![Sample our Economics, Finance,Business & Industry journals, sign in here to start your access, latest two full volumes FREE to you for 14 days]({"type":"image" , "src" : "/sda/5931/TOC-Subject-Sample-2021-Economics-Finance-Business-Industry.jpg"})
Abstract
We conduct two experiments to examine potential causes of the disposition effect. In Experiment 1, we rule out beliefs in mean reversion as a cause of the disposition effect. Although a belief in the mean reversion of stock prices should be independent of whether an investor owns or only follows the stock, we show only investors who own the stock behave as though prices will reverse. In Experiment 2, participants buy and sell securities over multiple periods. We find that self-regard and investing confidence (two types of self-esteem) have opposing influences on investors’ tendency to hold losing investments. Investors with lower self-regard hold losing investments longer than those with higher self-regard, and investors with higher confidence hold losing investments longer than those with lower confidence. We focus on investors’ tendency to hold losing stocks too long because prior research suggests the gain versus loss sides of the disposition effect are driven by different biases.
ACKNOWLEDGMENTS
The authors appreciate the programming assistance of Lisa Harris and Ron Harris.
FUNDING
This research benefited from a Dean's Research Grant from the Goizueta Business School at Emory University.
Notes
1. Weber and Camerer [1998] provide experimental evidence corroborating the findings of archival studies on the disposition effect. In their study, participants in two experimental conditions traded shares in six securities with different probabilistically determined price paths. In the first condition, participants could keep their shares from round to round. In the second condition, participants were forced to sell shares from the prior round at the start of a new round. However, in each new round in the second condition, participants could buy back stocks they were forced to sell in the previous round. Doing so would be economically equivalent to holding the security from one round to the next as participants were allowed to do in the first condition. Consistent with the disposition effect, the tendency to hold losing stocks (or buy them back after forced liquidation) was greater in the first condition than in the second. Because these results were produced in a controlled experiment without tax implications or other market influences, they provide strong support for the idea that this trading pattern cannot be explained by simple preferences for wealth maximization.
2. Hartzmark and Solomon [2010] study the trading decisions of “investors” betting on Tradesport.com, an online market for gambling contracts. They assume that investors in this contract market are less likely to be risk averse, and may actually be more risk seeking, than investors in more typically studied markets.
3. Our data allow for an investigation of both sides of the disposition effect. We provide supplemental analysis of the gain side of the disposition effect.
4. While common usage of “self-regard” implies a focus on one's own needs to the exclusion of others’ (i.e., egotism or narcissism), we follow the prior literature in defining self-regard as an evaluative view of the self (positive versus negative), without the negative connotations (e.g., see Cohen et al. [2000], Heine et al. [1999]).
5. We also manipulated whether participants in the current investor conditions had an opportunity for self-affirmation (Sherman and Cohen [2006]). Participants answered a set of questions designed either to make them reflect on their own values (affirmation condition) or on items they did not value very highly (no affirmation condition) (Sivanathan et al. [2008]). Manipulation checks indicate that this manipulation failed. Further, this manipulation had no effect on subsequent investment decisions. Thus, we do not discuss this manipulation further.
6. The full model containing all interactions involving gender did not converge to stable parameter estimates due to insufficient cell sizes. To address this issue, we deleted the three-way interaction of gender, price change, and investor status. We report results using the reduced model, which produces stable parameter values.
7. As in Experiment 1, we also included a between-participants manipulation of self-affirmation. This manipulation failed. Conclusions are not affected by the inclusion of this manipulated variable in our analyses, and so we omit it.
8. We could not use previously validated self-esteem scales to measure self-regard and confidence because our proposed division of self-esteem into self-regard and confidence is not contemplated by those scales. We relied on brief, single-item measures because they had successfully been used in prior research and because the timing of the measures (immediately prior to the trading task) necessitated that they be brief.
9. This assumption is similar to that used in microstructure models such as Kyle [1985] and in experimental trading markets (Bhojraj et al. [2009], CitationBloomfield, Tayler, and Zho [2009]).
10. This is a simple weighted average based on the price of the securities at the time they were purchased. For example, the participant may have bought 12 shares of Security A for $43 per share, then purchased 6 more shares at the current price of $49 per share. The weighted average would then be ([12 x $43] + [6 x $49])/18 = $45.
11. Participants could calculate gains and losses on traded securities based on a weighted-average method (as displayed in our trading interface), or based on first in first out (FIFO), last in first out (LIFO), or some other share tracking method. To reduce idiosyncratic differences in how a participant identifies a gain or loss, we provide weighted-average information and calculate realized and unrealized gains accordingly. If participants implement an inventory management method other than the one provided in the trading interface, this will work against our finding any evidence of the disposition effect. However, we have no reason to expect differences in (unobservable) individual inventory tracking methods to vary systematically with our other variables of interest.
12. Our predictions (Hypotheses 2 and 3) relate participants’ self-regard and confidence measures at the time they are making investment decisions to those investment decisions. Therefore, it was important that we collect their assessments after they experienced the practice round (and could assess the threat) and immediately before engaging in the trading task. Because self-regard is relatively stable, we did not expect participants’ experiences in the practice round would affect their self-regard. However, experience in the practice round could affect their confidence in their ability to do the task. This does not pose a problem for our theory, which is about task-specific confidence, regardless of the source of that confidence. Indeed, we found that performance in the practice round (as measured by assets held at the end of practice) was uncorrelated with self-regard (r = 0.04, p = 0.717) and modestly correlated with confidence (r = 0.20, p = 0.066).
13. Not revealing the exchange rate parameters (Y = −9400, X = 0.01) does not reduce the salience of our compensation (Smith [1982]) because participants are informed that the exchange rate is positive such that improved performance will lead to a greater final payoff. CitationBloomfield, O'Hara, and Saar [2009] demonstrate that a compensation scheme similar to the one used in our study generates behaviors almost identical to those observed with more traditional compensation schemes.