Thesis
Reference
Essays on Central Bank credibility and monetary policy predictability
MARKOV, Nikolay
Abstract
This thesis contributes to the literature through an empirical analysis of the first decade of single European monetary policy within a real-time approach and offers new specifications of the policy reaction function of Central Banks based on a novel econometric modeling technique. The results of the thesis show that the policy interest rate of the European Central Bank (ECB) has been broadly well predicted by the market participants. Second, there is evidence that the ECB's monetary policy has switched between a normal regime where the Central Bank focuses on its price stability commitment and a crisis regime where it puts a higher emphasis on stabilizing the economic outlook. Finally, the semi-parametric specifications of the policy reaction function permit to better predict the actual policy rate of several OECD Central Banks than the standard linear Taylor Rule and the policy rule with interest rate smoothing in the medium and longer terms.
MARKOV, Nikolay. Essays on Central Bank credibility and monetary policy predictability. Thèse de doctorat : Univ. Genève, 2012, no. SES 782
URN : urn:nbn:ch:unige-227848
DOI : 10.13097/archive-ouverte/unige:22784
Available at:
http://archive-ouverte.unige.ch/unige:22784
Disclaimer: layout of this document may differ from the published version.
1 / 1
Monetary Policy Predictability and
Thèse présentée à la Faculté des
sciences économiques et sociales de l'Université de Genève
Nikolay Markovpar
pour l'obtention du grade de
Docteur ès sciences économiques et sociales mention économie politique
Membres du jury de thèse:
Prof. Henri LOUBERGÉ, co-directeur de thèse, Université de Genève Prof. Charles WYPLOSZ, co-directeur de thèse, IHEID
Prof. Jaya KRISHNAKUMAR, présidente du jury, Université de Genève Prof. Jordi GALÍ, Universitat Pompeu Fabra de Barcelone
Prof. Ulrich KOHLI, Université de Genève
Thèse no 782
Genève, le 26 juin 2012
La Faculté des sciences économiques et sociales, sur préavis du jury, a autorisé l'impression de la présente thèse, sans entendre, par là, n'émettre aucune opinion sur les propositions qui s'y trouvent énoncées et qui n'engagent que la responsabilité de leur auteur.
Genève, le 26 juin 2012
Le doyen
Bernard MORARD
Impression d'après le manuscrit de l'auteur.
Résumé xi
Abstract xiii
Acknowledgments xv
General Introduction 1
1 Actual versus Perceived Taylor Rules 5
2 A Regime Switching Model for the ECB 59
3 Is the Taylor Rule Nonlinear? 117
General Conclusion 181
Bibliography 186
iii
Résumé xi
Abstract xiii
Acknowledgments xv
General Introduction 1
1 Motivation . . . 1
2 Overview . . . 2
1 Actual versus Perceived Taylor Rules 5 1 Introduction . . . 5
2 Theoretical framework . . . 8
3 Data and methodology . . . 10
4 Linear model estimations . . . 15
4.1 Estimations for the next calendar quarter and year horizons . . . 15
4.2 Estimations for the one-quarter and one-year horizons . . . 18
4.3 Actual and Perceived tted policy rules . . . 21
5 Rolling window estimations . . . 24
5.1 Coecient estimates of policy inertia . . . 25
5.2 Coecient estimates of expected ination . . . 26
5.3 Coecient estimates of expected GDP growth . . . 28
6 Estimation with time dummies . . . 30
7 Robustness analysis . . . 33
8 Conclusion . . . 36
9 Appendix . . . 38
9.1 Variables used in the estimations . . . 38
9.2 Unit root and stationarity tests . . . 40
9.3 Actual and Perceived Taylor Rules with a dummy variable for the ECB meetings . . . 42
9.4 Actual and Perceived Taylor Rules with Consensus Economics Fore- casts . . . 43
9.5 Actual and Perceived Taylor Rules with the Economic Sentiment Indicator . . . 44
9.6 Actual and Perceived Taylor Rules with Consensus ination forecasts and the Economic Sentiment Indicator . . . 45
9.7 Predicted re rate target, Investment Bank Forecasts . . . 46
9.8 Predicted re rate target, Consensus Economics Forecasts . . . 48
9.9 Rolling window estimates, one-year horizon . . . 50
9.10 Recursive window estimates, next calendar year horizon . . . 53
9.11 Recursive window estimates, one-year horizon . . . 56 v
2 A Regime Switching Model for the ECB 59
1 Introduction . . . 59
2 The Model . . . 61
3 Data and methodology . . . 63
4 Empirical evidence . . . 67
4.1 Actual regime switching Taylor Rule . . . 67
4.2 Perceived regime switching Taylor Rule . . . 71
4.3 Actual and Perceived ltered probabilities . . . 74
4.4 Actual and Perceived tted policy rates . . . 75
5 Robustness analysis . . . 80
5.1 Actual and Perceived Taylor Rules with consensus forecasts . . . 80
5.2 Actual and Perceived Taylor Rules with an alternative forecast horizon 84 5.3 Augmented Taylor Rules . . . 90
6 The Taylor Principle and determinacy . . . 94
7 Threshold regression analysis . . . 96
8 A three regimes switching model . . . 98
9 Conclusion . . . 103
10 Appendix . . . 105
10.1 Variables used in the estimations . . . 105
10.2 Unit root and stationarity tests . . . 107
10.3 Regime switching tests . . . 108
10.4 Fitted policy rates of the baseline model . . . 109
10.5 Augmented Taylor Rules with a one-year forecast horizon . . . 110
10.6 Threshold regression results . . . 112
10.7 A three regimes switching model with a one-year forecast horizon . . 114
3 Is the Taylor Rule Nonlinear? 117 1 Introduction . . . 117
2 Theoretical framework . . . 119
3 Data and methodology . . . 121
4 Empirical evidence . . . 122
4.1 Bivariate GAM . . . 123
4.2 Univariate GAM without interaction terms . . . 124
4.3 Univariate GAM with interaction terms . . . 134
5 Out-of-sample forecasting . . . 138
6 Robustness analysis . . . 141
7 Estimations with the consensus forecasts . . . 142
8 Simulation results with dierent forecast horizons . . . 144
9 Conclusion . . . 145
10 Appendix . . . 147
10.1 Data description . . . 147
10.2 Estimation tables, OECD forecasts . . . 152
10.3 Estimation tables, Consensus Economics Forecasts . . . 154
10.4 Simulation results with dierent forecast horizons . . . 156
10.5 Figures . . . 160
General Conclusion 181 1 Main ndings . . . 181
2 Policy issues . . . 182
3 Future research . . . 184
Bibliography 186
Chapter 1
1 The timing of the model . . . 10
2 The predictability of ECB's main policy rate by the professional forecasters 12 3 Actual and Perceived Taylor Rules, next calendar quarter forecasts . . . 22
4 Actual and Perceived Taylor Rules, next calendar year forecasts . . . 23
5 Policy inertia estimates, next calendar year forecasts (rolling) . . . 25
6 Ination coecient estimates, next calendar year forecasts (rolling) . . . 27
7 Output growth coecient estimates, next calendar year forecasts (rolling) . 29 A.1 Actual and Perceived Taylor Rules, one-quarter forecasts . . . 46
A.2 Actual and Perceived Taylor Rules, one-year forecasts . . . 47
A.3 Actual and Perceived Taylor Rules, next calendar year consensus forecasts . 48 A.4 Actual and Perceived Taylor Rules, one-year consensus forecasts . . . 49
