Financial Crisis and Foreign Exchange Exposure of Korean Exporting Firms

Transcription

1 Fnancal Crss and Foregn Exchange Exposure of Korean Exportng Frms Jae-Young Cho a, Ronald A. Ratt b*, Sung-Wook Yoon c a Mnstry of Plannng and Budget, 520-3, Banpo-dong, Seocho-gu, Seoul , Korea b Department of Economcs, Unversty of Mssour-Columba, Columba, MO c Mnstry of Fnance and Economy, Government Complex II, Gwacheon , Korea Abstract Durng the Asan crss, a rse n short-term for debt relatve to short-term debt denomnated n domestc currency results n a sgnfcant declne n negatve exposure of Korean frms, wth Chaebol frms able to beneft more. Exposure of non-chaebol frms s sgnfcantly affected by maturty of debt. Results are consstent wth recent work modelng frm level phenomena durng fnancal crss that emphaszes the mportance of credt constrants, and wth observatons that exportng frms have sgnfcantly hgher foregn debt ratos and eve of crss export and foregn debt ratos are strongly correlated wth growth n sales mmedately after the crss. JEL classfcaton codes: G15; F31; F34 Key Words: Foregn exchange rate exposure; Foregn debt; Short term debt; Fnancal crss; Credt ratonng * Correspondng author. Department of Economcs, Unversty of Mssour-Columba, MO Telephone: ; fax: E-mal address: RattR@mssour.edu 1

2 Fnancal Crss and Foregn Exchange Exposure of Korean Exportng Frms In theoretcal work, Lev (1994), Marston (2001), and Bodnar et al. (2002), amongst others, defne foregn exchange rate exposure of a frm as a measure of the senstvty of ts current and future cash flows to changes n exchange rates. Gven the dffculty of accountng for all cash flow effects n an emprcal model, most emprcal work follows the approach of Adler and Dumas (1984) and focuses on the mpact of the exchange rate on market value of the frm. Contrary to expectatons, most emprcal studes have found lttle or only modest connecton between exchange rate movement and company valuaton, an outcome that may n part be due to focus on large multnatonal companes capable of actve rsk management n advanced economes. 1 Ths paper nvestgates foregn exchange exposure of non-fnancal Korean corporatons contnually lsted on the Korean Stock Exchange before and after the onset of the Asan crss usng frmlevel monthly data over 1992: :12. Consderaton of exchange rate exposure for Korean corporatons s of nterest snce the economy s more open than that of major developed countres and the foregn debt rato of Korean frms s hgher than that of other Asan countres affected by the Asan fnancal crss. 2 The dramatc ncrease n the volatlty of exchange rate and stock prces durng the Asan fnancal crss s llustrated n Fgures 1 and For example, Joron (1990), Bodnar and Gentry (1993), Amhud (1994), Bartov and Bodnar (1994), Bartov et al. (1996), Grffn and Stulz (2001), and Dodge et al. (2002) fnd lmted evdence of exchange rate exposure. In contrast, Cho and Prasad (1995) fnd some evdence of exchange rate exposure for 15% of 409 U.S. multnatonal frms and Bodnar and Wong (2003) fnd evdence of market portfolo exchange rate exposure. Frberg and Nydahl (1999) fnd that the more open the economy, the stronger s the relatonshp between return on stock market and the exchange rate. Pantzals et al. (2001) provde evdence on the nfluence of operatonal hedges for the management of exchange rate rsk of U.S. multnatonal frms. 2 The (exports+mports)/gdp rato n 2000 for Korea was about 60%, compared to 18.8% and 16.6% for the U.S. and Japan, respectvely. Allayanns et al. (2003) report that n 1996 the largest corporatons n Korea had a foregn debt to frm value rato of 21.3% compared to an average rato of the largest corporatons n Indonesa, Malaysa, the Phlppnes, and Thaland of 12.6%. 3 The potental mportance of ncrease n the volatlty of the exchange value of a currency has been noted by Bartov et al. (1996) wth regard to an ncrease n the volatlty of U.S. frm value assocated wth the breakdown of Bretton 2

3 Examnaton of exchange rate exposure over ths tme perod also attracts attenton gven the emergence of a lterature on frm-level phenomena durng fnancal crss, that ncludes contrbutons by Krugman (1999), Aghon et al. (2001; 2004), Brs and Kosknen (2002), and Schneder and Tornell (2004), amongst others, n whch credt constrants, exportng frms, corporate leverage, and foregn debt levels play key roles. 4 As shown n Table I, for Korean corporatons over 1992: :12 the unweghted frm level foregn sales rato s 28 percent and the debt equty rato s over 300 percent. The foregn debt rato of 12 percent mples a foregn debt to equty rato of 38 percent and the short-term debt rato of 66 percent aggravated problems for credt constraned frms durng the Asan crss. Systematc relatonshps between leverage, foregn debt, and exportng frms (ndcated by fracton of sales exported), pre- and post-crss can be dscerned n the data. Shown n Table II, pror to crss, Korean exportng frms compared to non-exportng frms have sgnfcantly hgher foregn debt ratos and are less levered, and that, eve of crss frm level export and foregn debt ratos are strongly correlated wth growth n sales mmedately after the crss. In our work, nformed by contrbutons to the lterature on frm level characterstcs durng a fnancal crss, we focus on the effects of leverage, foregn debt, and of short-term debt on foregn exchange rate exposure. We estmate foregn exchange exposure of each frm from 1992:01 to 2001:12 and for two sub-perods, 1992:01 to 1997:06 and 1997:07 to 2001:12. We fnd that about 30 or 36 percent of Korean frms yeld sgnfcant exposure for the entre sample perod dependng on whether we use U.S. Woods fxed exchange rate system and consequent ncrease n currency rsk for U.S. corporatons. The mplcatons of the tme varyng nature of the exchange rate exposure has been consdered by Lev (1994), Gao (2000), Domnguez and Tesar (2001a), and Patro et al. (2002). 4 A lterature has recently emerged on the theory of behavor and of emprcal phenomena at frm-level durng fnancal crss. In Krugman (1999) and Aghon et al. (2001) deprecaton causes fnancal dstress at hghly levered corporatons wth sgnfcant levels of foregn debt. In contrast, Brs and Kosknen (2002) present a model n whch fnancal problems at exportng frms precede crss. Fnancal dstress at exportng frms s then releved by government nterventon va currency deprecaton. Brs et al. (2004) report evdence supportve of the model n Brs and Kosknen (2002). Schneder and Tornell s (2004) model lnks corporate rsk takng to foregn borrowng and to a crss equlbrum n whch there are bankruptces and credt crunch. They suggest an asymmetry n whch frms n the tradable sector far better relatve to frms n the non-tradable followng crss compared to the pre-crss perod. In related work on form-level phenomena durng crss, Forbes (2002) shows that performance of commodty frms followng deprecaton s sgnfcantly enhanced f ther captal labor rato s low and nterest rates do not ncrease substantally. If the cost of captal rses followng deprecaton ths could erode the advantage of lower labor costs. 3

