Identifying Unconventional Monetary Policy Shocks *

Transcription

1 Idenifying Unconvenional Moneary Policy Shocks * June 216 Kiyoaka NAKASHIMA Masahiko SHIBAMOTO Konan Universiy Kobe Universiy Koji TAKAHASHI Universiy of California, San Diego Absrac This paper proposes a novel mehod for idenifying unconvenional moneary policy shocks. Specifically, our idenifying mehod incorporaes he movemen in wo unconvenional moneary policy indicaors, namely he size and composiion of he cenral bank s balance shee, afer is policy decisions. Under some resricions imposed in he vecor auoregressive model, we idenify wo unconvenional policy shocks, quaniaive and qualiaive shocks, as news shocks ha bes porend he curren and fuure pahs of he unconvenional policy indicaors in response o he moneary policy shocks. We find ha qualiaive easing shocks have expansionary effecs on real economy, while quaniaive easing shocks have conracionary ones. Such differen effecs beween he wo unconvenional moneary policy shocks would arise from agen s risk aking. Keywords: quaniaive easing; qualiaive easing; convenional moneary policy; vecor auoregressive model; news shock. JEL classificaion: E52, E58. * We are graeful for he helpful commens and discussions by Kosuke Aoki and Kenneh Kuner. This sudy received financial suppor in he form of a Grans-in-Aid from he Japanese Minisry of Educaion and Science. Kiyoaka Nakashima, Faculy of Economics, Konan Universiy, 8-9-1, Okamoo, Higashinada, Kobe, , Japan, kiyoaka@cener.konan-u.ac.jp. Correspondence o Masahiko Shibamoo, Research Insiue for Economics and Business Adminisraion, Kobe Universiy, 2-1, Rokkodai, Nada, Kobe, , Japan, shibamoo@rieb.kobe-u.ac.jp. Koji Takahashi, Deparmen of Economics, Universiy of California, San Diego, 95 Gilman Dr., La Jolla, CA, , Unied Saes, kakahas@ucsd.edu.

2 1. Inroducion Cenral banks have several moneary policy opions, even under he zero ineres rae lower bound (see Bernanke and Reinhar (24)). For example, since February 1999, he Bank of Japan (BOJ) has newly developed is so-called unconvenional moneary policy. Then, by seing he argeed overnigh call rae (OCR) o almos %, he BOJ adoped a quaniaive easing policy in March 21. Under his policy framework, he moneary base (MB), or he size of he BOJ s balance shee, has expanded a an OCR of % by enlarging he excess reserves in he BOJ s curren accoun (see Figure 1). I was in June 26 ha he quaniaive easing policy was lifed. The argeed rae, however, has remained well below.5% since hen. In April 213, he BOJ inroduced is quaniaive and qualiaive easing policy, furher deepening he unconvenional policy framework no simply by enlarging he size of is balance shee, bu also by increasing he raio of unconvenional asses on is balance shee. 1 Such an unconvenional policy framework, characerized by he increase in he size of he cenral bank s balance shee and change in is composiion a exremely low policy-argeed ineres raes, has also been followed by cenral banks in indusrialized counries such as he Unied Saes, he Unied Kingdom, and hose in he Euro area. Alhough he acual implemenaion of he unconvenional moneary policy in many counries has simulaed empirical research on unconvenional policy effecs using he srucural vecor auoregressive (VAR) model, is effecs on he real economy are sill dispuable parly because of he difficuly in he idenificaion of exogenous unconvenional policy shocks. 2 A major difficuly for assessing unconvenional policy effecs by using VAR analysis involves he issue of which financial variables should be used as he moneary policy indicaors ha precisely reflec he cenral bank s policy decisions in he unconvenional moneary policy; in oher words, how we should associae he moneary policy indicaors wih he exogenous componens of he unconvenional moneary policy. Some previous research has assumed ha he MB 1 See Shirasuka (21) and Ueda (212) for a deailed explanaion of unconvenional asses in Japan. 2 See Joyce e al. (212) for a survey of he empirical research on unconvenional policy effecs. 1

3 represens is policy sance, hus uilizing he reduced form VAR innovaions of he MB as he exogenous componens of he unconvenional moneary policy (Honda e al. (213)). This empirical sraegy is a simple exension of he Cholesky decomposiion approach o esimae he effecs of he convenional moneary policy in which cenral banks conrol shor-erm nominal ineres raes (Bernanke and Blinder (1992)). The oher srand of empirical research for unconvenional policy effecs has pursued a sraegy ha does no require exracing exogenous policy componens from he cenral bank s policy indicaors (Kapeanios e al. (212) and Gambacora e al. (214)). This sraegy assumes ha unconvenional moneary policy shocks can be inegraed ino one unobserved policy shock and hereby imposes resricions on he impulse responses of macroeconomic variables including he MB o unobserved policy shocks. However, aking ino accoun he fac ha cenral banks conrol for some policy variables ogeher in a low ineres rae environmen, he above wo empirical sraegies would be based on oo srong assumpions o assess unconvenional policy effecs. Given ha he BOJ has used differen policy ools, namely he size and composiion of is balance shee as well as is policy rae, for he case of he Japanese unconvenional moneary policy discussed above, i has sough o implemen moneary policy in a low ineres rae environmen in erms of hree dimensions: quaniaive and qualiaive easing and convenional policy rae seing. In his paper, we inroduce a novel idenificaion approach o separaely idenify hese hree moneary policy shocks quaniaive and qualiaive easing and convenional moneary policy shocks o disenangle he causal effecs beween he BOJ s hree policy indicaors and oher macroeconomic variables. To his end, we firs confirm he movemen in he moneary policy indicaors afer he BOJ s policy decisions. Then, we consider he idenificaion sraegy used o overcome he issues of unconvenional moneary policy shocks in hree aspecs: he variaion of policy ools, he endogeneiy of he moneary policy indicaors, and unconvenional policy shocks as news shocks. 3 3 Noe ha our sraegy canno disinguish anicipaed shocks from unanicipaed ones. Raher, we incorporae anicipaed shocks ino our model wih unanicipaed ones, while previous sudies using a VAR model ofen inerpre all shocks as unanicipaed shocks inenionally or uninenionally. 2

