Journal of Business and Economic Development

2024; Vol. 9, No. 2, pp. 21-35

https://doi.org/10.11648/j.jbed.20240902.11

Research Article

Macroeconomic Currency Crisis Early Warning Indicators in

Emerging Countries
Siriki Coulibaly1, 2, 3, * , Aya Marouane3
1
DEECAF (Economic Destination Africa), Rennes, France
2
Department of Economics, University Péléforo Gon Coulibaly, Korhogo, Ivory Coast
3
Research and Development (R&D), Quantylix, Tunis, Tunisia

Abstract
This study investigates the relationships between exchange rate and the main macroeconomic variables as GDP, inflation and
unemployment on one hand and the ability of these variables in alerting about coming exchange rate crisis in emerging countries.
The three variables have significant coefficients with exchange rate in line with literature signs except unemployment rate. The
study uses signal approach, dealing specifically with the main macroeconomic variables, selected by system GMM method in
emerging markets. The study develops macroeconomic pressure indices from these selected macroeconomic variables using the
market pressure index methodology from Early Warning System literature. Based on the macroeconomic variables, a
combined macroeconomic pressure index has been built. The results of the non-parametric early warning system indicate that
the individual macroeconomic pressure indexes created are good warning tools of a currency crisis. The macroeconomic pressure
indexes are better early warning indicators than market pressure index built from international reserves, in emerging countries for
four quarters warning period window. Production pressure index appears more accurate followed by inflation but unemployment
pressure index is the most sensitive. However, the number of effective indicators and the accuracy of the indexes are not the same
for all the countries, changing from a country to another.

Keywords
Exchange Rate, Crisis, Macroeconomics, Warning System

1. Introduction
The world is in an era of growing globalization of business other from domination and contagions. Established states
activities and volatility in international financial markets. have their own policies in line with the sovereignty principle.
Globalization is in fact today what implicitly establishes Monetary sovereignty allows each country to have its own
bridges between countries made of rules that govern interac- currency. But not all countries have the same economic
tions and make sure that countries are protected from each weight allowing different values of currencies and establish-

*
Corresponding author:

Received: 21 March 2024; Accepted: 8 April 2024; Published: 10 May 2024

Copyright: © The Author(s), 2024. Published by Science Publishing Group. This is an Open Access article, distributed
under the terms of the Creative Commons Attribution 4.0 License (http://creativecommons.org/licenses/by/4.0/), which
permits unrestricted use, distribution and reproduction in any medium, provided the original work is properly cited.
Journal of Business and Economic Development http://www.sciencepg.com/journal/jbed

ing exchange rates. Globally, the volatility in the activities is undesired consequences thereof, (Du & Lai [17]).
reflected in market prices. The exchange value risk is reoc- The currency could stir volatility depending on the eco-
curring and overwhelming with growing uncertainty and nomic situation. Macroeconomic variables have direct or
volatility. indirect impact on exchange rate movements and fluctuations
According to Bartram and Karolyi [4], the management of in exchange rates may have an adverse effect on the economy,
foreign exchange rate risk, in this globalization era, gains in Abdoh [1]. This questions about the specification of the rela-
importance in financial and nonfinancial sectors. The focus tionship between exchange rate and macroeconomic variables,
has been first made on the relationship between exchange rate is it a dynamic one?
and the volatility of firms‘ cash flow, managers concerns. The This study focuses on selected main macroeconomic vari-
Basel Committee on Banking Supervision [6] defines market ables, GDP, inflation and unemployment, by determining first
risk as ‗the risk of losses in on and off-balance-sheet positions their relationship with exchange rate and second investigates
arising from movements in market prices‘ (BIS [5]). Ac- on their ability to alert on the occurrence of currency crises
cording to Swami et al. [54], market risk is the risk of losses to regarding different time horizons in emerging economies.
the bank arising from movements in market prices as a result The paper is organized as follows. The following section
of changes in interest rates, foreign exchange rates and equity describes exchange rate systems. In the third section we re-
and commodity prices. The focus source of market prices view first the literature macroeconomic variables and ex-
volatility here is Foreign exchange rate risk, which is the risk change rate nexus and second, the currency crisis early
that the value of the bank‘s assets or liabilities changes comes warning system. The fourth section presents methodological
from currency exchange rate fluctuations. Generally, banks aspects of the study. Empirical findings are given in the fifth
are vulnerable to three types of foreign exchange risk: trans- section and section six concludes and gives recommendations.
action (commitment), economic (operational, competitive or
cash flow) and translation (accounting) (Abor [2]). Transac- 1.1. Exchange Rate Systems
tion risk arises when the value of existing obligations is dete-
riorated by movements in foreign exchange rates (Abor [2]). According to Rose [47], the exchange rate is an unusual
Economic risk occurs due to impact of high unexpected vol- asset price and even the most heavily traded asset price in
atility in the exchange rate on equity/income for both domes- that it has official regimes of volatility. There are different
tic and foreign operations. Translation risk is associated with exchange rate regimes that governments choose from and
the assets or income derived from offshore activities (Abor that are managed by their central bank. A country should
[2]). Exchange rate risk appears large enough to destabilize an choose the regime best suited to meet its particular economic
economy. challenges, taking into account in its decision the implica-
In fact, policy makers are interested today in understanding tions of this choice for overall systemic stability, from
possible sources and capturing this kind of risk because in the Ghosh et al. [26]. In theory, if the right regime is adopted, it
long terms, sharp fluctuations of the exchange rate will result could facilitate better business climate (Mohammed et al.
in a currency crisis which can be economically fatal. 2017). There are two polar extreme, either freely floating
Since the early 1980s, researchers have been working on currency or hard peg such as a currency union or currency
approaches to analyze the vulnerability of exchange rates and board, Ilzetzki et al. [32]. However, intermediate regimes are
predict currency crises. The first theoretical models of exter- some popular options for many countries in line with polit-
nal shocks on asset valuation schemes had been developed a ical-economy considerations according to Ilzetzki et al. [32].
few years prior, pushing researchers to consider crises as The paper borrows Abdoh et al. [1] exchange rate systems.
rational processes. Important countries and regions faced
serious currency crises soon after. For instance, the crises in 1.2. Fixed Exchange Rate System
Latin America (e.g., Mexico 1994–95, Brazil 1999, Argentina
2001), Asia (e.g., Thailand, South Korea, Indonesia 1997–98), In a fixed exchange rate system, the exchange rate was
and Russia 1998 raised new issues. These crises were actually being allowed to volatility only within very narrow bounda-
followed by a banking and sovereign debt crisis. All of this ries. If the exchange rate begins to move extreme, the gov-
intensified the need for tools to predict or at least better un- ernments will intervene in order to maintain it within the
derstand crises among policymakers and the financial indus- boundaries. It is to ensure that the exchange rates movement is
try. It is intuitive and meaningful to ask the question whether drifted no more than one percent above.
we can identify a distinct pre-crisis regime before these crises
happened, so that these currency crises are not total surprises. 1.3. Freely Floating Exchange Rate System
The solution of this question may give policymakers of each
country enough time to take preemptive actions to deal with Under freely floating exchange rate system, the exchange
the coming currency crisis and even prevent or avoid the rate values would be determined by the market forces. The

