Over the past decade or so, a number of low-income developing countries (LIDCs)2 have issued sovereign bonds in the international capital markets, driven in part by African frontier markets.3 Since 2005, some 14 LIDCs have issued international sovereign bonds, 10 of which are in Sub-Saharan Africa. In 2013, LIDCs issued sovereign bonds amounting to US$4 billion, and this trend continued in 2014, with Côte d’Ivoire, Ethiopia, Ghana, Kenya, Senegal, Vietnam, and Zambia having issued bonds totaling about US$8 billion (Table 1). Given the declining trend of aid flows4, sovereign bond issues could contribute to the financing of investment projects.
|Issuer||Issue Date||Yield at issue||Tenor||Amt (USD mn)||Spread (bps)||S&P rating at issue|
|Republic of the Congo||12/7/07||8.77||22||480||458.0||Not rated|
|Côte d’Ivoire||4/8/10||17.35||23||2330||393.0||Not rated|
|Côte d’Ivoire||11/15/12||20||187||Not rated|
What determines the ability of LIDCs to issue bonds in the international capital markets? What are the factors that influence the spreads on these bonds? What can LIDCs learn from the experience of emerging markets and developing economies (EMDEs) that have market access? To address these questions, this paper examines the experience of 104 EDMEs during the period 1995-2013, including 49 that had issued sovereign bonds at least once in the international capital markets, with the remainder having never issued.
This paper builds on an extensive literature on access to international capital markets for EMDEs and contributes to this work in two ways.5 First, the analysis of sovereign bond issuance (SBI) by EMDEs is extended by jointly estimating the factors that influence the spread on these bonds. Second, the paper captures the experience in recent years when several LIDCs issued international sovereign bonds for the first time, thus extending the analysis on emerging markets by Eichengreen and Mody (2000) and Gelos and others (2011), as well as work done on first-time issues by EMDEs (Gregorian, 2003; Thomas, 2009; IMF, 2013; Guscina and others, 2014; Gueye and Sy, 2015).6
Simple comparisons of the data on EMDEs that issued sovereign bonds in global capital markets (issuers) and those that did not issue (non-issuers) provides useful insights. Issuers typically have higher per capita real GDP, deeper financial markets, stronger external positions, greater government effectiveness,7 and are more likely to have had an IMF program in place over the previous 3 years (compared to non-issuers). Among countries that have issued, we separate the sample between regular issuers (those that have issued in 5 or more years during 1995-2013) and occasional issuers. We find that regular issuers have higher per capita real GDP, stronger external reserve positions, and more effective governments (than occasional issuers).
The econometric analysis indicates that an EMDE is more likely to issue sovereign bonds in the global market when it is large and has higher per capita GDP (than those that are smaller and less developed), and has sound macroeconomic fundamentals (measured by the external debt-to-GDP ratio and fiscal position) and strong government effectiveness. Spreads on sovereign bonds are lower for countries with strong external position (reflected in external reserve position and current account balance) and fiscal positions, as well as robust economic growth and government effectiveness. There is evidence that once an EMDE issues for the first time, there is a high likelihood that it will reissue in the future and at lower spreads. We find a robust catalytic role of IMF programs for market access. With regard to global factors, we find that bond spreads are lower in periods of declining global market volatility (measured by VIX), though the US rate does not seem to have a robust relation with either issuances or spreads.
The rest of the paper is organized as follows. Section II discusses the data and some stylized facts about sovereign bond issuances over the period 1995-2013. Section III presents the empirical approach and the associated econometric results, while the section IV concludes.
II. Stylized Facts
A. The Dataset
The analysis is based on a sample of a set of 104 EDMEs during 1995-2013. Forty-nine of these countries have issued international sovereign bonds at least once since 1995.8 The other 55 countries are used as a “control” group; these are countries with per capita income level in the same range as the ones that issued bonds (i.e., countries with a GDP per capita lower than US$10,000 in 19959), but which did not issue bonds in global capital market. In contrast to some of the existing literature (Thomas 2009; Gelos and others 2011), the paper focuses on international bonds, excluding commercial bank syndicated loans to national governments.10 The motivation for focusing on sovereign bonds is twofold. First, the volume of syndicated loans is smaller than from sovereign bonds (Figure 1)11. Second, looking exclusively at sovereign bonds—for which information on size and spread at issue is available—allows a joint analysis of factors associated with the likelihood to issue and the associated price.12
Figure 1:International Sovereign Bonds and Syndicated Loans to the Public Sector
Notes: Based on annual data for 104 emerging markets and developing economies, over 1995-2013. GDP-weighted averages. Syndicated loans to the public sector include the central government and state/provincial authorities as borrowers.
