The Pricing of Foreign Exchange Risk Around the Asian Financial Crisis: Evidence from Taiwan's Stock Market

全文

(1)Journal of Multinational Financial Management 12 (2002) 223–238 www.elsevier.com/locate/econbase. The pricing of foreign exchange risk around the Asian financial crisis: evidence from Taiwan’s stock market Yuanchen Chang * Department of Finance, National Chengchi Uni6ersity, 64 Chi Nam Rd. Sec. 2, Mucha, Taipei 11623, Taiwan Received 13 October 2000; accepted 26 March 2001. Abstract The volatile exchange rate movement during the Asian financial crisis has led global investors to re-evaluate the importance of currency exposures in Asian stock markets. In this paper, we examine industry-level currency risk of Taiwan’s stock market around the Asian financial crisis. The results show that most export-oriented industries, except for the electronics industry, are positively affected by the depreciation of the New Taiwan Dollars (NTD) against the US Dollars (USD). We also find that the magnitude of currency risk is less for banking and electronics industries in the Taiwan Stock Exchange (TSE) than for those in the over-the-counter (OTC) security exchange. Our results are consistent with the findings of Chow et al. (J. Financial Res. 2 (1997b) 191) and have important implications for international investors with exposures in Taiwan’s stock market. © 2002 Elsevier Science B.V. All rights reserved. JEL classification: F3; F23; F31 Keywords: Foreign exchange risk; Asian financial crisis; Taiwan stock market. 1. Introduction The purpose of this paper is to examine industry-level exchange rate risk in Taiwan’s stock market around the Asian financial crisis. Prior researches on the * Fax: + 886-229-393-394. E-mail address: [email protected] (Y. Chang). 1042-444X/02/$ - see front matter © 2002 Elsevier Science B.V. All rights reserved. PII: S 1 0 4 2 - 4 4 4 X ( 0 1 ) 0 0 0 5 2 - 4.

(2) 224. Y. Chang / J. of Multi. Fin. Manag. 12 (2002) 223–238. exchange rate exposure of asset returns focus mainly on the U.S. and Japanese markets (Ceglowski, 1989; Jorion, 1990, 1991; Bodnar and Gentry, 1993; Chow et al., 1997a,b; Choi et al., 1998). The volatile exchange rate movement during the Asian financial crisis has led global investors to re-evaluate the importance of currency exposures in Asian stock markets. At the firm level, Jorion (1990) finds that the exposure of U.S. multinationals is positively related to the percentage of foreign sales. Chow et al. (1997b) point out that the exchange rate exposure of U.S. multinationals is significant related to firm size but not to the relative portion of foreign sales to total sales. They conclude that hedging activities exhibit economies of scales, so the magnitude of exchange rate exposure is less for larger firms than for smaller firms. Choi et al. (1998) show that the pricing of exchange rate risk in the Japanese market varies when different exchange rate measures are used. The exchange rate risk is priced in both weak and strong yen periods using the bilateral yen/USD rates, but the result is mixed using the trade-weighted exchange rates. At the industry-level, Ceglowski (1989) finds that the depreciation of the USD increases the sales of oil extraction, industrial machinery, instruments, transportation and hotel industries, but the construction and durable goods industries are adversely affected during the sample period. Bodnar and Gentry (1993) find that the percentages of industries having significant exchange rate exposure in Canada, USA and Japan are 21, 28 and 35%, respectively during 1979–1988. The variances of the estimated exchange exposure coefficients across industries are larger for Canada and Japan than for USA. They conclude that this result is consistent with the open-economy hypothesis, which indicates that the smaller and more open the economy is, the larger are the inter-industry differences in exchange rate exposure. The currency exposures of emerging stock markets in Asia have been the subject of recent interest on the part of academics and investors alike. This interest is attributable in part to the volatile currency movement during the Asian financial crisis, which has led global investors to realize that ignoring the currency risk can have substantial effects on their portfolio performances. International investors in Taiwan still suffer from few means to hedge local currency risk. The New Taiwan Dollar futures market does not exist, and the scope of the forward exchange market is limited in terms of annual turnover. The lack of currency hedging instruments is a unique feature of Taiwan’s foreign exchange market compared with that in East Asian countries. This may affect how foreign exchange risk is priced. Besides, the economy in Taiwan is typically controlled by small to median size firms with family at the core of the business. These firms may not be constantly on the alert for local currency risk because the Central Bank of China has a long track record of keeping the exchange rate stable. Therefore, they are particularly vulnerable to volatile NTD rates around the Asian financial crisis. Moreover, Taiwan’s stock market is one of the important markets both in Asia and in the world. By dollar value of transaction, the Taiwan stock market (US$1310.2 billion) ranked third in the world behind the New York Stock Exchange (US$5777.7 billion) and the London Stock Exchange (US$1989.5 billion) in 1997. Taiwan is in the process of continued opening of its equity market to foreign investment. The Morgan Stanley Capital.

