Determinants of emerging-market bond spreads: cross-country evidence

Determinants of emerging-market bond spreads:
cross-country evidence
Abstract
a
Hong-Ghi Min
a,
*
, Duk-Hee Lee
a
, Changi Nam
,
Myeong-Cheol Park
a
, Sang-Ho Nam
School of Management, Information and Communications University, 109-6 Moonji-dong,
Yusong-gu, Dajeon 305-732, Republic of Korea
b
Department of Economics, Seowon University, Cheong-Ju, Republic of Korea
b
This paper investigates the importance of liquidity and solvency variables in determining bond
spreads in emerging economies. First, we find that liquidity and solvency variables explain most of
the spread variations in 11 emerging economies during the 1990s. Second, the U.S. interest rate and
macroeconomic fundamentals play a significant role for the determination of bond spreads of
emerging economies. Third, it is shown that Latin countries have a negative yield –maturity
relationship.
D 2003 Elsevier Inc. All rights reserved.
JEL classification: E44; F34; G15
Keywords: Bond spread; Emerging market; Liquidity variables; Macroeconomic fundamentals
1. Introduction
Global Finance Journal 14 (2003) 271–286
The World Bank (1997) reports that many countries have taken decisive steps recently
to promote the development of their bond markets, and as a result, corporations are
floating growing amounts of fixed-income securities in international and domestic markets
while steadily reducing their dependence on bank financing. This change in the corporate
financing pattern is caused by the necessity of substantial investments in infrastructure and
capital-intensive projects that require long-term and fixed-rate debt capital.
* Corresponding author. Tel.: +82-42-866-6801; fax: +82-42-866-6806.
E-mail address: hmin@icu.ac.kr (H.-G. Min).
1044-0283/$ - see front matter D 2003 Elsevier Inc. All rights reserved.
doi:10.1016/j.gfj.2003.10.001
a
H.-G. Min et al. / Global Finance Journal 14 (2003) 271–286272
For this reason, the question of how spreads are determined for emerging-market bonds
merits a closer investigation in view of the ongoing turbulence in emerging markets and
the rapidly changing prospects of these countries.
Some of the important previous literatures that are relevant to our study include
Antzoulatos (2000), Edwards (1986), Eichengreen and Mody (1998), Haque, Kumar,
Mark, and Mathieson (1996), and Sachs (1985). Sachs investigated the role of various
macroeconomic policies and fundamentals for the debt crisis and provided the empirical
rationale for using certain economic fundamentals in the determination of the risk
premium in international capital markets. In particular, he emphasized the importance of
trade and exchange rate policy for a developing country’s performance. Edwards
compared the pricing of bonds and bank loans to test whether the two markets were
significantly different. He found that that there was a positive effect of high debt ratios on
the risk premium on less developed countries’ bonds traded in the secondary market. More
recently, Haque et al. investigated the economic determinants of the creditworthiness of
about 60 developing countries. They found that economic fundamentals—the ratio of
nongold foreign exchange reserves to imports and the ratio of the current account to GDP,
growth, and inflation—explained a large amount of variation in credit ratings. An
additional finding was that all developing country ratings were adversely affected by
increases in international interest rates, independent of domestic economic fundamentals.
While Eichengreen and Mody find that changes in fundamentals explain only a fraction of
the spread compression for foreign debt in developing countries, Antzoulatos finds the
important role of liquidity in international capital market for the emerging-market bond
spread determination.
Since previous studies focused on a certain limited number of explanatory variables, the
target of this paper is to test comprehensive list of economic determinants of yield spreads
on the U.S. dollar-denominated fixed-income securities for 11 emerging economies in
Latin America (Argentina, Brazil, Columbia, Mexico, Venezuela) and Asia (China,
Indonesia, Malaysia, Korea, and the Philippines) for the period of 1991–1999. By using
a more comprehensive set of explanatory variables and performing several robust tests, our
results clearly indicate that not only the liquidity and solvency variables but also
macroeconomic fundamentals do indeed matter in determining emerging-market debt
spreads. Finally, we provide evidence that international bond market has witnessed
negative association between maturity and bond yield spread during the period under
investigation for Latin countries.
This paper is organized as follows. In Section 2, we provide a model for the yield
spread determination and a list of liquidity and solvency variables and macroeconomic
fundamentals that may affect the yield spread of the fixed-income securities, as well as
their expected signs. In Section 3, the model is estimated using panel data and the volatility
of the yield spread is analyzed. Section 4 presents our conclusions.
2. The model: determinants of the yield spread
The probability of default is a function of the unsustainability of a given level of
external debt, arising either as a result of short-term illiquidity or long-run insolvency that
is reflected in liquidity problems (see, e.g., Hanson, 1974; Eaton, Gersovitz, and Stiglitz,
1986; Sachs, 1981, 1984). Assuming a risk-neutral lender and following the conventional
model of risk premium (see, e.g., Edwards, 1986), the following reduced-form model of
bond spread determination can be established as:
log s ¼ a þ
X
b
i
x
i
þ e
i
ð1Þ
a ¼ logð1 þ i*Þð2Þ
where s is the yield spread on fixed income securities (spread), i* is the risk-free world
interest rate, x
i
are the economic determinants of the probability of default, b
are the
corresponding coefficients, and e
i
is a stochastic error term. A number of variables for x
have been suggested by previous studies. These include economic variables that measure
the domestic and external economic performance of a country and exogenous shocks that
affect liquidity and solvency of developing countries (see Edwards, 1986; Haque et al.,
1996; Sachs, 1985). To be comprehensive, we selected 18 independent variables and
classified them into the following four groups: (i) liquidity and solvency variables, (ii)
macroeconomic fundamentals, (iii) external shocks, and (iv) dummy variables. See
Appendix A for data sources and definitions.