4 dollar or trade weghted exchange rate value of the Korean won. Most Korean frms are negatvely exposed n that stock prce declnes when the currency deprecates, a result drven by ncluson of the Asan crss perod. From cross-secton regresson analyss of the determnants of exposure we fnd that for samples ncludng the Asan crss perod, frms wth hgher foregn sales rato have sgnfcantly lower negatve exposure, and that the adverse mpact of currency deprecaton on stock returns s sgnfcantly greater for those frms wth hgh leverage and low proftablty. Over 1992:01 to 1997:06, possbly because there was lttle movement n the exchange rate, these varables are not statstcally sgnfcant n explanng exposure. Consderaton of short-term debt allows examnaton of the presence of a credt constrant. 5 For a credt constraned frm, expected change n present value of future net foregn currency revenues from a change n the exchange rate, wll be less f the frm cannot obtan short-term credt. Wth an ncrease n short-term debt, a credt constraned frm wll become more postvely exposed f that frm has postve foregn exchange rate exposure, or less negatvely exposed f that frm has negatve foregn exposure, as expected present value of future net foregn currency revenues from a change n the exchange rate rses. Before the Asan crss, non-chaebol frms are found to be credt constraned, and Chaebol frms are not credt constraned. For non-chaebol frms, whether debt s domestc or foregn doesn t matter, but maturty of debt matters. For Chaebol frms, debt beng foregn or domestc or long-term or short-term doesn t matter before the Asan crss. The null hypothess that a shft n the maturty of foregn debt has the same effect on exposure as a shft n the maturty of domestc debt s not rejected. For samples embracng the Asan crss, non-chaebol frms are agan credt constraned, but now debt beng foregn rather than domestc matters for both types of frm. Exposure of Chaebol frms s not affected by a fall n maturty of debt denomnated n ether domestc or foregn currency. However, a rse n short-term debt denomnated n foregn currency relatve to that denomnated n domestc currency 5 Aghon et al. (2001) show that n the presence of a credt constrant on a subset of agents, a real shock can be amplfed through a fnancal accelerator effect and can lead to a severe declne n output. Ths s certanly consstent wth the experence n Korea durng the Asan crss. In Aghon et al. (2004) frms beng credt constraned s ntegral to showng nstablty of open economes. 4

5 results n a statstcally sgnfcant declne n the negatve exposure of frms, wth Chaebol frms beneftng more. Thus, durng the Asan crss, frms, ncludng Chaebol frms, were credt constraned wth regard to ablty to assume debt denomnated n foregn currency. These results are consstent wth recent work modelng frm level phenomena durng fnancal crss that emphasze the mportance of credt constrants by Aghon et al. (2001; 2004) and Schneder and Tornell (2004). We hypothesze that corporate foregn debt s postvely assocated wth expected change n present value of a frm s net foregn currency revenues attendant on currency deprecaton, because hgher foregn debt ndcates capacty to captalze on compettve advantages bestowed by currency deprecaton. Ths explanaton s consstent wth the fndng by Allayanns, Brown, and Klapper (2003) that across a set of large East Asan frms, foregn currency debt n 1996 s postvely related to ncreased sales over We also fnd the same result across Korean lsted frms n our sample. These results seem to contrast wth the vew artculated by Mshkn (1999) and Krugman (1999) that hgher corporate foregn debt, durng devaluaton of the currency at the tme of the Asan crss s assocated wth declne n a frm s net worth. 6 The model of foregn exchange rate exposure s presented n Secton I. Data and descrptve statstcs appear n Secton II. Emprcal results on foregn exchange rate exposure are provded n Secton III. Secton IV concludes. I. Analytcal Framework and Hypotheses A. Measurng Foregn Exchange Exposure Snce Adler and Dumas (1984) n ther semnal artcle defned the foregn exchange exposure of a frm as the effect of exchange rate changes on the value of a frm, most researchers have measured the 6 We draw a dstncton between non-fnancal frms and fnancal nsttutons. Our fndngs only refer to the foregn debt holdng of non-fnancal corporatons. 5

