IMF-related news and emerging financial markets
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
Journal of International Money and Finance 24 (2005) 1126e1142
www.elsevier.com/locate/econbase
IMF-related news and emerging financial markets
Bernd Hayo a,b,*, Ali M. Kutan b,c,d,1
a
Faculty of Business Administration and Economics (FB 02), Philipps-University Marburg,
Universitätsstr. 24, 35037 Marburg, Germany
b
ZEI, University of Bonn, Germany
c
Department of Economics and Finance, Southern Illinois University, Edwardsville, IL 62026-1102, USA
d
Emerging Markets Group, London, UK
Abstract
We examine the reaction of financial market returns and volatility in a diverse group of six emerging mar-
kets to a set of IMF events during the Asian, Russian and Brazilian crises of 1997e1999. Focusing on stock
markets first, we find that on average, negative (positive) IMF news reduces (increases) daily stock returns by
about one percentage point. The most influential event is the delay of loans from the IMF. For foreign ex-
change market returns, we only observe significant effects of bad IMF news, and on bond markets neither
good nor bad news seems to affect interest rate spreads. IMF news does not have a significant impact on
the volatility of the financial markets. Further, both gains and losses resulting from IMF news tend to be neu-
tralized within one day after the announcement. Finally, we find that IMF news does not cause creditor panic.
Ó 2005 Elsevier Ltd. All rights reserved.
JEL classification: F300; G100
Keywords: IMF news; Financial market returns; Emerging markets; Bond spread; Creditor panic; Market efficiency
1. Introduction
The role of the International Monetary Fund (IMF) in the international monetary system has
been extensively studied. While most of the discussion centers on issues such as effectiveness
* Corresponding author. Faculty of Business Administration and Economics (FB 02), Philipps-University Marburg,
Universitätsstr. 24, 35037 Marburg, Germany. Tel.: C49 6421 2823091; fax: C49 6421 2823088.
E-mail addresses: hayo@wiwi.uni-marburg.de (B. Hayo), akutan@siue.edu (A.M. Kutan).
1
Tel.: C618 650 3473; fax: C618 650 3047.
0261-5606/$ - see front matter Ó 2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jimonfin.2005.08.007B. Hayo, A.M. Kutan / Journal of International Money and Finance 24 (2005) 1126e1142 1127
of IMF programs and IMF-induced moral hazard, little is known about the influence of the IMF
as a source of news on financial markets. In this paper, we analyze the impact of various cat-
egories of IMF actions on emerging financial markets by studying the reaction of investors to
announcements of IMF actions during the Asian, Russian and Brazilian crises of 1997e1999.
IMF-related news has two possible roles to play in emerging financial markets. First, this
news may convey some information to market participants, for instance, about the economic
situation of a country, which was not known before. Second, it may give some indication to
market participants of how the IMF is going to act in response to a country’s crisis. Unantic-
ipated good IMF news should increase returns on the day of announcement, while bad news
should result in lower returns. Whether returns continue to rise or fall subsequently depends
on how market participants perceive such news and whether they consider the IMF stabilization
program to be effective (Evrensel, 2002).
IMF policies during the Asian financial crisis have been severely criticized by some observ-
ers, and the debate over the effectiveness of IMF policies has intensified since the crisis (see, for
instance, Katz (1998) and Naim (2000)). Critics have argued that the Asian crisis, resulted
from, or was intensified by, a ‘‘creditor panic’’ resulting from the IMF’s statements that Asia
needed drastic financial reforms. Declining investor confidence in the region was therefore
an important factor in initiating and sustaining the crisis. For example, Sachs (1999) has argued,
‘‘.provocative IMF actions have probably contributed to the [creditor] panics’’ (p. 389).
A significant and sustained decline in stock market returns would be an evidence that unan-
ticipated negative IMF actions, such as unfavorable statements regarding a country’s economic
situation or a delay of IMF loan announcements, have undermined investor confidence in mar-
kets if the latter follows the former. In the extreme, this may cause investors to rapidly and mas-
sively sell off their assets, creating a self-fulfilling panic. It is well known that volatility
linkages among bond, money, and stock markets are strong (see Fleming et al. (1998) and
the references cited therein), and therefore IMF actions might affect many different local finan-
cial markets. The empirical approach initially utilized to study the effects of news on stock mar-
kets is therefore also applied to foreign exchange and bond markets in order to separate general
from idiosyncratic effects of IMF actions.2
This paper focuses on the response of stock, bond, and foreign exchange markets to a com-
prehensive set of IMF-related news to investigate whether market participants are able to ex-
tract information from IMF-related news and on the question of whether such news has any
effect on market returns. The paper also investigates whether IMF-related news is associated
with increasing volatility in financial markets because such volatility increases are a sign of
panic among market participants. Our findings refute the panic hypothesis, suggesting that fi-
nancial markets showed no excessive reaction to IMF announcements during the turbulent pe-
riod of 1997e1999.
2
Another heavily debated issue is that of the IMF causing moral hazard in financial markets (Lane and Phillips,
2000). According to this view, investors knowingly take very risky positions because they believe that the IMF will
bail them out (creditor moral hazard) or governments do not adjust inappropriate economic policies unless bailed
out by the IMF (debtor moral hazard). A thorough empirical test of the portfolio moral hazard hypothesis, among others,
is provided by Sarno and Taylor (1999). They show that the asset price bubbles observed before the crisis may be due to
moral hazard problems in financial intermediation. The empirical approach in this paper is not really suited to contribute
to this debate, as moral hazard may not be connected with specific IMF news announcements. Moral hazard may also
take place long before IMF announcements due to the expectation of IMF support in a country. In addition, there are
many forms of moral hazard, and it is not easy to detect what type of moral hazard is primarily related to IMF news.1128 B. Hayo, A.M. Kutan / Journal of International Money and Finance 24 (2005) 1126e1142
The remainder of the paper is organized as follows. In the next section, we summarize pre-
vious work and discuss our contribution. Section 3 describes the construction of IMF events,
and Section 4 reports our empirical results. The time pattern of returns surrounding IMF
news is analyzed in Section 5. To assess the robustness of our findings for stock markets, Sec-
tion 7 considers evidence from foreign exchange and bond markets. The last section summa-
rizes the paper and discusses its policy implications.
