DISPOSITION EFFECT IN GROUP VERSUS INDIVIDUAL FINANCIAL ...

1

DISPOSITION EFFECT IN GROUP VERSUS INDIVIDUAL

FINANCIAL DECISION-MAKING

MERCY OMUYOMA1, ROBERT MUDIDA2, GULNUR MURADOGLU3

1Strathmore Institute of Mathematical Sciences, Strathmore University.

2Institute for Public Policy, Strathmore Business School, Strathmore University.

3School of Business and Management, Queen Mary, University of London.

ABSTRACT

The study compares disposition effect for individual investors and investment groups

at the Nairobi Securities Exchange (NSE) and determines the influence of

sociodemographic characteristics on disposition effect of individual investors and

investment groups. The alpha measure Weber and Camerer’s (1998) is applied to

measure disposition effect. A tobit model is used to determine the influence of socio-

demographic characteristics on disposition effect of individual equity investors and

investment groups at the NSE.

Keywords: Disposition effect, group financial decision-making, .

1 INTRODUCTION Disposition effect refers to the tendency to sell investments that have appreciated in

value too soon while holding on to investments that have gone down in value for too

long (Shefrin & Statman, 1985). This behaviour is irrational since the logical course

of action would be to sell losing stocks as soon as possible to cut further losses, while

holding on to winning stocks to earn further gains. Possible explanations of disposition

effect are provided by prospect theory (Kahneman & Tversky, 1979), mental

accounting (Thaler, 1985) and emotions (Shefrin & Statman, 1985). Disposition effect

is one of the most robust findings about trading behaviour of individual investors

(Barberis & Xiong, 2009).

Previous research has documented evidence of disposition effect in individual retail

investor trading activity using brokerage firm databases (Aduda et al., 2012; Bashall

et al., 2018; Feng & Seasholes, 2005; Odean, 1998), stock exchange data (Grinblatt &

Keloharju, 2001; Brown et al., 2006; Barber et al., 2007) and mutual fund investor

databases (Bailey et al., 2011; Firth, 2015). Disposition effect has also been observed

in expert investors trading behaviour for instance among mutual fund managers

(Ammann et al., 2012; Cici, 2012) and futures traders (Locke & Mann, 2005; Choe &

Eom, 2009). Breitmayer et al. (2019) studied disposition effect of investors from 83

countries using brokerage firm data and found that disposition effect was higher for

investors from the Asia-Pacific region than for investors from Europe and Sub-Saharan

Africa. However, disposition effect has not been investigated for group investors. The

purpose of this paper is to contribute towards filling this research gap.

Investors from various parts of the world carry out joint investments through

investment groups, comprised of colleagues, friends or family that come together to

2

pool money for purposes of investing. Investment groups offer their members the

benefit of risk sharing, access to large scale profitable investments, social networks

and friendships, while at the same time learning about savings and investments.

Investment groups exist around the world such as in North America, Europe and some

countries in Africa. Investment groups in the USA and UK have been in existence

since the 1930’s and 1950’s respectively and invest in the stock market securities. In

Africa, investment groups exist in various countries such as in South Africa (stokvels),

Ethiopia (ekub), Kenya (chamas) and Uganda (Kedir & Ibrahim, 2011; Bisrat, Kostas,

& Feng, 2012; Ojijo, 2014). These investment groups are largely informal and are

usually formed to facilitate longterm investments in capital goods, property and

financial securities.

Group decisions are likely to differ from individual decisions. In the case of investment

groups, group decisions could potentially lead to better judgement, reduction in biases

and higher quality decisions as it facilitates aggregation of information, unique

perspectives and error checking which means a group has a larger pool of intellectual

resources to rely on when making decisions (Guzzo & Dickson, 1996; Kerr & Tindale,

2011). However, groups’ potential to make better decisions than individuals is

undermined by several factors such as shared information bias ((T. Chen & Sun, 2016;

Faulmüller et al., 2010; Stasser & Titus, 1985), groupthink (Janis, 1972), group

polarization (Moscovici & Zavalloni, 1969) and social influence (Wang et al., 2006).

Thus, due to these factors it is likely that there could be differences in disposition effect

exhibited by group investors compared to individual investors.

Investigating disposition effect under group decisions is important because research

evidence suggests that group decision making may have an adverse effect on investor

biases and investment performance. Studies done in the US based on investors’

transactional data and found that investment groups performed worse than the market

and individual investors (Barber & Odean, 2000), and also exhibited greater biases

than individuals (Barber et al., 2003).

However, disposition effect has not been investigated for group investors. One study

by Cici (2012) examined disposition effect in team versus individually managed

mutual funds and found that disposition effect was aggravated for team managed

mutual funds. However, mutual fund managers are a group of highly experienced and

sophisticated investors, hence the findings cannot be generalised to retail group

investors who are usually less experienced and sophisticated than mutual fund

managers. This is the first study that examines disposition effect in group compared to

individual retail investors.

