IMES DISCUSSION PAPER SERIES

Transcription

1 IMES DISCUSSION PAPER SERIES Deerminans of Households Inflaion Expecaions Kozo Ueda Discussion Paper No. 29-E-8 INSTITUTE FOR MONETARY AND ECONOMIC STUDIES BANK OF JAPAN 2-- NIHONBASHI-HONGOKUCHO CHUO-KU, TOKYO JAPAN You can download his and oher papers a he IMES Web sie: hp:// Do no reprin or reproduce wihou permission.

2 NOTE: IMES Discussion Paper Series is circulaed in order o simulae discussion and commens. Views expressed in Discussion Paper Series are hose of auhors and do no necessarily reflec hose of he Bank of Japan or he Insiue for Moneary and Economic Sudies.

3 IMES Discussion Paper Series 29-E-8 March 29 Deerminans of Households Inflaion Expecaions Kozo Ueda* Absrac In his paper, we invesigae he deerminans of households inflaion expecaions in Japan and he Unied Saes. We esimae a vecor auoregression model in which he four endogenous variables are inflaion expecaions, inflaion, he shor-erm nominal ineres rae and he oupu gap, wih energy prices and (fresh) food prices being exogenous. Shor-erm nonrecursive resricions are imposed aking accoun of simulaneous codependence beween realized inflaion and expeced inflaion. We find, firs, ha responding no only o changes in energy prices and food prices bu also o moneary policy shocks, inflaion expecaions adjus more quickly han does realized inflaion. This explains why Japanese and US daa indicae ha inflaion expecaions lead realized inflaion. Second, he effecs of changes in energy prices and food prices on inflaion and inflaion expecaions are large in he shor run in Japan, while in he Unied Saes, hey are no only large bu also long lasing. Third, shocks o expecaions occasionally flucuae grealy, and can have self-fulfilling effecs on realized inflaion. The self-fulfilling propery is more apparen in he Unied Saes han in Japan. Keywords: expeced inflaion; srucured vecor auoregression; moneary policy JEL Classificaion: C32, E3, E52 * Insiue for Moneary and Economic Sudies, Depuy Direcor and Economis, Bank of Japan ( kouzou.ueda boj.or.jp) The auhor is graeful o Tasuyoshi Okimoo, Feng Zhu and oher seminar paricipans a Bank for Inernaional Selemens and he Bank of Japan for helpful suggesions. I would also like o hank Ko Nakayama for providing me hisorical daa. Views expressed in his paper are hose of he auhor and do no necessarily reflec he official views of he Bank of Japan. All errors are my own.

4 . Inroducion Inflaion expecaions are imporan in conducing moneary policy for many reasons. Firs, heir self-fulfilling propery causes acual inflaion. I is ofen said ha an unconrollable increase in inflaion expecaions generaed he hyper-inflaion of he 97s. Second, inflaion expecaions maer because hey affec real ineres raes, and hereby he real economy and acual inflaion. The effec of moneary policy is hus grealy consrained by how expecaions are formed. Third, inflaion expecaions influence wage negoiaions beween employers and employees. A he same ime, inflaion expecaions are affeced by wages, which can induce price wage spirals. Fourh, uncerainy abou expecaions discourages real aciviies and resuls in inefficien resource allocaion. For hese reasons, inflaion expecaions are imporan for moneary policy. However, lile is known abou heir properies, as Federal Reserve Board chairman Bernanke (27) saes: Undoubedly, he sae of inflaion expecaions grealy influences acual inflaion and hus he cenral bank s abiliy o achieve price sabiliy. Bu wha do we mean, precisely, by he sae of inflaion expecaions? How should we measure inflaion expecaions, how should we use he informaion for forecasing and conrolling inflaion? I cerainly do no have complee answers o hose quesions, bu I believe ha hey are of pracical imporance (emphasis is added in ialic). In his paper, we aim o answer Bernanke s second quesion. We furher aim o answer he supplemenary quesions promped by he following reques by Bernanke (27): we mus undersand beer he hisorical variaion in inflaion expecaions, he effec of his variaion on acual inflaion and economic aciviy, and he relaionship beween policy acions and he formaion of inflaion expecaions. There are a number of empirical sudies of he formaion of inflaion expecaions. For example, Carlson and Parkin (975) argue ha pas inflaion affecs inflaion expecaions. Supposing perfec foresigh, Robers (998) and Oshima and Nakayama (999) argue ha no only pas inflaion bu also prediced fuure inflaion affecs inflaion expecaions. Mullineaux (98), Gramlich (983) and Pearce (987) examine he effecs on inflaion How o measure inflaion expecaions, which is he firs quesion raised by Bernanke (27), is beyond he scope of his paper. See Carlson and Parkin (975) and Berk (2) for deails of he mehods used o ransform qualiaive survey daa ino quaniaive esimaes. The papers by Toyoda (987) and Hiraa and Kamada (26) are available only in Japanese. For he comparisons of inflaion expecaions among differen ypes of agens, see Gramlich (983), Bachelor and Dua (989), Robers (998), Oshima and Nakayama (999), Thomas (999) and Mankiw e al. (23). The disribuion of inflaion expecaions is analyzed by Carlson and Parkin (975), Bachelor and Dua (989), Mankiw e al. (23) and Hiraa and Kamada (26).