4 The fundamenal issue o be considered in idenifying moneary policy shocks is when and where hey primarily appear in an economy. In he case of Japan, he BOJ decides is policy scheme a is moneary policy meeing (M), which used o be held once or wice a monh, and publicly saes is policy decision jus afer. In accordance wih his acual policy implemenaion, we exploi he idea ha he cenral bank s policy shocks are refleced in asse prices as marke paricipans expecaion revisions change immediaely following is public saemens; in oher words, we ake up he moneary policy surprises in financial markes, or he expecaion revisions of marke paricipans in financial markes, induced by he BOJ s saemens abou policy decisions. As long as we correcly obain he moneary policy surprises, we can use hem as he causally relevan facors of he reduced form VAR innovaions o idenify he causal relaionships beween he BOJ s moneary policy shocks and macroeconomic variables. From his poin, he convenional mehod in which one considers he reduced form VAR innovaions of he moneary policy indicaors as policy shocks would lead o an erroneous assessmen of moneary policy effecs because i has he poenial shorcoming of esimaing spurious causaion among he innovaions of policy indicaors and oher macroeconomic variables. Indeed, Gerler and Karadi (213) and Shibamoo (216) poined ou his ype of moneary policy effec produced by using he reduced form VAR innovaions of policy raes. Moreover, given he curren moneary policy of imposing and mainaining he sabilizaion of he financial sysem, he VAR innovaions of he unconvenional moneary policy indicaors (i.e., he size and he composiion of he cenral bank s balance shee) would endogenously respond o he underlying financial variables lef ou of he VAR sysem. From his analyical viewpoin, some recen empirical sudies of moneary policy effecs on an asse s marke prices use high-frequency daily rading daa on asse prices. For example, Kuner (21), Cochrane and Piazzesi (22), Gürkaynak e al. (25a), and Campbell e al. (212) consruced moneary policy surprises in federal funds or one-monh Eurodollar fuures ha occurred on Federal Fund Open Marke Commiee 3

5 daes. 4 To examine financial marke responses o exogenous moneary policy in Japan, Honda and Kuroki (26) consruced moneary policy surprises in hree-monh euro-yen fuures ha occurred on he BOJ s M daes from 1989 o This empirical sraegy is sill useful for idenifying he BOJ s moneary policy shocks under he unconvenional policy regime; however, we canno simply follow i because i uses shor-erm ineres rae fuures, which have hardly changed since he BOJ s inroducion of is unconvenional moneary policy. Based on he foregoing, in his paper we propose more comprehensive indicaors for moneary policy shocks consruced from high-frequency daily rading daa on he major financial markes jus before and jus afer he BOJ s public saemen abou he M. In conras o previous sudies, we do no focus on changes in a paricular asse marke, bu raher exploi all informaion on changes in he major financial markes as heir common facors obained using he principal componen analysis mehod. Anoher idenifying issue is how he moneary policy indicaors respond o policy changes. As discussed above, previous research on unconvenional policy effecs based on VAR analysis has eiher assumed he reduced form VAR innovaions of he MB as he unconvenional policy shock or imposed resricions on he impulse responses of he VAR variables o one unobserved underlying unconvenional policy shock. Irrespecive of he difference in mehodology, hese wo approaches share he common assumpion ha all moneary policy shocks o he MB are anicipaed and conclude ha unconvenional policy shocks yield favorable effecs on he macroeconomy. Their idenificaion is, however, implausible in erms of he essenial naure of he quaniaive and qualiaive easing policy. Concerning he dynamics of he unconvenional and convenional policy indicaors, he size and composiion of he cenral bank s balance shee do no reflec he policy changes of cenral banks immediaely afer an announcemen, while is policy rae basically does. In oher words, he size and 4 See also Sock and Wason (212) for a deailed survey of his empirical sraegy using moneary policy surpirses, namely changes in asse marke prices, ha occur on Federal Fund Open Marke Commiee. 5 Gagnon e al. (211) and Joyce e al. (211) used an even sudy analysis o invesigae moneary policy effecs on an asse s marke price in he Unied Saes and he Unied Kingdom. Ueda (212) esimaed financial marke responses o he BOJ s moneary policy by using an even sudy analysis. 4

6 composiion hardly respond immediaely following moneary policy changes, bu raher move so gradually ha heir arge levels are realized afer cenral banks publicly announce changes in he arge level. By conras, he policy rae moves so quickly and subsanially in response o policy changes ha is arge level is realized immediaely afer cenral banks public saemens. Given he naures of hese policy indicaors, we idenify wo unconvenional moneary policy shocks relaing o he size and composiion of he BOJ s balance shee as news shocks ha bes presage heir curren and fuure pahs. We also idenify one convenional moneary policy shock as a shock ha has an insananeous impac on he BOJ s policy rae, bu is no explained by unconvenional policy shocks. These idenifying resricions reflec ha cenral banks seek o gradually achieve heir arge levels for unconvenional policy measures afer policy decisions and implemen he unconvenional moneary policy when hey canno pursue he convenional policy opion of lowering shor-erm nominal ineres raes below zero (Bernanke and Reinhar (24)). By idenifying quaniaive, qualiaive, and convenional policy shocks, we provide evidence ha he quaniaive easing shock involving he increase in he size of he BOJ s balance shee does no have favorable effecs on real economic aciviy, alhough i significanly decreases he long-erm nominal ineres rae. On he conrary, he qualiaive easing shock involving he increase in he BOJ s unconvenional asse raio o is oal asses as well as he convenional policy shock involving an immediae decrease in he policy rae yield real effecs. The remainder of his paper is organized as follows. Secion 2 analyzes he movemen in each policy indicaor in response o he BOJ s acual policy changes and our moneary policy shock measures. Secion 3 proposes an idenificaion mehod considering he source of he origin of policy shocks and incorporaing he movemen in he moneary policy indicaors. Secion 4 repors he empirical resuls for he idenified moneary policy shocks and explores several implicaions given our empirical findings. Finally, Secion 5 provides concluding commens. The Appendix explains he deailed procedure for consrucing he moneary policy shocks. 5

