Investor Overconfidence and Trading Volume

Abstract

High market-wide returns make some investors overconfident because they incorrectly attribute the gains to their stock picking talents. Investors who are subject to biased self-attribution increase their trading in subsequent periods in models by Gervais and Odean (2001) and Odean (1998a). We use a vector autoregressive and impulse-response function methodology to investigate the trading volume implication of the overconfidence hypothesis. We find empirical support for the overconfidence hypothesis as well as the disposition affect of Shefrin and Statman (1985). Specifically, market-wide trading activity in NYSE/AMEX shares is positively correlated to past shocks in market return, with the turnover response lasting months and perhaps years. This increase (decrease) in market-wide trading activity subsequent to bull (bear) markets can be explained by either the overconfidence hypothesis or the disposition effect. Vector autoregressions on individual stocks indicate that volume responds to past shocks in individual security return, which prior researchers have interpreted as evidence of disposition effect trading. However, we show that individual security trading activity is even more responsive to past shocks in the market-wide return, which we interpret as evidence of the overconfidence hypothesis. The empirical lead-lag relationships between returns and trading activity as measured by turnover are both statistically and economically significant and an important empirical characteristic of the domestic equity market. The trading volume patterns we document are in addition to well-known volume-volatility relationships and turn-of-the-year effects, and are robust to a number of specification alternatives.