The disposition effect describes investors' tendency to sell winning stocks too quickly while holding on to losing stocks too long. An investor whose holding has risen in value rushes to lock in the gain. A stock trading below its purchase price instead gets held in hope of a rebound. Dhar and Zhu (2002) published this analysis as Up Close and Personal: An Individual Level Analysis of the Disposition Effect. They analyzed trading records of 7,965 individual investors at a large U.S. discount brokerage firm from January 1991 to November 1996. They found a mean individual-level disposition effect of 0.19, compared with a market-aggregate-level disposition effect of 0.068. Despite this average bias, 19.7 percent of investors showed no disposition effect or the opposite pattern.
What the Study Found
High-income investors showed a mean disposition effect of 0.189, compared with 0.211 for low-income investors, a difference significant at the 10 percent level (p=0.051). Professional investors exhibited a mean disposition effect of 0.2029, compared with 0.2450 for non-professional investors, a difference significant at the 5 percent level (p=0.028). Non-employed investors, who have a mean age of 61 and are mostly retired, had the lowest mean disposition effect at 0.1738. The Proportion of Gain Realized (PGR) averaged 0.38 while the Proportion of Loss Realized (PLR) averaged 0.19 across the sample. 4.7 percent of investors never realized their winning stocks during the period, while 21.5 percent never sold any losing stocks.
Methodology
Dhar and Zhu (2002) used trade, position, and demographics files from a large U.S. discount brokerage firm covering January 1991 to November 1996. The disposition effect was calculated for each investor as the Proportion of Gain Realized minus the Proportion of Loss Realized. The sample included 14,872 investors with calculable trading records and 7,965 investors with demographic data. Income categories (high-, medium-, and low-income) and occupation categories (professional, non-professional, non-employed) served as proxies for investor sophistication. Regressions controlled for the logarithm of investor age, the logarithm of number of trades, return of realized gains and losses, and portfolio size.
Key Statistics
| Metric | Finding | Context |
|---|---|---|
| Mean disposition effect (individual-level) | 0.19 | Full sample, Jan–Nov trades, 1991–1996 |
| Mean disposition effect (market-aggregate-level) | 0.068 | Computed by aggregating across all investors |
| High-income vs. low-income disposition effect | 0.189 vs. 0.211 | Difference significant at 10% (p=0.051) |
| Professional vs. non-professional disposition effect | 0.2029 vs. 0.2450 | Difference significant at 5% (p=0.028) |
| Investors with no or opposite disposition effect | 19.7% | Full sample with calculable DE |
| Disposition Effect formula | DE = PGR − PLR | PGR = Proportion of Gain Realized, PLR = Proportion of Loss Realized |
Why This Matters
Brokerage firms that help clients recognize the disposition effect may improve clients' after-tax portfolio performance, since realizing losses earlier can generate tax benefits. Policymakers evaluating self-directed retirement accounts should consider that behavioral biases are not uniformly distributed across the investor population. Financial advisors working with less sophisticated clients may need to provide more explicit guidance on when to sell underperforming positions.