fundamentals or market sentiment: what causes country risk?
TRANSCRIPT
This article was downloaded by: [University of Illinois Chicago]On: 10 November 2014, At: 10:04Publisher: RoutledgeInforma Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,37-41 Mortimer Street, London W1T 3JH, UK
Applied EconomicsPublication details, including instructions for authors and subscription information:http://www.tandfonline.com/loi/raec20
Fundamentals or market sentiment: what causescountry risk?Vladimir Kühl Teles a & Maria Carolina Leme aa Sao Paulo School of Economics , Getulio Vargas Foundation (EESP-FGV) , Sao Paulo, BrazilPublished online: 16 Mar 2009.
To cite this article: Vladimir Kühl Teles & Maria Carolina Leme (2010) Fundamentals or market sentiment: what causescountry risk?, Applied Economics, 42:20, 2577-2585, DOI: 10.1080/00036840801964518
To link to this article: http://dx.doi.org/10.1080/00036840801964518
PLEASE SCROLL DOWN FOR ARTICLE
Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) containedin the publications on our platform. However, Taylor & Francis, our agents, and our licensors make norepresentations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of theContent. Any opinions and views expressed in this publication are the opinions and views of the authors, andare not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon andshould be independently verified with primary sources of information. Taylor and Francis shall not be liable forany losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoeveror howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use ofthe Content.
This article may be used for research, teaching, and private study purposes. Any substantial or systematicreproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in anyform to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://www.tandfonline.com/page/terms-and-conditions
Applied Economics, 2010, 42, 2577–2585
Fundamentals or market sentiment:
what causes country risk?
Vladimir Kuhl Teles* and Maria Carolina Leme
Sao Paulo School of Economics, Getulio Vargas Foundation (EESP-FGV),
Sao Paulo, Brazil
The country risk indicator, as measured by the JP Morgan’s EMBI or
grades of rating agencies such as Standard & Poor’s (S&P’s) or Moody’s,
does not seem to truly reflect the fundamentals of an economy. Countries
that pursue sound economic policies are frequently placed on the same
level as countries with a populist orientation or with a recent history of
default or debt restructuring. Such circumstance generates a feeling of
unease with regard to these ratings. The objective of this article is to
investigate whether these indicators truly reflect market fundamentals or
whether some sort of prejudice, or intolerance towards certain countries,
can be identified. We use the Oaxaca–Blinder decomposition to analyse
the differences in country risk, measured as by EMBIþ, for a group of
emerging markets. This decomposition allows us to separate the ‘justified’
(differences in fundamentals) from the ‘unjustified’ differences (same
fundamental differently evaluated).
I. Introduction
With the recent integration of international capital
markets, country risk has become an important
factor within macroeconomics. The degree of free-
dom of monetary and fiscal policy is, at some level,
constrained by the country risk indicators given
that the economy’s external balance depends on this
variable. The most common ratings used are from
agencies like Standard & Poor’s (S&P’s), Fitch and
JP Morgan (EMBIþ). Ratings are used as a guide
to investors; a low rating is very costly to the
economy since long-term institutional investors are
not allowed to invest in a country not rated as
‘investment grade.’ In this article we use the
indicator EMBIþ by JP Morgan as the measure
of country risk, which reflects the average spread
governments papers pay over the US Treasury Bills.
This spread implies a direct cost to governments
and an indirect cost to firms given that, for most of
the countries, debt repayment depends on the
repayment policy of governments.From the theoretical point of view the determi-
nants of the country risk still are in their early
stages of development. However, there is a con-
sensus view about what determine this indicator: the
economy’s economic fundamentals. Nevertheless,
several countries with fragile fundamentals or with
a history of recent defaults have been rated as less
risky than countries with more solid economic
conditions. As a result, we intend to answer in this
article the following questions: Is country risk
determined by fundamentals or by sentiment? Are
the criteria used to determine country risk equally
applied for all countries or are there countries
treated with a certain level of intolerance? Which
countries suffer the most from discrimination when
their country risk is determined?
