The disposition effect: a survey

Abstract

The disposition effect is a well-recognized behavioral economic phenomenon which has been studied in numerous papers during the last three decades. Besides a literature review of the major theoretical, empirical, and experimental papers, the body of academic literature is subjected to a bibliometric analysis where all papers are taken into account which can be found via the search engine “EBSCOhost” (589 hits). This paper not only extensively describes the state of the art of research on the topic of the disposition effect. The combination of insights from a content analysis as well as a bibliometric analysis serves as a case study on the pattern of diffusion of research hypotheses and research methods. For example, it can be shown that the number of publications per year on the disposition effect remained relatively low subsequent to its discovery and only later started to substantially grow over time. On the other hand, the quality of papers, as approximated by the share of “A” journal articles per year, decreased over time.

1 Introduction

Investors often tend to sell assets that have gained in value while holding losing assets for longer periods. This “disposition effect”, a term coined by Shefrin and Statman (1985), has been analyzed in many countries all over the world. It has been empirically found to influence Australian (Brown et al. 2006), Chinese (Visaltanachoti et al. 2007; Feng and Seasholes 2005), Finnish (Grinblatt and Keloharju 2001), Israeli (Shapira and Venezia 2001), Taiwanese (Barber et al. 2007; Goo et al. 2010; Lai et al. 2010) and US-American investors in their decision making (e.g. Shefrin and Statman 1985; Odean 1998). Experimental evidence from Estonia (Talpsepp 2010) and Germany (Weber and Welfens 2006) points in the same direction.¹

Recently, the bias has even been found in sports betting on National Basketball Association (NBA) (Borghesi 2014) and National Football League (NFL) games (Hartzmark and Solomon 2012), leading to substantial deviations between prices of betting contracts and their respective values. Even individual home owners seem to be prone to the bias: Genesove and Mayer (2001) deliver empirical findings for loss aversion from Boston condominium owners in the 1990s and Case and Shiller (1988) find some evidence for this pattern in interviews with real estate owners during rising and falling real estate markets. Einiö et al. (2008) present comparable evidence from the Helsinki real estate market. Furthermore, the disposition effect has been illustrated in asset pricing (e.g., Shumway 1997) and portfolio choice models (Hung and Yu 2006).

Against this background, the disposition bias can be regarded as a well-recognized behavioral economic pattern which does not comply with normative strategies for investing and selling stocks which, e.g., focus on tax optimization (e.g., Constantinides 1983). Since there is a vast body of “standard” empirical and experimental literature on this topic already, recent studies focus more on genetic determinants (Cronqvist and Siegel 2014) and analyze the neurological factors underlying this bias (Frydman et al. 2014).

The disposition effect can have severe consequences for investors. When people do not sell underperforming stocks their trading performance is negatively affected. Moreover, the bias leads investors to pay more taxes on capital gains than necessary. Although the disposition effect is found to be significant in numerous studies, some critics cast doubt on its existence as well as on the theory it is based on.

The disposition effect is most relevant for private investors with little financial education. If they are prone to loss aversion, they will inevitably suffer from the disposition effect which then in turn harms their performance. In light of long periods of low and partially negative interest rates (Japan, Europe, USA), investing has become more important than mere saving. If private investors are made aware of this and other costly biases (e.g. the ostrich effect), there is a chance to improve their investment behavior and reduce the probability of old-age poverty.

The current paper essentially pursues two goals. First, it gives an overview of the vast body of literature on the disposition effect. The theoretical framework, empirical, and experimental findings as well as implications and consequences of the bias are presented. Second, with the aid of a bibliometric analysis, I give an impression of how the academic interest in this pattern has developed over time since its discovery in 1985. 589 papers concerning the disposition effect can be found via the well reputed search engine for academic literature “EBSCOhost”. They are categorized according to the underlying research method (empirical, experimental, model, etc.) and the journal ranking (Handelsblatt VWL Ranking 2013 and Handelsblatt BWL Ranking 2012) and are sorted by year of publication. 24 papers, secondary literature and recent working papers, have been manually added to the list. All papers are separated into “major” and “minor” contributions, depending on the importance of the phenomenon in the respective study. It can be shown that there was only little academic interest in the disposition effect subsequent to its discovery and that the majority of papers were published in the 2000s and 2010s with a peak in 2012. Furthermore, the share of articles which were published in “A” or “A$+$” journals on average decreases over time and takes an almost wavelike course. However, when taking into account the large number of publications in the last five years, the disposition effect is still a prominent topic in (behavioral) finance (see Figs. 3 and 5 for an overview of the number of publications on the disposition effect per year).²

Only few literature reviews for the disposition effect exist (see Kaustia 2010a and Amarnani 2010), and to my knowledge, a bibliometric analysis of the disposition effect has not yet been conducted. This paper aims to provide a new perspective on the vast body of literature, research hypotheses, and methods. The content analysis comprises the underlying theory as well as empirical and experimental evidence on drivers and consequences of the disposition effect. It also contains a comprehensive review of studies which utilize Odean’s (1998) frequently used method to measure the intensity of the disposition effect. In sum, the content analysis provides a structured overview of the vast number of theoretical and empirical studies on the disposition bias and its intensity. The combination of insights from a content analysis and a bibliometric analysis as the one described above serves as a case study on the pattern of diffusion of research hypotheses and research methods. Hence, the present study extensively describes the state of the art of the disposition effect and serves as a solid basis for future studies of the disposition effect.

The remainder of this paper is structured as follows. Section 2 deals with the theoretical background of the disposition effect. I set out the theoretical pillars of Shefrin and Statman’s (1985) pioneering work and also examine studies which cast doubt on this theoretical foundation. Having dealt with the theoretical basis which is essential for the understanding of the disposition effect, empirical and experimental studies are presented in Sect. 3. This section provides insights into early studies, potential drivers, recent research lines and implications of the disposition bias. The bibliometric analysis is presented in Sect. 4. In Sect. 5, the main insights and implications from the content analysis and the bibliometric study are discussed. Section 6 comprises limitations and gives suggestions for future research.

2 Theory of the disposition effect

The disposition effect labeled by Shefrin and Statman (1985) has been subject of numerous experimental, empirical, and theoretical studies. Shefrin and Statman (1985) attempted to embed this behavioral pattern into a theoretical framework which consists of four pillars: prospect theory, mental accounting, regret aversion, and self-control. These are briefly discussed in the following.

2.1 Prospect theory

In their path-breaking article from 1979, Kahneman and Tversky explain their prospect theory for decisions under uncertainty. First, they distinguish between an early phase of editing and a subsequent phase of evaluation. In the former phase, subjects choose a reference point as a mental starting point for the following decision. In the latter phase, people evaluate possible actions or “prospects” against this reference point. Compared to it, losses (gains) can be regarded as negative (positive) prospects. When people have to choose among risky prospects, they tend to behave risk-averse when prospects are framed as possible profits. They act risk-seeking when they are framed as possible losses. This framing-pattern in prospect theory is generally referred to as the reflection effect.

