WORKING PAPER NO. 192 IS THE EUROPEAN CENTRAL BANK (AND THE UNITED STATES FEDERAL RESERVE) PREDICTABLE? BY GABRIEL PEREZ-QUIROS AND JORGE SICILIA

Transcription

1 EUROPEAN CENTRAL BANK WORKING PAPER SERIES E C B E Z B E K T B C E E K P WORKING PAPER NO. 192 IS THE EUROPEAN CENTRAL BANK (AND THE UNITED STATES FEDERAL RESERVE) PREDICTABLE? BY GABRIEL PEREZ-QUIROS AND JORGE SICILIA November 2002

2 EUROPEAN CENTRAL BANK WORKING PAPER SERIES WORKING PAPER NO. 192 IS THE EUROPEAN CENTRAL BANK (AND THE UNITED STATES FEDERAL RESERVE) PREDICTABLE? BY GABRIEL PEREZ-QUIROS 1 AND JORGE SICILIA 2 November 2002 The authors would lke to thank G. de Bondt, G. Camba-Méndez, F. Drud, J.L. Escrvá, H.-J. Klöckers, an anonymous referee and partcpants at an nternal semnar for specfc comments, to S. Eusep for assstance and useful dscussons on an earler verson of the paper, and to R. Plegaard for assstance n the data collecton. Edtoral suggestons from C. Burns are gratefully acknowledged. All remanng errors are our own responsblty.the vews expressed heren are those of the authors and do not necessarly represent those of the European Central Bank or of the Eurosystem. 1 Banco de España. 2 European Central Bank.

3 European Central Bank, 2001 Address Kaserstrasse 29 D Frankfurt am Man Germany Postal address Postfach D Frankfurt am Man Germany Telephone Internet Fax Telex ecb d All rghts reserved. The vews expressed n ths paper do not necessarly reflect those of the European Central Bank. Reproducton for educatonal and non-commercal purposes s permtted provded that the source s acknowledged. ISSN

4 Contents Abstract 4 Executve Summary 5 1. Introducton 7 2. Heurstc approach to measure the predctablty of the monetary polcy decsons Monetary polcy shocks, surprses and monetary polcy decsons of the ECB What do we mean by monetary polcy shocks? Monetary polcy shocks n the euro area: whch rates could we use? Monetary polcy shocks n the euro area: applyng prncpal components (PC) An analyss of the monetary polcy shocks and the monetary polcy decsons of the ECB (and the US Federal Reserve) Has the daly pattern of the varance of these shocks changed wth the announcements of monetary polcy? Impact of the shocks on the term structure of nterest rates Monetary polcy shocks and the yeld curve Monetary polcy shocks and the mplct nterest rates at long horzons Conclusons 28 Bblography 29 Annex 1. Descrpton of the data 33 Annex 2. Prncpal components 35 Annex 3: estmaton of a Probt 37 Tables 38 Fgures 49 European Central Bank Workng Paper Seres 57 ECB Workng Paper No 192 November

5 Abstract The objectve of ths paper s to examne the predctablty of the monetary polcy decsons of the Governng Councl of the ECB and the transmsson of the unexpected component of the monetary polcy decsons to the yeld curve. We fnd, usng new methodologes, that markets do not fully predct the ECB decsons but the lack of perfect predctablty s comparable wth the results found for the Unted States Federal Reserve. We also fnd that the mpact of monetary polcy shocks on bond yelds declnes wth the maturty of the bonds, and that ths mpact s sgnfcantly lower when the shock stems from a monetary polcy meetng of the ECB. Usng mplct rates nstead of bond yelds, we fnd evdence that the market vews the ECB as credble. Keywords: Predctablty, monetary polcy shocks, prncpal components, transmsson of monetary polcy. JEL classfcaton: C22, E52 4 ECB Workng Paper No 192 November 2002

6 Executve Summary The objectve of ths paper s to examne the predctablty of the monetary polcy decsons of the Governng Councl of the ECB and the transmsson of the unexpected component of ts monetary polcy decsons to the yeld curve. Wth respect to the frst goal, the predctablty analyss, we apply a battery of tests and we conclude that the markets have predcted the monetary polcy decsons of the ECB rather well. However, the results do not accept the hypothess of perfect predctablty. To evaluate the magntude of the devatons from ths hypothess, applyng the same battery of tests, we draw a comparson of these results and those obtaned on the predctablty of the monetary polcy decsons of the Unted States Federal Reserve durng the same perod. We provde evdence that the predctablty of both central banks s broadly smlar. Wth respect to the second objectve, we analyse the mpact of the unexpected component of the monetary polcy decsons on the term structure of nterest rates n the euro area. We use seres of daly monetary polcy shocks n the euro area n whch the observatons on the days of the monetary polcy meetngs of the ECB are the unexpected component of the monetary polcy decsons. Ths allows us to dentfy the mpact of the surprse part of a monetary polcy decson on the yeld curve and compare t to the normal response of the yeld curve to other daly shocks. We show that the mpact of the daly monetary polcy shocks on bond yelds declnes wth the maturty of the bonds, and that ths mpact s sgnfcantly lower when the shock stems from a monetary polcy meetng of the ECB. Usng mplct rates nstead of bond yelds, we fnd evdence that the market vews the ECB as credble. In addton to the former contrbutons, the paper presents a new methodology to approach the problem of measurng monetary polcy shocks and predctablty of central bank decsons. The contrbutons can be summarse as follows: Frst, as a dfference to other standard papers n the lterature, we use daly data and consder all days, not only meetng days or T days before the meetngs. Our purpose wth ths approach s twofold. Frst, to have daly seres of monetary polcy shocks whch can be nterpreted as how market partcpants change the expected path of monetary polcy nterest rates on a daly bass (at dfferent horzons) as new nformaton becomes avalable. Second and takng advantage of ths seres, to test for the sgnfcance of the shocks assocated wth the monetary polcy meetngs compared to the shocks produced on any other day. Second, we gather nformaton about the shocks from dfferent money market nterest rates, avodng the lqudty (and potentally other) consderaton(s) unrelated to monetary polcy expectatons that affect the ndvdual seres. We comprse the nformaton of the dfferent ECB Workng Paper No 192 November

