You are on page 1of 19

Journal of International Money and Finance 26 (2007) 265e283

www.elsevier.com/locate/jimf

An empirical assessment of currency devaluation


in East Asian countries
Yoonbai Kim a,b, Yung-Hsiang Ying c,*
a
Department of Economics, University of Kentucky, Lexington, KY 40506, USA
b
Department of Economics, National University of Singapore, Singapore 117570
c
The Institute of Interdisciplinary Studies, College of Social Science, National Sun Yat-Sen University,
Kaohsiung 804, Taiwan

Abstract

The hypothesis of contractionary devaluation has received surprisingly strong empirical support,
especially in the context of Latin American countries. In this paper, we study whether it applies equally
well in seven East Asian countries. When we use the pre-1997 crisis data and the trade-weighted exchange
rate, we find no evidence of contractionary devaluations. In fact, currency devaluation appears strongly
expansionary in several countries. This is contrasted to the case of Chile and Mexico where the evidence
of devaluation is persistent. We find that the results are somewhat sensitive to the definition of the
exchange rate and the period of estimation.
Ó 2006 Elsevier Ltd. All rights reserved.

JEL classification: F31; F33

Keywords: Devaluation; Contractionary effects; East Asia

1. Introduction

Is devaluation contractionary on output? A typical textbook presents a model in which


devaluation of the domestic currency is unambiguously expansionary. The MarshalleLerner
elasticities’ condition for stability is considered to be met and thus the trade balance improves

* Corresponding author. Tel.: þ886 7 5252000x5515; fax: þ886 7 5255511.


E-mail address: yying@mail.nsysu.edu.tw (Y.-H. Ying).

0261-5606/$ - see front matter Ó 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jimonfin.2006.11.004
266 Y. Kim, Y.-H. Ying / Journal of International Money and Finance 26 (2007) 265e283

with devaluation. In the context of the AD-AS model, currency devaluation increases aggregate
demand by increasing net exports. The potentially adverse supply-side effects are either ignored
or assumed to be minor. (See Frankel, 1988; Goldstein and Khan, 1985 for more discussion and
empirical evidence.)
The possibility of contractionary devaluation has been studied in seminal papers by
Diaz-Alejandro (1963), Cooper (1971), and Krugman and Taylor (1978). A contractionary
devaluation may occur from the aggregate demand side: the trade balance may worsen if the
price elasticities of export and import demands are too low or if the initial trade balance is
in deep deficit. In the monetary model, devaluation reduces aggregate demand by raising the
domestic price level through higher prices of imported goods and thereby lowering the real
money balance (Frenkel and Johnson, 1976). Another channel is a redistribution of income
from low saving groups (wages) to high saving groups (profits) (Krugman and Taylor, 1978).
Devaluation can work through the supply-side channels as well. It raises the price of
imported intermediate goods and, under indexation, wages, resulting in an upward shift in
the aggregate supply. See, inter alia, Findlay and Rodriguez (1977), Sachs (1980), Marston
(1982), and Buffie (1989). Relative to the ambiguous demand-side effects, the supply-side
effects are more clear-cut and have been treated as the main channel in which devaluation
can be contractionary. (See Lizondo and Montiel, 1989, for a comprehensive model.)
With liberalization of financial markets, additional channels have emerged that may make
devaluations more likely to be contractionary, especially in developing countries. When domes-
tic financial and nonfinancial firms have liabilities in foreign currency, currency devaluation
will increase debt-servicing obligation and, similar to a negative supply shock, generate stag-
flationary effects (Gylfason and Risager, 1984). The deterioration of the balance sheet induced
by increases in the domestic currency value of foreign debt has been cited as one of the main
causes of the ‘‘twin’’ crises leading to the unprecedented recessions in East Asia. (See Krug-
man, 1999; Corsetti et al., 1999; Schneider and Tornell, 2000 for more discussion of the
‘‘balance sheet effects.’’) In developing countries, devaluation e often preceded by speculative
attacks e may trigger a loss of access to capital markets, especially if it is seen as breaking
a policy commitment. The resulting serious interruption of external financing constitutes
another supply-side shock (Reinhart, 2000).
Despite theoretical ambiguity and the prevailing assumption, empirical studies find that
devaluations tend to be contractionary in surprisingly many case studies beginning with
Edwards (1989) and Morley (1992). A review of recent econometric studies by Kamin and
Rogers (2000) and Upadhyaya et al. (2000) indicates that devaluations almost uniformly lead
to reduced output and there is little evidence for subsequent reversal. These empirical results
were mostly obtained using Mexican and other Latin American data. This can be contrasted
with relatively more positive views about the effects of devaluations in industrial countries. Ex-
amples are the devaluations in the United Kingdom after the exchange rate mechanism (ERM)
crisis in 1992 where growth accelerated with little inflationary consequence.1 Similarly, deval-
uations by the nations that broke away from the gold standard in the 1930s are considered to

1
Gordon (2000) shows that the experience of the United Kingdom was exceptional. A comparison of ERM leavers
(Finland, Italy, Portugal, Spain, Sweden, and the United Kingdom) and the stayers (Austria, Belgium, France,
Netherlands, and Switzerland) reveals that there was no excess growth of real GDP and inflation was far from being non-
existent in the leavers other than the UK. He points out that some country-specific factors such as fiscal constraints due to
Maastricht Treaty and an unusually flexible labor market may be responsible for such ‘‘free-lunch’’ like outcomes in the
UK.
Y. Kim, Y.-H. Ying / Journal of International Money and Finance 26 (2007) 265e283 267

have helped those economies to recover from severe recessions (Obstfeld and Rogoff, 1995;
Eichengreen and Sachs, 1985). Although exclusive studies for East Asian countries are rela-
tively scarce, it is commonly believed that devaluation against the yen boosts exports and is
generally expansionary (Kwan, 1994). Thus, one can be led to suspect that contractionary
devaluations are mainly a Latin American phenomenon.2 A panel study by Kamin and Klau
(1998), however, shows that contractionary devaluation applies equally to developed economies
and Asian countries as well as Latin American countries. Upadhyaya and Upadhyaya (1999)
find that with few exceptions a devaluation in Asian countries fails to make any effect on output
over any length of time e short run, intermediate run, or long run.
In this paper, we investigate the effects of currency devaluations in seven East Asian
countries e Indonesia, Korea, Malaysia, Philippines, Singapore, Taiwan, and Thailand e and
compare them with Mexico and Chile.3 We employ a fairly comprehensive vector auto-
regressive (VAR) model in which the endogeneity of the exchange rate is fully taken into
consideration. When we use the pre-1997 crisis data and the trade-weighted exchange rate,
we find no evidence of contractionary devaluations. In fact, currency devaluation appears
strongly expansionary in several countries. This is contrasted to the case of Chile and Mexico
where the evidence of contractionary devaluation is persistent.
The plan of the paper is as follows: Section 2 provides preliminary data analysis using cross
correlation and Granger causality analyses. Section 3 presents a VAR model for a small open
economy. Estimation results are reported in Section 3. The paper concludes in Section 4.

