A Study on the Early
Warning Indicators of Currency Crisis :
A Regional Perspective
In Soo Kang*
*
Professor, Department of Economics, Sookmyung Women’s University, Korea
|
This research is supported
by the Korea Research Foundation. This paper mainly focuses on the signals
approach to Korean currency crisis and compare the results with those of OLS and logit
estimation. Chronological crisis episode of Korea and related regional data
are used in the research. Both results tell us that indicators and
determinants of currency crisis should consider regional characteristics and
country specific results, rather than general results, are more useful to
policymakers and market participants. Since this is a first
draft, comments on this proposal would be really appreciated. If you have any
constructive comments on this proposal, please send me e-mail to the
following address. E-mail address : iskang@sookmyung.ac.kr |
<Contents>
Ⅰ. Background
and Objective
Ⅱ. Currency
crisis of Korea
III. Signals approach
IV. Regression analysis
on Korean currency crisis
V. Concluding Remarks
Ⅰ. Background and Objective
After the Asian currency crisis
of 1997 the distressed countries had to pay more than 10% of GDP to recover the
crisis. If the crisis could be detected in advance, the adjustment costs would
have been much lower. There has been a strong need for the improved warning
system from both academics and policymakers. As summarized in Kaminsky, Lizondo
and Reinhart(1997) and Goldstein(1998),
there are massive studies on leading indicators of currency crisis.
Theoretically, the most
representative explanations for the currency crisis can be summarized as
‘fundamental approach’ vs. ‘self-fulfilling expectations approach’. The former
approach is adopted by Krugman(1979) and Flood and Garber(1984): under a fixed
exchange rate, domestic credit expansion in excess of money demand growth leads
to a gradual loss of international reserves and eventually to a speculative
attack on the currency. The macroeconomic policies which are not consistent
with the fixed exchange rate system, and the weak structure of the economy are
the essential part of this approach. International reserves, domestic credit,
fiscal imbalances, real exchange rate, trade or current account balance, real
wages, and domestic interest rates could be used as leading indicators of
crises.
The latter approach is adopted by
Obsfeld(1994, 1996). It argues that crises may occur without any noticeable
change in economic fundamentals. The contingent nature of economic policies may
give rise to multiple equilibria and generate self-fulfilling crises. Mutual
interactions(circularity) between economic policies and expectations of
economic agents play an important role in this approach. A variety of
factors(such as, output level, interest rate differential, stock price of
banking sector, proportion of non-performing loans, political instability, etc.)
which may affect the objective function of government could be used as leading
indicators of the crises. Contagion effects are also considered as a main cause
of the crises in some papers(Gerlach and Smets(1994)).
Empirically, except a few
papers(such as, Cumby and Van Wijenbergen(1989) and Park and Choi(1999)), most
of the papers investigate the experience of various countries and try to
generalize the usefulness of the selected indicators. However, these types of
generalization have obvious limits in predicting the currency crises of a
specific country like Korea because substantial parts of the regional
characteristics including structural and institutional factors are neglected.
Regional differences seem to be equally important in explaining the currency
crises.
One of the most commonly used
methods in empirical works on currency crises is to compare the behavior of a
variable during the pre-crisis and tranquil period for the same group of
countries(Frankel and Rose(1996)). Even though this method is useful in finding
systematic differences between the two periods, the scope of the potential
indicators are mostly confined to major macroeconomic variables and the
usefulness of indicators in predicting a currency crisis is quite limited.
Other group of empirical works on currency crises(Blanco and Garber(1986))
estimate the probability of devaluation in advance by using the logit(or
probit) estimation. Since the dependent variable of this method is qualitative
one(1 for crisis and 0 for non-crisis), imposing a specific threshold value is
critical to the result. However, the determination of threshold value is
somewhat arbitrary and it is still very hard to predict the exact starting time
of the crisis. Also, it does not provide the relative importance of indicators.