The choice of the reference point can, e.g., explain the well-known endowment effect³ because owners incorporate the possessed item into their status quo and therefore choose a reference point including the item. On the contrary, people that are asked for their willingness to pay (WTP) opt for a basing point without item. The possibility of buying will be interpreted as a positive prospect whereas the selling corresponds to a negative prospect. Subjects face “mixed” prospects (Kahneman and Tversky 1979, p. 288) in case they are urged to choose between actions that might cause either losses or gains.

Beyond that, a second aspect comes into play: the value function and its curvature. Kahneman and Tversky propose a function that is (i) defined by deviations from the reference point, (ii) concave for gains and in general convex for losses and (iii) less steep for gains than for losses (Kahneman and Tversky 1979, p. 279). This implies that people value losses larger than commensurate gains and that people are thus loss averse (“loss aversion”). Hence, the value ($\Delta \hbox {V}_{\mathrm{G}})$ induced by a gain ($\Delta \hbox {G}$) is lower than a decrease in value ($\Delta$V$_{L})$ caused by a comparable loss ($\Delta \hbox {L}$). A value function that fits to these assumptions is illustrated in Fig. 1 (see Weber and Camerer 1998, p. 171 for a comparable figure).

Fig. 1

Value function in prospect theory

Weber and Camerer (1998) apply this framework to the disposition effect. They call the application of a reference point to determine losses and gains “reference point effect” and the distinction in risk attitudes concerning losses and gains “reflection effect” (see Weber and Camerer 1998 in the following). They try to show that a combination of these two effects leads to disposition effects in a simplified example where an asset has the same chances to gain or lose a certain amount ($\Delta \hbox {L}$ or $\Delta \hbox {G}$) in each period (Weber and Camerer 1998, p. 170f.). In Fig. 1, the intersection of the coordinate axes is the reference point (P) which comprises the original price the asset has been purchased for. If the subject maintains this reference point and the asset “wins” the amount $\Delta \hbox {G}$, she will sell the asset in the following period if she is risk averse. The asset might “win” the same amount again $(2\Delta \hbox {G})$ but the additional value $(\hbox {V}(2\Delta \hbox {G}) - \hbox {V}(\Delta \hbox {G}))$ is smaller than the value she has already gained $(\hbox {V}(\Delta \hbox {G}))$ from the perspective of the original reference point.

Vice versa, an investor will keep an asset which has already “lost” a certain amount $(\Delta \hbox {L})$. The subject can then choose to accept a negative value $(\hbox {V}(\Delta \hbox {L}))$ or to keep the asset with the possibility of breaking even with the reference point or of losing the same amount again $(2\Delta \hbox {L})$ with the respective decline in value $(\hbox {V}(2\Delta \hbox {L}))$. If the investor is risk seeking in losses with equal chances for winning or losing, she will not sell the asset because the additional loss in value $(\hbox {V}(2\Delta \hbox {L}) - \hbox {V}(\Delta \hbox {L}))$ is smaller than the amount she has already lost and might regain. This leads to the pattern referred to as the disposition effect.

The original prospect theory merely considers singular weighted probabilities whereas cumulative prospect theory (Tversky and Kahneman 1992) takes cumulative probabilities into account and might therefore be more suitable for the abovementioned case. This extension primarily differs from the original theory in three respects (Tversky and Kahneman 1992, p. 302): First, this theory can be applied to continuous distributions. Second, it can be used with probabilistic as well as with uncertain prospects. Third and last, it allows for distinct decision weights for losses and gains.

With regard to prospect theory, Thaler and Johnson consider the question of how risk-taking is affected by prior gains and losses (Thaler and Johnson 1990). They conducted one of the first real money experiments and succeeded in showing that subjects exhibit increased risk seeking behavior when prior gains are present (“house money effect”) and that subjects seem to favor outcomes that offer a possibility to break even when facing prior losses (“break-even effect”).

Weber and Camerer (1998) conclude that the disposition effect does not occur when investors change their reference points. If the actual price (and not the original purchase price or any other historic price) is always the reference point and the value function proposed by prospect theory applies, investors will always sell the asset when chances for gains and losses are equal. The possible additional value in case of a gain $(\hbox {V}(\Delta \hbox {G}))$ is smaller than the loss in value in case the asset loses the same amount $(\hbox {V}(\Delta \hbox {L}))$.

Although many studies embed the disposition effect in the theoretical framework of prospect theory, the number of studies which doubt that this theory is capable of explaining the disposition bias is growing. Explanations based on prospect theory might only be applied to some cases but might not hold true for all. E.g., Hens and Vlcek (2011) show in their model approach that investors who tend to sell “winner” assets and to hold “loser” stocks would not have invested in stocks at all. They do find an “ex-post disposition effect” when they assume that investors actually have a stock at their disposal but claim that prospect theory cannot explain “ex-ante disposition behavior”. Barberis and Xiong (2009) also doubt that prospect theory suffices to explain the disposition effect. When focusing on realized gains and losses, their model more steadily predicts the pattern of the disposition effect than in case of annual gain and loss data where their model often fails. They conclude that utility from realized gains and losses might therefore be more appropriate with regard to analyzing investors’ behavior. Furthermore, Kaustia (2010b) proposes that even converse behavior (“reverse disposition effect”) might occur under certain conditions.

2.2 Mental accounting

The second pillar of Shefrin and Statman’s (1985) theoretical framework refers to mental accounting. This term was coined by Thaler (1980). In his famous article “Toward a positive theory of consumer choice” he speaks of a “psychic accounting system”. He elaborates on the same thought in a working paper where he describes a framework in which people have the tendency to separate distinct kinds of gambles into several accounts and use prospect theory to solve decision problems for each account, thereby neglecting potential dependencies (Thaler 1984, cited in Shefrin and Statman 1985, p. 780). This might also help to explain why traders do not adjust reference points for stocks (Amarnani 2010). In this context, loss aversion and mental accounting help to explain the status quo bias coined by Samuelson and Zeckhauser (1988) in which having a choice or default option enhances preference for it. Their experimental results suggest that an alternative becomes significantly more popular when it has been designated as the status quo and that its advantage increases with the number of alternatives.

Thaler (1999) explains the three most relevant components of mental accounting in his paper “Mental Accounting Matters”. First, he focuses on how people perceive outcomes and how they make and assess decisions, before and after the respective decision. Second, he describes that people tend to allocate financial means and their sources to actual and mental accounts and that people categorize expenses while restraining their expenditures implicitly or explicitly. Third and last, he describes another component of mental accounting that comprises the frequency of evaluating the respective accounts (daily, weekly, monthly, etc.) and what Read et al. (1999) call “choice bracketing”. This implies that persons can bracket a bunch of decisions broadly (evaluating possible consequences of all choices together) or narrowly (assessing each decision separately). Thaler (1999, p. 185) concludes that all mentioned aspects influence choices, harm the economic principle of fungibility, and thus “matter”.