7 rates by usng prncpal components. Ths approach allows us to get a rch varety of conclusons on how the new daly nformaton affects the expected path of monetary polcy rates at dfferent horzons. For example, we show that the mpact of monetary polcy decsons (ether to change the key ECB nterest rates or to mantan them unchanged) can be consdered surprses when we use very short-term rates but not so when usng longer-term rates. We see ths as evdence showng that the surprses on monetary polcy decsons mght be more related to the tmng of the decsons than to the decson tself. Thrd, we measure the predctablty of the monetary polcy decsons of a central bank from dfferent ponts of vew by usng dfferent technques n order to check the robustness of our fndngs. These technques go from a graphcal ntuton to an EGARCH specfcaton for the prncpal components of the seres, gong through an heurstc approach based on a weghted average of the possble outcomes, an analyss of the probabltes of change based on a probt specfcaton and lnear regressons for the transmsson mechansm. Fnally, to our knowledge the paper presents the most comprehensve approach to compare the euro area and the US n terms of the amount of nformaton used, a prelmnary analyss of the seres n order to take nto account the dfferences due to maturty, lqudty, etc., the varety of technques used and the robustness of the results. 6 ECB Workng Paper No 192 November 2002

8 1. Introducton Not so long ago central banks gave lttle weght to beng transparent; provdng tmely, open and clear nformaton on ther mandate, strategy, assessment and decsons to the publc. 1 Ths has changed sgnfcantly n the recent past for good reasons and today transparency s vewed as a very mportant component of the monetary polcy framework of a central bank. One of these reasons s related to the noton of credblty. Credblty s ultmately drven by the ablty and track record of the central bank n fulfllng ts mandate, and can be defned as the belef on the sde of the publc that prce stablty wll be mantaned over the medum term. Transparency facltates the understandng of what the central bank does and by dong so, t helps central banks to foster ther credblty. Another mportant reason stems from the fndng that that forward-lookng economc agents have relevant methodologcal consequences for the monetary transmsson mechansm (see McCallum, 1999, 2001). If the market 2 fully understands the role of a central bank, the belef n the commtment to mantanng prce stablty over the medum term should anchor nflaton expectatons and nduce a rule lke behavour on the part of market partcpants. Ths would lead the market to react to the new nformaton changng ther expected path of monetary polcy rates n a way consstent wth the monetary polcy strategy of the central bank. By beng transparent, expectatons on the path of future monetary polcy decsons are formed more effcently and accurately. The polcy makers understand ths and have stressed ther commtment to stand up to the challenge. For example, n the words of a monetary polcy maker n the euro area, when the markets correctly antcpate that a new pece of nformaton wll lead to a change n offcal nterest rates they wll do much of the work themselves through a change n the term structure, Issng (1999). Has ths been the case? Ideally, t could be consdered that the relevant queston to be answered s to what extent the market expectaton on the future path of monetary polcy rates s broadly n lne wth the vew of the central bank at every pont n tme. Ths s however hard to test. What can be analysed nstead s to what extent a central bank has been predctable; whether market partcpants have antcpated ts monetary polcy decsons. By 1 2 There are many defntons of transparency n the lterature. In Kng et al (1998) t s defned t as a process by whch nformaton about exstng condtons, decsons, and actons s made accessble, vsble, and understandable. Ths defnton s broadly n lne wth Wnkler (2000), where transparency s ( broadly and loosely ) defned as the degree of genune understandng of the monetary polcy process and polcy decsons by the publc. Several authors (Ejffnger and Geraats (2002), Gerbach and Hahn (2002)) have useful dscussons about the dfferent aspects of transparency. Whle the dstncton between market partcpants and the publc at large s relevant for the communcaton of a central bank, gven the emprcal nature of the paper, we wll concentrate on market partcpants. ECB Workng Paper No 192 November

9 becomng more predctable, a central bank gans the ablty to nfluence nterest rates before the announcement of ts monetary polcy decsons. Predctablty s sometmes vewed as a necessary consequence of transparency. In ths ven, the degree of predctablty of a central bank s thus sometmes seen as a way of measurng whether t s transparent. 3 For example, Poole and Rasche (2001) argue that wth complete transparency, the monetary polcy decsons of a central bank should be fully predctable. In fact, they test the predctablty of the Unted States Fed by checkng to what extent monetary polcy decsons affect market rates, as ther vew s that polcy announcements should not provde nformaton to market partcpants, and thereby should not trgger any reacton of asset prces. It s clear that a hgher degree of transparency should be connected to a hgher degree of predctablty. However, t can also be argued that perfect predctablty mght not be fully attanable n a world of uncertanty. The decson makng process of monetary polcy s a complex one n whch all relevant peces of nformaton have to be assessed n the lght of ther mplcatons for the monetary polcy mandate. Gven that the outcome of the process of mappng all the nformaton on the state and the functonng of the economy (whch s nherently uncertan) to take monetary polcy decsons s based on judgement and s not done mechancally, t could be argued that a certan lack of predctablty mght not necessarly be related to a lack of transparency. Some authors also argue that when the decson s a collectve one, as n the case of the European Central Bank (ECB), full transparency (n fact, operatonal transparency) may not be reached. 4 In ths same ven, the precse tmng of monetary polcy decsons may be hard to antcpate perfectly, especally f monetary polcy meetngs are held very frequently, as was the case for the Governng Councl of the ECB before November Whlst n a world of uncertanty polcy actons wll most lkely never be fully predctable, from the pont of vew of central bank t s mportant to avod beng unpredctable (or perhaps more mportantly, to avod that market uncertanty ncreases because of an ncorrect nterpretaton of ts own behavour). Ths calls for the need for a contnuous effort to be transparent, communcate effectvely and provde actve gudance to the markets explanng Other consderatons are mportant determnants of predctablty, such as gradualsm n nterest rate decsons (Lange, Sack and Wcksell (2001)). See Cukerman (2000). In addton, Wnkler (2001) holds the vew that as the monetary polcy n the euro area s a relatvely new event the level of common language and understandng between the central bank and market partcpants stll needs to be fully tuned. Untl 8 November 2001, the Governng Councl of the ECB held monetary polcy dscussons at all of ts meetngs, generally every two weeks. Snce then, t has dscussed monetary polcy ssues only once a month. 8 ECB Workng Paper No 192 November 2002

