Risk Aversion in International Stock Exchanges:
Cultural Differences or Global Uniformity?
An empirical and comparative case study of the Disposition Effect in the
London Stock Exchange (LSE) and the Shenzhen Stock Exchange (SZSE)
Name Thomas Gabriël Baak Student number 10779132 Programme BSc Economics and Business Specialization Finance and Organization Coordinator dr. P.J.P.M. Versijp Supervisor I. Sakalauskaite Date January 2017 ABSTRACT The disposition effect, an anomaly encountered in behavioural finance, is researched through analysing data on the trade volumes of the individual stocks in the London Stock Exchange (LSE) and Shenzhen Stock Exchange (SZSE) from 2010–2014. According to the Disposition Effect, investors sell value winning stock while holding on to those that are losing value. In this panel data based empirical research, abnormal trade volumes are explained by evaluating changes in asset pricing respective to the stock exchange the stock is traded in, and whether Chinese stocks are classified as A- or B-shares. With the data at hand, evidence on the Disposition Effect is found. Results indicate that (i) the Disposition Effect is smaller in the SZSE than in the LSE, and (ii) the Disposition Effect is smaller for A-Shares than for B-Shares. With Prospect Theory and the Cushion Hypothesis in mind, there is indication that ChineseStatement of Originality This document is written by Student Thomas Baak, who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Table of Contents 1. Introduction 4 2. Literature Review 7 2.1 Prospect Theory 7 2.2 Disposition Effect 9 2.3 Cultural Differences in Risk Preferences 10 2.4 Hypotheses 11 3. Methodology 12 3.1 Choice of Stock Exchange 12 3.2 Data 12 3.3 Research Method 14 3.4 Capturing Cultural Diversity in Trade Volumes Analysis 15 a. Testing Hypothesis 1 15 b. Testing Hypothesis 2 16 c. Testing Hypothesis 3 17 4. Results and Analysis 18 4.1 Disposition Effect in the LSE and SZSE 18 a. Part 1: Finding the Disposition Effect 19 b. Part 2: Indication of Trade Volume Differences Exchanges 20 4.2 Disposition Effect in China 22 4.3 Reverse Disposition Effect in the LSE and SZSE 25 5. Conclusion and Discussion 27 6. Bibliography 29 7. Appendix 31 7.1 List of Variables 31 7.2 Creating the Summary Statistics 32 7.3 Disposition Effect in Separate and Combined Market Settings 33 7.4 Reverse Disposition Effect for Each Individual Exchange 36 7.5 Indication of Discrepancy Between LSE & SZSE 41 7.6 Discrepancy Between SZSE A- and B-Shares 42
1. Introduction
After centuries of explaining economic phenomenon and framing them as physical law, it has become apparent that human behaviour provokes anomalies that are difficult if not impossible to fit into assumptions of rationality. Key in making theories such as the Efficient
Market Hypothesis viable, rational expectations seem to not only be determined by rational
outlook, available information and past experience, but also by human psychology (Shiller, 2003). Therefore, expectations on the future state of the economy depend also on psychological, social, cognitive and emotional factors of those individuals making economic decisions in it. In the 90’s, according to Shiller, this realisation gave rise to Behavioural Finance, a scientific field that combines theory from conventional economics, finance and psychology to explain irrational financial decisions. One well-known anomaly observed in the financial market is known as the Disposition Effect. First labelled by Shefrin and Statman (1985), it describes the tendency of investors to sell value winning stock, and keep those stocks that experience a drop in value. This reluctance to realize losses appears to be incompletely explainable by rational explanations such as portfolio rebalancing or tax strategies (Odean, 1998). However, Kahneman and Tversky’s
Prospect Theory offers a solution (1979). Prospect Theory is about decision-making under
uncertainty and, in subsequent research, defines Diminishing Sensitivity: investors are risk-averse for gains and risk-seeking for losses (Tversky & Kahneman, 1992). It is Diminishing
Sensitivity that researchers often cite as the underlying mechanism of the Disposition Effect (Li & Yang, 2013). Despite the fact that there have been lots of different types of research confirming this phenomenon: empirical, experimental to purely theoretical. And that there are contributions on specific variables that affect the magnitude of the effect, such as gender (Rau, 2014), or the influence of tax on trading volumes (Odean, 1998). There is very limited research to be found in the domain of cultural differences having an impact on the disposition effect (Chen, Kim, Nofsinger & Oliver, 2007). In a world that is globalizing at unprecedented rates through the rise of connecting computer networks and the consequent inflow of foreign investors on stock exchanges, understanding the effects of culture appear to be essential in apprehending investor behaviour. Of course, with financial markets linked and integrated globally, it is very difficult to statistically prove that there is a component of the disposition effect that can be assigned to
a certain type of people, culture or way of living without data on the specific origin of the traders involved in a specific stock trade. Yet opportunities for research arise through financial markets that have been shielded from outside investors by their governments. China provides an interesting and useful stock exchange system in which only part of the stock market is open to foreign investors. On the Shenzhen Stock Exchange (SZSE), only B-shares are available to both foreign and domestic investors, while A-shares are limited to purchase by domestic investors only. If stock trading activity in respect to asset pricing is compared between those of the domestic-purchase-only-exchange, Chinese A-shares, and those of a globalized exchange, the London Stock Exchange (LSE), the cultural side of the disposition phenomenon could be researched. At the same time, through the connection the Disposition Effect and
Prospect Theory share, from data analysis China specific Risk Aversion relative to that of the
world could be inferred. This could prove vital in the future as China’s influence and participation in global affairs is every increasing. With this opportunity in mind, the following research question is to be answered in this bachelor’s thesis: In this empirical research, the Disposition Effect is evaluated by creating a combined sample of individual stocks from both the SZSE and LSE in the interval 2010-2014. With this data, abnormal trade volumes on individual stocks is determined by regression. In a second regression, the abnormal trade volumes will then be coupled with their respective price changes to find a disposition effect. This approach is similar to what Ferris, Haugen and Makhija used to analyse the Disposition Effect in their 1988 paper. To gather evidence on a possible cultural effect, dummies and interaction variables on specific stock exchange and stock type are added. Results found in this research indicate that (i) the Disposition Effect is smaller in the Shenzhen Stock Exchange than in the London Stock Exchange, and (ii) the Disposition Effect is Does a cultural difference in terms of Risk Aversion exist between the people of China and the globalized financial markets: Is there a difference in the Disposition Effect between the Shenzhen Stock Exchange (SZSE) and London Stock Exchange (LSE)?
