Woring Paper Series No 1077 / augus 009 The recepion of pubic signas in financia mares wha if cenra ban communicaion becomes sae? by Michae Ehrmann and David Sondermann
WORKING PAPER SERIES NO 1077 / AUGUST 009 THE RECEPTION OF PUBLIC SIGNALS IN FINANCIAL MARKETS WHAT IF CENTRAL BANK COMMUNICATION BECOMES STALE? 1 by Michae Ehrmann and David Sondermann 3 In 009 a pubicaions feaure a moif aen from he 00 bannoe. This paper can be downoaded wihou charge from hp://www.ecb.europa.eu or from he Socia Science Research Newor eecronic ibrary a hp://ssrn.com/absrac_id=1440673. 1 We woud ie o han Marin T. Boh, Phiippine Cour-Thimann, Rein Gropp, Bernd Wifing, an anonymous referee, and seminar paricipans a he for hepfu commens and suggesions. This paper presens he auhors persona opinions and does no necessariy refec he views of he European Cenra Ban. European Cenra Ban, Kaisersrasse 9, D-60311 Franfur am Main, Germany; e-mai: Michae.Ehrmann@ecb.europa.eu 3 Universiy of Münser, Deparmen of Economics, Am Sadgraben 9, D-48143 Münser, Germany; e-mai: david.sondermann@wiwi.uni-muenser.de
European Cenra Ban, 009 Address Kaisersrasse 9 60311 Franfur am Main, Germany Posa address Posfach 16 03 19 60066 Franfur am Main, Germany Teephone +49 69 1344 0 Websie hp://www.ecb.europa.eu Fax +49 69 1344 6000 A righs reserved. Any reproducion, pubicaion and reprin in he form of a differen pubicaion, wheher prined or produced eecronicay, in whoe or in par, is permied ony wih he expici wrien auhorisaion of he or he auhor(s). The views expressed in his paper do no necessariy refec hose of he European Cenra Ban. The saemen of purpose for he Woring Paper Series is avaiabe from he websie, hp://www.ecb.europa. eu/pub/scienific/wps/dae/hm/index. en.hm ISSN 175-806 (onine)
CONTENTS Absrac 4 Non-echnica summary 5 1 Inroducion 7 Modeing he recepion of differen pubic signas in financia mares 10.1 The heoreica mode 10. The daa and he economeric mode 13 3 Financia mare reacions o he reease of pubic signas 16 3.1 The uncondiiona recepion of pubic signas 17 3. Wha if cenra ban communicaion becomes sae? 18 4 Concusions 1 References Tabes and figures 4 European Cenra Ban Woring Paper Series 3 Woring Paper Series No 1077 Augus 009 3
Absrac How do financia mares price new informaion? This paper anayzes price seing a he inersecion of privae and pubic informaion, by esing wheher and how he reacion of financia mares o pubic signas depends on he reaive imporance of privae informaion in agens informaion ses a a given poin in ime. I sudies he reacion of UK shor-erm ineres raes o he Ban of Engand s infaion repor and o macroeconomic announcemens. Due o he quarery frequency a which he Ban of Engand reeases one of is main pubicaions, i can become sae over ime. In he course of his process, financia mare paricipans need o rey more on privae informaion. The paper deveops a syized mode which predics ha, he more ime has eapsed since he aes reease of an infaion repor, mare voaiiy shoud increase, he price response o macroeconomic announcemens shoud be more pronounced, and macroeconomic announcemens shoud pay a more imporan roe in aigning agens informaion se, hus eading o a sronger voaiiy reducion. The empirica evidence is fuy supporive of hese hypoheses. Keywords: pubic signas, infaion repors, moneary poicy, ineres raes, announcemen effecs, co-ordinaion of beiefs, Ban of Engand. JEL Cassificaion: E58, E43, G1, G14. 4 Woring Paper Series No 1077 Augus 009
Non-echnica summary This paper sudies he imporance of privae and pubic informaion in he price discovery process of financia mares. In paricuar, i ess wheher and how he reacion of financia mares o pubic signas depends on he reaive imporance of privae informaion in agens informaion ses a a given poin in ime. In order o pin down he reaive imporance of privae informaion, i sudies he exampe of a financia mare where one of he mos decisive pubic signas is issued ony a a ow frequency and can herefore become sae over ime. In he course of his process, financia mare paricipans need o rey more on privae informaion, such ha he reaive imporance of privae informaion is idenified by he ime ha has eapsed since he as updae of he pubic signa. Oher pubic signas, even if ess direcy reevan for he pricing of he underying asse han he ow frequency signa, wi be used by agens o updae heir informaion se in he meanime. The paper assumes ha pubic signas serve o homogenize he informaion se of raders, and as such shoud ead o a convergence of views. This effec shoud be magnified when privae signas are reaivey dominan, and raders informaion ses accordingy heerogeneous, suggesing a sronger mare reacion o he reease of a pubic signa. The empirica appicaion of his paper sudies he roe of cenra ban communicaion for shor-erm ineres raes, given ha hese depend cosey on he course of moneary poicy. The paper invesigaes wheher he price discovery process depends on he freshness of informaion ha has been provided by he cenra ban. Wih more recen, and hus more precise, pubic informaion abou he curren hining wihin he cenra ban, he imporance of privae informaion shoud be reaivey imied. Wih growing disance o he reease of he cenra ban s pubic signa, is informaion conen is becoming sae, hus eaving more room for heerogeneous beiefs and privae informaion. We wi use he Ban of Engand s infaion repor as a esing case. The infaion repor is reeased quarery, and conains an indeph anaysis of he curren economic siuaion in he UK, as we as forecass for GDP and infaion. Imporany, he Ban of Engand s infaion repor is signed off by he Moneary Poicy Commiee (MPC), i.e. he body ha decides on he Ban s poicy raes, and has been judged o be of very high quaiy. We herefore conjecure ha he infaion repor conveys a raher precise (abei infrequen) signa on he curren hining of he poicy maers, such ha i shoud be cosey wached by financia mare paricipans. In he inerim period beween wo infaion repors, financia mare paricipans can updae heir beiefs abou he course of he economy and he iey seing of moneary poicy based on he informaion conen of reeases of macroeconomic daa. As se ou above, our hypohesis is ha hese serve o homogenize agens views, bu ha financia mare reacions o heir reease vary in beween wo infaion repors, wih he effecs becoming arger over ime. In ine wih he hypohesis ha pubic signas serve o homogenize agens informaion ses, he paper finds ha condiiona voaiiy decines in response o boh he infaion repor and macroeconomic news. The increasing reiance on privae informaion wih growing disance o he as reease of an infaion repor is aso confirmed, as condiiona voaiiy generay rises over ime in beween wo repors. Furhermore, in response o macroeconomic reeases price adjusmens become more pronounced, and he reducion of condiiona voaiiy aso srenghens, suggesing ha hese pubic signas are given more weigh in he updaing of beiefs, he more disan he as cenra ban communicaion, and herefore he arger he roe payed by privae informaion. These findings have a number of imporan impicaions. In he ream of moneary poicy, cenra bans rouiney sudy he mare response o macroeconomic news. To appropriaey judge he magniude of such mare movemens, i is imporan o compare hese agains he correc benchmar, which is no he average response, bu raher he average response Woring Paper Series No 1077 Augus 009 5
