Taylor Rules for Sweden s Monetary Policy Committee *

Transcription

1 Taylor Rules for Sweden s Moneary Policy Commiee * Henry W. Chappell, Jr. Professor of Economics Universiy of Souh Carolina Phone: Fax: chappell@moore.sc.edu Rob Roy McGregor Professor of Economics Universiy of Norh Carolina a Charloe Phone: Fax: rrmcgreg@uncc.edu Todd A. Vermilyea Vice Presiden Federal Reserve Bank of Philadelphia Phone: Fax: Todd.Vermilyea@phil.frb.org JEL Caegories: E520 - Moneary Policy E580 - Cenral Banks and Their Policies Keywords: Moneary policy, cenral banking * The views expressed are hose of he auhors and do no necessarily reflec he views of he Federal Reserve Bank of Philadelphia or he Federal Reserve Sysem. November 21, 2011

2 Absrac Taylor Rules for Sweden s Moneary Policy Commiee We esimae Taylor rules for Sweden s cenral bank, he Riksbank, covering he 2000 o 2011 period. Sweden is ineresing because of he ransparency of is moneary policy deliberaions and also because of is unusual experience in he recen recession. Sweden lapsed ino a severe recession in 2008 bu, unlike oher counries, had a rapid and robus recovery. Prior o he recession, he Riksbank s moneary policy appears o have been highly inerial. This finding is compaible wih he hypohesis ha commiee decision-making ends o be inerial. However, he policy response o boh he recession and he recovery in Sweden was quick and subsanial, exhibiing less ineria han would have been prediced on he basis of pre-recession Taylor rule esimaes. A noable feaure of our economeric work is he use of a dynamic Tobi specificaion o accoun for he lower bound on nominal ineres raes encounered during he recession. 2

3 I. Inroducion The Taylor rule has become an increasingly acceped framework for empirically analyzing moneary policy choices (Taylor 1993). During he Grea Moderaion, he lack of macroeconomic volailiy may have made i difficul o ge reliable esimaes of Taylor rule parameers; if here is lile variaion in macroeconomic condiions, esimaes of he responses o hose condiions are likely o be imprecise. Wih he arrival of a global recession in 2007, his obsacle largely disappeared; large cyclical flucuaions produced noable policy responses from many cenral banks. For several reasons, Sweden provides an ineresing case for esimaing a Taylor rule in he pos-recession environmen. Firs, Sweden experienced a sharp recession, bu i recovered more quickly han many oher counries. We herefore observe policymaker responses boh o he downurn and o he subsequen recovery. Second, alhough Sweden flired wih he lower bound on ineres raes, i has subsequenly reurned o ineres rae levels ha exceed he bound. In our work, we employ dynamic Tobi specificaions ha are appropriae for analyzing policymaking in he presence of a lower bound on ineres raes. Third, alhough Sweden implemened some policy acions ha migh be characerized as quaniaive easing, hose acions did no consiue a comprehensive subsiue for radiional ineres rae policies. Fourh, he Riksbank is among he mos ransparen of cenral banks, and i has published imely accouns of he deliberaions of is moneary policy commiee (MPC). Specifically, discussions of he lower bound on ineres raes are described in deail in meeing minues. 3

4 II. The Taylor Rule The Taylor rule is convenionally wrien as i r y y e, (1) * * y where i is a seleced arge ineres rae, is he rae of inflaion, * r is he real rae of * ineres, is he arge rae of inflaion, y is he log of real GDP, and y is he log of poenial real GDP. The difference, y y, is usually referred o as he oupu gap. As originally proposed, his was a prescripive rule, wih he following parameer values specified: * r 2, 2, 0.5, and 0.5. * y In empirical applicaions, Taylor rule parameer values are esimaed raher han prescribed, and he specificaion is usually alered o permi ineria in policymaking. There have been variaions in he measuremen of he oupu gap and inflaion variables. For example, one can use curren or lagged measures, or forecass of fuure values. Mos sudies have concluded ha he Taylor rule is a useful framework for analyzing Federal Reserve decision-making during he Grea Moderaion and even over longer horizons (Clarida, Gali, and Gerler 1999, 2000; Judd and Rudebusch 1998; Orphanides 2001, 2003, 2004). III. A Criique of Economeric Taylor Rules From he mid-1980s unil 2007, he period of he Grea Moderaion, business cycles in he Unied Saes and much of he world were subdued. GDP growh raes were less volaile han in earlier periods, and inflaion raes were low and seady. Some have argued ha appropriae moneary policy choices, perhaps choices ha were broadly 4