Source: Dealogic Loan Analytics and Bloomberg.
Data on sovereign bonds are from Bloomberg and include information on the date of issuance, maturity, amount, yield and spread at issue. Data on macroeconomic and institutional variables are from the World Economic Outlook database and the World Development Indicators.
The set of domestic factors included in the model as potential determinants of sovereign bonds follows the existing empirical literature on market access.13 The analysis takes into account differences in economic development and macroeconomic performance by including per capita real GDP, GDP growth, inflation, current account balance as a ratio to GDP, international reserves in months of imports, fiscal balance as a ratio to GDP, and external public and publicly guaranteed debt as a ratio to GDP. The importance of the institutional setting is taken into account by incorporating institutional strength in the analysis, measured by an index of government effectiveness.
The paper also considers participation in IMF lending programs as a potential determinant of sovereign bond issuance and spread (Mody and Saravia, 2006). On the one hand, IMF-supported programs may act as a seal of approval, reassuring investors and catalyzing private capital flows; on the other hand, IMF-supported programs could generate moral hazard and signal high financial and sovereign risks (Bird, 2007). The hypothesis that sovereign bond issuance could be persistent is tested by including in both equations a dummy variable equal to one for countries that have issued at least once in the previous three years. Finally, the paper considers country size, measured by population, since it could affect the ability to borrow because of the fixed costs associated with market access and the potential punishment of a default (Faria and others, 2010; Gelos and others, 2011).
To mitigate the influence of extreme values, GDP growth, inflation, private credit, total debt, fiscal balance, current account, reserves have been winsorized at the 1st and 99th percentiles.14 Summary statistics, definition and sources of all variables are provided in Table 3.
|Bosnia & Herzegovina||0||Guyana||0||Rwanda||1|
|Burkina Faso||0||Jamaica||1||Sri Lanka||1|
|Burundi||0||Jordan||1||St. Kitts and Nevis||0|
|Cambodia||0||Kenya||0||St. Vincent & Grenadines||1|
|Comoros||0||Liberia||0||Sâo Tomé and Príncipe||0|
|Congo, Republic of||1||Libya||0||Tajikistan||0|
|Côte D’Ivoire||1||Macedonia, FYR||1||Togo||0|
|SBI (0/1)||Dummy equal to one if the country issued at least a sovereign bond in the year, and zero otherwise.||Bloomberg||0.109||0.312||0||1||1,749|
|SPREAD||Spread of the sovereign bond over the EMBI. If a country issues more than one bond in the year, SPREAD is the simple average of the single spreads.||Bloomberg||309.590||171.972||21.600||825.000||191|
|Real per capital GDP||Real per capita GDP in USD, in logarithms.||WDI||7.172||1.119||4.775||9.445||1,749|
|GDP growth||Real GDP growth rate.||WEO||4.329||3.621||−11.967||21.060||1,749|
|Inflation||Consumer rice index, annual percent change.||WEO||17.401||76.563||−4.148||1265.734||1,749|
|PPG external debt (%GDP)||Public and publicly guaranteed external debt (% of GDP).||WDI||58.491||45.712||2.203||304.203||1,749|
|Fiscal balance (%GDP)||General government net lending/borrowing (% of fiscal year GDP).||WEO and country reports||−2.287||4.223||−16.403||20.123||1,749|
|Current account (%GDP)||Balance on current account (% of GDP)||WEO||−5.369||8.086||−34.795||21.405||1,749|
|Reserves (in months of imports)||International reserves in months of imports||WDI, WEO and country reports||4.698||4.257||0.031||30.307||1,749|
|IMF program in the previous 3 year||Dummy equal to one if the country signed at least a loan agreement in the previous 3-year period||IMF historical data set||0.150||0.199||0||1||1,749|
|Resource rich dummy||Dummy equal to one for resource rich countries and zero otherwise.||IMF (2012)||0.200||0.400||0||1||1,749|
|Population (logs)||Total population, in logarithms. Government effectiveness index, ranging||WDI||15.516||1.843||10.642||20.923||1,749|
|Government effectiveness||from approximately −2.5 to 2.5, with higher values corresponding to better outcomes.||WGI||−0.482||0.599||−1.929||1.278||1,614|
|Total debt (%GDP)||Public debt (% of GDP).||WDI||65.306||50.900||5.474||328.583||1,747|
|Private credit (%GDP)||Domestic credit provided by banking sector (% of GDP).||Abbas and others (2011)||36.479||30.205||−40.464||149.421||1,745|
|Aid (% GDP)||Net official development assistance and official aid received (% of GDP)||WDI||7.532||9.032||−0.016||100.386||1,708|
|US rates||Yield on the 10-year US Treasury notes||Federal Reserve||4.320||1.310||1.800||6.570||1,749|
|VIX index||Chicago Board Options Exchange Volatility Index||CBOE||21.316||6.088||12.389||32.693||1,749|
B. Descriptive Analysis
Sovereign bond issuances by EMDEs in global markets have steadily increased over time, moving from an annual average of 8 per year in the late 1990s, to 12 in the 2000s and 20 since 2010 (Figure 2). During 1995-2013, LIDCs issued 20 sovereign bonds (8.4 percent of all SBIs in the sample) and, with the exception of Moldova, started issuing more recently (Table 1).15 Global sovereign bond issuances are concentrated geographically: 110 issuances are from Europe and Central Asia, 89 from Latin America and the Caribbean, 16 from Sub-Saharan Africa, 11 from Middle East and North Africa, 6 from South Asia and 5 from East Asia and Pacific region.
