Foreign Exchange Exposures, Financial and Operational Hedge Strategies of Taiwan Firms

Foregn Exchange Exposures, Fnancal and Operatonal Hedge Strateges of Tawan Frms AUTHORS ARTICLE INFO JOURNAL Y-Chen Chang, Hu-Ju Ln Y-Chen Chang and Hu-Ju Ln (7). Foregn Exchange Exposures, Fnancal and Operatonal Hedge Strateges of Tawan Frms. Investment Management and Fnancal Innovatons (open-access), 4(3) "Investment Management and Fnancal Innovatons (open-access)" NUMBER OF REFERENCES NUMBER OF FIGURES NUMBER OF TABLES The author(s) 17. Ths publcaton s an open access artcle. busnessperspectves.org

Investment Management and Fnancal Innovatons, Volume 4, Issue 3, 7 95 FOREIGN EXCHANGE EXPOSURES, FINANCIAL AND OPERATIONAL HEDGE STRATEGIES OF TAIWAN FIRMS Y-Chen Chang *, Hu-Ju Ln * Abstract Usng multple-horzon data of Tawan non-fnancal frms durng the perod of 1998-5, ths study examnes fnancal and operatonal hedge strateges of foregn exchange exposures smultaneously. Our emprcal fndngs show that the use of operatonal hedge strateges does not help reduce foregn exchange exposures for Tawan frms. Also, the use of foregn currency dervatves (FCD) s an effectve hedgng strategy n a one-month horzon, but t s less effectve when the horzon lengthens. In addton, the use of foregn currency-denomnated debts (FDD) always ncreases foregn exchange exposures. Key words: foregn exchange exposures, operatonal hedge, foregn currency dervatves, foregn currency-denomnated debt. JEL classfcatons: F31, G3. 1. Introducton Tawan s a small open economy. Frms n Tawan always have been forced to drect most of ther operatons toward foregn countres due to the scarcty of natural resources and the small home markets. Unexpected fluctuatons n foregn exchange rates have been an mportant concern to frms wth nternatonal busness operatons snce future cash flows, and therefore the value of frms wll be affected. Accordng to Marshall (), a total of 87% of Asa Pacfc respondent companes surveyed n hs research rank foregn exchange rsk management as equally or sgnfcantly mportant as busness rsk management. To mtgate the mpact of foregn exchange rate fluctuatons, t has been clamed that frms can employ fnancal hedge strateges through foregn currency dervatves (FCD) and foregn currency-denomnated debts (FDD) 1. Many emprcal studes have proven that frms use FCD or FDD for the purpose of hedgng. Geczy (1997) suggests that frms may use dervatves to reduce cash flow varaton. Allayanns and Ofek (1) fnd a strong negatve assocaton between FCD use and a frm s foregn exchange exposures. Nguyen and Faff (3) fnd that the use of FCD reduces short-term foregn exchange exposures. As the return horzon lengthens, FCD appear to be less effectve n hedgng foregn exchange exposures. Burgman (1996) nterprets the postve relaton between leverage and foregn currency rsk as multnatonal corporaton s (MNC s) use of FDD to hedge currency rsk. Chen, Cheng, He, and Km (1997) fnd that the debt rato s postvely assocated wth the level of foregn operatons, whch provdes the evdence of FDD hedgng foregn currency rsk. The results of Allayanns and Ofek (1) are that exposures through foregn sales are postvely and sgnfcantly related to a frm s decson to ssue foregn debts and the level of foregn debts. Ellott, Huffman, and Makar (3) fnd a postve relatonshp between foregn currency exposures and the level of FDD, ndcatng that FDD may be used as a hedge. Keda and Mozumdar (3) fnd strong evdence that frms ssue FDD to hedge ther exposures both at the aggregate and the ndvdual currency levels. Theoretcal papers argue that operatonal hedge strateges are more effectve n managng long-run exposures, whereas fnancal hedge strateges are more effectve n managng short-run exposures (Logue, 1995; Chowdhry and Howe, 1999). MNCs have the operatng flexblty to shft * Feng Cha Unversty, Tawan. * Feng Cha Unversty, Tawan. 1 Snce foregn currency-denomnated debts (FDD) represent cash outflows n foregn currences, they can be used as hedges when frms have foregn cash nflows, ether from operatons abroad or from exports. Y-Chen Chang, Hu-Ju Ln, 7.

