Accounting for the East Asian Crisis A Quantitative Model of Capital Outflows in Small Open Economies

Transcription

1 Accouning for he Eas Asian Crisis A Quaniaive Model of Capial Ouflows in Small Open Economies David Cook Deparmen of Economics Hong Kong Universiy of Science and echnology Michael B. Devereux 2 Deparmen of Economics Universiy of Briish Columbia Absrac o wha degree can he qualiaive and quaniaive aspecs of he Eas Asian crisis be accouned for wihin a dynamic general equilibrium model? his paper invesigaes ha quesion, in a framework where he crisis iself is modeled as an exogenous shock o he counry risk premium. he exercise has an empirical discipline because he scale of he shock can be measured by he behavior of he repored risk premium. We calibrae a quaniaive sicky-price dynamic general equilibrium model of a small open economy o mach he feaures of hree Eas Asian economies: hailand, Korea, and Malaysia. We idenify a shock o he counry risk premium using published daa from inernaional bond markes and idenify shor-run moneary policy using observed domesic ineres raes. We find ha he modeled response o he observed increase in exernal ineres raes subsanially maches macroeconomic daa on prices and quaniies a he aggregae and secoral level. he model has more difficuly explaining he large exchange rae devaluaions ha occurred in hose economies. Keywords: Eas Asian crisis, sicky prices, small open economy. JEL Code: F4 Macroeconomic Aspecs of Inernaional rade and Finance davcook@ushk.us.hk Phone: (82) Fax: (82) ; We would like o hank wihou implicaing Wai Ming Ho, Graeme Wells, and paricipans in he Canadian Macro Sudy Group and HKUS Brown Bag for helpful commens. Cook hanks he Universiy Grans Council of Hong Kong for financial suppor. 2 devm@inerchange.ubc.ca; Devereux hanks he SSHRC of Canada for financial suppor.

2 I. Inroducion In he middle of 997, a number of Eas Asian economies underwen a common financial crisis. Korea, Malaysia and hailand each experienced severe exchange rae devaluaions and curren accoun reversals. In each economy, he currency devaluaions were followed by large real conracions. Oher economies in he region, including Hong Kong, Indonesia, he Philippines, Singapore, and aiwan, were also affeced o varying degrees. his paper will examine some of he common feaures of hese conracions and examine he abiliy of sandard, bu modern, business cycle heory o quaniaively capure hese feaures. he purpose is o accoun for he Eas Asian crisis using a quaniaive small open economy model whose principles are drawn from he equilibrium sicky-price open economy macroeconomic (S-POEM) lieraure. We model he crisis as an exogenous rise in he counry risk premium on foreign lending ino a small open economy. Daa on bond yields show ha he premium on he US dollar bonds issued by eniies in hese counries rose dramaically during he crisis. In our model, we assume a rise in he counry risk premium o exacly mach he rise in he counry risk premium observed. We compare he dynamic response of he model o he macroeconomic impac of he financial crisis in hree Eas Asian economies (Korea, Malaysia and hailand) ha were among hose mos severely affeced by he crisis. In a sicky-price economy, however, he response of he economy o a given shock will depend on he sance of moneary policy. We noe ha iner-bank lending raes rose dramaically a he onse of he crisis. We model his seep rise in nominal ineres raes as deviaions from a moneary policy rule ha arges domesic inflaion and oupu. he deviaions are calibraed o exacly mach he rise in ineres raes in he daa. he disinguishing characerisic of he S-POEM lieraure (see Lane, 2, for a review) is ha an equilibrium includes he opimal decision rules of forward-looking agens wih raional expecaions, bu nominal prices converge only over ime o marke-clearing levels due o menu coss. A rise in he real ineres rae causes households o delay consumpion and firms o delay invesmen spending. In he absence of a sufficien moneary easing, he exernal shock o he economy generaes a conracion in nominal demand for domesic goods. he conracion in nominal demand ranslaes ino a conracion in real demand in he presence of sicky prices. In Eas Asia, he U.S. dollar is he currency of denominaion of much inernaional rade. In our model, boh impors and expor prices are se in a foreign currency, while goods produced for he domesic economy are priced in domesic currency. A rise in exernal ineres raes above domesic ineres raes leads o an exchange rae depreciaion. A nominal exchange rae depreciaion increases he price of impors relaive o domesic goods, leading o a swiching of expendiure oward domesic goods. Conversely, he exchange rae depreciaion does no reduce he real price of expors. his pricing model combines he producer currency pricing framework of Obsfeld and Rogoff (99), which emphasizes expendiure swiching, and he local currency pricing framework of Bes and Devereux (996, 2), in which exchange rae changes have no immediae expendiure-swiching 2

3 effecs. his model of pricing is effecive in maching he immediae response of expor and impor prices o he currency devaluaions observed in Eas Asia. Anoher aspec of he daa in which we are ineresed is he secoral response of oupu. A exbook response of a small open economy o an exernal ineres rae shock is a swich from he producion of non-raded goods for domesic use o he producion of raded goods needed o repay foreign debs (see for example Sachs and Larrain, 993). However, his does no mach he secoral response observed in he daa for Eas Asia. During he crisis we observe a sharp drop in producion of boh raded and non-raded goods. o deal wih his, we depar from he exbook model by incorporaing inermediae impored maerials ino producion. In our model, impored inermediae maerials are an imporan facor in he producion of raded goods. hese inpus mus be purchased in advance, subjecing hem o finance coss. he rise in he price of inernaional borrowing during he crisis resuls in higher coss of obaining he working capial needed o purchase impored maerials. his resuls in an equilibrium fall in oupu of raded goods. We find ha when we mach he shocks o inernal and exernal ineres raes in our model o he daa, he modeled responses of nominal and real macroeconomic aggregaes mach heir counerpars in he daa boh qualiaively and quaniaively. One conroversial aspec of he crisis was wheher he moneary response of he respecive cenral banks conribued o he severiy of he crisis. We examine a counerfacual case in which he moneary policy response o he crisis is described by a aylor (99) rule. In his case, he conracion in oupu and invesmen are subsanially less persisen. Our paper is relaed o a large lieraure on he causes of emerging marke crises of he mid- o lae-99s. Explanaions migh be divided ino wo broad caegories: inernal and exernal. Inernal crisis heories explain he crisis as he resul of eiher policy failings or limiaions of financial markes wihin he affeced economies. Exernal explanaions describe he crisis as he resul of imperfecions wihin inernaional financial markes hemselves. Our paper is essenially agnosic wih respec o his disincion. We measure he exogenous shock ha drives he crisis from observed counry risk spreads on U.S. dollar bonds. hese spreads could be driven by herding behavior among inernaional invesors (see Calvo and Mendoza, 2) or by an eliminaion of loan guaranees by domesic fiscal auhoriies (see Corsei, Peseni, and Roubini, 999). In his sense he shock may be consisen wih eiher an ulimaely inernal or exernal source. hus, our paper does no offer a fundamenal explanaion of he source of he crisis. Raher, we impose a mainained hypohesis ha, whaever he underlying source, he scale of he crisis iself can be represened by he movemen in observed risk spreads, and ha hese risk spreads are independen of he subsequen pah of he economy and he sance of moneary policy. Under his hypohesis, we ask how well he response o he crisis can be racked by a sicky-price muli-secor dynamic open economy model. A number of heories have explained crises as he oucomes of policy errors or marke failures inernal o he affeced counries. Chang and Velasco (998, 2a & b, 2) implicae bank runs 3