2. Preliminary data analysis

In this study, we use quarterly data for seven countries in East Asia: Indonesia, Korea, Ma-
laysia, Philippines, Singapore, Taiwan, and Thailand. For comparison, two Latin American
countries are chosen: Chile and Mexico. The sample period varies for each country, longest
for Malaysia (1970:1e2000:1) and shortest for Indonesia (1981:1e1999:2). Appendix A lists
the details on the source of the data and the sample period. The choice of countries has
been dictated by the availability of relevant data. Many countries in Latin America had to be
excluded since necessary quarterly data tend to be short due to the history of high inflation
in those countries.
We use, as the benchmark case, the trade-weighted effective exchange rate. While using the
effective exchange rate might be reasonable in most cases, choosing the weights for currencies
in the basket can be tricky. Weights based on exports, total trade, invoice currency, and debt
contracts can be very different from each other. Most East Asian countries appear to have fairly
diversified trade patterns by the 1990s, with the US, Japan and Western Europe taking roughly

2
In marked contrast to that of Latin American countries, the exchange rate policy in East Asia has been evaluated
positively at least until the onset of the financial crisis in 1997. Some of the main features would include (1) with
the emphasis on the competitiveness of the trade sector, the real exchange rate has been well maintained without severe
real exchange rate misalignments (World Bank, 1993); (2) thanks to moderate inflation, the nominal exchange rate has
not been used as a nominal anchor (Dornbusch and Park, 1999). Moreover, large devaluations have been relatively rare
and considered an important policy tool in times of economic downturn.
3
Devaluation may be a misnomer as the term usually applies to exchange rate changes under a fixed exchange rate
regime whereas most countries in this study maintained some sort of managed floating for substantial parts of the sam-
ple period. We use devaluation and depreciation interchangeably. In so doing, we treat symmetrically large discrete
changes in the exchange rate under pegged exchange rate systems and small daily movements under managed floats.
268 Y. Kim, Y.-H. Ying / Journal of International Money and Finance 26 (2007) 265e283

equal shares in trade in several countries. Another important characteristic of the trade patterns
of the East Asian countries is the share of trade with regional neighbors. For instance, in coun-
tries such as Korea, Malaysia and Singapore, the share of trade with East Asian neighbors is
almost as high as the sum of all the industrial countries combined. The trade patterns of the
two Latin American countries are noticeably different. In Mexico, the US is clearly dominant
while Japan and Europe are relegated to minor trading partners. On the other hand, Chile shows
the highest portion of trade with European countries among all countries in this study. Thus, in
the weighting for East Asian countries, the weights of industrial countries are assigned to the
US, Japan, and European Union according to their actual trade volumes while the weights of
developing countries are assigned to the three largest trading partners within East Asia as
long as necessary data are available.4 Due to data limitations, Hong Kong and China are omit-
ted among the three. (We do not include the trade weights of the United Kingdom in the EU due
to its explicitly different exchange rate policy from the rest of the Union.) For Chile, however,
we use only industrial countries in weighting because its main regional partners such as Brazil
and Argentina have relatively small shares in trade with Chile and their required macroeco-
nomic time series are very short as a result of the history of hyperinflation. For Mexico, we
use the bilateral exchange rate against the US dollar due to the preeminent role of the US
for the country.5 The foreign price level, used in defining the real exchange rate as well as for-
eign income, is constructed in a parallel manner using the same weighting system.6
Fig. 1 shows the real exchange rate (in real line) and output (in dotted line). The nominal and
real exchange rates are defined so that an increase denotes depreciation of the domestic
currency. We use trade-weighted exchange rates. The HodrickePrescott filter is used to elimi-
nate the strong trend in real income data (with l ¼ 1600). As indicated by the scale of the real
exchange rate shown on the left side of the figure, the real exchange rate is more stable in East
Asian countries with the exception of Indonesia.
The case of negative correlation shows up most clearly in Mexico where cyclical downturns
are almost invariably associated with real depreciation. The relative stability of the real ex-
change rate in East Asia makes it difficult to read correlation in other countries. In general,
unusually large real depreciation tends to be associated with unusually large decline in real
income, the most recent and visible case being the financial crisis of 1997e1998 in East Asia.
Theory does not make it clear whether the bivariate relationship is long-term or short-term.
The effects of exchange rate changes on income are expected to be temporary since, in the long
run, prices adjust proportionately to the change in the nominal exchange rate. The Balassae
Samuelson effect suggests, however, that there may be a negative correlation in the long run.
In fast-growing economies like those of East Asian countries, the relative price of nontraded

4
We use the trade data of 1992, which is roughly the midpoint of the sample period. All trade data are obtained from
Direction of Trade Statistics (CD ROM) except those of Taiwan which are obtained from Aremos provided by Ministry
of Education of Taiwan.
5
More realistic weighting involving other countries makes little qualitative difference since non-US countries have
small shares in trade.
6
The distribution of weights is as follows. Indonesia: US (0.15), Japan (0.31), EU (0.18), Korea (0.12), Singapore
(0.15), and Taiwan (0.08); Korea: US (0.27), Japan (0.23), EU (0.13), Indonesia (0.13), Singapore (0.15), and Taiwan
(0.11); Malaysia: US (0.20), Japan (0.23), EU (0.13), Korea (0.05), Singapore (0.32), and Taiwan (0.07); Philippines:
US (0.30), Japan (0.22), EU (0.14), Korea (0.10), Singapore (0.09), and Taiwan (0.15); Singapore: US (0.22), Japan
(0.17), EU (0.14), Korea (0.06), Malaysia (0.32), and Taiwan (0.10); Taiwan: US (0.31), Japan (0.23), EU (0.17), Korea
(0.10), Malaysia (0.08), and Singapore (0.15); Thailand: US (0.19), Japan (0.28), EU (0.17), Malaysia (0.08), Singapore
(0.19), and Taiwan (0.09); Chile: US (0.33), Japan (0.25), and EU (0.41).
Indonesia Korea Malaysia
-3.25 0.32 -4.06 0.10 -4.16 0.16
-3.50 0.24 0.12
-4.13 0.05 -4.32
-3.75 0.16 0.08
-4.20
-4.00 0.08 -0.00
-4.48 0.04
-4.25 0.00 -4.27
-0.05 0.00

Y. Kim, Y.-H. Ying / Journal of International Money and Finance 26 (2007) 265e283
-4.50 -0.08 -4.34 -4.64 -0.04
-4.75 -0.16 -0.10
-4.41 -0.08
-5.00 -0.24
-4.48 -0.15 -4.80
-5.25 -0.32 -0.12
-5.50 -0.40 -4.55 -0.20 -4.96 -0.16
81

83

89

91

93

95

99

81

83

89

91

93

95

99

81

83

89

91

93

95

99
97

97

97
85

87

85

87

85

87
19

19

19

19

19

19

19

19

19

19

19

19

19

19

19

19

19

19

19

19

19
19

19

19
19

19

19

19

19

19
Philippines Singapore Taiwan
-4.16 0.25 -4.30 0.15 -4.15 0.050
-4.24 0.20 -4.35 -4.20
0.10
-4.32 0.15 -4.40 -4.25 0.025
-4.40 0.05
0.10 -4.45 -4.30
-4.48
0.05 -4.50 0.00 -4.35 0.000
-4.56
0.00 -4.55 -4.40
-4.64 -0.05
-4.72 -0.05 -4.60 -4.45 -0.025
-0.10
-4.80 -0.10 -4.65 -4.50
-4.88 -0.15 -4.70 -0.15 -4.55 -0.050

81

83

89

91

93

99
93

95
99
81

83

89

91

93

95

99

81

83

89

91

95

97
85

87
97

97
85

87

85

87

19

19

19

19

19

19
19

19
19
19

19

19

19

19

19

19

19

19

19

19

19

19
19

19
19

19
19

19

19

19
Thailand Chile Mexico
-4.1 0.100 -4.0 0.15 -3.50 0.06

-4.2 0.075 0.04


-4.2 0.10
0.050 -3.75 0.02
-4.3
0.025 -4.4 0.05 0.00
-4.4 -0.000 -0.02
-4.6 0.00 -4.00
-4.5 -0.025 -0.04
-0.050 -4.8 -0.05 -0.06
-4.6 -4.25
-0.075 -0.08
-4.7 -5.0 -0.10
-0.100 -0.10
-4.8 -0.125 -5.2 -0.15 -4.50 -0.12
81