Recently, Eichengreen, Rose, and
Wyplosz(1995), Kaminsky and Reinhart(1996), Kaminsky, Lizondo and
Reinhart(1997), Goldstein and Reinhart(1998), Goldstein(2000) use a
nonparametric approach to evaluate the usefulness of several variables in
signaling an impending crisis. Unlike the previous approach, the signal
approach can tell the relative importance of individual indicators. This
approach uses an ‘index of exchange market pressure’ which is usually defined
as a weighted average of percentage changes in the exchange rate and percentage
changes in the gross international reserves and percentage changes of domestic
interest rates. Periods in which the index is above its mean by more than three
standard deviations are defined as crises in these papers. More than 100
indicators are considered in previous literatures and about 15 indicators
turned out to perform well as signals for crises in Goldstein(2000).
This approach is useful for the
analysis in aggregate level (such as, Asian crisis), however, as mentioned in
Park and Choi(1999) we could have different results for a specific country even
though the same methodology is applied. This implies that in addition to the
general results country specific investigation is necessary to improve the
predictability of indicators. As noted in Goldstein(1998), the poor
performances of existing signals might be due to the lack of credibility of
data on borrowers and the problems caused by the bail-out of official sector
for borrowers. These problems could be minimized if we use the country specific
reliable data. Also, since currency crises are likely to be preceded by
multiple economic problems, we need to consider a variety of indicators for an
effective warning system.
In this research country specific
indicators will be investigated by using the signal approach. We focus on Korea
and the regional characteristics of the crises and credibility of the data.
Ⅱ. Currency crisis of Korea
There has been a general
consensus regarding the fundamental causes of Asian crises in late 1990s. In
Latin American countries macroeconomic policy failures, such as excessive
fiscal deficits, excessive money supply, and strict exchange rate control, are
regarded as the main causes of currency crises. However, Asian crises stemmed
from a kind of bank-run in which the depositors are foreigners. Government-led
banking polices resulted in moral hazard of firms and the over-investment of
firms through excessive borrowing worsened the financial structure of banks. The
currency crises happened in the process of withdrawal by foreign investors who
found out the situation going too far. However, one could ask why it happened
at that time because such structural problems are not new at all during last
few decades. In order to find out the critical factors of the crisis which
affected the credibility of foreign investors, we need to pay attention to the
regional characteristics of individual country along with the structural
problems.
The currency crisis of
Korea began from the immature responses of both government and private sector
to the currency crises of Southeast Asian countries which started from Thailand
in July 1997. In ex post sense the
government guarantee of foreign debts of private banking sector and
maltreatment of Kia Motor company
made a credit crisis of private sector into a currency crisis of country. In
the middle of October it became worse by spending the reserves(dollars) to protect the exchange rate of
Korean won. Even after IMF loan in
December 1997, the conflicts with IMF during negotiation process and the
attempt to issue new bonds by Korean Development Bank(KDB) gave very bad
impression to foreign investors.
As shown in <Table
1> most of the macroeconomic indicators of Korea in 1990s are stable. During
1990-1996, GDP growth rates are 5.1%-8.9%(7.1% in 1996), inflation rates(CPI)
are 4.5%-9.3%(4.9% in 1996), interest rates are 11.6%-16.2%(11.9% in 1996), M2
growth rates are 15.5%-18.6%(16.2% in 1996), fiscal deficit/GDP ratios are
–0.7%-0.3%(0.3% in 1996), and reserves are 17.1
billion$-33.7billion$(33.2billion$ in 1996).These indicators tell us that the
causes of currency crisis are different from those of Latin American countries.
However, the external sector, especially current account balance/GDP ratio,
becomes worse in this period. The current account deficit keeps increasing and
current account balance/GDP reaches –4.9% in 1996. Since the capital inflow
makes overall balance positive, the government does not pay much attention to
the deterioration of current account.