96 Investment Management and Fnancal Innovatons, Volume 4, Issue 3, 7 ther sales and producton operatons among locatons to hedge foregn exchange rsk (Mller and Reuer, 1998; Debrun and Huffman, 1999; Pantzals, Smkns and Laux, 1). Large frms are more lkely to have multple operatons and thus beneft from the natonal hedges assocated wth geographc dversfcaton (Makar, DeBrun and Huffman, 1999). MNCs wth greater network breadth are less exposed to currency rsks, whereas frms wth more hghly concentrated networks (greater depth) are more exposed (Pantzals, Smkns, and Laux, 1).Allayanns, Ihrg, and Weston (1) examne fnancal and operatonal hedge strateges smultaneously. They fnd that frms fnancal hedge strateges are related to lower exposures, but operatonal hedge strateges do not reduce exposures. In ther studes, the use of FCD and FDD are combned as an ndcator varable of fnancal hedge strateges. In our opnon, however, FCD also can be used for speculatve purposes. Lkewse, FDD also can be used for other purposes, such as rasng funds or mprovng a frm s captal structure. Snce frms may use FCD or FDD for dfferent ncentves, we try to separate these two fnancal hedge strateges n our study. Usng multple-horzon data of Tawan non-fnancal frms durng the perod of 1998-5, ths study examnes fnancal and operatonal hedge strateges of foregn exchange exposures smultaneously. We create a measure of foregn exchange exposures for each frm usng a two-factor model as n Joron (199). The absolute value of the estmated exposure s then regressed on the use of fnancal and operatonal hedges, wth such addtonal control varables as the percentage of overall revenues from abroad and the frm sze. We use three proxes for operatonal hedge strateges: breadth, depth, and dsperson. FCD and FDD are separated nto two proxes of fnancal hedge strateges. Our emprcal fndngs show that the use of operatonal hedge strateges does not help reduce foregn exchange exposures for Tawan frms. Also, the use of FCD s an effectve hedgng strategy n one-month horzon, but t s less effectve when the horzon lengthens. In addton, the use of FDD always ncreases foregn exchange exposures. The remander of ths study s structured as follows. Secton descrbes the data. Secton 3 presents the emprcal framework and results. The results are then dscussed n secton 4. Fnally, secton 5 concludes ths study.. Data We select non-fnancal frms lsted on the Tawan Stock Exchange Corporaton (TSEC). Fnancal frms are excluded, as the focus of our study s on end-users rather than producers of fnancal servces. To be ncluded n the sample, frms must have monthly stock returns coverng the perod January 1998 through December 5. Ths selecton crteron results n 36 frms. It s wdely beleved that hgher foregn nvolvement accompanes the hgher foregn exchange exposures. Thus, the degree of hgh foregn nvolvement n our studes s defned as: (1) the frm s rato of foregn sales to total sales (FS/TS) s more than 1%, and () the frm s holdng shares of any foregn subsdares s more than % 1 durng the sample perod. There are 99 frms n our fnal data, accordng to these two standards. Table 1 summarzes the descrptve statstcs of the varables used n ths study. The data show that frms n our sample have szable foregn sales. Table presents the correlaton matrx of the ndependent varables used at the second stage of the regresson. 1 Under the generally accepted accountng prncples of Tawan, f a company holds a moderate shares (-5%) or more of the votng stocks n a foregn corporaton, the nvestment s consdered as a sgnfcant nfluence on the foregn corporaton.

