“Nothing would help improve standards more than if countries that met higher standards were rewarded with lower borrowing costs”
— Stanley Fischer (2002, p. 17)
A central objective of the IMF’s Data Dissemination Initiative is to improve data dissemination in support of the operation of international financial markets. The Mexican financial crisis of 1994–95 heightened awareness of the need to provide better information to the public and financial markets. The IMF responded by establishing the Data Dissemination Initiative to improve timely public release of economic and financial data and related information on compilation and release procedures. The Data Dissemination Initiative includes the Special Data Dissemination Standard (SDDS) to guide countries that have, or might seek, access to international capital markets, and the General Data Dissemination System (GDDS) to establish procedures to improve the quality of data of countries not yet aspiring to meet the SDDS requirements.
In recent years, empirical evidence has accumulated of lower borrowing costs for emerging market countries subscribing to the SDDS. Several secondary bond market studies have found an interest rate discount on bonds of emerging market countries subscribing to the SDDS. Recently, Cady (2005) found evidence of a similar discount for emerging market SDDS subscribers issuing bonds in the primary bond market. To our knowledge, the impact of GDDS participation on the cost of sovereign borrowing has yet to be examined.
This chapter seeks to fill this gap by examining the impact of the GDDS, along with that of the SDDS, on the borrowing costs of emerging market and developing countries that have issued sovereign bonds over the past decade and a half. It extends Cady’s (2005) paper in two ways: first, by expanding the empirical analysis to include GDDS participants with access to international capital markets and testing if GDDS participation influences sovereign borrowing costs; and second, by testing the influence of the data standards initiatives using sovereign credit ratings as an alternative specification for the fundamental macroeconomic determinants of sovereign borrowing costs.
Eichengreen and Mody (1998) provide evidence of a tendency for primary market launch spreads to follow secondary market spreads with a three-to-four-quarter lag. But they note that secondary market spreads can move differently over the short run due, in part, to market sentiment about emerging market debt as a distinct asset class. This chapter examines primary market launch spreads, in part to avoid the possibility of such disconnects influencing statistical inference, but principally to measure the cost of borrowing (abstracting from underwriting and legal costs) relevant to the sovereign issuer.
The analysis provides strong and consistent econometric evidence of discounts for sovereign issuers participating in the GDDS, as well as emerging market countries subscribing to the SDDS. Estimated launch spread discounts amount to about 9 percent for GDDS participants and 20 percent for SDDS subscribers, or the equivalent of 20 and 50 basis points, respectively. These results are consistent across the alternative specifications, stable over time, and broadly in line with estimates from other studies.1 Modeling spreads as a function of sovereign credit ratings reinforce the results of models specified with key macroeconomic variables.
Institutional Background
Following the Mexican financial crisis, the international community recognized the essential role of data transparency in meeting the challenges and risks of globalization and reducing the likelihood of financial crises. Hence the call for the timely dissemination of macroeconomic and financial data and an improved early warning system permitting a swifter response to financial shocks. The IMF endorsed the establishment of voluntary standards to guide member countries in the public dissemination of economic and financial data. The standards aimed to enhance the availability of timely macroeconomic and financial statistics, thereby contributing to the formulation and pursuit of sound macroeconomic polices, as well as improved functioning of financial markets. The SDDS was approved by the IMF Executive Board in March 1996 and the GDDS in December 1997.
The SDDS is a voluntary standard monitored by the IMF that focuses on dissemination of economic and financial data used principally by financial market participants. The standard identifies four dimensions of data dissemination: coverage, periodicity (frequency of compilation), and timeliness (speed of dissemination) of the data; access by the public; and the integrity and quality of the disseminated data. For each of these dimensions, the SDDS prescribes two to four monitorable elements. Subscribing countries must observe requirements for data categories covering four sectors of the economy2 and provide metadata (descriptions of the data), advance release calendars, and other information about their data dissemination practices. Subscribers must also agree to post this information on the IMF’s Dissemination Standards Bulletin Board (DSBB) and establish a national summary data page linked to the DSBB.
The GDDS is a framework to guide countries in the development of sound statistical systems and dissemination of economic and financial data to the public. It is built around the same four dimensions as the SDDS—data characteristics, quality, access, and integrity—and is intended to provide guidance for the overall development of macroeconomic, financial, and sociodemographic statistics. The data characteristics dimension prescribes coverage, periodicity, and timeliness for 19 data categories, including sociodemographic data.3 The GDDS calls for participating countries to prepare metadata and describe statistical practices and development plans over the short and medium term, along with any technical assistance requirements. Participating countries are expected to update metadata annually and describe how data compilation and dissemination activities are keeping pace with development plans and good international statistical practices as set forth in the GDDS. Complete information on both the SDDS and GDDS, including specific data categories and indicators, are available on the DSBB.4
About three-fourths of the IMF membership either subscribes to the SDDS or participates in the GDDS. As of November 2007, SDDS subscription stood at 64 countries, while 95 countries have participated in the GDDS, six of which have progressed to the SDDS (Figure 4.1). The IMF’s technical assistance program in statistics aims to promote graduation of GDDS participants to the SDDS. To maintain the credibility of the data standards initiatives, the IMF monitors observance of the SDDS, and aligns the structures of the SDDS and GDDS with the IMF’s Data Quality Assessment Framework (DQAF).5

Number of SDDS Subscribers and GDDS Participants
Source: IMF Statistics Department.Note: SDDS = Special Data Dissemination Standard; GDDS = General Data Dissemination System.
Number of SDDS Subscribers and GDDS Participants
Source: IMF Statistics Department.Note: SDDS = Special Data Dissemination Standard; GDDS = General Data Dissemination System.Number of SDDS Subscribers and GDDS Participants
Source: IMF Statistics Department.Note: SDDS = Special Data Dissemination Standard; GDDS = General Data Dissemination System.The IMF’s Data Dissemination Initiative was developed at a time when international investors were showing greater interest in emerging market and developing countries, reflecting a search for yield through international portfolio diversification. Such a search requires that investors compare the creditworthiness and investment risks of developed and emerging market countries using available economic and financial data. Public provision of a continuous flow of macroeconomic and financial data under predictable dissemination policies and release schedules is the central purpose of the Data Dissemination Initiative. Hence, improved timeliness of data provided by emerging market and developing countries borrowing on international capital markets should, all else being equal, allow easier access on better terms to global finance. The mechanism at work is well described by Eichengreen (1999, p. 27): “… [SDDS] subscription status provides an objective indicator of countries’ creditworthiness, providing an alternative to the judgments of commercial credit agencies. Investors might become reluctant to lend to countries that fail to subscribe to the standard or might use interest rate spreads to ration credit to them.”
