Measuring the Degree of Currency Misalignment Using Offshore Forward Exchange Rates: The Case of the Korean Financial Crisis

Transcription

1 Commens Welcome Measuring he Degree of Currency Misalignmen Using Offshore Forward Exchange Raes: The Case of he Korean Financial Crisis Daekeun Park*and Changyong Rhee** February, 2000 Absrac This paper proposes a new mehod of measuring he degree of currency misalignmen hrough he use of offshore forward exchange raes. Using defaul risk adjused no-arbirage condiions for forward exchange conracs, we calculae he spo exchange raes and he domesic ineres raes ha are implied from he observed forward exchange raes. The difference beween he implied and he observed spo exchange raes is our measure of currency misalignmen. Our mehodology is based on he presumpion ha, during a currency crisis, offshore forward exchange raes reflec marke fundamenals more closely han onshore spo and forward exchange raes. The laer are usually ighly regulaed and heavily affeced by governmen inervenion during a non-normal even such as a financial crisis. We apply he mehod o he Korean financial crisis in 1997 and discuss is implicaion for evaluaing he IMF adjusmen program and explaining foreign capial flows. ** We are graeful o Mark Bils, Jang Ok Cho, June Whan Im, In June Kim, Per Krusell, Marin Schneider, Alan Sockman and workshop paricipans a he Universiy of Rocheser and Seoul Naional Universiy for helpful commens and suggesions. We hanks Jong Min Kim for excellen research assisance. This research is financially suppored by he Korea Research Foundaion. * Hanyang Universiy, Seoul, Korea, parkdk@ .hanyang.ac.kr ** Seoul Naional Universiy & Universiy of Rocheser, rhee5@plaza.snu.ac.kr JEL Code: F3, F4 Key words: currency misalignmen, covered ineres pariy, non-deriverable forwards, Korean financial crisis 1

2 I. Inroducion This paper proposes a new mehod of measuring he degree of currency misalignmen hrough he use of offshore forward exchange raes; i hen applies his mehod o he 1997 Korean financial crisis. Unlike he Lain American deb crises of he 1980s, which were largely he resul of mismanaged moneary and fiscal policy, Korea's crisis originaed in he counry's background srucural problems including excessive deb financing and weak regulaion of financial inermediaries. (Corsei, Preseni and Roubini (1998), Fischer (1998), Krugman (1998), Park and Rhee (1998)) In paricular, many researchers believe ha he Korean won was no significanly overvalued in he monhs before he Korean financial crisis. (Chinn (1998), Goldsein (1998), Goldfajn and Baig (1998), Lane, e al. (1999), and Lee (1997)) 1 However, his judgemen has no been well received ouside academic circles. In he wo-monh period from Ocober o December in 1997, he Korean won depreciaed by 53 percen agains he US dollar. Such a drasic drop of currency values has led many o believe ha he Korean won was eiher significanly overvalued on he eve of he crisis or became irraionally undervalued immediaely afer he crisis began. In eiher case many nonacademics do no easily accep he academic verdic ha he foreign exchange marke provided accurae price signals during he crisis period. This sharp disagreemen is no surprising given ha here is no consensus on how o measure he degree of currency misalignmen even among economiss. The economic lieraure suggess hree disinc approaches o measuring currency misalignmen, summarized in Chinn (1988), Edward (1994), and Williamson (1994). The firs approach is based on a simple no-arbirage condiion such as purchasing power pariy. The second is based on a formal macroeconomic model of exchange rae deerminaion. The hird is based on he concep of long-run solvency and curren accoun susainabiliy. Even leaving aside he concepual concerns in selecing an accurae definiion of currency overvaluaion, however, here exiss a pracical problem in implemening any of hese esimaion mehods: during a non-normal even such as a financial crisis, he resuls are highly dependen on he assumpions and models used. 2 (Bayoumi e al (1994)) 1 In a very horough empirical sudy, Chinn (1998) concludes ha he Korean won was undervalued, no overvalued, prior o he crisis while he currencies of he oher crisis-hi Asian counries were slighly overvalued. 2 For example, we calculaed he degree of overvaluaion of he Korean won by varying he base periods in which we apply purchasing power pariy. The degrees of overvaluaion of he Korean won agains he U.S. dollar a he end of 1997 are esimaed o be 10 percen, percen, and 3.7 percen, respecively, depending on which base year is chosen. Three differen base periods are compared during which he curren 2

3 To address his issue, we inroduce a new mehod of measuring currency misalignmen, paricularly during and immediaely preceding a financial crisis. Using he observed offshore forward exchange raes, we calculae he spo exchange rae implied from he covered ineres pariy. The difference beween he implied and he observed spo exchange raes is hen our new measure of currency misalignmen: i measures he deviaion of he observed spo exchange rae from he level ha he paricipans in he offshore forward exchange marke expec o prevail. 3 Our mehod reas he forward and he spo exchange rae differenly in a no-arbirage condiion, since we believe ha during he Korean financial crisis he former refleced marke senimen more closely han he laer. The presumpion is based on our use of non-deriverable forwards (NDFs) which are raded in he offshore and no he domesic marke. During he currency crisis, he Bank of Korea inervened in he domesic spo and forward exchange markes in order o uphold he value of he won. However, as we will explain in secion II, NDFs for he Korean won were raded in Hong Kong and Singapore wihou he direc regulaion and inervenion of he Korean governmen. By ha accoun, we believe ha he forward exchange rae in he NDF marke could reflec marke senimen more closely han he domesic rae. One problem wih our approach is ha he covered ineres pariy in general does no hold during a financial crisis if foreign invesors fear ha he local governmen may impose regulaions ha preven he free movemen of foreign funds or if he defaul probabiliy of borrowers increases significanly. 4 To ake care of his problem, we adjus he covered ineres pariy by explicily considering he poliical or he defaul risk. These risks are esimaed using he yield spreads beween a dollar-denominaed Korean governmen bond and he U.S. Treasury bond of a similar mauriy, which is a generally acceped measure of sovereign risk premium. In addiion, our mehod calculaes no only he implied spo exchange raes bu also he implied ineres raes on won-denominaed bonds by solving no-arbirage condiions for forward conracs accoun of Korea was nearly balanced: he period from he firs half of 1985 o he second half of 1986, he period from he second half of 1989 o he firs half of 1990, and he year The resuls are also sensiive o oher alernaive assumpions such as he choice of price indices, including or excluding ime rends in measuring equilibrium exchange raes, ec. 3 I is imporan o undersand ha he Korean won is no an inernaional currency and here is no offshore spo exchange marke for he Korean won. Therfore, even if foreign invesors believe ha he Korean won is overvalued, here is a limi ha hey can exploi he arbirage opporuniy by selling won. As such, he implied spo exchange raes in our mehod can deviae from he observed spo exchange raes. However, when he Korean won is undervalued, i is rue ha arbirage ransacion can be made by buying won a he onshore exchange marke. 4 Aliber (1973) discusses he role of poliical and counry risk in he forward exchange marke. Durbin and Ng (1999) sudies he effec of counry risk in emerging marke bond pricing. 3