2. Previous literature and our contribution
There is a growing literature on the effects of IMF news on financial markets. Kho and Stulz
(1999) use an event study to examine the impact of IMF assistance on the value of bank stocks
during the Asian financial crisis. They conclude that the IMF program announcements in-
creased bank’s shareholder wealth. In a related study, Dong et al. (2000) examine the impact
of the announcement dates of IMF support programs on the abnormal returns on U.S. banks’
stocks during crises, and they report similar results in that these bank stocks tend to earn ab-
normal returns. Overall, both studies find that IMF news has a significant influence on returns
of stocks of banks that have credit exposure in countries seeking IMF assistance.
Brealey and Kaplanis (2000) look at a broad sample of IMF programs, beyond those imple-
mented during the Asian crisis, and they cover a wider range of financial assets than those in-
cluded in the two above-mentioned studies. They find a substantial decline in asset prices in the
weeks leading up to the announcement of the IMF programs but no evidence that the announce-
ment of the IMF support causes any part of these wealth losses to be reversed. Ganapolsky and
Schmukler (1998) examine the impact of the IMF agreement announcement during the Tequila
crisis in Argentina. They find that the announcement had a positive impact on stock and bond
returns. They also show that the announcement of the agreement played an instrumental role in
reversing the dynamics of the crisis.
All these studies focus on a particular IMF event or events. For example, Kho, Lee and Stulz
(1999) investigate IMF bailout news, while Kho and Stulz (1999) concentrate on IMF program
announcements. Kaminsky and Schmukler (1999) and Ganapolsky and Schmukler (1998) in-
vestigate, among other news, debtor country agreements with international organizations, in-
cluding IMF agreements. Besides IMF assistance, Brealey and Kaplanis (2000) also cover
the progress of negotiations with the IMF by separating news into ‘‘good’’ and ‘‘bad’’ news.3
Our analysis differs from those described above in several respects. First, because these stud-
ies focus on a particular event or selected events, they may not capture the overall impact of
IMF actions on investor behavior. In this paper, we collect, categorize, and use a wide spectrum
of available information about IMF-related actions and events that occurred during our sample
period (for details, see next section). Second, we provide evidence on the effects of IMF news
from three financial markets.
A third common thread in the earlier studies is that they all focus on returns, and they do not
investigate how IMF events affect the conditional volatility of markets. But if stock market vol-
atility is a measure of stock market risk or uncertainty, then a multi-country investigation of
how IMF actions affect volatility can broaden our general understanding of the determinants
of such risk and allows us to price risk more efficiently. Moreover, policy makers may act to
3
Brealey and Kaplanis (2000) also study returns on currencies and emerging bond indexes.B. Hayo, A.M. Kutan / Journal of International Money and Finance 24 (2005) 1126e1142 1129
reduce the possibility that IMF-related news will adversely affect stock market volatility. Thus,
a better knowledge of the determinants of the conditional second moments of asset returns is
crucial for improving our understanding of asset pricing and efficient asset allocation decisions.
Consequently, we provide evidence about the effect of IMF-related news on market volatility.
Finally, most of the previous papers use an event study whereas we use a panel regression
approach, following Kaminsky and Schmukler (1999), who utilize both approaches. Because
we cover a much broader range of IMF news, a regression approach is more effective and
also avoids some of the potential problems of an event study. We also attempt to overcome
some of the methodological shortcomings of the Kaminsky and Schmukler study, which uses
an OLS equation with pooled data. Such an estimation procedure ignores the typical finding
of time-varying volatility observed in emerging stock market returns.4 Moreover, Kaminsky
and Schmukler provide no information about the statistical validity of their model, and they
do not include a control variable to capture the general evolution of stock markets.
We employ a GARCH model with country-specific fixed effects to capture the autoregres-
sive conditional heteroscedasticity inherent in financial series and to take the panel nature of
the database into account. Moreover, we include returns based on the U.S. Standard & Poor’s
stock market index as a control variable. Finally, we provide a number of statistical tests to en-
sure the adequacy of our models.
3. Construction of IMF news
News about the interactions between the IMF and the countries searching for help in a re-
gional or global crisis or for reasons specific to each country was collected from the IMF web-
site, the Washington Post, and BBC News. The news was then categorized according to
different types of IMF actions. Because consistency is crucial for categorization, a substantial
effort was made to put similar news under the same heading. The categories are listed and de-
scribed in Table 1.
After completing the categorization process, the dates for which news was published were
matched with the dates for the daily stock return data of the corresponding country. A dummy
variable was created for each different category such that the variable was equal to one on the
day the particular news was announced, and zero otherwise. We only consider the impact effect
of news and not the continuation of an event. For instance, a visit by an IMF delegation will
typically last several days or weeks, but we only note the announcement of the visit. We
take time-zone differences into account, so that lags in the model correspond to time lags ex-
perienced by economic agents.
In addition to categorizing the news in a neutral way, it is also useful to group IMF events
into ‘‘good’’ and ‘‘bad’’ from the point of view of the country in crisis. The guiding idea is that
IMF news conveys information that is similar to either a positive or negative shock to the value
of stocks. There are two basic competing hypotheses. First, it could be the case that stockhold-
ers are neutral or even negative about an IMF intervention because the loan part of IMF pro-
grams has to be paid back later.5 The money to repay the loans has to be obtained by the
government either through higher taxes, by printing money or by selling bonds. The first
4
Bekaert and Harvey (1997) and Aggarwal et al. (1999) provide evidence that emerging market stock returns also
exhibit time-varying volatility, similar to more mature, developed markets.
5
So far every IMF loan has been paid back by the countries involved (see Aylward and Thorne, 1998).1130 B. Hayo, A.M. Kutan / Journal of International Money and Finance 24 (2005) 1126e1142
Table 1
Defining IMF news categories
Category Definition No. of
cases
Agreements Signing of agreements to implement a new economic program. 9
Delays Delay of loans or talks between the countries and the IMF as 6
a result of disagreements between the parties on issues such
as the reforms that should be made or the conditions that must
be fulfilled in order to achieve an agreement.
IMF-supportive Positive statements made by the country’s government officials 23
announcements regarding IMF policies or announcements of economic programs
that favor IMF policies or are necessary for eligibility for new loans.
Loan approval Credits and loans extended by the IMF through stand-by arrangements, 35
extended fund facility, reserve tranche policies, credit tranche policies
and policy on emergency assistance.
Request of funds Request for loans, including ‘‘bail-out’’ loans. Such requests 4
from the IMF usually come from the president of the country, usually in a
situation of financial or economic crisis, in search of support
funds to make reforms in the economy.
Talks/negotiations Negotiations that take place between a country and the IMF on a loan 14
approval or implementation of an economic program.
