Unconventional Monetary Policy and the Dollar: Conventional Signs, Unconventional Magnitudes

Transcription

1 FEDERAL RESERVE BANK OF SAN FRANCISCO WORKING PAPER SERIES Unconvenional Moneary Policy and he Dollar: Convenional Signs, Unconvenional Magniudes Reuven Glick and Sylvain Leduc Federal Reserve Bank of San Francisco November 2015 Working Paper hp:// Suggesed ciaion: Glick, Reuven, Sylvain Leduc Unconvenional Moneary Policy and he Dollar: Convenional Signs, Unconvenional Magniudes. Federal Reserve Bank of San Francisco Working Paper hp:// The views in his paper are solely he responsibiliy of he auhors and should no be inerpreed as reflecing he views of he Federal Reserve Bank of San Francisco or he Board of Governors of he Federal Reserve Sysem.

2 Unconvenional Moneary Policy and he Dollar: Convenional Signs, Unconvenional Magniudes Reuven Glick and Sylvain Leduc November 29, 2015 Economic Research Deparmen Federal Reserve Bank of San Francisco Absrac We examine he effecs of unconvenional moneary policy surprises on he value of he dollar using high-frequency inraday daa and conras hem wih he effecs of convenional policy ools. Idenifying moneary policy surprises from changes in ineres rae fuure prices in narrow windows around policy announcemens, we find ha moneary policy surprises since he Federal Reserve lowered is policy rae o he effecive lower bound have had larger effecs on he value of he dollar. In paricular, we documen ha he impac on he dollar has been roughly hree imes ha following convenional policy changes prior o he financial crisis. JEL classificaion: E5, E43, F31 Keywords: moneary policy, large scale asse purchases; quaniaive easing, dollar, exchange rae We hank Sefania D Amico for providing us wih her measure of federal funds rae policy surprises. We hank Daniel Molior and Jeremy Pearce for excellen research assisance. We also hank paricipans a he Peerson Insiue for Inernaional Economics symposium on Spillovers of Unconvenional Moneary Policy for commens. The views expressed here are hose of he auhors and do no necessarily represen hose of he Federal Reserve Bank of San Francisco or he Board of Governors of he Federal Reserve Sysem. Reuven.Glick@sf.frb.org, Sylvain.Leduc@sf.frb.org. 1

3 1. Inroducion During he financial crisis and is afermah, he Federal Reserve inroduced new moneary policy measures o sabilize financial markes and miigae he effecs of he crisis on economic aciviy. These so-called unconvenional policy ools have been necessary boh because of he exraordinary naure of he financial crisis and because he federal funds policy rae was quickly dropped o is effecive lower bound of near zero percen by he end of As a resul, he Federal Reserve urned o large-scale asse purchases (LSAPs) also commonly called quaniaive easing and o greaer forward guidance abou he fuure pah of moneary policy o achieve is dual mandae of price sabiliy and maximum employmen. These new policy ools come wih a significan amoun of uncerainy regarding heir effeciveness, paricularly wheher he sandard ransmission channels of moneary policy hrough financial asse markes work as well as hey did in he pas. An imporan channel hrough which changes in moneary policy affec he economy, paricularly when he policy rae is near is lower bound, is he value of domesic currency. There is much empirical evidence, for insance, documening ha he dollar ypically depreciaed following declines in he federal funds rae in he pre-crisis period (see, for insance, Clarida and Galì, 1994; Eichenbaum and Evans, 1996; Faus and Rogers, 2003; Scholl and Uhlig, 2008; and Bouakez and Normandin, 2010). In his paper, we examine how he U.S. dollar has reaced o changes in unconvenional moneary policy since he federal funds rae reached is zero lower bound in December 2008 and how his effec compares o hose following changes in moneary policy in he period before hen. In paricular, we analyze he impac of moneary policy announcemens beween 1994 and 2014, hus capuring he effecs of he hree waves of quaniaive easing. We use high-frequency inraday daa in panel regressions o sudy he dollar s movemens agains he currencies of 2

4 major U.S. rading parners in ime inervals immediaely following moneary policy announcemens by he Federal Reserve. The use of inraday daa enables us o beer isolae he response of he dollar o moneary announcemens from oher possible deerminans. To conrol for he likelihood ha marke paricipans anicipae policy changes, we consruc surprise changes in moneary policy using changes in shor-erm and long-erm ineres rae fuures around he ime of policy announcemens. We compue hree ypes of moneary policy surprises. We firs use changes in federal funds rae fuures around Federal Open Marke Commiee (FOMC) announcemens abou he federal funds rae arge o measure surprises in he policy arge, ermed arge surprises by Kuner (2001). 1 Clearly, arge surprises are only relevan during he pre-crisis period when he federal funds rae was above he zero lower bound. Second, as emphasized by Gürkayrkanak, Sack, and Swanson (2005), FOMC announcemens no only conain informaion abou he policy arge, bu also include communicaion abou he fuure pah of moneary policy. As a resul, we follow heir approach o isolae he surprise movemens in he expeced pah of he federal funds rae, as measured by he change in he one-year ahead euro-dollar fuures rae, which we label shor-erm pah surprises. Third, we consruc an addiional measure of policy pah surprises, ha we erm long-erm pah surprises, using long-erm Treasury fuures raes. The idea is ha hese surprises may capure he Federal Reserve s aemps o direcly influence long-erm Treasury raes via LSAPs and long-erm forward guidance (see Wrigh, 2012). Since he pre-crisis period was dominaed by he use of changes in he level and pah of he federal funds arge rae as he main ool of moneary policy, we refer o his period as he convenional policy period. Correspondingly, we denoe he crisis and pos-crisis period when 1 See also Bernanke and Kuner (2005), Fleming and Piazzesi (2005), Faus e al. (2007), and D Amico and Farka (2011) for oher analyses of he effecs of moneary policy arge surprises during he period before he financial crisis. 3