4 wih differen mauriies simulaneously. The difference beween he implied and he observed domesic ineres raes has some ineresing implicaions for evaluaing he high ineres rae policy of he IMF adjusmen program in Korea. The empirical resuls sugges ha he Korean won was overvalued agains he U.S. dollar by 4 o 11 percen on he eve of he counry's crisis in November These esimaes are significanly larger han hose in previous sudies. (Chinn (1998), Goldsein (1998), Goldfajn and Baig (1998), and Lee (1997)) Our resuls also show ha he sharp depreciaion of he won agains he dollar in December 1997 made he won undervalued in he beginning of However, conrary o popular belief, he degree of under-valuaion was no large enough o resume capial inflows ino Korea. Afer hen, he Korean won became significanly overvalued again in he middle of 1998, when he Korean governmen sared o aggressively ease is moneary policy sance and a new round of financial crises in emerging markes developed in Russia and Brazil. Only since he end of 1998 does our mehod find no significan misalignmen. The paper is organized as follows. Secion II explains our mehodology and he insiuional srucure of he NDF marke for he Korean won. In secion III, we repor he esimaed defaul probabiliy and he degree of currency misalignmen during he period bordering he Korean crisis in In addiion, he domesic ineres raes implied from he covered ineres pariy are compared wih he observed raes. Secion IV relaes our findings o he paerns of foreign capial flows during he Korean financial crisis, and secion V concludes. II. Mehodology If capial is perfecly mobile across borders, he covered ineres pariy should hold beween spo and forward exchange raes. 5 By reaing he observed forward exchange rae as exogenous, we calculae he implied spo exchange rae ha saisfies he covered ineres pariy. The raio of he implied spo exchange rae over he observed spo rae is our measure of currency misalignmen. Needless o say, spo exchange raes, forward exchange raes and ineres raes are all endogenous variables. Therefore, i generally does no make sense o rea one variable as exogenous and he oher as endogenous in he covered ineres pariy. In pracice, however, i is easy o find such cases. For example, currency raders ofen quoe he forward exchange raes by looking a curren ineres raes and spo exchange raes and using he covered ineres pariy formula. In 4

5 conras o heir approach, our mehod swiches he role of he spo and he forward exchange raes since we believe ha he laer refleced marke senimen beer han he former immediaely before and during he Korean crisis in This presumpion depends crucially on our use of nondeliverable forwards (NDFs) which are raded in he offshore foreign exchange marke. NDFs are derivaives ha allow paricipans o hedge exposure o currencies in which a naural forward marke does no exis or is no accessible because of conrols imposed by local auhoriies. These derivaives are called non-deliverable since NDF selemens a mauriy are made in US dollars a he prevailing exchange rae and no local currency changes hands. Currenly, wo of he bigges Asian NDF markes are for he Korean won and he New Taiwan Dollar, which are mosly raded in Hong Kong and Singapore. In 1997, he daily NDF ransacions were close o US$250 million for he Korean won and US$150 million for he New Taiwan Dollar, respecively. Even hough mauriy daes are sill fairly shor, usually a wo years or less, he marke size has seadily grown ogeher wih he growh of rade volume and capial flows wihin he region. During he currency crisis, he Bank of Korea acively inervened in he domesic spo and forward exchange markes and ighly regulaed capial ransacions o uphold he value of he won. However, since he NDF marke for he Korean won was no under he direc regulaion and inervenion of he Korean governmen, is forward exchange rae could have refleced marke senimen more closely han he spo and he forward raes in he domesic marke. Therefore, we would like o inerpre he spo exchange raes implied from he NDF forward exchange raes as he equilibrium raes ha would have prevailed had he governmen no inervened in he domesic marke. We can see supporing evidence in Figure 1, which depics he movemen of he one-monh forward exchange raes in he domesic forward marke and he NDF marke ogeher wih he spo exchange rae. During he firs half of 1997, he forward exchange raes in boh markes closely followed he spo exchange rae. Saring from July 1997 he NDF forward exchange rae began rising faser han he spo rae, indicaing ha he marke was anicipaing depreciaion of he won. 6 On he oher hand, he domesic forward exchange rae coninued o move closely wih he spo rae. This migh be in par due o he hinness and inefficiency of he domesic forward marke, bu was 5 McCormick (1979) and Clinon (1988) discuss empirical evidence supporing he covered ineres pariy. 6 However, he one-monh forward premium never exceeded 20 won per dollar from July o Ocober 1997, indicaing ha wha invesors anicipaed was no a currency crisis bu a smooh depreciaion of he won. Only in lae Ocober did he inernaional capial marke anicipae he upcoming crisis in Korea and he forward premium ake a big jump. 5