This study seeks to compare disposition effect for group and individual investors at

the Nairobi Securities Exchange. There are an estimated 300,000 investment groups

in Kenya managing a total of Ksh 300 billion (USD $3billion) in assets (KAIG

Handbook 2016). A survey done by FSD Kenya revealed that investment groups

savings are the third most popular solution used to invest in future goals in Kenya

(CBK et al., 2019). Thus, Kenya provides a laboratory for testing disposition effect for

groups compared to individual investors. Therefore, this study’s aim is to ascertain

whether there is a difference in disposition effect exhibited by group investors

compared to individual equity investors at the NSE and to examine the influence of

socio-demographic characteristics on disposition effect among investors at the NSE.

3

2 LITERATURE REVIEW

This section reviews the theoretical and empirical literature relevant to this study. It

begins by discussing the theoretical framework underpinning this study, followed by

a review of the empirical literature on relevant constructs. This section concludes with

a summary of empirical studies and the research gaps are identified.

2.1 Disposition Effect Theory

Shefrin and Statman (1985) identified that individuals have the tendency to hold on to

losers for too long and sell winners too early a behaviour they referred to as disposition

effect. One explanation for disposition effect is provided by prospect theory

(Kahneman and Tversky, 1979). Under prospect theory, an individual prone to

disposition effect will be inclined to sell a winning stock, whose price has appreciated

since purchase, if they are risk averse over gains and will be reluctant to sell a losing

stock, that is trading at a loss since purchase, if they are loss averse and risk seeking

over losses. Prospect theory proposes that people first edit gambles by determining

whether they are gains or losses based on deviations from a fixed reference point.

People then evaluate these gains and losses using an s-shaped value function that is

concave in the gains domain, convex in the losses domain and is steeper for losses than

for gains. Thus, the prospect theory function captures the experimental findings that

people are loss averse as opposed to being just risk averse hence they exhibit risk

aversion over gains and risk seeking over losses.

Other theories that explain disposition effect include mental accounting (Thaler, 1985)

regret aversion and lack of self-control (Shefrin & Statman, 1985). Mental accounting

proposes that decision makers exhibit disposition effect because they segregate gains

and losses from different stocks into separate mental accounts that causes them to have

difficulty in realising losses from stocks. Regret aversion predicts that investors will

be hesitant to sell losing stocks to avoid experiencing regret while the quest for pride

may incline investors to sell winning stocks too soon. Investors’ reluctance to realize

losses may also be seen as a self-control problem where investors may hold on to

losing investments contrary to logic due to a lack of self-control.

A limitation of Shefrin and Statman's (1985) disposition effect is that the theory did

not provide a precise definition of ‘selling too soon’ and ‘holding too long.’ Dacey and

Zielonka (2008) attempted to resolve the time imprecision concern in by providing

precise time-independent concepts that replace ‘sell too soon’ and ‘hold too long.’

Dacey and Zielonka (2008) compare investor behavior under prospect theory and EUH

and conclude that the Kahneman-Tversky investor sells when he should hold and holds

when he should sell in comparison to the von Neumann-Morgenstern investor. Thus,

‘selling too soon’ and ‘holding for too long’ can be defined as when an investor sells

stocks that they should hold and holds stocks that they should sell under EUH.

2.2 Group Decision Making Theory

Group decision making differs from individual decision making in that it entails

features such as information aggregation, groupthink, group polarization, social

influence and a variety of other factors relevant to group decision making.

The information aggregation feature in group decision making has the potential to

improve on individual decision making, as long as the relevant information is unevenly

distributed among group members. The potential information capacity of a group is

4

roughly equivalent to the product of the group size and the potential capacity of any

individual reduced by an overlap as a result of shared information (Kerr & Tindale,

2011). Thus, groups consisting of members with varied expertise and knowledge have

the potential to make better quality decisions with less biases than the mean of

individual group members. Additionally, group discussions allow members to

countercheck each other’s arguments and point out errors and inaccuracies (Laughlin,

1999; Laughlin & Hollingshead, 1995).However, groups’ potential to make better

decisions than individuals due to information aggregation is undermined by shared

information bias that limits the effectiveness of groups in fully utilizing the pooled

information to make decisions. Shared information bias refers to the tendency for

groups to discuss only commonly shared information by all members while ignoring

equally important information that is held by only a few of the members (T. Chen &

Sun, 2016; Faulmüller et al., 2010; Stasser & Titus, 1985).

Groupthink (Janis, 1972), another feature of group decisions, is the concept where the

desire of individuals to keep the group together overrides the need to assess all

alternative plans of action, leading to bad group decision-making. Janis (1972) noted

that groups suffering from groupthink were more likely to justify irrational decisions

by failing to evaluate all alternatives. For example, in cohesive groups where the leader

is more assertive and outspoken, the resulting group decisions may tend to be heavily

influenced by the leader.

Group decisions also exhibit group polarization that is said to occur when individuals

in a group setting engage in more extreme decisions than their original private

decisions (Moscovici & Zavalloni, 1969). The theory assumes that group members

polarize their opinions after a group discussion because they exchange arguments for

their preferences and end up with even stronger arguments in favor of their preferences

after the group discussion. For example, if a group of investors were deciding on

whether or not to sell particular company’s shares whose value has fallen since

purchase. One argument in favour of selling would be that the share value is expected

to continue declining in the near future. An argument against the sale could be that

there is a chance that the share could appreciate in value due to new management. If

the group begins the discussion leaning against the sale, this will increase the chances

that the second argument will be discussed more than the first.