condiiona on he freshness of cenra ban communicaion. An above average response o he reease of infaion daa shory prior o he pubicaion of a major cenra ban reease, for insance, is o be expeced, and shoud be assessed accordingy. Beyond a pure moneary poicy perspecive, hese findings migh hep expain why i is a imes difficu o esabish a in beween asse prices and macroeconomic fundamenas. Wha he curren paper suggess is ha a missing (or a mued) responsiveness need no auomaicay impy a broen in. Raher, mued responsiveness can very we arise if he informaion conained in anoher source (ha coud have enered he informaion se aready a an earier sage) is sufficieny precise o dominae he new signa. Finay, i is imporan o remember ha his paper has used he case of cenra ban communicaion as a esing case for is hypoheses, bu ha he impicaions are appicabe o a much broader se of siuaions. Gropp and Kadareja (006) show reaed effecs for soc mares, where he price discovery process depends on he disance from he as reease of commercia bans annua repors. Taen ogeher, hese findings impy ha ransparency by acors ha have he poenia o affec asse prices (be i poicy maers or pubicy raded companies) is generay beneficia. 6 Woring Paper Series No 1077 Augus 009
1. Inroducion How do financia mares price new informaion? This quesion is of fundamena imporance o financia economics, and has herefore received exensive aenion in he ieraure. The recepion of pubic signas in financia mares has been subjec o paricuar scruiny, iner aia because he news arriva process is reaivey easiy idenified in he case of pubic news. The ieraure on announcemen effecs (Ederingon and Lee 1993, Feming and Remoona 1999a, Andersen e a. 003, 007, and many ohers), for insance, has subsaniay improved our undersanding of he price discovery processes a wor in financia mares. Sudying he imporance of privae informaion, in conras, is by definiion much more difficu. Papers in his ieraure ypicay foow he approach of French and Ro (1986) and compare voaiiy of asse prices during exchange rading hours o is counerpar during nonrading hours, hoding he arriva process of pubic informaion consan. Io e a. (1998), for insance, sudy he roe of privae informaion in foreign exchange mares, by comparing is voaiiy before and afer he inroducion of unch-ime rading in Toyo (a change which shoud no have aered he fow of pubic informaion). Wih his idenificaion approach, privae informaion has been idenified as a major facor behind mare voaiiy. The curren paper aims o conribue o boh srands of he ieraure by sudying price seing a he inersecion of privae and pubic informaion. I ess wheher and how he reacion of financia mares o pubic signas depends on he reaive imporance of privae informaion in agens informaion ses a a given poin in ime. In order o pin down he reaive imporance of privae informaion, we wi oo a he exampe of a financia mare where one of he mos decisive pubic signas is issued ony a a ow frequency, such ha i can become sae over ime. In he course of his process, financia mare paricipans need o rey more on privae informaion, such ha he reaive imporance of privae informaion is idenified by he ime ha has eapsed since he as updae of he pubic signa. Oher pubic signas, even if ess direcy reevan for he pricing of he underying asse han he ow frequency signa, wi be used by agens o updae heir informaion se in he meanime. The paper assumes ha pubic signas serve o homogenize he informaion se of raders, and as such shoud ead o a convergence of views. This effec shoud be magnified when privae signas are reaivey dominan, and raders informaion ses accordingy heerogeneous, suggesing a sronger mare reacion o he reease of a pubic signa. This paper is herefore cosey reaed o Gropp and Kadareja (006), who sudy he response of reaized voaiiy of commercia bans socs o moneary poicy shocs, and show ha he magniude of his response increases wih he ime since he commercia ban has reeased is annua repor. In he curren paper, we wi oo a anoher exampe of his mechanism, foowing a somewha differen approach, which aows disenanging he response of he condiiona mean and he condiiona variance. Furhermore, we are ineresed in onger-erm effecs, and wi herefore use daiy raher han inraday daa. The exampe we sudy reaes o cenra ban communicaion as a pubic signa of direc and major reevance for shor-erm ineres raes (which are nown o cosey depend on he course of moneary poicy). We wi invesigae wheher he price discovery process depends on he freshness of informaion ha has been provided by he cenra ban. Wih more recen, and hus more precise, pubic informaion abou he curren hining wihin he cenra ban, he imporance of privae informaion shoud be reaivey imied. Wih growing disance o he reease of he cenra ban s pubic signa, is informaion conen is becoming sae, hus eaving more room for heerogeneous beiefs and privae informaion. We wi use he Ban of Engand s infaion repor as a esing case. The infaion repor is reeased quarery, and conains an in-deph anaysis of he curren economic siuaion in he UK, as we as forecass Woring Paper Series No 1077 Augus 009 7