2.3 Regret aversion

Having made an (ex post) bad investment decision, people may feel ashamed to admit this mistake because they tend to avoid regret (regret aversion) and to seek pride, leading to the described disposition to “realize gains and defer losses” (Shefrin and Statman 1985). Therefore, people who are prone to this behavior might make bad choices by trying to circumvent potentially regrettable circumstances.

Coricelli et al. (2005) conducted gambling experiments while observing participants’ brain activity via functional magnetic resonance imaging (fMRI). In their “regret condition” participants were informed about the outcome of the gamble they had not chosen. During the experiment, participants became more regret-averse while growing brain activity could be traced in the amygdala, an area which encodes incoming stimuli according to emotional factors rather than logical attributes. Furthermore, the medial orbitofrontal cortex was activated. This is a brain region where the emotional and motivational value of information is processed while foreknowledge is taken into account. Experimentees exhibited this brain activity pattern even before making a decision, suggesting that the same circuits seem to mediate experience and anticipation of regret.

Regret-aversive behavior is also closely linked to the status quo bias described above (Samuelson and Zeckhauser 1988). In the case of investing, this might lead investors to remain inactive and to neither realize gains nor losses. Shefrin and Statman (1985) give an example where investors sell a certain asset at a (small) gain and continue watching its price development. If the price rises, the original feeling of pride might be reduced by regret of having sold too early.⁴

2.4 Self-control

The disposition effect can also be seen as a problem of self-control. Thaler and Shefrin (1981) offer a model framework which helps to explain this issue. They assume that an individual (in their case an organization) is both a “myopic doer” and a “farsighted planner”. Shefrin and Statman (1985) presume that the “doer” utility is a function of the mental accounts the individual uses. They conclude that individuals want to accelerate their feeling of pride when having chosen the right asset in the past. Vice versa, they refuse to sell loser stocks because they try to postpone regret. The inner “doer” promotes such reactions and the “planner” may not be powerful enough to prevent such behavior, leading to the disposition bias.

3 Evidence

The first subsection covers early studies of the disposition effect, providing insight into the historical roots of this bias. Numerous empirical and experimental studies focus on determinants of the disposition effect which are described in second subsection. Here, Odean’s (1998) measure to quantify the disposition effect is introduced. Further, it gives a structured overview of prominent studies that focus on drivers of the disposition effect and how this bias can be alleviated. The third subsection comprises the latest innovative lines of research on the disposition effect: genetics and neuroeconomics. In the last subsection, consequences and implications of the disposition effect are described.

3.1 Early prominent studies of the disposition effect

Shefrin and Statman (1985) were the first to mention the disposition effect. They also empirically substantiated their findings and embedded them into a theoretical framework (see section above). They call attention to a former study conducted by Schlarbaum et al. (1978) who analyzed the behavior of 2,500 individual investors between 1964 and 1970 (Shefrin and Statman 1985, p. 785f.). Schlarbaum et al (1978) focus on realized gains data and find that the securities chosen by investors in their sample outperformed a market portfolio by on average 5.3 percentage points per year. They interpret their findings as “a reasonably favorable picture of the security-selection abilities of the individual investor” (Schlarbaum et al. 1978, p. 300). Shefrin and Statman doubt this and suggest that the materialized gains primarily stem from the successful stock selections whereas investors tend to hold on to their losing investments (Kaustia 2010a). Moreover, Shefrin and Statman (1985) analyze aggregate data from mutual funds trades from the Investment Capital Institute from 1961 to 1981. They point out that in weak stock market months substantially fewer redemptions can be traced than in good months and assume this to be an appropriate example of their disposition bias.

Ferris, Haugen and Makhija (1988) also present empirical findings which speak for the existence of the disposition effect in US stock markets although their data set merely consists of a small sample of stocks with low market capitalization. They use trading volumes of the respective stocks and not individual investors’ data. Another early empirical study by Heisler (1994) deals with professional traders who speculate in the US Treasury bond futures market. He shows that in his small sample better performing traders are less vulnerable to the disposition bias.

Individual stock market investors are nowadays the main subjects of empirical analyses on the disposition effect. This trend was at least partially triggered by Odean’s (1998) study. He analyzes trading records for 10,000 accounts at a broker house and finds that investors strongly tend to sell “winner shares” more often than “loser shares”. He controls for partial selling which might be motivated by portfolio rebalancing. When eliminating them, he still finds significant evidence of the disposition bias. Moreover, he finds the December effect in his data which implies a higher percentage of investors selling “losers” at the end of the year due to tax considerations. In this article, he develops the PGR–PLR-measure of the disposition effect which has been widely used (see Sect. 3.2).

One of the first experimental studies on the disposition effect that has been published in a prestigious journal was conducted by Weber and Camerer (1998). They find that student experimentees exhibit a noticeable disposition to sell “winner” stocks too early and to hold “loser” stocks too long in an experimental stock market. Subjects could buy and sell shares of six risky assets over several rounds while the prices of these assets changed over time and participants demonstrated a substantial disposition bias on average.

3.2 What drives and what reduces the intensity of the disposition effect?

3.2.1 Quantifying the disposition effect: PGR and PLR

Odean (1998) introduces two ratios which have been widely used in order to measure the disposition bias and its intensity: the Proportion of Gains Realized (PGR) and the Proportion of Losses Realized (PLR).⁵ Realized gains (losses) comprise the number of transactions that have been closed with a profit (loss). Nowadays, some studies also use the volume instead of the numbers (e.g., Frazzini 2006 who uses both measures). Odean’s original ratios are defined as follows (Odean 1998, p. 1782):

$$\begin{aligned} \mathrm{(i)}\quad {\textit{PGR}}=\frac{{\textit{Realized}}\,{\textit{Gains}}}{{\textit{Realized}}\, {\textit{Gains}} + {\textit{Paper}}\, {\textit{Gains}}}\\ \mathrm{(ii)}\quad {\textit{PLR}}=\frac{{\textit{Realized}}\,{\textit{Losses}}}{{\textit{Realized}}\, {\textit{Losses}} + {\textit{Paper}}\,{\textit{Losses}}} \end{aligned}$$

The commonly used measure of the disposition effect is the difference between PGR and PLR which can lie between −1 and 1. If PGR is greater than PLR, the disposition effect is present. Odean (1998) reports a PGR of 0.148 and a PLR of 0.098 and thus a disposition measure of around 0.051, indicating a (positive) disposition effect. PGR exceeds PLR by more than 50% ($\frac{{\textit{PGR}}}{{\textit{PLR}}}=\frac{0.148}{0.098}=1.510)$, indicating a strong bias. Until today, many studies have used Odean’s (1998) measure for quantifying the disposition effect. Table 1 gives an overview of such studies which are listed in alphabetical order. Empirical as well as experimental papers from different countries and investor classes have been summed up in this overview. Only ratios which were calculated with the Odean-method are displayed. Frazzini (2006), e.g., calculates these ratios for numbers as well as for volumes of losses and gains. The latter were not factored in. The ratios of Weber and Welfens (2006) only contain the empirical findings. For their paper, they also conducted experiments, the results of which are not included.