Capital market participants generally view the SDDS as useful. Mosely (2003) reports that a survey of U.S. and UK mutual fund managers conducted during 2000 indicated concerns with the availability and quality of information, especially for developing and emerging markets. While awareness of the SDDS was not high, with over 60 percent of respondents indicating that the SDDS played no role in their decision making, about 7 percent indicated that they would attach a smaller risk premium to countries subscribing to the SDDS. In principal, improved access to timely, high-frequency, and quality data should permit a more precise quantification of measurable risks and help reduce uncertainty in the subjective assessments of country risk typically made by market participants.6 According to a 2000 Financial Stability Forum (FSF) discussion of international standards and codes, for market incentives to work, market participants must be aware of the standard, judge it of relevance, and use it in forming their risk assessments. Further, this must be reflected in the pricing or allocation of credit or investment in a particular economy or institution operating in that economy, in the form of differentiated credit ratings, borrowing spread, or asset allocations (Financial Stability Forum, 2000, p. 4). Related FSF surveys found that market participants’ familiarity with 12 key international standards varied widely, but that the SDDS and the International Accounting Standard were the best known and viewed as particularly useful. While available survey material suggests that the Data Dissemination Initiative is helpful, our analysis looks for econometric evidence in the capital market data.
Data
The influence of SDDS subscription and GDDS participation on sovereign borrowing costs is examined using panel data models. The dataset consists of quarterly time-series data on new issues of sovereign bonds, denominated in U.S. dollars, Japanese yen, and euros,7 and key macroeconomic and credit indicators for a group of 26 emerging market and developing countries (see Appendix 4.1).
Bond characteristics and issuance data were drawn from the IMF’s bonds, equities, and loans (BEL) database. Spreads reported in the BEL database are measured as the annual yield to maturity at the time of the launch minus a “risk-free” benchmark yield, defined as the annual yield for an industrial country government bond of the same currency and maturity. Again, we focus on launch spreads (and yields) as they represent the actual cost of borrowing incurred by countries, in contrast to the well-known JPMorgan Emerging Market Bond Index family that measures spreads of existing securities traded in secondary markets.
The dataset is primarily comprised of some 320 sovereign bonds issued by the 26 emerging market and developing countries over 1989–2004.8 The dataset has an unbalanced time dimension, as the sample periods for countries vary according to their differing bond issuance histories and the availability of macroeconomic data and sovereign credit ratings (Table 4.A1). In general, the time frame extends approximately seven years prior to and following the opening of subscriptions to the SDDS in April 1996 and participation in the GDDS in December 1997. Over 2000–04, 24 of the 26 countries accounted for an average of 68 percent of the value of all new bond issues by emerging market and developing countries (IMF, 2005, Table 15).9 The maturity of bonds in the panel dataset ranged from one to 30 years with a median of seven years.
In addition to bond characteristics, the analysis accounts for country characteristics, including key macroeconomic performance indicators or sovereign credit ratings, as well as changes in institutional quality. The IMF’s International Financial Statistics and World Economic Outlook and the World Bank’s Global Development Finance serve as sources for the macroeconomic variables. Information on IMF financial arrangements, SDDS subscription, and GDDS participation were drawn from the IMF’s records, while country indicators of institutional quality have been derived from the International Country Risk Guide prepared by the Political Risk Services Group, Inc.10
Sovereign credit ratings were drawn from publications of the three principal credit rating agencies: Standard and Poor’s, Moody’s, and Fitch. Following several analysts, beginning with Horrigan (1966) and continuing through Montford and Mulder (2000), alphanumeric credit ratings were transformed into numerical ratings (Table 4.A2). When more than one agency provides a rating, the mean of the numerical ratings was used.
Model
The cost of issuing a sovereign bond is assumed to be related to borrower and bond characteristics in a log-linear model:
where the dependent variable for cost (C i,t) is measured as the natural logarithm of the launch spread (SP i,t) for country i in period t; χi,t is a vector of explanatory variables; and u i,t is a random error term. Specifically, χi,t is composed of issuer and bond characteristics, indicators for macroeconomic performance or credit ratings, and participation in the IMF’s data standards initiatives.
Following the empirical policy evaluation literature,11 the influences of SDDS subscription and GDDS participation on sovereign launch spreads are examined using dummy variables while controlling for bond characteristics and macroeconomic performance (and credit ratings in an alternative specification). The SDDS dummy variable equals zero prior to subscription and one in the quarter of subscription and thereafter. The GDDS dummy variable is similarly defined, and is based on the quarter that formal participation began (Table 4.A1).
The selection of appropriate macroeconomic variables was guided by the literature12 and includes the rate of real GDP growth (GDPDOT), inflation differentials vis-à-vis the United States (DPDOT), the change in the primary fiscal balance as a percentage of GDP (GPBAL), and the debt-export ratio of the borrowing country (DXR). In an alternative specification, these macroeconomic indicators are replaced by the country’s credit rating (CR), based on the view that ratings subsume the information content of the macroeconomic variables and may reflect additional information, such as social and political considerations, that could bear on country risk and the cost of borrowing.13 Given that credit ratings reflect additional factors, this specification should prove a stringent statistical test of the influence of the SDDS and GDDS on launch spreads. Finally in both specifications, the potential effects of IMF-supported programs are also examined using a dummy variable (IMF).14
The maturity of the bond (MAT), measured in years, is included as an exogenous variable. This follows the view that creditors take into account the risk of default, which generally increases with maturity, when determining the terms of a bond. Granger causality tests were carried out on spreads and maturities to investigate the possibility of endogeneity bias.15 The hypothesis of the exogeneity of maturity was not rejected in all but four cases, where the results were mixed and ambiguous. Estimation results proved robust to the exclusion of these four countries, diminishing the importance of the simultaneity issue as a practical matter.