The remainder of this thesis is structured as follows. In paragraph 2 existing literature is reviewed through the sections: 2.1 Prospect Theory, 2.2 Disposition Effect, 2.3 Cultural Differences in Risk Preferences, and 2.4 Hypotheses. After that, in paragraph 3, the research methodology is introduced through sections: 3.1 Choice of Stock Exchange, 3.2 Data, 3.3 Research Method, and 3.4 Capturing Cultural Diversity in Trade Volume Analysis. In paragraph 4 performed data analysis and results are discussed: 4.1 Disposition Effect in the LSE and SZSE, 4.2 Disposition Effect in China, and 4.3 Reverse Disposition Effect in LSE and SZSE. Finally, in paragraph 5, concluding remarks are made, placed into perspective and suggestions are made for further research.
2. Literature Review
In this paragraph, Prospect Theory as developed by Kahneman and Tversky (1979, 1992) is explained in section 2.1. With the foundations of Prospect Theory in mind, in section 2.2, investor trading behaviour is then clarified by examining work of Shefrin and Statman (1985) on the Disposition Effect and other reasoning for the sale and purchase of shares by investors will be evaluated. Notably taxes, as found by Odean (1998), are introduced as a possible cause for the Reverse Disposition Effect found in some months of the year. Afterwards, in section 2.3, research on risk preferences and loss aversion conducted at a national and cultural level are discussed to form expectations on the cultural side of the disposition effect. Finally, in section 2.4, hypotheses are formed combining existing literature and expectations. 2.1 Prospect Theory After Expected Utility Theory was under increased criticism as specific axioms proved to be systematically insatiable in reality, Kahneman and Tversky (1979) constructed Prospect Theory as an alternative. According to Prospect Theory, two phases are distinguished in the choice-making process: editing and evaluation. In the editing phase, a preliminary analysis is made of the offered prospects. Subsequently, in the evaluation phase, the prospects are valued and the one of highest value is chosen. The true insight the theory offers is through assigning value to gains and losses rather than to final wealth levels in the editing phase. Explained by Kahnemand and Tversky as
Coding, gains and losses are defined relative to some neutral Reference Point. For individuals active in the financial markets the reference point is often the purchase price of an asset (i.e. shares). Usually, the gains and losses coincide with the current asset position relative to the initial payment made to attain the asset. However, the reference point and subsequent coding of outcomes as gains or losses, can be affected by the formulation of the offered prospects, and thus effectively by the decision maker (Kahnemand & Tversky, 1979). Three further classifications in the editing phase are made: Combination, Segregation and Cancellation. Firstly, combination through realizing that some prospects have the same outcomes and thus can be combined. Secondly, segregation through considering that some prospects consist of a risky- and a riskless component. And finally, cancellation through an
Through experiments with participating students and university faculty staff, Kahneman and Tversky find that individuals indeed evaluate outcomes, according to their perception of gains and losses relative to a reference point. Furthermore, as summarized by Li and Yang (2013), it is found that individuals are more sensitive to losses than to gains of the same magnitude as per Risk Aversion. And finally, also labelled the Reflection Effect, that investors are risk-averse for gains and risk-seeking for losses.
Subsequent to these findings, Kahneman and Tversky (1979) propose a Value Function on deviations from the reference points that is (i) steeper for losses than for gains; and (ii) generally concave for gains and commonly convex for losses. Note that the S-shaped function, a consequence of these characteristics, is steepest at the reference point and displays insensitivity to gains and losses at the extreme ends.
In 1992, Tversky and Kahneman advance Prospect Theory by reviewing the way decision weights are formed. Decision weights are inferred from choices between prospects much as subjective probabilities. The implication of the separable weighing system in the original 1979 paper is that all unlikely events are overweighed, yet judging from experiments with physical participants, only extreme events appear to be overweight. The advanced version of Prospect Theory called Cumulative Prospect Theory, employs cumulative rather than separable decision weights. Mechanically similar to the system found in Rank-Dependent
Expected Utility, this allows for exclusively extreme events to be overweight by using the
cumulative probability distribution function (Quiggin, 1982).
With the weights subject to a different weighting system, also the shape of the value function is revisited by reviewing past experiments. Notably, the inequality of steepness in
gains and losses near the reference point is assigned to Loss Aversion. This is the tendency of individuals to assign less value to gains than they do to losses of the same magnitude. Meanwhile, the insensitivity to gains and losses at the extreme ends of the value function is labelled as Diminishing Sensitivity. New experiments furthermore point towards a “fourfold pattern of risk attitudes”: there appears to be risk-aversion for gains and risk-seeking for losses of high probability, while there is risk-seeking for gains and risk-aversion for losses of low probability (Tversky & Kahneman, 1992). 2.2 Disposition Effect Prospect Theory describes that individuals that made losses recently are seeking risks that would otherwise be unacceptable for them. Contrary to the controlled experimental settings in which Kahneman and Tversky assess Prospect Theory, Shefrin and Statman (1985) test whether this finding also holds true in real-world market settings. Through examining decisions to realize gains and losses in the financial markets, results indicate that investors appear to exhibit a reluctance to realize losses. This is visible through the tendency of investors to sell value winning stock, and keep those stocks that experience a drop in value. The “disposition” to ride losers longer than rationally explainable by non-behavioural theory, is named the Disposition Effect. Besides the explanatory properties of the Disposition Effect through Prospect Theory’s indication of an individual’s aversion to loss realization, Shefrin and Statman (1985) propose four additional explanations: Regret Aversion, Self-control and Tax Considerations. Firstly, investors may resist the realization of a loss because it stands as proof that their initial judgement at moment of purchase was wrong. Known as Regret Aversion, it is related to emotional regret that with ex post knowledge different past decisions would have fared better. The positive counterpart of regret aversion is to be found in the quest for Pride of selling a value winning stock. Together, Pride and Regret Aversion could theoretically lead to the disposition to realize gains and defer losses. Secondly, the Disposition Effect could be caused by a lack of Self-control through unwillingness to accept the loss out of regret, and selling too quickly to pursue pride. And finally, the Disposition Effect is explained by an investor’s Tax Considerations.
investor’s incentive to lower its tax liability. Lakonishok and Smidt (1986) expanded theory on tax related influences by testing for tax effects in all months of the year. However, only evidence is found on tax motivated trading in December and January. In later research by Odean (1998), exclusively tax considerations are found to cause Reverse Disposition in December, while explanations of the Disposition Effect based on portfolio rebalancing and trading costs are rejected.
