On the computation of returns in tests of the stock market overreaction hypothesis
A number of recent studies have attempted to find out whether investors overreact. We argue that, in many of these studies, the method used to compute cumulative returns-the arithmetic method-is flawed, and we show that estimates of portfolio performance can be affected. Our findings are relevant not only to tests of the overreaction hypothesis but to other areas in finance, where the arithmetic method and event-study approach are used. © 1994.