6 exposure of ndvdual frms as the regresson coeffcent of stock returns on changes n exchange rate. 7 We buld on ths work by estmatng the foregn exchange rate exposure of a frm by the followng tme seres regresson: Rt = β 0 + β x FX t + β m Rmt + ε t, t = 1,2,..., T, (1) where at tme t, R t s rate of return on the th corporaton s stock, FX t s rate of change (rate of deprecaton) of nomnal exchange rate, R mt s rate of return on a market portfolo, and ε t s the regresson error. In equaton (1) β x measures the senstvty of stock returns to changes n exchange rates for frm and wll be nterpreted as a measure of foregn exchange exposure for frm. A postve sgn for β x ndcates that a corporaton benefts from deprecaton of the Korean won, whle a negatve sgn for β x ndcates that a deprecaton of the Korean Won reduces the value of a Korean frm. A market portfolo return varable s ncluded on the rght-hand sde of the regresson model (1) to control for macroeconomc factors correlated wth the exchange rate that affect the valuaton of all frms n the market. Bodnar and Wong (2000) pont out that ncluson of R mt reduces the resdual varance of the regresson and mproves the precson of the exposure estmates. A zero foregn exchange exposure found n the regresson (1) s taken to mply that frm value s affected to the same degree as the market portfolo rather than the dea that the frm s value s ndependent of changes n exchange rates. 8 A.1. Selecton of Exchange Rate 7 For example, related approaches are taken by Joron (1990), Amhud (1994), Bartov and Bodnar (1994), Cho and Prasad (1995), Chamberlan et al. (1997), He and Ng (1998), Bodnar and Wong (2003), Allayanns and Ofek (2001), Domnguez and Tesar (2001a), Dodge et al. (2002), and Patro et al. (2002). 8 Ths argument has been made by a number of researchers, ncludng Bodnar and Wong (2003), Domnguez and Tesar (2001b), and Dodge et al. (2002). However, Domnguez and Tesar (2001b) argue that t s possble that exchange rate and market return are jontly determned, and thus the exposure estmated by equaton (3) mght understate the overall mpact of a change n the exchange rate on frm returns. Chamberlan et al. (1997) examne the exposures of U.S. and Japanese bankng ndustres and extend the estmatng equaton to nclude a bank portfolo return, whch provdes some control for other ndustry-specfc sources of varaton n stock returns. 6

7 In ths paper the nomnal exchange rate s measured by Korean won/u.s. dollar exchange rate. We wll also check robustness of results by consderng a trade-weghted exchange rate ndex for Korea. Chamberlan et al. (1997) fnd that results usng blateral exchange rate are qualtatvely robust to alternatvely usng a multlateral exchange rate. 9 We note that the correlaton between Korean won/u.s. dollar exchange rate and a trade-weghted exchange rate ndex of the Korean won s 94.6 percent over our sample perod Ths hgh correlaton s consstent wth the observaton that the Korean won/u.s. dollar and Korean won/japanese yen exchange rates have a correlaton rate of 88.6 percent and that U.S. and Japan account for over 40 percent of Korean nternatonal trade over ths perod. The hgh assocaton of the Korean won/u.s. dollar and Korean won/japanese yen exchange rates over s llustrated n Fgure 2. A.2. Return Horzon Ths paper follows most of the lterature n usng a monthly return horzon to obtan frm foregn exposure estmates. Use of weekly or daly data may understate the true extent of exposure. Bartov and Bodnar (1994) argue that stock prces do not respond mmedately to changes n exchange rates because of complexty n modelng the effect of exchange rate on frm s future cash flow. Chow et al. (1997a) present evdence that suggests that the exposure of frms becomes more detectable as the return horzon lengthens. 11 A.3. Frm-level versus Industry-level Analyss 9 Wllamson (2001) notes that use of a trade-weghted exchange-rate ndex to test foregn exchange exposure wll lack power f a frm s mostly exposed to one or two currences. Domnguez and Tesar (2001b) fnd that usng a trade-weghted exchange rate s lkely to understate the extent of exposure n some countres. They fnd, for example, that 61% of Italan frms and 18% of Japanese frms that exhbt sgnfcant exposure to movements n U.S. dollar do not have sgnfcant exposure to movements n trade-weghted exchange rate ndex. 10 Trade-weghted exchange rate ndex s constructed by takng the weghted average of 19 blateral exchange rates. The weghts represent each country s proporton of total trade wth Korea. The major tradng partners and weghts are: U.S. (24.4%), Japan (19.6%), Chna (14.3%), and Hong Kong (4.9%). 11 Chow et al. (1997a) dstngush between transacton exposure from economc exposure. The former s related to the rsk that exchange rates can change n the short run between the tme a foregn currency transacton s entered nto and when t s settled, whle the latter s related to the rsk that exchange rate changes wll alter long-term future cash flows of a frm. Therefore, they argued that a longer return horzon, such as a month or more, should be adopted to measure true economc exposure. In contrast, He and Ng (1998) show that lagged exchange rate changes have no explanatory power for Japanese multnatonal frms, and Domnguez and Tesar (2001a), usng weekly return horzon, fnd consderable exposure for eght non-u.s. countres. 7

8 In ths paper we focus on frm-level effects of foregn exposure for Korean frms. Cho and Prasad (1995) and Domnguez and Tsar (2001b) note that frms wthn an ndustry group are not necessarly homogeneous n ther operatonal characterstcs or n ther fnancal strateges and movement n exchange rates may lead to offsettng effects wthn an ndustry, and that exchange rate exposure may be more of a frm-specfc phenomenon than an ndustry-specfc one. Snce ndustry groups may nclude frms wth postve and negatve exchange rsk exposure, aggregatng across such frms may result n fndng an nsgnfcant exposure coeffcent for the ndustry group, and loss of understandng of the factors underlyng foregn exposure. 12 A.4. Stablty of Exposure: Tme-Varyng Nature Lev (1994) and Patro et al. (2002), amongst other researchers, note that a frm s exposure may vary over tme snce operatonal and fnancal strateges of ndvdual frms evolve as factors such as proftablty of foregn operatons, foregn currency assets and debts, hedgng polces, etc., change. 13 We nvestgate the tme-varyng nature of foregn exchange exposure by splttng full sample nto two sub-perods: before and after the start of the Asan crss. The volatlty of exchange rate and market portfolo stock return has ncreased dramatcally snce the Asan crss as llustrated n Fgure 1. The sample perod s dvded nto two sub-perods at the begnnng of July 1997, the tme at whch the Asan crss began wth currency crss nvolvng Thaland. The foregn exposure coeffcent, β x, n equaton (1) s examned to determne whether the response of frm value to fluctuatons n exchange rate s stable across the sub-perods. A.5. Exporters or All Frms 12 For example, Joron (1991), Bodnar and Gentry (1993), Cho and Prasad (1995) and Grffn and Stulz (2001), who all use ndustry level data, fal to fnd a sgnfcant level of ndustry exposure. Wth a dfferent emphass, Allayanns and Ihrg (2001) use ndustry-level data to assess ndustry structure (and compettveness) as a determnant of foregn exposure. 13 Lev (1994) argues that lack of evdence of sgnfcant exposure n the lterature s due to the dffculty n obtanng stable measures of exchange rate exposure. Gao (2000) propose an emprcal model relatng foregn exchange exposure to a multnatonal s foregn sales and producton poston notng the potental for nstablty of exposure. Patro et al. (2002) use weekly return horzon data rather than monthly observaton to allow for the exposure to change from year to year. 8