Anti-IMF Government-announced policies and statements that do not comply with or 7
policies/statements support IMF policies.
Resumptions The resumption of negotiations between a country and the IMF following 5
a period of delay of loans or talks.
Visits Each year a team of four or five IMF staff members travels to the capitals 12
of the countries that it is helping and spends time gathering information
and holding discussions with government officials about the country’s
economic policies. Statements, agreements or loan approvals usually
accompany these visits.
Favorable statements Statements made by IMF officials praising the reform measures taken by 28
the countries, the developments in the economic or financial situation of
the country.
Unfavorable Statements made by IMF officials usually about countries that did not fulfill 10
statements or that did not put enough effort into satisfying the restrictive conditions
required by an agreement or a loan approval.
Bad news Combines the above categories ‘‘delays’’ and ‘‘unfavorable statements’’, 17
and partially ‘‘talks’’.
Good news Combines the above categories ‘‘agreements’’, ‘‘loan approval’’, 90
‘‘resumptions’’, ‘‘visits’’, ‘‘favorable statements’’, and partially ‘‘talks’’.
alternative may reduce net dividends in the country, the second alternative may generate infla-
tion, and the third alternative may lead to a public sector debt problem. Because the second and
third alternatives are usually ruled out by the non-loan parts of IMF stabilization programs, it is
typically through taxation that governments service the loans. But it is not clear a priori that
IMF loans will have a positive impact on stock returns if investors know that they have to
pay back the money borrowed plus interest later.
Second, one can argue that a supportive IMF intervention will, in general, be beneficial to
the country in question and to asset holders. We subscribe to this view because countries asking
for help often have liquidity constraint and/or solvency problems, but their assets still outweigh
their liabilities. The cash inflow from IMF loans keeps the system working until liquidity has
been restored or solvency issues have been overcome, enabling the economy to maintainB. Hayo, A.M. Kutan / Journal of International Money and Finance 24 (2005) 1126e1142 1131
current levels of output. A similar view is suggested by Sachs (1999) who emphasizes liquidity
problems caused by the withdrawal of foreign assets as a major factor for the worsening of the
Asian crisis. Thus, by injecting liquidity into the markets, IMF actions can positively influence
market returns. Alternatively, a delay of IMF loans may signal future liquidity constraints, thus
causing a decline in current market returns. Other negative announcements by the IMF may
represent an unfavorable external evaluation of a country’s macroeconomic policies and may
signal unattractive prospects, such as solvency problems, in the future.
IMF programs are usually market-oriented, and the loan part of programs always comes with
obligations to de-regulate the economy, which, on average, is good for business. Recent empirical
evidence suggests (see Evrensel, 2002) that countries participating in IMF programs increase
their reserves, and improve their current account balances during program years. During post-
program years, improvements in reserves and current account tend to disappear. Moreover,
countries’ macroeconomic policies do not appear to change significantly during stabilization pro-
grams compared to the situation before. Therefore, since the IMF seems to be unable or unwilling
to implement the conditions outlined in the letters of intent, it is not clear that the business climate
is going to improve as a result of the official adoption of a stabilization program.
4. Data and econometric methodology
Our analysis is based on daily closing stock returns for six countries, Indonesia, South Ko-
rea, Argentina, Brazil, Pakistan, and Russia, over the time period from 1 July 1997 to 31 De-
cember 1999. The data for the closing stock prices were obtained from the Yahoo finance
website, http://finance.yahoo.com/m2. Countries were chosen to represent different regional
groups. The first two countries were affected by the Asian financial crisis, while the next
two were affected by the crisis in South America. The last two belong to a group of countries
whose crises were more idiosyncratic. Although Pakistan has not had a financial crisis recently,
it had significant interactions with the IMF. In addition, contagion from financial crises in other
countries is quite possible. For instance, creditor panics in one country with significant IMF
involvement that may spread to other emerging markets with similar involvement.
Our sample countries have had significant growth in their stock markets in the last decade.
For example, market capitalization of the markets as a percentage of GDP in 2001 ranged be-
tween 11 percent for Pakistan and 58 percent for Argentina. The corresponding figures for 1990
were 7.0 and 2.3 percent, respectively (World Development Indicators, Table 5.3, 2000). Ac-
cording to the same source, the number of listed domestic companies has also grown. For ex-
ample, listed domestic companies for Indonesia, Pakistan, and Russia increased from 125, 487,
and 13 in 1990 to 316, 747, and 236 in 2001, respectively. This suggests an increase in the role
of domestic investors in these markets. However, Argentina and Brazil have experienced a slight
decline in the number of listed domestic companies, suggesting a growing role for foreign in-
vestors. Nevertheless, private fixed investment as a percentage of total gross fixed investment in
these two countries increased from 67.4 percent in 1990 to 89.9 percent in 1999 for Argentina
and from 76.7 percent to 86.2 percent for Brazil (World Development Indicators, Table 5.1,
2000). In other countries, private fixed investments made up more than 60 percent of total gross
investment in 1999. These observations suggest that stock markets have played a relatively sig-
nificant role in the overall economic development of our sample countries.
Composite stock returns (Returns) are computed by taking the first differences of logged daily
stock price indices, multiplied by 100. As a control variable and to capture the impact of1132 B. Hayo, A.M. Kutan / Journal of International Money and Finance 24 (2005) 1126e1142
general global market developments, we also include the U.S. Standard & Poor’s Stock Price In-
dex returns in our estimations (S&P returns). To account for the time-zone differences, the first
lag of the S&P index is being used. Descriptive statistics for both variables are reported in
Table 2. The mean return is about four times as high in the U.S. as it is in the emerging countries
over our sample period, while the standard deviation is less than half as large. The differences in
means are not only due to extreme cases. Comparing the 50th percentile of the two series, we find
a median of 0.035 for emerging markets returns and 0.136 in the case of S&P returns.
The emerging markets returns series exhibits excess kurtosis but almost no skewness. Excess
kurtosis signals a typical problem with financial data, namely that the series do not conform to
a normal distribution. Another problem is that time periods with high volatility are followed by
periods of low volatility, a phenomenon called volatility clustering. The existence of volatility
clustering implies that classical methods of estimation are not efficient. ARCH models (Engle,
1982) have been developed to address this situation. To take volatility clustering into account,
we employ a more general GARCH specification based on Bollerslev (1986).