83

89

91

93

99

81

83

89

91

93

99

81

83

89

91

93

99
95

97

95

97

95

97
85

87

85

87

85

87
19

19

19

19

19

19

19

19

19

19

19

19

19

19

19

19

19

19
19

19

19

19

19

19
19

19

19

19

19

19

269
Fig. 1. Real exchange rate and output.
270 Y. Kim, Y.-H. Ying / Journal of International Money and Finance 26 (2007) 265e283

goods would rise faster and thus lead to appreciation of the real exchange rate over time.
Existing empirical studies seem to suggest that the above growth effects on the real exchange
rate are small or nonexistent in most Asian countries. (See, inter alia, Kamin and Klau, 1998;
Bahmani-Oskooee, 1998; Upadhyaya, 1999.)
We employ cointegration tests to investigate the nature of long-run relationships. Table 1
reports the results of unit-root and cointegration tests. Augmented Dickey Fuller (ADF) tests
with 4 lags are used. Variables are defined as follows: YD, domestic real income; RX, the
real exchange rate; EX, the nominal exchange rate; RP, the ratio of the domestic price level
to the foreign; RM, real money supply; KY, the capital accounteincome ratio; CY, the current
accounteincome ratio. (See Appendix A for the definition of each variable used in this study.)
The results indicate that the presence of unit roots is not rejected for most variables in most
countries. In a small number of exceptional cases such as Singapore’s real income and the
nominal exchange rate, the current accounteincome ratio of Malaysia and Thailand and several
variables of Chile and Mexico, there is evidence that the unit-root hypothesis is rejected at the
5% significance level.
The last three columns in Table 1 labeled CI2, CI3, and CI5 report the ADF test of cointegra-
tion for real income and the real exchange rate. Given the possibility of omitted variables in the
cointegration relations, we also report the test results adding foreign income in CI3, and foreign
income, the capital accounteincome ratio and real money supply in CI5. In no case (with a pos-
sible exception of Indonesia), the null hypothesis of noncointegration between output and the real
exchange rate is rejected at the 5% level. Based on these results, we proceed with the assumption
that variables employed here are nonstationary and noncointegrated. Given the low power of the
test and that there is some evidence suggesting the assumption may not be valid, one needs to be
cautious and a further check on the robustness of the results may be necessary.
The bivariate relationships between the real exchange rate and real income are summarized
using cross correlations with leads and lags up to four quarters. Since the results turn out to be
somewhat sensitive to the method of detrending, we employ three different filters: linear
detrending, first differencing, and the HodrickePrescott filter. The 1997 financial crisis in
most East Asian countries constitutes a major structural break in the data. Appendix B reports
the results of tests of structural break around the time of the financial crisis. No doubt, the

Table 1
Unit-root and cointegration tests
Unit-root tests Cointegration tests
YD RX EX RP RM KY CY CI2 CI3 CI5
Indonesia 2.56 1.03 2.36 1.42 1.84 2.82 1.25 1.79 1.84 3.98
Korea 2.48 1.90 2.30 2.68 1.41 2.79 2.77 0.96 3.05 3.00
Malaysia 2.91 0.13 1.83 0.54 2.28 2.58 3.78 3.12 2.92 3.15
Philippines 1.93 2.47 2.50 2.81 0.11 1.17 0.08 0.91 2.18 1.49
Singapore 3.30 3.35 2.86 3.78 3.45 1.02 1.26 0.08 2.79 2.68
Taiwan 1.70 1.56 1.16 2.20 1.49 1.90 1.59 0.67 2.54 2.30
Thailand 2.18 1.83 2.11 2.71 2.16 1.77 3.17 2.26 1.86 2.00
Chile 5.08 2.11 3.41 3.54 1.48 3.27 2.99 2.05 2.53 4.04
Mexico 3.12 3.29 2.46 2.01 2.57 2.26 2.14 0.49 3.31 2.91
Note: Reported are Augmented Dickey Fuller unit-root tests with 4 lags. CI2 and CI3 are cointegration tests with (YD,
RX) and {YD, RX, YF}, respectively. CI5 is an expanded model with {YD, RX, YF, KY, RM}. For the real and nominal
exchange rates, the effective exchange rate is used for the East Asian countries and the bilateral US dollar rate for Chile
and Mexico.
Y. Kim, Y.-H. Ying / Journal of International Money and Finance 26 (2007) 265e283 271

nominal exchange shows the strongest evidence of structural breaks. It is interesting to note that
both capital flows and the current account behave differently in the majority of the East Asian
countries. In contrast, the two Latin American countries show no evidence of structural break
around the same period. In the following, we present the results using only the pre-crisis data
until the second quarter of 1997 as the benchmark case although we report the case for the
whole period for completeness.
Table 2 reports three representative short-term cross correlations between the real exchange
rate and real income at lags 4, 0, and 4. A negative (positive) lag means the number of quar-
ters by which the real exchange rate leads (lags) real income. The effects of devaluation on
output can be gleaned from correlations at negative lags while correlations at positive lags
suggest the extent of reverse causation from real income to the real exchange rate. Correlations
at negative lags are clearly negative in the two Latin American countries. In other words,
depreciation of the real exchange rate is followed by a cyclical downturn, consistent with

Table 2
Cross correlations
Lag Pre-crisis Whole period
LT DIF HP LT DIF HP
Indonesia 4 0.08 0.08 0.15 0.23 0.01 0.31
0 0.24 0.09 0.00 0.21 0.09 0.10
4 0.45 0.17 0.14 0.33 0.07 0.05
Korea 4 0.46 0.12 0.34 0.11 0.13 0.31
0 0.11 0.08 0.11 0.33 0.11 0.28
4 0.29 0.05 0.53 0.54 0.05 0.35
Malaysia 4 0.43 0.10 0.02 0.31 0.07 0.04
0 0.27 0.16 0.22 0.17 0.07 0.27
4 0.15 0.10 0.11 0.16 0.04 0.08
Philippines 4 0.48 0.26 0.46 0.53 0.21 0.38
0 0.27 0.02 0.00 0.21 0.03 0.03
4 0.19 0.26 0.63 0.17 0.25 0.60
Singapore 4 0.24 0.24 0.62 0.30 0.19 0.55
0 0.14 0.12 0.06 0.08 0.16 0.11
4 0.53 0.29 0.61 0.47 0.28 0.58
Taiwan 4 0.10 0.13 0.37 0.35 0.15 0.36
0 0.53 0.07 0.25 0.71 0.08 0.25
4 0.63 0.02 0.31 0.73 0.05 0.30
Thailand 4 0.47 0.01 0.12 0.29 0.12 0.08
0 0.22 0.13 0.19 0.03 0.20 0.23
4 0.12 0.08 0.45 0.26 0.16 0.36
Chile 4 0.23 0.03 0.10 0.23 0.00 0.08
0 0.59 0.03 0.35 0.59 0.02 0.34
4 0.27 0.06 0.03 0.31 0.04 0.00
Mexico 4 0.35 0.04 0.21 0.27 0.05 0.20
0 0.76 0.21 0.59 0.66 0.20 0.58
4 0.16 0.07 0.12 0.10 0.07 0.11
Note: Reported are the cross correlations at lag k, corrðret ; ~ytk Þ, of real exchange rate, re, and detrended real income, ~y.
A positive (negative) lag represents the number of quarters by which real income leads (lags) the real exchange rate.
272 Y. Kim, Y.-H. Ying / Journal of International Money and Finance 26 (2007) 265e283