4 caused by a mauriy mismach in he domesic banking secor as he source of he crisis. Dooley (2) argues ha implici loan guaranees of privae-secor deb in Eas Asia would lead o a currency crisis if hose guaranees were called and he resuling debs were o be moneized. Krugman (998a) is an influenial early paper calling aenion o crony capialism ha characerized some Eas Asian economies. Burnside, Eichenbaum and Rebelo (2) consruc a model ha allows for a quaniaive sudy of he size of debs ha migh be incurred and he resuling effec on exchange raes. Corsei, Peseni and Roubini (999) argue ha governmen loan guaranees would lead o over-invesmen; when a crisis caused guaranees o be removed, he resuling dis-invesmen would resul in a real conracion. Our model absracs from he fiscal aspecs of he currency crisis, alhough in principle our approach is consisen wih he view ha he discovery of loan guaranees caused a deerioraion in he exernal borrowing environmen, so long as he subsequen pah of he economy can be represened as in he dynamic model. Many papers have sressed he aggregae imporance of balance-shee aspecs of he Eas Asian crisis. Many privae Eas Asian debors faced a mismach in he currency denominaion of heir asses and liabiliies. Krugman (998b) and Aghion, Bachea, and Bannerjee (2, 2) show how balance-shee consrains may lead o a self-fulfilling exchange rae crisis. Arrellano and Mendoza (22), and Gusavo, Roldos, and Chrisiano show ha balance-shee consrains may be imporan in generaing sudden sops, defined as a swich owards a binding naional consrain on capial inflows for a borrowing counry (see also Calvo and Reinhar, 999). Since any binding consrain on capial flows can by definiion be reinerpreed as a rise in he exernal borrowing rae, our paper is also consisen wih his lieraure. A few papers have sudied he quaniaive impac of ineres rae shocks, an exercise very similar o ours. Mendoza (2) sudies he business cycle effecs of sudden sops in a quaniaive business cycle model. McKibbin (998) sudies he effec of an ineres rae shock on a muli-counry model a annual frequencies. Neumayer and Perri (2) examine he business cycle behavior of emerging markes subjec o counry risk premium shocks calibraed o he ime-series behavior of Argenine spreads. Bursein, Eichenbaum and Rebelo (22) quaniaively examine he effec of ineres rae shocks on exchange raes and nominal prices. None of hese papers focus on he quaniaive accouning quesion ha is he cener-piece of our analysis. Kim (22) consrucs a dynamic muliple equilibrium model o examine wheher animal spiris shocks can explain quaniaive feaures of he Asian crisis. Gerler, Gilchris and Naalluci (2) explore wheher balance shee effecs help in explaining he conracion ha occurred following he Eas Asian currency crisis. he res of he paper is organized as follows. Secion II documens he key macroeconomic facs of he Eas Asian crisis in Korea, Malaysia, and hailand. Secion III ses ou a wo-secor sicky-price model of a small open economy. Secion IV explains how we calibrae he model. Secion V presens he resuls, and secion VI offers some conclusions. 4

5 II. he Eas Asian Crisis his secion documens he macroeconomic response of Korea, Malaysia and hailand o he Eas Asian crisis 3. Figure, Column A graphs he monh-by-monh movemens in he Korean won, he Malaysian ringgi, and he hai bah relaive o he U.S. dollar from 997 o 22. hrough early 997, hese currencies are sable agains he U.S. dollar. In he hird quarer of 997, each currency experiences a large depreciaion. Afer a large bu brief overshooing, he currencies sele a levels beween 3 and 4% weaker han pre-crisis levels. he iming of he depreciaions varied across counries. he bah begins depreciaing a he ouse of he hird quarer, while he ringgi and won do no depreciae unil lae in he hird quarer. Figure, Column B shows he response of he counry risk premium for each of hese economies. Since early 997, HSBC has calculaed yields on an index of U.S. dollar bonds issued in each of a number of Asian emerging markes. We measure he counry risk premium for Korea, Malaysia, and hailand as he spread beween he HSBC yield indices and conemporaneous yields on hree-monh U.S. reasury bills. In early 997, he observed spreads in each counry are beween and 2 annualized basis poins. During lae 997, spreads rise dramaically in Korea and hailand, approaching levels beween 6 and 8 annualized basis poins and reaching a peak in lae 998. Malaysia experiences a more moderae rise of 4 basis poins in lae 997 before rising o more han 2 basis poins in lae 998. Laer, spreads narrow bu remain beween 3 and basis poins even years laer. o idenify he macroeconomic effecs of he Eas Asian crisis, we uilize is large size and clear iming, which is refleced in he movemens in he counry risk premium. We assume ha he shock in he risk premium ha occurred in he hird quarer of 997 and beyond was responsible for he difference beween he acual realizaions in he economy and is previous pah. A downside of his idenificaion scheme is ha i disregards subsequen shocks. However, he unprecedened size of he crisis makes i possible o race is impac even in he presence of background noise. We repor resuls from seasonally adjused quarerly naional income accouns. he real variables we examine are GDP, Personal Consumpion Expendiure, Gross Fixed Capial Formaion, Expors, Impors, raded Goods Value Added (he raded goods secor being he sum of manufacuring, mining, and agriculure), and Non-raded Goods Value Added (GDP minus raded goods). he nominal variables are he deflaors of domesic absorpion, expors, impors, raded goods, and non-raded goods. We mus firs esimae he pre-crisis pahs of each economy s 3 Daa on he HSBC indices are from Bloomberg. he remaining Eas Asian naional income accouning, exchange rae and ineres rae daa used in his paper are from CEIC/DRI Asia daabase, which gahers published daa from official sources.