Daily data on the six countries are combined into one panel data set. We then specify
a GARCH model with country-specific fixed effects. A model that works well for our sample
with regard to eliminating ARCH effects is the following GARCH (1,1) model with student-t
distributed residuals (see Bollerslev, 1987):
Returnst ZmCd Returnst1 Cg S&Pt1 Cf DummiesCut ;
ut Z3t h1=2 2
t ; ht Za0 Ca1 ut1 Cb1 ht1 ; ð1Þ
where a, b, d, g are parameters, f is a vector of parameters, Dummies is a matrix containing
country and IMF news dummies, and 3tjUt1 Z t[n] with Ut1 capturing all information up to
t 1 and t[n] a standardized t-distribution with n degrees of freedom.
We chose the GARCH (1,1) model in view of the fact that it fits a wide variety of financial
time series very well (see Campbell et al., 1997, 479ff). Thus, whether or not a financial crisis
occurred, the equation should yield satisfactory results. The use of a panel framework is nec-
essary because it would not be possible to obtain a sufficient number of observations for each
type of IMF news if we restrict the sample to a single country. In particular, the aggregation of
IMF events in the panel allows us not only to test for the effects of good versus bad IMF news
but also to analyze the more detailed categorization of IMF-related events outlined in Table 1.
There are also costs to using a panel, in particular with respect to assuming similar-sized
coefficients for the variables of interest and a common error structure. In the next section,
we begin the statistical analysis by studying the effects of good and bad IMF-related news.
5. Effect of IMF news on stock returns
Before we report the econometric results, it is worthwhile comparing raw returns on days
with IMF news and without IMF news. Table 3 summarizes the average raw returns for the
Table 2
Descriptive statistics of returns and S&P
Mean St. Dev. Minimum Maximum Skewness Excess kurtosis
Returns 0.02 3.33 26.6 28.8 0.09 8.7
S&P 0.08 1.24 7.1 5.0 0.48 4.1B. Hayo, A.M. Kutan / Journal of International Money and Finance 24 (2005) 1126e1142 1133
Table 3
Average raw returns on days
No good or bad news IMF good IMF bad
0.0065 0.7493 1.9589
No IMF-related news Favorable IMF statement Request from the IMF
0.0087 0.6196 1.145
IMF-supportive gov. announcement Loan approval Visit by IMF delegation
0.1844 0.9192 0.4421
Unfavorable statement by the IMF Agreement Talks or negotiations
0.7834 1.5942 0.0609
Anti-IMF gov. announcement Delay of loans or talks Resumption of loans
1.4721 3.7788 1.8510
different news categories. On ‘‘no announcement’’ days, mean returns are slightly positive. On
days of good news, returns increase to 0.8% and on bad days they drop to 2%. Favorable
statements, loan approvals, agreements, and resumption of loans coincide with positive stock
returns. Requests from the IMF, IMF-supportive announcements by governments, visits by
IMF delegations, unfavorable statement by IMF, talks and negotiations, anti-IMF announce-
ments by governments, and delay of loans coincide with negative stock returns. Of these cat-
egories, IMF-supportive announcements by governments and especially talks and negotiations
have only small absolute effects on returns. It follows from Table 3 that most results correspond
to our priors.
Next, we model the relationship between IMF news and stock returns. We estimate the
benchmark model as a GARCH (1,1) model with student-t distributed residuals (see Eq. (1))
using quasi-maximum likelihood techniques over the time period from 1 July 1997 to 31 De-
cember 1999. The country dummies are then tested. They are not significantly different from
zero (Chi2(6) Z 11.1) and can be removed from the model. We continue the analysis with
the resulting more parsimonious model (Model 1) given in Table 4.
We find that all parameters relevant for the conditional variance are significant. A sufficient
condition for the conditional variance ht to be non-negative is that a0, a1, and b1 are non-neg-
ative, which is fulfilled here. Moreover, the sum of a1 and b1 is less than unity, ruling out that
the model is an integrated GARCH (see Nelson, 1990).6 The estimate of the student-t points
towards a distribution with four degrees of freedom that has fatter tails than a normal
distribution.
The diagnostic tests for Model 1 indicate that there is neither any trace of ARCH left nor is
the Portmanteau-type test for autocorrelation (for 60 lags) significant. The only problem is non-
normality of the residuals. However, a non-parametric estimate of the residual density indicates
that it is uni-modal and symmetric around zero, and therefore testing should not be too adversely
affected. Moreover, we use robust standard errors based on Bollerslev and Wooldridge (1992)
in Tables 4 and 5. In any case, for most variables, both types of standard errors are quite
similar.
6
Estimates of ARFIMA models show no evidence of fractional integration (results omitted). Thus, the model is both
strictly stationary as well as covariance stationary.1134 B. Hayo, A.M. Kutan / Journal of International Money and Finance 24 (2005) 1126e1142
Table 4
Explaining stock returns using a GARCH (1,1) model with t-student distribution
Model 1 Model 2
Coeff. SE Coeff. SE
a0 0.39** 0.10 0.41** 0.10
a1 0.23** 0.03 0.22** 0.04
b1 0.77** 0.03 0.77** 0.03
Student-t d.o.f. (n) 4.26 4.37
Returnst1 0.07** 0.02 0.07** 0.02
S&Pt1 0.21** 0.03 0.21** 0.03
Positive IMF news 0.55* 0.27
Negative IMF news 1.49** 0.71
Favorable IMF statement 0.84 0.69
Request from the IMF 0.007 1.62
IMF-supportive announcement by gov. 0.60 0.35
Loan approval 0.42 0.35
Visit by IMF delegation 0.35 0.77
Unfavorable statement by the IMF 1.43 1.03
Agreement 0.62 0.49
Talks or negotiations 0.007 0.68
Anti-IMF announcements by gov. 1.45 1.56
Delay of loans or talks 1.02* 0.48
Resumption of loans 1.42 2.23
Number of observations 3501 3501
Log-likelihood 8364.5 8363.1
Normality test Chi2(2) Z 2585** Chi2(2) Z 2449**
ARCH 1e2 test F(2, 3489) Z 0.58 F(2, 3480) Z 0.71
Portmanteau test Chi2(60) Z 78.2 Chi2(60) Z 79.1*
Notes: * (**) indicates significance at a 5% (1%) level. Standard errors are heteroscedasticity-consistent.
We also find that the lag of the dependent variable matters. This implies that stock markets
are not efficient, probably reflecting thin trading and the effects of crises on these markets.7 In
addition, Standard & Poor’s returns matter. In other words, S&P returns appear to be Granger-
causing emerging market returns. An increase in U.S. returns by one percentage point raises
emerging markets returns by 0.21 percentage points on the next day.