Figure 2:International Sovereign Bond Issuances and Global Conditions, 1995-2013
Notes: Based on annual data for 104 developing countries (49 issuers), over 1995-2013.
Spreads at issue declined somewhat during 2005-09, underpinned by strong global conditions, and picked up thereafter, possibly reflecting the expanding role of frontier markets. The latter were able to place international bonds at higher spreads than the average EMDE (Figure 3). The average primary spread for the 18 LIDCs that issued international bonds was 434 basis points, while the average spread of the bonds issued by EMDEs was 310 basis points (the difference between the two sample averages being statistically significant).
Figure 3:The Distribution of Bond Spread at Issue, 1995-2013
Notes: Based on data for 202 SBIs (by 49 countries), over 1995-2013.
In line with developments in the 1980s and 1990s, sovereign bond issuances have been influenced by global factors. The number of bond issuances slumped significantly during the peak of the global financial crisis, when the volatility of markets increased substantially and investors retreated from risky-asset classes. Starting in 2010, as risk appetite improved and global interest rates further declined, international investors, inclined to diversify their asset portfolio, resumed their search for yield in a low-interest rate environment and sovereigns took advantage of low global interest rates to finance themselves in international markets. As a result, bond issuances picked up considerably (Figure 2).
Before formally testing the correlation between country-specific characteristics in the run-up to a bond issuance in an econometric model, we undertake some simple cross-country analyses, aimed at determining possible systematic and persistent differences across countries that issued bonds regularly or occasionally and those that did not issue. The analysis follows Gelos and others (2011) and divides the sample into three sub-groups:
- No-issuance countries (NI): 55 countries that did not issue a sovereign bond in the sample period 1995-2013;
- Occasional issuers (OI): 31 countries that issued sovereign bonds for less than 5 years;
- Regular issuers (RI): 18 countries that issued sovereign bonds for 5 or more years over the time period 1995-2013.16
The analysis shows that access to international capital markets is correlated with country characteristics measuring the level of economic, financial and institutional development; fiscal and external balances; and external reserve position. Table 4 reports the average values of the macroeconomic and institutional indicators for the whole sample and for the three groups, over the entire sample period. A series of mean-comparison t-tests are conducted to assess whether differences across groups are statistically significant. The cross-country comparison highlights some interesting patterns.
|Variable||All countries||No-issuance countries||Occasional issuers||Regular issuers||Test of equality of means|
|Per capita real GDP||7.20||104||6.72||55||7.40||31||8.31||18||***||***||***|
|Private credit (% GDP)||37.43||104||31.38||55||40.86||31||49.99||18||**|
|PPG External debt (% GDP)||58.84||104||64.23||55||54.73||31||49.46||18|
|Fiscal balance (% GDP)||−2.30||104||−2.18||55||−2.14||31||−2.97||18|
|Current account (% GDP)||−5.69||104||−6.88||55||−4.91||31||−3.43||18||**|
|Reserves (in months of imports)||4.81||104||5.27||55||3.86||31||5.08||18||*|
|IMF program in the previous 3 year||0.14||104||0.12||55||0.15||31||0.17||18||*|
- In comparison with occasional and regular issuers, countries that did not issue at all have lower per capita real GDP, are less financially developed and have a worse external position, but, on average, grow at a faster rate.
- Sovereigns that issued bonds have more effective governments and are more likely to have had an IMF program than countries that did not issue.
- There are differences between occasional and regular issuers: the latter have higher per capita real GDP, stronger institutions, and more external reserves.
- There are no statistically significant differences across the three country groups with regard to inflation and fiscal position.
