Cheap labor meets costly capital: the impact of devaluations on commodity firms

Transcription

1 Journal of Developmen Economics 69 (2002) Cheap labor mees cosly capial: he impac of devaluaions on commodiy firms Krisin J. Forbes* MIT-Sloan School, 50 Memorial Drive, Cambridge, MA 02142, USA 1050 Massachuses Avenue, Cambridge, MA 02138, USA Absrac This paper examines how devaluaions affec he relaive coss of labor and capial and, herefore, influence producion, profiabiliy, invesmen and sock reurns for firms in he crisis counry, as well as compeiors in he res of he world. Afer developing hese ideas in a small, open-economy model, he paper performs a series of empirical ess using informaion for abou 1100 firms in 10 commodiy indusries beween 1996 and The empirical ess suppor he model s main predicions: (1) immediaely afer devaluaions, commodiy firms in he crisis counry have oupu growh raes abou 10 20% higher han compeiors in oher counries; (2) immediaely afer devaluaions, commodiy firms in he crisis counry have operaing profi growh raes abou 15 25% higher han compeiors in oher counries; (3) he effec of devaluaions on fixed capial invesmen and sock reurns (and, herefore, expeced long-run oupu and profis) is deermined by capial/labor raios and changes in he cos of capial. For example, crisis-counry firms have higher raes of capial growh and beer sock performance afer devaluaions if hey had lower capial/labor raios and here was no subsanial increase in heir ineres raes. D 2002 Elsevier Science B.V. All righs reserved. JEL classificaion: F1; F2; F3 Keywords: Currency crises; Devaluaions; Commodiies; Exchange-rae exposure 1. Inroducion In 1997, Indonesia abandoned is pegged exchange-rae regime and allowed he rupiah o depreciae subsanially. In 1998, Indonesia s economy conraced by abou 14%. Several companies and indusries, however, increased producion and profis improved * Tel.: ; fax: address: kjforbes@mi.edu (K.J. Forbes). URL: hp://web.mi.edu/kjforbes/www /02/$ - see fron maer D 2002 Elsevier Science B.V. All righs reserved. PII: S (02)

2 336 K.J. Forbes / Journal of Developmen Economics 69 (2002) significanly. For example, in 1998, he planaion secor (which includes coffee, cocoa, rubber, palm oil and ea) grew by 6.5%. One of he larges planaions in he counry repored ha profis increased by a muliple of four. 1 Over he same period, however, numerous firms complained abou a credi crunch and heir inabiliy o obain financing o increase producive capaciy and ake advanage of lower dollar expor prices. Many firms even claimed ha hey were unable o obain enough working capial o purchase inpus necessary o coninue producing a predevaluaion levels. Financial crises no only affec firms in he devaluing counry, bu can also impac compeiors around he world. Afer Brazil devalued he real in January of 1999, Argenine companies were forced o reduce heir expor prices for soya beans in order o compee wih cheaper Brazilian expors. During he Asian and Russian crises, sock reurns for firms ha compeed wih devaluing counry expors were significanly lower han for oher firms in he same counries. 2 This suggess ha he Asian and Russian devaluaions negaively impaced expeced earnings and profis for firms ha compeed wih expors from he crisis counries. This paper examines how devaluaions affec relaive coss, producion decisions and profiabiliy for firms wihin a crisis counry, as well as compeiors in oher counries (where a crisis counry is defined as a counry ha devalues is currency). More specifically, i analyzes how devaluaions influence firms oupu, profiabiliy, capial invesmen and sock reurns, as well as indusry prices and quaniies, in he shor and long run. In he heoreical model, firms use wo variable inpus (labor and maerials) and one fixed inpu (capial). Labor is priced in domesic currency. Maerials and capial are priced in dollars, and he price of capial also incorporaes any domesic capial marke developmens. The immediae impac of devaluaions is o lower he relaive cos of labor in he crisis counry. Firms in he devaluing counry increase oupu and profis, while compeiors decrease oupu and profis. In he longer erm, however, devaluaions raise he cos of capial for firms in he crisis counry (poenially by more han he exchangerae movemen). If his increase in he cos of capial is large enough and he firm s capial/ labor raio is high enough, more expensive capial could ouweigh he benefis of relaively cheaper labor. Therefore, in he long run, devaluaions could decrease oupu, profis and invesmen for firms in he devaluing counry, and increase hese variables for compeiors in oher counries. The empirical secion of he paper uses daa for over 1100 firms in 10 commodiy indusries o es he model s main predicions during a series of devaluaions beween 1997 and Resuls show ha immediaely afer devaluaions, commodiy firms in he crisis counry have oupu growh raes abou 10 20% higher and profi growh raes abou 15 25% higher han compeiors in oher counries. These growh effecs are shorlived and end o disappear wihin a year, alhough he levels of oupu and profis remain higher indefiniely. Moreover, invesmen growh and sock reurns (boh of which signal changes in expeced long-erm oupu and profiabiliy) are correlaed wih capial/labor raios and changes in he crisis counry cos of capial. For example, crisis-counry firms have larger increases in capial invesmen and beer sock performance afer devaluaions 1 Saisics from he McCarhy (1999). 2 Forbes (2000).