Group decision making is also characterized by social influence. Social comparisons

theory (SCT) advanced by Sanders & Baron's (1977) proposed that individuals change

their initial decisions after group discussions because of the need to seek social

approval and trying to avoid social disapproval. Group discussions reveal which views

are socially acceptable, so group members change their choices in the direction of the

group in order to gain approval with the group. Thus, if majority of the individual

group members are inclined to sell winners too soon while holding losers for too long,

individual members who initially did not exhibit this behavior will be socially

influenced to behave in the same way.

Finally, other factors that may have an effect on the effectiveness of the group decision

include the nature of the task (intellective versus judgmental), group size, the nature

of interaction between group members (face-to-face versus anonymous), the group

decision rule (unanimity versus majority rule) and the group’s ability to encourage

constructive conflict (Kerr & Tindale, 2011). Studies in social psychology report that

groups outperform individuals in intellective tasks that have a known correct solution,

5

than in judgmental tasks (Laughlin et al., 2002; Schultze et al., 2012). In addition,

majority rule tends to polarize group decisions and result in more extreme decisions

relative to the mean pre-discussion group member preferences especially in larger

groups.

The group decision making theories presented thus far suggest that disposition effect

may worsen under group decision making due to several reasons. First, since

investment decisions are more of judgmental tasks, and groups tend to perform worse

than individuals in such tasks compared to intellectual tasks, groups should display

greater disposition effect than individuals (Laughlin et al., 2002; Schultze et al., 2012).

Second, feelings of regret and pride may be enhanced after group discussion due to

group polarization which may aggravate disposition effect (Moscovici & Zavalloni,

1969). Disposition effect could also be heightened in group decisions due to

groupthink that leads to group members unwillingness to critically discuss losing

investments (Janis,1972). Based on these propositions, I hypothesize that dispostion

effect will be greater for group investors than individual investors. Moreover different

sets of investor characteristics have an influence on dispostion effect in the context of

individual versus group investors.

2.3 Empirical Literature

2.3.1 Disposition Effect in Investor Decisions

Disposition effect has been investigated for a variety of retail and professional

investors. Odean (1998) was among the earliest researchers to test for disposition

effect among investors with accounts at a large discount brokerage firm in the USA.

Odean computed the percentage of gains realized (PGR) and percentage of losses

realized (PLR) ratios for each account such that a positive difference between PGR

and PLR was interpreted as evidence of disposition effect. Bailey, Kumar and Ng

(2011) conducted a similar study of retail investors’ disposition effect using US mutual

fund investors’ shareholding data. Retail investors in the two studies above were found

to exhibit disposition effect which had an adverse effect on their investment

performance, even after accounting for alternative motivations such as tax incentives,

mean reversion expectation, portfolio rebalancing and trading costs.

Similarly, several studies have investigated how disposition effect differs across

investor categories in other varied contexts using investor transactional data from

brokerage firms. Brown, Chappel, Da Silva Rosa and Walter, (2006) analysed the

influence of the disposition effect across different categories of IPO and index stocks

investors at the Australian Stock Exchange. The study compared disposition effect

estimated through Odean’s (1998) PGR/PLR methodology for the different categories

of investors, that is, nominee companies, insurance companies, superannuation fund

companies, government, other companies, individuals and foreign investors. Brown et

al. (2006) reported that nominee and insurance companies exhibited the smallest

disposition effect suggesting that disposition effect may be lower for more

sophisticated investors.

Barber, Lee, Liu, and Odean (2007) use investors’ transactional data from the Taiwan

Stock Exchange to investigate the influence of disposition effect across different

categories of investors. The study found that individuals, companies and dealers

exhibited disposition effect with consistent results whether disposition effect was

aggregated across investors and overtime or averaged across investors. In contrast, the

6

foreign investors and domestic mutual funds did not exhibit disposition effect when

disposition effect was aggregated across investors and over time. The study also found

no difference in disposition effect exhibited by men and women.

Chen et al. (2007) examined the extent to which Chinese individual and institutional

investors exhibited disposition effect and how disposition effect was influenced by

investor characteristics. The study quantified disposition effect as the PGR-PLR

difference and regressed disposition effect on account age, frequent trading dummy,

account value and location. Chen et al. (2007) found that Chinese individual investors

were more affected by disposition effect than institutional investors and that middle

aged investors and those from cosmopolitan cities exhibited greater disposition effect.

Studies on disposition effect among Kenyan individual investors is scanty. One study

by Aduda et al. (2012) conducted a survey of investors from a sample of ten stock

brokerage firms in Kenya where the respondents were asked which shares, they bought

and sold in the previous six months. The study then used monthly stock return data

and financial statements from the Nairobi Securities Exchange (NSE) to determine the

profitability of shares sold by the investors. The study found that shares sold by

investors were highly profitable which the study concluded was an indication of

disposition effect. Although Aduda et al. (2012) examined disposition effect among

Kenyan investors, the study surveyed a small sample of investors and inferred

disposition effect from the sale of profitable shares by the investors. Hence the study

did not explicitly measure disposition effect. The proposed study will use transactional

data for equity investors at the NSE to compare disposition effect in investment groups

and individual equity investors.