22
Journal of Business and Economic Development http://www.sciencepg.com/journal/jbed

freely floating exchange rate systems are not intervention by non where more than 20 percent of the value of domestic
various governments in the country. Under this system also, a currency suddenly drops against the foreign currency
central bank is not forced to implement an intervention policy (Yazdani & Nikzad [56]. In literature, at least three theoretical
that may have an unfavorable effect on the economy just sources for currency crises are developed. These underlying
control exchange rates. theories include, in the following order of ranking, weak
macroeconomic fundamentals, economic actors‘ bad expec-
1.4. Managed Float Exchange Rate System tations and ineffective banking intermediary. These sources of
currency crises seem to be linked, being able to cascade one
Managed float exchange rate system is similar to the fixed after the other until the crisis. Macroeconomic results could
system. The managed float exchange rate system is allowed impact actors‘ expectations expressed in speculative attacks
the governments to intervene for preventing their currencies that could affect in turn, through financial panic, banking
from moving too much. This system shown that the currencies intermediation. That puts macroeconomic variables at the
have no explicit boundaries. But this will tie in with supply heart of the topic and puzzle out their relationship with ex-
and demand factors. change rate when studying countries currency crises. The
theoretical relationships between the exchange rate and
1.5. Pegged Exchange Rate System macroeconomic variables are known, but there is no empirical
consensus on their significance, direction and sign. Under
Some countries may use pegged exchange rate system these conditions, before using a variable as a warning tool, it
when their bone currency‘s value is pegged to a foreign cur- is necessary to determine the relevance of its link with the
rency. One of the best-known pegged exchange rate ar- exchange rate.
rangements was been established by the European Economic For Simon [52], exchange rate badly affects mainly small
Community knows as EEC in April 1977, when the EEC economies and it has direct and positive relationship with
members decided to maintain their currencies to be estab- inflation. Roubini [48] stated that changes in macroeconomic
lished with the limits of each other. The market pressure will phenomenon could cause changes in exchange rate move-
cause some of the currencies to fluctuate at outside their es- ments. Specifically, he shows that the positive change in
tablished limits. nominal interest at domestic level will cause the currency to
be appreciated and can be vice versa. Kasif [35] shows for
Pakistan that, when exchange rate increased and the inflation
2. Literature Review rate decreased, that does not occur simultaneously. Harberger
Since the early 1980s, researchers have been working on [30] investigated the impact of GDP growth on real exchange
approaches to analyze the vulnerability of exchange rates and rate. He found that there is no systematic relationship between
to predict currency crises. Global, regional and local financial economic growth and real exchange rate. Kamin [34] empir-
crises follow one another, and they attracted considerable ically found that the relationships between inflation and the
attention in literature where two main views regarding them real exchange rates in most countries of Asia and Latin
emerged according to Budsayaplakorn et al. [8]. The first America shown a negative relationship. Husain et al. [31]
view supported by some researchers as [35, 42, 11, 34, 19, 17] experienced no robust relationship between economic fun-
and others, argues that the sources of financial crises are re- damentals and exchange rate in developing countries. In de-
gional contagion, investors panic, market behavior and veloped economies higher economic growth is associated
changings in market expectations. [13, 16, 14] are among with lower inflation and lower exchange rate, determining
those holding the second point of view, which makes weak negative connexion between GDP and exchange rate. Moc-
economic performances and the poor quality of institutions, cero [41] have done an investigation to find out the link be-
the determinants of crises. Many authors stress the role of tween the real exchange rate volatility and the export in Ar-
deteriorating economic fundamentals prior to currency crises gentina and found that there are significantly negative rela-
according to Budsayaplakorn et al. [8]. That, from this paper tionships between those variables. In the study of Achsani [3],
point of view, makes macroeconomic determinants potential inflation gives the theoretical correct sign in its relationship
warning signs of currency crises for an economy, provided with exchange rate, which is negative. Mirchandani [40]
they are statistically consistent. investigated various macroeconomic variables leading to
exchange rate in India. Results show that exchange rate has a
negative correlation with interest rate and inflation rate. The
2.1. Exchange Rate and Macroeconomic connexion of exchange rate with GDP and foreign direct
Variables Nexus investment is positive. But there is no correlation between
current account and exchange rate. Ramasamy & Abar [46]
Economists mentioned that currency crisis is a phenome-
show that interest rate, Balance of Payment and inflation rates