10 ts polcy decsons. 6 In fact, central banks care about predctablty. Ths paper analyses to what extent the markets have antcpated the monetary polcy decsons of the ECB. There s not one sngle approach to measure predctablty n the emprcal lterature. A great deal of work has been done to measure the predctablty of monetary polcy decsons n the Unted States and some European countres pror to the Monetary Unon. 7 However, the predctablty of the monetary polcy decsons of the ECB has not been tested extensvely, partly due to the relatvely short perod of tme n whch the ECB has been conductng the sngle monetary polcy n the euro area. To our knowledge, two papers, Gaspar, Perez-Quros and Scla (2001), Hartman, Manna and Manzanares (2001) have analysed t and found evdence ndcatng that fnancal markets have generally understood and predcted the monetary polcy decsons of the ECB. 8 Interpretng the results s not easy. Whle perfect predctablty s the clearest benchmark that comes to our mnd, gven the above arguments t mght not be too realstc. For ths reason, we also provde some evdence on the predctablty of the Unted States Federal Reserve (Fed), whch allows for a rouge comparson between the degree of predctablty of the two central banks. As the lterature has typcally found that predctablty s an evolvng process, and that the market has mproved ts ablty to predct the monetary polcy decsons over tme, 9 perhaps not enough tme has passed yet for the ECB. We also analyse the transmsson of the unexpected component of the monetary decsons of the ECB to the term structure of nterest rates. The reacton of the yeld curve to the unexpected component of the monetary polcy decsons at the Federal Open Market Commttee (FOMC) has been used n the lterature (Roley and Sellon (1998), Poole and Rasche (2001), Kuttner (2001), Cochrane and Pazzes (2002)) to analyse the predctablty of the Unted States Fed. Besdes applyng ths analyss to the monetary polcy decsons of the ECB, takng advantage of the seres of daly monetary polcy shocks estmated to assess Not surprsng the markets cannot be an objectve tself of monetary polcy, followng what market partcpants expect, regardless of the vew the central bank holds on ts assessment of the lkelhood of reachng ts objectve. As Blnder puts t: markets tend to overreact, are susceptble to fads and speculatve bubbles, and seem to be have more short-term horzons than central bankers. Whle central banks should not have any nterest n surprsng the markets, t mght be unavodable on some occasons. For example, for the Fed, among others, Roley and Sellon (1998), Poole and Rasche (2001), Kuttner (2001), Poole, Rasche and Thornton (2002), Cochrane and Pazzes (2002); For the Bank of England, Haldane and Read (1999); for a seres of European countres pror to the Monetary Unon and the Unted States, see Favero et al (1998) and Buttglone et al (1998). Ross (2002) extends the analyss of Gaspar, Perez Quros and Scla (2001) for the ECB and compares the predctablty of the ECB wth the one of the Bank of England and the Federal Reserve. Bernhardsen and Kloster (2002) also compare the predctablty of several central banks usng changes n the three-month nterest rates. For the Unted States (see references n footnote 9) a common fndng s that the predctablty of Fed s actons ncreased after the decson to announce changes n Fed polcy rates mmedately after FOMC meetngs. In turn Haldane and Read (1999) show that the ntroducton of nflaton targetng n the Bank of England mproved the predctablty of ts monetary polcy decsons. ECB Workng Paper No 192 November

11 predctablty, our contrbuton s to study how the unexpected component of the monetary polcy decsons has affected the term structure of nterest rates compared to the normal mpact of shocks on other days wth no monetary polcy decsons. The paper s structured as follows: In secton 2, we present a smple heurstc approach to assess how well market partcpants have predcted the monetary polcy decsons of the ECB before the meetng of the Governng Councl. In secton 3 we defne seres of daly monetary polcy shocks n the euro area applyng prncpal components to an array of daly money market data. We consder ths approach a good way of summarsng all the nformaton contaned n the money market and we present t n a way n whch the predctablty can be analysed. These seres wll be of partcular mportance, as they wll allow us to measure to what extent monetary polcy decsons have moved short-term money market rates (.e. how have they surprsed the markets), as compared to the normal behavour of these rates. Secton 4 analyses, usng an EGARCH, how the monetary polcy meetngs of the Governng Councl have changed the volatlty pattern of these monetary polcy shocks. Throughout these sectons, to fnd a benchmark wth whch to compare the predctablty results for the ECB, we apply (the same battery of) measures of predctablty to the Fed. In Secton 5 we analyse the reacton of the term structure of the euro area to the daly shocks and to the unexpected component of the monetary polcy decsons of the ECB (the shocks on the days of the monetary polcy meetngs of the ECB). Secton 6 sums up and concludes. 2. Heurstc approach to measure the predctablty of the monetary polcy decsons A rather ntutve approach s to analyse to what extent market partcpants have predcted the monetary polcy decsons taken shortly before the meetng. Gaspar, Quros and Scla (2001) used the EONIA 12 to calculate the probablty attached to a change n the key ECB nterest rates before the meetngs of the Governng Councl. However, the hgh volatlty of the EONIA and the mpact of lqudty consderatons n ts pattern of behavour, lke when underbddng epsodes occur (Bndsel 2002), argue n favour of usng other short-term nterest rates to assess market expectatons. The very short end of the money market curve, and n partcular the EONIA swap rates, are good canddates. The money market data used n the remander of ths secton for the euro area s the onemonth and the two-week EONIA swap rate from 1 January 1999 to 7 June Followng Gaspar, Quros and Scla (2001), we consder that the short-term market rate can be seen as 12 The EONIA s an overnght ndex average rate (see Annex 1). 10 ECB Workng Paper No 192 November 2002