Another important bond characteristic considered in the model is the currency of denomination. The basic currency of denomination is the U.S. dollar, while dummy variables indicate yen and euro denominations (YEN and EURO, respectively). The dataset includes 55 bonds denominated in yen and 97 denominated in euros, respectively representing 17 and 30 percent of all the bonds considered.
The analysis also incorporates quality indicators for a country’s legal and bureaucratic framework. A combined index of country institutional quality (INST) is included in the model so that the effects of progress in these two areas over the sample period can be estimated separately from improvements in data transparency and dissemination practices represented by SDDS and GDDS participation.
Panel unit root tests16 permitted rejection of the hypothesis of nonstationarity at conventional levels of significance for all of the variables discussed above except credit ratings,17 obviating cointegration in the panel dataset.
Estimation Issues
This section outlines the econometric methodologies deployed to deal with characteristics of the dataset, including country-specific heterogeneity, cross-country heterosckedasticity, and contemporaneous correlation. When incompatibilities arose between models and estimators, we opted to attached the highest priority to consistency and efficiency considerations in support of statistical inference concerning SDDS subscription and GDDS participation.
There is a high likelihood that the panel dataset exhibits cross-section heterosckedasticity, meaning a differing residual variance for each cross-section (country). This is illustrated with a few descriptive statistics: the mean launch spread for Brazil is about 490 basis points, with a standard deviation of about 165, while for Korea spreads average about 100 basis points, with a standard deviation of about 40. On the basis of such differences, regression residuals for Brazil ought to be larger than those for Korea, and exhibit larger variances. Examination of ordinary least square (OLS) panel regression residuals confirmed this characteristic, pointing to the need to employ an estimator robust to cross-section heterosckedasticity. Further, market analysis tends to treat emerging market debt as a separate asset class, suggesting that changes in market sentiment toward the asset class could drive common trends, raising the potential for errors for different cross-sections to be contemporaneously correlated. International liquidity conditions, to the extent that they affect emerging market yields and spreads as a group, could also be a potential source of contemporaneous cross-section correlation. These considerations suggest the need for an estimator robust to contemporaneous correlation. Under these conditions and given these market characteristics, the most appropriate estimator is the feasible generalized least square (GLS) estimator allowing for residuals that exhibit cross-section heterosckedasticity and that are contemporaneously correlated.18
A primary concern in panel estimation is how to allow for unobserved heterogeneity that may be correlated with regressors. Hausman tests, formal tests of whether or not individual country effects are fixed, on both the macroeconomic and credit rating specifications failed to reject the null hypothesis of random effects. However, feasible GLS estimates of a random-effects model suffered from autocorrelation; unfortunately, this estimator when applied to a random effects model cannot be estimated with specifications containing autoregressive terms. That being the case, and considering autocorrelation the more significant problem, the reported model estimates have been derived using feasible GLS estimation corrected for cross-section heterosckedasticity, contemporaneous correlation, and autocorrelation, but without modeling cross-section heterogeneity.19 Panel OLS and weighted GLS estimation of random and fixed-effects models for both the macroeconomic and credit rating specifications (allowing for autoregressive terms where feasible) yielded coefficient estimates broadly similar in sign, size, and significance to those reported in this chapter, and in particular those attached to the SDDS and GDDS dummy variables. On this basis, we are confident that omission of modeling country heterogeneity presents minimal difficulties.
Results
From equation (1), the estimating equation specified as a function of macroeconomic variables becomes:
GLS estimation of equation (2) yields a coefficient estimate for GDDS participation with a negative sign that is statistically significant from zero at conventional confidence levels (Table 4.1, first column). This point estimate implies that GDDS participation reduces launch spreads by about 9 percent, or 23 basis points when evaluated using an illustrative spread of 250 basis points. The estimated coefficient for SDDS subscription is also statistically significant, negative, and of a magnitude very close to estimates obtained by Cady (2005). The point estimate implies that SDDS participation reduces launch spreads by close to 20 percent, or about 50 basis points on an illustrative spread of 250 basis points. Both the GDDS and SDDS coefficient estimates are quite stable when estimated over differing time periods (Figure 4.2).
Log-Spread Generalized Least Squares Regressions for 26 Emerging Market and Developing Countries
Global estimation range for the unbalanced panel; Table 4.A1 reports country-specific sample periods. Dependent variable is the natural logarithm of the yield spread; t-statistics, based on panel consistent standard errors, reported in parentheses. * significant at 10 percent; ** significant at 5 percent; *** significant at 1 percent. GDDS = General Data Dissemination System; SDDS = Special Data Dissemination Standard.
Log-Spread Generalized Least Squares Regressions for 26 Emerging Market and Developing Countries
(1) | (2) | ||
---|---|---|---|
Estimation range1 | 1991:4-2003:4 | 1989:2-2004:4 | |
Constant | 3.908 | 3.439 | |
(10.03)*** | (10.00)*** | ||
Real GDP growth (GDPDOT) | –0.277 | — | |
(-2.18)** | |||
Inflation differential (DPDOT) | 0.010 | — | |
(1.76)* | |||
Primary balance (ΔGPBAL) | –0.491 | — | |
(-2.34)** | |||
Debt-export ratio (In DXR) | 0.417 | — | |
(7.10)*** | |||
Credit rating (In CR) | — | 1.140 | |
(10.93)*** | |||
Maturity (In MAT) | 0.038 | 0.021 | |
(3.29)*** | (1.74)* | ||
Institutions (In INST) | –0.336 | –0.337 | |
(-5.16)*** | (-5.12)*** | ||
Yen-denominated issue (YEN) | –0.446 | –0.450 | |
(-19.33)*** | (-19.00)*** | ||
Euro-denominated issue (EURO) | –0.308 | –0.318 | |
(-18.03)*** | (-16.89)*** | ||
IMF arrangement (IMF) | –0.036 | –0.049 | |
(-1.78)* | (-2.26)** | ||
SDDS subscription (SDDS) | –0.194 | –0.139 | |
(-5.50)*** | (-4.17)*** | ||
GDDS participation (GDDS) | –0.093 | –0.076 | |
(-2.82)*** | (-2.75)*** | ||
Time trend (TIME) | 0.013 | 0.006 | |
(3.07)*** | (2.37)** | ||
Autocorrelation coefficient | 0.869 | 0.809 | |
(52.87)*** | (46.27)*** | ||
Adjusted R2 | 0.8204 | 0.8320 | |
Durbin-Watson statistic | 2.171 | 2.139 | |
Countries in panel | 26 | 26 | |
Observations | 778 | 852 | |
Mean of the dependent variable (basis points) | 262.4 | 265.7 | |
Memorandum items: | |||
Point estimate of discount (evaluated at an illustrative spread of 250 basis points): | |||
SDDS | 48.50 | 34.75 | |
GDDS | 23.25 | 19.00 |
Global estimation range for the unbalanced panel; Table 4.A1 reports country-specific sample periods. Dependent variable is the natural logarithm of the yield spread; t-statistics, based on panel consistent standard errors, reported in parentheses. * significant at 10 percent; ** significant at 5 percent; *** significant at 1 percent. GDDS = General Data Dissemination System; SDDS = Special Data Dissemination Standard.
