Working Paper. An Analysis of Emerging Market Spreads NO.3. Shin Oya. November 2001 JBIC INSTITUTE JAPAN BANK FOR INTERNATIONAL COOPERATION

Transcription

1 JBICI Working Paper An Analysis of Emerging Market Spreads Shin Oya NO.3 November 21 JBIC INSTITUTE JAPAN BANK FOR INTERNATIONAL COOPERATION

2 JBIC Working Paper is based on the research done by staffs of Japan Bank for International Cooperation (JBIC) and published by JBIC Institute. Views expressed in this paper are those of the author and do not reflect those of JBIC.

3 An Analysis of Emerging Market Spreads Shin Oya Shin Oya is at the Japan Bank for International Cooperation (JBIC). The views expressed are those of the author and do not necessarily represent those of the JBIC. 2 The author would like to thank Professor William C. Handorf, Associate Professor Hidenobu Okuda, Messrs. Yoshito Miyairi and Kaoru Hayashi for helpful comments and suggestions. Any errors that remain are the author s responsibility.

4 [Abstract] We regressed JP Morgan EMBI plus on GDP Growth, Per Capita GDP, Debt Service Ratio, External Balance, Capitalization/GDP. Except for Debt Service Ratio, these independent variables have statistically significant coefficients. A positive coefficient for external balance can be interpreted to suggest that less creditworthy countries are forced to maintain a positive external position because they cannot import capital. 1. Introduction The financial crisis of 1997 and 1998 demonstrated to us how significant the change in country risk has become. It also taught us the difficulty in predicting such changes. This paper tries to capture the macroeconomic aggregates that explain the level of country risk. We use secondary market spreads of emerging market sovereign bonds as an indicator of country risk. We choose several macroeconomic aggregates as independent variables. We regress emerging market spreads on these macroeconomic aggregates and analyze the results. While we believe that the characteristics of individual bonds (maturity, existence of call option, issue size) are important elements for an individual bond s spread, we disregard these characteristics to focus on general country risk. We used the JP Morgan EMBI Plus (Emerging Market Bond Index Plus) as dependent variables. 2. Literature Mauro, Sussman, and Yafeh (2) analyzes yield spreads on sovereign debt issued by emerging markets using modern data from the 199 s and newly-collected historical data on debt traded in London during They show that the co-movement of spreads across emerging markets is higher today than it was in the historical sample. They also show that sharp changes in spreads today tend to be mostly related to global events, whereas country specific events played a bigger role in They use the JP Morgan EMBI (Emerging Markets Bond Index) and EMBI Plus as the modern data on emerging market spread as is used in this analysis. However, while my analysis is based on regression, their analysis is expressed as the difference of spread co-movements between then versus now. Handorf and Amira (1999) analyze Eurobonds issued by Latin American sovereigns. After separating data between A-rated and Higher and Baa-rated and Lower, they use linear regression analysis. In other words, they basically keep credit risk constant for regression. Their dependent variable is the relative yield spreads of each bond. Independent variables are maturity, issue size, call option, repeated borrowing experience, currency and country. The study shows these characteristics affect relative yield spreads.

5 Eichengreen and Mody (1998) analyze nearly one thousands bonds issued by developing countries during Their dependent variable is log spread (launch spread). Their independent variables include credit rating residual, Debt/GDP ratio, Dummy for debt rescheduling, Debt Service Ratio, Reserve/GDP ratio and Deficit/GDP ratio. The result shows, they argue, that measures of higher credit quality lead to a higher probability of issue and to a lower spread. Observed change in fundamentals, however, explains only a small portion of the spread compression in the period leading to the recent crisis in emerging markets. Stone (1991) analyzes the influence of non-quantifiable information. He uses Institutional Investors country ratings as dependent variables and considers Debt/GDP, Reserve/Imports, Real GDP and log CPI as independent variables. He then regresses secondary loan market prices, dependent variables, using the same set of independent variables and including the residual of first regression. Comparison between the first and second regression shows that r-squared increases significantly, from 47% to 74%. Based on this, he concludes that market participants use both macroeconomic aggregates and unmeasurable information to price LDC loans. Haque, Kumar, Mark, and Mathieson (1996) analyze ratings by Institutional Investors, Euromoney, and the Economic Intelligence Unit. The dependent variable is rating and independent variables are many ratios which demonstrate economic fundamentals, such as growth of export, current account balance, international reserves as a ratio of imports, debt to GDP ratio, real exchange rate, GDP growth, inflation, etc. The resulting analysis is that economic fundamentals have played a key role in determining a developing country s credit rating. The study also concludes that the most important economic variables influencing a country s credit rating are the ratio of non-gold reserves to imports, the ratio of current account balance to GDP, GDP growth, and inflation. In comparison to this literature, features of our analysis are the following: (i) Our study focuses on country risk, not individual bonds characteristics. Therefore, we use EMBI Plus, and not individual bond prices, as dependent variables. We do not include bond characteristics, such as issue size, call feature or maturity, as independent variables. Readers who are interested in these features should consult Handorf and Amira (1999). (ii) We used secondary market spreads (EMBI Plus) rather than primary market spreads. This, we believe, makes our regression analysis more robust in analyzing country risk. (iii) We do not use country risk rating either as dependent or independent variables. While rating activity itself is the issue that deserves more attention by researchers, our study focuses on the relationship between country spread and macroeconomic aggregates. While we consider Haque, Kumar, Mark, and Mathieson (1996) an informative and excellent study, defining qualitative information is always difficult because there is a possibility that measurable information can be overlooked.

