Do market-based inflation expectations matter for interest rate decisions? 1

Transcription

1 Do marke-based inflaion expecaions maer for ineres rae decisions? 1 Jérôme Coffine 2, Jean-Séphane Mésonnier 3, Axel Lang 4 This version : 19 February 2009 Absrac We examine o wha exen cenral bankers pay aenion o marke-based privae secor inflaion expecaions when seing he policy raes. We esimae empirical reacion funcions for hree cenral banks (he ECB, he US Federal Reserve and he Bank of England) over he las decade using ordered-probi echniques. The resuls show noably ha marke-based inflaion expecaions have no significan impac on policy decisions by he ECB, while hey seem o play a significan role in he decision process in he UK and he US. Keywords: inflaion expecaions, inflaion-indexed securiies, moneary policy, orderedprobi model. JEL classificaion: E31, E44, E52, E58 1 The views expressed in his paper are hose of he auhors and do no necessarily reflec he views of he insiuions hey belong o. 2 Banque de France. Address: Banque de France 1, rue de la Vrillière Paris, France. Phone number: address: jerome.coffine@banque-france.fr. 3 Banque de France. Address: Banque de France 1, rue de la Vrillière Paris, France. Phone number: address: jean-sephane.mesonnier@banque-france.fr. 4 This paper was realized during he ime Axel Lang was an inern a he Moneary Policy Research Uni a he Banque de France. 1

2 1. Inroducion In modern economies, i is widely recognized ha he sae of inflaion expecaions grealy influences acual inflaion oucomes. 5 Unsurprisingly, cenral banks in developed economies do monior closely curren developmens in inflaion expecaions formed by privae agens, as hey can provide useful signals of emerging risks o price sabiliy. Among all measures available, marke-based measures of privae expecaions are scruinized wih paricular aenion by cenral bankers, noably he break-even inflaion raes (BEIR) compued as he difference beween yields of nominal and hose of inflaion-indexed bonds of comparable mauriies. This close monioring of marke-based inflaion expecaions is winessed by numerous speeches by cenral bank officials 6 as well as public minues of policy meeings. 7 Indeed, alhough such marke-based indicaors are known o be noisy indicaors of rue inflaion expecaions and can be disored by echnical facors affecing he liquidiy of he underlying securiies, heir availabiliy a high frequency and for various horizons have made hem of special ineres for policy. Hence, cenral banks have devoed imporan research resources in recen years o filer ou he relevan informaion abou expecaions ha hey conain. 8 Besides, recen empirical research shows ha, a leas for horizons longer han wo o hree years, BEIR provide a reliable basis for assessing he anchoring of inflaion expecaions a a medium erm horizon. 9 This said, i remains unclear o wha exen cenral banks use he informaion hey exrac from BEIR o define heir policy. In his paper, we aim o clarify his poin, esimaing 5 See e.g. Gali e al. (2001) for esimaes of he so-called New Keynesian Philips curve. 6 See e.g. Bernanke (2007), Triche (2007) and Vickers (1999) for speeches highlighing he role of his monioring a he US Federal Reserve, he ECB and he Bank of England respecively 7 See Bank of England (2006) for an example of references o BEIR in minues of is Moneary policy commiee. Anoher similar example is given in a recen inroducory saemen by ECB s Presiden J.C. Triche (8 November 2007), saing ha We are looking very carefully a all [ ] informaion we exrac from he financial markes. 8 See for insance Hördahl and Trisani (2007) and Adrian and Wu (2009) and references herein. 9 See Beechey e al. (2008), Coffine and Frappa (2008). 2

