Asymmetric Exchange Rate Exposure in Indonesian Industry Sectors

Transcription

1 MPRA Munch Personal RePEc Archve Asymmetrc Exchange Rate Exposure n Indonesan Industry Sectors Lestano Lestano Faculty of Economcs, Atma Jaya Catholc Unversty, Jakarta, Indonesa 2015 Onlne at MPRA Paper No , posted 15. May :34 UTC

2 Asymmetrc Exchange Rate Exposure n Indonesan Industry Sectors Lestano Faculty of Economcs, Atma Jaya Catholc Unversty, Jakarta, Indonesa Ths verson: May 1, 2015 Abstract Ths paper nvestgates asymmetrc exchange rate exposure on Indonesa ndustry s stock returns n both (non)lnear specfcatons and dfferent settng n exchange rate regmes and sub-sample perods usng the EGARCH model. The results reveal that negatve exchange rate exposure domnates over postve exposure n the lnear exposure settng, but there s no domnance sgn n nonlnear exposure effect specfcaton. The negatve exchange rate exposure s more pronounced n the epsodes of Asan and Global fnancal crss and largely reduces n tranqulty perod. In relaton to exchange rate arrangements, many ndustres experence statstcally sgnfcant negatve exposure to the US dollar wth managed floatng exchange rate regme than flexble regme. Keywords: exchange rate exposure, ndustry-level exposure, exponental GARCH-type model, Indonesa JEL-code: C22, F31, G15 Acknowledgements: The author would lke to thank a support of the Department of Economcs, Econometrcs, and Fnance, Unversty of Gronngen, The Netherlands. Correspondng author: Lestano, Faculty of Economcs, Atma Jaya Catholc Unversty, Jl. Jenderal Sudrman 51, Jakarta 12930, Indonesa. Tel.: + 62 (0) and + 62 (0) Moble: +62 (0) Fax: +61 (0) Emal: lestano@atmajaya.ac.d 1

3 1. Introducton Exchange rate exposure that measures the senstvty of frm s value to exchange rate changes s a crucal ssue n fnancal management, because unexpected exchange rate changes may affect a frm s prcng decsons, future cash flows, and frm valuaton as well as rsk management. Over the past three decades, frms and ndustres that were once natonal have become more global resultng n large ncreases n nternatonal actvty. Ths n turn makes exchange rate exposure management a key component of corporate strategy. The challenges for nvestors, credtors, and managers alke s then accurately access the sze and drecton of ths exposure, snce qute bascally the problem must be accessed before t can be managed. Consequently, many theoretcal models and emprcal studes on the exchange rate exposure on frm s or ndustral s valuaton n both ndustral and developng countres have been publshed over the last decades. Koutmos and Martn (2003), Bartram, Dufey, and Frenkel (2005), and Muller and Verschoor (2006) provde revews. Most emprcal studes of exchange rate exposure to date have focused on western ndustral countres and Asan emergng markets, and most have found only modest exposure. In ths paper, we focus solely on Indonesa ndustral sectors. We stress on how the returns for Indonesan ndustral sectors respond to exchange rate returns after takng nto account of returns on a market ndex. Ths paper extends the exstng lterature n several mportant ways. We nvestgate whether the exchange rate arrangement can mpact exchange rate exposure and whether the exchange rate exposure dffers across ndustral sectors n terms of how they experenced the Asan and Global fnancal crses. In addton, we use the EGARCH (Exponental Generalzed Autoregressve Condtonal Heteroskedastc) model wth (non)lnear 2

4 exposure specfcatons whch s enable us to reveal mportant propertes of hgh frequency data of fnancal markets nstead of usng the smple regresson model. Usng a sample of ten Indonesa ndustral sectors for the perod May 29, 1996 to March 3, 2012, our fndngs can be summarzed as follows. Frst, many Indonesa ndustres experence statstcally sgnfcant negatve exposure to fluctuatons n the US dollar. Second, we fnd that the number of ndustres exhbted sgnfcant negatve exposure ncreases from tranqulty perods to the onset of Asan and Global fnancal crss. Fnally, we fnd that ndustres under managed floatng exchange rate regme are more exposed to foregn currency movements than under floatng regme. The remander of the paper s organzed as follows. Secton 2 brefly hghlghts the lterature of exchange rate exposure both theoretcally and emprcally. Secton 3 presents model specfcaton and then t outlnes estmaton method for uncoverng the relaton between ndustry value and exchange rate exposure. Secton 4 dscusses the data and ts propertes. Secton 5 reports the emprcal results. Secton 6 reveals the senstvty and stablty of exchange rate exposure n the present of exchange rate regme changes and the Asan fnancal crss and Global crss Secton 7 concludes. 2. Lterature revew In the theoretcal lterature the concept of foregn exchange rate exposure descrbes the mpact of foregn exchange rate changes on corporatons. More precsely, economc foregn exchange rate exposure represents the senstvty of frm value n natonal currency reflected by changng the present value of ts expected future cash flows wth regard to unexpected foregn exchange rate changes (Adler and Dumas 1984). There are a number of channels through whch the exchange rate mght affect the value of a frm. Some theoretcal models put forward a role of nternatonal trade 3

5 as channel of exchange rate exposure. Shapro (1975), among others, concludes that a deprecaton of the domestc currency leads to an ncrease n sales revenue and profts as well as to hgher producton of a multnatonal frm. The degree of exposure may also relate to the level of foregn competton and the degree of substtutablty between local and mported factors of producton. Lev (1994) and Marston (2001) support that the proftablty of sales n foregn country and net foregn revenues are the man channel for a frm s exchange rate exposure. Grffn and Stulz (2001) and Wllamson (2001) argue that the ndustry competton level play key roles n exchange rate exposure, whlst Allayanns and Ihrg (2001) emphasze the role of ndustry markup besdes competton level. Bodnar, Dumas, and Marston (2002) demonstrate that an exportng frm s exchange rate exposure relates wth the extent t pass-through exchange rate changes to prces. Even frms that do no nternatonal busness, O'Bren (1994) and Marston (2001) stress that these type of frms may be nfluenced ndrectly by foregn competton and therefore they may have exchange rate exposure. Ths may happen n case of nternatonal market leaders located abroad are able to contest the local market wth mports (mport competton). Muller and Verschoor (2006) provde further detaled on the theoretcal foundatons of exchange rsk exposure. Emprcal studes to nvestgate exchange rate exposure are qute dverse n terms of geographcal covered, factors determned sgn and sze of the exchange rate exposure estmates, aggregaton level n terms of frm- or ndustry-level, and model specfcaton and estmaton method used to relate exchange rate and frm value. Most emprcal exposure studes focus on U.S. frms (e.g. Joron 1990; Bartov and Bodnar 1994; Fraser and Pantzals 2004; Prtaman, Shome, and Sngal 2005; Kolar, Moorman, and Sorescu 2008; Du and Hu 2012; Chaeb and Mazzotta 2013) or 4