the hypothesis of a contractionary devaluation. The statistical significance is strongest with the
linear trend and much weaker when first differencing is used. As for the East Asian countries,
evidence points to the opposite direction: in most countries, the correlation at negative lags is
positive, with Indonesia being the main exception. It is also interesting to note that the evidence
of contractionary devaluation is stronger (or the evidence of expansionary devaluation is
weaker) in the whole sample than in the pre-crisis period. In the cases of Indonesia, Korea,
and Taiwan, the difference appears to be significant.
The cross correlations for positive lags, in contrast, appear to be negative in both regions
with Malaysia being the sole exception. The negative correlation appears more significant in
East Asia than in Latin America. The negative correlation appears marginally significant in
Mexico. It suggests that robust economic growth tends to be followed by real appreciation,
and recession by real depreciation in these countries.
In sum, Mexico seems to be the only country where the hypothesis of contractionary devalu-
ation appears to be more plausible than the reverse causation hypothesis. For most East Asian
countries, the evidence reported in Table 2 is more consistent with the hypothesis in which
increases in real income cause appreciation in the real exchange rate than with one in which
devaluations cause economic contraction.
To explore the bivariate relationship further, we report in Table 3 the results of Granger cau-
sality tests. In all regressions, we use 4 lags. Capital flows and foreign income (defined using
the same weighting system mentioned above) are included in the regression to control external
influences that affect both real exchange rate and real income at the same time.
In Chile and Mexico, there is evidence that the causation is from the real exchange rate to real
income rather than the other way around. This causation shows up consistently in both sample
periods. In East Asia, Singapore seems to be the only country where the causation from the real
exchange rate to output is consistently significant. In this case, the causation from real income to
the real exchange rate seems equally strong. In other countries, however, evidence that the
observed bivariate correlation is due to causal factors does not seem to be robust.

3. The model

Instead of contractionary effects of devaluation, the negative correlation between real


income and the exchange rate may arise due to reverse causation and spurious correlation.
Reverse causation from real income to the real exchange rate is likely to be negative. This is
true especially in fast-growing developing countries where higher income is associated with
increases in the relative price of nontraded goods and accompanying appreciation of the real
exchange rate over time. Spurious correlation arises where both are affected by some third fac-
tors and devaluations usually occur in times of adverse economic conditions. Sharp changes in
oil prices or capital flows may drive both income and the real exchange rate in opposite direc-
tions. In several Latin American countries, exchange rate-based stabilization was a main cause
of large capital flows leading to a temporary boom and real appreciation and, as a result, a neg-
ative correlation in the two variables (Kiguel and Liviathan, 1992; Calvo and Vegh, 1993). It is
thus important to control for macroeconomic conditions and to separate exchange rate changes
that may be classified as exogenous policy shocks from those that are reactions to macroeco-
nomic events.
In this section, we develop a VAR model to shed some light on the issue raised in the
previous section. We consider a six-variable model that consists of the following variables:
capital inflows (kt), real income ( yt), the relative price (qt), real money supply (mt), the current
Y. Kim, Y.-H. Ying / Journal of International Money and Finance 26 (2007) 265e283 273

Table 3
Granger causality tests
Pre-crisis Whole period
LT DIF HP LT DIF HP
Indonesia A 0.18 (0.95) 0.30 (0.87) 0.25 (0.91) 4.74 (0.00) 4.49 (0.00) 5.92 (0.00)
B 1.06 (0.39) 1.10 (0.37) 0.54 (0.71) 1.41 (0.24) 0.42 (0.79) 0.20 (0.94)
Korea A 0.84 (0.51) 1.78 (0.15) 0.82 (0.52) 1.23 (0.31) 1.18 (0.33) 1.27 (0.29)
B 1.89 (0.13) 0.77 (0.55) 2.61 (0.05) 1.50 (0.22) 0.38 (0.82) 1.45 (0.23)
Malaysia A 1.05 (0.39) 0.33 (0.86) 1.47 (0.22) 3.04 (0.02) 1.63 (0.17) 3.42 (0.01)
B 0.94 (0.45) 0.56 (0.69) 1.79 (0.14) 0.44 (0.78) 0.30 (0.88) 0.95 (0.44)
Philippines A 2.47 (0.06) 2.26 (0.08) 2.74 (0.04) 1.80 (0.14) 0.94 (0.45) 1.80 (0.14)
B 2.32 (0.07) 2.48 (0.06) 2.94 (0.03) 0.68 (0.61) 1.97 (0.11) 3.12 (0.02)
Singapore A 2.59 (0.04) 3.11 (0.02) 7.19 (0.00) 2.81 (0.03) 3.61 (0.01) 6.17 (0.00)
B 4.96 (0.00) 5.28 (0.00) 8.11 (0.00) 4.00 (0.00) 4.49 (0.00) 5.53 (0.00)
Taiwan A 2.34 (0.07) 3.32 (0.01) 3.29 (0.02) 1.25 (0.30) 1.89 (0.12) 2.48 (0.05)
B 3.34 (0.02) 2.65 (0.05) 2.46 (0.06) 4.22 (0.00) 3.28 (0.02) 2.49 (0.05)
Thailand A 5.19 (0.00) 1.49 (0.21) 1.21 (0.32) 3.86 (0.01) 0.85 (0.50) 0.82 (0.52)
B 1.27 (0.29) 0.73 (0.57) 1.08 (0.37) 2.29 (0.07) 1.41 (0.24) 1.80 (0.14)
Chile A 6.01 (0.00) 3.39 (0.01) 4.73 (0.00) 5.55 (0.00) 3.19 (0.02) 4.69 (0.00)
B 1.24 (0.30) 1.00 (0.41) 2.93 (0.03) 1.47 (0.22) 1.39 (0.25) 3.03 (0.02)
Mexico A 3.54 (0.01) 4.79 (0.00) 5.01 (0.00) 5.12 (0.00) 6.05 (0.00) 5.90 (0.00)
B 0.40 (0.80) 0.14 (0.96) 0.31 (0.87) 0.50 (0.74) 0.14 (0.97) 0.39 (0.81)
Note: Reported are F-statistics with P values inside the parentheses. Row A tests the hypothesis that the real exchange
rate Granger causes real income. Row B tests the hypothesis that real income Granger causes the real exchange rate.

account balance (ct), and the nominal exchange rate (et). The model also includes two exoge-
nous variables: foreign real GDP ðyt Þ and the foreign interest rate ðit Þ. The six-variable model
with two exogenous variables is fairly comprehensive. The six endogenous variables are
ordered as listed. Given the evidence reported in the previous section, all variables are differ-
enced in the model.7 Eq. (1) summarizes the model in a compact form:

0 1 0 1 0 1 0 11
Dkt m1 Dkt1 3t
B Dyt C B m2 C B Dyt1 C B 32t C
B C B C B C    B 3C
B Dqt C B m3 C  B C   B C
B C ¼ B C þ Aij ðLÞ B Dqt1 C þ Bij ðLÞ Dyt1 þ B 3t4 C; ð1Þ
B Dmt C B m4 C B Dmt1 C Dit1 B3 C
B C B C B C B t5 C
@ Dct A @ m5 A @ Dct1 A @3 A
t
Det m6 Det1 36t

where Aij(L) and Bij(L) are 6  6 and 6  2 matrices of polynomials in lag operator L.

