The Effects of Exchange Rate Variability on Malaysia s Disaggregated Electrical Exports Koi Nyen Wong * Tuck Cheong Tang *,+ * School of Business, Monash University Malaysia, 2 Jalan Kolej, Bandar Sunway, 46150 Petaling Jaya, Selangor Darul Ehsan, Malaysia. + Corresponding author. E-mail:- tang.tuck.cheong@buseco.monash.edu.my 1
The Effects of Exchange Rate Variability on Malaysia s Disaggregated Electrical Exports Abstract This paper examines the influence of exchange rate variability on the demand for Malaysia's top five electrical exports by SITC (Standard International Trade Classification) product groups. The empirical results indicate that foreign income and prices are important determinants of export demand for all of the five electrical exports, in both the long run and the short-run over the period of 1990-2001. Most importantly, this paper supports the view that exchange rate variability has an adverse effect on Malaysia s electrical exports. This paper is important to policy makers for the design of both exchange rate and trade policies to enhance export growth that can lead Malaysia s transition towards hightechnology industrialization. Keywords: Exchange rate volatility; Electrical exports; Malaysia; JEL classification codes: C32; F14 I. INTRODUCTION The recent experience of the Asian currency crisis has generated some research interests whether exchange rate variability has any adverse effect on Malaysia s exports, which are dominated by electrical and electronics (E&E) sub-sectors. In 2005, the E&E industry contributed 66% and 53% to total manufactured exports and total gross exports respectively (Malaysia, 2006). According to the Ninth Malaysia Plan (Malaysia, 2006), the electrical and electronics industries are targeted to grow at an average annual rate of 4.1% and 7.7% respectively during the Plan period from 2006 to 2010. Hence, the exports of E&E products play a substantial role in the Malaysian economy in terms of the so-called export led growth strategy through the channels of foreign exchange earnings, employment generation and contribution to gross domestic product (GDP) growth. The objective of this paper is to estimate the disaggregated industry-based export demand function of the Malaysia s top five electrical exports by SITC (Standard International Trade Classification) product groups. More precisely, this paper contributes to the empirical literature by examining the influence of exchange rate variability on electrical exports since 2
the international empirical evidence is contentious (Arize, 1996a and 1996b; Abbott et al., 2001; Bredin et al., 2003). One potential deficiency in conventional export demand model is that it does not consider the effect of exchange rate variability on export flows. Lanyi and Suss (1982) pointed out that the uncertainty which exchange rate movements introduce to international transactions varies with the time dimension of economic decisions involved. For instance, firms involved in international transactions can minimize or avoid the uncertainty in the short run by taking out cover in the forward exchange markets. However, hedging against adverse exchange rate movements is not sufficient to eliminate the effect of variability in the long run because the maturity of the forward contracts is relatively short (Caporale and Doroodian, 1994) and firms are not able to predict their foreign exchange receipts and payments accurately (Bureau of Industry Economics, 1991). On the other hand, the available evidence is limited for both disaggregated studies and Malaysia s electrical export sector. The above studies highlight that failure to include exchange rate variability variable can potentially result in misspecification of the export demand model and hence, produce biased results. Following Abbott et al. (2001), the ARDL (Autoregressive Distributed Lag) approach for cointegration is employed in this study. The structure of this paper is as follows. Section II presents the specification of the export demand model using a cointegration approach advocated by Pesaran et al. (2001), which is based on ARDL specification. It also provides a description of the data. Section III reports and analyzes the econometric results. Policy implications and concluding remarks are presented in Section IV. II. MODEL SPECIFICATION AND DATA To reduce the possible risk of aggregation bias and to capture the sub-sectoral effects which may occur at the industry level, demand for electrical exports is specified based on the top five SITC product groups, which takes the following form (see Arize, 1996a and 1996b; Arize et al., 2000; Abbott et al. 2001): lnq SITC t = β 0 + β 1 lnrp t + β 2 lnyf t + β 3 lnev t + e t (1) 3
