Do large losses loom larger than gains? Salience, holding periods, and the disposition effect

• Published date: 01 May 2022
• Author: Vladimir Kotomin,Abhishek Varma
• Date: 01 May 2022
• DOI: http://doi.org/10.1111/fire.12288

DOI: 10.1111/fire.12288

ORIGINAL ARTICLE

Do large losses loom larger than gains? Salience,

holding periods, and the disposition effect

Vladimir Kotomin Abhishek Varma

Illinois State University, Normal, Illinois, USA

Correspondence

AbhishekVarma, Illinois State University,

Normal,IL 61790–5480, USA.

Email:avarma@ilstu.edu

Abstract

Individual investors are more likely to sell stocks with

nominal gains and losses that are large relative to their

brokerage portfolio value. The salience of nominal gains and

losses affects stock sales in both taxable and tax-deferred

accounts and across investor groups, but the effect of

nominal losses is weaker for stocks with high valuation

uncertainty. The effect has a time dimension: at short hold-

ing periods, individuals are more likely to sell stocks with

large nominal losses than gains of the same size, mitigating

the disposition effect. Investors may be compelled to revisit

their beliefs after incurring large losses quickly.

KEYWORDS

disposition effect, individual investors, recency, reverse disposition

effect, salience, valuation uncertainty, volatility

JEL CLASSIFICATION

G11, G41

1INTRODUCTION

How do gains and losses affect investors’ decisions to sell stocks? Do reactions to gains and losses change with the

passage of time? Researchers have extensively discussed investors’tendency to hold losers and sell winners, known

as the disposition effect (Shefrin & Statman, 1985). More recently,Kaustia (2010) shows that the probability of selling

a stock jumps at zero return, and Ben-David and Hirshleifer (2012) find that the magnitude of percentage returns

affects the sale decision. Lastly, Hartzmark (2015) reports that stocks with extreme (highest or lowest) percentage

returns within a portfolio are more likely to be sold.

Researchers mostly use percentage returns (usually since acquisition) to study the effects of gains and losses on

investors’ trading decisions. We propose that using nominal gains and losses is more appropriate. Nominal value

changes, unlike percentage returns, represent direct changes in the investor wealth. In addition, making selling deci-

Financial Review. 2022;57:397–427. wileyonlinelibrary.com/journal/fire ©2021 The Eastern Finance Association 397

398 KOTOMINAND VARMA

sions based on individual positions’ percentage returns requires investors to engage in an extreme form of narrow

framing (Kahneman & Lovallo, 1993; Tversky& Kahneman, 1981) or mental accounting (Thaler, 1985), wherein they

look at each position in isolation and ignore the rest of their portfolio. However,investors should reduce the tendency

to engage in extreme narrow framing when they simultaneously observe (e.g.,on an account statement or snapshot)

values, gains, and losses for the entire portfolio in addition to individual positions. Viewing this information together

primes investors to consider their total brokerage portfolio value in addition to or instead of percentage returns on

individual positions.1Thus, we hypothesize that nominal gains and losses can more robustly explain stock sales than

percentage returns. We construct measures of nominalgains and losses scaled by an investor’s portfolio value that we

label the scaled nominal gain (SNG) and the scaled nominal loss (SNL).

We establish two new stylized facts when we employ SNG and SNL as explanatory variables in studying individ-

ual investors’ stock-selling behavior.First, consistent with our hypothesis, stocks with large nominal gains and losses

(scaled by the portfolio size) are more likely to be sold. The effects of nominal gains or losses on the probability of

a stock sale are more robust than the effects of percentage returns. This is consistent with the experimental finding

of Shavit et al. (2010) that people spend more time looking at nominal value changes than percentage returns. An

increase in SNG (SNL) of one percent is associated with an increase of 0.33 (0.55) percent in the probability of selling

a stock by an investor on a day on which the investor sells at least one stock. The higher propensity to sell positions

with larger gains and losses reflects the effect of salience on judgment in general (Taylor & Thompson, 1982) and on

stock-selling decisions (Barber & Odean, 2007; Ben-David & Hirshleifer,2012). Investors likely pay more attention to

such positions and research them more.

Theimpact on SNG and SNL on selling decisions is observed across different account types and investor subsamples

and survives several other robustness checks.2When we control for ranks of SNGs and SNLs within a portfolio to

address the importance of the rank effect in the context of stock selling decisions (Hartzmark, 2015), unlike SNG,

SNL remains statistically significant. Thus, the size of a loss relative to the portfolio size affects stock sales beyond

the ranking of losses within the portfolio. Finally, the impact of SNL on sales is lower for stocks with high valuation

uncertainty,which is consistent with Kumar’s (2009a) observation of individuals’ preferences for stocks with lottery-

likefeatures. The higher perceived probability of price recovery for stocks with high valuation uncertainty may prompt

investors to hold on to such stocks evenwhen they accumulate large nominal losses.

Second, for stocks with short holding periods investors are more likely to realize large nominal losses compared

to gains of the same magnitude, mitigating the disposition effect. These effects are strong for holding periods ranging

from a few days to severalmonths and monotonically taper off as the holding period lengthens. They extend to about

15months in taxable a ccounts and up to 3 months in tax-deferred accounts. Forexample, for holding periods below 0.5

months, 0.5–1 months, and 3–6 months the probability of selling losers is higher than that of selling winners (i.e., the

disposition effect is eliminated) when SNL and SNG are above 2.68%, 3.90%, and 21.50%, respectively.While tax-loss

selling is important, it does not fully explain the elimination of the disposition effect for large nominal gains/losses at

shortholding periods as a less pronounced pattern is also present in tax-deferred a ccounts. While our observedeffects

for SNL and SNG over different holding periods are impervious to the valuation uncertainty of stocks, consistent with

our prior findings, the impact of SNL is notably lower for stocks with high valuation uncertainty.

The impact of holding periods on individual investors’ selling decision has also been considered in recent studies.

Specifically, Ben-David and Hirshleifer (2012) find that at short holding periods, investorsare more likely to sell big

(percentage) losers than small ones. Kaustia (2010) finds that not only is the propensity to sell approximatelyconstant

1People often make different decisions when considering information together rather than on a stand-alone basis as they start focusing on differences

betweenavailable alternatives (Bazerman et al., 1992; Hsee, 1996; Kahneman, 2003;List,2002).

2Forexample, one concern about using SNG and SNL is that investors may frame their gains and losses in the context of their total wealth rather than their

brokerageaccount value. To alleviate it, we run our tests for subsamples of investorswith different portfolio value ranges, likely representing different levels

of wealth. As robustness checks, we use an investor’s averagebrokerage portfolio value throughout the entire sample period or an investor’s self-reported

networth instead of the previous day’s portfolio value as the base for computing SNG and SNL. The results remain the same: the probability of a stock saleon

aday on which a given household sells at least one stock still increases with both SNG and SNL.