This paper provides empirical analyses of three explanations for the observed positive autocorrelation of short-horizon stock index returns, using NYSE/AMEX and NASDAQ data. Results indicate that index autocorrelation cannot be substantially explained by either autocorrelated, time-varying expected returns, or nonsynchronous trading. The third explanation for index autocorrelation, the nonsynchronous information transfer hypothesis, states that stocks incorporate market-wide information on a nonsynchronous basis due to information and transaction costs. Evidence from analyses of mean returns on various portfolios following large returns on the S 500 futures contract, as well as regressions of portfolio returns on current and lagged futures returns, support this explanation. Small (large) firms collectively require approximately 7 (1-2) weeks to fully incorporate new market information on average, and this delayed impoundment accounts for the bulk of the observed autocorrelation.
market efficiency information transfer index return autocorrelation