for GDP and infaion. Imporany, he Ban of Engand s infaion repor is signed off by he Moneary Poicy Commiee (MPC), i.e. he body ha decides on he Ban s poicy raes, and has been judged o be of very high quaiy (Fracasso e a. 003). We herefore conjecure ha he infaion repor conveys a raher precise (abei infrequen) signa on he curren hining of he poicy maers, such ha i shoud be cosey wached by financia mare paricipans. In he inerim period beween wo infaion repors, financia mare paricipans can updae heir beiefs abou he course of he economy and he iey seing of moneary poicy based on he informaion conen of reeases of macroeconomic daa. As deaied above, our hypohesis is ha hese announcemens serve o homogenize agens views, bu ha financia mare reacions o heir reease vary in beween wo infaion repors, wih he effecs becoming arger over ime. Modeing he mare response o signas by means of GARCH modes aows separaing effecs on he condiiona mean from hose on he condiiona voaiiy equaion. This approach is disinc from severa oher reaed papers, which sudy he response of eiher asse prices (wihou a simuaneous modeing of he condiiona variance, such as Gürayna e a. 006), or (reaized) voaiiy (such as Gropp and Kadareja 006). The separaion of he wo is imporan for he purposes of his paper: condiiona voaiiy shoud fa in response o a pubic signa ha homogenizes agens views, whereas reaized voaiiy migh acuay increase (due o a response of he condiiona mean equaion). As a maer of fac, we find very srong evidence ha he condiiona voaiiy of shor-erm ineres raes is significany reduced in response o he reease of he Ban of Engand s infaion repor, as we as in response o macroeconomic reeases, especiay when separaing he incidence of reeases from heir surprise componen. A he same ime, he paper provides evidence ha he mare reacion o he aer is srongy dependen on he ime ha has eapsed since he reease of he preceding infaion repor. Three findings are imporan in ha respec. Firs, condiiona voaiiy generay increases wih he disance o he preceding infaion repor, suggesing ha as cenra ban communicaion becomes sae, privae informaion pays a reaivey arger roe in mare paricipans informaion se. Second, he response of he mean equaion o macroeconomic reeases becomes more pronounced, and hird, he reducion of condiiona voaiiy in response o macroeconomic reeases is aso sronger, he arger he ime gap since he as updae of he infaion repor. These findings have a number of imporan impicaions. In he ream of moneary poicy, cenra bans rouiney sudy he mare response o macroeconomic news. To appropriaey judge he magniude of such mare movemens, i is imporan o compare hese agains he correc benchmar, which is no he average response, bu raher he average response condiiona on he freshness of cenra ban communicaion. An above average response o he reease of infaion daa shory prior o he pubicaion of a major cenra ban reease, for insance, is o be expeced, and shoud be assessed accordingy. Beyond a pure moneary poicy perspecive, hese findings migh hep expain why i is a imes difficu o esabish a in beween asse prices and macroeconomic fundamenas (see, e.g, he discussion on his issue in Andersen e a. 003). Wha he curren paper suggess is ha a missing (or a mued) responsiveness need no auomaicay impy a broen in. Raher, mued responsiveness can very we arise if he informaion conained in anoher source (ha coud have enered he informaion se aready a an earier sage) is sufficieny precise o dominae he new signa. Finay, i is imporan o remember ha his paper has used he case of cenra ban communicaion as a esing case for is hypoheses, bu ha he impicaions are appicabe o a much broader se of siuaions. Gropp and Kadareja (006) show reaed effecs for soc mares, where he price discovery process depends on he disance from he as reease of commercia bans annua repors. Taen ogeher, hese findings impy ha ransparency by acors ha have he poenia o affec asse prices (be i poicy maers or pubicy raded companies) is generay beneficia. 8 Woring Paper Series No 1077 Augus 009
The paper adds o various differen srands of he ieraure. A very recen ieraure (surveyed in Binder e a. 008) deas wih aspecs reaed o moneary poicy, ransparency and cenra ban communicaion. In paricuar, Morris and Shin (00) and Amao, Morris and Shin (00) dea wih he probem ha communicaion by cenra bans can reduce wefare if is signas are reaivey noisy, ye are used by mare paricipans o noneheess co-ordinae heir beiefs. Wheher his impies imis o cenra ban ransparency in pracice has been chaenged by Svensson (006), however, on he grounds ha cenra ban informaion may no be sufficieny noisy o ead o his resu. Ehrmann and Frazscher (007b) es he predicions of he mode empiricay, and find suppor for he Morris and Shin hypohesis. A reaed empirica ieraure assesses how cenra ban ransparency and communicaion affecs financia mares. There is ampe evidence ha communicaion exers a subsania impac on asse prices (Guhrie and Wrigh 000, Kohn and Sac 004, Andersson e a. 006, Ehrmann and Frazscher 007a). Reinhar and Sac (006) as we as Reeves and Sawici (007) show ha in paricuar communicaion on behaf of he enire poicy-maing commiee is a srong mare mover; for he case of he Ban of Engand, he infaion repor has been shown o be paricuary imporan. The paper aso is cosey reaed o he ieraure on announcemen effecs in financia mares. Ederingon and Lee (1993) demonsrae ha prices of ineres raes and foreign exchange fuures mares respond significany o macroeconomic news announcemens. Feming and Remoona (1999a) expand his anaysis by furhermore esing he effec on rading voume and bid-as spreads, and idenify wo disinc adjusmen processes. In he firs sage, prices adjus neary insananeousy o news, hus confirming he conjecure of French and Ro (1986) ha pubic informaion affecs prices before anyone can rade on i. In he second sage, which is somewha more persisen, rading voume surges and price voaiiy persiss, as residua disagreemens among he raders abou he inerpreaion of he news riggers rading. Noe ha hese wo periods span a coupe of hours, i.e. are ypicay concuded wihin he course of a rading day. Effecs of macroeconomic news on soc, bond and foreign exchange mares have subsequeny been found in various sudies (for ineres raes in paricuar see Baduzzi e a. 001, Godberg and Leonard 003, and Faus e a. 007). Andersen e a. (003 and 007) documen imporan asymmeries in ha respec, e.g. a greaer impac of bad news han of good news on exchange raes, or a differen response of equiy mares depending on he sage of he business cyce. Whie mos evidence has focused on he US, a number of conribuions aso provide evidence for Briish financia mares. Gürayna e a. (006) focus on he responsiveness of ongerm infaion compensaion o macroeconomic news in he US, he UK and Sweden. They find ha hey are insensiive o news reeases in Sweden and in he UK (since he Ban of Engand was graned independence), whereas hey are in he US and were in he UK prior o 1997. The auhors concude ha infaion argeing can anchor infaion expecaions. Imporany for he purposes of his paper, hey aso show ha zero coupon yieds (obained from he same source as in his paper, and esimaed over a simiar sampe, namey 1998-005) a he 1-year mauriy are responsive o macroeconomic news. A number of papers in his ieraure esimae condiiona mean and condiiona variance equaions: Andersen e a. (003 and 007), as we as Ehrmann and Frazscher (005) using a weighed eas squares approach, Jones e a. (1998) or Fannery and Proopapadais (00) using GARCH modes, and Goodhar e a. (1993) even using GARCH in mean modes. The ey insigh from hese sudies is ha he effecs on he condiiona mean migh very we differ from he effecs on he condiiona variance. Fannery and Proopapadais (00), for insance, find ha infaion news affec ony he condiiona mean of soc reurns, whereas some rea facors rigger responses in he condiiona variance, wihou affecing he condiiona mean. Woring Paper Series No 1077 Augus 009 9