Recently, research has also been conducted on the Disposition Effect in the absence of taxes. Firth (2015) finds that, without taxes, investors still have the tendency to hold on to value losing stock. This further emphasizes the claim that tax motivated trading does not explain the overall pattern in stock trading volumes which lead to the Disposition Effect (Shefrin & Statman, 1985).
2.3 Cultural Differences in Risk Preferences
Through the many explanations offered for the Disposition Effect, it has become apparent that abnormal trading activity, as summarized by Kanehiro and Shoji (2016), is due to a combination of emotional demeanour (Regret Aversion), fiscal considerations (Tax Aversion) and mostly an individual’s perception of risk (Loss Aversion). These elements seem to be highly correlated with an individual’s upbringing or culture. As researched by Hofstede (1980), a society’s culture affects values and behaviour of its members. To model a society’s effect on behaviour, Hofstede’s cultural dimensions theory was developed on the foundations of Power
Distance, Individualism, Masculinity, Uncertainty Avoidance, Long Term Orientation, and Indulgence.
In search of a possible culture related explanation of differing risk preferences
between Chinese and American investors, Hsee and Weber (1998) analysed the Collectivism-Individualism continuum as found in Hofstede’s cultural dimensions theory to propose the Cushion Hypothesis. In the Collectivism-Individualism continuum a distinction is made whether
people in a specific country prefer to be treated as individuals in their society or prefer to act as a collective (Hofstede, 2001). In collective societies, it is argued that people are more likely to receive help. With this in mind, Hsee and Weber propose the Cushion Hypothesis: the impact of adverse outcomes in risky options is lower for people in collective societies and thus perceived risk and risk adversity are also lower. Contrary, in individualistic societies losses are more financially disturbing and thus perceived risk and risk adversity are higher.
Conceptualized in finance by evaluating people’s willingness to pay (WTP) for a certain risky option, evidence was found that there appear to be cross-cultural variations in WTP (Hsee & Weber, 1998). While American respondents took less risk, Chinese respondents appeared to be risk neutral leading to statistical evidence on the Cushion Hypothesis for respectively individualistic and collectivistic societies. As per Hofstede’s theory (1980), Americans were assumed to be individualistic while Chinese were collectivistic.
2.4 Hypotheses
Judging from past literature on the subject, the Disposition Effect is directly connected to irrational human behaviour on the stock exchanges through aversion for losses, regret and taxes. Culture wise, there seems to be a strong relation between collectivism and risk neutrality. In this thesis it is therefore expected that, in a comparison of traders active on the London and Chinese Stock Exchanges: 1) A Disposition Effect is to be found in the Shenzhen Stock Exchange that is smaller than the one in the London Stock Exchange reflecting the apparent risk neutrality of Chinese Investors. 2) A Disposition Effect is to be found in the Shenzhen Stock Exchange that is smaller for A-Shares than for B-Shares reflecting the apparent risk neutrality of Chinese Investors. 3) A Reverse Disposition is to be found in both the Shenzhen Stock Exchange and London Stock Exchange in December due to tax motivations of investors worldwide.
3. Methodology In this paragraph, the methodology applied in this thesis is established. First, in section 3.1, the choice of the particular stock exchange and subsequent value to explaining the Disposition Effect through cultural effects is explained. In section 3.2, the origin of the data, different variables, conducted mutations and chosen sample format are discussed. In section 3.3 and 3.4, the method of research and model setup is performed. 3.1 Choice of Stock Exchange
In this empirical research, the Disposition Effect is evaluated on a possible culture-based difference in investor trading activity between the Shenzhen Stock Exchange (SZSE) and London Stock Exchange (LSE). To assess cultural differences, a comparison is made of abnormal trade activity in the LSE and SZSE.
The SZSE functions as a sample on Chinese-specific trading activity because of the unique setup of the exchange: stock is divided in A- and B-Shares. A-Shares are not purchasable by foreign investors, while B-shares are. Effectively, this creates a laboratory experiment in which outside influences (in this case trade behaviour of international investors) can be identified, tested and excluded from the sample population. At the same time, the LSE is treated as a sample on overall, world-average trading activity as it is accessible to all nations and associated citizens. 3.2 Data A combined sample of individual stocks is created from both the SZSE and LSE in the time interval 2010-2014 (60 months). Data (monthly) was obtained from the database Orbis on the following three variables: (i) Trade Volume, (ii) Outstanding Stock, and (iii) Closing Stock Price in US$. The resulting dataset consists of 60 observations for each of the 3 different variables over 1945 individual companies in the LSE, and 1827 individual companies in the SZSE. Through random selection, a sample is created consisting of 201 individual companies: 101 companies from the SZSE and 100 from the LSE. Of the 101 individual companies in the SZSE sample, 52 companies are of foreigner-protected A-Share status while the other 49 are of foreigner-infected B-Share status. To forgo any statistically precarious inconsistencies caused by missing data, the sample consists only of individual stocks with at least 30 observations per variable.
Throughout the research, three variables (trade volume, outstanding stock and closing stock price) are used in different sample-settings and configurations that can be best summarized in the following Summary Statistic: Based on the original three variables stated above, for each sample three new variables are created: (i) Total Trade Volume; the cumulative of Trade Volume of all stock for a given month. (ii) Total Shares Outstanding; the cumulative of Outstanding Stock of all stock for a given month.
Classification of ∆Stock Price Changes in %
Ranges Variable Name Interval
1 RisePrice [ 0 ; +5% ] 2 RisePrice2 [ +5% ; +15% ] 3 RisePrice3 [+15% ; +∞ ] 4 FallPrice [ -5% ; 0 ] 5 FallPrice2 [ -15% ; -5%] 6 FallPrice3 [ -∞ ; -15%] (iii) % (Stock) Price Change; stock price performance over the past month is computed through: !" $ !"%&
!"%& , where Pt = Closing Stock Price in month t and Pt-1 = Closing Stock
Price in month t-1.
