Taka-Dollar exchange rate and Bangladesh trade balance: evidence on J-curve or S-curve?

Rahman, Matiur ; Islam, Anisul M.

Abstract

This paper examines the dynamics of Taka-Dollar exchange rate and Bangladesh trade balance. The fairly standard bivariate cointegration procedure is implemented by employing nominal data from the first quarter of 1972 through the final quarter of 2003. The results indicate weak evidence of the J-curve with significant short-run trade balance deterioration and a sluggish improvement in the longer run. But these results require interpretations with extra cautions.

INTRODUCTION

In theory, exchange rate depreciations would reduce imports and increase exports thereby contracting a country's trade deficit provided the well-known Marshall-Lerner condition that the sum of the export and import demand elasticities are at least equal to unity holds (for details, see Kulkarni, 1994). Therefore, the effects of exchange rate depreciations on exports, imports, and hence on trade balance are neither guaranteed (depends at least partly on whether the Marshall-Lerner condition holds) nor instant.

The traditional theoretically expected result may not be guaranteed because the Marshall-Lerner condition or some of its underlying conditions (such as infinite supply elasticity) may not hold. Further, the expected result, even if it happens, may not be instant as the exports and imports may respond to exchange rate adjustments with lags and their response lag-structures may also be different. Magee (1973) pioneered the J-curve theory to describe the effects of exchange rate depreciations on trade balance. According to this theory, a country's trade deficit worsens just after its currency depreciates because price effects will dominate the effect on volume of imports in the short run (see also Kulkarni 1994). In other words, the higher cost of imports will more than offset the reduced volume of imports. Thus, the J-curve depicts that a decline in the value of home currency against a foreign currency should be followed by a temporary worsening in the trade deficit before its longer-term improvement.

A small-open-economy model that captures some important features of less-developed-country (LDC) economies can also reproduce the S-curve as they have limited access to international financial markets for capital formation and for smoothing effects of international terms of trade and other exogenous shocks (Senhadji, 1998).

Bangladesh has a long history of persistently widening trade deficit and a near-continual depreciation of the currency against major currencies since her inception (Figure 2). A casual inspection also provides some preliminary indication of the existence of long-run relationship between trade balance and exchange rate (Figure 2).In Bangladesh, exchange rate depreciation is used as a tool to rectify the country's trade deficit in lieu of directly invoking restrictive trade practices. In the international economic literature, the reactions of trade balance to currency depreciations pose an important question. For Bangladesh, this question is of high importance.

[FIGURE 1 OMITTED]

[FIGURE 2 OMITTED]

The broad objectives of this paper with a theoretical overtone include (i) examination of a possible long-run equilibrium relationship of trade balance with real effective exchange rate, foreign income and domestic income; (ii) investigation of the direction of causal relationship among the aforementioned variables; and (iii) application of impulse response analysis to ascertain whether shocks to the real effective exchange rate induce the trade balance to follow a J-curve or S-curve trajectory. By implementing cointegration and error-correction modeling, this paper offers valuable insights into the dynamics of Bangladesh trade balance.

The remainder of the paper is structured as follows. Section II briefly reviews some recent relevant literature. Section III discusses the model and the empirical methodology. Section IV reports empirical results. Section V offers conclusions and remarks.

BRIEF REVIEW OF SOME RECENT LITERATURE

The J-curve literature, in particular, is extensive and yet evolving shifting the focus to the short-run dynamics that trace the post-devaluation time-path of trade balance. The S-curve literature is rather relatively very scant. The empirical literature on the evidence of the J-curve is rather mixed.

Rose and Yellen (1989) studied the short-run dynamics between exchange rate and trade balance. They found no evidence of the J-curve for G-7 countries. Rose (1990) examined the relationship for a sample of developing countries and found no evidence of the J-curve. Wilson and Tat (2001) did not find any evidence of the J-curve for Singapore. Lai and Lowinger (2002) did not find any evidence of the J-curve for Japan. Bahmani-Oskooee and Ratha (2004) considered 18 major trading partners of the United States (Australia, Austria, Belgium, Canada, Demark, Finland, France, Germany, Ireland, Italy, Japan, Netherlands, New Zealand, Norway, Spain, Sweden, Switzerland and U.K.) and were unable to discover any J-curve pattern in the short run, although real depreciation of dollar revealed favorable effects on the U.S. trade balance in most cases.

