Regret, Pride, and the Disposition E ect

Transcription

1 Regret, Pride, and the Disposition E ect Alexander Muermann y Jacqueline M. Volkman z October 27 Abstract Seeking pride and avoiding regret has been put forward as a rationale for the disposition e ect, the tendency of investors to sell winning stocks and hold on to losing stocks. We develop a dynamic portfolio choice model that incorporates anticipated regret and pride in individual s preferences to formally investigate this explanation. We specify a setting for which this rationale is satis ed, but show that it is not robust to variations in the underlying assumptions. JEL Classi cation D8, G Keywords disposition e ect, regret theory, dynamic portfolio choice We wish to thank Neil Doherty, Olivia Mitchell, and Harris Schlesinger for their valuable comments. Muermann gratefully acknowledges nancial support of the National Institutes of Health National Institute on Aging, Grant number P3 AG2836, the Boettner Center for Pensions and Retirement Security at the University of Pennsylvania, and National Institutes of Health National Institute of Child Health and Development Population Research Infrastructure Program R24 HD-44964, all at the University of Pennsylvania. Volkman gratefully acknowledges nancial support from the Insurance and Risk Management Department at the Wharton School through a Bradley Foundation Scholarship. y Muermann: Institute of Risk Management and Insurance, Vienna University of Economics and Business Administration, Nordbergstraße 5, A-9 Wien, AUSTRIA, alexander.muermann@wu-wien.ac.at z Volkman: Schools of Business, Fordham University, 79 Broadway, Room 327, New York, NY 9, USA, volkman@fordham.edu

2 Introduction In nancial markets, there is a unique phenomenon where investors appear reluctant to realize losses and eager to realize gains; that is, investors seem to have a preference for selling winning stocks too early and holding losing stocks too long. This pattern has been labeled the disposition e ect by Shefrin and Statman (985) and cannot be explained by traditional trading theories. For instance, Odean (998) found this e ect even after accounting for portfolio rebalancing and trading costs. Similarly, Lakonishok and Smidt (986) and Ferris et al. (987) consider trading volume and nd that the disposition e ect dominates tax-related motives for selling stocks at a loss. The disposition e ect has also been discovered in the Finnish stock market (Grinblatt and Keloharju 2), the Finnish apartment market (Einio and Puttonen 26), the Taiwanese stock market (Barber et al. 26), in the exercise of company stock options (Heath, Huddart, and Lang 999), and in the sale of residential housing (Genesove and Mayer 2). Furthermore, it has been found for professional investors at an Israeli brokerage house (Shapira and Venezia 2) and for professional futures traders (Locke and Mann 2); although, Shapira and Venezia (2) and Dhar and Zhu (26) nd that investors with less trading experience exhibit a stronger disposition e ect. Experimental evidence has further supported the disposition e ect (Weber and Camerer 998, Andreassen 988). We refer to Barber and Odean (25) for a more in-depth review of this phenomenon. Shefrin and Statman (985) proposed several explanations for the disposition e ect, including loss aversion, mental accounting, seeking pride and avoiding regret, and self control. Much of the literature to date on the disposition e ect has focused on loss aversion as an explanation. Hens and Vlcek (25) and Barberis and Xiong (26) recently formally analyzed this reasoning and found that loss aversion mostly does not predict the disposition e ect. We describe their ndings in further detail below. In this paper, we formally model the anticipation of regret and pride in a dynamic setting to see if it can cause investors to optimally follow a strategy in which they sell winning stocks and hold losing stocks; that is, we examine if seeking pride and avoiding regret can explain the disposition e ect. In addition to Shefrin and Statman (985), several empirical papers documenting the disposition e ect argue that loss aversion explains this phenomenon (Odean 998, Lakonishok and Smidt 986, Ferris et al. 987, Grinblatt and Keloharju 2, Shapira and Venezia 2, Coval and Shumway 22, Dhar and Zhu 26). Kahneman and Tversky (979) rst proposed loss aversion as part of prospect theory, arguing that people make decisions considering gains and losses relative to some reference point rather than to wealth levels. Individuals who are loss-averse have preferences that are risk-seeking over losses and risk-averse over gains, and are more sensitive to losses than to equivalent gains. The intuition behind why loss aversion can explain the disposition e ect is as follows. Since a winning stock is considered a gain, and individuals are risk-averse in this domain, they will sell the stock. On the other hand, a losing stock is considered a loss, and being risk-seeking in this domain, investors will hold the stock. Most previous studies that consider the disposition e ect are empirical and list loss aversion 2

