Information about Western Hemisphere Hemisferio Occidental
Chapter

Chapter 4: A Framework for Assessing the Level of Public Debt

Author(s):
Marco Pinon, Alejandro Lopez Mejia, M. (Mario) Garza, and Fernando Delgado
Published Date:
July 2012
Share
  • ShareShare
Information about Western Hemisphere Hemisferio Occidental
Show Summary Details
Author(s)
Geoffrey J. Bannister and Luis-Diego Barrot 

The fiscal response to the 2008–09 global financial crisis led to an increase in the debt ratios of the economies in Central America, Panama, and the Dominican Republic (CAPDR),1 which raised concerns about debt sustainability and the available fiscal space to confront future shocks.2 In this context, a key question arises: What is the debt level that CAPDR countries could safely target going forward? Various approaches have been used to answer this question, including analyses of debt sustainability, of the impact of debt levels on growth and the external balance, and of the effectiveness of fiscal stimulus under different debt levels.

Building upon the debt intolerance literature, this chapter develops a revised approach to setting debt targets for the region. The measure of debt intolerance for CAPDR provides an indicator of debt safety that could be used as guidance for reducing fiscal and external risks. The chapter is organized as follows: the second section briefly reviews the literature on debt targets. The third section describes the features of the original debt intolerance approach. The fourth section presents the revised methodology for measuring debt intolerance. The fifth section applies the revised approach to CAPDR countries and develops an indicator of debt intolerance.

Background

Debt Targets in Theory

The theoretical literature provides little practical guidance for setting debt targets. General equilibrium models have been developed to explore the trade-off between the costs and benefits of government debt. Aiyagari and McGrattan (1998), Floden (2001), and Shin (2006), for example, illustrate the trade-off between the benefits of public debt to enhance the liquidity of households by providing an additional means to smooth consumption and effectively loosen their borrowing constraints, and the cost of future implied taxes that have adverse effects on wealth distribution and incentives, and the crowding out of private investment. Public debt thus improves the risk-sharing ability of households and provides a vehicle for diversifying idiosyncratic income risk by relaxing the borrowing constraint, but at the cost of allocative inefficiency and slower growth. Saint-Paul (2005) cites additional benefits of government debt markets for efficient financial markets, including financial innovation that comes from deep and liquid public debt markets, and the ability to use public debt as collateral for private sector borrowing. Few of these models have been implemented empirically. Aiyagari and McGrattan (1998) calibrate their model for the United States and calculate the optimal debt ratio at 66 percent of GDP; Weh-Sol (2010) uses the same model to calculate an optimal debt ratio for the Republic of Korea of 62 percent of GDP.

Debt Targets in Practice

More practical approaches to developing debt targets have relied on empirical evidence for the conduct of fiscal policy and the relationship between debt and growth (Table 4.1). IMF (2003) uses a debt sustainability approach to determine the level of debt to GDP that is compatible with a country’s past fiscal performance. If a country has been able to generate high primary surpluses in the past, then it should be able to tolerate a higher level of debt to GDP without encountering destabilizing or unsustainable debt dynamics. This conclusion amounts to assuming that past fiscal effort, as summarized by the average primary surplus, is the best guide to future fiscal effort. Under this scenario, assuming a historical differential between the real interest rate and the real growth rate, they determine that the average sustainable debt for emerging market countries is about 25 percent of GDP. A more recent analysis along similar lines for industrial countries is performed by Ostry and others (2010), but estimates a fiscal reaction function to summarize past behavior. They find that the debt limit above which debt dynamics become unsustainable ranges between 170 and 180 percent of GDP.

Table 4.1Empirical Studies of Debt Thresholds
Threshold (Percent of GDP)AuthorsApproachCountry coverage
25IMF (2003)Debt sustainability (historical primary surpluses)Emerging markets
25IMF (2008, 2009)Effectiveness of countercyclical fiscal policy conditional on initial level of debtEmerging markets
60-75IMF (2008, 2009)Effectiveness of countercyclical fiscal policy conditional on initial level of debtAdvanced economies
64Caner, Grennes, and Koehler-Geib (2010)Threshold least squares regressionDeveloping economies
77Caner, Grennes, and Koehler-Geib (2010)Threshold least squares regressionAdvanced economies
90Reinhart and Rogoff (2010)Histograms relating debt to growthAdvanced economies
90Reinhart and Rogoff (2010)Histograms relating debt to growthEmerging markets
90Kumar and Woo (2010)Panel growth equationAll countries
170-180Ostry and others (2010)Debt sustainability (fiscal-reaction function)Advanced economies
Source: Authors’ compilation.
Source: Authors’ compilation.

An alternative approach is to examine the effectiveness of countercyclical fiscal policy conditional on the starting level of debt to GDP. A number of studies have found that the power of fiscal policy to affect aggregate demand depends on the initial level of debt, with higher debt levels leading to a reduction in the multiplier or even a negative multiplier. IMF (2008, 2009), for example, find that the effectiveness of fiscal policy is smaller or negative in countries with high public debt, defined as more than 60–75 percent of GDP for industrial countries and 25 percent of GDP for emerging markets.