In contrast, Mahdavi and Sohrabian (1993) found evidence of a delayed J-curve for the USA. Demirden and Pastine (1995) also found evidence of the J-curve for the USA. Kale (2001) found evidence of the J-curve for Turkey. Lal and Lowinger (2002) found evidence of the J-curve for a group of 7 East Asian countries (Indonesia, Japan, Korea, Malaysia, the Philippines, Singapore and Thailand). Narayan (2004) concluded that New Zealand's trade balance exhibited a J-curve pattern following a depreciation of the New Zealand dollar. Kulkarni (1994) found the evidence of the J-curve phenomenon for Egypt and Ghana. In this study, Kulkarni also suggested the possibility of a shifting J-curve phenomenon for these countries over time. In another study, Kulkarni and Bhatia (2002) found the evidence of J-curve in six out of seven different countries the authors examined, the Philippines, Kenya, Japan, Indonesia, Mexico, China, and Spain (with the exception of China).

The dynamics of consumption smoothing and capital formation of small-open-economies of LDCs give rise to the S-curve in the presence of productivity shocks only (Senhadji, 1998). For these countries, the trade balance determines the net foreign exchange receipts while the terms of trade determine their purchasing power. Additionally, Bahmani-Oskooee (1986b) found evidence of a W-curve for the U.S. current account using quarterly data for 1973-1985. This describes that subsequent to depreciation of the dollar, the current account deteriorated for two quarters and then started improving for five quarters, again deteriorated and finally improved.

THEORETICAL MODEL AND EMPIRICAL METHODOLOGY

Following Rose and Yellen (1989), demand for imports is given by

[D.sub.m] = [D.sub.m] = [D.sub.m] (Y, [P.sub.m]) and [D.sup.*.sub.m], = [D.sup.*.sub.m] ([Y.sup.*], [P.sup.*.sub.m]) (1)

where [D.sub.m] ([D.sup.*.sub.m]) = the quantity of goods imported by home(foreign) country, Y([Y.sup.*]) = level of domestic (foreign) real income, proxied by the respective industrial output index, [P.sub.m] = relative price of imported goods to domestically produced goods, and [P.sup.*.sub.m] = analogous relative price of imports abroad.

The equations for the supply of exportables are as follows:

[S.sub.x] = [S.sub.x]([P.sub.x]) and [S.sup.*.sub.x] = [S.sup.*.sub.x]([P.sup.*.sub.x]) ... (2)

where [S.sub.x]([S.sup.*.sub.x]) = supply of home (foreign) exportables, [P.sub.x] = ratio of domestic price of exportables to the domestic price level, P ; [P.sup.*.sub.x] = ratio of foreign price currency price of exportables to the foreign price level, [P.sup.*]. Thus, the domestic relative price of imports can be written as

[P.sub.m] = E[P.sup.*.sub.x] / P = REX x [S.sup.*.sub.x] ... (3)

where, E = nominal exchange rate (number of domestic currency units per unit of foreign currency) and REX = real exchange rate. Likewise, the relative price of imports abroad is

[P.sup.*.sub.m] = [P.sub.x] / REX ... (4)

In equilibrium,

[D.sub.m] = [S.sup.*.sub.x] and [D.sup.*.sub.m], = [S.sub.x] ... (5)

The value of the home country's trade balance in real terms is the difference between the value of exports and the value of imports in domestic currency:

TB = [P.sub.x][D.sup.*.sub.m] - REX.[P.sup.*.sub.x] x [D.sub.m] ...(6)

Equations (1) through (6) yield the following reduced form:

TB : TB(REX, Y, [Y.sup.*]). ... (7)

A log-linear variant of equation (7) with time subscript :

In T[B.sub.t], = [[beta].sub.0] + [[beta].sub.1] ln[Y.sub.t] + [[beta].sub.2] ln[Y.sup.*.sub.t] + [[beta].sub.3], ln [REX.sub.t] + [u.sub.t]. ... (8)

where, [u.sub.t] is the random error term and In is the natural logarithmic transformation. Cointegration tests are applied to determine the existence of long-run relationships among the variables. If the series are nonstationary, integrated of order one I(1), and cointegrated, the following estimating error- correction model is specified based on (Engle and Granger, 1987):

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (9)

where, TB = ratio of exports to imports, and [EC.sub.t-1] = error-correction term.

Quarterly data from 1972. I through 2003. IV are employed in this paper. Data source includes IMF's IFS CD ROM, December 2004.

