Journal Issue
Share
Article

Colombia

Author(s):
International Monetary Fund
Published Date:
May 2010
Share
  • ShareShare
Show Summary Details

I. Determinants of Investment Grade Status and Implications for Colombia’s Public Debt1

A. Introduction

1. The Colombian authorities are interested in regaining investment grade status, which was lost during the financial crisis of 1998–99. Achieving investment grade status would help lower financing costs for the sovereign, and expand the pool of potential buyers of Colombian paper. It would also reduce borrowing costs for corporates with access to international capital markets.

2. Identifying the main determinants of investment grade status can help guide policies towards achieving this goal. Colombia’s debt levels as of end 2008 were broadly similar to the average for investment grade emerging markets, suggesting that other indicators are taken into account in rating agencies’ assessments. Gauging the weights given by rating agencies to alternative macroeconomic variables can be a helpful input for the authorities’ overall policy framework, including the role of a medium term debt goal.

3. The paper finds that, to a large extent, investment grade rating status can be explained by a small number of variables. The analysis is based on a random effects binomial logit model estimated with data for a sample of 48 emerging market economies during the period 1993–2008. The model identifies a set of five core variables that are relevant for the determination of investment grade status, namely external public debt, domestic public debt, political risk, exports, and broad money (all variables as a share of GDP, except for political risk).

4. Building on existing literature, the main contributions of this paper consist of: (i) defining the dependent variable in binary form (investment grade/speculative grade) as opposed to ordinal ratings; (ii) restricting the sample to emerging market economies only, to avoid industrial country bias on investment grade determinants; and (iii) presenting additional explanatory variables not included in previous studies, in particular a breakdown of debt indicators and a measure of financial depth.

B. Background and Literature Review

5. Sovereign debt ratings are intended to be forward-looking qualitative measures of the probability of default elaborated by rating agencies. They are summary assessments of a government’s ability and willingness to repay its debts in full and on time. The three major credit rating agencies—Moody’s Investor Services (Moody’s), Standard and Poor’s (S&P), and Fitch Ratings (Fitch)—indicate that their assessments of government risk are based on the analysis of a broad set of economic, social, and political factors, but are not explicit about the weights given to those variables in their final assessments.2 The ratings (and meaning) given by these agencies are summarized in Table 1.

Table 1.Sovereign Credit Ratings by Agency
S&PMoody’sFitch
Investment Grade
Highest quality, reliable, stableAAAAaaAAA
Quality, but a little more riskAAAaAA
Economic situation can affect financesAAA
Just making the gradeBBBBaaBBB
Speculative Grade
More sensitive to economic changesBBBaBB
Financial situation varies considerablyBBaB
Vulnerable, dependent on favorable economic conditions to meet paymentsCCCCaaCCC
Highly vulnerable, speculativeCCCaCC
Close to default, may be in arrearsCCC
Defaulted on obligationsDD
Note: Within rating categories, S&P and Fitch use plus (+) or minus (-) signs to show relative standing, with A+ being better than A or A-.Moody's uses a modifier of 1, 2, or 3 for the same purpose, with A1 being better than A2 or A3.
Note: Within rating categories, S&P and Fitch use plus (+) or minus (-) signs to show relative standing, with A+ being better than A or A-.Moody's uses a modifier of 1, 2, or 3 for the same purpose, with A1 being better than A2 or A3.

6. Sovereign credit ratings are important for at least three reasons. First, they are a key determinant of a country’s borrowing costs in international capital markets. Second, the sovereign rating generally sets a ceiling for the ratings assigned to domestic banks and companies, and therefore affects private financing costs.3 And third, some institutional investors have lower bounds for the risk they can assume in their investments and will choose their portfolio composition taking into account the credit risk signaled by the rating notations.

7. Since sovereign ratings summarize a vast amount of information, empirical studies have tried to predict country ratings based on a parsimonious set of economic variables. The seminal paper by Cantor and Packer (1996), based on a sample of 49 industrialized and developing countries, suggested that six variables were likely to explain ratings: per capita income, GDP growth, inflation, external debt, level of economic development, and default history. Using the same methodology, Afonso (2003) found that GDP per capita was the only relevant determinant of ratings of developed countries, while external debt played a key role for developing countries. In contrast, Mulder and Perrelli (2001) found that for emerging market economies the ratio of investment to GDP was the key variable explaining ratings. Results on the significance of political variables have been mixed: Archer et al. (2007) concluded that political factors had little effect on bond ratings; Mellios and Paget-Blanc (2006) found that indicators of corruption were an important determinant of ratings; and Afonso et al. (2007) found a significant coefficient for an indicator of government effectiveness. (See Table 2).