Debt levels have been found to be correlated with real GDP growth. Higher public debt levels can crowd out financing for more efficient private investment, or lead to a higher expected tax burden in the future (under the assumption that Ricardian equivalence does not hold), which would lower investment and growth. Alternatively, high levels of debt may hinder the ability of the fiscal authority to relax policies when appropriate, resulting in lower growth. Reinhart and Rogoff (2010), using histograms, show that for a sample of 44 countries over 200 years, the relationship between growth and debt seems to be weak at debt levels of less than 90 percent of GDP, but when debt rises to more than 90 percent of GDP the median growth rate falls by 1 percent. Caner, Grennes, and Koehler-Geib (2010) use an econometric approach—a growth regression estimated using threshold least squares (Hansen, 2000) that identifies a threshold in the relationship between the long-run average public-debt-to-GDP ratio (1980-2008) and long-run average growth. They find that the threshold level of the debt ratio is 77.1 percent, above which each additional percentage-point increase in the debt ratio lowers growth by 0.0174 percentage points. For developing countries the ratio is lower, at 64 percent of GDP. Kumar and Woo (2010) use a dynamic panel growth equation to investigate the relationship between the initial ratio of debt to GDP and growth, and estimate the threshold at which debt has a negative effect on growth to be about 90 percent of GDP.

The original debt intolerance approach

The literature on debt intolerance was largely developed by Reinhart, Rogoff, and Savastano (2003) and Reinhart and Rogoff (2009). These authors observe that certain countries seem to be intolerant of debt (i.e., prone to serial default) even at relatively low levels of external debt to GNP. Debt intolerant countries experience payment problems and default at external debt levels that would seem manageable by standards of advanced economies. The path typically involves a vicious cycle of loss of market confidence, spiraling interest rates, and political resistance to repaying foreign creditors. History seems to be a good guide to identifying countries that are debt intolerant because there is persistence in the probability of default, which is, at least in part, a reflection of persistent institutional weaknesses. Safe debt thresholds turn out to depend heavily on a country’s record of default and inflation.

Looking at the ratio of external debt to GNP for defaulters and nondefaulters in a sample of advanced and emerging market countries, Reinhart and Rogoff (2009) find that more than half of the observations for countries with solid credit histories have debt well below 35 percent of GNP, whereas more than half of the observations for countries with a history of default correspond to debt levels of more than 40 percent of GNP. They conclude that an external debt ratio higher than 35 percent of GNP can begin to increase the risk of default for debt intolerant countries.

To develop a measure of external debt intolerance Reinhart, Rogoff, and Savastano (2003) and Reinhart and Rogoff (2009) use the Institutional Investor Rating (IIR), a rating compiled twice a year by Institutional Investor magazine based on surveys of institutional investors, money managers, and economists.3 The two studies argue that a country’s IIR score can be explained by a very small number of variables, including (1) its initial level of external debt, (2) its repayment history (history of default from 1970 to 2009), and (3) its history of macroeconomic stability (proxied by episodes of inflation higher than 40 percent—a measure of domestic default). They use these variables in a two-step approach to find the marginal effect of an additional unit of debt on the IIR and hence on debt intolerance.

First, the authors divide their sample (53 developing and industrial countries) into “clubs” based on the countries’ average IIRs during the period. Club A comprises countries one standard deviation or more above the mean. Club C comprises countries one standard deviation or more below the mean. Club B comprises countries at the mean plus or minus one standard deviation. Club B is further divided into “regions” (I to IV) based on the level of external debt to GNP and further divisions within this IIR range (Figure 4.1). Using these clubs, the authors run a cross-section regression using averages across time periods for the variables, explaining the IIR as a function of debt, default history, and inflation, and using an interactive dummy for debt on those countries that belong to club A and another for those countries that do not. They find negative and significant coefficients on inflation and history of default, a negative and significant coefficient on debt for countries not in club A, and a positive and significant coefficient on debt for countries in club A. They also perform a panel estimation with qualitatively similar results. Inflation and default have negative effects on credit ratings (IIRs) as does not being in club A.

Figure 4.1Division of Countries into Clubs

Source: Reinhart, Rogoff, and Savastano (2003).

Note: IIR = Institutional Investor Rating; GNP = gross national product.

The second step is to calculate a country-specific debt threshold. Using the estimated coefficients from the first regression, together with actual values of the regressors, it is possible to predict the level of the IIR for varying levels of the debt ratio for a given country. Table 4.2 (from Table 9 in Reinhart, Rogoff, and Savastano, 2003) illustrates the exercise for Argentina and Malaysia.

Table 4.2Predicted Institutional Investor Rating and Debt Intolerance Regions for Argentina and Malaysia
ArgentinaMalaysia
External debt as a percentage of GNPPredicted Institutional Investor RatingRegion typePredicted Institutional Investor RatingRegion type
051.4I61.1I
549.3I59.0I
1047.3I57.0I
1545.2III54.9I
2043.2III52.9I
2541.1III50.8I
3039.1III48.8I
3537.0III46.7II
4034.9IV44.7IV
4532.9IV42.6IV
Source: Reinhart, Rogoff, and Savastano (2003).Note: Reinhart, Rogoff, and Savastano’s calculations are based on the coefficients from regression (1) in Table 8 of Reinhart, Rogoff, and Savastano (2003).
Source: Reinhart, Rogoff, and Savastano (2003).Note: Reinhart, Rogoff, and Savastano’s calculations are based on the coefficients from regression (1) in Table 8 of Reinhart, Rogoff, and Savastano (2003).