*Corresponding author. E-mail: [email protected]
Applied Economics ISSN 0003–6846 print/ISSN 1466–4283 online � 2010 Taylor & Francis 2577http://www.informaworld.com
DOI: 10.1080/00036840801964518
Dow
nloa
ded
by [
Uni
vers
ity o
f Il
linoi
s C
hica
go]
at 1
0:04
10
Nov
embe
r 20
14
Over the course of the last 20 years, the determi-
nants of emerging markets bonds spreads have been
analysed by a number of authors. The first to do so
was Sebastian Edwards in 1984. Edward analysed the
bond markets for 13 developing countries between
1976 and 1980 – a period when most Lesser
Developed Country (LDC) debt consisted of bank
loans. He found evidence that the debt to Gross
Domestic Product (GDP) ratio had a positive effect
on bond spreads. Recently, studies have benefit from
the much larger bond markets of these countries.Liquidity and solvency indicators play an impor-
tant role in bond spreads. For example, Min (1999)
found that ratios of external debt, international
reserves and debt service to GDP, along with
export and import growth rates were significant for
a group of Latin America and Asian countries
during the period of 1991 to 1995. Additional
macro variables such as inflation rate, terms of
trade, real exchange rate and net foreign asset
accumulation also helped to explain these spreads
in Min’s sample. On much the same lines as Min
(1999), Fiess (2003) separates contagion from
country-specific fundamentals for the joint determi-
nation of capital flows and country risk for four
Latin American countries in the 1990s. Fiess found
that capital flows were driven by both country risk
and global factors while country risk was solely
determined by a number of fundamental variables:
the ratio to GDP of primary balance and public
debt. Furthermore, Beck (2001), when estimating a
panel data, found that during the period immedi-
ately after the Asian crises the spreads were ‘almost
entirely explained by expectations of market funda-
mentals’. On the other hand, for developed
countries, international interest rates also played a
role but stock market volatility did not.Eichengreen and Mody’s (1998) analysis
attempted to separate the impact of changes in
fundamentals from changes in market sentiment
regarding LDC’s bond spreads. They controlled for
selective bias of new issuers and found that changes
in fundamentals explained the change in spreads.
However, the same explanatory variables had a
different impact in different periods of time. For
instance, market sentiment played a stronger role
during the Mexican crises (increasing spreads)
and during subsequent periods (declining spreads)
than during the periods just before the crises.
Moreover, the responses of spread to fundamentals
observed in Latin American countries were different
from the others in the sample.
In a similar way, Reinhart et al. (2003) analysedwhy it is more difficult for some countries to repaytheir debts at levels that would be consideredmoderate for other countries. They built a measureof debt intolerance based on an estimated riskregression, using ratings of the Institutional InvestorRatings as a function of fundamentals and the defaulthistory of the countries.
The present article is along the same lines asEichengreen and Mody’s (1998) and Reinhart et al.’s(2003). The estimated country risk equation is basedon very simple models as in Blanchard (2005) whoassumes that country risk is endogenously determinedby the country’s fiscal policy given that it reflects theprobability of default on the public debt. The debt,derived from the government flow budget constraint,is a function of: real rates of return of bondsdenominated in domestic and in foreign currency,primary surplus and real exchange rate.Correspondingly, the exchange rate is determinedby the relationship between the capital flow, theexchange rate and the trade surplus. That is, if capitalflow declines, the exchange rate must depreciate so asto generate the necessary trade surplus to compensatefor the capital account deficit. The solution to thesethree relationships is highly nonlinear, but a simpli-fied linear equation was estimated for Brazil.1
We also added another dimension to the problemby including the default history of countries inaddition to the fundamentals. We use the coefficientof this regression to compute how much of thedifference in spreads can be attributed to thedifference in country risk and how much tosentiment.
The article is structured as follows. The nextsection discusses some stylized facts regarding theeconomies being analysed. Section III lays out theestimates, and the last section presents the finalconsiderations. In particular, a specification with thenonlinearity of public debt–GDP ratio is included inthe tests of fiscal fundamentals and its significance isrejected.
II. Stylized Facts
JP Morgan’s EMBIþ measures the difference in bondspreads vis-a-vis US Treasury Bills of the samematurity, for countries that are ranked as Baal/BBBþ or below by Moody’s and S&P’s ratingagencies and some liquidity ranking rules. BetweenDecember 1998 and February 2006 a huge difference
1 Tests on a specification of nonlinearity of debt to output ratio in the fiscal fundamental was not significant.
2578 V. K. Teles and M. C. Leme
Dow
nloa
ded
by [
Uni
vers
ity o
f Il
linoi
s C
hica
go]
at 1
0:04
10
Nov
embe
r 20
14
in the spreads is exhibited by the countries in the
sample2: ranging from 3241 base points (Argentina)
to 111 (Malaysia) (Fig. 1).Looking at the relationship between the EMBIþ
indicator and a number of macro variables, we
observe that correlations are, for the most part, in
the predicted directions. In order to measure the
liquidity/solvency conditions of countries, we use
the ratio between exports and public debt which
measures the number of months of exports necessary
to repay the debt. The average EMBIþ and its
variance declines with this indicator (Fig. 2).In order to capture the fiscal fundamentals we
calculate the ratio of public debt to GDP. The higher
the ratio is the higher the EMBIþ result (Fig. 3).
