The event-study activity puzzle

Transcription

1 The even-sudy aciviy puzzle Jeffrey R. Campbell, Jonas D. M. Fisher, Alejandro Jusiniano, and Leonardo Melosi Inroducion and summary 1 If news of higher fuure ineres raes reflecs a shif in he sance of moneary policy, hen we expec i o lower projecions for real gross domesic produc (GDP) growh and raise projecions for unemploymen. However, he daa poin o quie differen conclusions. When marke paricipans raise heir esimaes of fuure ineres raes afer a meeing of he Federal Reserve s moneary policy commiee, he Federal Open Marke Commiee (FOMC), privae forecasers end o raise heir projecions of real GDP growh and lower heir projecions for unemploymen. Campbell, Evans, Fisher, and Jusiniano (2012) documened his fac, and we labeled i he even-sudy aciviy puzzle in our NBER Macroeconomics Annual chaper (Campbell, Fisher, Jusiniano, and Melosi, 2017). In his aricle, we summarize an empirical exploraion of he effecs of FOMC saemens on financial marke expecaions of fuure ineres raes and privae forecasers expecaions of fuure macroeconomic performance ha appeared in Campbell e al. (2017). Our work resolves he even-sudy aciviy puzzle by demonsraing ha he FOMC moves marke expecaions of fuure ineres raes in par by ransmiing is views of fuure macroeconomic fundamenals. When hese are srong, privae forecasers revise heir projecions accordingly, while marke paricipans mark up heir expecaions of fuure ineres raes. How does he FOMC ransmi macroeconomic expecaions? The FOMC releases a saemen afer each of is meeings ha announces any desired changes o he federal funds rae and oher policy ools, describes is assessmen of curren macroeconomic circumsances, and possibly describes how furher developmens could influence is fuure policy choices. This discussion of possible fuure acions is called forward guidance. Perhaps he mos prominen example of forward guidance from before he Grea Recession was he use of he Commiee believes ha policy accommodaion can Economic Perspecives 3 / 2017

2 be removed a a pace ha is likely o be measured in he FOMC s pos-meeing saemens from May 2004 hrough November Afer he FOMC se is arge for he federal funds rae o is effecive lower bound in December 2008, i urned o forward guidance as one way o deliver addiional moneary accommodaion. 1 As demonsraed in he heoreical work of Krugman (1998) and Eggersson and Woodford (2003), expansionary forward guidance can improve curren macroeconomic oucomes if i lowers households desired savings by increasing heir expecaions of fuure inflaion and oupu. According o he heory, such changes in expecaions can be induced by he FOMC making a credible commimen o keep he fuure federal funds rae lower han i would oherwise do given he prevailing macroeconomic condiions. Campbell e al. (2012) called his Odyssean forward guidance. 2 Afer is Augus 2011 meeing, he FOMC saed ha macroeconomic condiions are likely o warran excepionally low levels for he federal funds rae a leas hrough mid Similarly, i saed afer is December 2012 meeing ha ineres raes would remain excepionally low a leas as long as he unemploymen rae remains above 6-1/2 percen, inflaion beween one and wo years ahead is projeced o be no more han a half percenage poin above he Commiee s 2 percen longer-run goal, and longer-erm inflaion expecaions coninue o be well anchored. 4 Boh saemens could be viewed as Odyssean forward guidance. 