5 expecaions of governmen policy and he money supply. However, o our knowledge, Berk (2) represens he only aemp o sudy he effec on inflaion expecaions of shor-erm nominal ineres raes, which are considered he insrumen of moneary policy. 2 His wo-sep approach can be summarized as follows. Firs, using daa for Europe, he esimaes a vecor auoregression (VAR) model in which he five variables are he domesic money marke rae and is foreign equivalen, he inflaion rae, indusrial producion and he money sock. He hen obains an esimaed moneary policy shock from he residual of he ineres rae equaion in he VAR. Second, he uses a vecor error correcion model (VECM) of inflaion and expeced inflaion, in which he consruced moneary policy shock is included as an exogenous variable. Berk finds ha he moneary policy shock does no have a saisically significan impac on inflaion expecaions. However, Berk s mehod seems problemaic for wo reasons. Firs, here is no explanaion of his idenificaion sraegy for he VAR, and he moneary policy shock is unlikely o be properly idenified. Second, even if his is no he case, he should have inegraed his wo-sep procedure ino one by consrucing a five-variable VAR or VECM. In his paper, o invesigae he deerminans of inflaion expecaions, we use a srucured VAR (SVAR) model. By using boh Japanese and US daa, we consruc a four-variable VAR in which he endogenous variables are he oupu gap, he shor-erm nominal ineres rae, realized inflaion and inflaion expecaions. The exogenous variables are energy price changes and (fresh) food price changes. Then, aking accoun of simulaneous codependence beween realized inflaion and expeced inflaion, we impose a nonrecursive resricion o idenify srucural shocks. Our analyses shows ha, firs, responding no only o changes in energy prices and food prices, bu also o moneary policy shocks, inflaion expecaions adjus more quickly han does realized inflaion. This explains why Japanese and US daa indicae ha inflaion expecaions lead realized inflaion. I is also shown ha our SVAR procedure resolves he price puzzle; an ineres rae rise decreases prices even on impac. Second, in Japan and he Unied Saes, changes in energy prices and food prices have large effecs on inflaion and inflaion expecaions. However, hese are only shor-run effecs in Japan, while in he Unied Saes, hey are persisen. Third, shocks o expecaions occasionally flucuae grealy, and may have self-fulfilling effecs on realized inflaion. The self-fulfilling propery is more apparen in he Unied Saes han in Japan. 2 In addiion o sudies of he effecs of shor-erm nominal ineres raes, here are also sudies of he effecs of oher moneary policy insrumens. For example, Hori and Shimizuani (23) analyze a survey ha direcly asks respondens abou he effec on inflaion expecaions of he quaniaive easing of moneary policy implemened from 2 in Japan. From he perspecive of inflaion argeing, Orphanides and Williams (22) and Erceg and Levin (23) sudy people s learning processes and heir inflaion expecaions. 2

6 This paper is srucured as follows. In Secion 2, we provide an overview of inflaion expecaions. In Secion 3, we explain our mehodology. In Secions 4 and 5, respecively, we repor our esimaion resuls for Japan and he Unied Saes. In Secion 6, we examine he robusness of our resuls. Secion 7 concludes he paper. 2. Overview of Inflaion Expecaions Inflaion expecaions daa are divided ino wo main caegories: marke-based inflaion expecaions, which are refleced in Treasury inflaion-proecion securiies, and survey-based inflaion expecaions. In his paper, we focus on he laer. Table repors he survey-based inflaion expecaions daa ha are available in Japan. In he firs hree surveys, because respondens are households, one can examine inflaion expecaions from he perspecive of he buyers of goods. Of hese hree surveys, he Consumer Confidence Survey, which began in 97, covers he longes period. This survey is herefore useful for conducing ime-series analysis. However, a major limiaion is ha is informaion is qualiaive; ha is, households are asked if fuure price changes are expeced o be good (lower) or bad (higher). Thus, one has o ransform his qualiaive informaion ino quaniaive informaion by using an appropriae procedure, such as ha developed by Carlson and Parkin (975). TANKAN, he fourh survey in he able, is a survey of enerprises. Thus, he answers from reail indusries can be used o examine inflaion expecaions from he perspecive of he sellers of final goods. However, his survey is also qualiaive. In he final wo surveys in he able, respondens are economiss. Consensus Forecass publishes heir inflaion expecaions over horizons of one quarer o en years, and hese expecaions are quaniaive. However, his informaion from Consensus Forecass is only available from 99, and he respondens are no necessarily expers on he Japanese economy. In his paper, we focus on household inflaion expecaions. This is done for wo reasons. Firs, households are imporan economic agens. Their inflaion expecaions almos cerainly influence he real economy hrough real ineres raes and wage negoiaions. Moreover, heir inflaion expecaions affec realized inflaion by affecing households consumpion demands, and because some are price-seers. Second (as shown subsequenly), households inflaion expecaions appear o lead realized inflaion by one o hree quarers. We aemp o explain his finding, which is of paricular ineres. Of he available household surveys, we use he Consumer Confidence Survey because i covers he longes period. We consruc wo ypes of measures of inflaion expecaions. One is a diffusion index (DI), which is consruced by assigning scores of,.25,.5,.75 and, respecively, o respondens expecaions ha prices will be higher, slighly higher, no 3