Table 2.Main Studies on the Determinants of Sovereign Ratings
Study and Regression TechniqueData SampleSignificant explanatory variables
Afonso (2003)

OLS
81 countries, June 2001GDP per capita (+), GDP growth (+)

Inflation (-), External debt ratios (-)

Economic development (+), Default history (-)
Afonso, Gomes, Rother (2007)

Pooled ordered probit, random effects ordered probit
130 countries, 1970-2005 Moody’s, S&P, FitchGDP per capita (+), GDP growth (+)

Inflation (-), External debt ratios (-)

International reserves (+), Default history (-)

Government effectiveness (+)

EU countries (+)
Archer et al. (2007)

Linear regression with panel-corrected standard errors
50 developing countries, 1987-2003 Moody’s, S&P, FitchTrade (+), Inflation (-)

GDP growth (+), Default (-)
Cantor and Packer (1996)

OLS
49 countries, September 1995 Moody’s, S&PGDP per capita (+), GDP growth (+)

Inflation (-), External debt ratios (-)

Economic development (+), Default history (-)
Mellios and Paget-Blanc (2006)

Ordered logistic model
86 countries, December 2003 Moody’s, S&P, FitchPer capita GDP (+), Government income (+)

Real exchange rate changes (+), Inflation (-)

Default history (-), Corruption index (+)
Mulder and Perrelli (2001)

Pooled OLS and Feasible GLS
25 emerging market economies, 1992-1999 Moody’s, S&PDebt over exports (-), Rescheduling history (-)

Fiscal balance (+), Output growth (+)

Inflation (-), Investment to GDP (+)
Rowland (2004)

OLS regression
50 developing countries, July 2003 Moody’s, S&PGDP per capita (+), GDP growth (+)

Inflation (-), External debt ratios (-)

International reserves (+), Openness (+)
Rowland and Torres (2004)

Random effects GLS regression
16 emerging market economies, 1987-2001 Moody’s, S&PGDP growth (+), Inflation (-)

External debt ratios (-), International reserves (+)

Openness (+), Default history (-)

C. Empirical Model Specification

8. This paper will focus on the determinants of investment grade ratings for emerging market economies. All the studies listed in Table 2 transformed credit ratings into a linear scale and used this ordinal measure as the dependent variable. In contrast, this paper defines a binary dependent variable for investment grade status, based on ratings data from Moody’s, S&P, and Fitch. The rating for any given year is the end-December rating, and the dummy is made equal to 1 for countries that were assigned investment grade status by at least two out of the three agencies.4 A panel data framework is used to control for heterogeneity across countries. A random effects binomial logit model produces better results (from an econometric point of view) than those obtained from a pooled regression and a fixed effects regression. The advantage of this technique is that it uses information from all countries in the sample and the marginal effect of any independent variable on the probability is conditional on the values of all covariates.5

9. The model specification can be written as:

Where IGit is the binary variable equal to 1 for countries with investment grade status; Xit is a vector containing the time-varying explanatory variables described below; Zi is a vector of time invariant variables that include regional and default dummies; ai stands for the individual effects for each country i (that can either be modeled as an error term or as N dummies to be estimated) and μit represent disturbances that are independent across countries and across time.

10. Building on the evidence provided by the existing literature, the paper identifies a set of potential determinants of investment grade status. The selection of the explanatory variables is guided by the rating agencies’ reports and previous empirical evidence. Table 3 lists the explanatory variables that were included as regressors (Xit) in model.6

Table 3.Explanatory Variables and Expected Sign
Macroeconomic IndicatorsGovernment Sector
GDP per capita (US$ dollars)+Primary balance+
Real GDP growth+External public debt to GDP**-
Potential GDP growth+Domestic public debt to GDP**-
Inflation rate-
Unemployment rate-Financial Depth
Broad money to GDP**+
External SectorOther
Exports to GDP+Political risk index ICRG
Current account balance to GDP-(higher index implies lower risk)+
Private external debt to GDP**-Default history-
NIR to GDP+Regional dummies
Note: ** denotes variables not included in previous empirical studies
Note: ** denotes variables not included in previous empirical studies