3.3 Research Method
To determine the Disposition Effect, a similar approach to Ferris, Haugen and Makhija (1988) is used. First, monthly abnormal trade volumes are estimated using a “market”-model regression for volume: Determining abnormal trade volume 𝑉() = 𝐴( + 𝐵(𝑉(.+ 𝜀() where, Vit = turnover for stock i in month t = )5)9= 01.234 56 789437 56 7)5:; ( 51)7)90<(0> (0 .50)8 )01.234 56 789437 56 7)5:; ( )49<3< (0 .50)8 ) Vim = )5)9= 01.234 56 9== 7)5:;7 51)7)90<(0> (0 .50)8 )01.234 56 789437 56 9== 7)5:;7 )49<3< (0 .50)8 ) εit = abnormal turnover for stock i in month t
The abnormal turnover of stock i, the residual εit, can then be found by regressing Vit on Vim.
However, this does not prove any dispositional effect as a result of investor behaviour. Merely that in some months there is more trade activity. To research the relationship between abnormal stock turnover, εit, and stock performance, six price ranges are created. Specific stock price performance over the past month are identified in terms of percentage change. According to the magnitude of change in stock price, the stock is then assigned to a specific range in binary fashion {0, 1}.
H"#: β' > 0 for k = 1, 2, 3 and β' < 0 for k = 4, 5, 6
H"#: β'< 0 for k = 1, 2, 3 and β'> 0 for k = 4, 5, 6
Effectively, each range functions as a dummy variable on price changes of a certain magnitude. Value gaining stock is typically classified in the ranges 1 through 3, while value losing stock is inserted in the ranges 4 through 6.
To find evidence of the Disposition Effect, panel data analysis is performed on price changes, associated trade volumes of individual stock and specific month of trade. This amounts to the following regression with abnormal trade volume, εit, as the dependent
variable, and the six ∆Stock Price ranges as the independent explanatory variables:
In case of a Disposition Effect, investors sell stock value gaining stock (driving abnormal trade volumes, εit, up) and keep value losing stock (driving εit down). Therefore, it is expected
to find statistical evidence in the form of the following coefficient pattern under:
Under tax-motivated trading, it is expected to find a Reverse Disposition Effect in December-months statistically taking the following form: 3.4 Capturing Cultural Diversity in Trade Volumes Analysis Having established an empirical approach to measure the Disposition Effect through abnormal trade volumes, this section focuses on the specific regression analysis that is pursued in this thesis to make statistical inference on cultural effects possible. 3.4a Testing Hypothesis 1
As proposed in section 3.3, causes for abnormal trade volume are explained through six dummy variables on price changes of company specific stock. This approach is utilised in regression 1 to find statistical evidence of the Disposition Effect: the LSE and SZSE are analysed for abnormal trade anomalies separately and as a combined sample with stocks of both
ε#$= α'+ β*RisePrice + β2RisePrice2 + β4RisePrice3 + β6FallPrice + β:FallPrice2 + β;FallPrice3 + u#$
Regression 1
εit = α0+ β1RisePrice + β2RisePrice2 + β3RisePrice3 + β4FallPrice + β7FallPrice2 + β8FallPrice3 + ζ
Regression 2
εit = α0 + β1RisePrice + β2RisePrice2 + β3RisePrice3 + β4FallPrice +
β7FallPrice2 + β8FallPrice3 + β9RisePricexSZSE + β>RisePrice2xSZSE + β?RisePrice3xSZSE + β@AFallPricexSZSE + β@@FallPrice2xSZSE +
β@BFallPrice3xSZSE + β@CSZSE + ζ Statistical evidence of the Disposition Effect is found when coefficients on RisePrice (β {1, …, 3}) are > 0, and coefficients on FallPrice (β {4, …, 6}) are < 0.
Next, the combined sample of both LSE and SZSE stock is analysed further for exchange specific trading behaviour in regression 2. This is done through introducing both an indicatory interaction variable and control variable on exchange the stock was traded in: SZSE. In dummy variable SZSE, stock is assigned a value of 0 if it is in the LSE and 1 if it is a SZSE. The variable is then multiplied by the six stock price ranges RisePrice and FallPrice, creating six interaction variables (β {7, …, 12}).
For the Disposition Effect, coefficients on RisePrice ( β { 1, … , 3}) must be > 0, while coefficients on FallPrice (β {4, …, 6}) must be < 0. Indication of cultural differences at play arise when one or more of the interaction variables (β {7, …, 12}) in regression 2 are significantly different from zero. The coefficient on SZSE, β13, functions as a control variable. 3.4b Testing Hypothesis 2 Even in case cultural differences between LSE and SZSE are found after analysis of results in regression 2, the question remains whether any cultural differences are directly detectable between a sample adjusted to the influences of international investors. This can be tested by using the Chinese exchange mechanism of A- and B-Shares in the SZSE. Foreigner-protected A-Shares and foreigner-infected B-Shares are identified through an indicatory binary dummy variable {0, 1} on B-Shares: B. In dummy variable B, stock is assigned a value of 0 if it is an A-Share and 1 if it is a B-Share. The variable is then multiplied
Regression 3
εit = α0 + β1RisePrice + β2RisePrice2 + β3RisePrice3 + β4FallPrice + β7FallPrice2 + β8FallPrice3 + β9RisePricexB + β<RisePrice2xB + β=RisePrice3xB + β>?FallPricexB + β>>FallPrice2xB +
β>@FallPrice3xB + ζ
Regression 4
εit = α0 + β1RisePrice + β2RisePrice2 + β3RisePrice3 + β4FallPrice + β7FallPrice2 + β8FallPrice3 + ζ by the six stock price ranges RisePrice and FallPrice, creating six interaction variables (β {7, … , 12}). Indication of differences in stock trading behaviour arise when one or more of the interaction variables on B-shares (β {7, …, 12}) in regression 3 are significantly different from zero. As discussed earlier, for the Disposition Effect, coefficients on RisePrice (β {1, …, 3}) must be > 0, while coefficients on FallPrice (β {4, …, 6}) must be < 0.