However, we again observe that this variance also
increases with the debt ratio.For other variables capturing domestic fundamen-
tals there is a slight positive correlation between
domestic interest rates and the country risk, with
a declining variance (Fig. 4) and a negative correla-
tion between this variable and the GDP growth
rate (Fig. 5).The default history can be measured by the number
of times each country has defaulted. In this sample,
we generally find that countries that have never
defaulted (i.e. Korea and Malaysia) had an average
EMBI five times lower than countries which have
repudiated or restructured their debts at least once.
In addition to that, there is a positive relationship
between the number of defaults and the average
EMBI (Fig. 6) in spite of extreme cases such as
Venezuela and Ecuador (with seven registered
defaults each one) which have an average EMBI
lower than Argentina (with six registered defaults).However, as pointed out by Reinhart et al. (2003)
a better measure of the correlation of country risk
would be the length of time since the last default or
restructuring. Figure 7 shows the EMBI figures as of
December 2004 and the number of months since the
last default. Argentina, Ecuador, Russia and Ukraine
are countries with the most recent defaults.Still following Reinhart et al. (2003) approach, we
calculate debt intolerance as measured by the inverse
of the lapse (length) of time since the last default3
multiplied by the ratio of public debt to GDP. The
correlation of this index with the EMBI is highly
positive (Fig. 8). In regard to the bond market there is
a slight positive correlation between the average
spreads and average duration for countries in the
sample and that the variance increases with duration
(Fig. 9).Based on these facts, we can confirm the expecta-
tion that country risk is highly determined by
fundamentals. Nevertheless, some countries are
systematically over the trend line, while others are
frequently above it. Argentina’s EMBI, for instance,
is frequently over it, indicating that its country risk is
higher than the predicted by fundamentals. On the
other hand, Egypt has a country risk lower than the
predicted by its fundamentals given that it is
frequently below the trend lines.
0500
10001500200025003000
Malaysia
South Africa
PolandEgyp
tKorea
Mexico
Panama
Morocco
Ukraine
Bulgaria
Philippines
Peru
ColombiaTurke
yBrazil
Venezuela
Russia
Ecuador
Argentina
Fig. 1. Average EMBI by country
EQ
CO
AG
UK RSBR
EGPOPHPA TK
0
500
1000
1500
2000
2500
3000
0
Exports/Public debt
Embi
Fig. 2. Average EMBI versus average exports/public debt
EQ
CO
AG
UKRS BR
EGPH PA TKPO0
5001000
1500
20002500
3000
0 0 0 0 0 1
Public debt/GDP
Embi
Fig. 3. Average EMBI versus average public debt/product
2 The EMBIþ has been computed from 1993 on and many countries were excluded from the sample because they did not meetthe standards. The countries in Fig. 1, plus Qatar and Nigeria, make up the current sample.3We use the inverse of time since the last default so as not to exclude countries that never defaulted from the analysis.
Fundamentals or market sentiment 2579
Dow
nloa
ded
by [
Uni
vers
ity o
f Il
linoi
s C
hica
go]
at 1
0:04
10
Nov
embe
r 20
14
III. Estimation and Results
Aiming to measure the impact of fundamentals oncountry risk, we first estimate a panel of 19 countriesusing monthly data between December 1998 andFebruary 2006. Second, we estimate the same modelfor each one of the countries and then, using themethodology proposed by Oaxaca (1973) andBlinder (1973), we decompose the differences incountry risk into two components: one partexplained by fundamentals and another explainedby market sentiment. In Eichengreen and Mody’s(1998) study this methodology was used with theintention to capture the effects of this kind ofsentiment on different spreads over different periods
of time. In the present article we adopt the same
strategy, but with a different purpose, we use it to
compare country risks between economies. The
decomposition is based on a comparison of each
country to the pool of countries in the sample. This
benchmark is particularly opportune to our purpose
given that we are able to capture the differences in
criterion when comparing similar economies in
terms of fragility of their institutions and political
regimes.The decomposition is very simple: let X be the
observed explanatory variables of country risk Y;
therefore for country i we can estimate:
Yi ¼ Xi�i þ ui
EQ
CO
AG
UK BR
EGPHPA TK
PO0500
10001500200025003000
4 8 12 16 20 24
Domestic interest rates
Embi
RS
Fig. 4. Average EMBI versus average domestic interest
rates
CO
AG
UKRSBR
EGPHPA TK
PO0500
10001500200025003000
3 4 5 6 7 8
GDP growth
Embi
Fig. 5. Average EMBI versus average product growth
0
500
10001500
2000
2500
3000
0 1 2 3 4 5 6 7
Number of defaults
Embi
Fig. 6. Average EMBI by number of defaults
EQ
CO
AG
UKRS BR
EGPH PATK
PO0500
10001500200025003000
5 15 25 35 45 55 65 75
Years
Embi
Fig. 7. Average EMBI versus years since last default
EQ
CO
AG
UKRS BR
EGPHPATK
PO0500
10001500200025003000
2 12 22 32 42 52 62
DI* (PD/GDP)
Embi
Fig. 8. Average EMBI versus default index
Note: *Public debt/GDP.