A.5 Policy inertia estimates, one-year forecasts (rolling) . . . 50
A.6 Ination coecient estimates, one-year forecasts (rolling) . . . 51
A.7 Output growth coecient estimates, one-year forecasts (rolling) . . . 52
A.8 Policy inertia estimates, next calendar year forecasts (recursive) . . . 53
A.9 Ination coecient estimates, next calendar year forecasts (recursive) . . . . 54
A.10 Output growth coecient estimates, next calendar year forecasts (recursive) 55 A.11 Policy inertia estimates, one-year forecasts (recursive) . . . 56
A.12 Ination coecient estimates, one-year forecasts (recursive) . . . 57
A.13 Output growth coecient estimates, one-year forecasts (recursive) . . . 58
Chapter 2 1 Actual ltered probabilities, baseline model . . . 70
2 Perceived ltered probabilities, baseline model . . . 73
3 Actual and perceived ltered probabilities, baseline model . . . 74
4 Actual and perceived tted target rates, baseline model . . . 77
5 Ination and real GDP growth forecasts, baseline model . . . 77
6 Actual and perceived ltered probabilities, consensus forecasts . . . 84
7 Actual and perceived ltered probabilities, alternative model . . . 89
8 Actual and perceived ltered probabilities, augmented model . . . 92
9 Actual and perceived ltered probabilities, three regimes model . . . 101
A.1 MRS and linear model tted rates, ATR baseline model . . . 109
A.2 MRS and linear model tted rates, PTR baseline model . . . 109
A.3 Actual and perceived ltered probabilities, augmented model (one-year) . . 111
A.4 Actual and perceived ltered probabilities, three regimes model (one-year) . 116 Chapter 3 1 Estimated smooth functions for Australia . . . 125
2 Estimated smooth functions for Canada . . . 126
3 Estimated smooth functions for Japan . . . 127
4 Estimated smooth functions for New Zealand . . . 128 vii
5 Estimated smooth functions for Norway . . . 129
6 Estimated smooth functions for Sweden . . . 130
7 Estimated smooth functions for Switzerland . . . 132
8 Estimated smooth functions for USA . . . 133
9 Estimated Surface of the Policy Rate for Australia, output growth . . . 160
10 Estimated Surface of the Policy Rate for Canada, output growth . . . 160
11 Estimated Surface of the Policy Rate for Japan, output growth . . . 160
12 Estimated Surface of the Policy Rate for New Zealand, output growth . . . 160
13 Estimated Surface of the Policy Rate for Norway, output growth . . . 161
14 Estimated Surface of the Policy Rate for Sweden, output growth . . . 161
15 Estimated Surface of the Policy Rate for Switzerland, output growth . . . . 161
16 Estimated Surface of the Policy Rate for USA, output growth . . . 161
17 Estimated Surface of the Policy Rate for Australia, output gap change . . . 162
18 Estimated Surface of the Policy Rate for Canada, output gap change . . . . 162
19 Estimated Surface of the Policy Rate for Japan, output gap change . . . 162
20 Estimated Surface of the Policy Rate for New Zealand, output gap change . 162 21 Estimated Surface of the Policy Rate for Norway, output gap change . . . . 163
22 Estimated Surface of the Policy Rate for Sweden, output gap change . . . . 163
23 Estimated Surface of the Policy Rate for Switzerland, output gap change . . 163
24 Estimated Surface of the Policy Rate for USA, output gap change . . . 163
25 Estimated Surface of the Policy Rate for Canada, CEF output growth . . . 164
26 Estimated Surface of the Policy Rate for Japan, CEF output growth . . . . 164
27 Estimated Surface of the Policy Rate for Norway, CEF output growth . . . 164
28 Estimated Surface of the Policy Rate for Sweden, CEF output growth . . . 164
29 Estimated Surface of the Policy Rate for Switzerland, CEF output growth . 165 30 Estimated Surface of the Policy Rate for USA, CEF output growth . . . 165
31 Estimated Surface of the Policy Rate for Canada, CEF output gap change . 165 32 Estimated Surface of the Policy Rate for Japan, CEF output gap change . . 165
33 Estimated Surface of the Policy Rate for Norway, CEF output gap change . 166 34 Estimated Surface of the Policy Rate for Sweden, CEF output gap change . 166 35 Estimated Surface of the Policy Rate for Switzerland, CEF output gap change166 36 Estimated Surface of the Policy Rate for USA, CEF output gap change . . . 166
37 Regression plot with ination, OLS method . . . 167
38 Regression plot with output growth, OLS method . . . 167
39 Fitted values against Actual Policy Rate . . . 168
40 Out-of-sample forecasts of the policy rate in Australia . . . 172
41 Out-of-sample forecasts of the policy rate in Canada . . . 173
42 Out-of-sample forecasts of the policy rate in Japan . . . 174
43 Out-of-sample forecasts of the policy rate in New Zealand . . . 175
44 Out-of-sample forecasts of the policy rate in Norway . . . 176
45 Out-of-sample forecasts of the policy rate in Sweden . . . 177
46 Out-of-sample forecasts of the policy rate in Switzerland . . . 178
47 Out-of-sample forecasts of the policy rate in USA . . . 179
Chapter 1
1 The re rate adjustments . . . 12
2 Actual Taylor Rule, 2000-2009 . . . 16
3 Perceived Taylor Rule, 2000-2009 . . . 17
4 Actual Taylor Rule, 2000-2009 alternative model . . . 19
5 Perceived Taylor Rule, 2000-2009 alternative model . . . 20
6 Actual Taylor Rule, 2000-2009 crisis dummies . . . 31
7 Perceived Taylor Rule, 2000-2009 crisis dummies . . . 32
A.1 List of variables . . . 38
A.2 Summary statistics . . . 39
A.3 Unit root tests, Investment Bank Forecasts . . . 40
A.4 Unit root tests, Consensus Economics Forecasts . . . 41
A.5 Actual Taylor Rule, 2000-2009 meeting dummy . . . 42
A.6 Perceived Taylor Rule, 2000-2009 meeting dummy . . . 42
A.7 Actual Taylor Rule, 2000-2009 consensus forecasts . . . 43
A.8 Perceived Taylor Rule, 2000-2009 consensus forecasts . . . 43
A.9 Actual Taylor Rule, 2000-2009 ESI . . . 44
A.10 Perceived Taylor Rule, 2000-2009 ESI . . . 44
A.11 Actual Taylor Rule, 2000-2009 consensus forecasts and ESI . . . 45
A.12 Perceived Taylor Rule, 2000-2009 consensus forecasts and ESI . . . 45
Chapter 2 1 Actual MRS Taylor Rule, Baseline Model (BM) . . . 67
2 Actual MRS Taylor Rule, BM long-run parameters . . . 68
3 Perceived MRS TR, Baseline Model (BM) . . . 71
4 Perceived MRS Taylor Rule, BM long-run parameters . . . 72
5 Actual MRS Taylor Rule, CEF . . . 80
6 Actual MRS Taylor Rule, CEF long-run parameters . . . 81
7 Perceived MRS Taylor Rule, CEF . . . 82
8 Perceived MRS Taylor Rule, CEF long-run parameters . . . 83
9 Actual MRS Taylor Rule, Alternative Model (AM) . . . 85
10 Actual MRS Taylor Rule, AM long-run parameters . . . 86
11 Perceived MRS Taylor Rule, Alternative Model (AM) . . . 87
12 Perceived MRS Taylor Rule, AM long-run parameters . . . 87
13 Augmented Taylor Rules, linear model . . . 90
14 Actual and Perceived MRS Taylor Rules, augmented model . . . 91
15 Determinacy condition for equilibrium . . . 96
16 Actual MRS Taylor Rule, Three Regimes Model (TRM) . . . 99
17 Actual MRS Taylor Rule, TRM long-run parameters . . . 99
18 Perceived MRS Taylor Rule, Three Regimes Model (TRM) . . . 100
