Investor overconfidence describes how investors overestimate the precision of their private information after experiencing gains, leading them to trade more frequently. An investor who profits during a rising market may misattribute those gains to skill and trade more in subsequent months. In "Investor Overconfidence and Trading Volume," Statman, Thorley, and Vorkink (2003) analyzed CRSP data on 1,878 NYSE/AMEX securities from August 1962 to December 2002. A one standard deviation market return shock produced an 8.6% increase in market turnover the following month.
What the Study Found
Market turnover rose 8.6% in the month following a one standard deviation market return shock. It rose 7.3% in the second month, accumulating to a 30% increase over six months. The first-lagged market turnover coefficient was 0.284 (standard error 0.047). The first-lagged market return coefficient was 0.819 (standard error 0.133) in the same regression. At the individual security level, the first-lagged market return coefficient on turnover was 0.990, compared to 0.171 for the security's own first-lagged return. Contemporaneous market volatility carried a coefficient of 1.712 and return dispersion a coefficient of 5.024 in the market turnover regression. Both the overconfidence and disposition effects were more pronounced in small-capitalization stocks and in the earliest subperiod, 1963–1972, than in 1993–2002.
Methodology
The study used CRSP monthly and daily data on NYSE/AMEX common stocks, excluding closed-end funds, REITs, and ADRs, from August 1962 to December 2002. The sample included 485 months of market-wide data, plus 1,878 individual securities with at least 120 months of contiguous data. That yielded 530,608 monthly observations. The authors estimated vector autoregressions with impulse response functions, using 10 lags of endogenous variables and 2 lags of exogenous variables. Lag lengths were selected by the Schwarz Information Criterion. Market volatility and return dispersion served as exogenous controls; standard errors for individual-security coefficients came from a 5,000-iteration bootstrap.
Key Statistics
| Metric | Finding | Context |
|---|---|---|
| Market turnover response to 1 SD market return shock (month 1) | 8.6% increase | Full sample, Aug 1962–Dec 2002 |
| Accumulated market turnover response over 6 months | 30% increase | Full sample, Aug 1962–Dec 2002 |
| First-lagged market return coefficient (market VAR) | 0.819 (SE 0.133) | Full sample, Aug 1962–Dec 2002 |
| First-lagged market return coefficient on security turnover | 0.990 | 1,878 securities, full sample |
| First-lagged own-security return coefficient on security turnover | 0.171 | 1,878 securities, full sample |
| General VAR model | Y_t = μ + ΣA_k Y_(t-k) + ΣB_l X_(t-l) + e_t | Equation (1), Section 2.2 |
Why This Matters
Aggregate trading volume is not solely determined by liquidity needs or information events, but also by shifts in investor confidence following market gains. The lead-lag pattern between returns and turnover implies that strong market performance may foreshadow elevated trading activity. Elevated trading carries transaction costs that can erode investor returns over time. Separating the market-wide overconfidence effect from the security-specific disposition effect gives advisors a framework for diagnosing why clients trade more after gains.