3 K.J. Forbes / Journal of Developmen Economics 69 (2002) if hey had lower capial/labor raios and heir ineres raes did no increase subsanially. Therefore, alhough he empirical analysis is no a formal es of he heoreical model, he esimaes agree wih he model s cenral predicions. Even hough crisis-counry firms may benefi from cheaper labor immediaely afer devaluaions, hey may no benefi in he longer erm if hey use capial inensively and/or heir capial becomes more cosly. This paper focuses on firms ha produce commodiies (or any undiffereniaed produc) for hree reasons. Firs, commodiy expors are a large share of GDP in mos counries ha abruply devalue heir currency. Moreover, many of hese counries rely heavily on commodiy expors for ax revenues and foreign currency. Second, alhough here has been a subsanial amoun of research on he impac of currency movemens on differeniaedgoods firms (such as he pricing-o-marke lieraure), here has been relaively lile analysis for homogenous goods. Third and finally, he producion srucure for mos commodiies is exremely useful in isolaing some of he key effecs of devaluaions. For example, many commodiies require a large, fixed invesmen (such as planing rees or drilling mines) ha mus be made several years before he resuling oupu is sold. This srucure helps differeniae beween he shor- and long-run impac of devaluaions. In addiion, he producion of mos commodiies requires impored inpus and capial, as well domesic labor, so ha i is possible o capure how devaluaions inerac wih relaive inpu inensiies and inpu coss o affec oupu, profis and invesmen. These facors appear o have played an imporan role in deermining how recen devaluaions affeced firms around he world. The remainder of his paper is divided ino four secions. Secion 2 reviews several branches of relaed lieraure. Secion 3 presens a heoreical model of how devaluaions affec firms in he crisis counry and he res of he world in he shor and long erm. I esablishes several condiions under which devaluaions are more likely o increase firms oupu, profiabiliy and invesmen levels. Secion 4 uses firm-level daa for a series of devaluaions beween 1997 and 2000 o es four cenral predicions of he model. Secion 5 concludes. 2. Lieraure overview This paper is relaed o six diverse branches of lieraure, each of which is so exensive ha his overview does no make any aemp o discuss all of he relevan work. Insead, i simply highlighs he key quesions and refers o recen surveys and closely relaed aricles. 3 The firs branch of lieraure examines how devaluaions affec expors. Ghei and Priche (1999) provide a deailed summary of why devaluaions may or may no improve expor performance, as well as he empirical challenges in measuring hese effecs. They conclude ha expors ypically increase afer a devaluaion, and ha mos of his response occurs rapidly. A second and closely relaed branch of lieraure examines how devaluaions and crises affec no only expors, bu also oher variables such as oupu, invesmen 3 Forbes (2002) also provides a more deailed review of relaed lieraure.

4 338 K.J. Forbes / Journal of Developmen Economics 69 (2002) and inflaion. Agénor and Moniel (1996) develop a general-equilibrium model exploring hese effecs and survey relaed work. They conclude ha he evidence on wheher devaluaions are conracionary is mixed. 4 While hese branches of lieraure focus on he domesic impac of devaluaions and crises, a more recen and rapidly growing body of work examines conagion and how crises affec oher counries. Claessens and Forbes (2001) include recen surveys of his lieraure, as well as case sudies and empirical analyses. While almos all of his empirical work uses macroeconomic daa, one excepion closely relaed o his paper is Forbes (2000), who examines how he Asian and Russian crises affeced sock reurns for over 10,000 companies around he world. A fourh relevan branch of lieraure also focuses on sock reurns and measures he exen of exchange-rae exposure for various companies. Mos papers esimae reduced-form, marke models of how exchange-rae movemens affec sock reurns. Dominguez and Tesar s (2001) work is one of he mos recen and horough examples of his lieraure. They conclude ha abou 12 23% of firms are significanly exposed o exchange-rae movemens. A fifh lieraure relaed o his paper examines how exchange-rae movemens affec oupu prices. This lieraure includes he work on pricing-o-marke and pass-hrough and emphasizes he role of indusrial srucure. 5 Dornbusch (1987) develops hese ideas heoreically, and Goldberg and Kneer (1997) survey he empirical evidence of pricing-o-marke in specific indusries. 6 A final lieraure is he exensive work on commodiies. This work includes everyhing from he analysis of how commodiy price movemens affec macroeconomic saisics, o he role of carels and impac of El Nino. Cashin e al. s (1999) work is a recen paper ha surveys much of his work and analyzes commodiy-price cycles. Despie he range of heoreical frameworks and empirical ess used in hese six branches of lieraure, none of hese papers has explicily addressed he key quesion explored in his paper: how do devaluaions affec oupu, profiabiliy and invesmen decisions of firms around he world? The majoriy of his lieraure focuses on macroeconomic relaionships and counry-level evidence. The lieraure on exchange-rae exposure and pricing-o-marke uses firm-level models and daa, bu focuses on how exchange-rae movemens affec sock reurns or produc prices. The pricing-o-marke lieraure focuses on differeniaed goods in developed counries. Moreover, none of hese branches of lieraure has focused on he key radeoff analyzed in his paper: how devaluaions give firms a relaive cos advanage in erms of cheaper labor and disadvanage in erms of more expensive capial. This radeoff generaes a number of ineresing predicions for firms in he devaluing counry, as well as compeiors in he res of he world. 4 Edwards (1989) also provides lieraure surveys and deailed empirical sudies. 5 Several papers have exended his framework o examine how exchange-rae movemens affec micro-level variables for example, Campa and Goldberg (1999) examine he effec on secoral invesmen. 6 One sudy ha combines his approach wih he work on exchange-rae exposure is ha of Allayannis and Ihrig (2000). They examine how marke srucure, including expor and impor compeiiveness, affecs he exchange-rae exposure of a large sample of U.S. firms.

5 3. The heoreical model K.J. Forbes / Journal of Developmen Economics 69 (2002) This secion develops a model o show how devaluaions affec firms oupu, profis and capial invesmen in he shor and long run. Secion 3.1 models firms decisions in he shor run when heir level of capial is fixed. Secion 3.2 models he long run when firms can adjus heir capial levels. Secion 3.3 examines he shor- and long-run impac of devaluaions. This model and framework form he basis of he empirical ess in Secion 4. Before developing he model in deail, he nex few paragraphs highligh is key componens and cenral predicions. Each firm uses hree inpus (labor, maerials and capial) o produce he same commodiy. Firms are locaed in wo counries, he crisis counry (ha devalues is currency) and he res of he world (r.o.w.). Each firm produces a small share of global oupu and has no impac on global prices. There are no rade barriers or ransporaion coss. As a resul, he commodiy s price is deermined by global supply and demand. Firms make heir producion decisions in wo sages. In he shor run, each firm s level of capial is fixed. This fixed invesmen could include anyhing from planing rubber rees o drilling mines. Given his fixed capial, each firm chooses is opimal mix of labor and maerials o maximize shor-run profis. Labor is priced in domesic currency and can be inerpreed as any local componen of producion, while maerials are priced in r.o.w. currency and can be inerpreed as impored inpus. Firms have company-specific produciviy parameers. In he long erm, each firm can also adjus is capial level. Capial is priced in r.o.w. currency and includes a counry-specific componen (o capure differences in domesic risk and local capial markes). Each firm chooses is level of capial expecing relaive prices and exchange raes o remain consan (a leas unil he nex chance o inves). Then he crisis counry devalues is currency. In he shor run, he devaluaion reduces he relaive cos of labor in he crisis counry. This causes crisis-counry firms o increase oupu, lowering he global price of he commodiy. Firms in he r.o.w. respond by decreasing producion, alhough by less hen he oal increase by crisis-counry firms. Therefore, he aggregae shor-run impac of he devaluaion is o increase global producion and decrease he oupu price. The magniude of hese effecs is deermined by he crisis counry s share of global producion and he share of labor in producion. The devaluaion also decreases profis for firms in he r.o.w. and increases profis for firms in he crisis counry. Over longer periods, however, firms can adjus heir capial levels and he above predicions can be reversed. Since capial is priced in r.o.w. currency, devaluaions increase he cos of capial for firms in he crisis counry, possibly by even more han he exchange-rae movemen if here is a simulaneous increase in domesic risk or conracion in lending. More specifically, if he firm s capial/labor raio is large enough, or he increase in he cos of capial is large enough, he devaluaion could raise he oal cos of producion for crisis-counry firms and ouweigh he benefis of relaively cheaper labor. Crisis-counry firms would decrease oupu, invesmen and profis, raising he long-run price of he commodiy and causing compeiors o increase producion, invesmen and profis. Therefore, alhough he model s shor-run