19 Perceived MRS Taylor Rule, TRM long-run parameters . . . 100 ix
A.1 List of variables . . . 105
A.2 Summary statistics . . . 106
A.3 Unit root tests, Investment Bank Forecasts . . . 107
A.4 Unit root tests, Consensus Economics Forecasts . . . 107
A.5 Regime switching tests . . . 108
A.6 Augmented Taylor Rules, linear model (one-year) . . . 110
A.7 Actual and Perceived MRS Taylor Rules, augmented model (one-year) . . . 110
A.8 Threshold regression models, threshold variable time trend, Actual Taylor Rule . . . 112
A.9 Threshold regression models, threshold variable GDP growth forecast, Ac- tual Taylor Rule . . . 112
A.10 Threshold regression models, threshold variable time trend, Perceived Taylor Rule . . . 113
A.11 Threshold regression models, threshold variable GDP growth forecast, Per- ceived Taylor Rule . . . 113
A.12 Actual MRS TR, Three Regimes Model (TRM, one-year) . . . 114
A.13 Actual MRS TR, TRM long-run parameters (one-year) . . . 114
A.14 Perceived MRS TR, Three Regimes Model (TRM, one-year) . . . 115
A.15 Perceived MRS TR, TRM long-run parameters (one-year) . . . 115
Chapter 3 1 Summary statistics, OECD forecasts . . . 148
2 Unit root tests, OECD forecasts . . . 149
3 Summary statistics, Consensus forecasts . . . 150
4 Unit root tests, Consensus forecasts . . . 151
5 Bivariate GAM, OECD output growth . . . 152
6 Univariate GAM with no interactions, OECD output growth . . . 152
7 Univariate GAM with interactions, OECD output growth . . . 152
8 Root Forecast Mean Squared Error, OECD output growth . . . 152
9 Bivariate GAM, OECD output gap change . . . 153
10 Univariate GAM with no interactions, OECD output gap change . . . 153
11 Univariate GAM with interactions, OECD output gap change . . . 153
12 Root Forecast Mean Squared Error, OECD output gap change . . . 153
13 Bivariate GAM, CEF output growth . . . 154
14 Univariate GAM with no interactions, CEF output growth . . . 154
15 Univariate GAM with interactions, CEF output growth . . . 154
16 Root Forecast Mean Squared Error, CEF output growth . . . 154
17 Bivariate GAM, CEF output gap change . . . 155
18 Univariate GAM with no interactions, CEF output gap change . . . 155
19 Univariate GAM with interactions, CEF output gap change . . . 155
20 Root Forecast Mean Squared Error, CEF output gap change . . . 155
21 RFMSE with dierent horizons, OECD output growth . . . 156
22 RFMSE with dierent horizons, OECD output gap change . . . 157
23 RFMSE with dierent horizons, CEF output growth . . . 158
24 RFMSE with dierent horizons, CEF output gap change . . . 159
La crédibilité d'une banque centrale et la prévisibilité de ses décisions de taux d'intérêt sont deux éléments clés pour renforcer l'ecacité de la politique monétaire dans un envi- ronnement qui est orienté vers le futur.
Cette thèse contribue à la littérature premièrement à travers une analyse empirique de la première décennie de politique monétaire Européenne unique dans une approche en temps réel. En s'appuyant sur une nouvelle base de données, elle analyse la prévisibilité du taux d'intérêt directeur sous le prisme des participants au marché nancier. Deuxièmement, la thèse apporte un éclairage nouveau sur la nonlinéarité de la règle de taux d'intérêt de la Banque Centrale Européenne (BCE) dans le cadre d'un modèle avec changement de régime.
Finalement, en suivant une approche plus générale, elle propose de nouvelles spécications de la fonction de réaction des banques centrales qui se basent sur une nouvelle technique de modélisation économétrique.
Les résultats empiriques de la thèse montrent que le taux d'intérêt de la Banque Cen- trale Européenne a été globalement bien prédit par les participants au marché. Toutefois, ces-derniers ont prévu de manière plus précise les hausses de taux plutôt que les baisses du taux d'intérêt directeur. Deuxièmement, la politique monétaire de la BCE a changé entre un régime normal dans lequel la banque centrale se focalise sur son engagement de maintien de la stabilité des prix et un deuxième régime dans lequel elle met davantage l'accent sur la stabilisation de la conjoncture. Finalement, les spécications semi-paramétriques de la fonction de réaction permettent de mieux prédire le taux d'intérêt directeur de plusieurs banques centrales de l'OCDE que la règle de Taylor linéaire standard et la règle avec lissage du taux d'intérêt à moyen et à long termes.
xi
The credibility of a Central Bank and the predictability of its interest rate decisions are two key issues for enhancing the eectiveness of monetary policy within a forward-looking environment.
This thesis contributes to the literature rst through an empirical analysis of the rst decade of single European monetary policy within a real-time approach. Building on a new data set, it investigates the predictability of the main policy rate through the eyes of nancial market participants. Second, the thesis sheds new light on the nonlinearity of the policy rule of the European Central Bank (ECB) based on a regime switching framework.
Finally, following a more general approach, it provides new specications of the policy reaction function of Central Banks based on a novel econometric modeling technique.
The empirical results of the thesis show that the policy interest rate of the European Central Bank has been broadly well predicted by the market participants. Nevertheless, the latter have more accurately foreseen the policy rate hikes than the interest rate cuts.
Second, there is evidence that the ECB's monetary policy has switched between a normal regime where the Central Bank focuses on its price stability commitment and a crisis regime where it puts a higher emphasis on stabilizing the economic outlook. Finally, the semi-parametric specications of the policy reaction function permit to better predict the actual policy rate of several OECD Central Banks than the standard linear Taylor Rule and the policy rule with interest rate smoothing in the medium and longer terms.
xiii
In this section, I would like to express my profound gratitude and indebtedness to the people that have given to me their support and who have contributed to the achievement of this work. Their insightful comments and enlightened remarks have made possible the accomplishment of this thesis. I rst start with my thesis supervisors, Professor Henri Loubergé and Professor Charles Wyplosz. I would like to express my deep recognition to them for the complete condence they have had in me for carrying out this work and for their continuous support throughout the writing of the thesis. I am particularly indebted to them for the precious academic and professional advices they have given to me, which have helped me to properly execute the economic analysis and the statistical treatment of the data both in hard and in good times. Their illuminating views and encouragement have always kept me on the right track and have permitted to maintain a rm belief in the successful achievement of this endeavor.