7
VAR models have been employed in related contexts. Rogers and Wang (1995) show that devaluations lead to output
declines in Mexico; Hoffmaister and Vegh (1996) find that a permanent reduction in the exchange rate depreciation
leads to a long-lasting increase in output. See Kamin and Rogers (2000) for a comprehensive study with the Mexican
data as well as an excellent survey of recent literature.
274 Y. Kim, Y.-H. Ying / Journal of International Money and Finance 26 (2007) 265e283

Several features of the model can be noted. First, external shocks represented by foreign
income and foreign interest rate are explicitly incorporated.8 (The US 3-month Treasury bill
rate is used as foreign interest rate.) Second, capital flows are treated as a driving force of
the macroeconomic variables. The placement of the capital flows’ variable at the top of the
endogenous variables reflects the assumption that, in the short run, it is largely determined
by external factors (Calvo et al., 1993). Third, the real exchange rate is decomposed into the
nominal exchange rate and the relative price. While a country may have some control over
the nominal exchange rate in the short run, the real exchange rate is ultimately determined
by such factors as productivity growth and thus cannot be controlled in the long run. Moreover,
devaluation often fails as resulting increases in the price level erode competitive advantage gen-
erated by the change in the nominal exchange rate. With the setup, we can study the effects of
a nominal devaluation on the relative price and therefore on the real exchange rate. Fourth, the
real money balance is included since changes in the money supply may determine the short-
term results of devaluation. The effects of devaluation crucially depend on monetary and fiscal
policies that accompany it. The behavior of the real money supply can be an important key to
understanding the effects of devaluation.
Fifth, the current account is included as well as the capital account. Most countries included
in this study eliminated capital controls and liberalized capital account transactions in the
1990s. During the bulk of the sample period, shocks to the current account such as terms of
trade shocks were the main shocks to the balance of payments. Moreover, due to heavy inter-
vention in the foreign exchange market, the two accounts were strongly related with opposite
signs but their behaviors were not identical. By placing the current account above the nominal
exchange rate, we can also narrow the range of exchange rate shocks so that they do not include
contemporaneous reactions to developments in the balance of payments from either capital or
current account sources.
One of the most important differences between this model and the models adopted in the
previous studies would be that the nominal exchange rate is placed at the bottom. The ordering
is based on the observation that the exchange rate, as a forward-looking asset price, responds
quickly to macroeconomic shocks. For instance, the exchange rate authority may respond to
adverse shocks to the current account by letting the domestic currency depreciate. Similarly,
favorable productivity shocks may lead the authority to allow its appreciation. In this setup,
these contemporaneous (endogenous) changes in the exchange rate due to other variables in
the system are taken care of and excluded from exchange rate shocks. What remains after
accounting for those endogenous responses can be termed exchange rate shocks. They may
be called exogenous changes in the exchange rate that are the subject of a devaluation study.9
This is adapted from the methodology of Kim and Roubini (2000) who employed a similar
methodology in their study of monetary policy effects and accounted for various puzzles in
existing studies. The assumption appears reasonable in our case since most countries studied

8
We tried additional variables such as the terms of trade but dropped them as they made little qualitative differences
in results.
9
The exchange rate is often placed at or near the top of ordering on the grounds that empirical studies have found that
it is not well explained by other variables. For instance, Kamin and Rogers (2000) place the real exchange rate right
below the external variables such as the foreign (US) interest rate and the terms of trade. In such models, innovations
from the exchange rate equation are likely to include adverse shocks to the economy such as deterioration in terms of
trade, leading to spurious correlation.
Y. Kim, Y.-H. Ying / Journal of International Money and Finance 26 (2007) 265e283 275

in this paper have maintained some sort of managed floating in which the exchange rate is
adjusted in response to the economic condition.
Panels in Figs. 2 and 3 show the impulse responses to an increase in the nominal exchange
rate (shown in broken line with dots in wide gaps). In addition to the response of output (in
evenly broken line), responses of the real exchange rate (solid real line), the current account
(in broken line with dots in narrow gaps), and the real money supply (evenly dotted line) are
also listed.10 Although the model is estimated in the first difference form, the variables are con-
verted back to levels in impulse responses.
Due to the structural break in the series induced by the financial crisis, it is not clear which,
between the whole sample and the pre-crisis period, better represents the economies in East
Asia. There is strong evidence of structural breaks in several variables in the system in all cri-
sis-stricken economies (see Appendix B). We present the case with the pre-crisis period in
Fig. 2 and the one with the whole period in Fig. 3.
Panels in Fig. 2 suggest that devaluation of the effective exchange rate in the pre-crisis
period has been contractionary in no East Asian countries. On the contrary, it seems to be
strongly expansionary in all countries but Indonesia, Malaysia, and the Philippines, where
the output effects appear small and insignificant. Moreover, the expansionary effects seem to
be persistent and supported by persistent depreciation of the real exchange rate after devalua-
tion. In remarkable contrast, devaluation appears to be strongly contractionary in the two Latin
American countries. The contractionary effects seem stronger and to last for longer periods in
Mexico although they appear to be temporary in both countries.
According to Fig. 3, for the whole period, devaluation is unambiguously contractionary in
Mexico and, at least initially, in Chile as well. Fig. 3 suggests that devaluation leads to similar
results in three East Asian countries as well: Malaysia and, with some lags, Indonesia and the
Philippines. In the other four countries e Korea, Singapore, Taiwan, and Thailand e however,
there is no evidence that devaluation is recessionary. Instead, output rises strongly and signif-
icantly in all these cases.
It then follows that the possibility of contractionary devaluation is more general for Chile
and Mexico, whereas in East Asia, it is much more limited. The difference in the results in
the two estimation periods probably arises from the fact that during the 1997e1998 financial
crisis, currency devaluation and economic recession occurred at the same time on an unprece-
dented scale. At least for the pre-crisis period, there is no evidence that devaluations are con-
tractionary in East Asia.
A closer inspection of Figs. 2 and 3 reveals that sources of contractionary devaluations differ
across countries. In Mexico and less so in Chile, the real exchange rate falls below the pre-
devaluation level within a few quarters after devaluation. In these cases, appreciation of the
real exchange rate seems to be the main cause of the contractionary pressure after devaluation.
The two countries have employed the exchange rate as an anchor in a number of attempts to
reduce inflation, in which the rate of currency devaluation was kept at a level lower than the
differential between domestic and foreign inflation rates. With strong inflation inertia, the
real exchange rate starts appreciating soon after devaluation episodes, thereby building up
the recessionary pressure as the traded sector of the economy loses competitiveness. As can

10
Although the responses of the nominal exchange rate and the current account are shown in similar lines, the distinc-
tion is easy because the nominal and real exchange rates start at a non-zero level while the other three responses start at
zero.
276
Indonesia Korea Malaysia
1.50 1.4 1.5
1.25 1.2 1.0
1.00 1.0 0.5
0.75 0.0
0.8

Y. Kim, Y.-H. Ying / Journal of International Money and Finance 26 (2007) 265e283
0.50 -0.5
0.6
0.25 -1.0
0.4
0.00 -1.5
-0.25 0.2 -2.0
-0.50 0.0 -2.5
-0.75 -0.2 -3.0
5 10 15 0 5 10 15 0 5 10 15

Philippines Singapore Taiwan


1.5 2.0 3.5
1.0 3.0
1.5
0.5 2.5
1.0
0.0 2.0
-0.5 0.5 1.5
-1.0 1.0
0.0
-1.5 0.5
-0.5
-2.0 0.0
-2.5 -1.0 -0.5
0 5 10 15 0 5 10 15 0 5 10 15

Chile Mexico
Thailand
2.0 1.00 3

1.5 0.75 2

1
1.0 0.50
0
0.5 0.25
-1
0.0 0.00
-2
-0.5 -0.25 -3

-1.0 -0.50 -4
0 5 10 15 0 5 10 15 0 5 10 15

Fig. 2. Impulse responses to exchange rate change (pre-crisis period).