where Q SITC is export demand at SITC 3- and 4-digit level from the electrical industry; RP t is the relative price of Malaysia s export at SITC 3- and 4-digit level. The relative price is the domestic price e.g. of the electrical industry (P SITC ) deflated by the price of similar products produced by Malaysia's trading partners (PW); YF is foreign income; EV is the exchange rate variability which is constructed based on the moving-sample standard deviation of real effective exchange rate (REER); and β 1, β 2 and β 3 are the parameters for price, income and exchange rate variability elasticities. According to export demand theory, the estimated β 1 is expected to be negative because holding other things constant, the higher is the price of Malaysia s exports relative to the prices of similar goods made in other countries, and the lower is the quantity demanded for Malaysia s exports. Therefore, RP t is expected to be negatively related to Q SITC. On the other hand, β 2 is expected to be positive because other things being equal, if the income from the rest of the world has increased, the greater is the demand by foreign consumers for Malaysia s made goods. For example, an economic boom in the economies of Malaysia s major trading partners tends to increase their quantity demanded for its exports. Thus, YF tends to be positively related to Q SITC. The expected sign of β 3 is still contentious because there is mixed evidence on the impact of exchange rate variability on export flows reported by previous studies (Arize, 1996a and 1996b; Hassan and Tufte, 1998; Arize et al., 2000), while empirical findings by Needham (1986), Bailey et al. (1987), Asseery and Peel (1991), and Abbott et al. (2001) failed to detect a significant association between exchange rate variability and trade flows. Homogeneity is imposed on equation (1) by expressing export demand as function of relative prices rather than specifying two separate price terms because the former could reduce collinearity between the price terms and conserve the degrees of freedom. We choose to impose homogeneity as a maintained hypothesis because the economic prior restrictions on these variables cannot be tested as disaggregated unit export prices are not available from published sources for Malaysia's competitors in the electrical export market. 4
The sample period covers quarterly observations from 1990:1 to 2001:4, and has been determined largely by the availability of the unpublished data of electrical exports by SITC product group provided by the Department of Statistics, Malaysia. The data for YF, PW and EV have been collected from International Financial Statistics Yearbook (various issues) (see Appendix for the definition and transformation of the variables). Given the small sample size of 44 observations over the period 1990-2001, the bounds test approach developed by Pesaran et al. (2001), which is based on the estimation of an ARDL, is most appropriate for cointegration analysis and a set of new critical values for small sample as small as 30 observations is tabulated in Narayan (2005). Moreover, this testing procedure for cointegration or long-run relationship does not require pre-testing for the order of integration of each variable because it incorporates both I(0) and I(1) independent variables. According to Abbott, et al. (2001), the exchange rate variability is I(0). We find that EV is inclusive on its stationarity. KPSS (Kwiatkowski, Phillips, Schmidt and Shin) tests suggest that EV is stationary at 5 percent level, while PP (Phillips-Perron) tests reject this finding. Hence, as suggested by Abbott, et al. (2001) the ARDL approach is appropriate and its representation