The Thursday Effect of the Forward Premium Puzzle

Transcription

1 The Thursday Effec of he Forward Premium Pule This draf: February 8 Liang Ding Deparmen of Economics, Macaleser College, S.Paul, M, U.S.A Absrac: This paper examines he forward premium pule based on -week forward raes across weekdays. The paper finds ha Thursday consisenly appears o be a special day on which he pule disappears, while i is presen on oher weekdays. In addiion o Thursday, Monday is also found o be a similar special day for he Euro. The paper proposes ha his Thursday effec Monday effec for Euro is caused by moneary announcemens released consanly on Tuesday Thursday by he FOMC he European Cenral Bank. The empirical ess provide convincing evidence in favor of he proposed explanaion. Keywords: Forward premium pule; Day-of-he-week-effec; Cenral Bank; Moneary announcemens JEL Classificaion: F3; G4; G Corresponding auhor. Tel.: ; Fax: address: ding@macaleser.edu

2 . Inroducion One of he mos exensively examined opics in inernaional finance has been he efficiency of forward marke for foreign exchange. From early sudies such as Fama 984, o more recen ones such as Meredih and Ma, an enormous number of sudies have shown he failure of forward exchange rae o serve as an unbiased predicor of fuure spo rae. In lieraure, he robusness of he pule has been esed in muliple aspecs, including differen ime periods e.g. Zhou and Kuan, differen counries e.g. Bansal and Dahlquis, differen mauriies e.g. Chinn and Meredih 4, ec. This paper inspecs he pule from a new angle differen weekdays. I is noiceable ha he sudies using forward raes wih -week mauriy usually ake he observaions on a paricular weekday as he represenaive of ha week, bu no papers seem o concern if differen weekday would generae inconsisen resuls. This paper adds o he lieraure by examining he day-of-he-week effec in he relaionship beween curren forward rae and fuure spo rae. The paper firs ess he forward premium pule by using -week forward raes across weekdays. The resuls are ineresing ye inriguing: here does exis an inconsisency across weekdays. Specifically, for all currencies, Thursday consisenly appears o be a special day on which he pule disappears, while i is presen on oher weekdays. In addiion, for Euro, Monday is also found o be a similar special day. These empirical findings, which have no been repored in lieraure before, can be called he Thursday effec of forward premium pule. The furher robus es shows ha he effec is significan only when he span of sample period is long enough. The paper aribues he effec o he moneary policy announcemens made by he Federal Open Marke Commiee he similar moneary policy commiee in he European Cenral Bank consanly on Tuesdays Thursdays. The paper argues ha he marke expecaion abou fuure spo rae is dominanly affeced by such fundamenal informaion so ha expecaion error can be The paper will show ha Monday effec of Euro is acually he European version of he Thursday Effec.

3 reduced and he fuure spo rae follows he relaionship implied by he heory in he week when he informaion is released. Due o he delivery arrangemen ha he orders will be seled wo business days afer he ransacion dae in boh he money marke and foreign exchange marke, when he informaion is released on Tuesday Thursday in he European case, he arbirage ransacions conduced by he invesors who wan o speculae on he informaion will be compleed wo business days laer, i.e. Thursday Monday for Euro. Thus he relaionship beween fuure spo rae and curren forward rae on Thursdays Monday for Euro is closer o he scenario described by he heory, and he pule seems o disappear on Thursdays and also Monday for Euro. Bu since he FOMC does no mee every week abou eigh imes a year and once one monh in he European case, he effec emerges only when he sample period is long enough. The empirical ess provide evidence supporing such an explanaion. The conribuions of his paper are wofold. Firs, i provides new empirical evidence ha deepens he forward premium pule. This evidence shows ha he forward premium pule depends on he day of he week. Second, he paper proposes an explanaion ha has solid inuiive foundaion. As summaried in Engle 996, he explanaions for he pule basically can be divided ino wo caegories -- ime varying risk premium and irraional expecaions. And he findings in his paper end o suppor he second caegory. The paper is also relaed o he lieraure of exchange rae dynamics. A lo of sudies show ha here is a lack of connecion beween spo rae and macroeconomic variables, bu as shown in Simpson e. al, he spo rae does respond o fundamenal shocks such as he announcemens of ineres rae. The findings in he paper are consisen wih such sylied facs. The res of he paper is srucured as below: secion repors he new empirical evidences abou he forward premium pule, secion 3 proposes an explanaion, secion 4 ess he enaive explanaion and secion concludes. 3