6 macroeconomic aggregaes. We derend all variables excep he exchange and ineres raes wih a log-linear quadraic rend. In many cases, real variables are subsanially above rend in mid-997. We esimae AR() processes for each of he derended variables. Figures 2-4 show he difference beween each series and heir oucomes if hey had revered geomerically back o rend afer he hird quarer of 997. Each counry experiences a sharp real conracion in boh producion and absorpion of goods a he end of 997; he decline in GDP reaches a rough of more han % below he pre-shock pah. Korean and Malaysian GDP reurn o rend by early 2. hai GDP reurns o rend laer in 2. Personal consumpion expendiure and gross fixed capial formaion conrac in each counry. In hailand, he conracion in consumpion is of approximaely he same size as GDP. In Korea and Malaysia, he conracion in consumpion is slighly more severe han he conracion in GDP, reaching a rough more han % below he pre-crisis pah. he sharpes conracions occur in fixed invesmen. he rough of invesmen is 2% below he pre-crisis pah in Korea. In hailand and Malaysia, invesmen falls o more han % below he previous pah. he conracion in absorpion is much sharper han he conracion in oupu. In each counry, here is a sharp decline in impors o around 3% below he previous pah. By comparison, here is no obvious immediae impac of he crisis on expors. During 998, expors fall in hailand and Malaysia. wo years afer he crisis, however, here is a large surge in expors in each counry, rising o a level ha is 8-2% above he previous pah of expors. Conracions occur in boh raded and non-raded goods secors and are, iniially, of approximaely he same size in he wo secors. he recovery is much sharper in he raded secor and he conracion much more persisen in he non-raded secor. In each economy, he domesic currency prices of boh expors and impors rise conemporaneously wih he exchange rae depreciaion. he price increase in hese caegories is of comparable size o he exchange rae depreciaion. In each counry, here is a much smaller rise in he absorpion deflaor (perhaps reflecing he rise in he price of impors). In each counry, here is a rise in he price of raded goods (perhaps reflecing he rise in domesic currency expor prices). All of hese price spikes converge back o rend. In hailand, here is a mild rise in non-raded prices in he immediae afermah of he shock. Conemporaneous changes in he non-raded price level are minimal in Korea and Malaysia. Panels (M) (O) show he response of he nominal and real exchange rae wih he U.S. dollars and nominal ineres raes. Real exchange raes are calculaed using absorpion deflaors as relaive prices. o assess he response of he nominal and real exchange rae o he crisis, we examine deviaions from he mean exchange rae in he firs half of 997. Each economy experiences a sharp exchange rae depreciaion in mid-997. he size of hese depreciaions are large, wih he won and bah depreciaing by more han 6% and he ringgi by more han 4%. here is some overshooing in each counry, wih he currencies finding a persisen level somewhere near 4% of he original value. Malaysia adops a peg in lae 998, while Korea and hailand floa heir currencies. In each economy, 6

7 crisis-era inflaion was higher in he affeced economies han in he U.S., so he real exchange rae depreciaes by less han he nominal exchange rae. However, he price rise is no as large as he exchange rae depreciaion, so he real exchange rae depreciaions are subsanial. In each economy, he real exchange rae iniially depreciaes by 3 4% in lae 997. he real exchange rae depreciaion appears quie persisen. In boh Malaysia and hailand, he real exchange rae remains a more han 2% above he pre-crisis pah. he rapid disinflaion ha occurs in Korea leads o a real exchange rae ha is currenly % weaker han he pre-crisis level. o assess he response of he nominal ineres rae, we examine deviaions of he shor-erm iner-bank lending rae from he mean ineres rae in he firs half of 997. In each economy, he iner-bank lending rae rises sharply during he iniial periods of he crisis. In Korea, he ineres rae a peak is 2 basis poins above he pre-crisis level. In Malaysia and hailand, he rise in he ineres rae is no as sharp, reaching a peak 4 basis poins above he previous level. Afer he fall of 998, however, ineres raes in all hree counries drops sharply. In Korea and hailand, he nominal ineres rae is 8 basis poins below he pre-crisis level by 999. In Malaysia, he nominal ineres rae is 4 basis poins below he pre-crisis level by 999. III. he Model We model a small open economy populaed by a represenaive agen ha owns he facors of producion and borrows from he world economy a an exogenous ineres rae. A range of non-raded and raded goods are produced and sold by a range of monopolisically compeiive firms. A. Demand. he Represenaive Agen he infiniely-lived represenaive agen and has raional expecaions. In any period, he agen derives uiliy from consumpion, C, and disuiliy from labor, H. he agen s subjecive discoun funcion is a declining funcion of he agen s consumpion level. V β ( C ) E [ V ] = max [log( C ) ΓH ] + + C, H he subjecive discoun rae is a concave funcion (β < and β > ) of curren consumpion wih enough curvaure o ensure a saionary level of wealh. he agen issues foreign currency deb, D, a an exogenous rae +r. Secor-specific capial K and K N is rened o firms in he raded goods and non-raded goods indusries in compeiive markes a raes R and R N respecively. he agen supplies labor in compeiive markes a wage rae W, and receives profis from monopolisically compeiive firms ha sell raded and non-raded goods domesically as well as exporers; oal profis are Π =Π + Π N + Π EX. Agens purchase final goods a price P and allocae goods o consumpion and o raded and non-raded goods invesmen, I and I N. Lump-sum axes finance governmen spending, G. Define S as he spo exchange rae. he budge consrain is: 7