6 nternatonal ndustres (e.g. Bodnar and Gentry 1993; Grffn and Stulz 2001; Bodnar, Dumas, and Marston 2002; Bredn and Hyde 2011). A potental problem wth examnng only U.S. frms s that they may dffer wdely n ther exposure even after controllng for the level of foregn sales. Indeed, recent studes show more evdence of exposure n ndustres and frms for developed countres outsde the U.S. (e.g. He and Ng 1998; Grffn and Stulz 2001; Apergs, Artks, and Sorros 2011; Agye- Ampomah, Mazouz, and Yn 2012). More recent studes emphasse on emergng and developng countres (e.g. Domnguez and Tesar 2001, 2006; Dodge, Grffn, and Wllamson 2006; Muller and Verschoor 2007; Chue and Cook 2008; Aysun and Guld 2011). In terms of relaton between exchange rate exposure and ts determnants, the studes show a qute mxed result. Joron (1990) and He and Ng (1998) fnd foregn sales to be an mportant determnant of exposure for US and Japanese multnatonals. However, Bartov and Bodnar (1994) conclude that no evdence of contemporaneous exposure for U.S. multnatonals, although they do show that U.S. frms respond to past quarterly exchange rate movements, and Koch and Saporoschenko (2001) ndcate that Japanese keretsu fnancal frms have hgher than average market rsk but nsgnfcant exposure to exchange rate changes. Domnguez and Tesar (2001, 2006) fnd a lnk between foregn actvty and exposure n a sample of frms from eght non-u.s. countres, ncludng Japan. Usng frm-level data from 18 countres (developed and emergng markets), Dodge, Grffn, and Wllamson (2006) show that frm sze, the level of nternatonal sales, foregn ncome, and foregn assets are all sgnfcantly negatvely related to exposure. Chue and Cook (2008) reveal that deprecatons for 15 emergng markets tend to have a negatve mpact on emergng market stock returns n the pre-asan fnancal crss epsode, but ths tendency has 5

7 largely dsappeared n the post-asan crss. Focusng on frm-level data of seven Asan markets, Muller and Verschoor (2007) fnd that about a quarter of Asan frm experenced economcally sgnfcant exposure effects to the US dollar and the Japanese yen. The overall extent of exposure s not sample dependent; a deprecatng (apprecatng) Asan currency aganst foregn currences has a net negatve (postve) mpact on stock returns. Recent study by Ln (2011) concludes that exchange rate exposure n the Asan emergng markets exsts durng sample perod, especally the exposure more ntense durng the outbreak of the Asan fnancal crss and the 2008 global crss. Varous types of model specfcatons and estmaton methods exst n the lterature used to relate exchange rate and frm value. The studes dffer n terms of (non)lnear treatments of exchange rate exposure, (a)symmetrc exposure, estmaton method, and relaton to rsk management strateges. The majorty of emprcal studes test for a constant lnear relatonshp between stock returns and exchange rate changes, for nstance, Joron (1990), Bodnar and Gentry (1993), Grffn and Stulz (2001). In contrast, the theoretcal lterature on the relatonshp between the value of a frm and the exchange rate generally posts a nonlnear relatonshp, among others, outlned by Kogut and Kulatlaka (1994), Bartram (2004), Prestley and Ødegaard (2007), and Aysun and Guld (2011). Several arguments have been put forward to explan the exstence of nonlnear exposure. Future corporate cash flows are uncertan as reflected n the way prces and quanttes of sales changes to adjust exchange rate changes. Nonlnear exchange rate exposures result when corporatons react to exchange rate movements by be able to shft manufacturng or sourcng n dfferent countres or by ther flexblty n producton and export or mport decsons. In addton, a nonlnear exposure can also stem from the result of asymmetrc reactons 6

8 of frm value to exchange rate changes. Asymmetrc exposure s mpled n theoretcal models purportng to descrbe actual corporate behavor, such as, prcng-to-market (Marston 1990; Knetter 1994), hysteress (Baldwn 1988), and asymmetrc hedgng (Mller and Reuer 1998). Koutmos and Martn (2003, 2007) provde more detaled revew on asymmetrc exchange rate exposure. Most of emprcal studes use lnear regresson model both n panel data settng and usng nstrumental varable to uncover frm value and exchange rate exposure (e.g. Joron 1990; Domnguez and Tesar 2001, 2006; Chue and Cook 2008; Agye-Ampomah, Mazouz, and Yn 2012). Some studes employ event study, such as, Dewenter, Hggns, and Smn (2005). Recent studes use a varant type of GARCH models both n bvarate and multvarate settng n nvestgatng senstvty n both changes and volatlty of stock returns to changes and volatlty of exchange rate changes and the correlaton between volatltes of stock returns and exchange rate changes. Those studes, among others, are Koutmos and Martn (2003), Muller and Verschoor (2007), Jayasnghe and Tsu (2008) and Gounopoulos et al. (2013). In terms of rsk management strateges to deal wth exchange rate fluctuatons, Lev (1994) and Marston (2001) argue that cash flows of specfc commercal transactons that have already been booked s relatvely easy to dentfy through accountng systems over a gven tme perod. Exchange rate exposure to these transactons can be effectvely managed by well-structured hedgng strateges. Hageln and Pramborg (2006) and Bartram, Brown, and Mnton (2010) reveal that a large global manufacturng frms may be able to use pass-through, operatonal hedgng, and fnancal hedgng to reduce ther foregn exchange exposure. These strateges represent only a part of the relevant economc foregn exchange rate exposure of the frm. Whle exchange rate movements affect contractually fxed 7