Investment Management and Fnancal Innovatons, Volume 4, Issue 3, 7 97 Descrptve Statstcs of Varables Table 1 N Mnmum Maxmum Mean Std. Devaton Rus 96 -.457.73.16.143 Rm 96 -.15.5.19.816 FSALES 99.756.9915.6581.761 SIZE 99 9.8967 1.6943 1.8437.4653 BREADTH 99 1.45.3587.585 DEPTH 99.4374 1.315.83.15 DISPERSION 99.9688.493.869 Rus s the return on the exchange rate of NTD/USD1; Rm s the return on the market captalzaton-weghted ndex of Tawan, TAIEX; FSALES s foregn sales rato (foregn sales/total sales, FS/TS); SIZE s log (frm s total assets); Breadth s log (number of foregn countres n whch the frm has subsdares); Depth s (number of foregn subsdares n the top two foregn countres)/(number of foregn subsdares); Dsperson s geographc dsperson ndex: K ( Dsperson) 1[( No. subsdares) /( TotalNo. subsdares) ]. j 1 j The Correlaton Matrx between Varables Table FSALES SIZE HEDGE FCD FDD BREADTH DEPTH DISPERSION FSALES 1. SIZE -.75 1. HEDGE.156.33 1. FCD.154.37.968 1. FDD.145.667.398.56 1. BREADTH -.75.87.9.95 -.1 1. DEPTH -.316.34 -.16 -.14 -.46 -.13 1. DISPERSION.175.89.16.14.154.36 -.81 1. Rus s the return on the exchange rate of NTD/USD1; Rm s the return on the market captalzaton-weghted ndex of Tawan, TAIEX; FSALES s foregn sales rato (foregn sales/total sales, FS/TS); SIZE s log (frm s total assets); Breadth s log (number of foregn countres n whch the frm has subsdares); Depth s (number of foregn subsdares n the top two foregn countres)/(number of foregn subsdares); Dsperson s geographc dsperson ndex: K ( Dsperson) 1[( No. subsdares) /( TotalNo. subsdares) ]. j 1 j

98 Investment Management and Fnancal Innovatons, Volume 4, Issue 3, 7 3. Emprcal Framework 3.1. Estmaton of foregn exchange exposures Lke Joron (199) and many subsequent studes, the foregn exchange exposure ( estmated by usng the followng equaton 1 : Rt 1 Rmt Rxt ut, (1) where R t s the return on stock n perod t, and R xt s the percentage change n the exchange rate n perod t. We control for market movements by ncludng the return on the market portfolo n perod t, R 3. u s the error term. mt t The results of equaton (1) show that there are 54 frms postvely exposed ( ) s >) and 45 frms negatvely exposed ( <). The postve exposures mean that stock returns ncreased as NTD deprecated aganst USD. The negatve exposures mean that stock returns ncreased as NTD apprecated aganst USD. Table 3 reports the results of the mean dfferences tests between frms wth postve and negatve exposures. As the tests provde no sgnfcant dfference between them, we wll use full samples nstead of dstngushng them n the followng emprcal tests. Table 3 Comparson of Mean Values for Descrptve Statstcs between Frms wth Postve and Negatve Exposures Full Sample (n=99) Frms Postvely Exposed (n=54) Frms Negatvely Exposed (n=45) Dfference n Means t-test Exposure coeffcent ( ).754.734.685.37 FS/TS.6581.6981.586 1.87 Sze 1.8437 1.17 1.91-1.334 Breadth.3587.498.51 -.657 Depth.83.786.865 -.9983 Dsperson.493.571.4579.75 1. Ths table reports the t-statstc for the mean dfferences test between the samples consstng of frms wth postve and negatve exposures.. Sgnfcance levels are ndcated as follows: ***1%, **5%, *1%. 3.. Cross-sectonal regresson wth fnancal hedge strateges Once the foregn exchange exposure s estmated, the basc relatonshp between the absolute value of the exposure versus foregn nvolvement, proxed by foregn sales to total sales (FS/TS), and fnancal hedge strateges controlled wth the sze effect s then tested usng the cross-sectonal regresson framework 4. 1 One-month horzon data are used here. The exchange rate used here s the US Dollar (USD) n terms of the New Tawan Dollar (NTD). There are at least three reasons to use ths exchange rate. Frst, Tawan s a small and export-orented economy, and the Unted States s one of the largest trade partners of Tawan all the tme. Second, snce the US dollar s a leadng vehcle currency, prces of tradable goods are often denomnated n the US dollar, no matter whch countres Tawan frms trade wth (Chao, Hung and Nwanna, 1). Thrd, the correlaton coeffcent between NTD/USD and the effectve exchange rate for the NTD s about 9%. The effects of NTD/USD seem to domnate other exchange rates. 3 The market portfolo, TAIEX, s a market captalzaton-weghted ndex of Tawan that nvolves all currently lsted common stocks, except newly-ssued stocks and the stocks of fnancally dstressed frms n Tawan. 4 The ndependent varables are not hghly correlated accordng to the correlaton coeffcents shown n Table. In addton, multcollnearty s not a severe problem here snce the varance nflaton factors (VIF) are less than ten (Kennedy, 1998).