2.1. Liquidity and solvency variables
The first group of variables relates to a country’s liquidity and solvency problems. In
any given period, lower export earnings, EXG, or higher import expenditures, IMG, can
increase the likelihood of short-term liquidity problems and hence debt-service difficulties;
thus, we expect EXG to have a negative sign, whereas IMG should have a positive
coefficient. While a decline in the growth rate of output can contribute to a long-term
insolvency problem leading to higher spread, on the other hand, a decline in the GDP
growth rate may ameliorate an external liquidity constraint through lower imports and can
lead to a lower spread; therefore, the impact of GDP growth rate (GDPG) on spread is
uncertain.
In most theoretical models of foreign borrowing, the debt–output ratio plays a crucial
role (see, e.g., Edwards, 1984; Hanson, 1974; Harberger, 1980; Sachs, 1984).
Given the
adverse effect of the debt-to-GDP ratio (DGDP) on a country’s ability to service its debt,
ceteris paribus, we expect its coefficient to have a positive sign. The lower the
international reserves-to-GDP ratio (RGDP), the greater will be the threat of a sudden
liquidity crisis, and the riskier lending to this country becomes so we would expect the
coefficient on this variable to be negative
2
(see, e.g., Edwards, 1984; however, Gersovitz,
1985, claimed that the sign would be positive).
1
Two previous studies (e.g., Burton & Inoue, 1985; Sachs, 1981) found that the coefficient for this variable
in their analysis of the bank’s risk premia was insignificant.
2
H.-G. Min et al. / Global Finance Journal 14 (2003) 271–286 273
However, for countries with liberalized financial markets where financial flows are huge and derivatives
play the major role, this may not be a good indicator of liquidity. Furthermore, in a signaling model, a low reserve
ratio may reflect the confidence of borrowing country.
1
i
i
H.-G. Min et al. / Global Finance Journal 14 (2003) 271–286274
Conversely, if the current account balance-to-GDP ratio (CGDP) is positive and higher,
the yield spread will be lower (see, e.g., Sachs, 1981); thus, we expect the sign on CGDP
to be negative. In a given year, the current account deficit equals the increase in a country’s
net liabilities to foreigners, subject to an adjustment for capital gains and losses on
preexisting stocks of assets and liabilities. The cumulative deficit for years [net foreign
assets (NFAs)] should then approximately equal the increase in the country’s net liabilities
over the course of the decade (Sachs, 1985). Thus, an increase in the cumulative current
account deficit (represented by DGDP and NFA) implies an increase in the yield spread.
The debt service to export or debt–service ratio (DSX) measures possible liquidity (as
opposed to solvency) problems faced by a particular country. It is expected that higher
debt service ratios lower the degree of creditworthiness, resulting in a higher yield spread.
2.2. Macroeconomic fundamentals
The second group of variables represents macroeconomic fundamentals, which might
have an impact on the long-term insolvency problem of a country. In recent years, the
extent to which a country has been perceived to be well managed or well disciplined in its
macroeconomic policymaking has given an important influence on the changes of yields it
faces. The inflation rate (INF) can be regarded as a proxy for the quality of economic
management: the higher the INF, the worse the economic management, so the higher the
yield spread. The influence of international developments on a country’s creditworthiness
is examined through two variables that capture the effects of external shocks to a county’s
trade and financial flows. Shocks to a country’s trade flows are represented by changes in a
country’s terms of trade (TOT). We expect that improvements in the TOT, ceteris paribus,
would lead to a lower yield spread.
The real exchange rate (RXI) is also included to measure the trade competitiveness of
an economy. Sachs (1985) demonstrated the importance of exchange rate management and
the trade regime to the debt crisis. Cline (1983) also claimed that inappropriate exchange
rate policies in a number of LDCs were among the most important causes of the debt
crises. Sustained real appreciation of these countries’ currencies played a major role in the
process of overborrowing. A less competitive RXI or appreciation is expected to adversely
affect the yield spread. This negative effect would be especially pronounced in the case of
Latin American countries where overvalued currencies were one of the important causes of
capital flight.
2.3. External shocks
Antzoulatos (2000), Barr and Pesaran (1997), Calvo, Leiderman, and Reinhart (1993),
Dooley, Fernandez-Arias, and Kletzer (1996), and Frankel (1994) have suggested that
changes in international interest rates have been a key factor influencing capital flows to
developing countries in the 1990s. Since higher interest rates affect not only the cost of
3
From the national income identity, the current account balance is the sum of the fiscal balance and the
private saving-investment gap. However, fiscal balance is not included to avoid multicollinearity.