<Table 1>
Macroeconomic Indicators of Korea
(unit:%, billion$)
80-85
86-91
92
93
94
95 96
GDP
growth rate
6.3 9.9 5.1 5.8 8.6 8.9 7.1
Inflation
rate
10.9
6.1
6.3
4.8
6.2 4.5 4.9
3
year yields
19.0
15.1 16.2 12.6 12.9 13.8
11.9
M2 growth
rate
20.6
18.8 18.4 18.6 15.6 15.5
16.2
Fiscal
balance/GDP -2.5 -0.2 -0.7 0.3 0.5 0.4 0.3
Current
Ac/GDP
-3.8
3.0
-1.5
0.1 -1.2 -2.0 -4.9
Reserves
7.1
12.2 17.1 20.3 25.7 32.7
33.2
Source : Bank of Korea
•Current
account and exchange rate
One
could argue that appreciation of Korean won which might cause the current
account deficit is the main element of currency crisis. Real appreciation of
domestic currency just before currency crises and consequent current account
deficits are frequently observed in many cases. However, there are some
evidences that the increase in current account deficit of Korea in this period
is temporary phenomenon rather than structural problem. Especially, the sharp
increase of current account deficit in 1996 is mainly due to the deterioration
of terms of trade.[1] Also, the
increase of current account deficit could be partly explained by the increase
in investment since saving rate is quite stable(35.0%-37.4%) during this
period. Depending on the base year the degree of appreciation of exchange rate
will be different. Korean won(won/US$) has been substantially depreciated
before the currency crisis[2],
however, one may argue that Korean won has been appreciated during the same
period because Japanese yen has been depreciated even more against US dollar.
Korean won has been appreciated against Japanese yen during this period,
however, it does not seem to be a serious cause of currency crisis.[3]
•Foreign
debts
Asian
growth model contains a moral hazard problem from the beginning because
government-led resource allocations result in inefficient over-investment of
private sector despite rapid economic growth. Since most of banking sector
loans are implicitly guaranteed by government, after the partial liberalization
of capital market in 1990s domestic banking sector can borrow money from
foreign creditors without much difficulties. However, with failures of
over-investment and consequent deterioration of banking sector foreign
creditors begin to suspect the credibility of borrowers. With sudden massive
withdraws of foreign creditors the currency crisis has started. The official
foreign debts of Korea reaches its maximum(170.6 billion US$) in September
1997. However, the actual foreign debts turned out to be much more than the
official announcement.[4]
This gap plays a critical role in downgrading credibility of Korean government
in December 1997. During this period not only total debts but also the portion
of short-term debt had sharply increased. In 1996 foreign debt/GNP and portion
of short-term debt reached 22% and 58.2%, respectively. These indicators might
be taken as signals of currency crisis, however, not many people including
foreign creditors took these as serious signals. There are several reasons for
this. First, increase in foreign debt might be a temporary phenomenon which is
due to the deterioration of terms of trade. Second, increase in portion of
short-term debt does not necessarily mean worsening of term structure of
interest rate because interest rate is lower for the shorter loan in
international financial market. Third, since most of the short-term foreign
debts are delivered by banks in Korea, the increase in the portion of
short-term debts does not seem to be a problem due to long-term relationships
with foreign banks.
•Policy failures:
The Korean
government did not fully liberalize the domestic financial market even after
joining OECD. She did not fully open bond market and short-term financial
commodities, instead she allowed domestic banks to borrow from abroad. It was
expected that these measures would bring indirect effects of capital market
liberalization. However, the domestic
interest rate could not be used as a buffer for international capital flows
when credibility was downgraded. Korean government took sterilization policy
when she intervened in FX market in 1997 to avoid consecutive bankruptcies. You
may argue that Korean government should have abandoned this policy to avoid
currency crisis, however, incomplete
liberalization of financial market would have prevented the success of
contractionary monetary policy and high interest rate policy.
•Chronology
of Korean currency crisis
Development
and main events of Korean crisis can be summarized as follows.
January-February
1997: Strikes by labor union against labor reform legislation and dishonor of
Hanbo(14th ranked chaebol whose debts amount 6 billion dollars) .
Credit rates of 3 major banks(Korea Exchange Bank, Korea First Bank, Chohung Bank) are lowered.
However, credit rate of Korea does not change because rating agencies(Moody’s,
S&P) distinguish credit status of individual bank and country.
March-June
1997: Dishonor of Sammi and Jinro lowered the credit rate of Korea First Bank.