4 . Empirical findings The daa employed in his paper were exraced from DaaSream and conain dollar raes for several major currencies, including he Canadian Dollar CAD, Euro EUR and Japanese Yen JPY. Boh spo and -week forward exchange raes were colleced on a daily basis for each of hese currencies. The CAD and JPY daa cover he period beween /7/997 and //4, and he EUR daa sar from /4/999 and end a //4. All weekend raes were excluded. and Suppose s τ, s are he logarihms of he spo raes a ime τ and respecively f, τ is he logarihm of he forward exchange rae wih mauriyτ a ime. A ypical es for he unbiasedness hypohesis of he forward rae is: s,. τ s = β f τ s ε τ As argued by Breuer 996, wheher or no he selemen dae of he forward conrac maches he selemen dae of fuure spo rading should be examined for he es. In our case, since we are using forward conracs wih -week mauriy, he quesion is wheher or no he fuure spo rae in equaion should be exacly 7 days ahead of he curren spo rae. According o Grabbe 99, o deermine he one week forward selemen dae for forward rading ha akes place oday, firs deermine he spo selemen dae, and hen he same day of he nex week is he selemen dae of he forward conrac. For example, for a EUR/ USD -week forward conrac enered ino on Monday, he spo selemen dae will be wo business days afer i.e. Wednesday, and hus he delivery dae of he forward conrac should be nex Wednesday. To speculae on his conrac, a speculaor should rade currency in he spo marke nex Monday, wih he delivery dae being nex Wednesday. Due o his inuiional arrangemen, he forward rae on any day of he week is associaed wih he spo rae exacly one week laer. Therefore he fuure spo rae used in he es should be he spo rae exacly one week afer he ransacion dae. Spo selemen dae is wo business days afer he ransacion dae excep for Canadian dollar and Mexican Peso agains U.S. dollars in Norh America. 4

5 Thus, by consrucion, -week forward raes are regressed wih weekly spo reurns. Therefore he frequency of he daa and mauriy of he conracs do no overlap in he esimaion, and serial correlaion of error iems can be ignored here. However, i is well known ha spo reurns have ime-varying volailiy in he second condiional momens. To avoid he inefficiency of he esimaion caused by heeroscadiciy, he Newey-Wes esimaor for he sandard errors is applied in he esimaion. Table repors he resuls. For he CAD, he beas on all weekdays oher han Thursday are significanly negaive numbers around - in our sample. In conras, he bea on Thursday is -.46, and an insignifican -saisics -.78 suggess ha he bea is no saisically differen han ero. For he EUR, Thursday also generaes an insignifican resul. In addiion, he bea on Monday appears o be saisically insignifican and close o ero as well, while oher weekdays have significanly negaive beas around -6. For he JPY, alhough none of he weekdays show a significan resul, he inconsisency beween Thursday bea is -.6 and oher weekdays beas are from -.8 o -. is sill noiceable. Thus, a new paern never repored before can be found in he able: insead of a significanly negaive number like oher weekdays, he bea on Thursday is saisically equal o ero across currencies. Addiionally, Monday is a similar special day for he Euro. This paern can be called he Thursday effec of he forward premium pule 3 To es he robusness of he finding across various subperiods wihin he enire daa period, a rolling regression was applied for each weekday. In he regressions, a window is chosen ha is shorer han he enire daa coverage. Beginning a he firs insance of his window, equaion is used o regress he daa wihin he window and obain one resul. Then, moving he window one posiion forward and regressing he daa in he nex window generaes anoher resul. This process coninues unil he boom of he window reaches he end of he daa. Thus, muliple esimaes can be obained for each weekday o deermine wheher he new finding is consisen 3 The res of he paper will argue and show ha boh he Thursday effec across he currencies and Euro s Monday effec are caused by he similar reasons, and he laer is acually a European version of he Thursday effec.