9 transactons drectly, there are addtonal effects on future cash flows related to the compettve poston of the frm, whch can manfest themselves n prce as well as quantty effects. Consequently, Stulz and Wllamson (1997) emphasze that economc foregn exchange rate exposure provdes sgnfcant varablty n cash flows for most frms worldwde, and therefore t s qute dffcult to hedge effcently. In relaton to effects of dfferent settng of exchange rate regme to frm s hedge ncentves, Kaml (2012) fnds that corporatons may be able to dmnsh negatve exchange rate exposure by hedgng n envronment of flexble exchange rate regme than pegged regme. 3. Model specfcaton and estmaton method Snce ts ntroducton, the GARCH model, devsed by Bollerslev (1986), has been appled n countless emprcal studes. The GARCH model and ts subsequent extenson has turned out to be partcularly useful for modelng condtonal volatlty n a wde varety of fnancal market data. Bollerslev, Chou, and Kroner (1992) and Bollershev, Engle, and Nelson (1994) provde revews. In spte of the GARCH models elegantly capture fat tals and volatlty clusterng (hghly persstent perods of volatlty and tranqulty) n asset returns, ths feature of GARCH models neglect asymmetrc condtonal varance functon that may be approprate for modelng the volatlty of asset returns because t can represent a phenomena knows as leverage effect, whch s the negatve correlaton between volatlty and past returns. 1 Ths s not only observed for ndvdual assets, but volatltes of asset market also exhbt ths behavor over tme. Many varants of GARCH-type models that are capable of 1 The asymmetrc volatlty phenomenon explaned by leverage effect s put forward by Black (1976) and Chrste (1982). Other explanatons of the phenomenon are the volatlty feedback effect (Pndyck 1984; French, Schwert, and Stambaugh 1987; Campbell and Hentschel 1992; Bekaert and Wu 2000) and an ncreasng correlaton of stocks n down-markets (Conrad, Gultekn, and Kaul 1991; Longn and Solnk 2001; Ang and Chen 2002). 8

10 capturng volatlty asymmetry have been developed. A wdely accepted varant of such models that allows for asymmetrc effects s the Exponental GARCH (EGARCH) model ntroduced by Nelson (1991). In ths paper, we adopt an unvarate EGARCH(1,1) model to capture the exchange rate exposure on ndustral sector returns. We specfy the mean and varance structures n detals below. We use a commonly adopted formula for measurng economc exposure, that s, an asset prcng equaton, whch s usually the market model, a varaton of the Captal Asset Prcng Model (CAPM), augmented by the foregn exchange rate varable, as follows: R R R u (1) t, 0, 1, mt, 2, f, t t, where R t, s the daly return on the stock of ndustral sector on day t; R mt, s the return on a Indonesa market ndex on day t; R f, t equals the change on the exchange rate n perod t expressed n Indonesa rupahs per unt of US dollar; and u s a random error term. An ntercept s 0,. The senstvty of the return on ndustry to the market return s denoted by 1, and 2, s the senstvty of the return on ndustry to foregn exchange rate changes or return on foregn exchange. In order to take nto account nonlnear structures n the foregn exchange rate exposure, we follow Prestley and Ødegaard (2007) by addng the squared values of the exchange rate changes to the lnear exposure framework specfed n Equaton (1): R R R R u (2) 2 t, 0, 1, mt, 2, f, t 3, f, t t, where 2 R f, t are the squared values of exchange rate changes, and 3, measure the senstvty of stock to nonlnear effects R f, t and R f, t are by constructon to be hghly correlated. One may orthogonalze the exchange rate changes to avod multcollnearty. However, Sercu and Vandebroek (2006) clam that ths method s defected, n terms of the drop of the estmator s standard error s msleadng and the sgnfcance tests 9

11 The specfcaton for the condtonal varance of EGARCH model s: log( ) gz ( ) log( ) (3) 2 2 t, t, 1 t, 1 gz ( ) z ( z E z ) (4) t, t, t, t, where the terms and are real constants. The terms 2 t, s the condtonal varance, u t, s the nnovaton at tme t, and z t, s the standardzed nnovaton (.e. zt, ut, / t, ). It s assumed that z t, has Generalzed Error Dstrbuton (GED) wth 2 mean zero and varance ( t ) and u t, N(0, 2 t, ). Substtutng Equaton (4) nto Equaton (3), the condtonal varance of EGARCH model s gven by 2 2 log( t, ) zt, 1 ( zt, 1 E z t, 1 ) log( t, 1) (5) where and. An mportant advantage of EGARCH specfcaton s 2 that snce the equaton uses the log of, the varance tself wll always be postve. t, Therefore, unlke the GARCH model, no restrctons are mposed on the parameters n Equaton (5) to ensure that the condtonal varances are non-negatve. The coeffcent for zt, 1,, s the sgn effect whereby asymmetrc effects are present n response to a shock when 0 and the postve coeffcent of means the non-exstence of asymmetrc volatlty (leverage effect), whle a negatve, whch s also statstcally sgnfcant, represents the presence of leverage effect. The leverage effect asserts that negatve shocks to the condtonal mean equaton tend to have a larger effect on volatlty than postve shocks. The usual nterpretaton of the leverage effect s that a lower stock prce reduces the value of equty relatve to corporate debt and a sharp declne n stock prces ncreases corporate leverage and therefore the rsk of holdng stocks. doubtful. One may capture the essence of exchange rate exposure non-lnearty by testng the jont 2 effect of R f, t and R f, t on the market returns. 10