Table 1 Overview of studies using the PGR/PLR measure

Over all studies in Table 1, the average disposition-effect-measure comes to 0.083 which suggests that the disposition effect in this sample is on average positive and greater than in Odean’s (1998) study. Applying the relative measure (PGR/PLR) leads to an overall average of 1.475 which is comparable with Odean’s (1998) results: PGR exceeds PLR by almost 50% on average. The disposition effect has been found to be strongest in Taiwan and Australia whereas New Zealand investors exhibit only a small bias. US-American and international funds managers even exhibit a negative disposition effect in some studies.

The studies in Table 1 differ with regard to their research approaches (experimental vs. empirical), time periods and sample (sizes). For example, the range of subjects contains large groups of individual investors (e.g. Shu et al. 2005), small groups of undergraduate students (Goulart et al. 2013) or data from mutual fund managers. However, most of the studies in Table 1 have at least two things in common. On one hand, they use comparable measures to identify the disposition effect and its magnitude. On the other hand, these studies test the disposition bias in relatively homogenous groups with respect to ethnicity. It is generally accepted that cultural differences can account for differences in judgments and valuations. For example, the role of ethnicity has already been explored with regard to numerous biases. E.g., in a study focusing on the phenomenon of overconfidence, Yates et al. (1997) mentioned that on average, Asians are more prone to behavioral biases than Americans. However, there are not many studies testing for the influence of ethnic difference on the disposition effect. One study conducted by Gong and Wright (2013) has explicitly focused on the influence of ethics on the disposition effect. Analyzing more than 10,000 individual trading accounts in Australia, they find on average, investors with Chinese heritage more often realize gains than losses. On the other hand, Chen et al. (2007) also tested the disposition effect in American and Chinese, finding that American investors are comparably prone to the disposition effect.

Feng and Seasholes (2005) argue that the PGR–PLR-method is appropriate for analyzing the disposition bias at an average but not at an individual level. They state that PGR and PLR vary with demographic variables which might be associated to the bias. They give the example of diversification. When an individual investor has many stocks in her portfolio she might be regarded as more sophisticated in terms of financial literacy in comparison with an investor with few stocks. The number of stocks is thus directly linked to the PGR- and PLR-ratios which is an unfavorable statistical property. Nonetheless, these measures are the most prevalent ratios in empirical and experimental studies on the disposition effect which make comparisons across countries and different kinds of studies possible.

Since the introduction of Odean’s (1998) PGR–PLR-measure, numerous studies have tried to identify factors which contribute and which lessen the extent of the disposition bias. The subsequent paragraphs give a structured summary. Table 1 already provided an overview of studies from different countries, indicating that cultural differences might influence the intensity of the disposition effect although there exists only little empirical evidence which focuses this question. On the other hand, gender, age, information and trading experience have been found to be relevant determinants of the disposition bias. Technical trading rules help to alleviate the bias and its intensity could also be manipulated in experimental setups, e.g. by framing.

3.2.2 Gender and age

An empirical study over a six year period on the Taiwan Futures Exchange conducted by Cheng et al. (2013) indicates that men and younger traders, compared with female and mature individuals, show a less severe disposition bias. In his study of Estonian investors, Talpsepp (2010) shows that female investors generally perform better than male investors. When “controlling for different markets, trading, performance and investor sophistication related variables” he does not find differences concerning the susceptibility towards the disposition effect between men and women (Talpsepp 2010, p. 91). Da Costa et al. (2008) show that on average female subjects, in contrast to male subjects, do not keep losing stocks. They sell winners when the reference point changes from the original purchase price towards the previous price. They (speculatively) attribute this to differences between male and female brains in interpreting distinct reference points in general. Rau (2014) also posits that gender effects might exist. First, he finds that male subjects realize more capital losses on average than women. Second, men have substantially lower disposition effects and third, men are less loss averse than women in his experimental framework.

3.2.3 Information

The more information investors gather concerning a certain investment, the less they seem to be prone to the disposition bias. Kuo (2011) conducted a nationwide survey in Taiwan and collected 1,672 responses. The answers give evidence that when investors pay more attention towards making investment investigations they tend to keep losses shorter and thus exhibit a smaller disposition bias. Kuo (2011) concludes that if investors are willing to conduct the necessary research they might be likely to behave more rationally and be less controlled by regret aversion, thereby decreasing the magnitude of the disposition effect. Shapira and Venezia (2001) find evidence for the disposition effect in Israeli investors. They compare 1,642 independent investors and 2,688 investors who have been professionally advised and show that the disposition effect exists in both groups but is stronger for those without the additional information from professional advisors.

3.2.4 Experience

Frazzini (2006) reveals a disposition bias in the trading habits of experienced US-American fund managers which is, however, weaker than for individual traders. Da Costa et al. (2013) show that the disposition effect can at least be reduced by experience. In their experiment, the disposition bias noticeably decreases when investors have more than five years of experience in stock markets. Their findings are in line with findings by Coursey et al. (1987) and List (2003) who suggest that market anomalies, in their case the well-known endowment effect, decrease with rising market experience of buyers and sellers.

Dhar and Zhu (2006) analyze trading data from a big discount brokerage firm, containing information from more than 50,000 individual traders between 1991 and 1996. Besides trading data, they also utilize demographic and socioeconomic information in order to estimate traders’ literacy. They find that with rising trading experience, the degree of the disposition effect decreases. Further, individuals who are wealthier and who have professional occupations are on average less vulnerable to the disposition effect.

The experimental findings offered by Oehler et al. (2003) lead in the opposite direction. They test whether the disposition effect persists in dynamic market settings. When subjects assume the purchase price of the stock as a reference point, they tend to hold stocks longer even in “bad” times regardless of the market setting. The disposition effect is thus found to be a persistent pattern over time in their experiment. It can be reduced in a “dealer market” where the last price serves as a reference point.

3.2.5 Technical trading rules and automatic selling

Nöth (2005) proposes that the disposition effect can be substantially reduced by applying simple technical trading concepts which induce sell and buy signals, such as stop-loss mechanisms, and provides empirical evidence for his hypothesis. The results of a recent experimental study by Fischbacher et al. (2014) also indicate that stop-loss and take-gain rules can alleviate the disposition bias. While all participants of their experiment can buy and sell various assets, subjects in their treatment group are able to use the abovementioned automatic selling mechanisms. In this group, the disposition effect could be significantly reduced, primarily because losses were more often realized than in the baseline treatment without stop-loss and take-gain devices.