I know Professor Loubergé since the beginning of my bachelor studies in the fall of 2002 from the course of Economie politique générale. I have also followed with great interest his lectures in the courses of Economie du risque et de l'incertitude and Theory of nancial markets which have been an important source of insightful thinking for my understanding of key concepts in economics. Professor Loubergé has also transmitted to me the quest for perfectionism and assiduity in the pursuit of my passion for monetary economics and in my commitment to the economics profession. I think that my desire to follow up the doctoral studies has paved its way one day in the spring of 2007 when I asked Professor Loubergé whether I could envisage to apply for a Ph.D position and he replied that it would be a good idea to enroll in the doctoral program. I would like to thank Professor Loubergé for his thorough availability, his invaluable consideration of my dissertation and his strong encouragement throughout the writing of the thesis. I am also indebted to him for having accepted me as a Ph.D candidate and as his Teaching Assistant in the Department of Economics. His remarkable ideas and shrewd comments, as well as his rened care in the execution of any academic work will always be of the utmost example for my future career and for my commitment to the economics profession. I am also extremely grateful to him for the personal and nancial support he has given to me for the thesis which has made possible the presentation of my work at international conferences.
I had the great pleasure to meet Professor Charles Wyplosz in the course of Analyse monétaire in the fall of 2004. I have also followed with great interest the Monetary theory and Open Economy Macroeconomics courses he has taught in the master's program. He has always been a deep source of breathtaking inspiration and has transmitted to me his extraordinary passion for monetary economics. I am extremely indebted to him for his thorough availability, amazing ideas and strong encouragement and support since the fulllment of my master thesis in June 2006 throughout the achievement of my doctoral dissertation. The master thesis is entitled "L'inconséquence intertemporelle des banques centrales: application à la politique monétaire de la Banque Nationale de Bulgarie" and has been realized under his supervision. A few days after I started working as an Assistant
xv
in the Department of Economics, in October 2007, I met Professor Wyplosz and shared with him the desire to continue working in the eld of monetary policy and to write a doctoral thesis as a natural progression of my master thesis. I am particularly grateful to Professor Charles Wyplosz who has taught me that the quality of an academic paper depends heavily on the relevance of the research question that one raises. I am indebted to Professor Wyplosz for having accepted me as a Ph.D candidate and as his Teaching Assistant, as well as for his precious help in nding the data necessary for carrying through the thesis. I am also grateful to him for his condence and trust in the accomplishment of this work. His captivating advices and acute comments have always been of the utmost importance for the proper execution of the dissertation and will remain an invaluable source of inspiration for my future career and for my commitment to the economics profession.
I know Professor Ulrich Kohli since 2010. I am particularly grateful to him for his invaluable comments and insights he has given to me from the perspective of a distinguished Central Banker. His precious advices have substantially improved the quality of my work.
I also would like to thank Professor Jaya Krishnakumar for her lectures in econometrics which have helped me to understand and implement the techniques necessary for the appropriate execution of the statistical analysis in the thesis. The strong quantitative methods I have learned from her have particularly improved the technical solidness of my work. I am also indebted to Professor Jordi Galí for his fascinating lectures on the science of monetary policy he has given in the doctoral program in Gerzensee, which have greatly inspired me and have further deepened my knowledge of modern monetary policy.
Furthermore, I am also grateful to the participants of the T2M Conference of Macroe- conomics in Le Mans 2010, the30th CIRET Conference on Economic Tendency Surveys in New York 2010 and the Young Researchers Seminars of the Department of Economics of the University of Geneva. In particular, I would like to thank Paul Beaudry, Arnaud Chéron, Christian Conrad, Jaime de Melo, Jean-Paul L'Huillier, Alessandro Missale, Michel Nor- mandin, Marcelo Olarreaga, Syed Rizvi, Frédéric Robert-Nicoud, Elvezio Ronchetti, Stefan Sperlich, Jürgen von Hagen.
I also would like to acknowledge the valuable comments and remarks I have received from the colleagues of the Department of Economics and Econometrics of the University of Geneva, which have improved upon my work and the quality of the thesis presentation.
In particular, I would like to thank William Aeberhard, Estefania Amer, Gerda Cabej, Giovanni Ferro Luzzi, Jean-Marc Natal, Cyril Pasche, Julie Regolo, Emmanuel Rousseaux, Gisèle Schmid, Mathias Thoenig, Sylvain Weber. Special thanks are due to Anjeza Kadilli for her precious help in Latex and for her careful proofreading of this work. I am also grateful to Carlos de Porres and Anjeza Kadilli for their enlightening remarks and critical thinking which have improved upon my research.
Finally, I feel strongly indebted to my parents, grandparents and relatives for their kind eorts in supporting me throughout my Bachelor, Master and Ph.D studies and who have always trusted in me and have ensured a favorable environment for the accomplishment of my academic work.
tice is as much art as science. Nonetheless, while practicing this dark art, I have always found the science quite useful. And I came to believe that the Federal Reserve and other central banks could prot from more disciplined and systematic think- ing . . . "
Alan S. Blinder (1997) Vice Chairman
Board of Governors of the Federal Reserve System, 1994-1996
1 Motivation
Central Bank credibility and monetary policy predictability are two key factors for en- hancing monetary policy eectiveness in a forward-looking environment of policy making.
During the last two decades, central banking has considerably evolved towards greater transparency and monetary policy decisions have been taken within more explicit and more clearly dened institutional frameworks. In the past, the strategy of Central Banks has often been guided by the theoretical underpinning that only unanticipated policy has real eects. The fact that monetary policy does not exert a systematic eect on real variables and therefore that Central Banks should abstain from producing surprises is pio- neered by the seminal work of Lucas (1972). Assuming rational expectations, the Central Bank should not try to fool the public but instead focus on the design of an explicit com- munication policy to ensure that its actions are well understood by the public. However, the lack of a formal communication strategy was a main feature of the usual secrecy and mystique that have surrounded this esoteric art as it is exposed in Brunner (1981) and is explained in Goodfriend (1985) and in Blinder (1998). This practice is clearly reected in the Chairman Alan Greenspan's statement made in 1987: "Since I've become a central banker, I've learned to mumble with great incoherence. If I seem unduly clear to you, you must have misunderstood what I said.". It is only quite recently that the communica- tion policy of the Federal Reserve and of other major Central Banks has geared towards a greater openness within appropriately designed policy frameworks. As most Central Banks have adopted ination targeting as a framework for their monetary policy strategy, this improved transparency has helped to credibilize the low ination commitment of these Central Banks over the policy relevant medium to long term horizons.
The era of Central Bank's openness has come up in parallel with the recent advances in the monetary economics literature as highlighted in the seminal paper of Clarida, Galí and Gertler (1999) as well as in Woodford (2003) and in Galí (2008). The New Neoclassical Synthesis has emerged as a new paradigm in monetary economics and has revived the non neutrality of money in the short term stemming from the presence of nominal rigidities.