3 as an explanation for the e ect. preferences in portfolio choice problems. More recently, a few papers have formally modeled loss averse loss-averse investors would be consistent with the disposition e ect. Gomes (25) nds that the optimal portfolio choice for Kyle et al. (26) examine the liquidation decision of a project with loss-averse preferences and also nd optimal behavior that is consistent with the disposition e ect. Yet, neither paper considers the initial decision. That is, the investor is endowed with the stock or the project and whether or not a loss-averse investor would initially buy the stock or invest in the project is not considered. In contrast, Hens and Vlcek (25) and Barberis and Xiong (26) take the initial decision into consideration and mostly nd that loss aversion cannot explain the disposition e ect with short time horizons. The equity premium must be so high for loss-averse investors to initially invest in the stock that subsequent optimal behavior is not consistent with the disposition e ect. In fact, Barberis and Xiong (26) show that this often implies momentum trading by the investor, which is the opposite of the disposition e ect: keeping winning and selling losing stocks. Furthermore, Barberis and Xiong (26) demonstrate that investors are only willing to invest initially in the stock and exhibit optimal behavior that is consistent with the disposition e ect when stock returns are negatively skewed and/or the curvature of the value function is su ciently smaller than suggested by Tversky and Kahneman (992) and/or the number of trading periods is large and the equity premium is su ciently low. As the literature illustrates, it is only under speci c conditions that loss averse preferences imply investment behavior consistent with the disposition e ect. Another explanation for the disposition e ect suggested by Shefrin and Statman (985) and formally examined in this paper is the feeling of regret and pride, which has recently been supported with experimental evidence (O Curry Fogel and Berry 26). The intuition is that an investor who regrets an investment in a stock that has lost value will hold the stock because he hopes that the stock price will rise in the next period, enabling him to avoid regret. If the stock has risen in value, however, the investor wants to feel pride in having made such a good investment and therefore sells the stock; if he had held it and then the price fell, he would have foregone feeling pride. Wanting to feel pride and delay regret is what causes investors to realize gains more quickly than losses. 2 Although the explanation seems intuitive, as it seems with loss aversion, it is not as straightforward to argue that preferences including regret and pride would give rise to the disposition e ect in a dynamic setting. For instance, an investor may sell a stock after it rises over one period, but if the stock rises again over the following period, he will feel regret from having sold the stock too early. Therefore, anticipating regret over both periods, in this instance, could cause the investor to hold the stock after the rst period, or to not invest in the stock in the rst place. To the best of our knowledge, to date there has been no formal analysis of whether preferences including regret and pride can predict the disposition e ect. In this paper, we develop such a Hens and Vlcek (25) consider a two-period model with a myopic investor whereas Barberis and Xiong (26) examine a general multi-period model with a dynamically optimizing investor. 2 Barber et al. (24) nd empirical evidence and Weber and Welfens (26) nd experimental evidence of an investor s tendency to repurchase or purchase additional shares of stocks that have recently lost value. In both papers, the authors argue that repurchase decisions could be driven by counterfactuals which might give rise to feelings of regret. 3

4 model. Considering a dynamic setting with regret and pride raises some interesting questions and thus requires certain assumptions to be made. For instance, does the investor experience future regret or pride only for the current investment decision or do such feelings result from all decisions made in the past? When does the investor experience regret at the nal period or during intermediate periods? In a dynamic setting, some decisions will elicit regret and others pride. How do these feelings interact and compound over time? Furthermore, if the investor does not hold the stock, does he know how it performs, and can he experience regret then from not holding it if it does well (or pride if it performs poorly)? In what follows, we will state and explain the assumptions that underlie the model for how anticipation of regret and pride causes individuals to sell stocks that have gained recently and hold stocks that have lost. We conclude that, in a speci c setting, the disposition e ect can occur if investors experience regret and pride over their investment decisions; yet, as with loss aversion, this result is not robust to numerous variations of this setting. The paper is structured as follows. In the next section, we introduce the model, the assumptions, and preferences that allow individuals to consider regret and pride. In Section 3, we examine the optimal portfolio choice problem and provide su cient conditions for the investor s optimal strategy to be consistent with the disposition e ect. In Section 4, we discuss the robustness of our assumptions. Finally, we conclude in Section 5. 2 Model and Preferences Regret is an individual s ex-post feeling that his ex-ante decision turned out to be suboptimal with respect to the resolved uncertainty; that is, the individual s ex-post level of wealth could have been higher had he made a foregone alternative decision. Equivalently, pride is the ex-post feeling that the ex-ante decision turned out to be better than some foregone alternative decision. In this setting, an individual makes a decision considering the anticipated disutility or additional utility derived from regret or pride. Regret theory was initially formulated by Bell (982) and Loomes and Sugden (982) and has been shown in both the theoretical and experimental literature to explain individual behavior. Bell (982) depicted how regret could explain preferences for both insurance and gambling. More recently, the impact of regret on decision making has been examined in a static framework for di erent scenarios, including the demand for insurance (Braun and Muermann 24), portfolio choice (Muermann et al. 26), asset pricing and portfolio choice in an Arrow-Debreu economy (Gollier and Salanié 26), currency hedging (Michenaud and Solnik 26), and rst price auctions (Filiz and Ozbay 27). We contribute to this literature by considering a dynamic portfolio choice problem. This framework allows us to capture an investor s dynamically optimal response to realized portfolio returns and feelings of regret and pride implied by such realized returns. In the following, we introduce a model that is simple yet rich enough to capture these issues. There are two assets: a risk-free asset (bond) with a zero normalized return and a risky asset 4