Bashall et al. (2018) compared disposition effect of professionally advised retail

investors and those without professional advice in South Africa, an emerging market.

The study used retail investors’ transactional data form one of the largest stock

brokerages in South Africa to determine disposition effect based on Odean's (1998)

PGR/PLR analysis. Bashall et al. (2018) then tested the significance of the difference

between PGR and PLR for the professionally advised investors and those without

professional advise using z-tests. The study reported that professionally advised

investors exhibited lower disposition effect than those without professional investment

advice. This suggests that professional advice has the potential to reduce disposition

effect.

Hincapié-Salazar & Agudelo (2020) investigated disposition effect in both stock and

bond markets and for different types of investors in Colombia, another emerging

market economy. The study used transactional data from the Colombian Stock

Exchange to measure disposition effect using the PGR-PLR methodology and then

carried out cross sectional regressions at investor level to determine whether

experienced and sophisticated investors were less prone to disposition effect, while

controlling for other investor specific variables. The findings were that individuals

exhibited a stronger disposition effect than institutions and that experienced and

sophisticated stock investors exhibited less disposition effect.

The findings from the studies reviewed so far suggest that disposition effect is lower

for sophisticated investors such as mutual funds, insurance companies, foreign

investors, and professionally advised retail investors than for the less sophisticated

investors such as individuals. These findings are consistent for investors in both

7

developed market economies like Australia and emerging market economies such as

Taiwan, China, South Africa and Colombia (B. M. Barber et al., 2007; Bashall et al.,

2018; Brown et al., 2006; G. Chen et al., 2007; Hincapié-Salazar & Agudelo, 2020).

2.3.2 Disposition effect in group versus individual investors

Evidence on disposition effect in group compared to individual investor decisions is

limited. Cici (2012) studied disposition effect among team managed and individually

managed mutual funds in the USA using Odean (1998) PGR/PLR analysis. The study

ran a logit regression of disposition effect (a dummy variable equal to 1 if the mutual

fund’s PGR was more than the PLR and 0 otherwise) on a team dummy variable equal

to 1 if the mutual fund was team managed and 0 otherwise together with a variety of

other explanatory variables. Cici (2012) found that team managed funds exhibited

higher disposition effect than individually managed funds. However, Cici (2012) made

use of mutual fund managers’ quarterly holdings data to determine disposition effect.

Mutual fund managers are more experienced and sophisticated investors than retail

investors and thus their behavior is likely to differ from that of retail investors.

Barber & Odean (2000) and Barber et al. (2003) conducted related studies that

investigated the investment performance and biases of investment clubs in the US.

Barber & Odean (2000) used data from a discount brokerage firm in the US to compare

the mean return of investment clubs to market index returns and reported that clubs

underperformed the market. Barber et al. (2003) compared good-reason-based stock

purchase decisions of investment clubs and individual investors in the US using

brokerage firm data. They used “most admired companies” ranking, sales growth and

three-year return as proxies for good reasons and ran pooled time series regressions of

stock purchases of clubs and individuals on proxies for good reasons to buy. Barber et

al. (2003) found that investment clubs had a greater preference for admired firms with

more dramatic sales growth and stock returns than individual investors. This suggests

that groups were more biased and made worse stock purchase decisions than

individuals. However, none of these studies investigate disposition effect in group

investors. This study extends disposition effect literature by analyzing transactional

data to compare disposition effect for group and individual investors at the Nairobi

Securities Exchange.

2.3.3 Investor characteristics and disposition effect

A considerable amount of literature has been published on the influence of investor

sociodemographic and trading characteristics on disposition effect. Lehenkari and

Perttunen (2004) investigated the relationship between disposition effect and

investor’s age, gender and other characteristics, by analyzing the stock trading records

of all individual investors in the Finnish stock market. The study conducted GLS

regressions to determine how investors’ age, gender and other factors affect the selling

propensity as an indicator of disposition effect. The study found that selling propensity

was related to age, consistently reducing with higher age categories. However,

Lehenkari and Perttunen (2004) found that gender had no effect on the disposition

effect exhibited by the investors.

Dhar & Zhu (2006) used US brokerage firm data to analyse how investor

characteristics influence disposition effect for individual investors. The study

conducted an OLS regression of disposition effect (estimated using Odean’s (1998)

8

PGR/PLR methodology) on income categories, professional categories, number of

trades, investor’s age, return of realised gains, return of realised losses, mean number

of stocks held and the inverse of the number of stocks held. The study found that

wealthier, professional and high frequency traders exhibited lower levels of

disposition effect.

Another study by Bouteska & Regaieg (2018) sought to establish whether Tunisian

individual investors exhibit disposition effect and how investor individual

characteristics and market trends (bearish or bullish) influence investor disposition

effect. The study quantified investors’ disposition effect based on Odean's (1998)

methodology and then conducted OLS regressions to test for the effect of individual

investor characteristics and market trends on disposition effect using OLS regression

with White's heteroskedasticity correction of standard errors. The findings of the study

were that combinations of market trends and age and market trends and gender were a

key determinant of the strength of disposition effect. In a bear market, older investors

and male investors were less likely to exhibit disposition effect.