9 Economc theory and emprcal lterature results do not suggest that exposure wll be lmted to multnatonals or frms that hghly engage n nternatonal trade. Such frms mght be the least lkely to be exposed snce they are the most lkely to have access to operatonal and fnancal hedgng strateges. On the other hand, frms wth no foregn currency revenue could be nfluenced by fluctuatons n exchange rate through compettve structure of the markets. 14 In addton, frms that have hgh levels of foregn denomnated debt could be exposed to movements n foregn exchange rates, regardless of the extent of ther other nternatonal contacts. For these reasons we nclude all frms, ncludng purely domestc as well as export-orented frms, n our emprcal work. B. Hypotheszed Determnants of Foregn Exchange Exposure The lterature on foregn exchange rate exposure of the frm has stressed that a large number of factors can potentally nfluence the sze and drecton of foregn exchange rate exposure. To revew some of these effects t s useful (essentally followng the presentaton n Lev (1994)) to express the value of frm at tme t (for smplcty we consoldate foregn countres and currences nto one) as: 1 Vt = ( 1 ) Et τ ( TRs TCs ) DDt et FDt, = 1,2,..., n s t s= t (1 + ρ), (2) where E t represents ratonal expectaton gven nformaton avalable at tme t, revenues and total costs of frm at tme s t, DD t and TR s and TC s are total FD t are net monetary lablty postons n domestc and foregn currency, e t s unts of domestc currency per unt of foregn currency at tme t, τ s a tax rate, and ρ s rsk-adjusted shareholder opportunty cost of captal. 14 For work on competton effects at ndustry-level, see Marston (2001), Allayanns and Ihrg (2001) and Bodnar et al. (2002). 9

10 Total revenues and total costs wll nclude components, whch we wll refer to as foregn sales, FS s, and mports, IM s, respectvely, that are expressed n foregn currency and need to be converted to domestc currency. Thus, TR s s expressed n domestc currency and ncludes s e sfs, and TC s s expressed n domestc currency and ncludes e sims, where es s the exchange rate at tme s t. If the effects of the exchange rate on domestc sales and purchases are gnored, the tax rate and rsk-adjusted shareholder cost of captal are taken to be constant, and α s defned as the frm s hedgng rato, the effect of a change n the exchange rate on the value of frm at tme t s gven by: V t e t = (1 α )[(1 τ ) E t s= t 1 (1 + ρ) s t ( FS s IM s + e s ( FSs IM e s s ) e ) e s t FD ], =1,2,..., n. (3) On the rght hand sde of equaton (3), the term n braces captures the expected change n present value of future net foregn currency revenues from a change n the exchange rate nclusve of effects arsng from change n compettveness and of change n future value of the exchange rate. The last term nsde the brackets captures change n value of the frm due to net foregn debt when the foregn exchange rate changes. In equaton (3), frms can reduce the effect of the exchange rate on frm value by hedgng aganst foregn exchange rate movement by ncreasng α. The effect on the rate of return on frm s stock of a change n the exchange rate, the β x coeffcent appearng n equaton (1), s defned as V / e (n equaton (3)) dvded by V (defned n equaton (2)) and s gven by: Rt 1 Vt β x = =, = 1,2,..., n (4) e V e t t t 10

11 From equatons (3) and (4), we can see that foregn exchange exposure ( β ) wll depend upon the extent of hedgng operatons and the frm s rsk-adjusted shareholder opportunty cost of captal, and could be postve or negatve dependng on the relatve magntudes of a frm s net foregn debt and net foregn revenue outlook. The effects of the exchange rate on future net revenues n foregn currency are dffcult for the market to gauge when there s nstablty n the exchange rate and n the polces mplemented to stablze the exchange rate. 15 Abstractng from the effect of foregn debt on expected net foregn revenue, equatons (3) and (4) suggest that hgher foregn debt has a negatve assocaton wth exposure. In addton, wthout addtonal structure on the model, there s no partcular expectaton regardng the assocaton of short-term debt wth exposure. Credt constrants If frms have nadequate access to credt to the extent that operatons and proftablty are mpared, then the composton of frm debt wll play a role n determnng foregn exchange rate exposure. For a frm that s not credt constraned, varaton n short-term domestc debt or short-term foregn debt wll not be assocated wth expected net foregn revenue. For a frm that s credt constraned, an ncrease n short-term domestc debt or short-term foregn debt wll rase expected net foregn revenue, snce ths frm s now somewhat less lmted n access to credt. For a credt constraned frm the term n braces n equaton (3), expected change n present value of future net foregn currency revenues from a change n the exchange rate, wll be larger f short-term debt rses. A credt constraned frm wth lmted access to short-term debt markets wll become more postvely exposed f that frm has postve foregn exchange rate exposure ( β > 0 ), or less negatvely exposed f that frm has negatve foregn exposure ( β < 0). x x x 15 For example, the won deprecaton relatve to the dollar of 5.8% durng October 1997, was followed by deprecatons of 21% and 29% over the followng two months. After reachng a peak of 1960 won/dollar on December 23 rd, the exchange rate, after further ups and downs gradually fell (the won apprecated) to the 1400 range durng Aprl 1998 and the 1250 range durng July 1998, stll well above the rate of about 890 won/dollar n early July