Finally, Figure 4 shows the correlations between primary spreads and country characteristics in the run-up to issuing sovereign bonds. Macroeconomics variables are measured as the average of the three-year period prior to each issuance. The charts show that lower primary yield spreads are associated with faster growth, higher per capita real GDP, stronger institutions, current account surpluses, and higher international reserves. By contrast, there is limited evidence to a significant simple correlation between primary spreads and the fiscal position, either considering the government balance or external public debt.
Figure 4:Spreads and Country Performance Before Issuance
Notes: Based on data for 185 SBIs (by 49 countries), over 1995-2013.
III. The empirical Analysis
A. Empirical Specification
Access to international debt markets can be modeled as a two-step process. The first relates to the willingness of sovereigns to access global markets and of the latter to supply funds. The second concerns the price at which demand and supply are settled. If unobserved factors that determine bond issuance also affect spreads, a standard linear model for the determinants of sovereign spreads at issue would be biased because of sample selection, given that the dependent variable (the spread in the primary market) is observed only for a non-random subset of country-year observations. Empirically, this problem can be addressed by specifying a sample selection model à la Heckman (1979). According to this procedure, in the first stage (selection equation), the probability of sovereign bond issuance is estimated by a standard probit regression, and in the second stage (outcome equation) the primary spread is a linear function of the set of variables and of the inverse Mills ratio calculated from the first-stage regression, which corrects for sample selection bias.
Access to international capital markets is likely to be the result of demand and supply factors: disentangling these two would require an exogenous shock in the demand or in the supply schedule. Focusing exclusively on EMDEs (many of which are LIDCs) should help minimize the cases of voluntarily absence from the market (lack of demand), given that these countries generally need large amounts of external funds to finance domestic investment (Eaton and Gersovitz, 1980; Gelos and others., 2011). Countries, however, could still self-select out of international credit markets, especially in the case of sovereigns with sufficient access to grants and concessional loans. We take this possibility into account by controlling for participation in IMF-supported lending programs and, as robustness check, we control also for the amount of aid inflows. In addition, countries could base their demand for international bonds on expected borrowing costs. While it is difficult to fully control for these costs, especially for first-time issuers for whom there are no secondary market bond spreads, the information on expected borrowing costs is indirectly taken into account by inclusion of a comprehensive set of domestic and global controls in the selection equation. However, in the absence of an identification strategy, the empirical exercise focuses on factors associated with sovereign bond issuance without implying causality.
In the selection equation, the likelihood of a sovereign bond issuance (SBI) in the global market by a country is a function of: 1) the 10-year US treasury notes yield and the VIX index17 as measures of global liquidity and volatility (GLOBAL); 2) a set of j time-varying domestic factors (DOMESTIC); and 3) country size (POPULATION). The latter is measured by the logarithm of population, which is taken as excluding restriction under the assumption that it affects only the likelihood to issue but not the primary spreads, given that, in the presence of fixed costs for borrowing, smaller countries will access markets less frequently than larger countries (Faria and others 2010; Gelos and others 2011). The selection equation is as follows:
where the dependent variable is a binary indicator that takes the value of one when country i issued at least one sovereign bond in year t, and zero otherwise, and Φ(.) is the normal cumulative distribution function. In the outcome equation, the primary spread on the sovereign bond (SPREAD) issued by country i in year t (observed exclusively in country-year observations where SBI = 1)18 is a linear function of the same set of global and domestic factors, plus the inverse Mills ratio (IMR) which takes account of the selection bias:
In both equations, global conditions includes the contemporaneous values of the 10-year US treasury notes yield and the VIX index, which may affect both the supply of funds by investors and demand for financing by issuing countries. The set of domestic factors includes indicators of overall economic performance, past issuances and participation in IMF-supported programs, external sector position, liquidity, fiscal position, and government effectiveness, as discussed in Section II.A.19 All domestic explanatory variables are measured as averages in the 3-year period prior to the year of issuance. Taking a 3-year period rather than measuring independent variables in t-1 further mitigates possible endogeneity concerns (Gelos and others, 2011) and minimizes the incidence of outliers. In addition, this choice assumes that market access does not depend on domestic macroeconomic conditions in place just the year before the issuance, but it is influenced by what happened in the run-up of the issuance. Given that the inclusion of the dummy variable for lagged issuance—which accounts for persistence in market access—may create problems making the model dynamic, we start showing results for a model that excludes the dummy for past sovereign bond issuance.
The model also includes regional20 dummy variables to account for regional fixed effects and spillovers (Dell’Erba and others, 2013). The baseline analysis takes into account the role of global liquidity and volatility including the 10-year US treasury notes’ yield and the VIX index. Alternatively, we replace these two variables with year-fixed effects, which capture global (common) time-variant shocks and allow focus on the role of domestic factors.