3
new borrowing but also the interest charges on existing debt that is contracted at variable
rates, the 3-month U.S. Treasury bill rate (T-bill) is used (Antzoulatos, 2000) to capture the
effects of external financial developments. Due to the aforementioned adverse liquidity
effects, we expect a positive sign for its coefficient.
4
The real oil price (ROP) is also included in the analysis to represent the effect of
external events on the sample of countries (Raymond & Rich, 1977). During the late 1970s
and early 1980s, oil shocks caused a world recession and increased the demand for capital
in oil-importing countries. Hamilton (1983) observed that all but one postwar U.S.
recession was preceded by oil price increases and found a strong negative correlation
between oil price changes and GNP growth using a multivariate vector autoregression
system. Gisser and Goodwin (1986) and Dotsey and Reid (1992) largely confirmed
Hamilton’s findings. The higher is the ROP, the higher will be the yield spread since it may
cause a world recession and will adversely affect oil-importing countries.
2.4. Dummy variables
Several dummy variables are used in the empirical work presented in the following
section. To account for potential regional differences in spreads, a regional dummy
variable—LATIN for all Latin American countries
5
—is included. The Mexican peso crisis
in 1994 might have driven spreads for all LDCs to higher levels afterwards. To investigate
this potential adverse effect, a period dummy (Y5) is used to distinguish transactions
before 1995 from thereafter. To capture the different effect of issuer types, a private issuer
dummy (IS3) is included to categorize issuer types into public issuer and private issuer. In
sum, the yield spread of fixed-income securities (spread) is a function of 18 independent
variables.
3. Panel estimation
Using pooled data, first, we estimated parameters of the original model with all
explanatory variables; second, we tested the joint hypotheses of zero coefficients on
various sets of variables. Since we could not reject the null hypotheses of zero
restrictions on the different sets of coefficients, the model was reestimated without
these variables.
3.1. Data and summary statistics
The sources and definitions of the data used in this study are reported in Appendix
A. Table 1 presents the summary statistics of several of the economic variables that are
4
World bond issue can be a good proxy for the international liquidity as suggested by Antzoulatos. However,
it does not cover an important item like cross-border lending and is excluded in the analysis.
5
H.-G. Min et al. / Global Finance Journal 14 (2003) 271–286 275
Country dummy can be used to investigate country specific effect; however, since the goal of this research
is cross-country investigation, we do not cover this issue here.
H.-G. Min et al. / Global Finance Journal 14 (2003) 271–286276
Table 1
Summary statistics of key variables
Variable name Region Mean S.E. Skewness Kurtosis
Spread Latin America 378.59 164.9 0.002 0.360
Asia 324.84 115.4 0.571 0.566
Maturity Latin America 5.303 3.94 3.068 14.54
Asia 5.763 3.12 1.439 4.194
Inflation Latin America 416.38 784.85 1.631 0.85
Asia 12.36 18.09 3.123 9.02
DSX Latin America 49.18 116.26 11.48 154.5
Asia 34.9 37.66 2.90 14.42
TOT Latin America 102.7 11.63 0.266 1.768
Asia 103.05 6.83 0.251 0.108
GDP growth rate Latin America 3.54 3.15 0.111 0.348
Asia 6.39 3.58 3.154 15.947
Import growth rate Latin America 4.61 0.92 2.201 81.55
Asia 4.59 1.41 1.622 34.792
Export growth rate Latin America 11.32 8.56 0.434 2.66
Asia 15.62 11.76 0.476 0.52
Number of observations Latin America 870
Asia 176
used to explain bond spreads in our sample of emerging economies from 1991 to
1999.
To identify the regional difference between Latin America and Asia, statistics for both
regions are reported. It is interesting to note that the mean value of the yield spread for the
Latin American countries is higher than that for the Asian countries by 54 basis points and
their standard errors are twice as large as the difference.
6
The DSX of the Latin American
countries is about 40% higher than that of the Asian countries, indicating that the average
Latin American country has a greater potential of liquidity problem. The most notable
difference between these regions is the average INF: The INF of the Latin American
countries is about 35 times greater than that for the Asian countries. The average GDP
growth rate of the Asian countries is almost double the rate of the Latin American
countries and the export growth rate is about 40% higher in the Asian countries. However,
the average maturity of bonds, TOT, and import growth rates are not much different
between the two groups.
Table 2 shows the total bond issues in the international bond market by issuer type from
1990 to 1999.
We can see that in every year, private issuers are responsible for a much greater share of
the world bond market. Additionally, the shares of private issuers display higher growth
rates than the shares of public issuers.
6
Chile, Columbia, Mexico, and Venezuela have relatively lower spreads that other Latin American countries,
which contributed to the lower mean value of the overall Latin American yield spread.