7 2. Convenional and Unconvenional Moneary Policy Indicaors In his secion, we examine he movemen in he moneary policy indicaors in response o policy shocks. By doing so, we show he necessiy of our new mehod o idenify he moneary policy shocks relaing o each of he policy indicaors, using he srucural VAR approach. The convenional policy indicaor is he OCR, while he unconvenional policy indicaors are he MB and he composiion raio of he BOJ s unconvenional asses o is oal asses (ermed CO hereafer). The MB, or he size of he BOJ s balance shee, is he quaniaive policy indicaor and he composiion raio is he qualiaive policy indicaor in he framework of he BOJ s unconvenional moneary policy (Shirasuka (21) and Ueda (212)) Basic Saisics of he Moneary Policy Indicaors In his secion, we begin by overviewing he basic ime series properies of he moneary policy indicaors. Table 1 repors he es resuls for he persisence of he moneary policy indicaors on he basis of he following regressions: p I = I, ι + I, p p + p= 1 I a a I v Q I = b I, ι + b I, q P q + q= 1 I M I v,, (1) where I denoes he moneary policy indicaors. The auoregressive componens in each equaion involve he persisence of each policy indicaor and he augmened Dickey Fuller es. The augmened Dickey Fuller es saisics indicae ha boh he MB and he CO are nonsaionary variables, while he OCR is a saionary variable, albei wih long persisence. The variance raio es saisics indicae ha he MB has he saisical feaure ha once i changes in one direcion, i coninuously moves in he same direcion. By conras, for he CO and OCR, he ransiory and permanen componens play an imporan role in deermining heir movemen. 6

8 Table 2 shows he esimaed correlaions among he moneary policy indicaors. The OCR and MB have a negaive correlaion in a heoreically consisen manner. The OCR and CO also have a negaive correlaion, possibly because he BOJ increases no only convenional asses bu also unconvenional ones when lowering he shor-erm ineres rae in he unconvenional moneary policy. The level variables of he MB and CO have a posiive correlaion, indicaing ha he wo variables share an increasing rend, as shown in Figure 1. This finding implies ha he BOJ sough o increase he purchases of unconvenional asses, especially long-erm Japanese governmen bonds, in he process of achieving is quaniaive arge for moneary policy. 6 By conras, he log-differenced variables have a negaive correlaion, indicaing ha he BOJ increases convenional asses o enlarge he size of is balance shee in he shor-run adjusmen of he MB and CO under he unconvenional moneary policy Even Sudy Illusraion In his subsecion, we conduc an even sudy analysis of he response of he policy indicaors o he inroducion of he new moneary policy schemes. The aim here is o illusrae more concreely he degree of variaion in he response of he moneary policy indicaors o acual policy changes and he imporance of inroducing a srucural analysis. Figures 2 o 5 show he ime pah of he MB, CO, and OCR during he following four periods: 1) he inroducion of he zero ineres rae policy in February 1998, 2) he inroducion of he quaniaive and qualiaive easing policy in April 213, 3) he end of he zero ineres rae policy and inroducion of quaniaive easing in March 21, and 4) he end of quaniaive easing and recurrence of he convenional moneary policy hrough shor-erm ineres rae conrol in July 26. All he figures share he common feaure ha he shor-erm ineres rae (OCR) immediaely responded o he BOJ s moneary policy enering he new regimes, while he MB and CO, namely he quaniaive and qualiaive indicaors, responded more 6 In fac, he BOJ has ofen faced an underbidding of securiies purchases, where bids fall shor of offers, in is open marke operaions o injec liquidiy, which may be because of he lack of marke demand for liquidiy. 7

9 coninuously and gradually. Indeed, as shown in Figure 2, he inroducion of he zero ineres rae policy in 1998 immediaely decreased he OCR, whereas i coninuously increased he MB and CO afer heir immediae, bu small responses o his policy change. 7 This even sudy illusraion is consisen wih hose obained in he previous subsecion in ha he unconvenional indicaors move slowly and laer in response o policy changes, while he policy rae moves immediaely. Figure 3 indicaes ha he inroducion of he quaniaive and qualiaive easing policy in 213 saw he MB and CO coninuously and gradually increasing, whereas i had almos no effec on he OCR because his shor-erm ineres rae had been subsanially in he zero lower bound since December 28 when he BOJ lowered is arge rae for he OCR from.3% o.1%. The end of he zero ineres rae policy in 2 and inroducion of quaniaive easing in 21, as shown in Figure 4, quickly raised and lowered he policy rae, respecively. As for he response of he unconvenional policy indicaors, he CO immediaely and coninuously decreased and increased in response o he end of he zero ineres rae policy and inroducion of quaniaive easing. On he conrary, he MB did no respond immediaely following hese policy changes. These responses of he unconvenional policy indicaors following he inroducion of quaniaive easing in March 21 reflec ha he BOJ began purchasing long-erm governmen bonds as an unconvenional asse, as he MB had already me is arge level. In oher words, quaniaive easing was a phase of he qualiaive easing policy, raher han he quaniaive easing one, in he early sages. From Augus o December 21, afer he end of quaniaive easing, however, he BOJ expanded quaniaive easing by coninuously increasing he MB and consisenly lowering he OCR. In his period, he BOJ hus implemened he unconvenional moneary policy hrough quaniaive easing. Noe also ha, as shown in Figures 2 and 4, he inroducion and end of he zero 7 Under he convenional moneary policy framework, he CO decreases because cenral banks purchase convenional asses o lower shor-erm nominal ineres raes. From he esimaed increase in he CO following he decrease in he OCR, we can infer ha he BOJ purchased unconvenional asses so aggressively ha he CO would rise in he unconvenional moneary policy framework afer he inroducion of he zero ineres rae policy. 8