This new strand of the literature postulates that within a forward looking environment, the equilibrium values of key macroeconomic aggregates depend heavily on expectations about the future policy actions of the Central Bank. In this framework, the credibility of the Central Bank's actions becomes essential and monetary policy operates mainly through agents' expectations of the future levels of macroeconomic fundamentals. In the current state of the art, most of the Central Banks control the short term nominal interest rate while the consumption and investment decisions depend on the current and on the future level of real interest rates. Therefore, the art of monetary policy lies in the Central Bank's ability to shape the expectations of the private agents on the way interest rates, prices and income are likely to evolve in the future as pointed out by Woodford (2003). An alignment of the private sector's expectations with the price stability objective announced by the monetary authority will foster the predictability of the Central Bank's decisions on the
1
future policy interest rates and will help to achieve a low ination credibility. Thereby, the more predictable and credible an independent Central Bank is the more eective will be monetary policy in stabilizing key macroeconomic aggregates and the smaller should be the disination costs. While central banking was once considered as more successful when maintaining an inherent secrecy and mystique about the policy strategy and actions, it has become the science of managing market expectations on key macroeconomic variables as highlighted by Woodford (2003). Hence, the science has gradually overtaken the practice of this mystical art which is intended for the initiated elite using the words of Brunner (1981).
Therefore, in theory and in practice, Central Bank's credibility is a crucial issue for enhancing monetary policy eectiveness. The latter requires a clear commitment of the monetary authority to maintain price stability as the ultimate policy goal and necessitates the adoption of an explicit strategy as a framework for monetary policy decisions. The rst main goal of the present thesis is to investigate the predictability of the key policy rate of the European Central Bank (ECB), the re rate, by the relevant market participants within a forward-looking environment and in a real-time setting. The focus on the ECB is motivated by the fact that it is a recently established Central Bank that is entrusted with the challenging task to conduct a single monetary policy within the euro area member states since1stJanuary 1999. The successful achievement of its price stability mandate de- pends crucially on the predictability of its policy actions and on its low ination credibility among the market participants. The second goal is to study more thoroughly the behavior of the ECB, and in particular to determine whether its responsiveness to macroeconomic fundamentals has changed along with the state of the economy within an appropriately designed regime switching model. The third core objective of the thesis is to challenge the standard linear Taylor Rule specication by adopting a more general approach to empir- ically model the policy reaction function. The aim is to better understand the behavior of Central Banks and to accurately forecast their main policy rate. The general nonlinear modeling of the policy rules is performed with a novel econometric technique in monetary economics.
2 Overview
The rst chapter of the thesis investigates the predictability of the main ECB's policy rate through the eyes of the professional forecasters from a large investment bank before each meeting of the Governing Council. The analysis is performed within forward-looking Ac- tual and Perceived Taylor Rules which are estimated in real-time with a newly constructed database over the period April 2000-November 2009. The former policy rule is based on the main renancing operations rate (the re rate) set by the ECB, while the latter is estimated with the re rate point forecast of the investment bank's economists one week ahead of the Governing Council meeting. The data are taken from the weekly economic projections reported by the economists from the bank. Using the forecasts for the next calendar quarter and the next calendar year, we examine to what extent the responsive- ness of the Central Bank to the macroeconomic projections changes along the forecasting horizon. Furthermore, we also investigate whether the estimated coecients of the policy rules change over time within some rolling window regressions. Given the magnitude of the nancial crisis in 2008, the responsiveness of the ECB to macroeconomic fundamentals might have been altered in that period. This analysis is important as the time-varying Taylor Rules could potentially impair the predictability and eectiveness of monetary pol- icy decisions. This in turn might aect the communication strategy and credibility of the Central Bank. We also perform some sensitivity analysis of the empirical results using a recursive window methodology and Consensus Economics Forecasts of ination and real
GDP growth in the estimations.
The second chapter builds on the theoretical framework of the Actual and Perceived Taylor Rules for the European Central Bank presented in the rst part of the thesis. In light of the evidence from the rolling and recursive window regressions, the goal of this chapter is to account for a possible change of regime in the re rate within a regime switching specication for the ECB's policy rate and for the re rate point forecast of the investment bank's economists. This analysis permits to investigate the presence of nonlinearities in the forward-looking policy reaction functions in a real-time setting over the period April 2000- June 2010. The estimations of the Actual and Perceived Taylor Rules help to determine whether the economists have foreseen the policy regimes of the ECB and to what extent they have accurately predicted the timing of the regime switches. This approach provides a more detailed study of the predictability of the European monetary policy and of its implication for the credibility of the ECB. The presence of several regimes undermines the relevance of a linear policy rule for describing and predicting the behavior of the Central Bank. Furthermore, we investigate the importance of the monetary aggregate M3 and of the nominal eective exchange rate in the setting of the re rate within the augmented regime switching policy rules. The latter also bring some evidence on the robustness of the baseline Actual and Perceived Taylor Rules and of the estimated regimes. As another sensitivity analysis, we also estimate the reaction functions with the Consensus Economics Forecasts to determine to what extent the regimes are sensitive to the data used in the regressions.
The third chapter oers a more in-depth analysis of the empirical specication of the Taylor Rule. Building on the New Keynesian paradigm, we derive a general nonlinear specication for the optimal policy rule that nests the standard linear Taylor Rule as a particular case. Using a novel semi-parametric modeling approach, we estimate both a standard and an augmented specication of the Taylor Rule using a Generalized Additive Model (GAM) for the main policy rate of 8 OECD countries. The analysis is conducted over the period 1976-2010 within a quarterly frequency using the OECD projections of ination and real GDP growth. The semi-parametric estimations are performed with three specications of the policy rule: a bivariate GAM, a univariate GAM and an augmented GAM. The rst model is used to study the general functional form of the Taylor Rule, while the second specication is the closest analog to the original Taylor Rule and it aims at providing more intuition on the Central Banks' behavior along the support of the explanatory variables considered in the nonlinear modeling framework. The goal of the augmented GAM is to estimate and more accurately forecast the policy rate of Central Banks and to circumvent the possible specication problem that might arise when using interest rate smoothing in the policy rule. The augmented GAM includes an interaction term between economic fundamentals and accounts explicitly for the interaction between economic fundamentals and the business cycle. In order to corroborate the relevance of the semi-parametric specications for modeling the policy rules, we perform an out-of- sample forecasting of the interest rates using two dierent techniques and two measures of the economic outlook. The latter are compared to the forecasts from a linear model and the forecast gain is computed for the best forecasting semi-parametric model. As another sensitivity analysis, we estimate the Taylor Rules with the Consensus Economics Forecasts over the period 1989-2010 within a monthly frequency. This investigation permits to further assess the forecasting performance of the semi-parametric specications in comparison to a linear Taylor Rule using forward-looking data provided in real-time.
As the nonlinearity issue of the Taylor Rule is at the heart of the present thesis, it is important to clarify what this concept refers to. In a rst step, the nonlinearity of the policy rule could be related to a structural break occurring in a period of nancial and/or
economic crisis for instance. In this case the nonlinearity of the Taylor Rule refers to dierent slope coecients over specic time periods. Another source of nonlinearity could be a change of responsiveness of the Central Bank vis-à-vis macroeconomic fundamentals depending on the state of the economy. In a period of economic euphoria, the Central Bank might react more strongly to ination, while in a state of crisis it could focus more on stabilizing economic activity. Finally, even without observing a structural break and assuming that the preferences of the Central Bank do not change, the Taylor Rule could be nonlinear because of a possible nonlinearity in the structure of the economy, reected either in the NKIS curve and/or in the NKPC. In the more general case, the nonlinearity of the policy rule could reect a combination of the three possible sources of nonlinearity previously mentioned. In the Markov regime switching approach the nonlinearity of the Taylor Rule is reected in the change of the estimated coecients, while with the semi- parametric approach the nonlinearity is observed in the shape of the estimated functions of macroeconomic fundamentals (ination and output growth). The latter provides a convenient and exible method for modeling a general specication of the policy rule within the support of the explanatory variables.