8 N N N () SD = ( + r ) SD + P C + I + I + G ( WH + R K + R K +Π) Consumpion, invesmen, and governmen goods are defined as a nesed CES combinaion of domesically produced non-raded goods, X N, domesically produced raded goods, X d, and impored raded goods, X m. Domesic and foreign raded goods are combined ino a quaniy of raded goods absorbed, X. he price index is he cos-minimizing marginal cos of acquiring hese goods. N φ φ φ N (2) C + I + I + G = X = a { X } + ( a) { X } µ d µ m (3) X = b { X } + ( b) { X } µ µ he non-raded aggregae is a Dixi-Sigliz aggregae of a uni range of differeniaed goods, he domesic raded aggregae is a combinaion of differeniaed goods, domesic raded and non-raded goods are denoed by P d and P N. N N ξ ξ N N = ξ {, i} = {, i} o o X x di P p di d d ξ ξ d d = ξ {, i } = {, i } o o µ d x i ξ ξ φ φ N x i.. he price index of X x di P p di he price index of impored goods is S P *. he price of he cos-minimizing combinaion of domesic raded goods and impors is hen P, defined in a sraighforward manner from he aggregaor X. he represenaive agen accumulaes capial in each secor hrough invesmen, subjec o adjusmen coss (4) () e 2 K = ( δ ) K + I 2 ( δ ) + K I K K I K N N N N e 2 N = ( δ) 2 ( ) + δ + N K he home agen maximizes uiliy subjec o he budge consrain and he accumulaion equaions for secoral capial. he firs-order condiions ha characerize he opimal plans are: ' (6) + β ( C ) V + P Ω C (7) Γ= WΩ I K (8) Λ Ω P I e δ K 8

9 (9) Λ Ω P N I e δ K N N φ φ () { } { } N φ N = Pa X X P () p x = N N i, i, N N X P ξ φ φ (2) ( ) { } { } φ = P a X X P (3) p x = d d i, i, d d X P ξ µ (4) { } { } d µ µ d = Pb X X P µ () ( ) { } { } m µ µ * = P b X X S P Ω = C E Ω + r + β ( ) + S (6) ( ) (7) Ω = ( C ) E Ω ( + i ) (8) β ( ) β + S I + I + Λ Λ + E + ( δ e δ ) + Ω + R+ K + K + C Λ Λ + I I +Ω N N N β N N C E ( δ e ) N N R K K δ (9) ( ) where Ω is he shadow value of domesic currency and Λ and N Λ are he shadow values of secorspecific capial. Equaions (6) and (7) describe he opimal choice of consumpion and labor supply. Equaions (8) and (9) characerize he choice of invesmen in he wo secors. Equaions ()-() and (2)-() respecively describe he demand for aggregae non-raded goods and raded goods, and he individual demands for each good. Equaions (6) and (7) describe he Euler equaions for inernaional borrowing, and equaions (8) and (9) describe he opimal choices for capial accumulaion in each secor. 2. Expors 9

10 Some raded goods are sold overseas by exporing reailers. he reailers sell differeniaed raded goods a a foreign currency price, p $. Define aggregae expors, EX, as a Dixi-Sigliz combinaion of hese differeniaed goods and an associaed price index, P $. ξ ξ $ $ ξ { i} { i } o o ξ EX = ex, di P = p, di, < ξ < he foreign demand for hese expors is deermined in parallel o he domesic demand for impors. * * µ * µ $ = µ (2) ( ) { } { } P b EX X P (2) p ex = $ i, i, $ EX P ξ * where X represens oal foreign demand for raded goods. B. Producion. raded Goods raded goods are produced wih a combinaion of secor-specific capial, labor H, and impored maerials M. γ γ γ γ γ { } ( ) { } θ V { } { } A K H Y = A d V + d M 2 θ where V is domesic value added in he raded goods secor. raded goods manufacurers are price akers in oupu and inpu markes. he manufacurers purchase he impored maerials, one period in advance wih foreign currency borrowed a rae r. he firs-order condiions of he firms profimaximizaion problem are (22) γ V γ V ( θ) = H Y PPI d W γ V γ V θ = K Y PPI d R γ M ( ) [ ] ( ) γ * (23) d E PPI = + r P E [ S ] Y where PPI * is he producer price of raded goods and P is he foreign currency price of impored maerials. Condiions (22) characerize he opimal choice of labor and capial in raded goods producion, while (23) represens he opimal choice of impored maerials. Noe ha he oupu of raded goods will be limied wihin any period by he pre-se quaniy of impored maerials. In addiion, he foreign ineres rae will adversely affec he purchases of impored maerials. 2. Non-raded Goods Price-aking non-raded goods manufacurers use labor and capial o produce goods, Y N, sold a price PPI N.

11 θn { } { } Y = K H N N N he firs-order condiions of he firms profi-maximizaion problem, for he choice of labor and capial, are N N Y Y (24) PPI ( θn) = W N PPI θn = R N H K θn N N N C. Sicky Prices In his model, here are hree ypes of sicky prices: reail prices of non-raded goods; domesic reail prices of raded goods; and foreign currency prices of expored goods. he dynamics of each of hese prices follow a similar framework. For each, here is a range of monopolisically compeiive price seers who purchase an undiffereniaed inpu good (i.e. raded or non-raded) from manufacurers a a scale invarian marginal cos, generically referred o as MC, and face a consan elasiciy demand curve. In each period, a randomly disribued fracion ( κ ) of price seers ge a chance o adjus prices, as described by Calvo (983) and Yun (996). he price-seing mechanism can be described as follows. Define p and q as he price and quaniy of he individual generic price seer, and he generic aggregae oupu and price indices as Q and P. Q ξ q, idi ξ ξ = ξ P = pi, di p i, ξ qi, = Q P Since he model se ou above assumes ha he elasiciy of subsiuion of he final goods purchaser (wheher domesic households or foreigners) for boh raded expor goods, and non-raded goods are all equal and consan a level ξ, hen his descripion of price and quaniy aggregaors applies o each of hese aciviies equally. ( ) Generic profis are Π p MC q. Price seers wih an opporuniy o change heir price maximize he sum of expeced discouned profis over he expeced life of he price. We assume shocks occur afer he price seers have chosen heir prices. hus, he price index is quasi-fixed in any period, as described in Roemberg and Woodford (997). j j ξ κ κ ξ ξ ξ max E ( + i ) max l Π j = E ( + il) P j Q j p p MCj p p j= l= j= l= Opimal price * p is choice ha maximizes expeced profis as