for equation (1) can be written as lnq SITC t = a 0 + p b 1j lnrp t-j + j= o p b 2j lnyf t-j + j= 1 p b 3j EV t-j j= 1 + p b 4j lnq SITC t-j-1 + c 1 lnrp t-1 + c 2 lnyf t-1 + c 3 EV t-1 j= 1 + c 4 lnq SITC t-1 + u t (2) In order to determine whether there is a long-run relationship in equation (2), we can test the null hypothesis of no long-run relationship (i.e. H 0 : c 1 = c 2 = c 3 = c 4 = 0) against the alternative hypothesis of long-run relationship (i.e. H A : c 1 0, c 2 0, c 3 0, c 4 0) using F-statistic (Wald test). If the computed F-statistic is greater than the upper critical bound as tabulated in Nayaran (2005), then the null hypothesis can be rejected. On the other hand, if the computed F-statistic is less than the lower critical bound, the test fails to reject the null hypothesis. If the test statistic lies within the critical bounds, conclusive inference can only be made once the order of integration of the underlying regressors is known. 5
III. EMPIRICAL RESULTS As general rule of using quarterly data, a maximum lag length of 4 quarters was imposed on equation (2) initially. The parsimonious ARDL was selected based on Hendry s general-tospecific modeling strategy using the t-statistic on the estimated coefficients all those variables that have relative small absolute t-value (less than one) were dropped sequentially. The main advantage of this modeling strategy is that the final model is supported by data and is free of omitted variable bias (see Gilbert, 1989). Table 1 reports the results of parsimonious ARDL for each SITC product group. The estimation results suggest that the fit of each regression is reasonably good and behave well in terms of diagnostic tests for ordinary least squares assumptions relating to error processes such as absence of serial correlation, normality and homoscedasticity. The RESET statistics indicate no evidence of incorrect functional form, and the plots of CUSUM and CUSUMSQ tests suggest regressions are stable at 5 per cent significance level (refer to Figure 1). Table 1. Final ARDL estimates of Malaysia s top five electrical exports by SITC product group Regressor: Q 772 Q 7415 Q 773 Q 716 Q 775 lnq SITC t-1-1.127*** (-4.333) 0.044 (0.187) -0.266* (-1.818) -0.017 (-0.195) -0.631*** (-5.396) lnrp t-1-1.014*** (-4.185) 1.017 (0.952) -0.602*** (-2.981) -0.146 (-1.142) -0.777*** (-4.968) lnfy t-1 6.616*** (4.027) -4.198*** (-2.94) 0.896 (1.37) -1.33*** (-3.016) 2.215*** (4.051) EV t-1-3.652*** (-4.606) -1.338 (-0.515) -1.897** (-2.79) -1.39*** (-3.951) -1.908*** (-3.183) lnq SITC t-1 0.613** (2.732) -1.034*** (-4.441) -0.159 (-1.539) -0.349** (-2.208) lnq SITC t-2 0.032 (1.457) -1.118*** (-5.523) -0.297* (-1.833) -0.445*** (-3.158) lnq SITC t-3 0.0427*** (2.844) -0.892*** (-4.905) -0.064 (-1.474) lnq SITC t-4 0.046 (1.003) -0.222 (-1.464) 0.122*** (3.029) 0.066* (1.896) 0.287*** (4.455) lnrp t -1.031*** (-41.694) -2.545*** (-4.334) -1.105*** (-12.445) -1.027*** (-14.264) -0.975*** (-11.448) lnrp t-1 0.485** (2.452) -3.354*** (-3.233) -0.412** (-2.454) lnrp t-2-3.204*** -0.229-0.6*** 6
(-3.56) (-1.327) (-4.486) lnrp t-3-2.132*** (-2.96) -0.141* (-1.814) lnrp t-4 0.059 (1.218) -0.911* (-1.76) 0.225*** (3.663) lnyf t 2.99*** (3.444) -3.84* (-1.862) lnyf t-1-3.647*** (-3.217) lnyf t-2 3.174** (2.747) -1.289** (-2.121) -3.508** (-2.502) lnyf t-3 6.217*** (4.376) 1.947** (2.101) -1.683** (-2.148) -3.234** (-2.472) lnyf t-4 3.838*** (3.981) 2.172 (1.569) 1.159 (1.362) -1.058 (-1.69) -1.938* (-2.062) EV t 15.952** (2.215) -2.921** (-2.356) EV t-1-2.281* (-1.937) -32.164*** (-3.872) -7.361** (-2.226) -5.477*** (-4.280) EV t-2 8.579 (1.266) 2.334* (1.914) EV t-3-8.071 (-1.227) -2.993 (-1.333) EV t-4 5.759*** (3.575) 4.028 (1.47) 5.113 (1.445) 5.155** (2.376) Constant -5.76** (-2.644) 13.083*** (3.988) 1.33 (0.8124) 5.233*** (4.284) -0.836 (-0.582) R-squared 0.996 0.959 0.935 0.955 0.904 F-stat [p-value] 290.68 [0.000] 26.146 [0.000] 22.219 [0.000] 26.197 [0.000] 16.184 [0.000] Jargue-Bera [p-value] 0.202 [0.904] 4.4829 [0.106] 0.516 [0.773] 1.803 [0.406] 3.16 [0.206] LM test [p-value] 0.612 [0.736] 0.745 [0.689] 0.943 [0.624] 1.195 [0.55] 1.527 [0.466] ARCH test [p-value] 1.431 [0.489] 1.713 [0.425] 1.121 [0.571] 