The Impact of Salience on Investor Behavior: Evidence from a Natural Experiment

Abstract

We test whether the display of information causally affects investor behavior in a high-stakes trading environment. Using investor-level brokerage data from China and a natural experiment, we estimate the impact of a shock that increased the salience of a stock's purchase price but did not change the investor's information set. We employ a difference-indifferences approach and find that the salience shock causally increased the disposition effect by 17%. We use microdata to document substantial heterogeneity across investors in the treatment effect. A previously documented trading pattern, the "rank effect," explains heterogeneity in the change in the disposition effect.

How Do Taxes Affect the Trading Behavior of Private Investors? Evidence From Individual Portfolio Data

“…As recently suggested by Frydman and Wang (2020), the causes of the disposition effect are still subject to debate, and we are able to add to this debate in that we show that capital-gains taxes have an impact on the disposition effect.…”

Measuring the disposition effect

“…In this simple case, the OLS estimate of β 1 in Equation 5 will be fully equivalent to Odean's (1998) DE measure 2 , while the intercept estimate corresponds to the probability of selling a losing position (i.e., P LR i in Equation 3). This simple OLS technique has been used by Chang et al (2016) to study how the DE is affected by whether the asset held is a stock or a fund and by Frydman & Wang (2020) to analyze the effects of salience shocks on the DE.…”

“…These characteristics make them quite suitable for analyzing the disposition effect, and it is thus natural that they are used in several papers published in top journals. For example, hazard models are used in Seru et al (2010) to examine the effect of experience on the DE and in Heimer (2016) to test the relation with social interaction, as well as in some other papers [e.g., Coval & Shumway (2005), Frydman & Wang (2020)].…”

“…The first one is a logistic regression, which allows the addition of variables and controls on the right-hand side of the equation. In analyzing the relationship between a given variable and the DE, it is natural to include this variable in the model, but it is also necessary to include interaction term(s), as in Dierick et al (2019) or Frydman & Wang (2020). However, with logistic regressions, a choice again has to be made between using a daily frequency or a sale-based frequency.…”

“…In addition to these ratio-based approaches, econometric techniques have been proposed to analyze the DE and provide interesting alternatives, especially when the goal is to understand which factors (such as financial literacy or portfolio size) may affect the individual DE or when there is a need to control for such factors when comparing individuals. Frydman & Wang (2020) use ordinary least squares regressions to check whether the salience of a stock purchase price affects the disposition effect in a natural experiment. Logistic regression models allow Grinblatt & Keloharju (2001) to check whether the DE interacts with past returns to modify the propensity to sell.…”