7 change, slighly lower, and lower. 3 The scores are convered ino a figure for expeced inflaion by applying he mehod developed by Carlson and Parkin (975), which is he mos widely known and used mehod. This mehod assumes ha respondens repor a variable o rise or fall if heir evaluaion lies above or below a cerain hreshold; i is also assumed ha heir answers are symmeric and normally disribued. Furhermore, i is assumed ha he average value of pas realizaions equals he average value of expecaions. Given hese assumpions, one obains he hreshold value and inflaion expecaions for each period. For daa on prices, we use he Consumer Price Index (CPI) for all goods excep fresh food. CPI for all goods excep fresh food and % rimmed mean CPI are said o show beer performances han oher core inflaion indicaors in erms of racking he underlying rend of inflaion and forecasing he fuure direcion of headline inflaion (Shirasuka [26]). CPI for all goods excep fresh food is he mos widely used index in Japanese moneary policy analysis. Figure illusraes he movemen of households inflaion expecaions. In he graph, he inflaion rae for each period indicaes he price change from he previous year. The inflaion expecaion for each period represens he difference beween he curren price and he expeced fuure price. We find ha realized and expeced inflaion are highly correlaed, and ha inflaion expecaions change more quickly han does realized inflaion. Table 2 suppors hese findings. The correlaion beween he Carlson Parkin measure of expeced inflaion and realized inflaion is as large as.7, and he expeced inflaion leads realized inflaion by wo o hree quarers. This finding is confirmed when using daa from 99, when adjusing for he inflaionary effec of he consumpion ax hikes of April 989 and April 997, and wheher using he CPI including or excluding food and energy. Nex, we compare he inflaion expecaions of households wih hose of reailers and professional economiss. Table 3 shows ha he lead lengh of households inflaion expecaions exceeds hose of reailers and economiss inflaion expecaions. Similar findings apply o he Unied Saes. Table 4 summarizes he main surveys of inflaion expecaions. The Michigan Survey provides households inflaion expecaions, while he Livingson Survey, Survey of Professional Forecasers and Consensus Forecass provide professional economiss expecaions. In conras wih Japan, all four US surveys are quaniaive. Thus, he Carlson Parkin mehod is no needed. As Figure 2 and Table 5 show, households inflaion expecaions lead realized inflaion for boh he CPI and he core PCE deflaor by one quarer and hree quarers, respecively. Table 6 compares he inflaion expecaions of households wih hose of economiss. We find ha economiss inflaion expecaions do no lead realized inflaion. These findings are consisen wih hose of 3 For some survey periods, he number of choices is hree (raher han five), o which we assign scores of,.5 and, respecively. 4

8 Thomas (999). A quesion ha arises a his sage is why households inflaion expecaions change more quickly han does realized inflaion. To answer his quesion, in he following secions, we examine he deerminans of households inflaion expecaions. 3. Mehodology We use an SVAR o examine he deerminans of inflaion expecaions. We assume ha he rue model can be wrien as A X = A L) X + e, I = E[ e e' ], () ( n where X is a vecor of n endogenous variables, A and A(L) are coefficien marices, and L is he lag operaor. The srucured shocks, e, are assumed o be muually orhogonal, and heir variance covariance marix is an n n ideniy marix. A sandard VAR mehod is described by he following reduced form: X In his case, he obained shocks, = B L) X + ε, Σ = E[ ε ε' ]. (2) ( ε, are muually dependen, and hus canno be regarded as srucural. Therefore, one needs o impose resricions, and idenify muually independen srucural shocks ha cause he endogenous variables o flucuae. In general, he number of resricions needed is n(n-)/2. We impose a nonrecursive resricion. Of he proposed approaches, he simples is o impose a recursive resricion. This mehod assigns a cerain ime ordering o he endogenous variables. I is he mos widely used mehod in macroeconomic sudies, paricularly for he analyses of moneary policy (see, for example, Sims [98, 992], Bernanke and Mihov [998] and Chrisiano, Eichenbaum and Evans [999]). However, applying his mehod o inflaion expecaions is inappropriae because inflaion expecaions affec, and are simulaneously affeced by, inflaion. Thus, we impose a nonrecursive resricion, as developed by, for example, Kim (999) and Sims and Zha (25). This enables us o incorporae he ineracion beween inflaion and inflaion expecaions and assess he deerminans of inflaion expecaions. Before moving o our nonrecursive resricion, we explain our VAR specificaion. We esimae a VAR model ha has four endogenous variables. These are he oupu gap (y), he shor-erm nominal ineres rae (i), he inflaion rae from he previous quarer (π ) and he e curren expeced inflaion rae ( π ). Shor-erm nominal ineres raes are conraced ineres raes on loans and discouns (shor erm and sock) available from he Bank of Japan. We use hese daa, raher han daa on overnigh call raes, because he laer had a zero lower bound around 2. The oupu gap is calculaed by Hara e al. (26) using a producion funcion approach. The inflaion rae is he change in CPI, excluding fresh food, from he previous quarer. Adjusmens are made for seasonaliy and for he effecs of consumpion axes. These 5

9 measures of he oupu gap and inflaion are used because he Bank of Japan appears o assign hem he mos weigh; hence, heir use helps us o idenify a moneary policy shock. Moreover, CPI excluding fresh food is said o show beer performances han oher core inflaion indicaors in erms of racking he underlying rend of inflaion and forecasing he fuure direcion of headline inflaion (Shirasuka [26]). Inflaion expecaions are calculaed using he Carlson Parkin mehod. Inflaion expecaions are no adjused for he effec of consumpion axes because i is difficul o idenify when and how much households change heir inflaion expecaions responding o announced changes in consumpion axes. We also use he following exogenous variables: oil price changes from he previous quarer (dpoil), fresh food price changes from he previous quarer (dpfresh) and consumpion ax dummies. Fresh food prices are seasonally adjused. The firs wo variables have no more han one lag. Oil price daa are aken from he Nikkei oil index. Consumpion ax dummies ake he value uniy from 988:Q4 o 989:Q and from 996:Q4 o 997:Q, and zero oherwise. Our sample period ranges from 975:Q o 27:Q4. 4 The number of lags seleced by he Hannan Quinn informaion crierion is wo. All daa are in logarihms muliplied by. We use a similar VAR specificaion for he Unied Saes, for which he four endogenous variables are he oupu gap, he federal fund rae, he inflaion rae and he expeced inflaion rae. The oupu gap is obained using he Hodrick Presco filer. The price index used is core PCE deflaor excluding food and energy and is seasonally adjused. We use hese measures of he oupu gap and inflaion because hey seem o be assigned he mos weigh by he Federal Reserve; hence, heir use helps us o idenify a moneary policy shock. Inflaion expecaions are he mean of households one-year-ahead forecass of inflaion, obained from he Michigan Survey. 5 The wo exogenous variables are he change in energy prices (dpenergy) and he change in food prices (dpfood). Boh variables have no more han one lag. Food prices are seasonally adjused. The sample period ranges from 97:Q o 27:Q4. The number of lags seleced using he Hannan Quinn informaion crierion is hree. To find he rue srucured model, we impose he zero resricion described in equaion (3) below. Given he definiion X e = { y, i, π, π }', he coefficien marix A in equaion () is: x A = x x x x x x x. x x (3) Because here are four variables, we impose 4(4 )/2 = 6 zero resricions. From equaion (), 4 Daa on he oupu gap are available from 975:Q. 5 We use mean (raher han median) inflaion expecaions hese daa are available over a longer period. 6