Argentina would have to lower its external debt ratio to less than 15 percent to get to region I within club B. Malaysia is able to remain in region I with a debt ratio of up to 30 percent of GNP. This approach has been used in a number of other papers to establish debt thresholds or targets (Table 4.3).

Table 4.3Debt Thresholds Using the Debt Intolerance Approach
AuthorCountryTarget debt as a percentage of GDPClub
Di Bella (2008)Dominican Republic25BII
Everaert (2008)Kenya41BI
Topalova and Nyberg (2010)India49BI
Source: Authors’ compilation.
Source: Authors’ compilation.

Revised approach to measuring Debt intolerance

There are a number of problems with the traditional approach to estimating the debt intolerance equation. First is the possibility of endogeneity of regressors (debt, inflation, and default) to the dependent variable (IIR), which may lead to biased estimates of the coefficients. Reinhart, Rogoff, and Savastano (2003) recognize this and use instrumental variable estimations, achieving the same general results as in their original estimation. However, there may also be endogeneity of IIR groupings (clubs). Because the groupings are based on partitions of the dependent variable (as opposed to the more traditional dummies based on partitions of independent variables), the clubs may also be correlated with the error term, resulting in biased estimates of the coefficients. Second, the static cross-section estimation does not take into account changes in the IIR and debt over time, and the linear relationship between IIR and debt may be restrictive. In addition, the definition of “clubs” depends on the sample of countries chosen for the estimation, which may lead to bias when defining IIR targets and the corresponding debt targets. Third, previous estimations use external debt only, whereas a broader definition that includes domestic debt (i.e., general government debt) might give a more complete picture of the importance of debt levels for debt intolerance.

The revised approach of this chapter seeks to address some of these problems by

  • using a dynamic panel data framework to estimate changes in the IIR with respect to changes in debt, which facilitates application to different countries with different levels of debt and debt intolerance;
  • using generalized method of moments estimation in a panel error correction model framework (Arellano and Bond, 1991; and Arellano and Bover, 1995), correcting for endogeneity of regressors, and introducing country-specific fixed effects;
  • estimating one equation for all countries, avoiding potential endogeneity of the debt groupings and the error term, instead of dividing the sample into ad hoc clubs based on the mean and standard deviation of the IIR;
  • basing the IIR target and corresponding debt threshold calculation on credit ratings issued by major rating agencies; and
  • using a new dataset on general government debt (Abbas and others, 2010), instead of external debt only as in Reinhart, Rogoff, and Savastano (2003).

The sample includes 120 countries, developed and developing, from 1989 through 2009. The panel is unbalanced except for the last 15 years, which are fairly complete. To eliminate noise, and following standard practice, the analysis takes five-year averages over four periods (the six-year period 1989–94; and the five-year periods of 1995–99, 2000–04, and 2005–09).

The result is the estimation of a debt intolerance equation that can be applied flexibly to the calculation of debt targets for most countries in the IIR list. First, after benchmarking the relationship between the IIR and debt at the 2010 level for each country, it is possible to use the coefficients of the econometric estimation to describe a country-specific functional relationship between the IIR and debt. Second, mapping the IIR with the credit rating makes it possible to choose a reference level for the IIR (an IIR target) that corresponds with a given credit-rating target (say, investment grade), which then permits the direct calculation of a corresponding debt target that would allow a country to reach the target IIR, all other things equal.

The Institutional Investor Rating and Debt

Figure 4.2 presents a scatter plot of the average IIR and debt-to-GDP ratio for the 120 countries in the sample over the four periods between 1989 and 2009. The plot suggests a C-shaped relationship between the IIR and the debt ratio, and shows a marked difference between countries with IIRs higher than 65 and those with IIRs less than 65. For countries whose IIRs are higher than 65, there seems to be either no relationship or a slightly positive relationship between debt and the IIR, as the simple linear trend in the figure suggests. For countries with IIRs less than 65, the relationship seems to be negative.

Figure 4.2Institutional Investor Rating and Debt

Sources: Institutional Investor; and IMF staff estimates.

The scatter plot suggests that those countries with IIRs higher than 65 are tolerant to additional levels of debt (at least, in the perception of institutional investors), whereas countries with IIRs less than 65 are less so. To illustrate the point, notice the two points at the top of Figure 4.2 to the right corresponding to Japan in 2000-04 with an IIR of 85 and a debt-to-GDP ratio of 160 percent, and 2005-09 with an IIR of 88 and a debt-to-GDP ratio of 197 percent. Despite an average increase in the debt ratio of almost 40 percentage points, the average IIR increased by three points.

The Debt Intolerance Equation

The modeling of the debt intolerance equation begins with the standard one-way error correction fixed effects model (Baltagi, 2005):

IIRit is the Institutional Investor Rating, Xit is a matrix of k explanatory variables, vi is the unobserved country-specific fixed effect, and uit is the error term, all indexed over i countries and t time periods. Following Reinhart, Rogoff, and Savastano (2003), the analysis uses the ratio of debt to GDP (Dit), a dummy for periods of inflation over 10 percent (IDUAQ, and a dummy for periods of debt restructuring or default (Defaultit as explanatory variables. Unlike Reinhart, Rogoff, and Savastano, these variables do not pick up the historical track record, but the contemporaneous episodes. The model introduces two additional explanatory variables:

  • the ratio of debt to GDP squared (D2it), to capture the nonlinear effects of debt on the IIR; and
  • the level of per capita GDP (CGDPit), which captures a group of attributes (the quality of institutions, endowments, economic structure, and political stability, among others) that allow countries to transform an additional unit of debt into higher income more efficiently.