EQ
CO
AG
UKRSBR
EGPHPATK
PO 0500
10001500200025003000
0 1 2 3 4 5 6 7
Maturity
Embi
Fig. 9. Average EMBI versus average maturity
2580 V. K. Teles and M. C. Leme
Dow
nloa
ded
by [
Uni
vers
ity o
f Il
linoi
s C
hica
go]
at 1
0:04
10
Nov
embe
r 20
14
For the pool of countries in the sample we have thesame estimation
Y ¼ X�þ u
Therefore, the difference in averages between eachcountry and the pool of countries is:
�Y� �Yi ¼ �X�� �Xi�i ¼ ð �X� �XiÞ�þ �Xið�� �iÞ
The first term in the right-hand side would be thecomponent explained by fundamentals while thesecond would be explained by market sentiment.Certainly the two effects can be quite differentdepending on which population is chosen as thebase group. Also, as pointed out by Fournier (2005),one can get valuable information by contrasting bothresults. In our analysis we are comparing eachcountry with the pool of countries in the samplethis group is, by construction, our nondiscriminatorybase. This strategy is similar to Neumark’s (1988)who, looking at wage discrimination, estimated thecoefficient of the nondiscriminatory wage structurefor the whole sample.4
We measured country risk by the EMBIþ index.As discussed by Blanchard (2005) the EMBI spreadreflects the probability of default as well as the riskaversion of foreign investor. This last componentprobably varies over time but is not country specific;therefore the difference between the indexes of thecountries is free from it.
The country risk equation was estimated as afunction of: (i) the fiscal fundamentals captured bythe ratio of public debt to GDP (PD/GDP); (ii) thegeneral fundamentals of the economy: the GDPgrowth over the last 12 months (GROWTH), inflationin the last 12 months (INF) and the annual nominalinterest rates (IR) as well as the ratio of internationalreserves to GDP (RES/GDP); (iii) the liquidity andsolvency conditions as captured by the ratio betweenexports and public debt (X/GDP) and (iv) the debtintolerance to defaulters which is obtained by multi-plying public debt to GDP by the index of the inverseof time since the last default (DI). We also controlledfor the debt profile including the average duration ofthe bonds (MAT). The reference currency is theUS dollar and inflation rates were represented bythe Consumer Price Index (CPI). All variables wereobtained from the International Financial Statistics(IFS) and Government Financial Statistics (GFS)of the International Monetary Fund or from theMoody’s database.
Some of the data were only available on an annualbasis. A common approach in the literature is the use
of industrial production to transform the GDP into a
monthly measure (e.g. Mitra, 2007; Teles and
Andrade, 2008). The reason behind this approach is
based on the following factors: (i) within short-term
movements in GDP, output indicators are usually
regarded as being better guides than expenditure
indicators (Mitchell et al., 2005) – industrial produc-
tion is the only monthly output indicator for all
countries in samples with a lot of countries (like
ours); (ii) some studies have found that in several
countries indicators based on industrial activity are
useful when forecasting real GDP growth rates in the
short run (Mourougane and Roma, 2003). We
reconstructed the monthly series for the debt series
using the public deficits/surpluses available in the
IMF/GFS statistics.It can be expected that country risk would increase
with the ratio of public debt to GDP; decrease with
GDP growth and increase with interest rates since
both variables indicate the long-term conditions of
public debt path; and increase with inflation and
decrease with the ratio of international reserves to
GDP. We also expect spreads to increase with the
ratio of exports to public debt and with the inverse of
the amount of time since the last default by PD/GDP
and decline with duration. This variable is itself a
measure of country risk: the riskier the country the
shorter the duration of its bonds and, if a country
wants to issue a bond with a longer duration it will
probably have to pay a higher premium.Before proceeding with the model estimations we
tested the stationarity of the variables used. The null
hypothesis of nonstationarity was tested using the
panel unit root tests of Im et al. (2003), Maddala and
Wu (1999) and Choi (2001). These tests allow each
member of the cross section to have a different
autoregressive root and different autocorrelation
structures under the alternative hypothesis.