Indonesia Korea Malaysia
1.5 1.50 2

1.0 1.25 1
1.00
0.5 0

Y. Kim, Y.-H. Ying / Journal of International Money and Finance 26 (2007) 265e283
0.75
0.0 0.50 -1
0.25 -2
-0.5
0.00
-1.0 -3
-0.25
-1.5 -0.50 -4
0 5 10 15 0 5 10 15 0 5 10 15

Philippines Singapore Taiwan


1.6 1.6 2.00
1.75
0.8 1.2
1.50
0.8
-0.0 1.25
0.4 1.00
-0.8
-0.0 0.75
-1.6 0.50
-0.4
0.25
-2.4 -0.8 0.00
-3.2 -1.2 -0.25
0 5 10 15 0 5 10 15 0 5 10 15

Thailand Chile Mexico


1.05 1.00 3

0.70 0.75 2
1
0.35 0.50
0
0.00 0.25
-1
-0.35 0.00
-2
-0.70 -0.25 -3
-1.05 -0.50 -4
-1.40 -0.75 -5
0 5 10 15 0 5 10 15 0 5 10 15

Fig. 3. Impulse responses to exchanger rate change (whole period).

277
278 Y. Kim, Y.-H. Ying / Journal of International Money and Finance 26 (2007) 265e283

be seen in Fig. 3, Indonesia e perhaps the only East Asian country that has had inflation prob-
lems similar to that of Chile and Mexico e seems to follow this pattern. In all the other coun-
tries, the real exchange rate and the nominal exchange rate follow similar paths, suggesting that
currency devaluation has not been undermined by price increases for a significant length of
time.
In Malaysia, the real money supply sharply declines with devaluation. In this country, con-
tractionary devaluation seems to arise mainly from conservative (unaccommodating) monetary
policy that accompanies devaluation. Panels in Fig. 3 suggest, however, that tight monetary
policy was not necessarily responsible for the outcome given that devaluation leads to output
expansion despite sustained reduction in real money supply in several other countries such
as Korea, Singapore and Thailand.
It is important to note that, in both sets of figures, whether output increases after a devalua-
tion or not has little to do with the fact whether the current account improves or not. In Fig. 2,
where devaluation turns out to be expansionary in East Asia, several countries including the
Philippines and Singapore seem to experience a moderate decline in the current account. In
Fig. 3, the current account improves in all three countries e Malaysia, Indonesia, and the
Philippines e where there is evidence of contractionary devaluation. It suggests that, whether
devaluation is contractionary or expansionary may have little to do with the trade elasticities’
condition.
In Tables 4 and 5, we present two sets of variance decomposition for the sample periods.
3i denotes innovations from the equation for variable i. For space reasons, variance decompo-
sitions of only three variables are reported: output, the relative price, and the nominal exchange
rate.
For real income, own shocks are the most important source in all countries. In the pre-crisis
East Asia (Table 4), output is little affected by the exchange rate in any East Asian countries
except Taiwan. Also, capital flows play a small role in all but the Philippines. These results sug-
gest that East Asian countries during the pre-crisis period may well be described as ‘‘insular.’’
Even though the export-driven economic growth has significantly opened up the economy and
raised the extent of dependence on foreign economic conditions, interest rate and foreign
exchange controls and various other impediments on capital flows seem to have rendered
them largely insensitive to external shocks.11
For the exchange rate, Table 4 suggests that own shocks are most important with the excep-
tion of the Philippines. Shocks to income are the distant second in explaining the exchange rate.
Other than shocks to income, no variables seem to be consistently related to exchange rate
shocks. The results suggest that the exchange rate authorities of these countries have used
the exchange rate as a stabilization tool, allowing it to adjust to business conditions while trying
to keep the effective exchange rate stable.
As for the whole period (shown in Table 5), capital flows take a major role in several cases.
The increase in the importance of capital flows is clear in Indonesia, Korea, and Thailand. Ma-
laysia and Chile are interesting exceptions. Both are well known to have used capital controls in
times of exchange rate crises, which have presumably helped to maintain stability in the middle
of financial turmoil. As a consequence, variations in capital flows have minimal effects on the

11
Note that our model includes two exogenous variables, foreign output and the foreign interest rate. Small effects of
capital flows on output may mean that the two exogenous variables explain most of the variations in capital flows. Thus
when their effects are taken into consideration, little is left to capital flows.
Y. Kim, Y.-H. Ying / Journal of International Money and Finance 26 (2007) 265e283 279

exchange rate in these cases. The role of capital flows is also small in the relatively stable coun-
tries of Singapore and Taiwan.

4. Conclusion

In this paper, we investigated whether currency devaluation is likely to be contractionary on


domestic output in East Asian countries as it is in Latin American countries as suggested in

Table 4
Variance decomposition (pre-crisis)
3KY 3YD 3RP 3RM 3CY 3EX
(4) (12) (4) (12) (4) (12) (4) (12) (4) (12) (4) (12)
Indonesia
YD 2.2 2.5 79.1 70.3 0.2 2.4 15.0 21.0 2.6 2.6 1.0 1.1
RP 2.6 3.9 17.6 21.0 64.3 51.3 5.1 7.7 8.5 11.8 1.9 4.3
EX 13.1 10.9 10.5 10.1 11.1 17.8 10.6 19.3 21.5 17.6 32.3 24.4
Korea
YD 1.2 6.0 82.7 66.6 1.8 8.5 6.6 6.9 5.9 10.1 1.9 1.8
RP 7.6 7.3 2.0 5.3 66.0 58.8 4.9 5.2 17.3 20.6 2.3 2.9
EX 2.7 5.3 12.4 17.2 2.7 8.8 12.0 14.7 6.5 9.4 63.6 44.6
Malaysia
YD 4.2 5.4 84.2 78.2 0.4 2.0 5.5 4.9 5.2 6.9 0.5 2.7
RP 18.6 18.8 0.8 1.6 68.3 64.3 3.5 4.0 1.4 2.7 7.4 8.6
EX 2.5 3.2 3.7 4.7 12.0 11.7 8.8 9.9 3.4 4.2 69.6 66.1
Philippines
YD 12.2 22.2 55.4 39.2 13.6 12.6 2.9 6.8 12.6 15.0 3.2 4.3
RP 5.8 10.7 38.0 31.2 33.4 26.4 9.3 10.5 1.9 12.3 11.5 8.0
EX 23.4 22.3 21.6 30.4 19.8 14.6 22.3 9.5 2.8 12.9 10.0 10.3
Singapore
YD 3.2 6.0 83.5 69.8 0.3 3.0 9.6 17.5 0.5 0.8 2.9 2.9
RP 5.3 6.8 2.9 5.6 84.3 77.2 1.3 3.8 5.9 5.9 0.4 0.7
EX 3.4 4.0 12.7 16.7 6.9 7.4 1.9 8.8 3.6 6.3 71.6 56.9
Taiwan
YD 2.0 8.4 60.6 44.8 2.9 4.8 18.6 18.7 0.5 1.6 15.4 21.7
RP 7.1 6.3 1.3 1.6 74.4 70.1 2.4 3.6 6.3 7.3 8.5 11.0
EX 6.7 10.0 18.2 17.4 3.5 5.7 5.8 6.2 6.1 8.0 59.8 52.8
Thailand
YD 2.5 2.7 89.9 80.6 1.5 3.7 2.4 5.1 0.3 1.3 3.4 6.6
RP 2.8 9.3 9.1 12.0 80.3 58.5 6.6 11.5 0.5 3.1 0.7 5.5
EX 2.4 4.8 1.3 2.5 1.3 3.2 3.7 4.9 4.3 5.6 87.1 79.1
Chile
YD 0.1 0.6 73.4 48.2 6.6 11.0 7.7 15.1 10.3 20.3 1.8 4.7
RP 0.9 1.1 1.3 5.4 56.7 54.3 5.0 5.3 11.5 10.3 24.5 23.6
EX 1.9 2.0 1.3 5.9 34.1 32.4 4.9 7.8 21.2 20.1 36.7 31.9
Mexico
YD 31.9 26.8 40.9 29.5 0.1 0.5 0.2 3.1 0.4 1.9 26.6 38.2
RP 25.4 18.4 3.7 6.7 16.5 10.8 1.6 1.6 12.9 9.4 40.0 53.1
EX 19.8 18.9 3.3 6.9 9.6 9.0 0.2 0.5 5.6 5.4 61.5 59.4
280 Y. Kim, Y.-H. Ying / Journal of International Money and Finance 26 (2007) 265e283

