Loading...

Experimental Evidence on the Disposition Effect

Bachelor Thesis 2013 39 Pages

Business economics - Investment and Finance

Excerpt

Contents

List of Figures

List of Tables

List of Acronyms

List of Symbols

Zusammenfassung

1 Introduction

2 What is the Disposition Effect?

3 Empirical Evidence of the Disposition Effect

4 Theoretical Explanation of the Disposition Effect
4.1 Matching Theories to Empirical Data
4.2 Prospect Theory
4.3 Psychology and Realization Utility Model

5 Experimental Evidence of the Disposition Effect
5.1 The Experiment as a Method
5.2 Experiments and Disposition Effect
5.3 Evidence in Experiments
5.3.1 Disposition Effect
5.3.2 Additional Experimental Testing
5.3.3 Evidence of the Realization Utility Model

6 Threats for Valid Conclusion Making in Experimental Re­search
6.1 Threats to Statistical Conclusion Validity
6.2 Threats to Internal Validity
6.3 Threats to External Validity
6.4 Nuisance Variables
6.5 Neuroscientific Tools in Economic Research

7 Judging Scientific Methods

Appendix
A Additional Tests
A.1 Tax Motivated Selling

B Additional Figure

References

List of Figures

1 V(+G) < —V(-L) (Weber and Camerer, 1998)

2 Developing Theories (see Reiß and Sarris, 2012)

3 Testing the RUM (Frydman et al., 2011)

4 PGR for each month (Odean, 1998) VII

5 Neural and Behavioural Loss aversion (Tom et al., 2007) . . VII

List of Tables

1 Empirical findings in Weber and Welfens (2007)

2 Results for disposition effect variables

List of Acronyms

illustration not visible in this excerpt

List of Symbols

illustration not visible in this excerpt

Zusammenfassung

Das Thema dieser Bachelorthesis ist die Experimentelle Evidenz des Dis­positionseffektes.

Der Dispositionseffekt beschreibt den Umstand, dass Investoren tenden­ziell ihre Gewinneraktien zu schnell verkaufen und ihre Verliereraktien zu lange halten. Zum Dispositionseffekt existiert ein breites Spektrum an em­pirischer Evidenz sowie an deskriptiven Theorien, die den Effekt erklaren können.

Bei der Suche nach den Ursachen des Dispositionseffektes nimmt das Ex­periment als Forschungsmethode eine besondere Stellung ein. Experimente können die theoretischen Erklärungsansatze separieren und sie wiederholt auf ihre Plausibilitat testen.

Ein besonderer Fokus dieser Thesis liegt auf der ’’Realization Utility Hy­pothesis” des Modells von Barberis und Xiong (2012) als jöngste the­oretische Grundlage des Dispositionseffektes. Die Grundannahmen des Modells wurden mit Hilfe von Gehirnstrommessungen in einem neurowis­senschaftlichen Experiment von Frydman, Barberis, Camerer, Bossaerts und Rangel (2011) getestet und unterstötzt.

Die Nutzung solcher Experimente in der (okonomischen) Forschung birgt allerdings Gefahren, denn die Ergebnisse koönnen in vielerlei Hinsicht verfaölscht oder verzerrt sein. Elementar fur die Anwendung der experimentellen Methode ist also ein gut ausgeprägtes Bewusstsein fur ihre Risiken, be­zogen auf den Gegenstand der Untersuchung.

Die Aufgabe dieser Thesis wird sein, den zusaötzlichen Erkenntnisgewinn von Experimenten zum Dispositionseffekt darzustellen, aber auch die Gefahren der angewandten Methoden abzuwagen.

1 Introduction

The topic of this thesis is Experimental Evidence on the Disposition Effect. Accordingly, the following text refers to two superordinated domains of economic science by reviewing the most recent and important literature to the topic:

1. In general, the research on individual decision biases and heuristics (simple decision rules) in behavioural economics. Specifically, the disposition effect as the biased trading behaviour of investors to sell their winning shares and to hold their losing shares, although contrary behaviour would be rational.

2. In general, the advantages and disadvantages of methodical instru­ments. Specifically, the experiment as a research method to analyse the disposition effect. As well as its different economic, psychological and neuroeconomic tools.