10 ineres rae policy in 1998 and 2 yielded differen associaions beween he responses of he hree policy indicaors. This finding suggess ha changes in he policy rae are associaed wih he differen unconvenional policy operaions of he MB and CO; hence, we should separaely idenify each policy shock involving he OCR, MB, and CO. As repored in Figure 5, he end of quaniaive easing in July 26 led o a coninuous decrease in he MB and increase in he CO, hereby working as quaniaive ighening and qualiaive easing policies. By conras, he OCR did no respond o he policy change because he BOJ mainained he zero ineres rae policy despie ending quaniaive easing. These resuls indicae ha even when lifing quaniaive easing, he BOJ, leaving he policy rae in he zero lower bound, conduced he unconvenional moneary policy by conrolling he MB and CO (i.e., he argeed variables). We should also noe ha he OCR immediaely rose following he recurrence of shor-erm ineres rae conrols in July 26. In his monh, he MB slighly decreased, bu he CO drasically decreased. In February 27, when he BOJ raised he arge rae for he OCR, boh he OCR and he CO wen up, while he MB did no change. Hence, he periods of June 26 and February 27 saw policy ighening in erms of he convenional moneary policy framework as well as policy ighening and easing in erms of he unconvenional qualiaive policy framework. Given he above even sudy analysis, we provide wo insighs o idenify he moneary policy shocks delivered in he unconvenional moneary policy regime. Firs, he responses of he hree policy indicaors do no have a paricular relaionship; ha is, i is difficul o find simple associaions among heir responses o similar ypes of policy changes such as changes in he policy rae and in he size of he cenral bank s balance shee. This observaion suggess ha we canno inegrae moneary policy shocks in he unconvenional moneary policy regime ino a single policy shock; pu differenly, we mus idenify separaely he hree moneary policy shocks corresponding o he moneary policy indicaors. Second, he unconvenional policy indicaors (MB and CO) do no respond immediaely following moneary policy changes, whereas he shor-erm nominal ineres 9

11 rae (OCR) changes quickly and subsanially in response o policy changes. Such a difference in heir responses o policy changes is aribued o he naure of each policy indicaor. The unconvenional policy indicaors are essenially quaniaive financial variables for which he arge level is realized laer afer cenral banks sae heir policy changes for he size and composiion of heir balance shee. In his sense, we mus consider unconvenional policy shocks relaing o he MB and CO as anicipaed shocks or news shocks ha porend he fuure pahs of hese wo indicaors. By conras, he policy rae is essenially he marke price of he reserve deposis for which he arge level is quickly realized hrough an immediae reacion of he reserves marke Moneary Policy Surprises and he Moneary Policy Indicaors As discussed above, he fundamenal issue o be considered in idenifying moneary policy shocks relaing o he policy indicaors is when cenral banks sae heir policy decisions. In his subsecion, we discuss he source of he origin of moneary policy shocks. The BOJ decides is policy scheme in he M held abou wice per monh and publicly saes is policy decision jus afer. Therefore, we consider ha agens expecaion revisions following public saemens are he source of he origin of he BOJ s moneary policy shocks. Hence, we use moneary policy surprises in asse markes or an agen s expecaion revisions in asse markes, as he causally relevan facors o conrol for he endogenous responses of he hree policy indicaors o he variables remaining in and ou of he VAR. We consruc he moneary policy surprises from high-frequency daily rading daa on he major financial markes jus before and jus afer he BOJ s public saemen. More concreely, we employ he principal componen analysis of Bernanke e al. (24) and Gürkaynak e al. (25b) and prepare for he moneary policy surprises as common facors of unanicipaed changes in he major financial marke variables following he 1

12 public saemen. 8 In paricular, we use welve financial marke variables: one fuures rae (he hree-monh euro-yen TIBOR fuures rae), four yen ineres swap raes (he one-year, wo-year, five-year, and en-year yen ineres swap raes), one forward rae (he hiry-year forward rae of Japanese governmen bonds), one shor-erm spo rae (he hree-monh euro-yen TIBOR rae), wo spo exchange raes (he Yen-U.S. dollar spo exchange rae and he Yen-AUS dollar spo exchange rae in he Tokyo marke), wo sock price indexes (TOPIX and Nikkei Jasdaq index), and bank reserve deposis. Then, we compue he hree common facors as moneary policy shocks on he welve financial markes. In he Appendix, we explain furher deails on our procedure used o consruc he hree common facors. Below, we examine he saisical relevance among he moneary policy surprises and moneary policy indicaors by running he following disribued lag regression of he policy indicaors on he curren and lagged moneary policy surprises: 9 3 H I, h pc I, S = ι, I, S + I, pc h + pc= 1 h= y r r S e, (2) I where y denoes each of he moneary policy indicaors: OCR, MB, and CO. For he MB, we examine no only is logarihmic value, bu also is monhly growh raes pc of log-differenced values. S h indicaes he h lagged values for he hree moneary policy surprises generaed by using he facor analysis. r ι, I, S and I, e S denoe consan erms and sochasic disurbances, respecively. Table 6 repors he esimaion resuls for he disribued lag regression. We find 8 Unlike his paper, Bernanke e al. (24) and Gürkaynak e al. (25b) aggregaed informaion on shor-erm fuures wih differen mauriies. We do no use he same ype of informaion because, as discussed above, he nominal ineres raes in shor-erm fuures in Japan have sayed in he zero lower bound since he BOJ s inroducion of he unconvenional moneary policy. 9 h This regression corresponds o he firs componen ( R = Φ ξ + h ) of regression (9) inroduced in subsecion

13 ha moneary policy surprises are saisically correlaed wih he moneary policy indicaors in he following manner: hey are significanly associaed wih he OCR a he horizon of h =, alhough mos of our sample includes periods of % ineres raes. This finding indicaes ha moneary policy surprises have informaion on he movemen in he BOJ s policy rae. By conras, moneary policy surprises are no significanly associaed wih eiher he MB or he CO a h =, bu hey are significanly associaed wih he MB a he horizon of h 12 and wih he CO a h 2. These esimaion resuls imply ha hese wo unconvenional policy indicaors move slowly and laer in response o he BOJ s policy changes. Our finding on he responses of he unconvenional moneary policy indicaors clearly indicaes ha moneary policy surprises acually have subsanial informaion on heir fuure movemens, bu no heir conemporaneous ones. In oher words, he public saemens made jus afer he M abou changes in he wo unconvenional policy indicaors have he feaure of a news shock ha porends fuure changes in hem. In he nex secion, we incorporae hese medium- and long-erm findings among he moneary policy surprises and wo unconvenional policy indicaors ino an idenifying resricion on he ineremporal relaions among he unconvenional moneary policy shocks and indicaors. Furher, noe ha each of he moneary policy indicaors has a differenial associaion wih he moneary policy surprises, which forces us o separaely idenify he hree moneary policy shocks relaing o he hree policy indicaors: one convenional moneary policy shock ha aims o exogenously change shor-erm nominal ineres raes and wo unconvenional moneary policy shocks, each of which aims o exogenously change he size and composiion of he cenral bank s balance shee. 3. Idenifying Moneary Policy Shocks under he Unconvenional Policy Regime In his secion, we describe our empirical sraegy used o idenify he effecs of he convenional and unconvenional moneary policy shocks in he srucural VAR analysis. Firs, we consider ha he public saemens made jus afer he M are he 12