Corruption scandal of Hanbo imprisoned the son of ex-president. However, credit
rate of Korea remains at the same level and the situation becomes stabilized in
May and June, which leads the government to underestimate urgent structural
adjustments.
July-September
1997: Financial reports on the performances of banking sector disappointed
foreign investors. Currency crisis of Thailand and Kia Motor’s request for dishonor prevention in July lowered the
credit rate of Korea. Government announced purveyance guarantee for the debts
of banking sector in August, which becomes a momentum of magnifying the credit
crisis of private sector into that of country. However, not many people including
Moody’s are sure about the currency crisis.
October-November
1997: Government decided to make Kia Motor a public company and right after
this decision stock market of Hong Kong crashed. These two events caused
massive capital outflow from Korea and foreign creditors denied renewal of
short-term loans. As a consequence, demand for dollar sharply increased in FX
market and the government spent a lot of foreign reserves(dollar) to prevent
the depreciation of Korean won.[5]
December
1997: Even after the emergency loan from IMF(55 billion dollars in early
December) the credibility of Korea worsened due to the unrevealed short-term
foreign debts and disharmony with IMF after the president election. Also, the
attempt of government to issue additional bonds(2 billion dollars) through
Korean Development Bank after the signing with IMF made the situation even
worse.
III. Signals approach
•Definition of currency crisis
A crisis is defined as a situation in which an attack on the currency leads to a sharp depreciation of the currency, a large decline in international reserves, or a combination of the two (Goldstein(2000) p.19).
•Index
of currency crisis
An
index of currency market turbulence is made as a weighted average of exchange
rate changes and reserve changes.
where De/e= percentage changes in the exchange rate
(defined as units of domestic currency per U.S. dollar)
DR/R= percentage
changes in international reserves
s = standard deviation of each variable
The
ratio of standard deviations are chosen as weights so that each component of
the index has the same conditional variance. In Kaminsky, Lizondo and
Reinhart(1997) and Goldstein(2000), periods in which the index(INDEX) is above its mean by more than
three standard deviations are defined as crises. Sensitivity analysis is
required for this definition because high threshold value reduces the
possibility of false signal but increases the possibility of no signal when
there is a crisis. Also, extremely high threshold value is inappropriate for
the analysis of a country with few crisis experiences like Korea. <Figure
3> shows the trend of the index . If we define crises as periods in which
the index is above its mean by more than 1.5 standard deviation, there are only
two crises out of 156 periods. To avoid this limited crises episode, we define
crises as periods in which the index is above its mean by more than 1.1
standard deviation.
•Indicators
Most
of indicators which are considered in the previous researches are summarized in
Kaminsky, Lizondo and Reinhart(1997). The selection of indicators is very
important in signals approach, however, the data availability and credibility
limit the range of selection, especially in international comparison. Even if
each indicator has economic linkage with currency crises, the theoretical
justifications for the linkage cannot be strong in general. Selection of
indicator starts from subjective judgement and considers as many indicators as
possible, then tests the predictive power of each indicator. Most of the
previous research including Goldstein(2000) focus on the international
comparison and find general indicators. However, these types of general results
in which regional characteristics are not reflected cannot be applied to an
individual country.
•Signaling
window
An
indicator sends a signal if it departs from normal behavior. If an indicator
sends a signal that is followed by a crisis within a plausible time frame we
call it a good signal. If the signal is not followed by a crisis within that
interval, we call it a false signal(or noise). The signaling window for
currency crises is set at 24 months preceding the crisis as in Goldstein(2000).
•Threshold
In
order to test the hypotheses H0
(The economy is in a state of tranquility) versus H1 (A crisis will occur sometimes in the next 24 months)
on an indicator-by-indicator basis, we need to select a threshold (or critical value)
that divides the probability distribution of that indicator into normal region
and rejection region as in any hypothesis test. If the observed value of an
indicator falls into the rejection region, that indicator is said to be sending
a signal. To select the optimal threshold for each indicator, we allow the size
of the rejection region to oscillate between 1 percent and 20 percent. For each
choice, the noise-to-signal ratio is tabulated and the optimal set of
thresholds which minimize the noise-to signal ratio is chosen.