Thus, the main target of this thesis is to illustrate the additional research findings, which the experimental approach can acquire on disposition effect and to discuss these experiments on their usefulness and threats.

First, some basic explanations are necessary. The logic is as follows:

To analyse the disposition effect in experiments, it has to be understood in detail (section 2). If the disposition effect is convincingly observed in real security markets (section 3), it makes sense to build theories that match these empirical data. In contrast to traditional economic theories, these new descriptive theories can explain how and why the disposition effect arises (section 4 ). Now the theories explain the causes of the effect differently. Thus, researchers use experiments to distinguish between the different explanations and to test separately its possible causes (section 5). Unfortunately, the results of experiments can be biased in several ways. Therefore, it is necessary to look at the problems that can distort experimental findings and cause difficulties to the reasoning of consistent relations between cause and output as well as to the generalization of the experimental findings (section 6).

At the end, the text should have shown that the disposition effect exists in real security markets as well as in several experimental settings. It should illustrate the most important theories to the disposition effect, its possible causes and the discussion of the experimental method that it able to analyse the effect.

There are several subsequent questions that cannot be treated here due to the given standards. For example, whether different investor groups exhibit different strength of the effect or how it could be overcome. Additionally, the disposition effect is not limited to the trading behaviour of individual investors. It also plays a role in other domains as the housing market. There are far more theories as mentioned in this thesis, which give possible explanations to the effect. These theories could be combined or tested. Moreover, there are several ways to vary the assumptions of the mentioned theories or to explain the mathematical models much more detailed as this thesis can do. At last, it should be noticed that there are three fundamental questions in economic science that the text refers to and which are more extensive as it could possibly be shown here:

- Should economic science analyse the behaviour of individual market participants in regard of their psychological profiles and individual decision biases as behavioural economics do? Or should it continue to assume rational behaviour of ”homo oeconomicus”-traders in com­plete markets?
- Should economic science use new research methods like experiments or is the observation of whole, real markets the only way to gain solid economic research findings?
- Should economic science be combined with other disciplines like psy­chology and even neuroscience to gain information on problems that cannot be solved by traditional economic theory alone?

In noticing the importance of these underlying questions to future economic science, this thesis will illustrate the additional value of experimental ap­proaches, always regarding the disposition effect, but also not omitting the criticism that this method has to face.

The following text uses a linguistic distinction between ’’empirical evi­dence” and ’experimental evidence”.[1] In this context, empirical evidence means the research of field data (like in Odean, 1998). More precisely it means the naturalistic observation of individuals in their natural setting as a passive form of research, without any manipulating influence of the researcher on the data (Kirk, 2013). In this context, experimental evidence means the laboratory testing of the disposition effect (like in Frydman et al., 2011). This linguistic distinction is given in order to simplify the sep­aration and comparison of the two methods.

2 What is the Disposition Effect?

Suppose you invest $200 in a portfolio of two equally expensive securities (”A” and ”B”, $100 each). In the subsequent period, security A has fallen to $50, while security B has risen to $150. You will get the chance to realize assets out of your portfolio at the end of the period. Which one would you rather sell: The loser asset, $50 under purchase price, or the winner asset, $50 over purchase price?

This kind of investment decision has recently been paid much attention to in economic science. Field data has been analysed regarding how investors in real capital markets behave[2] and subjects have been asked these kind of questions in several experiments.[3] Researchers found out that investors who have to decide under uncertainty heavily tend to sell their winners and immediately realize a gain. In contrast, they delay selling their losers if possible.[4] In this context, winner-assets are those whose current price is higher than the investor’s individual reference point, which is mostly as­sumed to be the purchase price. Thus, loser-assets are those whose current price is lower than their purchase price. Investors tend to sell their winners too fast and hold their losers too long.

”Too fast” means that in order to achieve higher expected profits (and tax advantages), investors should hold their winners longer in terms of rational decision making. ”Too long” means that overall loss could be decreased by selling losers quicker (Jungermann et al., 2010). Shefrin and Statman (1985) named this phenomenon the ’’disposition effect”.

Another way to put the disposition effect into words:

Investors have a higher propensity to sell their winners than to sell their losers.