Investment Management and Fnancal Innovatons, Volume 4, Issue 3, 7 99 Fnancal hedge strateges nclude foregn currency dervatves (FCD) and foregn currency-denomnated debt (FDD) 1. Lke Allayanns, Ihrg, and Weston (1), we construct an ndcator varable Hedge that sets equal to 1 f frms use FCD or FDD and for non-users. Snce frms may use FCD or FDD for dfferent ncentves, we use another equatons to separate these two fnancal hedge strateges. In equaton (b), FCD s an ndcator varable that sets equal 1 f frms use FCD and for non-users. In equaton (c), FDD s an ndcator varable that sets equal 1 f frms use FDD and for non-users. In addton, we use equaton (d) to consder FCD and FDD smultaneously. A negatve value on the estmated coeffcent for the dummy suggests that fnancal hedges reduce exposures. 1( FS / TS) ( Hedge) 3( Sze), (a) ) 1( FS / TS) ( FCD) 3 ( Sze, (b) ) 1( FS / TS) ( FDD) 3( Sze, (c) ) 1( FS / TS) ( FCD) 3( FDD) 4 ( Sze. (d) Table 4 shows the results of the four equatons 3. Consstent wth the evdence of Allayanns and Ofek (1), we fnd a negatve relatonshp between foregn exchange exposures and Hedge, ndcatng that frms use fnancal strateges as hedges. The negatve and sgnfcant coeffcent on FCD n equaton (b) also ndcates that frms use FCD as hedges. But the postve and sgnfcant coeffcent on FDD n equaton (c) ndcates that foregn exchange exposures ncrease when frms use FDD. When FCD and FDD are separated n equaton (d), the results reman the same. Therefore, FDD s not an effectve nstrument for currency rsk management. The postve and sgnfcant coeffcents on FS/TS ndcate that for a gven exposure, an ncrease n revenue from foregn operatons ncreases foregn exchange exposures. The negatve and sgnfcant coeffcents on the frm sze ndcate that greater frm sze s sgnfcantly assocated wth lower foregn exchange exposures. Table 4 Cross-Sectonal Regresson wth Fnancal Hedge Strateges Dependent Varable: OLS (Ordnary Least Square) (a) (b) (c) (d) Intercept 3.517** 3.131** 4.9546*** 4.864*** FS/TS.458*.4616*.361.3746 Hedge -.14 FCD -.346** -.754* FDD.856***.931*** Sze -.873** -.75** -.471*** -.498*** Adj R-Square.889.895.139.1351 F-Value 3.5** 3.47** 4.1*** 3.58** 1. Ths table reports the cross-sectonal regresson results of equatons (a) to (d) usng one-month horzon data. The absolute value of s used as the dependent varable.. Sgnfcance levels are ndcated as follows: ***1%, **5%, *1%. 1 Compared to the relatvely large percentage of frms that use FCD (79%), only 31% of the frms n our sample use FDD. We use ndcator varables to measure FCD and FDD usages because the reported notonal prncpal amounts are mssng or are just the aggregate data. Detal ed data are dffcult to get. 3 Accordng to the Ramsey Reset test, there s no functonal form msspecfed. Accordng to the Whte test and the Generalzed Durbn-Watson test, there s no heteroskedastcty and autocorrelaton n the resduals of the OLS regresson. After the Kolmogorov-Smrnov test, the resduals are not rejecton of normalty.

1 Investment Management and Fnancal Innovatons, Volume 4, Issue 3, 7 3.3. Cross-sectonal regresson wth fnancal and operatonal hedge strateges We use three proxes for a frm s operatonal hedgng: (1) the number of countres n whch t operates (breadth), () country concentraton (depth), and (3) the geographc dsperson of ts subsdares across countres (dsperson). Breadth s the logarthm of the number of foregn countres n whch the frm has subsdares. Depth s calculated as the (number of foregn subsdares n the top two foregn countres) / (number of foregn subsdares) 1. Dsperson s constructed wth the Hrschman- Herfndahl concentraton ndex over all the countres n whch a frm operates. The geographc dsperson for frm s calculated as: K ( Dsperson) 1[( No. subsdares) /( TotalNo. subsdares) ], (3) j 1 j where K s the total number of countres n whch frm operates. Ths measure has a value close to 1 f the frm has subsdares n many countres and a value of f the frm has subsdares n only one country. We now add the operatonal hedges to equatons (a) and (d) and