23
Journal of Business and Economic Development http://www.sciencepg.com/journal/jbed

should influence the exchange rate positively as per theory, behaviors prior to a currency crisis and estimate the proba-
but the results show the opposite. bility of crisis event within 24 months before. Kaminsky [33]
built a foreign exchange market pressure index (MPI) to
2.2. Currency Crisis Prediction quantify the financial crisis, which is the weighted average of
percentage change of exchange rate and international reserves.
According to Yazdani & Kikzad [56], the theoretical litera- In this study we choose to use signal approach, dealing spe-
ture has usually focused on the pre-crisis periods to investigate cifically with the main macroeconomic variables GDP, infla-
the reasons of currency crises. Mainly two major approaches tion and unemployment in emerging markets. In this study,
are expressed in the empirical literature on currency crises. The we develop macroeconomic pressure indices by selecting
first group of studies focus on crisis prediction [9, 25, 39], and relevant macroeconomic variables using the market pressure
the second group of studies consider the outcome of currency index methodology. Instead of basing on literature nexus we
crises and particularly output effects [29, 12]. We are interested choose to calculate the indices from the empirical nexus be-
in the first group dealing with currency crises prediction. Many tween exchange rate and macroeconomic variables. This is
authors stress the role of deteriorating some indicators prior to based on the non-consensus phenomenon in literature about
currency crises. They are various statistical and econometric these relationships (no relationship, bidirectional and unidi-
methods in literature that have been used to predict crisis. The rectional from one to other and vice versa). So, the first step is
definition of crisis, models utilized, and explanatory variables the estimation of empirical relationships between exchange
have varied from one study to another, (Sevim et al. [50]). To rate and GDP, inflation rate and unemployment rate. The
deal with currency crises prediction, three types of research are second step is macroeconomic pressure indices determination
known in literature. The earliest category refers to the regres- and the third is the estimation of their ability to alert or to give
sion models such as Logit–Probit models estimating crises crisis event signal.
ahead of time via leading indicators. This category is hold by
researchers as [19, 23, 27, 10, 49, 15, 44]. The second category 3.1. Macroeconomic Determinants
uses potential early warning indicators and is associated with
the Kaminsky et al. [33] (KLR) Model, which is also known as This study focuses on three main macroeconomic variables,
the signaling approach. The third category focuses on machine GDP, inflation and unemployment. Literature review shows that
learning applications, which are relatively new in forecasting there is possible bidirectional relationship between exchange rate
financial crises. It is a popular predictive tool used by [57, 36, and the study‘s macroeconomic variables. There may be simul-
21, 24]. taneity in terms of influence that is one of the sources of en-
dogeneity among unobserved heterogeneity, measurement errors,
endogenous sample and serial autocorrelation. To deal with
3. Methodological Aspects endogeneity (reverse causality), we use a dynamic framework,
performing the Blundell and Bond [7] system GMM model and
Efforts to anticipate currency crises systemically have cre-
compare it to pooled and fixed effect models to ensure its ro-
ated a monitoring instrument known as the Early Warning
bustness. This model is designed for situation with few time
System, Sutrisno et al. [53], among others. According to Shi
periods and many individual units, linear functional relationship,
and Gao [51], in the mainstream models for the financial crisis
endogenous variable depending on its own past realizations,
early warning, the KLR has better performance. A large mass
independent variables that are not strictly exogenous (correlated
of research constructed early warning system for currency
with errors), unobserved heterogeneity and heteroskedasticity
crises. In this line Kaminsky et al. [33] (KLR) introduced the
and autocorrelation within individual units‘ errors but not accross
signal approach. They have proposed watching the evolution
them. The model is specified as:
of some macroeconomic variables that experience unusual

(1)

Where the variable Exchge is exchange rate, the endoge- interest rate (DInterest) and terms of trade (Trade).
nous variable. There are two groups of independent variables.
The first one represents the focuses macroeconomic variables 3.2. Currency Crisis
chosen for signal ability analysis, GDP growth (GDP), infla-
tion rate (Inf) and unemployment (Uemp). The second are This study uses the Frankel-Rose definition of a currency
control variables, international reserves (Reserv), broad crisis. It is based on two main conditions. The first condition
money (Money), lending interest rate (LInterest), deposit identifies crisis when a country‘s nominal bilateral dollar

24
Journal of Business and Economic Development http://www.sciencepg.com/journal/jbed

exchange rate depreciates by 25 % or more in a given quarter the depreciation in the preceding quarter. The two inclusive
compared to the previous year‘s value. The second condition two conditions lead to a binary variable computed as follow:
sets that the depreciation has to be at least 10 % greater than

( )
{ ( ) ( ) ( ) (2)

The double condition was set so that reoccurring currency of financial pressure index, the macroeconomic and market
crashes due to the first condition could be avoided. We need to pressure indexes in this study, is exceeded. We called this
make sure that the crisis variable captures only sudden crashes threshold, the pressure index value at risk (PIVaR) beyond
of currency. Furthermore, we will be using a crisis window of which the corresponding indicator gives a crisis warning as
four quarters which is called a combined crisis period. This follows:
simplifies the model by avoiding repeated crises by treating
clusters of recurring crises as a particular event. { (8)

3.3. Early Warning System Where, is a factor that takes values between 1 and 3 in
In general, the features of the currency crisis include ex- financial crisis literature. According to Sevim et al. [50], it is a
change rate depreciation due to the decrease in international It is a country-specific and heuristic value that seeks to im-
reserves mainly. The signal approach commonly uses Market prove the signal performance. The factor is usually an
Pressure Index (MPI) built by Kaminsky et al. [33] (KLR) arbitrary coefficient. In this study we estimate the optimal
from the two variables, exchange rate and international re- factor that minimizes forecast errors following Kaminsky et al.
serves. This study follows KLR method by building Financial [33]. They determine the value ok that is the minimum and
Pressure Indexes (FPI). The first FPI is the MPI, second the the best threshold (Shi and Gao, [51]) for individual indicators
macroeconomic pressure indexes specifically Production through noise-to-signal ratio (NSR) from the summarized
Pressure Index (PPI), Inflation Pressure Index (IPI) and matrix of their performances as follows:
Unemployment Pressure Index (UPI) and the combine Mac-
roeconomic Pressure Index (MePI).
Table 1. Indicator signal matrix.

(3) Crisis in signaling No crisis in signal-

window ing window

(4)
Signal issued A B
No signal issued C D
(5)

(6) In this matrix, A is the number of quarters where the indi-

cator issued a good signal, B is the number of quarters where
the indicator issued a bad signal (type II error), C is the
number of quarters where the indicator failed to issue a good
signal (type I error), and D is the number of quarters where the
(7) indicator refrained from issuing a bad signal. The NSR of
each indicator is calculated following Goldstein et al. [28].
The are standard deviations and denote percentage According to Phadan and Prabheesh [43], they determine the
NSR from the unconditional probability of a crisis P(crisis) =
changes in study‘s variables (x). The symbol in equations (A+C) / (A+B+C+D) and the conditional probability of a
(4) to (7) indicate that this study doesn‘t consider the a priori crisis P(crisis | Signal) = A / (A+B) as follows:
theoretical relationship‘s direction between exchange rate and
the leading macroeconomic indicators. Instead, the study ⁄( )
( | ) ( ) (9)
based on the empirical relationship from model (1). Accord- ⁄( )
ing to Sevim et al [50], financial crisis arises when a threshold

25
Journal of Business and Economic Development http://www.sciencepg.com/journal/jbed

From Shi and Gao [51], an effective early warning indicator and World Bank open online database from 1990 to 2020. In
has a NSR > 0 and the indicator that NSR is larger than 1 fact, two types of data sets have been used during this work.
cannot act as leading indicators. Yearly data are used to estimate model (1) and quarterly data
in all other models. The variables collected are Exchange Rate
3.4. Warning Period Window (Dollar per Local Currency), Gross Domestic Product (Per
Capita), Consumer Price Index, Unemployment (Measured in
When talking about forecasting or prediction, the time percentages), Real lending rate, Real deposit rate, Terms of
window has a significant effect on the quality of the predic- trade, Broad Money, International reserves (in US dollars).
tion and in our case the quality of the signal. There is The study‘s countries are emerging ones specifically Brazil,
therefore also interest in the study to assess the quality of the Chile, China, Colombia, Czech Republic, Hungary, Republic
indicators over four quarters. From equation (8) the perfect of Korea, Mexico, Poland, South Africa, Thailand.
signal takes a value of 1 if a crisis is expected to occur within
the upcoming window period and a value of 0 if otherwise.
For instance, if a crisis is assumed in January of 2022, the 4. Findings and Interpretations
perfect signal indicator always takes a value of 1 before
12-months period of January 2022 for the chosen warning 4.1. Stationarity
period window.
To ensure models stability the study performs stationarity
tests and results table shows that only GDP growth and un-
3.5. Data Description
employment rate are stationary at first difference I(1). The
The data collection in this research study is derived from other variables are I(0).
secondary sources, the International Monetary Fund (IMF)

Table 2. Stationarity.

