A Note on Carry Trade and the Related Financial Variables

Transcription

1 Inernaional Journal of Economics and Finance Vol. 3, No. 3; Augus A Noe on Carry Trade and he Relaed Financial Variables Takvor H. Muafoglu Deparmen of Economics, Huner College, CUNY 95 Park Avenue, New York, NY 5 Tel: muafog@huner.cuny.edu Received: Ocober 5, Acceped: November, doi:.5539/ijef.v3n3p9 Absrac A recen sudy, employing srucural vecor auoregression (SVAR) mehodology, finds ha he U.S. sock marke performance has a posiive impac on he direcion of he speculaive yen carry rade aciviy using monhly posiioning daa of non-commercial raders in currency fuures. However, I illusrae ha his finding is no robus when weekly posiioning daa is inroduced o he same mehodology and sample period. Insead, I find ha i is he flucuaions in he Japanese yen agains he U.S. dollar exchange rae, raher han he U.S. sock marke performance, ha deermines he direcion of he yen carry rade. This represens ha conclusions drawn from emporally aggregaed daa should be inerpreed cauiously since emporal aggregaion loses informaion abou he underlying daa processes. Keywords: Carry rade, Exchange raes, SVAR, Granger causaliy. Inroducion A recen aricle by Nishigaki (7) invesigaes he relaionship beween he yen carry rade aciviy and he relaed financial variables in he U.S. and Japan. Carry rade is commonly known as he pracice of borrowing money in a currency wih low ineres raes funding currency in order o inves in a currency wih higher ineres raes arge currency. The dynamics of his sraegy is relaively sraighforward. An invesor iniiaes by borrowing a given amoun in a low-ineres-rae currency, hen convers he funds ino a high-ineres-rae currency, and finally lends he resuling amoun in he arge currency a he high ineres rae. The yen carry rade has been he focus of he financial markes since he official lending raes has been less han percen for over a decade in Japan. Therefore, invesors have been reaing he yen as a popular funding currency. Nishigaki uses he ne posiions of non-commercial raders in he Japanese currency fuures as a measure for he yen carry rade aciviy. Alhough, hese posiions are available a weekly frequency from he U.S. Commodiy Fuures Trading Commission, Nishigaki chooses o aggregae hese weekly posiions and use monhly observaions in his analysis. His resuls are hreefold. Firs, he discovers ha he ineres rae differenial beween Japan and he U.S. has an insignifican impac on he carry rade aciviy. Second, he finds ha he U.S. sock marke performance has a posiive impac on he direcion of he yen carry rade aciviy. And lasly, he observes ha he speculaive carry rade has a significan impac on he nominal exchange rae. Alhough, his findings are noeworhy, hey should be aken wih some cauion. In his noe, I illusrae ha while he empirical conclusions drawn from Nishigaki (7) can be successfully replicaed using monhly daa, cerain findings fail he robusness es when weekly frequency is inroduced o he idenical model he employs. The organizaion of he paper is as follows; he nex secion presens he daa and mehodology used in Nishigaki (7), empirical resuls are discussed in Secion 3, and concluding remarks are provided in Secion.. Daa and Mehodology In order o beer undersand he effec of financial facors on he speculaive yen carry rade aciviy beween he U.S. and Japan, Nishigaki (7) employs he following variables; he ineres rae differenial beween he U.S. and Japan (IRD), he global volailiy indicaor (VIX), he dollar agains he yen exchange rae (FER), he U.S. sock price (SUS), and he Japanese sock price (SJP). The abbreviaions used here are he same as Nishigaki s. The U.S. ineres rae is he Federal Funds Targe Rae whereas he Japanese ineres rae is he Bank of Japan arge policy rae which is also known as he Uncollaeralized Overnigh Call Rae. VIX is he Chicago Board Opions Exchange (CBOE) Volailiy index. I is an opion implied volailiy index conveyed by he S&P 5 sock index opion prices and is known as he fear index among he marke paricipans. SUS is he S&P 5 index whereas SJP is he Tokyo Sock Price Index which is also commonly known as he TOPIX. Carry rade aciviy is measured by he ne open posiions of non-commercial raders in he yen currency fuures marke raded on he Chicago Mercanile Exchange (CME). I is abbreviaed as IMR and measured as he raio of long o shor conracs in yen currency Published by Canadian Cener of Science and Educaion 9