Tables A1 and A2 in the appendix show the
hypothesis that all variables considered in the
estimations do not contain a unit root at 5% for at
least one of these tests.We estimated two alternative specifications:
EMBIi¼ �iþ�iPDi
GDPiþ �iMATiþ�i
DI�PDi
GDPi
� �þ’iX
PDii
ð1Þ
The equations were estimated for the entire sample
of countries as well as for each country separately,
using Generalized Least Square (GLS) and
Instrumental Variables (IV) methods (to control for
4 There are other methods to estimate the nondiscriminatory structure such as in Oaxaca and Ramson (1994) and Shresthaand Sakellariou (1996).
Fundamentals or market sentiment 2581
Dow
nloa
ded
by [
Uni
vers
ity o
f Il
linoi
s C
hica
go]
at 1
0:04
10
Nov
embe
r 20
14
the endogeneity of maturity and interest rates) and
the lagged values of the variables as instruments.For the panel of countries we estimated a fixed
effect model, since the Hausman and redundant
fixed effect tests argue in favour of this model
(Table A3 in the appendix).5 This is the appropriated
model since we want to control for unobserved effects
that capture features of each country possibly
correlated with the explanatory variables and that
can biases their coefficients. The fixed effect model
controls for possible idiosyncratic characteristics of
countries that may affect investors’ risk perceptions.
We assume that these characteristics are time
invariant, a reasonable assumption given the short
interval considered within the analysis. For example,
South Korea may have a higher than expected risk (as
measured by economic fundamentals) based on the
hostility with North Korea, a very unstable regime.As we can see in the first model, the estimated
coefficients exhibited the expected sign and are
significant, except for the ratio of exports to public
debt in both estimations.Omitting the variable X/PD improves the overall
explanatory power of the model without changing
the coefficients of the remaining variables (Model 2).
To test for nonlinearities in the relationship between
country risk and the fiscal variable, the model
was estimated once again including the square of
the public debt to GDP ratio which showed no
significance (Model 3). The results using GLS and IV
are quite similar.The second specification estimated using the IV
method is as follows:
EMBIi ¼ �i þ�iPDi
GDPiþ �iMATi þ �i
DI�PDi
GDPi
� �þ �Z
ð2Þ
where Z is a vector of variables that captures theeconomy’s general fundamentals, as explained above.The ratio of exports to public debt ratio was droppedin the estimation and the results are in Table 2.
The impact of the general fundamentals on countryrisk was first investigated including one variable at atime. We could see that GDP growth is notsignificant, inflation is but collinear with maturityand both, interest rates and the ratio of internationalreserves to GDP are also significant. We can assumethat a model which incorporates all variables, exceptfor the inflation and GDP growth, is the mostparsimonious and was chosen to be used in thedecomposition for the pool of countries – this will beour benchmark.
The first specification and the most parsimoniousversion of the second specification were estimated foreach country separately. To verify the impact of thedifference in coefficients (or discrimination, if youprefer) we used the Oaxaca–Blinder decomposition(Table 3).
The second column of Table 3 shows the actualaverage EMBI for each county. The third and fourthcolumns are the predicted EMBI using coefficientsfrom the panel data estimations in the two specifica-tions with the variables of each country, respectively.In other words, the EMBI that would prevail ifcountries’ fundamentals were evaluated by themarket using the same parameters. The two specifica-tions produce similar results. The last two columnsare the differences between the EMBI’s predicted andthe actual values.