Weber and Camerer (1998) find that the disposition bias can be lowered by means of automatic selling at the end of a holding period.⁶ Weber and Camerer (1998) state that rational decision makers can be expected to behave similarly with or without automatic selling, at least in the absence of transaction costs. In their experiment the disposition effect could be reduced when all stocks within an experimental portfolio were automatically sold at the end of a holding period and participants got the opportunity to buy them back at the beginning of the next period. While the findings presented by Nöth (2005) and Fischbacher et al. (2014) have practical relevance for individual as well as professional investors, automatic selling at the end of each year would generate enormous transaction costs and possibly high taxes for realized capital gains which in turn reduce profitability of investments. Weber and Camerer (1998) interpret their findings as an indicator of the presence of the disposition effect and do not recommend to (automatically) sell out portfolios at the end of each year.

3.2.6 Experimental manipulations: framing and saliency

Kirchler et al. (2005) conducted experiments on the disposition effect with 64 students in Vienna and also found the disposition pattern in their participants. Individuals were told that dividends of their asset would randomly be drawn from a normal distribution with a mean of 95 experimental currency units (ECU) with a variance of 20 ECU. The negatively framed subjects received the (correct) information that the next dividend would be smaller or equal to 56 ECU with a probability of 5% whereas the positively framed participants were (truly) told that the dividend would be larger or equal to 134 ECU with the same probability. Those who received the negative information sold their assets earlier than the positively framed subjects.

Another experimental study conducted by Frydman and Rangel (2014) shows that the disposition bias can be reduced when the original purchase price is not salient. In a “high-saliency” condition, subjects see the purchase price as well as the development of the stock in the current period and are asked whether they are willing to sell whereas the purchase price is not shown on the screen of the experimental trading software in their “low-saliency” treatment. In the latter, the disposition effect is found to be 25% smaller. Rubaltelli et al. (2005) offer experimental evidence that the disposition effect can be substantially reduced when the loss of an asset is displayed as a percentage of the buying value and the current value instead of an absolute monetary difference.

3.3 Recent research lines—genetics and neuroeconomics of the disposition effect

Cronqvist and Siegel (2014) empirically study the “genetics of investment biases” by analyzing the trading behavior of Swedish twins. Their data stem from the Swedish Twin Registry which they combine with extensive information on their respective trading habits. They try to divide distinctions between individuals into environmental and genetic factors. Having controlled for observable individual patterns, they find that genetic distinctions may explain approximately 45% of the variation between individual traders with respect to the disposition effect, a lack of diversification and excessive trading. Another comparable twin-study has been conducted by Cesarini et al. (2012). They posit that loss aversion might at least partially have genetic reasons. Cronqvist and Siegel (2014) point out that the disposition pattern is related to framing, loss aversion and prospect theory and give a brief literature review on these topics. They mention a study by Zhong et al. (2009) who find genes that have an effect for the curvature of the value function proposed by prospect theory as well in gain as in loss regions.

Lakshminarayanan, Chen and Santos tried to examine the origins of investment biases by experimentally testing them in capuchin monkeys (Cebus apella). In their first study (Chen et al. 2006), they succeeded in showing that the monkeys are prone to loss aversion, suggesting that this pattern is innate and evolutionary stable. They also find evidence for the endowment effect (Lakshminaryanan et al. 2008) and framing effects (Lakshminarayanan et al. 2011) which are closely related to the disposition bias.

Frydman et al. (2014) tested investors’ behavior experimentally and measured their brain activity with the aid of fMRI. Like in most experimental settings, participants exhibited a substantial disposition effect in trading stocks which the researchers explain with the realization utility hypothesis (e.g., Barberis and Xiong 2012). This hypothesis claims that subjects obtain direct utility from the physical act of materializing gains or losses. The neural data substantiate their idea: First, they find that gains of possible transactions correlate with brain activity in an area (ventromedial prefrontal cortex) where the value of options during choices is encoded. Second, they show that the experimentees’ inclination towards the disposition bias is associated with the neural measure of realization utility. Third, when participants make profits, a brain area (ventral striatum) reveals positive reactions where data about variances in the present value of experienced utility are processed.

3.4 Implications and consequences of the disposition effect

In general, biases describe irrational patterns which do not comply with the concept of the “homo oeconomicus”. On the aggregate level, this can lead to reductions in market efficiency (see, e.g., the endowment effect (Kahneman et al. 1990)) or even total market failures (Bühren and Pleßner 2014) as well as welfare losses. On the individual level, biases can induce suboptimal behavior, especially for inexperienced private investors prone to the disposition bias, resulting in a loss of money due to underperformance in asset trading. Another cause of monetary losses for private investors is because they do not fully consider tax-advantages. A profit maximizing investor investing in a tax-conscious way, would avoid realized gains before receiving a long-term tax status, lowering the tax burden for capital gains (Kaustia 2010a). Investment behavior according to the disposition effect is completely contrary, and therefore against investors’ material interests. Their behavior can be attributed to the following potential drivers.

3.4.1 Suboptimal strategy with respect to taxes

Kaustia (2010a) gives an overview of the implications of the disposition bias. Among others, he lists and explains welfare costs and describes that the disposition effect increases investors’ capital gain taxes. He also refers to studies which show the disposition effect to be harmful to investors’ returns in different ways. For example, Odean (1998, p. 1790) finds that for “winners that are sold, the average excess return over the following year is 3.4% more than it is for losers that are not sold.” Thus, investors in his sample could have raised their performance by getting rid of losers and holding on to winners.

Locke and Mann (2005) analyze around 300 professional futures traders. They find that these traders on average hold “losers” longer than “winners” and that the average size for losing trades exceeds the respective size of winner trades. Surprisingly, the disposition bias does not lead to further costs in their analysis and thus does not negatively affect trading performance. Cici (2012) also posits that disposition-driven behavior does not have an observable effect on the performance of the funds in his sample. He examines US-American equity mutual funds and finds that they on average prefer to realize losses rather than gains. Managers who exhibit investment behavior influenced by the disposition bias do not perform differently to a substantial degree. Cici (2012) claims that learning effects due to academic research have reduced the bias noticeably, at least in many funds managers who are familiar with the respective academic literature.

In both studies, the authors focus on professional traders and investors. Professionals and mutual funds trade high volumes on a regular basis, and thus, their transaction costs per trade are negligible compared to private investors’ costs. If transaction costs are high, they will raise (reduce) realized losses (gains) which will adversely affect performance. Consequently, private investors whose transaction costs are relevant for their (dis)investment decisions may behave differently than professional ones. In order to cover the transaction costs, their prospective gains have to be greater and small gains are necessary to at least “break even”. On the other hand, losses become greater when transaction costs are taken into account. Already small losses might be enough to induce loss averse private investors to hold on to their losing investments longer than they should in order to prevent severe portfolio losses.

3.4.2 Disposition behavior and momentum effect

Jegadeesh and Titman (1993) were the first to document return momentum. They found that buying stocks with a good past performance and selling stocks that reveal a poor past performance generates significant gains over a holding period of 3 to 12 months. Traders who are prone to the disposition bias trade against the momentum effect, thereby leaving gains.