•Signals,
noise and crises probabilities
The
possible outcomes as summarized by the following 2x2 matrix where A,B,C,D
represent the number of observations in each case.
<Table
2> Possible outcome matrix
Crisis occurs in the No crisis occurs in the
following 24 months following 24 months
Signal
A
B
No Signal
C
D
The
unconditional and conditional probability of crisis can be calculated as
follows;
P(C) = (A+C)/(A+B+C+D) : unconditional
probability of crisis
P(C｜S) = A/(A+B) : conditional
probability of crisis
If
the indicator is not noisy, there are relatively few entries in cell B and
P(C｜S)≒1. Hence, if an indicator sends a signal
and that indicator has a reliable track record, we can expect that the
probability of a crisis, conditional on a signal, P(C｜S), is greater
than the unconditional probability, P(C). The difference between the two
probabilities (P(C｜S) - P(C))
is used as a criteria of good indicator, i.e. the bigger this difference is,
the better the predictive power of an indicator is.
The
noise-to signal ratio is defined as
N/S = [B/(B+D)]/[A/(A+C)]
where
A/(A+C) is a proportion of signal when there is a crisis and B/(B+D)
is a proportion of signal when there is no signal.
•Empirical
results
<Table
3> summarizes the performance of monthly indicators in Korean currency
crisis. The variables are shown in descending order based on their marginal
predictive power. For each indicator, the first column of the table shows the
noise-to signal ratio(N/S). The second column shows the probability of a
crisis conditional on a signal from the indicator( P(C｜S)) and the
third column lists the difference between the conditional and unconditional
probabilities(P(C｜S) - P(C)).
The fourth column shows the threshold values of each indicator and the last
column shows the ranking that the indicator received in Goldstein(2000). The
indicators’ rankings are based on their marginal predictive power, P(C｜S) - P(C). The
better the indicator, the higher the probability of crisis conditioned on its
signaling.
As shown in last column
the ranking there is much difference between the ranking of this paper and that
of Goldstein(2000). In Goldstein(2000) and Kaminsky(1988) real exchange rate,
stock prices, M2 multiplier, output, exports, real interest rate are the major
predictive indicators. However, in this paper export concentration ratio,
change in S&P rating, foreign debt of banking sector/reserves, M2
multiplier, industry inventory index/output index turn out to be reliable
indicators in Korean crisis case. This difference results from the sample
size(Goldstein(2000) uses 25 country sample) and data availability. From this
result we can verify that for a practical purpose indicators should be selected
country by country. The better performances of indicators with multi-country
sample do not necessarily mean the better performance for an individual
country.
•Lead
time of signals
The
ranking of indicator represents the ability of anticipating crises while
producing few false alarms. Such criteria, however, do not speak to the lead
time of the signal. To the policymakers or financial market participants, not
only the well performances of indicators but also lead time of the signals are
important. From the time series of indicators we can calculate the average
number of months from when the first signal is issued to the crisis month, i.e.
lead time of the signal. In case of Korean currency crisis, 15 indicators whose
noise-to-signal ratios(N/S) are less than 1 send the first signal
anywhere between 12 months and 18 months before the crisis erupts. The average
lead time for these early signals is 14 months in Korea, hence all these
indicators are considered as leading rather than coincident, which is
consistent with the spirit of an early warning system.