5 consisen wih Taylor rule prescripions, were parly responsible for hese desirable oucomes. However, if policymakers are successful in suppressing flucuaions, ha success may make i more difficul o deermine he parameers of empirical Taylor rules. Consider he exreme case where cycles are compleely eliminaed and inflaion is seady. The explanaory variables in he Taylor rule regression specificaion become consans, and heir coefficiens canno be idenified. Some have argued ha he macroeconomic variables in he Taylor rule should be forecass raher han acual values. However, a cenral bank ha successfully arges inflaion will insure ha, over a sufficienly long period, expeced inflaion will be equal o he inflaion arge. Replacing acual inflaion wih expeced inflaion in an empirical Taylor rule could herefore exacerbae he idenificaion problem even if acual inflaion flucuaes, expeced fuure inflaion will be more sable, and is coefficien will be difficul o esimae precisely. 1 If observed values of inflaion are used in he Taylor rule, esimaion is sill problemaic. Suppose ha underlying inflaion rends are low and seady. Furher, suppose ha a one-ime shock affecs he price level and he curren measured inflaion rae. If underlying inflaion remains low and seady, a good argumen can be made for ignoring he shock for purposes of moneary policymaking. The implicaion is ha in a scenario when underlying inflaion is generally on arge, observed variaions in inflaion are precisely hose variaions ha should be ignored. Even hough policymakers would 1 Esimaes could be biased as well as imprecise. Consider a scenario where inflaionary pressures are srong. Even when he cenral bank reacs o figh inflaion, he wo-year forecas of average inflaion saring from he curren period will probably be above he arge by a small amoun. Tha is, even if inflaion will be back o a arge pace in wo years, above-arge inflaion will affec he average in he inervening wo-year period. This argumen suggess ha small deviaions of he wo-year inflaion forecas from is arge could be associaed wih large policy responses, and a large inflaion coefficien esimae could resul. This is because a small inflaion deviaion from arge is an indicaion of wha would have been a large deviaion, absen he endogenous response. 5

6 reac srongly o a persisen change in inflaion, esimaed Taylor rules migh indicae a mued response o observed inflaion. For much of he period we sudy, he Riksbank calculaed expeced fuure inflaion raes under he assumpion ha he repo rae, is policy arge, would be unchanged. Such a forecas is probably mos appropriae for inclusion in he Taylor rule i measures inflaion pressures ha are purged of an expeced endogenous policy response. IV. The Zero Lower Bound on Nominal Ineres Raes The recession ha began in 2007 produced large drops in real GDP in he Unied Saes and oher counries. Inflaion raes also fell, alhough less dramaically. In response, cenral banks reduced heir ineres rae arges. In he Unied Saes, he federal funds rae fell from 5.25% in Sepember 2007 o near-zero in December The rae has remained a is effecive lower bound since hen. Given he impossibiliy of lowering raes furher, he Taylor rule has become largely irrelevan as a descripion of U.S. moneary policy since ha ime. Evens have unfolded somewha differenly in Sweden. Oupu began o plunge in lae 2008, he GDP gap reached -7.5% in Sepember 2009, and he arge repo rae fell o 0.25%. Alhough some members of he Riksbank s MPC argued for a furher reducion o zero, he majoriy of he commiee believed ha 0.25% was an effecive lower bound. For economeric purposes, we will repor esimaions ha assume a lower bound on he repo rae a ha level. 6