5 LSAPs and relaed policies were he main ools of moneary policy as he unconvenional policy period. Our resuls show ha he exchange rae channel of he ransmission of moneary policy is highly effecive during boh he convenional and unconvenional policy periods, bu ha he effecs are significanly larger in he laer period. In paricular, we firs documen ha during he convenional period he U.S. dollar depreciaed significanly in response o boh arge and shor-erm pah surprises, hough no in response o long-erm surprises. Specifically, we find ha a one sandard deviaion surprise easing leads o a oal decline of 17 basis poins in he value of he dollar in he hour afer announcemens. In comparison, during he unconvenional policy period, he U.S. dollar depreciaed significanly in response o boh shor-erm and long-erm pah surprises, wih arge surprises no longer a feasible ool of moneary policy as long as he federal funds rae was expeced o remain a is effecive lower bound. Since he end of 2008, we find ha a one sandard deviaion surprise easing in unconvenional policy leads o a oal decline of 51 basis poins in he value of he dollar wihin 60 minues, a magniude roughly hree imes ha during he convenional period. Our paper adds o a growing and acive lieraure on he effecs of unconvenional moneary policy. Saring wih Gagnon e al. (2011), several papers have aemped o analyze he effeciveness of recen moneary policy acions wih even sudies of Federal Reserve announcemens; see, for insance, Neely (2010), Krishnamurhy and Vissing-Jorgensen (2011), D Amico e al. (2012), Glick and Leduc (2012), Hamilon and Wu (2012), and Li and Wei (2012). By emphasizing he effecs on he U.S. exchange rae, our work relaed o ha of Neely (2010) and Wrigh (2012) who look a he impac of announcemens of large-scale asse purchases and oher announcemens by he Federal Reserve on he dollar. However, our focus is 4

6 differen, as we seek o compare he effec of surprise changes in unconvenional policy, including boh shor-erm and long-erm pah surprises, on he exchange rae o hose during he convenional period. While our approach parly follows Wrigh s mehodology in consrucing moneary policy pah surprises, we also make an addiional disincion beween shor-erm and long-erm pah surprises. In addiion, our work differs from Neely in ha i conrols for marke expecaions of possible changes in moneary policy, which is imporan o precisely idenify he surprise componen of policy announcemens. We also have he benefi of working wih a longer sample ha includes policy announcemens during he firs, second, and hird rounds of largescale asse purchases beween 2008 and Neely s sample covers only he firs round of LSAPs beween November 2008 and November 2009, while Wrigh s sample of 28 observaions exends o Sepember 2011 o encompass he second round, bu no he hird round. 2 Finally, our approach here differs from ha followed in previous work of ours (Glick and Leduc, 2013) which absraced from he ransmission of moneary policy via pah surprises during he convenional period. Taking hese surprises ino accoun alers he comparison of he effecs of moneary surprises on he dollar across regimes, which we now find o be subsanially larger during he unconvenional period. In addiion, our mehodological approach differs in ha in his paper we employ a pooled panel ha includes observaions from boh he convenional and unconvenional periods. This enables nesed ess o direcly compare he effeciveness of policies across periods. The paper is organized as follows. In Secion 2 we describe our daa and measures of moneary surprises. Secion 3 presens he benchmark empirical resuls for he effecs of 2 See also Rogers, Scoi, and Wrigh (2014) who examine he effecs of unconvenional policies by he Federal Reserve, he ECB, he Bank of England, and he Bank of Japan on bond yields and sock prices, in addiion o hose on exchange raes. Bowman, Londono, and Sapriza (2014) examine he effecs of unconvenional U.S. moneary policies on asse prices in emerging markes, including exchange raes. 5

7 unconvenional and convenional moneary policy on he value of he dollar. Robusness exercises are repored in Secion 4. Secion 5 concludes. 2. Idenificaion of Moneary Policy Evens and Surprises 2.1 Idenifying moneary policy surprises We examine he effecs of moneary policy surprises on he value of he U.S. dollar during he recen period when policymakers relied heavily on unconvenional policy ools, such as large-scale asse purchases and communicaions abou fuure policy acions he unconvenional policy period, and conras hese effecs o hose following policy surprises when he federal funds rae arge was above he zero lower bound he convenional policy period. The ransiion beween hese wo periods is somewha blurred since convenional policy acions were sill being employed while he Federal Reserve s inenions o adop unconvenional measures were being signaled. For insance, while he FOMC lowered he federal funds rae o is effecive lower bound on December 16, 2008, he fuure use of unconvenional policy ools had already been indicaed by Chairman Bernanke in speeches in November and early December ha year. In our benchmark specificaion, we assume ha he convenional period ends in Ocober As a resul, hese speeches, which provided imporan informaion o marke paricipans abou he ype of unconvenional policies ha migh be pursed in he fuure, are included in he se of policy announcemens during he unconvenional period. One advanage of using his sample spli is ha i makes our sample of unconvenional policy announcemens more comparable o ha ypically used in he lieraure, as we discuss below. Neverheless, we also conduc sensiiviy analysis o alernaive sample breaks. Thus, our sample period for convenional moneary policy acions exends from February 1994, when he FOMC began issuing a press release afer every meeing and every change in 6