6 340 K.J. Forbes / Journal of Developmen Economics 69 (2002) predicions of how devaluaions affec oupu and profis are fairly clear, he long-run predicions depend on firm s capial/labor raios and relaive changes in inpu coss. Before developing his model in deail, i is worh menioning wha he model does no consider. Firs, i does no allow for any sor of sraegic pricing behavior. Firms are assumed o ake he global oupu price as given and are unable o affec his price by adjusing producion or forming carels. Similarly, exchange-rae movemens are fully passed hrough ino maerial inpu prices. Second, alhough exogenous shocks can affec global demand for he commodiy in each period, he model assumes ha firms expor mos of heir oupu, so here is no direc impac of he devaluaion on global demand. Finally, all prices are in real erms and he real impac of he devaluaion on relaive prices is no eroded by differences in inflaion raes across counries The shor-run wih a fixed level of capial 7 In order o produce mos commodiies, firms mus make a large, upfron invesmen in fixed capial. Afer making his iniial invesmen, here is ofen a subsanial ime lag before he firs uni of oupu is sold. For example, here is abou a 6-year lag afer coffee is planed unil he beans can be sold. To capure his, I model firms decisions in wo sages. In he shor run, defined as he periods from =1,...T, a firm s level of capial is fixed. In he long run when >T (which is modeled in Secion 3.2), a firm is able o choose is level of capial. 8 Beginning wih his shor-run scenario, each firm i has a fixed level of capial k i >0. Oupu in each period is deermined by he choice of wo variable inpus: domesic labor (l i, ) and impored maerials (m i, ). Firm oupu is given by a Cobb Douglas producion funcion ha has decreasing reurns o scale 9 : q i; ¼ A i k a i lb i; mc i; wih a þ b þ c < 1: ð1þ A i is a echnology parameer (which varies across firms). The cos of labor is w and of maerials is s. The firm chooses variable inpus o maximize shor-run profis (p i, SR )in each period: max l;m psr i; ¼ P q i; w l i; s m i; ; ð2þ where P is he sale price per uni of oupu. Each firm produces idenical goods and here is no differeniaion beween markes (i.e. no barriers o rade or ransporaion coss). Therefore, he global price of he good can be 7 This shor-run version of he model is loosely based on Dornbusch (1987). 8 Alhough he model does no explicily include enry and exi, firms can se oupu and capial o zero in he long run and, herefore, exi. Enry is capured by allowing firms o increase heir capial levels. 9 Decreasing reurns o scale ensures ha he mos efficien firm does no produce all of global oupu.

7 K.J. Forbes / Journal of Developmen Economics 69 (2002) expressed as a funcion of global oupu ( Q ), and o simplify he model s soluion, assume ha his global price is deermined by a consan-elasiciy demand funcion: P ¼ Z Q u ð3þ where Z is any period-specific shock o global demand and 1/u is he elasiciy of demand. 10 In addiion, assume ha each firm produces a relaively small fracion of global oupu and, herefore, akes inpu coss and he indusry price as given. These assumpions are fairly accurae descripions of compeiion in mos commodiies. Wihou loss of generaliy, assume ha firms are locaed in wo counries : he crisis counry (which devalues is currency) and he res of he world (r.o.w.). In he noaion ha follows, all variables for he crisis counry ha differ from he r.o.w. are wrien wih a f, and firms in he crisis counry are indexed by j. The oupu price ( P) and cos of impored maerials (s) are boh expressed in r.o.w. currency, which can be inerpreed as dollars. In he shor run, he only price ha differs beween he wo counries is he domesic wage. The exchange rae is he relaive cos of labor in he wo counries, e ¼ w w ; ð4þ so ha a devaluaion in he crisis counry is a decrease in e. For a fixed invesmen and given oupu price, each firm chooses is opimal combinaion of he wo variable inpus in order o maximize profis in period. If he r.o.w. wage is normalized o equal one, hen he opimal oupu levels for firms in he r.o.w. and crisis counry are: " q SR i; ¼ P bþc A i ki a b b # 1 c c 1 b c and w q SR j; ¼ P bþc A j ka j b e s c " b # 1 c s 1 b c : ð5þ Nex, assume ha here are n firms in he r.o.w, and ñ firms in he crisis counry. Global oupu can be expressed as: Q ¼ Z n i¼1 q i; þ Z ñ j¼1 q j; : Appendix A presens he full soluions for he global price and quaniy The long-run wih variable capial In periods longer han T, firms can adjus heir capial levels. Capial is priced in r.o.w. currency, bu prices vary across counries due o facors such as domesic risk, capial marke liquidiy, capial conrols, ec. Firms chose heir opimal levels of capial (k i z 0 and 10 The commodiy is a normal good, so ha u>0. ð6þ