With regard to our variables of interest, we detect significant coefficient estimates for both pos-
itive and negative IMF events. In line with our expectations, positive (negative) news has a positive
(negative) effect on returns. The size of the point estimates suggests that there is some asymmetry,
namely that negative IMF news (1.49) has a larger absolute impact than positive ones (0.55).
Testing the absolute size of the coefficient for equality, however, reveals that the data does not re-
ject such a hypothesis (Chi2(1) Z 1.25). Thus, we cannot reject the hypothesis that IMF news ef-
fects on stock market returns are symmetric.
Given that the size of the coefficients for both good and bad news is close to unity, one may
conjecture that the economic impact on stock returns of these types of news is very similar (one
7
Omitting the lagged dependent variable from our estimations did not qualitatively change the results with respect to
the parameters of interest: IMF good news: 0.58*, IMF bad news: 1.40*. Since this omission brings about a significant
degree of autocorrelation in returns, we have decided to leave the lagged dependent variable in the equation.B. Hayo, A.M. Kutan / Journal of International Money and Finance 24 (2005) 1126e1142 1135
Table 5
Explaining stock returns with IMF news also in the variance equation
Model 3 Model 4
Coeff. SE Coeff. SE
a0 0.39** 0.10 0.41** 0.10
a1 0.23** 0.03 0.22** 0.04
b1 0.77** 0.03 0.77** 0.03
Student-t d.o.f. (n) 4.25 4.23
Returnst1 0.07** 0.02 0.07** 0.02
S&Pt1 0.21** 0.03 0.21** 0.03
Positive IMF news 0.56* 0.27
Negative IMF news 1.52* 0.71
Delay of loans or talks 1.12* 0.52
IMF news aggregate based on good 0.70 0.97
and bad in variance equation
IMF news aggregate based on all items 0.96 0.87
in variance equation
Number of observations 3501 3501
Log-likelihood 8364.1 8367.1
Normality test Chi2(2) Z 2631** Chi2(2) Z 2661**
ARCH 1e2 test F(2, 3488) Z 0.58 F(2, 3480) Z 0.58
Portmanteau test Chi2(60) Z 77.4 Chi2(60) Z 77.2
Notes: * (**) indicates significance at a 5% (1%) level. Standard errors are heteroscedasticity-consistent.
percentage point). The null hypothesis that the coefficient for good news is equal to one cannot
be rejected at a 5% level (Chi2(1) Z 3.02). In the case of bad news, we can neither reject that
the coefficient is equal to unity in absolute terms (Chi2(1) Z 0.38), nor do we have to reject the
hypothesis when testing the hypotheses of an absolute influence of unity jointly for good and
bad news (Chi2(2) Z 3.40). This is not necessarily the result of a weak testing procedure, as
testing the joint hypotheses that both news coefficients are equal to zero in absolute terms
can be rejected (Chi2(2) Z 8.04*). Although there remain some indications that negative shocks
tend to have an absolutely greater effect, we conclude that the data are consistent with the state-
ment that positive (negative) IMF news on average increases (decreases) stock returns by about
one percentage point.
Next, we replace the simple dichotomy of good versus bad IMF news with the detailed clas-
sification of IMF-related news items. These categories include signed agreements, delay of
loans or talks between the countries and the IMF, IMF-supportive announcements, loan appro-
vals, request of funds, talks/negotiations, resume of negotiations, visits, anti-IMF government
policies, and favorable and unfavorable IMF statements about the countries’ economic perfor-
mance. A detailed explanation of these categories is given in Table 1.
Again, we start with the general model outlined above and are able to remove all country
dummies in a consistent testing-down process (Chi2(6) Z 11.3), the outcome of which is Model
2 in Table 4. The diagnostic tests remain generally unchanged, except for a slight violation of
the null hypothesis in the autocorrelation test ( p-value: 0.0496).
The impacts of different news categories turn out to be insignificant except for ‘‘delay of
loans’’. We cannot reject the hypothesis that all IMF news items have a zero coefficient apart
from ‘‘delays’’ (Chi2(10) Z 9.86). Because we cannot reject the hypothesis that the impact of
delay news is 1 (Chi2(1) Z 0.20), we conclude that an announcement of a delay of loans on1136 B. Hayo, A.M. Kutan / Journal of International Money and Finance 24 (2005) 1126e1142
average reduces stock returns by about one percentage point. The finding that the ‘‘delay’’ news
leads to a stock market decline can indicate evidence of investor concerns about liquidity con-
straints or solvency problems. Alternatively, it could suggest poor progress on, or failure of,
government reform policies, thus signaling bad prospects for the economy.
Finally, there is the question of whether IMF news has an impact on the volatility of stock
markets or, more technically, whether news variables enter the variance equation significantly
(see Bollerslev and Ghysels, 1996).8 Negative IMF news may signal poor economic perfor-
mance and thus higher risk in the future. However, this risk should not necessarily lead to
a higher volatility in the very short run. If IMF news leads to a convergence of market partic-
ipants’ expectations, then volatility might even decline. In other words, the news may also af-
fect volatility by reducing the information asymmetry in the market, thus resulting in a decline
in short-term volatility.9 If IMF news is, to a large extent, expected, we should not expect to see
a dramatic change in volatility.
We construct two aggregate news categories. The first is based on the categories good and
bad news, and the second one on the other items displayed in Table 1. Table 5 lists the estima-
tion results, the diagnostics of which are very similar to those in Table 4. We find that the first
news aggregate, based on good and bad news events, which was included in Model 3, has a pos-
itive impact on the conditional variance. This effect, however, is not statistically significant.
The same conclusion holds for the news variable computed using all available IMF news cat-
egories, displayed in Model 4. In addition, relaxing the restriction that all news items contribute
to the same extent to volatility does not yield any significant parameter estimates (results
omitted).10
Moreover, neither the statistical nor the economic significance of IMF events on stock re-
turns is affected by including the news variables into the variance equation. To summarize,
in our framework, IMF news does not contribute significantly to a cluster of higher or lower
stock market volatility.
6. Dynamic aspects of IMF news
So far the analysis has concentrated on the impact of news on the day of announcement. It is
interesting to investigate the time pattern of news effects. In particular, we want to analyze re-
turns on the day before the announcement and the day after. This gives us information about the
economic situation that exists before the news comes in. We can also draw inferences about the
longevity of gains and losses. For instance, the presence of a major investor panic in financial
markets requires sustained changes in stock returns in response to specific IMF news.