3.4c Testing Hypothesis 3
Lastly, the Reverse Disposition Effect is tested. By regressing only December-month related trade activity, inference can be made about tax-motivated trading around the world. The following samples are tested separately: LSE, SZSE, SZSE A-Shares, and SZSE B-Shares.
For the Reverse Disposition Effect, coefficients on RisePrice (β {1, …, 3}) must be < 0, while coefficients on FallPrice (β {4, …, 6}) must be > 0.
4. Results and Analysis
In this paragraph the results of the regressions specified in the Methodology section are analysed on the basis of three tables containing this research’s most important findings. Each table is discussed in a separate section: 4.1 Disposition Effect in the LSE and SZSE, 4.2 Disposition Effect in China, and 4.3 Reverse Disposition Effect in LSE and SZSE. Subsequent implications to the hypotheses are discussed.
Tables are constructed in the following fashion: the first column of each table contains the name of the coefficients regressed, while the first row specifies the sample group tested and dependent variable Abnormal Trade Volume (ATV). Following the coefficient’s approximated value sometimes an asterisk can be found: a single asterisk behind the coefficient value indicates a significance level of 10%, while two and three asterisks respectively indicate a 5% and 1% level of significance. Between brackets the individual standard deviations are noted. Furthermore, at the bottom of each table, a short descriptive statistic is given on the number of observations (NObservations), number of individual companies
with shares in the sample (NGroups), and the amount of variance in abnormal trade volume
explained by the model (R2). Additional statistics, direct Stata output, and experimental instructions can be found in the Appendix section (paragraph 7). 4.1 Disposition Effect in the LSE and SZSE In this section, statistical evidence on the Disposition Effect is first analysed on the basis of stock price increases and decreases through the variables RisePrice and FallPrice. In the second part, interaction variables are added on origin of exchange the stock is traded in. Summarized in Table 1 (page 21), it contains results on regression 1 and 2 as specified in the methodology section. In regression 1, evidence of the Disposition Effect is researched in both a separate and combined sample of the LSE and SZSE. In regression 2, a distinction in a combined sample of LSE and SZSE is made for company stock inside the SZSE to research possible differences in Disposition Effect between the exchanges.
To better structure the analysis, each sample is referred to according to a certain number (also visible in the top row of Table 1). For regression 1: LSE (1), SZSE (2), and LSE & SZSE combined (3). Regression 2: LSE & SZSE combined (4).
4.1a Part 1: Finding the Disposition Effect
For statistical evidence on the Disposition Effect, coefficients on RisePrice must be > 0, while coefficients on FallPrice must be < 0.
Starting with results found in regression 1, unexpectedly in a sample of only the LSE (1) no significant coefficients on the variables RisePrice or FallPrice have been found. This indicates that changing prices in the LSE do not explain any abnormal trade volume, and thus no statistical evidence for the presence of the Disposition Effect is found. However, results from the SZSE-sample (2) do partly support the dispositionary effect of price on abnormal trade volume. In particular, the variable FallPrice2 (which indicates trade volume on assets that have fallen between 5% and 15% in value) is with a T-score of -2.12 significant at the 5% level and has a negative effect -873.065 on abnormal trade volume. Additional evidence of the Disposition Effect is found in a combined sample of LSE and SZSE (3) shares in variables RisePrice2, FallPrice, and FallPrice2. These variables follow the pattern of positive coefficients in RisePrice and negative coefficients in FallPrice: in RisePrice2 a positive coefficient is exposed of 277.471, with a T-score of 1.75, and significant at the 10% level. FallPrice has a negative coefficient value of -638.767, a T-score of -3.19, and is significant at the 1% level. Lastly, in FallPrice2 a negative coefficient is found of -578.934, a T-score of -2.39, and significance at the 5% level. Contrary to expectations set in Literature Review and Methodology, coefficients on FallPrice3 (which indicates trade volume on assets that have fallen over 15% in value) seem to follow a different pattern than expected. In both a sample group of stock in the SZSE (2) and a combined sample group consisting of LSE & SZSE stock (3), FallPrice3 coefficients are positive and significant. In SZSE (2), FallPrice3 takes a value of 2572.363, with a T-score of 1.65 significant at the 10% level. In LSE & SZSE (3), FallPrice3 is 3673.385, with a T-score of 2.95 and significant at the 1% level. These positive coefficients indicate a higher level of abnormal trade volume, instead of expected lower trade volume, at price drops of over 15%. As this phenomenon occurs in both the LSE and SZSE, culture does not seem a likely causal determinant. A possible explanation could be that investors do start selling value losing shares once the losses become of a certain magnitude. There might be a point at which an individual’s utility of holding on to a value losing stock does no longer satisfy the utility of
4.1b Part 2: Indication of Trade Differences Between the Exchanges Introduction of the interaction variables on RisePrice, FallPrice and whether stock is traded in the SZSE gives some valuable insights on the existence of trade differences between the LSE and SZE. Firstly, R2 increases significantly when compared to samples without SZSE-interaction variables: from 0.009 in (3) to 0.013 in (4). Secondly, there is evidence of the Disposition Effect in both RisePrice and FallPrice variables. At the 1% significance level, positive coefficients are found in RisePrice2 and RisePric3: RisePrice2 has a coefficient of 369.309 and T-score of 7.55. RisePrice3 at the same time shows a coefficient of 369.284 and T-score of 7.44. Meanwhile, negative coefficients at the 1% level are present in FallPrice and FallPrice2: FallPrice has a coefficient of -1442.527 and T-score of -5.81. FallPrice2 shows a coefficient of -1458.277 and a T-score of -3.46. Contrary to earlier regressions, the variable FallPrice3 is in this case insignificant.