12 The coeffcent on ( zt, 1 E z t, 1 ),, represents the sze effect whch ndcates how much volatlty ncreases rrespectve of the drecton of the shock. A coeffcent value for 0, mples that condtonal volatlty would tend to rse (fall) when the absolute value of the standardzed resdual s larger (smaller). The condtonal varance specfcaton of EGARCH model captures the degree of volatlty persstence measured by. The uncondtonal varance s fnte f 1. If 1, then the uncondtonal varance does not exst and the condtonal varance follows an ntegrated process of order one. A sgnfcant suggests that the effects of an nformaton shock persst for some tme. The persstence of volatlty may also be quantfed by an examnaton of the half lfe (HL), whch ndcates the tme perod requred for the shocks to reduce to one half of ther orgnal sze. Followng Koutmos, Lee, and Theodossu (1994), the HL of the volatlty persstence can be calculated as HL = ln(0.5) / ln. The z t s assumed to follow the Generalzed Error Dstrbuton (GED) wth probablty densty functon vexp( (1/ 2) z / ) f( zt), zt,0v t (1 v) / v 2 (1/ v) t v where () s the gamma functon, v s a tal-thckness parameter, and s defned as 2/ v (1/ v) /(2 (3/ v)) 1/2 (6) Under the GARCH specfcaton, the contrbuton of a sngle observaton to the sample loglkelhood and ts analytcal scores are gven as (1 ) / 2 log( ) (1/ 2) / v log 2 v v t (1/ ) (1/ 2)log( t ) L v z v (7) Substtutng Equaton (6) nto Equaton (7) and usng the fact that zt, ut, / t,, the smplfed loglkelhood functon s gven by 11

13 v /2 2 t t t L(1/ 2) log( v (3 / v) / 4 (1/ v) ) u (3 / v) / (1/ v) (1/ 2) log( ) 4. Data sources and propertes We use daly data coverng the perod from May 29, 1996 to March 3, 2012 consstng of 4,122 observatons for nvestgatng the potental exchange rate exposure at the ndustry level n Indonesa. All data are collected from Thomson- Datastream. Sample perods cover the Asan fnancal crss and the recent Global crss. We employ country-specfc market portfolo returns proxed by Jakarta stock exchange (JSX) composte prce ndex, an overall stock ndex n Indonesa. Foregn exchange rates are quoted n terms of domestc currency per unt of US dollar. 3 Our choce of the exchange rates s supported by the fact that the US s one of the most mportant tradng partners of the Indonesa market. In addton, usng blateral rates may facltate the detecton of exposures. Bartov and Bodnar (1994) and Koutmos and Martn (2003) emphasze that t may be more dffcult to detect exposure to an exchange rate ndex, such as a trade-weghted rate or a weghted average of multple blateral exchange rate, f frms or ndustres have dfferent relatve lnkages than those reflected n the ndex. Our dataset on the stock of ndustral sectors focuses on the three-dgt ndustry level. It comprses the daly ndustral ndexes of ten sectors n Indonesa. The ndustral sectors nclude agrculture, basc ndustry, consumer goods, fnance, manufacture, mnng, mscellaneous ndustres, constructon and property, trade and servce, and utlty nfrastructure. We nclude all ndustres n our emprcal work wthout lmtng to ndustres that actvely engage n nternatonal trade or have multnatonal operaton. Ths emprcal strategy allows the data to nform us about 3 We use nomnal exchange rate for a number of reasons. We deal wth monetary assets n whch ts value of home currency s senstve to changes n the nomnal exchange rate. Unlke non-monetary assets vares wth changes n the real rate. By usng the nomnal rates we assume that partcpants n fnancal markets cannot nstantaneously observe the nflaton rates (whch s lkely to be the case). In addton, t s well documented n the lterature that a hgh correlaton between the changes n nomnal and real exchange rates exsts. 12

14 whch ndustres are more or less lkely to be exposed. Fgure 1 plots the returns (daly dfferences of natural logarthms) on the ndustral sectors, the market ndex, and on the exchange rate. The fgure clearly hghlghts that most of the returns have a huge varablty durng the Asan fnancal crss and relatvely modest varablty durng the Global crss [Fgure 1 about here] Table 1 presents the descrptve statstcs of daly stock ndvdual ndustry and market returns, and exchange rate returns. The table shows that all of seres have postve average returns. The table mples that the changes n the exchange rates and the stock returns of fnancal sector exhbt the lowest and the hghest dsperson (measured as the absolute value of the percentage coeffcent of varaton). The sample skewness coeffcents ndcate that all of the seres are postvely skewed, except for basc ndustres, mnng, and market stock returns yeld negatve skewness. All returns are leptokurtc (peaked relatve to the normal dstrbuton and fat tals). Consequently, all seres dsplay strong evdence of non-normalty as s also llustrated by the Jarque-Bera statstc. [Table 1 about here] Table 2 reports test results of seral correlaton and ARCH effects for all seres at 12 and 24 lags. The Ljung-Box Q-statstcs at 12 and 24 lags are computed for both the seres and squared seres. The results of the Q tests show that there s sgnfcant autocorrelaton n the resduals. Ths s seen as evdence for lnear and nonlnear dependences. Lnear dependences may be caused by some form of market neffcency or market structure, and nonlnear dependences may lead to autoregressve condtonal heteroscedastcty. The ARCH tests also ndcate that the 13

15 null hypothess of no autoregressve condtonal heteroscedastcty s rejected at the 5% sgnfcance level. [Table 2 about here] 5. Emprcal results The estmaton results of EGARCH(1,1) model outlned n Equaton (1) or (2) and (5) wth the Generalzed Error Dstrbuton (GED) are shown n Table 3. Two specfcatons of exchange rate exposure, lnear and nonlnear, are estmated. Table 3 reveals that two of the 10 ndustral sectors for both lnear and nonlnear exposure specfcaton exhbt statstcally sgnfcant leverage effects, makng other ndustres wth 0 better nvestment n antcpaton of bad news for rsk-averse nvestors. The leverage effects for the lnear exposure model are exhbted n ndustral sectors,.e., consumer goods and constructon and property. For non lnear exposure model, manufacture, and constructon and property ndustres show the same effect. The leverage effect seems more pronounced for manufacture ndustry. The stronger leverage effect mght be due a greater concentraton on commercal actvtes that have greater mplcatons for the volatlty of those ndustry s stock prces. [Table 3 about here] All ndustres exhbt a sgnfcant postve. Ths mples postve relaton between condtonal volatlty and the absolute value of the standardzed resdual. The persstence n volatlty s measured by the parameter. The values of are less than one for all ndustres, whch s a necessary condton for the volatlty process to be stable. The magntudes of the parameter are close to unty for all ndustres n both (non)lnear specfcaton showng evdence of a hgh level of persstence of volatlty shocks over tme whch n turn mght have been due to the use of daly data. 14