12 p he price index evolves over ime as * j κ ξ E ( il ) P + j Q jmcj j= l= = j ξ κ ξ E ( + il ) P j Q j j= l= ( κ) P P p ξ * = κ ξ ξ + Each of he price seers in he hree areas follows his paern. A uni measure of non-raded (raded) reailers purchase undiffereniaed goods in he non-raded (raded) manufacuring secor and sell differeniaed non-raded (raded) goods o he home counry household. Likewise, a uni measure of reailers purchase undiffereniaed raded goods and sell hem as differeniaed expor goods o foreign residens. In he non-raded secor, reailing firms purchase goods a marginal cos, PPI N, and sell o he household a price p N. In he raded goods secor, reail firms buy a marginal cos, PPI, and sell a price p (domesic raded goods reailers) or Sp $ (expor reailers). See able for a one-o-one maching of he elemens of he paricular price seer and he basic sicky-price framework. D. Moneary Policy Define real GDP as he sum of he value added in he raded and non-raded secors. he former is measured as he sales o domesic consumers, plus he sales o foreigners, less he cos of maerial impors. Real GDP is hen implicily defined as PGDP = P X + S P EX ( + r ) S P M + P Y d d $ * N N he moneary policy rule has wo componens. he cenral bank deermines a long-erm nominal growh rae ha implies a long-erm nominal ineres rae, i, and CPI inflaion rae, π. he auhoriies also se shor-erm ineres raes o respond o flucuaions in inflaion and GDP from heir saionary seady saes. i i = P ( GDP GDP) ε ρ π π + ρy + P where ε is a saionary funcion ha allows he moneary auhoriy o arge ineres raes for a finie ime period in response o shocks. E. he Crisis and Equilibrium We sudy he response of he economy o a single shock ha occurs a ime =. he single shock generaes a dynamic pah for he counry risk premium hrough finie ime period,{ r }, and = moneary policy hrough ime period L, { } ε L =. 2

13 Define Ξ as he hisory of he economy up o ime. An equilibrium is a se of policy funcions of he represenaive agens, manufacurers and price seers: C(Ξ ), I (Ξ ), I N (Ξ ), X(Ξ ), X (Ξ ), X N (Ξ ), X d (Ξ ), EX(Ξ ), IM(Ξ ), Y (Ξ ), Y N (Ξ ), M(Ξ ), H(Ξ ), H (Ξ ), H N (Ξ ), D(Ξ ), K (Ξ ), K N (Ξ ), w (Ξ ), w N (Ξ ), p (Ξ ), p N (Ξ ), p $ (Ξ ); and price funcions: P(Ξ ), P (Ξ ), W(Ξ ), R (Ξ ), R N (Ξ ), PPI (Ξ ), PPI N (Ξ ), S(Ξ ), i(ξ ); which solve he firs-order condiions of he agens opimizaions problems and labor and goods markes clear: H + H = H N x i, di + ex i, di = Y x di = Y IV. Calibraion Lacking a closed-form soluion, we log-linearize and solve he approximae linear model using he soluion algorihm in King and Wason (99). We calibrae he dynamics of adjusmen a sandard values from he calibraed business cycle lieraure. Following Baxer and Crucini (993), he depreciaion rae of capial is se a δ=.2 and he capial adjusmen cos is se so he seady-sae elasiciy of he invesmen-capial raio wih respec o marginal obin s q, (eδ) - =. Following Gali and Gerler (2), we se κ=.7, so ha prices change on average once per year. he elasiciy of subsiuion beween domesic producion and impors (and he demand elasiciy of expors) is se a.6 ( µ ) =, which is an average of he esimaes from Asia, repored in Reinhar (99). Kollmann (2) uses he same elasiciy for a model of developed economies. he elasiciy of subsiuion beween raded and non-raded goods is.66 ( φ ) =, based on a GMM esimae using pooled daa from five Asian counries as described by Osry and Reinhar (992). he elasiciy of subsiuion beween maerials and value added,.7 ( γ ) =, is se a.7 following Roemberg and Woodford (996). We will examine hree separae numerical cases, one for each of Korea, Malaysia, and hailand. We calibrae he long-erm macroeconomic raios of each case o he naional income daa of he corresponding counry from he ime period in Korea and hailand and for Malaysia. Lane and Milesi-Ferei (2) consruc counry-level daa on ne inernaional invesmen posiions. Row [A], able 2 shows he average of ne exernal asses posiions as a share of GDP. We calibrae he subjecive discoun rae funcion, β(.), so seady-sae deb as a share of GDP maches Row [A]. Expors as a share of GDP are repored in Row [B], able 2. he share of impors ha is inermediae maerials is repored in Row [C]. We calibrae b and d in each case o mach Rows [B] and [C]. Non-raded goods as a share of GDP are repored in Row [D]. We calibrae a in each case o mach hese shares. Row [E] shows governmen spending as a share of GDP. We calibrae he seady-sae governmen consumpion-gdp raio o mach Row [E]. Sarel (997) uses cross-counry daa o esimae he capial inensiy of various one-digi ISIC code indusries. For , we calculae weighed averages of Sarel s capial inensiy N i, N 3

