LONG-TERM MARKET OVERREACTION OR BIASES IN COMPUTED RETURNS

We show that the returns to the typical long-term contrarian strategy implemented in previous studies are upwardly biased because they are calculated by cumulating single-period (monthly) returns over long intervals. The cumulation process not only cumulates ''true returns but also the upward bias in single-period returns induced by measurement errors. We also show that the remaining ''true'' returns to loser or winner firms have no relation to overreaction. This study has important implications for event studies that use cumulative returns to assess the impact of information events.