2 While he heoreical possibiliy for appropriae forward guidance o improve curren macroeconomic oucomes is clear, he evidence does no unambiguously suppor his conjecure. Ben Zeev, Gunn, and Khan (2015) and D Amico and King (2016) boh use economeric models and money marke fuures prices o idenify unexpeced movemens in expeced fuure ineres raes caused by FOMC forward guidance; hey boh find ha oupu and inflaion rise presenly and in he fuure when expeced fuure raes fall. This is exacly as he sandard New Keynesian model predics when changes o expeced fuure ineres raes represen exogenous adjusmens o he sance of moneary policy. On he oher hand, Campbell e al. (2012) measure unexpeced movemens in fuure ineres raes wih changes in money marke fuures raes on he daes of FOMC meeings. When hey relae revisions in privae secor forecass of oupu growh and unemploymen o hese forward-guidance shocks, hey find ha decreases in fuure expeced ineres raes lower expecaions of oupu and raise expeced unemploymen. This is he even-sudy aciviy puzzle menioned above. Quanifying he puzzle To quanify he even-sudy aciviy puzzle, we relaed changes in privae secor forecass for GDP growh and unemploymen over monhs conaining an FOMC meeing wih he change in he expecaion of he federal funds rae four quarers ino he fuure, on he day of he FOMC announcemen. By consrucion, his even-sudy of how moneary policy announcemens affec expecaions of fuure moneary policy conrols for all informaion available o marke paricipans a he close of business on he previous day. Since FOMC announcemens are usually he mos imporan financial news on he day hey occur, hey should dominae our measured changes in expecaions of ineres raes. Our measures of privae secor forecass come from he Blue Chip Economic Indicaors, a monhly survey of privae forecasers. We use is consensus forecas, which averages all submied forecass. Is respondens provide quarerly forecass for real GDP growh and he unemploymen rae for he curren and nex calendar years. Therefore, he survey repors privae forecass for he curren and nex four quarers wihin one monh of every FOMC meeing. For each of hese observables, we esimae he coefficiens of a simple linear model, 4 e = α+ β i + ε m d Economic Perspecives 3 / 2017

3 Here, m e is he change in he Blue Chip forecas for some variable over he monh conaining a paricular FOMC meeing, d i 4 is he change in expecaion of he federal funds rae four quarers hence on he day of he FOMC meeing, and ε is a residual ha accouns for all movemens in m e ha are unrelaed o d i 4. We esimaed he model s unknown coefficiens α and β using daa from all FOMC meeings beween 1994 and 2010 inclusive. 5 For each of he privae expecaions, we esimaed he unknown coefficiens α and β using he mehod of ordinary leas squares. This mehod makes he sum of he squared implied values for ε as small as possible. Table 1 repors he esimaed coefficiens muliplying d i 4. When hese are posiive, he daa indicae ha increases in he ineres rae expeced four quarers in he fuure are associaed wih increases in he given expecaion. Of course, sampling error could make any of our coefficien esimaes non-zero even hough he coefficien s rue value is zero. For each regression, we esed he null hypohesis ha he rue value of β equals zero agains he alernaive ha i is non-zero using mehods described in Campbell e al. (2017). When he es rejecs he null hypohesis a he convenional levels of saisical significance of 5 percen and 1 percen, we mark he associaed esimae of β wih ** and ***, respecively. A smaller level of saisical significance means ha he chances of a model wih a rue coefficien of zero generaing he esimaed non-zero coefficien are lower. Nex o each esimaed coefficien is he model s R 2 measure. This varies beween 0 and 1 and gives he fracion of variance in m e ha he model assigns o movemens in d i 4. 3 The resuls in able 1 illusrae he even-sudy aciviy puzzle. Raising he expecaion of he federal funds rae four quarers hence raises privae secor forecass of GDP growh and lowers unemploymen forecass for he nex four quarers. For example, a 1 percenage poin increase in d i 4 raises he expecaion of annualized GDP growh in he nex quarer by 66 basis poins and lowers he expecaion of nex quarer s unemploymen rae by 21 basis poins. Only he changes in expecaions for he nex quarer are saisically significan a convenional levels. For expeced GDP growh, he coefficiens esimaed magniudes a horizons of wo o four quarers vary from 27 o 6 basis poins. For unemploymen, he coefficien esimaes for hose longer horizons are comparable