12 a lnear combnaton (β, 1-β) of two events, a decson not to change nterest rates from ther prevalng level ( 0 ) or to change them by 25 bass ponts ( 25 ). t = β (1) 25 + ( 1 β ) 0 β can thus be nterpreted as the probablty of at least a 25 bass pont change (postve when the expectaton s of an ncrease and negatve otherwse), aganst the alternatve of not changng the key rate. 13 At these maturtes there seems to be no need to control for the rsk prema, as t s estmated to be zero. 14 However, to take account of the natural spread between the market rate and the MRO rate (whch s a collateralsed rate wth lower credt rsk than the nterbank market rate), we apply a spread of 5 bass ponts (bp) between the market rate and the MRO rates. 15 We mpose a (rather arbtrary) benchmark for ß to assess the extent to whch the market has predcted the monetary polcy decsons taken by the ECB. We assume that f ß s above 12.5 bp n absolute value, whch corresponds to a probablty of 50% attached to a change of 25 bp n the key rates, the market expected the ECB to change ts key nterest rates. We calculate ß for each meetng of the Governng Councl usn g the two-week and onemonth EONIA swap money market rates one day before the meetng. We then evaluate the percentage of tmes n whch fnancal markets have antcpated the monetary polcy decsons of the ECB. Smlar to the graphc analyss n Robertson and Thornton (1997) and Ross (2002), Fgures 1 and 2 show the results for all the meetngs of the Governng Councl. [Insert Fgures 1 and 2 about here] The monetary polcy decsons of the ECB have been accurately predcted 87% (94%) of the tmes when the one-month rate (two-week rate) s used to assess the expectatons of market partcpants. The two-week rate s better than the one-month rate for assessng the predctablty of the monetary polcy decsons n the euro area before November 2001, when the ECB dscussed monetary polcy decsons bmonthly. Gven that t then swtched to monetary polcy dscussons once a month, t s probably more accurate to use snce then the one-month rate. In any case, the results snce November 2001 are smlar usng both rates. The decsons are analysed n more detal n Table 1. Usng the two-week rate, the market has antcpated wth a smlar probablty the decsons to change nterest rates (92%) and to The ECB consders as key ECB nterest rates the MRO rate (the fxed rate under fxed rate tenders and the mnmum bd rate under varable rate tenders) and both the margnal and lendng faclty rates. For the sake of clarty, n the remander of the paper we use MRO rate or key rate nterchangeably. It cannot be rejected that the rsk prema s sgnfcantly dfferent from zero n the short -term nterest rates n the EONIA swap market. See Durre, Evjen and Plegaard (2002) for a thorough analyss on estmates for the rsk prema across the maturty spectrum for the euro area EONIA swaps. ECB Workng Paper No 192 November

13 mantan them unchanged (94%). On the slghtly more negatve sde, the relablty of changes, defned as the percentage of tmes n whch the model sgnals a rate change and t actually happens, has been 80%. Gven the frequent meetngs of the Governng Councl of the ECB before November 2001, the markets may have found some dffcultes antcpatng the decson on a partcular day. Fgure 1 shows how the majorty of occasons n whch a monetary polcy decson was expected and dd not occur are mostly concentrated on the meetngs shortly before the ones n whch the actual change was mplemented. Whle t may be consdered that the decson to swtch to monthly dscussons of monetary polcy may have affected for the better the predctablty of the monetary polcy decsons of the ECB, t s too soon to tell. [Insert Table1 about here] The results fall short of the "perfect predctablty" benchmark. As already noted, ths may however be too an extreme benchmark by whch to judge a central bank. To see to what extent ths result s comparable wth other smlar central banks we apply the same analyss to the monetary polcy decsons n the Unted States, usng the one-month Lbor dollar rate n a sample spannng from 4 January 1999 to 6 June Fgure 3 (and also Table 1) presents the results for the Fed. As can be seen, the smlartes are large. The percentage of tmes n whch the decsons were antcpated was 90%. Whle the number of changes antcpated s lower than for the ECB (81%), the Fed changed rates on a larger number of occasons than the ECB. The percentage of hts for the cuts (82%) and ncreases (100%) n nterest rates mplemented are also smlar. The man dfference s that, n the sample, markets have never antcpated a change that the Fed faled to delver and thereby the hgh score n the relablty of changes (100%). Ths could be due to the fewer meetngs held by the FOMC n the sample, or perhaps to the fact that markets may have had better gudance, e.g. through speeches. Moreover, there are many more announcements of changes than tmes when the FOMC decded to keep the Fed Fund rate unchanged. As Fgure 3 shows, Alternatve estmatons applyng a natural spread of 3 and 7 bass pont yeld smlar results. Whle the results cannot be completely comparable as the operatonal framework n whch the two central banks operate are dfferent, the use of the one-month rate to measure the predctablty of the monetary polcy decsons of the Fed mnmse the lack of comparablty, as the FOMC hold scheduled meetngs approxmately every sx weeks. Yet, some mportant caveats need to be consdered. The FOMC met on fewer occasons than the Governng Councl of the ECB n that perod, so the market had fewer opportuntes to bet on the outcome of a meetng. In addton, three monetary polcy decsons n the sample were taken at scheduled meetngs (3 January, 18 Aprl, and 17 September 2001), for only one for the ECB. Whle the model could have been appled to a longer sample for the US, we would rather not draw comparsons from dfferent samples. An estmaton or t,t+1 = α + β*e t-1 ( t,t+1 ) + ε t, where t,t+1 s the one-month dollar Lbor rate at tme t and E t-1 ( t,t+1 ) s the expected one month rate for at tme t calculated at t-1, whch are contegrated varables, yelded a rsk prema of 13 bass ponts wth a standard devaton of 4.4 bass ponts. Dfferng from the calculatons carred for the euro area, the rsk prema s sgnfcantly dfferent form zero. 12 ECB Workng Paper No 192 November 2002