Table 2
World bond issuance in 1990s (in millions of U.S. dollars)
Year Public issuance
a
3.2. Estimation and test of zero restrictions on the model
Private issuance
b
Total
1990
c
63,210 141,168 204,379
1991 102,751 189,831 292,582
1992 119,456 191,601 311,057
1993 205,808 280,777 486,584
1994 196,642 307,929 504,571
1995 201,958 360,541 562,499
1996 270,164 538,728 808,892
1997 297,195 657,970 955,164
1998 467,171 765,592 1,232,763
1999 573,968 1,150,337 1,724,304
Total 2,498,322 4,584,475 7,082,797
Source: Euromoney Bondware.
a
Public issuance includes the following entities: the central government, local authorities, public banks,
public corporations, public finance companies, public utilities, and others.
b
c
H.-G. Min et al. / Global Finance Journal 14 (2003) 271–286 277
Private issuance includes private banks, private corporations, private firms, private utilities, and others.
Yearly breakdown is based on the closing date of bond issuance.
For pooled data, numerous estimation techniques have been developed (Baltige &
Griffin, 1997). Because of the short panel in our data set (11 countries with 19 regressors),
we use a dummy variable model (Judge, Griffiths, Hill, Luckpohl, & Lee, 1985; Taylor,
1980). The model is estimated by OLS and White’s (1980) heteroscedasticity-consistent
standard errors are reported in the parentheses of Table 3.
Some of the estimated coefficients are insignificant; specifically, the regional (LATIN)
and period (Y5) dummy variables, GDPG, ROP, and CGDP are all insignificant.
To examine the robustness of the estimation results, we estimated several different
specifications of the model. The results are reported in the second and third columns of
Table 3. First, we tested the joint hypothesis of zero restrictions on the coefficients of the
CGDP and ROP variables. Using an F test for the zero restrictions on the coefficients of
two variables, we found that F(2,431) = 17,926. Since the P value of this test is .167, we
cannot reject the joint hypothesis that the estimated coefficients of these two variables are
not significantly different from zero. Excluding these two variables, the model is
reestimated and the result is reported in the second column of Table 3. From the first
and second columns of Table 3, we can see that all the estimated coefficients change
within one standard error.
Second, we tested the joint hypothesis of zero restrictions on the coefficients of the
period dummy (Y5), GDPG, and NFA. Using an F test for the zero restrictions on the
coefficients of three variables, we get F(3,431) = 2.199. Since the P value of this test is
.087, we cannot reject the joint hypothesis that the estimated coefficients of three variables
are not significantly different from zero at the 5% critical level. Excluding these variables,
the model is reestimated and the results are reported in the third column of Table 3. We can
see that all the estimated coefficients change within one standard error.
H.-G. Min et al. / Global Finance Journal 14 (2003) 271–286278
Table 3
Pooled estimation of the model
Independent
variable
From the adjusted R
Regression (1) Regression (2) Regression (3)
Constant 6.009** (0.514) 6.421** (0.40) 5.676** (0.470)
IS3 0.276** (0.045) 0.279** (0.046) 0.296** (0.044)
LATIN 0.062 (0.068) 0.086 (0.068) 0.103 (0.081)
Y5 0.067 (0.045) 0.077 (0.044) –
Liquidity and solvency variables
DGDP 0.005** (0.001) 0.007** (0.002) 0.005** (0.001)
RGDP 0.026** (0.005) 0.003** (0.007) 0.024** (0.005)
CGDP 0.021 (0.023) – 0.015 (0.023)
DSX 0.030** (0.003) 0.029** (0.003) 0.030** (0.003)
IMG 0.039** (0.012) 0.039** (0.012) 0.030** (0.012)
EXG 0.011** (0.003) 0.011** (0.003) 0.009** (0.003)
GDPG 0.017 (0.011) 0.009 (0.010) –
NFA 0.022* (0.008) 0.017** (0.009) –
Macroeconomic fundamentals
INF 0.016** (0.003) 0.015** (0.003) 0.017** (0.003)
TOT 0.019** (0.004) 0.019** (0.004) 0.015** (0.004)
RXI 0.164* (0.065) 0.120* (0.053) 0.145* (0.068)
Exogenous shocks
ROP 0.005 (0.005) – 0.003 (0.005)
T-bill 0.039** (0.006) 0.029** (0.005) 0.049** (0.003)
Maturity, etc.
MT 0.015** (0.004) 0.016** (0.004) 0.014** (0.004)
AMT 0.032** (0.009) 0.033** (0.009) 0.031** (0.009)
Adjusted R
2
.649 .647 .646
Regression FF(18,851) = 191.344,
P value=.0000
F(16,853) = 145.745,
P value=.0000
OLS is used for the estimation, and figures in the parentheses are White’s heteroscedasticity-consistent standard
errors.
*Significant estimated coefficient at the 5% critical level.
**Significant estimated coefficient at the 1% critical level.
2
, it is reasonable to conclude that the linear model used in this study
has a good support from the panel data set so that statistical inferences based on this model
estimation are valid. Since the two different specifications of the original model both have
statistical support based upon F tests and since all estimated coefficients change within one
standard error across specifications, we can conclude that the estimation results are robust.
3.3. Estimation results and inference
3.3.1. Dummy variables
From Table 3, we can see that the estimated coefficient of the issuer type dummy
variable (IS3) has the expected positive sign, implying that private sector issuers pay
F(14,855) = 123.214,
P value=.0000
a higher yield spread than public sector issuers.
7
The insignificance of the estimated
coefficient of the regional dummy variable (LATIN) can be attributed to the lower
spread levels of Chile, Columbia, Mexico, and Venezuela, whose transactions
dominated both in frequency and amount during the 1990s.