14 origin of moneary policy shocks. Second, we consider he idenifying resricions ha incorporae he feaures of he moneary policy indicaors discussed in Secion 2. Specifically, we impose resricions on which of he unconvenional moneary policy shocks emerge as hose ha capure curren and fuure changes in he size and composiion of he BOJ s balance shee, while he convenional moneary policy shock is defined as a shock ha has an immediae impac on he cenral bank s policy rae (OCR). In addiion, we assume ha unconvenional moneary policy shocks have no conemporaneous effecs on he level of he OCR. This fac reflecs ha, as discussed by Bernanke and Reinhar (24) among ohers, he cenral bank implemens he unconvenional moneary policy when convenional policy opions such as conrolling he shor-erm ineres rae are no longer viable. Our procedure for VAR idenificaion is based on he following wo-sep approach. In he firs sep, we use he moneary policy surprises as he causally relevan facors of he reduced form VAR innovaions of he hree policy indicaors and oher macroeconomic variables. In his sep, we consruc an impac marix for he insananeous responses of he VAR variables by disenangling he causal relaionships among he moneary policy shocks and VAR variables. In his sage, he impac marix does no incorporae he movemen in he unconvenional policy indicaors following policy changes. Hence, in he second sep, we impose resricions o idenify he unconvenional moneary policy shocks (i.e., he quaniaive and qualiaive shocks), which are defined as shocks ha bes explain he revisions of an agen s expecaions abou he curren and fuure pahs of he size and composiion of he cenral bank s balance shee, bu have no conemporaneous effecs on he OCR. To his end, we employ Francis e al. s (214) maximum forecas error variance (MFEV) approach, which builds on he work of Faus (1998) Srucural VAR Analysis Leing y denoe a K 1 vecor of ime-varying observables, his sochasic srucure can be expressed in erms of he vecor moving average represenaion: 13

15 y = Φ( L), (3) where denoes he K 1 vecor of he reduced form VAR innovaions. The MB, CO, and OCR are given by he firs, second, and hird elemens of srucural vecor moving average represenaion can be wrien as follows: y. Then, he y = Ψ ( L) ξ, (4) where ξ denoes he K 1 vecor of he srucural shocks. Le ξ be he 3 1 policy shock vecor [ UQN, UQL, CSR ξ = ξ ξ ξ ], where UQN ξ, UQL ξ, and CSR ξ denoe unconvenional quaniaive, qualiaive, and convenional shor-erm moneary policy shocks, respecively. The space spanned by he policy shock vecor ξ is disenangled from he space spanned by oher possible shocks of X he ( K 3) 1 vecor ξ in he following linear relaion beween he reduced form VAR innovaions and srucural shocks ξ : = Rξ = R ξ + R ξ, X X (5) R = [ R, R ], ξ = [ ξ, ξ ], X X ( K K) ( K 3) ( K ( K 3)) ( K 1) (3 1) (( K 3) 1) where R denoes he impac marix for he responses of he VAR variables y o he moneary policy shocks. The variance-covariance marix of he space spanned by he moneary policy shocks can be represened as Σ = R E( ξ ξ ) R = R R, (6) where he variance of moneary policy shocks is normalized o one. To disinguish a convenional policy shock from an unconvenional one, we impose a resricion in which 14

16 unconvenional policy shocks have no conemporaneous effecs on he level of he policy rae. More specifically, his resricion can be expressed in he following manner: R ( K 3) UQN UQL CSR rmb rmb r MB UQN UQL CSR rco r r CO CO CSR R 1:3,1:3 r CR = =. UQN UQL CSR R r 4: K,1:3 XK 4 rxk 4 r XK 4 UQN UQL CSR rxk rxk r XK (7) 3-2. Conrolling he Endogeneiy of he Moneary Policy Indicaors We use he moneary policy shocks exraced from he changes in he welve major financial markes on M days as he confounding facors of he reduced form VAR innovaions, ε. Thus, we aim o conrol for he endogeneiy of he moneary policy indicaors and disenangle he causal effecs of he policy shocks on he VAR variables a he shock arrival ime. More concreely, we conduc he following sysem regression: e = R S + e, (8) S ( K 3) (3 1) where S denoes he 3 1 vecor of he hree moneary policy surprises a a monhly frequency. The sysem regression yields he insananeous responses of he VAR variables o he moneary policy shocks in he form of fied values Rˆ S S. Then, we obain he variance-covariance marix involving he conemporaneous impacs of he moneary policy shocks o he VAR variables as follows: ˆ ˆ S ˆ S S = R S S R. (9) 15