The maximum likelihood estimates of the 2-equation system are shown in Table 5, which reports the coefficients and the associated robust standard errors (correcting for heteroskedasticity). For each model specifications, the first column reports the estimates of outcome equation (2) and the second refers to the selection equation (1).
|Real per capita GDP||−121.760***||0.873***||−82.269***||0.707***||−77.397***||0.573***||−12.059||0.480***||−9.217||0.489***|
|PPG external debt (%GDP)||0.323||−0.004**||0.558||−0.005**||0.512||−0.004**||1.033**||−0.005**||1.220***||−0.005**|
|Fiscal balance (%GDP)||9.699*||−0.085***||−88.350***||0.356**||−87.485***||0.325**||−85.466***||0.336**||−72.149**||0.354**|
|Current account (%GDP)||−5.631***||−0.008||−5.353***||−0.007||−4.740**||−0.007||−9.020***||−0.005||−10.038***||−0.005|
|Reserves (in months of imports)||−11.813***||−0.027*||−11.796***||−0.025*||−11.273***||−0.014||−7.975**||−0.021||−7.799***||−0.021|
|IMF program in the previous 3 year||94.664**||0.594**||92.074**||0.607**||92.465**||0.568**||19.193||0.619**||−0.289||0.633**|
|Fiscal balance (%GDP) * real pc GDP||12.006***||−0.056***||11.507***||−0.049***||10.459***||−0.050***||8.964**||−0.052***|
|SBI in the previous 3 year||−74.553**||0.727***||−70.154**||0.652***||−72.560***||0.685***|
|Wald test (p-value)||0.018||0.034||0.014||0.025||0.020|
|Test area FE||0.000||0.000||0.000||0.000||0.000|
|Test year FE||1.000||1.000||1.000||1.000||0.000|
Overall, the analysis shows that sovereign bond issuance and spreads depend on global conditions, macroeconomic performance (the fiscal position and indicators of liquidity), and institutional strength in the run-up to bond issuances. The results indicate that country size is a significant predictor of sovereign bond issuance, as there would be fixed costs for borrowing through issuance (Gelos and others, 2011), validating the exclusion restriction.21 Regional dummy variables and global factors are statistically significant, suggesting the presence of spillovers in market access and the relevance of global demand- and supply-side factors in the sovereign bond market. As in Eichengreen and Mody (2000), the coefficient ρ on the estimated inverse Mills ratio in the outcome equation is negative and statistically significant,22 suggesting that participation in the international debt market is actually non-random, and that the unobservable political, institutional and economic characteristics of countries affecting the likelihood SBI also inversely influence bond spreads.
The baseline specification (columns 1-2) suggests that global factors matter for market access. When market volatility increases, so do primary spreads: a one standard deviation of the VIX index corresponds to a 29 basis points change in spread. The effect of global interest rates is twofold: lower US rates are associated with a higher probability that countries issue sovereign bonds, but also with higher spreads. The estimated coefficients indicate that a 1 percentage point reduction in the US rate translates into a 29 basis-point increase in the primary spread. This result is consistent with that of Eichengreen and Mody (2000) and, as suggested when discussing the stylized facts (see above, Section II.B), it is likely to be the result of frontier markets accessing international bond markets in a period of declining global interest rates, but placing bonds at spreads higher than the average spread on bonds placed by EMDEs (Figure 3).23
Moving to domestic factors, the baseline specification shows that countries with higher per capita real GDP levels are more likely to issue sovereign bonds. The coefficient on real GDP growth shows that low-growth countries are penalized when issuing bonds as they tend to do so at higher spreads (Gelos and others, 2011; Eichengreen and Mody, 2000). The coefficients on inflation are generally not statistically significant. Countries that had an IMF-supported lending arrangement in the previous three years are more likely to issue, supporting the catalytic role of IMF lending (Mody and Saravia 2006).
The external sector position and liquidity matter for the terms of the issuance. Countries with lower current account deficits and higher international reserves face lower spreads (than those with higher external deficits and lower reserves). By contrast, there is no evidence of a robust association between the external position and the probability of bond issuance. The negative correlation between reserves and the probability of sovereign bond issuance, even if not significant, is consistent with the findings of Gelos and others (2011) and could suggest that a higher level of reserves insures sovereigns against exclusion from credit markets.