14 on two of the three occasons n whch the markets faled to antcpate a move from the Fed n the sample, nterest rates were changed at unscheduled meetngs. [Insert Fgure 3 about here] To sum up, usng a very smple approach to assess the predctablty one day before the monetary polcy meetngs, we fnd that the monetary polcy decsons of the Governng Councl of the ECB have been very predctable. These results are broadly comparable to the ones obtaned for the Unted States Federal Reserve. 3. Monetary polcy shocks, surprses and monetary polcy decsons of the ECB What do we mean by monetary polcy shocks? Market rates summarse the vast amount of nformaton used by the central bank to reach the monetary polcy decsons. In fact, these rates change as a reacton to the nformaton that arrves to the market. 18 In ths secton, we defne the daly changes of a set of short-term nterest rates as monetary polcy shocks. These daly changes, f devod of lqudty consderatons, are almost deal measures of how unexpected news changes market s expectatons of future monetary polcy decsons durng the maturty of the nterest rate consdered. On the days of monetary polcy meetngs, these shocks reflect the surprse assocated wth the monetary polcy decson. Very short-term nterest rates (from nstruments whch mature before the next meetng of the central bank) wll reflect the short-term surprses of the monetary polcy decson, that s f the decson was expected to take place at that precse meetng. Daly changes n other longer-term money market rates (from nstruments whch mature only after the next meetng of the central bank) allow for analyss f the surprse has also changed the short-term expected path of monetary polcy rates. Ths defnton of monetary polcy shocks s not new n the lterature. Roley and Sellon (1998) Kuttner (2001), Poole and Rasche (2001), Cochrane and Pazzes (2002) have used the daly change n some money market nterest rates as a measure of the monetary polcy shocks (the surprse or unexpected component of the monetary polcy decson). 19 Most of the prevous papers, however, defne the monetary polcy shocks as daly changes n market rates on the days n whch the central bank took a monetary polcy decson (and only as a prevous step to analysng the mpact of these shocks on the yeld curve). In our vew, defnng the shocks on a daly bass, rather than only on monetary polcy meetng days makes Daly changes n rsk premum can be consdered very low at these short horzons. In any case, the rsk prema n the euro area s estmated not to be sgnfcantly dfferent from zero. See footnote number 14. Favero et al (1998) defne the movement n the overnght rate as polcy shocks and defne monetary polcy surprses as the dfference between observed overnght rates and expected overnght rates. ECB Workng Paper No 192 November

15 sense, as t permts the comparson of the shocks on the days of the meetngs to other news or events that have affected the perspectve of future monetary polcy decsons. It allows to quantfy the mpact of monetary polcy decsons from the normal nose n the market. Besdes extendng the defnton of shocks to daly changes n market nterest rates, what s new n ths paper s the way we calculate monetary polcy shocks n the euro area. The nsttutonal framework matters a lot n the analyss of what the changes n money market rates mean. Whle n the Unted States there s a strong consensus n the lterature that the Fed Fund rates should be used to assess expectatons 20, t s not easy to fnd such a consensus n the euro area Monetary polcy shocks n the euro area: whch rates could we use? Every nterest rate may have ts own advantages and dsadvantages. Usng daly changes n EONIA, for example, provdes a measure of shocks hghly nfluenced by lqudty ssues, rather than (solely) by monetary polcy consderatons. EONIA swap rates (whch span out to one year) mght be a better alternatve as they are not as affected as the EONIA by lqudty ssues, especally for maturtes larger than two weeks. However, they are not completely free of the characterstcs of the specfc operatonal framework. Let us take a (rather) extreme example to clarfy ths. Assume that we use the two-week EONIA rate to gauge market expectatons. If at the begnnng of a mantenance perod market partcpants receve a pece of news that changes the expectaton of nterest rates movements by the ECB only for a meetng takng place n the next mantenance perod, the two week rate may not change at all. If, however, ths same event occurs less than two weeks before the end of a mantenance perod, the effect wll be partally covered by the two-week rate, and the more so as the end of the mantenance perod approaches. 21 All ths suggests that, to the extent that ths type of effects exsts, by measurng shocks wth the short-term money market rates we could be underestmatng the monetary polcy shock f the shock occurs that day. In addton, we may also be measurng as a shock the mpact of nformaton that became avalable at the begnnng of the mantenance perod See Thornton (1995). The fact that the US monetary polcy mplementaton mples daly open market operatons allows the Fed Funds rate to have more nformaton about market expectatons than the nformaton contaned n the EONIA where weekly and monthly patterns exst due to bank s lqudty management consderatons. For a recent comparson on the approprateness of the dfferent rates to measure expectatons of monetary polcy, see Gürkaynak, Sack and Swanson (2002). The behavour of daly rates n the mantenance perod s explaned n Perez Quros and Rodrguez (2001). 14 ECB Workng Paper No 192 November 2002