A hird srand of he ieraure ha is of reevance for his paper aemps o disenange he roe of privae and pubic informaion in financia mares, and o undersand he drivers for mare voaiiy. Shaen (1993) deveops a mode where agens dispersion of expecaions is causa for excess mare voaiiy. Simiary, in he mode by Harris and Raviv (1993), raders receive common informaion, bu inerpre i in differen ways. Boh modes are cosey reaed o our paper, where he heerogeneiy in agens beiefs is a funcion of he quaiy of he pubic signa. The oder he signa, he more iey i is ha agens inerpre i differeny, which in urn shoud ead o an eevaed voaiiy in mares. Empirica ess for he imporance of privae informaion in generaing mare voaiiy ypicay rey (foowing he approach of French and Ro 1986) on idenifying differences in voaiiy beween rading and non-rading imes, whie eeping he fow of pubic signas consan. In paricuar changes in insiuiona seings have been appied as a esing vehice. Io and Lin (199) show ha he reducion in mare voaiiy during unchime is smaer for he New Yor Soc Exchange (which does no brea for unch) han for he Toyo Soc Exchange (which does brea for unch). As menioned above, Io e a. (1998) mae use of he aboishmen of he unch brea in he Toyo foreign exchange mare. Furhermore, Barcay e a. (1990) anayze voaiiy before and afer he haf-day rading in he Toyo Soc Exchange on Saurdays was abandoned. The ieraure srongy suggess ha privae informaion is paying a arge roe in generaing mare voaiiy. The presen paper aemps o conribue o hese hree ines of research. I deveops a shor, syized mode of he recepion of differen pubic signas in financia mares, and describes he daa and he mehodoogy underying our empirica anaysis in Secion. Secion 3 provides he empirica resus, firs ooing a he uncondiiona recepion of pubic signas, and subsequeny asing how his percepion changes if cenra ban communicaion becomes sae. Secion 4 concudes.. Modeing he recepion of differen pubic signas in financia mares The purpose of his secion is o presen a simpe modeing framewor, iusraing he effec of he reease of infaion repors and of macroeconomic news and heir ineracion, 1 as we as he seup of our subsequen empirica anaysis..1 The heoreica mode Every period, agen i forms an expecaion Ei (, h ) abou he fuure pah of ineres raes, h (where he expecaion is formed over he horizon from period o period +h). The agen has a uiiy funcion of he form U i E i (1), h, h suggesing ha he agen wans her expecaion o cosey refec and be consisen wih he aer reaizaion of, h Each agen receives a privae signa abou, h each period. This signa is unbiased, ye noisy, and can be modeed as xi, h i, where i is i.i.d.-norma wih zero mean and variance, and E( i j ) 0 for i j. For simpiciy, we assume ha he variance of his signa remains fixed over ime. In he absence of oher signas, he 1 The noaion and modeing approach are borrowed from Morris and Shin (00). 10 Woring Paper Series No 1077 Augus 009
agen s expecaion wi be E i (, h ) generae heerogeneiy in expecaions. x. Noe ha i is hese privae signas which i To expand he mode, we can assume ha agens aso receive a second signa, namey a pubic signa (such as he cenra ban s infaion repor), y, h, where is aso normay disribued wih zero mean and variance, and independen of. The pubic signa is updaed ony infrequeny. Wih evoving over ime, he pubic signa becomes sae., h We mode his by means of a ime-varying variance of he pubic signa, whereby monoonicay increasing wih he eapsed ime since he pubic signa was issued. is Figure 1 iusraes hese assumpions. Each period, a privae signa is received (indicaed by he gray verica ines), wih heir precision being consan. Occasionay, a pubic signa, namey he infaion repor (depiced by means of he dashed ine), is issued. 3 Given he infrequen updae of his pubic signa, is precision erodes over ime, uni a new pubic signa is reeased. The char assumes ha he infaion repor, when being issued, provides a ceaner picure of he cenra ban s inenions han he privae signa. Noe, however, ha he resus of he mode do no depend on he reaive precision of he various signas. FIGURE 1 HERE Having avaiabe wo signas abou he fundamena, he agen now needs o decide on he weighs ha shoud be aached o each when forming her expecaions. The deermining facor for he aocaion of weighs is he reaive precision of he wo signas, wih a arger weigh being aocaed o he more precise signa. Defining he precision of he privae signa as 1/ and of he pubic signa as 1/ impies ha he expeced vaue of he fundamena is, h y xi i (, h ) h. () E i, This expression aows us o anayze he heerogeneiy in expecaions. The variance of individua expecaions across agens is given by This is a simpificaion, which does no affec our resus quaiaivey. As a maer of fac, i represens a ower bound of he oss of informaion conained in he pubic signa. If we assume a daa generaing process for he fundamena of he ype 1, h1, h 1, as period s signa eners he expecaion abou he fundamena (in he absence of oher signas) as E1( 1, h1) E1(, h 1) E1( y 1). Accordingy, he signa o noise raio deerioraes over ime. If he fundamena is a random wa, i.e., 1, he modeing shorcu of an increasing variance is correc. If he fundamena is ess persisen, i.e. 1, he signa o noise raio deerioraes even faser over ime. In he exreme case of a whie noise process, a signa wi no be informaive for any subsequen period any more. Excep for his exreme case, which we do no beieve o be reaisic when i comes o moneary poicy, our modeing shorcu eads o he waned resu ha agens pu ess weigh on he pubic signa over ime. 3 The char aso conains a second pubic signa (e.g., a macroeconomic announcemen, indicaed by he doed ines), which is reaivey more frequen, bu simiar in naure o he firs pubic signa. The mode wi be expanded wih his second pubic signa beow. Woring Paper Series No 1077 Augus 009 11