The fiscal position is a key determinant of market access. The coefficient on the public external debt-to-GDP ratio in the selection equation indicates that more indebted countries are less likely to issue sovereign bonds (Gelos and others, 2011) and when they do, they tend to pay higher prices: a 10 percent increase in the external debt-to-GDP ratio is associated with 10-12 basis-point increase in the primary spread (columns 7-9). The reinforcing effect of the debt-to-GDP ratio in the selection and outcome equations would suggest the presence of demand-side effects: high debt ratios would discourage demand from international investors and this shift would reduce the price of bonds, with a corresponding increase in primary spreads (Eichengreen and Mody, 2000). The negative coefficient on the government budget balance in the selection equation suggests that the demand for external borrowing is higher when fiscal deficit is larger, supporting the hypothesis that countries in the sample are capital scarce (Grigorian, 2003).
In the following specification (columns 3-4), the analysis of the effect of the fiscal deficit on sovereign bond issuance allows for a non-linear effect of the fiscal balance based on countries’ income level. Incorporating an interaction term between the fiscal balance-to-GDP ratio and per capita real GDP, we find that fiscal discipline is especially important for poorer countries, for which a stronger fiscal position is associated with a higher likelihood to issue sovereign bonds and at lower primary spreads than their richer counterparts. In the selection equation, the positive coefficient on the government budget balance (as a share of GDP) and the negative one on its interaction with per capita real GDP indicate that the fiscal balance is positively associated with the probability to issue, but this effect becomes weaker for richer countries (actually, the coefficient turns positive for sufficiently high level of per capita real GDP).
In addition, when we allow for heterogeneity (captured by differences in per capita income), a stronger budget balance is found to be associated with lower spreads, especially for lower income countries. A one percentage point improvement in the fiscal balance translates, on average, into a 3 basis-point drop in spreads, but this effect is more than five times larger for lower income countries (with a per capita GDP of about US$400). Modeling this non-linearity does not alter the results of the other variables.
In columns 5-6, the model is augmented to account for the fact that sovereign bond issuance is likely to be persistent. The analysis finds that after its first issuance, a country is more likely to issue again in the future, and when it does, spreads are lowered: the reduction in spreads for repeat issuers is about 75 basis points compared to countries that did not issue in the previous three years. This result lends further support to the hypothesis of the presence of fixed costs in accessing international capital markets.24
Other model specifications (in columns 7 and 8) control for institutional strength, measured by the index of government effectiveness25, which the existing literature has identified as a key driver of market access (Thomas 2009; Gelos and others 2011). Greater government effectiveness increases the likelihood of sovereign bond issuance as well as lowers its cost: a one standard deviation increase in the index of government effectiveness is associated with 98 basis-point reduction in primary spreads. Moreover, per capita real GDP and the effect of past IMF programs are no longer significantly associated with bond spreads. In addition, once we control for institutional quality, the coefficient on the US rate is less precisely estimated, while we still find evidence that spreads increase in periods of market volatility.
Finally, in columns 9 and 10, we model global shocks with time fixed effects rather than with the inclusion of the US rates and the VIX index. The results on the domestic factors are broadly unaffected (the negative correlation between inflation and the probability of bond issuance becomes significant and the coefficient on debt in the outcome equation is larger and more precisely estimated than in the alternative models) and time fixed effects are jointly significant, lending further support to the hypothesis that common global factors matter for market access by EDMEs.
The robustness analysis focuses mainly on the potential role of additional domestic factors in affecting market access. We take the last specification of Table 5 (columns 9-10), including the dummy variable for past bond issuances, the interaction between the fiscal balance and per capita real GDP, government effectiveness and year fixed effects, as reference. The robustness of our findings is tested by including a set of alternative domestic macroeconomic control variables. The results are reported in Table 6. We use 3 rather than 6 regional dummy variables to avoid convergence issues when estimating the model. It is reassuring that the baseline results (Table 6, columns 1-2) are very similar to the ones reported in Table 5 (columns 5-6).