EMPIRICAL RESULTS

Since quarterly world GDP data are not available for the sample period and there are no effective real exchange rate data for Bangladesh, the variables y and [y.sup.*] are excluded from the final estimating regressions (8 and 9). Instead of effective real exchange rate, Taka-Dollar Exchange rate is used for practical conveniences. All the reported estimation results thus involve only two variables (trade balance and bilateral exchange rate).

To begin with, the time series property of trade balance and exchange rate for nonstationarity (unit root) is investigated by implementing the fairly standard Augmented Dickey-Fuller (ADF) and Phillips-Perron tests (Dickey and Fuller, 1979; Phillips and Perron, 1988). The results are reported in Table 1 as follows:

By comparing the estimated numerical values of both ADF and Phillips-Perron test statistics ([MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] and [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII], respectively) with the McKinnon critical values at 5 per cent and higher levels of significance, the null hypothesis of unit root (nonstationarity) cannot be rejected for both trade balance and exchange rate. Moreover, each variable in levels becomes stationary on first-differencing. Consequently, it is inferred that each variable depicts nonstationarity and I (1) behavior.

Next, the bivariate cointegration relationship is investigated by implementing the Engle and Granger (1987) procedure. The results are reported in Table 2 as follows:

The associated t-values of the intercept term and the slope coefficient in the cointegration regression are statistically highly significant as reflected in their respective P-values. The estimated error terms are retrieved from the above cointegration regression to specify the ADF regression for detecting possible bivariate long-run equilibrium relationship as ([MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]). The numerical value of [[??].sub.0] and the associated pseudo t-values are statistically significant. This signifies the existence of a long-run equilibrium relationship between Bangladesh trade balance and TakaDollar exchange rates. For Phillips-Perron test to search for a cointegrating relationship, [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] is considered. This test also clearly confirms the existence of a cointegrating relationship between trade balance and exchange rate (Table 3).

As both variables are nonstationary with first-order integration and cointegrated, the associated error-correction model following Engle and Granger (1987) is estimated. The results capturing the long-run and short-run dynamics are reported as follows (Table 4).

The coefficient of the error-correction term ([EC.sub.t-1]) has the expected negative sign and is statistically highly significant in terms of the associated t-value. This confirms a unidirectional long-run equilibrium (cointegrating) relationship stemming from exchange rate changes to changes in Bangladesh trade balance. The lagged responses of the changes in trade balance and exchange rate reveal bidirectional short-run dynamics. This is confirmed by the F-statistic, as shown above. To add further, the 4-quarter optimum lag length is determined by Akaike information criterion (AIC) as developed in (Akaike, 1969).

Finally, impulse response function as shown below in the Fig. 3 (top right part of the graph) reveals that a depreciation of Bangladesh Taka against U.S. Dollar leads to an initial decline in the export-to-import ratio indicating a deterioration of the trade balance for first two quarters followed by an improvement since the fourth quarter. But the rate of improvement beyond the 8th quarter seems sluggish.

The above finding based on the top right side of Figure 3 weakly suggests the tendency of a J-curve pattern of Bangladesh trade balance with a near-flat-top indicating that the effect of a one-time depreciation gradually diminishes in 8 quarters. This is somewhat consistent with the delayed J-curve phenomenon.

[FIGURE 3 OMITTED]

CONCLUSIONS AND REMARKS

The weak evidence of J-curve for Bangladesh suggests that depreciation of Taka against U.S. dollar causes substantial trade balance deterioration in the short-run relative to the long-run trade balance improvement. So, depreciation of the external value of Taka remains in question to bring a long-term benefit following a significant short-term loss from foreign trade. Perhaps, frequent small doses of depreciation of Taka without appropriate timing and tepid response of trade has something to do with this observed J-curve pattern. To add, there is considerable uncertainty surrounding the J-curve as the size and the timing of the aggregate adjustment of the trade balance depend on the size of the change in the exchange rate (Meade, 1988). However, an understanding of the relationship between the trade balance and the exchange rate is helpful to a pragmatic trade policy formulation.

Evidently, this study does not exhibit any evidence of S-curve despite Bangladesh being a small-open-economy with a very limited access to international financial market for capital formation. Perhaps, this implies that Bangladesh is not unduly vulnerable to external shocks due to an erratic exodus of short-term foreign capital that was in part, responsible for the 1997-98 Southeast Asian Financial Crisis.