5 (stock) with a stochastic return ~x t per period. In this paper, we consider only one risky asset in order to be consistent with the mental accounting framework noted by Thaler (985) and supported by Gross (982): the idea is that decision makers di erentiate gambles into separate accounts, applying their preferences to each account, and ignoring the interaction between them. In this manner, investors would view each stock they hold individually, and therefore, we only consider one. We assume that the risky returns are independent and identically distributed across periods and take the two values x + > > x with probability p and p in each period. The individual is endowed with initial wealth w and can only invest all of his wealth in one of the two assets. There are two periods. At t = the investor decides whether to invest his wealth, w, in the stock or bond. At t = the investor observes his realized level of wealth, w, and decides again whether to invest it in the stock or bond. At t = 2 all assets are liquidated and the investor observes and consumes his nal level of wealth, w 2. The restriction that the individual cannot split his wealth between the two assets is consistent with the discussion and analysis in Shefrin and Statman (985) and Odean (998), which are based on stock trading records of individual investors. Alternative settings include the purchase and sale of an indivisible asset such as housing, or the investment in and liquidation of a project (e.g., Kyle et al. 26, who consider such decisions with loss-averse preferences). Our model thus speaks to the empirical evidence of the disposition e ect in the real estate market provided by Genesove and Mayer (2) and Einio and Puttonen (26). We follow Bell (982, 983) and Loomes and Sugden (982) by implementing the following two-attribute utility function to incorporate regret and pride in investors preferences v (w) = u (w) kg(u(w alt ) u (w)). () The rst attribute represents the individual s level of risk-aversion and is characterized by the individual s utility function of actual level of wealth, w. We assume that the utility function u () exhibits CRRA preferences, i.e. u (w) = w (2) for 6= and u (w) = ln (w) for =, where is the coe cient of relative risk aversion. This implies that the time horizon has no e ect on the optimal portfolio allocation of an individual who does not consider regret and pride in his decision. That is, the individual makes his decision as if he was myopic. Additionally, we assume that stock returns satisfy p ( + x + ) + ( p) ( + x ) > if < p ( + x + ) + ( p) ( + x ) < if > ( + x + ) p ( + x ) p > if = (3) This assumption implies that the risk premium is high enough such that an individual who does not consider regret and pride nds it optimal to invest in the stock in all periods. Thereby we 5

6 exclude portfolio rebalancing as an explanation for the disposition e ect which allows us to focus on how regret and pride in uence the optimal portfolio allocation. The second attribute represents the individual s feeling of regret or pride toward the ctitious level of wealth, w alt, the individual would have obtained from a foregone alternative. If the actual level of wealth, w, falls below the alternative level of wealth, w alt, the individual regrets his decision; otherwise the individual feels pride. The function g () measures the amount of regret and pride that the investor experiences, and we assume that it is increasing and convex with g () = ; that is, the individual weighs the disutility incurred from regret relatively more than the additional utility derived from pride. This assumption is supported in the literature (Thaler 98, Kahneman and Tversky 982) and has recently found experimental support (Bleichrodt et al. 26). The parameter k measures the relative importance of the second attribute, regret, to the rst attribute of the utility function. We assume that the individual incurs the disutility or additional utility from regret or pride only in the nal period. Similar to the assumption that there is no intermediate consumption, we assume that the individual does not incur regret or pride in intermediate periods. The investor thus makes his portfolio choice by maximizing his expected utility of terminal wealth using the value function v () given in equation (). We make the following two additional assumptions, which turn out to be crucial for predicting that regret and pride cause individuals to behave according to the disposition e ect. In Section 4, we will discuss how deviations from these assumptions impact our results. Assumption The individual only observes the realized stock return if he holds the stock. This assumption is relevant for regret-averse individuals as foregone alternatives and their resolution can impact decisions. In our setting, this assumption implies that the individual has the option to avoid regret or forego pride by investing in the bond and not observing the realized return of the stock (by not reading the newspaper). This relates to Bell (983), who shows that it can be optimal for a regret-averse individual to not have a foregone alternative lottery resolved. In fact, we will show in Section 4 that observing stock returns after selling the stock yields a lower level of expected utility for the investor. This implies that if the individual has the choice to observe or not observe stock returns, then it is optimal in our setting for him not to observe them. Assumption 2 If the individual s decisions turn out to be ex-post optimal (i.e., they imply the maximum level of wealth with respect to the realized returns), then he experiences pride toward the foregone worst alternative (FWA), the lowest level of wealth he could have obtained with respect to the realized returns. If the individual s choices turn out to be ex-post suboptimal, then he incurs regret toward the foregone best alternative (FBA), the level of wealth he would have obtained from the ex-post optimal choices. We assume the investor feels regret/pride for all past decisions including the current one; that is, the FWA and FBA are derived with respect to all decisions up to and including the current one. 6

7 This assumption addresses the issue of how the feeling of regret and pride interact and accumulate over time. A decision rule might turn out to be optimal over the rst period but suboptimal over the second period. Here, we assume that the feeling of regret is stronger than pride in the sense that the individual incurs regret as long as one decision turns out to be ex-post sub-optimal; put di erently, the individual incurs pride only if all decisions turn out to be ex-post optimal. In that case, we assume that his additional utility from pride is measured in reference to the FWA. 3 Optimal Portfolio Choice and the Disposition E ect In this section, we examine how an individual who is prone to feelings of regret and pride makes decisions in a dynamic portfolio choice problem. In the rst subsection, we investigate the optimal decision at t = under the assumption that the individual invested in the stock at t =. show that, under certain conditions, the disposition e ect can emerge as the optimal strategy: conditional on the stock gaining value over the rst period, it is optimal to sell the stock at t =, and conditional on the stock losing value over the rst period, it is optimal to hold the stock over the second period. In the second subsection, we solve for the dynamically optimal choice including the initial decision at t = and show that, under the same conditions, the disposition e ect can emerge as an optimal strategy. That is, it can be optimal for the investor to buy the stock at t =, and then sell it at t = if it went up or hold it if it went down over the rst period. 3. Portfolio Choice at t = We assume that the investor bought the stock at t =, i.e. his level of wealth at t = is given by ~w = w ( + ~x ), which can take the two values w + = w ( + x + ) > w or w = w ( + x ) < w depending on whether the stock went up or down over the rst period. The following proposition determines the condition under which it is optimal for the individual to follow the disposition strategy. Proposition Suppose the individual bought the stock at t =. We It is then optimal for the individual with 6= at t = to sell the stock if it went up and to keep the stock if it went down over the rst period if and only if stock returns satisfy the following two conditions < pkg (w ( + x + )) w p + x + + ( p) + x + kg + x + 2( )! + ( p) kg (w ( + x + )) w + x +! + x! (4) 7

