The author adds to the literature on the disposition bias by analyzing US mutual fund managers’ propensity to sell winning stocks early and hold losing stocks too long under various circumstances, confirming that a disposition bias exists among professional money managers. In addition, the author studies the cross-sectional variation in holdings to determine what, if any, stock-level characteristics correlate with this bias. The results indicate that the bias is stronger when a stock is riskier (high beta) or hard to value and that high-bias managers tend to purchase attention-grabbing stocks, such as those that pay attractive dividends. The author also provides suggestions for combating disposition bias.

Behavioral finance literature tends to support the notion that through increased awareness, individuals can combat their own biases. According to the author, professional investors can “de-bias” their results by remaining aware of the stock characteristics and other variables that influence disposition bias. The author also presents evidence of irrational behavior among fund managers as a possible challenge to the concept of market efficiency.

Although the author does not question whether more-aware managers might take advantage of other traders’ biased behavior, in theory, a manager could build a list of stocks with the characteristics known to correlate with disposition bias and take positions designed to benefit from expected trading mistakes. At times during the study period, managers appear to have learned from their mistakes, which means the bias may not be stable in all markets and thus such a strategy is probably not straightforward.

One insight from this study reveals that less biased managers may focus their analytical energies on smaller companies. Accordingly, a manager might consider widening a fund’s mandate to allow smaller stocks in the portfolio to help boost returns. This strategy may be wise if the small-cap effect holds up in the future.

Finally, when dividends correlate strongly with the disposition bias, managers’ screens should include moderating data points to limit the extent to which dividends drive trading behavior. A secondary qualitative review could be required, or cash flow measures (instead of dividends) could proxy for yield.

To gauge the disposition effect, the author follows prior research to compute a weighted average purchase price (WAPP) for each stock in a fund, which is then used to compute the assumed realized and unrealized gains or losses.

Because funds do not publish daily holdings data, periodic holdings reports are used instead, and the trades are assumed to have taken place at each quarter-end. This limitation should be kept in mind when interpreting the study’s results, because the author does not know for sure at what price managers traded.

If holdings have changed, a trade is assumed and the realized gain or loss is calculated against the WAPP. If holdings have not changed, the paper gains or losses are calculated against the WAPP. Finally, the author calculates a proportion of realized-to-unrealized gains or losses to arrive at a “spread” that measures the disposition bias.

After computing each manager’s bias in terms of the spread, the author runs various regressions against manager bias deciles to determine which stock characteristics and market conditions best correlate with the bias. Characteristics include, among others, book-to-market ratio; earnings per share; earnings volatility; price-to-earnings ratio; cash flow volatility; dividend yield (and a dividend dummy); intangible assets; property, plant, and equipment; market cap; firm size; mean return; turnover; shares outstanding; volume; and market return. Regressions include such controls as manager experience and team versus individual manager fund structure.

The data, from Thomson Financial, CRSP, and Compustat, cover the period 1980–2010. Results are benchmarked against the S&P 500 Index using the three-month T-bill as the risk-free rate.

The author presents a very thorough analysis of the disposition effect and stock-level characteristics that correlate with it. Given that the research shows the effect can wane in certain periods, indicating that managers learn over time, opportunities to apply the study’s findings in a real-world portfolio may be limited. But as a self-check, the findings could be very useful in helping an active stock manager compensate for known behavioral pitfalls. If actual daily trade data were available, with gain and loss information from a sample of high- and low-bias managers, the author could further evaluate the accuracy of these research results.