6 across subperiods. The resuls are illusraed in hisogram figure hrough 3. Each sub-figure of he figure illusraes he disribuion of he beas associaed wih various weekdays row of he figure marix and he choice of window periods column of he marix. As we can see from he las column of he figures, in a long window period 6-year for CAD and JPY, -year for EUR, he beas on Thursdays fall in a narrow inerval beween -. o., while he resuls on oher weekdays are significanly negaive -3 o - for CAD, -4 o -6 for EUR, - o - for JPY. In conras, in shorer window periods lef wo columns of he figure, he beas are mosly disribued around ero and he differences beween Thursday s disribuion and oher weekdays are no clearly idenifiable. In addiion o Thursday, Monday also appears o be a special day for Euro. On ha day, he majoriy of beas are clusered around raher han some significan negaive numbers in he long window period, and such an inconsisency can no be seen eiher in shorer window periods. Therefore, wo conclusions can be drawn from hese figures. Firs, he Thursday Monday effec is robus across subperiods; second, he effec emerges only when he esimaion period is long enough. In summary, he new findings abou he pule discovered by he paper are: Thursdays generae differen resuls han oher weekdays across currencies. Mondays behave similarly for he Euro, 3 he Thursday and Monday effec do no show up unil he window period is long enough. 3. Explanaions The proposed explanaion in his paper arises from he effor o answer he key quesion: under wha condiion can bea be close o ero 4. I derived he condiion based on a simple facor 4 Iniial effor o find an explanaion focuses naurally on wheher abnormal evens or informaion in he foreign exchange marke ake place consanly on Thursday, bu i did no succeed 6

7 model of exchange rae ha follows Backus e. al.. In his model, he forward premium, according o CIP, equals ineres rae differenial. i.e., f s = r r where r, r are he logarihm of foreign and domesic ineres raes respecively. Preliminary daa examinaion shows ha he wo ineres raes are highly correlaed. Also, Anh 4 found ha common facors can explain over 9% of he variaion of he ineres raes. So we can assume ha boh counries yield raes are affeced by some common facors, bu he way he raes are affeced is differen. Suppose,,, are wo laen common facors and,, are consan coefficiens, hus domesic and foreign ineres raes can be wrien as: r,, = 3 = 4 r,, Only wo common facors are included in his seup because of he following reasons. Firs, Ding 7 found ha wo common facors would accoun for almos all variaion of he ineres rae explained by he common facors; Second, inuiively here are wo ypes of informaion in he marke: fundamenal and non-fundamenal informaion, which can be represened by he wo facors respecively; Third, according o Baillie and Bollerslev 994, boh properies of high persisency and random walk can be found in he forward premium. Usually, fundamenal informaion such as macroeconomic variable is highly persisen and prey sable, while nonfundamenal informaion is independen across ime and much more volaile. So his seup can also accommodae he saisical feaures of he forward premium. Wihou losing generaliy, I assume ha,, represens fundamenal informaion and reflecs he non-fundamenal informaion. An ineres rae differenial can be obained from equaions 3 and 4: Exensive sudies show ha CIP generally holds. In fac, dealers quoe he forward raes based on he CIP formula. 7

8 8 r r,, = Therefore, forward premium can be wrien as, s f,, = 6 The fuure spo exchange rae is srucured as forward rae plus risk premium, i.e.: premium risk ε = f s where ε is an error erm wih sandard normal disribuion. The risk premium is cerainly affeced by he laen facors,,, as well. Assume,, premium risk λ λ = where,λ λ are consan coefficiens. Therefore, we have: f ε λ λ =,, s 7 Subracing curren spo rae on he boh sides of equaion 7 resuls in: s f s s ε λ λ =,, 8 Plugging equaion 6 ino equaion 8 resuls in: s s ε λ λ =,,,, 9 Thus, based on equaion 6 and 9, he OLS esimaed slope coefficien of equaion can be expressed as: V V V V λ λ β = which can be rewrien as: λ λ β = V V V V