D. Data and Estimation Results

Data

11. The regression analysis is based on a sample of 48 emerging market countries. The ratings data are obtained from the three main rating agencies for the 1993–2008 period. Table 4 provides some descriptive statistics of the dependent variable.7 Unless otherwise specified, the macroeconomic variables are drawn from the WEO database. The political risk variable is based on the political risk index published by the International Country Risk Guide, where a higher value indicates lower risk. For all the time-varying regressors, lagged values of the explanatory variables are used to avoid endogeneity problems.

Table 4.Descriptive Statistics of Countries in the Sample and Investment Grade Status
Latin AmericaEuropeEast AsiaOtherTotal
Number of countries151671048
Number of observations218222112139691
Percent of observations in the total sample31.333.314.620.8100.0
Percent of observations with investment grade status17.463.157.927.040.7
Percentage of countries with a switch in status40.050.057.130.043.8
Percentage of countries with at least one upgrade to investment grade33.337.542.930.037.5
Percentage of countries with at least one downgraded to speculative grade20.018.857.110.022.9

12. Tests of means and medians show that investment grade countries tend to outperform speculative grade countries on most of the economic dimensions captured by the regressors. Welch tests were used to test for equality of means across the two groups of observations (investment and speculative grade), and Wilcoxon rank-sum tests were used to test whether the distribution is independent across the two groups of observations.8Table 5 shows that both tests yielded similar results, with investment grade countries showing “better” values for the indicator than speculative grade countries on most accounts in almost all cases.

Table 5.Test of Equality of Means and Medians of Country Characteristics, by Investment Grade Rating
MeanMedian
VariableColombia 2008Investment gradeSpeculative gradeWelch testInvestment gradeSpeculative gradeWilcoxon test
Macroeconomic Variables
GDP per capita (US$ dollars)4,7638,0833,766**4,9502,800**
Real GDP growth2.54.83.8**5.14.4**
Potential GDP growth4.04.43.7**3.93.8*
Inflation7.05.725.9**4.47.7**
Unemployment10.68.310.5**7.59.8**
External Sector
Exports to GDP17.745.730.4**40.929.1**
Current account balance to GDP(2.8)(3.0)(2.4)(2.9)(2.4)
Private external debt to GDP6.942.419.6**32.314.3**
NIR to GDP9.718.714.5**17.111.2**
Government Sector
Primary balance2.60.40.8*(0.1)0.8**
External public debt to GDP12.212.830.6**9.925.3**
Domestic public debt to GDP19.921.727.3**15.219.6**
Financial Depth
Broad money to GDP35.864.354.6**50.442.6**
Other
Political risk index ICRG
(+ is lower risk)58.573.064.6**74.065.5**
Sources: Authors’ calculations based on data from IMF, World Bank, and International Country Risk Guide.Note: ** stands for statistical significance at the 1 percent level; * stands for statistical significance at the 5 percent level.
Sources: Authors’ calculations based on data from IMF, World Bank, and International Country Risk Guide.Note: ** stands for statistical significance at the 1 percent level; * stands for statistical significance at the 5 percent level.

Regression results

13. Table 6 shows the results of estimating equation (1), with three different techniques: pooled, random effects and fixed effects. For each technique, the first column reports the unrestricted model (i.e. columns A, C, and E), whereas the second shows the results for the restricted model (i.e. columns B, D and F). The unrestricted model incorporates all the variables listed in Table 3, whereas the restricted model contains only the variables which were found to have a statistically significant impact.9

14. The random effects model (column D) is found to be the preferred specification. The likelihood-ratio test rejected the null hypothesis of no variation in the country specific errors of the pooled regression (columns A and B), indicating the need to control for country specific effects. At the same time, Hausman specification tests did not reject the null hypothesis that the random effects (column D) and consistent fixed effects coefficients (column F) were the same, suggesting that the random effects model is appropriate.