16 Usng the HL (half-lfe) parameter, the volatlty persstence can be expressed n day terms. Based on the HL results for lnear exposure model, fnance ndustry takes the longest to reduce the mpact from ts shocks by half (76.67 days) and agrculture sector takes the least tme (18.38 days), whch suggests that agrculture sector has the lowest level of volatlty persstence out of all ndustres. For non lnear exposure model, agrculture ndustry stll has the shortest HL persstence n volatlty (16.15 days), whlst fnance and mnng ndustres has the longest HL (76.69 days). Table 3 reports the estmates of the exposure coeffcents for the ndustres wth lnear and non lnear exposure. The estmates of the lnear exposure coeffcents show that all of ndustres, except agrculture and mnng ndustres, have a statstcally sgnfcant exposure coeffcent relatve to the US dollar. Seven ndustres exhbt negatve sgn of the US dollar exposure wth consumer goods sector yelds the hghest negatve exposure (-1.17), only utlty nfrastructure ndustry have postve sgn. The negatve 2, coeffcent suggests that an apprecaton of the US dollar aganst the rupah has a negatve mpact on Indonesa s ndustral sector returns. On the other hand, the postve 2, coeffcent obtaned suggests that Indonesa ndustres experence a benefcal valuaton effect when the US dollar apprecates. The negatve exchange rate exposure domnates over postve exposure across ndustral sectors whch are consstent wth Parsley and Popper (2006) and Muller and Verschoor (2007). Ths evdence ndcates that most Indonesa ndustres experence gans (relatve to the market) when ther domestc currency apprecates the value of Indonesa ndustry share values benefts (s hurt) when the domestc currency apprecates (deprecates). The effect of exchange rate to ndustry s stock returns may relate to the followng channels: foregn-currency debt exposure, nternatonal trade exposure, and 15

17 the depth of fnancal system (Chue and Cook 2008). Apprecaton of US dollar may dmnsh frm s value n the presence of hgh foregn-currency domnated debt. As external debt n Indonesa markets s prmarly denomnated n foregn currences, exchange rate devaluatons (apprecatons) lead to an ncrease (decrease) n foregn currency debt repayments, thereby causng negatve exchange rate exposure. In the case of ndustres that actvely engaged n nternatonal trade, exporters and domestc frms facng mported competton suffer under a large dollar apprecaton aganst the rupah. Operatonally, ths s consstent wth a noton that Indonesa ndustres are netmporters, and/or have foregn operatons for export to the world market. The apprecaton US dollar that gve a benefcal valuaton effect to Indonesa ndustres would be the case f these ndustres are net-exporters, and/or f they set up foregn operatons for the purpose of local sales. In terms of the depth of fnancal development channel, less-developed fnancal markets (especally wth mnor access to both fnancal and operatonal hedgng strateges) should hgher frms cost of hedgng and decrease ther desre to hedge. On the other hand, the shorter range of avalable fnancal products may also decrease frms desre to speculate n the foregn exchange market. The former effect tends to hgher, whle the latter effect tends to lower, the negatve magntude of a frm s exchange rate exposure. The results from runnng nonlnear exchange rate exposure model for those ndustres reveals that four ndustral sectors experence statstcally sgnfcant nonlnear foregn exchange rate exposure based on z-statstc (ndvdual effect). There are two ndustres that have negatve and postve exposure coeffcents, respectvely, to the US dollar, that are statstcally sgnfcant. Mnng ndustry and trade and servce sector have the hghest postve (+0.47) and negatve (-0.48) nonlnear exposure, respectvely. The negatve coeffcent on the nonlnear effects 16

18 would ndcate that exporters are affected snce as the US dollar apprecates stock returns wll fall. However, gven the convex relatonshp, the effect s reduced relatve to the lnear case. Recall that the lnear coeffcent was negatve and suggested that exporters and domestc frms that have mport competton suffered. The negatve sgn on the nonlnear term suggests that the loss they ncur s offset as the exchange rate change gets larger. Many of not sgnfcant of ndvdual effects may due to the exstence of multcollnearty. When we test the jont effect of exchange rate changes and ts squared values on the market returns usng the Wald test based on F-statstc, we obtan all ndustres sgnfcantly exhbted non-lnearty n terms of jont effect, except agrculture ndustry. Ths evdence shows us that together exchange rate changes and ts squared values nfluence the market returns, but t s dffcult to sort out the effect of each. 6. Exposure senstvty and stablty Havng gauged the foregn exchange exposure, we next examne ts emprcal lnk to structural change n the Indonesa ndustral s exposure to exchange rate fluctuatons durng two major fnancal crss,.e. the Asan fnancal crss and the Global fnancal crss ; and exchange rate arrangements. We employ separately two estmatons strategy to address the two ssues. 4 Frst, n dealng wth the frst ssue, we follow Muller and Verschoor (2007) to mplement a subsamplebased analyss n order to take account the mportance of structural breaks owng to the Asan and Global crses. The sample s dvded nto fve sub-perods: May 29, 1996 May 30, 1997 (pre-crss perod); June 2, 1997 December 31, 1999 (Asan crss perod and ts aftermath); January 1, 1999 July 31, 2007 (between crss 4 We focus only on lnear exchange rate exposure specfcaton for uncoverng the relaton of exchange rate exposure to exchange rate arrangement and the Asan and Global fnancal crss. As noted n the prevous secton that nonlnear exposure s not decsve enough n terms of statstcally sgnfcant of nonlnear exposure coeffcents of ndvdually effect. 17