9 Since he non-fundamenal variable is believed o be much more volaile han he fundamenal one, so he raio of wo variances can be ignored. Then, he slope coefficien can be simplified as: λ β = According o he equaion above, he ero bea suggess: λ which can be rewrien as: λ Plugging his relaionship ino equaion 9, he expeced fuure spo exchange rae can be expressed as: λ Es = s,, In he equaion above, non-fundamenal informaion is canceled ou, which suggess ha if expeced spo exchange rae change depends mainly on fundamenal informaion, hen he bea ends o be ero. Meanwhile, an imporan insiuional facor has o be considered. Equaion is derived from CIP and UIP, while boh CIP and UIP are based on he no-arbirage condiion. Recall he arbirage process for he CIP condiion by using a -week forward conrac. Firs, borrow domesic currency in he money marke. Second, buy foreign currency in he spo marke. Third, deposi foreign currency o earn ineres and fourh sell he foreign currency a he rae specified in he forward conrac sell he foreign currency in he spo marke for UIP one week laer. According o Grabbe 99, he delivery dae of shor-erm money marke insrumens is wo business days afer he ransacion dae. Suppose a speculaor observes informaion and decides o speculae on i on weekday T T could be any number from hrough. The money will be delivered o his accoun on weekday T means wo business days, and hen he can sar 9

10 rading in he foreign exchange marke. Therefore, informaion released on weekday T of his week should significanly affec he relaionship beween spo rae on weekday T of he nex week and forward rae on weekday T his week. According o he heoreical model, ero bea on Thursdays means ha Thursday s expeced spo rae is basically deermined by fundamenal informaion. The insiuional arrangemen previously discussed suggess ha he fundamenal informaion ha dominanly influences invesors expecaion of nex Thursday s spo rae should be released his Tuesday. Thus, he cause of he Thursday effec should be he answer o he following quesion: is here any major fundamenal informaion ha is released consanly on Tuesdays and affecs he nex Thursday s spo rae significanly? Afer screening a lo of fundamenal informaion, he Federal Open Marke Commiee FOMC s moneary announcemen abou he federal fund arge rae is found o be he mos likely candidae. The FOMC, responsible for open marke operaions in he U.S., mees regularly o deermine he adjusmen of he base ineres rae. Ineresingly, i is almos always on Tuesdays when hey mee and announce he adjusmen during he coverage period of our daa. See able 4. When such major fundamenal informaion comes o he marke on Tuesday, marke paricipans are likely o form heir expecaions of he spo raes for nex Thursday mainly based on his informaion. This process explains how fundamenal informaion dominaes non-fundamenal informaion in deermining he spo raes on hese special Thursdays. I is his scenario which I propose as he explanaion of he Thursday Effec. For he Euro, in addiion o Thursday, Monday also possesses similar properies as Thursdays. According o our logic, we expec o see ha some fundamenal informaion such as moneary policy announcemens is released consanly wo business days before Monday, i.e. Thursday. The fac is ha he moneary policy commiee in he European Cenral Bank does mee and announce he base ineres rae mosly on Thursday, as shown in able. So our proposed explanaion can also explain he Monday effec for he Euro.

11 This logic can also explain why his effec only shows up in long window periods: no every Thursday has an announcemen. The FOMC mees abou eigh imes a year, and abou once a monh for is counerpar in he European Cenral Bank. Therefore, he effec is no significan enough o emerge in a relaively shor window period, as i conains fewer observaions of he announcemen weeks. As for oher cenral banks, he moneary policy commiee in he Bank of Canada also mees on Tuesdays wih he same frequency as he FOMC. So he effec of heir moneary policy announcemen should be mixed wih ha of FOMC and does no cause a differen day-of-heweek effec. The meeing dae of he similar commiee in he Bank of Japan is fairly random and no fixed a one specific weekday. Thus, he Japanese moneary announcemen does no cause a differen day-of-he-week effec eiher. The proposed explanaion argues ha when moneary policy announcemens are made, as opposed o non-fundamenal informaion usually privae informaion, marke paricipans consider he news as common knowledge and have similar inerpreaions of he news. Therefore, irraional expecaion behavior will be reduced in he foreign exchange marke. In his sense, our explanaion fis in he lieraure ha aribues he pule o irraional expecaions or expecaion errors. 4. Empirical ess The proposed explanaion has wo componens: he FOMC s he European Cenral Bank moneary announcemen is he direc cause of he Thursday effec Monday effec for Euro, and he ransmission mechanism is ha people expec fuure spo rae mainly based on he fundamenal informaion in he announcemen weeks. These wo aspecs can be used o es he proposed explanaion. A naural way o es he firs componen is o compare he beas obained wih and wihou observaions in announcemen weeks. The logic of he es is simple: if he effec is caused by he