Log-Spread Generalized Least Squares Regressions for 26 Emerging Market and Developing Countries
(1) | (2) | ||
---|---|---|---|
Estimation range1 | 1991:4-2003:4 | 1989:2-2004:4 | |
Constant | 3.908 | 3.439 | |
(10.03)*** | (10.00)*** | ||
Real GDP growth (GDPDOT) | –0.277 | — | |
(-2.18)** | |||
Inflation differential (DPDOT) | 0.010 | — | |
(1.76)* | |||
Primary balance (ΔGPBAL) | –0.491 | — | |
(-2.34)** | |||
Debt-export ratio (In DXR) | 0.417 | — | |
(7.10)*** | |||
Credit rating (In CR) | — | 1.140 | |
(10.93)*** | |||
Maturity (In MAT) | 0.038 | 0.021 | |
(3.29)*** | (1.74)* | ||
Institutions (In INST) | –0.336 | –0.337 | |
(-5.16)*** | (-5.12)*** | ||
Yen-denominated issue (YEN) | –0.446 | –0.450 | |
(-19.33)*** | (-19.00)*** | ||
Euro-denominated issue (EURO) | –0.308 | –0.318 | |
(-18.03)*** | (-16.89)*** | ||
IMF arrangement (IMF) | –0.036 | –0.049 | |
(-1.78)* | (-2.26)** | ||
SDDS subscription (SDDS) | –0.194 | –0.139 | |
(-5.50)*** | (-4.17)*** | ||
GDDS participation (GDDS) | –0.093 | –0.076 | |
(-2.82)*** | (-2.75)*** | ||
Time trend (TIME) | 0.013 | 0.006 | |
(3.07)*** | (2.37)** | ||
Autocorrelation coefficient | 0.869 | 0.809 | |
(52.87)*** | (46.27)*** | ||
Adjusted R2 | 0.8204 | 0.8320 | |
Durbin-Watson statistic | 2.171 | 2.139 | |
Countries in panel | 26 | 26 | |
Observations | 778 | 852 | |
Mean of the dependent variable (basis points) | 262.4 | 265.7 | |
Memorandum items: | |||
Point estimate of discount (evaluated at an illustrative spread of 250 basis points): | |||
SDDS | 48.50 | 34.75 | |
GDDS | 23.25 | 19.00 |
Global estimation range for the unbalanced panel; Table 4.A1 reports country-specific sample periods. Dependent variable is the natural logarithm of the yield spread; t-statistics, based on panel consistent standard errors, reported in parentheses. * significant at 10 percent; ** significant at 5 percent; *** significant at 1 percent. GDDS = General Data Dissemination System; SDDS = Special Data Dissemination Standard.

Recursive SDDS and GDDS Coefficient Estimates
Source: Authors’ calculations.Note: SDDS= Special Data Dissemination Standard; GDDS = General Data Dissemination System.
Recursive SDDS and GDDS Coefficient Estimates
Source: Authors’ calculations.Note: SDDS= Special Data Dissemination Standard; GDDS = General Data Dissemination System.Recursive SDDS and GDDS Coefficient Estimates
Source: Authors’ calculations.Note: SDDS= Special Data Dissemination Standard; GDDS = General Data Dissemination System.Coefficient estimates of other variables in equation (2) are all statistically significant, with the expected signs, and are broadly in line with previous studies. The estimate of the coefficient for real GDP growth implies that spreads are lower by 35 basis points when growth is a ½ percentage point higher. If the primary fiscal balance improves by a ½ percentage point of GDP, the estimated reduction in spreads is about 60 basis points. A decline in the debt-export ratio from 50 to 40 percent is estimated to reduce spreads by 23 basis points. The significance of these macroeconomic indicators is consistent with cited studies, starting with Cantor and Packer (1996).
Improvement in the legal and bureaucratic framework of a country was found to lower spreads. Most of the countries included in the sample improved not only the transparency of their data and public dissemination practices, but also the quality of their institutional framework over the sample period. The variable constructed to measure institutional quality from indicators for the legal and bureaucratic framework has coefficient estimates whose magnitude and statistical significance are robust to differing specifications and estimation techniques. A one-standard deviation increase in the institutional quality variable around its mean was estimated to reduce spreads by 35 basis points. Including an institutional quality variable in the model represents an effort to control the estimated influences of GDDS participation and SDDS subscription for simultaneous improvements in institutional quality.
The longer the maturity of a bond, the higher the spread. Longer maturity increases the likelihood that the creditworthiness or repayment capacity will change over the term of the bond. The higher repayment uncertainty is estimated to increase spreads by about six to seven basis points for an increase in maturity from five to 10 years.
The negative coefficient estimates for the yen and euro dummy variables are highly significant, possibly reflecting regional currency-of-issue preferences on the supply and demand sides or the fact that average yields on yen- and euro-denominated bonds were significantly lower than on dollar-denominated bonds throughout the sample period. This issue is open for discussion and, since it is not critical to the topic at hand, is left to future research.