8 and > pkg (w ( + x )) (w ( + x + )) p + x + + ( p) + x + kg + x! + ( p) kg For =, i.e. u (w) = ln (w), the conditions are ln and ln w w + x! + x 2( )!. (5) + x + p + x p +kg ln + x + < pkg 2 ln + x + +( p) kg ln + x + x + p + x p +kg ln + x > pkg ln + x +( p) kg 2 ln + x. Proof. See Appendix A.. Condition (4) implies that selling the stock after it went up over the rst period is optimal. depends on the equity premium and the relative strength of the certain feeling of pride when selling the stock versus the uncertain feeling of additional pride or regret when holding the stock over the second period. period is optimal. Condition (5) assures that keeping the stock after it goes down over the rst Again, this condition depends on the equity premium and the relative strength of the certain feeling of regret when selling the stock versus the uncertain feeling of additional or less regret when holding the stock over the second period. with the disposition e ect if both conditions hold. (6) (7) Optimal behavior is thus consistent In the following lemma we show that condition 3 on stock returns does not further restrict the set of stock returns that are consistent with the disposition e ect. It Lemma Suppose that stock returns satisfy condition 5. Then condition 3 holds. Proof. See Appendix A.2. We provide intuition for why both conditions can hold by focusing on the e ects of the following two interrelated changes to the optimal choice: increasing the equity premium and adding anticipated regret and pride to the decision making, i.e. the second attribute of the utility function v () in equation (). Considering the rst attribute of the utility function, u (), increasing the equity premium makes keeping the stock at t = more attractive, independent of the stock s movement over the rst period. For a xed equity premium, adding the second attribute, kg(u(w alt ) u (w)), makes selling the stock at t = more attractive, again independent of the stock s movement over the rst period. This is due to the convexity of g (), as the individual prefers a certain level of regret and pride by selling the stock to an uncertain exposure to regret and pride by keeping the stock. In other words, an individual who considers anticipated regret and pride in his decision 8

9 making requires a higher equity premium for keeping the stock than an individual whose preferences do not include these psychological factors. The crucial e ect that implies di erent behavioral responses to the stock s movement over the rst period arises from the di erent e ects that a marginal increase in the equity premium has on regret and pride, i.e. on the second attribute of the utility function. If the stock went down over the rst period, then the individual will always feel regret and never pride, as he already made a suboptimal decision at t = (see Assumption 2). If the stock went up, however, the individual will either only feel pride if he sells the stock or he will expose himself to either regret or even greater pride if he keeps the stock. Marginally increasing the equity premium will impact the di erences in the marginal e ects between selling and keeping the stock. If this di erence is larger after an up-move of the stock over the rst period than after a down-move, then selling the stock would be marginally more attractive after an up-move, which can imply optimal behavior consistent with the disposition e ect. To focus on those e ects, suppose that =, i.e. u (w) = ln (w), p = =2, and that we increase the equity premium by marginally increasing the positive return, x +, while keeping the negative return, x, xed. 3 If the stock went down over the rst period, there is no e ect on the impact of the second attribute, as it solely depends on x (see lower limit in condition (7)). Conversely, if the stock went up, the convexity of g () implies that the marginal increase in pride when selling the stock is larger than the potential marginal increase in pride when keeping the stock (see condition (6)). Thus, increasing the equity premium makes selling more attractive than buying when the stock has gone up over the rst period, while it has no e ect when the stock has gone down. As a result, there arises a situation in which anticipated regret and pride induces behavior that is consistent with the disposition e ect. These marginal e ects on the psychological factors have to be traded-o against the marginal e ect that an increase in the equity premium has on the rst attribute, i.e. the increased attractiveness of keeping the stock. Condition (4) implies that the risk premium is low enough such that the bene t of securing pride at t = by selling the stock outweighs the risky prospect of feeling even more pride at the cost of potentially incurring regret when keeping the stock. Condition (5) implies that the risk premium is high enough to compensate the individual for the risky prospect of feeling less regret at the cost of feeling even more regret when keeping the stock. In Section 3.3, we will show with an illustrative example that both conditions (4) and (5) can be satis ed and how they vary with the degrees of risk aversion and regret. 3.2 Dynamic Portfolio Choice In this section, we examine the dynamically optimal behavior of the individual with the preferences speci ed above. We thus endogenize the decision at t =. Let us emphasize again that this proved to be crucial for explaining the disposition e ect by loss aversion. Although loss aversion 3 Equivalent results obtain for 6= and/or increasing the equity premium by marginally increasing x while keeping x + xed. 9