8 policy, unil Ocober The period characerized by unconvenional moneary policy acions spans he period from November 2008 o he end of our sample in December The exen o which an announcemen affecs he currency when i is released o he public depends on how much marke paricipans expec he announcemen. If marke paricipans fully anicipae he conen of an announcemen, hen no addiional informaion is revealed a he ime of he announcemen s release and he value of he dollar should no move as a resul. Therefore, conrolling for marke paricipans expecaions is crucial for our analysis. To idenify surprise changes in moneary policy, we use changes in ineres rae fuures in a igh ime inerval around moneary policy news. For he convenional policy period, given ha moneary policy is conduced via changes in he arge for he federal funds rae, we follow he approach proposed in Kuner (2001), and use he change in federal funds rae fuures consruced by D Amico and Farka (2011) o idenify moneary policy surprises in he arge for he federal funds rae. 4 We refer o hem as arge surprises. To beer isolae he influence of changes in moneary policy, he procedure uses inraday ick daa o measure he change in federal funds rae fuures from 10 minues before a 3 Our benchmark sample includes unscheduled inermeeing announcemens on April 18, 1994, January 3, 2001, April 18, 2001, January 22, 2008, and Ocober 8, 2008, and excludes unscheduled announcemens made on Ocober 15, 1998, Sepember 17, 2001, as well as hose on Augus 10, 2007, Augus 17, 2007, and March 11, The Ocober 15, 1998 even followed he Russian ruble devaluaion and he near collapse of Long-Term Capial Managemen, and governmen securiies markes were closed a he ime of he FOMC announcemen ha day. The Sepember 17, 2001 even was excluded as well, on he grounds ha asse marke responses a ha ime reflec no jus he effecs of he FOMC announcemen bu also he fac ha i was he firs day ha he federal funds rae marke was open afer he Sepember 11 erroris aack. Bernanke and Kuner (2005) and D Amico and Farka (2011) also exclude Ocober, 15, 1998 and Sepember 17, The unscheduled meeings of Augus 10, 2007, Augus 17, 2007, and March 11, 2008 are excluded because he FOMC merely communicaed awareness of evens afer hese meeings and did no announce policy changes. 4 Following Kuner (2001), we assume ha he federal funds fuures rae can be expressed as a weighed average of he rae prevailing so far in he monh and he expeced rae for he res of he monh, plus a risk premium. Assuming a consan risk premium implies ha our moneary surprise measure can be defined as he change in he fuures rae, adjused by he scale facor, D/(D-d), where D is he number of days in he monh and d is he day in he monh of he moneary policy announcemen. We use his definiion as long as he announcemen occurs earlier han he las seven days of he monh. If he announcemen falls in he las seven days, he surprise is compued as he unadjused change in he nex-monh federal funds fuures conrac o avoid unduly large adjusmen facors. 7

9 policy announcemen o 20 minues afer. 5 This sraegy provides a good measure of moneary policy shocks if possible ineres risk premia remain relaively consan around policy announcemens. However, as Gürkaynak, Sack, and Swanson (2005) have highlighed, FOMC announcemens during he convenional policy period no only conain informaion abou he curren arge for he federal funds rae, bu also include informaion abou he fuure pah of moneary policy. Following Gürkaynak eal, w define he shor-erm pah surprises as he change in he one-year eurodollar fuures rae around he ime of policy announcemens ha are orhogonal o he arge surprises. 6 For he pos-crisis period, idenifying moneary policy surprises wih he changes in federal funds rae fuures is no a feasible empirical sraegy as long as he federal funds rae is expeced o remain a is effecive lower bound and moneary policy is conduced hrough unconvenional means. However, pah surprises of he kind suggesed by Gürkaynak, Sack, and Swanson (2005) can be used o idenify policy surprises associaed wih forward guidance or LSAPs during he unconvenional policy period. In addiion, given he Federal Reserve s emphasis on direcly lowering long-erm ineres raes hrough unconvenional means, we differeniae beween shor-erm and a long-erm pah surprises by also examining he change in longer-erm fuures raes around policy announcemens. More specifically, we define long-erm pah surprises as he change in he principal componen of he wo-, five-, en, and hiry-year Treasury rae fuures, again measured during a 30-minue window, from 10 minues before an announcemen o 20 minues afer (see Wrigh, 2012)). 7,8 We examine he effec of long-erm 5 This window represens he narrow window in D Amico and Farka (2011). They also considered wider windows, exending o 60 minues afer announcemens. We use he wider 60 minue windows as a robusness check. 6 Specifically, we use ransacion prices for he eurodollar conrac wih mauriy closes o one year. 7 We use he neares dae fuures conracs on Treasuries from Tickdaa. The surprises were consruced from changes in he reurns on he on he wo-, five-, en-, and hiry-year bond fuures conracs, divided by he duraion of he cheapes-o-deliver securiy in he fuures baske, as gahered from Bloomberg. In our principal componens 8

10 pah surprises during he convenional as well as he unconvenional periods since policy announcemens during he convenional period may also conain informaion abou he fuure pah of policy ha is no capured by he shor-erm pah surprises. For he convenional period, we isolae he separae effecs of arge, shor-erm pah, and long-erm pah surprises by orhogonalizing (1) he shor-erm pah surprises wih respec o he arge surprises, and (2) he long-erm pah surprises wih respec o boh he arge surprises and he shor-erm pah surprises. For he unconvenional period, we orhogonalize he long-erm pah surprises wih respec o he shor-erm pah surprises.. All policy surprises are demeaned, scaled o have a sandard deviaion of 1, and defined such ha surprises wih a posiive sign indicae moneary easing, while surprises wih a negaive sign indicae moneary ighening. Overall, he news evens in he convenional policy period consis of 124 FOMC announcemens, 119 following scheduled meeings and 5 following unscheduled inermeeing communicaions. The series includes unscheduled meeings during his period only if he announcemens included a change in he federal funds arge. For insance, he measure excludes he unscheduled meeings in 2007 because he Federal Reserve did no announce a change in he federal funds rae a hose meeings. 9 For he period characerized by unconvenional moneary policy, we use all FOMC announcemens beween December 2008 and December 2014 including boh regularly scheduled and some unscheduled meeings. We also include seleced speeches and esimonies given by Board of Governors Chairman Bernanke in which he signaled possible policy changes, analysis of hese duraion-adjused yield changes, we ake he eigenvecor corresponding o he larges eigenvalue, i.e., he firs principal componen, and muliply each yield change by is respecive eigenvecor componen. I should be noed ha he bulk of Federal Reserve asse purchases during he hird LSAP round involved morgage-backed securiies. However, we do no have inraday daa on hese securiies since hey ypically are raded over he couner. 8 Wrigh (2012) uses a baseline surprise window from 15 minues before a given Federal Reserve announcemen unil 1 hour and 45 minues afer. Our surprise window (-10, +20) was chosen o mach ha of he narrow measure of D Amico and Farka (2011) for federal fund surprises employed below. A wider surprise window is considered as a robusness exercise. 9 The unscheduled meeings included in he measure are April 18, 1994, January 3, 2001, April 18, 2001, January 22, 2008, and Ocober 8, See foonoe 3 for more deails. 9