10 imposing his resricion yields: 6 y = A ( L) X e e y e i = a π + A ( L) X e i e π = a y + a π + A ( L) X 4 + e p π = a y a i + a π + A ( L) X + e pe. (4) The raionale for his resricion is as follows. The firs equaion implies ha he oupu gap does no respond conemporaneously o he oher variables. The corresponding shock is inerpreed as a demand shock. As is shown subsequenly, his inerpreaion is jusified by looking a he impulse response o a posiive demand shock ha raises he oupu gap, inflaion and expeced inflaion. The second equaion implies ha a cenral bank canno monior he curren inflaion rae or he curren oupu gap, bu ha i is forward-looking and akes accoun of households inflaion expecaions o proxy is own expecaions. The coefficien a is expeced o be posiive. The corresponding shock is inerpreed as an ineres rae (moneary policy) shock by he cenral bank. Third, inflaion is no conemporaneously responsive o he ineres rae because of he lagged effec of moneary policy. This equaion is comparable o he New-Keynesian Phillips curve. The coefficiens a 2 and a 3 are expeced o be posiive, and a 3 is expeced o be less han uniy because of price sickiness. The corresponding shock is inerpreed as an unexpeced inflaion shock. In he fourh equaion, no zero resricion is imposed. This is because we aim o idenify he deerminans of inflaion expecaions, abou which lile is known. Anoher reason is ha people are assumed o form heir expecaions by aking accoun of all he currenly available informaion. We expec he coefficiens a 4, a 5 and a 6 o be posiive. The corresponding shock is inerpreed as an inflaion expecaions shock. In his resricion, we assume simulaneous ineracions beween inflaion expecaions and inflaion and beween inflaion expecaions and he ineres rae. The inflaion expecaions shock conemporaneously and indirecly affec inflaion and he ineres rae, because in he second and hird equaions, he inflaion expecaions shock conemporaneously influence inflaion expecaions, and in urn, inflaion and he ineres rae. In an opposie direcion, here is a feedback from inflaion and he ineres rae o inflaion expecaions. Such simulaneous ineracions are inrinsic o nonrecursive resricions (Kims [999] and Sims and Zha [25]). For example, Kim (999) uses he world expor commodiy price index insead of inflaion expecaions, and imposes a nonrecursive resricion ha has a simulaneous ineracion beween he world expor commodiy price index and he ineres rae. He hen idenifies five srucural shocks including he moneary 6 For simpliciy, we omi he exogenous variables from his expression. 7

11 policy shock, and sudies he conribuion of he srucural shocks o oupu flucuaions. 4. Esimaion Resuls (Japan) 4. Esimaed Coefficiens From he esimaed reduced-form VAR, he coefficiens, a o a 6, are derived as shown in Table 7. To compue he corresponding confidence inervals, we use he Mone-Carlo simulaions of, draws based on he following disribuion for he reduced-form parameers and covariance marix vec( B) ~ N[ vec( Bˆ), Σ ( X ' X ) ] Σ ~ IW[( TΣˆ ), T p], where Bˆ and Σˆ are he esimaed parameers and covariance marix, N[] denoes he normal disribuion, IW[] denoes he invered Wishar disribuion, and p and T are he number of explanaory variables and observaions, respecively. By using he covariance marix from each draw, we calculae he coefficiens a o a 6 and sum hem up o yield he figures repored in Table 7. 7 As expeced, all coefficiens are posiive excep for a 3, which is insignifican. This validaes our VAR resricions. 4.2 Impulse Responses Figure 3 illusraes he impulse responses of he four endogenous variables o four srucural shocks. Each column represens a srucural shock whose magniude is one sandard error, and each row represens he responses of he endogenous variables. Solid lines represen he means, and doed lines represen he 6h and 84h perceniles. The firs column shows ha a posiive shock increases he oupu gap, inflaion, inflaion expecaions and he ineres rae. Thus, his shock is inerpreed as he demand shock. The second column shows he responses o a posiive ineres rae shock in he moneary policy. A rise in ineres raes causes he oupu gap and inflaion expecaions o decrease on impac. Inflaion decreases on impac, oo. This resul suggess ha he price puzzle is resolved because we use he daa of inflaion expecaions for esimaion and we impose he nonrecursive resricion described as equaion (3). We also find ha inflaion expecaions respond more quickly han does inflaion. This resul suggess ha here is price sickiness and ha households correcly forecas he fuure direcion of price changes following a moneary policy shock. The response of inflaion on impac is no large, bu is persisen. The 7 Because of nonrecursive resricions, in he Mone-Carlo simulaion, we did no always obain finie coefficiens. In such cases, we ook anoher draw. For his and subsequen esimaions, he probabiliy of having o redraw was abou % o 3%. 8