(3) Y. Chang / J. of Multi. Fin. Manag. 12 (2002) 223–238. 225. International Inc. (MSCI) increased the inclusion weight of Taiwan in the MSCI Emerging Markets Free (EMF) Index from 65 to 80% in November 2000. Since currency risk is an important concern for foreign investors with exposures in emerging markets, our results should provide them with valuable information on how the exchange risk was priced in Taiwan’s stock market. The New Taiwan Dollar (NTD), which was pegged at 40 NTD per USD before 1985, was transferred to a managed float after 1985. The NTD/USD exchange rate was volatile during our sample period, 01/January/1996 – 31/October/1998. It fell from 27.30 to near 27.10 and stayed within 27.10– 28.00 in the beginning of 1996. When the Asian financial crisis started to hit the Taiwanese currency market in July/1997, it rose from 28.00 to 35.00 in 6 months, then gradually fell to near 32.50 in October/1998. It is interesting to investigate the relationship between exchange rate risk and industry characteristics in Taiwan’s stock market during this period. The results can be compared with studies focusing on the U.S. and Japanese markets (Bodnar and Gentry, 1993; Chow et al., 1997b; Choi et al., 1998). Besides, the capitalization requirement for firms listed in the Taiwan Stock Exchange1 (TSE) is higher than that in the over-the-counter (OTC) security exchange. It allows us to test the link between firm size and exchange rate exposure by examining industrylevel currency risk in these two exchanges. This paper has three major contributions to the current literature. First, the exchange rate risk in the TSE, covering the period of the Asian financial crisis, is examined and results can be compared with studies using the U.S. and Japanese data (Bodnar and Gentry, 1993; Chow et al., 1997a,b; Choi et al., 1998). Second, both bilateral and trade-weighted NTD exchange rates are used to test the effects of different currency measures on the pricing of currency risk. Finally, comparison between TSE and OTC industry-level currency exposure in Taiwan’s stock market can shed light on whether firm size is significantly related to currency risk. The rest of the paper is arranged as follows. Data description is provided in Section 2. The empirical design between exchange rate exposure and industry returns is explained in Section 3. Section 4 analyses the empirical results. Section 5 offers concluding remarks.. 2. Data description The data used in this paper consists of daily NTD/USD bilateral exchange rates, trade-weighted effective NTD rates and industry-level stock indexes from 01/January/1996 to 31/October/1998. Returns are defined as the percentage changes in the price level multiplied by 100, [(pt −pt − 1)/pt − 1]× 100. The bilateral NTD/USD exchange rates quoted by the Bank of Taiwan are retrieved from the Taiwan Economic Journal. It is defined as the NTD costs of per USD so that a decrease of the variable represents an appreciation of the NTD. The trade-weighted 1 A minimum capital of 300 million NTD is required for companies to be listed in the TSE, while the minimum requirement for companies to be listed in the OTC market is 50 million NTD..

(4) 226. Y. Chang / J. of Multi. Fin. Manag. 12 (2002) 223–238. effective NTD, calculated from exchange rates of 17 countries,2 is provided by the Council for Economic Planning and Development. It is defined as the foreign exchange cost of per NTD so that an increase of the variable represents an appreciation of the NTD. The daily stock returns, retrieved from the Taiwan Economic Journal, consist of firms listed in the TSE and the OTC securities exchange. Industry returns are market capitalization weighted average returns of firms in each industry. For the TSE market, the 18 industries used in this paper are cement, food, plastics, textiles, electrical machinery, electric wire and cable, chemicals, glass, pulp and paper, iron and steel, rubber, automobile, electronics, construction, shipping, tourist, banking and insurance, and department store. For the OTC market, only returns of the electronics industry and banking and insurance industry are used due to data constraint.3 Summary statistics of the return series are presented in Table 1. It is seen that the variance of TSE market returns is much smaller than that of the OTC market returns. The variances of TSE electronics industry is the highest, followed by the plastic industry in the TSE market. Both the bilateral and trade-weighted exchange returns have large excess kurtosis. The excess skewness and kurtosis of industry returns indicate that they are not normally distributed. Some industries have slow declining autocorrelations for the return series, and a strong serial dependent conditional heteroskedasticity of the return series for all the industries can be seen from the highly significant Ljung and Box (1978) statistics, Q1(12). These statistics test for up to the 12th order serial correlation in the squared returns. They are distributed as 212 under the null hypothesis of i.i.d. returns and have a 95 percentile of 21.03. It is crucial to take into account the statistical properties of these industry returns when we investigate the impact of the exchange rate risk in Taiwan’s stock market. A generalized autoregressive conditional heteroskedasticity (GARCH) regression framework with Student’s t distribution is used to model the strong heteroskedasticity and deviation from the normal distribution of the return series.. 3. Empirical design This paper focuses on the measurement of industry-level exchange risk in Taiwan’s stock market around the Asian financial crisis. A two-factor model, similar to that used by Jorion (1990), Bodnar and Gentry (1993), is employed to estimate the exchange rate risk of industry portfolios. To measure the exchange rate. 2. The countries in the currency basket include the USA, Japan, Germany, Britain, The Netherlands, France, Italy, Canada, Malaysia, Hong Kong, Korea, PRC, Singapore, Indonesia, Australia and Philippine. 3 Most of the industry indexes in the OTC market started from June 1998, except for the electronics industry and banking and insurance industry indexes, which started from June 1996..