An IMF-supported program has a measurable effect on launch spreads. When IMF support becomes effective, launch spreads are estimated to decline by about 10 basis points, perhaps reflecting market expectations that IMF-supported programs help to restore macroeconomic stability. This is consistent with the findings of Eichengreen, Kletzer, and Mody (2005) and Cady (2005), providing additional evidence that IMF-supported programs are considered positively in financial markets, perhaps conveying information about a country’s economic policies or capacity and willingness to repay.
The small positive estimated coefficient attached to the time trend (TIME) reflects the net effects of any trends or time-related factors not explicitly accounted for in the model. Possible factors include an increasing investor base interested in emerging market countries, developments in global liquidity and financial markets throughout much of the sample period, as well as the effects of the Mexican, Asian, and Russian crises on emerging market securities as an asset class and their subsequent dissipation. Our concern is to obtain the best possible unbiased estimators for the SDDS and GDDS coefficients, and the time trend is therefore included. That said, all coefficient estimates are robust to the exclusion of the time trend or inclusion of individual country time trends, including the SDDS and GDDS coefficients, which remain consistently negative about the same magnitude and statistically significant.
A country’s commercial credit rating performs well as a substitute for key macroeconomic indicators, as found by other studies.20 As previously noted, this alternate specification provides a check on the parameter estimates of the basic model, and perhaps a more stringent statistical test of the influence of the SDDS and GDDS on launch spreads. When specified as a function of credit ratings, the estimation equation becomes:
Coefficient estimates for equation (3) are highly significant and stable (Table 4.1, second column). The GLS estimates for the full panel imply a 38 basis point reduction in launch spreads when the borrower’s credit rating is upgraded one full notch from adequate payment capacity into the range of strong payment capacity.21 Other studies that examined the stability of spread equations over different time periods also found a highly stable and significant credit rating impact, while coefficient estimates for macroeconomic indicators were less significant and varied in magnitude.22
The coefficient estimates for the GDDS and SDDS variables using a country’s credit rating did not differ significantly from the estimates using macroeconomic indicators. The reduction in launch spreads owing to GDDS participation was estimated to be about 20 basis points, while that for SDDS participation was 35 basis points. As with the specification using macroeconomic indicators, these estimates controlled for institutional quality. The estimated reduction of 35 basis points in spreads for a one-standard deviation increase in the institutional quality variable around its mean was highly statistically significant.
Concluding Comments
The policy implications of the findings in this chapter are straightforward. Although macroeconomic performance and solvency considerations are fundamental in determining access to international capital markets on favorable terms, participation in the IMF’s Data Dissemination Initiative can provide significant cost savings to sovereign borrowers. Our empirical findings indicate that sovereign borrowers have financial incentives to participate in the GDDS and even larger incentives to subscribe to the SDDS. For the IMF, maintaining the credibility of the SDDS as a monitored standard is critical, since the credibility of the standard and continued financial benefits to subscribers depend on their observance of all provisions of the standard.
The 11 GDDS participants considered in this chapter had borrowed in international capital markets prior to the launch of the GDDS. Consequently, our findings should not be construed as implying that GDDS participation alone contributes to market access. Generally, the aim of GDDS participation is to improve statistical practices rather than to gain market access. That said, however, previously established creditworthiness and access appear to be enhanced by GDDS participation, perhaps by reducing uncertainty in the view of investors sufficiently enough to warrant a small interest rate discount.
With regard to the SDDS, it is difficult to distinguish if it is the content of the standard or the fact that observance is monitored by the IMF that is most relevant to investors. Further, investors may not fully distinguish between the GDDS as a statistical development system and the SDDS as a monitored standard. Investors could view both SDDS subscription and GDDS participation as a signal of lower uncertainty about the reliability and serviceability of economic and financial data. This may enable investors to make better-informed assessments, which, in turn, could warrant lower risk premiums for emerging market and developing countries. Our estimates indicate that the SDDS discount is larger than the GDDS discount, which is consistent with the fact that the requirements of the SDDS, the monitored standard, are significantly more stringent than those of the GDDS, the developmental system.
This chapter found evidence of lower sovereign borrowing costs for emerging market and developing countries subscribing to the SDDS or participating in the GDDS. This financial incentive can, in turn, help improve data quality and dissemination standards in the virtuous cycle alluded to by Stanley Fischer in 2002.
Appendix 4.1. Data Sources
Data used in this study have been drawn from the following sources:
The IMF’s bonds, equities, and loans (BEL) database (sourced from Dealogic) for the spreads, yields, maturity, and currency of denomination (U.S. dollars, Japanese yen, or euros) of sovereign bonds issued by 26 emerging market and developing countries during 1989–2004. Table 4.A1 presents the countries considered in this chapter, their dates of GDDS participation or SDDS subscription, respective sample periods, and the number of bonds issued during this period.
The IMF’s International Financial Statistics and World Economic Outlook for quarterly economic growth and inflation rates, and the annual primary deficit in percent of GDP. Debt-export ratios were drawn from the World Bank’s Global Development Finance.
