Empirical tests of the overconfidence hypothesis

The conventional asset pricing models and the theoretical notions of efficiency have a difficult time in explaining several stylized facts observed in securities markets. Overconfidence has been advanced as an explanation for these observed anomalies. We derive four testable empirical hypotheses of the overconfidence hypothesis from recent theoretical works. First, overconfident investors overreact to private information and underreact to public information. Second, market gains (losses) increase (decrease) investors' overconfidence, and consequently they trade more (less) aggressively in subsequent periods. Third, as investors become overconfident, they underestimate risk and, as such, trade more of the riskier securities. Fourth, excessive trading of overconfident investors in securities markets makes a contribution to observed excessive volatility. In this paper we conduct various tests to empirically examine the validity of the above-mentioned testable hypotheses of the overconfidence hypothesis.; First, by identifying private and public information based on a restricted VAR framework in which a restriction imposed on the model is based on theoretical considerations, we find that stock prices tend to overreact to private information with an initial short-horizon underreaction, while tending to underreact to public information.; Second, we find strong evidence that an increase in stock returns is always followed by an increase in trading volume, as well as an increase in the Index of Consumer Sentiment, as a proxy for investors' confidence. In addition, the Index of Consumer Sentiment contemporaneously increases with stock returns.; Third, we form the portfolios on the basis of different measures of risk simultaneously to disentangle the overconfidence hypothesis and the disposition effect, and to investigate the relationship between overconfidence and risk taking.; Fourth, we find that the volatility of stock return contemporaneously increases with trading volume, the volatility of trading volume, the Index of Consumer Sentiment, and the volatility of the Index of Consumer Sentiment. In addition, by linking the contemporaneous relation between conditional volatility and volume to the lead-lag relation between return and volume, we find some evidence that the observed excessive volatility effectively stems from the excessive trading of overconfident investors.; Overall, our test results provide empirical evidence in support of the overconfidence hypothesis except for the relationship between overconfidence and risk taking. (Abstract shortened by UMI.)...