Table 5
Variance decomposition (whole period)
3KY 3YD 3RP 3RM 3CY 3EX
(4) (12) (4) (12) (4) (12) (4) (12) (4) (12) (4) (12)
Indonesia
YD 5.1 22.2 81.8 47.9 3.9 11.9 5.5 5.4 2.6 8.3 1.1 4.4
RP 80.6 72.1 1.7 4.6 11.4 12.4 0.8 3.0 2.3 2.8 3.1 5.2
EX 53.6 32.7 0.5 3.9 14.3 32.7 2.3 4.8 15.4 15.0 14.0 11.0
Korea
YD 31.2 36.0 45.5 30.1 4.5 5.5 15.5 20.4 1.4 3.1 1.8 5.0
RP 20.2 21.2 3.2 5.2 55.9 47.4 2.1 4.8 14.4 17.5 4.1 4.0
EX 48.6 44.1 3.3 4.7 7.7 8.0 3.0 5.9 4.4 8.3 33.1 29.0
Malaysia
YD 3.8 3.8 80.1 68.5 0.1 0.8 2.4 2.4 1.0 2.1 12.6 22.4
RP 15.7 16.2 0.6 1.2 64.6 60.2 3.4 4.6 0.5 2.4 15.2 15.5
EX 0.5 0.9 2.5 2.9 5.7 5.6 4.7 5.1 3.1 4.3 83.5 81.3
Philippines
YD 9.4 5.3 72.6 50.9 4.1 21.3 3.2 3.3 5.2 8.7 5.5 10.6
RP 2.7 3.0 11.9 12.4 59.3 55.3 4.8 4.4 4.2 5.7 17.1 21.9
EX 11.1 12.1 7.0 14.4 32.3 28.8 10.3 7.1 2.4 7.9 37.0 29.6
Singapore
YD 1.1 4.2 85.2 69.9 0.4 4.6 11.3 18.1 0.2 0.3 1.9 2.8
RP 6.6 7.6 2.1 4.8 81.4 74.1 1.8 5.1 7.1 6.9 0.9 1.4
EX 5.6 7.0 13.8 17.1 2.1 2.1 10.6 12.0 5.0 7.9 62.9 53.9
Taiwan
YD 0.9 3.0 78.6 62.0 3.4 6.0 8.9 12.9 0.8 1.8 7.3 14.4
RP 6.1 6.2 2.1 2.7 74.9 70.2 1.9 2.6 6.6 9.1 8.4 9.1
EX 3.3 6.1 11.3 12.1 11.0 11.3 4.6 5.3 5.5 10.5 64.3 54.6
Thailand
YD 20.4 36.9 78.0 55.5 0.9 2.0 0.1 1.7 0.5 0.6 0.0 3.3
RP 14.0 14.4 7.6 7.8 74.1 56.3 2.4 15.3 1.0 2.4 0.9 3.9
EX 25.1 29.1 4.4 5.6 2.8 3.2 9.0 8.8 7.3 12.1 51.4 41.1
Chile
YD 0.2 1.0 75.8 51.9 4.1 9.3 11.4 19.9 5.2 13.7 3.2 4.2
RP 0.8 1.1 2.4 4.3 61.9 61.4 3.8 4.7 7.6 7.6 23.5 20.9
EX 1.6 1.7 4.5 8.9 26.5 26.7 3.7 8.6 13.7 13.7 50.1 40.5
Mexico
YD 35.4 29.4 41.1 28.7 0.1 0.6 0.4 3.4 1.0 2.0 21.9 35.8
RP 25.1 17.4 2.1 2.2 19.7 12.7 1.3 1.2 9.9 7.0 41.9 59.5
EX 19.9 17.6 1.6 2.3 12.8 11.6 0.4 0.7 3.6 3.7 61.6 64.1

many previous studies with surprising consistency. We employed quarterly data for the last two
decades. Summarizing the main results: Historically, devaluation is closely associated with reces-
sion in both groups of countries. A closer inspection reveals, however, that the negative correlation
may be due to reverse causation, especially in East Asian countries in which strong income growth
has preceded appreciation of the real exchange rate. With the pre-crisis data alone, there is no
indication that devaluation is contractionary. In contrast, we find that output increases strongly
in the majority of cases. This is sharply contrasted to the case of Mexico and, somewhat weakly,
Chile, where there is strong evidence that a devaluation leads to temporary reduction in output.
Y. Kim, Y.-H. Ying / Journal of International Money and Finance 26 (2007) 265e283 281

The results also suggest that contractionary devaluation cannot be ruled out in East Asia.
They indicate that devaluation could be contractionary in East Asian countries just as well
as in Mexico and Chile, especially when the post-crisis data are included in estimation. How-
ever, given the evidence of structural breaks around the financial crisis, one needs to exercise
due care in reading the results based on the mixture of the pre- and post-crisis periods. No doubt
the inclusion of the dramatic episode tilts the results towards such a finding. Our exercise sug-
gests the crisis period was indeed different. With financial liberalization and improvement in
information technology, devaluation may be more likely to be contractionary than before as
it worsens the balance sheet of financial and nonfinancial business firms with heavy foreign-
currency liabilities and results in serious interruption of external financing through a loss of
credibility with international financial investors. It is too early to determine whether this new
paradigm is here to stay for East Asian countries. One must exercise due care in interpreting
the results for the period including financial crises since the exchange rateeoutput linkage
can be complicated by many abnormal factors such as contagion and self-fulfilling
expectations.

Acknowledgements

We thank two anonymous referees and the coeditor for their helpful comments. We also
appreciate Shirley Fong and Reiny Iriana for excellent research assistance. This research was
partially supported by University Research Grant (RP 122-000-036-112) at the National Uni-
versity of Singapore. National Science Council of Taiwan also provided generous funding
(NSC 91-2415-H-l 10-002) in the early stage of this research. This paper is also supported
by ‘‘Aim for the top University plan’’ of National Sun Yat-Sen University and Ministry of
Education, Taiwan, R.O.C.

Appendix A. Data and definitions

The domestic price level is measured by the consumer price index (CPI) and real income by
industrial production or, in Taiwan, real GDP. Most real income series are deseasonalized by
removing the effects of seasonal terms in a simple regression containing a linear trend and sea-
sonal dummies. For the money supply, the M2 concept is used. The capital (current) accounte
income ratio, KY (CY), is obtained as the ratio of the capital (current) account balance to the
trend nominal GDP. If the balance of payments statistics are unavailable, the current account
balance is approximated by the difference between merchandise exports and imports. The cap-
ital account, in turn, is approximated by first obtaining the difference in net foreign assets at the
central bank (which is considered the official reserve balance) and then subtracting the current
account from it. The log of trend nominal GDP is assumed to be linear in trend. For nominal
income series, we use nominal GDP if available. Otherwise, it is approximated by the product
of real income and the domestic price level, both defined above.
Foreign real income and the foreign price level are measured by real GDP and the CPI of the
US. The US 3-month Treasury Bill interest rate is used as foreign interest rate. For EER,
the simple geometric average of the US and Japan (the US, Japan, and Germany) is used as
the foreign country in East Asian countries (Chile). All variables are in logarithm except the
interest rate and the CY and KY ratios.
282 Y. Kim, Y.-H. Ying / Journal of International Money and Finance 26 (2007) 265e283

The sample periods are as follows: Indonesia (1981:1e1999:2), Korea (1980:1e2000:3),


Malaysia (1970:1e2000:1), Philippines (1981:1e2000:2), Singapore (1970:1e1999:4), Taiwan
(1981:1e2000:3), Thailand (1976:1e2000:2), Chile (1972:4e2000:3), and Mexico (1981:1
e2000:2). Taiwan’s data are extracted from Aremos databank. All other data are obtained
from International Financial Statistics (CR-ROM).