Such behaviour is meaningless in terms of rational decision making, due to five arguments:

1. The difference of a stock’s price to his purchase price is supposed to not matter (Camerer, 2000), because purchase prices are sunk costs. You paid them and you cannot bring this money back, so it does not matter for rational decision making.
2. A rational investor does not account the gain or loss of a single se­curity. He rather seeks to maximize the overall profit of his entire portfolio.
3. Rational decision making also implies future orientation. Rational decisions should only depend on future consequences of the given alternatives. A rational investor weights the expected future values (and not the past ones) of his security portfolio with his preferences and maximises them.
4. Concerning these values, research finds that recent winners will per­form better in the future as recent losers do.[5] Thus, investors are losing money in exhibiting a disposition effect.[6]
5. And finally, selling winners instead of losers is suboptimal for tax­able investments, because realising gains increases the amount of tax payment while realising losses decreases it (Odean, 1998).

Neoclassical economic theory assumes ”homo oeconomicus”-investors, which always behave rationally. The next section will show that this assumption is to be doubted for real environments like security markets.

3 Empirical Evidence of the Disposition Ef­fect

Odean (1998) shows that investors in real security markets do not exhibit behavioural patterns that are in line with rational decision making but ex­hibit a disposition effect instead. He uses the method to observe empirical field data, namely the trading records of 10, 000 accounts at a large dis­count brokerage house. As all empirical methods have their strengths and weaknesses, this method has several advantages: It is able to analyse the specific population, which the researcher wants to observe (individual in­vestors). It also observes a long time interval of 7 years, where the subjects are unaware that their behaviour is studied and therefore act in a natural way (Weber and Welfens, 2007; Kirk, 2013).

Odean (1998) has found that individual investors significantly prefer to sell winners and hold losers, except in December.

These findings are consistent with the hypothesis that investors experience a disposition effect from January to November and conduct tax-motivated- trading in December.[7] By definition, disposition effect in investment be­haviour exists if ’’individual investors have a greater propensity to sell stocks trading at a gain relative to the purchase price rather than stocks trading at a loss” (p2, Frydman et al., 2011). Thus, to proof the existence of a disposition effect it is needed to show that the number of actually real­ized gains relative to the total number of potential gains is bigger than the number of realized losses relative to the number of potential losses (Odean, 1998).[8] The ratios are named PGR and PLR with:

illustration not visible in this excerpt

PGR : Proportion of Gains Realized and PLR : Proportion of Losses Realized.

The potential gains and losses consist of the realized gains and losses plus paper gains and losses. In this context, ’paper gains and losses” are defined as potential gains that the investor has refused to realize.

In this notation, the disposition effect is the empirical fact that PGR is significantly higher than PLR (Hi).

To find the empirical data to be consistent with Hi, Odean (1998) tests whether the null Hypothesis (PGR is less or equal than PLR) can be rejected.[9] A look at the formulas should clarify the statements:

Hi : PGR > PLR & H0 : PGR < PLR (3)

The null hypothesis is highly significantly rejected with statistical t-tests (Odean, 1998). Statistically, this means that we can assume Hi: The PGR is greater than the PLR. In other words: We assume that investors in the

illustration not visible in this excerpt

Table 1: Empirical findings in Weber and Welfens (2007).

sample exhibit a disposition effect.

To define a concrete variable that measures the height of disposition effect that is found in the data, the variable DE is defined as

DE = PGR — PLR. (4)

With PGR and PLR being ratios between 0 and 1, the value of DE lies in the interval of [-1,1]. If DE is equal to 1, it means that all gains are realized immediately, while no loss is ever realized (Weber and Welfens, 2007). A DE of —1 is the analogue value of the other extreme. A DE of 0 indicates that investors do not exhibit any disposition effect.

The values of DE are found to be positive in recent empirical research: DE is equal to 0.05 in Odean (1998) and 0.21 in Dhar and Zhu (2006). Table 1 illustrates the findings of Weber and Welfens (2007) who observed the behaviour of 2614 individual investors in a set of field data from a Ger­man online broker.