The Mexican peso crisis in 1994 might have caused a structural shift of the
yield spread to a higher level. According to the J.P. Morgan emerging local market
index, the average yield spread of Mexico rose after early 1994, reaching a peak in
March of 1995 due to unfavorable market sentiment (International Monetary Fund
[IMF], 1996) and then trended downwards till early 1997 until it equaled the level
that had prevailed in early 1994. However, Table 3 indicates that there was no
significant difference in the spread levels before and after the 1994 Mexican crisis
(Y5).
8
This finding is consistent with Antzoulatos (2000), who found that global bond
issuance was not affected by the Mexican peso crisis. This can be explained by two
factors. First, from the end of 1994, the world’s major bond markets witnessed one of the
greatest rallies supported by an environment of declining interest rates, which reflected
optimism about the prospects of U.S. budget deficit reduction. Second, the high levels of
volatility in bond markets, which had emerged with the onset of the turbulence in early
1994 and were sustained by developments during the crisis in emerging markets in early
1995, started to decline to a more normal level during the summer and fall of 1995 (IMF,
1996–2000).
3.4. Liquidity and solvency variables
All the estimated coefficients of the liquidity variables are significant and have
the expected signs. The total DGDP is significant and has the expected sign. A
1% increase in DGDP increases the yield spread by 1.005%. The nongold RGDP
has a significant and negative sign. The growth rate of exports, EXG (imports,
IMG), is negatively (positively) related to the yield spread of fixed-income
securities, with the estimated coefficients being significant at the 1% critical level,
implying that increased export income lessens the liquidity constraint on the
economy.
The estimated coefficient of DSX is significant and has the expected positive sign. This
confirms that the yield spread of developing countries increases with a higher DSX, which
is a measure of the liquidity problem of a country. A 1% increase in the DSX will increase
the yield spread by 1.03%. Finally, NFAs, as measured by the cumulative current account
(NFA), are significant and have the expected negative sign. A 1% increase in NFAs lowers
the spread by 1.022%.
7
Public sector issuers include the central government, local authorities, public banks, public corporations,
public finance entities, public utilities, state/local authorities, and supernational institutions. Private issuers
include private banks, private corporations, private finance entities, and private utilities.
8
H.-G. Min et al. / Global Finance Journal 14 (2003) 271–286 279
We set different time dummy from 1- to 12-month period starting from October 1994, but estimation result
did not change and the coefficient was not significantly different from zero.
H.-G. Min et al. / Global Finance Journal 14 (2003) 271–286280
3.4.1. Macroeconomic fundamentals
The three most important macroeconomic fundamentals determining the yield spread
are the domestic INF, TOT, and RXI. High inflation (INF) in a country implies an
unhealthy macroeconomic situation and causes an increase in the yield spread. The
estimated coefficient is significant and has the expected positive sign. Based on the
estimated coefficient, a 1% increase in the domestic INF is associated with 1.016%
increase in the yield spread.
An improvement in the TOT implies an increase in export earnings and better
repayment capacity, which reduce the yield spread. The estimated coefficient is significant
and has the expected negative sign, with a 1% improvement in the TOT reducing the yield
spread by 1.02%.
The estimated coefficient for the RXI is significant and has the expected positive sign.
This finding might imply that certain countries have maintained their RXIs at a too
competitive level, which caused high inflation and contraction in the economy, thus
increasing the yield spread (see, e.g., Kamin & Rogers, 1997).
3.4.2. External shocks
Like Antzoulatos (2000), who find that U.S. interest rates were an important
determinant of bond flows to the Latin American countries during the 1990s, we find
that the world interest rate, when 3-month U.S. T-bill rate is used as a proxy, is highly
significant. This implies that favorable international capital market, by reducing cost of
borrowing, encourages bond issues because significant portions of bond issuance are tied
to the short-run dollar interest rates. However, the estimated coefficient of the ROP is
insignificant in explaining the determination of the yield spread of the fixed-income
securities during the 1990s.
3.4.3. Maturity
The estimated coefficient of maturity is significant and negative, implying a negative
yield curve. An inverted yield curve occurs when a surge in demand for short-term credit
drives up short-term rates on instruments like T-bills and money-market funds while long-
Table 4
The ratio of short-term debt to total debt for selected emerging economies (%)
Country Brazil Mexico Korea Thailand All developing
countries
1991 21.8 (26.3) 19.2 (21.9) 43.9 (17.2) 33.1 (12.5) 17.9 (280)
1992 18.7 (24.1) 21.9 (24.5) 43.2 (18.5) 35.2 (14.7) 19.2 (313)
1993 21.3 (30.6) 27.6 (36.3) 43.7 (19.2) 31.3 (13.4) 18.8 (335)
1994 20.7 (31.4) 28.1 (39.3) 53.4 (30.4) 29.2 (14.0) 17.9 (345)
1995 19.2 (30.5) 22.5 (37.3) 57.7 (45.3) 32.3 (18.3) 18.3 (428)
1996 19.6 (35.4) 18.9 (29.8) 57.5 (66.6) 39.5 (42.6) 20.7 (464)
1997 17.3 (34.4) 18.7 (27.8) 39.3 (53.8) 34.4 (37.8) 20.1 (468)
1998 12.3 (30.2) 16.5 (26.3) 20.2 (28.1) 28.3 (29.7) 16.0 (410)
1999 12.1 (29.5) 14.4 (24.1) 26.8 (34.7) 24.3 (23.4) 15.9 (406)
The figures in the parentheses are short-term debt in billions of U.S. dollars. Source: World Bank, Global
Development Finance, 1997– 2001.