16 Whle longer-term money market rates provde a pcture of how the market vew the path of key ECB nterest rates, they mght not be devod of these specfc problems ether. Take the monthly rate. Whle ts changes are clearly more related to monetary polcy expectatons over longer horzons, some lqudty consderatons, such as the end-of-month and end-of-year effects may also matter. Other long-term nstruments, such as EURIBOR future contracts, whle they are not affected by these consderatons and form a very deep market, may have other problems. As the contracts apply to a fxed perod of tme, the maturty of the nstrument changes as tmes passes, whch does not happen wth EONIA swap rates. All n all, there are reasons to use an array of nterest rate data to measure the monetary polcy shocks n the euro area. Obvously, there s a wde pool of rates from whch we can extract the nformaton. Before that decson, however, we should test f, on average, all the varables contan the same amount of nformaton, abstractng from the mpact of lqudty consderatons n very shortterm money market rates. It s of partcular nterest to test f mplct or forward rates and the actual realsaton of rates present a long-term relaton showng a stable behavour of the spreads. If ths were the case, mxng nformaton from mplct rates and actual rates would be approprate to solve the problem of contamnaton of the nformaton that comes from dfferent lqudty consderatons. The best way of testng for the long-term relaton between actual and mplct rates s to check f these varables present a unt root but that a lnear combnaton between the actual and the mplct rates are statonary,.e. a contegraton relaton exsts between them. In partcular we check for contegraton n the followng set up: t = α + β j * E t-j ( t ) + ε j t+k (2) where t s the one month nterest rate, and E t-j ( t ) represent the one-month rate n one, two and three months as ndcated by the value of j=1,2,3. In all cases, for both, the euro area and the US, the seres show contegraton and the β j can be accepted to be equal to one. In ths set up, the ε j t+k represent, not only the spread but also the shock to the nformaton set n t-1. It seems that there s a long-term equlbrum (markets do not make mstakes on average) and that devatons from ths equlbrum are statonary. We can therefore wden our set of money market nterest rate rates and combne them n order to acheve a better specfcaton for the monetary polcy shocks Whle an approach usng ths lne has been proposed n the lterature to measure predctablty over longhorzons, and our analyss show that overall the decsons have been predcted on average up to three-months n advance, t has the problem that the nformaton set s not the same. Whle the expectatons are calculated wth the set of nformaton at t-j (for j=1, 2, and 3 months), the actual realsaton of the one-month rate uses nformaton up to t. Results are avalable upon request. ECB Workng Paper No 192 November

17 3.3. Monetary polcy shocks n the euro area: applyng prncpal components (PC) We propose to use the daly changes of several money market nterest rates and add them up daly. However, nstead of assgnng ad hoc weghs to each of the nterest rates used, we let the data speak by extractng ther prncpal component, wthout dong any type of nterventon n the seres. The objectve s to capture the man common component that shapes the evoluton n all these rates. The partcular consderatons that mght affect only one seres (and that should not be related to monetary polcy consderatons) would n the ma jorty of cases not play an mportant role n the seres obtaned through the prncpal component. We are also nterested n measurng shocks wth rates of dfferent maturtes. Daly changes n longer-term nterest rates wll reflect better how the expected short-term path of offcal nterest rates changes. For example, f after a monetary polcy decson of the ECB market partcpants are only surprsed by the tmng, say because they expected the change a fortnght after, longer-term nterest rates mght not change much. However, we do not want to use very long money market rates, as ther lqudty, and therefore ther nformaton content dmnshes progressvely. 23 We use daly changes n the EONIA, changes n the EONIA-swap wth maturtes of one-week, one, two and three-months, and the change n the closest three-month EURIBOR futures. 24 We defne dfferent measures of monetary polcy shocks usng prncpal components (PCj), accordng to the maturtes of the nterest rates. PCall s calculated applyn g prncpal components to the daly changes of all the above mentoned money market rates. PCshort uses the market nstruments up to and ncludng the one-month rate (EONIA, the one-week and the one-month rate). PClong uses the two and three-month EONIA swap rate and the three-month EURIBOR future. Fnally, PCnoe s PCall wthout the EONIA rate, whch s very volatle and could affect the results. 25 Whle we would expect that PCshort could stll be nfluenced by lqudty consderatons (due to the weght of EONIA), we would expect that the other defnton of shocks to be devod of lqudty consderatons An analyss of the monetary polcy shocks and the monetary polcy decsons of the ECB (and the US Federal Reserve) We now have daly seres of monetary polcy shocks for the euro area n whch the shocks generated by the monetary polcy decsons of the ECB are only observatons of that seres See ECB (2001a). Annex 1 presents a detaled descrpton of all the nterest rates used n the paper. We dd not use longer-term rates, as those rates mght reflect other consderatons dfferent other than the expectatons of monetary polcy. Annex 2 analyses n detal the prncpal component technque used and the calculated weghts for each defnton of shock. 16 ECB Workng Paper No 192 November 2002

18 These daly shocks (at dfferent maturtes) provde a benchmark wth whch we can compare the monetary polcy announcements of the ECB. We defne a monetary polcy surprse as a shock bgger than two tmes ts standard devaton. Of the 78 meetngs of the Governng Councl (n a sample of 878 observatons) only between 7 and 10 (dependng on the defnton of the shock used) were surprses. 26 That s, only between 18-24% of the surprses n the sample have been caused by monetary polcy decsons of the ECB (ncludng decsons to change rates and to keep them unchanged). That s, other peces of nformaton have an mportant affect on the expected path of key nterest rates. Of all the meetngs of the Governng Councl the markets have not been surprsed n 87% of them (usng the shocks measured by PCshort). The percentage ncreases slghtly to 90-91% when the other measures of shocks are used. These results, together wth the meetngs of the Governng Councl of the ECB n whch a surprse occurred (accordng to the four measures of shocks), are presented on Table 2. Table 3 n turn lsts the shocks on the other days of the sample, and ponts to possble determnants. [Insert Tables 2, 3a-3b about here] In turn, Fgure 4 plots for all the monetary polcy meetngs of the ECB the changes n the key ECB rates and the monetary polcy shocks on those days. [Insert Fgure 4a-4d about here] By defnton, these shocks capture the surprse assocated wth the tmng of the monetary polcy decsons. In fact, t s easy to see why ths holds. For every shock, we can defne the expected change n the key ECB rates one day before the meetng as E t-1 ( k t ) = k t - PC t (3) where k s the level of the MRO or key nterest rate. As a major dfference to the approach taken n Secton 2, the sze of the changes n the key ECB nterest rates now matters. For example, f the market expects a cut n key ECB rates of 50 bass ponts and rates are only lowered by 25 bass ponts, the shock would adjust by some 25 bass ponts 27. In fact, Fgure 4 shows how some of the changes of 50 bass ponts that were not consdered surprses n the analyss conducted n Secton 1, now appear as surprses The total number of surprses oscllated between 32 and 55, dependng on the shock (see Table 2). Care needs to be taken when nterpretng these results as the shocks are constructed wth rates that span more than one meetng. These expected rates, however, are good sgnals of the monetary polcy expectatons. Annex 3 explots these seres of expected rates to show, estmatng a Probt, that ths s a good measure of expectatons of changes n the key ECB nterest rates. ECB Workng Paper No 192 November