19 perod); August 1, 2007 March 31, 2009 (Global crss perod and ts aftermath); and Aprl 1, 2009 March 15, 2012 (post-crss perod). 5 We complement ths analyss by ntroducng crses dummy varables both n mean and condtonal varance equatons of EGARCH(1,1) lnear exchange rate exposure model specfed n Equaton (1) and (3), respectvely. The dummy varable takes value of one to ndcate the presence of the Asan and Global crses, and otherwse t equals zero. Next, we test jont hypothess of the presence of the Asan and Global crses usng the F- statstc based of Wald test. Table 4 provdes the estmated EGARCH(1,1) model wth mean equaton, as defned n Equaton (1), for the fve sub-sample perods. The table provdes evdence n support of tme-varaton n exposure at the ndustral level. The number of sgnfcant 2, coeffcents ncreases substantally from tranqulty sub-sample perods (pre-crss, between crss, and post-crss) to fnancal turmol sub-perods (Asan fnancal crss and Global crss). It seems lkely that the ncrease n exchange rate fluctuatons durng the perod of Asan and Global fnancal crss faced by Indonesa ndustres generally has led to a hgher percentage of frms wth sgnfcant exchange rsk exposure. However, sgnfcant exchange exposure s hgher durng Asan fnancal crss than Global crss. Consstent wth the results of Table 3, almost all Indonesa ndustres experence a predomnantly negatve foregn currency exposure effect. It should be notced that the estmated parameters are statstcally sgnfcant and reveal a hgh degree of volatlty persstence for all ndustres, except constructon and property, utlty nfrastructure, and trade and servce ndustres n the pre-crss perod. Based on the HL parameters, the longest HL persstence n volatlty 5 The crss datng s based on offcal tmelnes for the Asan crss dated by Bag and Goldfajn (1999) and Nagayasu (2001) and for the Global crss by Federal Reserve Board of St. Lous (2009) and Flardo et al. (2010). 18

20 ntensfes durng the epsode of Asan and Global crss. Moreover, a number of ndustres that sgnfcantly have leverage effect escalate partcularly n the onset of Asan crss. Table 5 gves further evdence that jont hypothess of the role of dummy varables of the crss s not rejected for all ndustres. In addton, the dummy varables play a role n varance of exchange exposure model for all ndustres but not n the mean equaton. Meanng that Asan and Global crss ntensfy exchange rsk exposure of all ndustres. Overall, the emprcal results provded by Table 4 and 5 suggest that the Asan fnancal crss n has dramatcally affected and altered the characterstcs of Indonesa ndustres. Ths results also supports the fact that the spllovers from the Global fnancal turmol orgnated from the Unted States are too strong to avod by Indonesa, despte the fact that the regon have an ncreased reslence relatve to the Asan crss. 6 [Table 4 about here] [Table 5 about here] Second, to lnk exchange rate arrangements and the foregn exchange exposure, we focus separately at exposure under the alternatve arrangements that exst n our sample. Specfcally, we look at the exposure under a managed float exchange rate regme, and we look at the exposure wthout one. To do so, we estmate EGARCH(1,1) model wth the followng mean equaton: R R D R D R u (8) t, 0, 1, mt, m, mt, f, t n, nt, f, t t, where D mt, equals one when the frm s home currency s managed float aganst another currency and equals zero otherwse, and Dnt, 1 Dmt,. The parameters m, and n, provde separate managed float and floatng exposure estmates, respectvely, 6 Dder, Heva, and Schmukler (2012), Gournchas and Obstfeld (2012), and Frankel, Vegh, and Vuletn (2013) conclude that the strength of fnancal sector, ablty to conduct countercyclcal monetary and fscal polces, and the role of flexble exchange rate regme adopted snce the Asan crss, are key reasons behnds relatve reslence of Asa economes to the effects of the Global crss. 19

21 for the ndustres that have had experence both wth and wthout a managed float durng the sample perod. We test H0 : m, n, n order to reveal whether nformaton about the exchange rate regme s useful for asymmetrc mpacts of the exchange rate returns. The dates of the exchange rate managed float are dentfed by relyng on the offcally reported arrangement n whch Indonesa announced to adopt floatng exchange rate regme started from August 14, by m, Table 6 summarzes the fndngs from estmatng Equaton (8). As exhbted and n,, far more ndustres show statstcally sgnfcant exposure to the US dollar wth managed floatng exchange rate regme than wthout one. Therefore, ndustres under dependent floatng currences may have to be accustomed to hedgng, and hedgng may be less costly f there s an access to both fnancal and operatonal hedgng strateges. Most of ndustres exposed by the US dollar under managed floatng regme yeld negatve exchange exposure. That s, n most of the ndustres, when the US dollar deprecates aganst the home currences, returns go up, not down. One may argue that a dollar managed floatng would dscourage frms/ndustres nto takng on too much dollar denomnated debt, makng them effectvely more postve exposed by the foregn exchange rate movements. However, ths effect does not seem to occur snce there are so many of the ndustres are essentally negatve exposed by the US dollar shown by Table 6. Ths evdence reveals that a managed floatng exchange rate regme provdes less benefcal envronment for busness n terms of provdng stablty and removng the need for expensve hedgng. 7 It s nterestng to note that only two ndustres (fnance and utlty nfrastructure ndustres) shows 7 Kaml (2012) reveals that freely floatng exchange rate regme dmnshes currency msmatches n corporate's balance sheet and transton from managed to a floatng regme renforces corporate to reduce porton of foregn currency domnated debt and use export revenues and foregn currency domnated asset to gradually elmnate unhedged foregn currency exposures and therefore offset rsks of foregn currency domnated debt. 20

22 statstcally sgnfcant evdence of leverage effect and all of ndustres exhbts relatvely comparable magntude of statstcally sgnfcant persstence n volatlty compared to the results wthout augmented exchange rate arrangements exhbted n Equaton (1) and Table 3. The null hypothess whether m, and n, are the same s rejected for more than a half of ndustres. Ths evdence reveals that asymmetrc effects of exchange rate on those ndustres stock returns are sgnfcantly matter n dfferent settng of exchange rate arrangement. [Table 6 about here] 7. Concluson The purpose of ths paper s to explore the level of exchange rate exposure on stock returns across Indonesa ndustral sectors for the perod from May 29, 1996 to March 3, The EGARCH model wth (non)lnear exposure specfcatons were used for the purpose of ths study. In addton, the extenson of the EGARCH model allow us to observe the stablty and senstvty of exchange rate exposure on ndustral s value n dfferent settng of exchange rate arrangements and sub-sample perods settng between tranqulty and fnancal crss. Analyzng the (non)lnear exchange rate exposure across Indonesa ndustral s stock returns, we fnd that negatve exchange rate exposure domnates over postve exposure across ndustral sectors n the lnear exposure settng. Nevertheless, there s no domnance sgn n nonlnear specfcaton. In both (non)lnear models, only few ndustres exhbt statstcally sgnfcant leverage effects, but all ndustres exhbt hghly persstence n volatlty. Usng sub-perods analyss, we detect changes n the exchange rate exposure of Indonesa ndustral s stock returns over tme. In comparson to tranqulty sub-perods (pre-crss, between crss, and post-global crss), mmedately followng the Asan and Global fnancal crses, Indonesa 21

