Paper 02/10 (Washington:International Monetary Fund).

Braun, Miguel, and Luciano Di Grescia, 2003, “TowardEffective Social Insurance in Latin America: TheImportance of Countercyclical Fiscal Policy,” IADBWorking Paper No. 487 (Washington: Inter-American Development Bank).

Buiter, Willem H., 1985, “Guide to Public Sector Debtand Deficits,” Economic Policy: A European Forum,Vol. 1 (November), pp. 13–79.

Burnside, Craig, Martin Eichenbaum, and SergioRebelo, 2001, “Prospective Deficits and the AsianCurrency Crisis,” Journal of Political Economy, Vol. 109(December), pp. 1155–97.

Cashin, Paul, C. John McDermott, and Alasdair Scott,2002, “Booms and Slumps in World CommodityPrices,” Journal of Development Economics, Vol. 69(October), pp. 277–96.

Chalk, Nigel, and Richard Hemming, 2000, “AssessingFiscal Sustainability in Theory and Practice,” IMFWorking Paper 00/81 (Washington: InternationalMonetary Fund).

Debrun, Xavier, and Charles Wyplosz, 1999, “OnzeGouvernements et une Banque Centrale,” Revued’Economie Politique, Vol. 109 (May–June),pp. 387–424.

Economic Policy Committee, 2001, “BudgetaryChallenges Posed by Ageing Populations,”EPC/ECFIN/655/01-EN final (Brussels: EuropeanCommission, October).

European Commission, forthcoming, “Public Financesin EMU,” in European Economy, No. 3/2003.

Fatàs, Antonio, and Ilian Mihov, forthcoming, “TheCase for Restricting Fiscal Policy Discretion,”Quarterly Journal of Economics.


86

Favero, Carlo A., 2002, “How Do European Monetaryand Fiscal Authorities Behave?” CEPR DiscussionPaper No. 3426 (London: Center for EconomicPolicy Research).

Frederiksen, Niels, 2001, “Fiscal Sustainability in theOECD: A Simple Method and Some PreliminaryResults,” Working Paper No. 3/2001 (Copenhagen:Finansministeriet).

Galí, Jordi, and Roberto Perotti, 2003, “Fiscal Policyand Monetary Integration in Europe,” CEPRDiscussion Paper No. 3933 (London: Center forEconomic Policy Research).

Gavin, Michael, Ricardo Hausman, Roberto Perotti,and Ernesto Talvi, 1996, “Managing Fiscal Policy inLatin America and the Caribbean: Volatility,Procyclicality, and Limited Creditworthiness,” IADBWorking Paper No. 326 (Washington: Inter-American Development Bank).

Goldfajn, Ilan, and Eduardo Refinetti Guardia, forth-coming, “Fiscal Rules and Debt Sustainability inBrazil,” in Rules-Based Fiscal Policy in EmergingMarkets: Background, Analysis, and Prospects, ed. byGeorge Kopits (London: Palgrave).

Hemming, Richard, Michael Kell, and Selma Mahfouz,2002, “The Effectiveness of Fiscal Policy inStimulating Economic Activity—A Review of theLiterature,” IMF Working Paper 02/208(Washington: International Monetary Fund).

Hemming, Richard, Michael Kell, and AxelSchimmelpfennig, 2003, Fiscal Vulnerability andFinancial Crises in Emerging Market Economies, IMFOccasional Paper No. 218 (Washington:International Monetary Fund).

Hemming, Richard, and Teresa Ter-Minassian, 2003,“Public Debt Dynamics and Fiscal Adjustment,”Chapter 6 in Managing Financial Crises: RecentExperience and Lessons for Latin America, ed. byCharles Collyns and G. Russell Kincaid, IMFOccasional Paper No. 217 (Washington:International Monetary Fund), pp. 65–84.

International Monetary Fund, 2002, “Assessing Sus-tainability” (Washington). Available via the Internet:http://www.imf.org/external/np/pdr/sus/2002/eng/052802.htm.

———, 2003, “Sustainability—Review of Applicationand Methodological Refinements” (Washington).Available via the Internet: http://www.imf.org/external/np/pdr/sustain/2003/061003.htm.

Kaufmann, Daniel, Aart Kraay, and Pablo Zoido-Lobatón, 1999, “Aggregating GovernanceIndicators,” World Bank Policy Research

Working Paper No. 2195 (Washington: WorldBank).

Klingen, Christoph, Beatrice Weder, and JerominZettelmeyer, 2003, “How Private Creditors Fared inEmerging Debt Markets: 1970–2000” (unpublished;Washington: International Monetary Fund).

Kopits, George, 2001, “Fiscal Rules: Useful PolicyFramework or Unnecessary Ornament?” IMFWorking Paper 01/145 (Washington: InternationalMonetary Fund).

Kose, M. Ayhan, Eswar S. Prasad, and Marco E.Terrones, 2003, “Financial Integration andMacroeconomic Volatility,” IMF Staff Papers, Vol. 50(Special Issue), pp. 119–42.