11 paricularly hose suggesing modificaions o he Federal Reserve inenions o buy long-erm asses. The major announcemens ha refer o large-scale asse purchases and forward guidance news are lised in Table 1. The complee sample for he unconvenional policy period, which includes hese LSAP announcemens as well as oher announcemens following FOMC meeings, consiss of a oal of 56 observaions. 10 Our sample hus encompasses announcemens used in oher sudies on he effecs of large-scale asse purchases. For insance, our announcemens associaed wih he firs round of large-scale asse purchases (LSAP1) beween December 16, 2008, and March 18, 2009, largely overlap wih hose used by Gagnon e al. (2011) and Neely (2010). Similarly, he five announcemens for he second round of asse purchases (LSAP2) from Augus 10 o November 3, 2010, are similar o hose used by Wrigh (2012), Krishnamurhy and Vissing-Jorgensen (2011), and Glick and Leduc (2012). In addiion, our analysis encompasses several major announcemens associaed wih he hird round of asse purchases (LSAP3), which was iniiaed in Sepember 2012 and ended in Ocober This round of announcemens also includes he Congressional esimony of Chairman Bernanke on May 22, 2013, which led o he so-called aper anrum. 2.2 Inraday exchange rae movemens We conduc our analysis using inraday daa on currency fuures prices from Tickdaa for he days in our announcemen sample. The daa se conains minue-by-minue ick ransacion prices on foreign exchange conracs involving he U.S. dollar wih several currencies, including he, Briish pound, Canadian dollar, euro, and yen. 11 In 2010, hese four currencies accouned 10 In addiion o he LSAP-relaed speeches by Chairman Bernanke cied in Table 1, our sample also includes a speech on Augus 26, 2011, when he Chairman saed he Fed was considering all of is opions, hough he was no explici abou addiional policy acions. We do no separaely break ou FOMC announcemens relaed o he Mauriy Exension Program involving he sale of shor-erm Treasuries o purchase longer-erm asses for he Federal Reserve s balance shee. For he Bernanke speeches on November 25, 2008 and December 1, 2008, we impued values of 0 for he arge surprise measure, since here were no announcemens regarding he policy arge. 11 These daa are based on conracs raded on he Chicago Board of Trade. We use he price of he neares, mos heavily raded fuures conrac on each announcemen day. In he case of he euro, we use he deuschmark before he euro s inroducion in

12 for over 70 percen of all spo dollar ransacions 12 and over 60 percen of all swap and fuures dollar ransacions (BIS, 2010), while he counries issuing hese currencies accouned for abou 40 percen of U.S. bilaeral rade ransacions. One advanage of using inraday daa ha is paricularly relevan for moneary policy announcemens is ha i enables us o beer isolae heir effecs. For insance, many sudies of large-scale asse purchases by he Federal Reserve since 2008 have relied on daily daa o assess he effec of unconvenional moneary policy on he price of financial asses (see, for insance, Gagnon e al., 2011)). This approach assumes ha he marke effecs from a moneary announcemen will dominae effecs from any oher informaion released ha day. However, his assumpion may be paricularly roublesome for asse prices such as exchange raes, which reac naurally o news from around he world. Hence, i is more difficul o precisely uncover poenial links beween moneary policy announcemens and movemens in currency values using daily daa, as he effecs of oher news evens on he U.S. dollar are likely o confound hose from moneary policy. For insance, sudying he effecs of he European Cenral Bank (ECB) Securiies Marke Programme on sovereign yields, Ghysels e al. (2013) found ha he use of iner-day daa masks he significan effecs ha he ECB s inervenions had on sovereign yields ha only could be deeced using higher frequency inra-day daa. Consequenly, we look a movemens in he value of he U.S. dollar agains foreign currencies in relaively narrow ime inervals. Consisen wih our idenificaion of moneary policy surprises, we use response windows around moneary policy announcemens of 30 minues (10 minues before, unil 20 minues afer) and 70 minues (10 minues before, unil 60 minues afer). Using igh ime inervals helps us isolae he effecs of he moneary announcemens from oher possible deerminans of currency values, assuming hese 12 The euro, yen, pound, and Canadian dollar accouned for 39, 15, 12, and 7 percen of spo ransacions, respecively. 11

13 announcemens rapidly influence he views of marke paricipans and are quickly refleced in he value of he dollar. For comparison, we also repor resuls exending he response surprise windows o 1440 minues, i.e., 24 hour, afer announcemens. 3. Resuls 3.1 Changes in value of he dollar during LSAP rounds We begin our analysis by reporing he raw, i.e., acual, changes in he value of he dollar during he hree rounds of LSAPs. Figure 1 illusraes he inraday behavior of bilaeral exchange raes on seleced LSAP announcemen days. As shown in panel A, he dollar depreciaed sharply agains all four currencies on December 16, 2008, immediaely afer he 2:15pm FOMC announcemen abou he deails of LSAP1. The dollar depreciaion was smaller following he seleced FOMC announcemens abou LSAP2 and LSAP3. In conras, he dollar appreciaed sharply during he aper anrum following Chairman Bernanke s congressional esimony on May 22, 2013, as markes evidenly inerpreed his discussion abou he fuure lifoff of he federal funds rae as a surprise moneary ighening. Table 2 repors changes in he value of he dollar vis-à-vis he pound, Canadian dollar, euro, and yen in response o he major announcemens during he hree LSAP rounds idenified in Table 1, wih he laer round separaed ino subsamples associaed wih he ramp-up of asse purchases, he aper anrum episode and subsequen announcemens relaed o apering of hese purchases. The response windows sar 10 minues before announcemens and end 20 minues afer. Observe ha he dollar depreciaed agains hese currencies in response o announcemens during all hree LSAP rounds, and appreciaed during he aper anrum episode, and subsequen aper-relaed announcemens. (The appreciaion of he dollar agains he yen during early phase of LSAP3 is an excepion, possibly because of he yen s srong appreciaion in he week before 12