23 ndustres are mostly negatvely exposed to exchange rate changes. Ths negatve exposure s more frequent n the aftermath of Asan crss than Global crss. Ths result supports the evdence of relatvely reslence of Asan emergng economes to Global crss. We go on to nvestgate the exposure n dfferent settng of exchange rate arrangements, managed float or floatng regme. The results reveal that many ndustres experence statstcally sgnfcant negatve exposure to the US dollar wth managed floatng exchange rate regme than wthout one. Fnally, our emprcal results lead to the followng mplcatons. The exstence relatonshp between exchange exposure rsks and frm value leads to even more crucal of the mplcatons of rsk management. Therefore, the exposure profle and behavor determne the choce of hedgng tools n order to hedge properly. In the lght of ncreased fnancal nstablty, t s mportant to account for renforcng a sound flexble exchange rate regme, other than to mantan fnancal sector strength and sound countercyclcal monetary and fscal polcy, to stablze exchange rate and to manage crss. References Adler, M., and B. Dumas Exposure to currency rsk: defnton and measurement. Fnancal Management 13: Agye-Ampomah, S., K. Mazouz, and S. Yn The foregn exchange exposure of UK non-fnancal frms: A comparson of market-based methodologes. Internatonal Revew of Fnancal Analyss 29: Allayanns, G., and J. Ihrg Exposure and markups. Revew of Fnancal Studes 14: Ang, A., and J. Chen Asymmetrc correlatons of equty portfolos. Journal of Fnancal Economcs 63: Apergs, N, P. Artks, and J. Sorros Asset prcng and foregn exchange rsk. Research n Internatonal Busness and Fnance 25:

24 Aysun, U., and M. Guld Exchange rate exposure: A nonparametrc approach. Emergng Markets Revew 12: Bag, T., and I. Goldfajn Fnancal market contagon n the Asan crss. IMF Staff Papers 46/2, Internatonal Monetary Fund, Washngton, D.C. Baldwn, R Hysteress n mport prces: the beach head effect. Amercan Economc Revew 78: Bartov, E., and G.M. Bodnar Frm valuaton, earnngs expectatons and the exchange-rate exposure effect Journal of Fnance 49: Bartram, S.M Lnear and nonlnear foregn exchange rate exposures of German nonfnancal corporatons. Journal of Internatonal Money and Fnance. 23: Bartram, S.M., G. Dufey, and M.R. Frenkel A prmer on the exposure of nonfnancal corporatons to foregn exchange rate rsk. Journal of Multnatonal Fnancal Management 15: Bartram, S.M., G.W. Brown, and B. Mnton Resolvng the exposure puzzle: the many facets of exchange rate exposure Journal of Fnancal Economcs 95: Bekaert, G., and G. Wu Asymmetrc volatlty and rsk n equty markets. Revew of Fnancal Studes 13: Black, T Studes of stock prce volatlty changes, Proceedngs of the 1976 Meetngs of the Amercan Statstcal Assocaton, Busness and Economc Statstcs Secton, Bodnar, G., B. Dumas, and R.C. Marston Pass-through and exposure. Journal of Fnance 57: Bodnar, G., and W. Gentry Exchange rate exposure and ndustry characterstcs: evdence from Canada, Japan, and the U.S. Journal of Internatonal Money and Fnance 12: Bollershev, T., R.F. Engle, and D.B. Nelson ARCH models, In Handbook of Econometrcs, R.F. Engle and D.L. McFadden (Eds.), North Holand: Amsterdam, Bollerslev, T Generalzed autoregressve condtonal heteroscedastcty. Journal of Econometrcs 31: Bollerslev, T., R.Y. Chou, and K.F. Kroner ARCH modelng n fnance: a revew of the theory and emprcal evdence. Journal of Econometrcs 52: Bredn, D., and S. Hyde Investgatng sources of unantcpated exposure n ndustry stock returns. Journal of Bankng & Fnance 35:

25 Campbell, J.Y., and L. Hentschel No news s good news: an asymmetrc model of changng volatlty n stock returns. Journal of Fnancal Economcs 31: Chaeb, I., and S. Mazzotta Uncondtonal and condtonal exchange rate exposure. Journal of Internatonal Money and Fnance 32: Chrste, A.A The stochastc behavor of common stock varances. Journal of Fnancal Economcs 10: Chue, T.K., and D. Cook Emergng market exchange rate exposure. Journal of Bankng and Fnance 32: Conrad, J., M. Gultekn, and G. Kaul Asymmetrc predctablty of condtonal varances, Revew of Fnancal Studes 4: Dder, T., C. Heva, and S.L. Schmukler How reslent were emergng economes to the global crss? Journal of Internatonal Money and Fnance 31: Dewenter, K.L., R.C. Hggns, and T.T. Smn Can event study methods solve the currency exposure puzzle?, Pacfc-Basn Fnance Journal 13: Dodge, C., J. Grffn, and R. Wllamson Measurng the economc mportance of exchange rate exposure. Journal of Emprcal Fnance 13: Domnguez, K., and L. Tesar A reexamnaton of exchange rate exposure. Amercan Economc Revew 91: Domnguez, K., and L. Tesar Exchange rate exposure. Journal of Internatonal Economcs 68: Du, D., and O. Hu Exchange rate rsk n the US stock market. Journal of Internatonal Fnancal Markets, Insttutons & Money 22: Federal Reserve Board of St. Lous The fnancal crss: A tmelne of events and polcy actons, Federal Reserve Board of St. Lous, St. Lous. Flardo, A., J. George, M. Loretan, G. Ma, A. Munro, I. Shm, P.Wooldrdge, J. Yetman, and H. Zhu The nternatonal fnancal crss: tmelne, mpact and polcy responses n Asa and the Pacfc, n Bank for Internatonal Settlements, edtor, The nternatonal fnancal crss and polcy challenges n Asa and the Pacfc, Bank for Internatonal Settlements, Basel, Swtzerland, Frankel, J.A., C.A. Vegh, and G. Vuletn On graduaton from fscal procyclcalty. Journal of Development Economcs 100: Fraser, S.P., and C. Pantzals Foregn exchange rate exposure of US multnatonal corporatons: a frm-specfc approach. Journal of Multnatonal Fnancal Management 14:

26 French, K.R., G.W. Schwert, and R. Stambaugh Expected stock returns and volatlty. Journal of Fnancal Economcs 19: Gounopoulos, D., P. Molyneux, S.K. Stakouras, J.O.S. Wlson, and G. Zhao Exchange rate rsk and the equty performance of fnancal ntermedares. Internatonal Revew of Fnancal Analyss 29: Gournchas, P.O., and M. Obstfeld Stores of the twenteth century for the twenty-frst. Amercan Economc Journal: Macroeconomcs 4: Grffn, J.M., and R.M. Stulz Internatonal competton and exchange rate shocks: A cross-country ndustry analyss of stock returns. Revew of Fnancal Studes 14: Hageln, N., and B. Pramborg Emprcal evdence concernng ncentves to hedge transacton and translaton exposure. Journal of Multnatonal Fnancal Management 16: He, J., and L. Ng Foregn exchange exposure, rsk, and the Japanese stock market. Journal of Fnance 53: Jayasnghe, P., and A.K. Tsu Exchange rate exposure of sectoral returns and volatltes: Evdence from Japanese ndustral sectors. Japan and the World Economy 20: Joron, P The exchange rate exposure of U.S. multnatonals. Journal of Busness 63: Kaml, H How do exchange rate regmes affect frms ncentves to hedge currency rsk? Mcro evdence for Latn Amerca, IMF Workng Paper No.12/69, Internatonal Monetary Fund, Washngton, D.C. Knetter, M.M Is export prce adjustment asymmetrc? Evaluatng the market share and marketng bottlenecks hypothess. Journal of Internatonal Money and Fnance 13: Kogut, B., and N. Kulatlaka Operatng flexblty, global manufacturng, and the opton value of a multnatonal network. Management Scence 40: Kolar, J.W., T.C. Moorman and S.M. Sorescu Foregn exchange rsk and the cross-secton of stock returns. Journal of Internatonal Money and Fnance 27: Koutmos, G., and A.D. Martn Asymmetrc exchange rate exposure: theory and evdence. Journal of Internatonal Money and Fnance 22: Koutmos, G., and A.D. Martn Modelng tme varaton and asymmetry n foregn exchange exposure. Journal of Multnatonal Fnancal Management 17:

27 Koutmos, G., U. Lee and P. Theodossu Tme-varyng betas and volatlty persstence n Internatonal Stock markets. Journal of Economcs and Busness, 46: Ln, C.H Exchange rate exposure n the Asan emergng markets. Journal of Multnatonal Fnancal Management 21: Lev, MD (1994), Exchange rates and the valuaton of frms, In Exchange Rates and Corporate Performance, Y. Amhud and RM Levch (Eds.), Irwn: New York, Longn, F., and B. Solnk Extreme correlaton of nternatonal equty markets. Journal of Fnance 56: Marston, R.C Prcng to market n Japanese manufacturng. Journal of Internatonal Economcs 29: Marston, R.C The effects of ndustry structure on economc exposure. Journal of Internatonal Money and Fnance 20: Mller, K.D., and J.J. Reuer Asymmetrc corporate exposures to foregn exchange rate changes. Strategc Management Journal 19: Muller, A., and W.F.C. Verschoor Foregn exchange rsk exposure: Survey and suggestons. Journal of Multnatonal Fnancal Management 16: Muller, A., and W.F.C. Verschoor Asan foregn exchange rsk exposure. Journal of Japanese Internatonal Economes 21: Nagayasu, J Currency crss and contagon: evdence from exchange rates and sectoral stock ndces of the Phlppnes and Thaland. Journal of Asan Economcs, 12: Nelson, D.B Condtonal heteroskedastcty n asset returns: a new approach. Econometrca 59: O'Bren, T.J Corporate measurement of economc exposure to foregn exchange rsk. Fnancal Markets, Insttutons and Instruments 3: Parsley, D.C., and H.A. Popper Exchange rate pegs and foregn exchange exposure n East and South East Asa. Journal of Internatonal Money and Fnance, 25: Pndyck, R,S Rsk, nflaton and the stock market. Amercan Economc Revew 74: Prestley, R., and B.A. Ødegaard Lnear and nonlnear exchange rate exposure, Journal of Internatonal Money and Fnance 26: Prtaman, M., D. Shome, and V. Sngal Exchange rate exposure of exportng and mportng frms. Journal of Appled Corporate Fnance 17:

28 Shapro, A.C., Exchange rate changes, nflaton and the value of multnatonal corporatons. Journal of Fnance, 30: Stulz, R.M., and R., Wllamson Identfyng and Quantfyng Exposures, In Fnancal Rsk and the Corporate Treasury: New Developments n Strategy and Control, R. Jameson (Ed.), London: Rsk Publcatons, Wllamson, R., Exchange rate exposure and competton: Evdence from the automotve ndustry. Journal of Fnancal Economcs 59:

29 Fgure 1. The underlyng data Agrculture Basc Industry Consumer good Fnance Manufacture Mnng Mscellaneous ndustres Constructon and property Trade and servce Utlty nfrastructure Market stock returns Exchange rate Note: Shadng area shows the tmng of the Asan fnancal crss: June 2, 1997 to December 31, 1998; and the Global crss: August 1, 2007 to March 31, The crss datng s based on offcal tmelnes for the Asan crss dated by Bag and Goldfajn (1999) and Nagayasu (2001) and for the Global crss by Federal Reserve Board of St. Lous (2009) and Flardo et al. (2010). 28