19 Ths same analyss can be appled to the Unted States Federal Reserve. Followng Poole and Rasche (2001), we use the change n the one-month-ahead federal fund future rate as our measure of shocks (PR from now on). 28 We also use the two-month-ahead change n the Fed fund future (PR1) as a shock, to see f the results are senstve to the horzon (ts maturty ranges between 2 and 3 months, whle PR spans only between 1 and 2 months dependng on the day of the month). For the 877 observatons n the sample, and the 30 meetngs of the Fed n that perod 29 only 8 of the surprses (both accordng to the measure of PR and PR1) were on days n whch the FOMC met. That s, only between 22-23% of the surprses n the sample (agan, defned as 2 tmes the standard devaton of each seres) have stemmed from the meetngs of the FOMC, a smlar rato to the one obtaned for the euro area. However, gven the lower number of meetngs, the percentage of tmes n whch the market has not been surprsed by the monetary polcy decsons s 73%. Table 4 shows these and also lsts the meetngs of the FOMC n whch a surprse was estmated to have occurred (accordng to the two measures of the shocks whch provde very smlar results). Smlar to the euro area, an ndcatve (and noncomprehensve) table whch lsts all the shocks and the events whch happened those days s provded n Table 5. [Insert Table 4 and Tables 5a-5b about here] Fgure 5 plots for all the meetngs of the FOMC the change n the Fed Funds rate and the correspondng shock PR on that day (the results wth PR1 are very smlar). [Insert Fgure 5 about here] Overall, ths secton has shown that usng a more demandng measure of the predctablty of the monetary polcy decsons of a central bank, the markets have not been surprsed on 87-91% of the monetary polcy meetngs of the ECB, a result whch s slghtly better than for the FOMC. 4. Has the daly pattern of the varance of these shocks changed wth the announcements of monetary polcy? In ths secton we analyse to what extent the volatlty pattern of the seres of shocks change on the days of the meetngs. Ths s a good measure of how the monetary polcy decsons have surprsed the markets. Tables 3a-3b (5a-5b) lst all the surprses n the euro area (n the 28 Poole, Rasche and Thornton (2002) show that ths measure of shock s broadly smlar to the measure used by Kuttner (2001), that uses the change n the Fed Fund rate of the current month. 18 ECB Workng Paper No 192 November 2002

20 Unted States) n the sample. The last column ndcates the peces of news that were cted from market sources (Bloomberg) to be the major movers that day. As already analysed n the prevous sectons, besdes the monetary polcy meetngs, the nformaton that arrves to the market on a daly bass changes the expected path of monetary polcy rates. After an examnaton of the lst, the natural varables to check seem to be related to releases of money data, nflaton and leadng ndcators for actvty. We use an EGARCH specfcaton for the analyss of the dfferent factors on the volatlty. The EGARCH model, ntroduced by Nelson (1991) and wdely used n the fnance lterature allows a flexble dynamc specfcaton for the varance that easly solves the nonnegatve constrant assocate wth the GARCH models. The estmated model s: PCj =β 0 +ε t where ε t N t ( 0, ht ) n ε ε t j t j 2 ln( h ) t) = λ ' Vt + δ j,1(ln( ht j) λ' Vt j) + δ j,2 + δ j,3( (4) j= 1 ht j h π t j where PC j represents the prncpal component (the change n a set of money market nterest rates). The rest of the varables are: V V V V V V V V V 1t 2t 3t 4t 5t 6t 7t 8t 9t = Cons tan t = End MP Dummy = Begnnng MP Dummy = M 3 Publcaton Dummy = End Year Dummy = EndMonthDummy = IPC PublcatonDummy = IFOPublcatonDummy = Meetng DayDummy We can rewrte the volatlty equaton as: 29 There s no need to take out the meetng on 29 December 2001, as our measure of shock s not affected by the end-of-year effect. ECB Workng Paper No 192 November