test fnancal and operatonal hedge strateges smultaneously. Equatons (4a) to (4f) are used to test as followng: ( Breadth), (4a) 1( FS / TS) ( Hedge) 3 ( Sze) 4 FS TS FCD FDD 1( / ) ( ) 3 ( ) 4 ( ) 5 ( ), (4b) 1( FS / TS) ( Hedge) 3 ( Sze) 4 Sze Breadth ( Depth), (4c) FS TS FCD FDD 1( / ) ( ) 3( ) 4 ( ) 5 ( ), (4d) Sze FS TS Hedge Sze Dsperson) Depth 1( / ) ( ) 3 ( ) 4 (, (4e) ( Dsperson). (4f) 1( FS / TS) ( FCD) 3 ( FDD) 4 ( Sze) 5 The OLS regresson results of equatons (4a) to (4f) are shown n Table 5 3. The coeffcents on fnancal strateges, FS/TS and frm sze reman the same sgn as n Table 4. In addton, the postve coeffcents on both Breadth and Dsperson ndcate that frms that are geographcally dspersed have hgh exposures. The negatve coeffcent on Depth, however, ndcates that frms concentratng n few countres have low exposures. In addton, the coeffcents on Breadth, Depth and Dsperson are all nsgnfcant. Overall, our results suggest that operatonal hedgng s not an effectve tool for currency rsk management. Cross-Sectonal Regresson wth Fnancal and Operatonal Hedge Strateges Dependent Varable OLS (Ordnary Least Square) Table 5 (4a) (4b) (4c) (4d) (4e) (4f) Intercept 3.686** 4.9356*** 3.49** 5.135** 3.78** 4.845*** FS/TS.4681*.3754.385.31.4178*.345 Hedge -.94 -.169 -.115 FCD -.857* -.898* -.86* 1 See Pantzals, Smkns and Laux (1). See Allayanns, Ihrg and Weston (1). 3 The ndependent varables are not hghly correlated accordng to the correlaton coeffcents shown n Table. In addton, multcollnearty s not a severe problem here snce the varance nflaton factors (VIF) are less than ten. After the Ramsey Reset test, there s no functonal form msspecfed. After the Whte test and the Generalzed Durbn-Watson test, there s no heteroskedastcty or autocorrelaton n the resduals of the OLS regresson. After the Kolmogorov-Smrnov test, the resduals are not rejecton of normalty.

Investment Management and Fnancal Innovatons, Volume 4, Issue 3, 7 11 Table 5 (contnued) (4a) (4b) (4c) (4d) (4e) (4f) FDD.953***.848**.816*** Sze -.856** -.445 -.697* -.478 -.847** -.4385 Breadth.443.57 Depth -.645 -.568 Dsperson.1557.1398 Adj R-Square.951.154.968.974.956.115 F-Value.36*.57**.49**.64**.4*.56** 3.4. Robust test 1. Ths table reports the cross-sectonal regresson results of equatons (4a) to (4f) usng one-month horzon data. The absolute value of s used as the dependent varable.. Sgnfcance levels are ndcated as follows: ***1%, **5%, *1%. Instead of the absolute value of estmated exposures n prevous sectons, we now use the raw value of estmated exposures from equaton (1) as the ndependent varables, and run equatons (4a) to (4f) agan. The results are shown n Table 6. The coeffcents on Hedge are postve now. But f fnancal hedge strateges are separated nto FCD and FDD, we fnd the same results as shown n Table 5. The coeffcents on FCD are stll negatve but nsgnfcant. The coeffcents on FDD are stll postve and sgnfcant. Ths means that foregn exchange exposures decrease when frms use FCD, but ncrease sgnfcantly when frms use FDD. The postve and sgnfcant coeffcents on both Breadth and Dsperson emphasze that frms that are geographcally dspersed have hgh exposures. The negatve and sgnfcant coeffcents on Depth, however, emphasze that frms concentratng n few countres have low exposures. The coeffcents on FS/TS and frm szes reman the same sgn as n Table 5. Cross-Sectonal Regresson wth Fnancal and Operatonal Hedge Strateges Dependent Varable: OLS (Ordnary Least Square) Table 6 (5a) (5b) (5c) (5d) (5e) (5f) Intercept.345 5.489**.871 5.8654**.471 5.956** FS/TS.465.379.41.956.354.1657 Hedge.437.7.9 FCD -.89 -.41 -.345 FDD.5343**.5146*.4948* Sze -.631 -.5756** -.89 -.5334** -.95 -.5781** Breadth.451.678 Depth -.895* -.8667* Dsperson.7435**.76** Adj R-Square.367.75.697.159.8.154 F-Value.89 1.56 1.78.5*.6*.49** 1. Ths table reports the cross-sectonal regresson results of equatons (4a) to (4f) usng one-month horzon data. The raw value of s used as the dependent varable.. Sgnfcance levels are ndcated as follows: ***1%, **5%, *1%.