( ) E. (3) Taing he firs derivaive of (3) wih respec o yieds E( ) ( ) 3 0. (4) This impies ha heerogeneiy in expecaions varies wih he freshness of he pubic signa, as i is smaer, he more receny he pubic signa has been issued. Accordingy, heerogeneiy in expecaions shoud increase wih he ime ha has eapsed since he as updae of he pubic signa and drop in he ime period when he pubic signa ges updaed (and hus experiences a subsania increase in is precision). Le us furher assume ha whie he pubic signa y becomes sae, anoher pubic signa (such as a macroeconomic announcemen) z, h, is observed. In anaogy o he oher signas, is assumed o be normay disribued wih zero mean and variance, and independen of boh and. Lie in he case of he oher pubic signa, he precision of his signa aso erodes over ime, eading o he variance being monoonicay increasing wih he eapsed ime since he pubic signa z was issued. The precision of he signa is defined as 1/. The informaiona assumpions regarding his signa are aso iusraed in Figure 1. In his case, he decision rue of agen i is E y x z i i i (, h ), h. (5) Equaion (5) aows us o anayze wo furher ses of quesions. Firs, we can es he reacion of expecaions o a changing pubic signa z, and how his depends on he precision of he oher pubic signa. Second, we examine o wha exen heerogeneiy of expecaions evoves in response o he reease of he pubic signa z, and again how his depends on he precision of he oher pubic signa. Looing a he firs case, we find ha E i (, h z ) 0, (6) ( E i ( ) / z ), h ( ) 0. (7) Hence, expecaions wi reac o a reease of he pubic signa z (equaion 6), and his reacion wi depend on he precision of he firs pubic signa y (equaion 7): he more precise is he pubic signa, he ess pronounced wi he expecaions reacion be. Wih regard o he heerogeneiy in expecaions, we find ha 1 Woring Paper Series No 1077 Augus 009
( ) E, (8) E( ) ( ) 3 0, (9) ( E( ) / ) 6 4 ( ) 0. (10) As wih he firs pubic signa, heerogeneiy wi be smaer, he higher he precision of he pubic signa, and herefore drops whenever he signa ges updaed (equaion 9). A he same ime, however, his reaionship wi cruciay depend on he precision of he oher pubic signa: he more precise is y, he smaer he heerogeneiy reducion in response o an updae of he pubic signa z (equaion 10).. The daa and he economeric mode The syized mode deveoped in he preceding secion has ed o a number of hypoheses ha we woud ie o pu o an empirica es. The mode is formuaed in erms of agens expecaions abou he fuure pah of ineres raes and heir heerogeneiy. A direc es of he mode s hypoheses woud herefore require daa on expecaions, e.g. as derived from surveys. Unforunaey, here are no surveys avaiabe wih a sufficieny high frequency. Insead, we wi need o exrac he reevan informaion from financia mare daa. We wi mae wo assumpions in his regard. Firs, a change in expecaions abou he fuure pah of ineres raes shoud be refeced in financia mare prices, mos direcy in mare ineres raes; he more iquid he mare, he beer he price response wi refec he change in expecaions. Second, as Shaen (1993) and Harris and Raviv (1993) have shown, differences in expecaions rigger mare voaiiy. A changing heerogeneiy shoud herefore be mirrored in variaions of voaiiy. Aso in ha regard, mare ineres raes seem mos promising, given heir cose in wih moneary poicy. We wi herefore use Briish zero coupon governmen bond yieds for differen mauriies ranging from o 1 monhs. Our dependen variabe is defined as he firs difference of he daiy yieds, as is common pracice in he announcemen ieraure. These daa have been provided by he Ban of Engand. The choice of hese daa is based on he assumpion ha shor-erm ineres raes shoud be paricuary sensiive o news reaed o moneary poicy. A possiby preferabe aernaive woud have been o use inerban raes, as hese are arguaby more ighy ined o moneary poicy; unforunaey, he avaiabe series do ypicay no vary a he daiy frequency, especiay a he beginning of our sampe, such ha we had o use governmen bond yieds insead. The sampe spans a rading days from March 1997 uni December 008. Whie he sar of he sampe is owed o he unavaiabiiy of earier zero coupon yieds daa, i coincides roughy wih he independence of he Ban of Engand, and as such consiues a meaningfu choice. The saring poin is aso jus prior o he 1998 Ban of Engand Ac, which requires he MPC, i.e. he Ban of Engand s rae-seing body, o sign off he Ban s infaion repor. We use he infaion repor as our measure of he Ban of Engand s pubic signa, despie he fac ha here are a arge number of oher imporan communicaion evens. Moneary poicy decisions and he accompanying saemens or he subsequen reease of he minues of MPC meeings are obvious exampes, as are speeches by MPC members. We oped for he infaion Woring Paper Series No 1077 Augus 009 13
repor for a number of reasons: The reguar pubicaion schedue (compared o speeches), he reaivey ow pubicaion frequency (compared o minues and poicy saemens) and he indeph anaysis provided herein (aing he engh of he documens as a proxy, infaion repors are a around 50-60 pages more han 5 imes as ong as minues, which ypicay conain around 10 pages). Imporany, he infaion repor aso conains he forecass of infaion and GDP, and as such an imporan forward-ooing eemen of he Ban s economic assessmen, which direcy feeds ino poicy deiberaions. According o he Ban of Engand (009), he infaion repor serves wo purposes. Firs, is preparaion provides a comprehensive and forward-ooing framewor for discussion among MPC members as an aid o our decision maing. Second, is pubicaion aows us o share our hining and expain he reasons for our decisions o hose whom hey affec. According o Lomax (005), he infaion repor pays a cenra roe in he MPC s communicaion. As furher indicaions for is prominence, noe ha he infaion repors are he firs on he is of he Ban s Main pubicaions on is websie (www.banofengand.co.u), and ha i is he ony pubicaion ha is accompanied by a reguar press conference, where he Governor engages in a Q&A session wih journaiss. 