Breitmayer et al. (2019) examined the influence of culture and other investor

characteristics on disposition effect using data for retail investors from 83 countries

sourced from a UK stock brokerage firm. The study used Odean’s (1998) PGR/PLR

analysis to estimate disposition effect of investors and then conducted OLS regression

of disposition effect on culture (six culture dimensions from Hofstede’s dimensions),

demographics (gender and age dummy variables), economic (GDP per capita) and

region (dummy for region of investor). They report that men exhibited lower

disposition effect than women and that disposition effect seemed to increase with age.

The study also found that investors from the Asia-Pacific region exhibited higher

levels of disposition effect than those from Europe and Sub-Saharan Africa.

Gender differences in disposition effect have been investigated through experimental

studies with inconclusive findings. One such study by Jr et al. (2008) investigated the

role that gender plays in disposition effect through an experiment in Brazil using a

different measure of disposition effect by Weber and Camerer (1998). Jr et al. (2008)

found no difference in disposition effect across genders for sale decisions using the

purchase price as a reference point. However, when the previous period price is used

as a reference point the disposition effect vanishes for females. This implies that

women interpret reference points differently from men.

Similarly, Rau (2014) replicated the experimental methodology of Weber and Camerer

(1998) to study gender differences in disposition effect among German subjects. Contrary to findings by Jr et al. (2008) their results showed that women subjects

exhibited higher levels of disposition effect than men attributed to a higher loss

aversion in women than men. Braga and Fávero (2017) also conducted experiments to

test whether there were gender differences in disposition effect among Brazilian

subjects. In line with Jr et al. (2008) but in contrast to Rau (2014), Braga and Fávero

(2017) found no gender differences in disposition effect. The lack of conclusive results

in regard to gender differences in disposition effect motivates inclusion of gender as

an explanatory variable for disposition effect.

Overall, these studies highlight various investor characteristics that influence

disposition effect for individual investors. Disposition effect seems to diminish with

investor’s age (Bouteska & Regaieg, 2018; Lehenkari & Perttunen, 2004) and with

9

greater wealth, professionalism and frequency of trade (Dhar & Zhu, 2006). There

seems to be inconclusive evidence on the influence of gender on disposition effect

where some studies find no difference in disposition effect in men versus women

investors (Barber et al., 2007) while another study found that men exhibited lower

disposition effect than women (Breitmayer et al., 2019). Similarly there is mixed

evidence on the influence of gender on disposition effect from experimental studies

where Jr et al. (2008) and Braga and Fávero (2017) find no gender differences in

disposition effect while Rau (2014) reports that women subjects exhibit higher levels

of disposition effect than men. Although these studies provide evidence of the

prevalence of disposition among investors across different markets, that is, the USA,

Australia, China, Taiwan, Colombia, UK, Brazil, Tunisia, South Africa and Kenya,

they do not examine disposition effect in the context of group investors who make

joint investments.

2.4 Summary of research gaps

Evidence on disposition effect in group decisions is limited. Most of the previous

studies examined disposition effect in individual investor trades (Aduda et al., 2012;

Bailey et al., 2011; Bashall et al., 2018; G. Chen et al., 2007; Odean, 1998). There is

a scarcity of studies that examine disposition effect for group investors. Studies

examining the performance and biases of investment clubs in the US found that

investment clubs performed worse than the market (Barber & Odean, 2000) and were

more biased than individual investors (Barber et al., 2003). Although the two studies

do not specifically examine disposition effect of the investment clubs, their findings

suggest that group investors could be exhibiting greater biases than individual

investors. One study by Cici (2012) showed that disposition effect was higher for team

managed funds versus individually managed mutual funds in the US. However, mutual

fund managers are highly sophisticated investors and therefore these results may not

be generalised for retail investors.

This study digs into group decision making and disposition effect literature to draw

three reasons why disposition effect may differ under group decision making. First,

due to shared information bias that is present in group decision making, groups may

fail to critically examine all points of view and thus may still make the same mistakes

as individuals. This coupled with the fact that investment decisions are more of

judgmental tasks, and that groups tend to perform worse than individuals in such tasks

compared to intellectual tasks, implies that groups could potentially display greater

disposition effect than individuals. Second, group polarization causes enhanced

feelings of regret and pride after group discussion which may aggravate disposition

effect. Finally, the groupthink, phenomenon present in group decision making could

lead to heightened disposition effect in groups due to group members’ unwillingness

to critically discuss losing investments in an effort to conform. Based on these reasons,

it is possible that investors’ disposition effect may worsen under group decision

making.

In addition, this study compares the influence of investor characteristics on disposition

effect for individuals and group investors. Previous studies argue that disposition

effect is lower for more sophisticated investors (Barber et al., 2007; Bashall et al.,

2018; Brown et al., 2006; G. Chen et al., 2007; Hincapié-Salazar & Agudelo, 2020)

and diminishes with investor’s age (Bouteska & Regaieg, 2018; Lehenkari &

Perttunen, 2004), with greater wealth, professionalism and frequency of trade (Dhar

10

& Zhu, 2006). However, all these studies only examine these investor characteristics

at the individual level. This is the first study to compare the influence of investor

characteristics on disposition effect for individuals and for group investors. Apart from

examining the influence of investor characteristics such as age, gender and location on

disposition effect, the study also examines additional variables for individuals that

could act as a proxy for investor sophistication and experience such as the number of

stocks traded, number of trades and account value. These additional variables have

been previously used in studies by Chen et al. (2007) and Hincapié-Salazar & Agudelo

(2020). This study also examines the influence of group size, gender composition of

groups and location composition as variables that could influence disposition effect

for group investors.