1 Investment Management and Fnancal Innovatons, Volume 4, Issue 3, 7 3.5. Estmaton of foregn exchange exposure over dfferent horzons Bodnar and Wong (3) menton that exposures may be more accurately estmated over longer horzons, due to the complexty of factors nfluencng exposures and the nose n hghfrequency observatons of exchange rates relatve to the persstence of low-frequency movements. Thus, we examne foregn exchange exposures wth longer horzon (three-month, sx-month and twelve-month) returns of stocks and market portfolos as n equaton (1) 1. Usng the MLE estmated exposures over dfferent return horzons as dependent varables, we regress equatons (4a), (4b), (4c), (4d), (4e) and (4f) agan. When the three-month horzon estmated exposure s used as the dependent varable, we get results (6a), (6b), (6c), (6d), (6e) and (6f), as shown n Table 7. When the sx-month horzon estmated exposure s used as the dependent varable, we get results (7a), (7b), (7c), (7d), (7e) and (7f), as shown n Table 8. When the twelve-month horzon estmated exposure s used as the dependent varable, we get results (8a), (8b), (8c), (8d), (8e) and (8f), as shown n Table 9. Accordng to Tables 7, 8, and 9, the sgns of coeffcents on FCD are dfferent from those of the one-month horzon. They are postve but not sgnfcant now, ndcatng that foregn exchange exposures do not decrease when frms use FCD n a longer horzon. The coeffcents on other ndependent varables almost reman the same sgn as those of the one-month horzon. The postve coeffcent on FS/TS exhbts that for a gven exposure, a rase n revenue from foregn operatons ncreases foregn exchange exposures. The negatve coeffcent on the frm sze ndcates that greater frm sze s sgnfcantly assocated wth lower foregn exchange exposures. Agan, there are nsgnfcantly postve coeffcents on both Breadth and Dsperson and nsgnfcantly negatve coeffcent on Depth. To sum up, most of the results from multple return horzons are the same as those of the one-month horzon, except that FCD s not an effectve hedge strategy as the horzon lengthens. Table 7 Cross-Sectonal Regresson wth Fnancal and Operatonal Hedge Strateges over 3-month horzon MLE (Maxmum Lkelhood Estmates) (6a) (6b) (6c) (6d) (6e) (6f) Intercept.9856 4.943** 3.1489 4.931** 3.54 4.7543* FS/TS.9485***.838**.861**.7659**.8779**.773** Hedge.416*.415*.4167* FCD.398.33.313 FDD.3865.367.3591 Sze -.66 -.4557* -.491 -.438* -.756 -.4431* Breadth.654.845 Depth -.917 -.848 Dsperson.691.554 1. Ths table reports the cross-sectonal regresson results of equatons (4a) to (4f) usng three-month horzon data. The absolute value of s used as the dependent varable.. Sgnfcance levels are ndcated as follows: ***1%, **5%, *1%. 1 The resduals of equaton (1) for multple return horzons are homoscedastcty, but wth hgh-order autocorrelaton. After usng the stepwse autoregresson nstructon n the SAS program to remove nsgnfcant autoregressve parameters, the maxmum lkelhood estmators (MLE) are produced. After Whte test and Generalzed Durbn-Watson test, there s no heteroskedastcty and autocorrelaton n the resduals.