To investigate these questions, we include dummy variables for the days before and after the
announcement in our model (see Table 6). We find that there are different time patterns for good
and bad news. On the days before good news, returns are above average but not significantly so.
After the announcement day, returns fell, but the coefficient on the dummy variable is only
8
As another robustness check, we also considered modeling with an EGARCH framework (Nelson, 1991). However,
for our sample, estimating an EGARCH with a non-normal distribution did not lead to converging estimates. The
EGARCH with a normal distribution turned out to be inferior to the model in Table 4 with regard to log-likelihood value
and Akaike information criterion. In any case, point estimates were of similar magnitude.
9
We are grateful to a referee for suggesting these two interpretations.
10
In addition, leaving out the IMF-related news categories in the returns equation when testing the effect of news on
volatility leads to the same conclusion.B. Hayo, A.M. Kutan / Journal of International Money and Finance 24 (2005) 1126e1142 1137
Table 6
Stock returns GARCH model with before and after announcement dummies
Model 5
Coeff. SE
a0 0.41** 0.10
a1 0.22** 0.03
b1 0.77** 0.03
Student-t d.o.f. (n) 4.41
Returnst1 0.07** 0.02
S&Pt1 0.21** 0.03
Day before positive IMF news 0.12 0.23
Positive IMF news 0.56* 0.27
Day after positive IMF news 0.44 0.26
Day before negative IMF news 1.23* 0.61
Negative IMF news 1.52* 0.72
Day after negative IMF news 0.66 1.13
Number of observations 3501
Log-likelihood 8361.3
Normality test Chi2(2) Z 2476**
ARCH 1e2 test F(2, 3485) Z 0.72
Portmanteau test Chi2(60) Z 78.6
Notes: * (**) indicates significance at a 5% (1%) level. Standard errors are heteroscedasticity-consistent.
significant at a 10% level. We can now test whether the sum of coefficients on these three days
is zero. The test statistics is Chi2(1) Z 0.33, which is not significant. Even looking only at the
event day and the day after reveals no significant result (Chi2(1) Z 0.10). Thus, good news has
no sustained effect on the stock market.
In the case of bad news, average returns on the day before the IMF events are significantly
above normal. We cannot reject the hypothesis that the coefficient is equal to unity
(Chi2(1) Z 0.09). There is a drop in returns on the event day and a recovery resulting in (non-sig-
nificant) above average returns on the day after. We cannot reject the hypothesis that the sum of
the three coefficients is zero (Chi2(1) Z 0.07). When concentrating on the event day and the fol-
lowing day, neither can we reject the hypothesis that the effect is zero (Chi2(1) Z 0.59).
The following conclusions emerge from these results. First, our IMF news seems to be news
indeed, as the changes in stock market returns are not foreshadowed by the trading on the day
before the event. Second, over a three-day period bracketing the IMF news, or a two-day period
of event day and aftermath, no significant deviations from average returns are realized. In other
words, there is no sustained impact of IMF actions on returns. Hence, IMF-related news does
not cause creditor panic. Third, the time pattern of good and bad returns is different on these
days. Good news enhances a somewhat positive stock market situation. Bad news occurs, on
average, after significantly positive returns on the day before. Fourth, the point estimates sug-
gest that fluctuations in returns surrounding bad events are stronger than those for good events,
but this conclusion cannot be maintained when taking statistical uncertainty into account.
7. Further evidence from foreign exchange and bonds markets
After drawing out a number of conclusions about the influence of IMF news on stock mar-
kets, we ask whether similar effects can be detected in other financial markets. Here we1138 B. Hayo, A.M. Kutan / Journal of International Money and Finance 24 (2005) 1126e1142
concentrate on foreign exchange markets and interest rate spreads in bond markets.11 The find-
ings in this section should be interpreted cautiously because of the potential effects of govern-
ment intervention in the market.
Starting with exchange rates, we compute the daily returns (growth rates) in percentage
points. Exchange rates are measured in price notation, which means an increase (decrease) re-
flects a depreciation (appreciation) of the domestic currency. We estimate a model similar to the
one for stock market returns. To somewhat control for international portfolio shifts, we also in-
clude lagged emerging market stock market returns, as well as S&P returns. These variables
enter with a lag of one period to avoid simultaneity problems. However, they do not survive
the testing-down process (Chi2(6) Z 4.7).12 The results of the modeling procedure are dis-
played in Model 6 of Table 7. The diagnostic tests indicate no trace of autocorrelation or
ARCH. Non-normality is even worse than in the case of stock market returns. In addition,
the model would not be stationary if estimated unrestricted (IGARCH). However, to avoid these
complications, we restrict the sum of the coefficients of a1 and b1 to be less than one, which
ensures stationarity.
This time we find significant country differences, with Brazil and Russia showing an above
average depreciation rate. There is also a significant influence of previous day returns. Finally,
we analyze the time pattern of effects surrounding IMF news again. On the news day, good
news has no noticeable effect on exchange rates. However, bad news leads to a depreciation
of the currency. Further testing reveals that we cannot reject the hypothesis that the loss in ex-
ternal value amounts to 0.05 percentage points (Chi2(1) Z 0.42). Looking at the time pattern
again, we find no significant coefficients for good news. For bad news, losses on the announce-
ment day are completely offset by gains on the previous day and the day after (Chi2(1) Z 0.12).
Thus, bad IMF news stops an appreciation of the exchange rate.
An interpretation of this finding is that market participants were hoping for an intervention by
the IMF to support the local currency. Bad IMF news then signals that this will not be the case, and
the fear of further exchange rate devaluations in the future prompts investors to sell the local cur-
rency. However, an effect of five basis points on exchange rate returns is not very dramatic in eco-
nomic terms. These results provide no strong evidence in favor of the creditor panic hypothesis.
Results for the disaggregated categories in Model 7 of Table 7 reveal that the results in stock mar-
kets only partially carry over to foreign exchange markets.13 In particular, while the delay of loans
was the most significant effect on stock markets, this has no effect on currency markets. Instead, oth-
er types of news seem important. Requests for assistance from the IMF lead to an appreciation of the
exchange rates, as do talks with the IMF and the resumption of loan payments. However, we also find
that anti-IMF announcements by government have the same effect, which is rather unexpected. A
depreciation is caused by unfavorable comments by the IMF about the country.
Regarding the impact of news on volatility, neither do we find a significant influence of the
sum of good and bad news on volatility, nor do we find an effect when considering the impact of
all categories (results omitted).
11
Data are obtained from Datastream. Bond spread data reflect the difference between yields of local bonds and U.S.