3 policy reacion funcions for hree major cenral banks: he US Federal Reserve, he ECB and he Bank of England. Compared wih he abundan empirical lieraure on moneary policy rules, he specificiy of our approach is hreefold. Firs, we use ordered probi models of ineres rae decisions in order o be consisen wih he discree naure of changes in policy raes. Second, we base our esimaions on reconsruced real-ime daa se, so as o reproduce he rue condiions which were prevalen a he ime when observed ineres rae decisions were effecively aken. Third, whereas several sudies on moneary policy reacion funcions already combine real ime daa and survey-based measures of privae inflaion expecaions 10, we focus on cenral banks reacion o BEIR. We find ha marke-based inflaion expecaions have no significan impac on policy decisions by he ECB, while hey seem o play a significan role in he decision process in he UK and he US. The res of his paper is organized as follows. Secion 2 presens he daa and he empirical sraegy. Secion 3 commens he resuls and provides robusness checks. Secion 4 concludes. 2. Model and daa We aim o model moneary policy decisions regarding he level of he policy rae in hree developed economies (he euro area, Unied Kingdom and he Unied Saes). Since decisions of policy ineres rae changes in he hree considered economies are aken in small seps of usually 25 bp and occasionally 50 bp each, i is inappropriae o fi a Taylor (1993) rule ype model of he policy rae using OLS. An alernaive way o proceed is o esimae he probabiliy of a change in he policy rae condiionally on a se of sae 10 See e.g. Gerdesmeier and Roffia (2005) and Gorer and De Haan (2008) for Taylor rule esimaions using survey based expecaions daa in he case of he ECB. 3

4 variables observed by he cenral bank, using an ordered-probi model (see Gerlach, 2007). Le i denoe he repo rae and i T he relaed cenral bank s arge. In line wih he Taylor rule lieraure, we can assume ha he arge rae is summarized as a simple linear funcion of a few macroeconomic sae variables, colleced in a vecor X, such ha i T = α + βx. Allowing for gradual adjusmen of he nominal rae o is arge as for insance in Judd and Rudebusch (1998), we ge an equaion describing he required change in he policy rae: i * T ( i i 1) + ρ i 1 ε = θ + (1) where ε is a residual. Subsiuing he Taylor rule expression for he arge rae and rearranging, we have hen: i * ~ = AX θ i + ρ i + ε 1 1 Noe ha since changes in policy ineres raes are implemened in discree seps, he coninuous variable * i is an unobservable, or laen, variable. For he purpose of esimaion, we consider a period beginning eiher in 1997, in 1998, 2001, depending on he availabiliy of BEIR daa 11 and ending in July Indeed, we choose o cu our daase jus before he financial crisis ha sared in he summer of 2007, as decisions aken since hen were likely o be moivaed by excepional financial 11 Indexed-linked bonds were inroduced laely in mos developed economies, e.g. in 1997 for he US TIPS and 2001 for he French OAT i indexed on euro area inflaion. An excepion is he UK were inflaionlinked governmen bonds were issued as early as We prefer o limi o he period poserior o 1997, when he Bank of England gained formal independence, boh for comparabiliy of resuls across counries and o resrain he risk of ignoring changes in policy regimes. 12 Noe ha resricing he observaion period o a sample common o all four counries does no affec qualiaively he resuls. Since we had no reason o suspec regime shifs in moneary policy in eiher he US or he UK beween 1997 and 2001, we preferred o use a longer sample for boh counries. 4

