Measurement and Management of Exchange Rate Exposure: New Approach and Evdence Taek Ho Kwon a, Sung C. Bae b,*, Rae Soo Park c January 2013 * Correspondng author; Tel) 419-372-8714; E-mal) bae@bgsu.edu a Kwon s Professor at the Department of Busness Admnstraton, Chungnam Natonal Unversty n Daejon, Korea; Tel) 82-42-821-5533; E-mal) thk5556@cnu.ac.kr. b Bae s the Ashel G. Bryan/Huntngton Bank Professor at the Department of Fnance, College of Busness Admnstraton, Bowlng Green State Unversty n Bowlng Green, OH, USA; Tel) 419-372-8714; E-mal) bae@bgsu.edu. c Park s Professor at the College of Busness Admnstraton, Sookmyung Women s Unversty n Seoul, Korea; E-mal) rsoo@sookmyung.ac.kr. The authors acknowledge helpful comments from Wendy Rotenberg. The work was supported by the Korea Research Foundaton Grant funded by the Korean Government (KRF-2009-32A-B00083). The usual dsclamer apples.
Measurement and Management of Exchange Rate Exposure: New Approach and Evdence Abstract We examne the effectveness of frms operatng and fnancal actvtes n managng ther exchange rate exposure. Unlke prevous studes, we measure the expected exchange rate exposure whch reflects exchange rate rsk assocated wth frms basc busness actvtes pror to the usage of exposure management actvtes and compare t wth the observed exchange rate exposure whch reflects the effects of frms exposure management actvtes as well. Our results show that the effectveness of exposure management actvtes depends on the underlyng characterstcs (e.g., drecton) of the frm s nherent expected exchange rate exposure. Whle frms wth postve expected exchange rate exposure reduce ther exposure through currency dervatves, nternal transactons wth foregn subsdares, and the ssuance of foregn currency denomnated debt, frms wth negatve expected exchange rate exposure do so only through exchange rate pass-through actvtes. Our analyss of frms wth sgnfcant observed exchange rate exposure further offers evdence that structured currency forward contracts (e.g., currency KIKO contracts) deterorate frms exchange rate exposure. JEL Classfcaton: F31; G15 Key words: Expected exchange rate exposure; Observed exchange rate exposure; Exposure management; Korean manufacturng frms 2
1. Introducton A large body of studes has examned the measurement of exchange rate exposure and ts effect on frm value wth nconclusve evdence on the exstence of exchange rate exposure. Adler and Dumas (1984) show that a frm s exchange rate exposure can be measured from a regresson model of the frm s exposed asset prces expressed n the domestc currency aganst the foregn currency. Joron (1990, 1991) measures the exchange rate exposure usng stock returns but fals to show evdence that exchange rsk s an mportant determnant of frm value. Subsequent studes have estmated exchange rate exposure usng regresson models of stock returns and exchange rate changes. Whle a few of these studes report relatvely sgnfcant levels of exchange rate exposure (e.g., He and Ng, 1998, Domnguez and Tesar, 2006) and exchange rsk premum (e.g., Dumas and Solnk, 1995; Kwon, Bae, and Chung, 2004), the majorty of these studes offer evdence that s not n supportve of the exstence of exchange rate exposure or the need to manage exchange rate rsk. As possble explanatons for the lack of strong evdence on exchange rate exposure, earler studes offered several factors ncludng the measurement problems assocated wth the estmaton of exchange rate exposure and the delayed effects of exchange rate changes (see, e.g., Bartov and Bodnar, 1994). Bartram and Bodnar (2007) demonstrate, however, that correctng and consderng these factors stll do not lead to an ncrease n the sgnfcance levels of the estmated regresson coeffcents of exchange rate exposure. Instead, they note that the evdence of nsgnfcant regresson coeffcents of exchange rate exposure reported n prevous studes s not surprsng at all. They argue that t s not due to the estmaton problem but manly due to frms management of exchange rate exposure. Bartram, Brown, and Mnton (2010) offer evdence supportng the Bartram and Bodnar s vew by reportng that frms ndeed manage a large porton of ther exchange rate exposure. In ths paper, we examne two man ssues pertanng to exchange rate exposure the measurement of exchange rate exposure and the effectveness of frms managng actvtes of exchange rate exposure. Our paper, however, takes a very dfferent approach to examne these ssues from 3
prevous studes. As Bartram and Bodnar (2007) and Bartram, Brown, and Mnton (2010) pont out, an effort to measure a frm s exchange rate exposure usng the frm s observed stock returns would fal to uncover the exstence and the true level of the frm s exchange rate exposure. Because a frm s observed stock returns would already have reflected the outcome of the frm s actvtes for exposure management, an examnaton of observed stock returns s more lkely to reveal nsgnfcant exchange rate exposure for the frm. In order to overcome ths measurement problem, we separately measure expected exchange rate exposure, whch reflects a frm s exposure level related to the frm s basc busness actvtes pror to the frm s usage of any exposure management actvty, and observed exchange rate exposure, whch reflects the effects of a frm s other actvtes for exposure management as well. We then compare the observed exchange rate exposure wth the expected exchange rate exposure and analyze ther dfference. Our logc s that because ths dfference captures the outcome of a frm s exposure management actvtes, ts analyss wll reveal the true effectveness of actvtes frms use to manage exchange rate exposure by controllng external factors that mght affect the frm s exchange rate exposure. That s, f the observed exchange rate exposure s excluded from the expected exchange rate exposure, one can solate the exchange rate exposure assocated wth the frm s exposure management actvtes. Hence, the sgnfcantly large expected exchange rate exposure relatve to the observed exchange rate exposure would ndcate a substantal reducton n the frm s exchange rate exposure and thus the effectveness of an exposure management actvty. Our paper s the frst of ts knd to emprcally measure expected exchange rate exposure for a comparson to observed exchange rate exposure. Although Bodnar and Marston (2002) develop a theoretcal model of the expected exchange rate exposure usng a frm s mport rato as one of ts determnants, no exstng study has emprcally measured the expected rate exposure. 1 Ths s prmarly because data on a frm s mport ratos are regarded as the frm s trade secrets and thus are not publcly 1 A frm s mport rato s the rato of expenses of mported materals for fnal products to total operatng expenses. 4