14 he Sepember 13, 2012, FOMC meeing and marke alk abou possible Bank of Japan inervenion.) On a rade-weighed basis, he dollar depreciaed by an average of 62, 23, and 13 basis poins (bps) afer announcemens abou LSAP1, LSAP2, and he ramp up of LSAP3, respecively. 13 The relaively small effec under LSAP3 does no necessarily imply ha he Fed s LSAP3 moneary policy acions were ineffecive, since he markes may have anicipaed hese announcemens and incorporaed hem ino asse prices. This moivaes he need o conrol for he exen o which he announcemens were surprises o he marke. During he aper anrum episode, when markes inferred a greaer likelihood of Federal Reserve ighening in he near erm, he dollar appreciaed by 60 bps. The dollar also appreciaed in subsequen pre-aperrelaed announcemens by an average of 1.3 bps, hough here is a fair degree of variabiliy across currencies. For comparison, he able also shows oal changes in he inerday value of he dollar agains major currencies, as calculaed by he Board of Governors over he 24-hour period from he end of floor rading on he day prior o each announcemen (usually 2:30pm EST) and he end of floor rading on he announcemen day. 14 Noe ha he inerday changes have he same signs bu are generally larger han he inraday changes measured over he even window periods. 3.2 Pooled Effecs of convenional and unconvenional moneary policy surprises We esimae he effecs of surprise moneary policy announcemens on he value of he dollar agains he Briish pound, Canadian dollar, he DM/euro, and he yen using he following panel specificaion: 13 ST LT a 2, i 2PS 2PS i, S (1) ST LT u i,, w a1, i 1TS 1PS PS D 13 We consruc rade weighs from IMF Direcion of Trade daa in 2011 on U.S. bilaeral expors and impors wih he U.K., Canada, Eurozone, and Japan, wih calculaed weighs of 0.07, 0.41, 0.39, and 0.13, respecively. Resuls from aking simple averages are comparable. 14 Noe ha all of he LSAP evens repored in Table 2 occurred before he end of rading on he day of announcemen.

15 where S i,, w is he (log) change in he exchange beween currency i and he Unied Saes a ime during a ime window w. TS is he federal funds rae arge surprise, PS ST is he shor-erm pah surprise, PS LT is he long-erm pah surprise, a i is a currency fixed effec, and is an error erm. u D is a dummy variable ha is equal o one for he unconvenional period and is zero oherwise. The parameers α 1, β 1, and γ 1 represen he effecs of arge surprises, shor-erm pah surprises, and long-erm pah surprises on he dollar during he convenional period, respecively. Shifs in he impac of he shor- and long-erm pah surprises on he dollar during he unconvenional compared o he convenional one are capured by he parameers β 2 and γ 2. The effecs of he shor- and long-erm pah surprises on he dollar during he unconvenional period are hus given by (β 1 + β 2 ) and (γ 1 + γ 2 ). Noe ha we assume ha here are no arge surprises during he unconvenional period, since he arge for he federal funds raes was a is effecive lower bound. 15 As discussed in Secion 2, posiive values of he moneary policy surprises are defined o indicae moneary easing surprises, while he exchange rae is defined as unis of foreign exchange per U.S. dollar, so ha a decrease in S indicaes a depreciaion of he dollar. Hence, negaive coefficien esimaes are consisen wih he finding ha moneary policy easing leads o a depreciaion of he dollar. To illusrae he relaionship beween he change in he exchange rae and he differen moneary surprises, Figure 2 repors scaer plos of he change in he value of he dollar agains arge surprises, shor-erm pah surprises, and long-erm pah surprises for he convenional and unconvenional periods. To convey he informaion compacly, we rade-weigh he dollar 15 Federal funds rae fuures were hinly raded during his period. Their movemens more han likely did no represen expecaions of fuure policy changes, since given he amoun of excess reserves held by banks, he federal funds rae ceased o be an effecive moneary policy ool. Therefore, we absrac from arge surprises alogeher during he unconvenional policy period. 14

16 exchange raes agains he four currencies included in our analysis he Briish pound, Canadian dollar, euro, and yen. Firs, observe ha he sample includes boh negaive, i.e., unexpeced ighening, as well as posiive, i.e., unexpeced easing, moneary surprises. The scaers indicae a clear negaive relaionship beween he dollar and moneary surprises, paricularly for he arge and he shorerm pah surprises during he convenional period and for he shor- and long-erm pah surprises during he unconvenional period. Thus, surprise moneary loosening (ighening) are associaed wih dollar depreciaion (appreciaion), he more so he greaer he surprise. In addiion, we noe ha he dollar appeared o move subsanially more in response o moneary surprises during he unconvenional period han during he convenional period. A more formal empirical analysis confirms his assessmen. Table 3 repors coefficien esimaes of equaion (1) from regressions of changes in he value of he dollar on our measure of policy surprises, using response windows of lenghs ranging from 10 minues before he announcemen o w = 20, 60, 1440 minues (i.e., 24 hours) afer. Consans are included in he regressions, bu are no repored in he able for breviy. We firs concenrae on he effec of policy surprises on he dollar during he convenional period. Table 3 indicaes ha he dollar is affeced via wo channels. Firs, a one sandard deviaion surprise easing in he federal funds rae arge leads o a 6.96 basis poin decline in he value of he dollar 60 minues afer a policy announcemen and 8.32 basis poins a day afer. However, he dollar is also impaced by surprise informaion abou he fuure pah of moneary policy. Specifically, we find ha a one sandard deviaion easing in he shor-erm pah surprise during he convenional period leads he dollar o depreciae 8.58 basis poins 60 minues following announcemens and basis poins one day afer. These effecs are saisically significan a 1% or lower. In conras, long-erm pah surprises did no much impac 15