12 The lterature on foregn exchange rate exposure, for example Lev (1994) and Martson (2001), emphaszes that a large number of factors, ncludng foregn sales, proftablty, the debt equty rato, and foregn debt, determne β. In emprcal work, however, very lttle has appeared on the role of ether x short-term debt or foregn corporate debt on frm level foregn exchange rate exposure. 16 In ths paper, we wll focus on the roles short-term debt and foregn debt play n foregn exchange rate exposure. In dscusson of the lkely effects of dfferent varables on exchange rate exposure below we bear n mnd that Korean corporatons dd not have sophstcated foregn currency hedgng operatons, smlar to those at U.S. or Japanese multnatonal frms, snce the Korean government tghtly managed exchange rates untl the mddle of Allayanns et al. (2003) pont out that Korean frms were prevented by statute from hedgng ther foregn exchange rate rsk before the Asan fnancal crss. However, we note that although opportunty for formal hedgng s lmted, t s stll possble for a frm to reduce the absolute value of foregn exchange rate exposure ( β x ) by more closely matchng the expected effect of a change n the exchange rate on the present value of net foregn revenues and net foregn debt. Several operatonal and fnancal varables are dentfed as potental determnant varables for foregn exchange exposure. These varables can be placed n four categores: foregn operatons, frm sze, proxy varables for frm s hedgng ncentves, and ndcators for credt and/or lqudty constrants such as short-term foregn and short-term domestc debt n nteracton wth a dummy varable ndcatng membershp n a Chaebol. B.1 Foregn Operatons Most studes on the determnants of exposure demonstrate, wth a few exceptons, that a frm s foregn exchange exposure s sgnfcantly related to the level of ts foregn operatons ndcated by foregn sales and foregn trade. There are two offsettng hypotheses concernng the effect of foregn 16 Chung (1999) adopted foregn debt as one of determnants of foregn exposure of 206 Korean manufacturng frms between January 1987 and December 1992, but he found no sgnfcant effect of foregn debt on foregn exposure. Ths s also our result for the pre-asan crss subperod. 12

13 operatons on foregn exposure. One hypothess s that prmarly multnatonal frms exhbt a foregn exchange exposure due to ther nternatonal actvtes. Joron (1990), He and Ng (1998), Gao (2000), and Wllamson (2001) present emprcal results supportve of ths vew, but Chow et al. (1997b) and Domnguez and Tesar (2001a, 2001b) fnd no relatonshp between foregn operatons and exposure. It s argued for a number of reasons that companes prmarly wth nternatonal busness are less lkely to be exposed to exchange rate rsk. Allayanns and Weston (2001) note that such companes may be aware of ther foregn exchange exposures and take steps to dversfy foregn currency rsk operatonally (n equaton (4), a frm can reduce foregn exposure by mantanng net foregn debt and expected present value of net foregn revenues such that they cancel out each other) or use fnancal hedgng nstruments. Pantzals et al. (2001) demonstrate that U.S multnatonal frms that have many branches overseas, are dversfed operatonally and tend to have less exposure. In ths paper, followng most studes, foregn sales n rato to total sales ( FS ) wll be employed to measure level of foregn operatons of each frm. 17 From equaton (4) we expect a rse n FS to rase β x, whether β x s postve or negatve. Wth a currency deprecaton, f β x s negatve, FS serves as a hedge and exposure s reduced, and f β x s postve, FS serves to ncrease exposure. B.2. Frm Sze and Large Busness Groups (Chaebol) The relaton between frm sze and exposure mght be nfluenced by a frm s hedgng ncentves as well as a frm s level of engagement n foregn operatons. Two opposte hypotheses have been proposed to explan ths relatonshp. Nance et al. (1993) argue that larger frms have easer access to rsk management expertse than smaller frms and can take advantage of economes of scale n hedgng costs. Therefore, larger frms would be more lkely to hedge a currency rsk and less lkely to be exposed. Under 17 The effect of exchange rates on stock prces should be proportonal to net revenues denomnated n foregn currency that s, foregn currency denomnated revenues mnus foregn currency denomnated expenses not gross revenues. Korean frms, however, report only foregn sales and provde no useful nformaton about foregn expenses ncludng amount of mports. Nevertheless, the use of the rato of foregn sales to total sales could be a good proxy of the percentage of net foregn revenues (out of total revenues). That s, we can assume that the rato of foregn sales to total sales s proportonal to the rato of foregn net revenues to total net revenues. 13

14 ths hypothess, a large frm has small α n equaton (4). Emprcal results presented by Chow et al. (1997b) support ths vew. Domnguez and Tesar (2001b) note that snce large frms are more lkely to be multnatonal corporatons that engage n a greater level of nternatonal busness they are the most aware of foregn exchange rsk. However, n ther emprcal work, Domnguez and Tesar (2001b) fal to fnd a clear relatonshp between frm sze and exposure. The other hypothess wth regard to frm sze effect on exposure s that bgger corporatons should be more exposed to foregn exchange rsk. Warner (1977) argues that snce potental bankruptcy costs of small frms are greater, small frms have a greater ncentve to hedge than bg corporatons, and they should be less exposed to exchange rate rsk. Under ths hypothess, α s small for large frms. He and Ng (1998) show that large Japanese multnatonal frms have more exposure compared to small multnatonals. The connecton between sze, measured by logarthm of total sales (LSale), and foregn exchange rate exposure of Korean corporatons s a pror not obvous. Busness Groups and Exposure: Chaebol Effect The effect of a frm s belongng to a large busness group, or a Chaebol, on foregn exchange exposure wll be examned. 18 He and Ng (1998) state that a frm belongng to large busness group has a stronger lqudty poston and a lower probablty of bankruptcy, and thus t would tend to hedge less aganst exchange rate rsk than an ndependent frm. A dummy varable s employed to dstngush Chaebol frms from ndependent frms. The dummy varable takes the value of one f a frm belongs to a Chaebol and zero otherwse. We hypothesze that frms not belongng to a Chaebol have a greater 18 In Korea, a large busness group s often called a Chaebol. The Korea Far Trade Commsson (KFTC) defnes a Chaebol as a group of companes of whch more than 30 percent of shares are owned by the group s controllng share holders and ts afflated companes. Accordng to Bae et al. (2002), major characterstcs of Korean Chaebols are as follows: Frst, the Chaebols operate n many dfferent ndustres. Second, ther economc power s greater than ndependent frms n that the largest thrty Chaebols represent only 24.3 percent of all frms lsted n the Korean Stock Exchange (KSE), but they account for as much as 45.8 percent of the KSE s total market captalzaton as of Thrd, member frms wthn the Chaebols are connected by an extensve arrangement of recprocal shareholdng agreements. Therefore, cross-debts guarantees among member frms have been common practce n Korea. Ferrs et al. (2003) emphasze that Korean Chaebols use explct centralzed control n that they make nvestment and fnancng decsons as a group, rather than as separate ndvdual frms. 14