1.033 [0.597] 0.564 [0.754] Ramsey RESET [p-value] 2.547 [0.28] 2.83 [0.093] 9.489 [0.0087] 3.211 [0.201] 0.359 [0.549] Notes: (1) SITC 772 = Electrical apparatus, resistors, other than heating resistors; printed circuits; switchboard and control panels; (2) SITC 7415 = Air-conditioning machinery comprising a motor-drive fan and elements of changing temperature and humidity, parts; (3) SITC 773 = Equipment for distributing electricity; (4) SITC 716 = Rotating electric plants and parts; and (5) SITC 775 = Domestic electrical and non-electrical equipment. *, **, and *** denote significance at 10%, 5% and 1%, respectively. (.) is t- statistic. 7
Figure 1: Plots of CUSUM and CUSUMSQ tests Q 772 15 1.6 10 1.2 5 0.8 0 0.4-5 0.0-10 -15 1997 1998 1999 2000 CUSUM -0.4 1997 1998 1999 2000 CUSUM of Squares Q 7415 15 1.6 10 1.2 5 0.8 0 0.4-5 0.0-10 -15 1997 1998 1999 2000 CUSUM -0.4 1997 1998 1999 2000 CUSUM of Squares Q 773 15 1.6 10 1.2 5 0.8 0 0.4-5 0.0-10 -15 1996 1997 1998 1999 2000 CUSUM -0.4 1996 1997 1998 1999 2000 CUSUM of Squares Q 716 15 1.6 10 1.2 5 0.8 0 0.4-5 0.0-10 -15 1996 1997 1998 1999 2000 CUSUM -0.4 1996 1997 1998 1999 2000 CUSUM of Squares Q 775 15 1.6 10 1.2 5 0.8 0 0.4-5 -10 0.0-15 1996 1997 1998 1999 2000-0.4 1996 1997 1998 1999 2000 CUSUM CUSUM of Squares 8
The results of bounds tests are presented in Table 2. The computed F-statistics for all electrical exports are greater than the upper bound of critical value at 1 per cent level of significance implying that there is a long-run relationship between the demand for Malaysia s electrical exports and their key determinants, viz. relative prices, foreign income and exchange rate variability, with the exception for exports of equipment for distributing electricity (Q 773 ), which lies within inconclusive region. Further cointegration tests using both the trace and maximum eigenvalue test statistics suggest one or at least one cointegrationg vector(s) for the five largest Malaysia s electrical exports, thus corroborating our conclusions based on the bounds tests. The results of Johansen s multivariate cointegration tests are not reported here but are available from authors upon request. Table 2. Bounds tests for cointegration Q 772 Q 7415 Q 773 Q 716 Q 775 F-statistics 6.315 8.082 3.585 14.633 8.298 Critical value:- 1% 5% 10% Lower bound 4.983 3.535 2.893 Upper bound 6.423 4.733 3.983 The critical values are from Narayan (2005, p. 1988), Appendix: Critical values for the bounds test: case III: unrestricted intercept and no trend. Table 3 provides the OLS long-run estimates of price, foreign income and exchange rate variability elasticities. 1 Firstly, both the estimated coefficients for relative prices and foreign income have the expected sign negative and positive respectively, but the price elasticity s magnitude varies among the five electrical product groups ranging from 0.37 (i.e. export demand for domestic electrical and non-electrical equipment - Q 775 ) to 4.22 (i.e. export demand for air-conditioning machinery - Q 7415 ). Given the export demand of the former is price inelastic, exporting firms can adopt various forms of non-price competition such as product quality improvement and product innovation to enhance their export growth. On the other hand, the policy implications of price elastic export demand for the latter are that exporting firms can use price competition (including real devaluation) to maintain or even to 1 According to Abeysinghe and Tan (1999), in small samples OLS may still be the best choice among the six estimation techniques viz. OLS, unrestricted error correction model or autoregressive distributed lag model (ARDL), a fully modified least square, 3-step estimator, OLS regression augmented by leads and lags of the differenced explanatory variables, and Johansen s estimator. 9