avalable. In ths paper, we proxy a frm s mport rato by the nput-output rato of mported goods derved from the frm s nput-output tables. Frms management of exchange rate exposure generally nvolves both operatng and fnancal actvtes. The operatng actvtes nclude domestc-currency nvocng, matchng and offsettng, and exchange rate pass-through, among others. In partcular, several studes offer evdence supportng that frms pass through ther exchange rate exposure to sales prces wth and wthout the adjustment of mark-ups, especally n the face of decreasng local currency value (Allayanns and Ihrg, 2001; Campa and Goldberg, 1999; Gagnon and Knetter, 1995; Knetter, 1989, 1993, 1994; Krugman, 1987; Manon, 1995). 2 The fnancal actvtes for exposure management nclude the use of currency dervatves, fnancngs through foregn-currency denomnated debt, and nternal transactons wth foregn subsdares, among others. Exstng studes offer nconclusve evdence on the effectveness of the foregn currency denomnated debt n managng exchange rate exposure (see, e.g., Bae and Kwon, 2011; Bartram, Brown, and Mnton, 2010; Clark and Judge, 2009). Smlarly, the exstng lterature shows mxed evdence on the effectveness of currency dervatves. Whle several studes (e.g., Allayanns and Weston, 2001; Allayanns and Ofek, 2001; Bartram, Brown, and Mnton, 2010; Clark and Judge, 2009; Graham and Rogers, 2002; Guay, 1999) document postve hedgng effects of fnancal dervatves, a good number of other studes cast doubt on ther effectveness n exposure management (e.g., Bal, Terrence, and Martell, 2007; Guay and Kothar, 2003; Hentschel and Kothar, 2001). In ths paper, we examne both operatng and fnancal actvtes of exposure management and argue that the mxed evdence on the effectveness of exposure management actvtes s largely attrbuted to the falure to consder the underlyng characterstcs of exchange rate exposure for each ndvdual 2 Through a survey of large euro-area companes, Döhrng (2008) shows that many frms make use of operatonal hedges ncludng matchng foregn currency revenues wth expendtures n the same currency and the nternatonal dversfcaton of sourcng, producton, and sales. 5
frm. 3 Because exchange rate exposure can be ether postve or negatve, whch would requre dfferent exposure management actvtes, t s mportant to dentfy the correct drectons of a frm s exchange rate exposure for the effectveness of exposure management actvtes. We focus on manufacturng companes n Korea, one of the premer developng countres. Korean manufacturng frms have long depended on nternatonal trades and foregn captal wth steady ncreases n ther foregn busness operatons and foregn drect nvestments over the last decade or so, whch have made ther frm values hghly senstve to exchange rate changes. 4 In addton, followng the adopton of a new accountng system n early 2000 s that requres Korean frms to report gans and losses n asset values assocated wth exchange rate changes n the current year s balance sheet, the fnancal performance of Korean frms s affected to a greater extent by the changes n exchange rates n the same year. Consequently, wth the larger swngs n the exchange rates recently, Korean frms have had much greater needs to manage ther exchange rate rsk than ever. It s well documented that Korean frms manage ther exchange rate exposure usng varous actvtes ncludng exchange rate pass-through, nternal transactons wth foregn subsdares, and dervatves products (see, e.g., Jung and Kwon, 2007). In ths regard, Korean manufacturng frms offer an nvaluable opportunty to nvestgate the relatonshp between exchange rate exposure and varous actvtes of exposure management. The results of our paper support the noton that the effectveness of exposure management actvtes depends on the underlyng characterstcs (e.g., drecton) of the frm s nherent exchange rate exposure, that s, the expected exchange rate exposure n our paper. To be more specfc, frms wth postve expected exchange rate exposure are shown to reduce ther exposure through the sell transactons of currency forwards/futures, nternal transactons wth foregn subsdares, and the ssuance of foregn currency denomnated debt, whereas frms wth negatve expected exchange rate exposure do so only 3 Other factors that may contrbute to the mxed evdence nclude dfferences n sample frms, tme perods, and/or measurement model specfcatons. 4 Snce the Asan fnancal crss n late 1997, Korean frms have been exposed to an unprecedented level of foregn exchange rsk prmarly due to the adopton of the flexble exchange rate system n late 1990s. 6
through exchange rate pass-through (wth and wthout mark-ups) assocated wth exchange rate changes. In addton, our analyss of frms wth sgnfcant observed exchange rate exposure offers evdence that structured currency forwards deterorate frms exchange rate exposure. Our results also show that whle some frms tend to over-manage exchange rate exposure, the majorty of frms properly use hedgng actvtes to manage exchange rate exposure. The overall results of our paper suggest that for the effectve management of exchange rate exposure frms need to consder both the underlyng characterstcs (e.g., drecton) of exchange rate exposure and the plausble condtons n the product market that they may face before they engage n exposure management actvtes. Relyng smply on the regresson results from a pooled sample wthout consderng ths noton would lkely yeld very lmted and unrelable emprcal mplcatons. Our paper s organzed as follows. Secton 2 presents the research desgn and data ncludng methods to measure the expected exchange rate exposure and the observed exchange rate exposure. Secton 3 reports the emprcal results, wth the summary and conclusons n Secton 4. 2. Research Desgn and Data 2.1. Measurement of expected exchange rate exposure We measure the expected exchange rate exposure nherent n the frm s basc busness actvtes by the estmaton models used by Bodnar and Marston (2002). A frm s value can be expressed usng current and future cash flows as follows: V CF t (1 t t 1 ) (1) where V = frm value; CF t = expected future cash flows, measured as the dfference between ncome after taxes and net nvestment); ρ = dscount rate. Assumng constant cash flows and zero net nvestment for smplcty, the frm value wll be represented by: 7