6 mosly aribuable o he acive forward inervenion by he Bank of Korea. 7 < Figure 1: The won-dollar Spo and Forward Exchange Raes > We propose four differen mehods of compuing he implied spo exchange rae from he covered ineres pariy. II.1 Mehod I The firs mehod uses he observed ineres raes on won-denominaed and dollar-denominaed bonds o calculae he implied spo exchange raes from he NDF forward exchange raes. Le S denoe he spo exchange rae of he Korean won expressed as he unis of he Korean won per U.S. dollar. Le F n, sand for is n-monh forward exchange rae a he NDF marke, i n, he annualized ineres rae on won-denominaed bonds wih n-monh mauriy, and i * n, he annualized ineres rae on dollar-denominaed bonds wih n-monh mauriy a ime, respecively. If capial is perfecly mobile across borders, he following covered ineres pariy should hold: F n, = S (1 + i n, n M )/(1 + i * n, n M ), where n M = n/12 (1) 8 However, i is imporan o remember ha invesors are covered only from he exchange rae risks in he covered ineres pariy. The domesic ineres rae in he pariy is only a promised ineres rae. I does no necessarily reflec he expeced reurn for foreign invesors especially during a currency crisis when poliical and defaul risks ge significanly large. Therefore, we have o adjus he covered ineres rae pariy by explicily considering hese risks before applying our mehod. Le p denoe he probabiliy ha a won-denominaed bond will be on defaul wihin a year from ime and le χ denoe he salvage value expressed as proporion of iniial invesmen in case of defaul. χ is a number beween 0 and 1. In general, χ is no equal o zero since here exiss 7 Park and Rhee (1998) discuss he iming and he magniude of foreign exchange marke inervenion by he Bank of Korea in his period. 8 n is inroduced o conver annualized ineres ras o an effecive n-monh raes. N-monh ineres raes are M 6

7 possibiliy of deb rescheduling or deb resrucuring in pracice. In secion III.1, we will explain how p and χ are esimaed bu for he ime being, le us assume ha hey are known parameers. If we assume ha foreign invesors are risk neural and require he same expeced rae of reurn from domesic and foreign invesmen, he following defaul risk-adjused covered ineres pariy should hold: 9 F n, = S {(1+ i n, n M )(1- p n M )+ χ p n M }/{1+ i * n, n M }, where n M = n/12 (1') By reaing he forward exchange raes and he ineres raes as exogenous, we can calculae he implied spo exchange rae, Ŝ, which saisfies he risk adjused pariy condiion (1'): ^ S [(1 + i n, (1 + i )(1 p n ) * = n, n, M (2) n F M n M ) + χp n M ] We measure he degree of currency misalignmen by he raio of he implied spo exchange rae o he observed rae, ˆ S. If S ˆ / S S / >1, he observed spo exchange rae is lower han he level ha paricipans in he NDF marke expec o prevail and i indicaes ha he Korean won is overvalued agains he dollar. 10 On he oher hand, if undervalued agains he dollar. S ˆ / S <1, we inerpre ha he Korean won is II.2 Mehod II Mehod I uses he observed and foreign ineres raes o compue he implied spo exchange raes. Thus, i implicily assumes ha foreign invesors have unresriced access o he domesic securiy marke and ha he observed domesic ineres raes properly reflec he rae of reurn ha foreign invesors expec o earn by invesing in won-denominaed bonds in Korea. However, he wondenominaed bond marke in Korea had no been fully open o foreigners unil afer he crisis, so he approximaed as n/12 of he annualized ineres raes. 9 Since i is he risk-free foreign ineres rae, equaion (1) is rue only when foreign invesors are risk * n, neural. 10 Noe ha all exchange raes are in erms of he Korean won agains he US dollar. 7

8 observed domesic ineres raes may no properly reflec he rae of reurn ha foreign invesors could earn by invesing in Korea. 11 Moreover, during he Korean crisis, he domesic ineres raes were heavily regulaed and a severe credi crunch widened he ineres rae differenial beween he curb marke and he insiuionalized marke. Therefore, he observed domesic ineres raes could no properly represen he marke rae of reurn for domesic invesors le alone for foreign invesors. To he exen ha domesic invesors could have demanded higher curb marke premium a ha ime, we believe foreign invesors could have done he same in he won-denominaed bond marke. Since he observed domesic ineres raes seem o be a poor measure of he rae of reurn from invesing in won-denominaed bonds, Mehod II ries o infer he domesic ineres raes direcly from he covered ineres pariy. 12 In oher words, Mehod II calculaes simulaneously he implici spo exchange rae and he domesic ineres raes ha saisfy he covered ineres pariy. To do ha, we need o use more han one no-arbirage condiion and hereby rely on he covered ineres pariies ha hold for forward exchange conacs wih differen mauriies. For example, equaions (3) and (4) are he risk adjused covered ineres pariies beween he spo and he forward exchange raes wih 3 and 6 monh mauriies, respecively. F 3, = S {(1+ i 3, 3 n )(1- p 3 n )+ χ p 3 n }/{1+ *, i3 3 n }, where n 3 = 1/4 (3) F 6, = S {(1+ i 6, 6 n )(1- p 6 n )+ χ p 6 n }/{1+ *, i6 6 n }, where n 6 = 1/2 (4) If he yield curve of he domesic ineres raes is fla, i.e., if he condiion i 6, = i, we can calculae he implied spo exchange rae, Ŝ, and he implied ineres rae, 3 holds, hen ˆ (n = 3, 6) from equaions (3) and (4). As in Mehod I, he raio of he implied spo exchange rae o he observed rae, S ˆ / S, is he measure of currency misalignmen. We will also analyze he behavior i n, of he difference beween he implied and he observed domesic ineres raes, ˆ. i n, / in, 11 This does no necessarily mean ha arbirage was no possible beween he NDF and he domesic forward exchange marke prior o he crisis. As we will see shorly, he covered ineres rae pariy was saisfied reasonably well prior o he crisis. 12 Mehod II coninues o assume ha observed ineres raes on dollar-denominaed deposis properly reflec foreign invesors' opporuniies. 8