8 342 K.J. Forbes / Journal of Developmen Economics 69 (2002) Table 1 Model predicions for he impac of devaluaions Sign Magniude Caveas Shor-run predicions dq SR de <0 C b e ð ˆQ W;SR ˆQ C;SR uðbþcþ 1 b cþuðbþcþ Þ dp SR de >0 CV b e ˆQ C;SR ð ˆQ W;SR Þ 1 b cþu 1 b cþuðbþcþ dq SR i; de >0 q SR i; 1 b c ðb þ cþ P SR dp SR de d q SR j; de <0 q SR j; 1 b c ðb þ cþ P SR dp SR de b e dp SR i; de >0 p SR i; 1 b c 1 P SR dp SR de d p SR j; de <0 p SR j; 1 b c 1 dp SR b P de e if ð1=uþ > 1 ˆQ C;SR and=or ˆQ W;SR 1 b c þ uðb þ cþ < u Long-run predicions dq LR? de K b e þ ã r d r de ð ˆQ W;LR ˆQ C;LR uðbþcþaþ I b c aþuðbþcþaþ Þ < 0if b > ã d r e r de dp LR de? KV b e þ ã r d r de ð ˆQ W;LR ˆQ C;LR 1 b c aþu 1 b c aþuðbþcþaþ Þ > 0 if b > ã d r e r de dq LR i; de? q LR i; 1 b c a ðb þ c þ aþ P LR dp LR de > 0 if b > ã d r e r de dk LR i; de? ki; LR 1 b c a 1 P LR dp LR de R > 0 if b > ã d r e r de

9 K.J. Forbes / Journal of Developmen Economics 69 (2002) Table 1 (coninued) dp LR i; de? Sign Magniude Caveas p LR i; 1 dp LR > 0 if 1 b c a P LR de R b > ã d r e r de d q LR j; de? q LR j; 1 b c a ðb þ c þ aþ P LR dp LR de b ã d r e r de < 0 if b > ã d r e r de? klr j; de 1 b c a d k LR j; 1 P LR dp LR de b ð1 b cþ d r e r de < 0 if b e > ã d r r de and 1 u > H and=or ˆQ C;LR ˆQ W;LR < HV d p LR j; de? p LR j; 1 b c a 1 dp LR b ã d r P de e r de < 0if b e > ã r 1 u > X and=or ˆQ C;LR ˆQ W;LR < XV d r de and Noes: ˆQ R;SR ¼ n Ā k a 1 1 b c w b s c ; ˆQ C;SR ¼ ñ 1 1 b c a à k a e b s c 1 1 b c ; ˆQ W;SR 1 1 b c a Ā à ¼ n w b s c r a ; ˆQ C;LR ¼ ñ e b s c r a C; CV; K; KV; H; HV; X; XV¼ f ða; b; c; T; Z; u; e; r; d r=deþ: ¼ ˆQ R;SR ; ˆQ W;LR þ ˆQ C;SR ; ˆQ R;LR ¼ ˆQ R;LR þ ˆQ C;LR and k j z 0) a cos r and r o maximize long-run profis (p LR T ) unil he nex opporuniy o inves. To simplify he algebra, assume ha here is no discouning. The profi-maximizaion equaion for r.o.w. firms is: max p LR i; ¼ E k Z T ¼1 ðp q i; w l i; s m i; Þ rk i : ð7þ Nex, assume ha companies expec inpu prices and demand shocks o be consan (so ha E[w ] =w, E[e ]=e, E[s ]=s and E[Z ]=Z). Then prices from =1oTare expeced o be consan and E[ P ]=P. Subsiuing he shor-run soluions ino he long-run profimaximizaion equaion, he opimal capial invesmen for firms in he r.o.w. and crisis counry are: 1 A i P k i ¼ f w b s c r 1 b c 1 b c a and kj ¼ f 1 1 b c a A j P e b sc r 1 b c ð8þ

10 344 K.J. Forbes / Journal of Developmen Economics 69 (2002) wih f = f(a, b, c, T). As a resul, firms will chose higher capial levels if: hey are more producive, he expeced oupu price is higher, or any of he inpu prices are lower. Using Eq. (8), i is possible o solve for each firm s opimal long-run oupu. As long as he price realizaions are equal o heir expeced values, he long-run oupu soluions are: 1 1 b c a q LR i; ¼ W A ip bþcþa w b s c r a 1 1 b c a and q LR j; ¼ W A jp bþcþa e b s c r a wih W = f(a, b, c, T). Appendix A repors he long-run global quaniy and price soluions The impac of devaluaions on firms around he world Using he model developed above, i is possible o predic he impac of devaluaions on firms in he crisis counry and r.o.w. in boh he shor run (when capial is fixed) and he long run (when capial is adjusable). Assume ha here is an unexpeced devaluaion ha has no impac on global demand for he commodiy (i.e. ha dz /de = 0). 11 In he shorrun, he main impac is o reduce he relaive cos of labor in he crisis counry. The op of Table 1 repors he prediced impac of devaluaions on global oupu and prices and shows ha devaluaions (a decrease in e) cause global oupu o increase and prices o fall. These effecs are proporional o he share of labor in producion and he share of global oupu produced by crisis-counry firms. The nex secion of Table 1 shows he shor-run effec of devaluaions on firm oupu and profis. 12 Exchange-rae movemens affec firms in he r.o.w only hrough movemens in he global commodiy price. Therefore, devaluaions cause r.o.w. firms o reduce oupu and profis. On he oher hand, exchange-rae movemens affec crisis-counry firms hrough wo channels in he shor run: he global commodiy price and he relaive cos of labor. The firs effec is he same as for firms in he r.o.w. Couneracing his oupu price effec, however, is an inpu price effec. Devaluaions reduce he cos of labor in he crisis counry relaive o he cos of oher inpus, as well as relaive o firms in he r.o.w. Algebraic manipulaion shows ha his second inpu price effec always dominaes he firs oupu price effec, so ha devaluaions unambiguously cause crisis-counry firms o increase oupu in he shor run. Finally, since global oupu increases and oupu by he r.o.w. decreases, oupu by he crisis counry mus increase by even more han he increase in global producion. Therefore, firms in he crisis counry will sell o cusomers/markes ha were previously serviced by firms in he r.o.w. Devaluaions will also increase shor-run profis for firms in he crisis counry, wih one addiional cavea: if he global elasiciy of demand for he produc is greaer han one and/ or if he crisis counry has a small enough share of global oupu. In oher words, if he price decline leads o a large enough increase in global demand, or if he impac on global prices is fairly small (since he crisis counry only produces a small fracion of global 11 This assumpion is realisic for commodiy firms in mos emerging markes since he majoriy of producion is expored o developed counries, and mos devaluaions have minimal impac on heir demand. In selec cases, such as he 1998 Russian crisis, his assumpion may no apply. I is sraighforward o exend he model and allow exchange-rae movemens o affec global demand. 12 Shor-run profis do no include he cos of capial invesmen and are defined in Eq. (2). ð9þ