SDDS Subscription and GDDS Participation Dates, Sample Periods, and Numbers of Bonds Issued
SDDS Subscription and GDDS Participation Dates, Sample Periods, and Numbers of Bonds Issued
Data Initiative and Country | Date of Subscription or Participation | Sample Period with Macro Variables/Credit Ratings | Number of Bonds Issued During the Sample Period |
---|---|---|---|
SDDS | |||
Argentina | August 16, 1996 | 1994:2 to 2002:4 | 24 |
1992:3 to 2002:4 | 24 | ||
Brazil | March 14, 2001 | 1995:3 to 2002:4 | 16 |
1995:3 to 2002:4 | 16 | ||
Colombia | May 31, 1996 | 1995:2 to 2002:4 | 19 |
1995:2 to 2002:4 | 19 | ||
Costa Rica | November 28, 2001 | 1998:3 to 2003:4 | 7 |
1998:3 to 2004:4 | 8 | ||
Croatia | May 20, 1996 | 1997:2 to 2001:4 | 8 |
1997:2 to 2001:4 | 8 | ||
Hungary | May 24, 1996 | 1996:1 to 2001:2 | 7 |
1992:3 to 2002:4 | 7 | ||
Korea, Rep. of | September 20, 1996 | 1990:3 to 2002:4 | 27 |
1990:3 to 2002:4 | 27 | ||
Lithuania | May 30, 1996 | 1996:1 to 2001:4 | 9 |
1996:4 to 2002:4 | 9 | ||
Malaysia | August 21, 1996 | 2000:4 to 2002:4 | 2 |
2000:4 to 2002:4 | 2 | ||
Mexico | August 13, 1996 | 1991:2 to 2002:4 | 24 |
1991:2 to 2002:4 | 24 | ||
Philippines | August 5, 1996 | 1993:3 to 2002:4 | 8 |
1993:4 to 2002:4 | 8 | ||
Poland | April 17, 1996 | 1996:2 to 2002:4 | 7 |
1995:3 to 2002:4 | 7 | ||
South Africa | August 2, 1996 | 1990:2 to 2002:4 | 13 |
1994:4 to 2002:4 | 13 | ||
Tunisia | June 20, 2001 | 1995:2 to 2002:4 | 6 |
1995:4 to 2002:4 | 6 | ||
Turkey | August 8, 1996 | 1990:2 to 2002:4 | 34 |
1992:3 to 2002:4 | 34 | ||
Uruguay | February 12, 2004 | 1992:3 to 2001:4 | 12 |
1994:1 to 2002:4 | 12 | ||
GDDS | |||
Barbados | May 22, 2000 | 1994:3 to 2003:4 | 4 |
1995:1 to 2004:4 | 2 | ||
China, People’s Republic of | April 15, 2002 | 1994:1 to 2000:4 | 12 |
1994:1 to 2002:4 | 13 | ||
Guatemala | December 6, 2004 | 1997:3 to 2003:4 | 3 |
1997:4 to 2004:4 | 3 | ||
Jamaica | February 28, 2003 | 1997:3 to 2002:4 | 8 |
1998:2 to 2004:4 | 9 | ||
Kazakhstan | May 22, 2000 | 1997:1 to 2002:4 | 7 |
(SDDS March 2003) | 1997:1 to 2004:4 | 7 | |
Lebanon | January 16, 2003 | 1994:4 to 2003:4 | 22 |
1997:2 to 2004:4 | 24 | ||
Panama | December 28, 2000 | 1997:2 to 2003:4 | 10 |
1997:2 to 2004:4 | 12 | ||
Romania | February 14, 2001 | 1996:3 to 2001:2 | 7 |
(SDDS May 2005) | 1996:3 to 2002:4 | 7 | |
Trinidad and Tobago | September 30, 2004 | 1993:1 to 2001:4 | 6 |
1993:2 to 2004:4 | 6 | ||
Venezuela | March 29, 2001 | 1989:3 to 2001:4 | 15 |
1989:3 to 2002:4 | 15 | ||
Totals: 26 countries | Macro variable sample | 317 | |
— | Credit rating sample | 322 |
SDDS Subscription and GDDS Participation Dates, Sample Periods, and Numbers of Bonds Issued
Data Initiative and Country | Date of Subscription or Participation | Sample Period with Macro Variables/Credit Ratings | Number of Bonds Issued During the Sample Period |
---|---|---|---|
SDDS | |||
Argentina | August 16, 1996 | 1994:2 to 2002:4 | 24 |
1992:3 to 2002:4 | 24 | ||
Brazil | March 14, 2001 | 1995:3 to 2002:4 | 16 |
1995:3 to 2002:4 | 16 | ||
Colombia | May 31, 1996 | 1995:2 to 2002:4 | 19 |
1995:2 to 2002:4 | 19 | ||
Costa Rica | November 28, 2001 | 1998:3 to 2003:4 | 7 |
1998:3 to 2004:4 | 8 | ||
Croatia | May 20, 1996 | 1997:2 to 2001:4 | 8 |
1997:2 to 2001:4 | 8 | ||
Hungary | May 24, 1996 | 1996:1 to 2001:2 | 7 |
1992:3 to 2002:4 | 7 | ||
Korea, Rep. of | September 20, 1996 | 1990:3 to 2002:4 | 27 |
1990:3 to 2002:4 | 27 | ||
Lithuania | May 30, 1996 | 1996:1 to 2001:4 | 9 |
1996:4 to 2002:4 | 9 | ||
Malaysia | August 21, 1996 | 2000:4 to 2002:4 | 2 |
2000:4 to 2002:4 | 2 | ||
Mexico | August 13, 1996 | 1991:2 to 2002:4 | 24 |
1991:2 to 2002:4 | 24 | ||
Philippines | August 5, 1996 | 1993:3 to 2002:4 | 8 |
1993:4 to 2002:4 | 8 | ||
Poland | April 17, 1996 | 1996:2 to 2002:4 | 7 |
1995:3 to 2002:4 | 7 | ||
South Africa | August 2, 1996 | 1990:2 to 2002:4 | 13 |
1994:4 to 2002:4 | 13 | ||
Tunisia | June 20, 2001 | 1995:2 to 2002:4 | 6 |
1995:4 to 2002:4 | 6 | ||
Turkey | August 8, 1996 | 1990:2 to 2002:4 | 34 |
1992:3 to 2002:4 | 34 | ||
Uruguay | February 12, 2004 | 1992:3 to 2001:4 | 12 |
1994:1 to 2002:4 | 12 | ||
GDDS | |||
Barbados | May 22, 2000 | 1994:3 to 2003:4 | 4 |
1995:1 to 2004:4 | 2 | ||
China, People’s Republic of | April 15, 2002 | 1994:1 to 2000:4 | 12 |
1994:1 to 2002:4 | 13 | ||
Guatemala | December 6, 2004 | 1997:3 to 2003:4 | 3 |
1997:4 to 2004:4 | 3 | ||
Jamaica | February 28, 2003 | 1997:3 to 2002:4 | 8 |
1998:2 to 2004:4 | 9 | ||
Kazakhstan | May 22, 2000 | 1997:1 to 2002:4 | 7 |
(SDDS March 2003) | 1997:1 to 2004:4 | 7 | |
Lebanon | January 16, 2003 | 1994:4 to 2003:4 | 22 |
1997:2 to 2004:4 | 24 | ||
Panama | December 28, 2000 | 1997:2 to 2003:4 | 10 |
1997:2 to 2004:4 | 12 | ||
Romania | February 14, 2001 | 1996:3 to 2001:2 | 7 |
(SDDS May 2005) | 1996:3 to 2002:4 | 7 | |
Trinidad and Tobago | September 30, 2004 | 1993:1 to 2001:4 | 6 |
1993:2 to 2004:4 | 6 | ||
Venezuela | March 29, 2001 | 1989:3 to 2001:4 | 15 |
1989:3 to 2002:4 | 15 | ||
Totals: 26 countries | Macro variable sample | 317 | |
— | Credit rating sample | 322 |
Fitch Ratings, Fitch—Complete Sovereign Rating History, March 2, 2005, Moody’s Investors Service, Sovereign Ratings History, March 4, 2004, and Standard & Poor’s Ratings Services, Sovereign Ratings History Since 1975, March 3, 2005, for sovereign credit ratings. Alphanumeric credit ratings are transformed into numerical ratings according to Table 4.A2. All three ratings agencies qualify their ratings with outlook and review/watch qualifications to signal a possible upgrade or downgrade. To take account of these signals, the basic numerical ratings are decreased by 0.2 for positive outlook and watches/review qualifications while negative outlook or watches/review are increased 0.2 each. For example, a sovereign with an A+ rating from both S&P and Fitch would be assigned a numerical value of 5; A+ ratings with a positive outlook would be assigned 4.8 and a positive review 4.6. A+ rating with a negative outlook would be assigned a value of 5.2 and a negative review 5.4.
