复旦大学投资学课件 Section6 Random Walks.ppt

Stock market prices do not follow random walks: evidence from a simple specification test Andrew W.Lo A.Craig MacKinlay The review of financial studies 1988 Abstract Menu 1. The Specification Test 2. The random walk hypothesis for weekly returns 3. Spurious autocorrelation induced by nontrading 4.The mean-reverting alternative to the random walk 5.Conclusion Introduction Fama(1970): most studies were unable to reject the EMH of common stocks Forecastability of common stock returns: Random walk theory. But some papers find predictable stock returns components Fama and French(1987): negatively serially correlated returns This article provide evidence that stock prices do not follow random walks by using a simple specification test based on variance estimators. Results : RW not consistent with weekly returns. But positive serial correlation for weekly and monthly returns Data: 1962-1985 weekly, CRSP return index AR(1)=0.3 This result may not imply inefficiency Economic model of price generating: other models This test: the variance of the increments of a random walk is linear in the sampling interval 1. The specification test Deviate from normality Develop a test statistic which is robust to many forms of heteroscedasticity and nonnormality 1.1 Homoscedastic increments Main estimators Some intuition of these variance ratios 1.2 Heteroscedastic increments Volatilities do change over time Need to develop a specification test of the random walk model that is robust to changing variances 2. The random walk hypothesis for weekly returns 1962-1985 CRSP weekly data (Wednesday) Test both equal and value-weighted CRSP indexes, q=2 to 16 The ratios should be 1+AR(1) The rejection of RW is weaker for the value-weighted index. But the general pattern persist The variance ratios exceed 1 and the z(q) decline as q increases Four week data Cannot reject the RW model even for the equal-weighted index. Using monthly data will have similar results previously 2.2 results