Actual versus Perceived Taylor Rules. How Predictable is the European Central Bank? ∗
1 Introduction
It has often been argued in the recent monetary economics literature that modern Central Banks follow a Taylor Rule essentially as a guiding principle for the implementation of their monetary policy strategy rather than as a strict rule of thumb. In a forward-looking environment, monetary policy is mainly considered as the art of managing the expectations of private agents, as has well emphasized Michael Woodford (2003). In the same spirit, the President of the Governing Council of the European Central Bank (ECB), Jean-Claude Trichet, has dened the monetary policy strategy during a press conference on 31 August 2006: "Our concept is very simple: we do what -in our opinion- is necessary to counter the inationary risks that we see, to deliver price stability over the medium term and to be credible in the delivery of price stability [...] I do not want to qualify in any other manner what we will do in the future, because, once again, we will do what we have to do in order to deliver price stability, be credible in the delivery of price stability over the medium term and continue anchoring inationary expectations."
The previous statement points out that the ECB's strategy is based on the rm commit- ment to deliver price stability in the policy relevant medium to long term horizons within a exible monetary policy framework.1 In this perspective, the European monetary policy can be characterized as a constrained discretion framework which features the absence of pre-commitment to future policy decisions. Thus, while being fully transparent about the ination objective, the Governing Council remains silent on the future orientation of mone- tary policy. Hence, with regard to the latter its communication with the relevant economic agents breaks down. This limited procedural transparency of the Central Bank might ham- per the predictability of the future monetary policy stance which would adversely aect the credibility of the ECB.2 De Haan, Amtenbrink and Waller (2004) point out that the
∗This paper was presented at the Annual Conference in Macroeconomics T2M in Le Mans (France) in March 2010 and at the 30th CIRET Conference on Economic Tendency Surveys organized by the Conference Board and the KOF Swiss Economic Institute in New York (USA) in October 2010.
1The Governing Council has dened price stability as a yearly rate of ination below 2%. More precisely, members of the Council have declared that their ination objective corresponds to an ination rate that is below but close to 2%. The latter embodies inationary expectations remaining in the range of 1.7-1.9% over the policy relevant horizon. The ECB has also claried that the medium term horizon corresponds to a period of approximately 2 years.
2For an extensive treatment of the relation between Central Bank transparency, communication and credibility one can refer to Blinder (2000), Blinder et al. (2001 and 2008), Demertzis and Hallett (2007),
5
ECB does not seem to be perceived as transparent and credible based on a survey they have conducted among professional economists in the rst years of the common monetary policy. However, the policy statement of Mr. Trichet emphasizes the importance of being credible in order to shape the expectations of economic agents in line with the price sta- bility goal of the ECB. An anchoring of the private sector expectations with those of the Central Bank reduces the overall macroeconomic uncertainty and enhances the predictabil- ity of monetary policy in the medium and longer terms. The alignment of expectations improves monetary policy eectiveness and fosters a low ination credibility in the ECB.
This strategy can be implemented by a transparent Central Bank that features excellent communication skills. In such a transparency enhanced framework the ECB could achieve a better anchoring of ination expectations, thus contributing to sustainable growth and employment creation in the medium to long terms. Monetary policy is hence more eective at stabilizing key macroeconomic aggregates.
Given that, on the one hand the ECB is fully committed to the delivery of price stability, while on the other it is never pre-committed to its future policy decisions, it is important to analyze the implications of the current monetary policy framework for the predictability of the European monetary policy stance.3 Most of the previous literature has studied this issue on the basis of the predictability of the money market interest rates derived from futures contracts, as for instance in Kuttner (2001) and in Ross (2002). The latter have found that future contracts provide broadly accurate predictions of the money market interest rates the Central Banks seek to inuence. Swansson (2006) has found that an improvement of the Federal Reserve transparency has gradually lead to more accurate private sector forecasts of the fed funds since the 1990's. Rosa and Verga (2007) argue that the forecast accuracy for the forthcoming re rate based on the ECB communication is similar to the market based expectations for the main policy rate. Gerlach (2007) points out that the changes in the ECB's re rate can be explained by subjective measures of the economic outlook based on the monthly bulletin of the Central Bank along with the growth rate of M3 and the nominal eective exchange rate. Conversely, the ination rate and the output gap do not seem to play a role in the ECB's interest rate setting.
More recently, using a new approach Hamilton, Pruitt and Borger (2009) have specied a market-perceived monetary policy rule to measure how market participants understand the Federal Reserve policy rule using macroeconomic news at a monthly frequency. Their model relates the change in the forecast of the future fed funds to a change in the forecasts of macroeconomic variables through a perceived Taylor Rule. Within this framework, the authors nd that the market participants indeed extract information from macroeconomic news in order to predict the future fed funds. Their ndings also point out that the estimated policy rule has changed over time. Sauer and Sturm (2003) have estimated Taylor Rules for the ECB and have found that the Central Bank implements an ination destabilizing policy for ination when using a backward-looking specication, however, its policy is ination stabilizing when including forward-looking variables in the policy rule.
Along the same lines, Gorter, Jacobs and de Haan (2008) show that the ECB responds to the expectations of ination and real GDP growth using Consensus Economics Forecasts in real-time. The authors also point out that the estimates from a contemporaneous Taylor Rule do not imply an ination stabilizing policy of the Central Bank because of the lack of a forward-looking nature in the variables. Therefore, one should prefer a specication with expectations data in order to more properly model the preemptive behavior of Central Banks.
Consistently with this new line of research, this chapter contributes to the literature
Geraats (2007, 2008 and 2009) and Woodford (2005).
3It is important to highlight that a communication of the expected interest rate path does not imply that the Central Bank is committed to a future course of policy actions since the expected path might need to be adjusted as new shocks develop and new data become available.
to the extent that it adopts an alternative approach to the analysis of monetary policy predictability. In contrast with the previous methodologies, we introduce a direct mea- sure of the ECB's key policy rate point forecasts performed by the professional forecasters of a large investment bank. Thus, we examine whether the relevant market participants have accurately predicted the main policy rate of the Central Bank before each meeting of the Governing Council. An alignment between the private forecasts of the policy rate with the actual interest rate would point out that the Central Bank has used a similar information set on macroeconomic fundamentals when setting the re rate as the profes- sional forecasters. In that perspective, Berger, Ehrmann and Fratzscher (2009) analyze the forecast accuracy of the ECB's main interest rate using the Reuters' poll of economic forecasters over the period September 2000-January 2005. Within a panel approach, the researchers show evidence in favor of an important degree of forecast heterogeneity which they explain by geographic factors, information clustering in important nancial areas and country specic macroeconomic variables. Finally, Boeckx (2011) uses monthly interpo- lated data from the quarterly ECB's Survey of Professional Forecasters (SPF) to estimate a policy reaction function for the ECB using ination and real output growth forecasts within a discrete choice approach. He nds that the predictions from an ordered probit model are in general in line with the observed policy rate and that the Central Bank has responded less strongly to the economic forecasts since the height of the nancial crisis in October 2008. However, a comparison of the results obtained from the probit model with the survey information from the Reuters' poll of forecasters points to some misalignments in the predictions. In contrast with the previous approaches, we assume that close to the monetary policy meetings, the market participants and the Central Bank should have broadly similar views on ination and on the economic outlook, which are fostered by the economic transparency of the ECB and its desire to align the private sector expectations with its macroeconomic forecasts.