The mean values of PGR[10], PLR and DE are given, furthermore the num­bers of subjects, which exhibit a DE > 0 or a DE < 0. Weber and Welfens (2007) find DE to be approximately 0.09. Although 9% seems to be a low number for the amount of additional propensity to sell winners more than losers, it corresponds to the statement that winner stocks are over 50 per­cent more likely to be sold than loser stocks are (Both Odean, 1998 and Weber and Camerer, 1998 find this). This follows from an average ratio of PGR to PLR equal to 1.5 (pR = 1.5) for the entire year.[11] Effects that are found on the aggregate level in empirical field data can often be explained differently. Section 5.2 shows that, in contrast, the ex­perimental method is able to distinguish the different drivers of trading behaviour and to observe their influence separately.

Neoclassical economic theory that assumes rational decision making cannot explain the data Odean (1998) and Weber and Welfens (2007) found in naturalistic environments. Therefore, a plenitude of new theo­ries evolved, which try to accurately describe people’s behaviour (positive) more than to be solely prescriptive (normative) (Shefrin, Statman, 1985). The previous section shortly explains why researchers have to build theo­ries before testing the disposition effect in experiments and introduces two theories, which can possibly explain how the disposition effect arises.

4 Theoretical Explanation of the Disposi­tion Effect

4.1 Matching Theories to Empirical Data

To understand the data that is found in naturalistic observations, it is nec­essary to build theories, which can explain the circumstances and predict these outputs. Theories are therefore build in a way that their output matches empirical data in the best possible way. The developed theory is an internally consistent and general system of statements, which implies the important facts of reality and does not display the non-relevant facts of reality (Wolf, 2008). Theories therefore illustrate what happens, how it happens and why something happens, in a measurable and concrete way. After a theory was build to illustrate reality, it can be tested, whether it predicts empirical findings and whether its hypotheses are supported. It is therefore essential to develop or to take an existing theory on the disposi­tion effect first, before testing the effect and its causes in an experiment, which means testing the hypotheses that were deducted from the theory. This thesis only reviews two possible theoretical explanations for the dispo­sition effect: The ’’prospect theory” (PT) of Kahneman and Tversky (1979) and the ’realization utility model” (RUM) of Barberis and Xiong (2012) in combination with ’mental accounting” of Thaler (1999). The prospect theory is reviewed because of its prominence (Barberis and Xiong, 2009) and because it is typically used in research papers to explain the disposi­tion effect (Kaustia, 2010). The realization utility model is the most recent explanation that extends PT, combines economic and psychological sci­ence and was also recently tested in experiments that found broad support (Frydman, 2011).

4.2 Prospect Theory

Taking the example from the beginning, the investor buys a share to a purchase price of $100. It raises to $150 during the first period. Suppose his reference point is the purchase price. Instead of evaluating his shares with the entire wealth that they implement ($150 in this case), he will always compare the current price to the reference point and therefore frames the share as a momentarily gain of +$50 ($150-$100). This way to evaluate the worth of a share is called the reference point effect (Jungermann et al., 2010). The investor now faces a decision problem in which he has to choose between two alternatives (A1 and A2):

A1: Selling the share to a price of $150 and realizing a gain of:

illustration not visible in this excerpt

With E(A): Expected value of alternative A.

A2 : Holding the share and thus accepting a risky lottery with a proba­bility of a that the share rises again to an amount of $150 + x and a probability of 1 — a that the share falls to $150 — y. The expected value how the investor frames A2 is:

illustration not visible in this excerpt

Notice that A1 is a risk-free alternative, because the investor earns $150 surely out of his sale. On the other hand, A2 is an uncertain/risky al­ternative, meaning that the concrete value of the output is unknown but different values can appear in different states, which have certain probabil­ities to arise.[12] To reduce complexity, we assume that the expected value of A1 is equal to the expected value of A2 (E(A1) = E(A2) = 50$). The investor’s choice now depends on his preferences: Is he supposed to be risk­averse, he will sell the share. If he is risk-seeking, he will accept the risky lottery and hold the share. In the domain of gains, investors normally be­have risk-averse (Wagner, 2011). Kahneman and Tversky (1979) observed that people’s behaviour in the domain of losses seems to be a mirror image of the behaviour in the domain of gains. They called this phenomenon the reflection effect. Out of their findings resulted the prospect theory that is visualised in Figure 1 and shows that investors are risk-averse (with a con­cave utility function) in the domain of gains and risk-loving (with a convex utility function) in the domain of losses, always framing a share as a loss or a gain by comparing the current price to the reference point. Thus, the prospect theory predicts that investors will sell their risky shares in the domain of gains and hold their shares (and accept to face a risky lottery) in the domain of losses.[13] In other words, they experience a disposition effect. Another possibility is that the investor has to decrease the future expected return of the share after the first period. In the domain of losses, he could find that the share’s expected return still justifies its risk, while in the domain of gains, it could not justify it (Odean, 1998). Once more, the disposition effect results.