Finally, the analysis introduces the lagged value of the IIR (IIRit-1) on the right-hand side of the equation to capture the persistence of the level of the IIR over time. Intuitively, past evaluations of creditworthiness and debt intolerance should have a large bearing on the current evaluation. In addition, to the extent that history matters (i.e., for episodes of inflation and default before the sample period) this effect would be captured in the initial level of the IIR. A trend variable is also introduced to capture time-specific effects. The equation is thus

in which

IIRit= Institutional Investor Rating

Dit= general government debt/GDP

dit= matrix of inflation and default dummies (contemporaneous)

CGDPit= per capita GDP

τt= time trend

Vi= country-specific fixed effects.

The Estimation Method

Estimating this equation presents a number of challenges. Even without the lagged value of the IIR on the right-hand side, there is endogeneity between the IIR and the level of debt that could lead to bias in the error term. A second problem is the estimation of the country-specific fixed effect. To overcome these challenges, equation (4.2) is estimated using the generalized method of moments estimator of Arellano and Bond (1991), also known as difference GMM, and an augmented version (system GMM) developed by Arellano and Bover (1995).

These estimators are designed for dynamic panel models with (1) a “small T” (time dimension), and “large N” (observations—in this case, countries); (2) a linear functional relationship; (3) a single left-hand-side variable that is dynamic, depending on its own past realizations; (4) independent variables that are not strictly exogenous, meaning that they are correlated with past and possibly current realizations of the error; (5) fixed individual effects; and (6) heteroscedasticity and autocorrelation within individuals but not across them (Roodman, 2006). These GMM estimators take the first difference of the estimated equation to eliminate the fixed effects term and then use the lagged value (or future value in the forward orthogonal case) of the right-hand side variables as instruments to estimate the coefficients.

Estimation Results

Table 4.4 presents estimations for the panel using ordinary least squares, fixed effects, the Arellano-Bond difference GMM estimator, system GMM, and system GMM forward orthogonalized. This estimation drops countries with an IIR of less than 25, considering these to not have access to private international debt markets. This reduces the sample to 102 countries from 120. The estimation is performed on a panel for which the annual observations have been averaged over four multiyear periods between 1989 and 2009.

Table 4.4Estimation of the Debt Intolerance Equation
(1)(2)(3)(4)(5)
Model specificationOrdinary least squaresFixed effectsArellano-Bond 1-stepSystem GMMSystem GMM orthogonalized
IIR lagged0.795***0.398***0.555***0.535***0.522***
[34.17][7.231][7.564][8.077][6.361]
Debt/GDP0.045−0.150***−0.337***−0.318***−0.347***
[-1.363][-2.847][-3.075][-3.410][-3.163]
Debt/GDP squared6.85e-050.0001660.00116*0.00114***0.00129***
[-0.308][0.559][1.865][2.584][2.661]
Inflation−0.865***−1.182***−1.952***−2.247***−2.412***
[-3.521][-3.308][-4.733][-4.678][-4.011]
Default−1.444***−0.156−1.487***−1.462***−1.106**
[-3.233][-0.271][-2.648][-2.772][-2.134]
Per capita GDP0.000194***0.000227**0.000502***0.000550***0.000594***
[4.054][2.462][3.321][4.468][4.040]
Trend0.2262.209***−0.722−1.031*−1.293**
[0.593][3.655][-1.467][-1.923][-2.167]
Constant16.94***34.89***38.74***39.06***40.87***
[9.515][12.15][8.113][7.537][6.136]
Observations271271271271271
R20.9430.731
Number of countries102102102102
Number of instruments323232
Hansen test p-value0.6070.6070.518
Arellano-Bond AR(1) test p-value0.01630.01640.0241
Arellano-Bond AR(2) test p-value
Source: IMF staff calculations.Note: Robust t-statistics in brackets. AR(1) = autoregressive of order 1; AR(2) = autoregressive of order 2; GMM = generalized method of moments; IIR = Institutional Investor Rating.

p < 0.01,

p < 0.05,

p < 0.1.

Source: IMF staff calculations.Note: Robust t-statistics in brackets. AR(1) = autoregressive of order 1; AR(2) = autoregressive of order 2; GMM = generalized method of moments; IIR = Institutional Investor Rating.

p < 0.01,

p < 0.05,

p < 0.1.

The coefficients in the estimation are significant and of the expected sign. The coefficient on the lagged IIR is positive 0.5, signaling a significant degree of persistence in the IIR. The coefficient on the debt ratio is negative, signaling the effect of higher debt on the perception of creditworthiness, and the coefficient on debt squared is positive and significant. Coefficients on inflation and default variables are negative and the coefficient on per capita GDP is positive, as expected. The number of instruments in the estimation is kept at 32 for all GMM estimations. The Arellano-Bond AR(1) test for autocorrelation of the residuals rejects the hypothesis that the errors are not autocorrelated, which is expected because first differencing should result in autocorrelation of order 1. The test of AR(2) is not presented because the four periods do not allow sufficient degrees of freedom for its calculation. The Hansen test for overidentification restrictions on the instruments is significantly above zero, suggesting that the number of instruments is appropriate.