H.-G. Min et al. / Global Finance Journal 14 (2003) 271–286 281
Table 5
Spread-maturity regressions for Latin countries
Latin American countries: private and public issuers; number of observations =870
Spread = 418.92** (8.12) 0.51**(MT)
Latin American countries: private issuers; number of observations =661
Spread = 470.14** (13.56) 1.34* (MT)
Latin American countries: public issuers; number of observations = 209
Spread = 312.16** (13.19) 0.17**(MT)
2
2
2
(0.15) Adjusted R
(0.54) Adjusted R
(0.054) Adjusted R
=.026
White’s heteroscedasticity-consistent covariance matrix estimation is used for the standard errors.
* Significant estimated coefficient at the 5% critical level.
**Significant estimated coefficient at the 1% critical level.
term rates move up more slowly since borrowers are not willing to commit themselves to
paying high interest rates for many years in the future. This happened in the late 1970s and
early 1980s (Edwards, 1986) when short-term interest rates were around 20% whereas
long-term rates rose to only about 17%.
Also, there was a sustained surge in short-term borrowing by Korea, Mexico, and
Thailand throughout the 1990s, whose transactions dominated the world bond market
in both their frequency and amounts. Table 4 shows that the ratios of short-term debt
to total debt for these three countries are much higher than those of the developing
countries as a whole. This increased demand for short-term capital brought about a
Table 6
Spread-maturity regressions for five Latin American countries
Argentina
Number of observations = 170
Spread = 462.4** (14.60) 1.88** (MT)
Brazil
Number of observations = 343
Spread = 519.77** (13.18) 1.48** (MT)
Columbia
Number of observations = 53
Spread = 193.71** (25.46) 0.08 (MT)
Mexico
Number of observations = 249
Spread = 327.69** (10.97) 0.22 **(MT)
Venezuela
Public issuers; number of observations = 55
Spread = 277.26** (21.66) + 0.10 (MT)
2
2
2
(0.51) Adjusted R
(0.51) Adjusted R
(0.37) Adjusted R
2
2
(0.04) Adjusted R
(0.21) Adjusted R
=.034
White’s heteroscedasticity-consistent covariance matrix estimation is used for the standard errors.
**Significant estimated coefficient at the 1% critical level.
2
2
2
=.06
=.096
2
2
2
2
2
=.148
=.062
=.045
=.083
H.-G. Min et al. / Global Finance Journal 14 (2003) 271–286282
Table 7
Correlation matrix of spreads and independent variables
SPD DGDP RGDP ROP INF NFA TOT DSX RXI MT AMT GDPG
A. Level
SPD 1.000
DGDP .561 1.000
RGDP .529 .142 1.000
ROP .192 .386 .036 1.000
INF .520 .213 .737 .132 1.000
NFA .154 .114 .421 .106 .338 1.000
TOT .387 .291 .764 .096 .637 .657 1.000
DSX .324 .188 .087 .034 .179 .149 .348 1.000
RXI .020 .277 .085 .500 .100 .097 .159 .247 1.000
MT .460 .129 .290 .215 .244 .123 .215 .209 .042 1.000
AMT .392 .066 .263 .085 .206 .079 .162 .211 .005 .374 1.000
GDPG .141 .336 .051 .236 .108 .433 .138 .193 .007 .123 .103 1.00
B. S.D. (time varying)
SPD 1.000
DGDP .638 1.000
RGDP .537 .944 1.000
ROP .183 .574 .523 1.000
INF .613 .765 .512 .453 1.000
NFA .025 .001 .062 .447 .107 1.000
TOT .475 .751 .804 .418 .392 .254 1.000
DSX .119 .243 .385 .088 .123 .421 .667 1.000
RXI .174 .267 .076 .451 .534 .033 .006 .454 1.000
MT .288 .276 .107 .308 .514 .003 .219 .570 .299 1.000
AMT .148 .036 .164 .103 .426 .037 .305 .349 .008 .665 1.000
GDPG .005 .095 .071 .486 .093 .199 .230 .280 .276 .602 .097 1.000
S.D. is calculated using 24 observations each time. Bold figures are insignificant at 5% critical level. The
variables are defined in Appendix A; in addition, SPD = spread.
negative yield curve in the international bond market in the 1990s. Since 1992, the
average yield spread has been decreasing while the average maturity has been
increasing.
9
These changes reflect the increased supply of funds into the emerging
economies and the resulting decreased time-varying liquidity premia (Mankiw &
Summers, 1984). Consequently, some emerging economies overborrowed—even those
whose rate of return on investment was quite low.
9
10
Average yield spread decreased from 409.1 in 1993 to 298.9 in 1999, while average maturity has increased
from 4.19 years in 1993 to 8.38 years in 1999.