17 3-3. Idenifying Unconvenional Policy Shocks Here, we describe he second-sep procedure o idenify he convenional and unconvenional moneary policy shocks. Specifically, we consider resricions o exrac he hree ypes of srucural shocks associaed wih he convenional and unconvenional policy indicaors from variance-covariance marix (9) in which he endogeneiy of he policy indicaors is conrolled for by he moneary policy surprises. When idenifying he unconvenional moneary policy shocks, we uilize he policy indicaors of he MB and CO, which gradually move and for which he arge levels are achieved soon afer he BOJ s public saemens on M days. To incorporae his feaure ino our idenificaion of he unconvenional moneary policy shocks, we define hem as he surprise componens of moneary policy ha bes explain he curren and fuure pahs of he MB and CO. This idenificaion sraegy involves modeling revisions in agens expecaions abou he curren and fuure pahs of he unconvenional policy indicaors. To his end, we employ he MFEV approach proposed by Faus (1998) and Francis e al. (214). 1 This approach allows us o specify he revisions in agens expecaions in erms of maximizaion problems for he conribuions of he unconvenional policy shocks o he forecas error variances of he unconvenional policy indicaors. To explain he MFEV approach, we begin by expressing he h sep ahead forecas error condiioning on he srucural shocks ξ : y h h h X X + h E 1y+ h= ΦRξ+ h = ΦR ξ+ h + ΦR ξ+ h = = =, (1) where he firs and second equaliies use equaions (4) and (6). Then, he h sep ahead forecas error due o moneary policy shocks ξ + can be expressed as 1 Uhlig (23), Barsky and Sims (211), and Kurmann and Orok (213) employed he MFEV approach o idenify news shocks, or anicipaed shocks, abou fuure echnology. Zeev e al. (215) idenified anicipaed moneary policy shocks in he Unied Saes by using his approach. 16

18 h h ΦR ξ+ h = ΦR D ξ+ h = =, (11) where R denoes he following K 3 orhogonalizaion marix: r 11 r r R r r r, = r r r K1 K2 K3 and D denoes he 3 3 orhonormal marix ( ' D D I = ). 11 Thus, he share of he h sep ahead forecas error variance of moneary policy indicaor i aribuable o, moneary policy shock ξ j is expressed in he following form of a variance decomposiion: i, j h h 1i( Φ 2 je τ 2 j Φτ) e1 i Φi, R d jd j R τ Φi, τ τ = τ = = h h e1 i( ΦΣ τ Φτ ) e1 i Φi, τσ Φi, τ τ = τ = e R D e D R Ω ( h) =. (12) This variance decomposiion models he revisions in an agen s expecaions abou he curren and fuure pah of policy indicaor i a he ime of he emergence of policy shock j, where i ( i = 1, 2, 3 ) indicaes he place of he moneary policy 11 In pracice, we obain an orhogonalizaion marix R as follows. Firs, we perform a Cholesky ˆ decomposiion of he 3 3 upper lef submarix Σ of variance-covariance marix (1), such ha 1:3,1:3 ˆ ' Σ = R R. Then, we calculae an orhogonalizaion marix 1:3,1:3 1:3,1:3 1:3,1:3 ' S 1 ' S R = R R R. 1:3,1:3 1:3 ( ˆ ) ˆ 17

19 indicaors (MB, CO, OCR) in vecor variable place of moneary policy shocks, e 1, i and UQN ξ, UQL ξ, and y, while j ( j = 1, 2, 3 ) indicaes he CSR ξ, in policy shock vecor ξ. e 2 j denoe he K 1 and 3 1 selecion vecors wih one in he i h and j h place and zeros elsewhere. orhonormal marix Noe ha D. d j is he 3 1 vecor indicaing he j h column of he R d j is he K 1 vecor corresponding o he j h column of a possible orhogonalizaion in equaion (12) and hus inerpres he conemporaneous impac of he j h moneary policy shock on he VAR variables. Therefore, if we have esimae d ˆ j, we can generae he impulse responses of he VAR variables o he j h moneary policy shock by using he esimaed impac vecor R d in equaions (4) and (6). We apply he MFEV approach o variance decomposiion (13) o idenify he quaniaive, qualiaive, and convenional moneary policy shocks. More concreely, we begin by idenifying he quaniaive moneary policy shock,, saisfying he following condiions: ˆ j UQN ξ dˆ = arg ma x Ω ( h) = UQN MB, UQN h τ = Φ R MB, τ UQN UQN MB, τ h τ = Φ d Σ d Φ R MB, τ MB, τ Φ, (13) s.. R (3,1) d (1,1) + R (3, 2) d (2,1) + R (3,3) d (3,1) =, (14) UQN UQN UQN d d UQN UQN = 1. (15) Consrain (15) imposes ha he quaniaive shock, ξ on he OCR. UQN, has no conemporaneous effec This consrain reflecs ha he cenral bank implemens he unconvenional moneary policy when he convenional policy opion of conrolling he shor-erm ineres rae is unavailable. Consrain (16) ensures ha d UQN is he firs 18

20 column vecor belonging o orhonormal marix D. We obain d ˆUQN by solving he above maximizaion problem and calculae he impulse responses of he VAR variables o he quaniaive moneary policy shocks by using esimaed impac vecor policy shocks. R d. ˆU QN Nex, we describe he idenificaion of he qualiaive and convenional moneary Specifically, we idenify he qualiaive moneary shocks ξ UQL by solving he following maximizaion problem: dˆ = arg max Ω ( h) = UQL CO, UQL h τ = Φ CO, τ UQL UQL CO, τ h τ = R d d R Φ Σ Φ CO, τ CO, τ Φ, (16) s.. R (3,1) d (1,1) + R (3, 2) d (2,1) + R (3,3) d (3,1) =, (17) UQL UQL UQL d UQN = dˆ, (18) UQN d d UQL UQL = 1. (19) As wih consrain (15), consrain (18) imposes ha he qualiaive policy shock has no conemporaneous impac on he OCR. Consrain (19) ensures ha he qualiaive shock is orhogonal o he quaniaive shock idenified in advance; in oher words, he qualiaive shock is idenified so ha he second column in R ξ is orhogonal o he firs column obained in he maximizaion problem (14) hrough (16). We can compue he impulse responses o he qualiaive moneary policy shocks by using esimaed impac vecor R dˆ UQL. CSR Convenional moneary policy shock ξ is idenified so ha he hird column in R ξ is orhogonal o he firs and second columns obained hrough he above maximizaion problems and he surprise componen of moneary policy has a conemporaneous impac on he OCR. 19