9 II.3 Mehod III I seems unrealisic o assume ha he erm premium was zero especially during such a urbulen period like a financial crisis. Credi crunches, increased uncerainy, and changes in expeced inflaion raes make shor-erm ineres raes and he slope of he yield curve change wildly. Mehod III inroduces a non-zero bu consan erm premium α o mehod II. i 6, = i, 3 + α (5) Since anoher unknown parameer α is inroduced, we need one more condiion; we assume ha he sample mean of he implied hree monh ineres raes on won-denominaed bonds is equal o he sample mean of he observed hree monh ineres raes: ˆ = i 3, i 3, (6) To check he robusness of our empirical resuls, differen sample periods will be examined in applying equaion (6). 13 In summary, Mehod III calculaes he spo exchange rae, he domesic ineres raes, and he erm premium from equaions (3), (4), (5) and (6). II.4 Mehod IV To calculae he consan erm premium α, Mehod III has o choose a specific sample period o equae he sample mean of he implied ineres raes wih ha of he observed raes. Insead of relying on his arbirary assumpion, Mehod IV calculaes he unknown erm premium by using addiional covered ineres pariy for a differen mauriy. For example, in addiion o he covered ineres pariies for 3-monh and 6-monh mauriies, Mehod IV uses one more pariy condiion for 13 Alernaively, we can use equaion (6'), which explicily considers he sovereign defaul probabiliy and equalizes he expeced reurns. The esimaed resuls, however, are very similar. 9

10 a forward conrac wih 1-monh mauriy. F 1, = S {(1+ i 1, 1 n )(1- p 1 n )+ χ p 1 n }/{1+ i * 1, n 1}, where n 1 = 1/12 (7) Like Mehod III, we sill assume ha he slope of he yield curve is consan a ime as expressed in equaions (8) and (9). Bu, in Mehod IV, he monhly erm premium, β, is imevarying. i i = i 2 β (8) 3, 1, + 6, i1, + = 5 β (9) From equaions (3), (4), (7), (8) and (9), Mehod IV calculaes he implied spo exchange rae, he domesic ineres raes, and he erm premium while reaing he forward exchange raes and he foreign ineres raes as exogenous. Among he four mehods, we hink Mehod I is inferior o he oher hree since he observed domesic ineres raes did no properly reflec he rae of reurn ha foreign invesors could ge from invesing in won denominaed asses. Considering he severiy of he dual srucure of he Korean financial marke during he crisis, we prefer o esimae he domesic ineres raes raher han use he heavily regulaed observed domesic ineres raes. Also, Mehod IV seems superior o Mehods II and III since i allows ime varying erm premiums. However, Mehod IV has is own weakness. Our mehodology relies on he presumpion ha NDF exchange raes were relaively unaffeced by he exchange marke inervenion of he Korean governmen. This presumpion migh be less valid for he forward exchange raes wih shorer mauriy if invesors believe inervenion policy could be effecive a leas in he shor run; herefore, Mehod IV migh be less valid han Mehods II and III. In inerpreing he empirical resuls, we will rea Mehods II, III and IV equally and use Mehod I as a benchmark for comparison. III. Empirical Resuls ( 1+ iˆ 3, n3 )(1 pn3 ) + χ pn3 = (1 + i 3, n3 ) (6') 10

11 III.1 Defaul Probabiliy In his secion, we explain how we esimae he defaul probabiliy of won denominaed bonds, p, and he proporion of salvage value from iniial invesmen, χ. Before he crisis eruped, he won-denominaed governmen bond marke in Korea was compleely closed o foreigners, so ha heir yield could no properly reflec he rae of reurn ha foreign invesors could earn by invesing in Korea. Therefore, in our covered ineres pariy condiion, we consider he foreign invesors who invesed on he won-denominaed privae bonds issued by large Korean conglomeraes (Chaebols) or big financial insiuions. We assume ha he defaul probabiliies of hese large conglomeraes or financial insiuions were equal o he sovereign defaul probabiliy of Korea. 14 Sovereign defaul in general means he governmen's inabiliy o pay back is foreign currency denominaed debs, and i has concepually nohing o do wih he defaul of he local currency denominaed debs. However, by assuming ha he defaul probabiliies of he won denominaed and foreign currency denominaed debs are equal, we are implicily assuming ha large conglomeraes or financial insiuions will bankrup and defaul heir privae debs if sovereign defaul occurs. This assumpion seems consisen wih a long-sanding radiion of he credi raing indusry o observe he sovereign ceiling, ha is, no corporae deb has a raing higher han ha of he home governmen. The sovereign defaul probabiliy of Korea can be inferred from he yield spread beween a dollar denominaed Korean governmen bond and he U.S. Treasury bond of a similar mauriy, which is a generally acceped measure of sovereign risk premium. However, since he Korean governmen began o issue dollar-denominaed bonds only in he beginning of 1998, we use he Korean Developmen Bank (KDB)'s global bonds o esimae he sovereign defaul probabiliy. The KDB is owned by he governmen and is bonds have been reaed as a near-sovereign benchmark securiy for Korea in he inernaional capial marke for a long ime. Figure 2 shows ha he yield spread beween he KDB bond and he U.S. Treasury bond was less han 1 percen in he firs half of However, afer he Thai crisis eruped in July 1997, i began rising sharply and reached 5 percen a he peak of he Korean crisis in December Afer he deb resrucuring deal was made beween he Korean governmen and he credior group in New 14 Given he exisence of a few dominan conglomeraes in he Korean economic srucure and he implici 11