6 3. Data and Framework (1) Feature of Data As explained above, we chose JP Morgan EMBI Plus as the dependent variable. Although some EMBI countries have only limited years of data, this analysis uses EMBI data from the end of 1993 to beginning of 21. We used spreads traded at the end of each year, more specifically, trading dates of December 31, 1993; December 3, 1994; December 29, 1995; December 31, 1996; December 31, 1997; December 31, 1998; and December 31, Regarding independent variables, we consider some macroeconomic aggregates that may affect EMBI spreads. We construct time-series and cross sectional linear regression. Independent variables were GDP Growth, Per Capita GDP, Debt Service Ratio, External Balance and Capitalization/GDP. These independent variables are annual data, and excerpted from the World Bank s World Development Indicators 21. Appendix 1 is attached to show these data. As you notice from Appendix 1, some countries have data for seven years, but other countries have only one-year data. This is due to the limitation of either spread data (dependent variable), or World Bank data (independent variables). Specific country-year data that lacks either (i) the spread data, or (ii) two or more independent variables data was not used. If specific country-year data lacks only one independent variable, we compensated by our best estimate. Fortunately, the only data determined by this method is Bulgaria market capitalization data for We assumed that Bulgaria s 1994 market capitalization level was the same as 1995 (.5% of GDP). The number of observations is 81. For exact definitions of independent variables, please see Appendix 2. Please also see Table 1 below depicting the mean, minimum, maximum, and standard deviation of the dependent and independent variables. (Table 1 is the same as the last five rows of the Appendix 1.) Table 1 Feature of Data Spread Growth Capita DSR Exter Capitali Mean Minimum Maximum Standard Deviation

7 (2) Economic Rationale The economic rationale for choosing these data as independent variables is based on the following: (i) GDP Growth (Annual %) GDP Growth rate is the single most important indicator of strength of the economy. Economic growth generally accompanies increased tax revenue and this strengthens a government s capacity for debt repayment. Growth usually (if not always) also accompanies increase in exports and this makes a country s ability to make external payments more solid. Growth indicates that countries economic management is efficient and economic policy is appropriate. This reduces risk perception from the point of view of the external financiers and reduces required rate of return (and reduces spread). We expect high growth rate reduce spread. (ii) Per Capita GDP It is well known that developed countries with high per capita GDP have a smaller spread than developing countries with low per capita GDP. High per capita GDP means a country is less vulnerable to external shock and that it has political stability mainly due to broad middle class as well as efficient economic activity. We expect that the correlation coefficient of per capita GDP is negative. (iii) Debt Service Ratio Ex ante, high debt service ratio means that a county has to pay a high ratio of interest and principal in comparison to exports. This may cause difficulties in continuing external payments, and that the market require a high spread for purchasing bonds issued by these countries. However, debt service ratio measures actual external payments vis-à-vis actual exports. Therefore, if a country stops external payments or agrees with debt rescheduling, the debt service ratio declines. It is difficult to say which effect is more influential as economic theory. (iv) External Balance External balance is export minus import, divided by GDP. Common sense is that high positive external balance shows high ability for external payment, and therefore a smaller spread. However, if a country is hit by financial crisis and forced to be a capital export country, that country has no choice but to contract its imports in order to make its external balance positive and meet external payment obligations. It is difficult to predict the sign for the correlation coefficient of an external position. (v) Capitalization/GDP

