The rank effect, which is investors’ tendency to sell extreme winning and losing positions in their portfolios, is identified as a decision rule in behavioral finance. It provides evidence for a holistic choice framework in which asset choices are based on comparisons with the combined holdings of a given portfolio.
As an extension of behavioral finance theory regarding the joint evaluation of decisions, the author conducts a series of carefully controlled analyses of retail traders and mutual fund managers and determines that individuals are more likely to sell the most extreme (winning and losing) portfolio positions. This rank effect is not related to firm-specific information, holding period, or the actual level of returns but instead to the noteworthiness of the outlier holdings. It can also be influenced by marketwide characteristics, such as stocks highlighted by recent high returns or trading volume. The effect appears to be independent of tax law incentives.
How Is This Research Useful to Practitioners?
Any new evidence on the fundamental processes by which investment choices are made may justify practitioner attention. The author notes that the disposition effect, which is the tendency to sell gains more than losses, has been well established as arguably the most studied trading behavior in finance, but the rank effect can also have real performance implications. And its impact can be sharply distinguished from any operative disposition effect because evidence shows that the stocks most likely to be sold by mutual funds are the worst-ranked positions in their portfolios; for retail investors, those holdings have just a slightly lower probability of sale. When the investor decides to sell, there is a 26%–31% chance of selling either the worst or best performing stock in the portfolio. So, the impact of losses to propel portfolio moves seems to be a tangible influence on trading.
The rank effect appears to have basic universality, although it may vary slightly by degree of investor sophistication (e.g., novices are more likely to hold losing positions). For example, even the alphabetical order in portfolios by company name can induce a bias toward sales of those holdings higher on an alphabetical list, regardless of the underlying appeal of a stock’s fundamentals determined through economic analysis. The rank effect is a reminder that investment assessment is based in part on context, namely through comparisons across holdings in a given portfolio. This view is in stark contrast to the typical assumption of narrow framing, which holds that each stock is evaluated in isolation on its own merits regardless of its companion assets.
How Did the Author Conduct This Research?
The analysis is based on two datasets that focus on decisions to sell stocks. For individual investors, daily trades for 10,619 unique accounts at a large discount brokerage firm for the period January 1991 to November 1996 are matched with corresponding portfolios. For mutual funds, over the period of 1980–2000, Thompson Reuters holdings data are combined with fund price and volume information from CRSP to monitor 4,730 funds. Stocks are ranked best to worst by return from purchase with statistical analysis performed on the likelihood of sales. The demonstrated tendency to sell extreme performing positions is not simply a matter of rank substituting for firm-specific information, such as changes in volatility or increased news coverage. This finding is confirmed by comparisons between otherwise similar stocks that are simultaneously ranked extreme in one investor’s portfolio but not so in another’s. So, the stocks differ only by rank among various investors.
To determine whether the rank effect is proxying for the disposition effect, the author controls for stock performance after purchase to show that investors are not disproportionately inclined to hold extreme losing positions. He controls for return levels to show that relative performance is driving the rank effect. Middle-ranked holdings are confirmed to have the lowest likelihood of sale by a linear probability model. Tax-related effects are ruled out by showing that tax deferred and taxable accounts share a comparable rank effect. Attention-related effects, such as alphabetical ordering, are also included in the controls.
Having won industry recognition for its research quality, the author’s thorough canvas of applicable psychological theory provides a convenient “one-stop shop” for thinking on a previously under-investigated type of decision-making shortcut related to the noteworthiness of extreme positions. The study’s methodology of isolating the rank effect from potential confounding factors is creative. It also offers a foundation for a potentially profitable investment strategy, namely using the publicly available information on mutual fund positions to anticipate holdings that will be under selling pressure and thus targeting trades to exploit the resultant price effects.