(5) Y. Chang / J. of Multi. Fin. Manag. 12 (2002) 223–238. 227. exposure of U.S. industry portfolios, Bodnar and Gentry (1993) add the contemporaneous changes in the exchange rate to the domestic market model and treat the unexpected changes in the exchange rates as news to the stock market. Under the hypothesis of efficient equity market, exchange rate changes will affect realized returns since they provide information about the economic conditions. Another strand of literature shows that conditional covariance of the U.S. stock market is variable over time (Bollerslev et al., 1988, 1992). Bollerslev et al. (1988), estimating a multivariate GARCH model, show that the conditional covariance of the U.S. stock market is time-varying, and it is a significant determinant of the Table 1 Summary statistics of variables Mean. Variance. Skewness. Kurtosis. Q(12). Q1(12). 2.2318. 21.72*. 112.77*. Market and exchange returns, January 1996–October 1998 0.0455 1.9048 Rm (market excess −0.1994 returns-TSE) Rm (market excess 0.0841 3.3657 0.2935 returns-OTC) FX (bilateral) 0.0219 0.1572 1.3880 FX (weighted) −0.0106 0.2148 −0.8622. 1.0943. 23.74*. 169.39*. 23.7559 16.6165. 47.49* 20.53. 271.21* 72.97*. TSE-industry excess returns (Ri ), January 1996–October 1998 Cement −0.0023 1.8831 0.2567 Food 0.0801 1.8691 −0.3246 Plastics 0.0462 3.3528 0.4224 Textiles 0.0260 2.3221 0.1103 Electrical machinery 0.0295 1.6189 −0.2048 Electric wire and cable 0.0315 2.4456 −0.0035 Chemicals 0.0296 2.3298 −0.2366 Glass −0.0131 1.8696 0.2016 Pulp and paper −0.0091 2.6278 0.1513 Iron and steel 0.0210 1.7794 0.4690 Rubber 0.0719 3.1152 −0.2619 Automobile 0.1088 2.1861 0.1386 Electronics 0.1378 5.0912 −0.1905 Construction 0.0413 2.6576 0.2623 Shipping 0.0092 2.2988 0.2068 Tourist 0.0409 2.6897 0.1727 Banking and insurance 0.0094 2.7865 0.5807 Department store 0.0640 1.9820 −0.2177. 2.2980 1.7656 1.8886 1.5766 1.9630 1.7633 1.6619 1.9076 1.5865 2.8015 1.4932 1.9304 0.7249 2.0100 2.2547 1.2488 2.7757 2.0561. 17.68 23.96* 17.38 23.64* 26.85* 12.35 25.51* 18.95 14.99 11.62 14.51 12.52 25.89* 23.45* 12.18 18.05 21.78* 18.34. 59.56* 26.05* 64.65* 23.93* 48.00* 85.22* 111.02* 38.86* 73.14* 82.47* 142.77* 106.04* 317.09* 72.99* 34.61* 71.44* 116.08* 93.88*. OTC-industry excess returns (Ri ), June 1996–October 1998 Electronics 0.1104 3.4469 0.1732 Banking and insurance −0.0400 1.7291 1.3821. 1.1560 5.9407. 38.33* 33.09*. 68.29* 55.88*. Rm is the value-weighted market excess return and Ri is the industry excess return (18 industry indexes for TSE market and two for OTC market). FX is the foreign exchange returns. Skewness and kurtosis statistics are excess skewness and kurtosis values. These values should be zero if series is i.i.d. normal. Q(12) and Q1(12) is the Ljung–Box test statistic for 12 lags returns and squared returns (distributed as 212). * Statistical significance at the 5% level..

(6) 228. Y. Chang / J. of Multi. Fin. Manag. 12 (2002) 223–238. stock return risk premiums. To capture the pervasive conditional heteroskedasticity in the daily observations used in this paper, the impact of exchange rate exposure on index returns is discussed within a GARCH regression framework. The model, where the return of an industry i at time t is a linear function of the return on the market factor and the exchange rate risk factor, is described below: Ri,t =h1 + h2Rm,t +h3FXt +mi,t, mi,t T(0, hi,t ), hi,t = i1 + i2hi,t − 1 +i3FXt +i4m 2i,t − 1.. (1). Estimation is based on the orthogonalized exchange rate return, FXt, which is achieved by running a side regression of the actual percentage change in the exchange rate on the market factor. Rm,t is the value-weighted market excess return and Ri,t is the industry excess return (18 industry indexes in the TSE and two industries in the OTC securities exchange). The excess returns are calculated by subtracting the risk free rate, which is calculated from the interbank over-night rate converted to daily bases. h1 is the constant term and h2 is the market factor beta. h3 is the exchange rate (bilateral NTD/USD or trade-weighted effective exchange rates) coefficient in the mean equation. i1/[1−(i2 + i4)] is the unconditional variance and i3 is the exchange rate coefficient in the conditional variance equation. The specification for the conditional variance simplifies to a GARCH (1, 1) model when the coefficient of the exchange risk factor is equal to zero. The error terms in Eq. (1) are assumed to follow a Student’s t distribution to account for the apparent deviation of return series from the normal distribution. It is equivalent to a normal distribution as the degree of freedom is approaching infinite. Following Choi et al. (1998), both bilateral and trade-weighted exchange rates are used to examine the effects of different currency measures on the pricing of currency risk. Besides, it is suggested in the literature that firm size is significantly related to currency risk (Nance et al., 1993; Chow et al., 1997b). Nance et al. (1993) argue that large firms are more likely to hedge exchange rate risk so the degree to which a firm is exposed should be negatively related to its size. We compare industry-level currency exposures in the TSE and OTC exchanges and test the robustness of the empirical results over the entire sample period and the two sub-periods. The first sub-period is when the currency market was less volatile and the second sub-period is when the Asian currency crisis became more acute.. 4. Estimation results The results of maximum likelihood estimation of the GARCH regression models for the TSE industries over the entire sample period are summarized in Table 2. Tables 3– 5 report results for the comparison of exchange rate risk between the electronics industry and banking and insurance industry over the entire sample period and two sub-periods..