2 Inernaional Journal of Economics and Finance Vol. 3, No.3; Augus fuures. Nishigaki (7) employs monhly observaions of he above menioned variables from January 993 o January 7. We acquire all daa from Thomson Daasream excep IMR which is calculaed from he Commimen of Traders (COT) repor provided by he U.S. Commodiy Fuures Trading Commission (CFTC). The mehodology used in Nishigaki (7) is srucural vecor auoregression (SVAR) model. An SVAR represenaion of he relaionship beween he six variables can be wrien as: BX BLX ( ) () where X he column vecor ( IRD, VIX, IMR, FER, SUS, SJP) B a marix of coefficiens, reflecing conemporaneous relaionships among he six variables IRD VIX IMR FER SUS SJP he column vecor of srucural error erms (,,,,, ) BL ( ) a marix wih elemens equal o he polynomials Bij ( L ) L a lag operaor Pre-muliplying all erms by B, we obain he reduced-form VAR represenaion: X ALX ( ) e () where A( L) Bo B( L) and e B Noice ha B represens he marix of conemporaneous responses of X o he srucural shocks, e B so Be o. The SVAR uilizes he following expression for he variance-covariance marix of he reduced-form VAR residuals, denoed by, in order o idenify he elemens of B. ' ' e B Eee B E( )( B ) B ( B )' (3) The oal number of parameers o be esimaed is 93. The righ-hand side of equaion (3) has 7 unknown parameers. However, 57 conemporaneous resricions are provided by he assumed orhogonaliy of he shocks, normalizaion of he diagonal elemens of B o equal uniy, combined wih he esimaed variance-covariance marix of he reduced-form VAR ( e ), leaves 5 resricions for exac idenificaion of he unknown parameers. Therefore, o idenify he srucural shocks from he reduced-form shocks, a number of resricions have been imposed.. Shocks o oher variables in he sysem have no effecs on he ineres rae differenial beween he US and Japan. Thus, IRD is o be he mos exogenous variable in he sysem. This assumpion leads o he resricions b ( L) b3( L) b ( L) b5 ( L) b ( L).. Chicago Board Opions Exchange (CBOE) Volailiy Index, VIX, is assumed o be affeced only by shocks o IRD. This resricion is incorporaed as b3( L) b ( L) b5( L) b ( L). 3. Currency carry rade proxy IMR is influenced by shocks o IRD, and VIX, which yields he resricions b3 ( L) b35( L) b3 ( L).. S&P 5 sock index, SUS, and Tokyo Sock Price Index, SJP, have no conemporaneous effec on he dollar agains yen exchange rae, FER. This assumpion is illusraed as b5( L) b ( L). 5. SUS is assumed o be affeced by shocks o IRD, VIX, IMR, and FER. This resricion is inroduced as b ( L). 5. Shocks o all oher variables are assumed o affec Tokyo Sock Price Index, SJP, and hence i is deermined endogenously in he sysem. Wih he above menioned 5 resricions, he sysem is exacly idenified. The sysem of equaions arising from hese resricions can be expressed as follows: IRD b u VIX b u b u IMR b u b u b u FER b u b u b u b u SUS b u b u b u b u b u SJP b u b u b u b u 5 bu IRD IRD VIX IRD VIX IMR IRD VIX IMR FER 3 IRD VIX IMR FER SUS IRD VIX IMR FER SUS 3 b u SJP 9 ISSN 9-97X E-ISSN 9-97