12 York in January 1998, he yield spread declined emporarily. Bu i soared again as soon as he inernaional financial urmoil worsened and a new round of financial crises in Russia and Brazil sared in he second half of The yield spread peaked in Ocober 1998, reaching almos 9 percen. < Figure 2: The Yield Spread of he KDB Global Bond > From he yield spread on KDB bonds, we can compue he implied probabiliy of he sovereign defaul of Korea as follows. Le p, i KDB and i TB be he sovereign defaul probabiliy of Korea wihin a year, and he annualized ineres raes of he KDB and he U.S. Treasury bonds, respecively. If foreign invesors are risk neural and demand he same expeced rae of reurn from domesic and foreign invesmen, he following condiion should be saisfied. n n n n n 1+ i ) = (1 + i ) (1 p ) + (1 + i ) χ [1 (1 p ) ] (10) ( TB, KDB, KDB, In equaion (10), n denoes he mauriy of he bonds and we implicily assume ha he bonds are kep unil mauriy. 15 By plugging in he observed yield spread of he KDB bond and choosing an appropriae number for χ, we can calculae he implied probabiliy of sovereign defaul from equaion (8). Figure 3 shows he esimaed defaul probabiliies when χ is se o 0.0, 0.25, and 0.5. Judging from he recen deb resrucuring oucomes beween he Korean governmen and he foreign crediors, i seems realisic o assume ha χ is no larger han No surprisingly, he esimaed probabiliies show he same paern as ha of he yield spread on he KDB bond presened in Figure 2. I is noeworhy ha he foreign invesors' esimaes of he defaul probabiliy reached is peak when a new round of emerging marke financial crises sared in Russia and Brazil in he governmen guaranees on financial insiuions, his assumpion does no seem o be unreasonable. 15 Under an alernaive assumpion ha he bonds are sold before mauriy, we can use equaion (10'). Equaion (10') implicily assumes ha he yield o mauriy of a long erm bond is equal o he rae of reurn ha an invesor expecs o earn even if he bond is sold before mauriy. The esimaed sovereign defaul probabiliies urn ou o be no sensiive o he use of equaion (10) or (10'), bu are sensiive o he choice of χ. (1 + i TB, ) = (1 + i KDB, )(1 - p ) + χ p (10') 16 For example, on January 22, 2000, he Korean Resrucuring Coordinaion Commiee (KRCC) and Daewoo foreign crediors seering commiee agreed on a recovery rae of 39 o 40 percen of loans made o Daewoo corporaion wihou any collaeral. 12

13 middle of 1998, long afer he financial crisis eruped in Korea. In he following, we are going o assume ha he esimaed sovereign defaul probabiliy is equal o he defaul probabiliy of he won denominaed bonds. < Figure 3: Sovereign Defaul Probabiliy > III.2 Observed and Implied Spo Exchange Raes Figure 4 depics he raios of he implied spo exchange raes o he observed spo raes, S ˆ / S, from Augus 1996 o March These raios are calculaed by he four mehods inroduced in secion II. S ˆ / S >1 indicaes ha he Korean won is overvalued agains he U.S. dollar. Our daa se consiss of he NDF exchange raes, LIBORs and he won-denominaed CD ineres raes wih 1, 3, and 6 monh mauriies. 17 To esimae he erm premium α in Mehod III, he pre-crisis sample period from Augus 1996 o July 1997 is used in equaing he sample means of he observed domesic ineres raes wih ha of he implied ineres raes. 18 July 1997 was he poin a which he crisis in Thailand sared o spread o oher Asian counries. In Figure 4, we show he resuls when χ is equal o 0.25 as a represenaive case. For he oher values of χ, he resuls are quie similar; we repor he corresponding figures and ables when χ is equal o 0 in he appendix. < Figure 4: The Degree of Currency Misalignmen > Several feaures of he figure are worh noing. Firs, he movemen of he S ˆ / S raio shows similar paerns across all four mehods. In paricular, he raio S ˆ / S had been quie close o one 17 The qualiaive aspecs of he empirical resuls are quie robus o he choice of ineres raes. Use of he U.S. T-bill raes insead of LIBOR and use of oher domesic ineres raes such as he call raes, he moneary sabilizaion bond yields, ec., provides similar resuls. 18 Two oher sample periods - he whole sample period and he pos crisis sample period - are examined also, bu he resuls in Figure 4 do no change much. This does no mean ha he esimaed erm premium α is close o zero. To he conrary, he esimaed erm premium is sensiive o he choice of sample periods. However, he difference in α does no grealy affec he magniude and he behavior of he implied spo exchange rae. 13

14 unil he Thai crisis sared in July 1997, indicaing ha he de faco arbirage beween he onshore and he offshore forward exchange markes was effecive and he covered ineres pariy held reasonably well prior o he crisis. 19 Therefore, adoping a mehodology based on he covered ineres pariy is no oally unrealisic even hough he domesic bonds marke in Korea was no fully open o foreign invesors prior o he crisis. Second, Figure 4 shows ha he Korean won was sharply overvalued during he period beween he erupion of he Thailand crisis in July 1997 and Korea's signing of a sand-by arrangemen wih he IMF on December 1, The degree of overvaluaion agains he U.S. dollar as of November 1997 is esimaed o be as large as 8 o 11 percen according o Mehod I, II and III. Mehod IV presens a smaller degree of overvaluaion of 4 percen. I urns ou ha he smaller degree of overvaluaion from Mehod IV is largely aribuable o is use of he one monh forward exchange raes raher han he difference in he mehodology per se. When one and hree monh forward raes are used insead of hree and six monh forward raes in Mehod II and III, he degree of overvaluaion decreased from 8 o 6 percen. As previously noed, he NDF exchange raes wih shorer mauriy were more likely o be affeced by he anicipaion of inervenion in he onshore exchange markes by he Korean governmen han he NDF raes wih longer mauriy. In oher words, compared wih he forward raes wih longer mauriy, he one monh forward exchange raes may have depreciaed less han hey should have on he eve of he crisis. If his is he case, he one monh forward raes make he spo exchange raes look relaively less overvalued han do he forward raes wih longer mauriy. The fac he difference of he degree of currency misalignmen is no large across Mehods II, III and IV excep on he eve of he crisis indirecly suppors our inerpreaion. In summary, he esimaed degree of currency overvaluaion on he eve of he Korean crisis is significanly larger han hose found in oher sudies, which do no exceed 4 percen (Chinn (1998), Goldsein (1998), Goldfajn and Baig (1998), and Lee (1997)). We believe ha he overvaluaion of he won was mosly aribuable o he heavy foreign exchange marke inervenion by he Bank of Korea (BOK) from Ocober o November Despie is effor o uphold he value of he won, Figure 4 shows ha he invesors in he NDF marke sared o anicipae immediae and sharp depreciaion of he Korean won from Ocober Third, he Korean won was undervalued in he beginning of A he peak of he counry's 19 For he pre-crisis period, he covered ineres pariy held reasonably well wihou adjusing for he sovereign defaul probabiliies. 14