( 1 ) V CF (2) where τ = corporate tax rate; π = ncome before taxes. If we denote the exchange rate of local currency relatve to foregn currency as Err, then a change n frm value relatve to a change n the exchange rate, that s, exchange rate exposure, can be expressed as dv/derr. Wth a constant corporate tax rate of τ and a constant dscount rate of ρ, the exchange rate exposure can be expressed as: dv / derr [(1 )/ ] d / derr (3) Changes n exchange rates typcally affect frms engagng n the followng types of busness actvtes: (1) producng domestcally and exportng to foregn countres; (2) producng or sellng fnal products usng mported raw materals; and (3) producng same or dfferent products n foregn subsdares. As a way to consder all these types of busness actvtes, we take the case of a multnatonal enterprse that produces and sells both domestcally and n foregn countres products made of domestc or foregn raw materals. Then, the elastcty of ts frm value n terms of operatng profts to a change n the exchange rate s measured by: 1 h1 ( h1 h2 )( 1) (4) r where r = operatng proft margn; h 1 = the proporton of operatng profts desgnated n foregn currences out of total operatng profts; h 2 = the proporton of operatng expenses desgnated n foregn currences out of total operatng expenses. Equaton (4) shows that the expected exchange rate exposure nherent n the frm s basc busness actvtes can be measured usng the frm s export rato, mport rato, and operatng proft margn. In equaton (4), the relatonshp between foregn sales rato and the rato of expenses of mported materals to total operatng expenses s a basc factor n determnng the exstence of a frm s exchange rate exposure, and the frm s operatng proft margn plays a role n the determnaton of the magntude of 8
exchange rate exposure. In equaton (4), f a frm s foregn sales relatve to total sales are greater than the expenses of mported materals relatve to total operatng expenses, the frm wll be exposed to greater exchange rate exposure where ts value ncreases (decreases) as the exchange rate goes up (down). It s also shown that ths expected exchange rate exposure becomes bgger as the frm s ablty to generate profts, or the operatng proft margn, s lower. 2.2. Measurement of observed exchange rate exposure We measure the observed exchange rate exposure present n the frm s stock returns by employng the followng market model (Joron, 1990): Sr, t 0, 1, Mrt 2, Errt, t (5) where Sr = a frm s stock returns; Mr = market returns; Err = a change n won/usd rate. In equaton (5), the estmated model coeffcent of α 2, captures the observed exchange rate exposure. 2.3. Analyss of dfference n expected and observed exchange rate exposure The observed exchange rate exposure measured from the relaton between a frm s stock returns and changes n the exchange rate represents the combned outcome of the expected exchange rate exposure that results from the frm s basc busness actvtes and the effects of other actvtes (e.g., usng foregn currency denomnated debt and currency dervatves) on exchange rate exposure. Therefore, t would be mpossble to analyze separately what level of exchange rate exposure s nherent n the frm s basc busness actvtes and whether the frm s other actvtes to manage exchange rate exposure are effectve by examnng the observed exchange rate exposure alone. In order to correctly measure the effects of the frm s managng actvtes of exchange rate exposure, t s necessary to frst estmate the expected exchange rate exposure nherent n the frm s basc busness actvtes separately. Then, f the observed exchange rate exposure s excluded from the expected exchange rate exposure, one can solate the exchange rate exposure assocated wth the frm s managng actvtes of exchange rate exposure. 9
Hence, f one analyzes the dfference between the expected exchange rate exposure and the observed exchange rate exposure, called exchange rate exposure dfference (EDIF), one should be able to fnd the relatonshp between managng actvtes of exchange rate exposure and the frm s exchange rate exposure. Because EDIF reflects the outcome of the frm s managng actvtes of exchange rate exposure, an analyss of EDIF by controllng external factors that mght affect the frm s exchange rate exposure wll reveal the effectveness of the types of actvtes frms use to manage exchange rate exposure. In Fgure 1, the two types of exchange rate exposure are presented n relaton to the frms exposure management actvtes. Frms wth nsgnfcant observed exchange rate exposure are lkely to have managed effectvely ther exchange rate exposure by consderng external condtons (Bartram and Bodnar, 2007; Bartram, Brown and Mnton, 2010). Hence, f one measures EDIF of frms wth nsgnfcant observed exchange rate exposure and estmates regresson models of EDIF as dependent varable and frms managng actvtes of exchange rate exposure as explanatory varables, one would be able to fnd the effectveness of the types of actvtes used for managng exchange rate exposure. Based on ths dscusson, we estmate the followng regresson model (for smplcty we omt the subscrpt t for year): EDIF b b FWD b NSFWD J b NFCFIN b Ln( FSIZE ) b RND b IDUMMY b YDUMMY 7 0 1 8 2 3 8 j j, j 1 y 1 b PASS Y 4 8 J Y b INTR 5 y, b DIVER 6 (6) where EDIF = exchange rate exposure dfference, measured by the dfference between expected and observed exchange rate exposure. In regresson model (6), FWD, NSFWD, NFCFIN, PASS, INTR are testng varables, representng frms actvtes of exposure management, and the other varables are used as control varables. A bref descrpton of explanatory varables along wth ther measurements s gven below: FWD represents the transacton (buy and sell) amount of currency dervatves ncludng currency 10
