A Non-Random Walk Down Wall Street

Transcription

1 A Non-Random Walk Down Wall Street Andrew W. Lo A. Craig MacKinlay Princeton University Press Princeton, New Jersey

2 list of Figures List of Tables Preface xiii xv xxi 1 Introduction The Random Walk and Efficient Markets The Current State of Efficient Markets Practical Implications 8 Parti 13 2 Stock Market Prices Do Not Follow Random Walks: Evidence from a Simple Specification Test The Specification Test Homoskedastic Increments Heteroskedastic Increments The Random Walk Hypothesis for Weekly Returns Results for Market Indexes Results for Size-Based Portfolios Results for Individual Securities Spurious Autocorrelation Induced by Nontrading The Mean-Reverting Alternative to the Random Walk Conclusion 39 Appendix A2: Proof of Theorems 41 vii

3 viii 3 The Size and Power of the variance Ratio Test in Finite Samples: A Monte Carlo Investigation Introduction The Variance Ratio Test The IID Gaussian Null Hypothesis The Heteroskedastic Null Hypothesis Variance Ratios and Autocorrelations Properties of the Test Statistic under the Null Hypotheses The Gaussian IID Null Hypothesis A Heteroskedastic Null Hypothesis Power The Variance Ratio Test for Large q Power against a Stationary AR(1) Alternative Two Unit Root Alternatives to the Random Walk Conclusion 81 4 An Econometric Analysis of Nonsynchronous Trading Introduction A Model of Nonsynchronous Trading Implications for Individual Returns Implications for Portfolio Returns Time Aggregation An Empirical Analysis of Nontrading Daily Nontrading Probabilities Implicit in Autocorrelations Nontrading and Index Autocorrelations Extensions and Generalizations 105 Appendix A4: Proof of Propositions When Are Contrarian Profits Due to Stock Market Overreaction? Introduction A Summary of Recent Findings Analysis of Contrarian Profitability The Independently and Identically Distributed Benchmark Stock Market Overreaction and Fads Trading on White Noise and Lead-Lag Relations Lead-Lag Effects and Nonsynchronous Trading A Positively Dependent Common Factor and the Bid-Ask Spread An Empirical Appraisal of Overreaction 132

4 ix 5.5 Long Horizons Versus Short Horizons Conclusion 142 Appendix A Long-Term Memory in Stock Market Prices Introduction Long-Range Versus Short-Range Dependence The Null Hypothesis Long-Range Dependent Alternatives The Rescaled Range Statistic The Modified R/S Statistic The Asymptotic Distribution ofj^ The Relation Between Q^ and Q, The Behavior of Q^ Under Long Memory Alternatives R/S Analysis for Stock Market Returns The Evidence for Weekly and Monthly Returns Size and Power The Size of the R/S Test Power Against Fractionally-Differenced Alternatives Conclusion 179 Appendix A6: Proof of Theorems 181 Part II Multifactor Models Do Not Explain Deviations from the CAPM Introduction Linear Pricing Models, Mean-Variance Analysis, and the Optimal Orthogonal Portfolio Squared Sharpe Measures Implications for Risk-Based Versus Nonrisk-Based Alternatives Zero Intercept F-Test Testing Approach Estimation Approach Asymptotic Arbitrage in Finite Economies Conclusion Data-Snooping Biases in Tests of Financial Asset Pricing Models Quantifying Data-Snooping Biases With Induced Order Statistics Asymptotic Properties of Induced Order Statistics Biases of Tests Based on Individual Securities

5 x Biases of Tests Based on Portfolios of Securities Interpreting Data-Snooping Bias as Power Monte Carlo Results Simulation Results for 0 p Effects of Induced Ordering on F-Tests F-Tests With Cross-Sectional Dependence Two Empirical Examples Sorting By Beta Sorting By Size How the Data Get Snooped Conclusion Maximizing Predictability in the Stock and Bond Markets Introduction Motivation Predicting Factors vs. Predicting Returns Numerical Illustration Empirical Illustration Maximizing Predictability Maximally Predictable Portfolio Example: One-Factor Model An Empirical Implementation The Conditional Factors Estimating the Conditional-Factor Model Maximizing Predictability The Maximally Predictable Portfolios Statistical Inference for the Maximal B? Monte Carlo Analysis Three Out-of-Sample Measures of Predictability Naive vs. Conditional Forecasts Merton's Measure of Market Timing The Profitability of Predictability Conclusion 283 Partffl An Ordered Probit Analysis of Transaction Stock Prices Introduction The Ordered Probit Model Other Models of Discreteness The Likelihood Function The Data Sample Statistics The Empirical Specification 307

6 xi 10.5 The Maximum Likelihood Estimates Diagnostics Endogeneity of At k and IBS k Applications Order-Flow Dependence Measuring Price Impact Per Unit Volume of Trade Does Discreteness Matter? A Larger Sample Conclusion Index-Futures Arbitrage and the Behavior of Stock Index Futures Prices Arbitrage Strategies and the Behavior of Stock Index Futures Prices Forward Contracts on Stock Indexes (No Transaction Costs) The Impact of Transaction Costs Empirical Evidence Data Behavior of Futures and Index Series The Behavior of the Mispricing Series Path Dependence of Mispricing Conclusion Order Imbalances and Stock Price Movements on October 19 and 20, Some Preliminaries The Source of the Data The Published Standard and Poor's Index The Constructed Indexes Buying and Selling Pressure A Measure of Order Imbalance Time-Series Results Cross-Sectional Results Return Reversals Conclusion 387 Appendix Al A12.1 Index Levels 389 A12.2 Fifteen-Minute Index Returns 393 References 395 Index 417