First level First difference

Variables Conclusion
LLC IPS LLC IPS

Exchge -6,599*** -1,684** S -9,438*** -6,128*** S I(0)

(0.00) (0.04) (0.00) (0.00)
GDP 1,213 -3,604*** NS -1,488* -6,859*** S I(1)
(0.88) (0.00) (0.06) (0.00)
Inf -6,777*** -6?277*** S -9,726*** -9,009*** S I(0)
(0.00) (0.00) (0.00) (0.00)
Uemp -1,924** 0,915 NS -3,049*** -4,347*** S I(1)
(0.02) (0.81) (0.00) (0.00)
Reserv -3,777*** -5,885*** S -17,467*** -9,224*** S I(0)
(0.00) (0.00) (0.00) (0.00)
Money -3,495*** -5,671*** S -8,610*** -8,581*** S I(0)
(0.00) (0.00) (0.00) (0.00)
Trade -6,488*** -3,856*** S -8,919*** -7,084*** S I(0)
(0.00) (0.00) (0.00) (0.00)
Linterest -6,191*** -3,491*** S -5,209*** -5,853*** S I(0)
(0.00) (0.00) (0.00) (0.00)
Dinterest -6,103*** -4,186*** S -5,350*** -6,247*** S I(0)

26
Journal of Business and Economic Development http://www.sciencepg.com/journal/jbed

First level First difference

Variables Conclusion
LLC IPS LLC IPS

(0.00) (0.00) (0.00) (0.00)

Notes: S determines Stationarity and NS is non stationarity

nomic variables and the exchange rate one by one respectively.

4.2. Granger Causality Granger Non-Causality is an econometric test used to verify
the usefulness of one variable to forecast another. Granger
In the study, this step is important so that one could have a causality is not an absolute one. It is based on 2 principles; the
preliminary point of view of where the data set is standing cause happens prior to its effect and the cause has unique
compared to what history or literature says. Thus, the study information about the future values of its effect.
will conduct a Granger Non-Causality test on the macroeco-

Table 3. Granger causality test of exchange rate on macroeconomic variables.

Variables p-values Conclusions

GDP 0.09615 Exchange rate does not Granger cause GDP

Inflation 0.2999 Exchange rate does not Granger cause Inflation
Unemployment 2.351e-09 Exchange rate Granger cause Unemployment

Table 4. Granger causality test of macroeconomic variables on exchange rate.

Variables p-values Conclusions

GDP 0.2601 GDP Granger cause Exchange rate

Inflation 0.2999 Inflation does not Granger cause Exchange rate
Unemployment 1.531e-14 Unemployment Granger cause Exchange rate

From tables 3 and 4, it appears that there is bidirectional and the lower shows diagnostic statistics. Starting by models‘
causality between exchange rate and unemployment, a uni- diagnostic, it appears that the determination coefficients of
directional causality from GDP to exchange and no causality pooled-OLS and fixed effect models are low. The models
relationship between exchange rate and inflation in emerging barely explain a quarter of variations in exchange rate. GMM
countries. The analysis needed deeper techniques and bidi- regression is two-step and the Windmeijer [55] finite sample
rectional case involves taking into account endogeneity bias. correction for standard errors is employed. Instrument matrix
is collapsed in SYS-GMM regression. For first and second
4.3. Exchange Rate Macroeconomic order autocorrelation, the null hypothesis of over-identifying
restrictions validity and instruments validity necessary for
Determinants
system GMM robustness, respectively the study reports the
Model (1) is estimated three times with different methods p-values of AR(1), AR(2), Sargan and Hansen tests. The
(pooled OLS, fixed effect and system GMM) to deal with results of the Arellano-Bond tests indicate that there is no
endogeneity and to ensure results robustness. Table 5 gives second-order serial correlation. The null hypotheses of Sar-
results divided in two parts where the upper gives coefficients gan's test and Hansen's J test cannot be rejected. It therefore

27
Journal of Business and Economic Development http://www.sciencepg.com/journal/jbed

appears that the test statistics show an appropriate specifica-

POOLED-OL FIXED EF-
tion. SYS-GMM gives robust results and determines the ex- SYS-GMM
S FECTS
change rate and macroeconomic variables relationships in
emerging markets. In the robust estimations (i.e. the system Constant 10,607*** -9,732*** -14,349***
GMM) the lags (1 & 2) of exchange rate and the focuses
(0.00) (0.00) (0.00)
macroeconomic variables are significant. Lags 1 & 2 ex-
change rate variable have opposite effects on its current value, 0.25 0.23
respectively significantly positive at 1% level and signifi- AR (1) -1,87*
cantly negative at 10% level. The differential effect from the
AR (2) -0,84
two lags remains significantly positive. The exchange rate
tends to potentially follows its recent momentum in emerging Sargan 2,94
economies. Hansen 3,22