10 can explain the disposition e ect conditional on having bought the stock at t = (Gomes 25, Kyle et al. 26), it can only explain the disposition e ect when the initial purchasing decision is endogenized for certain settings (Hens and Vlcek 25, Barberis and Xiong 26). We show in the following proposition that regret and pride can explain dynamic trading behavior which is consistent with the disposition e ect; that is, buying the stock at t = and selling (keeping) the stock after it went up (down) over the rst period. Furthermore, the necessary and su cient conditions on stock returns for the disposition e ect to hold in this dynamic setting are equivalent to the necessary and su cient conditions, (4) and (5), derived in the previous section. Proposition 2 It is optimal for the individual at t = to buy the stock and at t = to sell the stock if it went up and to keep the stock if it went down over the rst period if and only if stock returns satisfy conditions (4) and (5) when 6=. Proof. See Appendix A.3. For =, conditions (6) and (7) must hold. Therefore, in a dynamic portfolio choice problem, for a certain range of stock returns, i.e. under conditions (4) and (5), it is optimal for an investor who is prone to feeling regret and pride to follow the disposition e ect strategy. Note that the individual s behavior is time-consistent as he optimally plans at t = to follow the disposition e ect strategy at t = (Proposition 2) and at t = optimally executes this strategy (Proposition ). 3.3 Illustrative Example The objective of providing an illustrative example is to show how the set of stock returns that satisfy the necessary and su cient conditions (4) and (5) varies with the degrees of risk aversion and regret. Suppose that returns are not skewed, i.e. p = 2, w =, and that the function g () is given by g(x) = exp(x). Figure plots for k = three pairs of lines associated with three levels of coe cient of relative risk aversion: = thin lines, = 2 dashed lines, = 3 thick lines. Figure 2 plots for = three pairs of lines associated with three levels of degree of regret: k = thin lines, k = 2 dashed lines, k = 3 thick lines. 4 The lower line of each pair represents all levels of stock returns y = x + and x = x such that the lower constraint is binding. Analogously, the upper line of each pair represents the upper constraint. Thus, for any pair of stock returns (y; x) that falls between those two lines, the individual optimally follows the disposition strategy. 5 Otherwise, for any pair of stock returns (y; x) that is below the lower line it is optimal at t = to sell the stock independent of the stock s movement over the rst period. Equivalently, for any pair of stock returns (y; x) that is above the upper line it is optimal at t = to buy the stock independent of the stock s movement over the rst period. 4 Alternatively, the intensity of regret can be varied by the convexity of g. We obtain similar results for the less convex function g (x) = x 2. 5 We did not plot condition 3 as it does not further restrict the set of stock returns that are consistent with the disposition e ect (see Lemma ).

11 We note that increasing the degree of risk aversion and the intensity of regret has opposing e ects on the necessary and su cient conditions (4) condition (5). Condition (4) for selling the stock after it went up becomes less stringent while condition (5) for keeping the stock after it went down becomes more stringent. The intuition is that increasing the degree of risk aversion increases the risk premium required for keeping the stock and increasing the intensity of regret increases the compensation for the risky prospect of regret and/or pride when keeping the stock. Therefore, selling the stock becomes more attractive x y Figure : This graph plots for p = =2, w =, k =, g(x) = exp(x), and various coe cients of relative risk aversion ( = thin lines, = 2 dashed lines, = 3 thick lines) the constraints on stock returns in conditions (4) and (5) which are neccesary and su cient for the disposition e ect to hold.

12 x y Figure 2: This graph plots for p = =2, w =, =, g(x) = exp(x), and various degrees of regret (k = thin lines, k = 2 dashed lines, k = 3 thick lines) the constraints on stock returns in conditions (4) and (5) which are neccesary and su cient for the disposition e ect to hold. 4 Discussion of Assumptions In this section, we discuss the importance of the assumptions made to explain the disposition e ect and give intuition about why deviations from those assumptions change the predictions. We focus on the portfolio choice at t =, as this is a necessary step in explaining the disposition e ect in a dynamic setting. The disposition e ect at t = would be reinforced by those changes in the assumptions that make selling the stock more attractive after it went up and make holding the stock more attractive after it went down over the rst period. Assumption The rst assumption considers whether the individual observes a stock s returns even if he does not hold it in his portfolio. By comparing the levels of expected utility as in the proof of Proposition, it can be shown that observing stock returns implies the opposite optimal decision after the stock went up, i.e. it 2

13 becomes optimal for the individual to keep the stock. This holds for any deviations in Assumption 2 that we discuss below. The intuition behind this result is as follows. Suppose the individual observes stock returns after selling the stock. Since he will observe the realization of the foregone alternative, he is exposed to a spread in feelings of regret and pride over the next period. As the function g () is convex, the individual s level of expected utility is lower when he is exposed to this spread compared to the situation in which he does not observe stock returns after selling and is thereby not exposed to this spread. Note that when holding the stock, the individual necessarily observes stock returns as they impact his level of wealth. Hence, observing stock returns makes selling less attractive and leads to the opposite optimal decision after the stock went up over the rst period, i.e. it is not optimal to follow the disposition strategy. This also implies that if the individual has the choice to observe stock returns or not, then it is optimal in our setting for him to not observe them and follow the disposition strategy under conditions (4) and (5). This relates to the result of Bell (983), who shows that it can be optimal for a regret-averse individual, i.e. with a convex function g (), to not have a foregone alternative lottery resolved. Remark 3 The importance of Assumption implies an interesting empirical prediction. Since it is more di cult to avoid news about the performance of stocks that receive media attention, our model predicts that the disposition e ect is more pronounced in the trading of stocks that receive little media coverage compared to those that receive a lot of media coverage, such as major stock market indices. Assumption 2 The second assumption relates to the reference level of wealth, w alt, toward which the individual feels regret or pride. In a dynamic setting, some decisions will elicit regret and others pride. This raises the interesting question how those feelings interact and aggregate. We assume that the individual only incurs feelings of pride if he has made choices that are all optimal after the fact. He then feels pride toward the FWA, which includes all decisions in the past and the current one. If one decision, either in the past or the current one, is sub-optimal, then the individual incurs regret toward the FBA. We discuss the following two deviations from Assumption 2 under both Assumption and its deviation. First, suppose the individual only considers regret in his decision making but not pride. Sugden (993) and Quiggin (994) provide an axiomatic foundation for regret in which the individual s disutility from regret depends only on the actual level of wealth and the level of wealth associated with the FBA. This change in assumption only potentially e ects the decision after the stock went up over the rst period as it is only then that the individual can incur pride. By comparing the levels of expected utility, it can be shown that by not considering pride, selling the stock becomes relatively less attractive compared to keeping the stock. Furthermore, this e ect implies that it is then never optimal to follow the disposition strategy at t =. 3