11 K.J. Forbes / Journal of Developmen Economics 69 (2002) oupu), hen devaluaions increase shor-run profis for crisis-counry firms. There is a lenghy debae in he developmen lieraure on wheher he elasiciy condiion (1/u>1) holds for mos commodiies. Esimaes of he price elasiciy of demand are highly dependen on an individual commodiy s characerisics, how narrowly i is defined, and he lengh of ime consiuing shor run. In mos cases, however, he oupu-share condiion (Qˆ Crisis,SR /Qˆ World,SR < X) is saisfied because he producion of mos commodiies is widely dispersed across counries. 13 Moreover, even for commodiies wih exremely low price elasiciies of demand, his condiion should be saisfied. 14 Therefore, devaluaions are expeced o increase shor-run profis for firms in he crisis counry. Over he longer erm, however, he shor-run impac of devaluaions on oupu and profis can be reversed. Devaluaions also affec he relaive cos of capial, and when z T, firms can adjus heir invesmen accordingly. More specifically, assume ha: dr =de ¼ 0 and d r =de < 0: ð10þ In oher words, devaluaions have no impac on he cos of capial for r.o.w. firms bu increase he cos for crisis-counry firms. This increase could occur because if capial invesmen is financed in r.o.w. currency and/or impored, hen he crisis-counry ineres rae would move in proporion o he devaluaion. If devaluaions also raise domesic ineres raes by increasing he counry risk premium, conracing bank lending, and/or decreasing propery and collaeral values, hen ineres raes in he crisis counry could increase by significanly more han he exchange-rae movemen. The middle of Table 1 shows he prediced long-run impac of an exchange-rae movemen on global oupu and prices. Devaluaions can eiher increase or decrease global oupu and prices based on wo condiions: he relaive share of labor and capial in producion and he impac on he crisis-counry s cos of capial. If labor is a more imporan componen of producion han capial and/or if he impac of he devaluaion on ineres raes is small, hen he devaluaion is more likely o increase global producion and decrease global prices. The equaions also show he inuiive resul ha he impac of devaluaions on global oupu and prices is greaer when he crisis counry produces a larger share of global oupu. The long-run impac of devaluaions on firm oupu, capial and profis is shown a he boom of Table 1. Exchange-rae movemens coninue o affec oupu and profis for r.o.w. firms, as well as invesmen levels, only hrough movemens in he global commodiy price. Therefore, a devaluaion could cause r.o.w. firms o eiher increase or decrease heir oupu and invesmen, based on wheher he devaluaion leads o a long-run increase or decrease in he global price as deermined by he above crieria. If producion is relaively more capial inensive, or if he devaluaion causes a large increase in ineres 13 Even when a counry is heavily specialized in a specific commodiy, i rarely has a dominan share of global producion. Agénor and Moniel (1996) make his poin and documen ha only 16 developing counries have as much as 10% of he world marke for any commodiy (based on hree-digi SITC groups). 14 For example, if u = 5 so ha he price elasiciy of demand for he commodiy is 0.2 (which is unrealisically low), and b + c = 0.5 (which is also unrealisically low for an emerging marke), hen X = In oher words, even using exreme parameer values which make i more difficul o saisfy his condiion, a counry would have o produce over 60% of global producion in order for Qˆ Crisis,SR /Qˆ World,SR < X.

12 346 K.J. Forbes / Journal of Developmen Economics 69 (2002) raes, hen here is a greaer chance ha he global price increases and r.o.w. firms subsequenly increase oupu and invesmen. Finally, as shown a he boom of Table 1, devaluaions affec oupu, profis and capial for crisis-counry firms hrough he same global price effec as for r.o.w. firms, as well as hrough changes in he relaive coss of labor and capial. Algebraic manipulaion, however, shows ha he sign of he impac depends on he same wo crieria: capial labor raios and changes in he crisis-counry cos of capial. Devaluaions only increase oupu, invesmen and profis for crisis-counry firms in he long run if: labor s share is relaively larger han capial s share and he increase in he cos of capial is no oo large. 15 This is inuiive. Devaluaions only cause crisis-counry firms o increase oupu, invesmen and profis if he cos advanage gained from cheaper labor is greaer han he cos disadvanage from more expensive capial. Therefore, in he long run devaluaions have he opposie impac on firms in he crisis counry han firms in he r.o.w. 4. Empirical ess This secion ess four of he model s key predicions. Secion 4.1 describes he daa se, devaluaion episodes and commodiy groups used for he analysis. Secions 4.2 and 4.3 examine how devaluaions affec oupu growh and profi growh, respecively. Secions 4.4 and 4.5 assess he expeced long-run impac by examining changes in firms invesmen and sock reurns. These wo secions also consider how capial/labor raios and changes in he cos of capial deermine he impac of devaluaions on differen groups of firms. Alhough his empirical analysis is no a formal es of he full heoreical model, he resuls suppor he model s cenral predicions for how devaluaions impac oupu, profiabiliy and invesmen for firms around he world The evens, daa and commodiy groups The empirical analysis focuses on major devaluaions in eigh counries beween January 1997 and January Major devaluaions are defined as episodes where he local currency/u.s. dollar exchange rae increased by 15% or more wihin any 4-week period. Table 2 liss he counries wih major devaluaions in chronological order, as well as he monh of he devaluaion. 17 This lis of devaluaion episodes includes he sandard evens ypically analyzed in he currency-crisis lieraure: several Asian counries in , Souh Africa and Russia in 1998 and Brazil in The impac of devaluaions on capial and profis in crisis-counry firms also depends on he elasiciy and/ or global oupu share crieria. As discussed above, however, hese condiions are expeced o be saisfied since commodiy producion is rarely concenraed in an individual counry. 16 This period was chosen o correspond wih he available firm-level daa. Exchange-rae daa is from Daasream. 17 Afer a devaluaion, he nex 4 weeks are excluded so ha here can be, a mos, one devaluaion even wihin any 4-week period. The only major devaluaion ha is no included in his analysis is for Ecuador saring in January of This even is excluded because here is no firm-level daa for Ecuador and Ecuador expors less han 1% of global expors for each of he commodiy groups in he sample.

