Sunteți pe pagina 1din 14

Hidden Markov Models

Advanced Studies in Theoretical and Applied Econometrics


Volume 40

Managing Editor: J. Marquez, The Federal Reserve Board, Washington, D.C., U.S.A.

Editorial Board: F.G. Adams, University of Pennsylvania, Philadelphia, U.S.A. P. Balestra, University of Geneva, Switzerland M.G. Dagenais, University of Montreal, Canada D. Kendrick, University of Texas, Austin, U.S.A. J.H.P. Paelinck, Netherlands Economic Institute, Rotterdam, The Netherlands R.S. Pindyck, Sloane School of Management, M.I.T., U.S.A. H. Theil, University of Florida, Gainesville, U.S.A. W. Welfe, University of Lodz, Poland

The titles published in this series are listed at the end of this volume.

Hidden Markov Models


Applications to Financial Economics

by

Ramaprasad Bhar
School of Banking and Finance, The University of New South Wales, Sydney, Australia

and

Shigeyuki Hamori
Graduate School of Economics, Kobe University, Japan

KLUWER ACADEMIC PUBLISHERS


NEW YORK, BOSTON, DORDRECHT, LONDON, MOSCOW

eBook ISBN: Print ISBN:

1-4020-7940-0 1-4020-7899-4

2004 Springer Science + Business Media, Inc. Print 2004 Kluwer Academic Publishers Dordrecht All rights reserved No part of this eBook may be reproduced or transmitted in any form or by any means, electronic, mechanical, recording, or otherwise, without written consent from the Publisher Created in the United States of America

Visit Springer's eBookstore at: and the Springer Global Website Online at:

http://www.ebooks.kluweronline.com http://www.springeronline.com

To Rajiv, Mitra, Hitoshi, Makoto, and Naoko

Contents

Dedication Acknowledgments List of Figures List of Tables 1. INTRODUCTION

v xi xiii xvii 1

1 2 3 4 5 6 7 8 9 10

Introduction Markov Chains Passage Time Markov Chains and the Term Structure of Interest Rates State Space Methods and Kalman Filter Hidden Markov Models and Hidden Markov Experts HMM Estimation Algorithm HMM Parameter Estimation HMM Most Probable State Sequence: Viterbi Algorithm HMM Illustrative Examples

1 1 5 6 11 13 16 18 22 24 29 29 31 31 32 33 33

2. VOLATILITY IN GROWTH RATE OF REAL GDP 1 2 Introduction Models 2.1 GARCH Model 2.2 Markov Switching Variance Model Data Empirical Results

3 4

viii 5 Conclusion 38 41 41 44 44 45 46 46 51 55 55 58 62 63 76 81 81 81 83 85 85 86 87 91 107

3. LINKAGES AMONG G7 STOCK MARKETS 1 2 Introduction Empirical Technique 2.1 Markov Switching Stock Return Model 2.2 Concordance Measure Data Empirical Results Conclusion

3 4 5

4. INTERPLAY BETWEEN INDUSTRIAL PRODUCTION AND STOCK MARKET 1 2 3 4 5 Introduction Markov Switching Heteroscedasticity Model of Output and Equity Data Empirical Results Conclusion

5. LINKING INFLATION AND INFLATION UNCERTAINTY 1 Introduction 1.1 Ination and Ination Uncertainty 1.2 Ination Uncertainty and Markov Switching Model Empirical Technique 2.1 Markov Switching Heteroscedasticity Model of the Ination Rate 2.2 Non-Nested Model Selection using Vuong Statistic Data Empirical Results Conclusion

3 4 5

6. EXPLORING PERMANENT AND TRANSITORY COMPONENTS OF STOCK RETURN 1 Introduction

117 117

ix 2 3 4 5 Markov Switching Heteroscedasticity Model of Stock Return Data Empirical Results Conclusion 119 120 121 125

7. EXPLORING THE RELATIONSHIP BETWEEN COINCIDENT FINANCIAL MARKET INDICATORS 1 Introduction 2 Markov Switching Coincidence Index Model 3 Data 4 Empirical Results 5 Conclusion References Index

127 127 129 131 131 139 145 153

Acknowledgments

We dedicate this monograph to express our thanks to our families, friends, and colleagues for their invaluable support in the writing of this volume. Without the excellent editorial guidance of Mrs. Cathelijne van Herwaarden and Mrs. Herma Drees, the kind and helpful comments of Mr. Akira Tokihisa, and the love of our family members Rajiv, Mitra, Hitoshi, Makoto, and Naoko, we never would have succeeded in completing this work. This research was partly supported by a grant-in-aid from the Japan Society for the Promotion of Science.

xi

List of Figures

1.1 2.1 2.2 2.3 4.1 4.2 4.3 4.4 4.5 4.6 4.7 4.8 4.9 4.10 4.11 4.12 4.13 4.14 4.15 4.16 4.17 4.18