Shumway and Wu (2006) investigate whether the disposition effect contributes to the momentum effect. Using a sample of 13,460 Chinese investors and firms, they found the Chinese investors prone to the disposition bias. The higher the intensity of the disposition effect, the less often investors trade and the smaller the trade sizes turn out to be. When sorting the stocks of the Shanghai Stock Exchange, using unrealized net losses or gains of investors who were vulnerable to the disposition bias, they revealed a “winner/loser spread” of seven percent per annum. They therefore posit that the disposition effect contributes to momentum. In their model approach, Grinblatt and Han (2005) also assume that the disposition bias might facilitate momentum. They succeed in empirically substantiating their model approach, which implies that momentum is a consequence of an underreaction to news.⁷ Birru (2015) also analyzes the disposition effect in connection with the momentum effect. His results “suggest that the disposition effect may slow the incorporation of news, but not to the extent that it alone explains momentum” (Birru 2015, p. 1849).

3.4.3 Lack of diversification

Goetzmann and Kumar (2008, p. 456) state, “when investors are more reluctant to sell their ‘loser‘ and exhibit a stronger disposition effect, they end up with a relatively diversified portfolio of losers“. However, losing assets sometimes have similar traits in certain periods, for example with regard to the industry (e.g. dot-com-stocks). The disposition effect might therefore also contribute to a lack of diversification, especially in private investors’ portfolios.

4 Bibliometric analysis

Bibliometric analyses generally can add value (see, e.g., van Raan 2001) and can generate insights in case there are a large number of academic contributions, which holds true for the disposition bias. The basis for this bibliometric analysis is composed of all academic contributions to the disposition effect that can be found via EBSCOhost (test day: 6th January 2015).⁸ The short descriptions of the articles in EBSCOhost primarily contain the title, abstract, author(s), subjects, and year of publication. For the disposition effect, 589 hits remain after EBSCOhost automatically excludes exact duplicates. 24 articles have been manually added to the list: An 2014; Annaert et al. 2008; Case and Shiller 1988; Choi 2014; Dorn and Strobl 2011; Einiö et al. 2008; Fischbacher et al. 2014; Fu and Chen 2012; Gerke and Bienert 1993; Gong and Wright 2013; Goulart et al. 2013; Grinblatt and Keloharju 2001; Heisler 1994; Jin and Scherbina 2011; Krause et al. 2009; Lakonishok and Smidt 1986; Lakshminarayanan et al. 2011; Locke and Mann 2000; Meng 2010; Ploner 2014; Shumway and Wu 2006; Stracca 2004; van der Sar 2004; Weber and Welfens 2006. These papers were found as secondary literature which deals with disposition-like behavior in asset markets or are recent working papers which have not been published in an academic journal yet. Unfortunately, some relevant papers that deal with the pattern of selling “winners” too early and holding “losers” for too long might still be missing in this selection in case the disposition effect is not directly mentioned, on one hand. On the other hand, one article (Marquis and Filiatrault 2003) was excluded although it carries the term “disposition effect” in its title because it does not deal with the disposition effect in behavioral finance.⁹ Nonetheless, the most relevant papers that focus on the subject have been mentioned in the sections above.

From these altogether 612 hits, 35 working papers were taken out of the sample manually because the article that was finally published had the same or a very similar title. 56 EBSCOhost hits were also excluded because they were unusable (e.g., front and back matters of the respective journals, summaries of papers of the respective issues, table of contents, summaries of abstracts etc.). The remaining 521 articles are the foundation for the following bibliometric analysis of literature on the disposition effect.

As a next step, the most relevant articles were filtered out. Those contributions where the disposition effect is mentioned in the (sub)title, the abstract, the keywords or the subjects-section are categorized as most relevant (“major”) whereas the articles in which the disposition effect is merely mentioned in the continuous text are categorized as less relevant (“minor”). Altogether, 206 contributions are categorized as “major”. Among them are 163 articles published in academic journals, 18 working papers, 14 dissertations, 4 periodicals, 3 conference papers, 2 books, and 2 collective volume articles. Within the “major” contributions, the articles are clustered according to the underlying research method (empirical, experimental, model, survey, review and comment, empirical/model, and “miscellaneous” (e.g., structural interviews)). Figure 2 gives an overview of the composition of the “major” articles that have been published on the disposition effect.

Fig. 2

Categorization of papers on the disposition effect into major and minor contributions. (For a more condense overview and for reasons of clarity, the category “experimental/empirical” (altogether three contributions, e.g., Weber and Welfens 2006) has been matched with the category “experimental”)

Figure 2 visualizes the vast majority of empirical studies of the disposition effect within the “major” contributions. This result strikes the eye but it is intuitive since the disposition effect is an empirical phenomenon.¹⁰ More than 53% (111 of 206) of all major papers published on this topic chose an empirical research method. With approximately 20%, experimental papers come second place. 20 model-based approaches (10%) deal with the disposition effect as a major topic and it has been discussed in 13 review articles (6%). With the aid of (primarily online) surveys, 9 papers (4% of all major papers) have analyzed this behavioral economic pattern. The category “miscellaneous” comprises, e.g., structural interviews and simulations and articles which could not be clearly assigned to the abovementioned categories.

Fig. 3

Development of the number of papers on the disposition effect in research categories per year

Figure 3 gives an impression of the developments in research on the disposition effect over time. Corso et al. (2014) find that Joseph de la Vega describes investor behavior consistent with the disposition effect already in 1688 in his famous book “Confusion de Confusiones” which is generally regarded as one of the oldest written contributions on stocks exchanges. However, Shefrin and Statman (1985) were the first who coined the term “disposition effect” for holding “losers” too long and selling “winners” too early. It is categorized as “empirical/model” since they provide the theoretical basis for this pattern and also give empirical support for its existence. It can easily be seen that the disposition effect was not immediately picked up on as a topic of research subsequent to its discovery. Three years after Shefrin and Statman’s article in the Journal of Finance the first publications explicitly took up this issue. Ferris and his coauthors presented empirical evidence of the “tax-loss selling hypothesis” and the disposition effect (Ferris et al. 1988). The latter is presented as a relevant factor throughout the year and not merely a determinant of year-end-volume. Their article, which was also published in the Journal of Finance, was thoroughly discussed by Lawrence Harris in a comment in the same issue (Harris 1988).

In the following years from 1989 until 1992, merely “minor” articles dealt with the disposition effect. According to Kaustia (2010a) the period without considerable growth in number of publications was on one hand due to the shortage of relevant information. On the other hand, “individual investors’ behavior was simply deemed uninteresting” at that time (Kaustia 2010a, p. 173).

The first “major” articles dealing with the disposition effect after 1988 were written by Gerke and Bienert (1993, 1994). They conducted computer based tests with student subjects and hereby substantiate the disposition effect experimentally. In 1998, two major studies were published: One empirical investigation by Odean (1998) and one experimental study by Weber and Camerer (1998) (see Sects. 2 and 3 for details). From the year 2000 onwards, the number of major academic contributions to the disposition effect substantially increases and takes almost a wavelike course. In 2007, 2011 and 2013 downturns can be observed when comparing the number of papers with those of the respective preceding year. A clear peak concerning the number of publications per year can be observed in 2012: 28 major studies and 28 minor studies on the disposition effect were published in this year.