Table 5. Exchange rate and macroeconomic variables relationships. Notes: (*), (**) and (***) indicate significant level at 10%, 5% and 1%
respectively.
POOLED-OL FIXED EF-
SYS-GMM
S FECTS
However, some country structural forces could enhance or
Exchge(-1) 0,353***
inhibit this trend. It appears that GDP and unemployment are
inhibition forces as they have negative relationship with ex-
(0.00) change rate respectively (-0.018) and (-0.074) at 5% and 10%.
Exchge(-2) -0,266* GDP has the expected literature negative sign, but unemploy-
ment has the unexpected negative sign. It implies that with
(0.06)
growth or growing unemployment rate, study‘s emerging
GDP -0,003** -0,004** -0,018** countries‘ currency grows in value. If GDP or unemployment
(0.01) (0.01) (0.02) increases by 100%, exchange rate drops by 2% and 7% re-
spectively. Inflation has the expected sign, which if increases
Inf 0,006*** 0,006*** 0,053**
by one the exchange rate will go up by 0.053, defining a posi-
(0.00) (0.00) (0.05) tive relationship. GDP, inflation and unemployment are ex-
Uemp 0,010* 0,010* -0,074*
change rate determinants with the expected sign for GDP and
inflation but not for unemployment. A decrease in GDP growth
(0.06) (0.07) (0.08) or in unemployment rate and an increase in inflation demon-
Reserv -0,099*** -0,09*** -0,098 strate the depreciation of exchange rate of national currency. As
seen in financial pressure indexes equations (3) to (7), in-
(0.00) (0.00) (0.13)
creased index will raise pressure of national currency to be sold.
Money -0,0002 -0,0002 -0,0009
(0.50) (0.52) (0.11) 4.4. Early Warning System
Trade -0,0004 -0,0004 -0,001 Table 5 gives empirical relationship between exchange rate
(0.15) (0.36) (0.49) and macroeconomic variables in emerging markets that al-
lows determining the financial pressure indexes and their
Linterest 0,0002 0,001 0,00001
ability to alert on currency crises through the signal-to-noise
(0.73) (0.59) (0.99) ratio (lower than 1) and the optimal factor ( ).
Dinterest 0,005*** 0,0009 -0,003 From table 5 the symbol in the financial pressure indexes
formulas is determine negative for PPI and UPI, positive for
(0.00) (0.79) (0.66)
IPI. The study started building the EWS model looking for the
Year dummy 0,005*** 0,004*** 0,007*** optimal threshold for each indicator given in table 6.
(0.00) (0.00) (0.00)

28
Journal of Business and Economic Development http://www.sciencepg.com/journal/jbed

Table 6. Optimal threshold factor.

Indicators Financial pressure indexes Optimal threshold

International reserves MPI 1.7

GDP PPI 1.7
Inflation IPI 1.2
Unemployment UPI 2.6
GDP-Inflation-Unemployment MePI 1.3

In literature, the critical value of FMP index that leads to mation on indicators for each country. The column FPI shows
the crisis has been mainly calculated by 1-3 standard devia- the indicators and that with asterisk (*) are the helpful ones in
tions. In this study the factor is in line with literature as predicting currency crises. It appears that none of the mac-
. From equation (2) and (8) the study deter- roeconomic and market indicators is significant for Chile,
mines effective currency crises on one hand and the financial Colombia, Cezch Republic and Hungary. For the other coun-
and macroeconomic indexes crisis warnings on the other hand. tries, on average three indicators out of five are useful in
This allows determining confusion matrices for each country signaling currency crises.
and indexes performances, given in table 7. It gives infor-

Table 7. Performances of indicators under signal approach.

BRAZIL
FPI A B C D ( | ) ( ) ( )
MPI 3 9 21 49 63.41 0.125 0.155 1.24 0.25 0.29 -0.04
PPI 1 4 23 54 67.07 0.041 0.068 1.65 0.2 0.29 -0.09
IPI 3 9 21 49 63.41 0.125 0.155 1.24 0.25 0.29 -0.04
UPI* 3 4 21 54 69.51 0.125 0.068 0.55 0.43 0.29 0.14
MePI 3 9 21 54 65.52 0.125 0.155 1.14 0.25 0.29 -0.04
CHILE
FPI A B C D ( | ) ( ) ( )
MPI 3 10 35 34 45.12 0.07 0.22 2.87 0.23 0.46 -0.23
PPI 0 1 38 43 52.44 0 0.02 - 0 0.46 -0.46
IPI 0 0 38 44 53.66 0 0 - - 0.46 -0.46
UPI 3 4 35 40 52.44 0.07 0.09 1.15 0.42 0.46 -0.04
MePI 2 5 36 39 50.00 0.05 0.11 2.15 0.28 0.46 -0.18
CHINA
FPI A B C D ( | ) ( ) ( )
MPI* 7 8 22 45 63.41 0.24 0.15 0.62 0.46 0.35 0.11
PPI* 4 4 25 49 64.63 0.13 0.07 0.54 0.5 0.35 0.15
IPI* 7 8 22 45 63.41 0.24 0.15 0.62 0.46 0.35 0.11
UPI* 4 6 25 47 62.20 0.13 0.11 0.82 0.4 0.35 0.05

29
Journal of Business and Economic Development http://www.sciencepg.com/journal/jbed

MePI* 4 6 25 47 62.20 0.13 0.11 0.82 0.4 0.35 0.05

COLOMBIA
FPI A B C D ( | ) ( ) ( )
MPI 2 8 27 45 57.32 0.06 0.15 2.18 0.2 0.35 -0.15
PPI 0 3 29 50 60.98 0 0.05 - 0 0.35 -0.35
IPI 2 8 27 45 57.32 0.06 0.15 2.18 0.2 0.35 -0.15
UPI 3 12 26 41 53.66 0.10 0.22 2.18 0.2 0.35 -0.15
MePI 3 12 26 41 53.66 0.10 0.22 2.18 0.2 0.35 -0.15
CEZCH REPUBLIC
FPI A B C D ( | ) ( ) ( )
MPI 4 5 36 37 50.00 0.1 0.11 1.19 0.44 0.48 -0.04
PPI 0 1 40 41 50.00 0 0.02 - 0 0.48 -0.48
IPI 4 5 36 37 50.00 0.1 0.11 1.19 0.44 0.48 -0.04
UPI 3 7 37 35 46.34 0.07 0.16 2.22 0.33 0.48 -0.15
MePI 3 7 37 35 46.34 0.07 0.16 2.22 0.33 0.48 -0.15
HUNGARY
FPI A B C D ( | ) ( ) ( )
MPI 3 5 28 46 59.76 0.09 0.10 1.01 0.375 0.378 -0.003
PPI 0 1 31 50 60.98 0 0.02 - 0 0.378 -0.378
IPI 3 5 28 46 59.76 0.09 0.10 1.01 0.375 0.378 -0.003
UPI 2 12 29 39 50.00 0.06 0.23 3.64 0.14 0.378 -0.238
MePI 2 12 29 39 50.00 0.06 0.23 3.64 0.14 0.378 -0.238
KOREA REPUBLIC
FPI A B C D ( | ) ( ) ( )
MPI* 5 4 27 46 62.20 0.15 0.08 0.52 0.55 0.39 0.16
PPI 1 2 31 48 59.76 0.03 0.04 1.28 0.33 0.39 -0.06
IPI* 5 4 27 46 62.20 0.15 0.08 0.51 0.55 0.39 0.16
UPI* 7 7 25 43 60.98 0.21 0.14 0.64 0.5 0.39 0.11
MePI* 7 7 25 39 58.97 0.21 0.14 0.69 0.5 0.39 0.11
MEXICO
FPI A B C D ( | ) ( ) ( )
MPI 3 8 30 41 53.66 0.09 0.16 1.79 0.27 0.40 -0.13
PPI* 6 7 27 42 58.54 0.18 0.14 0.78 0.46 0.40 0.06
IPI 3 8 30 41 53.66 0.09 0.16 1.79 0.27 0.40 -0.13
UPI* 4 5 29 44 58.54 0.12 0.10 0.84 0.44 0.40 0.04
MePI* 4 5 29 44 58.54 0.12 0.10 0.84 0.44 0.40 0.04
POLAND
FPI A B C D ( | ) ( ) ( )