12 announcemens, hen he sample wihou he observaions of he announcemen weeks should generae a bea consisen wih oher weekdays. Accordingly, I esimaed equaion by using hree differen samples: all Thursday observaions, Thursdays excluding announcemen weeks and Thursdays of announcemen weeks only. The Panel A of able repors he beas of hree ess for each currency. The resuls are consisen across currencies. In he sample of all Thursdays, he bea for all currencies are prey close o ero CAD: , EUR: -.88 and JPY: -.6 and have low corresponding -saisics CAD: -.78, EUR: -.39, JPY: -.3. Once he observaions of Thursdays in announcemen weeks are aken ou, he remaining sample generaes dramaically differen resuls. The beas become more negaive and much furher away from ero CAD:-.43 EUR:-6.743, JPY:-. and heir -saisics are much more significan CAD:-.6 EUR:-.3, JPY:-.93. The similariy beween he Thursdays wihou he announcemens and oher weekdays can be seen by comparing he preceding beas wih he average beas of oher weekdays repored in he las column of he able. Apparenly, he forward premium pule reappears on Thursday if he observaions on announcemen weeks are excluded. Moreover, in he es using only announcemen week Thursdays, insignifican -saisics sugges ha he ero bea hypohesis canno be rejeced. For EUR and JPY, he bea even ends o move oward he posiive side. All hese resuls suppor he argumen ha i is announcemen week Thursdays ha significanly change he bea on Thursdays overall and cause he inconsisency wih previous forward premium pule findings. Similar ess are conduced for he Euro s Monday effec and he resuls are repored in panel B of able. In his case, he hree samples consis of all Monday observaions, all nonannouncemen Mondays, and only he Mondays wih he announcemens. The bea based on all he Monday observaions is -.48 wih an insignifican -saisic -., while he sample wihou he observaions in announcemen weeks generaes a negaive bea wih a significan -saisic -.7. This laer finding is consisen wih oher weekdays weekday

13 average is All hese resuls suppor he claim ha he moneary announcemen made by he European Cenral Bank consisenly on Thursday is he direc cause of he Monday effec for he Euro. Finally, we es he second componen of he explanaion. If our proposed mechanism of he Thursday effec is correc, he spo rae should respond significanly o changes in he effecive federal fund rae in he announcemen weeks, bu no in oher weeks. A simple way o es his would be o regress spo rae reurn on he federal fund rae along wih a dummy variable indicaing he announcemen weeks. However, since i is very likely ha marke paricipans inerpre he informaion in a nonlinear way, such a linear or even quadraic es may no be conclusive. To avoid he complicaion of specifying he specific relaionship beween hese wo variables, I used a sign es proposed by Campbell and Dufour 99 o deec he response of exchange raes o he moneary announcemen. Le g denoe any condiioning variable ha is a measurable funcion of informaion se I, and le b denoe he median of y and define n- = sign y b g = S b, g where sign u = for u > and ero oherwise. Under he null hypohesis ha y has consan median b and is independen of g, Campbell and Dufour 99, proposiion show ha S g b has an exac binomial finie sample disribuion wih n rials and a probabiliy of success equal o.. To apply he sign es in our case, le y = s s and g i i =, where s, s are he logarihm of he spo raes, and i, i are he effecive federal fund rae and is foreign counerpar. The sign saisic is given by n- S b = sign s i, where b is he median spo g = s b i exchange rae change. If he ineres rae does no have a significan impac on spo rae reurn, 3