8 Market capitalization is the share price times the number of shares outstanding. We divide this by GDP. This indicates the development of stock market. Development of the domestic stock market means the county can utilize domestic savings and is less vulnerable to external shock. We believe that the correlation coefficient is negative. Annual data for these dependent variables are available through the World Bank s World Development Indicators. For exact input data, please see Appendix 1. Our linear regression model is as follows: EMBI Spread = function (GDP Growth, Per Capita GDP, Debt Service Ratio, External Balance, Capitalization/GDP) 4. Results (1) Result The result of regression is below. Please see Table 2 for t-value, p-value, f-value, r-square, and comparison with expected sign and actual sign. EMBI Spread = GDP Growth.63 Per Capita GDP.73 Debt Service Ratio External Balance 5.68 Capitalization/GDP The Valiance Inflation Factor is small and there is no sign of high multicollinearity. Durbin Watson d statisitic is close to two (it is 1.832) and there is no sign for autocorrelation. Sign Expected Actual Table 2 Regression Results Spread Coefficient t-value p-value (%) Statistical Significance Intercept clearly yes GDP Growth clearly yes Per Capita GDP yes Debt Service Ratio? weak External Balance? clearly yes Capitalization/GDP clearly yes f-value 6.98 p-value of f (%).1 r-square.289 Adj r-square.2416 Durbin Watson d Sources: JP Morgan and the World Bank, World Development Indicators 21

9 (2) Interpretation GDP Growth, Per Capita GDP, and Capitalization/GDP have the same sign as expected. Coefficients for GDP Growth and Capitalization/GDP are statistically significant at 5% confidence level. The statistical significance for Per Capita GDP is a little weak, but still significant at 1% confidence level. Debt Service Ratio has very small t-value and large p-value. We cannot deny a null hypothesis that the actual coefficient is zero. As previously discussed, it is possible to interpret this as an ex post effect compensating for an ex ante effect. The sign for External Balance is positive and statistically significant. This shows that less creditworthy countries are forced to maintain a positive external position because they cannot import capital. This result may be related to the fact that our data period covers the time when investors were extremely sensitive to country risk following the 1994 peso crisis and 1997 and 1998 Asia financial crises. Our r-square is.29. This means that 29% of the spread can be explained by GDP Growth, Per Capita GDP, Debt Service Ratio, External Balance and Capitalization/GDP. While the question of whether 29% is small or not is a judgement call, it is true that there are many other elements that decide country spread. This may indicate that unmeasurable non-quantifiable information is important, as suggested by Stone (1991). Our sample period coincides with the period for which many commentators argue an increase of comovement (Mauro, Sussman, and Yafeh, 2) and the existence of a self-fulfilling prophecy. This herd behavior by investors can be a reason that reduces the explanatory power of simple macroeconomic aggregates. 5. Conclusion We regressed JP Morgan EMBI plus on GDP Growth, Per Capita GDP, Debt Service Ratio, External Balance, Capitalization/GDP. Except for Debt Service Ratio, these independent variables have statistically significant coefficients. Given a positive coefficient for external balance, it can be interpreted that less creditworthy countries are forced to keep positive external position, because they cannot import capital.

10 Selected References Eichengreen, Barry and Ashoka Mody What explains Changing Spreads on Emerging Market Debt: Fundamentals or Market Sentiment? RMC Discussion Paper Series 123, The World Bank Eichengreen, Barry and Ashoka Mody. 2 Lending boom, reserves and sustainability of short-term debt: inferences from the pricing of syndicated bank loans Journal of Development Economics Vol. 63 (2): Handorf, William C and Khaled Amira Sovereign Latin America Eurobonds Economia Aplicada, Vol. 3, No. 4, October-December 1999: Haque, Nadeem U, Manmohan S. Kumar, Melson Mark and Donald J. Mathieson The Economic Content of Indicators of Developing Country Creditworthiness IMF Working Paper WP/96/9 Lensink, Robert, Niels Hermes and Victor Murinde. 2 Capital flight and political risk Journal of International Money and Finance 19 (2): Mauro, Paolo, Nathan Sussman, and Yishay Yafeh. 2 Emerging Market Spreads: Then Versus Now IMF Working Paper WP//19 Stone, Mark R On the Information Content of LDC Secondary Loan Market Prices Unpublished manuscript, IMF 1991 World Bank World Development Indicators 21