(7) h2. Panel B: multilateral trade-weighted Cement −0.0655 (−2.17)** Food 0.0587 (2.21)** Plastics −0.0332 (−0.88) Textiles −0.0231 (−0.85) Electrical −0.0134 (−0.58) machinery. exchange rates 0.6952 (35.76)** 0.7957 (41.75)** 0.9539 (37.66)** 0.9361 (48.37)** 0.7471 (43.25)**. Panel A: bilateral NTD/USD exchange rates Cement −0.0562 (−1.85)* 0.6967 (34.00)** Food 0.0618 (2.36)** 0.7929 (40.90)** Plastics −0.0338 (−0.89) 0.9522 (37.65)** Textiles −0.0230 (−0.85) 0.9343 (48.82)** Electrical −0.0124 (−0.54) 0.7510 (42.45)** machinery Electric wire −0.0317 (−1.06) 0.8695 (42.13)** and cable Chemicals −0.0031 (−0.09) 0.8649 (38.24)** Glass −0.0780 (−2.39)** 0.6356 (27.43)** Pulp and −0.0694 (−1.85)* 0.7571 (28.70)** paper Iron and steel −0.0308 (−1.14) 0.6426 (32.01)** Rubber −0.0001 (−0.00) 1.0034 (42.70)** Automobile 0.0238 (0.69) 0.6378 (25.37)** Electronics 0.0477 (1.02) 1.1970 (38.82)** Construction −0.0493 (−1.50) 0.8543 (38.67)** Shipping −0.0216 (−0.71) 0.8157 (38.71)** Tourist −0.0440 (−1.12) 0.7809 (28.81)** Banking and −0.0949 (−4.39)** 0.9805 (54.61)** insurance Department 0.0127 (0.49) 0.8176 (45.06)** store. h1. 0.0355 0.0203 0.0277 0.0687 0.0234 0.0485 0.3674 0.0039. (2.34)** (1.87)* (1.68)* (2.08)** (1.82)* (2.57)** (2.64)** (0.71). −0.0204 0.0084 −0.0864 −0.0254 −0.1114. (−0.40) (0.17) (−1.17) (−0.48) (−2.49)**. 0.0713 0.0302 0.1061 0.0281 0.0481. (2.21)** (2.05)** (2.25)** (2.14)** (2.49)**. −0.0732 (−4.67)** 0.0846 (2.73)**. (1.91)* (0.67) (0.11) (−2.38)** (1.12) (1.25) (2.94)** (1.66)*. 0.0384 (1.86)* 0.1351 (1.93)* 0.0342 (1.71)*. (1.95)* (3.14)** (2.25)** (2.07)** (2.22)**. 0.2021 (2.58)** 0.1012 (1.39) 0.1763 (2.05)**. 0.0511 0.0599 0.1056 0.0251 0.0398 0.0454 (2.07)**. (0.01) (1.80)* (0.51) (2.55)** (1.70)**. 0.0657 (0.81). 0.0012 0.1108 0.0503 0.1935 0.1074. i1. (17.08)** (6.07)** (15.16)** (23.56)** (16.12)**. (20.63)** (29.57)** (33.42)** (24.82)** (33.78)** (19.32)** (5.10)** (43.82)**. 0.8129 0.8505 0.7882 0.8511 0.7873. (14.76)** (20.16)** (15.18)** (22.74)** (14.34)**. 0.7252 (10.94)**. 0.8172 0.8581 0.9108 0.8605 0.9013 0.7952 0.5975 0.9035. 0.8808 (21.95)** 0.7804 (8.95)** 0.9031 (30.10)**. 0.8478 (21.17)**. 0.8513 0.7710 0.7883 0.8572 0.8177. i2. hi,t = i1+i2hi,t−1+i3FXt+i4m 2i,t−1. 0.1391 0.0666 0.0111 −0.2765 0.0865 0.1169 0.2769 0.1374. h3. Ri,t = h1+h2Rm,t+h3FXt+mi,t, mi,tT(0, hi,t ),. Table 2 The impact of exchange rate risk on the Taiwanese industries, 01/January/1996–31/October/1998. (0.10) (0.81) (0.08) (−0.03) (0.39). (−0.77) (0.33) (0.53) (−1.03) (−0.13) (−0.27) (0.77) (−0.54). (2.95)** (3.95)** (3.93)** (3.67)** (3.49)**. 0.1564 0.1368 0.0707 0.1075 0.0803 0.1744 0.1636 0.1369. (3.94)** (4.33)** (3.25)** (4.12)** (3.65)** (4.46)** (3.03)** (3.23)**. 0.0802 (2.98)** 0.0948 (2.44)** 0.0764 (3.18)8*. 0.1154 (3.46)**. 0.0948 0.1576 0.1576 0.1167 0.1142. i4. 0.0961 −0.0277 −0.0155 0.0090 −0.0496. (2.51)** (−0.71) (−0.14) (0.19) (−1.18). 0.1174 0.1132 0.1571 0.1188 0.1300. (3.25)** (3.44)** (3.95)** (3.71)** (3.64)**. −0.1077 (−2.32)** 0.1644 (3.36)**. −0.0417 0.0033 0.0277 −0.1244 −0.0084 −0.0241 0.1281 −0.0425. −0.0522 (−1.31) −0.0627 (−0.69) −0.0194 (−0.21). 0.0167 (0.26). 0.0070 0.0434 0.0105 −0.0014 0.0203. i3. (4.42)** (3.11)** (4.22)** (2.95)** (4.35)**. 5.8031 (3.82)**. 6.7242 9.9546 6.0585 8.4873 6.5188. (4.46)** (3.41)** (3.71)** (0.10) (3.61)** (3.55)** (3.54)** (7.25)**. 6.1684 10.1074 6.1079 8.6058 6.8655. (4.51)** (3.08)** (4.28)** (2.95)** (4.20)**. 8.0705 (3.31)**. 6.6399 7.9277 5.9991 207.9306 6.7052 7.5037 5.9002 3.0813. 13.3487 (1.89)* 5.6432 (4.45)** 7.3110 (3.85)**. V. Y. Chang / J. of Multi. Fin. Manag. 12 (2002) 223–238 229.