15 ncentve to reduce exposure to movements n foregn exchange rate, and thus reduce the absolute value of foregn exchange rate exposure ( β x ). B.3. Fnancal Soundness or Hedgng Incentves Accordng to the optmal hedgng theory, a fnancally weak frm has more hedgng ncentve n order to reduce the chance of bankruptcy. Drect emprcal nvestgaton of hedgng effect on a frm s exposure had not been conducted due to lack of data. 19 The relatonshp between frms s hedgng operatons and exposure has been nvestgated usng several proxy varables for hedgng ncentves of a frm n the lterature. We wll comment on the use of these varables n the context of Korean corporatons. Fnancal Leverage Smth and Stulz (1985) noted that hedgng operatons can reduce the probablty that a frm wll go bankrupt and thereby lower the expected costs of fnancal dstress. He and Ng (1998) and Bongn et al. (2000) argue that snce a frm s fnancal leverage would be an ndcator of fnancal dstress, frms wth hgher leverage have a greater desre to engage n hedgng actvtes. Furthermore, Chow and Chen (1998) argue that frms wth hgher leverage are more lkely to hedge exchange rate rsk to decrease the varance n frm value so that the cost of ssung debt can be reduced. Therefore, fnancal leverage would be a good ndcator for a frm s hedgng ncentves. wrtten as We note that equatons (3) and (4) mply that our exchange rate exposure coeffcent, β x, can be 19 In the U.S., for example, untl the begnnng of the 1990s, a frm s postons n dervatves were not dsclosed, because ths nformaton was consdered an mportant component of strategc compettveness. Smlarly, t s only snce the begnnng of 2001 that Korean corporatons have been requred to report n ther fnancal statements the notonal amount of dervatves beng used. 15

16 (1 τ ) = β x (1 α )[ Et TDt s= t 1 (1 + ρ) s t ( FS s IM s + e s ( FSs IM e t s ) e ) e s t FDt TDt ], = 1,2,..., n, (4 ) TD V t t where TD t s total debt of frm at tme t and the last varable on the rght hand sde of equaton (4 ), TD t / V t, s the frm s debt-to-equty rato ( DE ). If the effect of DE on the ncentves to hedge (and hence on α ) and the lkely negatve assocaton of DE wth the effect of the exchange rate on expected future net revenues s gnored, equaton (4 ) suggests that β x / DE has the same sgn as β x and that a rse n the debt equty rato s assocated wth greater foregn exchange rate exposure. If β x s negatve, the lkely negatve assocaton of DE wth the effect of a deprecaton on expected future net revenues renforces the negatve effect of DE on β x. If ncentves to hedge provded by a larger value of DE lead to a rse n α an addtonal effect on β x arses, that has the opposte sgn to x β gven by ( α / DE)[ βx /(1 α)] 16. If β x s negatve, ths hedgng effect f large enough could offset or overcome the other effects and lead to a rse n β x followng a rse the debt equty rato. Gven practce n the lterature, our prmary emprcal measure of leverage wll be taken to be the debt asset rato DA, computed as total debt/total assets. Lqudty Poston Nance et al. (1993) post that corporatons can mtgate expected costs of fnancal dstress by mantanng short-term lqudty. Froot et al. (1993) also provde a smlar argument that a frm s lqudty poston s negatvely related to hedgng actvtes. He and Ng (1998) show that Japanese multnatonals havng a strong lqudty poston tend to be more exposed to foregn exchange rsk due to less hedgng ncentves. Chow and Chen (1998) argue that frms can reduce default probablty by reservng more lqud assets and thus they have less ncentve to hedge. Therefore, a frm s lqudty poston would be a

17 good ndcator for hedgng ncentves and thus one of the potental determnants of exposure. To measure a frm s lqudty poston, each frm s current rato (CR) s employed. CR s a measure of how quckly assets can be converted nto cash or cash equvalents, and calculated as a rato of current assets to current labltes. Proftablty Optmal hedgng theory suggests that less fnancally constraned frms are more lkely to be exposed snce these frms have less ncentve to engage n hedgng actvtes. It s assumed that a frm s proftablty s also closely related to ts fnancal condton. A frm s proftablty s measured usng return on equty (ROE), whch s calculated as a rato of net ncome to total equty. B.4. Credt Constrants Corsett et al. (1999), Chang and Velasco (2000), Aghon et al. (2001; 2004), and Schneder and Tornell (2004) have emphaszed that fnancal constrants on frms were partcularly severe durng the Asan crss. Fnancal constrants wll be examned by consderng the nfluence of short-term foregn and domestc debt on exposure. To confrm that these varables are ndcators of credt constrant, the nteracton of a dummy varable ndcatng membershp n a Chaebol wth the short-term debt ratos wll also be consdered. Membershp n a Chaebol wll mply a more relaxed fnancal constrant, wth the result that the effects of deprecaton are less severe on expected net foregn revenues for a Chaebol frm than for a non-chaebol frm. B.5. Determnants of Foregn Exchange Exposure and Emprcal Model The nfluence of frm characterstcs, ncludng extent of foregn operatons, on foregn exchange exposure s examned usng cross-sectonal regresson analyss. Foregn exchange exposure beta ( ) β ) for each frm obtaned from equaton (1) are regressed on varables lnked to exposure. Followng the approach wdely taken n lterature, by representatve studes by Joron (1990), He and Ng (1998), and 17 x