7 Afer Sepember 2009, he economy rebounded, and real GDP grew a a rae exceeding 7% in calendar year The repo rae was adjused upward beginning in June 2010, and reached a level of 1.75% by April Alhough Sweden underook measures ha migh be described as quaniaive easing during he recession, he repo rae remained he primary ool for moneary policy. Because of he rapid recession and recovery, and because complicaions associaed wih quaniaive easing were minimal, esimaion of a Taylor rule for he Riksbank using daa from he recen recession should be feasible and informaive. V. Economeric Issues The Taylor rule can be regarded as an empirical model for describing a cenral bank s moneary policy choices. In an empirical conex, equaion (1) is viewed as a regression specificaion in which a cenral bank s ineres rae arge is he dependen variable, and prevailing inflaion and oupu gap measures are independen variables. Empirical Taylor rules normally include lagged dependen variables o accoun for ineria and o lessen problems wih serial correlaion (Orphanides 2004). For he Riksbank, we wrie our base empirical Taylor rule specificaion as * i i i y y e. (2) y The dependen variable is a shor-erm ineres rae, he repo rae. As a measure of inflaion, we employ he inflaion forecas ha is developed by he Riksbank saff and presened o he MPC. This is a wo-year-ahead forecas of annual inflaion according o he consumer price index. The inflaion forecasing mehod used by he Riksbank changed several imes during he period we sudy. We will address complicaions 7

8 resuling from hese changes in secion VII. The oupu gap used in esimaion is published by Sweden s Naional Insiue of Economic Research (NIER). The original gap daa were quarerly, bu we have inerpolaed o obain oupu gap esimaes a commiee meeing daes. 2 Our sample of Riksbank daa begins wih he May 2000 MPC meeing and exends o he April 2011 meeing. In ha period, he MPC me 82 imes. In five of he meeings, from July 2009 hrough April 2010, he repo rae was se a he boundary level of 0.25%. In he presence of a lower bound on he dependen variable, he appropriae specificaion is a Tobi model, as described in (3): wih i i i y y e, (3) * * y i i if i * 0.25 * and i 0.25% if i * The dependen variable in his specificaion is a desired ineres rae, i *. The desired rae is no bounded; he MPC could have a desired ineres rae ha is negaive, even hough he acual rae canno be. The model specifies ha when i * is below is lower limi of 0.25%, i is unobserved, and he acual ineres rae, i, akes he boundary value of 0.25%. Noe ha in (3) we have specified ha he lagged ineres raes on he righ-hand side are he acual observed raes. We refer o his as he Tobi I specificaion. 2 Jansson and Vredin (2003) and Berg, Jansson, and Vredin (2006) have previously analyzed Swedish moneary policy experience using esimaed Taylor rules. 8

9 I is plausible ha lagged values of he underlying desired ineres raes (no acual ineres raes) should appear on he righ-hand side, as in (4): i i i y y e. (4) * * * * y We refer o his as he Tobi II specificaion. In his model, unobserved variables appear on boh sides of equaion (4), making he esimaion problem more difficul. To esimae his model, we use he maximum simulaed likelihood (MSL) mehod. For each observaion in he sample, any missing values of he unobserved righ-hand side variables are simulaed repeaedly according o he process specified o deermine hem. The likelihood funcion is hen calculaed for each simulaed observaion, and he likelihood for an observaion is calculaed as he average over he simulaed values. See Chang (2011) for deails abou various simulaion mehods for esimaing dynamic Tobi models. VI. Model Esimaes We iniially esimae he Taylor rule model for daa spanning he period preceding he recession in Sweden, i.e., for he period from May 2000 o December We hen esimae he model for he period including he recession and recovery, covering meeings from February 2008 hrough April The laer period includes he meeings in which he ineres rae encounered a presumed lower bound a 0.25%, so ha he Tobi specificaion is appropriae. For purposes of comparison, we esimae via boh OLS and Tobi for he laer period. Two versions of he Tobi model are employed, as described by (3) and (4). In he firs version, observed lagged values of he ineres rae appear on 9