A number of sensitivity tests and alternative specifications were attempted, but most resulted in qualitatively similar results. Country-specific dummies were used; however, they were not significant.4 A further attempt was made to modify the sample by excluding or controlling for countries that belong to the European Union (EU). An examination of the data showed that a number of countries had substantially improved IIRs while at the same time substantially increasing their levels of indebtedness. Among these are countries that improved their policy frameworks significantly during the period, including India, China, Singapore, and Korea. Aside from these, most others are EU countries that received boosts in their perceived creditworthiness from the creation of the union or from subsequently joining it. Most of these countries have both higher IIRs and higher debt at the end of the period compared with the beginning. The prominent example is Greece, whose IIR increased by 30 points between 1989-93 and 2005-09, and whose debt ratio rose by 24 percentage points. Excluding EU countries from the sample reduces the number of countries to 88 from 102, but does not significantly change the qualitative results of the estimation. An EU member dummy, on its own or interacted with debt, does not turn out to be significant.5

An estimation was also run on the annual data series. The coefficient on the lagged IIR was higher, as expected (0.78), because abrupt changes are less likely from year to year, and the coefficient on debt lower (-0.2); both had the correct sign and were significant. Coefficients on inflation, default, and per capita GDP were similar to the baseline estimation. Tests reject second-order autocorrelation and confirm that the number of instruments is appropriate.

Application to capdr

To operationalize the model, the analysis takes the first difference of the estimated equation (4.2) and assumes no changes in the inflation and default dummies and per capita GDP so that these elements, the constant, and the country-specific fixed effects term drop out. Time-specific effects are also assumed to be zero. This yields the following equation for the determination of the IIR with respect to debt:

The path of the IIR is then calculated for changes in debt starting from the benchmark of the 2010 level.6 For example, the 2010 IIR of 55.1 and debt-to-GDP ratio of 37.5 percent are used for Costa Rica, yielding the curve in Figure 4.3. The curve has the expected properties, in that the marginal effect of debt on the IIR goes to zero as debt reaches about 135 percent of GDP.

Figure 4.3Costa Rica: Relationship between Institutional Investor Rating and Debt

Source: IMF staff calculations.

The next step is to establish thresholds for the IIR to calculate reference target levels of debt. Rather than dividing the country sample into clubs with reference to the mean and the standard deviation of the IIR, this analysis investigates the correspondence between the 2010 IIR and country sovereign debt ratings issued by the three major rating agencies (Moody’s, Fitch, and Standard & Poor’s). The sample falls naturally into three groups: (1) those countries that are clearly rated investment grade, (2) those countries that are clearly rated noninvestment grade, and (3) those that have an uncertain grade (borderline). The thresholds are found by ranking countries by their IIRs in 2010, then moving down the list to find the country with the highest IIR for which at least one rating agency has issued a non-investment-grade rating. This forms the upper threshold. The lower threshold is found by starting at the bottom of the list and moving up to the first country for which at least one rating agency has issued an investment-grade rating. Countries between these two thresholds can have mixed ratings, but some are uniformly rated investment grade and others are uniformly rated noninvestment grade.7 This method yields an upper threshold IIR of 58.7, above which all countries are clearly investment grade, and a lower threshold IIR of 51.3, below which all countries are clearly noninvestment grade (Table 4.5).

Table 4.5Institutional Investor Rating and Credit Ratings
CountryFitchMoody’sStandard & Poor’s2010 IIR
IrelandA+Aa2AA-67.5
Unambiguously investment grade
SpainAA+Aa1AA66.7
Russian FederationBBBBaa1BBB66.4
BotswanaA2A-66.3
IndiaBBB-Baa3BBB-64.6
PortugalAA-A1A-62.2
South AfricaBBB+A3BBB+62.0
ThailandBBBBaa1BBB+60.2
EstoniaAA1A59.5
ColombiaBB+Ba1BBB-58.7
TunisiaBBBBaa2BBB58.5
Borderline
Mauritius0Baa2054.1
NamibiaBBB-0052.8
PeruBBB-Baa3BBB-58.3
HungaryBBBBaa1BBB-56.9
PanamaaBBB-Baa3BBB-56.9
IndonesiaBB+Ba2BB56.2
CroatiaBBB-Baa3BBB55.6
TurkeyBB+Ba2BB55.3
MoroccoBBB-Ba1BBB-55.2
Costa RicaaBBBaa3BB55.1
BulgariaBBB-Baa3BBB54.4
LithuaniaBBBBaa1BBB54.3
KazakhstanBBB-Baa2BBB-54.0
LibyaBBB+A-53.7
RomaniaBB+Baa3BB+51.4
PhilippinesBBBa3BB-51.3
EgyptBB+Ba1BB+51.0
Unambiguously noninvestment grade
UruguayBBBa3BB50.1
VietnamBB-Ba3BB47.9
LatviaBB+Ba3BB47.3
JordanBa2BB45.8
El SalvadoraBBBa1BB45.5
GuatemalaaBB+Ba1BB45.3
GreeceBB+Ba1BB+43.9
GabonBB-BB-42.2
LesothoBB-42.1
AngolaB+B1B+41.3
IcelandBB+Baa3BBB-41.0
Dominican RepublicaBB1B40.8
UgandaBB+35.0
Sources: Fitch; Moody’s; Standard & Poor’s; and Institutional Investor.Note: Shaded ratings in borderline area are not investment grade. IIR = Institutional Investor Rating.