(8) 0.8568 0.6265 0.7595 0.6392 1.0050 0.6411 1.2056 0.8573 0.8137 0.7871 0.9825. 0.8134 (43.02)**. (−0.07) (−2.41)** (−1.85)* (−1.15) (−0.03) (0.72) (0.99) (−1.44) (−0.72) (−1.17) (−4.41)**. (0.37). (37.76)** (27.74)** (28.84)** (32.04)** (43.04)** (26.36)** (38.90)** (39.46)** (38.51)** (29.32)** (55.26)**. 0.8629 (41.63)**. (−1.19). h2. (−0.54) (−4.79)** (−1.04) (−2.02)** (0.84) (−0.35) (1.57) (0.16) (−2.11)** (0.23) (−0.57). −0.0147 (−0.26). −0.0368 −0.1173 −0.0757 −0.1111 0.0630 −0.0240 0.1384 0.0111 −0.1252 0.0150 −0.0338. −0.0030 (−0.05). h3. (1.69)* (2.55)** (1.96)** (2.38)** (1.83)* (1.58) (2.30)** (1.80)* (2.59)** (3.46)** (1.06). 0.0744 (2.46)**. 0.0289 0.2613 0.0418 0.0392 0.0187 0.0182 0.0805 0.0213 0.0510 0.4310 0.0055. 0.0831 (2.63)**. i1. (26.00)** (5.59)** (28.89)** (19.49)** (30.17)** (41.89)** (23.94)** (35.23)** (18.75)** (5.94)** (43.12)**. 0.7563 (11.16)**. 0.8965 0.6363 0.8920 0.8081 0.8619 0.9274 0.8529 0.9059 0.7913 0.5537 0.8995. 0.7831 (14.81)**. i2. hi,t =i1+i2hi,t−1+i3FXt+i4m 2i,t−1. (0.09) (−2.66)** (−0.79) (0.25) (0.51) (1.14) (0.02) (1.29) (0.16) (−2.36)** (−0.14). 0.0339 (0.69). 0.0043 −0.2954 −0.0698 0.0151 0.0318 0.0679 0.0022 0.0713 0.0143 −0.3292 −0.0069. 0.0525 (0.87). i3. (3.08)** (2.71)** (3.27)** (3.92)** (4.31)** (3.21)** (3.96)** (3.54)** (4.39)** (3.08)** (3.31)**. 0.1428 (3.03)**. 0.0743 0.1283 0.0826 0.1632 0.1334 0.0610 0.1093 0.0776 0.1768 0.1795 0.1381. 0.1618 (3.60)**. i4 4.7694 (5.55)**. 8.1491 (3.29)**. 15.8719 (1.55) 5.3356 (4.76)** 7.3046 (3.82)** 6.3542 (4.52)** 8.1539 (3.23)** 6.0919 (3.54)** 240.73 (0.09) 6.7726 (3.59)** 7.0680 (3.74)** 5.3180 (4.03)** 3.1460 (7.15)**. V. Estimation is based on the orthogonalized exchange rate return, FXt. Rm,t is the value-weighted market excess return and Ri,t is the industry excess return (18 industry indexes). h1 is the constant term and h2 is the market factor beta. h3 is the exchange rate coefficient for the mean equation. The bilateral NTD/USD exchange rates are defined as the NTD costs of per USD so that a decrease of the variable represents an appreciation of the NTD. The trade-weighted effective NTD are defined as the foreign exchange cost of per NTD so that an increase of the variable represents an appreciation of the NTD. i1/[1−(i2+i4)] is the unconditional variance and i3 is the exchange rate coefficient for the conditional variance equation. t-Statistic is in the parentheses. V is the degree of freedom for the Student’s t distribution. It is equivalent to a normal distribution as V is approaching infinite and gets very close for V higher than 30. * Statistical significance at the 10% level. ** Statistical significance at the 5% level for two-tailed tests.. Electric wire −0.0354 and cable Chemicals −0.0026 Glass −0.0782 Pulp and paper−0.0690 Iron and steel −0.0311 Rubber −0.0012 Automobile 0.0248 Electronics 0.0470 Construction −0.0469 Shipping −0.0216 Tourist −0.0451 Banking and −0.0960 insurance Department 0.0098 store. h1. Ri,t =h1+h2Rm,t+h3FXt+mi,t, mi,tT(0, hi,t ),. Table 2 (Continued).. 230 Y. Chang / J. of Multi. Fin. Manag. 12 (2002) 223–238.

(9) h2. i1. 0.1213 (1.28) 0.0015 (0.02). 0.1384 (1.57) −0.0338 (−0.57). 0.1018 (2.40)** 0.6084 (1.45). 0.0805 (2.30)** 0.0055 (1.06). −0.1222 (−5.39)** 0.0936 (2.39)** 0.0786 (1.48) 0.0567 (1.06). −0.2765 (−2.38)** 0.0687 (2.08)** 0.1375 (1.66)* 0.0039 (0.71). h3. 0.8093 (19.31)** 0.5760 (5.70)**. 0.8529 (23.94)** 0.8995 (43.12)**. 0.8159 (20.79)** 0.9587 (53.33)**. 0.8605 (24.82)** 0.9035 (43.82)**. i2. 0.1394 (1.33) −0.4652 (−1.56). 0.0022 (0.02) −0.0069 (−0.14). −0.1817 (−2.01)** −0.3136 (−1.59). −0.1244 (−1.03) −0.0425 (−0.54). i3. 0.1608 (3.87)** 0.4881 (1.44). 0.1093 (3.96)** 0.1381 (3.31)**. 0.1574 (3.99)** 0.0590 (1.31). 0.1075 (4.12)** 0.1369 (3.23)**. i4. 7.2965 (3.77)** 2.3472 (8.45)**. 240.73 (0.09) 3.1460 (7.15)**. 7.3838 (3.75)** 2.2708 (11.22)**. 207.93 (0.10) 3.0813 (3.31)**. V. Estimation is based on the orthogonalized exchange rate return, FXt. Rm,t is the value-weighted market excess return and Ri,t is the industry excess return (18 industry indexes). h1 is the constant term and h2 is the market factor beta. h3 is the exchange rate coefficient for the mean equation. i1/[1−i2+i4)] is the unconditional variance and i2 is the exchange rate coefficient for the conditional variance equation. t-Statistic is in the parentheses. V is the degree of freedom for the Student’s t distribution. It is equivalent to a normal distribution as V is approaching infinite and gets very close for V higher than 30. Sample period for the OTC market is from 01/June/1996 to 31/October/1998 and for the TSE market is from 01/January/1996 to 31/October/1998. * Statistical significance at the 10% level. ** Statistical significance at the 5% level for two-tailed tests.. OTC market Electronics 0.0118 (0.23) 0.8870 (29.13)** Banking and −0.2395 (−8.47)** 0.4416 (21.68)** insurance. Panel B: multilateral trade-weighted exchange rates TSE market Electronics 0.0470 (0.99) 1.2056 (38.90)** Banking and −0.0960 (−4.41)** 0.9825 (55.26)** insurance. OTC market Electronics 0.0064 (0.12) 0.8725 (28.66)** Banking and −0.2455 (−8.51)** 0.4621 (24.02)** insurance. Panel A: bilateral NTD/USD exchange rates TSE market Electronics 0.0477 (1.02) 1.1970 (38.82)** Banking and −0.0949 (−4.39)** 0.9805 (54.61)** insurance. h1. Table 3 The comparison of exchange rate risk between industries in the TSE and OTC markets over the entire sample period. Y. Chang / J. of Multi. Fin. Manag. 12 (2002) 223–238 231.