o hose for he nex quarer, bu he coefficiens are measured relaively imprecisely. The R 2 measures in able 1 all indicae ha mos variance in expecaions revisions is due o facors oher han changes in expeced ineres raes. Tha is, hese esimaes indicae ha FOMC forward guidance was no a key driver of privae expecaions during he sample period. TABLE 1 The even-sudy aciviy puzzle Change in privae forecas of β (coefficien muliplying d i 4 ) R 2 GDP growh Nex quarer 0.66*** 0.03 Two quarers hence Three quarers hence Four quarers hence Unemploymen Nex quarer 0.21** 0.05 Two quarers hence Three quarers hence Four quarers hence Noe: ** and *** indicae ha a es rejecs he null hypohesis a he 5 percen and 1 percen level, respecively. Sources: Auhors calculaions based on daa from he Board of Governors of he Federal Reserve Sysem and Blue Chip Economic Indicaors. Economic Perspecives 3 / 2017

4 Campbell e al. (2012) speculaed ha heir resuls arose from a failure of heir measured shocks o represen ruly exogenous changes in he sance of moneary policy. In general, he FOMC could surprise marke paricipans in wo disinc ways. Firs, i migh ruly change he sance of moneary policy based on is members judgmens abou commonly observed macroeconomic fundamenals. Second, i could signal ha i has received informaion abou fuure macroeconomic developmens ha is unavailable o marke paricipans. Alhough FOMC paricipans have very lile access o macroeconomic daa ha are unavailable o he public, he Federal Reserve Sysem s saff migh give he FOMC proprieary inerpreaive informaion. If his is so, hen ransmiing informaion ha he economy would run ho in he near fuure, for insance, would lead marke paricipans o increase heir forecass of boh ineres raes and oupu. Campbell e al. (2012) labeled such forward guidance ha simply repors informaion on fuure macroeconomic developmens and discussed possible FOMC reacions o hose developmens as Delphic. 6 If mos FOMC forward guidance is Delphic, hen he even-sudy aciviy puzzle can be undersood as he opimal reacions of marke paricipans and privae forecasers o prognosicaions abou fuure macroeconomic developmens in FOMC saemens. Alhough his Delphic resoluion o he even-sudy aciviy puzzle seems plausible, Campbell e al. (2012) provided no evidence o suppor i. In Campbell e al. (2017), we subjeced his hypohesis o empirical scruiny using measures of he FOMC s privae informaion abou shor-run macroeconomic performance. For his, we proceeded in wo seps. Firs, we inferred he change in ineres raes ha can be direcly explained by hese measures of privae informaion, and we aribued his esimae o FOMC shor-run Delphic forward guidance. Second, we quanified he exen o which shor-run Delphic forward guidance explains he evensudy aciviy puzzle by esimaing how his measure influenced changes of privae secor expecaions of GDP growh and unemploymen. 4 We consruced our measures of FOMC privae informaion by subracing consensus privae secor forecass for Consumer Price Index (CPI) inflaion, real GDP growh, and he unemploymen rae for he curren and nex four quarers from analogous forecass prepared by Federal Reserve Board saff for he FOMC. These forecass are made available o FOMC paricipans (Governors of he Federal Reserve Sysem and he 12 Reserve Bank presidens), bu hey are kep ou of he public domain for five years afer heir creaion. Therefore, any measurable effec hese have on innovaions o financial markes expecaions of fuure ineres raes should operae only hrough he conen of he FOMC saemen. To examine he influence of hese forecas differences on changes in financial marke expecaions of he federal funds rae, we esimaed he coefficiens of a simple linear model: y y u u d i 4 π π = µ + γ d + γ d + γ d + η. y u Here d, d, and d π represen he differences beween he Board saff and Blue Chip forecass for oupu growh, unemploymen, and inflaion, respecively. These include differences for he curren meeing and from he meeing ha preceded i. We included his seemingly sale informaion because FOMC paricipans migh require ime o form he consensus abou he meaning of new informaion required for is discussion in a meeing saemen. 