Central America, Panama, and the Dominican Republic.

Sources: Fitch; Moody’s; Standard & Poor’s; and Institutional Investor.Note: Shaded ratings in borderline area are not investment grade. IIR = Institutional Investor Rating.

Central America, Panama, and the Dominican Republic.

Applying these thresholds to the information in Figure 4.3 determines the debt ratio that Costa Rica, for example, would have to achieve to be unambiguously investment grade (in this case, 25.4 percent) or how much space it has to increase debt before becoming unambiguously noninvestment grade (in this case, raising debt to about 50 percent of GDP; Figure 4.4), holding all other determinants of debt intolerance constant.

Figure 4.4Costa Rica: Relationship between Institutional Investor Rating and Debt Thresholds

Source: IMF staff calculations.

Table 4.6 presents similar calculations for the other CAPDR countries. Like Costa Rica, Panama would target a lowering of the debt ratio from 40 percent of GDP to 32.7 to move into the unambiguously investment-grade category. El Salvador, Guatemala, and the Dominican Republic begin with IIRs in 2010 that are already unambiguously noninvestment grade (below 51.3), so would target this lower threshold. Contrasting El Salvador and Guatemala, the two have almost identical IIRs for 2010, despite having very different levels of debt to GDP. Guatemala’s debt-to-GDP ratio, at half that of El Salvador, clearly indicates that Guatemala has a higher level of debt intolerance. The same is true for the Dominican Republic. For these countries, reaching the intermediate group (where they could begin to be considered candidates for investment grade) would require a significant debt-reduction effort, signaling that the market perceives significant structural issues affecting their levels of debt tolerance that would have to be solved for them to take on higher debt comfortably.

Table 4.6CAPDR: Debt Thresholds Based on the Debt Intolerance Equation
RatingIIR indexDebt (Percent of GDP)
2010Target2010Target2010Target
PanamaBorderlineInvestment grade56.958.740.032.7
Costa RicaBorderlineInvestment grade55.158.737.525.4
El SalvadorNoninvestment gradeBorderline45.551.351.734.4
GuatemalaNoninvestment gradeBorderline45.351.324.111.4
Dominican RepublicNoninvestment gradeBorderline40.851.336.914.3
HondurasHighly speculativeSpeculative30.938.626.18.6
NicaraguaSubstantial riskHighly speculative23.925.866.331.3
Sources: Institutional Investor; and IMF staff calculations.Note: CAPDR = Central America, Panama, and the Dominican Republic.; IIR = Institutional Investor Rating.
Sources: Institutional Investor; and IMF staff calculations.Note: CAPDR = Central America, Panama, and the Dominican Republic.; IIR = Institutional Investor Rating.

Honduras and Nicaragua represent special cases because they have very limited domestic debt markets and much (if not all) of their external debt is taken on concessional terms from international financial institutions such as the World Bank and the Inter-American Development Bank. In these cases, the IIR and credit-rating thresholds are chosen as the next rung on the ladder of international investment ratings that would allow them to eventually qualify for access to international bond markets and nonconcessional lending. Nicaragua’s IIR in 2010 was classified as a “substantial risk” (CCC+), which is just a notch above default. Moving up to the next credit-rating category—”highly speculative” (single B)—would require a substantial debt-reduction effort to bring the debt ratio from 66 percent of GDP to 31 percent, all other things equal. Similarly, for Honduras to move from its current rating of “highly speculative” to “speculative” (from B to BB) would require a reduction in the debt ratio from its current 26 percent of GDP to 9 percent. Table 4.7 illustrates the different credit-rating levels that could be targeted by countries, with their corresponding 2010 IIR ratings.

Table 4.7Credit-Rating Classifications and the 2010 Institutional Investor Ratings
FitchMoody’sStandard & Poor’sClassIIR range
AAAAaaAAAPrime88.6-100
AA+Aa1AA+
AAAa2AAHigh grade80.6-88.5
AA-Aa3AA-
A+A1A+Upper medium grade
AA2A70.3-80.5
A-A3A-
BBB+Baa1BBB+Lower medium grade
BBBBaa2BBB58.7-70.2
BBB-Baa3BBB-
BB+Ba1BB+Noninvestment grade speculative
BBBa2BB42.1-58.6
BB-Ba3BB-
B+B1B+Highly speculative
BB2B25.0-42.0
B-B3B-
CCC+Caa1CCC+Substantial risk
CCCCaa2CCCExtremely speculative
CCC-Caa3CCC-0-24.9
CCCaCC
CCCIn default
RDSD
DD
Sources: Fitch; Moody’s; Standard & Poor’s; Institutional Investor, and IMF staff calculations.
Sources: Fitch; Moody’s; Standard & Poor’s; Institutional Investor, and IMF staff calculations.