T-bill. We are grateful to Michael Melvin for his generous help with the data.
12
It should be noted that when the analysis is performed with contemporary values, stock market returns are signif-
icant, while contemporary S&P returns are not. The conclusions for the variables of interest are robust to this change in
the specification of the model.
13
The removal of some country dummies and domestic and foreign stock return variables cannot be rejected
(Chi2(6) Z 4.59).B. Hayo, A.M. Kutan / Journal of International Money and Finance 24 (2005) 1126e1142 1139
Table 7
Exchange rate returns GARCH model
Model 6 Model 7
Coeff. SE Coeff. SE
a0 0.0002** 0.00003 0.0002** 0.00003
a1 0.34 0.34
b1 0.66** 0.02 0.66** 0.02
Student-t d.o.f. (n) 3.14 3.14
Brazil 0.03** 0.004 0.03** 0.004
Russia 0.04** 0.01 0.04** 0.01
Exchange rate returnst1 0.14** 0.02 0.14** 0.02
Day before positive IMF news 0.0001 0.005
Positive IMF news 0.0004 0.005
Day after positive IMF news 0.01 0.007
Day before negative IMF news 0.03** 0.007
Negative IMF news 0.04** 0.007
Day after negative IMF news 0.0006 0.009
Favorable IMF statement 0.008 0.01
Request from the IMF 0.30** 0.02
IMF-supportive announcement by gov. 0.06* 0.03
Loan approval 0.008 0.01
Visit by IMF delegation 0.006 0.02
Unfavorable statement by the IMF 0.045 0.01
Agreement 0.011 0.01
Talks or negotiations 0.567* 0.02
Anti-IMF announcements by gov. 0.053 0.01
Delay of loans or talks w0.052 0.05
Resumption of loans 0.160** 0.032
Number of observations 3381 3381
Log-likelihood 1400.8 1391.1
Normality test Chi2(2) Z 2 914 000** Chi2(2) Z 2 941 500**
ARCH 1e2 test F(2, 3358) Z 0.002 F(2, 3359) Z 0.002
Portmanteau test Chi2(60) Z 1.24 Chi2(60) Z 1.13
Notes: * (**) indicates significance at a 5% (1%) level. Standard errors are heteroscedasticity-consistent. The model was
forced to be stationary by restricting the sum of a1 and b1 to be slightly below one.
Next we consider bond markets, where we analyze the effect of IMF news on bond spreads.
Since we were unable to obtain consistent bond spread data for all countries in our sample, the
analysis continues without Brazil and Pakistan.14 As in the case of the other variables, we look
at the daily growth in bond spreads in percent. The testing-down process (Chi2(4) Z 2.79) re-
sults in Model 8 of Table 8. Domestic and foreign stock returns show up significantly negative,
as does the first lag of the dependent variable. The latter indicates non-efficiency of markets. As
in the case of exchange rates, we have to restrict GARCH coefficients to ensure stationarity.
Regarding our variables of interest, we do not find any significant effect of IMF news on
bond spreads, indicating no evidence of creditor panic.
14
When estimating the stock market returns model for this smaller data set, we get similar point estimates. However,
the coefficient on good news is no longer significant. This probably reflects the loss of information for the estimation of
parameters. We also lose significance of the negative effect of delay on stock returns. Instead, the only (positive) sig-
nificant influence is now support for IMF policies by emerging market governments.1140 B. Hayo, A.M. Kutan / Journal of International Money and Finance 24 (2005) 1126e1142
Table 8
Bond spreads growth GARCH model
Model 8 Model 9
Coeff. SE Coeff. SE
a0 2.40 1.78 2.63 1.65
a1 0.39** 0.41
b 0.61** 0.15 0.59** 0.13
Student-t d.o.f. (n) 2.99 2.99
Bond spreads growtht1 0.11** 0.02 0.11** 0.02
Stock market returnst1 0.08** 0.03 0.08** 0.03
S&P returnst1 0.28** 0.05 0.28** 0.05
Day before positive IMF news 0.14 0.39
Positive IMF news 0.40 0.39
Day after positive IMF news 0.27 0.37
Day before negative IMF news 0.07 0.62
Negative IMF news 1.08 1.20
Day after negative IMF news 0.34 1.49
Favorable IMF statement 1.58* 0.72
Request from the IMF 0.77 4.36
IMF-supportive announcement by gov. 0.24 0.93
Loan approval 0.08 0.49
Visit by IMF delegation 0.79 0.46
Unfavorable statement by the IMF 2.06** 0.72
Agreement 3.05 5.45
Talks or negotiations 0.73 0.85
Anti-IMF announcements by gov. 0.06 0.68
Delay of loans or talks 5.66 2.92
Resumption of loans 2.78* 1.32
Number of observations 2270 2270
Log-likelihood 6383.9 6381.0
Normality test Chi2(2) Z 4727** Chi2(2) Z 4581**
ARCH 1e2 test F(2, 2253) Z 1.47 F(2, 2249) Z 1.55
Portmanteau test Chi2(46) Z 59.6 Chi2(46) Z 59.8
Notes: * (**) indicates significance at a 5% (1%) level. Standard errors are heteroscedasticity-consistent. The model was
forced to be stationary by restricting the sum of a1 and b1 to be slightly below one.
Model 9, after the reduction process (Chi2(4) Z 2.37), looks at the disaggregated categories.
There are four significant effects, three of which are in accordance with our expectations. Fa-
vorable announcements by the IMF about the respective country and the resumption of loan
payments lower interest rate spreads, while the delay of loans causes an increase in bond
spreads.15 The remaining significant coefficient is rather counterintuitive because unfavorable
announcements by the IMF lead to a decrease in bond spreads. One explanation is that market
participants did not find the news as unfavorable as they expected. Finally, volatility is not af-
fected by the occurrence of IMF-related events (results omitted).
To conclude, the impact of IMF news is also apparent in exchange rate data, but less so in bond
data. In addition, the impact of particular categories is not fully robust. As we move from one
15
The coefficient on ‘‘delay’’ is strictly speaking not significant but we consider a p-value of 0.053 to be reasonably
close.B. Hayo, A.M. Kutan / Journal of International Money and Finance 24 (2005) 1126e1142 1141
market to another, there is a shift of significance from one variable to another, while the individual
signs do seem to be in line with a priori expectations. The differences in results between the
stock market and the other two markets may be explained by the greater role of the government
in the latter markets. Finally, news does not affect volatility in any of the equations considered
above.