10 he righ-hand side; in he second version, lagged values of he model s unobserved desired raes appear. Resuls are provided in Table 1. The resuls show some noiceable differences across samples, esimaion mehods, and specificaions. For he pre-recession sample, OLS esimaes show significan posiive coefficiens for boh he oupu gap and expeced inflaion. The coefficiens on hese variables are small; however, he sum of he lagged ineres rae coefficiens is high, a These resuls indicae a slow, inerial response of he repo rae o changes in economic condiions. For he recession sample, he gap variable remains significan and posiive in all esimaions. The inflaion variable remains marginally significan only in he OLS resuls. In he recession sample, and especially in he Tobi specificaions, inerial effecs are considerably weaker he sum of he coefficiens of he lagged dependen variables is in he Tobi I specificaion and in he Tobi II specificaion. Wih a larger gap coefficien and less ineria, he esimaed reacion o he recession appears o be much sronger han would have been prediced by he Taylor rule esimaed wih pre-recession daa. Oher resuls are similar for he wo versions of he Tobi model; sandard errors for mos coefficiens are smaller for Tobi I, he specificaion including observable lagged ineres raes, bu qualiaive resuls are similar, as are overall fis (he log-likelihoods are for Tobi I and for Tobi II). VII. Expeced Inflaion in he Taylor Rule Unil Ocober 2005, he Riksbank calculaed inflaion forecass under he assumpion ha he curren repo rae would remain unchanged over he forecas horizon. 10

11 However, from Ocober 2005 onward, differen forecasing assumpions were employed. In Ocober 2005, he inflaion forecas in he Moneary Policy Repor was calculaed under he assumpion ha he repo rae would develop in line wih expecaions in he financial markes as hey are expressed in implied forward raes. 3 Beginning wih he February 2007 meeing, he forecas was insead based on he ineres rae pah ha he Riksbank currenly considers will provide a well-balanced moneary policy. 4 In December 2010, he forecasing assumpion changed again. This ime he forecas was based on he pah of ineres raes ha would gain he suppor of a majoriy of he members of he Execuive Board. 5 The discussion above implies ha four disinc mehods were used in consrucing he Riksbank s expeced inflaion measure over our sample period. For our analysis, we will assume ha here have really been jus wo regimes: he consan repo rae regime and hree essenially equivalen endogenous repo rae regimes. If a majoriy of he MPC favored wha is also a well-balanced policy, and if markes anicipaed such a policy, hen he hree regimes would be equivalen. There is no indicaion given in minues or Moneary Policy Repors ha any subsanive differences were expeced under he hree endogenous rae regimes. Furher, he hree regimes were each in effec for shor periods, so disinguishing hem empirically is no likely o be pracical. The forecass ha assume endogenous ineres rae movemens are subjec o he problem idenified in our Secion III criique. If ineres raes are adjused in an 3 Source: Separae Minues of he Execuive Board meeing for Ocober 19, Source: Moneary Policy Repor, 2007:1. 5 Source: Separae Minues of he Execuive Board meeing for December 14,

12 appropriaely balanced fashion, hen he Riksbank should come close o hiing is inflaion arge, and he forecas of inflaion should rouinely be close o he arge value. To invesigae wheher he Riksbank s changing forecasing mehods affecs Taylor rule esimaes, we have alered our models o permi parameer shifs relaed o he forecasing regime. To do so, we define a dummy variable, DNEW, o indicae meeings held afer Augus This corresponds o he regime under which ineres raes were assumed o be adjused endogenously. We also creae an ineracion of he dummy variable wih he inflaion forecas,. Table 2 repors esimaes of Taylor rules for he pre-recession period ha add hese wo variables o he original specificaion. The dummy variable iself has an esimaed coefficien ha does no differ significanly from zero. However, he ineracion of he dummy variable wih expeced inflaion is negaively signed and significanly differen from zero. The resuls sugges ha he coefficien of he inflaion forecas goes o zero when he forecas is made endogenous (we are unable o rejec he hypohesis ha he coefficiens of and DNEW * sum o zero). Therefore, in he prerecession period, he change in forecasing mehod appears o explain he disappearance of an expeced inflaion effec on he repo rae. This also suggess ha he absence of an esimaed inflaion effec in he recession period could be a consequence of he forecasing mehod raher han a shif in he behavior of he MPC. 12