In order to investigate the predictability of the key policy rate, we estimate a forward- looking specication of the Taylor Rule for the ECB, as well as a perceived Taylor Rule for the investment bank professional forecasters. As a dependent variable for the Actual Taylor Rule, we use the ECB's re rate set on the corresponding monetary policy meeting, while the Perceived Taylor Rule is based on the professional point forecasts of the re rate for the upcoming meeting. Any dierence in the estimated coecients or in the implied predictions of the policy rules might point to a limited predictability of the monetary policy stance.
Such a nding could hinder the anchoring of private agents' expectations with the Central Bank's objective and would impair the eectiveness of monetary policy. Modeling private agents' perceptions of the future policy stance is particularly valuable for the appropriate design of monetary policy, as pointed out by van der Cruijsen and Eijnger (2008) in a survey they have conducted on the perceived transparency of the ECB among the Dutch households. The researchers emphasize that the perceived transparency is more important than the actual one as the former permits to ensure a public support for the Central Bank.
The authors show that the transparency perceptions can be further enhanced by improving the public knowledge about the Central Bank and increasing its degree of procedural and operational transparency. However, the ECB has an incentive to highlight its transparency strengths and not to disclose its weaknesses in order to foster public trust in the monetary institution.
The empirical results rst suggest that the ECB's monetary policy is broadly well pre- dicted by the professional forecasters within the estimated linear policy rules. However, the economists have foreseen more accurately the policy rate hikes than the interest rate cuts. There is also evidence for increasing Central Bank's reaction to ination and out- put growth expectations along the forecast horizon. The evidence from the rolling and recursive window regressions indicates that there are some gaps between the actual and
perceived point estimates over time, in particular since the height of the nancial crisis in 2008. Finally, the results point out that the ECB has responded less strongly to key macroeconomic fundamentals and has adjusted faster the re rate to the desired target level after the bankruptcy of Lehman Brothers. The time dummies show that the broad- ening of the nancial crisis has exerted an important negative and signicant eect on the estimated policy rules. Thereby, the evidence reveals that the ECB might have switched to a new monetary policy regime that occurs in periods of deep nancial and economic slump.
The structure of the chapter is the following. Section 2 presents the theoretical frame- work, while the data and methodology used are outlined in section 3. Section 4 contains the linear model estimations of the Actual and Perceived Taylor Rules, while the rolling window regressions are performed in section 5. Section 6 presents the estimation results of the policy rules with some relevant time dummies and section 7 conducts a sensitivity analysis of the baseline results. The nal section provides some concluding remarks on the main empirical ndings of the chapter.
2 Theoretical framework
Following most of the recent literature, we specify a linear forward-looking policy reaction function.4 This approach can be justied by the long and variable lags that are necessary for monetary policy to aect real economic activity. As in this perspective monetary policy becomes the art of managing the expectations of economic agents, the Central Bank is concerned about the predictability and credibility of its strategy. We assume that at the forthcoming monetary policy meeting, occurring in periodt+ 1, the Central Bank sets the main interest rate based on the following Taylor Rule specication:
¯it+1=r∗+π∗+βπEt{πt+k−π∗|Ωt}+βyEt{yt+k−y∗|Ωt}+ηt+1 (1) where¯it+1 denotes the Central Bank's target for the policy interest rate in period t+ 1, r∗ and π∗ are the equilibrium real interest rate and the ination objective respectively.
Et{πt+k−π∗|Ωt} and Et{yt+k−y∗|Ωt} are the ination and real output growth expec- tations respectively made in period t for a horizon t+k, in deviation from the ination objectiveπ∗ and the trend real output growth rate y∗.5 Ωt denotes the available informa- tion set in periodtandηt+1is a stochastic disturbance term.6 Notice that this specication of the Taylor Rule is in line with the speed limit policy recommended by Walsh (2003). The latter has shown that a Central Bank that responds to the change in the output gap rather than to its level can deliver the optimal pre-commitment policy outcome. This policy is also welfare improving compared to ination targeting if the ination process is forward- looking. We also introduce interest rate smoothing in the specication to account for the fact that usually Central Banks adjust gradually their policy rate to the desired level in order to avoid excess volatility in nancial markets.7 The partial adjustment equation that
4The theoretical specication is based on Gorter, Jacobs and de Haan (2008), Clarida, Galí and Gertler (1998, 1999 and 2000) and Woodford (2001). This policy rule diers from the original Taylor Rule (1993) in that the former is forward-looking and the Central Bank responds to the growth rate of real GDP and not to the output gap.
5In the empirical part of the rst and second chapters,k is set to one period but it is also relevant to consider longer forecast horizons, two or three years ahead for instance, as monetary policy aects key macroeconomic variables with long and variable lags. Unfortunately, these longer forecast horizons are not available in our data set.
6In the empirical part of the chapter, we use dierent forecast horizons in the estimation of the policy rules to determine to what extent the responsiveness of the Central Bank to macroeconomic fundamentals is sensitive to the measures of expectations and how it changes along the forecast horizon.
7In the empirical sections of the rst and second chapter, we nd that there is a high level of interest rate smoothing in the estimated policy rules. The latter is consistent with the empirical evidence reported in
accounts for this practice is the following:
it+1=ρit+ (1−ρ)¯it+1+ξt+1 (2) where it+1 is the observed interest rate in period t+ 1, ρ is the interest rate smoothing parameter andξt+1 is a stochastic disturbance. This equation points out that the Central Bank implements a fraction (1−ρ) of the desired policy rate target at each meeting of the policy committee. Combining equations (1) and (2) yields the nal specication to be estimated:
it+1 =ρit+ (1−ρ)[α+βπEt{πt+k|Ωt}+βyEt{yt+k|Ωt}] +1t+1 (3) whereα=r∗+π∗(1−βπ)−y∗βy.
Equation (3) is the Actual Taylor Rule specication in which the dependent variable is the observed re rate as set at the Governing Council meeting in periodt+ 1.
We also estimate a Perceived Taylor Rule which is based on the professional point forecast of the ECB's re rate to be set at the forthcoming meeting of the policy committee performed by the economists of the investment bank one week ahead:
Et{it+1}=ρit+ (1−ρ)[α+βπEt{πt+k|Ωt}+βyEt{yt+k|Ωt}] +2t (4) In the specication of the Taylor Rules, we have assumed that both the Central Bank and the market participants share the same information set about macroeconomic funda- mentals one week ahead of the Governing Council meetings. This is a reasonable assump- tion given that the ECB Council often refers to the ination and real GDP growth forecasts of market participants when explaining monetary policy decisions during the press confer- ences. Alternatively, one could also regress the actual policy rate on the one expected by the economists one period ahead. The advantage of this approach is that we can test for the rationality of the professional forecasts. However, the goal of this chapter is to investi- gate how the economists understand the interest rate decision process of the ECB prior to the monetary policy meetings in order to shed new light on the degree of transparency of the Central Bank. For that purpose it is more suitable to design a Taylor Rules framework assuming that the Central Bank's behavior can be accurately described by a simple rule of thumb as it is often argued in the literature.