21 4. Empirical Resuls for Unconvenional Moneary Policy Shocks In his secion, we discuss he empirical resuls obained by using he moneary policy shocks idenified by he mehod presened in he previous secion. In paricular, we focus on unconvenional moneary policy shocks. When consrucing he VAR, in addiion o he hree moneary policy indicaors (MB, CO, OCR), we include six macroeconomic variables: wo asse marke prices, hree real economic variables, and one price indicaor. The wo asse marke prices are chosen as he sock price index, SP, and he 1-year governmen bond yield, 1YJGB. The hree real economic variables are he one-year real ineres rae, 1YREAL; he bank s safe asse raio o risky asses, SAFE/RISK; and he index of indusrial producion, IIP. The consumer price index, CPI, is included as he price indicaor. As discussed below, we also include oher candidaes for he VAR variables o conduc robusness checks. We se he lag lengh o one in he reduced form VAR esimaion based on he Bayesian informaion crieria Convenional and Unconvenional Moneary Policy Shocks Here, we repor he saisical relevance beween he reduced form VAR innovaions and moneary policy shocks. Table 7 shows he esimaion resuls for he sysem regression of he reduced form VAR innovaions on he hree moneary policy shocks: 3 pc, k ˆk = rk, pcs + pc= 1. (2) The moneary policy surprises seem o explain subsanial proporions of he reduced form VAR innovaions. In paricular, he asse prices including SP and 1YJGB appear o quickly respond o he moneary policy surprises. On he conrary, he moneary policy surprises explain lile of he moneary policy indicaors (MB, CO, OCR) a he shock arrival ime. The insignificance of he convenional policy 2

22 indicaor, or he policy rae, can be aribued o is inacion for mos of our sample period from 1998 owing o he exremely low ineres rae regime in Japan. Table 8 repors he resuls for he variance decomposiion of he hree moneary policy indicaors aribuable o he moneary policy shocks for h =, 12, 24, 36, and 48 monhs ahead: Ω MB, j ( h), Ω CO, j ( h) and Ω CR, j ( h), where j indicaes he quaniaive, qualiaive, and convenional policy shocks. As clearly shown in his able, he quaniaive shock subsanially accouns for variaions in he MB conrolled by our causally relevan facors, or he moneary policy surprises, alhough he convenional policy shock also somewha accouns for hem. 13 The quaniaive and convenional moneary policy shocks appear o have major conribuions o he variaion in he fuure CO adjused for he moneary policy shocks. The convenional policy shock explains all he variaion in he OCR for h = according o our idenificaion of his shock. However, afer he firs impac, he quaniaive shock appears o increase is conribuion o he variaion in he OCR gradually Impulse Response Analysis In his subsecion, we repor he esimaed impulse responses o he exogenous moneary policy shocks. Figures 2 o 4 repor he esimaed impulse responses o he quaniaive, qualiaive, and convenional moneary policy shocks of one sandard deviaion, respecively. As shown in Figure 2, he quaniaive easing shock leads o a gradual and coninuous increase in he MB, albei i has no immediae impac on i. This finding illusraes ha he quaniaive shock is idenified as a news shock abou he expansion of he balance shee (i.e., agens expec he arge level for he MB o be achieved soon afer he BOJ announces is new arge). The quaniaive easing shock also leads o a slow 13 Noe ha mos of he variance in he MB a h= is aribuable o qualiaive easing shocks. However, he sum of he conribuion of he hree ypes of moneary policy shocks o he MB is economically negligible. Hence, his finding implies ha he forecas variance in he MB a h= canno be explained by moneary policy shocks. 21

23 increase in he CO, clearly indicaing ha he BOJ ends o increase unconvenional asses han convenional asses in he process of raising he size of is balance shee. For he esimaed responses of nominal ineres raes, he long-erm nominal ineres rae, 1YJGB, immediaely falls, while he OCR gradually decreases and reaches is boom one year laer. This resul implies ha a quaniaive easing shock has a policy duraion effec, which works as a signal abou he fuure pah of policy raes, hereby decreasing long-erm ineres raes immediaely. The quaniaive easing shock does no yield favorable effecs on SP or on SAFE/RISK (he former declines and he laer rises in response o he quaniaive easing shock). From he esimaion resuls, we can infer ha he quaniaive easing shock did no induce a porfolio rebalance, in which he financial insiuions wih more safe asses are expeced o increase lending and purchase relaively risky asses including socks. Raher, he quaniaive easing shock would merely aler supply/demand relaionships in he Japanese governmen bond marke or change he marke s expecaions abou he duraion of he zero ineres rae policy. Consisen wih his inference, he quaniaive easing shock also does no have a favorable effec on IIP or CPI. Taking ino accoun ha his shock significanly decreases he long-erm nominal ineres rae, we can infer ha he ineres rae channel hrough he decrease in he long-erm nominal ineres rae due o quaniaive easing does no work well under he unconvenional moneary policy regime in Japan. As shown in Figure 3, he qualiaive easing shock has a significan effec on he CO, alhough i has no conemporaneous impac. More concreely, he CO peaks almos six monhs laer. On he conrary, he MB does no significanly respond o he quaniaive easing shock. In conras o he quaniaive easing shock, he qualiaive easing shock leads o a subsanive increase in SP, a decrease in 1YJGB, and a decrease in SAFE/RISK. These findings imply ha qualiaive easing causes financial insiuions o increase he purchase of boh safe and risky asses and o lend more. The BOJ s larger purchases of unconvenional asses under qualiaive easing resuled in a igh supply/demand balance in he long-erm Japanese governmen bond marke and he rise in long-erm Japanese governmen bond prices. In addiion, a porfolio rebalance 22

24 occurred due o he qualiaive easing shocks. For he esimaed responses of he oher real economic variables, 1YREAL coninuously decreases in response o he qualiaive easing shock, implying ha qualiaive easing raises inflaion expecaions when i has lile influence on he shor-erm nominal ineres rae. Boh IIP and CPI significanly increase. From he resuls for he quaniaive and qualiaive easing shocks, we can infer ha he credi channel of he qualiaive easing policy works, whereas he ineres rae channel of he quaniaive easing policy does no. Figure 4 shows ha he convenional policy shock basically produces similar impulse responses o hose demonsraed in he VAR lieraure (Bernanke and Blinder (1992), Chrisiano e al. (1996), and Bernanke and Mihov (1998) for U.S. moneary policy, and Miyao (2; 22), Fujiwara (26), Nakashima (26), Inoue and Okimoo (28), and Shibamoo (216) for Japanese moneary policy), alhough he iniial responses of some variables seem o differ. The convenional policy shock is idenified so ha he OCR subsanially falls following is arrival. This convenional policy shock leads o an increase in he MB and a decrease in he CO. The esimaed impulse responses of SP are no significan in abou one year, bu coninuously increase from one year o hree years following he convenional policy shock. The long-erm nominal ineres rae, 1YJGB, appears no o significanly respond o he convenional policy shock. 1YREAL begins o fall afer he OCR, reurning o he pre-shock level abou one year laer, hus reflecing ha he inflaion rae sars o increase from ha period. From he negaive responses of SAFE/RISK, i can be inferred ha he convenional policy shock causes a porfolio rebalance even during his low ineres rae period. The convenional policy easing shock leads o increases in boh IIP and CPI, alhough he former shows a decrease in he firs few periods. 14 IIP peaks afer CPI begins o increase abou wo years following he convenional policy shock. 14 The firs decline in IIP can be explained by he expecaion revisions of marke paricipans and firms abou he fuure economic condiions, which are caused by convenional moneary policy shocks as Romer and Romer (2) poined ou. 23