3 METHODOLOGY

This section discusses the research design, population of the study and data used in the

study. In addition, this section provides a detailed description of the variables of the

study and the data analysis techniques.

3.1 Research design

To establish whether there is a difference in disposition effect of group investors and

individuals, I performed a two-sample t-test to test the statistical significance of the

difference in the mean disposition effect of groups and individuals. I then applied logit

regression analysis to examine the relationship between disposition effect and investor

characteristics where these variables were measured from actual investor transaction

records. The dependent variable (PGR – PLR) is a binary variable that takes that value

of 1 if positive and 0 otherwise. Furthermore, there is bunching of PGR-PLR around

+1, 0 and -1. Thus, logit regression analysis was preferred because it does not require

a linear relationship between disposition effect (PGR-PLR) and the explanatory

variables. A similar approach to examining the influence of investor characteristics on

disposition effect was used by Cici (2012).

3.2 Population of the study

The population of study is equity investors at the Nairobi Securities Exchange (NSE)

that consist of Kenyan and foreign individuals and corporates. The equity investor base

at the NSE stood at 1,936,529 as of 30 September 2020 (CMA, 2020). There are

approximately 65 listed companies under eight industry sectors at the NSE with a total

market capitalisation of USD 23 billion. The equity shareholder base stood at

1,936,529 as of 30 September 2020 (CMA, 2020). Kenyan individual investors form

95% of the equity investors in the NSE while the rest of equity investors comprise of

Kenyan corporates (3.4%), foreign individuals (0.7%), foreign corporates (0.07%),

East African individuals (0.4%), East African corporates (0.04%), junior investors and

brokers (0.1%).

The unit of analysis for this study is group investors and individual equity investors at

the NSE. Group investors constitute joint account investors where more than one

investor trades shares on the same account. The study compares disposition effect of

group investors and individual equity investors at the NSE.

11

3.3 Data description

Data is sourced from the Central Depository and Settlement Corporation (CDSC)

which is a limited liability company that is in charge of providing central clearing,

settlement and depository services for all securities listed on Kenya’s capital market,

the NSE. This is exclusive data that is not in the public domain and thus is valuable in

providing insights on the behaviour of equity investors. The data set comprises the

daily equity stock trades data for all investors at the Nairobi Securities Exchange

(NSE) for a period from 1 January 2016 to 31 December 2016. It contains information

on the security traded, date of trade, nature of trade (buy or sell), a unique transaction

reference number, investor type (local, East African or foreign companies and

individuals), gender of investor (male, female or neutral for companies), date of birth

or incorporation, the town where the investor is located, the stock traded, quantity of

stocks traded and price at which the stock was traded.

The complete data set consists of 41,630 investor accounts but not all accounts are

useful for this study. I deleted all 3 junior accounts and 2 broker accounts since are run

by representatives. I also deleted 1,258 corporate investor accounts because they are

not part of the study to leave a final sample of 40,367 investor accounts. CDSC allows

investors to open joint accounts. I classified accounts based on the date of birth or

incorporation. Investor accounts transacting with singular dates of birth or

incorporation and with no missing dates of birth were considered individual accounts

(22,367) that include both individuals and corporates. Accounts with missing dates of

birth or incorporation (2,078) or those with some transactions having a date of

birth/incorporation and others missing (6,632) were classified as unknown (for further

processing. The accounts classified as unknown, were then categorised based on

gender. Those accounts with only one gender transacting were classified as individual

investors (7,323), while those with multiple genders transacting through the same

account (1,387) were classified as group investors. I then dropped individual investors

that were corporates (1,883) based on the CDSC investor type classification. This

process is as represented in figure 1 in the appendix.

Data on the stock’s daily high and low prices of stocks for a period from 1 January

2016 to 31 December 2016 was also obtained from Thompson Reuters Financial

DataStream. These daily stock prices were used to determine the appropriate reference

points for calculating disposition effect.

The cleaned data set comprised of stock transaction data for a 1-year period for 26,549

individual investors and 11,935 group investors. For purposes of determining the

disposition effect, I only considered accounts with two or more stocks traded and with

at least one sale over the 1-year period. Stocks included must have a known purchase

date and price. The final dataset consisted of 2,515 individual accounts and 3,711

group accounts.

3.4 Data analysis

3.4.1 Measurement of disposition effect

In most previous studies, Odean’s (1998) PGR/PLR analysis is used to measure

disposition effect (Barber et al., 2007; Bailey et al., 2011; Cici, 2012; Firth, 2015).