(10) h2. 0.2385 (2.33)** −0.2404 (−5.13)**. 0.5278 (6.14)** 0.4701 (11.58)**. 0.1660 (1.88)* −0.2735 (−6.23)**. 0.5339 (6.66)** 0.4702 (11.75)**. 0.5520 (1.66)* −0.0208 (−0.14). 0.2841 (1.19) −0.0936 (−1.19). 0.9319 (1.09) 0.1246 (0.34). −0.4337 (−1.10) 0.3964 (1.68)*. h3. 0.1118 (1.39) 0.0533 (0.74). 0.0698 (1.74)* 0.0093 (1.14). 0.2161 (1.73)* 0.0996 (1.59). 0.0688 (1.73)* 0.0137 (1.31). i1. 0.7533 (12.41)** 0.9645 (52.90)**. 0.8244 (15.33)** 0.8580 (20.35)**. 0.7472 (10.72)** 0.9614 (56.35)**. 0.8258 (15.15)** 0.8600 (21.56)**. i2. 0.8035 (1.46) −0.5035 (−0.93). −0.0149 (−0.09) 0.0093 (0.17). 1.4956 (1.03) 2.1111 (1.99)**. 0.1188 (0.23) 0.1898 (0.89). i3. 0.2471 (3.19)** 0.0745 (1.06). 0.1476 (3.15)** 0.1936 (2.42)**. 0.2208 (2.89)** 0.0296 (1.01). 0.1481 (3.10)** 0.2144 (2.18)**. i4. 9.6767 (1.87)* 2.2263 (10.81)**. 95.1312 (0.15) 3.1291 (5.31)**. 13.9052 (1.11) 2.4248 (7.17)**. 107.14 (0.13) 2.8576 (5.74)**. V. Estimation is based on the orthogonalized exchange rate return, FXt. Rm,t is the value-weighted market excess return and Ri,t is the industry excess return (18 industry indexes). h1 is the constant term and h2 is the market factor beta. h3 is the exchange rate coefficient for the mean equation. i1/[1−(i2+i4)] is the unconditional variance and i3 is the exchange rate coefficient for the conditional variance equation. t-Statistic is in the parentheses. V is the degree of freedom for the Student’s t distribution. It is equivalent to a normal distribution as V is approaching infinite and gets very close for V higher than 30. The first sub-period for the OTC market is from 01/June/1996 to 30/June/1997 and for the TSE market is from 01/January/1996 to 30/June/1997. * Statistical significance at the 10% level. ** Statistical significance at the 5% level for two-tailed tests.. OTC market Electronics Banking and insurance. Panel B: multilateral trade-weighted exchange rates TSE market Electronics 0.1022 (1.67)* 0.8834 (17.98)** Banking and −0.0761 (−3.15)** 1.0623 (46.10)** insurance. OTC market Electronics Banking and insurance. Panel A: bilateral NTD/USD exchange rates TSE market Electronics 0.1003 (1.59) 0.9075 (17.86)** Banking and −0.0723 (−3.04)** 1.0316 (41.84)** insurance. h1. Table 4 The comparison of exchange rate risk between industries in the TSE and OTC markets in the first sub-period. 232 Y. Chang / J. of Multi. Fin. Manag. 12 (2002) 223–238.

(11) h2. 0.0617 (1.29) 0.2038 (1.48). i1. 0.0786 (0.86) −0.0330 (−0.49). 0.1747 (1.72)* −0.0246 (−0.31). 0.0952 (1.40) 0.4065 (1.66)*. 0.0745 (1.40) 0.2919 (1.90)*. −0.0995 (−2.41)** 0.1709 (1.88)* 0.0641 (3.42)** 0.3729 (1.68)*. −0.1663 (−1.53) 0.0340 (0.43). h3. 0.8556 (12.20)** 0.5083 (2.94)**. 0.8839 (18.26)** 0.6369 (4.39)**. 0.7749 (8.83)** 0.5398 (3.31)**. 0.8912 (18.79)** 0.7257 (4.94)**. i2. −0.0509 (−0.42) 0.1075 (0.70). −0.0834 (−0.68) −0.3642 (−3.67)**. −1.1711 (−1.96)** −0.1788 (−1.27). −0.0852 (−0.77) −0.0533 (−0.42). i3. 0.0786 (1.85)* 0.3352 (1.68)*. 0.0747 (2.50)** 0.1411 (1.81)*. 0.1208 (2.12)** 0.3127 (1.72)*. 0.0758 (2.44)** 0.1268 (1.80)*. i4. 5.7640 (3.30)** 2.7056 (6.05)**. 79.6599 (0.21) 3.7178 (3.84)**. 5.4091 (3.66)** 2.7346 (6.07)**. 29.8112 (0.67) 3.5662 (3.87)**. V. Estimation is based on the orthogonalized exchange rate return, FXt. Rm,t is the value-weighted market excess return and Ri,t is the industry excess return (18 industry indexes). h1 is the constant term and h2is the market factor beta. h3 is the exchange rate coefficient for the mean equation. i1/[1−i2+i4)] is the unconditional variance and i3 is the exchange rate coefficient for the conditional variance equation. t-Statistic is in the parentheses. V is the degree of freedom for the Student’s t distribution. It is equivalent to a normal distribution as V is approaching infinite and gets very close for V higher than 30. The second sub-period for the OTC market is from 01/July/1997 to 31/October/1998 and for the TSE market is from 01/July/1997 to 31/October/1998. * Statistical significance at the 10% level. ** Statistical significance at the 5% level for two-tailed tests.. OTC market Electronics −0.0682 (−1.20) 0.9810 (30.12)** Banking and −0.2192 (−5.96)** 0.4356 (18.52)** insurance. Panel B: multilateral trade-weighted exchange rates TSE market Electronics 0.0442 (0.61) 1.4337 (32.69)** Banking and −0.1687 (−3.79)** 0.8421 (31.42)** insurance. OTC market Electronics −0.0690 (−1.23) 0.9583 (29.55)** Banking and 0.2124 (−5.76)** 0.4442 (18.95)** insurance. Panel A: bilateral NTD/USD exchange rates TSE market Electronics −0.0411 (−0.77) 1.4176 (33.26)** Banking and 0.0036 (0.08) 0.8528 (30.97)** insurance. h1. Table 5 The comparison of exchange rate risk between industries in the TSE and OTC markets in the second sub-period. Y. Chang / J. of Multi. Fin. Manag. 12 (2002) 223–238 233.