In terms of variables on the origin of exchange the stock is traded in: the control variable SZSE, significant at the 5% level and T-score of 2.38, also has a positive coefficient of 268.151. This indicates that a higher amount of abnormal trade volume is detectable in the Chinese stock market. At the same time four out of six interaction variables entered show significant results: SZSExRisePrice has a negative coefficient of -288.375, T-score of -1.83, and is significant at the 10% level. SZSExFallPrice, SZSExFallPrice2, and SZSExFallPrice3 are significant at the 1% level with positive coefficients of respectively: 1611.397, 1365.83, and 4573.051. T-scores are correspondingly: 4.23, 2.70, and 2.79. Notice that the interaction variables give indication that in comparison with the LSE, in the SZSE less abnormal trade activity occurs when share prices increase and more when share prices fall. The Disposition Effect therefore appears to be smaller in the SZSE: interaction coefficients are negative in RisePrice and positive in FallPrice. Following this result, statistical inference can be made that in a combined sample of LSE and SZSE stock, the Shenzhen Stock Exchange indeed displays a smaller Disposition Effect than the one found in the London Stock Exchange (Hypothesis 1). However, it also has to be taken into account that in a sample of only the LSE (1) all variables on RisePrice and FallPrice were insignificant and thus no evidence of any Disposition Effect was found. This could be due to the amount of observations being inadequate.
(LSE) (SZSE) (LSE & SZSE) (LSE & SZSE)
ResidualLSE ResidualSZSE Residual Residual
RisePrice -0.712 165.483 3.423 117.696 (0.675) (144.551) (79.703) (110.999) RisePrice2 0.007 205.193 277.471* 369.309*** (1.320) (253.024) (158.203) (48.944) RisePrice3 -1.412 547.383 720.830* 369.284*** (1.979) (463.522) (401.595) (49.662) FallPrice -0.015 50.564 -638.767*** -1442.527*** (0.861) (295.828) (200.058) (248.139) FallPrice2 0.535 -873.065** -578.934** -1458.277*** (1.666) (411.460) (242.529) (421.606) FallPrice3 1.282 2572.363* 3673.385*** 109.223 (1.792) (1557.450) (1245.987) (248.989) SZSExRisePrice -288.375* (157.930) SZSExRisePrice2 -243.106 (292.349) SZSExRisePrice3 287.324 (503.291) SZSExFallPrice 1611.397*** (380.541) SZSExFallPrice2 1365.83*** (506.731) SZSExFallPrice3 4573.051*** (1638.486) SZSE 268.151** (112.439) Constant 0.104 -252.568** -146.189*** -251.970*** (0.238) (109.555) (50.507) (48.874) NObservations 5970 5708 11679 11679 NGroups 100 101 201 201 R2 0.001 0,004 0,009 0.013 (1) (2) (3) (4) TABLE 1: DISPOSITION EFFECT IN THE LSE AND SZSE ATV-LSE ATV-SZSE ATV-LSE & SZSE ATV-LSE & SZSE
4.2 Disposition Effect in China
In this section, statistical evidence on the Disposition Effect in specifically a sample of the SZSE is further analysed. First, the effects of price increases and decreases on abnormal trade volume are shown in samples of SZSE: A- Shares, SZSE: B-Shares, and SZSE comprising both stock types. This is done through the variables RisePrice and FallPrice. Thereafter, a distinction is made between A- and B-Shares by introducing interaction variables on respectively RisePrice, FallPrice and B: BxRisePrice and BxFallPrice.
Summarized in Table 2 (page 24), it contains results on regression 1 and 3 as specified in the methodology section. To better structure the analysis, each sample is referred to according to a certain number (also visible in the top row of Table 2): SZSE: A-Shares (1), SZSE: B-Shares (2), SZSE (3) which contains both type of shares, and SZSE (4) which contains both type of shares and makes use of B-Share indicatory interaction variables.
Again, for statistical inference on the Disposition Effect, coefficients on RisePrice must be > 0, while coefficients on FallPrice must be < 0.
Out of the samples (1), (2), and (3), only the SZSE: B-Shares sample (2), offers substantial evidence on the Disposition Effect. Results show that: RisePrice has a positive coefficient of 438.156, T-score of 1.88, and is significant at the 10% level. FallPrice2 has a negative coefficient of -2305.602, T-score of -2.77, and is significant at the 1% level. Meanwhile, FallPrice3 has a negative coefficient of -1630.729, a T-score of -1.80, and is significant at the 10% level.
Contrary, results from the SZSE: A-Shares sample (1) do not support the Disposition
Effect. FallPrice3 is the only significant coefficient found, positive in nature with a value of 5196.503. Associated T-score is 2.16 at a significant level of 5%.
Premature statistical inference from results between SZSE: A-Shares (1) and SZSE: B-Shares (2) do point towards the possibility that Chinese investors are indeed more Risk
Neutral. However, question remains whether this inference can also be made when both A- and B-Shares are in a combined sample. Partly supportive and contradictory evidence of the Disposition Effect is found in the SZSE sample (3). The variables FallPrice2 and FallPrice3 are significant: FallPrice2, as expected, has a negative coefficient of -873.065, a T-score of -2.12, and is significant at the 5% level. Conflict with expectations arises in FallPrice3m which has a positive coefficient of 2572.363 at the significance level of 10%, and associated T-score of 1.65.
Having found inconsistent results in sampling the SZSE in both A- and B-Shares settings, interaction variables on B-Shares are entered into the regression (4) to determine the abnormal trade volume caused by B-Shares. Contradictory to the Disposition Effect, FallPrice3 has a positive coefficient of 5475.945, T-score 2.32, and is significant at the 5% level. At the 1% significance level, variables BxFallPrice2, and BxFallPrice3 are found with respectively negative coefficient values of -2888.698 and -7354.581. Associated T-scores amount to -3.57 and -2.96. BxRisePrice3 has a negative coefficient of -1453.528, T-score of -1.78, and is significant at the 10% level.
Inferred from these results can be that variables RisePrice up to and including FallPrice3 are unsupportive of any Disposition Effect. Negative coefficients in BxFallPrice2 and BxFallPrice3 indicate that abnormal trade volume in B-Shares is less when stock prices drop, as per Disposition Effect. Contrary to the Disposition Effect, BxRisePrice3 expresses a significant negative coefficient where theory predicts a positive one.
Despite the ambiguous results through a negative coefficient BxRisePrice3, overall the interaction variables seem to emphasize results found earlier: trade activity in the B-Shares part of the SZSE supports the Disposition Effect, whereas in the A-Shares part it does not. Furthermore, this confirms Hypothesis 3 which states that inside the Shenzhen Stock Exchange a smaller Disposition Effect is to be found in A-Shares compared to B-Shares.