12 hird column suggess ha a posiive inflaion shock raises inflaion, and hereby ineres raes, bu is impac on inflaion expecaions is small. Oupu decreases. Therefore, his shock resembles a negaive supply shock. The fourh column illusraes he responses o a posiive inflaion expecaions shock. Such a shock raises inflaion expecaions on impac, and his increase is accompanied by an increase in ineres raes. However, is effec on inflaion is insignifican. To sum up, hese impulse responses appear reasonable, which validaes our VAR resricions. Figure 4 illusraes he impulse responses of he four endogenous variables o one-sandard-error changes in he exogenous variables. Inflaion expecaions promply reac o changes in oil prices and fresh food prices. These effecs are ransiory. Realized inflaion reacs relaively slowly o changes in oil prices and fresh food prices, bu hese effecs are persisen. The slow response of inflaion o fresh food prices is parly explained by he fac ha our price index excludes fresh food. In summary, hese resuls sugges ha households inflaion expecaions lead realized inflaion because households expecaions reac o boh moneary policy changes and changes in oil prices and fresh food prices more quickly han does realized inflaion. This reflecs he sluggish response of realized inflaion o expeced inflaion. 4.3 Variance Decomposiions Table 8 repors variance decomposiions showing he conribuions made by srucural shocks and exogenous variables o he forecas error variances of realized and expeced inflaion a horizons of, 4, 8 and 2 quarers. 8 This able shows, firs, ha he exogenous variables have large shor-run effecs on realized and expeced inflaion. Their conribuion is almos 3% a horizons of and 4 quarers. However, heir effecs are no persisen. 9 Our second finding is ha he moneary policy shock has a large effec. The conribuion of he moneary policy shock o expeced inflaion is abou 6% in he shor run. Is conribuion o inflaion is negligible in he shor run probably because of price sickiness, bu increases o 3% wihin wo o five years. Third, he conribuion of he demand shock o realized and expeced inflaion is smaller han 5% a horizons of and 4 quarers, bu larger han 5% a a horizon of 2 quarers. Fourh, he inflaion expecaion shock has a limied effec. These findings sugges ha changes in 8 Because here is apparenly no convenional way o calculae he conribuion of he exogenous variables, we simply add heir one-sandard-error deviaions o our VAR model, while we add one-sandard-error deviaions of he srucural shocks o calculae he conribuion of he srucural shocks. This mehod would be accurae if he exogenous variables were whie noise. In our model, because we use changes in (raher han levels of) he exogenous variables, our exogenous variables are saionary and no srongly auocorrelaed. 9 The conribuion of fresh food prices o realized inflaion is sill abou 3% a a horizon of 8 quarers. This is largely caused by he sample of he 97s. 9

13 realized inflaion are largely caused by changes in oil prices and fresh food prices in he shor run, bu in he medium o long run, are mainly changed by fundamenals such as oupu and moneary policy. This is consisen wih Shirasuka (26). The same resul is applied o changes in expeced inflaion. The second and fourh of hese findings for Japan differ from hose for he Unied Saes (see Secion 5). 4.4 Srucural Shocks Figure 5 shows he hisorical movemens of he wo srucural shocks, he inflaion shock and he expeced inflaion shock. The mos noable and ineresing propery is ha he inflaion expecaion shock is occasionally unsable. During he oil price shocks of he 97s and early 98s, and during he asse price bubble and burs of he lae 98s and he early 99s, he expecaion shock exhibied large flucuaions. This implies ha an expecaion shock occasionally affec inflaion and he real economy. In he lae 99s and early 2s, here were concerns abou a deflaionary spiral, bu he figure does no jusify hese concerns. Up unil he early 98s, when he inflaion shock was sabilized, here were large swings in he inflaion shock. 5. Esimaion Resuls (he Unied Saes) In his secion, we repor our esimaion resuls for he Unied Saes. We also confirm ha our mehodology is validaed by he US daa. Table 9 shows ha, as expeced, all he esimaed coefficiens are posiive excep for a 2, which is insignifican. The coefficien a 3 lies beween zero and uniy, which suggess price sickiness and ha households are forward-looking. The coefficien a 6 is posiive bu insignifican. Figures 6 and 7 illusrae he impulse responses. They show ha inflaion expecaions respond quickly o a moneary policy shock. Exogenous changes in energy prices and food prices affec inflaion wih a lag, bu inflaion expecaions reac quickly o changes in hese prices. These resuls are consisen wih hose for Japan, and explain why inflaion expecaions lead realized inflaion. An ineresing difference beween he US and Japanese resuls is ha realized inflaion reacs significanly o an inflaion expecaions shock in he Unied Saes. This implies ha, in he Unied Saes, expecaions shocks are self-fulfilling. Table repors he variance decomposiions, which convey he effecs of moneary policy shocks. For all ime horizons, heir conribuion o inflaion expecaions amouns o abou 5%. The fac ha he conribuion of a moneary policy shock o realized inflaion increases o up o 4% as he ime horizon lenghens reflecs price sickiness. The exogenous variables (changes in energy prices and food prices) have large effecs, wih conribuion reaching abou 2%. A conras wih Japan is ha he effecs are persisen in he Unied Saes bu only ransiory in Japan. This may explain why Japan experienced sharper