In absolute terms and taking into account onlyfiscal fundamentals, Argentina was the countrymost discriminated against among the countriesof the sample (see fifth column of Table 3). Theaverage EMBI for Argentina was approximately 1500base points higher than the expected considering
Table 1. Effects of debt fundamentals on country risk
1 2 3
Variables GLS IV GLS IV GLS IV
C 133.39 (125.59) 204.52 (128.27) 128.75 (125.34) 200.56 (128.04) �20.91 (188.27) �5.42 (197.24)PD/GDP 311.34 (36.55) 308.62 (36.80) 313.93 (36.31) 311.03 (36.55) 407.54 (95.08) 437.67 (99.04)MAT 249.43 (20.96) 243.90 (21.06) 249.33 (20.95) 243.80 (21.06) �64.94 (14.67) �78.79 (15.79)(PD/GDP) *DI �66.52 (14.56) �78.81 (15.19) �67.07 (14.53) �79.40 (15.16) 249.43 (20.95) 237.76 (21.34)X/PD �0.67 (1.08) �0.62 (1.08)(PD/GDP)2 �12.85 (12.06) �13.73 (12.42)
R2 0.56 0.20 0.56 0.56 0.56 0.56Adj. R2 0.55 0.20 0.55 0.56 0.55 0.56
Note: SDs in parentheses.
5All significant fiscal fundamentals were used to perform the tests (Model 2 in Table 1).
2582 V. K. Teles and M. C. Leme
Dow
nloa
ded
by [
Uni
vers
ity o
f Il
linoi
s C
hica
go]
at 1
0:04
10
Nov
embe
r 20
14
its fundamentals. At the other extreme is Panama,which received what could be referred to as a ‘VIP’treatment. If only fundamentals were taken intoaccount Panama’s average EMBI would be morethan 1000 base points higher than actually is. Whengeneral fundamental are included major changes takeplace. Argentina, for instance, still receives the mostunfavourable treatment. On the other side, Venezuelabecomes the next in the rank just above Russiaand Ecuador. Malaysia is the most remarkable case.This country has the highest ratio of internationalreserves to GDP (about three times above theaverage ratio of the pool) but is valued at a muchlower rate: its coefficient is about 0.60 of the averagecoefficient. Therefore, the inclusion of this variable
changes completely the predicted EMBI which shouldin fact be negative when compared to the pool ofcountries. Peru and Korea are in the break evenpoint.
It is worth noting that, in some cases, the inclusionof variables like reserves–GDP ratio and interestrates brings the actual EMBI closer to the predicted.As a result, these general fundamentals are useful toexplain country risk only in these cases. For theothers they do not explain intolerance. In somecountries the inclusion of these fundamentalsincreases the difference between actual and predictedEMBIs, indicating an intolerance with respect tothese variables.
Table 4 shows the differences in the country-riskratings as a consequence of market sentiment.The first column is the actual ranking risk. Thesecond and third columns are the predicted rankingsas indicated by the fundamentals. The last twocolumns are the differences between the actual andthe predicted results. A positive value in the last twocolumns means that market sentiment is significantenough to change the country’s ranking position.
Although, in absolute terms, Argentina is the mostdiscriminated country, it remains in the first spot ofthe ranking even when market sentiment is elimi-nated. Argentina is indeed the highest risk country ofthe sample. On the other side, Russia had the thirdhighest average EMBI during the period underanalysis and it was placed in the second spot whenonly fiscal fundamental are taken into account.However, when more general fundamentals areincluded, Russia occupies again the third place.Malaysia is the country with the major change inraking position for the reasons discussed above.
Table 5 shows the relative market sentiment for thetwo specifications. Some changes can be noted regard-ing the previous case of absolute market sentiment.
Table 2. The effects of general fundamentals on country risk
Models
Variables 1 2 3 4 5
C 379.91 (132.59) �600.45 (125.74) �316.97 (167.60) 746.00 (127.57) 889.32 (67.51)PD/GDP 296.66 (38.38) 384.38 (33.74) 400.06 (41.46) 338.29 (34.67) 34.17 (18.59)MAT �104.97 (15.60) 1.38 (14.79) �63.74 (16.80) �23.82 (14.98) �35.43 (7.33)(PD*GDP) *DI 200.71 (21.89) 52.20 (20.57) 187.29 (24.58) 207.69 (20.03) 311.64 (25.37)GDP GROWTH 17.21 (24.51)INFLATION 28.00 (1.43)INTEREST RATES 23.00 (4.94) 13.65 (1.32)RESERVES/GDP �1156.44 (87.97) �436.18 (41.64)
R2 0.59 0.68 0.53 0.61 0.64Adj. R2 0.59 0.68 0.52 0.61 0.64
Note: SDs in parentheses.