forwards, currency futures, and currency rsk nsurance. 5 We construct FWD by separatng buy transactons (equvalent to long postons) and sell transactons (equvalent to short postons), yeldng two varables of FWD-Buy and FWD-Sell. NSFWD represents the net poston n structured forward contracts, measured by the dfference between short and long poston n structured forward contracts. The short (long) poston n structured forward contracts combnes buyng (sellng) put optons and sellng (buyng) call optons, makng a poston dentcal to sellng (buyng) currency forward. A currency KIKO (knock-n knock-out) opton s an example of the structured forward contract and was one of the nstruments wdely used for hedgng exchange rate (n partcular, US dollars) rsks n the Korean fnancal markets untl late 2008. 6 NFCFIN represents net amount of foregn currency fnancng, measured by the dfference between total foregn currency fnancng and swap transacton amount. Total foregn currency fnancng s the sum of foregn-currency denomnated short-term debt and long-term debt, lqudty long-term debt, and foregn currency denomnated bonds, whereas swap transacton amount ncludes currency swaps and currency nterest swaps denomnated n foregn currences. PASS represents a frm s exchange rate pass-through rato. Because data on a frm s exchange rate pass-through rato are not publcly avalable, no exstng studes have attempted to measure ths rato emprcally. In ths paper, we estmate each frm s pass-through rato usng the pass-through ratos of sales tems for the ndustry to whch the frm belongs n response to the exchange rate changes as reported 5 Currency rsk nsurance s a part of export nsurance system offered by Korea Trade Assurance Corporaton (KTAC) snce 2000 for exportng frms and mportng frms of raw materals and works n a smlar way to the currency forward contract offered by fnancal nsttutons. It recovers a frm s losses and collects the frm s profts assocated wth exchange rate changes by comparng the exchange rate guaranteed by KTAC and the actual exchange rate at the settlement tme. 6 The KIKO opton s desgned to offer postve payoffs to the opton holder when the Korean won moderately apprecates up to a certan predetermned rate aganst USD; n exchange, the opton holder s oblgated to take negatve payoffs when the won value deprecates sgnfcantly (see Khl and Suh, 2010). As the won deprecated unexpectedly durng the global fnancal crss n 2007 and 2008, however, the KIKO opton ncurred substantal losses to the opton holders, known as KIKO dsaster. Accordng to the Korean Fnancal Servces Commsson, as of June 2008, 519 frms held the KIKO optons n the outstandng amount of USD10.1 bllon, and 68 frms holdng overhedged KIKO postons (amounts of KIKO optons exceedng ther export amounts) reported fnancal losses of USD384 mllon, whch far exceeded fnancal gans of USD142 mllon from ther USD export revenues. 11
by the Bank of Korea. 7 As an example, let s assume that a frm produces two products, A and B wth sales of $10 mllon and $20 mllon, respectvely. If the pass-through ratos of sales tems A and B n the ndustry are 50% and 20%, respectvely, (as reported by the Bank of Korea based on 3-dgt KSIC ndustry classfcatons), then the frm s pass-through rato s computed as 30% (=100/300 x 50% + 200/300 x 20%). INTR represents the degree of nternal transactons of each frm wth foregn subsdares, measured by the amount of nternal transactons relatve to the frm s total sales. DIVER represents the degree to whch a frm s operatons are dversfed nto other lnes of busness. DIVER s ncluded n the regresson model to consder the frm s potental effect on the relaton between a frm s dervatves transactons and ts rsk level (Bartram, Brown, and Conrad, 2009). For DIVER, we employ the wdely-used Caves weghted ndex of dversfcaton based on the frm s sales (Caves et al., 1981). A hgher value of DIVER ndcates a greater dversfcaton of a frm s operatons. FSIZE represents frm sze, and s measured by the sum of the market values of common stock and preferred stock and the book value of debt. FSIZE enters the regresson n a natural log form. RND s a frm s research and development ntensty rato, measured by the rato of total research and development expenses to sales. IDUMMY s ndustry dummes, spannng twenty-two ndustres from food and beverage (KSIC 10) to publshng (KSIC 58). IDUMMY s ncluded to consder the potental dfferences n ndustres wth respect to the rsk level and the ease of hedgng (Jn and Joron, 2006). The ndustry classfcaton s based on two-dgt KSIC. Snce frms may change ther ndustry classfcatons as tme goes by, the sales tem wth the hghest actual sales s used to dentfy the frm s major ndustry. YDUMMY s year dummes and ncluded to control for fxed-tme effects such as sudden rebounds and adjustments n exchange rates durng the sample perod. The varables of FWD (FWD-Buy and FWD-Sell), NSFWD, and NFCFIN are measured as 7 Ths measurement approach s smlar to the one for the measurement of a frm s mport rato. 12
relatve to frm sze, whereas PASS, INTR, DIVER, and RND are measured as relatve to the frm s sales. 2.4 Data The sample frms n our paper consst of all Korean ndustral frms except for fnancal frms lsted on the Korea Exchange for the perod of 2007-2009. 2007 year was when the exchange rates of Korean won aganst USD steadly decreased (.e., the value of won ncreased). In contrast, the exchange rates of Korean won aganst USD ncreased rapdly due to the global fnancal crss throughout the second half of 2008, contnued to ncrease n early 2009, and then declned n late 2009. The data for export ratos and operatng margn ratos necessary to estmate the expected exchange rate exposure are collected from TS2000, the database of Korean Lsted Companes Councl. The data on stock returns necessary to estmate the observed exchange rate exposure are collected from the KCMI-SD database of Korean Captal Market Insttute. We estmate the observed exchange rate exposure on the annual bass usng daly stock returns. For exchange rate changes, we use changes n nomnal exchange rates, rather than changes n real exchange rates, consderng the relatvely smaller changes n daly nflaton rates. For each frm s dervatves transactons, we collect and examne related nformaton from the frm s busness reports and audt reports. Regardng the frm s exchange rate pass-through rato, we rely on the frm s sales reports and the ndustry report on pass-through ratos classfed by sales tems as reported by the Bank of Korea. 3. Emprcal Results 3.1. Measures of expected exchange rate exposure and observed exchange rate exposure Table 1 reports mean values of expected and observed exchange rate exposures, estmated from equatons (4) and (5), respectvely, and other varables of nterest, by ndustry. For each frm, the exposure s measured on an annual bass and then pooled together to produce an average value over the 13