Table 6.Regression Results
Pooled regressionRandom effectsFixed Effects
VARIABLESABCDEF
GDP per capita (US$)0.00**0.00***0.000.00
(0.00)(0.00)(0.00)(0.00)
Potential GDP growth0.12-0.34-0.59
(0.12)(0.42)(0.53)
Inflation-0.05-0.02-0.01
(0.03)(0.04)(0.01)
Unemployment0.06*0.19-0.03
(0.04)(0.14)(0.17)
Exports to GDP0.02*0.010.100.08***0.23*0.10**
(0.01)(0.01)(0.07)(0.03)(0.12)(0.04)
Current account balance to GDP0.06*0.01-0.02
(0.04)(0.12)(0.16)
Private external debt to GDP0.02-0.02-0.05
(0.02)(0.03)(0.04)
Net international reserves to GDP0.000.010.160.29
(0.03)(0.02)(0.12)(0.18)
Primary balance to GDP0.050.03-0.31
(0.08)(0.18)(0.26)
External public debt to GDP-0.09***-0.09***-0.38***-0.25***-0.45***-0.22***
(0.02)(0.01)(0.10)(0.04)(0.11)(0.04)
Domestic public debt to GDP-0.06***-0.05***-0.18***-0.09***-0.24***-0.08***
(0.01)(0.01)(0.05)(0.02)(0.06)(0.03)
Broad money to GDP0.02**0.02***0.11**0.04***0.24***0.07**
(0.01)(0.01)(0.06)(0.02)(0.08)(0.03)
ICRG political risk index0.09**0.10***0.090.15***0.11*0.15**
(0.04)(0.03)(0.08)(0.05)(0.06)(0.06)
Latin America and Caribbean dummy-1.06-1.69***1.60
(0.72)(0.38)(3.81)
Europe dummy0.354.56
(0.66)(4.28)
East Asia dummy-0.25-0.27
(0.73)(4.40)
Default dummy-1.82***-3.27
(0.56)(4.85)
Constant-7.53**-6.44***-10.55-10.10***
(2.96)(1.80)(8.25)(3.83)
Observations483536483540248284
Number of countries35371719
Robust standard errors in parentheses*** p<0.01, ** p<0.05, * p<0.1
Robust standard errors in parentheses*** p<0.01, ** p<0.05, * p<0.1

15. The results in Table 6 suggest that investment grade status can be modeled parsimoniously with a handful of regressors.10

  • In line with other studies (e.g. Afonso (2003)), the results show that the level of debt matters for determining investment grade. However, the findings suggest that rating agencies do distinguish between types of debt. They tend to see risk in high public debt indicators, but do not seem to assign a significant weight to private external debt. Furthermore, rating agencies seem to attach greater risk to external public debt than to domestic public debt, with the coefficients of the former being more than 2½ times bigger.
  • The political risk index was found to be significant and positively related to the investment grade rating. This is in line with the findings of Mellios and Paget-Blanc (2006), though contrary to the findings of Archer et al. (2007) who used as regressors different proxies for democratic rule (such as executive party tenure, undivided government, and election cycles).
  • Exports to GDP and broad money to GDP were also found to be significant. The positive effect of exports on investment grade is line with the findings by Rowland (2004). The significant and positive impact of broad money on the determination of investment grade is a new finding as previous studies had not included measures of financial depth as regressors.

16. The random effects model provides a good fit, as it performs well in predicting both investment grade status and “switches”. Using the overall in-sample probability of being investment grade (0.4) as the cut-off point, the number of times the models correctly predicts zeroes and ones was computed. The model correctly classifies 86 percent of all observations, with Type I error (failing to predict investment grade status) of 8.9 percent, and Type II error (failing to predict speculative grade status) of 5.2 percent. In terms of “switches”, the model correctly predicts 6 out of 9 downgrades (68 percent), failing to predict the downgrades of Korea, Malaysia and Thailand—which happened in the context of the Asian crisis and were reversed within two years. The model also predicts correctly 11 of the 17 upgrades in the sample (65 percent). Among the upgrade cases that the model fails to predict are those of Mexico and South Africa (which are predicted with a three year lag) and those of Colombia and Uruguay (which lost their investment grade status within 4–5 years of the upgrade). Figure 1 depicts the behavior of the regressors in some of the cases that experienced an upgrade to investment grade during the sample period.

Figure 1.Emerging Markets: Value of Explanatory Variables at Time of Upgrade to Investment Grade and Five Years Earlier 1/

1/ Domestic debt to GDP was not graphed as there was not a significant change in this variable for the particular country examples shown.