14 disinflaion in he 98s han did he Unied Saes. Table also shows ha expecaions shocks accoun for abou % variaions in inflaion and inflaion expecaions. This is wice as large as he corresponding conribuion for Japan. Figure 8 illusraes he movemens of wo srucural shocks, namely, he inflaion shock and he inflaion expecaions shock. The inflaion shock exhibied large flucuaions during he wo periods of he oil price shocks, since when i gradually sabilized. The inflaion expecaions shock was paricularly volaile during he period of he second oil price shock. 6. Robusness 6. Japan Figures 9 o 8 examine he robusness of our esimaes in differen ways. Firs, we look a Japan. In addiion o our baseline sample, 975:Q~27:Q4, we use differen samples, ha is, 975:Q~989:Q4, 984:Q~27:Q4, and 975:Q~999:Q. The firs wo samples divide our full sample ino early and lae subperiods. The hird sample ends in 999:Q, when he Bank of Japan lowered is policy ineres rae o almos zero. Figure 9 shows ha our esimaes are quie robus. However, for he 975:Q~989:Q4 sample, he moneary policy shock does no seem o be properly idenified, judging from he observaion ha inflaion rises following a posiive ineres rae shock. Figure shows ha if he esimaing sample sars a 983:Q, he effecs of he exogenous variables on inflaion diminish. This is consisen wih inflaion having become more sable since he early 98s. However, i is worh noing ha he effecs of he exogenous variables on inflaion expecaions have no changed much. Second, we esimae our model by exending he number of lags from wo o four, using he compued oupu gap obained from applying Hodrick Presco filer, and excluding he exogenous variables. Figure shows ha our resuls are robus o hese modificaions. Third, we impose he SVAR resricions shown in Figure 2. Under he firs resricion (idenificaion resricion ), in equaion (4), we allow he curren oupu gap o be affeced by he curren change in inflaion expecaions. Under he second resricion (idenificaion resricion 2), we allow he ineres rae o respond no only o inflaion expecaions bu also e o curren inflaion. Tha is, we relax he zero resricion on π in he y equaion and he zero resricion on π in he i equaion. However, because we had o impose exacly six resricions o idenify four srucural shocks, we mus inroduce an addiional resricion. To do so, we assume ha curren inflaion is no affeced by curren oupu. Figure 2 demonsraes ha imposing hese idenificaion resricions does no change our resuls much. However, imposing he idenificaion resricion 2 speeds up he response of inflaion o a Raher han using he convenional smoohness facor λ =,6, we used λ = 2, because his value generaes cyclicaliy ha is beer suied o Japanese business cycles.

15 moneary policy shock, compared wih our baseline model. If we impose a recursive e resricion on he ordering of (y, π, i, π ) from he exogenous variables (which seems o be he mos reasonable ordering), hen he impulse responses change dramaically (idenificaion resricion 3). However, his resricion does no seem appropriae because, as he second column shows, a posiive ineres rae shock increases boh inflaion and expeced inflaion. This suggess ha such a recursive resricion canno be imposed o correcly idenify a moneary policy shock. Fourh, we examine he robusness of our esimaes o using differen measures of inflaion expecaions. In ransforming qualiaive survey daa ino quaniaive inflaion expecaions, we used he simples and mos widely used mehod, developed by Carlson and Parkin (975). However, as shown in Figure 3, here are many variaions o his mehod. For example, Hiraa and Kamada (26) adjus he mehod o correc for an upward bias ha seems inrinsic o households inflaion expecaions. This adjusmen yields much lower inflaion expecaions han ours for he whole sample period. Having argued ha survey quesions relae o changes in inflaion raes raher han o changes in prices, Toyoda (987) calculaes inflaion expecaions by using realized inflaion raes for each period. This mehod generaes inflaion expecaions ha are similar o realized inflaion raes. We reesimae he SVAR by using hese alernaive measures of inflaion expecaions. As Figure 4 shows, he resuling impulse responses are similar o hose from our baseline model. However, because of srong correlaion wih realized and expeced inflaion, using Toyoda s (987) inflaion expecaions causes he responses of inflaion and expeced inflaion o be similar o each oher. Fifh, we esimae he VAR in levels. We use real GDP insead of he oupu gap, price levels insead of inflaion raes, and expeced prices, ha is, price levels muliplied by inflaion expecaions, insead of expeced inflaion raes. We impose he same nonrecursive resricions as before. Figure 5 illusraes ha our esimaes are robus o his modificaion. 6.2 The Unied Saes For he Unied Saes, Figures 6 and 7 illusrae he impulse responses for he wo subperiods. Alhough he figures are similar, he effecs of he moneary policy on inflaion and inflaion expecaion are slighly larger in he 97:Q~983:Q4 period han in he 984:Q~27:Q4 period. The same is rue for he effecs of he exogenous variables on inflaion and inflaion expecaions. Figure 8 illusraes he impulse responses produced by increasing he number of lags from wo o four, excluding he exogenous variables, and using he CPI insead of he core PCE deflaor. Our resuls are robus o hese modificaions. 2

16 7. Concluding Remarks In his paper, we invesigae he deerminans of households inflaion expecaions in Japan and he Unied Saes. We esimae a vecor auoregression in which he four endogenous variables are inflaion expecaions, inflaion, he shor-erm nominal ineres rae and he oupu gap, wih changes in energy prices and (fresh) food prices being he exogenous variables. We impose shor-erm nonrecursive resricions aking accoun of he simulaneous co-dependence beween realized inflaion and expeced inflaion. Our findings can be summarized as follows. Firs, responding no only o changes in energy prices and food prices bu also o moneary policy shocks, inflaion expecaions move more quickly han does realized inflaion. This explains why inflaion expecaions appear o lead realized inflaion according o Japanese and US daa. I is also shown ha our SVAR procedure resolves he price puzzle; an ineres rae rise decreases prices even on impac. Second, in Japan and he Unied Saes, changes in energy prices and food prices have large effecs on inflaion and inflaion expecaions. However, in Japan, hese effecs are ransiory, whereas in he Unied Saes, hey are persisen. Third, expecaions shocks occasionally flucuae grealy, and could have self-fulfilling effecs on realized inflaion. This self-fulfilling propery is more apparen in he Unied Saes han in Japan. In implemening moneary policy, we need o forecas underlying inflaion, and our findings sugges ha expeced inflaion is a useful indicaor. However, since he effecs of energy prices and food prices on realized and expeced inflaion are only emporary, i is pracically very imporan o adjus hese effecs and exrac underlying expeced inflaion movemens caused by fundamenals such as changes in oupu and ineres raes. I is also imporan o pay aenion o he risk ha inflaion expecaions flucuae grealy and have self-fulfilling effecs on realized inflaion. To invesigae he properies of inflaion expecaions in more deail, we may need o subjec our esimaes o furher robusness checks. This may include imposing oher resricions, making inernaional comparisons and refining he ime-series analysis. Alhough we focused on households, i is imporan o analyze firms and professionals. I is also imporan o sudy long-erm inflaion expecaions. These are our fuure research asks. 3