15 crisis in December 1997, he Korean won los 70 percen of is value agains he dollar in a monh. 21 Figure 4 indicaes ha his sharp nominal depreciaion was beyond he expecaion of NDF marke paricipans. However, he degree of under-valuaion is no large compared wih he degree of overvaluaion prior o he crisis. In January 1998, he Korean won was esimaed o be abou 1-2 percen undervalued depending on he mehods. Fourh, from he middle of 1998 he Korean won became overvalued again. This was a period in which he Korean governmen sared o lower domesic ineres raes aggressively and a new round of financial crises in emerging markes sared in Russia and Brazil. However, he degree of won overvaluaion during his period differs significanly among he four mehods. I is esimaed o be 6 percen in June 1998 according o Mehod I. Bu he esimaes from Mehods II, III and IV are less han 2 percen. The discrepancy indicaes ha he observed and he implied ineres raes on wondenominaed bonds differ significanly, as we will see shorly. From July 1998 he degree of overvaluaion sared o decline and, a he end of 1998, he implied spo exchange raes in Mehods II, III, and IV became close o he observed raes again as hey were before July On he oher hand, Mehod I indicaes ha he spo exchange raes were sill slighly overvalued a he end of III.2 Covered Ineres Rae Differenial The fac ha S ˆ / S is no equal o 1 implies ha here are unrealized opporuniies for arbirage. Even hough he informaional conens are he same, i would be helpful o express our measure of currency misalignmen in erms of he covered ineres rae differenial. Equaion (11) defines he covered ineres rae differenial (CID ) when he forward exchange raes wih n monh mauriies are used. S = n, where n M = n/12. (11) * CID [(1 + in, nm ) {(1 + in, nm )(1 pnm ) + χpnm } ]/ Fn, M The CID measures he annualized exra rae of reurn ha invesors could make by swiching 20 Park and Rhee (1998) summarize he chronology and he magniude of he BOK inervenion in his period. 21 The spo exchange raes depreciaed from 1,112 o 1,965 won per U.S. dollar beween November 28 and 15

16 invesmen from won-denominaed bonds o dollar-denominaed bonds. Figure 5 shows he movemen of he CID and Table 1 presens is mean values for various sample periods. The CID labeled as Mehod I is derived using he observed 3 monh CD rae as he domesic ineres rae in equaion (11). The CIDs labeled as Mehods II, III and IV are compued using he implied domesic ineres raes wih 3 monh mauriies from Mehod II, III and IV, respecively. 22 < Figure 5: Covered Ineres Rae Differenial> < Table 1: Covered Ineres Rae Differenial> Needless o say, he movemen of he covered ineres rae differenial in Figure 5 should be idenical o ha of he S ˆ / S raio in Figure 4. Bu he degree of currency misalignmen is expressed in erms of he annualized ineres raes in Figure 5 and Table 1. Before July 1997, he CID was close o zero and he covered ineres pariy seemed o hold reasonably well. However, i increased sharply saring from July and came o be as large as 18 o 40 percen in November I is no surprising ha here were large capial ouflows from Korea a ha ime. In he beginning of 1998, due o he drasic depreciaion of he Korean won and he sharp increase in domesic ineres raes, he covered ineres differenial urned negaive, which means ha invesing in won-denominaed bonds became more aracive. However, he magniude of he covered ineres differenial varies widely across differen mehods. Beween January and March 1998, Mehod I indicaes ha invesing in won-denominaed bonds was abou 6.4 percen more profiable han invesing in dollar-denominaed asses. However, Mehods II, III and IV indicae ha he profi margins are smaller a only 3.1, 5.9, and 2.4 percen, respecively. In he middle of 1998, he Korean won came o be overvalued again, and he covered ineres rae differenial became posiive. According o Mehod 1, he ineres rae differenial was 16 percen during his period. Tha esimae was as high as he esimaes of CID in November 1997, he monh jus before he crisis eruped. However, he esimaes for CID by Mehods II, III, and IV are only 4.9, 2.1 and 2.6 percen, respecively. A he end of 1998, he covered ineres rae differenial declined and became close o zero again according o Mehods II, III and IV. In summary, he movemen of he S ˆ / S raio and he covered ineres differenial imply he December 23, In boh cases, he observed spo exchange raes are used in calculaing he CID in equaion (11), and no he 16

17 following facs. [1] Before he Asian crisis sared in July 1997 and afer he Asian financial markes showed signs of sabiliy a he end of 1998, he covered ineres rae pariy held reasonably well. [2] A he end of 1997 when he Korean crisis was a is peak, he Korean won was significanly overvalued; he covered ineres rae differenial indicaes ha invesing in dollar-denominaed bonds was more profiable. [3] In he beginning of 1998, he Korean won was undervalued and invesing in won-denominaed bonds was more profiable. [4] In he middle of 1998, he Korean won came o be overvalued again and invesing in dollar-denominaed bonds was more profiable. Alhough he four mehods show qualiaively similar paerns as summarized above, here sill exis significan differences in heir esimaes, and he discrepancy seems relaively large especially for facs [3] and [4]. Since he differences across he mehods are largely due o heir choices of he domesic ineres raes, we will compare he movemen of he observed domesic ineres raes wih he implied ineres raes derived from he covered ineres pariies. III.3 Observed and Implied Domesic Ineres Rae Figure 6 compares he behavior of he observed ineres raes on won-denominaed bonds wih he implied values calculaed by Mehods III and IV. Mehod III is differen from Mehod II in ha i allows a non-zero erm premium for domesic ineres raes. However, since he implied ineres raes of Mehods II and III urn ou o be quie similar, he resul from Mehod II is no repored separaely in Figure 6. < Figure 6: Observed and Implied Domesic Ineres Rae > As can be seen in Figure 6, he pre-crisis average 3-monh CD rae was 14 percen bu i soared sharply afer November A he peak of he crisis in January 1998, i recorded 23 percen. Following he gradual shif in moneary policy from auseriy o easy sance in he middle of 1998, i declined seadily and became lower han he pre-crisis level from Augus However, he implied ineres raes from Mehod III and IV show a very differen paern. They increased sharply from Sepember o November 1997 long before he surge of he observed rae. Besides, hey decreased, no increased, in he beginning of 1998 and soared again in he middle of 1998, even hough he observed ineres raes sared o decline sharply during his period. Only a he end of implied spo exchange raes. If implied spo exchange raes are used, CID should be zero by definiion. 17