Information on GDDS participation and SDDS subscription was drawn from the IMF’s Dissemination Standards Bulletin Board (DSBB) website, and IMF records for information on the effective dates of financial arrangements.
The International Country Risk Guide (ICRG), from the Political Risk Services Group, Inc., served as the source for indicators of law and order and bureaucratic quality. The institutional quality index used in this study is the sum of these two components of the ICRG’s overall political risk rating. The law and order indicator ranges from 1 to 6 and bureaucratic quality from 0 to 4. In this chapter’s sample of countries and time period, the institutional quality variable varies from zero to 10, with a mean of 5.3 and standard deviation of 2.2. We experimented with other components of the ICRG’s overall political risk rating, but found that the indicators for law and order and bureaucratic quality produced the most robust estimates of the effect of institutional quality. Estimation with additional components of the political risk rating produced negligible changes in the estimated coefficients of the GDDS and SDDS dummy variables.
Alphanumeric Credit Ratings and Assigned Numerical Ratings
Alphanumeric Credit Ratings and Assigned Numerical Ratings
Standard and Poor’s | Moody’s | Fitch | Description | Assigned Numerical Value |
---|---|---|---|---|
AAA | Aaa | AAA | Highest quality | 1 |
AA+ | Aa1 | AA+ | High quality | 2 |
AA | Aa2 | AA | — | 3 |
AA- | Aa3 | AA- | — | 4 |
A+ | A1 | A+ | Strong payment capacity | 5 |
A | A2 | A | — | 6 |
A- | A3 | A- | — | 7 |
BBB+ | Baa1 | BBB+ | Adequate payment capacity | 8 |
BBB | Baa2 | BBB | — | 9 |
BBB- | Baa3 | BBB- | — | 10 |
BB+ | Ba1 | BB+ | Likely to fulfill obligations | 11 |
BB | Ba2 | BB | — | 12 |
BB- | Ba3 | BB- | Ongoing uncertainty | 13 |
B+ | B1 | B+ | High-risk obligations | 14 |
B | B2 | B | — | 15 |
B- | B3 | B- | — | 16 |
CCC+ | Caa1 | CCC+ | Current vulnerability to default | 17 |
CCC | Caa2 | CCC | — | 18 |
CCC- | Caa3 | CCC- | — | 19 |
C | Ca | DD | In bankruptcy or default | 20 |
SD | D | DDD | — | 21 |
Alphanumeric Credit Ratings and Assigned Numerical Ratings
Standard and Poor’s | Moody’s | Fitch | Description | Assigned Numerical Value |
---|---|---|---|---|
AAA | Aaa | AAA | Highest quality | 1 |
AA+ | Aa1 | AA+ | High quality | 2 |
AA | Aa2 | AA | — | 3 |
AA- | Aa3 | AA- | — | 4 |
A+ | A1 | A+ | Strong payment capacity | 5 |
A | A2 | A | — | 6 |
A- | A3 | A- | — | 7 |
BBB+ | Baa1 | BBB+ | Adequate payment capacity | 8 |
BBB | Baa2 | BBB | — | 9 |
BBB- | Baa3 | BBB- | — | 10 |
BB+ | Ba1 | BB+ | Likely to fulfill obligations | 11 |
BB | Ba2 | BB | — | 12 |
BB- | Ba3 | BB- | Ongoing uncertainty | 13 |
B+ | B1 | B+ | High-risk obligations | 14 |
B | B2 | B | — | 15 |
B- | B3 | B- | — | 16 |
CCC+ | Caa1 | CCC+ | Current vulnerability to default | 17 |
CCC | Caa2 | CCC | — | 18 |
CCC- | Caa3 | CCC- | — | 19 |
C | Ca | DD | In bankruptcy or default | 20 |
SD | D | DDD | — | 21 |
References
Cady, John, 2005, “Does SDDS Subscription Reduce Borrowing Costs for Emerging Market Economies?” IMF Staff Papers, Vol. 52, No. 3, pp. 503–17.
Blundell, Richard, and Monica Costa Dias, 2000, “Evaluation Methods for Non-Experimental Data,” Fiscal Studies, Vol. 21, No. 4, pp. 427–68.
Cantor, Richard, and Frank Packer, 1996, “Determinants and Impact of Sovereign Credit Ratings,” Economic Policy Review, Federal Reserve Bank of New York, Vol. 2, (October), pp. 37–53.
Christofides, Charis, Christian Mulder, and Andrew Tiffin, 2003, “The Link Between Adherence to International Standards of Good Practice, Foreign Exchange Spreads, and Ratings,” IMF Working Paper 03/74 (Washington: International Monetary Fund).
Edwards, Sebastian, 1984, “LDC’s Foreign Borrowing and Default Risk: An Empirical Investigation,” American Economic Review, Vol. 74, No. 4, pp. 726–34.