17. An analysis of marginal effects provides further insights on the impact of the regressors on the probability of investment grade status. Table 7 shows the average partial effects11 for alternative levels of debt, openness, financial development and political risk. As expected in a binomial logit model, the marginal effects of each variable are nonlinear and are therefore larger for values in the middle of the distribution. For example, a 10 percentage point decrease in external public debt to GDP would increase the probability of investment grade by 16 percentage points on average, with the effect being almost double for those countries in the second quartile of the distribution. Similarly, a 10 point increase in the political risk index (implying an improvement in risk perception) would increase the probability of investment grade by 10 percentage points on average, with somewhat greater impact for countries in the middle of the distribution. Table 7 also shows that there are investment grade observations throughout the distribution of each variable, implying that there is not necessarily a minimum threshold for each of the regressors.

Table 7.Average Partial Effects on the Probability of Investment Grade Status
Average

Partial Effects
APE by Quartile
1234
External public debt to GDP, 10 percentage point decline0.160.090.300.200.04
Domestic public debt to GDP, 10 percentage point decline0.060.050.070.060.05
Political risk index, 10 point increase0.100.050.120.140.08
Exports to GDP, 10 percentage point increase0.050.040.080.080.02
Broad money to GDP, 10 percentage point increase0.030.020.030.040.02
Memorandum items:
Quartile cut-off points
External public debt to GDP26122137120
Domestic public debt to GDP26101933141
Political risk index6763697387
Exports to GDP36233245120
Broad money to GDP59344567279
Percentage of investment grade observations in each quartile
External public debt to GDP56.427.78.27.7
Domestic public debt to GDP39.527.217.415.9
Political risk index7.221.024.647.2
Exports to GDP12.817.427.742.1
Broad money to GDP11.326.233.828.7
Percentage of speculative grade observations in each quartile
External public debt to GDP7.625.336.131.0
Domestic public debt to GDP16.123.728.831.3
Political risk index35.428.824.111.7
Exports to GDP32.930.421.215.5
Broad money to GDP32.325.919.022.8

E. Implications for Colombia

18. The above results suggest that Colombia’s efforts to increase the likelihood of a rating upgrade should focus on improving its debt indicators. Colombia’s public debt indicators are not too far away from the median of investment grade emerging market countries. However, there is a significant difference in terms of the other indicators between Colombia and other emerging markets (see Figure 2). For example, Colombia’s ratio of exports to GDP (17.7 percent) is significantly below that of investment grade countries, and is in the lowest quartile for emerging market countries overall. Similarly, Colombia’s political risk index and broad money to GDP are well below those for investment grade countries, and also in the lower quartiles of emerging market countries. While these indicators have been improving in recent years, further progress is likely to be gradual and not directly linked with macroeconomic policies. In contrast, a strong process of fiscal consolidation could result in a steady reduction in debt levels.

Figure 2.Colombia: Factors Affecting Investment Grade Status

19. A marginal effects analysis for Colombia shows that reducing public debt, in particular external public debt, would increase substantially the probability of an upgrade. Based on the marginal effect analysis reported in Table 7 and holding other values at their 2008 levels, a decline in Colombia’s public debt ratio to 20 percent (including by halving its external public debt to 6 percent of GDP) would increase the probability of attaining investment grade status to 40 percent (the sample cut-off point).

F. Summary and Conclusions

20. The paper finds that investment grade ratings by the three major credit agencies can be explained by a small number of variables. The panel random effects framework identifies a set of five core variables that are relevant for the determination of investment grade status, in particular external public debt, domestic public debt, political risk, exports to GDP, and financial depth. Overall, the specification correctly predicts 86 percent of investment grade status of all observations, and two thirds of the upgrades and downgrades from and to investment grade.

21. The findings suggest that Colombia’s efforts to increase the likelihood of an upgrade to investment grade should focus on a faster pace of debt reduction. While Colombia’s public debt figures are not too far away from those of emerging market investment grade countries, it faces weaknesses on other determinants that appear to have significant weight in credit rating agencies decisions, namely political risk, financial depth, and exports to GDP. A stronger process of fiscal consolidation that results in a significant decline in public sector debt, in particular external public debt, could help compensate these structural weaknesses and improve the prospects for an upgrade in the near term.