three-year perod. Equaton (4) used to estmate the expected exchange rate exposure assumes a frm s maxmzaton of ts operatng profts. Hence, t s reasonably expected that for low-performng frms n a gven year, the estmated expected exchange rate exposure would have a lttle practcal meanng. Accordngly, we exclude frms whose operatng proft margns n a gven year are less than 25% of the average of all sample frms n the same year. 8 In Table 1, we nclude frms whose frm-year observatons are less than sx nto the ndustry classfcaton of Other. Panel A (B) of Table 1 shows results for frms whose estmated coeffcents of the observed exchange rate exposure are statstcally nsgnfcant (sgnfcant) at the 10% level. In each Panel, sub-panel 1 (2) reports frms wth postve (negatve) expected exchange rate exposure. The frms n Panel A.1 are characterzed by the relatonshp that ther values ncrease (decrease) when the exchange rate goes up (down) based on the expected exchange rate exposure. Lookng at the currency forward/futures postons (FWD-Buy and FWD-Sell), sell postons are larger by about fve tmes than buy postons (0.057 vs. 0.011) n all ndustres except for the three ndustres of KSIC 17 (pulp, paper, and paper goods manufacturers), KSCI 21 (medcne and medcal supples manufacturers) that engage n buyng currency dervatves only, and KSIC 46 (wholesalers and merchandse brokers) that has the same rato of buy and sell postons of currency dervatves. The largest sell transactons of currency dervatves are done by KSIC 31 (other transportaton equpment manufacturers) wth ts transactons of sell postons beng about 9 tmes those of buy postons (0.884 vs. 0.101). The foregn currency fnancng amount s on average about 3.6% of frm sze n all ndustres, wth ts hghest rato of 12.1% n KSIC 13 (textle manufacturers). The magntude of currency swaps and currency nterest swaps for the sample frms s about 37% (= 0.013/0.036) of foregn currency fnancng amount. KSIC 17 shows the largest net foregn currency debt of 11.6%, whle KSIC 31 has the largest net foregn currency assets of -10.2%, relatve to frm sze. 8 We perform robustness tests later to examne the effects of dfferent thresholds of operatng proft margn on our results. 14
When ndustres are ranked by ther effectveness of managng exchange rate exposure (that s, the hgher the expected exchange rate exposure, the greater the effectveness), KSIC 46 captures the top place, followed by KSIC 13, KSIC 31, and so on. To be more specfc, for KSIC 46, the transacton amount of buyng currency forward/futures s the same as that of sellng currency dervatves, ndcatng that currency forward/futures are lttle effectve n managng exchange rate exposure for frms n ths ndustry. On the contrary, ths ndustry has a lower rato of swap transactons to foregn currency fnancng amount and possesses net foregn currency debt. In addton, the observed exchange rate exposure has the same postve sgn as the expected exchange rate exposure, suggestng no over-management of exchange rate exposure. Frms n KSIC 13 possess substantally greater sell transacton amount of currency forward/futures than ther buy transacton amount, and relatvely large amount of foregn currency fnancng, satsfyng the condton to reduce the postve exchange rate exposure. On the other hand, the possesson of relatvely larger foregn currency assets weakens these frms ablty to reduce the postve exchange rate exposure. Combned together, these actvtes lead to negatve observed exchange rate exposure for frms n KSIC 13, mplyng an over-management of exposure. For frms n KSIC 31, the sell transacton amount of currency forward/futures s almost 9 tmes ther buy transacton amount, and the swap transacton amount relatve to foregn currency fnancng s low, leadng to a favorable condton to reduce the expected exchange rate exposure. On the other hand, ths ndustry holds relatvely large foregn currency assets, whch weaken the reducton of exchange rate exposure. Combned wth the negatve observed exchange rate exposure, the results suggest an over-management of exchange rate exposure for frms n KSIC 13. Panel A.2 shows results for frms that are characterzed by decreases (ncreases) n ther frm values when the exchange rate goes up (down). Overall, these frms have substantally larger transacton amount of buyng currency forward/futures than sellng currency forward/futures (25 tmes), engage swap transactons equal to about 47% of foregn currency fnancng amount, and possess net foregn currency 15
debt. Among ndustres, KSIC 35 (electrcty, gas, steam, and ar condtoner suppler) has the smallest average value of the expected exchange rate exposure (-20.571), engage n buy transactons of currency dervatves only, and trade larger amount of swap contracts than foregn currency fnancng. In summary, Panels A.1 and A.2 show mportant fndngs that frms wth postve expected exchange rate exposure engage n larger sell transactons of currency forward/futures, use larger foregn currency fnancng, and carry net foregn currency assets, whle frms wth negatve expected exchange rate exposure have larger transacton amount of buyng currency dervatves and carry net foregn currency assets. On the other hand, there s lttle dfference n the swap transacton amount between frms wth postve exchange rate exposure and frms wth negatve expected exchange rate exposure. Vewng from the noton that frms n Panel 1 have nsgnfcant observed exchange rate exposure, the results ndcate that frms use currency forward/futures properly n managng exchange rate exposure. Because frms wth postve expected exchange rate exposure possess foregn currency fnancng, they consder management of exchange rate exposure when usng foregn currency fnancng. In contrast, frms wth negatve exchange rate exposure also carry foregn currency