17 he exchange rae during he convenional period, as he esimaed magniude of 1 is small and barely significan a he 10% level only in he firs 20 minues. The ransmission of moneary policy o he exchange rae operaes differenly during he unconvenional policy period, wih he effecs of long-erm as well as shor-erm surprises boh being significanly larger han during he convenional period. More specifically, as Table 3 indicaes upon summing he coefficien esimaes β 1 and β 2, a one sandard deviaion shor-erm pah surprise leads o a (44.27) basis poins depreciaion 60 minues (24 hours) afer an announcemen, an effec far larger han during he convenional policy period. In addiion, we find ha long-erm pah surprises have effecs of a similar magniude as he shor-erm pah surprises during he unconvenional policy period: summing he coefficiens esimaes γ 1 and γ 2 implies ha, in response o a one sandard deviaion in he long-erm pah surprise, he dollar depreciaes by (22.89) basis poins, 60 minues (24 hours) afer announcemen. Convering our resuls ino basis poin erms, a arge surprise of one percenage poin (100 bp) causes an esimaed 0.87 percen decline in he value of he dollar wihin 60 minues. Similarly, a one percenage poin shor-erm (long-erm) pah surprise during he convenional period leads he dollar o depreciae by 1.3 percen, while a long-erm pah surprise leads o a negligible depreciaion of 0.4 percen. During he unconvenional period, a one percenage poin shor-erm pah surprise causes a 4.6 percen dollar depreciaion; a long-erm pah surprise leads o a depreciaion of comparable magniude, of 3.8 percen. 16 Table 3 indicaes noable and inuiive differences beween he differen channels hrough which moneary policy announcemens can affec he exchange rae in he convenional and unconvenional periods. Our resuls are comparable o oher findings in he lieraure. For example, Hausman and Wongswan (2011) also found ha during he pre-crisis period he dollar 16 The conversion ino basis poins changes uilizes he fac ha he sandard deviaions over he sample period of shor-erm and long-erm pah surprises are 6.34 bps and 5.41 bps, respecively, and ha he sandard deviaion of arge surprises is 7.75 bps. 16

18 is affeced no only by arge surprise announcemens, bu also by surprise announcemens abou he fuure pah of policy. 17 We augmen his finding wih he resul ha his forward-guidanceype channel during he convenional policy period is capured solely by shor-erm pah surprises, as he dollar barely reacs o long-erm pah surprises. In conras, during he unconvenional period he dollar responds significanly o longerm pah surprises as well as o shor-erm pah surprises. This is consisen wih he Federal Reserve s objecive of lowering long-erm ineres raes by purchasing long-erm asses in large amouns. 18 In urn, he absence of such a program during he convenional period is consisen wih our finding ha he dollar did no reac significanly o long-erm pah surprises before he crisis. Our finding ha long-erm pah surprises maer significanly during he unconvenional period is broadly in line wih he resuls repored in Wrigh (2012) and Rogers, Scoi, and Wrigh (2014) for he pound, he euro, he yen, and he Canadian dollar. 19 To compare he magniude of he effecs of policy surprises during he convenional and unconvenional periods, we assume ha a ypical FOMC announcemen during boh periods includes informaion abou he arge for he federal funds rae and also language abou he fuure pah of moneary policy. Thus, o calculae he effecs of moneary policy surprises beween 1994 and 2008, we sum he coefficien esimaes on he arge surprise and shor-erm and long-erm pah surprises ( ), which implies a basis poin dollar depreciaion, 60 minues afer 17 As noed above, we find ha he dollar depreciaed on average by 1.3 percen in response o a 100bps shor-erm pah surprise. Hausman and Wongswan (2011) examine he effecs of U.S. arge and shor-erm pah surprises on daily exchange rae changes for a panel of advanced and emerging economies during he convenional rae period from 1994 o Like us, hey find foreign currency values ypically respond more o pah han o arge surprises and repor ha a 100bps pah surprise leads on average o a 1.6 percen depreciaion of he dollar, comparable o our finding. 18 Of course, he long-erm pah surprise could be due o he Fed s forward guidance in addiion o is large-scale asse purchases. Our approach does no allow us o disinguish beween hese wo channels. 19 Bowman e al (2014) analyze he effecs of U.S. long-erm pah surprises on asse markes prices, including exchange raes, in emerging markes, during he unconvenional period. In an even sudy hey find evidence ha emerging marke currencies responded over 2 day windows around U.S. moneary policy announcemens. In a panel sudy hey find ha hese effecs are smaller (and no saisically significan) afer conrolling for counry-specific characerisics ha affec vulnerabiliy o changes in U.S. moneary policy. 17