On a regular basis, the Council compares the ECB's sta macroeconomic projections with the forecasts performed by the OECD, the IMF, the European Commission and Con- sensus Economics. In addition, the Central Bank conducts its own survey of ination and real output growth expectations among a wide range of professional forecasters in the euro zone at a quarterly frequency (the ECB SPF). Using all macroeconomic projections from dierent agencies and surveys the ECB often highlights the fact that the forecasts are broadly well aligned. This evidence brings further support to the decisions taken on the main renancing operations rate (the re rate) of the ECB and should enhance their pre- dictability. The current practice of the ECB points to a high level of economic transparency of the Central Bank even though the economic forecasts should be performed at the fre- quency of the Governing Council meetings. This practice would foster the predictability of the ECB's interest rate decisions by the relevant market participants. Another possibility would be to reduce the frequency of the monetary policy decisions and turn to quarterly policy meetings for instance.
Gerlach and Lewis (2010), Smets (2009), Sturm and Wollmershäuser (2008) for instance. The econometric tests performed do not point to serial correlation problems.
3 Data and methodology
The aim of this chapter is to assess the predictability of the ECB's main policy rate (the re rate) by the professional forecasters of a large investment bank before each meeting of the Governing Council. This analysis is important as it will bring new light on determining whether the policy strategy of the Central Bank is well understood by the relevant economic agents and to what extent the monetary policy stance is accurately predicted in a real- time setting. While this issue is crucial for the communication strategy and transparency of the ECB with the markets, collecting data that are available to the market participants in real-time and at the frequency of the meetings of the monetary policy committee is a particularly dicult and challenging task. After a long and comprehensive exploration of numerous relevant data, we have collected the macroeconomic projections of a large investment bank that are available on a weekly frequency and in real-time. Based on these data, we have constructed a database that contains the real-time point forecasts of the re rate for the upcoming policy meeting, ination and real GDP growth for the euro area as reported by the economists of the bank in their weekly economic research publications.8 The macroeconomic forecasts are performed for the current calendar quarter and the next calendar quarter, as well as for the current calendar year and the next calendar year.
This framework is particularly well suited for assessing the predictability of the European monetary policy stance based on the Actual and Perceived Taylor Rules derived in the previous section.
Since there are no forecasts of the output gap performed at the frequency of the Gov- erning Council meetings we use the projections of real GDP growth in the regressions9. The approach we adopt is consistent with the methodology used by Gorter, Jacobs and de Haan (2008), Sturm and Wollmershäuser (2008) for instance who assume a constant growth rate of potential output.10 Orphanides and van Norden (2002) have highlighted the important revisions in the output gap estimates which render the use of real-time output gap forecasts unreliable. According to the authors the main problem lies in the end-of-period estimates of potential output which are not completely trustworthy with the methods applied to extract potential output, and hence could lead to mistaken policy rec- ommendations. This is unlikely to be the case for the real GDP growth forecasts used in this chapter which are not revised. Figure 1 displays the timing of the model.
Figure 1: The timing of the model
As pointed out previously, the investment bank's economists report the most likely
8We have built-up the database from their weekly economic reports made in general one week before the upcoming monetary policy meeting of the ECB. The forecasts are provided in real-time and thus are not subject to the Orphanides' critique (2001). The latter has shown that one should use real-time data for actual policy making because revised data could yield misleading outcomes. Further evidence on the use of real-time rather than revised data is provided in Molodtsova, Nikolsko-Rzhevskyy and Papell (2008).
9This method is based on the speed limit policy described by Walsh (2003). The latter has emphasized that a policy rule in which the Central Bank responds to the change in the output gap is welfare improving especially in the case of imperfect observation of the output gap.
10Within this more general approach it is reasonable to assume that the euro area potential output has most likely not changed within the relatively short time period investigated.
point forecast for the re rate set by the ECB Governing Council at the upcoming monetary policy meeting. The forecasts are usually made in the week before the corresponding policy decision and are based on the economists' projections of key macroeconomic variables for the euro area. The database spans the period from April 2000 until November 2009. The frequency of the observations corresponds to the meetings of the Governing Council of the ECB which are in general monthly.11 A detailed description of the variables used is provided in table A.1 in the appendix. Table A.2 reports some summary statistics of the series used in the estimations. At a rst look at the table one can see that the average of the ination expectations of the economists for the next calendar year is 1.807% which is fully in line with the ination objective of the ECB. This result is corroborated by the consensus average ination forecast which is 1.814% for the next calendar year. This evidence suggests that the Central Bank has been successful at anchoring the ination expectations of the market participants. Data on the key policy interest rate are taken from the ocial website of the ECB.
The approach adopted in the chapter is also consistent with Sturm and Wollmershäuser (2008) who have estimated monetary policy rules for the euro area using expectations data at the frequency of the Governing Council meetings. The goal of their paper is to determine the adequacy of the ECB's monetary policy for the euro area countries. They have found evidence that there are important stress levels in the disaggregated analysis between the monetary union member countries and that the ECB has actually put a higher political weight on the smaller states when deciding on the appropriate level of the re rate. The methodology is also in line with the approach adopted in Berger, Ehrmann and Fratzscher (2009) and Boeckx (2011) who have analyzed the forecast accuracy of the ECB's re rate. In addition, Coibion and Gorodnichenko (2011) have recently shown that the Federal Reserve has responded to the ination and output growth forecasts from the Greenbook data set they have used in estimating the monetary policy rules at the frequency of the FOMC meetings.
Figure 2 compares the actual re rate set by the ECB with the re rate point forecast from the economists of the investment bank before the corresponding monetary policy meeting. This analysis provides a rst evidence for understanding the forecasting of the main policy rate of the ECB. The gure points to a possibly hampered predictability of the re rate in the period from 2000 to the end of 2003 as indicated by the recurrent prediction errors. During the cycle of policy easing the forecasters have often expected higher re rate cuts than the ones implemented by the Central Bank. Even though the economists have not predicted well the magnitude of the policy rate adjustments, they have managed to perceive broadly well the pattern of the policy rate in that period. Since June 2003 and until December 2005 the economists' point forecasts of the re rate have remained at 2% in line with the actual policy rate. There has been only one exception in March 2004, when the economists have perceived a further decline of the re rate to a level of 1.75% which the ECB has not implemented. The gradual monetary policy tightening that the Central Bank has started in December 2005 has been remarkably well predicted by the market participants one meeting ahead. Finally, the professional forecasters have fallen short into perceiving the timing and magnitude of the re rate cuts during the recent nancial crisis,
11The Governing Council has taken monetary policy decisions twice a month until October 2001. In the empirical part of the chapter we present the estimation results of the policy rules that contain a dummy variable for the period when the Council has met more than once within a month. It is also important to emphasize that the frequency of the monetary policy meetings reects the time period that is required to assess the monetary policy stance and is not necessarily associated with a change in the underlying monetary policy rule. A change in the policy reaction function can be explained with the regime switching specication presented in the second chapter. Moreover, Berger, Ehrmann and Fratzscher (2009) do not nd evidence that the number of meetings of the Governing Council has altered the forecast accuracy of the re rate.