8. Conclusions
We have analyzed the impact of IMF-related events on six emerging stock market returns
during recent financial crises and IMF bailouts. In such difficult times, the IMF can be viewed
as a lender of last resort, and investors trading in these markets form their actions based on their
anticipation of the actions of the IMF in the near future. Our hypothesis was that investors’ re-
sponses to IMF news are directly reflected in stock market returns. To test this hypothesis, we
constructed a panel data set covering six emerging markets and found that good and bad news
have statistically significant effects on stock market returns. This finding suggests that IMF ac-
tions have noticeable wealth effects for investors. On average, bad (good) IMF news decrease
(increase) stock returns by one percentage point. These results are consistent with earlier stud-
ies that also report significant effects of IMF announcements on asset returns.
Among the different categories of IMF actions, we find that stock markets react most signif-
icantly to delays of IMF loans or talks, suggesting that liquidity concerns, or solvency issues
play a very significant role in emerging markets during financial crises. As a rule of thumb,
delays of loans reduce stock market returns by about one percentage point. The robustness
of these results is assessed using data for foreign exchange and bond markets. A number of re-
sults carry over, especially that bad news causes a depreciation of the exchange rate. We also
look at the time pattern of market returns and find that there are differences between good and
bad news and across financial markets. As a general conclusion, we cannot detect any longer-
term impact of IMF news induced gains or losses.
In our opinion, these findings contradict Sachs’ (1999) view that a creditor panic took place
after IMF announcements. We do find that returns fall as a result of negative IMF news, but the
economic effect of IMF news is quite modest compared to the overall variation of returns in the
data. For example, in our sample, stock returns fluctuate between 26.6 and C28.8 percent
while the effect of IMF news is about 1 percent. Moreover, these effects do not last very
long. Thus, we do not find support for the hypothesis of IMF-induced investor panics. Even
stronger conclusions in this respect emerge from the analysis of foreign exchange and bonds
markets.
Further, we did not find strong evidence that IMF news increases market volatility. Hence,
although IMF news causes wealth effects, they do not appear to contribute significantly to over-
all market uncertainty. These results suggest that markets had largely discounted the effects of
IMF actions on volatility and functioned relatively smoothly even during this crisis period.
Based on these results, we would argue that IMF actions and events primarily have an effect
on returns, but not on risk.16
16
This finding that IMF actions in general do not affect market volatility is consistent with our evidence that IMF
actions do not appear to cause investor panic. Investor panic, resulting in a sudden increase in selling activity, would
cause higher market volatility.1142 B. Hayo, A.M. Kutan / Journal of International Money and Finance 24 (2005) 1126e1142 Acknowledgements We would like to thank two anonymous referees, Josef C. Brada, Volker Clausen, Radcliffe G. Edmonds Jr., Aysxe Y. Evrensel, and seminar participants at the Universities of Bonn and Hannover for many useful comments and suggestions. The usual disclaimer applies. References Aggarwal, R., Inclan, C., Leal, R., 1999. Volatility in emerging stock markets. The Journal of Financial and Quantitative Analysis 34, 33e35. Aylward, L., Thorne, R., 1998. An Econometric Analysis of Countries’ Repayment Performance to the International Monetary Fund. IMF Working Paper WP/98/32. Bekaert, G., Harvey, C.R., 1997. Emerging market volatility. Journal of Financial Economics 43, 29e77. Brealey, R.A., Kaplanis, E.C., July 2000. The Impact of the IMF Programs on Asset Values. Working Paper, London Business School. Bollerslev, T., 1986. Generalized autoregressive conditional heteroscedasticity. Journal of Econometrics 31, 307e327. Bollerslev, T., 1987. A conditional heteroscedastic model for speculative prices and rates of return. The Review of Eco- nomics and Statistics 69, 542e547. Bollerslev, T., Ghysels, G., 1996. Periodic autoregressive heteroscedasticity. The Journal of Business and Economics Statistics 14, 139e151. Bollerslev, T., Wooldridge, J.M., 1992. Quasi-maximum likelihood estimation and inference in dynamic models with time varying covariances. Econometric Reviews 11, 143e172. Campbell, J.Y., Lo, A.W., MacKinlay, A.C., 1997. The Econometrics of Financial Markets. Princeton University Press, Princeton. Dong, L., Kho, B.C., Stulz, R., 2000. U.S. banks, crises and bailouts: from Mexico to LTCM. The American Economic Review 90, 28e37. Engle, R., 1982. Autoregressive conditional heteroscedasticity with estimates of the variance of United Kingdom infla- tion. Econometrica 50, 987e1008. Evrensel, A.Y., 2002. Effectiveness of IMF-supported stabilization programs in developing countries. Journal of Inter- national Money and Finance 21, 656e687. Fleming, J., Kirby, C., Ostdiek, B., 1998. Information and volatility linkages in the stock, bond, and money markets. Journal of Financial Economics 49, 111e137. Ganapolsky, E.J.J., Schmukler, S.L., 1998. Crisis Management in Argentina during the 1994e95 Mexican Crisis: How did markets react? World Bank Policy Research Working Paper No. 1951. Kaminsky, G.L., Schmukler, S.L., 1999. What triggers market jitters? A chronicle of the Asian crisis. Journal of Inter- national Money and Finance 18, 501e514. Katz, S.S., 1998. The Asian crisis, the IMF and the critics. Eastern Economic Journal 25, 421e439. Kho, B.C., Stulz, R., August 1999. Banks, the IMF and the Asian Crisis Working Paper. Ohio State University. Kho, B.-C., Lee, D.W., Stulz, R.M., 1999. U.S. Banks, Crises, and Bailouts: From Mexico to LTCM. European Corpo- rate Government Institute. Lane, T., Phillips, S., October 2000. Does IMF Financing Result in Moral Hazard. IMF Working Paper WP/00/168. Naim, M., 2000. Washington consensus or Washington confusion. Foreign Policy 118, 87e103. Nelson, D., 1990. Stationarity and persistence in the GARCH (1,1) model. Econometric Theory 6, 318e334. Nelson, D., 1991. Conditional heteroscedasticity in asset pricing: a new approach. Econometrica 59, 347e370. Sachs, J., 1999. Creditor panics: causes and remedies. Cato Journal 18, 377e390. Sarno, L., Taylor, M.P., 1999. Moral hazard, asset price bubbles, capital flows, and the East Asian crisis: the first tests. Journal of International Money and Finance 18, 637e657. World Bank, 2000. World Development Indicators. Washington.
You can also read