Table 3. Actual EMBI and EMBI predicted by market
fundamentals
Actual Predicted 1 Predicted 2
(A) (B) (C) B�A C�A
Argentina 3023 1499 1475 �1524 �1549
Russia 1084 381 816 �703 �268
Ecuador 1484 1152 1246 �332 �238
Ukraine 548 265 874 �282 327
Venezuela 979 739 506 �240 �473
Philippines 468 281 632 �186 165
Colombia 536 367 710 �169 174
Mexico 350 188 655 �162 305
Peru 495 553 487 57 �8
Korea 184 255 185 70 1
Poland 171 319 533 148 362
Bulgaria 474 698 528 224 54
South Africa 148 427 825 279 676
Brazil 819 1105 1050 286 231
Malaysia 113 507 �129 394 �242
Egypt 239 704 694 465 454
Turkey 569 1071 1275 501 705
Morocco 398 1192 691 794 293
Panama 387 1536 748 1149 362
Fundamentals or market sentiment 2583
Dow
nloa
ded
by [
Uni
vers
ity o
f Il
linoi
s C
hica
go]
at 1
0:04
10
Nov
embe
r 20
14
Now, in terms of fiscal fundamental, Russia is themost
discriminated against: its EMBI should be roughly
65% lower than actually is. Meanwhile Panama and
Malaysia remain as themost privileged countries in our
sample. When other macro variables are included, this
scenario changes a bit:Malaysiamoves from the last to
the first position in the ranking.
IV. Conclusions
An analysis of country risk for a pool of emergingmarkets shows that its determinants reflect bothfundamentals and sentiment features. If the onlyfactors taken into consideration were economicfundamentals and past behaviour in relation to debtrepayment then a number of countries would exhibit adifferent level of risk than they actually do. Based onour analysis, the only country in the sample receiving a‘fair evaluation’ is South Korea, but, this is so, onlywhen macro variables are included. In absolute terms,Argentina is the most discriminated country in thesample and even after correcting for such tendency it isstill the riskiest country. Ecuador, Venezuela andRussia change positions but are always placed in thetop five of the ranking of risky countries. Turkey andBrazil considerably drop positions and are negativelyaffected and impacted by such ratings. After correct-ing for market sentiment, Brazil drops from the fifthto the 14th place of the riskiest countries in terms offiscal fundamentals or to 11th place in more generalterms. On the other hand, Turkey drops from the sixthto the 17th and 19th place, respectively. Malaysia is aspecial case given that in terms of fiscal fundamentalsit has a better position than it really should; but whenthe ratio of international reserves to GDP is included,Malaysia is severely discriminated and occupies theworst relative position.
References
Blanchard, O. (2005) Fiscal dominance and inflationtargeting: lessons from Brazil, in Inflation Targeting,Debt, and the Brazilian Experience, 1999 to 2003 (Eds)F. Giavazzi, I. Goldfajn and S. Herrera, MIT Press,Cambridge, MA, pp. 49–84.
Beck, R. (2001) Do country fundamentals explain emergingmarket bond spreads?, CFS Working Paper No.2001/02, Centre for Financial Studies, Frankfurt.
Choi, I. (2001) Unit root tests for panel data, Journal ofInternational Money and Finance, 20, 249–72.
Edwards, S. (1984) LDCs’ foreign borrowing and defaultrisk: an empirical investigation, 1976–1980, AmericanEconomic Review, 74, 726–34.
Eichengreen, B. and Mody, A. (1998) What explainschanging spreads on emerging-market debt:fundamentals or market sentiment?, NBER WorkingPaper Series W6408.
Fiess, N. (2003) Capital flows, country risk and contagionpolicy, #Research Working Paper No. 2943, TheWorld Bank.
Fournier, M. (2005) Exploiting information from pathdependency in Oaxaca–Blinder decompositionprocedures, Applied Economics Letters, 12, 669–72.
Im, K. S., Pesaran, M. H. and Shin, Y. (2003) Testing forunit roots in heterogeneous panels, Journal ofEconometrics, 115, 53–74.