14 hen S g b should have an exac binomial finie sample disribuion wih n rials and a probabiliy of success equal o., which can be verified by a sandard proporional -es. If he ineres rae does have a significan impac, hen a higher federal fund rae reduces he ineres rae differenial foreign minus domesic ineres rae, and according o UIP, should cause he domesic currency o depreciae i.e spo reurn should be higher. Therefore, if he spo rae responds significanly o he announcemens, he raio should be significanly less han.. The ineres raes used in he es were downloaded from he websie of S. Louis Fed and he cenral banks of he foreign counries. The spo reurns came from he DaaSream. Table 3 repors he resuls of his sign es. The raio in he able is calculaed as S g b divided by he number of observaions in each sample. For he sample wihou announcemens, he raios are very close o. CAD:., EUR:.3, JPY:.48. Low -saisics CAD:, EUR:.9, JPY: -.4 sugges ha for each currency, we canno rejec he hypohesis ha he raio equals.. For he sample wih he announcemen weeks only, he raios are significanly less han. CAD:.37, EUR:.8, JPY:. suppored by he high -saisics. These resuls imply ha he federal fund rae does no significanly affec he spo rae in he weeks wihou he moneary announcemens while i does in he announcemen weeks. The resuls obained from his es are hus supporive of he second componens of our proposed explanaion.. Conclusions. This aricle ess he unbiasedness hypohesis of forward raes across weekdays and finds ha he pule seems o disappear on Thursdays across currencies Monday for Euro as well. More specifically, insead of being significan negaive number, he bea is saisically equal o ero on he special weekdays. Furhermore, he paper finds ha his phenomenon only emerges when he window period is long enough. This new empirical finding can be referred o he Thursday Effec of he forward premium pule. 4

15 The paper proposes an explanaion based on a simple facor model of exchange raes. According o he model, he bea being ero suggess ha expeced fuure spo rae depends on fundamenal informaion dominanly. Hence he paper argues ha when U.S. moneary policy announcemens are released regularly on Tuesdays, he speculaors would mainly focus on his fundamenal informaion forming heir expecaions of he fuure spo rae. Meanwhile, due o he delivery arrangemen ha he orders will be seled wo business days afer he ransacion dae in boh he money marke and foreign exchange marke, he arbirage ransacions conduced by he speculaors will be compleed wo business days laer, i.e. Thursday Monday for Euro. Thus he relaionship beween fuure spo rae and curren forward rae is closer o he scenario described by he heory on Thursdays Monday for Euro. Similarly, he Monday effec for Euro can be aribued o European Cenral Bank s regular release of moneary policy informaion on Thursdays, which are wo working days ahead of Mondays. Empirical es of he paper suppor he proposed explanaion. The paper idenifies a calendar effec in he foreign exchange marke. So for fuure research, wheher a rading sraegy can be designed based on his marke inefficiency should be evaluaed. And also, since he paper suggess ha he spo rae responds o he fundamenal informaion such as he announcemen of ineres raes, he abiliy of hese fundamenal variables o forecas fuure spo rae in hese special weeks should be examined.

16 REFERE CES Ahn, Dong-Hyun, 4. Common facors and local facors: implicaion for erm srucures and exchange raes., Journal of Financial and Quaniaive Analysis 39, 69-. Backus, D, Foresi, S, Telmer, C... Affine erm srucure models and he forward premium anomaly. Journal of ffinance Bansal, R. and Dahlquis, M. The forward premium pule: differen ales from developed and emerging economies, Journal of Inernaional Economics -44 Baillie and T. Bollerslev, 994, The long memory of he forward premium, Journal of Inernaional Money and Finance, Elsevier, vol. 3, pp.6-7. Beruer, J.B., and Wohar, M.E., 996. The road less raveled: insiuional aspecs of daa and heir influence on empirical esimae wih an applicaion o ess of forward rae unbiasedness. Economic Journal. 6, Campbell, B., Dufour, J.M., 99. Exac nonparameric orhogonaliy and random walk ess. The Review of Economics and Saisics. 77, 6. Chinn, Menie,, Meredih, G. 4. Moneary policy and long-horion unconvered ineres pariy., IMF Saff papers, Ding, L. 7. The erm srucure of he forward premium pule. Typescrip. Engel, C The forward discoun anomaly and he risk premium: A survey of recen evidence. Journal of Eempirical Finance 3, 3-9. Fama, E., 984. Forward and spo exchange raes. Journal of Moneary Economics. 4, Gourinchas, Pierre-Olivier, and Tornell, A. 4. Exchange rae pules and disored beliefs., Journal of Inernaional Economics 64, Grabbe J. Orlin, 99 Inernaional Financial Markes, 3rd Ediion Prenice Hall; 3 ediion McBrady, M. Carry,. Traders, dealers and he forward discoun anomaly., Manuscrip, Universiy of Virginia, Darden Graduae sschool of Business. Meredih Guy and Ma Yue, The Forward Premium Pule Revisied IMF Working Paper No. /8 Wincoop, E.V. 7. Random walk expecaions and he forward discoun pule. Typescrip. Zhou, S., Kuan, A. M.,. Does he forward premium anomaly depend on he sample period used or on he sign of he premium? Inernaional Review of Economics and Finance., 4, 7-. Simpsona, M. W., Ramchanderb, S and Chaudhryc, M., The impac of macroeconomic surprises on spo and forward foreign exchange markes, Journal of Inernaional Money and Finance 4, Issue, Pages