4 We have coeced he reease daes of he Ban of Engand s infaion repor from he Ban of Engand websie, and creae a dummy variabe ha is equa o one on reease daes, and zero oherwise. The ime eapsed since he as repor is measured by couning he days since he preceding reease, and dividing his number by he oa number of days in beween he preceding and he subsequen reease. To faciiae he inerpreaion of coefficiens, we have normaized he resuing variabe by subracing is mean. The variabe does herefore range from -0.5 o +0.5. The as se of variabes reaes o he reease of macroeconomic daa. I is a we nown fac ha asse prices, a he ime of a reease, reac ony o he surprise componen conained herein (Kuner 001); as we wi show in his paper, his is no he case for mare voaiiy, hough. To consruc he surprise componen we foow he sandard in he announcemen ieraure, and deduc he expecaion of he announcemen from he acua announcemen vaue of he variabe. To aow a comparison of he magniude of he financia mare responses, we furhermore sandardize he surprises by heir own sandard deviaion. This sandardizaion removes differences in he uni of measuremen across variabes, and regression coefficiens for each series can hen be inerpreed as a response per one sandard deviaion surprise. We obained daa on financia mare expecaions of he various macroeconomic daa reeases from wo sources, namey Money Mare Services (MMS) and Boomberg Financia Services, and use he median response of he respecive pos as our measure of mare expecaions. 5 The macroeconomic announcemens conained in our daase reae o he Consumer Price Index (monh on monh), he Manufacuring Purchasing Managers Index, reais saes (monh on monh), unempoymen, and he rade baance. By oping for nomina and rea indicaors, we aim o cover a reaivey broad specrum of he economy; by choosing forward-ooing ones such as he PMI, we inend o capure indicaors ha migh hep mares o infer news abou he fuure course of moneary poicy. Given ha he bu of he announcemen ieraure has sudied mare responses o US reeases, we furhermore coeced US announcemen daa, again covering nomina and rea aspecs as we as eading indicaors, namey he Conference Board s composie index of eading indicaors, indusria producion, core CPI and non-farm payros. The US daa wi provide a basis for 4 The imporance of infaion repors for financia mares has aso been shown by Fracasso e a. (003). Based on indicaors of he quaiy of infaion repors for 19 counries, hey show ha higher quaiy repors are associaed wih smaer poicy surprises. Imporany, he Ban of Engand s infaion repors have ouperformed hose of a oher cenra bans in heir quaiy raning. 5 Ehrmann e a. (007) noe ha he informaion conen of he MMS and Boomberg series is very simiar. In paricuar, when boh surveys co-exis, he reease daa are idenica, and he expecaions agree amos perfecy. The quaiy of hese daa as measures of expecaions has been verified by, e.g., Baduzzi e a. (001) and Andersen e a. (003) for MMS, and by Ehrmann and Frazscher (005) for Boomberg. 14 Woring Paper Series No 1077 Augus 009
esing he robusness of our resus. Noe ha we can be agnosic on why US surprises move Briish yieds, given ha we are ony ineresed in how financia mares reac o pubic signas condiiona on he ime eapsed since he preceding infaion repor, regardess on he underying perceived ransmission mechanism from he US o he Briish economy. As menioned above, we are ineresed in how expecaions abou he fundamena, as we as heir heerogeneiy across agens evove. Our empirica counerpars o hese wo conceps are he price response and he voaiiy response, respecivey. I is imporan o noe, however, ha we are ineresed in he voaiiy response condiiona on he price response. The reease of a pubic signa can have wo effecs: firs, a change in he expecaion abou he fuure pah of ineres raes of he average mare paricipan. This wi be refeced in he price response. Second, aso he heerogeneiy of expecaions migh be affeced. Once he mare has seed o he new average expecaion, we woud expec o see a change in voaiiy compared o he ime prior o he reease of he signa. The immediae price response o he signa shoud herefore no be considered par of he voaiiy response. We do herefore need o esimae an economeric mode ha aows esing for he effec of news on boh he condiiona mean and he condiiona variance of asse prices. We esimae an exponenia GARCH (EGARCH) mode, foowing Neson (1991). 6 An EGARCH(1,1) mode is sufficien o address he nonnormaiy of he daa, in paricuar he seria correaion and heerosedasiciy of he daiy ineres rae series. We wi use differen varians of he economeric mode. In is mos exended version, he condiiona mean equaion is formuaed as r c s s g g r 1 1, (11) d wih r as he change in he daiy UK zero coupon raes, r -1 as he agged change, and d as a se of conros conaining day-of-he-wee effecs. s denoes he sandardized surprise componen conained in macroeconomic announcemen. Finay, g sands for he ime ha has eapsed since he preceding reease of he infaion repor. 7 Condiioned on he informaion se of as period ( I 1 ), we assume he disribuion of he disurbance o be I 1 ~ (0, h ). Hence, we express he condiiona variance of UK ineres rae changes, h, as og h c 1 1 h og 1 3 1 h h 1 1 IR a IR a a g g d. (1) 6 In order o es for asymmeries in voaiiy, we appy he Enge and Ng (1993) sign and size bias es. I examines wheher sign and size of he shocs impac differeny upon he condiiona variance. The join hypohesis of no asymmery is rejeced for a mauriies, suggesing he use of Neson s (1991) EGARCH. The resus are no dispayed bu are avaiabe from he auhors upon reques. 7 The mean equaion does no conain a variabe conroing for he reease of he infaion repor. In conras o he macroeconomic announcemens, we do no have a measure of mare expecaions avaiabe, such ha we canno consruc a variabe for he surprise componen conained in an infaion repor. Conros for he reease of an infaion repor do herefore ony ener he condiiona variance equaion, where a dummy variabe indicaing he reease dae is sufficien for our purposes. To faciiae he inerpreaion of coefficiens, we have normaized he variabe g o range from -0.5 o +0.5. This impies ha he variabe has a mean of zero, such ha he coefficiens can be inerpreed as he average effec of announcemen on ineres raes. Woring Paper Series No 1077 Augus 009 15