17 References Bachelor, R.A. and P. Dua (989), Household versus economis forecass of inflaion: A reassessmen, Journal of Money, Credi and Banking, 2(2), Berk, J.M. (2), Consumers inflaion expecaions and moneary policy in Europe, De Nederlandsche Bank Working Paper. Bernanke, B.S. (27), Inflaion expecaions and inflaion forecasing, Remarks a he NBER Moneary Economics Workshop. Bernanke, B.S. and I. Mihov (998), Measuring moneary policy, Quarerly Journal of Economics, 3, Carlson, J.A. and M. Parkin (975), Inflaion expecaions, Economica, 42, Chrisiano, L.J., M. Eichenbaum and C.L. Evans (999), Moneary policy shocks: wha have we learned and o wha end? in J.B. Taylor and M. Woodford (eds.), Handbook of Macroeconomics Vol.A, Amserdam: Elsevier Science, Erceg, C.J. and A.T. Levin (23), Imperfec credibiliy and inflaion persisence, Journal of Moneary Economics, 5, Gramlich, E.M. (983), Models of inflaion expecaions formaion: A comparison of household and economis forecass, Journal of Money, Credi and Banking, 5(2), Hara, N., N. Hirakaa, Y. Inomaa, S. Io, T. Kawamoo, T. Kurozumi, M. Minegishi and I. Takagawa (26), The new esimaes of oupu gap and poenial growh rae, Bank of Japan Review, E 3. Hiraa, W. and K. Kamada (26), Measuring inflaion expecaions wih a bias in survey daa, mimeo. Hori, M. and S. Shimizuani (23), Wha changes deflaionary expecaions? evidence from Japanese household-level daa, ESRI Discussion Paper Series, No.65. Kim, S. (999), Do moneary policy shocks maer in he G-7 counries? Using common idenifying assumpions abou moneary policy across counries, Journal of Inernaional Economics, 48, Mankiw, N.G., R. Reis, and J. Wolfers (23), Disagreemen abou inflaion expecaions, NBER Macroeconomic Annual, Mullineaux, D.J. (98), Inflaion expecaions and money growh in he Unied Saes, American Economic Review, 7(), Oshima, I. and K. Nakayama (999), On inflaion expecaions, Bank of Japan Working Paper Series, 99 7 (in Japanese). Orphanides, A. and J.C. Williams (22), Imperfec knowledge, inflaion expecaions, and 4

18 moneary policy, FRB San Francisco Working Paper. Pearce D.K. (987), Shor-erm inflaion expecaions: Evidence from a monhly survey: noe, Journal of Money, Credi and Banking, 9(3), Robers, J.M. (998), Inflaion expecaions and he ransmission of moneary policy, FRB Working Paper. Shirasuka, S. (26), Core Indicaors of Japan s Consumer Price Index, Bank of Japan Review, E-7. Sims, C.A. (98), Macroeconomics and realiy, Economerica, 48, 48. Sims, C.A. (992), Inerpreing he macroeconomic ime series facs: he effecs of moneary policy, European Economic Review, 36, 975. Sims, C.A. and T. Zha (25), Does moneary policy generae recessions? mimeo. Thomas, L.B.Jr. (999), Survey measures of expeced U.S. inflaion, Journal of Economic Perspecives, 3(4), Toyoda, T. (987), Inflaion: formaion of inflaion expecaions and he Phillips curve, in K. Hamada e al. (eds.) Macroeconomic analysis of he Japanese economy, Tokyo-daigaku-shuppankai, (in Japanese). 5

19 Name Respondens Organizaion Sar dae Descripion Consumer Households Cabine 97 Price changes in he following year Confidence Survey (abou 5, cases; abou 75% response Office (qualiaive, quarerly) (or six monhs depending on he surveyed periods) rae) 24 Three (or five) qualiaive choices (quaniaiv e, monhly) unil 24 and seven quaniaive choices from 24. Quarerly unil 24, monhly from 24 Opinion Households Bank of 999 Price changes in he following one Survey on he (abou 4, cases; Japan (qualiaive) and five years General abou 5% response 24 Quarerly Public s Views and Behavior rae) (quaniaiv e) Consumer Households Nippon 98 Price changes in he following year Senimen (abou,2 Research Three qualiaive choices Index Survey responses Insiue Every wo monhs TANKAN Enerprises Bank of 974 Oupu price changes in he (abou, cases Japan (no daa in following hree monhs including 7 reailers; March Three qualiaive choices abou % response 975) Quarerly rae) ESP Forecas 36 economiss who Economic 24 Inflaion raes in he following one Survey work in Japan (May Planning quarer o wo years 27) Associaion Monhly Consensus More han 24 Consensus 99 Inflaion raes in he following one Forecass economiss in he G7 Economics quarer o en years and Wesern counries Quarerly, bu long-erm forecass are semiannual. Table : Surveys of Inflaion Expecaions (Japan) 6