7 We esimaed he model s unknown coefficiens γ y, γ u, and γ π wih he mehod of ordinary leas squares, as described earlier. Table 2 gives he resuls of his esimaion. For he sake of parsimony, i repors he resuls of exclusion ess raher han he esimaed coefficiens hemselves. These es he null hypohesis ha he coefficiens muliplying each variable in a group all equal zero. For example, he exclusion es for all FOMC privae informaion abou GDP growh uses he null hypohesis ha all elemens of γ y equal zero. Jus as before, he superscrips ** and *** indicae ha a es rejecs he null hypohesis a he 5 percen and 1 percen level, respecively. 8 The exclusion es rejecs he null hypohesis ha all of he forecas differences are irrelevan a Economic Perspecives 3 / 2017

5 TABLE 2 Revelaion of Greenbook forecass: Exclusion ess Variables excluded Tes saisic All 33.33** Curren meeing s Previous meeing s 19.34*** GDP forecas differences 19.25*** Unemploymen forecas differences 7.79 Inflaion forecas differences 1.78 Noe: ** and *** indicae ha a es rejecs he null hypohesis a he 5 percen and 1 percen level, respecively. Sources: Auhors calculaions based on daa from he Board of Governors of he Federal Reserve Sysem and Blue Chip Economic Indicaors. he 5 percen level. Ineresingly, he es does no rejec he null hypohesis ha he forecas differences associaed wih he curren meeing are irrelevan. Insead, i is he previous meeing s forecas differences ha help explain he change in expeced fuure ineres raes following he curren meeing. As suggesed earlier, we inerpre his resul as reflecing he ime required for FOMC paricipans o process new informaion. Among he hree macroeconomic variables included, he exclusion ess indicae ha i is differences beween he Board saff and Blue Chip forecass for near-erm GDP growh ha have he mos influence on financial marke paricipans expecaions of fuure ineres raes. 9 These empirical resuls srongly sugges ha he FOMC rouinely ransmis he Board saff s view of fuure macroeconomic condiions o he public hrough FOMC saemens. Addiionally, hey provide he abiliy o conrol for hese ransmissions when assessing he influence of saemen-induced changes in expeced fuure ineres raes on privae macroeconomic forecass. For his, we use he esimaed values of γ y, γ u, and γ π o consruc he change in privae ineres rae expecaions expeced given he differences beween he Board saff and Blue Chip forecass. Call his esimae of he FOMC s shor-run Delphic forward guidance f ˆ : f ˆ y y u u µ ˆ + γˆ d + γˆ d + γˆ π d π. Here, he has indicae ha he linear model s coefficiens are esimaed from daa. Wih his, we can wrie 5 4 = ˆ + ˆ. d i f η The model s esimaed residual erm is ηˆ. By consrucion, is correlaion wih f ˆ in our sample equals zero. I represens all influences on d i 4 ha are no due o shor-run Delphic forward guidance. These include long-run Delphic forward guidance and any Odyssean forward guidance. We can now assess how hese wo componens of d i 4 shape revisions o privae secor expecaions. There is no good ex ane reason o expec he shor-run Delphic forward guidance in f ˆ o influence privae macroeconomic expecaions in he same way as he various messages embodied in ηˆ do. To accoun for hese possible differences, we can generalize he linear model used o creae able 1 by allowing hese wo componens of d i 4 o have disinc coefficiens. Tha is, e = α+ β fˆ η + β ηˆ + ε. m f Table 3 repors he resuls of esimaing his model wih he mehod of leas squares. Each row repors he model s esimaed coefficiens and R 2 measure for one macroeconomic expecaion. The superscrips ** and *** indicae he saisical significance of he coefficien esimaes a he 5 percen and 1 percen level, respecively. The esimaes clearly indicae ha he FOMC s shor-run Delphic forward guidance moves ineres rae expecaions and projecions of GDP growh in he same direcion. A 1 percenage poin increase in f ˆ is associaed wih a 2.88 percenage poin increase in expeced GDP growh nex quarer, a 1.75 percenage poin increase wo quarers hence, and a 0.66 percenage poin increase hree quarers Economic Perspecives 3 / 2017