14 The intuition is that when keeping the stock, the individual only incurs pride if the stock went up over the second period. When selling the stock the individual incurs a certain level of pride (if he does not observe returns) or he incurs pride if the stock goes down over the second period (if he observes returns). In both cases, the convexity of g () implies that the ex-ante value of foregone pride is smaller when keeping the stock compared to the ex-ante value when selling the stock. Note that in the latter case in which the individual observes all stock returns, it is more valuable to incur pride when the stock goes down compared to when it goes up over the second period. Therefore, not considering pride makes selling the stock relatively less attractive compared to keeping it. Second, suppose that past decisions do not matter with respect to the anticipated feeling of regret or pride, i.e. at t = the individual only considers the current decision when evaluating those feelings and not his decision at t =. By comparing the levels of expected utility, it can be shown that if the individual only considers the current decision, selling the stock becomes relatively less attractive after it goes up but relatively more attractive after it goes down over the rst period compared to keeping the stock. Furthermore, this e ect implies that it is then never optimal to follow the disposition strategy at t =. The intuition behind this result is similar to above. After the stock goes up over the rst period, if not considering the pride from the initial decision at t =, relatively more pride is taken away when selling the stock compared to keeping it. As argued above, this is implied by the convexity of g (). However, after the stock goes down, the disutility from regret is larger when keeping the stock compared to selling it. Not considering regret from the initial decision at t = thus makes selling relatively more attractive. We conclude that these deviations from Assumptions and 2 make selling the stock less attractive after it goes up and potentially make keeping the stock less attractive after it goes down over the rst period. Those e ects work against the disposition strategy and imply its non-optimality. Assumptions and 2 are thus crucial for explaining the disposition e ect with investors feelings of regret and pride. Furthermore, the convexity of g () is also a crucial assumption in explaining how anticipated regret and pride can lead to the disposition e ect. 5 Conclusion Prior empirical analyses have shown that trading patterns in capital markets exhibit the disposition e ect, and current theoretical work has shown that loss aversion can only provide a rationale for this e ect in particular settings. In this paper, we formally investigate whether the avoidance of regret and the quest for pride leads to trading behavior that is consistent with the disposition e ect. As with loss aversion, we do nd speci c conditions under which the feelings of regret and pride are compatible with the disposition e ect; yet, this result is not robust to variations of the model. Extensions of this model include considering multiple time periods and/or divisible assets. Doing so, we believe, will add other e ects, but we argue that even with such extensions, investors e orts to avoid regret and seek pride will only explain the disposition e ect under certain conditions. 4

15 Understanding how regret and pride a ect investors trading behavior and possibly, the disposition e ect enables us to learn more about the potential costs these investors may incur, which is especially relevant for the current debate surrounding the introduction of Personal Retirement Accounts (PRAs) to the Social Security system. Shapira and Venezia (2) and Dhar and Zhu (26) have shown that investors with less trading experience and/or lower income exhibit a stronger disposition e ect, which may lead to lower after-tax returns. The introduction of PRAs would thus lead to a much more pronounced disposition e ect in capital markets and provides a rationale for policymakers to protect investors with such demographic characteristics. It is important to understand individuals trading behavior and the extent to which di erent factors a ect it, which we do here with regard to regret and pride. 5