fnancng, suggestng that the purpose of usng foregn currency fnancng s not lmted to the management of exchange rate exposure, and ther foregn currency assets and debt are not set up to help frms manage ther exchange rate exposure. Panel B reports results for frms wth sgnfcant observed exchange rate exposure. As can be seen n Panel B.1, frms wth postve expected exchange rate exposure carry negatve observed exchange rate exposure, whch suggests over-management of exchange rate exposure. Whle the ratos of buyng and sellng currency forward/futures for these frms are smlar to those for frms wth nsgnfcant observed exchange rate exposure (Panel A.1), ther magntudes of the net poston of structured forward contracts are much larger. These results suggest that frms havng negatve exchange rate exposure s related to the transactons of structured forward contacts ncludng the currency KIKO contracts. When the expected exchange rate exposure s negatve, frms on average have hgher foregn 16
currency fnancng rato and swap transacton rato than frms wth nsgnfcant observed exchange rate exposure (Panel A.2). The net foregn currency debt s about twce that of frms n Panel A.2. The results show that compared to frms wth nsgnfcant observed exchange rate exposure (Panel A), frms wth sgnfcant observed exchange rate exposure (Panel B) do not properly use exposure-related actvtes to reduce the negatve expected exchange rate exposure. Because the expected exchange rate exposure estmates the effects of exchange rate changes on frm value by assumng that the expected exchange rate exposure regards the overall exchange rate changes as permanent, the expected exchange rate exposure tends to estmate the degree of exchange rate exposure larger than the observed exchange rate exposure does. Hence, t may be unreasonable to drectly compare the magntudes of coeffcents of the expected and observed exchange rate exposures. However, the drecton of the expected exchange rate exposure provdes useful nformaton on the frm s management of exchange rate exposure. The results n Panel A of Table 1 suggest that frms consder the drecton of the expected exchange rate exposure n managng exchange rate exposure. In other words, the expected exchange rate exposure we estmate and report n our paper s ndeed a useful tool to separate and explan the characterstcs and behavor of frms management of exchange rate exposure. Table 2 further shows mean values of varables by year. Across the three years of 2007-2009, the expected exchange rate exposure does not exhbt a sgnfcant change, regardless of ts postve or negatve sgn. In contrast, the observed exchange rate exposure reveals much greater swngs across the three years. For the case of postve observed exchange rate exposure, both the number of frms and the average exchange rate exposure declne substantally n 2008 and then ncrease n 2009 but to a much less extent. On the other hand, whle the number of frms wth negatve observed exposure ncreases to the hghest level of 299 n 2008, the average exposure declnes gradually over the three years. Hence, durng 2008 when the won/usd exchange rate soared substantally, more Korean frms experence losses n ther frm values, but the average exchange rate exposure for these frms s lower than n 2007. Hence, these fndngs show evdence contradctory to the general belef that the magntude of value losses to 17
frms wth exchange rate exposure would ncrease when exchange rates go up. Table 2 also shows that more frms engage n buy- and sell-transactons of currency dervatves (FWD-Buy and FWD-Sell), structured forward contracts (NSFWD), and swap contracts (SWAP) n 2008 than n 2007 or 2009. The overall results n Table 2 ndcate that unlke the expected exchange rate exposure and varables related to exposure management, the observed exchange rate exposure exhbts sgnfcant swngs wth respect to both number of frms and average exchange rate exposure year by year, suggestng that as the observed exchange rate exposure shows dfferent changes than the expected exposure, t s necessary to examne these two exchange rate exposures separately. Ths noton also suggests the possblty that the changes n the observed exchange rate exposure are caused by factors other than the wdely-known varables related to exposure management. 3.2. Characterstcs of frms wth nsgnfcant observed exchange rate exposure Table 3 reports mean and medan values of the expected and observed exchange rate exposures, frm characterstcs, and varables representng exposure management actvtes for frms wth nsgnfcant observed exchange rate exposure. The results are shown for three samples, a full sample and two subsamples consstng of frms wth postve exchange rate exposure and negatve exchange rate exposure. For the IMPORT (nput-output rato of mported goods or mport rato) varable, we use the estmated nput-output rato of mported goods, whch s computed by comparng each frm s current sales (collected from TS2000) to the nput-output ratos of mported ntermedary goods (that appear on the nput-output statement reported by the Bank of Korea). For example, f a frm produces goods belongng to the manufacturers of pulp, paper, and paperboard (KSIC 171), we use the relevant ndustry s nput-output rato of mported ntermedary goods, 25.56% (2007 year bass), as a proxy for the frm s nput-output rato of mported goods. If a frm s dversfed and thus produces multple goods, we compute and employ the weghted average of the nput-output ratos of mported ntermedary goods for the multple goods as the frm s IMPORT rato. For example, f a frm produces two products, A and B, 18