Eichengreen, Barry, 1999, Toward a New International Financial Architecture (Washington: Institute for International Economics).
Eichengreen, Barry, Kenneth Kletzer, and Ashoka Mody, 2005, “The IMF in a World of Private Capital Markets,” IMF Working Paper 05/84 (Washington: International Monetary Fund).
Eichengreen, Barry, and Ashoka Mody, 1998, “What Explains Changing Spreads on Emerging-Market Debt: Fundamentals or Market Sentiment?” NBER Working Paper No. 6408 (Cambridge, Massachusetts: National Bureau of Economic Research).
Erbaş, S. Nuri, 2005, “Comment on ‘Do Credit Rating Agencies Add to the Dynamics of Emerging Market Crises?’ by Roman Kraussl,” Journal of Financial Stability, Vol. 1, pp. 438–46.
Financial Stability Forum (FSF), 2000, “Report of the Follow-Up Group on Incentives to Foster Implementation of Standards,” paper presented at the Meeting of the Financial Stability Forum, September 7–8.
Fischer, Stanley, 2002, “Financial Crises and Reform of the International Financial System,” NBER Working Paper No. 9297 (Cambridge, Massachusetts: National Bureau of Economic Research).
Fitch Ratings, 2005, Fitch—Complete Sovereign Rating History, March 2, 2005.
Gelos, Gastón, and Shang-Jin Wei, 2002, “Transparency and International Investor Behavior,” IMF Working Paper 02/174 (Washington: International Monetary Fund).
Glennerster, Rachel, and Yongseok Shin, 2003, “Is Transparency Good for You, and Can the IMF Help?” IMF Working Paper 03/132 (Washington: International Monetary Fund).
Horrigan, J., 1966, “The Determination of Long-Term Credit Standing with Financial Ratios,” Empirical Research in Accounting 1966, Journal of Accounting Research, Vol. 4 (Supplement), pp. 44–62.
Institute of International Finance (IIF), 2002, Does Subscription to the IMF’s Special Data Dissemination Standard Lower a Country’s Credit Spread? (Washington: IIF).
International Monetary Fund (IMF), 2005, Global Financial Stability Report, World Economic and Financial Surveys (Washington), September.
Kamin, Steven B., and Karsten von Kleist, 1999, “The Evolution and Determinants of Emerging Market Credit Spreads in the 1990s,” BIS Working Paper No. 68 (Basel: Bank for International Settlements).
Montfort, Brieuc, and Christian Mulder, 2000, “Using Credit Ratings for Capital Requirements on Lending to Emerging Market Economies: Possible Impact of a New Basel Accord,” IMF Working Paper 00/69 (Washington: International Monetary Fund).
Moody’s Investors Service, 2004, Sovereign Ratings History, March 4, 2004.
Mosely, Layna, 2003, “Attempting Global Standards: National Governments, International Finance, and the IMF’s Data Regime,” Review of International Political Economy, Vol. 10 (May), pp. 321–62.
Standard & Poor’s Ratings Services, 2005, Sovereign Ratings History Since 1975, March 3, 2005.
For example, secondary bond market studies reported by Christofides, Mulder, and Tiffin (2003) and Glennerster and Shin (2003); however, our estimates are much lower than the 200 to 300 basis point decline in spreads for SDDS subscription reported by the Institute of International Finance (2002, Appendix D).
The real, fiscal, financial, and external sectors. Data categories include national accounts, labor markets, price and production indices, general and central government financial operations and debt, central and commercial bank accounts, interest rates, stock markets, balance of payments, international reserves, merchandise trade, international investment position, external debt, and exchange rates.
Includes population, health, education, and poverty indicators.
The DSBB internet address is http://dsbb.imf.org/.
The DQAF is the assessment framework for the preparation of the data module of the IMF’s Report on the Observance of Standards and Codes (also know as “IMF data ROSCs”).
For further discussion of data quality, risk, and uncertainty, see Erbaş (2005).
Prior to the introduction of the euro in 1999, bonds denominated in deutsche marks are considered.
The countries chosen include those subscribing to the SDDS and participating in the GDDS that launched a significant number of foreign currency-denominated bonds during the period under consideration, and for which adequate quarterly macroeconomic data are available. Certain large emerging market countries, including India and Singapore, did not issue sovereign foreign-currency-denominated bonds between 1989 and 2004; certain other countries began issuing bonds following SDDS subscription, thereby providing no basis for before-and-after subscription comparisons, and have not been considered. In addition, the Republic of South Korea’s limited sovereign issues have been supplemented by considering Korean Development Bank bonds.
Bond issuances for Barbados and Panama are not included in the IMF 2005 data, but their inclusion would not significantly change the reported share.
Available via the Internet: http://www.icrgonline.com.
For example, see Blundell and Costa Dias (2000).
For example, see Edwards (1984), Eichengreen and Mody (1998), and Kamin and von Kleist (1999).
Cantor and Packer (1996) provide a concise explanation of this view.
The dummy variable for an IMF-supported program is set equal to one in all quarters that an arrangement was in effect, and zero otherwise. The influence of Paris Club rescheduling histories were similarly investigated, but found to be insignificant.
See the earlier working paper version of this chapter at http://www.imf.org/external/pubs/ft/wp/2006/wp0678.pdf.
Ibid., footnote 15.
Annual data for external debt (public and publicly guaranteed) stock-to-exports ratios, drawn from the World Bank’s Global Development Finance database, were converted to a quarterly frequency (same value for all quarters) then smoothed with the Hodrick-Prescott filter with standard quarterly parameters prior to testing the order of integration.
This estimator is sometimes referred to as the Parks estimator. The procedure employs residuals from a first-stage regression to form an estimate of the variance-covariance matrix and uses this, in a second stage, to perform feasible GLS.
To ensure asymptotic efficiency of estimated standard errors and to facilitate statistical inference, all t-statistics reported in this chapter are derived using panel consistent standard errors robust to cross-section heterosckedasticity and contemporaneous correlation.
Eichengreen and Mody (1998) also find credit rating measures a highly significant explanatory variable in spread equations.
For example, from Baal to A3, according to Moody’s rating system (Table 4A.2).