With the Cushion Hypothesis in mind, in this research evidence is found that the Chinese people’s apparent collectivistic culture and risk neutrality also influences stock trading behaviour. As a result of not having the tendency to hold on to losing stock due to risk neutrality, the Disposition Effect would not show up.
TABLE 2: DISPOSITION EFFECT IN CHINA
(SZSE: A-Shares) (SZSE: B-Shares) (SZSE) (SZSE)
ATV-SZSE ATV-SZSE ATV-SZSE ATV-SZSE
RisePrice -138.579 438.156* 165.283 140.864 (164.742) (232.961) (144.551) (123.653) RisePrice2 10.094 332.678 205.193 289.536 (373.292) (308.970) (253.024) (354.861) RisePrice3 941.608 15.430 547.383 1221.05 (807.029) (347.955) (463.523) (785.442) FallPrice 156.180 -102.167 50.564 435.622 (457.554) (366.818) (295.828) (437.021) FallPrice2 55.747 -2305.602*** -873.065** 335.189 (281.773) (833.852) (411.460) (259.505) FallPrice3 5196.503** -1630.729* 2572.363* 5475.945** (2401.084) (907.679) (1557.450) (2361.955) BxRisePrice 49.385 (175.105) BxRisePrice2 -204.765 (423.294) BxRisePrice3 -1453.528* (818.889) BxFallPrice -785.697 (521.432) BxFallPrice2 -2888.698*** (808.356) BxFallPrice3 -7354.581*** (2487.675) Constant 26.875 -500.475*** -252.568** -252.568** (156.272) (143.714) (109.555) (109.613) NObservations 2946 2762 5708 5708 NGroups 52 49 101 101 R2 0.010 0.011 0.004 0.013 (1) (2) (3) (4)
4.3 Reverse Disposition Effect in LSE and SZSE
In this section, statistical evidence on the Reverse Disposition Effect in December-months is analysed on the basis of stock price increases and decreases through the variables RisePrice and FallPrice. Summarized in Table 3 (page 26), it contains results on regression 4 as specified in the methodology section.
To better structure the analysis, each sample is referred to according to a certain number (also visible in the top row of Table 3): LSE & SZSE combined (1), LSE (2), SZSE (3), SZSE: A-Shares (4), and SZSE: B-Shares (5). Each sample only takes data on the December-months into account.
For statistical inference on the Reverse Disposition Effect, coefficients on RisePrice must be < 0, while coefficients on FallPrice must be > 0.
Some evidence of the Reverse Disposition Effect is found in a combined sample of LSE & SZSE stock (1). RisePrice2 and RisePrice3 show negative coefficients that are significant at the 1% level: respectively, -116.334 and -77.918. Associated T-scores amount to -3.94 and -2.82. Negative coefficients in RisePrice point towards less abnormal trade volume with stock price increases as is to be expected by theory. Similarly, positive significant coefficients are expected to be found in FallPrice variables in order for evidence on the Reverse Disposition Effect. While no coefficients of significant value are found in LSE & SZSE (1), positive coefficients on FallPrice2 are found in samples on the LSE (2), SZSE (3) and SZSE: A-Shares (4). Significant at the 10% level, coefficients are respectively: 10.141, 112.021 and 162.573. Associated T-scores amount to 1.86, 1.75, and
1.82. Inference can be made that FallPrice2 is indeed positive under data from December-months.
Although throughout the samples tested there is some evidence supporting the
Reverse Disposition Effect (and thus tax-motivated trading as proposed in Hypothesis 3), there
cannot be spoken of a definitive confirmation. Only a limited couple of coefficients follow the pattern of being negative under RisePrice and positive under FallPrice. Additionally, the coefficients under FallPrice2 that are significant are only significant at the 10% level.
As is visible in the regression results in Table 3, proportional variance of the variables does decrease when more observations are used. This is a possible explanation why LSE (2),
(LSE & SZSE) (LSE) (SZSE) (A-Shares) (B-Shares)
Residual ResidualLSE ResidualSZSE ResidualSZSE ResidualSZSE
RisePrice 23.626 -2.792 49.727 203.689 -25.142 (18.483) (1.791) (53.136) (139.315) (41.777) RisePrice2 -116.334*** 2.375 -45.539 -19.105 -42.628 (29.557) (6.054) (40.161) (68.960) (42.379) RisePrice3 -77.918*** -8.071 -3.635 9.572 -1.820 (27.672) (7.944) (37.659) (77.476) (27.764) FallPrice -28.038 7.018 21.062 59.048 7.065 (30.300) (5.865) (57.014) (104.630) (32.812) FallPrice2 15.229 10.141* 112.021* 162.573* 80.689 (48.676) (5.458) (64.148) (89.377) (57.065) FallPrice3 10.745 -5.396 69.509 110.467 72.687 (45.708) (5.383) (57.075) (74.613) (91.559) Constant 8.645 -0.774 -73.02* -220.941** 64.338*** (22.018) (1.045) (43.891) (86.445) (22.448) NObservations 984 498 486 255 231 NGroups 201 100 101 52 49 R2 0,038 0.024 0,003 0,014 0,020 (1) (2) (3) (4) (5) TABLE 3: REVERSE DISPOSITION EFFECT IN LSE AND SZSE ATV-LSE&SZSE ATV-LSE ATV-SZSE ATV-SZSE ATV-SZSE
The amount of data points used to compute the results in Table 3 are significantly lower in comparison to earlier regressions. This is the case because only the December-months are taken into account to find out about Tax-motivated trading activity.
5. Conclusion and Discussion
In this thesis, the Disposition Effect (tendency of investors to hold on to value losing stock while selling value winning stock) is investigated for cultural influences. In particular, it is attempted to solve whether stock trade behaviour is universally the same or whether an investor’s cultural background has influence. This is possible through comparison of the London Stock Exchange and the unique Chinese Shenzhen Stock Exchange which consists of A- and B-Shares. A-Shares are purchasable only by Chinese investors, while B-Shares are purchasable by both foreign and Chinese investors creating the perfect real-life laboratory experiment. This culminates in the following research question:
Through Kahneman and Tversky’s Prospect Theory, human stock trading behaviour leading to the Disposition Effect is linked to Risk-, Regret-, and Tax Aversion. Hofstede’s
cultural dimensions lead to expectations on the collectivistic culture of the Chinese people.