Filtering Algorithm Estimated variance from the Markov switching heteroskedasticity model: Japan Estimated variance from the Markov switching heteroskedasticity model: UK Estimated variance from the Markov switching heteroskedasticity model: USA Industrial production growth: Canada Excess return: Canada Estimated ltered probability, Pr(St = 2): Canada Industrial production growth: France Excess return: France Estimated ltered probability, Pr(St = 2): France Industrial production growth: Germany Excess return: Germany Estimated ltered probability, Pr(St = 2): Germany Industrial production growth: Italy Excess return: Italy Estimated ltered probability, Pr(St = 2): Italy Industrial production growth: Japan Excess return: Japan Estimated ltered probability, Pr(St = 2): Japan Industrial production growth: UK Excess return: UK Estimated ltered probability, Pr(St = 2): UK xiii

13 36 37 37 67 67 67 68 68 68 69 69 69 70 70 70 71 71 71 72 72 72

xiv
4.19 4.20 4.21 5.1 5.2 5.3 5.4 5.5 Industrial production growth: USA Excess return: USA Estimated ltered probability, Pr(St = 2): USA Ination rate: Germany Ination rate: Japan Ination rate: UK Ination rate: USA Relationship among the probability of high variance state for permanent shocks, Pr(St = 1), the probability of high variance state for transitory shocks, Pr(St = 2), and ination rates: Germany Relationship among the probability of high variance state for permanent shocks, Pr(St = 1), the probability of high variance state for transitory shocks, Pr(St = 2), and ination rates: Japan Relationship among the probability of high variance state for permanent shocks, Pr(St = 1), the probability of high variance state for transitory shocks, Pr(St = 2), and ination rates: UK Relationship among the probability of high variance state for permanent shocks, Pr(St = 1), the probability of high variance state for transitory shocks, Pr(St = 2), and ination rates: USA Probability of high variance state for permanent shocks: Germany Probability of high variance state for permanent shocks: Japan Probability of high variance state for permanent shocks: UK Probability of high variance state for permanent shocks: USA Probability of high variance state for transitory shocks: Germany Probability of high variance state for transitory shocks: Japan Probability of high variance state for transitory shocks: UK Probability of high variance state for transitory shocks: USA

73 73 73 88 88 89 89

92

5.6

92

5.7

93

5.8

93 96 96 97 97 98 98 99 99

5.9 5.10 5.11 5.12 5.13 5.14 5.15 5.16

xv
5.17 5.18 5.19 5.20 5.21 5.22 5.23 5.24 7.1 7.2 7.3 7.4 7.5 7.6 Components of forecast variance at k = 2: Germany 103 Components of forecast variance at k = 4: Germany 103 Components of forecast variance at k = 2: Japan 104 Components of forecast variance at k = 4: Japan 104 Components of forecast variance at k = 2: UK 105 Components of forecast variance at k = 4: UK 105 Components of forecast variance at k = 2: USA 106 Components of forecast variance at k = 4: USA 106 Estimated coincident indicator and excess return: Japan 136 Estimated coincident indicator and excess return: UK 136 Estimated coincident indicator and excess return: USA 137 Inferred probability of low variance state: Japan 137 Inferred probability of low variance state: UK 138 Inferred probability of low variance state: USA 138

List of Tables

2.1 2.2 2.3 2.4 3.1 3.2 3.3 3.4 3.5 3.6 3.7 4.1 4.2 4.3 4.4

Growth rate in real GDP: GARCH(1,1) estimation Diagnostics using standardized residuals from GARCH(1,1) model Growth rate in real GDP: Markov switching heteroskedasticity estimation Diagnostics using standardized residuals from Markov switching heteroskedasticity model Summary statistics on stock return Correlation of stock returns Parameter estimates: Markov switching heteroscedasticity model of stock return Diagnostics using standardized returns from Markov switching heteroscedasticity model Estimated Markov probabilities of staying in the same state for the G7 countries Correlation statistics between probability of high variance state Concordance statistics between probability of high variance state Summary statistics: monthly excess return Summary statistics: monthly industrial production growth Parameter estimates: bivariate Markov switching heteroscedasticity model of output and equity Diagnostics using standardized residuals from excess return equation xvii

34 34 35 35 47 47 48 49 49 50 50 62 63 64 66

xviii
4.5 4.6 Correlation between probabilities of expansion states Concordance statistics, probabilities of expansion states or low volatility states Concordance statistics, expected probabilities in three months for expansion states Summary statistics Unit root test Parameter estimates: Markov switching heteroscedasticity model of ination rate Residual diagnostics and model adequacy tests Regime classication measure (RCM) Vuong statistics for non-nested model selection: Markov switching model against two dierent GARCH models Summary statistics Permanent and transitory components of equity return: Markov switching heteroscedasticity framework Residual diagnostics and model adequacy tests Correlations: return, permanent and temporary components Parameter estimates for the Markov switching coincidence index model Diagnostic statistics for the residuals of measurement equations Correlation between the coincident nancial indicator and its components Granger causality tests

74

75

4.7

75 87 90

5.1 5.2 5.3

94 100 101

5.4 5.5 5.6

101 121

6.1 6.2

122 123

6.3 6.4

123

7.1

132

7.2

133

7.3

133 134

7.4

S-ar putea să vă placă și