30
Journal of Business and Economic Development http://www.sciencepg.com/journal/jbed

MPI* 5 4 23 50 67.07 0.17 0.07 0.41 0.55 0.34 0.21

PPI 2 7 26 47 59.76 0.07 0.12 1.81 0.22 0.34 -0.12
IPI* 5 4 23 50 67.07 0.17 0.07 0.41 0.55 0.34 0.21
UPI* 4 7 24 47 62.20 0.14 0.12 0.90 0.36 0.34 0.02
MePI* 4 7 24 47 62.20 0.14 0.12 0.90 0.36 0.34 0.02
SOUTH AFRICA
FPI A B C D ( | ) ( ) ( )
MPI* 5 6 28 43 58.54 0.15 0.12 0.80 0.45 0.40 0.05
PPI* 2 2 31 47 59.76 0.06 0.04 0.67 0.5 0.40 0.10
IPI* 5 6 28 43 58.54 0.15 0.12 0.80 0.45 0.40 0.05
UPI 3 8 30 41 53.66 0.09 0.16 1.79 0.27 0.40 -0.13
MePI 3 8 30 41 53.66 0.09 0.16 1.79 0.27 0.40 -0.13
THAILAND
FPI A B C D ( | ) ( ) ( )
MPI 4 10 28 40 53.66 0.12 0.20 1.6 0.28 0.39 -0.11
PPI* 1 1 31 49 60.98 0.03 0.02 0.64 0.50 0.39 0.11
IPI 4 10 28 40 53.66 0.12 0.20 1.6 0.28 0.39 -0.11
UPI* 4 4 28 46 60.98 0.12 0.08 0.64 0.50 0.39 0.11
MePI* 4 4 28 46 60.98 0.12 0.08 0.64 0.50 0.39 0.11

Notes: * indicates valid financial pressure index a Percentage of observations correctly called under signal approach [(A+D)/(A+B+C+D). b
Good signals as percentage of possible good signals; A/(A + C). c Bad signals as percentage of possible bad signals; B/(B + D). d Adjusted
noise signal ratio; [B/(B + D)]/[A/(A + C)]. e Percentage of signals that were followed by at least one crisis within the subsequent window (4
quarters) = A/ (A + B). f The unconditional probability of a crisis—that is, (A + C)/(A + B + C + D).

The column ―correct‖ indicates the percentage of observa- tional probability should be higher than the unconditional one.
tions correctly called under the signal approach. The per- The variables with a positive sign in the last column are the
centages seem relatively acceptable. However, it appears that effective indicators that are classified in table 8. All the in-
for each country there is a threshold percentage between 58% dicators are effective for China and production pressure index
and 60% from which the indicator becomes efficient. The is better than literature market pressure index that is as better
good signals as the percentage of possible good signals are as inflation pressure index. In Korea Republic and Poland,
ranging from 0 percent to 24 percent. This is another measure only PPI is not effective and the best indicator is IPI. South
of the propensity of the indicators to issue good signals unlike Africa, Thailand and Mexico experienced three effective
the bad signals percentage one would like to be small. NSR is indicators where IPI seems to be the best.
the adjusted noise-to-signal ratio, which measures the noisi- It appears that, when effective the macroeconomic indica-
ness of indicators. The lower the NSR, the better the signal. It tors mainly inflation and GDP are as better as or better than
is the main criterion in crises prediction probability. A finan- market pressure index based on international reserves. This is
cial pressure index with a NSR equal or greater than 1 intro- opposed to Budsayaplakorn et al. [8] study‘s results which
duces too much noise and therefore cannot be an effective instead used twice the warning period window of this study.
indicator. The last three columns show the probability of a Unemployment based indicator is the one that appears effec-
crisis conditional on a signal from the indicator, the uncondi- tive more times than the others, followed by the combined
tional probability of a crisis and their difference, respectively. macroeconomic index, MePI. The GDP indicator is the most
The difference between the probability of a crisis conditional accurate one but unemployment indicator is the most sensitive.
on a signal and the unconditional probability measures the GDP is the main macroeconomic variable with the most ac-
noisiness of indicators. If the indicator is useful, the condi- curate data available. Unemployment variable is that experi-

31
Journal of Business and Economic Development http://www.sciencepg.com/journal/jbed

encing in model (1) results, unexpected or opposite theoretical crisis in emerging countries.
sign, indicating greatest unusual behavior before currency

Table 8. Effective indicators by country.

MPI PPI IPI UPI MePI

China 2nd 1st 2nd 3rd 3rd

Korea Rep. 2nd Not effective 1st 3rd 4th
Poland 1st Not effective 1st 3rd 3rd
South Africa 2nd 1st 2nd Not effective Not effective
st st
Thailand Not effective 1 Not effective 1 1st
Mexico Not effective 1st Not effective 2nd 2nd
Brazil Not effective Not effective Not effective 1st Not effective
Chile Not effective Not effective Not effective Not effective Not effective
Colombia Not effective Not effective Not effective Not effective Not effective
Czech Rep. Not effective Not effective Not effective Not effective Not effective
Hungary Not effective Not effective Not effective Not effective Not effective

but unemployment pressure index is the most sensitive.