References

    Archer, Candace C., GlenBiglaise, and KarlDeRouenJr., 2007, “Sovereign Bonds and the Democratic Advantage: Does Regime Type Affect Credit Rating Agency Ratings in the Developing World?” International Organization: 61, Spring2007, pp. 341–365.

    • Search Google Scholar
    • Export Citation

    Afonso, Antonio, 2003, “Understanding the Determinants of Sovereign Debt Ratings: Evidence for the Two Leading Agencies,”Journal of Economics and Finance,Vol. 27, Number 1, Spring2003.

    • Search Google Scholar
    • Export Citation

    Afonso, Antonio, P.Gomes and P.Rother, 2007, “What Hides Behind Sovereign Debt Ratings?”European Central Bank Working Paper Series No. 711.

    • Search Google Scholar
    • Export Citation

    Borensztein, Eduardo, KevinCowan, and PatricioValenzuela, 2007, “Sovereign Ceilings “Lite:”? The Impact of Sovereign Ratings on Corporate Ratings in Emerging Market Economies,”IMF Working Paper 07/75 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

    Cantor, Richard, and F.Packer, 1996, “Determinants and Impact of Sovereign Credit Ratings,” FRBNY Economic Policy Review, October.

    Mellios, Constantin, and E.Paget-Blanc, 2006, “Which Factors Determine Sovereign Credit Ratings?”The European Journal of Finance,Vol. 12, No. 4 (June), pp. 361–377.

    • Search Google Scholar
    • Export Citation

    Mulder, Christian, and R.Perrelli, 2001, “Foreign Currency Credit Ratings for Emerging Market Economies:”IMF Working Paper 01/191, November2001 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

    Rowland, Peter, 2004, “Determinants of Spread, Credit Ratings and Creditworthiness for Emerging Market Sovereign Debt: A Follow-up Study Using Pooled Data Analysis,” Borradores de Economía No. 296, Banco de la República de Colombia.

    • Search Google Scholar
    • Export Citation

    Rowland, Peter, and J.Torres, 2004, “Determinants of Spread, Credit Ratings and Creditworthiness for Emerging Market Sovereign Debt: A Panel Data Study,” Borradores de Economía No. 295,Banco de la República de Colombia.

    • Search Google Scholar
    • Export Citation

    Wooldridge, Jeffrey M., 2001, Econometric Analysis of Cross Section and Panel Data. Cambridge, Massachusetts: MIT Press Books, edition 1, Vol. 1, December.

    • Search Google Scholar
    • Export Citation
Appendix 1. Explanatory Variables

Macroeconomic variables

  • Per capita income in U.S. dollars. Higher per capita income tends to suggest a larger potential tax base and a greater ability to repay debt. It also serves as a proxy for the level of economic development, which might influence default risk.
  • Real GDP growth and potential GDP growth. Higher economic growth tends to decrease the relative debt burden and may help in avoiding insolvency.
  • Inflation rate. A low inflation rate reveals sustainable monetary and exchange rate policies. It can also be seen as a proxy of the quality of economic management.
  • Unemployment rate. A country with low unemployment tends to have more flexible labor markets making it less vulnerable to changes in the global environment.

External sector variables

  • Exports to GDP. A higher ratio suggests a greater capacity to obtain hard currency to repay foreign currency denominated debt. Unlike most previous studies, where exports are included only as a metric for external debt, this paper introduces it as an independent regressor.
  • External current account to GDP. A large current account deficit suggests a high dependence on foreign capital, which can be a source of risk to macroeconomic stability.
  • Private and public external debt to GDP. The higher the external indebtedness, the higher the risk of fiscal or balance of payments stress. In contrast to existing literature, this paper distinguishes between private and public external debt to allow for differences in the weight assigned by rating agencies to each one.
  • Net international reserves to GDP. The higher the ratio, the more resources are available to service foreign debt. It reduces a country’s vulnerability to liquidity shocks.

Government sector variables

  • Primary balance to GDP. A low primary balance indicates that the government lacks the ability or the will to increase taxes to cover current expenses. A weak fiscal position also implies a higher likelihood that external shocks result in a default. In contrast to previous studies, the regressions use the primary balance instead of the overall balance to avoid possible endogeneity with the credit rating.
  • Public debt to GDP. The higher the debt burden, the larger the transfer effort the government will have to make over time to service its obligations, and therefore a higher risk of default. In contrast to existing literature, this paper distinguishes between domestic and external public debt to allow for differences in weights.