n the amount of $10 mllon and $20 mllon, respectvely, and the costs for the mported ntermedary goods to produce A and B are 30% and 20%, respectvely, relatve to the total producton costs, then the frm s nput-out rato, IMPORT, s computed as 23.3% (= 100/300 x 30% + 200/300 x 20%). Lookng frst at the mean values for the full sample, the expected and observed exchange rate exposure carry postve and negatve sgns, respectvely. Whle the expected exchange rate exposure s a large number of 2.540, the observed exchange rate exposure has a very small absolute value of 0.020. Consderng that the observed exchange rate exposure reflects the outcome of exposure management, however, the fndng that the observed exchange rate exposure has a dfferent sgn than the expected exchange rate exposure suggests an over-management of exchange rate exposure. Most varables representng frm characterstcs such as export rato, mport rato, exchange rate pass-through rato, frm sze, R&D expenses, nternal transactons wth foregn subsdares, and dversfcaton ndex have hgher ratos for frms wth postve exposure than for frms wth negatve exposure. For example, whle the mean export rato s 26.5% for the full sample, there s substantal dfference n ths rato between frms wth postve expected exchange rate exposure (50.3%) and frms wth negatve exposure (3.5%). There s lttle dfference n the nput-output rato of mported goods between the two samples. Regardng currency dervatves, frms wth postve (negatve) exchange rate exposure on average engage n more transactons of sellng (buyng) currency forward/futures. For the full sample, frms use more transactons of sellng currency dervatves than buyng such contracts. The net poston of structured forwards s shown to be a net sell poston for the three samples; for example, frms wth postve expected exchange rate exposure hold about 0.4% of such contracts relatve to frm sze. Foregn currency fnancng amount s about 3.6% of frm sze for frms wth postve expected exchange rate exposure and about 2.3% for frms wth negatve expected exchange rate exposure. There s lttle dfference n the magntude of swap contracts based on the drecton of expected exchange rate exposure. Whle frms n the full sample own about 1% of net foregn currency debt relatve to frm sze, frms wth 19
postve expected exchange rate exposure have net foregn currency assets but frms wth negatve expected exchange rate exposure possess net foregn currency debt. Comparng frms wth sgnfcant observed exchange rate exposure to frms wth nsgnfcant observed exchange rate exposure, the followng characterstcs can be obtaned. Frst, for frms wth postve expected exchange rate exposure, frms wth sgnfcant observed exchange rate exposure have greater coeffcents of expected exchange rate exposure and smaller coeffcents of observed exchange rate exposure than frms wth nsgnfcant observed exchange rate exposure, suggestng the possblty of over-managng exchange rate rsk. Whle there s lttle dfference n currency dervatves transactons, frms wth sgnfcant observed exchange rate exposure tend to have more foregn currency fnancng, use less swap transactons, and hold a net poston of structured forwards about 2.75 tmes that of frms wth nsgnfcant expected exchange rate exposure. These dfferences explan at least n part the large negatve observed exchange rate exposure. Second, when frms have negatve expected exchange rate exposure, frms wth sgnfcant observed exchange rate exposure have smaller coeffcents of observed exchange rate exposure than frms wth nsgnfcant observed exchange rate exposure. Ths result confrms the noton that there s lack of an effectve management of exchange rate exposure. For frms wth sgnfcant exposure, the sze of ther buy and sell transactons of currency dervatves s the same, ndcatng no effects of exposure management through currency dervatves. In addton, these frms possess relatvely larger amount of net foregn currency debt. These results are consstent wth our fndngs of large negatve observed exchange rate exposure. Table 4 reports mean values of expected (EX) and observed (OB) exchange rate exposures for several categores classfed by dfferent relatons between these two exposures by year. Smlarly to Table 3, the results n Table 4 are for frms wth nsgnfcant observed exchange rate exposure whch thus manage ther exposure relatvely well. The categores of +EX>+OB and EX<-OB are where observed exposure s less than expected exposure (n absolute value), hence ndcatng an effectve exposure 20
management. The total number of frms belongng to these categores s 136 (36.3% out of 375) n 2007, 115 (29.4% out of 264) n 2008, and 155 (40.3% out of 385) n 2009. It s partcularly nterestng to see that the number of frms for the category of +EX>+OB declnes substantally to 32 n 2008, less than half of 69 n 2007. The categores of +EX, -OB and EX, +OB are where frms overmanage ther exchange rate exposure. Whle the number of frms gradually ncreases for +EX, -OB, t declnes substantally to 26 n 2008 (from 101 n 2007). Hence, as the exchange rate ncreases sgnfcantly n 2008, the degree of overmanagement of exchange rate exposure by Korean frms seems to have been eased to some extent. The fndngs n Table 4 ndcate that the characterstcs of the observed exchange rate exposure reflected n the frms stock returns can vary sgnfcantly dependng on the drecton of the expected exchange rate exposure. Smlarly to Table 2, the results n Table 4 also suggest that t would be necessary to nvestgate not only the observed and expected exposures separately but also the drectons of these exposures as well. 3.3 Pearson correlaton coeffcents Table 5 shows Pearson correlaton coeffcents of varables of nterest for frms wth nsgnfcant observed exchange rate exposure. Panels A and B report results for frms whose expected exchange rate exposure s postve and negatve, respectvely. The results n Panel A show that the expected exchange rate exposure s hghly postvely correlated wth EP (export rato), IT (nternal transacton rato), and FS (sell transactons of currency forward/futures). The hghly postve coeffcent (0.645) of EP ndcates that a frm s export rato explans a large porton of the postve expected exchange rate exposure, whereas the hghly postve coeffcent (0.229) of FS suggests that the sell transactons of currency dervatves are related to the frms exposure management actvtes. IP (mport rato) s hghly postvely correlated wth PT (pass-through rato) wth the correlaton coeffcent of 0.837, mplyng that frms usng relatvely large amount of 21