<Table 3> Marginal
predictive power of indicators
Indicator
N/S
P(C｜S) P(C｜S) - P(C) Threshold Rank
•Export
concentration ratio 0.067 92.9 59.1 91
-
•Change
in S&P rating
0.093
90.3
53.6 20
-
•Foreign
debt of banking
0.12
88.2
53.5 89 -
sector/reserves
•M2
multiplier@
0.126 85.7 42.7 95 3
•Industry
inventory index/
0.163
81.8
39.5 92 -
output index@
•Price
of service sector/
0.174 87.5 32.5 94 -
price of manufacturing
sector
•Reserves@
0.238 75.0 33.3 3
12
•Operation
rate of
0.468 65.2 29.7 15 -
manufacturing sector
•Terms
of trade@
0.482
60.2
27.2 14 13
•Capital
account balance/GDP 0.547 61.9 26.6
14 -
•M2/reserves
0.429
66.7
20.5 96 9
•Fiscal
deficit/GDP
0.456 66.7 19.0
2
-
•Real
effective exchange rate@ 0.476 60.0 18.3 93 1
•Real
interest rate differential 0.857 50.0 3.8 99 7
•Bankruptcy
rate
0.878
50.0 3.2 99 -
______________________________________________________________________
•Stock
prices@
1.142
38.4 4.7
18 2
•Industrial
production index@ 1.317 35.7 1.2
20
4
•Exports@
1.064
40.7 6.4 19 5
•Bank
deposits@
1.899
28.6
-14.6 5 8
•Excess
real M1 balances 1.041 45.
8.6 80 10
•Imports@
1.512
33. 0.3
87
15
•Current account/GDP
2.92
23.3
-13.6 20 -
Note:
1. 12-month growth rates are used for the indicators with @ mark.
2. The numbers in columns, P(C｜S), P(C｜S) - P(C), Threshold are
percentiles(%).
•Vulnerability
and signals
Kaminsky(1998)
shows how to construct a composite index to gauge the probability of a crisis
conditional on multiple signals from various indicators. The composite index is
made so that the more reliable indicators receive a higher weight. In weighting
individual indicators, we eliminate the indicators from the list if the
noise-to-signal ratio(N/S) of the indicator is bigger than 1. Seven
indicators (stock prices, industrial production index, exports, bank deposits,
excess real M1 balances, imports, and current account/GDP) are eliminated from
<Table 3>. All these indicators are considered to be good indicators and
included in composite index in Goldstein(2000). For the remaining 15 indicators
with noise-to-signal ratios below unity, we weighted the signals by the inverse
of the noise-to-signal ratios. Formally, we construct the following composite
indicator.
WINDEXt = ∑n j=1
Sjt /ωj
In
this equation, n represents the number of indicators considered. Each
indicator has a differentiated ability to forecast crises and this ability can
be summarized by the noise-to signal ratio, here denoted by ωj . Sjt is a dummy variable that
is equal to one if the univariate indicator, Sj crosses its
critical threshold and is thus signaling a crisis and is zero otherwise. As
before, noise-to-signal ratio is calculated under the assumption that an
indicator issues a correct signal if a crisis occurs within the following 24
months. All other signals are considered false alarms.
<Figure 4> shows
the trend of composite index in Korea. It reaches its maximum, 38.8, in July
1997. There are only three times when the score of composite index is over 30,
July 1997, November 1997, and August 1997. With the time series scores of
composite index, we can choose a critical value for the composite indicator so
that when the composite indicator crosses this threshold, a crisis is deemed to
be imminent. In Korean case the critical value of composite indicator turns out
to be 19, which minimizes the noise-to-signal ratio.[6]
There are only 16 months when the composite indicator crosses this critical
value, and half of these signals alarm in 1997.
IV. Regression analysis on Korean currency crisis
In this section the relations between the index of currency market turbulence and the indicators are investigated. Causes(or determinants) of crisis can be found through the general regression or logit(or probit) analysis(Frankel and Rose(1996), etc.). We analyze the determinants in both ways.
· Linear Regression analysis:
where
CRt = currency crisis
index(INDEX) at time t, and xit = ith indicator at time t.
Among
22 indicators, seven indicators show statistical significance in the
regression. Unlike the results of Park and Choi(1999), the real effective
exchange rate does not have a statistical significance.
All indicators in
<Table 4> have expected signs and are statistically significant except
real effective exchange rate in the first equation.[7]
In Sachs, Tornell, and Velasco(1996) which considers 10 variables of 20
developing countries, only real exchange rate, domestic credit/GDP, M2/Reserves
have statistically significant explanatory power on exchange rate market
pressure(currency crisis index). However, <Table 4> tells us that not
only financial sector variables(M2/reserves and credit rating) but also real
sector variables have significant explanatory power on Korean currency crisis.