The 163 “major” articles on the disposition effect that have been published in academic journals were undertaken a more detailed analysis with regard to their journal ranking. Basis for this analysis is the “Handelsblatt VWL Ranking 2013”. This ranking methodology for economic journals categorizes them into “A$+$”, “A”, “B$+$”, “B”, “C$+$”, “C” and “D” journals. The “weight” of journals and thus their ranking is primarily based on Combes and Linnemer (2010) who developed a novel consistent index of all EconLit journals. In their approach, they first establish an index which ranks all 304 journals of the Thomson Reuters (JCR) database the citations of which are counted. They extend this citation-based index to all almost 900 EconLit non JCR journals and develop a model which explains the index via the score of the authors of the respective journal as well as the journal’s citations in Google Scholar. In the end, they prognosticate the score of the index for the non-JCR-journals and thus receive a ranking index which is consistent for all EconLit journals.The analysis by Combes and Linnemer (2010), which is the basis for the Handelsblatt journal ranking, is the most comprehensive assessment of journals to date (Wohlrabe 2011, p. 66). Its highly regarded reputation and high-ranking appointment procedure for professorships in the German-speaking area, are additional reasons the Handelsblatt ranking was chosen. The “Handelsblatt-points” are the basis of evaluation for habilitations at German universities and also for doctor’s degrees, if the candidate chooses a cumulative dissertation consisting of several papers. As result, every economist who strives for a professorship in German speaking areas aims to publish in journals with high rankings according to this method.

The Handelsblatt ranking comprises more than 1300 journals but, like most other rankings, is not applicable for monographies. Like all other rankings, it is not infallible and its creators are receptive to improvements (Bannert et al. 2011).

Since not all journals are contained in the “Handelsblatt VWL Ranking 2013”, the “Handelsblatt BWL Ranking 2012” for papers on business administration was used to complete the missing information. Altogether, 145 articles could be categorized with regard to the quality of the journal they were published in. The journals the remaining 18 articles were published in were not part of the rankings and were thus not used.

Fig. 4

Categorization of academic journal articles on the disposition effect according to research method and journal ranking. (For a more condensed overview and for reasons of clarity, the categories “A$+$” and “A” were merged to “A”. This approach is also applied to “B$+$” and “B” as well as “C$+$” and “C”)

In Fig. 4, the 145 journal articles are clustered according to the underlying research method (see Figs. 2, 3) and ranking of the journal they were published in. The preponderance of empirical analyses strikes the eye, as in the figures above. 40% of all empirical studies were published in “A” or “B” journals (15 “A”, 20 “B”). In the experimental section, more than 42% of the articles fall in the category of B or better. 50% of the academic contributions which comprise the development of a model were published in journals with an “A” ranking. 3 of 4 papers which focus on the development of a model which was also empirically tested (empirical/model) were published in “A” or “B” journals. Although the sample of such studies is relatively small, this pattern is intuitive since this research method is the most sophisticated and work-intensive. Because almost every second article on the disposition effect which comprises a model approach is published in an “A” journal, it seems understandable that the addition of empirical evidence even enhances the quality and thus the share of articles in “A” and “B” journals. This pattern might also be partially due to self-selection. Economists who developed a model which they succeeded to base on empirical findings might be more confident with regard to their research achievement. Thus, they might be more likely to approach an “A” journal for publication than economists who merely focused on either a model or empirical research.

Fig. 5

Development of the number of major journal articles on the disposition effect in ranking categories per year

Figure 5 displays the development of the number of papers on the disposition effect per year in ranking categories. It can be seen that the first publications (Shefrin and Statman 1985; Ferris et al. 1988; Harris 1988) were exclusively published in “A” journals (Journal of Finance). In 1998, one of two major articles was published in an “A” journal. The number of top-tier-publications slightly increases over time with small peaks in 2005 (3), 2006 (4), 2009 (4) and 2012 (5). However, their relative share decreases due to a considerable increase in other publications. This might be the result from top-tier-journals’ increasing capability to select high-quality papers, especially in times where many papers on the disposition effect have been published.¹¹ The percentage of articles per year in an “A” journal almost takes a wavelike course and on average declines over time. In 2005, the share of “A” publications amounted to 25% (3 of 12 publications) and rose to 36% in 2006 (4 of 11 publications). The next peaks can be traced to 2009 and 2012 where 24% of all journal articles on the disposition effect were published in “A” journals. In 2014, this share merely amounted to 6%. Thus, the quality of “major” publications on the disposition effect in academic journals has deteriorated over time, on average. Moreover, the chance of successfully publishing an article in an “A” ranked journal seems to be higher the newer a research topic is.

The almost wavelike but declining development regarding the share of “A” journal articles on the disposition effect might be explained by three factors. First, the pioneering articles on the disposition effect were all exclusively published in “A” journals which might be due to the fact that they addressed a topic that had not been dealt with before. The wave that started in 2005 might have been due to the rise of behavioral economics which might also hold true for the number of publications on the disposition effect in general. An opposite pattern might have also occurred. Sometimes path-breaking articles are at first rejected by “A” or “A$+$” journals, which might lead authors to seek for lower ranked journals. E.g., the well known article “Market for ‘Lemons’: Quality Uncertainty and the Market Mechanism” by George A. Akerlof (Akerlof 1970) had first been rejected by three “A” journals (Gans and Shepherd 1994, p. 171). Eventually, it was published in a comparably ranked journal (the Quarterly Journal of Economics), but if Akerlof had exhibited a lack of persistence, he might have submitted his article in a less popular journal to increase the likelihood of publishing his article.

Second, articles which have been published in an “A” journal arouse more academic interest than articles published in “B” or “C” journals and might thus entail more “A” journal publications. In general, this might hold true for virtually any research topic in finance, not just for the disposition effect. Researchers might have become interested in the disposition effect in the years where many articles or a high percentage of papers on the disposition effect were published in “A” journals, e.g., 2005 and 2006.

Third, besides the actual time researchers spend on writing an article, the papers in general have to undergo a peer review process which is time-consuming. Thus, there is a time-lag between the completion of an academic paper and its publication which can lie between a few months and several years.