mported goods for ther producton tend to pass through the effects of exchange rate exposure to ther prcng polces. The hghly postve coeffcent of 0.243 between FS and FL (buy transactons of currency forward/futures) mples that currency dervatves are not merely used for the purpose of managng frms exchange rate exposure. The hghly postve coeffcent of 0.503 between FF (foregn currency fnancng) and FD (net foregn currency debt) suggests that foregn currency fnancng s mostly done wth foregn currency debt. Turnng to results n Panel B, the expected exchange rate exposure s hghly negatvely corrected wth IP and PT. Consderng the negatve sgn of expected exchange rate exposure, these results ndcate that the greater the expected exchange rate exposure, hgher nput-output ratos of mported goods and/or the more transactons of exchange rate pass-through lead to greater expected exchange rate exposure. Unlke frms wth postve expected exchange rate exposure reported n Panel A, EP s hghly postvely correlated wth SF (net sell poston of structured forwards) (correlaton coeffcent = 0.347). Hence, frms whose frm value declnes n response to an ncrease n the exchange rate tend to take greater short postons n (that s, sellng) structured forwards when they ncrease export actvtes. Ths evdence suggests that frms do not use structured forwards solely for the purpose of managng ther exchange rate exposure. 9 FD and SW (swap transactons) are hghly postvely correlated to each other (correlaton coeffcent = 0.340), showng that frms wth relatvely large amount of net foregn currency debt also engage n large transactons of swap contracts. In summary, comparng the results n Panels A and B of Table 5, the expected exchange rate exposure and the varables necessary for the estmaton of expected exchange rate exposure have n general the expected relatonshps postulated n equaton (5), but the relatonshps among these varables vary sgnfcantly based on the drecton (postve or negatve) of the expected exchange rate exposure. Vewng from these fndngs, f one analyzes the varables altogether wthout takng nto account the 9 Durng our sample perod, the majorty of structured currency forward contracts are the currency KIKO opton contracts. 22
drecton of the expected exchange rate exposure, the analyss would not produce meanngful results on the relatonshp between the expected exchange rate exposure and ts related varables. 3.4. Regresson results of dfference n exchange rate exposure Table 6 presents regresson results of dfference n exchange rate exposure (EDIF) as dependent varable estmated from equaton (6) by two subgroups of frms based on the drecton of the expected exchange rate exposure and a pooled group. We offer the results for pure manufacturng frms spannng from KSIC 10 (food manufacturers) to KSIC 31 (transportaton equpment manufacturers) by excludng frms n KSIC 35 (electrcty, gas, etc.) through KSIC 58 (publshers), as well as for the full sample. It s mportant to note that because EDIF s computed as the dfference between the expected and observed exchange rate exposure, EDIF wll have a larger value when the expected exchange rate exposure s postve and the observed exchange rate exposure s smaller (that s, the more effectve the exposure management actvty s). In contrast, when the expected exchange rate exposure s negatve and the observed exchange rate exposure s smaller, EDIF wll have a smaller value (or a larger absolute value). In other words, when the expected exchange rate exposure s postve, the larger the EDIF s, the more effectve the exposure management actvty s. When the expected exchange rate exposure s negatve, the smaller the EDIF s (that s, the larger the absolute value of the negatve EDIF), the more effectve the exposure management actvty s. If the sgns of the observed exchange rate exposure change are dfferent from those of the expected exchange rate exposure, t would be dffcult to explan the relatonshp between EDIF and the exposure management actvtes n a consstent manner, but one stll should be able to examne the relatonshp between frms exchange rate exposure and actvtes that frms use to manage ther exposure. In addton, as we postulate n ths paper, when the observed exchange rate exposure s not sgnfcant, the analyss of the relatonshp between EDIF and ts related varables allows us to examne whether frms properly use varous exposure management actvtes. 23 Accordngly, the results on Table 4 can be
nterpreted as the evdence on the relatonshp between the exchange rate exposure of frms that have performed ther exposure management properly and ther exposure management actvtes. For frms wth postve expected exchange rate exposure, EDIF s sgnfcantly postvely related to FWD-Sell, NFCFIN, and INTTR at least at the 5% level for the full sample, suggestng effectve managements of these actvtes n reducng exchange rate exposure. Frm sze s sgnfcantly negatvely related to EDIF, ndcatng that a larger frm s more passve n managng exchange rate exposure. On the other hand, NSFWD s postvely related to EDIF, but the relatonshp s not sgnfcant at the 10% level. Ths fndng offers nterestng evdence that lke exportng frms, frms whose values declne n response to a decrease n the exchange rate do not gan much from the usage of the structured currency forward/futures contracts n managng exchange rate exposure. Vewng from the noton that the postve expected exchange rate exposure means takng a long poston n foregn currency contracts, ths result s dffcult to nterpret because the short potons n structured forwards should result n the reducton n frms exchange rate exposure. One possble explanaton s that among the frms that use structured forwards, some frms use such contracts wthout consderng ther overall postons n the exchange rate exposure. Another possblty s that because the structured forward contracts (ncludng KIKO contracts) Korean frms use are desgned to be cancelled when the exchange rate declnes below a certan level, these types of dervatves are not effectve n managng exchange rate exposure resultng from long postons of a foregn currency. If the market responds by expectng ths noton, then the usage of the structured forwards would have lttle effect on the exchange rate exposure. The results for frms wth negatve expected exchange rate exposure show that NFCFIN has a postve and sgnfcant (at the 10% level) effect on EDIF, hence reflectng the postve relatonshp between exchange rate exposure dfference and net foregn currency debt. NSFWD s sgnfcantly postvely related to EDIF at the 1% level, hence ncreasng the exchange rate exposure. Ths result s not surprsng because the net poston of structured forwards has a smlar beneft-loss structure to the 24