13 K.J. Forbes / Journal of Developmen Economics 69 (2002) Table 2 Major devaluaion evens a Counry b Devaluaion period Thailand 7/97, 12/97, 1/98 Indonesia 10/97, 12/97, 1/98, 5/98, 6/98, 1/99 Philippines 12/97 Souh Korea 12/97, 1/98 Malaysia 1/98 Souh Africa 7/98 Russia 8/98, 10/98, 12/98, 1/99 Brazil 1/99 a Major devaluaions are episodes when he counry s local currency/u.s. dollar exchange rae increased by 15% or more wihin a 4-week period. Afer a devaluaion even, he nex 4 weeks are excluded so ha here can be, a mos, one devaluaion even wihin any 4-week period. Exchange rae daa is from Daasream. b Counries in he sample which do no have any major devaluaion evens are: Argenina, Ausralia, Ausria, Bahrain, Bangladesh, Belgium, Briain, Canada, Chile, China, Colombia, Czech Republic, Denmark, Finland, France, Germany, Greece, Hong Kong, Hungary, India, Ireland, Israel, Ialy, Japan, Jordan, Kuwai, Lebanon, Luxembourg, Mala, Mexico, Morocco, Neherlands, New Zealand, Norway, Poland, Porugal, Saudi Arabia, Singapore, Slovakia, Spain, Sri Lanka, Sweden, Swizerland, Taiwan, Turkey, Unied Arab Emiraes, Uruguay and Venezuela. Ecuador is he one counry ha has a major devaluaion bu is no included in he lis since firmlevel daa is no available for he counry and Ecuador expors less han 1% of global expors for each of he commodiies sudied. The empirical ess focus on 10 commodiy groups: naural rubber and relaed fores producs; mining for silver and gold ores; naural gas and crude peroleum; preserved fruis and vegeables; edible oils and fas; cigarees; indusrial inorganic chemicals; plasics, maerials and synheics; indusrial organic chemicals; and ferilizers. 18 The firm-level informaion includes daa for over 1100 firms and is from he Worldscope daabase published by Disclosure (2001). Table 3 shows he disribuion of firms (and average firm size) by counry and region. Alhough he sample includes informaion from 51 counries and 9 regions, coverage of many developing economies is exremely limied. Table 4 provides addiional informaion on he disribuion of firms by commodiy group in he devaluing counries, as well as corresponding hree-digi SIC codes. The remainder of his secion uses he 10 commodiy groups lised in Table 4 o examine he impac of he devaluaion evens lised in Table 2 on oupu, profis, invesmen and sock reurns in firms around he world. I focuses on he model s predicions for firms, raher han indusries, because global producion and prices are affeced by numerous shocks o supply and demand (oher han devaluaions) ha are exremely difficul o measure. For example, an unusually cold winer in he Norhern hemisphere can increase demand for naural gas and oil; a severe monsoon in Asia could decrease he global supply of edible oils; and echnological advances (such as fiberopics) can reduce he demand for specific minerals (such as copper). By focusing on wihin- 18 These commodiy groups were chosen based on hree crieria: (1) firm-level daa exised for a leas 20 companies in he hree-digi indusry; (2) a leas 5% of he firms in he indusry were locaed in devaluing counries; (3) he indusry roughly fis he characerisics of a commodiy as described in he model. Several of he resuling indusry groups are no ypical commodiies, bu as shown in he sensiiviy ess, resuls are robus o excluding differen commodiies.

14 348 K.J. Forbes / Journal of Developmen Economics 69 (2002) Table 3 Firm sample informaion Region Counry No. of firms Average asse value a Africa Egyp 2 106,735 Morocco 1 884,222 Souh Africa ,165 Zimbabwe 1 30,146 Ausralasia Ausralia ,031 New Zealand 3 893,242 Eas Asia Hong Kong 5 347,789 Japan 140 1,799,298 Korea 43 1,094,628 Malaysia ,508 Singapore ,846 Taiwan ,444 Easern Europe Czech Republic 9 221,110 Hungary 2 365,357 Poland 3 229,511 Russia 10 3,802,581 Lain America Argenina 9 2,529,668 Brazil 26 2,700,113 Chile 4 584,754 Colombia 2 62,207 Mexico 8 1,403,361 Peru 1 389,694 Venezuela 2 11,820,064 Middle Eas Israel 6 1,177,402 Jordan 1 481,423 Turkey 5 75,959 Norh America Canada ,109 Unied Saes 151 3,940,778 Souh and Wes Asia China 9 1,138,498 India ,866 Indonesia ,264 Pakisan 20 71,352 Philippines ,038 Thailand ,326 Wesern Europe Ausria 3 408,174 Belgium 8 3,192,222 Denmark 5 497,253 Finland 3 1,376,643 France 15 3,593,548 Germany 13 9,909,534 Greece 15 99,360 Ireland 7 139,395 Ialy 10 8,175,550 Luxembourg 1 8,643,201 Neherlands 7 2,933,540 Norway 8 3,812,535 Porugal 4 91,514 Spain 7 1,046,979

15 K.J. Forbes / Journal of Developmen Economics 69 (2002) Table 3 (coninued) Region Counry No. of firms Average asse value a Wesern Europe Sweden 5 1,368,383 Swizerland 6 3,371,978 Unied Kingdom 63 2,782,784 Toal sample ,777,259 a Asses measured in housands of U.S. dollars in 1996 (or he closes year available). indusry differences in firm performance, raher han on aggregae indusry rends, i is possible o conrol for hese exogenous shocks o global supply and demand and beer idenify he direc impac of devaluaions on firms around he world Tes of predicion 1: he shor-run impac of devaluaions on firm oupu Predicion 1: Immediaely afer devaluaions, commodiy-exporing firms in he devaluing counry increase oupu and compeing firms in oher counries decrease oupu. To es his predicion, Table 5 begins by lising average oupu growh for firms in counries ha devalued (in ha year) compared o compeiors in counries ha did no devalue. Oupu growh is measured as he annual percen change in ne sales and revenues (measured in local currency) for each year from 1996 hrough The firs row of he able repors mean oupu growh for he enire sample. The lower rows disaggregae average oupu growh ino he 10 commodiy groups in Table 4. The able also repors sandard deviaions of oupu growh (in parenheses) and z-saisics from a Wilcoxon Rank Sum Tes of he null hypohesis ha he wo independen samples of firms are from populaions wih he same disribuion. Table 5 shows ha average annual oupu growh for firms in devaluing counries was 21%, while oupu growh for firms in nondevaluing counries was 8%. When oupu growh is disaggregaed by commodiy, firms in devaluing counries have higher oupu growh in all 10 indusries. For example, in he edible oils and fas indusry, oupu growh for firms in devaluing counries averaged 16%, while oupu growh for firms in nondevaluing counries averaged only 4%. Moreover, he z-saisics are significan (a he 10% level) for he enire sample of firms and for 9 of he 10 commodiy groups. These ess indicae ha oupu growh raes for firms in devaluing counries are significanly higher han for firms in nondevaluing counries. Nex, in order o formalize his analysis and conrol for inflaion, annual shocks o oupu, and any forward and/or lagged impac of devaluaions, I esimae he model: Dq i; ¼ h 0 þ h 1 Devalue i; 1 þ h 2 Devalue i; þ h 3 Devalue i;þ1 þ h 4 Inflaion i; þ g þ e i; ð11þ where Dq i, is oupu growh for firm i in period ; Devalue i, 1 is a dummy variable equal o 1 if he counry where firm i is locaed will have a devaluaion in he 19 More specifically, ne sales and revenues are defined as gross sales and oher operaing revenues less discouns, reurns and allowances.