The regression results from pooling data have general implications on currency
crisis, but it cannot be directly applied to an individual country’s experience.
Regression results do not provide indicators, which is more important to
policymakers and market participants. However, the results confirm the
statistical importance of some indicators as a determinant.
<Table 4> Determinants
of Korean currency crisis
Indicators
Equation (1)
Equation (2)
Coefficients t-statistic Coefficients t-statistic
Constant
0.237
(0.63)
0.242 (0.64)
Change
in S&P rating
-3.016***
(-5.81) -2.917*** (-5.63)
Terms of trade@ -0.099** (-1.98) -0.106** (-2.14)
Capital
account balance/GDP -0.976*** (-12.79) -0.979*** (-12.74)
Bankruptcy
rate
5.296***
(3.69) 4.792*** (3.41)
Current
account/GDP
-1.499***
(-6.67) -1.618*** (-7.59)
M2/Reserves
0.055***
(3.26) 0.049*** (2.98)
Stock prices@ -0.019*** (-2.65) -0.013** (-2.11)
Real
effective exchange rate@ 0.051 (1.60)
Adjusted R2
0.764
0.762
Note:
1. 12-month growth rates are used for the indicators with @ mark.
2. **, *** denote
statistical significance at 5%, 1%, respectively.
· Logit analysis
Instead
of using quantitative index, a qualitative analysis is also possible through the
following method. Since the currency crisis index is composed of exchange rate
and reserves, we can mitigate the direct link between the index and indicators
by using logit model.
where
Crisis
index(CR) is assumed to be a function
of indicators(x) and Pr(yt=1) means the probability
of occurrence of currency crisis. In general the results are sensitive to the
threshold value(q), however,
in Korean currency crisis case there are not many crisis episode even if we
lower the threshold. As we did in signaling analysis, threshold of currency
crisis index is set at 1.1 standard deviation from the mean. The results of
logit estimation are reported in <Table 5>.
<Table 5> Logit
estimation on Korean currency crisis
Indicators
Equation (3)
Equation (4)
Coefficients t-statistic coefficient t-statistic
Constant
-12.32* (-1.92) -12.27** (-2.23)
Terms
of trade@
-0.514*
(-1.73)
-0.504* (-1.82)
Capital
account balance/GDP
-2.162**
(-1.98)
-2.164**
(-2.12)
M2/reserves 0.385* (1.63) 0.375** (1.97)
Stock
prices@
-0.046**
(-1.99) -0.041* (-1.89)
Real
effective exchange rate@ 0.101 (0.82)
McFadden R2
0.756 0.733
Note:
1. 12-month growth rates are used for the indicators with @ mark.
2. *, ** denote
statistical significance at 10%, 5%, respectively.
In
logit estimation statistical significance of each indicator has been decreased,
however, as in OLS estimation the real effective exchange rate does not have an
explanatory power.
<Figure
5>~<Figure
10> show the trends of major variables which are used in this analysis.
V. Concluding Remarks
As mentioned in
Goldstein(1998), the regression approaches can tell the statistical
significance of each variable, however, these cannot identify threshold values
for monitoring purposes. Also, due to the difficulties in forecasting
explanatory variables, the accurate forecasting of crisis is very hard in
regression analysis. To the contrary, the signal approach could forecast crises
by using currently available indicator data. However, the robustness of the
results needs to be checked by both approaches.
In case of Korea both
financial and real variables, such as terms of trade, capital account
balance/GDP, M2/reserve, stock prices turn out to be good indicators in signal
approach as well as determinants of currency crisis in regression analysis.
Recently, the relation with banking
crises becomes more important in recovery period after currency crises in East
Asian countries. Weak banking sector which might be due to institutional
characteristics, such as policy-directed lending, ownership structure of the
banking system, weak accounting, high levels of connected lending, incentive
incompatible official safety nets is likely to cause currency crises. Special
attention needs to be paid to these kinds of institutional or regional
characteristics because currency crises tend to follow banking crises. In the
country like Korea, portion of non-commercial banking sector(secondary banking
sector) in total loan or exhaustion rate of credit line(or overdraft) by
corporative sector are used as the most direct indicator of crises by the
policy makers.