14 esimaes of he raded secors (Agriculure, Manufacuring, and Mining), where he weighs are he share of he one-digi indusry in raded value added. We hen esimae θ as he average raded goods capial inensiy across ime. We repea he process for he remaining one-digi secors ha are classified as non-raded o esimaeθ N. he capial inensiy parameers are repored in able 2, Rows [F] and [G]. We calibrae he echnology parameers, A, A 2, and A 3 plus he parameers of expor demand funcion so ha, () he seady-sae marginal cos of domesic value added in he raded secor is equal o he marginal cos of impored maerials; (2) he seady-sae marginal cos of domesic final goods is equal o he marginal cos of impored final goods; (3) he seady-sae marginal cos of raded goods is equal o he seady-sae marginal cos of non-raded goods; and (4) he real exchange rae beween he foreign CPI and he domesic raded good price is equal o. We calibrae he economy near he case where seady-sae profis are arbirarily near zero. he shock o he exernal ineres rae is calibraed o average quarerly spreads beween HSBC s counry-level U.S. dollar bond yield indices (for Korea, Malaysia, and hailand) and U.S. reasury bills. he seady-sae ineres rae, r, is se equal o he average yield in he firs half of 997. he modeled shock is a projeced pah for he counry risk premium beginning in period =. For periods = 9, his pah exacly follows he observed spread beween he HSBC index and he 3-monh U.S. reasury rae over he period 997:4 o 22:2 (see Figure, Column B). For he periods = 2 8, he counry risk premium follows he average over he iniial 9 periods of he shock. Subsequenly, he counry risk premium revers o zero. he immediae response o he crisis in each of he hree counries was a sharp increase in shor-erm nominal ineres raes (see Figure, Column C). Following he crisis he moneary policies of he counries diverged, wih Korea and hailand adoping some form of inflaion argeing and Malaysia adoping an exchange peg wih he U.S. dollar. However, we ake he posiion ha predicing long-erm moneary policy was difficul following he large exchange rae devaluaions. As a defaul, we assume ha moneary policy converges o a benchmark aylor (996) rule, ρ π =., and L= ρ Y =.. In he shor-run, we se he moneary policy shock, { } 2 ε =, so ha he rise in nominal ineres raes exacly maches he rise in ineres raes observed in he daa. he increase in he nominal ineres rae is calibraed as he difference beween he observed ineres rae and he average ineres rae observed in he firs half of 997. V. Impulse Responses One advanage of examining he Eas Asian crisis is ha he singular, unprediced naure of he even makes i possible o idenify is macroeconomic effecs. However, he same singular naure makes i more difficul o use sandard saisical analysis o formally es he model. Our main 4

15 analysis will consis of visually and numerically comparing he impulse responses of an exernal risk premium shock o he acual pahs of oupu and oher variables. A. Qualiaive Oucome Here we briefly summarize he qualiaive effecs of he crisis. he rise in he counry risk premium will raise he cos of repaying ne foreign deb. his will have boh subsiuion and income effecs. he subsiuion effec causes he represenaive agen o delay consumpion and reduce invesmen in capial goods. As all hree economies are ne debors, he ineres rae rise has a negaive income effec, reducing opimal consumpion a all price levels. A given relaive prices, he conracion in real demand would reduce he componens of domesic absorpion: impors, domesic raded goods, and non-raded goods. However, he response o he shock will also involve an adjusmen of relaive prices. Given ha he shock induces a persisen rise in he exernal ineres rae above he domesic nominal rae, he ineres pariy condiion will imply an immediae depreciaion in he nominal exchange rae followed by a persisen expeced appreciaion. Since here is full exchange rae passhrough ino impor prices, he relaive price of impors rises sharply. Home agens subsiue domesic goods for impors, which bear he brun of he decline in demand. Domesic goods, however, are relaively imperfec subsiues for foreign goods, so changes in he relaive price generae only small subsiuion effecs. Hence, in equilibrium here is a decline in demand for domesic goods. Because exporers pracice local currency pricing, he exchange rae depreciaion does no immediaely increase he quaniy of expors. Only over ime, as exporers lower heir prices o reflec he weakening exchange rae, do expors increase. he impac of he risk premium shock on producion depends on he sance of moneary policy. Wih a passive or expansionary moneary policy, he auhoriies could leave domesic nominal ineres raes unchanged, or even aemp o reduce ineres raes, in response o he risk premium shock. Given he presence of uncovered ineres rae pariy, his would involve a large immediae real and nominal exchange rae depreciaion. he depreciaion would lead o a cushioning of he demand effecs of he shock on domesic raded and non-raded goods. Bu he conracionary moneary policy, observed in he daa and mached by our model, involves an increase in domesic ineres raes. his miigaes he immediae real exchange rae depreciaion, and hence leads o a greaer decline in absorpion and oupu in domesic raded and non-raded goods. he decline in domesic absorpion induced by a rise in he exernal risk premium can poenially have differen effecs according o he secor. All oupu of he non-raded secor is absorbed by he domesic economy, while he oupu of he raded secor will over ime face an increasing demand from he expor channel. Moreover, because domesic raded goods are beer subsiues for impored raded goods han are non-raded goods, he nominal depreciaion will elici a bigger subsiuion response owards domesic raded goods, whose prices are sicky. hus, we anicipae a bigger fall in