16 Table 4 Sample informaion for devaluing counries Indusry and SIC code Naural rubber and gums/ fores producs 083 No. firms Percen of oal Mining for silver and gold ores 104 No. firms Percen of oal Naural gas and crude peroleum 131 No. firms Percen of oal Preserved fruis and vegeables 203 No. firms Percen of oal Edible oils and fas 207 Brazil Indonesia Korea Malaysia Philippines Russia Souh Africa Thailand Toal crisis firms Toal firms in sample Indusry SIC code Cigarees 211 No. of firms Percen oal Indusrial inorganic chemicals 281 No. of firms Percen oal Plasics, maerials and synheics 282 No. of firms Percen oal Indusrial organic chemicals 286 No. of firms Percen oal No. firms Percen of oal Ferilizer/agriculural chemicals 287 Brazil Indonesia Korea Malaysia Philippines Russia Souh Africa Thailand Toal crisis firms Toal firms in sample Crisis firms are firms in counries wih a major devaluaion even beween 1996 and 2000 as defined in Table 2. No. of firms Percen oal 350 K.J. Forbes / Journal of Developmen Economics 69 (2002)

17 K.J. Forbes / Journal of Developmen Economics 69 (2002) Table 5 Trends in firm oupu growh a : devaluing counries versus res of world Firms in devaluing counries b Firms in res of world z-saisic for difference in oupu growh c Full sample Mean oupu growh ** Sandard deviaion (0.50) (0.46) Rubber planaions Mean oupu growh ** and fores producs Sandard deviaion (0.30) (0.34) Silver and gold ores Mean oupu growh Sandard deviaion (1.00) (0.76) Naural gas and Mean oupu growh ** crude peroleum Sandard deviaion (0.96) (0.66) Preserved fruis Mean oupu growh * and vegeables Sandard deviaion (0.44) (0.21) Edible oils and fas Mean oupu growh ** Sandard deviaion (0.40) (0.24) Cigarees Mean oupu growh ** Sandard deviaion (0.20) (0.19) Indusrial inorganic Mean oupu growh ** chemicals Sandard deviaion (0.13) (0.17) Plasics, maerials Mean oupu growh ** and synheics Sandard deviaion (0.34) (0.35) Indusrial organic Mean oupu growh ** chemicals Sandard deviaion (0.34) (0.24) Ferilizer Mean oupu growh ** Sandard deviaion (0.23) (0.32) a Oupu growh measured as percen change in ne sales and revenues measured in local currency. b Devaluing counries are counries wih a major devaluaion (as defined in Table 2) in he curren year. c z-saisic from a wo-sample Wilcoxon Rank Sum Tes if he wo groups of firms are from populaions wih he same disribuion. * Significance a he 10% level. ** Significance a he 5% level. nex period; Devalue i, is a dummy variable equal o 1 if he counry where firm i is locaed devalued in period ; Devalue i, +1 is a dummy variable equal o 1 if he counry where firm i is locaed devalued in he previous period; Inflaion i, is he inflaion rae for he counry where firm i is locaed in period ; g is a vecor of period dummy variables (for 1996 hrough 1999, wih 2000 he excluded year); and e i, is an error erm. The Devalue i, 1 variable capures wheher firms had higher or lower oupu growh in he year before a devaluaion, while Devalue i, and Devalue i, +1 capure wheher firms had significanly differen oupu growh in he year of he devaluaion or year immediaely following. The inflaion variable capures any impac of changes in he price level on repored oupu growh. 20 The period dummy variables capure any global shifs in supply or demand ha affec all firms in he sample in any period. 20 Inflaion is measured as he annual percen change in he consumer price index as repored in line 64..xzf of Inernaional Moneary Fund (2001).

18 352 K.J. Forbes / Journal of Developmen Economics 69 (2002) Table 6 Regression resuls: growh in firm oupu Devaluaion dummies Inflaion No. of R Pooled cross-secion 0.106** 0.109** (0.054) (0.038) (0.078) Fixed indusry effecs 0.082** 0.133** (0.040) (0.037) (0.051) Random indusry effecs 0.106** 0.109** (0.040) (0.036) (0.051) Rubber planaions 0.108* 0.266** and fores producs (0.065) (0.095) (0.317) Silver and gold ores (0.135) (0.272) (0.272) Naural gas and crude 0.950* peroleum (0.503) (0.296) (0.594) Preserved fruis ** and vegeables (0.074) (0.121) (0.081) Edible oils and fas 0.131** 0.129* (0.043) (0.077) (0.051) Cigarees ** (0.047) (0.060) (0.042) Indusrial inorganic 0.098** chemicals (0.034) (0.038) (0.066) Plasics, maerials and synheics (0.046) (0.054) (0.042) Indusrial organic 0.089* 0.150** chemicals (0.054) (0.050) (0.059) Ferilizer * (0.049) (0.063) (0.130) 0.005** 0.005** 0.005** 0.009** (0.002) (0.008) (0.003) 0.007** 0.008** 0.006** (0.002) (0.003) 0.005** 0.009** (0.002) 0.006** observaions Sandard errors are in parenheses and are Whie-adjused for heeroscedasiciy. R 2 is he wihin-r 2 for he fixedeffecs esimaes and he overall R 2 for he random-effecs esimaes. Period dummy variables are included in each specificaion and are always joinly significan a he 1% level. * Significance a he 10% level. ** Significance a he 5% level. Table 6 repors esimaes of Eq. (11). The firs row shows resuls when he model is esimaed as a pooled cross-secion of firms in all 10 commodiy groups. The second and hird rows repor resuls when he model is esimaed wih fixed or random indusry effecs. The boom par of he able repors esimaes when Eq. (11) is esimaed separaely for each of he commodiies. When he model is esimaed for he enire sample, Devalue i, is always posiive and significan a he 1% level. This suggess ha when a counry devalues is currency, domesic firms have significanly higher oupu growh (in ha year) han firms in nondevaluing counries. The size of his effec can be large. The pooled cross-secion esimaes indicae ha annual oupu growh is 11% greaer, on average, for firms in he devaluing counry. Moreover, when Eq. (11) is