Under PGR/PLR analysis, disposition effect is determined by comparing an investor’s

proportion of gains realised (PGR) to their proportion of losses realised (PLR). If the

12

PGR exceeds the PLR, then the investor exhibits disposition effect. The gains and

losses on stocks are computed based on a reference price such as the purchase price of

the stocks. Several studies found that disposition effect is robust to different reference

points such as the first purchase price, highest purchase price, mean purchase price

and most recent purchase price (Feng & Seasholes, 2005; Odean, 1998). Thus,

following Barber et al. (2007), I use the weighted mean purchase price as a reference

point for estimating gains and losses. This approach to measuring disposition effect is

preferred as it assumes that the purchase price is the most logical reference point for

investors when determining whether a stock has gained or lost value over time. The

main disadvantage of using Odean’s (1998) PGR/PLR analysis is that it requires

information on the purchase date and price of securities held and sold by the investor.

The purchase price and data may not be available for stocks bought before the start of

the sampling period.

Following Odean’s (1998) methodology, in order to determine whether an investor

exhibits disposition effect, each investor’s trading record is arranged in chronological

order and a portfolio of individual stocks is constructed for stocks whose purchase date

and price is known. On each day that a sale takes place for investors with portfolios of

two or more stocks, the selling price of the stock is compared to its weighted mean

purchase price to establish whether the stock was sold at a gain or loss. This is referred

to as a realised gain or realised loss. For every other stock that was in the investor’s

portfolio at the beginning of a day but was not sold, we compare the stock’s high and

low price for that day to its weighted mean purchase price. If the both the stocks high

and low prices are above or below its weighted mean purchase price, we recognise a

paper gain or a paper loss respectively. Where the mean purchase price falls in between

the stock’s high and low price, neither a loss nor a gain is recognised.

For example, suppose than an investor has five stocks in his portfolio: A, B, C, D and

E. A and B are worth more than their purchase price, while C, D and E are worth lower

price than their purchase price. Suppose for a three-day period the investor sells stock

B on day 2 and stock E on day 3. On day 1, stock A, B and C are realised gains, stock

D and E are paper losses. On day 2 stock A is a paper gain and stocks C, D and E are

paper losses and stock B is a realised gain. On day 3, stock A is a paper gain, stocks

C, D and E are paper losses while stock E is a realised loss. The value of realised gains,

paper gains, realised losses and paper losses are then summed for each individual

account and then summed across all accounts.

The proportion of gains realised (PGR) and the proportion of losses realised are then

calculated as follows.

𝑃𝐺𝑅 = 𝑟𝑒𝑎𝑙𝑖𝑠𝑒𝑑 𝑔𝑎𝑖𝑛𝑠

𝑟𝑒𝑎𝑙𝑖𝑠𝑒𝑑 𝑔𝑎𝑖𝑛𝑠+𝑝𝑎𝑝𝑒𝑟 𝑔𝑎𝑖𝑛𝑠 (1)

𝑃𝐿𝑅 = 𝑟𝑒𝑎𝑙𝑖𝑠𝑒𝑑 𝑙𝑜𝑠𝑠𝑒𝑠

𝑟𝑒𝑎𝑙𝑖𝑠𝑒𝑑 𝑙𝑜𝑠𝑠𝑒𝑠+𝑝𝑎𝑝𝑒𝑟 𝑙𝑜𝑠𝑠𝑒𝑠 (2)

In order to compute these ratios, the denominators of both PGR and for PLR must be

nonzero and at least one stock must be sold.

The values for disposition effect lie between -1 and +1, where +1 indicates that an

investor sells all stocks with gains and holds all stocks with losses while a value of -1

indicates an investor sells all stocks with losses and holds all stocks with gains. If

13

disposition effect is 0 this means that the investor’s proportion of gains realised is

equal to the proportion of losses realised. Disposition effect is present if there is a

significant positive difference between PGR and PLR of the investor (values above 0),

which would indicate that an investor is more inclined to sell stocks with a gain than

a loss. The higher the positive value the greater the disposition effect.

To determine whether disposition effect is statistically significant, I use two

approaches. In the first approach, I sum all the realised gains, realised losses, paper

gains and paper losses across all accounts and over time to arrive at the total realised

gains, realised losses, paper gains and paper losses for each investor category

(individuals and group investors). PGR and PLR are then calculated for each investor

category as a whole. The difference between PGR and PLR (that is, the disposition

effect) is then calculated for each investor category. In the second approach, for each

investor, I calculate PGR and PLR and determine the difference between PGR and

PLR. Then, I calculate the average difference across investors within each investor

category (individual investors and group investors). To test the statistical significance

of the difference between PGR and PLR, I perform one-tailed t tests with the null

hypothesis that PGR-PLR is not significantly greater than zero.

To ascertain whether there is a statistically significant difference in disposition effect

exhibited by group investors compared to individual equity investors at the Nairobi

Securities Exchange, I test the statistical significance of the difference in means by

performing a two-sample t-test.