a and a IR are announcemen dummies ha ae he vaue one on a days a macroeconomic announcemen or an infaion repor is reeased, and zero oherwise. A oher variabes are as described in equaion (11). The mode is esimaed via maximum ieihood, using he BHHH agorihm for opimizaion. Noe ha he mode is esimaed for a business days in he sampe, i.e. aso for days when neiher an infaion repor is issued, nor a macroeconomic announcemen is made. The corresponding variabes are equa o zero on such days. Taing he infaion repor as he pubic signa y, and he reease of a given macroeconomic daa as he second pubic signa z, he syized mode eads o he foowing hypoheses: 1) The pubicaion of an infaion repor shoud ead o a reducion in he heerogeneiy of beiefs, and hus o a reducion in condiiona voaiiy in financia mares (firs IR derivaive of equaion (4)); H0 : 0. ) Aso he reease of macroeconomic daa shoud ower condiiona voaiiy (equaion 9)); H0 : 0. 8 3) Wih passage of ime, he informaion conained in a given infaion repor becomes sae, which shoud ead o an increase in condiiona voaiiy uni he subsequen infaion repor is reeased (firs derivaive of equaion (4)); H 0 : 0. 4) The voaiiy reducion in response o macroeconomic reeases shoud be more pronounced, he more ime has eapsed since he preceding reease of an infaion repor (equaion (10)); H0 : 0. 5) The updae of beiefs, and hus he reacion of asse prices o macroeconomic news in he mean equaion shoud be sronger, he more ime has eapsed since he preceding reease of an infaion repor (equaion (7)); H : sign ( ) sign ( ). Hypoheses 1 o 4 are iusraed in Figure. The soid ine shows he evouion of he heerogeneiy in expecaions (proxied by mare voaiiy) ha is prediced by he mode of Secion.1. Imporany, he figure aso conains he doed ine, which shows ha he esimae of he voaiiy increase over ime depends cruciay on he incusion of macro announcemens in he regression mode: wihou hese, he sope of he ine ges underesimaed. This is imporan o bear in mind when inerpreing our resus: whie our mode conros for a few macroeconomic announcemens, i wi never be abe o incude a reevan news, such ha we woud expec o find a reaivey subdued esimae for. FIGURE HERE 3. Financia mare reacions o he reease of pubic signas This secion presens he empirica resus. We wi firs as how financia mares respond o he reease of pubic signas in genera, ooing a he response o infaion repors as we as o macro reeases. Subsequeny, we wi es wha happens over ime foowing he reease of an infaion repor, wih he informaion becoming sae. 0 8 In conras o his hypohesis, a number of sudies find posiive reacion coefficiens. This resu is pausibe if he surprise componen conained in an announcemen dominaes he incidence of he announcemen, and is endency o homogenize agens informaion ses. This issue is ouched upon in Kim and Sheen (000), Kim e a. (004) and DeGennaro and Shrieves (1997). Which effec dominaes remains an empirica quesion. In our anaysis, we wi herefore disi he pure effec arising from he incidence of he pubic signa on he harmonizaion of beiefs by conroing for he effec of he surprise componen in he signa. 16 Woring Paper Series No 1077 Augus 009
3.1 The uncondiiona recepion of pubic signas Our firs es reaes o he reacion of yieds o he reease of pubic signas, i.e. hypoheses 1 and. They sugges ha pubic signas provide an opporuniy for financia mare paricipans o updae heir beiefs, and as such shoud ead o more homogeneous views abou he fuure evouion of ineres raes. Mare voaiiy shoud herefore be reduced in response o pubic signas. 9 Tabe 1 shows he resus of a mode ha conains, in he variance equaion, dummies ha are equa o one on reease daes for infaion repors and he UK macro announcemens. 10 The mode furhermore conros for he effec of macroeconomic reeases on he mean equaion (as expained above, given he unavaiabiiy of a measure conaining mare expecaions abou he conen of infaion repors, we canno conro for he reease of infaion repors in he mean equaion). In he nex secion, we wi focus more on he effecs in he mean equaion. For he ime being i sha suffice o noe ha a number of macroeconomic reeases ead o significan responses. The direcion of hese effecs is as one woud expec, wih ineres raes rising in response o higher han expeced infaion and rea deveopmens, and wih ineres raes faing in response o higher han expeced unempoymen (ahough his effec is no saisicay significan). TABLE 1 HERE The imporan hing o noe in Tabe 1 is ha condiiona voaiiy is reduced in response o neary a announcemens. The imporance of he infaion repor o mare paricipans is confirmed by he fac ha he voaiiy-reducing effec in response o is reease is one of he sronges across he differen announcemens conained in his mode. Anoher ineresing fac is he consisency of hese resus across mauriies. A a mauriies, he announcemens reduce voaiiy, wih he effec becoming smaer, he onger he horizon. The singe excepion in he abe reaes o he UK rade baance. Is reease seems o heighen raher han dampen mare uncerainy. To undersand he possibe reasons for his finding, i is usefu o concepuay separae wo simuaneous evens ha are conained in he reease of macro news. Firs, he news reease ypicay conains a surprise componen, which he mares need o price. The incusion of he surprise componen in he mean equaion aemps o conro for his effec. However, if he surprise componen is paricuary arge, or creaes confusion for insance because i is a odds wih he informaion conained in oher recen reeases, uncerainy in mares migh we increase, couneracing he second facor ha is a pay, namey he homogenizing of agens informaion ses, and hus he co-ordinaion of beiefs. To separae hese wo effecs, we have expanded he curren regression mode by furhermore conroing for he absoue surprise in he condiiona voaiiy equaion. Tabe repors he corresponding resus, and very convincingy suppors he conjecure. Firs, voaiiy 9 The reducion in condiiona voaiiy shoud happen once iniia disagreemen abou he inerpreaion of he news has subsided. In ha conex, i is imporan o remember ha our anaysis is based on daiy daa, hereby no capuring shor-erm phenomena as idenified, e.g., in Feming and Remoona (1999a). 10 For breviy, he (E)GARCH coefficiens are no dispayed in Tabe 1 and he foowing abes. Throughou a esimaions, he GARCH erm is srongy significan and cose o bu beow one, indicaing a high persisence of he condiiona voaiiy. Boh EGARCH parameers are equivaeny significan in a esimaions. The sign asymmery parameer exhibis a negaive sign, ahough sma in size. The esimaes confirm he asymmery assumpion proposed by he Enge and Ng (1993) es resus and can be read as a negaive innovaion increasing voaiiy more han a posiive innovaion of equa magniude. The resus are avaiabe from he auhors upon reques. Woring Paper Series No 1077 Augus 009 17