16 References [] Andreassen, P. B Explaining the price-volume relationship: the di erence between price changes and changing prices. Organizational Behavior and Human Decision Processes [2] Barber, B. M., Y. Lee, Y. Liu, and T. Odean. 26. Is the aggregate investor reluctant to realize losses? Evidence from Taiwan. Working paper. [3] Barber, B. M. and T. Odean. 25. Individual investors. In Advances in Behavioral Finance Volume II. Ed. Richard H. Thaler. New York: Princeton UP, [4] Barber, B., T. Odean, and M. Strahilevitz. 24. Once burned, twice shy: naive learning, counterfactuals, and the repurchase of stocks previously sold. Working paper. [5] Barberis, N. and W. Xiong. 26. What drives the disposition e ect? An analysis of a long-standing preference-based explanation. Working paper. [6] Bell, D. E Regret in decision making under uncertainty. Operations Research 3(5) [7] Bell, D. E Risk premiums for decision regret. Management Science 29() [8] Bleichrodt, H., A. Cillo, and E. Diecidue. 26. A quantitative measurement of regret theory. Working paper. [9] Braun, M. and A. Muermann. 24. The impact of regret on the demand for insurance. Journal of Risk and Insurance 7(4) [] Coval, J. and T. Shumway. 25. Do behavioral biases a ect prices? Journal of Finance 6() -34. [] Dhar, R. and N. Zhu. 26. Up close and personal: investor sophistication and the disposition e ect. Management Science 52(5) [2] Einio, M. and V. Puttonen. 26. Disposition e ect in the apartment market. Working Paper. [3] Ferris, S. P., R. Haugen, and A. Makhija Predicting contemporary volume with historic volume at di erential price levels: evidence supporting the disposition e ect. Journal of Finance 43(3) [4] Filiz, E. and E.Y. Ozbay. 27. Auctions with anticipated regret: theory and experiment. American Economic Review 97(4) [5] Genesove, D. and C. Mayer. 2. Loss aversion and seller behavior: evidence from the housing market. Quarterly Journal of Economics [6] Gollier, C. and B. Salanié. 26. Individual decisions under risk, risk sharing, and asset prices with regret. Working paper. 6

17 [7] Gomes, F. 25. Portfolio choice and trading volume with loss-averse investors. Journal of Business 78(2) [8] Grinblatt, M. and M. Keloharju. 2. What makes investors trade? Journal of Finance 56(2) [9] Gross, L The art of selling intangibles: how to make your million ($) by investing other people s money. New York Institute of Finance, New York. [2] Heath, C., S. Huddart, and M. Lang Psychological factors and stock option exercise. Quarterly Journal of Economics 4(2) [2] Hens, T. and M. Vlcek. 25. Does prospect theory explain the disposition e ect? Working paper. [22] Kahneman, D. and A. Tversky Prospect theory: an analysis of decision under risk. Econometrica [23] Kahneman, D. and A. Tversky The psychology of preferences. Scienti c American [24] Kyle, A., H. Ou-Yang, and W. Xiong. 26. Prospect theory and liquidation decisions. Journal of Economic Theory 29() [25] Locke, P. and S. Mann. 2. Do professional traders exhibit loss realization aversion? Working paper. [26] Lakonishok, J. and S. Smidt Volume for winners and losers: taxation and other motives for stock trading. Journal of Finance 4(4) [27] Loomes, G. and R. Sugden Regret theory: an alternative theory of rational choice under uncertainty. Economic Journal 92(368) [28] Michenaud, S. and B. Solnik. 26. Applying regret theory to investment choices: currency hedging decisions. Working paper. [29] Muermann, A., O. S. Mitchell, and J. Volkman. 26. Regret, portfolio choice, and guarantees in de ned contribution schemes. Insurance: Mathematics and Economics 39(2) [3] O Curry Fogel, S. and T. Berry. 26. The disposition e ect and individual investor decisions: the roles of regret and counterfactual alternatives. Journal of Behavioral Finance 7(2) 7-6. [3] Odean, T Are investors reluctant to realize their losses? Journal of Finance 53(5) [32] Quiggin, J Regret theory with general choice sets. Journal of Risk and Uncertainty 8(2) [33] Shapira, Z. and I. Venezia. 2. Patterns of behavior of professionally managed and independent investors. Journal of Banking and Finance 25(8)

18 [34] Shefrin, H. and M. Statman The disposition to sell winners too early and ride losers too long: theory and evidence. Journal of Finance 4(3) [35] Sugden, R An axiomatic foundation of regret. Journal of Economic Theory 6() [36] Thaler, R. 98. Toward a positive theory of consumer choice. Journal of Economic Behavior and Organization [37] Thaler, R Mental accounting and consumer choice. Marketing Science 4(3) [38] Tversky, A. and D. Kahneman Advances in prospect theory: cumulative representation of uncertainty. Journal of Risk and Uncertainty 5(4) [39] Weber, M. and C. F. Camerer The disposition e ect in securities trading: an experimental analysis. Journal of Economic Behavior and Organization [4] Weber, M. and F. Welfens. 26. The repurchase behavior of individual investors: an experimental investigation. Working Paper. 8

19 A Appendix: Proofs A. Proof of Proposition Suppose the stock went up over the rst period such that the individual s level of wealth at t = is w + = w ( + x + ) > w. If he sells the stock then Assumptions and 2 imply that the individual incurs additional utility at t = 2 from pride about his decision at t =. Note that Assumption implies that the individual does not observe the realization of the stock at t = 2 and thereby foregoes potential regret or additional pride over the second period. The FWA would have been to not invest in the stock at t = which yields w alt = w. His nal level of utility from selling the stock is thus (w ( + x + )) kg w! (w ( + x + )). If the individual keeps the stock at t = he either incurs additional pride if the stock went up again over the second period or regret if it went down. In the rst case, Assumption 2 implies that the individual incurs pride towards the FWA which is not to have invested at all, i.e. w alt = w. individual made the optimal choice at t = but the sub-optimal choice at t =. In the latter case, the Assumption 2 implies that, in aggregate, the individual incurs regret towards the FBA which is to have invested in the stock at t = and sold it at t = yielding w alt = w ( + x + ). w ( + x + ) 2 B + ( p) B w (w ( + x + ) ( + x )) His nal level of expected utility is then w ( + x + ) 2 kg CC AA (w ( + x + )) Selling the stock at t = is preferred by the individual if and only if < pkg (w ( + x + )) w p + x + + ( p) + x + kg + x + 2( )! + ( p) kg (w ( + x + ))!! (w ( + x + ) ( + x )) w + x +! + x!. (8) Now suppose the stock went down over the rst period. If the individual sells the stock at t = he incurs regret about his decision at t = which leads to a nal level of utility (w ( + x )) kg w! (w ( + x )). 9