(12) 234. Y. Chang / J. of Multi. Fin. Manag. 12 (2002) 223–238. 4.1. GARCH regression results during the period 01 /January/1996 – 31 /October/1998 The GARCH regression results of 18 industries over the entire sample are presented in Table 2. Parameter estimates when bilateral NTD/USD is used in Eq. (1) are reported in Panel A of Table 2. The bilateral NTD/USD exchange rates are defined as the NTD costs of per USD so that a decrease of the variable represents an appreciation of the NTD. It is seen that the electronics and rubber industries have market factor betas (h2) that are higher than one. Sixteen out of 18 industries have positive exchange rate parameters (h3) during the sample period. Eight industries, including food, textiles, electrical machinery, chemical, pulp and paper, iron and steel, tourist, and banking and insurance industries, have statistically significant exchange rate coefficients (h3) at the 10% level. Positive h3’s indicate that these industries benefit from a depreciation of the NTD against the USD. On the other hand, the electronics and department store industries have negative and significant exchange rate coefficients (h3) in the mean equation, which indicates that these industries suffer from a depreciation of the NTD against the USD. It could be the results of higher costs of imported inputs caused by the depreciation of the NTD. The sums of all the variance parameters, i2 + i4, are close to one, which indicate a near-integrated conditional variance process. The degree-of-freedom parameters for the Student’s t distribution are small, except for those in the electronics industry. It suggests that most of the underlying distributions of the industry returns are not normally distributed. Though most of the GARCH (1, 1) conditional variance parameters are highly significant, the exchange rate coefficients (i3) in the conditional variance equations are small and insignificant. The only exception is the conditional variance of the department store industry, which is significantly reduced by the exchange rate factor. Panel B of Table 2 shows parameter estimates when the trade-weighted exchange rate is used in Eq. (1). The trade-weighted effective NTD is defined as the foreign exchange cost of per NTD so that an increase of the variable represents an appreciation of the NTD. Four industries, including electrical machinery, glass, iron and steel, and shipping industries, have negative and significant exchange rate coefficients (h3) at the 10% level in the mean equations. Because the definition of the trade-weighted exchange rate is the foreign exchange cost of per NTD, negative coefficients indicate that these industries benefit from a depreciation of the NTD against a basket of currencies. It is seen that both electrical machinery industry and iron and steel industry benefit from a depreciation of the NTD regardless of the exchange rate measures used. The conditional variances for the glass and tourist industries are significantly reduced by exchange rate changes. The conditional variance of the cement industry is significantly increased by exchange rate changes when the trade-weighted exchange rate is used..

(13) Y. Chang / J. of Multi. Fin. Manag. 12 (2002) 223–238. 235. Results for the two sub-periods,4 01/January/1996 –30/June/1997 and 01/July/ 1997–31/October/1998, show that fewer industries (two out of 18 industries) are significantly affected by the NTD rates during the first sub-period when the currency market was less volatile. The results during the second sub-period, when the Asian currency crisis became more acute, provide more evidence on the impact of currency changes in Taiwan’s stock market. It is seen that the textiles, chemical, pulp and paper, and tourist industries are positively affected by a depreciation of the NTD against the USD, but the department store industry is adversely affected by it. When the trade-weighted exchange rate is used as currency risk factor, the plastics, electrical machinery, electric wire and cable, glass, pulp and paper, iron and steel, shipping industries benefit from a depreciation of the NTD, but the electronics industry suffers from it. In short, for the entire sample and the second sub-period, export-oriented industries, such as textile and tourist, benefit from a depreciation of the NTD against the USD. Import-oriented industries, such as department store, suffer from the same circumstance. Electronics industry also seems to be adversely affected by the NTD depreciation against the USD. It could be the results of higher costs of imported components due to NTD’s depreciation. In addition, it is seen that the construction and banking and insurance industries, which mainly focus on the domestic market, are less affected by exchange rate changes in the sample period. These results support the hypothesis that industry’s link with the international environment affects the exchange rate exposure of an industry in Taiwan’s stock market.. 4.2. Comparisons between TSE and OTC industry-le6el currency risk To further explore the determinants of the exchange rate risk in Taiwan’s stock market, we applied the same methodology to industry returns in the OTC market to shed light on whether the magnitude of exchange rate exposure is less for larger firms than for smaller firms. The empirical results over the entire sample period are summarized in Table 3, followed by results in Tables 4 and 5 for the two sub-periods. Panel A of Table 3 shows the results over the entire sample period when bilateral NTD/USD rates are used as currency risk factor. It can be seen that exchange risk coefficients are statistically significant at the 5% level for the electronics and banking and insurance industries in both markets. In addition, the exchange risk coefficients for the conditional mean of electronics industry is negative while the exchange risk coefficients for the conditional mean of banking and insurance industry are positive. This indicates that the electronics industry is adversely. 4. Results for the sub-periods are not reported but are available upon requests..