21 ln( h ) = λ' V t λ X t + t j= 1 n εt j t j 2 [ δ j,1( λ' Vt j) ] + δ j,1 ln( ht j) + δ j,2 + δ j,3( ) n t j t j δ j,1 ln( ht j) + δ j,2 + δ j,3( j= 1 ht j ht j + n ε j= 1 t j t j ε 2 π ) h ε h π = (4a) where X t nclude the varables n V t and n lags of those and λ1 s a vector that ncludes the k coeffcents of Vt and (k-1)*n coeffcents that affect the lags of the dummy varables. We do not mpose the non-lnear restrctons mpled by (5) allowng a dfferent transmsson of the volatlty assocated to the specal days but not constranng (as would be the case f we dd not consder the lagged dummes) that these specal days transmt the varance n full as f the ncrease or decrease varance assocated to a calendar or meetng effect was due to a shock. Fnally, we test for the optmal value of the number of lags obtanng n=1. Lookng at Table 6, the results of the dfferent prncpal components specfcatons and the EONIA confrm that short-term rates are affected by lqudty needs and that ths s not true n the case of the long term rates. Dummy varables related wth perods assocated wth excess demand or supply of lqudty are clearly sgnfcant n the volatlty equaton for the shorterterm shocks and not sgnfcant for the longer-term shocks. Also, a prncpal component model that ncludes both short and long term rates seems to also avod ths lqudty problem. Ths result gves us some motvaton for the use of the prncpal component methodology. It allows us to, ncorporatng some nformaton on the short rates, avod the lqudty problem that could hde mportant volatlty movements. [Insert Table 6 about here] What are the results that we obtan for the volatlty assocated to the meetng? To start wth from all the events tested, the meetngs are the man drvers of the volatlty of the seres. Interestngly, economc varables do not seem to play a major role n the pattern of volatlty. Ths could be due to the fact that when euro area data comes out, data for ndvdual countres has already been publshed, reducng ts nformaton content. Whle we use CPI and the IFO for Germany (other euro area data has been found to be not sgnfcant), other country data (n the case of the IFO) and provsonal data for nflaton for the German Länder (n the case of the CPI) whch are publshed n advance of the data ncorporated n V mght explan ths result. Second, there s a greater varance on the days of the meetngs of the Governng Councl compared to the days n whch no meetngs took place. In partcular, the varance on the days of the meetng s between 1.6 and 2 tmes bgger on meetng days. As the volatlty s hgher 20 ECB Workng Paper No 192 November 2002

22 the shorter the horzon, ths result could be seen as ndcatng that the market s less surprsed over longer horzons after a meetng of the Governng Councl. However, as n the prevous sectons, we want to compare these results wth the ones obtaned for the FOMC to analyse how much that volatlty s. Table 7 compares t wth the results of the euro area. As wth other measures of predctablty, we obtan ndcatons that the varance added on days of the meetngs of the monetary authorty has smlar values n the Unted States and the euro area for the sample checked. [Insert Table 7 about here] The results of ths secton ndcate that the monetary polcy decsons of the ECB ncrease the volatlty of nterest rates, compared to the normal volatlty of the seres. Ths ncrease s smlar to the one observed to the one assocated n the Unted States to the meetngs of the FOMC. At the same tme, the results seem to ndcate that the market s less surprsed over longer-term measures of shocks. 5. Impact of the shocks on the term structure of nterest rates As noted n the ntroducton, several papers have analysed the mpact of the monetary polcy shocks from the days of the monetary polcy meetngs of the central bank to the yeld curve. Ths allows to measurng how the unexpected component of the monetary polcy decson s transmtted to the term structure of nterest rates. Dfferently from these papers, however, we are not only nterested n the mpact of these monetary polcy shocks on the days of the meetngs on the term structure of nterest rates, but also n the mpact of these specfc shocks compared to the shocks on any other day. Monetary polcy s conventonally vewed as runnng from short-term nterest rates managed by central banks to longer-term rates. Abstractng from default rsk consderatons, the expectaton theory of the term structure of nterest rates mples that (unexpected) monetary polcy decsons affect the prces of bonds to the extent that they lead nvestors to revse ther expected path of the monetary polcy rate. The mpact of the surprse change n the key ECB nterest rates on longer-term bond yelds wll depend on the percepton of the persstence of the surprse. Accordng to the expectaton hypothess, a surprse change n the key rates that s expected to last for the term of the bond wll ncrease the yeld on ths bond by the same amount. However, f monetary polcy decsons are perceved to have only a temporary effect, the mpact of a change n the key ECB nterest rates would be smaller the longer the maturty horzon of the bond. ECB Workng Paper No 192 November

23 The expectaton hypothess mght not be the only force shapng the move n the term structure. Gven the commtment of modern central banks to keep nflaton low over the medum term, a credble monetary polcy affects long-term bond yelds by anchorng nflaton expectatons over the long run (the Fscher effect). 30 If a central bank s credble, ts actons should be seen as compatble wth the mantenance of prce stablty over the medum term. We can see the movement n the term structure of nterest rates as the net effect of two forces, the expectaton theory and the Fsher effect. The mpact of a monetary polcy decson on the term structure depends on the mpact of such a decson on the future path of short-term nterest rates and on the expected effect of the monetary polcy decson on expected nflaton over long horzons. The former effect s lkely to domnate the short-to-medum term of the yeld curve, whle the latter s lkely to domnate the medum to long-end of the term structure Monetary polcy shocks and the yeld curve An extensve stream of the lterature has measured the mpact of monetary polcy decsons on the yeld curve. An early work of Cook and Hahn (1989) examned the one-day response of bond rates n the Unted States to changes n the target Fed Funds rate from 1974 to They regressed the change n the Treasury Bll and several bond rates ( R, where stands for the maturty of the bond) on the change n the target Fed funds rate (target rate or key rate, k). The sample conssts only of the days n whch the Fed changed the Fed Funds target rate. R = α + β k + ε (5) t t In more recent papers Kuttner (2001), Poole and Rasche (2001) and Poole, Rasche and Thornton (2002) have perfected ths approach, usng the Fed Funds Futures to dentfy the expected and unexpected component of the monetary polcy decson (the shock) 32. Once dentfed, they estmate the response of market rates to the expected and unexpected shocks on days n whch the Fed funds rate was changed. In these studes, the change n the rate of the current (Kuttner) or the one-month ahead (PR and PRT) federal funds futures contract t The prmary objectve of the monetary polcy of the ECB s the mantenance of prce stablty over the medum term. Prce stablty s, n turn, defned, as year-on-year ncreases of the HICP of below 2%. An updated estmaton of the approach of Cook and Hahn (1989) s developed n Roley and Sellon (1998) and Kuttner (2001). See Favero et al (1996) and Buttglone et al (1996) for further work on the mpact of monetary polcy decsons on the term structure of nterest rates conducted for several countres n Europe, and also for the Unted States. 22 ECB Workng Paper No 192 November 2002