Furthermore, Cushion Hypothesis creates the expectation that collectivistic cultures do present risk neutrality. On basis of the literature reviewed, the following hypotheses are formed: 1) A Disposition Effect is to be found in the Shenzhen Stock Exchange that is smaller than the one in the London Stock Exchange reflecting the apparent risk neutrality of Chinese Investors. 2) A Disposition Effect is to be found in the Shenzhen Stock Exchange that is smaller for A-Shares than for B-Shares reflecting the apparent risk neutrality of Chinese Investors. 3) A Reverse Disposition is to be found in both the Shenzhen Stock Exchange and London Stock Exchange in December due to tax motivations of investors worldwide.
Does a cultural difference in terms of Risk Aversion exist between the people of China and the globalized financial markets: Is there a difference in the Disposition Effect between the Shenzhen Stock Exchange (SZSE) and London Stock Exchange (LSE)?
Hypothesis 1 and Hypothesis 2. Research on the Reverse Disposition Effect, Hypothesis 3, due to tax-motivated trading was also conducted, however no definitive conclusion can be drawn. Research with more data points on December-months will probably prove more fortunate.
To answer the research question: With Hofstede’s cultural dimensions and the Cushion Hypothesis in mind, in this research evidence is found that the Chinese people’s apparent collectivistic culture and risk neutrality also influences stock trading behaviour. This conclusion is drawn because in the case that investors do not have the tendency to hold on to losing stock due to risk neutrality, the Disposition Effect would not show up. Some suggestions for further research:
1. Compare trade volumes of countries that are collectivistic with those that are individualistic in nature. Only then a true conclusion can be drawn on whether the
Disposition Effect solely exists in investors with a culturally individualistic background.
2. Research whether share prices of an abnormally high magnitude change abnormal trade volumes in different ways then expected by the Dispostion Effect. 3. Additional research with more data points on December-month specific trade activity should prove more fruitful than the attempt made in this research.
6. Bibliography
Chen, G.; Kim, K.A.; Nofsinger, J.R.; Rui, O.M. (2007). Trading performance, disposition effect, overconfidence, representativeness bias, and experience of emerging market investors. Journal of Behavioral Decision Making, 20(4), 425-451.
Dyl, E.A. (1977). Capital Gains Taxation and Year-End Stock Market Behavior. The Journal of
Finance, 32(1), 165-175.
Ferris, S.P.; Haugen, R.A.; Makhija, A.K. (1988). Predicting Contemporary Volume with Historic Volume at Differential Price Levels: Evidence Supporting the Disposition Effect. The Journal of Finance, 43(3), 677-697.
Firth, C. (2015). The disposition effect in the absence of taxes. Economics Letters, 136,
55-58.
Hofstede, G. (1980). Culture’s Consequences: International Differences in Work-Related
Values (1st ed.). USA, Beverly Hills, CA: Sage Publication.
Hofstede, G. (2001). Culture’s Consequences: Comparing Values, Behaviors, Institutions and
Organizations across Nations (2nd ed.). USA, Beverly Hills, CA: Sage Publication.
Hsee, C.; Weber, E.U. (1998). Cultural Differences in Risk Perception, but
Cross- Cultural Similarities in Attitudes Towards Perceived Risk. Management Science, 44(9),
1205-1217.
Kahneman, D.; Tversky, A. (1979). Prospect Theory: An Analysis of Decision under Risk.
Econometrica, 47(2), 263-292.
Kanehiro, S.; Shoji, I. (2016). Disposition effect as a behavioral trading activity elicited by Investors' different risk preferences. International Review of Financial Analysis, 46,
104-112.
Lakonishok, J.; Smidt, S. (1986). Volume for Winners and Losers: Taxation and Other Motives
for Stock Trading. The Journal of Finance, 41(4), 951-974.
Li, Y.; Yang, L. (2013). Prospect theory, the disposition effect, and asset prices. Journal of
Financial Economics, 107(3), 715-739.
Odean, T. (1998). Are Investors Reluctant to Realize Their Losses? The Journal of Finance,
53(5), 1775-1798.
Rau, H.A. (2014). The disposition effect and loss aversion: Do gender differences matter? Economics Letters, 123(1), 33-36. Shefrin, H.; Statman, M. (1985). The Disposition to Sell Winners Too Early and Ride Losers Too Long: Theory and Evidence. The Journal of Finance, 40(3), 777-790. Shiller, R.J. (2003). From Efficient Markets Theory to Behavioral Finance. Journal of Economic Perspectives, 17(1), 83-104. Tversky, A.; Kahneman, D. (1992). Advances in Prospect Theory: Cumulative Representation of Uncertainty. Journal of Risk and Uncertainty, 5(4), 297-323.
7. Appendix
7.1 List of Variables
7.2 Creating the Summary Statistics
Summary statistics were created on the basis of: (i) Shares Outstanding, (ii) Trade Volume, (iii) Share Price (Closing). Multiple residual variables were also used throughout the regressions as dependent variables and are therefore mentioned. Furthermore, five different sample groups are considered in this research: (i) LSE & SZSE Combined, (ii) LSE, (iii) SZSE: both type of shares, (iv) SZSE: A-Shares, and (v) SZSE: B-Shares. LSE & SZSE Combined LSE SZSE: Both type of shares SZSE: A-Shares
SZSE: B-Shares Residuals used in separate sample groups based on both stock exchanges: 7.3 Disposition Effect in Separate and Combined Market Settings Stata output on the disposition effect for each separate sample group. LSE & SZSE Combined
LSE SZSE: Both type of shares
SZSE: A-Shares SZSE: B-Shares
7.4 Reverse Disposition Effect for Each Individual Exchange Stata output on the (reverse) disposition effect for each separate sample group. Upper graph shows Stata output based only on data from the December months, the lower one accounts for data from all months but December. LSE & SZSE Combined DECEMBER WITHOUT DECEMBER
LSE
SZSE: Both type of shares
SZSE: A-Shares
SZSE: B-Shares
7.5 Indication of Discrepancy Between LSE & SZSE
7.6 Discrepancy Between SZSE A- and B-Shares