illustration not visible in this excerpt

Moreover, people experience ”loss aversion” (Kahneman and Tversky, 1979; Camerer, 2000). They try to avoid making losses if possible and as long as possible. They even dislike losses a lot more than they like gains! This aspect is illustrated in prospect theory as you can see in Figure 1 : although G = L (with V(x): value of x, L: amount of loss, G: amount of gain). It follows that investors delay realizing their losses, because they crave any possibility to ”break-even”, to avoid facing their losses, and again they are willing to take the risk and hold the losers.[14]

illustration not visible in this excerpt

[...]


[1] Tt follows the denotation of e.g. Weber and Welfens (2007).

[2] See Odean, 1998; Dhar and Zhu, 2006; Weber and Welfens, 2007; Ben-David and Hir- shleifer, 2012

[3] See Weber and Camerer, 1998; Weber and Welfens, 2007; Frydman et al., 2011

[4] Shefrin and Statman, 1985; Odean, 1998; Weber and Camerer, 1998; Weber and Welfens, 2007; Frydman et al., 2011

[5] This finding at least counts for short term predictions. Odean (1998) finds that recent winner’s average return over the following year is 3.4% higher than recent loser’s return.

[6] Weber and Welfens (2007) observe in their experiment subjects’ average income to be €12.32, while the income is €13.28, if subjects would follow the simple decision rule to always hold the shares with the highest price.

[7] Because tax motivated trading is not supposed to be the topic of this thesis, it concen­trates on pure testing of the disposition effect in the following text. Nevertheless, it is important to recognize that disposition effect is only one motive amongst others that drives trading behaviour.

[8] Suppose that the stock market value rises and there are much more winners than losers in investors’ portfolio. And suppose that on average more winners are realized than losers are. This data is not yet sufficient to show that a disposition effect exists, because it is still possible that investors tend to sell winners and losers with the same rate or even sell losers with a higher probability (Odean, 1998).

[9] After Popper’s (1935) "critical rationalism”, theories cannot be verified. Hypotheses, which are deducted from theories should be falsified instead to sort out false theories. Rejecting the null hypothesis in statistics therefore supports researchers’ hypothesis i.

[10] Weber and Welfens (2007) denote PGR as ”PWR” in their paper. The two ratios are equally defined.

[11] This ratio though is not constant over time. Its value falls from January to December like it is displayed in figure 5 (appendix). A possible explanation for the ratio to fall is tax motivated selling that is also investigated by Odean (1998). Because his findings are not correctly reviewed without mentioning this driver of trading behaviour at the end of the year, it is reviewed in the appendix.

[12] Hence, the definition of risk in economics differs from everyday parlance, which always aims at the negative value of the output.

[13] Prospect Theory as an explanation is also reviewed in Shefrin and Statman (1985), Odean (1998), Weber and Camerer (1998), Jungermann et al. (2010), Frydman et al. (2011).

[14] Moreover, it follows that when the share increased in the last episode and the investor’s reference point heightens with it, he is such afraid of the share’s value decreasing in the next period and of realizing it for a lower price than the current price is, that he prefers selling the winner immediately and assuring the current value.

Details

Pages
39
Year
2013
ISBN (eBook)
9783656404453
ISBN (Book)
9783656404682
File size
934 KB
Language
English
Catalog Number
v212299
Institution / College
Christian-Albrechts-University of Kiel – Institut für Quantitative Betriebs- und Volkswirtschaftliche Forschung
Grade
1,3
Tags
disposition effect Dispositionseffekt BWL Experiment experimental evidence neural bias rational irrational methodical Methode method validity Validität nuisance Störvariablen Theorie Empirie empirical theoretical prospect theory realization utility Odean Frydman Weber Camerer

Author

Share

Previous

Title: Experimental Evidence on the Disposition Effect