5 Discussion and implications

The recent paper focused on two goals. On one hand, it aimed to give a structured overview of the vast body of literature on the disposition effect. Theory, evidence, and implications of the bias are described. On the other hand, it provides a bibliometric analysis on this bias to classify related literature into major and minor contributions, depending on the importance of the phenomenon in the respective study. Furthermore, “major” papers were categorized according to research method and journal ranking. I analyzed the academic interest in the disposition effect over time and the development of the quality of major articles. Therefore, this paper extensively describes the state of the art of the disposition effect. To my knowledge, a content analysis and a bibliometric analysis on the disposition effect have not been combined before. This combination leads to interesting insights concerning research strategy. Referees for “A” journals more often ask for a motivating model. As a result, purely empirical or experimental studies are not often published in such journals. Besides journal editors’ and referees’ preferences, this observation could also be due to self-selection: Economists who chose a model approach and who empirically substantiated their findings might be more confident concerning their research achievement and might more likely approach an “A” journal for publication. With regard to its underlying theory, most papers explain the disposition effect with the aid of prospect theory, mental accounting, regret aversion and (lack of) self-control. Recent studies, however, claim that prospect theory can only explain the occurrence of the effect in some cases and not in all circumstances (see Barberis and Xiong 2009; Hens and Vlcek 2011).

The disposition effect has been found to influence private as well as professional investors all over the world, even though the latter seem to be less prone to the bias. Many studies have applied Odean’s PGR–PLR-measure to quantify the disposition effect. In Cici’s (2012) study, the effect in funds managers is found to be negative (PGR–PLR$\,=\,-$0.015, PGR/PLR$\,=\,$0.965), meaning that they tend to sell losers earlier than winners. Annaert et al. (2008) find comparable results for international funds managers in their sample. In Australian individual investors (see Brown et al. 2006), the effect is particularly strong in comparison with other studies in this sample (PGR–PLR$\,=\,$0.280, PGR/PLR$\,=\,$2.217). Shu et al. (2005) also deliver empirical evidence that Taiwanese individual investors are strongly biased (PGR–PLR$\,=\,$0.210, PGR/PLR$\,=\,$2.500). In contrast, individual investors from New Zealand do not seem to be strongly biased (PGR–PLR$\,=\,$0.006, PGR/PLR$\,=\,$1.055). In the light of the findings offered by Gong and Wright (2013), this evidence gives a hint that cultural differences might also lead to differences in the extent of the disposition effect.

Many studies analyze what drives and what lessens the disposition bias’ intensity. Some authors find that the disposition effect is lower for subjects with higher trading experience and literacy (Dhar and Zhu 2006) whereas the direction of gender effects is not really clear (see results of Rau 2014 and Talpsepp 2010). Mechanisms such as stop-loss trading can help to alleviate the bias (Nöth 2005; Fischbacher et al. 2014). Recent research lines focus more on neuroeconomic (Frydman et al. 2014) and genetic (Cronqvist and Siegel 2014) foundations of the disposition bias.

The disposition effect can have negative consequences for investors. On one hand, it leads to higher capital gains taxes. On the other hand, performance can be negatively affected (e.g., Odean 1998) although some studies reject this assumption (Cici 2012). Moreover, the disposition effect facilitates underreaction to news which in turn contributes to momentum effects (Grinblatt and Han 2005). Kaustia (2010a) presents implications for financial advice in this context. He states that people often tend to regard unrealized losses merely as paper losses which they hope to regain in the future and that they thus act contrary to the well-known investment advice “cut your losses and let your profits run”. Some investments might rise in value again whereas some remain at the low level for a long time. Merely regarding the losses as paper losses can thus hinder people from conducting the necessary research on their holdings and might contribute to the “ostrich effect”, a term coined by Galai and Sade (2006). This effect implies that people have the tendency to avoid negative information with regard to their investment portfolio. Therefore, investors who are prone to this bias do not look up their portfolios in bad market times and might miss relevant data for trading decisions.

The disposition bias was not taken note of directly after it had first been mentioned because, according to Kaustia (2010a), individual investors were not interesting enough at that time. It received promotion by the rise of behavioral economics and behavioral finance, especially since the beginning of the 2000s. Downturns concerning the number of publications per year can be traced to the years 2007, 2011, and 2013. Moreover, it could be shown that the average quality of major academic contributions has deteriorated over time. When taking into account the ranking of the journal the respective article was published in as the primary indicator of quality, it can be seen that the share of articles published in “A” journals according to the “Handelsblatt VWL ranking 2013” and the “Handelsblatt BWL ranking 2012” declines over the course of time (Fig. 5). The almost wavelike development of the share of “A” journal articles might be due to the fact that articles of high quality trigger interest in academic research and thus might entail a follow-up effect. Because of a time-lag in the peer-review process between completion and publication of an academic contribution several years might lie between the “peaks” in the development over time. Therefore, the pattern displayed in Fig. 5 might have looked completely different. In its actual form, it seems as if the chances of successfully publishing an article in an “A” ranked journal are higher the newer a research topic is.

6 Limitations and future research

The bibliometric analysis makes use of a “rule of thumb” when discriminating major from minor academic contributions. When the term “disposition effect” is neither mentioned in the title nor the abstract, the paper is automatically categorized as “minor”. This might not hold true for all contributions and is thus a limitation. Although some relevant contributions might be missing in the “major” section, they can easily be found by conducting the necessary intensive research. This “automatic” selection approach serves the purpose to quickly get an overview of the most relevant literature on a certain topic. It can be applied to all topics that can be searched via an academic search engine and might help authors in their description of the state of the art of their respective research subject. Besides EBSCOhost search engines like Google Scholar can be used and the approach can be applied analogously. As a first step it is less work-intensive than, e.g., the (unsystematic) snowball principle.

To my knowledge, a comparable bibliometric analysis has not yet been conducted for economic topics. A bibliometric study of other topics in (behavioral) economics could shed light on the pattern of diffusion of research hypotheses and research methods in general. It would also allow for comparisons with the patterns found for the disposition effect in the present study.

As of today, the disposition effect has mostly been tested in (at least ethnically) homogenous groups. As stated in Sect. 3.2 (see also Table 1), only few studies test the effect of ethnicity with regard to the intensity of the disposition effect (e.g., Gong and Wright 2013; Chen et al. 2007). This research gap is worth exploring in the future.

In future research, the disposition effect and the abovementioned ostrich effect could be tested together with regard to their impact on investment performance. Brown and Kagel (2009) analyzed the disposition effect, the status quo bias and the ostrich effect experimentally in simplified stock markets. They neither find evidence of the disposition effect nor of the ostrich effect and their subjects act completely contrary to these patterns. They attribute their findings with regard to the disposition effect to the structure of the experiment where stocks are exchanged one-for-one and gains or losses are therefore not directly realized. Using an experimental framework comparable to the one introduced by Weber and Camerer (1998), it would be possible to test which effect has worse consequences on trading performance and whether they reinforce each other. The results of an experiment conducted by Frydman and Rangel (2014) show that the disposition bias can be reduced when participants do not see the original purchase price. However, when subjects do not look up their portfolios in bad times and are vulnerable to the disposition effect, this might worsen trading performance even more. In that case, investors may even hold on to losers until they almost reach zero because they may tend to dismiss all negative information which might in turn cloud their (ex post) wrong expectations. Burying one’s head in the sand might help to shield from unfavorable information in bad times. But it is hard to utter optimal trading decisions with a mouth full of sand (Zweig 2008). Thus, it is worthwhile testing these biases in combination and to make investors aware of the consequences.