19 announcemens (noe ha we include he effec of he long-erm pah surprises even hough hey are no saisically significan in mos cases). For he unconvenional period, we assume ha a ypical surprise announcemen is composed of boh shor-erm and long-erm pah surprises (by summing 1, 2, 1, and 2 ). In his case, we find ha he dollar depreciaed basis poins 60 minues following a surprise easing announcemen, an effec which is abou hree imes larger han is effec following a convenional policy surprise easing. The relaively greaer impac of unconvenional policy surprises sill remains one day afer announcemens. 20 Thus, our resuls sugges ha moneary policy remained effecive in affecing he exchange rae even afer reaching he zero lower bound, which is in line wih he finding of Swanson and Williams (2014). However, while hey find ha he sensiiviy of he pound/dollar and euro/dollar exchange raes o economic news remained abou he same before and afer he zero lower bound was reached, our analysis shows ha he dollar responded by more following unconvenional policy surprises. 21 The findings from our panel regressions also hold for he individual currencies underlying our panel resuls. Table 4 presens individual resuls for he U.S. dollar exchange rae agains he Briish pound, Canadian dollar, euro, and yen. For breviy, we repor only he effecs wih he 60-minue response window afer policy announcemens. As for he pooled resuls during he convenional period, boh he arge and shor-erm pah surprises affec he dollar s 20 In Glick and Leduc (2013) we repored ha he impac of policy surprises on he dollar during he convenional and unconvenional periods were of similar magniudes. This difference in resuls is due o he fac ha our previous work absraced from he presence of boh shor-erm and long-erm pah surprises in addiion o he arge surprises during he convenional period. To compare he effecs on he dollar across periods, we convered he effecs of (long-erm) pah surprises during he unconvenional period ino equivalen arge surprise effecs during he convenional period, using an esimae of heir correlaion beween 1994 and Our curren approach differs since we include (orhogonalized) pah surprises along wih federal funds rae arge surprises. In addiion, here we employ a panel ha pools observaions from boh he convenional and unconvenional periods. This enables nesed ess o direcly compare he effeciveness of policies across periods. In a laer robusness resul we show ha he difference in coefficiens sill remains when working wih nonnesed samples. 21 The difference beween our resuls and hose of Swanson and Williams (2014) may be due o he differen news measures considered. We focus on news in he form of surprise policy announcemens, while hey examine he effecs of news in he form of macroeconomic daa releases. 18

20 value agains hese individual currencies, while he effecs from he long-erm pah surprises are insignifican (excep for he yen). During he unconvenional period, boh shor- and long-erm pah surprises affec he dollar exchange rae o a similar exen for all currencies. 22 In addiion, for he four currencies considered, he dollar depreciaes by a magniude several imes larger during he unconvenional moneary policy period han during he convenional period. 3.2 LSAP1 and he aper period Given ha financial markes were subsanially impaired during he end of 2008 and early par of 2009, he unconvenional moneary policy decisions aken during ha ime could have had effecs on he value of dollar ha differed quie subsanially from hose during less urbulen imes in he pos-crisis period. Similarly, large movemens in asse prices also occurred during he apering period as he FOMC signaled and hen iniiaed a gradual decline in Treasury and morgage-backed securiies purchases ha ulimaely ended he LSAP3 program. In paricular, Chairman Bernanke s remarks on May 22, 2013 appear o have caugh marke paricipans off guard and led o subsanial flucuaions in bond, equiy, and currency markes worldwide. In his secion, we examine he exen o which our resuls are driven by he key announcemens during LSAP1 and he apering period of LSAP3 (i.e., from May 22, 2013 o he end of he program on Ocober 29, 2014) by adding dummy variables (D LSAP1, D Taper ) o equaion (1) ha isolae he effecs of hese wo periods of ineres. Specifically, we run he following regression: ST LT u ST LT S a TS PS PS D a PS PS i,, w 1, i , i 2 2 D a PS PS D a PS PS LSAP1 ST LT Taper ST LT 3, i 3 3 4, i 4 4 i, (2) 22 Wrigh (2012) repors ha he Canadian dollar, pound, and euro appreciae inerday by 0.56, and 1.09 percen, respecively, in response o a sandardized long-erm moneary surprise, implying a similar rank response order as we repor in Table 4. We canno make a direc numerical comparison o our resuls for individual currencies, since we do no have he informaion o conver his surprise measure ino bp erms. 19

21 Table 5 repors he resuls. Noe ha he coefficiens 1, 1, 1 reflec he effecs of arge, shor-erm pah, and long-erm pah surprises during he convenional period, while 2, 2 reflec he addiional effecs of shor-erm and long-erm pah surprises in he unconvenional period, while excluding he effecs of surprises during he LSAP1 and aper periods; he laer effecs are capured by he coefficiens 3, 3 and 4, 4, respecively. The main message is ha he shor- and long-erm pah surprises had larger effecs during LSAP1 and he apering period, wih hose following LSAP1 announcemens being paricularly persisen during he 24 hour response window. While he large effecs of unconvenional moneary policy during LSAP1 when financial markes were impaired has been addressed by ohers (e.g., Gagnon e al. 2011) and Krishnamurhy and Vissing-Jorgensen 2011, among ohers), our resuls indicae ha large effecs occurred during he apering period as well, when financial markes were operaing more normally. 23 Neverheless, Table 5 also indicaes ha, alhough he effecs are aenuaed and less persisen, he dollar sill responded significanly o policy surprises ouside of he LSAP1 and apering periods. In paricular, during hese oher phases of he Federal Reserve s purchasing programs during he unconvenional period he dollar depreciaed on average by bps in he 60 minues following shor-erm pah surprises, while he long-erm pah surprises led o a decline in he dollar s value of 6.44 bps. Thus, absracing from he LSAP1 and apering periods, he dollar depreciaed abou 1.3 imes more in he 60 minues following a moneary easing during he unconvenional period han during he convenional period. 23 I should be noed ha he policy surprises are ypically negaively signed during he aper period, indicaing moneary ighening. Hence he negaive coefficiens on he variables D Taper *PS ST, D Taper *PS LT are consisen wih posiive effecs on he exchange rae, i.e., an appreciaion of he dollar. 20