18 1998 did he implied raes converge o he observed rae. Concepually he implied domesic ineres raes represen he raes of reurn ha foreign invesors demand from won-denominaed bonds in Korea. As explained in secion II, hey need no be equal o he observed domesic CD raes ha were heavily regulaed and whose paymen was guaraneed by he governmen during he crisis. To he exen ha domesic invesors could demand higher curb marke premiums over he regulaed ineres raes a ha ime, we believe foreign invesors could have done he same in he won-denominaed bond marke. Thus i is no surprising o see ha he implied domesic ineres raes in Figure 6 show he same paern as ha of he sovereign defaul probabiliy in Figure 3. As he sovereign risk sharply increased a he end of 1997, foreign invesors could have asked for a higher risk premium for invesing in won-denominaed bonds. The severe credi crunch problem and he excess demand for foreign funds mus have made i easier for foreign invesors o selecively choose blue chip companies and ask simulaneously for higher premiums. The behavior of he implied ineres rae in 1998 can be explained in a similar way. Despie he sharp decrease in he regulaed ineres raes due o expansionary moneary policies from he middle of 1998, he implied ineres raes soared again as he inernaional financial urmoil wen ino anoher urmoil. As he sovereign risk on Korea rose ogeher wih oher emerging marke counries, foreign invesors could have demanded a higher premium on heir won denominaed invesmen. 23 The validiy of he above inerpreaion can be checked if we can observe he curb marke ineres raes charged on large conglomeraes or financial insiuions. However, given he lack of appropriae daa, we ry o esimae he upper and lower bounds of he curb marke premium by considering he wo exreme cases. In Korea, 3 monh CD ineres rae was a good proxy for he risk free ineres raes prior o he crisis and he average of he 3 monh CD ineres rae was abou 14 percen prior o he crisis. Our firs exreme assumpion is ha he risk free ineres raes did no change afer he crisis eruped and ha risk neural invesors asked for he same 14 percen expeced rae of reurn from heir curb marke invesmen. Then he curb marke risk premium, φ 1,, had o saisfy equaion (12), where he defaul probabiliy of he hypoheical won-denominaed bonds is assumed o be equal o he sovereign defaul probabiliy. ( ) = ( φ 1, )(1- p ) + χ p (12) 23 We have o admi ha his hypohesis canno explain why he implied ineres raes are lower han he acual CD raes in he beginning of

19 In calculaing φ 1, from equaion (12), χ is assumed o be 0.25 or 0.5 since we do no know wheher he proporion of salvage value should be higher or lower in case of privae corporaion's defaul compared wih sovereign defaul. The risk premium adjused domesic ineres raes, φ 1,, are ploed in Figure 7. The series I and II correspond o he raes when χ is equal o 0.25 and 0.5, respecively. So far, we implicily assume ha he pre-crisis risk free ineres raes did no change even afer he erupion of he crisis. However, considering igh moneary policy and severe credi crunch, ha assumpion seems unrealisic. As he oher exreme case, assume ha he observed CD ineres rae was a good proxy for he risk free ineres raes even afer he crisis eruped and he curb marke invesors asked for he same expeced rae of reurn as he observed CD rae. Then he curb marke risk premium, φ 2,, saisfies he following condiion. (1 + i ) = (1 + i + φ 2, )(1- p ) + χ p (13) In equaion (13), i denoes he observed CD ineres rae and i + φ 2, is our second measure of he risk premium adjused domesic ineres rae. I is ploed in Figure 7 and he series III and IV are he raes when χ is equal o 0.25 and 0.5, respecively. We believe he rue risk premium adjused ineres raes should be beween he wo exreme cases shown in Figure 7. We find he resuls in Figure 7 quie comforing. The risk premium adjused ineres raes show a paern similar o ha of he implied ineres raes in Figure 6. Even hough hey canno fully explain he firs hump of he implied ineres raes a he end of 1997, hey seem o explain quie well he second hump in he middle of In paricular, when χ is se o 0.5, he risk premium adjused ineres raes explain a significan porion of he difference beween he observed and he implied domesic ineres raes. 24 This is a supporing evidence for our inerpreaion ha he observed CD raes canno properly reflec he rae of reurn ha foreign invesors expeced o earn by invesing in won denominaed bonds, while he implied ineres can. < Figure 7: Observed and Risk Adjused CD Raes > 24 The magniude of he implied ineres raes in Figure 6 does no change much if we vary χ from 0.25 o 19

20 IV. Currency Misalignmen and Capial Inflows The degree of currency misalignmen and he covered ineres rae differenial indicae which direcion foreign capial should flow. This secion examines he paern of capial flows during he Korean crisis o see wheher hey moved as prediced. Figures 8 and 9 show he monhly capial flows from January 1997 o January Since invesmen abroad by domesic residens in Korea was more or less resriced during ha period, we only look a invesmen by foreigners. 25 We also exclude borrowings from inernaional financial insiuions and foreign governmens in order o focus on privae capial flows. Figure 8 shows foreign capial flows by invesmen ype: foreign direc invesmen, porfolio invesmen, and oher invesmen. Figure 9 breaks down porfolio invesmen ino equiy and bond invesmen. Oher invesmen consiss of largely shor-erm and long-erm exernal borrowings of domesic financial insiuions and accouns for he lion's share of capial ouflows a he peak of he Korean crisis. The figure shows ha currency speculaion played only a limied role; he failure o roll over exernal deb of domesic financial insiuions was he main cause of capial ouflows in he crisis. < Figure 8: Foreign Invesmen by Type > < Figure 9: Foreign Porfolio Invesmen > Figure 8 shows ha he size of foreign direc invesmen was negligible and did no change significanly during he crisis period. In conras, porfolio invesmen was volaile, wih ne inflows in he middle of 1997 and in he firs half of 1998 and ne ouflows in he second half of 1997 and in he middle of If we consider porfolio capial flows alone, he degree of currency misalignmen and he covered ineres rae differenial repored in secion III are consisen wih he direcion of capial flows. However, he oher invesmen caegories ha had larger shares in oal capial flows did no fully conform o he changes in he covered ineres rae differenial. There were ouflows of oher invesmen a an acceleraing pace afer July 1997, reaching a peak in December Unlike porfolio invesmen, here was no reversal during he firs half of Including ne foreign invesmen by domesic residens does no change he resuls in his secion. 20