6 TABLE 3 Blue Chip forecass responses o decomposed moneary policy Delphic componen (ˆf ) Residual (ˆη ) R 2 GDP growh Nex quarer 2.88** Two quarers hence 1.75*** Three quarers hence 0.66** Four quarers hence Unemploymen Nex quarer 0.88*** Two quarers hence 1.08*** Three quarers hence 1.28*** Four quarers hence 1.18*** Noe: ** and *** indicae ha a es rejecs he null hypohesis a he 5 percen and 1 percen level, respecively. Sources: Auhors calculaions based on daa from he Board of Governors of he Federal Reserve Sysem and Blue Chip Economic Indicaors. hence. All of hese responses are saisically significan a convenional levels. Similarly, he FOMC s shor-run Delphic forward guidance moves expeced fuure ineres raes and unemploymen projecions in he opposie direcions. For example, a 1 percenage poin increase in f ˆ decreases he expeced unemploymen rae in four quarers by 1.18 percenage poins. The R 2 measures for hese linear models are beween 0.17 and Tha is, he FOMC s shor-run Delphic forward guidance accouns for abou one-fifh of he monhly revisions o unemploymen expecaions. 6 The esimaed residual erm ηˆ should be free of (measurable) shor-run Delphic forward guidance. The esimaed coefficiens indicae ha his is indeed so. All bu one of hem has he inuiive sign associaed wih exogenous changes o he sance of moneary policy. Tha is, hese esimaes are consisen wih he predicion ha expecaions of economic aciviy fall when he FOMC signals a ighening in fuure moneary policy beyond wha he privae secor would ordinarily forecas given he incoming macroeconomic daa. However, hey are relaively small in magniude and none of hem are saisically significan. Therefore, here remains considerable scope for fuure research o invesigae his hypohesis. Conclusion In Campbell e al. (2017), we presen a more deailed analysis of hese resuls, including a discussion of wheher forward guidance subsanially affecs Treasury erm premiums. Overall, we find ha accouning for he FOMC s ransmission of shor-run Delphic forward guidance in is meeing saemens resolves he even-sudy aciviy puzzle of Campbell e al. (2012). This resul is imporan in iself for designing cenral bank communicaions policy, and i clears he way for fuure research ino Odyssean forward guidance. Economic Perspecives 3 / 2017

7 NOTES 1 In a speech a he London School of Economics on January 13, 2009, Chairman Ben Bernanke was explici abou he value of forward guidance when policy is consrained by he effecive lower bound: Alhough he federal funds rae is now close o zero, he Federal Reserve reains a number of policy ools ha can be deployed agains he crisis. One imporan ool is policy communicaion. Even if he overnigh rae is close o zero, he Commiee should be able o influence longer-erm ineres raes by informing he public s expecaions abou he fuure course of moneary policy. To illusrae, in is saemen afer is December meeing, he Commiee expressed he view ha economic condiions are likely o warran an unusually low federal funds rae for some ime. To he exen ha such saemens cause he public o lenghen he horizon over which hey expec shor-erm raes o be held a very low levels, hey will exer downward pressure on longer-erm raes, simulaing aggregae demand. Read he complee ex of he speech a hps:// 2 Odyssean forward guidance consiss of cenral bankers saemens ha bind hem o fuure courses of acion. Jus as Odysseus bound himself o his ship s mas so he could enjoy he sirens song wihou succumbing o he empaion o drown himself while swimming oward hem, a cenral banker can improve economic oucomes by publicly commiing o a plan ha uses expecaions of subopimal fuure oucomes o improve curren economic condiions. 