Appendix B. Tests of structural break

DF1 DF2 Y EX RP RM KY CY
Indonesia 8 37 0.49 20.5** 9.99** 1.94 103** 3.48*
Korea 9 42 1.39 17.1** 1.99þ 3.18* 9.58** 3.04*
Malaysia 11 84 1.19 8.73** 0.28 1.63 1.17 2.09þ
Philippines 11a 22a 9.83** 3.20* 5.14** 0.96 4.34** 1.20*
Singapore 10 83 0.83 10.2** 0.23 3.85* 1.34 2.05
Taiwan 11 40 1.15 2.62* 1.03 0.69 0.59 0.6811
Thailand 11 63 1.78þ 6.38** 1.77þ 0.96 7.62** 4.14*
Chile 13 61b 0.57 1.65 0.95 1.61 0.30 0.14
Mexico 12 41 0.56 0.50 0.43 0.48 0.38 0.50
Note: Reported are the F-statistics for the Chow predictive tests. DF1 and DF2 are the degrees of freedom for the nu-
merator and the denominator, respectively. ‘þ’, ‘*’ and ‘**’ denote significance at the 10, 5, and 1% level, respectively.
a
Denotes that the DF1 for KY and CY and DF2 for RM, KY, and CY for the Philippines are lower by 1.
b
In the case of Chile, the DF2 for RM and KY is 60.

References

Bahmani-Oskooee, M., 1998. Are devaluations contractionary in LDCs? Journal of Economic Development 23 (1),
131e144.
Buffie, E.F., 1989. Imported inputs, real wage rigidity and devaluation in the small open economy. European Economic
Review 33 (7), 1345e1361.
Calvo, G., Vegh, C.A., 1993. Exchange rate based stabilization under imperfect credibility. In: Frisch, H., Worgotter, A.
(Eds.), Open Economic Macroeconomics. MacMillan, London, pp. 3e28.
Calvo, G., Leiderman, L., Reinhart, C.M., 1993. Capital inflows and real exchange rate appreciation in Latin America.
IMF Staff Papers 40 (1), 108e151.
Cooper, R., 1971. Currency depreciation in developing countries. In: Princeton Essays in International Finance, 86.
Princeton University.
Corsetti, G., Pesenti, P., Roubini, N., 1999. What caused the Asian currency and financial crisis? Japan and the World
Economy 11 (2), 305e373.
Diaz-Alejandro, C.F., 1963. A note on the impact of devaluation and the redistributive effects. Journal of Political
Economy 71 (6), 577e580.
Dornbusch, R., Park, Y.C., 1999. Flexibility or nominal anchor? In: Collignon, S., Pisani-Ferry, J., Park, Y.C. (Eds.),
Exchange Rate Policies in Emerging Asian Countries. Routledge, London, pp. 3e34.
Edwards, S., 1989. Real Exchange Rates, Devaluation, and Adjustment. MIT Press, Cambridge, MA.
Eichengreen, B., Sachs, J., 1985. Exchange rates and economic recovery in the 1930s. Journal of Economic History
44 (4), 925e946.
Findlay, R., Rodriguez, C.A., 1977. Intermediated imports and macroeconomic policy under flexible exchange rates.
Canadian Journal of Economics 10 (2), 208e217.
Frenkel, J.A., Johnson, H.C., 1976. The monetary approach to the balance of payments: essential concepts and historical
origins. In: Frenkel, J.A., Johnson, H.C. (Eds.), The Monetary Approach to the Balance of Payments. University of
Toronto Press, Allen & Unwin, pp. 21e45.
Frankel, J.A., 1988. Ambiguous policy multipliers in theory and in empirical models. In: Bryant, R.C., et al. (Eds.),
Empirical Macroeconomics for Interdependent Economies. Brookings Institution, Washington, DC, pp. 17e26.
Y. Kim, Y.-H. Ying / Journal of International Money and Finance 26 (2007) 265e283 283

Goldstein, M., Khan, M.S., 1985. Income and price effects in foreign trade. In: Jones, R.W., Kenen, P.B. (Eds.), Hand-
book of International Economics, vol. II. North-Holland, Amsterdam, pp. 1041e1105.
Gordon, R.J., 2000. The aftermath of the 1992 ERM breakup: was there a macroeconomic free lunch? In: Krugman, P.
(Ed.), Currency Crises. National Bureau of Economic Research, Chicago, pp. 241e284.
Gylfason, T., Risager, O., 1984. Does devaluation improve the current account? European Economic Review 25 (1),
37e64.
Hoffmaister, A.W., Vegh, C., 1996. Disinflation and the recession-now-versus-recession-later hypothesis: evidence from
Uruguay. IMF Staff Papers 43 (2), 355e394.
Kamin, S.B., Rogers, J.H., 2000. Output and the real exchange rate in developing countries: an application to Mexico.
Journal of Development Economics 61 (1), 85e109.
Kamin, S.B., Klau, M., 1998. Some multi-country evidence on the effects of real exchange rates on output. Board of
Governors of the Federal Reserve System, International Finance Discussion Papers No. 611.
Kiguel, M., Liviathan, N., 1992. The business cycle associated with exchange rate based stabilization. World Bank
Economic Review 6 (2), 279e305.
Kim, S., Roubini, N., 2000. Exchange rate anomalies in the industrial countries: a solution with a structural VAR
approach. Journal of Monetary Economics 45 (3), 561e586.
Krugman, P., 1999. Balance sheets, the transfer problem, and financial crises. Unpublished, MIT.
Krugman, P., Taylor, L., 1978. Contractionary effects of devaluation. Journal of International Economics 8 (3),
445e456.
Kwan, C.H., 1994. Economic Interdependence in the Asia-Pacific Region. Routledge, London.
Lizondo, S., Montiel, P.J., 1989. Contractionary devaluation in developing countries: an analytical overview. IMF Staff
Papers 36 (2), 182e227.
Marston, R.C., 1982. Wages, relative prices and the choice between fixed and flexible exchange rates. Canadian Journal
of Economics 15 (1), 87e103.
Morley, S.A., 1992. On the effect of devaluation during stabilization programs in LDCs. Review of Economic &
Statistics 74 (1), 21e27.
Obstfeld, M., Rogoff, K., 1995. The mirage of fixed exchange rates. Journal of Economic Perspectives 9 (4), 73e96.
Reinhart, C.M., 2000. The mirage of floating exchange rates. American Economic Review 90 (2), 65e70.
Rogers, J.H., Wang, P., 1995. Output, inflation, and stabilization in a small open economy: evidence from Mexico.
Journal of Development Economics 46 (2), 271e293.
Sachs, J., 1980. Wages, flexible exchange rates, and macroeconomic policy. Quarterly Journal of Economics 94 (4),
31e47.
Schneider, M., Tornell, A., 2000. Balance sheet effects, bailout guarantee and financial crises. NBER Working Paper
8060.
Upadhyaya, K.P., 1999. Currency devaluation, aggregate output, and the long run: an empirical study. Economics
Letters 64 (2), 197e202.
Upadhyaya, K.P., Upadhyaya, M.P., 1999. Output effects of devaluation: evidence from Asia. Journal of Development
Studies 35 (6), 89e103.
Upadhyaya, K.P., Dhakal, D., Mixon, F.G., 2000. Exchange rate adjustment and output in selected Latin American
countries. Economia Internazionale 53 (1), 107e117.
World Bank, 1993. The East Asian Miracle. Oxford University Press, New York.

You might also like