The findings of this paper indicate that exchange rate depreciation to cure chronic trade deficit may not be a very useful policy tool for Bangladesh as trade is affected by a host of other factors. Moreover, a sectoral J-curve might portray a better picture of the trade dynamics as compared to the aggregate J-curve since all segments of a country's export sector are not equally responsive to exchange rate adjustments. In closing, this study can be extended to other countries for different sample periods for which comprehensive data in line with the theoretical model are readily available. As a shortcoming, the results of this paper should be interpreted with due cautions.

ACKNOWLEDGEMENT

Thanks to the Editor of this journal and anonymous referees for their helpful suggestions and comments. However, the usual disclaimer applies.

REFERENCES

Akaike, H. (1969), Fitting Autoregression for Prediction, Annals of the Institute of Statistical Mathematics, 21, 243-47.

Bahmani-Oskooee, M. and Ratha, A. (2004), The J-curve Dynamics of U.S. Bilateral Trade; Journal of Economics and Finance, 28, 32-38.

Demirden, T. and Pastine, I. (1995), Flexible Exchange Rates and the J-curve, Economics Letters, 48, 373-7.

Dickey, D.A. and Fuller, W. A. (1979), Distributions of the Estimators for Autoregressive time Series with a Unit Root, Journal of the American Statistical Association, 74, 427-31.

Engle, R. F. and Granger, C. W.J. (1987), Cointegration and Error Correction: Representation, Estimation and Testing, Econometrica, 55, 251-76.

Engle, R. F. and Yoo, B. S. (1987), Forecasting and Testing in Cointegrated Systems, Journal of Econometrics, 35, 143-59.

Kale, P. (2001), Turkey's Trade Balance in the Short and the Long Run: Error Correction Modelling and Cointegration, International Trade Journal, XV, 27-56.

Kulkarni, Kishore (1994), The J-Curve Hypothesis and Currency Devaluation: A Test of Kulkarni Hypothesis with Egypt and Ghana, The Journal of Applied Business Research, reprinted in Kulkarni, Kishore (ed), Readings in International Economics, Serials Publications, New Delhi, India, 2004, pp. 25-38.

Kulkarni, Kishore and Bhatia, Alpana (2002), Empirical Evidence of the J-Curve Hypothesis, Indian Economic Journal, reprinted in Kulkarni, Kishore (ed), Readings in International Economics, Serials Publications, New Delhi, India, 2004, pp. 39-54.

Lal, A.K. and Lowinger, T. C. (2002), The J-curve: evidence from East Asia, Journal of Economic Integration, 17, 397-415.

Magee, S.P. (1973), Currency Contracts, Pass through and Devaluation, Brooking Papers and Economic Activity, 1, 303-25.

Mahdavi, S. and Sohrabian, A. (1993), The Exchange Value of the Dollar and the U.S. trade balance: An Empirical Investigation based on Cointegration and Granger causality tests, Quarterly Review of Economics and Finance, 33, 343-58.

Meade, E. E. (1988), Exchange Rates, Adjustment and the J-curve, Federal Reserve Bulletin, October, 633-44.

Narayan, P. K. (2004), New Zealand's Trade Balance: Evidence of the J-curve and Granger Causality, Applied Economics Letters, 11, 351-54.

Pesaran, M. H. and Shin, Y. (1995), An Autoregressive Distributed lag Modeling Approach to Cointegration Analysis, in Centennial Volume of Rangar Frisch (Eds) S. Strom, A. Holly, and P. Diamond, Cambridge University Press, Cambridge.

Pesaran, M. H., Shin, Y. and Smith, R. J. (2001), Bounds Testing Approaches to the Analysis of Level Relationships, Journal of Applied Economics, 16, 289-326.

Phillips, P.C.B. and Perron, P. (1988), Testing for a Unit Root in time Series Regression, Biometrika, 75, 335-59.

Rose, A. K. and Yellen, J. L. (1989), Is there a J-curve?, Journal of Monetary Economics, 24, 53-68. Rose, A. K. (1990), Exchange Rates and the Trade Balance: Some Evidence from Developing Countries, Economics Letters, 3, 271-5.

Senhadji, A. S. (1998), Dynamics of Trade Balance and the terms of trade in LDCs: the S-curve, Journal of International Economics, 46, 105-31.

Singh, T. (2002), India's Trade Balance: the Role of Income and Exchange Rates, Journal of Policy Modeling, 24, 437-52.

Wilson, P. and Tat, K. C. (2001) Exchange Rates and the Trade Balance: the Case of Singapore 1970 to 1996., Journal of Asian Economics, 12, 47-63.