20 If he keeps the stock then Assumption 2 implies that he will incur regret independent of the stock movement over the second periods as he made a sub-optimal choice once at t =. The level of expected utility is then!! (w ( + x + ) ( + x )) (w ( + x + )) (w ( + x + ) ( + x )) p kg w ( + x ) 2 w ( + x ) 2 B + ( B w CC AA. Keeping the stock at t = is preferred to selling it if and only if > pkg (w ( + x )) (w ( + x + )) p + x + + ( p) + x + kg + x! + ( p) kg w w + x! + x 2( )!. (9) Note that the right-hand side of inequality (9) is positive as g is increasing. Both conditions (8) and (9) must be satis ed for the disposition strategy to be optimal at t =. For =, the proof is equivalent. A.2 Proof of Lemma Suppose 6=. Condition 5 is equivalent to > pkg (w ( + x )) kg (w ( + x + )) w p + x + + ( p) + x + x! + ( p) kg + x!. The right-hand side of this inequality is positive as > g > g pkg w (w ( + x + )) w + x! + ( p) kg w w + x p + x + + ( p) + + x! + x!. The rst inequality holds as g is convex and the second inequality holds since p + x + + ( p) + + x > + x 2( )! + x 2( )! 2

21 for all 6=. Therefore, the left-hand side of the above inequality must be positive, i.e. p ( + x + ) + ( p) ( + x ) > which is equivalent to condition 3. For =, the proof is equivalent. A.3 Proof of Proposition 2 Following the disposition strategy yields a level of expected utility!! p (w ( + x + )) w (w ( + x + )) kg!! +p ( p) (w ( + x + ) ( + x )) (w ( + x + )) (w ( + x + ) ( + x )) kg w ( + x ) 2 w ( + x ) 2 + ( p) 2 B w CC AA. Next, examine all possible other strategies and compare their level of expected utility with the one derived from the disposition strategy.. The individual invests in the stock only once which yields a level of expected utility p (w ( + x + )) + ( p) kg (w ( + x )) w kg!! (w ( + x + )) w The disposition strategy is preferred to this strategy if and only if > pkg (w ( + x )) (w ( + x + )) which is equivalent to condition (5). p + x + + ( p) + x + kg + x! + ( p) kg!! (w ( + x )). w w + x! + x 2( )! () 2

22 2. The individual invests twice into the stock which yields a level of expected utility p 2 w ( + x + ) 2 + ( p) 2 +2p ( p) w ( + x ) 2 B w (w ( + x + ) ( + x )) B w w ( + x + ) 2 kg CC AA w ( + x ) 2 (w ( + x + )) The disposition strategy is preferred to this strategy if and only if < pkg (w ( + x + )) w which is equivalent to condition (4). p + x + + ( p) + x + kg + x + 2( )! + ( p) kg (w ( + x + )) CC AA!! (w ( + x + ) ( + x )). w 3. The individual does not invest in the stock at all which yields a level of utility w. The disposition strategy is preferred to this strategy if and only if w > pkg + x +! + x! () + ( p) + x p + x + + ( p) + x w +p ( p) kg + x +! + ( p) 2 kg (w ( + x + )) w + x 2( )! + x! (2) which is equivalent to (w ( + x )) > + x k "pg +p ( p) g + ( p) + x p + x + + ( p) + x w (w ( + x + )) + x +! + ( p) 2 g + x!#. w + x 2( )! 22

23 Condition () is equivalent to > (w ( + x )) + ( p) + x p + x + + ( p) + x + ( p) + x " (w ( + x + )) k pg + x! + ( p) g w + x 2( )! g w + x!# Condition () implies condition (2) if + ( p) + x " pg + ( p) g w > + x "pg +p ( p) g which is equivalent to pg (w ( + x + )) + x 2( )! g w (w ( + x + )) (w ( + x + )) + ( p) + x g w + x +! + ( p) 2 g + x!# + x! + ( p) g w + x! + x!# w w + x! + x 2( )! + x 2( )! > p + x w g + x +!. (3) 23

24 The right-hand side of this inequality is negative. For the left-hand side, we deduce pg (w ( + x + )) + ( p) + x g > g p (w ( + x + )) + ( p) + x g + x! + ( p) g w w + x! + x 2( )! + x + ( p) w + x! 2( ) w + x! = g p w + x + + x > g > + + ( p) + x w + x! + ( p) + x g w + x! + ( p) + x w + x! + ( p) + x g w + x! The rst inequality follows from the convexity of g. The second inequality holds since p w and g is increasing. The last inequality holds since and + x + + x > + ( p) + x > w + x >, and since g (ab) > ag (b) for all a >, b >, and g convex. 6 The left-hand side of inequality (3) is therefore positive and thus satis ed which implies that condition () implies condition (2). 4. The individual invests in the stock at t = and at t = keeps the stock if it went up or sells the stock 6 Proof. De ne h (b) = g (ab) ag (b). Then h () = (as g () = ) and h (b) = a (g (ab) g (b)) > for all a >, b >, and g convex. Thus h (b) > for all a >, b >, and g convex. 24