19 K.J. Forbes / Journal of Developmen Economics 69 (2002) esimaed separaely for each of he 10 commodiy groups, he coefficiens on Devalue i, remain posiive for each of he indusries (and are significan a he 10% level in 40% of he indusries). This suggess ha wihin mos indusries, firms in devaluing counries had higher oupu growh han compeiors in nondevaluing counries. For example, esimaes for he firs commodiy group indicae ha oupu growh for rubber planaions was 27% greaer in devaluing counries han for compeiors locaed elsewhere. When he model is esimaed for he enire sample, he coefficiens on he forward and lagged devaluaion variables are always negaive. The esimaes in he lower secion of he able, however, show ha he sign of hese esimaes flucuaes across indusries. Moreover, coefficien esimaes for he naural gas and crude peroleum group are such large negaive ouliers ha hey could be driving he large negaive coefficiens for he enire sample. 21 To es his, he firs row of Table 7 repeas he base resuls (using random effecs) from he op of Table 6 and hen row 2 excludes firms in he naural gas and crude peroleum indusry. The coefficiens on Devalue i, 1 and Devalue i, +1 are now posiive and insignifican, while he coefficien on Devalue i, remains posiive and highly significan. Nex, o examine if any oher indusry groups affec hese esimaes, I reesimae he model excluding one commodiy group a ime (boh wih and wihou he naural gas and crude peroleum indusry). In each case, he coefficien on Devalue i, is posiive and highly significan, and he signs and significance of he coefficiens on Devalue i, 1 and Devalue i, + 1 depend on he indusries included in he analysis. The remainder of Table 7 repors a sample of addiional sensiiviy ess. Row 3 adds a conrol variable for firm size and row 4 measures inflaion using an index of producer prices insead of consumer prices. Rows 5 and 6 include regional and counry dummy variables, respecively. Row 7 excludes OECD counries. Rows 8 and 9 repor resuls for only middle- or low-income counries, as defined in World Bank (2001). Row 10 only includes Asia. Row 11 repors resuls when he equaion for oupu growh is joinly esimaed wih he equaion for profi growh (discussed in Secion 4.3) using a seemingly unrelaed regressions model in order o conrol for any correlaion in he error erms across models. The op of Table 8 repors resuls when he devaluaion dummies are ineraced wih he size of he exchange-rae movemen in he given period (including squared erms o capure any nonlineariies). These resuls sugges ha he impac of devaluaions on sales growh is nonlinear and increases (a a decreasing rae) wih he size of he depreciaion. I also examine he impac of including conrols for ineres raes, as well as excluding each counry, each region, each devaluaion even, he inflaion variable, he period dummy variables, and/or he forward and lagged effec of devaluaions. 21 The naural gas and crude peroleum indusry is frequenly an oulier in he range of ess repored below. This could parially reflec he unique aspecs of his indusry (such as he power of OPEC and he concenraion of supply in a urbulen region). This may also reflec he unusual price volailiy during 1997 and Rapid increases in producion, an increase in OPEC quoas, a mild winer and hen an unexpeced decrease in global growh, all generaed an unusually rapid decline in oil prices during he period when mos of he counries in he sample experienced major devaluaions.

20 354 K.J. Forbes / Journal of Developmen Economics 69 (2002) Table 7 Sensiiviy analysis: growh in firm oupu Devaluaion dummies Inflaion No. of R observaions Base esimaes 0.106** 0.109** ** (random effecs) (0.040) (0.036) (0.051) Exclude naural gas ** ** and crude peroleum a (0.035) (0.032) (0.045) Conrol for firm size b 0.108** 0.108** ** (0.040) (0.036) (0.051) Inflaion based on 0.117** 0.142** ** producer-price index c (0.041) (0.036) (0.051) (0.000) Include regional 0.088** 0.136** ** dummy variables d (0.042) (0.038) (0.053) Include counry 0.116* 0.107* ** dummy variables (0.061) (0.066) (0.072) (0.002) Only non-oecd ** ** counries (0.052) (0.051) (0.070) Only middle-income 0.102* 0.108* 0.154* 0.004** counries e (0.060) (0.062) (0.087) Only low-income ** 0.004* counries e (0.138) (0.095) (0.002) Only Asia f 0.074* 0.115** 0.090* 0.005** (0.042) (0.036) (0.050) SUR join esimaion wih profi growh g 0.056* (0.033) 0.096** (0.030) (0.045) 0.005** Sandard errors are in parenheses and are Whie-adjused for heeroscedasiciy. All esimaes are random (indusry) effecs. R 2 is he overall R 2. Period dummy variables are included in each specificaion. a Excludes naural gas and crude peroleum (SIC group 131). b Firm size measured as oal asses in U.S. dollars. c Producer price index based on line 63..zf in Inernaional Moneary Fund (2001). Monhly daa corresponds o company s reporing period. d Regional dummy variables are based on he nine regions lised in Table 3. e Definiions of middle- and low-income counries are based on classificaions in World Bank (2001). f Includes Eas Asia and Souh Wes Asia, as defined in Table 3. g Resuls from a seemingly unrelaed regressions model including boh oupu growh and profi growh. Coefficien esimaes for he profi growh equaion are repored in Table 11. * Significance a he 10% level. ** Significance a he 5% level. In each of hese sensiiviy ess, he cenral resuls do no change. In he year of devaluaions, crisis-counry firms have oupu growh raes abou 10 20% higher han compeiors in oher counries. On he oher hand, he negaive coefficiens on he forward and lagged devaluaion variables are no robus. This suggess ha firm oupu growh in he years preceding and following devaluaions is no significanly differen. Therefore, devaluaions appear o generae a one-ime increase in firm oupu growh during he year of he devaluaion and, herefore, a permanen increase in firm oupu levels.