5 sabiliy concerns ha may no be refleced in he sandard reacion funcions of he cenral banks. 13 Table 1: summary of ineres rae decisions for hree cenral banks Sample Up Down Sau quo Obs. Euro area 11/08/2001 o 07/05/ (11.6%) 4 (5.8%) 57 (82.6%) 69 UK 07/10/1997 o 07/05/ (13.9%) 17 (13.9%) 88 (72.1%) 122 US 05/19/1998 o 08/07/ (27.8%) 16 (20.2%) 41 (51.9%) 79 As shown in Table 1, saus quo decisions are a frequen oucome of moneary policy councils meeings, ranging from 52 % of all decisions for he US o almos 83 % in he euro area case. Besides, changes of 50 bp or more are rare and ofen relaed o excepional evens, such as he Sepember 11 shock in To avoid weak idenificaion problems, we focus on explaining he direcion of he policy decisions (i.e. ups, downs, saus quo), raher han heir size. 14 We relae he sign of he observed (discree) changes in he policy rae o he noional level changes * i required by equaion (1) assuming ha ineres rae decisions are characerized by hreshold behaviour: he decision variable denoed D, changes values whenever he laen desired change in he policy rae passes wo unobservable hresholds γ 1 and γ 2 : 1: i * γ 1 D = 0 : γ i * γ 2 (2) 1 + 1: i * γ 2 Equaions (1) and (2) consiue an ordered-response model of policy rae decisions by he cenral bank j. Assuming ha ε follows a sandard normal disribuion, we can wrie he probabiliies of he differen oucomes as: 13 For similar reasons and as a robusness check, we also excluded from our samples he few excepional ineres rae cus ha were decided in he afermah of he erroris aacks of Sepember 2001 in he Unied Saes. Resuls are qualiaively unchanged. 14 Jansen and De Haan (2006) make a similar choice. 5

6 Pr Pr[ D = 1 z ] = Φ( γ 1 z ' β ) [ D = 0 z ] = Φ( γ 2 z ' β ) Φ( γ 1 z ' β ) Pr[ D = + 1 z ] = 1 Φ( γ z ' β ) 2 (3) where φ denoes he cumulaive normal disribuion and z is a vecor of explanaory variables: z =[X, i -1, D -1 ]. In line wih he empirical lieraure on cenral banks reacion funcions as aking he form of simple rules 15, official sraegies of he various cenral banks considered here, as well as how hey communicae abou he analysis underlying heir decisions, we include in X, for each cenral bank: - he laes available business climae indicaor as a measure of domesic real aciviy, - he laes available measure of domesic consumer price inflaion, - break-even inflaion raes as a measure of medium o long run inflaion expecaions, using an average of daily daa over he las wo weeks up o he policy meeing. As a robusness check, we also examine varians of he baseline regressions were we conrol eiher for alernaive measures of privae inflaion expecaions as aken from cenral banks Surveys of Professional Forecasers, or for inernal inflaion projecions a a one-year horizon, - he laes available hree-monh moving average growh in he relevan large moneary aggregae, - he laes observed value of he year on year change in he nominal effecive exchange rae (NEER), aken as a measure of exernal inflaion pressures. Noe ha we ook special care in consrucing our daase so as o reproduce he informaion base of he respecive cenral banks a he ime heir moneary policy councils me o ake ineres rae decisions. We considered several alernaive measures of 15 See noably Clarida e al. (1997) for a classical example of Taylor rules esimaion. 6

7 real aciviy and BEIRs for each counry. This did no aler qualiaively he resuls. 16 See he appendix for deails of daa used and sources. Char 1: policy raes and measures of privae inflaion expecaions (Noe: he shaded area corresponds o pos 9/11 policy rae cus) 4.5 Euro area (Nov May 2007) 8 Unied Kingdom (Oc May 2007) SPF 1 year ahead BEIR France 2012 Repo rae SPF 2 years ahead BEIR 2013 Bank rae 7 Unied Saes (May July 2007) SPF 1 year ahead BEIR 10 years Fed funds arge In an ordered probi model, he signs of coefficiens can be inerpreed regarding he probabiliies of being in he exreme cases only. For insance here, a significan posiive 16 The corresponding ables are no reproduced here o save space bu are available upon reques o he auhors. 7