17 Table : resuls for he es of he forward premium pule across weekdays Currency Coefficiens Mon Tue Wed Thu Fri CAD/USD EUR/USD JPY/USD β β β Noes: he numbers above parenhesis are esimae of alpha and bea in he es equaion s s = β f, s ε., number in parenhesis are -saisics for he ero null hypohesis. τ τ τ 7

18 Panel A currency CAD/USD EUR/USD JPY/USD Panel B currency EUR/USD Table : Tes of samples wih and wihou he announcemens All Thursdays All Mondays Thursday wihou announcemens Mondays wihou meeing Thursdays wih announcemens Mondays wih meeing Average of oher weekdays Average of oher weekdays Noes: numbers above parenheses repored in his able are he esimae of bea in he es equaion s s = β f, s ε. Numbers in he parenheses are -saisics based on τ τ τ ero hypoheses. The equaion is esed by hree samples respecively, including he observaions on all Thursdays Monday for Euro as well, he observaions on Thursday Monday for Euro excluding announcemen weeks and observaions on Thursdays Monday for Euro wih he announcemen only. Wha repored under he average of oher weekdays are he mahemaical average of bea esimae and -saisics on oher weekdays. Bu for Euro, i does no include he esimaes on Monday. 8

19 Table 3: sign es of samples wih and wihou he announcemens CAD/USD EUR/USD JPY/USD Raio for nonannouncemen weeks Raio for announcemen weeks Noes: numbers above parenheses repored in his able are he raio of he sign saisics n- S b = sign s over he sie of each sample. Numbers in he parenheses g = s b i i are he -saisics of he raio es. Null hypohesis is he raio equal.. The es is conduced by wo samples for each currency, including he observaions in weeks wih and wihou he announcemens respecively. 9

20 Table 4: The FOMC meeing daes Dae Day Dae Day Dae Day //997 Wednesday 6/7/ Tuesday 9/4/ Tuesday /6/997 Tuesday 8// Tuesday /6/ Wednesday /3/998 Tuesday /3/ Tuesday // Tuesday 3/3/998 Tuesday // Wednesday /8/3 Tuesday /9/998 Tuesday /9/ Tuesday 3/8/3 Tuesday 6/3/998 Tuesday /3/ Wednesday /6/3 Tuesday 8/8/998 Tuesday /3/ Tuesday 6/4/3 Tuesday 9/9/998 Tuesday 3// Tuesday 8//3 Tuesday /7/998 Tuesday 4/8/ Wednesday 9/6/3 Tuesday //998 Tuesday // Tuesday /8/3 Tuesday //999 Tuesday 6/6/ Tuesday /9/3 Tuesday 3/3/999 Tuesday 8// Tuesday /7/4 Tuesday /8/999 Tuesday 9/7/ Monday 3/6/4 Tuesday 6/9/999 Tuesday // Tuesday /4/4 Tuesday 8/4/999 Tuesday /6/ Tuesday 6/9/4 Tuesday //999 Tuesday // Tuesday 8//4 Tuesday /6/999 Tuesday /9/ Tuesday 9//4 Tuesday //999 Tuesday 3/9/ Tuesday //4 Wednesday // Tuesday /7/ Tuesday /4/4 Tuesday 3// Tuesday 6// Tuesday /6/ Tuesday 8/3/ Tuesday Informaion resources: websie of federal reserve FOMC hp://