10
All major recipients of capital flows saw a dramatic surge in private capital inflows during the 1990s, and
this surge was extremely large in relation to the size of the economies. Total net capital inflows into the all
developing countries were US$8.8 billion for the period of 1983– 1989, while total net capital inflows for the
period of 1990–1999 were US$204.9.
Table 5 presents results from regressions that examine the relationship between a
bond’s maturity and the yield spread. The Latin American countries have significant and
negative yield–spread relationships (see Eqs. 1–3 in Table 5).
Table 6 presents regressions that examine the spread–maturity relationship for
individual Latin American countries. Argentina, Brazil, and Mexico have a significant
negative yield curve.
3.4.4. Volatility analysis
11
Table 7 presents a correlation matrix that was used to investigate whether the
volatility of bond spreads is systematically affected by certain factors. This table
reports correlation coefficients of levels of spread (in the upper panel A) and standard
deviations of spread (in the lower panel B) with the variables that proved significant
in the regressions of Table 3. The lower section shows that liquidity and macroeconomic
fundamentals affect spread volatility. Other than NFA and the GDPG, all the
estimated correlation coefficients measured by standard deviations are significant. In
addition, the volatility of spread is highly correlated with the DGDP, RGDP, and
domestic INF. These results imply that not only the level but also the volatility of the
bond spread is significantly and positively affected by these three economic
fundamentals.
4. Concluding remarks
This paper has investigated the determinants of bond spreads for emerging markets.
Our key finding is that emerging economies’ liquidity-related variables play very
important role for the bond spread determination based on test of zero restrictions.
We identify other groups of important explanatory variables for the cross-country
differences in bond spreads. Several macroeconomic fundamentals like domestic INF,
NFAs (as measured by the cumulative current account), TOT, and RXI are found to be
significant for the determination of the yield spreads. However, external shocks when
measured by the ROP are found to be insignificant while international interest rate is
significant in determining yield spreads. This finding implies that the variation in
benchmark rates does matter. Finally, the Latin American countries appear to have an
inverted yield curve.
The lessons for developing economies seeking better access to the international bond
market are clear. Sound management of their liquidity-related variables, especially the
RGDP, and containing the domestic INF to a low level make difference for their
creditworthiness in the international bond market. Potential extension of this paper
includes predicting the business cycle of specific countries using their yield spreads
(Gertler & Lown, 2000).
11
H.-G. Min et al. / Global Finance Journal 14 (2003) 271–286 283
Spread and maturity can be both endogenous and can be determined by a third variable, like world bond
issuance.
H.-G. Min et al. / Global Finance Journal 14 (2003) 271–286284
Acknowledgements
Earlier version of this paper was prepared when the first author was an economist
at the World Bank. We are grateful to James Barth, Gerard Caprio, L. Hernandez,
Ronald Johannes, Frederick Kilby, and other seminar participants at the World Bank
for their helpful comments and to C. Michelitsh and Eung J. Kim for providing bond
spread data. Especially, we are grateful to an anonymous referee for his/her invaluable
comments.
Appendix A
Dependent variable
Spread or SPD Yield spread data are from Euromoney Bondware. The spread is defined as the number of
basis points that a fixed-rate issue yields above or below a comparable (in duration)
government bond at its launch price.
Dummy variables
IS3 Issuer type dummy; 1 if private issuer, 0 otherwise.
LATIN Regional dummy; 1 if Latin American countries, 0 otherwise.
Y5 Issue period dummy; 1 if issues in 1995, 0 otherwise.
Liquidity and solvency variables
DGDP Ratio of total external debt (World Debt Tables) to GDP (IMF’s International Financial
Statistics, IFS, line 99.b) converted to U.S. dollars by the exchange rate in IFS
line ae/we.
RGDP Ratio of international reserves (IFS line 1l.d) to GDP.
CGDP Ratio of current account (IFS line 77.ad) to GDP.
DSX Debt service (World Debt Tables) to export (IMF’s IFS line 70 converted to U.S.
dollars by the exchange rate in IFS line ae/we).
IMG Growth rate of imports (IFS line 71).
GDPG Growth rate of GDP (IFS line 99.b).
EXG Growth rate of exports (IFS line 70).
NFA Net foreign assets measured by the cumulated current account deficit/surplus with a
benchmark figure from 1989.
Macroeconomic fundamentals
TOT Terms of trade calculated by dividing export price (IFS line 76) by import price
(IFS line 76.x). For those countries whose value is missing in IFS, we obtained
the export price by dividing current export (import) of goods and nonfactor services by
1987 constant price export (import) of goods and nonfactor
services in the World Bank database.
INF Annual inflation rate measured by CPI (IFS line 64).
RXI Nominal exchange rate (IFS line ae/we) adjusted by CPI (IFS line 64).
External shocks
ROP Real oil price, which is the average crude oil price (IFS line 001) deflated by G-7 inflation
rates (from the World Bank database).
T-bill 3-month U.S. Treasury bill rate (IFS line 60.c).
Appendix A (continued)
Debt-related variables
MT Maturity of a bond (Euromoney Bondware).
AMT Amount of a bond (Euromoney Bondware).