18 Allayanns and Ofek (2001), we estmate cross-sectonal regressons of the exposure betas on the potental determnants of exposure as follows: ˆ β = a + a FS + a LSale + a DA + a CR + a ROE + a FD + a SD + v, 1,2,..., n (5) x = where, ) β x : frm s foregn exchange exposure FS : foregn sales to total sales rato of frm LSales : logarthm of total sales of frm DA : total debt to total asset rato of frm CR : current rato of frm ROE : return on equty of frm FD : foregn currency debt to total debt rato of frm, SD : short-term to total debt rato of frm, In order to nvestgate whether the rght-hand-sde varables n equaton (5) have smlar effects on both the negatve and postve exposure betas, we follow a procedure n He and Ng (1998). The sample of frms s dvded nto two groups: frms that yeld postve foregn exchange exposure; and frms that yeld negatve foregn exchange exposure. To acheve estmaton of the effect of the explanatory varables on both postve and negatve values of ) β x, we estmate the equaton: ˆ β = + x a0 D + a1dfs + a2dlsale + a3 ( D) DE + a4dcr + a5droe + a6dfd a7 DSD + d D d D FS d D LSale d D DA d D) CR (6) 0 ( 1 ) + 1 (1 ) + 2 (1 ) + 3 (1 ) + 4 (1 + d 5 ( 1 D) ROE + d 6 (1 FD + d 7 (1 SD + vd, n = 1,2,... n, 18

19 where D s a dummy varable that takes the value of one f ) β x > 0, and zero otherwse. II. Data and Descrptve Statstcs of Korean Frms Sample frms are lmted to non-fnancal companes lsted on the Korea Stock Exchange (KSE). In the emprcal work n ths secton on foregn exchange exposure, the sample comprses 447 companes wth complete nformaton on stock prces and fnancal statements, provdng data on the varables we use, for the entre perod from 1992:01 to 2001: We wll consder the whole perod and two subperods for analyss. The sub-perods are 1992:01 to 1997:06, a perod capturng events before the Asan crss, and 1997:07 to 2001:12, a perod coverng the Asan crss. Data on stock returns and fnancal statements of ndvdual frms are obtaned from the Korea Informaton Servce, Incorporated (hereafter KIS). 21 Each frm reports ts fnancal statement to the Korea Securtes Supervsory Board. Upon recevng the fnancal data from the Board, the KIS checks the ntegrty of the data. Snce Korean frms are not requred to report ther foregn sales rato n ther fnancal statements, the foregn sales rato data are obtaned from the collecton on foregn sales of Korean lsted frms by Natonal Informaton and Credt Evaluaton, Inc. Data frequency for stock return and exchange rate s monthly. Accountng ndcators are based on annual fnancal statements ncludng balance sheets and ncome statements. Table I Part A reports descrptve statstcs of a sample of 447 Korean frms. The descrptve statstcs nclude the mean and standard devaton (S.D.) values. The statstcs are average values of varables for each sample perod. The debt equty rato s over 300 percent over Korean frms have sgnfcant amounts of foregn debt and short-term debt, wth the average foregn debt and short- 20 About 640 non-fnancal companes were lsted on the Korean Stock Exchange at the end of December KIS s a prvately run Seoul-based frm that mantans databases on prvate and publc sector economc actvtes n Korea. Ever snce KIS was frst establshed n 1985 as the frst credt ratng company n Korea, KIS has provded comprehensve credt related servces assocated wth all fnancal and commercal transactons among corporatons and consumer ndvduals. KIS s apponted by the Korean government as a credt nformaton suppler pursuant to the Act for The Uses and The Safeguardng of Credt Informaton (July 1995). In August 1998, Credt Ratng Department was separated to establsh Korea Investors Servce, Inc. by jont venture wth Moody s Investors Servce. Informaton Busness Department changed company name to Korea Informaton Servce, Inc. 19

20 term debt representng 12.0 percent and 65.7 percent of total debt, respectvely, durng the entre sample. The current rato, gven by the rato of current assets to current labltes, averaged 1.42 over the entre sample. The average of frm total sales ncreases from 439 bllon won durng to 807 bllon won durng The rato of foregn sales to totals sales has mean value of 28.4 percent. The foregn sales rato s 3.0 percent at the frst quartle, 20.7 percent at the second quartle, and 47.3 percent at the thrd quartle. Chaebol and Non-Chaebol Frms Table I Part B reports comparatve statstcs for Chaebol and non-chaebol frms. The number of Chaebol frms s 85 n the sub-sample and 96 n the entre sample perod and n the subsample Chaebol frms compared to non-chaebol frms tend to be larger, more proftable, and to have hgher export ratos and debt asset ratos. Chaebol frms lqudty s less than non-chaebol frms lqudty reflectng the lower relatve chance of bankruptcy due to group membershp. The gap between Chaebol and non-chaebol frms leverage ratos has narrowed over tme, consstent wth the fndng of Borenszten and Lee (2002) that after the crss dstrbuton of fnancal resources s more on the bass of each frm s compettveness rather than accordng to sze or Chaebol membershp. Foregn Sales and Corporate Leverage We wll now brefly to examne bvarate relatonshps between some of the varables appearng n Table I n the lght of models on frm-level behavor around fnancal crss recently appearng n the lterature. In Parts A and B of Table II smple correlaton coeffcents are reported for frm-level export sales rato and leverage and debt varables. Schneder and Tornell (2004) argue that frms n the nontradable sector enjoy balout guarantees and fnance rsky nvestment wth foregn debt n the lead up to crss. From Part A of Table II t can be seen that the correlaton between frm-level foregn sales rato and foregn debt rato s postve and statstcally sgnfcant at the 0.01 level of confdence each year over 22 We classfed Chaebol frms based on the Chaebol lst n 1992 n the sub-sample , 1997 n the subsample , and 1997 n the entre sample