21 Table : The European Cenral Bank moneary policy commiee meeing daes Dae Day Dae Day Dae Day Dae Day 3/4/999 Thursday 3/6/ Thursday 4/6/ Thursday // Thursday 3/8/999 Thursday 3/3/ Thursday // Thursday /9/3 Thursday 4/8/999 Thursday 4/3/ Thursday /3/ Wednesday /6/3 Thursday 4//999 Thursday 4/7/ Thursday 6/7/ Thursday 3/6/3 Thursday /6/999 Thursday // Thursday 6// Thursday 4/3/3 Thursday //999 Thursday // Thursday 7// Thursday /8/3 Thursday 6//999 Wednesday 6/8/ Thursday 7/9/ Thursday 6//3 Thursday 6/7/999 Thursday 6// Wednesday 8// Thursday 7//3 Thursday 7//999 Thursday 7/6/ Thursday 8/3/ Thursday 7/3/3 Thursday 7//999 Thursday 8/3/ Thursday 9/7/ Monday 9/4/3 Thursday 7/9/999 Thursday 8/3/ Thursday // Thursday //3 Thursday 8/6/999 Thursday 9/4/ Thursday // Thursday /6/3 Thursday 9/9/999 Thursday // Thursday /8/ Thursday /4/3 Thursday 9/3/999 Thursday /9/ Thursday /6/ Thursday /8/4 Thursday /7/999 Thursday // Thursday /3/ Thursday //4 Thursday //999 Thursday /6/ Thursday /7/ Thursday 3/4/4 Thursday /4/999 Thursday /3/ Thursday 3/7/ Thursday /6/4 Thursday /8/999 Thursday /4/ Thursday 4/4/ Thursday 6/3/4 Thursday //999 Thursday /8/ Thursday // Thursday 7//4 Thursday //999 Wednesday // Thursday 6/6/ Thursday 8//4 Thursday // Wednesday // Thursday 7/4/ Thursday 9//4 Thursday // Thursday 3// Thursday 8// Thursday /7/4 Thursday /3/ Thursday 3// Thursday 9// Thursday /4/4 Thursday /7/ Thursday 3/9/ Thursday // Thursday //4 Thursday 3// Thursday 4// Wednesday /7/ Thursday //4 Thursday The forma of dae: monh/day/year. Daa resources: websie of European Cenral Bank hp://

22 Figure : Disribuion of he slope coefficien for CAD/USD on differen weekdays 6 Monday.-year Monday 3-year 3 Monday 4.-year 6 Monday 6-year Tuesday.-year Tuesday 3-year 3 Tuesday 4.-year 6 Tuesday 6-year Wednesday.-year 3 Wednesday 3-year 3 Wednesday 4.-year 6 Wednesday 6-year Thursday.-year 6 Thursday 3-year 6 Thursday 4.-year 4 Thursday 6-year Friday.-year Friday 3-year 3 Friday 4.-year 6 Friday 6-year Noes: Each sub-figure of he figure illusraes he disribuion of he beas across weekdays row of he figure marix for he choice of window periods column of he marix. Window periods used here are., 3, 4. and 6 years.

23 Figure : Disribuion of he slope coefficien for EUR/USD on differen weekdays 6 Monday.-year 3 Monday.-year 4 Monday 3.7-year Monday -year Tuesday.-year 6 Tuesday.-year 3 Tuesday 3.7-year 6 Tuesday -year Wednesday.-year 6 Wednesday.-year 3 Wednesday 3.7-year 6 Wednesday -year Thursday.-year 6 Thursday.-year Thursday 3.7-year Thursday -year Friday.-year 6 Friday.-year 4 Friday 3.7-year 4 Friday -year Noes: Each sub-figure of he figure illusraes he disribuion of he beas across weekdays row of he figure marix for he choice of window periods column of he marix. Window periods used here are.,., 3.7 and years. These windows are shorer han he windows for CAD and JPY due o he smaller sie of EUR sample. 3

24 Figure 3: Disribuion of he slope coefficien for JPY/USD on differen weekdays 4 Monday.-year 4 Monday 3-year Monday 4.-year 6 Monday 6-year Tuesday.-year 3 Tuesday 3-year Tuesday 4.-year Tuesday 6-year Wednesday.-year 4 Wednesday 3-year Wednesday 4.-year Wednesday 6-year Thursday.-year 3 Thursday 3-year Thursday 4.-year Thursday 6-year Friday.-year 3 Friday 3-year Friday 4.-year Friday 6-year Noes: Each sub-figure of he figure illusraes he disribuion of he beas across weekdays row of he figure marix for he choice of window periods column of he marix. Window periods used here are., 3, 4. and 6 years. 4