3 See hps:// 4 See hps:// 5 Our sample begins wih 1994 because he FOMC issued saemens only inermienly before hen, and i ends in 2010 because he Greenbook forecass we used o measure he FOMC s privae informaion (discussed in he ex) were no in he public domain hereafer. 6 Like he oracle of Delphi, he forward guidance predics he fuure bu promises no specific policy acions. 7 7 For each FOMC meeing, we compare Board saff and Blue Chip forecass for he curren quarer and he nex four quarers. Since we also include forecas differences from he previous meeing in our model, his would make for 30 coefficiens o esimae. There were only 148 FOMC meeings during our sample period, so including all of hese variables would pu us in danger of overfiing he model. To avoid his danger, we reduced he forecas differences for each meeing and each variable o wo variables using a saisical procedure called facor analysis. We give he deails of his procedure in Campbell e al. (2017). 8 See Campbell e al. (2017) for he deails. 9 As a poin of comparison, we showed in Campbell e al. (2017) ha our measures of FOMC privae informaion have no saisically deecable influence on he curren policy rae. REFERENCES Ben Zeev, Nadav, Chrisopher Gunn, and Hashma Khan, 2017, Moneary news shocks, Carleon Economic Papers, Carleon Universiy, CEP 15-02, revised February 27, 2017 (originally issued March 2015). Campbell, Jeffrey R., Charles L. Evans, Jonas D. M. Fisher, and Alejandro Jusiniano, 2012, Macroeconomic effecs of Federal Reserve forward guidance, Brookings Papers on Economic Aciviy, Spring, pp Campbell, Jeffrey R., Jonas D. M. Fisher, Alejandro Jusiniano, and Leonardo Melosi, 2017, Forward guidance and macroeconomic oucomes since he financial crisis, in NBER Macroeconomics Annual 2016, Marin Eichenbaum and Jonahan A. Parker (eds.), Vol. 31, Chicago: Universiy of Chicago Press, forhcoming. Economic Perspecives 3 / 2017

8 D Amico, Sefania, and Thomas B. King, 2015, Wha does anicipaed moneary policy do?, Federal Reserve Bank of Chicago, working paper, No , November. Eggersson, Gaui B., and Michael Woodford, 2003, The zero bound on ineres raes and opimal moneary policy, Brookings Papers on Economic Aciviy, No. 1, pp Krugman, Paul R., 1998, I s baaack: Japan s slump and he reurn of he liquidiy rap, Brookings Papers on Economic Aciviy, No. 2, pp Jeffrey R. Campbell is a senior economis and research advisor, Jonas D. M. Fisher is a vice presiden and he direcor of macroeconomic research, Alejandro Jusiniano is a senior economis and research advisor, and Leonardo Melosi is a senior economis in he Economic Research Deparmen a he. In addiion, Campbell is affiliaed wih CenER, Tilburg Universiy; and Jusiniano is affiliaed wih he Paris School of Economics Economic Perspecives is published by he Economic Research Deparmen of he Federal Reserve Bank of Chicago. The views expressed are he auhors and do no necessarily reflec he views of he Federal Reserve Bank of Chicago or he Federal Reserve Sysem. Charles L. Evans, Presiden; Daniel G. Sullivan, Execuive Vice Presiden and Direcor of Research; David Marshall, Senior Vice Presiden and Associae Direcor of Research; Spencer Krane, Senior Vice Presiden and Senior Research Advisor; Daniel Aaronson, Vice Presiden, microeconomic policy research; Jonas D. M. Fisher, Vice Presiden, macroeconomic policy research; Rober Cox, Vice Presiden, markes eam; Anna L. Paulson, Vice Presiden, finance eam; William A. Tesa, Vice Presiden, regional programs; Lisa Barrow, Senior Economis and Economics Edior; Helen Koshy and Han Y. Choi, Ediors; Julia Baker, Producion Edior; Sheila A. Mangler, Ediorial Assisan. Economic Perspecives aricles may be reproduced in whole or in par, provided he aricles are no reproduced or disribued for commercial gain and provided he source is appropriaely credied. Prior wrien permission mus be obained for any oher reproducion, disribuion, republicaion, or creaion of derivaive works of Economic Perspecives aricles. To reques permission, please conac Helen Koshy, senior edior, a or Helen.Koshy@chi.frb.org. ISSN Economic Perspecives 3 / 2017