<References>
Blanco,
Hermino, and Peter M. Garber, 1986. “Recurrent Devaluation and Speculative
Attacks on the Mexican Peso,” Journal of
Political Economy 94 (February): 148-166.
Cerra,
Valerie and Sweta C. Saxena, 2000, “Contagion, Monsoons, and Domestic Turmoil
in Indonesia: A Case Study in the Asian Currency Crisis,” IMF Working Paper 00/60.
Cumby,
Robert, and Sweder van Wijnbergen, 1989. “Financial Policy and Speculative Runs
with a Crawling Peg: Argentina 1979-81,” Journal
of International Economics 17 (August):111-127.
Eichengreen,
Barry, Andrew Rose, and Charles Wyplosz, 1995. “Exchange Market Mayhem: The
Antecedents and Aftermath of Speculative Attacks,” Economic Policy 21 (October): 249-312.
Eichengreen,
Barry, Andrew Rose, and Charles Wyplosz, 1996. “Contagious Currency Crises,” NBER Working Paper No.5681.
Frankel,
Jeffrey, and Andrew K. Rose, 1996. “Exchange Rate Crises in Emerging Markets,” Journal of International Economics 41,
No.3/4 (November): 351-368.
Fratzscher,
Marcel, 2000. “On Currency Crises and Contagion,” Institute for International Economics Working Paper No.00-9.
Gerlach, Stefan, and Frank Smets, 1995. “Contagious
Speculative Attacks,” European Journal of
Political Economy, Vol.11: 5-63.
Goldstein,
Morris, 1998a. “Early Warning Indicators and the Asian Financial Crisis,” Memio, Washington: Institute for
International Economics.
Goldstein,
Morris, 1998b. The Asian Financial
Crisis: Causes, Cures, and Systematic Implications, Policy Analyses in International
Economics 55, Washington:
Institute for International Economics.
Goldstein,
Morris, Graciela Kaminsky, and Carmen M. Reinhart, 2000. Assessing Financial Vulnerability: An Early Warning System for Emerging
Markets, Washington: Institute for International Economics.
Kaminsky,
Graciela, 1998. “Currency and Banking Crises: The Early Warnings of Distress,” International Finance Discussion Papers No.629,
Washington: Board of Governors of the Federal Reserve.
Kaminsky,
Graciela, Saul Lizondo, and Carmen M. Reinhart, 1997. “Leading Indicators of
Currenvy Crises,” IMF Working Paper 97/79.
Kaminsky,
Graciela, Saul Lizondo, and Carmen M. Reinhart, 1999. “The Twin Crises: The
Causes of Banking and Balance-of-Payments Problem, American Economic Review 89, No.3: 473-500.
Kaminsky,
Graciela, and Carmen M. Reinhart, 1998. “Financial Crises in Asia and Latin
America: Then and Now,” American Economic
Review 88, No.2 (May): 444-448.
Kaminsky,
Graciela, and Carmen M. Reinhart, 1999. “The Twin Crises: The Causes of Banking
and Balance-of-Payments Problems,” American
Economic Review 89, No.3 (June): 473-500.
Krugman,
Paul, 1979. “A Model of Balance of Payments Crises,” Journal of Money, Credit, and Banking 11(August): 311-325.
Obstfeld,
Maurice, 1994. “The Logic of Currency Crises,” NBER Working Paper No.4640.
Obstfeld,,
Maurice and Kenneth Rogoff, 1995. “The Mirage of Fixed Echange Rates,” Journal of Economic Perspectives 9
(Fall): 73-96.
Park, Wonam, and Gongpil Choi, 1999. “Empirical
Analysis on Causes of Korean Currency Crises,” International Economics Research(in Korean), Vol.5, No.3: 1-26.
Park,
Daegeun and Changryong Rhee, 1998. “Currency Crisis of Korea : Development and
Lesson,” Memio, Seoul National
University.