22 4. Robusness Analysis In his secion we subjec our benchmark resuls o several robusness checks. In paricular, we assess he role of he window size used o consruc our moneary surprise measures, he exclusion of long-erm pah surprises, alernaive break daes beween he convenional and unconvenional periods, he exclusion of unscheduled meeing announcemens, and nonnesed regressions for he convenional and unconvenional periods. 4.1 Wider surprise windows We firs consider he implicaions of using a wider window o consruc he convenional and unconvenional policy surprises, going from 10 minues before announcemens unil 60 (raher han 20) minues afer. The resuls are given in Table 6. Given his wider window for surprises, we repor he exchange rae effecs only for +60 minue and +24 hour response windows. Overall, we find ha our resuls are broadly robus o his alernaive measure of policy surprises, as he effecs of he shor- and long-erm pah surprises sill are much larger during he unconvenional period han during he convenional period. Looking more closely, i should be noed ha he shor-erm pah surprises during he convenional period end o have a much sronger impac on he exchange rae compared o hose repored in Table 3 wih he narrow window. This narrows he difference beween he exchange rae effec of moneary policy across periods. Overall, moneary easing during he unconvenional period leads o a depreciaion of he dollar afer 60 minues ha is abou wo and a quarer imes as large as ha during he convenional moneary period (where he laer includes he effec of arge surprises); his is lower han he roughly hree imes difference wih he narrow window resuls repored in Table 3. 21

23 4.2 Exclusion of long-erm pah surprises For he convenional period, he lieraure has generally emphasized wo ypes of moneary policy surprises, hose associaed wih unexpeced changes in he policy rae arge surprises -- and hose capuring he influence of FOMC communicaion on he fuure pah of relaively shor-erm ineres raes, such as he one-year eurodollar rae wha we erm shorerm pah surprises (see, e.g. Gürkayrkanak, Sack, and Swanson, 2005; Hausman and Wongswan, 2011)., By including long-erm pah surprises in our benchmark specificaion we allow for he possibiliy ha policymakers can direcly influence longer-erm ineres raes eiher via long-erm asse purchases or via forward guidance. However, given ha policymakers have less direc conrol over long-erm ineres raes, he idenificaion of policy surprises via his channel is possibly more uncerain. We now assess he robusness of our main finding by removing he long-erm pah surprises from he benchmark model alogeher. Ineresingly, we sill find ha moneary policy s impac on he dollar is wo o hree imes larger during he unconvenional han convenional period, as shown in Table 7. A convenional-period moneary announcemen ha included a one sandard deviaion decline in boh he arge and shor-erm pah surprises leads o a basis poin decline in he value of he dollar one hour following announcemen, whereas during he unconvenional period a one sandard deviaion fall in he shor-erm pah surprise generaes a dollar depreciaion of basis poins. Thus, our main finding is robus o he more ypical measuremen of policy surprises. 4.3 The December 16, 2008 announcemen As discussed earlier, he ransiion beween he convenional and unconvenional periods is difficul o pin down precisely. In our benchmark specificaion, we reaed Ocober 2008 as he end of he convenional period, before he FOMC s decision o fully lower he federal funds rae 22

24 o is effecive lower bound as i announced on December 16 laer ha year. As such, he December 16, 2008 FOMC saemen is somewha special as i conained informaion on forward guidance and on he FOMC s inenions o buy long-erm asses, in addiion o is decision o bring he federal funds rae down o is effecive lower bound. Thus, i is likely ha he announcemen s effecs on he dollar reflec he use of convenional as well as unconvenional policies. As a robusness check, we coninue o end he convenional period in Ocober 2008, bu add a dummy variable for he December 16, 2008 announcemen. As can be noed from panel A of Figure 1, he movemen in he dollar was large ha day, implying ha is inclusion in he convenional period of our baseline specificaion may affec he comparison of policy effeciveness across regimes. Our resuls in Table 8, however, indicae ha his is largely no he case. While we find ha he December 2008 FOMC saemen indeed had a very large impac on he exchange rae, he gis of our resuls go hrough under his alernaive specificaion. In paricular, we sill find ha he dollar depreciaed by roughly hree imes more following surprise easing during he unconvenional period as compared o he convenional period. We have also examined he robusness of our resuls o saring he unconvenional period in January 2009, hus eliminaing he announcemens regarding unconvenional policy in November and December 2008 and found similar resuls as wih our baseline specificaion. 24 All old, while he exac break beween he convenional and unconvenional periods is no clear cu, reasonable variaions leave our resuls essenially unchanged. 4.4 Unscheduled FOMC meeings Examining he effecs of he arge surprises on he dollar in he op panel scaers of Figure 2, i is apparen ha here are several large posiive surprises, roughly 40 basis poins in magniude, which had a much more mued impac on he exchange rae. These announcemens 24 These resuls are available upon reques. 23

25 can be raced o hree unscheduled FOMC meeings ha occurred on January 3, 2001, April 18, 2001, and January 22, The differenial effecs of inermeeing announcemens on asse markes have been noed in oher sudies. Fleming and Piazzesi (2005), for example, analyze moneary policy effecs over he period February 1994 o December 2004 using a sample ha includes hree of he episodes we examine April 18, 1994, January 3, 2001, and April 18, 2001 as well as Sepember 17, 2001, and Ocober 15, They find ha Treasury raes responded paricularly slowly o he announcemens on hese days. They sugges several reasons why inermeeing moves migh be imporan in explaining he marke s weak response: inermeeing arge rae easing surprises end o occur in relaively uncerain environmens, end o be larger, and may have a larger signaling componen han oher announcemens abou economic weakness, hereby dampening bond demand and he easing of long-erm raes, or alernaively hey may ake a longer ime o be digesed and processed by markes. Consequenly, he effec of policy surprises on he dollar during he convenional period, and hence our comparison wih he effecs during he unconvenional period, may be affeced by FOMC announcemens following unscheduled meeings. Therefore, as anoher robusness exercise, we remove he unscheduled meeings from he convenional period sample and repor he resuls in Table 9. As expeced, removing he unscheduled meeings implies a greaer impac of he arge surprises on he dollar, which now depreciaes by abou 14 basis poins in he 60 minues following announcemens compared o roughly 7 basis poins for our benchmark case in Table 3. However, he effecs of he shor-erm pah surprises are now somewha smaller. Taking hese wo effecs ino accoun, Table 9 indicaes ha he impac of a ypical moneary easing on he dollar during he unconvenional period sill remains several imes larger han ha during he convenional period. 24