increase exports with the emergence of cheaper production bases such as People s Republic of China (PRC) and India, and also when the economies of Malaysia s major trading partners are sluggish in the long run. Table 3. Ordinary Least Squares (OLS) estimates of long-run coefficients Regressor: Q 772 Q 7415 Q 773 Q 716 Q 775 lnrp t -1.016*** (-25.232) -4.220*** (-6.667) -1.220*** (-7.546) -0.546* (-1.991) -0.367* (-1.916) lnfy t 4.247*** (6.233) 6.863*** (6.703) 3.748*** (7.417) 4.223*** (6.840) 2.524*** (4.333) EV t -4.500*** (-4.459) -5.657*** (-3.910) -2.34*** (-3.359) -0.086 (-0.086) 1.041 (0.840) Constant 0.509 (0.168) -8.991*** (-3.337) -3.287* (-1.814) -7.488*** (-2.869) -6.029** (-2.172) *, **, and *** denote significant at 10%, 5% and 1%, respectively. (.) is t-statistic. Secondly, the empirical estimates of long-run foreign income elasticities are high implying the adoption of outward-oriented policy can promote the rapid growth of electrical exports. This evidence is consistent with previous studies by Goldstein and Khan (1985), Margues and McNeilly (1988), Feenstra (1994) and Abbott et al. (2001). Finally, the estimated results strongly indicate that exchange rate variability has a negative effect on electrical exports in the long run especially for Malaysia s top three electrical exports (e.g. electrical apparatus, air-conditioning machinery, and equipment for distributing electricity). From a policy perspective, the central bank of Malaysia can reduce the variability of exchange rate by adopting a crawling-peg exchange rate system, which allows the pegged exchange rate to adjust in order to maintain the purchasing power parity (i.e. to maintain international price competitiveness). Exporters can minimize their exposure to exchange rate risk by (a) decreasing their production activities (b) switching sources of demand and supply (c) increasing the proportion of exports invoiced in home currency and (d) switching some of the borrowings from offshore to onshore. IV. CONCLUSIONS 10
This paper has provided an examination of the long-run relationship between Malaysia s disaggregated electrical exports at 3- and 4-digit SITC level and exchange rate variability over the quarterly period 1990-2001 using the bounds test developed by Pesaran et al. (2001). The cointegration results suggest that there is unique long-run equilibrium relationship for each of the estimated regression. The vector of variables that is cointegrated with the demand for Malaysia s top five electrical exports are relative prices, foreign income and exchange rate variability. However, there are some variations of price elasticity of export demand among the electrical products. For instance, the export demand for Malaysia s three largest electrical exports, namely electrical apparatus, air-conditioning machinery, and equipment for distributing electricity, are price elastic implying that the improvement of price competitiveness is an appropriate strategy if Malaysia wants to increase electrical exports in the light of the emergence of relatively more price-competitive electrical exporters from PRC and India. On the hand, the export demand for Malaysia s fourth and fifth largest electrical exports such rotating electric plants and parts, and domestic electrical and non-electrical equipment, is price inelastic, suggesting that Malaysian producers need to be responsive to non-price factors such as improving the product quality and constantly upgrading the products in order to enhance export growth. With reference to high long-run income elasticities of export demand, they imply that Malaysia has a high degree of exposure to these traditional export markets such as the U.S., Japan and Singapore if they experience a recession. To mitigate these adverse effects, the Malaysian government can set up trade promotion activities to develop new export markets. The findings also strongly indicate the adverse effect of exchange rate variability on Malaysia s top three electrical exports (e.g. electrical apparatus, air-conditioning machinery and equipment for distributing electricity) in the long run. Policy makers can reduce the variability of exchange rate by adopting a crawling-peg exchange rate system while firms involved in international transactions can reduce their exposure to adverse exchange rate movements by using the forward exchange markets. 11