25 4-3. Exogeneiy of Moneary Policy Shocks In he previous secions, we showed ha qualiaive easing shocks have favorable effecs on he real economy, while quaniaive shocks do no. However, our resuls may arise from correlaions beween our moneary policy surprises and oher deerminans of he real economy. To demonsrae he plausibiliy of our moneary policy surprises, we examine he associaions among he nine variables in he VAR and he global economic variables ou of he VAR. Because he Japanese economy is linked o he global economy, alhough our idenified policy shocks can be exogenously deermined from global economic facors, he reduced form VAR innovaions of he policy indicaors can be endogenously deermined. From his analyical viewpoin, we conduc he following regression of he VAR forecas errors on he global economic facors: (21), ˆk x l xk = rkl, x +, l where ˆk denoes he VAR innovaions esimaed from he nine-variable VAR, while denoes he global economic variables expeced o have subsanive effecs on he Japanese economy. Table 9 shows ha he oil price, OIL, is posiively associaed wih he forecas errors of he shor-erm nominal ineres rae (OCR), 1YREAL, and CPI. This finding indicaes ha he OCR rises in endogenous response o increases in inflaion raes and inflaion expecaions. The U.S. index of indusrial producion, USIIP, is negaively associaed wih he VAR innovaions of he long-erm nominal ineres rae (1YJGB) and SAFE/RISK. From hese esimaes, i can be inferred ha he decline in he U.S. economy is significanly associaed wih he increase in demand for Japanese governmen bonds. For he inerrelaion among sock marke prices, he U.S. sock price index (USSP) is posiively associaed wih he VAR innovaions of he Japanese sock price index (SP) in a subsanive manner. l x 24

26 In erms of he endogenous response of moneary policy o global financial fragiliy, he TED spread is posiively associaed wih he forecas errors of he BOJ s CO. In addiion, he federal funds rae is posiively associaed wih he OCR. This finding implies ha U.S. financial fragiliy is associaed wih he BOJ s qualiaive easing and a decrease in he policy rae. In he nex exercise, we examine wheher our idenified moneary policy shocks are deermined from he above global economic variables using he following auxiliary regression: x = r x + x (22) ˆ, j x l, j jl,, i where ξ indicaes each of he qualiaive, quaniaive, and convenional moneary ˆ, j policy shocks. As shown in Table 1, none of he moneary policy shocks is significanly associaed wih he global economic variables a he 5% significance level. This esimaion resul ensures ha our moneary policy shocks are exogenous o global economic shocks. 15 The above analysis suggess ha Japanese moneary policy endogenously responds o he global financial condiions. Hence, he simple use of he reduced form VAR innovaions of he moneary policy indicaors causes us o erroneously esimae he policy effecs. For his reason, we used our moneary policy shocks o disenangle he exogenous responses of he economy o moneary policy shocks from he endogenous ones Robusness Check We also conduced a robusness check on he impulse responses o 15 As an excepion, he U.S. sock price index, USSP, is posiively associaed wih he quaniaive policy easing, CO, and he convenional policy easing, OCR, a he 1% significance level. This finding implies ha Japanese moneary policy shocks have a favorable effec on U.S. sock prices under he unconvenional moneary policy regime. 25

27 unconvenional moneary policy shocks, namely he quaniaive and qualiaive moneary policy shock, esimaed in he previous subsecion. We included he all-indusry aciviy index, unemploymen rae, and index for he shipmen of invesmen goods, insead of IIP. The use of hese alernaive variables did no provide favorable real effecs for he quaniaive easing shocks, bu yielded favorable real effecs for he qualiaive easing shocks. Our robusness checks for he impulse responses o he wo unconvenional policy shocks herefore srongly suppor he empirical resuls repored in he previous subsecion Unconvenional Moneary Policy Effecs We have hus far found ha quaniaive easing shocks do no have a favorable effec on he real economy, alhough hey do lead o a decrease in he long-erm nominal ineres rae, as expeced by he BOJ. On he conrary, qualiaive easing shocks cause favorable effecs no only on he long-erm ineres rae, bu also on real economic aciviy. In his subsecion, on he basis of hese findings, we provide wo hypoheses abou he unconvenional policy effecs for fuure research. One hypohesis for he ineffeciveness of quaniaive easing shocks is ha hey resul in raising concern abou he fuure fragiliy of he real economy. According o Romer and Romer (2), Ellingsen and Sodersrom (21), Claus and Dungey (212), and Campbell e al. (212), moneary policy acions provide he public wih signals of he cenral bank s informaion. If he quaniaive easing by he BOJ worked as a signal ha presages fuure decreases in oupu and inflaion, his signal would resul in suppressing firm invesmen and wage growh. Anoher hypohesis involves economic uncerainy and is cauious effec on real economic aciviy. By using sock marke opion-based implied volailiy daa, or VIX daa, Bekaer e al. (213) found ha convenional policy easing by lowering he shor-erm ineres rae resuls in he decrease in economic uncerainy and ha his decreased uncerainy also leads o favorable effecs on he real economy (see also Aasvei e al. (213) and Creal and Wu (216)). If he quaniaive easing shock resuls in elevaing economic uncerainy, while he qualiaive easing shock conribues o 26