However, the number of effective indicators and the accu-
5. Conclusion racy of the indexes are not the same from a country to others.
The study can be improve estimating indicators relationships
This study has examined the probability of currency crisis
with exchange rate by country and considering different
based on macroeconomic variables, GDP, inflation and un-
warning period window from 3 months to 24 months.
employment from the KLR signal approach method in
emerging countries. The study compared its created macroe-
conomic pressure indexes with the literature‘s market pres- Abbreviations
sure index based on international reserves. Instead of using
theoretical relationships to determine financial pressure in- AR: Auto Regressive
dexes, the study based on macroeconomic variables‘ empiri- BCBS: Basel Committee on Banking Supervision
cal signs with exchange rate. BIS: Bank for International Settlement
The results indicated from system GMM method that GDP, EEC: European Economic Community
inflation and unemployment have significant relationships EWS: Early Warning System
with exchange rate. GDP and inflation coefficients have the FMP: Financial Market Index
expected negative and positive signs respectively contrary to FPI: Financial Pressure Index
unemployment having the unexpected negative sign in GDP: Gross Domestic Product
emerging countries. The study has determined market pres- GMM: Generalized Methods of Moments
sure index, production pressure index, inflation pressure in- IMF: International Monetary Fund
dex, unemployment pressure index and a combine macroe- IPI: Inflation Pressure Index
conomic index. It appears that they are effective early warning IPS: Im-Pesaran-Shin
indicators if they correctly called at least 60 percent of ob- KLR: Kaminsky Lizondo Reinhart
servations providing them low noise-to-signal ratio (lower LLC: Levin-Lin-Chu
than 1). The macroeconomic pressure indexes are better early MPI: Market Pressure Index
warning indicators than market pressure index in emerging MePI: Macroeconomic Pressure Index
countries for one-year warning period window. Production NSR: Noise-to-Signal Ratio
pressure index appears more accurate followed by inflation OLS: Ordinary Least Square

32
Journal of Business and Economic Development http://www.sciencepg.com/journal/jbed

PIVAR: Pressure Index Value at Risqk [10] Canbas S., Cabuk A. and Kilic S. B. (2005). Prediction of
PPI: Production Pressure Index commercial bank failure via multivariate statistical analysis of
SYS-GMM: System Generalized Methods of Moments financial structures: The Turkish case. European Journal of
Operational Research, 166(2), 528-546.
UPI: Unemployment Pressure Index
[11] Chang, R., and A. Velasco. 1999. ―Liquidity Crises in
Emerging Markets: Theory and Policy.‖ National Bureau of
Author Contributions Economic Research Working Paper no. 7272, Cambridge, MA.

Siriki Coulibaly: Conceptualization, Formal Analysis, [12] Cerra, V., Saxena, S. C. 2008, Growth dynamics: the myth of
Supervision, Validation, Methodology, Visualization, Pro- economic recovery, American Economic Review, 98 (1),
ject administration, Writing - review & editing 439e457.
Aya Marouane: Data curation, Software, Writing - orig- [13] Corsetti, G., P. Pesenti, and N. Roubini, ―What Caused the
inal draft, Methodology, Visualization Asian Currency and Financial Crisis?‖ NBER Working Paper,
National Bureau of Economic Research, Massachusetts, 1998.

Conflicts of Interest [14] Davidson, Sinclair. ―The 1997–98 Asian Crisis: A Property
Rights Perspective,‖ Cato Journal, l. 25, no. 3, Fall 2005, pp.
The authors declare no conflicts of interest. 567-582.

[15] Davis, E. P., & Karim, D. (2008). Comparing early warning

References systems for banking crises. Journal of Financial stability, 4(2),
89-120.
[1] Abdoh, W. M. Y. M., Yusuf, N. H. M., Zulkifli, S. A. M., Bulot,
[16] Dooley, M. P. (2000). A Model of Crises in Emerging Markets.
N., & Ibrahim, N. J. (2016). Macroeconomic factors that influence
Economic Journal, 256-272.
exchange rate fluctuation in ASEAN countries. International
Academic Research Journal of Social Science, 2(1), 89-94. [17] Du, J., Yu, R., & Lai, K. K. (2020). Identification and predic-
tion of currency crisis: Markov Switching-Based approach.
[2] Abor, Joshua. (2005). Managing foreign exchange risk among
The Singapore Economic Review, 65(06), 1667-1698.
Ghanaian firms. The Journal of Risk Finance, 6(4), 306–318.
[18] Edison H. J. (2003). Do indicators of financial crises work ? An
[3] Achsani, N. A. (2010). The Relationship Between Inflation
evaluation of an early warning system. International Journal of
And Real Exchange Rate: Comparative Study Between
Finance and Economics, 8(1), 11-53.
ASEAN, The EU And North. European Journal Of Economics,
Finance And Administrative Sci., 18: 69- 76. [19] Eichengreen, B., Rose, A. K., Wyplosz, C., Dumas, B., &
Weber, A. (1995). Exchange Market Mayhem: The Anteced-
[4] Bartram, S. M., & Karolyi, G. A. (2006). The impact of the
ents and Aftermath of.
introduction of the Euro on foreign exchange rate risk expo-
sures. Journal of empirical finance, 13(4-5), 519-549. [20] Feridun, M., 2009. Determinants of exchange market pressure
in Turkey: An econometrics investigation. Journal of emerging
[5] Bank for International Settlements (BIS). (2005), Amendment
markets Finance and Trade, 45(2), pp. 65-81.
to the Capital Accord to Incorporate Market Risks. Retrieved
from http://www.bis.org/publ/bcbs119.htm [21] Fethi M. D. and Pasiouras F. (2010). Assessing bank efficiency
and performance with operational research and artificial intel-
[6] Basel Committee on Banking Supervision (BCBS). (2006),
ligence techniques: A survey. European Journal of Opera-
Basel II: International Convergence of Capital Measurement and
tional Research, 204(2), 189-198.
Capital Standards: A Revised Framework – Comprehensive
Version. Available at http://www.bis.org/publ/bcbs128.htm [22] Ford, J. L., Santoso, B. and Horsewood, N. J., 2007. Asian
Currency Crises: Do fundamentals still matter? A Mar-
[7] Blundell R. and Bond S. (1998). Initial conditions and moment
kov-Swiching approach to cause and timing. Department of
restrictions in dynamic panel data models. Journal of Econo-
Economics Discussion Paper, No. 07-07, University of Bir-
metrics, 87(1), 115-143.
mingham.
[8] Budsayaplakorn, S., Dibooglu, S., & Mathur, I. (2010). Can
[23] Frankel J. A. and Rose A. K. (1996). Currency crashes in
macroeconomic indicators predict a currency crisis? Evidence
emerging markets: An emprirical treatment. Journal of Inter-
from selected Southeast Asian countries. Emerging Markets
national Economics, 41(3-4), 351-366.
Finance and Trade, 46(6), 5-21.
[24] Garcia-Alonso C. R., Torres-Jiménez M. and Hervás-Martínez
[9] Bussière, M., Fratzscher, M. 2006, Towards a New Early
C. (2010). Income prediction in the agrarian sector using
Warning System of Financial Crises, Journal of International
product unit neural networks. European Journal of Opera-
Money and Finance, 25 (6), 953- 973.
tional Research, 204(2), 355-365.