16 oupu in he non-raded secor. On he oher hand, he raded goods secor employs impored inermediae inpus, and he rise in he exernal real ineres rae direcly increases he price of hese inpus. he paricular pricing srucure ha we assume generaes some resuls for oucomes of nominal prices. he local currency pricing ha we assume for expors implies ha, in he shor-run, expor prices denominaed in he domesic currency will rise, along wih he exchange rae devaluaion. Since domesically produced raded goods are a combinaion of expors and domesically raded goods whose prices are quasi-fixed, he raded goods price will also rise o reflec devaluaion. Over ime, as exporers and domesic reailers change heir prices, he expor and raded goods deflaor will fall. Since imporers pass he full effecs of changes in he exchange rae hrough o domesic prices, he exchange rae depreciaion leads o a one-for-one rise in he price of hese goods, a leas in he shor-run. oal home absorpion is a combinaion of non-raded goods (which have quasi-fixed prices), domesic raded goods, and foreign impors. As a resul, he absorpion deflaor rises wih he exchange rae and impor prices in he shor-run. Over ime, conracionary moneary policy will reduce he price of domesically produced goods. his disinflaion will iniially be mued by sicky nominal prices. B. Quaniaive Comparison: Model vs. Daa We examine hree differen cases, corresponding o Korea, Malaysia or hailand, ha differ in erms of he calibraion of he parameers and he dynamic shocks o domesic and exernal ineres raes. Figures, 6, and 7 illusrae he heoreical impulse responses for macroeconomic series for Korea, Malaysia, and hailand, respecively. hese series correspond wih he variables from he daa repored in Figures 2-4. he benchmark responses are represened by he solid line wih sars. he counerfacual response, had policy followed a aylor rule, is illusraed by he dashed line. he modeled response of real macroeconomic aggregaes is close o he daa along qualiaive and quaniaive dimensions. A V-shaped conracion in GDP is observed in boh he daa and he models. In each economy, GDP conracs sharply in he period of he shock and converges back o he mean in subsequen periods before ulimaely achieving a level slighly higher han he iniial seady sae. In he daa, GDP for all hree economies reached a rough approximaely % below he precrisis pah. In he models, GDP reaches a rough of approximaely % below he seady sae. Because he shocks are persisen, he GDP conracion also displays subsanial persisence. However, he models do no fully mach he persisence of he oupu conracions observed in he daa. Acual oupu is below he previous pah for a leas eigh quarers. By conras, oupu has subsanially revered back o seady sae afer four quarers in all modeled economies. his is o be expeced, given ha we are using a relaively parsimonious dynamic general equilibrium specificaion wihou oher mechanisms for generaing endogenous propagaion. 6

17 Panels (B) and (C) show he responses of invesmen and consumpion o he shock in he calibraed models. A rise in he exernal ineres rae leads o large conracions in consumpion and invesmen. In he daa, in each counry, gross fixed capial formaion conraced far more sharply han GDP. he models mach his qualiaive feaure. Moreover, we observe ha invesmen fell beween 2 and % in he hree economies. he model s invesmen response varies beween 3 and %. However, he paern of invesmen response across counries is slighly counerfacual: he model predics a higher invesmen response (% decline) in Korea and hailand, wih a smaller response in Malaysia, while in realiy Malaysia and hailand experienced he larges response (abou %), wih Korea having a lower response (2%). Empirically, personal consumpion expendiure fell by approximaely he size of he conracion in oupu. In he daa, his implies conracions in consumpion of more han 4% below seady sae. In he model economies, consumpion also falls sharply, varying beween a fall of % in he model based on Korea, and approximaely % in he hailand and Malaysia cases. he modeled consumpion conracions are very persisen, due o permanen income effecs. Panels (D) and (E) display he responses of expors and impors. In each of he model economies, local currency pricing prevens he exchange rae depreciaion from generaing any immediae increase in expors. However, over ime as expor firms adjus heir prices, modeled expors increase by beween 3% (Korea) and 8% (hailand). In he daa, afer some delay, here is a mild increase in expors in each counry prior o he weakening of he U.S. business cycle afer 2. Each modeled economy faces a sharp decline in impors. Following a rough conracion in he firs period of he crisis ranging beween approximaely % (in he Korea and Malaysia model) and more han 2% (in he hailand), impors converge almos geomerically back o seady sae. In he daa, each of he economies faced a sharp conracion in impors early in he crisis period. In each case, he rough conracion was approximaely 3% below he previous pah. In all cases, impors reurned o heir long-run level afer approximaely hree years. In he daa, boh he raded and non-raded secors conraced sharply in he iniial par of he crisis. In each economy, raded goods converged back o he pre-crisis pah almos wo years before non-raded goods producion recovered. In he model, he presence of shor-erm price sickiness in he domesic componen of raded goods and he slow adjusmen of expor quaniies combine o cause a fall in he oupu of raded goods in he immediae afermah of he shock. In he model calibraed o Korea, raded goods producion conracs by more han % in he period of he shock. In he models calibraed o Malaysia and hailand, raded goods producion conracs beween 7-%. As in he daa, he conracion in raded goods producion is more ransiory han ha in non-raded goods. By he hird period of he crisis, raded goods producion is slighly above seady sae in each case, and expands in subsequen periods boh because exporers are able o reduce foreign currency prices of expors in hose periods and because raded goods producers are able o subsiue cheaper impored maerials for domesic value added. 7

18 Qualiaively, he model generaes a large and persisen depreciaion in he nominal exchange rae. Quaniaively, he exchange rae depreciaion is in he range of -2% compared wih persisen depreciaions of a leas 4% observed in he daa. Moreover, he model has none of he overshooing seen in he daa. he failure of he model o capure he shor-run dynamics of exchange rae growh is ineresing in ha he model is calibraed o exacly mach domesic and foreign ineres raes. In each version of he model, he impor and expor price level immediaely rises in he period of he shock. Qualiaively, his dollar-currency pricing maches he daa. Quaniaively, he rise in impor and expor prices is small relaive o he daa, jus as he exchange rae depreciaion is small relaive o he daa. In he model, he dynamic response of he impors deflaor is equal o he response of he nominal exchange rae, since foreign prices are exogenous. he impor price rise is persisen in he model; in he daa, impor prices converges back o he pre-crisis pah as if foreign producers were cuing heir foreign currency prices. In he model, he rise in expor prices in he period of he shock exacly maches he exchange rae depreciaion. Over ime, he foreign currency price of expor goods is adjused downward. he daa also follows his paern. In he model, he rise in impor prices resuls in a small rise in he absorpion deflaor, varying beween.% and %. Over ime, in each case, prices converge back o he iniial seady sae as domesic prices fall. On his coun, our model seems o mach he daa quie closely. he peak rise in he absorpion deflaors is in he range of 6-8%. In he model, raded goods prices rise iniially, as par of raded goods is used for expors. Over ime, raded prices fall as domesic prices are reduced. his paern is refleced in he daa. he rise in raded prices ranges beween 2 and 6% in he daa and beween 4 and 2% in he model, hus reflecing he relaively small depreciaion in he model. In he model, he non-raded price level is quasi-fixed. Afer he iniial period, he decline in nominal demand resuls in a fall in he non-raded price level. he empirical behavior in non-raded prices in Korea seems o mach his paern, boh qualiaively and quaniaively. In hailand and Malaysia, he non-raded price level is subjec o high-frequency movemens, again making he effec of he shock difficul o delineae. In hese counries, he quaniaive decline in he non-raded price level is small in he daa relaive o he model. By consrucion, modeled nominal ineres raes exacly mach he daa in he firs hree periods of he crisis. In subsequen periods, when moneary policy follows a aylor rule, he ineres rae declines in boh he model and he daa. he empirical size of he decline in he ineres rae in Korea was larger han in he model. he empirical declines in hailand and Malaysia seem comparable wih he oucomes in he model. A naural quesion o ask of his model is how alernaive moneary policies would have changed he oucome of he crisis. he dashed lines in each panel show he pah of he variables under a sric aylor rule wih no exogenous rise in ineres raes (i.e.{ ε = } ). In hailand, which has a = 8

