An assessment of stock market overreaction

Abstract

Neo-classical finance builds its underpinnings on the Efficient Market Hypothesis (EMH). Amongst the plethora of evidence against the EMH, the overreaction hypothesis postulates that investors are poor Bayesian decision makers. This study investigates whether there is overreaction in the stock market. An event study approach was undertaken to explore overreaction in the stock market, focusing on U.S. data from the S&P 500 index spanning from 1980 to 2020. The days considered in the event sample were those when the stock market overreacted, with this defined as a change in the daily closing price which falls within the top one percentile of the dataset, with days of positive overreaction and days of negative overreaction given equal weight in the sample. A single linear regression model was then employed, with the aim of assessing whether a relationship exists between the daily return during days of overreaction and the daily return the following trading day. The results suggest that during days of overreaction, a one percentage point increase in daily returns leads to a 0.19 percentage point decrease in the returns of the next trading day. This result is robust along two important dimensions. The model was then used for out-of sample forecasting and was able to beat a naïve, random walk forecast. The results therefore imply preliminary evidence of overreaction, contradicting the EMH.