|Real per capita GDP||−1.962||0.516***||−3.476||0.491***||−2.932||0.535***||−2.318||0.537***||10.725||0.513***||58.600*||0.479***||−39.002**||0.695***|
|PPG External debt (%GDP)||1.278***||−0.005**||1.301***||−0.004*||1.279***||−0.005**||1.497***||−0.005**||1.738***||−0.006**||1.097***||−0.003|
|Fiscal balance (%GDP)||−96.671***||0.447***||−84.551***||0.432***||−96.813***||0.437***||−95.102**||0.422***||−78.661**||0.439***||−157.236***||0.455***||−0.731||−0.058**|
|Fiscal balance (%GDP) * real pc GDP||11.704***||−0.061***||10.582***||−0.061***||11.753***||−0.061***||11.506**||−0.058***||9.144**||−0.060***||19.040***||−0.062***|
|Current account (%GDP)||−7.438***||−0.007||−6.876***||−0.007||−7.583***||−0.007||−7.346***||−0.008||−7.727***||−0.007||−5.801**||−0.003||−5.513***||−0.008|
|Reserves (in months of imports)||−9.386***||−0.030*||−11.049***||−0.026*||−9.276***||−0.035**||−9.376***||−0.029*||−8.615**||−0.032**||−7.292**||−0.035**||−9.273***||−0.035**|
|IMF program in the previous 3 year||24.596||0.807***||45.524||0.731***||21.114||0.681**||25.163||0.824***||66.511||0.765***||−16.378||0.786***||63.401||0.797***|
|SBI in the previous 3 year||−58.861**||0.747***||−55.071||0.726***||−62.719**||0.734***||−57.966*||0.752***||−36.257||0.734***||−88.637**||0.783***||−43.947||0.703***|
|Total debt (%GDP)||0.938**||−0.006**|
|Natural resource-rich (0/1)||(0.419)||(0.003)||−7.100||0.212|
|Capital account openness||(48.756)||(0.199)||−24.673***||0.060|
|Wald test (p-value)||0.296||0.733||0.130||0.344||0.993||0.111||0.540|
|Test area FE||0.000||0.000||0.000||0.000||0.000||0.000||0.000|
|Test year FE||0.000||0.000||0.000||0.000||0.000||0.000||0.000|
We start by controlling for the level of total public debt, rather than measuring exclusively public and publicly guaranteed external debt, and we find that the two debt indicators have similar effects, as higher public debt ratios are associated with a lower likelihood to issue and higher bond spreads (columns 3-4). We also control for the level of financial development and find a negative correlation between the ratio of credit to the private sector to GDP and the probability of issuance, suggesting that sovereigns with less developed credit markets rely more on external financing (column 5-6). The inclusion of a dummy variable for resource-rich economies26 does not alter the baseline results (columns 7-8). Further, the Chinn and Ito (2006) de jure measure of financial openness shows that countries with more open capital accounts can issue at a significant lower spread than less financially open sovereigns. We also augment the model by adding the ratio of aid inflows to GDP, in order to control for the possibility that countries with easier access to grants and concessional loans could have a smaller demand for non-concessional financing (columns 11-12); our results indicate that countries depending more on foreign aid face larger spreads which may suggest that aid recipients could be perceived as riskier by markets.
Finally, in columns 13 and 14 all domestic variables are measured at time t-1 (rather than averaged over the three-year period before the issuance). Results are almost identical to the baseline, with a notable exception being the significant positive association between inflation and bond spreads.
Our analysis offers the following messages for low-income developing countries, drawing on the experience of emerging markets and developing economies that have gained market access.
- Bond spreads. Countries with stronger budget balance positions, as well as external positions (measured by the current account balance and level of international reserves) experience lower spreads (than those with less strong fiscal and external positions). Spreads are lower in periods of declining global market volatility.
- Bond issuance. The fiscal position, as reflected in the fiscal balance and public debt level, matters importantly for market access. The likelihood of issuance increases with an improved government budget balance. Countries with larger debt burdens are less likely to issue bonds than their peers and, if they do, they do so at a premium. Strong government effectiveness is associated with both a higher likelihood of issuance and lower spreads. Finally, countries that are larger and have higher per capita GDP seem to have greater market access (than those that are smaller and less developed).
AbbasS.A.BelhocineN.El-GanainyA. and HortonM. (2011). “Historical Patterns and Dynamics of Public Debt—Evidence From A New Database” IMF Economic Review59(4): 717–742.
AfDB OECD and UNDP (2014). “African Economic Outlook” African Development BankOrganization for Economic Co-operation and Development, United Nations Development ProgramTunis, Paris and New York.
AlessandroM.SandlerisG. and Van der GhoteA. (2011). “Sovereign Defaults and the Political Economy of Market Re-access,” Universidad Torcuato Di Tella Working Paper No. 08/2011.
AroraV. and CerisolaM. (2001). “How Does U.S. Monetary Policy Influence Sovereign Spreads in Emerging Markets?” IMF Staff Papers48(3): 474–498.
BellasD.PapaioannouM. G. and PetrovaI. (2010). “Determinants of Emerging markets sovereign bond spreads: fundamentals vs financial stress” IMF Working Paper No. 10/281.
BaldacciE.GuptaS. and MatiA. (2011). “Political and fiscal risk determinants of sovereign spreads in emerging markets” Review of Development Economics15(2): 251–263.
BirdG. (2007). “The IMF: A Bird’s Eye View of Its Role and Operations” Journal of Economic Surveys21(4): pp. 683–745.