3 Inernaional Journal of Economics and Finance Vol. 3, No. 3; Augus Alhough, a deailed heoreical reasoning of he above deemed ordering is no included here, i could be found a Nishigaki (7). Boh he Augmened Dickey-Fuller (ADF) and Phillip-Perron (PP) ess indicae ha all he variables are inegraed of order excep IMR. Therefore, all he variables are inroduced o he sysem in firs differences, denoed by, excep for IMR. Also, all he variables, excep IRD, are enered ino naural logarihms. 3. Empirical Resuls Our aemp o replicae Nishigaki (7) findings based on monhly observaions is me wih success. Through impulse response funcions, variance decomposiions and VAR Granger causaliy es, we were able o observe his key conclusions such as he insignifican effec of he ineres rae differenial beween he U.S. and Japan on he yen carry rade, he dominan and posiive impac of he U.S. sock marke performance on he carry rade aciviy, and finally he significan impac of he speculaive yen carry rade on he nominal exchange rae. We se ou o es he robusness of hese resuls by changing he frequency of he daa used in he SVAR represenaion. The COT repor is calculaed from Wednesday o Tuesday and released o he public he following Friday by he CFTC. Since ne posiions now represen he oucome of weekly adjusmens of rading sraegies by he raders, he res of he variables in he sysem are averaged over he Wednesday-Tuesday inerval o mach he COT daa. Alhough, he opion o expand he sample period exiss, we chose no o do so due o he recenly experienced financial crisis which had creaed significan insabiliy in he markes. Therefore, he sample period remains from January 993 o January 7. Furhermore, we did no aler he above menioned ordering in he SVAR sysem. Once again all he variables are esed for uni roo using he ADF and PP ess and he resuls are repored a Table. As before, all he variables are found o be I() a he percen significance level excep for IMR which is inegraed of I(). Therefore, all he variables are firs differenced excep he IMR. Firs, we sudy he impulse response funcions. Figure displays he response of IMR o a one-sandard deviaion innovaion of a paricular shock on all he variables over a -week period range and conains ± sandard error bands. Our resuls indicae ha he only variable which has a significan impac on he behavior of he speculaive carry rade aciviy is he nominal exchange rae. Specifically, we observe ha when he Japanese yen depreciaes agains he U.S. dollar, he yen carry rade raises. None of he oher variables seem o have a significan effec on he speculaive yen carry rade aciviy. Noe ha while Nishigaki (7) observes ha a posiive shock in he S&P 5 index leads o an increase in he yen carry rade aciviy, here we observe he opposie: ha is, when a posiive shock is inroduced o he U.S. sock price, yen carry rade aciviy decreases. Therefore, we fail o validae his eminen finding. Neverheless, when we look a Figure, which reveals he impulse responses of all he variables o IMR shocks, we sill do see he saisically significan impac of a sudden reversal of carry rade posiions on he nominal dollar agains he yen exchange rae. We observe ha he Japanese yen will appreciae agains he U.S. dollar when invesors unwind heir posiions. This finding is consisen wih Nishigaki s. Table repors he variance decomposiions. The variance decomposiion of IMR sugges ha shocks o he dollar agains he yen exchange rae explain more han percen of he variance in IMR six weeks afer a shock and in fac he explanaory power of shocks o FER increases o 5.3 percen afer weeks. On he oher hand, evidence suggess ha he U.S. sock price has no impac on he variance of he yen carry rade aciviy. However, we coninue o see he abiliy of invesors carry rade posiions o influence he dollar agains he yen exchange rae. Table displays ha a shock o IMR explains abou 5 percen of he variance in FER. We perform a mulivariae version of he Granger causaliy es because Granger causaliy es is of limied usefulness in a VAR sysem. This es is also known as a block-exogeneiy es. Block-exogeneiy ess generalize Granger causaliy, indicaing wheher he lagged independen variables joinly affec a paricular dependen variable. Table 3 provides he VAR Granger causaliy es resuls. The null hypohesis is ha he coefficiens are zero, no Granger causaliy, and a significanly low p-value indicaes a rejecion of he null. According o he able, we observe ha only he changes in he nominal dollar agains he yen exchange rae, FER, significanly Granger-causes he non-commercial posiions, IMR. None of he oher variables have any causal effec on he speculaive yen carry rade aciviy. This finding is in oal conradicion wih he findings of Nishigaki (7). I illusraes ha he mos imporan variable o influence he direcion of he carry rade aciviy is he exchange rae flucuaions in he marke.. Conclusion In his noe, we esed he robusness of he resuls obained by Nishigaki (7). We conclude ha his key finding, which is he dominan impac of he U.S. sock marke performance on he direcion of he yen carry rade aciviy, canno be observed when weekly, raher han monhly, observaions are used. Insead, we observe he significan Published by Canadian Cener of Science and Educaion 93