13 VIII. Conclusions and Discussion We have esimaed Taylor rules for Sweden s Riksbank over a period exending from 2000 o This period is ineresing because i includes he lae years of he Grea Moderaion, a severe recession, and a robus recovery. Our resuls show ha he policy response o he recessionary oupu gap was sronger and faser han would have been prediced from Taylor rules esimaed using only pre-recession daa poins. Policy appeared o be highly inerial before he recession, bu much less so aferward. Our esimaes also sugges ha he repo rae was adjused in response o expeced inflaion in he pre-recession period, bu no in he recession period. This may be a consequence of a change in he mehod for forecasing inflaion raher han a fundamenal change in policy behavior. In fuure work, we plan o invesigae his issue more carefully. We will also invesigae alernaive Taylor rule specificaions, including hose proposed by Driffill and Roondi (2007). Driffill and Roondi have suggesed ha much of apparen policy ineria is driven by ineria in macroeconomic condiions, raher han he policymaking process per se. The preliminary resuls repored here offer some suppor for his conjecure. 13

14 Table 1. Taylor Rule Esimaes: Riksbank OLS: May 2000 o December 2007 Variable Coefficien Sandard Error -saisic P-value Consan [.428] i [.000] i [.440] y y [.001] [.019] OLS: February 2008 hrough April 2011 Variable Coefficien Sandard Error -saisic P-value Consan [.007] i [.000] i [.125] y y [.001] [.078] Tobi I: February 2008 hrough April 2011 Parameer Esimae Sandard Error -saisic P-value Consan [.005] i [.109] i [.786] y y [.000] [.416] Tobi II: February 2008 hrough April 2011 Parameer Esimae Sandard Error -saisic P-value Consan [.094] * i [.583] * i [.735] y y [.035] [.612] 14

15 Table 2. Taylor Rule Esimaes under Alernaive Inflaion Forecass: Riksbank OLS: May 2000 o December 2007 Variable Coefficien Sandard Error -saisic P-value Consan [.939] i [.000] i [.985] y y [.078] [.003] DNEW [.145] DNEW* [.043] 15

16 References Berg, Claes, Per Jansson, and Anders Vredin How Useful are Simple Rules for Moneary Policy? The Swedish Experience. Sveriges Riksbank Research Paper Series No. 12. Chang, Sheng-Kai Simulaion Esimaion of Dynamic Panel Tobi Models. Deparmen of Economics, Wayne Sae Universiy, working paper. Clarida, Richard, Jordi Gali, and Mark Gerler The Science of Moneary Policy: A New Keynesian Perspecive. Journal of Economic Lieraure 37(4): Clarida, Richard, Jordi Gali, and Mark Gerler Moneary Policy Rules and Macroeconomic Sabiliy: Evidence and Some Theory. Quarerly Journal of Economics 115(1): Driffill, John, and Zeno Roondi Ineria in Taylor Rules. Birkbeck Working Papers in Economics and Finance, Birkbeck, Universiy of London, BWPEF Jansson, Per, and Anders Vredin Forecas-Based Moneary Policy: The Case of Sweden. Inernaional Finance 6(3): Judd, John P., and Glenn D. Rudebusch Taylor s Rule and he Fed: Federal Reserve Bank of San Francisco Economic Review 1998(3): Orphanides, Ahanasios Moneary Policy Rules Based on Real-Time Daa. American Economic Review 91(4): Orphanides, Ahanasios Hisorical Moneary Policy Analysis and he Taylor Rule. Journal of Moneary Economics 50(5): Orphanides, Ahanasios Moneary Policy Rules, Macroeconomic Sabiliy, and Inflaion: A View from he Trenches. Journal of Money, Credi, and Banking 36(2): Taylor, John B Discreion versus Policy Rules in Pracice. Carnegie-Rocheser Conference Series on Public Policy 39 (December):