Investment Management and Fnancal Innovatons, Volume 4, Issue 3, 7 13 Table 8 Cross-Sectonal Regresson wth Fnancal and Operatonal Hedge Strateges over 6-month horzon MLE (Maxmum Lkelhood Estmates) (7a) (7b) (7c) (7d) (7e) (7f) Intercept.5468 4.834.63 5.91.6459 4.995 FS/TS 1.11* 1.45** 1.756**.961* 1.791**.984** Hedge.71.869.871 FCD.61.931.967 FDD.491.473.491 Sze -.5 -.4737 -.165 -.458 -.337 -.4789 Breadth.1145.14 Depth -.134 -.14 Dsperson.151.178 1. Ths table reports the cross-sectonal regresson results of equatons (4a) to (4f) usng sx-month horzon data. The absolute value of s used as the dependent varable.. Sgnfcance levels are ndcated as follows: ***1%, **5%, *1%. Table 9 Cross-Sectonal Regresson wth Fnancal and Operatonal Hedge Strateges over 1-month horzon MLE (Maxmum Lkelhood Estmates) (8a) (8b) (8c) (8d) (8e) (8f) Intercept -.9489.157 -.6781.5589 -.8634 1.3656 FS/TS.6475.5489.51.461.4856.373 Hedge.8589.847.783* FCD.8561.793.683* FDD.6561.678.491 Sze.1561 -.173.1674 -.134.1145 -.159 Breadth.9 -.353 Depth -.58 -.573 Dsperson.8451.756 1. Ths table reports the cross-sectonal regresson results of equatons (4a) to (4f) usng twelve-month horzon data. The absolute value of s used as the dependent varable.. Sgnfcance levels are ndcated as follows: ***1%, **5%, *1%. 4. Dscusson In our emprcal study, two fnancal hedge strateges have very dfferent effects. The use of FCD s an effectve hedgng strategy n a one-month horzon, but t s less effectve when the horzon lengthens. Our results are consstent wth those of Nguyen and Faff (3). The use of FDD, however, always ncreases foregn exchange exposures. It seems that Tawan frms ssue FDD for other ncentves nstead of foregn exchange exposure hedges. Our results contradct to those of the prevous studes. Many Tawan frms ssue offshore convertble bonds (also known as Euro-convertble bonds, ECB) denomnated n US dollars. The ECB holders have an opton to convert nto stocks or not. If they convert, there are no US dollar debts anymore, and frms do not bear any foregn exchange exposures from ssung ECB. If they do not convert, however, frms wll have a short poston n US dollars and bear foregn exchange exposures from

14 Investment Management and Fnancal Innovatons, Volume 4, Issue 3, 7 ssung ECB. Snce t s a contngent exposure and not easy to control, frms ssung ECB always ncrease foregn exchange exposures 1. Meanwhle, the use of operatonal hedge strateges cannot reduce foregn exchange exposures, the same result as Allayanns, Ihrg, and Weston (1). The foregn nvolvements of Tawan frms often concentrate n Asa. As we know, Asan currences have a hgh correlaton wth each other. They often move n the same drecton. Hence, Tawan frms cannot get the advantage of currency dversfcaton even operatng n several countres, and operatonal hedge strateges do not functon well. 5. Conclusons and Research Restrctons Foregn currency rsk management s of consderable nterest to theoretcans and practtoners n corporate fnance. Our study contrbutes to the lterature by usng multple-horzon data of Tawan non-fnancal frms durng the perod of 1998-5 to examne fnancal and operatonal hedge strateges of foregn exchange exposures smultaneously. Our emprcal fndngs show that the use of operatonal hedge strateges does not help reducng foregn exchange exposure for Tawan frms. Also, the use of FCD s an effectve hedgng strategy n a one-month horzon, but t s less effectve when the horzon lengthens. In addton, the use of FDD always ncreases foregn exchange exposure. Emprcal examnaton of hedgng theores has been dffcult due to the general unavalablty of data on hedgng actvtes. Untl the Fnancal Accountng Standards Board (FASB) ssued SFAS 119 (1994), the corporatons were requred to dsclose the notonal amount of dervatves and other fnancal nstruments n footnotes of ther annual reports. Securtes and Futures Commsson (SFC) n Tawan publshed smlar rules n 1996. However, there s stll no standard form for footnotes of the annual fnancal reports used n Tawan. They are often dsclosed as aggregate data. Hence, we can only use dummy varables nstead of true values for fnancal hedges. Ths s the research restrcton of ths study. References 1. Adler, M. and B. Dumas, 1984, Exposure to currency rsk: defnton and measurement, Fnancal Management 13, 41-5.. Allayanns, G. and E. Ofek, 1, Exchange rate exposure, hedgng, and the use of foregn currency dervatves, Journal of Internatonal Money and Fnance, 73-96. 3. Allayanns, G., J. Ihrg, and J.P. Weston, 1, Exchange-rate hedgng: fnancal versus operatonal strateges, Amercan Economc Revew 91, 391-395. 4. Bodnar, G.M. and W.M. Gentry, 1993, Exchange rate exposure and ndustry characterstcs: evdence from Canada, Japan, and the USA, Journal of Internatonal Money and Fnance 1, 9-45. 5. Bodnar, G.M. and M.H.F. Wong, 3, Estmatng exchange rate exposures: ssues n model structure, Fnancal Management 3, 35-67. 6. Burgman, T.A., 1996, An emprcal examnaton of multnatonal corporate captal structure, Journal of Internatonal Busness Studes 7, 553-57. 7. Chen, C.J.P., C.S.A. Cheng, J. He and J. Km, 1997, An nvestgaton of the relatonshp between nternatonal actvtes and captal structure, Journal of Internatonal Busness Studes 8, 563-577. 8. Chen, C.C., and R.W. So,, Exchange rate varablty and the rskness of US multnatonal frms: evdence form the Asan fnancal turmol, Journal of Multnatonal Fnancal Management 1, 411-48. 9. Chao, C., K. Hung and G.I. Nwanna, 1, The mpact of market lberalzaton on frm s exchange rate exposure, Compettveness Revew 11, 4-5. 1 Currency optons are effectve hedge tools of contngent exposures (Clck and Coval, ). However, there are stll no exchange-traded currency optons n Tawan fnancal markets. Data Source: Industral Development & Investment Center Mnstry of Economc Affars n Tawan.