8 coefficien for he real aciviy variable implies ha higher records for real aciviy increase he probabiliy of an ineres rae hike and decrease he probabiliy of an ineres rae cu, bu nohing can be inferred abou he probabiliy of a saus quo. This being said, he expeced signs of he esimaed coefficiens in (3) are posiive for real aciviy, curren inflaion, expeced inflaion and money growh, negaive for he effecive exchange rae (an increase in he NEER means here an appreciaion of he currency, which may relieve inflaionary pressures). We also expec a negaive sign for he lagged level of he policy rae (reflecing some mean revering behaviour of he shor erm ineres rae). The sign of he lagged ineres rae decision can be eiher posiive, reflecing possible ineres rae smoohing by he cenral bank, or negaive, if he cenral bank moves raes pre-empively o clear he air. 3. Resuls Esimaion resuls of he ordered-probi model for each cenral bank are summarized in Table 2. The firs wo columns refer o he ECB, while columns (3)-(5) and (6)-(8) refer o alernaive specificaions of he model for he Bank of England and he US Fed respecively. Noe ha according o he R² as well as he share of correc predicions (see Table 3), he models do a relaively good job in explaining mos cenral banks decisions, noably as regards he ECB and he US Fed. 8

9 Table 2: esimaion resuls Euro area Euro area UK UK UK US US US (1) (2) (3) (4) (5) (6) (7) (8) Real Aciviy 4.20** 2.16** 0.03** ** 1.53*** 1.55*** 0.72 Inflaion Money 1.22* 1.93*** -0.08* ** NEER -0.39*** -0.37*** 0.16*** 0.18** 0.15*** BEIR *** 9.10*** 5.05*** 2.94*** 2.94*** 4.03** SPF ** * Inernal Forecass Previous Decision -5.15** -4.92*** -0.71*** -0.98** -0.75*** 2.10*** 2.09*** 1.36** Lagged policy rae *** -1.80*** -1.1*** -0.92*** -0.94*** Observaions R² Noe: *, ** and *** denoe significan coefficiens a he 10%, 5% and 1% levels respecively. Model in column (8) is esimaed over only, due o daa availabiliy on real-ime inernal projecions by he Fed 9

10 We now commen on he signs of he esimaed coefficiens. Remember ha he signs of coefficiens in Table 2 can be inerpreed bu no heir absolue values, since hey are no elasiciies. We find ha he probabiliy of an ineres rae hike is significanly increased by beer prospecs for real aciviy in all counries, alhough his effec is miigaed in he UK and he US when addiional conrols for projeced inflaion are added, hining ha hese are likely o be correlaed o our indicaors of he business cycle. Curren inflaion appears o have no significan impac on policy decisions. Ineres rae decisions in face of acceleraing money are diverse. The ECB ends o raise is key rae in face of higher money growh, consisenly wih is wo pillar approach. In conras, he US Fed does no seem o care, as does he Bank of England according o he bes model specificaion (wih SPF daa, column 4). Nominal appreciaion of he NEER has no impac in he US, while i significanly increases he probabiliy of a rae cu by he ECB, bu on he conrary increases he probabiliy of a rae hike in he UK. The direcion of he laes decision aken has a significan impac on he curren one, bu he ECB and he Bank of England seem o follow a clear he air paern, while he US Fed exhibis more of a smoohing behaviour. Finally, he lagged level of he policy rae acs as mean-revering force on curren decisions in boh he US and he UK. Focusing on he role of inflaion expecaions as derived from indexed-linked securiies, we find ha hey had no impac on ineres rae decisions by he ECB. Conversely, he US Fed and he Bank of England ended o ighen heir moneary policy sance in 10