(14) 236. Y. Chang / J. of Multi. Fin. Manag. 12 (2002) 223–238. affected by the depreciation of the NTD, while the banking and insurance industry benefits from it. Panel B in Table 3 shows the results over the entire sample period when trade-weighted exchange rates are used as currency risk factor. There is no significant exchange rate coefficient in the conditional mean and variance equations at the 10% level for both markets. The signs of currency risk factor confirm the results in Panel A that the electronics industry is adversely affected by the depreciation of the NTD while the banking and insurance industry benefits from it. Because the definition of the trade-weighted exchange rate is the foreign exchange cost of per NTD, positive coefficients indicate that this industry suffer from a depreciation of the NTD against a basket of currencies. The less significant results could be caused by the fact that some Asian countries included in the currency basket were also affected by the Asian financial crisis; therefore, the trade-average effective exchange rates were less volatile than the bilateral NTD/USD rates during the sample period. The findings in Table 4 indicate that there are fewer cases with statistically significant currency risk coefficients in the first sub-period when the NTD rates were less volatile. Panels A and B show that the exchange risk coefficients for the TSE banking and insurance industry and the OTC electronics industry are significant at the 10% level but not at the 5% level. The signs of risk factor confirm that electronics industry is adversely affected by the depreciation of the NTD, while the banking and insurance industry benefits from the depreciation of the NTD. We also find that the conditional variance of the OTC banking and insurance industry is significantly affected by the exchange rate risk at the 5% level. Table 5 presents the results during the period 01/July/1997 – 31/October/1998, when the Asian currency crisis became more acute. Panel A of Table 5 reports the parameter estimates when the bilateral NTD/USD rates are used as currency risk factor. Though currency risk has no statistically significant effects on the conditional means of both industries at the 5% level in the TSE market, it has strong and significant impact on both industries in the OTC market. This evident is consistent with the hypothesis that the magnitude of exchange rate exposure is less for larger firms than for smaller firms. The signs of risk factor also confirm the results that electronics industry is adversely affected by the depreciation of the NTD while the banking and insurance industry benefit from it. Panel B of Table 5 shows the results in the second sub-period when tradeweighted exchange rates are used as currency risk factor. It is seen that only the conditional mean of the TSE electronics industry is significantly affected by exchange rate risk at the 10% level. Because the definition of the trade-weighted exchange rate is the foreign exchange cost of per NTD, positive coefficients indicate that this industry suffer from a depreciation of the NTD against a basket of currencies. The signs of the currency risk factor are consistent with the results in Panel A of Table 5..

(15) Y. Chang / J. of Multi. Fin. Manag. 12 (2002) 223–238. 237. 5. Summary and conclusions The Asian financial crisis has led global investors to realize that ignoring currency risk in Asian stock markets can have important effects on their portfolio performances. This paper examines industry-level exchange rate exposure of Taiwan’s stock market and the results can be compared with studies focusing on the U.S. market (Bodnar and Gentry, 1993) and on the Japanese market (Choi et al., 1998). It is seen that over 50% of industries have statistically significant currency exposure over the entire sample period when bilateral NTD/USD rates are used as currency risk factor. Our results show that the department store and electronics industries are adversely affected by the depreciation of the NTD. On the other hand, the textiles, chemical, pulp and paper and tourist industries benefit from a depreciation of the NTD during the Asian financial crisis. The less significant results based on trade-weighted exchange rates could be caused by the fact that some Asian countries included in the currency basket were also affected by the Asian financial crisis. Therefore, the trade-average effective exchange rates are less volatile than the bilateral NTD/USD rates during the sample period. It is consistent with the results reported by Choi et al. (1998) using data in the Japanese stock market. By examining the exposures of the electronics and banking and insurance industries in the TSE and OTC exchanges, we show that there is a negative relationship between firm size and currency exposures in Taiwan’s stock market. The empirical findings are in accordance with the hypothesis that the exchange risk is less for larger firms than for smaller firms (Nance et al., 1993; Chow et al., 1997b). The implication of our results for global investors is to reduce the weights of industries such as department store and electronics, especially the OTC electronics industry, in their portfolio when the NTD depreciated dramatically.. Acknowledgements I thank Robert C. Fowk and Chinese Finance Association (1999) conference participants for their helpful comments.. References Bodnar, G.M., Gentry, W.M., 1993. Exchange exposure and industry characteristics: evidence form Canada, Japan, and the USA. Journal of International Money and Finance 12, 29 – 45. Bollerslev, T., Chou, R.Y., Kroner, K.F., 1992. ARCH modeling in finance: a review of the theory and empirical evidence. Journal of Econometrics 52, 5 –59. Bollerslev, T., Engle, R.F., Wooldridge, J.M., 1988. A capital-asset pricing model with time-varying covariances. Journal of Political Economy 96, 116 – 131. Ceglowski, J., 1989. Dollar depreciation and U.S. industry performance. Journal of International Money and Finance 8, 233 –251..

(16) 238. Y. Chang / J. of Multi. Fin. Manag. 12 (2002) 223–238. Choi, J.J., Hiraki, T., Takezawa, N., 1998. Is foreign exchange risk priced in the Japanese stock market? Journal of Financial and Quantitative Analysis 33, 361 – 381. Chow, Edward H., Lee, W., Solt, M., 1997a. The exchange-rate risk exposure of asset returns. Journal of Business 70, 105 –123. Chow, Edward H., Lee, W., Solt, M., 1997b. The economic exposure of U.S. multinational firms. Journal of Financial Research 2, 191 –210. Jorion, P., 1990. The exchange-rate exposure of U.S. multinationals. Journal of Business 63, 331 – 345. Jorion, P., 1991. The pricing of exchange rate risk in the stock market. Journal of Financial and Quantitative Analysis 26, 363 –374. Ljung, G.M., Box, G.E.P., 1978. On a measure of lack of fit in time series models. Biometrica 66, 297 – 303. Nance, D.R., Smith, C.W. Jr., Smithson, C.W., 1993. On the determinants of corporate hedging. Journal of Finance 48, 391 –405..

(17)