increases in response o he surprise componen of viruay a announcemens. Second, he resus wih regard o he announcemen dummies (now ceany measuring he effec of he incidence of a reease) improve in severa ways: he coefficien esimae wih regard o he rade baance dummy swiches sign, and urns negaive in ine wih a oher coefficiens; a he - and 3-monh mauriies, he previousy insignifican coefficien on he CPI dummy is now aso significan and negaive; a coefficiens are subsaniay arger han heir correspondens in Tabe 1. TABLE HERE Based on his evidence, we concude ha pubic signas do effecivey co-ordinae he beiefs of mare paricipans, hus owering condiiona voaiiy. This effec can possiby be overshadowed by he voaiiy-enhancing effec of he announcemen s surprise componen. However, in he majoriy of cases, he voaiiy-reducing facor urns ou o be dominaing. Taen ogeher, hese resus provide srong evidence in favor of hypoheses 1 and. 3. Wha if cenra ban communicaion becomes sae? To es hypoheses 3 o 5, we wi now urn o measuring he mare reacion condiiona on he freshness of cenra ban communicaion. For ha purpose, we expand he economeric mode by incuding he variabe ha measures how much ime has eapsed since he preceding reease of an infaion repor, and by ineracing his variabe wih he macroeconomic reease daa. The resus are provided in Tabe 3. TABLE 3 HERE Looing a he direc effec of he eapsed ime, i is apparen ha voaiiy does indeed increase wih he disance from he preceding infaion repor reease. This effec is sronges a he -monh mauriy, bu aso presen and saisicay significan up o a horizon of 6 monhs. Despie is saisica significance, i has o be noed ha he magniude of he effec is sma, however. A he same ime, Figure has shown ha he omission of macroeconomic announcemens wi end o depress esimaes of. Ineresingy, his conjecure is confirmed by he incusion of 4 US announcemens in a subsequen robusness es, which ends o doube or even ripe he esimaes of. The voaiiy-increasing effec disappears a he one-year horizon, and consiseny so in a furher modes ha we wi anayze. A he same ime, as shoud be expeced, he eapsed ime does no maer for he mean of yieds, as shown by he very sma and saisicay highy insignifican coefficiens for his variabe in he mean equaion. This resu is in cear suppor of hypohesis 3. Wha abou he voaiiy reducion in response o macroeconomic reeases? Hypohesis 4 saed ha his effec shoud be more pronounced, he more ime has eapsed since he preceding reease of an infaion repor. As a maer of fac, wih very few excepions, he ineracion erms are ypicay negaive, supporing he hypohesis. Ahough a number of hese are no esimaed o be significan, he picure ha emerges is fairy consisen: wih one excepion, a significan coefficiens are negaive. Furhermore, he effecs are sizabe. To ae he exampe of he CPI reease: is effec on average is esimaed o be -0.7 for he - monh mauriy. However, he firs CPI reease afer a given infaion repor has viruay no effec on voaiiy, wih is coefficien esimaed a 0. 005, whie he hird CPI reease has a coefficien esimae of 0. 608. 11 As before, he effecs are paricuary pronounced a shorer mauriies. 11 These numbers are cacuaed as foows: he average vaue of he ime eapsed variabe g is cacuaed for he firs and he hird CPI reeases. These vaues amoun o -0.60 and 0.319, 18 Woring Paper Series No 1077 Augus 009
The resus srenghen by imposing he spi of announcemen effecs ino he announcemen incidence and he absoue surprise componen, as shown in Tabe 4. In his specificaion, a saisicay significan coefficiens on he non-ineraced dummies are negaive, as are a he significan coefficiens on he ineracion erms. A he same ime, a significan coefficiens on he absoue surprise as we as on he respecive ineracion erms are posiive. Conroing for he surprise componen, in he exampe of CPI he differences beween he firs and he as announcemen are even more sriing, wih an effec of 0. 39 for he firs and of 0.916 for he as, suggesing ha he voaiiy reducion of he as announcemen is neary four imes as arge as for he firs announcemen. The evidence is herefore ceary in suppor of hypohesis 4. TABLE 4 HERE The as hypohesis o be esed, number 5, saed ha he updae of beiefs, and hus he reacion of asse prices o macroeconomic news shoud be sronger, he more ime has eapsed since he preceding reease of an infaion repor. To see wheher his is he case, we need o chec he coefficiens in he mean equaion. As menioned above, he non-ineraced coefficiens in he mean equaion show he sign ha shoud be expeced according o economic reasoning. The quesion o be seed is wheher he coefficiens on he ineraced variabes show he same sign, and are saisicay significan. Whie i is generay rue ha he coefficiens of he ineraced variabes have he same sign as hose of he non-ineraced variabes, heir saisica significance (agains zero) is raher wea. Ony in he case of he CPI announcemen do we find such a paern consiseny across he differen mauriies. One ineresing observaion in ha conex reaes o he evouion of coefficiens across he mauriy specrum. Conrary o wha was observed for he voaiiy equaion, effecs are now arger wih increasing mauriies. This finding is in ine wih Feming and Remoona (1999b), who had aso found an increasing magniude over he mauriies sudied in he curren paper. A he 1-year horizon, he effecs are reaivey sar, hough: he response of 1-year yieds o a one sandard deviaion surprise in he firs CPI reease foowing an infaion repor is esimaed a 1.9 basis poins, whereas he response o he as CPI reease is subsaniay arger a 3.4 basis poins. TABLE 5 HERE We have esed he robusness of hese resus in a number of ways. Firs, Tabe 5 repors he esimaed coefficiens in a mode incuding US macro reeases. The overa picure is confirmed: some, abei ie evidence ha he price response o news becomes arger wih furher disance from he previous infaion repor; srong evidence ha he announcemens reduce condiiona voaiiy and ha, as cenra ban communicaion becomes sae, his effec becomes more pronounced, whie mare voaiiy as such increases. 1 Imporany, he voaiiy increase wih he ime eapsed since he as infaion repored is esimaed o be wo o hree imes as arge as in Tabe 3. This ceary shows ha conroing for he voaiiy reducion of addiiona news has subsania effecs on his parameer esimae. Second, we esed for robusness o he incusion of oher Ban of Engand communicaion evens, which provide an occasion for financia mare paricipans o updae heir respecivey. The average effec of he firs CPI reease on voaiiy is hen given by CPI CPI ( 0.60) 0.70 0.60*1.059 0.005 ; he average effec for he hird CPI reease is CPI CPI 0.319 0.70 0.319 *1.059 0.608. 1 We refrain from aso incuding he absoue surprise componen and is ineracion wih he eapsed ime in his mode, as esimaing an EGARCH mode of his size woud no be feasibe. Woring Paper Series No 1077 Augus 009 19