Other data series
Emerging local market index for Mexico was obtained from the J.P. Morgan
Web address: http://www.jpmorgan.com.
References
H.-G. Min et al. / Global Finance Journal 14 (2003) 271–286 285
Antzoulatos, A. (2000). On the determinants and resilience of bond flows to LDCs, 1990– 1995. Journal of
International Money and Finance, 19, 399– 418.
Baltige, B., & Griffin, J. (1997). Pooled estimators vs. their heterogeneous counterparts in the context of dynamic
demand for gasoline. Journal of Econometrics, 77, 303–327.
Barr, D. G., & Pesaran, B. (1997). An assessment of the relative importance of real interest rates, inflation, and
term premium in determining the prices of real and nominal U.K. bonds. Review of Economics and Statistics,
79, 362– 366.
Burton, F., & Inoue, H. (1985). The influence of country risk factors on the interest differentials on international
lending to sovereign borrowers. Applied Economics, 17, 491– 507.
Calvo, G., Leiderman, L., & Reinhart, C. (1993). Capital inflows and real exchange rate appreciation in Latin
America: The role of external factors. IMF Staff Papers, 40, 108– 151.
Cline, W. (1983). Interest and debt: Systematic risk and policy response, Institute for International Economics.
Washington, DC: MIT Press.
Dooley, M., Fernandez-Arias, E., & Kletzer, K. (1996). Is the debt-crisis history? Recent private capital inflows to
developing countries. World Bank Economic Review, 10, 27– 50.
Dotsey, M., & Reid, M. (1992). Oil shocks, monetary policy, and economic activity. Federal Reserve Bank of
Richmond Economic Review, 78, 14– 27.
Eaton, J., Gersovitz, M., & Stiglitz, J. (1986). The pure theory of country risk. European Economic Review, 30,
481– 513.
Edwards, S. (1984). LDC foreign borrowing and default risk: An empirical investigation, 1976– 1980. American
Economic Review, 74, 726– 734.
Edwards, S. (1986). The pricing of bonds and bank loans in international markets: An empirical analysis of
developing countries’ foreign borrowing. European Economic Review, 30, 565– 589.
Eichengreen, B. & Mody, A. (1998). What explains changing spreads on emerging market debt: Fundamentals
or market sentiment? NBER Working Paper No. 6408.
Frankel, J. (1994). Why haven’t interest rates in Latin American and Asian countries converged to world levels?
Recent portfolio capital inflows and stabilization. Unpublished manuscript.
Gersovitz, M. (1985). Banks’ international lending decisions: What we know and implications for future research.
In Smith G., & Cuddington J. (Eds.), International debt and the developing countries. Washington,
DC: World Bank.
Gertler, M. & Lown, C. (2000). The information in the high yield bond spread for the business cycle: Evidence
and some implications. NBER Working Papers No. 7549.
Gisser, M., & Goodwin, T. (1986). Crude oil and the macroeconomy: Tests of some popular notions. Journal of
Money, Credit and Banking, 18, 94– 103.
Hamilton, J. (1983). Oil prices and the macroeconomy since the World War II. Journal of Political Economy, 91,
228– 248.
Hanson, J. (1974). Optimal international borrowing and lending. American Economic Review, 64, 616–630.
Haque, N., Kumar, M., Mark, N., & Mathieson, D. (1996). The economic contents of indicators of developing
country creditworthiness. IMF Staff Papers, 43, 688– 724.
Harberger, A. (1980). Vignettes on the world capital market. American Economic Review, 70, 331– 337.
H.-G. Min et al. / Global Finance Journal 14 (2003) 271–286286
International Monetary Fund (1996 – 2000). International capital markets: Developments, prospects and policy
issues. Washington, DC: IMF.
Judge, G., Griffiths, W., Hill, R., Luckpohl, H., & Lee, T. (1985). The theory and practice of econometrics. New
York: Wiley.
Kamin, S. & Rogers, J. (1997). Output and the real exchange rate in developing countries: An application to
Mexico. International Finance Discussion Paper No. 580, Board of Governors of the Federal Reserve System,
Washington, DC.
Mankiw, N., & Summers, L. (1984). Do long-term interest rates overreact to short-term interest rates? Brookings
Papers on Economic Activity, 1, 223–242.
Raymond, J., & Rich, R. (1977). Oil and the macroeconomy: A Markov state-switching approach. Journal of
Money, Credit and Banking, 29, 193– 213.
Sachs, J. (1981). The current account and macroeconomic adjustment in the 1970s. Brookings Papers on
Economic Activities, 1, 201– 268.
Sachs, J. (1984). Theoretical issues in international borrowing. Princeton studies in International Finance, 54.
Sachs, J. (1985). External debt and macroeconomic performance in Latin America and East Asia. Brooking
Papers on Economic Activities, 2, 523– 564.
Taylor, W. (1980). Small sample considerations in estimation from panel data. Journal of Econometrics, 13,
203–223.
White, H. (1980). A heteroscedasticity-consistent covariance matrix estimator and a direct test for heteroscedas-
ticity. Econometrica, 48, 817–838.
World Bank (1997). Private capital flows to developing countries. Washington, DC: World Bank.