REFERENCES Abbott, A., Darnell, A. and Evans, L. (2001) The Influence of Exchange Rate Variability on UK Exports, Applied Economics Letters, 8, 47-9. Abeysinghe, T. and Tan, K. B. (1999) Small Sample Estimation of a cointegrating vector: an empirical of six estimation techniques, Applied Economics Letters, 6, 645-8. Arize, A.C. (1996a) A Reexamination of the Demand for UK Exports: Evidence From an Error Correction Model, The International Trade Journal, 10, 501-25. Arize, A.C. (1996b) Real Exchange-Rate Volatility and Trade Flows: The Experience of Eight European Economies, International Review of Economics & Finance, 5, 187-205. Arize, A.C., Osang, T. and Slottje, D.J. (2000) Exchange-Rate Volatility and Foreign Trade: Evidence from Thirteen LDC s, Journal of Business and Economic Statistics, 18, 10-7. Asseery, A. and Peel, D.A. (1991) The Effects of Exchange Rate Volatility on Exports, Economics Letters, 37, 173-7. Bailey, M., Tavlas, G.S. and Ulan, M. (1987) The Impact of Exchange-Rate Volatility on Export Growth: Some Theoretical Considerations and Empirical Results, Journal of Policy Modelling, 9, 225-43. Bredin, D., Fountas, S. and Murphy, E. (2003) An Empirical Analysis of Short-run and Longrun Irish Export Functions: Does Exchange Rate Volatility Matter? International Review of Applied Economics, 17, 193-208. Bureau of Industry Economics (1991) Consequences of Exchange Rate Variability: Survey Evidence from Trade-exposed Manufacturers, Research Report 37, Canberra: Australian Government Publishing Service. Caporale, T. and Doroodian, K. (1994) Exchange Rate Variability and the Flow of International Trade, Economic Letters, 46, 49-54. Feenstra, R.C. (1994) New Product Varieties and the Measurement of International Prices, American Economic Review, 84, 157-77. Gilbert, C.L. (1986) Professor Hendry's Econometric Methodology, Oxford Bulletin of Economics and Statistics, 48, 283-307. 12
Goldstein, M. and Khan, M.S. (1978) The Supply and Demand for Exports: A Simultaneous Approach, Review of Economics and Statistics, 60, 275-86. Hassan, M. K. and Tufte, D.R. (1998) Exchange Rate Volatility and Aggregate Export Growth in Bangladesh, Applied Economics, 30, 189-203. Lanyi, A. and Suss, E.C. (1982) Exchange Rate Variability: Alternative Measures and Interpretation, IMF Staff Papers, 29, 527-60. Malaysia (2006) Ninth Malaysia Plan, 2006-20010, Kuala Lumpur: Percetakan Nasional Malaysia Berhad. Marques, J. and McNeilly, C. (1988) Income and Price Elasticities for Exports of Developing Countries, Review of Economics and Statistics, 70, 306-14. Narayan, P. (2005) The Saving and Investment Nexus For China: Evidence From Cointegration Tests, Applied Economics, 37, 1979-90. Needham, N. (1986) The Effect of Exchange Rate Volatility on the Price and Volume of Australia Bilateral Trade, Paper contributed to the 15 th Conference of Economists, Monash University, August. Pesaran, M., Shin, Y. and Smith R. (2001) Bounds Testing Approaches to the Analysis of Level Relationships, Journal of Applied Econometrics, 16, 289-326. 13
APPENDIX The data are expressed in logarithmic terms for the estimation regressions. The data are in indices (i.e.1995 = 100) to ensure all variables are unit free, except for exchange rate variability (EV). The definition of the variables is follows. Q SITC = Volume of electronic export at 3- and 4-digit SITC level. P j SITC = Unit price of electronic export at 3-digit SITC level, which is constructed based on the followings: Unit price of SITC 772 exports = Current FOB value (RM) of SITC 772 exports Quantity of SITC 772 exports where FOB is the abbreviation for free on board, and the export quantity of each SITC product group is measured in common unit. PW t = 3 w n P n t n=1 3 where w n = x n / x n, trade share of Malaysia's n th major trading partners i.e. U.S. n=1 Japan and Singapore, x n = Malaysia's electrical exports to the n th trading partners, and P n t = wholesale price index of Malaysia's n th major trading partners. YF t = 3 w n YF n t n=1 where YF n t = real gross domestic product of Malaysia's n th trading partner, and the weights used for constructing foreign income variables are similar to those used in the construction of the relative price variables. 1 m EV = [ (REER t+i-1 REER t+i-2 ) 2 ] ½ m = i 1 where REER = index of real effective exchange rate (1995=100), and m = 4, which is the order of the moving average. 14
15