Table 4.6 illustrates how different countries with different levels of debt can have the same IIR (Guatemala and El Salvador) and how countries with the same level of debt (for example, Costa Rica and the Dominican Republic) can have significantly different IIRs. These juxtapositions illustrate the market perceptions of different levels of debt intolerance for these countries. To capture these differences Reinhart, Rogoff, and Savastano (2003) propose using an index of (external debt/GDP)/IIR and (external debt/exports)/IIR. Because the present case uses general government debt, an alternative indicator could be the level of the IIR consistent with a level of debt to GDP of 50 percent. This method could objectively rank countries by debt intolerance. Table 4.8 presents the IIR ranking for the CAPDR countries using a debt target of 50 percent of GDP. The IIR rankings are not surprising overall, with Panama 5 points higher than Costa Rica and 10 points higher than El Salvador. Guatemala and the Dominican Republic seem to be in a different category of debt intolerance, ranked similarly about 20 points below Panama. The rankings for Nicaragua and Honduras are in a third category of countries with very high debt intolerance. The higher ranking for Nicaragua can be attributed to the higher initial debt level for the given IIR. Given that this debt is entirely concessional, whether this ranking can be compared with the others with no adjustment for the grant element of the debt is questionable.8

Table 4.8Ranking of Countries by Debt Intolerance

(Institutional Investor Rating consistent with debt level of 50 percent of GDP in 2010)

CountryInstitutional Investor Rating
Panama55.6
Costa Rica51.1
El Salvador45.3
Guatemala37.4
Dominican Republic36.3
Nicaragua28.4
Honduras21.1
Source: IMF staff calculations.
Source: IMF staff calculations.

Conclusion

This chapter presents an alternative method for calculating debt targets starting with the debt intolerance literature of Reinhart, Rogoff, and Savastano (2003) and Reinhart and Rogoff (2009). The methodology improves on the previous studies by using a dynamic panel approach, correcting for endogeneity in the regressors, and basing the calculation of debt targets on credit ratings. In addition, this chapter uses a database on general government debt (both external and domestic) that includes data for 120 countries for 21 years. Benchmarking the calculations on the 2010 IIR and debt for each country, econometric estimations can be used to calculate a reference IIR for different levels of debt. Choosing the IIR that corresponds to a desired credit-rating target then allows for the calculation of the debt level that would yield the desired IIR and credit rating, all other things equal. Choosing the IIR for all countries that is consistent with a given debt level (i.e., 50 percent of GDP) yields an index of debt intolerance, allowing a comparison to be made across countries.

The application to CAPDR provides reasonable target debt levels to get to different desired credit-rating equivalents. The goal for Panama and Costa Rica is to reach an IIR level that would put them in the unambiguously investment-grade category. These countries would have to reduce their 2010 debt-to-GDP levels by 7 and 12 percentage points, respectively, to reach this goal. The goal for El Salvador, Guatemala, and the Dominican Republic would be to reach an IIR level that would allow them to be considered candidates for an upgrade to investment grade (ambiguously investment grade). This goal would require significant debt-reduction efforts by the Dominican Republic (22.6 percentage points) and El Salvador (17.3 percentage points), but less so for Guatemala (12.7 percentage points). Honduras and Nicaragua have little to no access to private international sovereign credit markets, so the goal is to get to a rating that would allow them to consider some kind of limited market access. However, the reference numbers calculated here would have to be adjusted for the level of concessionality of their general debt to make them comparable to the other countries in the region.

This analysis holds all other factors that affect debt intolerance fixed, and relies solely on changes in the level of debt to GDP to improve the market’s perception of debt intolerance as embodied in the IIR. It is advisable that countries work on these other aspects as well to improve their debt tolerance. In these econometric equations, the other aspects are embodied in the inflation (macroeconomic stability), default, and per capita GDP variables, which are highly aggregated indicators for complex factors, such as institutional capacity and technical efficiency, that affect a country’s capacity to convert debt into growth and enhance its repayment capacity. A first step would be to develop an index of debt intolerance that would be comparable across countries for a particular point in time. This chapter proposes such an index, calculated as the level of the IIR in each country consistent with a given debt ratio (in this case 50 percent). When implemented for the CAPDR region, it shows that countries with divergent debt levels and IIRs can have similar levels of debt intolerance. Further work is required to explain the differences in these debt intolerance rankings, to see which other countries might be considered CAPDR’s peers with regard to debt intolerance, and to investigate the characteristics that affect debt intolerance most directly.

References

    Abbas, S.A., N.Belhocine, A.ElGanainy, and M.Horton, 2010, “A Historical Public Debt Database,” Working Paper 10/245 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

    Aiyagari, S.R., and E.R.McGrattan, 1998, “The Optimum Quantity of Debt,” Journal of Monetary Economics, Vol. 42, No. 3, pp. 447—69.

    • Search Google Scholar
    • Export Citation

    Arellano, M., and S.Bond, 1991, “Some Tests of Specification for Panel Data: Monte Carlo Evidence and an Application to Employment Equations,” Review of Economic Studies, Vol. 58, No. 2, pp. 277—97.

    • Search Google Scholar
    • Export Citation

    Arellano, M., and O.Bover, 1995, “Another Look at the Instrumental Variables Estimation of Error Components Models,” Journal of Econometrics, Vol. 68, No. 1, pp. 29—51.