11 response o higher levels of BEIRs. Noe ha our esimaes for he ECB are in line wih earlier findings of Gerlach (2007) who follows a similar empirical sraegy. As a robusness check, we re-esimaed he same models adding eiher measures of inflaion expecaions aken from surveys of professional forecasers (SPF) as colleced by cenral banks, or inernal forecass when hey were publicly available (see columns 2, 4-5 and 7-8 in able 2). Regarding he marginal informaion conen of BEIRs for cenral bank decisions, resuls are qualiaively unchanged. Ineresingly enough, one-year ahead inflaion projecions aken from he SPF sources maer for ineres rae decisions by he ECB 17 and he Bank of England, bu no he US Fed. 18 Besides, inernal one-year ahead projecions by he Fed of he Bank of England do no seem o have any bearing on heir conemporaneous decisions. Anoher quesion is wheher he consideraion of marke-based inflaion expecaions improves he predicive power of he model. As a comprehensive approach makes he comparabiliy of resuls difficul, we decided o focus on model (2) for he euro area, model (5) for he UK and model (7) for he US. The Table 3 below presens he correspondence beween he acual ineres rae changes and hose prediced by he model. For insance, one can see ha over he sample considered, he ECB decided 8 ineres rae hikes, 6 of which were correcly prediced by he model and 2 were no. In general, our models do a good job in predicing he ineres rae decisions: abou 90% of decisions in he euro area, 85% in he US and 84% in he UK are correcly prediced. The resuls are especially impressive for he US, where 20 ou of 22 ineres rae hikes are correcly prediced by he model (7). 17 Our esimaes also corroborae he resuls of sudies running linear esimaions of policy rules for he euro area, which ofen exhibi a significan posiive response of he shor erm money marke rae o surveybased measures of inflaion expecaions. See e.g. Gorer e al. (2008), Gerdesmeier and Roffia (2005). 18 Regarding he Bank of England, we used wo-years ahead inflaion projecions, in line wih he horizon of he inflaion argeing sraegy of he Bank. 11

12 Table 3: acual and prediced ineres rae changes (baseline models) Ups Downs Saus quos % correc Obs. Cor. Inc. Obs. Cor. Inc. Obs. Cor. Inc. Euro area % UK % USA % Looking more precisely a he informaive conen of marke-based inflaion expecaions, as measured by he break-even inflaion raes, we would like o measure he predicive power of he same models wihou he BEIR as explanaory variable. Thus, we reesimae he models (2), (5) and (7) 19 wih exacly he same endogenous and exogenous variables as before, bu he BEIR measures ha are dropped. The resuls are presened in he Table 4. Table 4: acual and prediced ineres rae changes (models wihou BEIR) Ups Downs Saus quos % correc Obs. Cor. Inc. Obs. Cor. Inc. Obs. Cor. Inc. Euro area % UK % USA % The resuls are consisen wih hose of he Table 2. In paricular, hey show ha disregarding BEIR does no aler he predicabiliy of ineres rae decisions by he ECB. On he conrary, BEIR seem o play a very imporan role in he UK, especially for ineres rae hikes and cus, as he predicive power of he model for he wo ypes of decisions deerioraes much when BEIR are removed from he esimaion. BEIR seem o play a minor role in he US as regards he predicabiliy of ineres raes decisions. 4. Conclusion We esimae a simple model of ineres rae decisions by hree major cenral banks over using ordered-probi echniques o invesigae wheher cenral banks do reac 19 Resuls are available upon reques. 12

13 o inflaion expecaions as derived from inflaion-linked governmen bonds. Ineresingly, he resuls do no depend on wheher he cenral bank has adoped a formal inflaion argeing sraegy, as Bank of England, or no, as eiher he US Fed or he ECB. In paricular, we find ha he ECB does no reac o BEIRs, bu he US Fed and he Bank of England do. We also find ha curren inflaion has generally no significan impac on ineres rae decisions, bu indicaors of curren business condiions, pas decisions and he level of he policy rae before he meeing are decisive o explain policy moves. 13