Investment Management and Fnancal Innovatons, Volume 4, Issue 3, 7 15 1. Chow, E.H., and H.L. Chen, 1998, The determnants of foregn exchange rate exposure: evdence on Japanese frms, Pacfc-Basn Fnance Journal 6, 153-174. 11. Chowdhry, B. and J. Howe, 1999, Corporate rsk management for multnatonal corporatons: fnancal and operatonal hedgng polces, European Fnancal Revew, 9-46. 1. Clck, R.W. and J.D. Coval,, The theory and practce of nternatonal fnancal management (Prentce Hall). 13. Ellott, W.B., S.P. Huffman, and S.D. Makar, 3, Foregn-denomnated debt and foregn currency dervatves: complements or substtutes n hedgng currency rsk? Journal of Multnatonal Fnancal Management 13, 13-139. 14. Fok, R.C.W., C. Carroll, M.C. Chou, 1997, Determnants of corporate hedgng and dervatves: A revst, Journal of Economcs and Busness 49, 569-585. 15. Geczy, C., B.A. Mnton, C. Schrand, 1997, Why frms use currency dervatves, The Journal of Fnance 5, 133-133. 16. Gujarat, D.N., 3, Basc econometrcs. Fourth ed. (McGraw-Hll /Irwn, Sngapore). 17. He, J. and L. Ng, 1998, The Foregn exchange exposure of Japanese multnatonal corporatons, The Journal of Fnance 53, 733-753. 18. Ioro, A.D. and R. Faff,, An analyss of asymmetry n foregn currency exposure of the Australan equtes market, Journal of Multnatonal Fnancal Management 1, 133-159. 19. Joron, P., 199, The Exchange-rate exposure of U.S. multnatonals, The Journal of Busness 63, 331-345.. Kennedy, P., 1998, A gude to econometrcs, 4 th ed. Blackwell, Oxford. 1. Kymaz, H., 3, Estmaton of foregn exchange exposure: an emergng market applcaton, Journal of Multnatonal Fnancal Management 13, 71-84.. Logue, D., 1995, When theory fals: globalzaton as a response to the hostle market for foregn exchange, Journal of Appled Corporate Fnance 8, 39-48. 3. Makar, S.D., J. DeBrun, and S.P. Huffman, 1999, The management of foregn currency rsk: dervatves use and the natural hedge of geographc dversfcaton, Accountng and Busness Research, 9, 9-37. 4. Marshall, A.P.,, Foregn exchange rsk management n UK, USA and Asa Pacfc multnatonal companes, Journal of Multnatonal Fnancal Management 1, 185-11. 5. Mller, K.D., 1998, Economc exposure and ntegrated rsk management, Strategc Management Journal 19, 497-514. 6. Nance, D.R., C.W. Smth, and C.W. Smthson, 1993, On the determnants of corporate hedgng, The Journal of Fnance 48, 67-84. 7. Nguyen, H., and R. Faff, 3, Can the use of foregn currency dervatves explan varatons n foregn exchange exposure? Evdence from Australan companes, Journal of Multnatonal Fnancal Management 13, 193-15. 8. Pantzals, C., B.J. Smkns, P.A. Laux, 1, Operatonal hedges and the foregn exchange exposure of U.S. multnatonals corporatons, Journal of Internatonal Busness Studes 3, 793-81. 9. Sm K. and A. Mozumdar, 3, Foregn currency-denomnated debt: An emprcal examnaton, Journal of Busness 76, 51-546.