33
Journal of Business and Economic Development http://www.sciencepg.com/journal/jbed

[25] Gerdesmeier, D., Reimers, H.-E., Roffia, B. 2009, Asset Price 28(8), 1317-1330.
Misalignments and the Role of Money and Credit, Working
Paper 1068. European Central Bank. [40] Mirchandani, A. (2013), «Analysis of macroeconomic
determinants of exchange rate volatility in India», Interna-
[26] Ghosh, M. A. R., Ostry, M. J. D., & Tsangarides, M. C. G. tional Journal of Economic and Financial Issues, 3 (1),
(2011). Exchange rate regimes and the stability of the interna- 172-179.
tional monetary system. International Monetary Fund.
[41] Moccero, D. A. (2006). Real Exchange Rate Volatility And
[27] Glick, R., & Rose, A. K. (1999). Contagion and trade: why are Exports: Argentine Perspectives. 4 Th Annual Conference Of
currency crises regional?. Journal of international Money and The Euro-Latin Study Network On Integration And Trade.
Finance, 18(4), 603-617. Paris: 20- 21 October.

[28] Goldstein, M., Kamisnky G. L., and Reinhart C., (2000). As- [42] Mohammed, T., Retia, M., Gaidi, K., & Boudeghdegh, A.
sessing Financial Vulnerability: An Early Warning System for (2017, July). The Impact of Exchange Rate Regimes on Eco-
Emerging Markets. Institute for International Economics, nomic Growth. In International Conference on Applied Eco-
Washington DC. nomics (pp. 415-427). Springer, Cham.

[29] Gupta, P., Mishra, D., Sahay, R. 2007, Behavior of Output [43] Padhan, R. and Prabheesh, K. P., (2019). Effectiveness of
during Currency Crises, Journal of International Economics, Early Warning Models: A Critical Review and New Agenda
72, 428-450. for Future Directions. Bulletin of Monetary Economics and
Banking, 22(4), 457–484.
[30] Harberger, A. (2004). Economic Adjustment and the Real
Exchange Rate, in S. Edwards and L. Ahamed (eds.), ―Eco- [44] Premachandra I. M., Bhabra G. S., Sueyoshi T. (2009). DEA as
nomic Adjustment Exchange Rates in Developing Countries‖, a tool for bankruptcy assessment: A comparative study with
University of Chicago Press, 10, 308-321. logistic regression technique. European Journal of Opera-
tional Research, 193(2), 412-424.
[31] Husain, A. M., Mody, A., Rogoff, K. S., (2004), ―Exchange
Rate Regime Durability and Performance in Developing Ver- [45] Radelet, S., and J. Sachs. 1988. ―The Onset of the East Asian
sus Advanced Economies‖, Journal of Monetary Economics, Financial Crisis.‖ National Bureau of Economic Research
52(1), 35-64. Working Paper no. 6680, Cambridge, MA.

[32] Ilzetzki, E., Reinhart, C. M., & Rogoff, K. S. (2021), « Re- [46] Ramasamy, R. (2015). Influence Of Macroeconomic Variables
thinking Exchange Rate Regimes » (No. w29347). National On Exchange Rate. Journal Of Economic Business And
Bureau of Economic Research. Management, 277.

[33] Kaminsky, G., Lizondo, S., Reinhart, C., 1998. Leading Indi- [47] Rose, A. K. (2011). Exchange rate regimes in the modern era:
cators of Currency Crises. International Monetary Fund Staff fixed, floating, and flaky. Journal of Economic Literature,
Papers 45, 1-48. 49(3), 652-72.

[34] Kamin, S. A. (2003). A Multi-Country Comparison Of The [48] Roubini, K. S. (2000). Exchange Rate Anomalies In The Indus-
Linkages Between Inflation And Exchange Rate Competi- trial Countries. Journal Of Monitory Economics,, 560-587.
tiveness. International J. Finance And Economics, 8(2):
167-184. [49] Sanchis A., Segovia M. J., Gil J. A., Heras A. and Vilar J. L.
(2007). Rough sets and the role of the monetary policy in fi-
[35] Kashif, S. (2000). The Relationship Between Exchange Rate nancial stability (macroeconomic problem and the prediction
And Inflation. International Academic Research Journal of of insolvency in insurance sector (microeconomic problem).
Social Science 2(1) 2016 Page 89-94. European Journal of Operational Research, 181(3),
1554-1573.
[36] Kumar, P. R., & Ravi, V. (2007). Bankruptcy prediction in
banks and firms via statistical and intelligent techniques–A [50] Sevim, C., Oztekin, A., Bali, O., Gumus, S., & Guresen, E.
review. European journal of operational research, 180(1), (2014). Developing an early warning system to predict cur-
1-28. rency crises. European Journal of Operational Research,
237(3), 1095-1104.
[37] Lahiri, A. and Vegh, C. A., 2000. Delaying the inevitable:
Optimal interest rate policy and BOP crises. unpublish manu- [51] Shi, J., & Gao, Y. (2010). A study on KLR financial crisis
script, Birbeck College. early-warning model. Frontiers of Economics in China, 5(2),
254-275.
[38] Marshall, D. 1998. ―Understanding the Asian Crisis: Systemic
Risk as Coordination Failure.‖Economic Perspectives (Third [52] Simon W. L. S. (1997), ―Is There Life Outside the ERM? An
Quarter): 13–28. Evaluation of the Effects of Sterling's Devaluation on the UK
Economy‖, International Journal of Finance and Economics, 2,
[39] Melvin, M., Taylor, M. P. 2009, The Crisis in the Foreign 199-216.
Exchange Market, Journal of International Money and Finance,

34
Journal of Business and Economic Development http://www.sciencepg.com/journal/jbed

[53] Sutrisno, H., Wulansari, D., & Handoyo, R. D. (2022). Mac- [56] Yazdani, M., & Nikzad, M. (2021). Output Losses from Cur-
roeconomic Indicators as A Signal of The Currency Crisis in rency Crises and the Role of Central Bank. Journal of Central
The Indonesian Economy. International Journal of Economics Banking Theory and Practice, 3, 79-97.
& Management, 16(1).
[57] Zhang G., Hu M., Patuwo B. E. and Indro D. C. (1999). Arti-
[54] Swami, O. S., Pandey, S. K., & Pancholy, P. (2016). Val- ficial neural networks in bankruptcy prediction: General
ue-at-risk estimation of foreign exchange rate risk in India. framework and cross-validation analysis. European Journal of
Asia-Pacific Journal of Management Research and Innovation, Operational Research, 116(1), 16-32.
12(1), 1-10.

[55] Windmeijer, F. (2005) A Finite Sample Correction for the

Variance of Linear Efficient Two-Step GMM Estimators.
Journal of Econometrics, 126, 25-51.
http://dx.doi.org/10.1016/j.jeconom.2004.02.005

35