20 Quarers (leads of inflaion expecaions) Sample 97:Q 97:Q 99:Q 97:Q 97:Q 27:Q4 27:Q4 27:Q4 27:Q4 27:Q4 CPI coverage Excl. fresh Excl. fresh Excl. fresh Excl. food All food food food and energy CPI Consumpion ax nonadjused adjused nonadjused nonadjused nonadjused Table 2: Correlaions beween Inflaion and Inflaion Expecaions (Japan) Noe: Yellow highlighs represen he highes correlaion. 7

21 Respondens Quarers (leads of inflaion Households Reailers Economiss expecaions) Sample 97:Q 27:Q4 975:Q2 27:Q4 99:H2 27:H (semiannual) Consumer Survey source Confidence Survey TANKAN Consensus Forecass (Carlson Parkin (Reailers, DI) (nex year forecass) mehod) CPI coverage Excl. fresh food Goods excl. fresh food Excl. fresh food CPI consumpion ax non-adjused adjused nonadjused Table 3: Comparison of Inflaion Expecaions Surveys in erms of Correlaion wih Inflaion (Japan) 8

22 Name Respondens Organizaion Descripion Survey of Consumers (Michigan Survey) Livingson Survey Survey of Professional Forecasers Consensus Forecass Households Universiy of Price changes in he following year (5 elephone Michigan Quaniaive inerviews) Quarerly Abou 4 economiss FRB Philadelphia CPI and PPI changes in he following six monhs o en years Quaniaive Semi-annual Abou 5 economiss FRB Philadelphia CPI and PCE deflaor changes in he following year and in he long run Quaniaive Quarerly More han 24 Consensus Inflaion raes in he following one economiss in G7 and Economics quarer o en years Wesern counries Quarerly, bu long-erm forecass are semiannual. Table 4: Surveys of Inflaion Expecaions (he Unied Saes) 9

23 Quarers (leads of inflaion expecaions) Sample 97:Q 97:Q 99:Q 99:Q 27:Q4 27:Q4 27:Q4 27:Q4 Price index CPI Core PCE CPI Core PCE Table 5: Correlaions beween Inflaion and Inflaion Expecaions (he Unied Saes) Respondens Quarers (leads of inflaion Households Economiss expecaions) Sample 97:Q 27:Q4 97:H 27:H2 (semiannual) Survey source Michigan Survey Livingson Survey Price index CPI CPI Table 6: Comparison of Inflaion Expecaions Surveys in erms of Correlaion wih Inflaion (he Unied Saes) 2

24 Coefficiens 5% median 95% a a a a a a Table 7: Esimaed Coefficiens (Japan) () Inflaion Rae e_y e_i e_p e_pe dpoil dpfresh T = T = T = T = (2) Inflaion Expecaions e_y e_i e_p e_pe dpoil dpfresh T = T = T = T = Table 8: Variance Decomposiions (Japan) 2

25 Coefficiens 5% median 95% a a a a a a Table 9: Esimaed Coefficiens (he Unied Saes) () Inflaion Rae e_y e_i e_p e_pe dpenergy dpfood T = T = T = T = (2) Inflaion Expecaions e_y e_i e_p e_pe dpenergy dpfood T = T = T = T = Table : Variance Decomposiions (he Unied Saes) 22

26 (%, y/y) (DI inverse) 25 2 CPI (excluding fresh food, consumpion ax nonadjused, lef axis) Inflaion expecaions (Carlson Parkin mehod, lef axis) 5 Inflaion expecaions (DI, righ axis) Figure : Movemen of Inflaion Expecaions (Japan) 6 (%, y/y) 4 2 CPI Core PCE Michigan Survey Figure 2: Movemen of Inflaion Expecaions (he Unied Saes) 23

27 Impulse responses e_y e_i e_p e_p e Y Responses of I pi pie e_y e_i e_p e_p e Figure 3: Impulse Responses o Srucural Shocks (Japan) Impulse responses o exogenous variable changes dpoil dpfresh Y I pi pie dpoil dpfresh Figure 4: Impulse Responses o Exogenous Variables (Japan) 24

28 5 4 e_p 3 e_pe Figure 5: Srucural Shocks (Japan) 25

29 Impulse responses e_y e_i e_p e_pe Y Responses of I pi pie e_y e_i e_p e_pe Figure 6: Impulse Responses o Srucural Shocks (he Unied Saes) Impulse responses o exogenous variable changes Y dpnergy dpfood 5 5 I pi pie dpnergy dpfood Figure 7: Impulse Responses o Exogenous Variables (he Unied Saes) 26

30 6 5 e_p 4 e_pe Figure 8: Srucural Shocks (he Unied Saes) 27

31 Y e_y e_i e_p e_pe ( % Baseline 984:-27:4 975:-999: 975:-989: I Pi Pie Figure 9: Robusness.: Differen Samples (Japan) dpoil dpfresh Y Baseline 984:-27:4 975:-999: 975:-989: I Pi Pie Figure Robusness.2: Differen Samples (Japan) 28

32 Y e_y e_i e_p e_pe ( % Baseline Lags = 4 No exogenous variables HP filered gap I Pi Pie Figure : Robusness 2: Miscellaneous Modificaions (Japan) e_y e_i e_p e_pe Y Baseline Idenificaion Idenificaion 2 Idenificaion I Pi Pie Figure 2: Robusness 3: Oher Resricions (Japan) 29

33 3 (%, y/y) CPI (excluding fresh food) Carlson and Parkin (975) Hiraa and Kamada (26) Toyoda (987) DI (righ axis) Figure 3: Various Measures of Inflaion Expecaions (Japan) Y e_y e_i e_p e_pe ( % Baseline Kamada and Hiraa (26) Diffusion Index Toyoda (987) I Pi Pie Figure 4: Robusness 4: Various Measures of Inflaion Expecaions (Japan) 3

34 e_y e_i e_p e_pe Y Baseline VAR in levels I Pi Pie Figure 5: Robusness 5: VAR in Levels (Japan) 3