Table 5. Relative market sentiment
Country Rel. 1 Rel. 2
Russia �0.65 �0.25Ukraine �0.52 0.60Argentina �0.50 �0.51Mexico �0.46 0.87Philippines �0.40 0.35Colombia �0.32 0.33Venezuela �0.25 �0.48Ecuador �0.22 �0.16Peru 0.12 �0.02Brazil 0.35 0.28Korea 0.38 0.00Bulgaria 0.47 0.11Poland 0.86 2.12Turkey 0.88 1.24South Africa 1.88 4.56Egypt 1.94 1.90Morocco 2.00 0.74Panama 2.97 0.94Malaysia 3.49 �2.14
Table 4. Rank
Actualrank
RankEMBI 1
RankEMBI 2
Country A B C B–A C–A
Argentina 1 1 1 0 0Ecuador 2 3 5 1 3Russia 3 2 3 �1 0Venezuela 4 5 2 1 �2Brazil 5 14 11 9 6Turkey 6 17 19 11 13Ukraine 7 4 14 �3 7Colombia 8 7 10 �1 2Peru 9 9 6 0 �3Bulgaria 10 12 8 2 �2Philippines 11 6 9 �5 �2Morocco 12 18 12 6 0Panama 13 19 15 6 2Mexico 14 8 13 �6 �1Egypt 15 16 17 1 2Korea 16 10 7 �6 �9Poland 17 11 16 �6 �1South Africa 18 13 18 �5 0Malaysia 19 15 4 �4 �15
2584 V. K. Teles and M. C. Leme
Dow
nloa
ded
by [
Uni
vers
ity o
f Il
linoi
s C
hica
go]
at 1
0:04
10
Nov
embe
r 20
14
Maddala, G. S. and Wu, S. (1999) A comparative studyof unit root tests with panel data and a new simple test,Oxford Bulletin of Economics and Statistics, 61, 631–52.
Min, H. G. (1999) Determinants of emerging market bondspread: do economic fundamentals matter?, WorkingPaper Series, The World Bank.
Mitchell, J., Smith, R. J., Weale, M. R., Wright, S. andSalazar, E. J. (2005) An indicator of monthly GDP andan early estimate of quarterly GDP growth, TheEconomic Journal, 115, 108–29.
Mitra, S. (2007) Is the quantity of government debt aconstraint for monetary policy?, IMF Working PaperNo. WP/07/62.
Mourougane, A. and Roma, M. (2003) Can confidenceindicators be useful to predict short term real GDPgrowth?, Applied Economics Letters, 10, 519–22.
Neumark, D. (1988) Employers discriminatory behaviorand estimation of wage discrimination, Journal ofHuman Resources, 23, 279–95.
Oaxaca, R. (1973) Male–female wage differentials in urbanlabor markets, International Economic Review, 14,693–709.
Oaxaca, R. and Ramson, M. R. (1994) On discriminationand decomposition of wage differentials, Journal ofEconometrics, 61, 5–21.
Reinhart, C., Rogoff, K. and Savastano, M. (2003) Debtintolerance, NBER Working Paper No. W9908.
Shrestha, K. and Sakellariou, C. (1996) Wagediscrimination: a statistical test, Applied EconomicsLetters, 3, 649–51.
Teles, V. K. and Andrade, J. (2008) Monetary policy andcountry risk, Applied Economics, 40, 2021–8.
Appendix
Table A1. Unit root tests
Levin, Lin and Chu t* Im, Pesaran and Shin W-stat** ADF-Fisher �2** PP-Fisher �2**
Variables Statistic Prob. Statistic Prob. Statistic Prob. Statistic Prob.
EMBI �2.37 0.01 �1.93 0.03 50.61 0.05 104.22 0.00Public debt/GDP �3.67 0.00 �2.34 0.01 63.94 0.00 67.27 0.00Maturity 2.41 0.99 0.87 0.81 46.76 0.11 56.70 0.02Default index*(PD/GDP) 0.67 0.75 �2.19 0.01 61.17 0.00 61.81 0.00Public debt/exports �4.44 0.00 �6.61 0.00 185.54 0.00 242.99 0.00GDP growth 0.02 0.51 �3.94 0.00 79.46 0.00 103.17 0.00Inflation �3.47 0.00 �4.60 0.00 89.27 0.00 89.76 0.00
Notes: *Assumes common unit root process. **Assumes individual unit root process.
Table A2. Unit root tests with trend
Levin,Lin and Chu t*
Breitungt-stat*
Im, Pesaran andShin W-stat** ADF-Fisher �2** PP-Fisher �2**
Variables Statistic Prob.** Statistic Prob.** Statistic Prob.** Statistic Prob.** Statistic Prob.**
Public debt/GDP 2.37 0.99 �1.32 0.09 �4.83 0.00 98.66 0.00 82.87 0.00Maturity 0.38 0.65 1.01 0.84 �3.57 0.00 76.31 0.00 105.55 0.00
Notes: *Assumes common unit root process. **Assumes individual unit root process.
Table A3. Fixed versus random effects tests
Test Statistic p-value
Hausman test 44.4076 0.0000Redundant fixed effectsCross-section F 35.5863 0.0000Cross-section chi-square 536.7126 0.0000
Fundamentals or market sentiment 2585
Dow
nloa
ded
by [
Uni
vers
ity o
f Il
linoi
s C
hica
go]
at 1
0:04
10
Nov
embe
r 20
14