CeruttiE.Hale.G. and Minoiu.C. (2014). “Financial crises and the composition of cross-border lending” IMF Working Paper WP/14/185.
ChinnM. D. and ItoH. (2006). “What matters for financial development? Capital controls, institutions, and interactions” Journal of Development Economics81(1): 163–192.
ComelliF. (2012). “Emerging markets sovereign bond spreads: Estimation and back testing” Emerging Markets Review13(4): 598–625.
CrucesJ. C. and TrebeschC. (2013). “Sovereign Defaults: The Price of Haircuts” American Economic Journal: Macroeconomics5(3): 85–117.
CsontóB. (2014). “Emerging markets sovereign bond spreads and shifts in global market sentiment” Emerging Markets Review20: 58–74.
DasU. S.PapaioannouM. G. and PolanM. (2008). “Strategic Considerations for First-Time Sovereign Bond Issuers” IMF Working Paper No. 08/261.
Dell’ErbaS.BaldacciE. and PoghosyanT. (2013). “Spatial spillovers in emerging markets spreads” Empirical Economics45: 735–756.
EatonJ. and GersovitzM. (1980). “LDC participation in international financial markets: Debt and reserves” Journal of Development Economics7(1): 3–21.
EatonJ. and TaylorL. (1986). “Developing countries finance and debt” Journal of Development Economics22(1): 209–265.
EatonJ. (1993). “Sovereign debt: A primer” World Bank Economic Review7(2): 137–172.
EichengreenB. and ModyA. (2000). “What Explains Changing Spreads on Emerging Market Debt?.” NBER Chapters in: Capital Flows and the Emerging Economies: Theory Evidence and Controversies pp. 107–136National Bureau of Economic Research
FariaA.MauroP. and ZaklanA. (2010). “The external financing of emerging markets – Evidence from two waves of financial globalization” Review of Finance15(1): 207–243.
GelosG.R.SahayR. and SandlerisG. (2011). “Sovereign borrowing by developing countries: What determines market access?” Journal of International Economics83: 243–254.
GrigorianD. (2003). “On the determinants of first-time sovereign bond issues” IMF Working Paper No. 03/184.
GueyeC.A. and SyA.N.R. (2015). “Beyond Aid: How Much Should African Countries Pay to Borrow?” Journal of African Economies24(3): 352–366.
GuscinaA.Pedras. G. and Presciuttini. G. (2014). “First-Time International Bond Issuance – New Opportunities and Emerging Risks” IMF Working Paper WP/14/127.
HeckmanJ. J. (1979). “Sample selection bias as a specification error” Econometrica47(1): 153–161.
International Monetary Fund (2012). Macroeconomic Policy Framework for Resource-Rich Developing CountriesBoard paperAugustWashington, DC.
International Monetary Fund (2013). Regional Economic Outlook – Sub-Saharan AfricaMayWashington, DC.
International Monetary Fund (2014). Macroeconomic Developments in Low-Income Developing Countries: 2014 ReportBoard paperOctoberWashington, DC.
KaminS.B. and von KleistK. (1999). “The evolution and determinants of emerging market credit spreads in the 1990s” International Finance Discussion Papers No. 653Federal Reserve Board.
KaufmannD.KraayA. and MastruzziM. (2010). “The Worldwide Governance Indicators: Methodology and Analytical Issues” World Bank Policy Research Paper No. 5430.
KennedyM. and PalermA. (2014). “Emerging market bond spreads: The role of global and domestic factors from 2002 to 2011” Journal of International Money and Finance43: 70–87.
ModyA. and SaraviaD. (2006). “Catalysing Private Capital Flows: Do IMF Programmes Work as Commitment Devices?” The Economic Journal116(513): 843–67.
MuY.PhelpsP. and StotskyJ.G. (2013). “Bond markets in Africa” Review of Development Finance3(3): 121–135.
RochaK. and MoreiraA. (2010). “The Role of Domestic Fundamentals on the Economic Vulnerability of Emerging Markets” Emerging Markets Review11: 173–182.
SiklosP. (2011). “Emerging Market Yield Spreads: Domestic, External Determinants, and Volatility Spillovers” Global Finance Journal2283–100.
Standard and Poor’s (2013). “The growing allure of Eurobonds for African sovereigns” RatingsdirectMay.
te VelteD.W. (2014). “Sovereign Bonds in Sub-Saharan Africa. Good for Growth or Ahead of Time?” ODI Briefing No. 87London.
ThomasA. (2009). “Access to market financing for IDA-eligible countries – The role of external debt and IMF-supported programs” IMF Working Paper No. 09/217.