14 References Adrian, T., Wu, H., The erm srucure of inflaion expecaions. Federal Reserve Bank of New York Saff Repors No. 362, February. Beechey, M., Johannsen, B., Levin, A., Are long-run inflaion expecaions anchored more firmly in he Euro area han in he Unied Saes? Finance and Economics Discussion Series , Board of Governors of he Federal Reserve Sysem (U.S.). Bernanke, B., Inflaion Expecaions and Inflaion Forecasing. Speech a he Moneary Economics Workshop of he Naional Bureau of Economic Research Summer Insiue, Cambridge, Massachuses, 10 July. Clarida, R., Galí, J., Gerler, M., The science of moneary policy: A New Keynesian perspecive. Journal of Economic Lieraure (37), pp Coffine, J., Frappa, S., Macroeconomic Surprises and he Inflaion Compensaion Curve in he Euro Area. Banque de France Working Paper Series No Gali, J., Gerler, M., Lopez-Salido, J., European inflaion dynamics. European Economic Review, vol. 45, Gerdesmeier, D., Roffia, B., The Relevance of Real-Time Daa in Esimaing Reacion Funcions for he Euro Area. Norh American Journal of Economics and Finance 16, Gerlach, S., Ineres Rae Seing by he ECB, : Words and Deeds. Inernaional Journal of Cenral Banking, mars, Gorer, J., De Haan, J., Taylor Rules for he ECB Using Expecaions Daa ( ). Scandinavian Journal of Economics, Vol. 110, Issue 3, pp Hördahl, P., Trisani, O., Inflaion Risk Premia in he Term Srucure of Ineres Raes. European Cenral Bank Working Paper Series 734. Jansen, D.-J., De Haan, J., Does ECB Communicaion Help in Predicing is Ineres Rae Decisions? CESifo Working Paper. Forhcoming in Applied Economics. 14

15 Judd, J., Rudebusch, G., Taylor's rule and he Fed, Economic Review, Federal Reserve Bank of San Francisco, pages Taylor, J., Discreion versus Policy Rules in Pracice. Carnegie Rocheser Conference series on Public Policy 39, 1993, Norh Holland, pp Triche, J.-C., Inroducory Saemen. 8 November. Vickers, J., Speech a he Glasgow Trades House. Lecure a Srahclyde Universiy, 26 May. 15

16 Appendix: Deails on daa definiions and sources Noe: whenever several possible measures of a variable have been esed, he indicaor seleced for our baseline esimaion (see able 2 in ex) appears in bold. Real Aciviy Euro area Unied Kingdom Unied Saes of America Business Climae Indicaor (European Commission) Economic Senimen Indicaor (European Commission); Business Climae Indicaor (OECD) Inflaion (y-o-y%) HICP (ECB) RPI (Office for Naional Saisics); CPI (Office for Naional Saisics);RPI ill 12/10/2003 hen CPI Non Farm Payrolls (Federal reserve, Sain Louis); Manufacuring PMI; Non-Manufacuring PMI CPI (Bureau of Labor Saisics); PCE (Bureau of Labor Saisics); CPI ill 12/17/2000 hen PCE Broad money (y-o-y%) M3 (ECB) M4 (BoE) M2 (Federal Reserve); M3 (no publicaion afer 03/2006) Nom. Eff. Exc. Rae (y-o-y%) Agains a 26- currencybaske (BIS) Break Even Inf. Rae Survey of Prof. Forecasers Inernal projecions 2012 (saring dae: 11/25/2001); 2014 (saring dae: 02/13/2004); 2020 (saring dae: 01/23/2004) SPF 1 year ahead; Agains a 26- currency-baske (BIS) 2009 (saring dae: 04/23/1986); 2011 (saring dae: 05/25/2001); 2013 (saring dae: 03/31/1993); 2015 (saring dae: 01/26/1995); 2020 (saring dae: 02/29/1996) SPF 2 years One-year ahead median of RPIX inflaion projecions (up o 2003), hen of CPI inflaion projecions (source Inflaion repors) Agains a 26-currencybaske (BIS) 10-year TIPS-derived expeced inflaion SPF 1 year One-year ahead inflaion projecions from he US Fed Green book, CPI ill 2000, hen PCE up o 2003 (source: ALFRED daabase, FRB of S- Louis). 16