Financial depth

  • Broad money to GDP. Countries that have access to a deep and diversified pool of finance are in a better situation than those whose private savings are low and whose financial system is repressed. For this reason, financial depth is a useful indicator of government financial flexibility. High levels of financial intermediation, as proxied by broad money to GDP, can be associated with a greater capacity to sustain a given domestic debt burden. Existing literature has not included this variable.

Other regressors

  • Political risk. Rule of law and respect for property rights provide confidence that political (and civil) institutions have a strong commitment to honoring financial obligations. As summarized by the International Country Risk Guide (ICRG), the political risk index is used as a proxy to measure a country’s willingness to repay.
  • Default history. A country’s default history tends to influence its rating. A binary variable is set equal to 1 when the sovereign has defaulted on its external debt at least once in the previous ten years.
  • Regional dummies. Dummies are included for the Latin America and the Caribbean, Europe, and East Asia.
Appendix 2. Robustness Analysis

The results reported in Table 6 were robust to alternative measures of the explanatory variables. In order to compare the results with those of previous studies, the model was estimated using debt scaled by exports as a regressor (as opposed to having debt and exports to GDP enter separately in the regression). Table A1 (column A) shows that the coefficient for the debt to exports ratio was found to be negative and significant, without altering the sign or significance of the other coefficients; however, this specification deteriorates the fit of the model. The political stability index of the World Bank Governance Indicators was used as an alternative measure to the ICRG political risk indicator. This coefficient was positive and significant, did not affect the sign or significance of the other coefficients, but the fit of the model deteriorated (column B). Alternative measures of financial depth, in particular net credit to the private sector and market capitalization of listed companies as a percent of GDP, did not prove to be significant, implying that broad money to GDP is the variable that best captures rating agencies’ concerns regarding financial depth.

Table A1.Alternative Specifications
Random effects
AB
External public debt to GDP-0.41***
(0.08)
External public debt to exports-0.04***
(0.01)
Domestic public debt to GDP-0.10***-0.10***
(0.02)(0.04)
Exports to GDP0.07***
(0.03)
Broad money to GDP0.04***0.07***
(0.01)(0.03)
ICRG political risk index0.14***
(0.05)
WGI political stability index3.90***
(1.15)
Constant-8.11**-7.68**
(3.45)(3.88)
Observations540553
Number of ccode3741
Robust standard errors in parentheses.*** p<0.01, ** p<0.05, * p<0.1
Robust standard errors in parentheses.*** p<0.01, ** p<0.05, * p<0.1

The model also yielded similar results when the dependent variable was defined based on the investment grade ratings of each individual credit agency. The results in Table A2 suggest that S&P, Fitch and Moody’s broadly share the same criteria for determining investment grade, although they weigh some variables differently. The broad similarity in criteria is not surprising given that the agencies agree on the classification of countries across investment/speculative grades in 94 percent of the cases. In all cases, the agencies attribute more weight to external public debt over domestic public debt. However, while Moody’s seems to put more weight on the political risk variable than the average, this variable was not significant for S&P. The ratio of exports to GDP was significant for Moody’s and with a lower coefficient for S&P, and broad money to GDP was significant for S&P and with a lower coefficient for Fitch.