4 Inernaional Journal of Economics and Finance Vol. 3, No.3; Augus impac of he nominal dollar agains he yen exchange rae on he speculaive yen carry rade aciviy. We find ha when he Japanese yen depreciaes agains he U.S. dollar, he yen carry rade raises. The oucome of he Granger causaliy es confirms our findings. Perhaps, one reason why Nishigaki s finding is no robus o using higher frequency daa has o do wih emporal aggregaion. I is known ha aggregaing daa from higher o lower frequencies, as Nishigaki implicily does, can imply some risks. For insance, Rossana and Seaer (995) and Marcellino (999) have empirically shown ha emporal aggregaion causes a severe loss of informaion abou he ime series processes driving many economic variables. Therefore, one mus be cauious while inerpreing resuls based on emporally aggregaed daa since emporal aggregaion loses informaion abou he underlying daa processes. References Marcellion, M. (999).Some Consequences of Temporal Aggregaion in Empirical Analysis.Journal of Business and Economic Saisics 7, No., 9-3. Nishigaki, H. (7). Relaionship beween he Yen Carry Trade and he Relaed Financial Variables. Economics Bullein 3, No., -7. Rossana, R. and J. Seaer. (995). Temporal Aggregaion and Economic Time Series.Journal of Business and Economic Saisics 3, No., -5. Table. Uni Roo Tess ADF PP ADF PP Variable Inercep Inercep and Trend IRD Firs Difference -7.9*** -.*** -7.*** -.39*** ln(vix) -3.** -.99** Firs Difference -3.*** -.3*** -3.*** -.9*** ln(imr) -.9*** -.*** -.3*** -.3*** Firs Difference -.*** -3.*** -.39*** -3.3*** ln(fer) Firs Difference -.53*** -.53*** -.5*** -.5*** ln(sus) Firs Difference -3.*** -3.*** -3.*** -3.*** ln(sjp) Firs Difference -7.3*** -7.3*** -7.9*** -7.9*** Noe: The lags in he ess were esimaed hrough he Schwarz Informaion Crierion. ***H o of a uni roo is rejeced a he %, **5%, and *% level.ird is he ineres rae differenial beween he U.S. and Japan, VIX is he CBOE volailiy index, IMR is he raio of long o shor posiions in JPY fuures, FER is he Japanese yen per U.S. dollar, SUS is he S&P 5 sock index, and SJP is Tokyo sock price index. 9 ISSN 9-97X E-ISSN 9-97

5 Inernaional Journal of Economics and Finance Vol. 3, No. 3; Augus Table. Variance Decomposiions Explained Horizon Variable (weeks) D(IRD) D(VIX) IMR D(FER) D(SUS) D(SJP) D(IRD) D(VIX) IMR D(FER) D(SUS) D(SJP) Noe: The able repors he forecas error variance decomposiions which indicae he amoun of informaion each column variable conribue o he row variables. Table 3. VAR Granger Causaliy Tes (P-values) D(IRD) D(VIX) IMR D(FER) D(SUS) D(SJP) D(IRD) D(VIX) IMR D(FER) D(SUS) D(SJP) Noe: Table shows he marginal probabiliies associaed wih he Granger-causaliy es. The forma is such ha he rows reflec he Granger-causal impac of he row-variable on he column-variable. Published by Canadian Cener of Science and Educaion 95

6 Inernaional Journal of Economics and Finance Vol. 3, No.3; Augus Accumulaed Response o Srucual One S.D. Innovaions ± S.E. Accumulaed Response of IMR o D(IRD) Accumulaed Response of IMR o D(VIX) Accumulaed Response of IMR o IMR Accumulaed Response of IMR o D(FER) Accumulaed Response of IMR o D(SUS) Accumulaed Response of IMR o D(SJP) Figure. Impulse Responses of IMR Accumulaed Response o Srucural One S.D. Innovaions ± S.E. Accumulaed Response of D(IRD) o IMR Accumulaed Response of D(VIX) o IMR Accumulaed Response of IMR o IMR Accumulaed Response of D(FER) o IMR Accumulaed Response of D(SUS) o IMR Accumulaed Response of D(SJP) o IMR Figure. Impulse Responses o IMR 9 ISSN 9-97X E-ISSN 9-97