Abstract
This paper provides an extensive empirical investigation into the sources of index return autocorrelation, focusing on the relation between autocorrelation in individual stock returns and autocorrelation in index returns. The study uses daily data from the Stockholm Stock Exchange over the period 1980-1995 and reports three main empirical findings. Daily Swedish stock index returns exhibit strong, and consistently positive, first order autocorrelation throughout the sample period. Positive autocorrelation is observed for return frequencies between 1 day and 3 months. The most liquid stocks exhibit strong positive return autocorrelation. Less liquid stocks exhibit weak or negative return autocorrelation. Autocorrelation is asymmetric, high after days of above average performance of the stock market, low after days of below average performance. When compared to the other days of the week, both index returns and individual stock returns exhibit the strongest autocorrelation following on Friday returns. The transaction cost hypothesis was tested using the Swedish stock market transaction tax. Results indicate lower precision of stock prices during the transaction tax period, but no direct effect on return autocorrelation. The paper concludes that at least three sources contribute to observed return autocorrelation. For daily and short-term returns, profit taking and nonsynchronous trading are the probable causes of observed autocorrelation. For monthly and longer term returns, time-varying expected returns best describe the empirical results.