Lindert, Peter, and Peter Morton, 1989, “HowSovereign Debt Has Worked,” in Developing CountryDebt and Economic Performance, ed. by Jeffrey Sachs(Chicago: University of Chicago Press for theNational Bureau of Economic Research),pp. 39–106.

Lucas, Robert E. Jr., 2003, “Macroeconomic Priorities,”American Economic Review, Vol. 93 (March),pp. 1–14.

Manasse, Paolo, Nouriel Roubini, and AxelSchimmelpfennig, 2003, “Predicting SovereignDebt Crises” (unpublished; Washington:International Monetary Fund).

Mélitz, Jacques, 1997, “Some Cross-Country Evidenceabout Debt, Deficits and the Behaviour of Monetaryand Fiscal Authorities,” CEPR Discussion Paper No.1653 (London: Center for Economic PolicyResearch).

Mendoza, Enrique G., and Pedro Marcelo Oviedo,2003, “Public Debt Sustainability under Uncer-tainty,” Research Department, IADB (unpublished;Washington: Inter-American Development Bank).

Mendoza, Enrique G., Assaf Razin, and Linda L. Tesar,1994, “Effective Tax Rates in Macroeconomics:Cross-Country Estimates of Tax Rates on FactorIncomes and Consumption,” Journal of MonetaryEconomics, Vol. 34 (December), pp. 297–323.

Obstfeld, Maurice, and Kenneth Rogoff, 1996,Foundations of International Macroeconomics(Cambridge, Massachusetts: MIT Press).

Özler, Sule, 1993, “Have Commercial Banks IgnoredHistory?” American Economic Review, Vol. 83 (June),pp. 608–20.

Pattillo, Catherine, Helene Poirson, and Luca Ricci,2002, “External Debt and Growth,” IMF WorkingPaper 02/69 (Washington: International MonetaryFund).

REFERENCES

87

Persson, Torsten, and Lars E.O. Svensson, 1989, “Whya Stubborn Conservative Would Run a Deficit:Policy with Time-Inconsistent Preferences,”Quarterly Journal of Economics, Vol. 104 (May),pp. 325–45.

Polackova Brixi, Hana, and Allen Schick, 2002,Government at Risk: Contingent Liabilities and FiscalRisk (Washington: World Bank).

Ramey, Garey, and Valerie A. Ramey, 1995, “Cross-Country Evidence on the Link Between Volatilityand Growth,” American Economic Review, Vol. 85(December), pp. 1138–51.

Reinhart, Carmen, Kenneth Rogoff, and MiguelSavastano, 2003, “Debt Intolerance,” in BrookingsPapers on Economic Activity: 1, Brookings Institution,pp. 1–62.

Rodrik, Dani, 1998, “Why Do More Open EconomiesHave Bigger Governments?” Journal of PoliticalEconomy, Vol. 106 (October), pp. 997–1032.

Rogoff, Kenneth, 1990, “Equilibrium Political BudgetCycles,” American Economic Review, Vol. 80 (March),pp. 21–36.

Rose, Andrew, 2002, “One Reason Countries Pay TheirDebts: Renegotiation and International Trade,”NBER Working Paper No. 8853 (Cambridge, Massa-chusetts: National Bureau of Economic Research).

Sachs, Jeffrey D., 1985, “External Debt and Macro-economic Performance in Latin America and East

Asia,” Brookings Papers on Economic Activity: 2,Brookings Institution, pp. 523–64.

———, and Andrew Warner, 1995, “Economic Reformand the Process of Global Integration,” BrookingsPapers on Economic Activity: 1, Brookings Institution,pp. 1–118.

Standard & Poor’s, 2002a, “Global Financial SystemsStress.”

———, 2002b, “Sovereign Defaults: Moving HigherAgain in 2003?”

Talvi, Ernesto, and Carlos Végh, 2000, “Tax BaseVariability and Procyclical Fiscal Policy,” NBERWorking Paper No. 7499 (Cambridge, Massachu-setts: National Bureau of Economic Research).

Turner, Dave, Claude Giorno, Alain De Serres, AnnVoutc’h, and Pete Richardson, 1998, “TheMacroeconomic Implications of Ageing in a GlobalContext,” OECD Economics Department WorkingPaper No. 193 (Paris: Organization for EconomicCooperation and Development).

von Hagen, Jurgen, 1992, “Budgeting Procedures andFiscal Performance in the European Communities,”Directorate General for Economic and FinancialAffairs, Economic Papers No. 96 (Brussels:European Commission), pp. 1–74.

———, and Ian Harden, 1995, “Budget Processes andCommitment to Fiscal Discipline,” EuropeanEconomic Review, Vol. 39 (April), pp. 771–79.


88

Documents

PUBLIC DEBT IN EMERGING MARKETS: CHAPTER III IS · PDF filethe suggestion that—despite the currently benign ... budgetary funds, ... cies between the debt and the fiscal flows data