21 Our measures of currency misalignmen and he covered ineres differenial show a mixed performance in explaining he paern of foreign capial flows. As summarized in Table 1, Mehod I esimaes ha he incenives for capial ouflows in he middle of 1998 were as srong as ha in he second half of The excess reurn from swiching won-denominaed invesmen o dollardenominaed invesmen is esimaed as 16.3 percen in he middle of 1998 and 18 percen a he end of 1997, respecively. However, in conras o his predicion, capial ouflows in he form of equiy invesmen and oher invesmen were much larger in he second half of 1997 han in he middle of Only capial ouflows in he bond invesmen caegory show similar magniude in hese wo periods. Also, here was no sign of large capial inflows in he beginning of 1998 despie he fac ha invesing in won-denominaed asses was esimaed o be abou 6.4 percenage poin more profiable han invesing in dollar-denominaed deposis. The finding ha capial flows during he Korean crisis were no sensiive o ineres rae differenials, and in paricular ha here was no reversal in capial flows in he beginning of 1998 despie high domesic ineres raes, is frequenly used as a case agains he IMF adjusmen program in Korea. The criics argue ha, insead of bringing back foreign capial and sabilizing exchange raes, he igh macroeconomic policies of he IMF and he consequen high ineres raes had a negaive effec on a highly leveraged counry such as Korea by deepening credi crunch problems. Such criicism, however, is no jusified by he resuls of Mehods II and III. These wo mehods demonsrae ha here were srong incenives for capial ouflows in he second half of However, he esimaed covered ineres differenials in he beginning of 1998 and in he middle of 1998 are much lower if we use Mehods II, III or IV. For example, Mehod IV esimaes he excess reurn from invesing in won-denominaed bonds o be 2.4 percen in he beginning of 1998, whereas Mehod I esimaes he excess reurn o be 6.4 percen. The excess reurn from invesing in dollar-denominaed deposis in he middle of 1998 is only 2.6 percen, no 16.3 percen, if we use Mehod IV insead of Mehod I. In oher words, Mehods II, III and IV sugges ha he incenives for capial inflows or ouflows in 1998 were no as srong as he incenives indicaed by Mehod I. Thus, i is no surprising o find ha capial inflows in early 1998 and capial ouflows in he mid-1998 were much smaller han he capial ouflows in he second half of This implies ha he problem did no lie in he low ineres rae elasiciy of foreign capial flows per se. The real problem was ha, judging from he foreign invesor's pessimisic perspecive on Korea, he Korean won was depreciaed oo lile, no 21

22 oo much, and he Korean domesic ineres raes increased oo lile o spark resumpion of foreign invesmen. As emphasized by Sigliz (1998), once a crisis sars, only unrealisically high and poliically unaccepable high ineres raes would be effecive in comforing foreign invesors and in defending exchange raes. 26 V. Conclusion This paper proposes a mehod of measuring he degree of currency misalignmen using offshore forward exchange raes. By reaing he observed offshore forward exchange raes as exogenous, we calculae he spo exchange raes and he domesic ineres raes implied from he covered ineres pariies and compare hem wih he observed ones. Our mehodology is based on he presumpion ha, during a currency crisis, offshore forward exchange raes may reflec marke fundamenals more closely han onshore ones ha are usually ighly regulaed and heavily affeced by governmen inervenion. 27 The implied spo exchange raes and domesic ineres raes are inerpreed as he ones ha would have prevailed if here were no governmen inervenion. Differenly from wha ohers have concluded (Chinn (1998), Goldsein (1998), Goldfajn and Baig (1998)), our mehod indicaes ha he Korean won was significanly overvalued on he eve of he counry's crisis in I also finds ha he Korean won became undervalued in he beginning of 1998 due o sharp nominal depreciaion, bu ha he degree of under-valuaion was no large enough o spark he resumpion of foreign capial inflows. I is rue ha our measure of currency misalignmen canno address he imporan issue of opimaliy. I simply measures he deviaion of he observed exchange rae from he level ha he paricipans in he NDF marke expec o prevail. If he NDF marke is conaminaed by an irraional bubble, herd behavior, ec, finding ha he exchange rae is no misaligned according o our mehod does no imply ha i is a an opimal level. However, for a small open economy wih 26 However, his argumen does no imply ha he IMF's high ineres rae policy in Korea was a misake. Oher raionales for high ineres rae policy exis apar from reversing foreign capial ouflows. For example, Park and Rhee (1998) argue ha a he peak of he crisis he high ineres policy was necessary o sop he domesic financial insiuions from buying dollars a he domesic exchange marke wih he won hey borrowed from he cenral bank. 27 In his respec, our mehodology can be applied only o a counry in which an onshore forward marke is under-developed or does no exis and consequenly has is currency raded heavily in offshore markes. Among he crisis-hi Asian counries, only Korea seems o saisfy his condiion. The currencies of Indonesia, Malaysia, and Thailand are no raded in he Asian NDF marke, whereas Chinese renminbi, Indian rupee, New Taiwan dollar, Philippine peso and Korean won are. I would be ineresing o apply our mehod o some 22