11 Appendix1 Data with Mean, Min, Max spread Growth capita dsr exter capitali Argentina Argentina Argentina Argentina Argentina Argentina Argentina Brazil Brazil Brazil Brazil Brazil Brazil Bulgaria Bulgaria Bulgaria Bulgaria Bulgaria Bulgaria Colombia Ecuador Ecuador Ecuador Ecuador Ecuador Korea, Rep Korea, Rep Mexico Mexico Mexico Mexico Mexico Mexico Mexico Morocco Morocco Morocco Morocco Morocco Morocco Morocco Nigeria Nigeria

12 Nigeria Nigeria Nigeria Nigeria Nigeria Panama Panama Panama Panama Peru Peru Peru Philippines Philippines Philippines Philippines Philippines Philippines Poland Poland Poland Poland Poland Poland Russian Federation Russian Federation Russian Federation Turkey Venezuela, RB Venezuela, RB Venezuela, RB Venezuela, RB Venezuela, RB Venezuela, RB Venezuela, RB South Africa South Africa South Africa spread Growth capita dsr exter capitali Mean Minimum Maximum Standard Deviation

13 Appendix 2 Definition of Independent Variable Following definitions are from the World Bank World Development Indicators 21 CD-ROM. GDP growth (annual %) Annual percentage growth rate of GDP at market prices based on constant local currency. Aggregates are based on constant 1995 U.S. dollars. GDP is the sum of gross value added by all resident producers in the economy plus any product taxes and minus any subsidies not included in the value of the products. It is calculated without making deductions for depreciation of fabricated assets or for depletion and degradation of natural resources. For more information, see WDI table 4.1. GDP per capita (constant 1995 US$) GDP per capita is gross domestic product divided by midyear population. GDP is the sum of gross value added by all resident producers in the economy plus any product taxes and minus any subsidies not included in the value of the products. It is calculated without making deductions for depreciation of fabricated assets or for depletion and degradation of natural resources. Data are in constant U.S. dollars. For more information, see WDI table 1.1. Total debt service (% of exports of goods and services) Total debt service is the sum of principal repayments and interest actually paid in foreign currency, goods, or services on long-term debt, interest paid on short-term debt, and repayments (repurchases and charges) to the IMF. For more information, see WDI table External balance on goods and services (% of GDP) External balance on goods and services (resource balance) equals exports of goods and services minus imports of goods and services (previously nonfactor services). For more information, see WDI table 4.9. Market capitalization of listed companies (% of GDP)

14 Market capitalization (also known as market value) is the share price times the number of shares outstanding. Listed domestic companies are the domestically incorporated companies listed on the country's stock exchanges at the end of the year. For more information, see WDI table 5.3.

15 Appendix 3 Change in Spread (A Group of Countries) Spread bps 年 12 月 1994 年 3 月 1994 年 6 月 1994 年 9 月 1994 年 12 月 1995 年 3 月 1995 年 6 月 1995 年 9 月 1995 年 12 月 1996 年 3 月 1996 年 6 月 1996 年 9 月 1996 年 12 月 1997 年 3 月 1997 年 6 月 1997 年 9 月 1997 年 12 月 1998 年 3 月 1998 年 6 月 1998 年 9 月 1998 年 12 月 1999 年 3 月 1999 年 6 月 1999 年 9 月 1999 年 12 月 2 年 3 月 2 年 6 月 2 年 9 月 2 年 12 月 21 年 3 月 Argentina Brazil Mexico Nigeria Poland Russia

16 Change in Spread (Individual Country) Appendix 4

17 Argentina Argentina Bulgaria Brazil Bulgaria Brazil Colombia Colombia

18 Ecuador Ecuador Korea Morocco Morocco Mexico Mexico Korea

19 Nigeria Nigeria Panama Peru Panama Peru Philippine Philippine

20 Poland Poland Qatar Russia Qatar Russia Turkey Turkey

21 Venezuera Venezuera South Africa South Africa 1993 年 12 月 1994 年 3 月 1994 年 6 月 1994 年 9 月 1994 年 12 月 1995 年 3 月 1995 年 6 月 1995 年 9 月 1995 年 12 月 1996 年 3 月 1996 年 6 月 1996 年 9 月 1996 年 12 月 1997 年 3 月 1997 年 6 月 1997 年 9 月 1997 年 12 月 1998 年 3 月 1998 年 6 月 1998 年 9 月 1998 年 12 月 1999 年 3 月 1999 年 6 月 1999 年 9 月 1999 年 12 月 2 年 3 月 2 年 6 月 2 年 9 月 2 年 12 月 21 年 3 月