19 relaively high share of impors in final goods and hus a relaively sharp modeled response of inflaion o an exchange rae devaluaion, he aylor rule proscribes a sharp immediae rise in he nominal ineres rae. hailand s acual nominal ineres rae was closes o ha modeled under a aylor rule. hus, in hailand, of he hree counries, he benchmark resuls under he acual ineres raes are closes o he resuls under he pure aylor rule. he conracion in hailand, in erms of each of he depiced real variables, is slighly milder and noiceably less persisen under a pure aylor rule han under observed ineres raes. he nominal exchange rae depreciaes slighly more in hailand under a aylor rule response han under observed ineres raes. However, his effec is small. In he model of Korea, he moneary policy response is a mild bu persisen cu in he nominal ineres rae. In he model of Malaysia he moneary policy response in he firs period is a small rise in he ineres rae followed by a persisen, bu mild cu in nominal ineres raes. hese more expansionary moneary policies resul in much less sharp and persisen conracions in GDP and invesmen relaive o he benchmark case. he rise in he exernal ineres rae sill implies a persisen decline in consumpion under each moneary policy. he exchange rae depreciaion is sharper under he aylor rule, especially in Korea. However, his exchange rae depreciaion is sill much smaller han ha observed in he daa. V. Conclusion Our inerpreaion of hese resuls is ha here is sufficien mach beween models and daa o build confidence in he sicky-price open macroeconomic framework. In his sense, our model does an adequae job of accouning for he experiences of hese hree counries following he Eas Asia crisis, when we measure he shocks using he observed rise in foreign currency risk premium, calibrae he srucural models of hese economies o reflec he observed feaures of hese economies, and specify a moneary reacion rule ha mirrors he behavior of domesic ineres raes in he afermah of he crisis. In erms of maching he response of mos quaniies, including oupu, consumpion, invesmen, impors, and expors, as well as he secoral composiion of oupu, our model seems quie successful. he dollar-currency pricing mechanism in our model allows us o undersand he simulaneous observaion of weak expor response wih a large fall in impors purely hrough subsiuion and income responses o he crisis, and wihou recourse o exogenous credi or financial consrains. he key shorcomings of he model include he following. ) Lack of Persisence of Oupu. he model generaes conracions in oupu ha a rough are comparable in size wih he daa. However, he conracion is less persisen in he model han in he daa along wo dimensions. Firs, he rough of he conracion occurs in he period of he shock in he model bu several periods laer in he daa. Second, producion converges back o seady sae faser in he model han in he daa. Our model simplifies his by assuming informaion abou he fuure dynamic pah of he counry risk premium ha was 9

20 revealed in he iniial period. A richer dynamic of informaion revelaion would be likely o generae richer dynamic responses. 2) Exchange Rae Devaluaions are oo small. he model generaes oo small an exchange rae depreciaion relaive o he daa, despie he consrucion of exac maches of domesic and foreign ineres raes in he periods following he shock. he model embeds an uncovered ineres rae pariy assumpion ha has been rejeced in he paper and seems o display some shorcomings here. Bibliography Aghion, P., P. Bacchea, and A. Banerjee, 2, A Simple Model of Moneary Policy and Currency Crises, European Economic Review 44, Aghion, P., P. Bacchea, and A. Banerjee, 2, Currency Crises and Moneary Policy in an Economy wih Credi Consrains, European Economic Review 4, 2-. Arellano, C. and E.G. Mendoza, 22, Credi Fricions and Sudden Sops in Small Open Economies: An Equilibrium Business Cycle Framework for Emerging Markes Crises, NBER Working Paper No Backus, D.K., P.J. Kehoe, and F.E. Kydland, 992, Inernaional Real Business Cycles, Journal of Poliical Economy, Baxer, M. and M. Crucini, 993, Explaining Saving-Invesmen Correlaions, American Economic Review 83, Bernanke, B., M. Gerler, and S. Gilchris, 2, he Financial Acceleraor in a Quaniaive Business Cycle Framework, in J. aylor and M. Woodford, eds. Handbook of Macroeconomics, Amserdam: Elsevier. Bes, C. and M.B. Devereux, 996, he Exchange Rae in a Model of Pricing-o-Marke, European Economic Review 4, 7-2. Bes, C. and M.B. Devereux, 2, Exchange Rae Dynamics in a Model of Pricing-o-Marke, Journal of Inernaional Economics, Burnside, C., M. Eichenbaum, and S. Rebelo, 2a, Hedging and Financial Fragiliy in Fixed Exchange Rae Regimes, European Economic Review 4, -93. Burnside, C., M. Eichenbaum, and S. Rebelo, 2b, Prospecive Deficis and he Asian Currency Crisis, Journal of Poliical Economy Bursein, A. M. Eichenbaum, and S. Rebelo, 22, Why are Raes of Inflaion so Low afer Large Devaluaions? NBER Working Paper No. w8748. Calvo, G.A., 983, Saggered Prices in a Uiliy-Maximizing Framework, Journal of Moneary Economics 2 (3),