21 Also, n Part B of Table II, we see that there s a postve margnally sgnfcant connecton between change n frm foregn debt rato between 1992 and 1996 and frm foregn sales rato n 1996.Thus, pror to crss, frms n the nontradeables sector have a lower foregn debt rato than exportng frms, and also at the very least, dd not have a relatve tendency to rase ther foregn debt ratos. The result n Part C of the Table II that frm growth n sales would be postvely correlated wth fracton of sales exported s consstent wth the predcton of the Schneder and Tornell (2004) model that frms n the nontradeable sector would grow more slowly followng exchange rate crss. In the Brs and Kosknen (2002) model, exportng frms have hgh leverage and a debt overhang problem pror to deprecaton, and deprecaton s seen by government as a way of relevng fnancal dstress on exportng frms. Ths setup s consstent wth the observaton already made that there s a statstcally sgnfcant postve correlaton between foregn sales rato and leverage each year and there s weak evdence of a postve assocaton of change n foregn debt rato and foregn sales rato pror to crss. However, as shown n Table II, contrary to the Brs and Kosknen (2002) model there s a statstcally sgnfcant negatve correlaton between leverage (ether debt/asset rato or debt/equty rato) and the foregn sales rato at the frm level every year up untl 1997, and there s no statstcally sgnfcant negatve correlaton between change n leverage and the foregn sales rato at the frm level pror to crss. Fnally, Allayanns, Brown, and Klapper s (2003) result that frm foregn currency debt n 1996 s postvely correlated wth ncreased sales over for large East Asan frms, s replcated n Panel C of Table II for all frms lsted on the Korean stock exchange. In concluson, we note that pror to crss, exportng frms (ndcated by fracton of sales exported) compared to non-exportng frms have sgnfcantly hgher foregn debt ratos, are less levered, have no statstcally sgnfcant ncrease n leverage, and have only margnally sgnfcant ncreases n foregn debt rato. Post crss, exportng frms compared to non-exportng frms contnue to have sgnfcantly hgher foregn debt ratos (the correlaton s smaller), but no longer have sgnfcantly lower leverage. 21

22 III. Emprcal Results on Foregn Exchange Exposure and Its Determnants A. Foregn Exchange Exposure of Korean Frms Ths secton reports detaled emprcal fndngs of foregn exchange exposure of non-fnancal Korean companes contnuously lsted from 1992 to Table III shows cross-sectonal dstrbuton of the 447 Korean corporatons estmated foregn exchange exposure ( ) β ), as defned n equaton (1), for the full perod and two sub-sample perods. The frst seven columns n Table II report quartles of ) β x together wth the number of postve and negatve coeffcents for each sample perod. The numbers of frms wth sgnfcant postve and negatve exposure are presented n the last two columns of Table II. Panel A and panel B report the estmates from regresson equaton (1) when the change n exchange rate ( FX ) s measured by won/dollar exchange rate and by trade-weghted exchange rate respectvely. x A.1. Results for Full Sample Perod For the full sample perod, Korean frms are overwhelmngly (more than 80 percent) negatvely exposed ( ) β x s negatve). About 30 percent of frms yeld statstcally sgnfcant exposure coeffcents at the 5 percent level when the won/dollar exchange rate s adopted, and most of them exhbt a negatve sgn (128 of 136 frms). Only 8 frms have sgnfcant postve exposure. Ths fndng of sgnfcant negatve effect of foregn exposure s relatvely hgh snce most studes fnd that about 8 percent to 25 percent of ther sample frms are sgnfcantly exposed 23. The basc result s the same n the tradeweghted exchange rate exposure n Panel B, although the tendency s more pronounced. The deprecaton 23 In a study of 18 countres Dodge et al. (2002) fnd that about 8% of frms have sgnfcant exposure. Domnguez and Tesar (2001a) fnd exposure to a trade weghted exchange rate of about 14% for frms n 8 countres (the U.S. not ncluded). Bodnar and Wong (2003) fnd that about 23% of U.S. frms have sgnfcant foregn exposure. He and Ng (1998) found that 25 percent of 171 Japanese multnatonal frms yeld sgnfcant postve exposure, whle only 2 percent of ther sample frms exhbt sgnfcant negatve exposure. However, Domnguez and Tesar (2001a) fnd that exposure coeffcents are roughly evenly splt between postve and negatve values for most of the countres consdered by them and Chow and Chen (1998) argue that, wth a modfed model, most Japanese frms are negatvely exposed to foregn exchange rsk. 22

23 of the Korean won aganst the U.S. dollar has an adverse mpact on the stock returns of most Korean companes. There are at least three possble explanatons for ths result. Frst, as Chow and Chen (1998) argue for the Japanese case, negatve exposure may reflect the fact that Korea lacks natural resources and thus Korean frms rely heavly on mported materals for the producton of exports and domestc consumpton. Deprecaton of the won reduces purchasng power over foregn goods prced on nternatonal markets and ths outweghs the postve mpacts of deprecaton for exports. Second, many Korean frms have operated under a large amount of foregn debt. Currency deprecaton that mght ncrease a company s compettveness n terms of prces of export goods can be offset f short-term foregn debt servcng oblgatons of a company ncrease 24. Aghon et al. (2001) argue that foregn debt oblgatons are lkely to be partcularly deleterous durng a currency crss snce frms are lkely to fnd themselves credt constraned. Thrd, Chang and Velasco (2000) have argued that durng the Asan crss heavly ndebted frms dd not have access to new fnancng wth detrmental effects on frm value. A.2. Before and After the Start of Crss The results n Table III ndcate that before the Asan crss, over the perod 1992: :06, more than half of frms (239 of 447 frms) exhbt postve exposure, although most of these effects are not statstcally sgnfcant. 25 Conversely, Korean frms are overwhelmngly (more than 80 percent) negatvely exposed to foregn exchange rsk after the onset of the Asan crss, over the perod 1997: :12, wth about one quarter of these negatve exposures beng statstcally sgnfcant at the 5 percent 24 Domnguez and Tesar (2001a), when they try to nterpret the hgh percent of negatve exposure n Thaland (80%), also proposed the possblty that large foregn currency denomnated labltes n Thaland mght hurt Tha frms when the currency deprecates. 25 Over 1992: :06 only 3 percent of frms have statstcally sgnfcant won/dollar foregn exchange exposure at the 5 percent level. The result s slghtly hgher for trade-weghted foregn exchange rate measure. At a 10 percent sgnfcance level, only 6.0 percent of frms yeld sgnfcant foregn exchange exposure pror to the Asan crss. 23