Table A2.Regression Results by Rating Agency
Moody’sStandard and Poor’sFitch Ratings
VARIABLESABCDEF
GDP per capita (US$)0.000.00*0.00
(0.00)(0.00)(0.00)
Potential GDP growth0.18-1.88**-1.57**
(0.51)(0.75)(0.74)
Inflation0.00-0.18*-0.11**-0.11
(0.01)(0.10)(0.06)(0.08)
Unemployment-0.55**0.170.06
(0.23)(0.21)(0.22)
External public debt to GDP-0.39***-0.29***-0.55***-0.44***-0.54***-0.26***
(0.09)(0.08)(0.11)(0.08)(0.17)(0.05)
Domestic public debt to GDP-0.15**-0.06*-0.20***-0.16***-0.32***-0.13***
(0.06)(0.03)(0.07)(0.04)(0.12)(0.03)
Primary balance to GDP-0.97***0.330.21
(0.35)(0.26)(0.25)
ICRG political risk index0.36***0.28***-0.070.020.15***
(0.13)(0.07)(0.12)(0.12)(0.06)
Private external debt to GDP0.060.05**0.06-0.04
(0.04)(0.02)(0.05)(0.05)
Net international reserves to GDP0.000.160.43**0.17***
0.14(0.15)(0.21)(0.06)
Current account balance to GDP-0.040.02-0.05
(0.14)(0.19)(0.19)
Exchange rate volatility-0.23**0.040.03
(0.11)(0.12)(0.13)
Broad money to GDP0.13**0.20***0.15***0.27**0.06**
(0.06)(0.06)(0.03)(0.12)(0.02)
Exports to GDP0.25**0.10***0.030.06**0.13
(0.11)(0.03)(0.08)(0.02)(0.13)
Latin America and Caribbean dummy-0.46-0.60-5.13-2.78*
(4.97)(4.10)(5.89)(1.55)
Europe dummy2.698.64**8.02**-1.53
(4.89)(4.32)(3.21)(4.76)
East Asia dummy-9.13-4.02-24.70-5.81**
(5.82)(5.61)(15.88)(2.82)
Default dummy-2.55-10.38-4.61
(6.48)(8.89)(6.50)
Constant-26.42**-17.21***11.990.474.65-7.61*
(10.34)(5.05)(11.00)(2.55)(12.38)(4.06)
Observations479595431487404436
Number of ccode354135383335
Robust standard errors in parentheses.*** p<0.01, ** p<0.05, * p<0.1
Robust standard errors in parentheses.*** p<0.01, ** p<0.05, * p<0.1

Beyond the set of core regressors, credit agencies appear to rely on a few additional variables. In the case of Moody’s, private external debt was found to have a positive significant effect on the investment grade rating, reflecting the view that steady private access to international markets serves as an indicator of market confidence in the soundness of the corporate sector (column B). In the case of S&P, inflation was found to have a negative significant effect, which serves as a proxy for the quality of macroeconomic management (column D). Finally, in the case of Fitch, net international reserves were found to have a significant and positive effect, while the export variable was not significant (column F).

1

Prepared by Laura Jaramillo.

3

Borensztein et al. (2007) find that sovereign ratings have a significant effect on private ratings even after controlling for country specific macroeconomic conditions and firm-level performance indicators.

4

The ratings do not differ significantly across the three agencies. Investment/speculative grade status coincided across the three rating agencies for 94 percent of all observations in the sample.

5

The pooled estimation does not control for unobserved country effects, while the fixed effects logit model has the disadvantage that only countries where the dependent variable “switches” (from 0 to 1 and vice versa) can be included in the estimation—which in this case would lead to a sizeable number of cases being dropped. In addition, the fixed effects estimations cannot assess the impact of non-time varying country characteristics.

6

See Appendix 1 for a description of the explanatory variables and rationale for their inclusion in the regression analysis.

7

The sample of countries consists of: Argentina, Brazil, Chile, Colombia, Costa Rica, Dominican Republic, Ecuador, El Salvador, Guatemala, Jamaica, Mexico, Panama, Peru, Uruguay, Venezuela, Bosnia & Herzegovina, Bulgaria, Croatia, Czech Republic, Estonia, Hungary, Iceland, Israel, Latvia, Lithuania, Poland, Romania, Russia, Serbia, Turkey, Ukraine, China, Indonesia, Korea, Malaysia, Philippines, Thailand, Vietnam, Egypt, India, Jordan, Kazakhstan, Lebanon, Morocco, Pakistan, South Africa, Sri Lanka, and Tunisia.

8

A comparison of means only could be misleading in the presence of large outliers.

9

Variables that did not reveal any explicative power were dropped based on Wald tests. The restricted models are robust to alternative exclusion procedures. Furthermore, the variables found to be significant in the unrestricted model generally remain significant with the same sign in the restricted model.

10

Although rating agencies may assign substantial weight to other factors in determining specific rating assignments to a particular country at a given point in time, no systematic relationship between those variables and investment grade status was detected in the sample.

11

Calculated as the partial effects averaged across the population. See Wooldridge (2002).

Other Resources Citing This Publication