MATIUR RAHMAN

McNeese State University, Lake Charles

ANISUL M. ISLAM

University of Houston-Downtown, Houston

Table 1
Results of the Unit Root Tests

A. Augmented Dickey-Fuller Tests ([DELTA][y.sub.t] = a + [b.sub.0]
[y.sub.t-1] + [k.summation over (j=1)] [b.sub.j] [DELTA][y.sub.t-j]
+ [e.sub.tk])

Variables Lags Level First Difference

EI 4 -2.88 (I,T) -4.83 * (I)
TB 4 -0.91 (I,T) -7.11 * (I)

B. Phillips-Perron Tests ([DELTA][Y.sub.t] = [[mu].sub.0] +
[[mu].sub.1][y.sub.t-1] + [k.summation over (j=1)] [[mu].sub.j]
[DELTA][y.sub.t-j] + [beta]D + [e'.sub.ik] with D as a dummy variable)

Variables Lags Level First Difference

FI 4 -2.50 (I,T) -7.81 * (I)
TB 4 -1.99 (I,T) -32.94 * (N)

Notes: (1) The McKinnon critical values (with intercept and trend)
are: (a) 1% = -4.07; (b) 5% = -3.46; and (c) 10% = -3.16 respectively.
(2) The McKinnon critical values (with intercept) are: (a) 1% = -3.48;
(b) 5% = -2.88; and (c) 10% = -2.58. (3) The McKinnon critical values
(without intercept and trend) are: (a) 1% = -2.59; (b) 5% = -1.94; and
(c) 10% = -1.62 respectively. (4) (a) * = significant at 1% level; (b)
** = significant at 5% level; and (c) *** = significant at 10% level.
(5) Letters in parentheses after the coefficients represent the
following characteristic included during the unit root tests and in
determining the McKinnon critical values as appropriate: I = Intercept;

T = Trend; and N = None (No Intercept or Trend).

Table 2
Long Run Engle-Granger Bivariate Cointegration Regression of LTBRI
on LEI: (LTB = a + b LEI + e,) or equation (8) with exclusions of
2nd and 3rd terms
 Std.
Variable Coefficient Error t-Statistic Probability

C 3.144406 0.139287 22.57503 0.0000
LEI 0.290471 0.035014 8.295770 0.0000
R-square 0.353248 Mean dependent var 4.286968
Adjusted R-square 0.348116 S.D. dependent var 0.291164
S.E. of regression 0.235084 Akaike info criterion -2.880126
Sum squared resid 6.963308 Schwarz criterion -2.835563
Log likelihood 4.703911 F-statistic 68.819790
Durbin-Watson stat 1.062738 Prob (F-statistic) 0.000000

Table 3
Results of the ADF and Phillips-Perron Tests for Cointegration

A. Augmented Dickev-Fuller Tests *

Variables Lags Coefficients

RES (-1) 4 -4.06

B. Phillips-Perron Tests

Variables Lags Level
RES (-1) 4 -7.94

* The critical values are -3.37, -3.73 and -4.22,
respectively at 10%, 5% and 1% significance levels
(Engle and Yoo, 1987).

Table 4
Short-run Dynamics and the Error Correction Model Based on the
Engle-Granger Cointegration Regression (Estimates of equation 9
with exclusions of 3rd and 4th terms)

 Std.
Variable Coefficient Error t-Statistic Prob.

C 0.002979 0.019461 0.153088 0.8786
DLTBRI(-1) -0.345431 0.115369 -2.994141 0.0034
DLTBRI(-2) -0.195592 0.102073 -1.916200 0.0579
DLTBRI(-3) -0.192990 0.091998 -2.097755 0.0382
DLTBRI(-4) 0.107781 0.081573 1.321289 0.1891
DLEI(-1) -1.217743 0.409561 -2.973286 0.0036
DLEI(-2) -0.468905 0.441390 -1.062339 0.2903
DLEI(-3) 0.112009 0.440499 0.254278 0.7997
DLEI(-4) 1.438840 0.427726 3.363929 0.0010
[EC.sub.t-1] -0.375088 0.111062 -3.377293 0.0010
R-square 0.482640 Mean dependent var 0.004788
Adjusted R-square 0.441434 S.D. dependent var 0.236839
S.E. of regression 0.177007 Akaike info criterion -3.385327
Sum squared resid 3.540458 Schwarz criterion -3.156694
Log likelihood 43.668150 F-statistic 11.712960
Durbin-Watson stat 2.179224 Prob (F-statistic) 0.000000