    • Search Google Scholar
    • Export Citation

    Baltagi, B.H., 2005, Econometric Analysis of Panel Data (New York: Wiley).

    Caner, M., T.Grennes, and F.Koehler-Geib, 2010, “Finding the Tipping Point—When Sovereign Debt Turns Bad,” Policy Research Working Paper No. 5391 (Washington: World Bank).

    • Search Google Scholar
    • Export Citation

    Di Bella, G., 2008, “A Stochastic Framework for Public Debt Sustainability Analysis,” Working Paper 08/58 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

    Everaert, G., 2008, Kenya: Selected Issues, IMF Country Report No. 08/337 (Washington: International Monetary Fund).

    Floden, M., 2001, “The Effectiveness of Government Debt and Transfers as Insurance,” Journal of Monetary Economics, Vol. 48, No. 1, pp. 81—108.

    • Search Google Scholar
    • Export Citation

    Hansen, B.E., 2000, “Sample Splitting and Threshold Estimation,” Econometrica, Vol. 68, No. 3, pp. 575—603.

    International Monetary Fund, 2003, “Public Debt in Emerging Markets: Is It Too High?” Chapter 3 in World Economic Outlook, World Economic and Financial Surveys (Washington, September).

    • Search Google Scholar
    • Export Citation

    International Monetary Fund, 2008, “Fiscal Policy as a Countercyclical Tool,” Chapter 5 in World Economic Outlook, World Economic and Financial Surveys (Washington, October).

    • Search Google Scholar
    • Export Citation

    International Monetary Fund, 2009, “From Recession to Recovery: How Soon and How Strong?” Chapter 3 in World Economic Outlook, World Economic and Financial Surveys (Washington, April).

    • Search Google Scholar
    • Export Citation

    Kumar, M.S., and J.Woo, 2010, “Public Debt and Growth,” Working Paper 10/174 (Washington: International Monetary Fund).

    Ostry, J.D., A.R.Ghosh, J.I.Kim, and M.S.Qureshi, 2010, “Fiscal Space,” IMF Staff Position Note No. 10/11 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

    Reinhart, C.M., and K.S.Rogoff, 2009, This Time Is Different: Eight Centuries of Financial Folly (Princeton, New Jersey: Princeton University Press).

    • Search Google Scholar
    • Export Citation

    Reinhart, C.M., and K.S.Rogoff, 2010, “Growth in a Time of Debt,” American Economic Review, Vol. 100, No. 2, pp. 573—78.

    Reinhart, C.M., K.S.Rogoff, and M.A.Savastano, 2003, “Debt Intolerance,” Brookings Papers on Economic Activity, Vol. 2003, No. 1, pp. 1—74.

    • Search Google Scholar
    • Export Citation

    Roodman, D., 2006, “How to Do xtabond2: An Introduction to ‘Difference’ and ‘System’ GMM in Stata,” Working Paper No. 103 (Washington: Center for Global Development).

    • Search Google Scholar
    • Export Citation

    Saint-Paul, G., 2005, “Fiscal Policy and Economic Growth: The Role of Financial Intermediation,” Review of International Economics, Vol. 13, No. 3, pp. 612—29.

    • Search Google Scholar
    • Export Citation

    Shin, Y., 2006, “Ramsey Meets Bewley: Optimal Government Financing with Incomplete Markets,” Department of Economics, University of Wisconsin, unpublished.

    • Search Google Scholar
    • Export Citation

    Topalova, P., and D.Nyberg, 2010, “What Level of Public Debt Could India Target?” Working Paper 10/7 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

    Weh-Sol, M., 2010, “Korea’s Optimal Public Debt Ratio,” SERI Samsung Economic Research Institute Quarterly, April.

This chapter is based on G. Bannister and L. Barrot, 2011, “A Debt Intolerance Framework Applied to Central America, Panama, and the Dominican Republic,” IMF Working Paper 11/220, September (Washington: International Monetary Fund).

1

CAPDR comprises Costa Rica, El Salvador, Guatemala, Honduras, Nicaragua, Panama, and the Dominican Republic.

2

Chapters 2 and 5 discuss the features of stimulating fiscal policies and the options to regain fiscal space in the region in more detail.

3

The IIR grades more than 166 countries from 0 (least creditworthy) to 100 (most creditworthy). The IIR has been published since the mid-1980s.

4

A regional dummy for CAPDR, interacting with debt, had a small, positive, and significant coefficient, which could be picking up the consistent improvement in debt tolerance in the region during the sample period.

5

Period dummies instead of a time trend were also used with no effect on the estimations. The dummies were not significant in the GMM estimation. Similarly, a variable for revenues/GDP was used to proxy debt repayment capacity but also turned out not to be significant. Finally, external debt was used instead of total debt, but the coefficient, although significant, was very small (0.02).

6

For the first period, it is assumed that IIRt-1 = IIRt-2 to maintain the smoothness of the debt intolerance curve.

7

An additional way to verify that countries belong in the intermediate range would be to examine the outlooks for the ratings of countries on the border between investment and noninvestment grade.

8

One way to adjust for the grant element would be to convert the debt into a market-equivalent nominal amount that would yield the same net present value as current debt at market interest rates.

    Other Resources Citing This Publication