3.4.2 Investor Characteristics and Disposition Effect

To determine the influence of investor characteristics on disposition effect, I use cross

sectional logit regressions at investor level. The dependent variable (disposition effect)

the difference between PGR and PLR coded as a binary variable that takes the value

of 1 if positive and 0 otherwise

Explanatory variables include several investor characteristics. Previous studies argue

that sophisticated and more experienced investors are likely to be more skilled in

trading and understand the market better than the mean investor and thus may exhibit

less disposition effect (Bouteska & Regaieg, 2018; Dhar & Zhu, 2006; Hincapié-

Salazar & Agudelo, 2020). Therefore, this study includes an explanatory variable for

investor sophistication measured as the number of distinct stocks traded by the investor

over the sample period (see Cici, 2012; Hincapié-Salazar & Agudelo, 2020; Seru et

al., 2010). Another explanatory variable for investing experience is the total number

of trades an investor has placed for the sample period. Investors may gain experience

from actively trading securities and learning from the outcome of each trade

(Hincapié-Salazar & Agudelo, 2020; Seru et al., 2010).

To determine the influence of investor characteristics on disposition effect the study

estimates cross sectional logit regressions separately for individual and group

investors.

For individual investors the study estimates the following cross sectional logit

regression model.

14

(𝑃𝐺𝑅 − 𝑃𝐿𝑅)𝑖=𝛼 + 𝛽1(𝐷𝑖𝑠𝑡𝑖𝑛𝑐𝑡 𝑆𝑡𝑜𝑐𝑘𝑠 𝑇𝑟𝑎𝑑𝑒𝑑)𝑖+ 𝛽2(𝑁𝑜. 𝑜𝑓 𝑇𝑟𝑎𝑑𝑒𝑠)𝑖+

+ 𝛽3(𝑉𝑎𝑙𝑢𝑒 𝑜𝑓 𝑆𝑡𝑜𝑐𝑘𝑠 𝑇𝑟𝑎𝑑𝑒𝑑)𝑖+𝛽4𝐴𝑔𝑒 𝐶𝑎𝑡𝑒𝑔𝑜𝑟𝑦𝑖 + 𝛽5𝐺𝑒𝑛𝑑𝑒𝑟𝑖 + 𝛽6(𝐿𝑜𝑐𝑎𝑡𝑖𝑜𝑛)𝑖

𝛽7(𝐼𝑛𝑣𝑒𝑠𝑡𝑜𝑟 𝑁𝑎𝑡𝑖𝑜𝑛𝑎𝑙𝑖𝑡𝑦)𝑖 + 𝜖𝑖

For group investors the study estimates the following cross sectional logit regression

model.

(𝑃𝐺𝑅 − 𝑃𝐿𝑅)𝑖= 𝛼 + 𝛽1(𝐷𝑖𝑠𝑡𝑖𝑛𝑐𝑡 𝑆𝑡𝑜𝑐𝑘𝑠 𝑇𝑟𝑎𝑑𝑒𝑑)𝑖+ 𝛽2(𝑁𝑜. 𝑜𝑓 𝑇𝑟𝑎𝑑𝑒𝑠)𝑖

+ 𝛽3(𝑉𝑎𝑙𝑢𝑒 𝑜𝑓 𝑆𝑡𝑜𝑐𝑘𝑠 𝑇𝑟𝑎𝑑𝑒𝑑)𝑖+𝛽4(𝐴𝑣𝑒𝑟𝑎𝑔𝑒 𝐴𝑔𝑒 𝐶𝑎𝑡𝑒𝑔𝑜𝑟𝑦 )𝑖

𝛽5(𝐺𝑒𝑛𝑑𝑒𝑟 𝐶𝑜𝑚𝑝𝑜𝑠𝑖𝑡𝑖𝑜𝑛)𝑖 +𝛽6(𝐿𝑜𝑐𝑎𝑡𝑖𝑜𝑛 𝐶𝑜𝑚𝑝𝑜𝑠𝑖𝑡𝑖𝑜𝑛)𝑖 + 𝜖𝑖

Table 1 provides a detailed description of the variables of the study.

4 RESULTS AND DISCUSSION This section presents the results of the study. It begins by outlining the descriptive

statistics of the data analysed, followed by the results for the comparison of disposition

effect for individuals and group investors. It concludes by presenting the results of the

regression analysis of disposition effect on investor characteristics for individuals and

group investors.

4.1 Descriptive Statistics

Table 2 presents descriptive statistics for 26,549 individual and 11,922 group investors

at the NSE that were tested for disposition effect for the period 1 January 2016 to 31

December 2016. Individuals constitute about 69% while groups form 31% of the

investors in the sample. Above 93% of both the individual and group investors are

Kenyans while the rest are from other East African countries and from the rest of the

world. About 80% of the individual and group investors are located in urban towns

with the rest of the investors are located in rural and foreign towns. Males form 67.5%

while females form 29.8% of the individual investors. Similarly, 51% of the group

investors are males, 34.6% are female while 14.4% of group investors do not specify

their gender. The mean age for individual investors (58.35) is slightly higher than that

for group investors (56.73).

The mean number of stocks traded by group investors was 5.23 which was about 2.5

times higher than that for individual investors (2.13) over the 1-year period. This is

consistent with the findings by Barber et al. (2003) who reported that the mean number

and value of stocks held by investment clubs were higher than that for individual

investors in the US. Overall, group investors traded more frequently (25.33) than

individual investors (6.61) and had larger volumes of stocks transacted per account

than individual investors. For both individual and group investors the mean number of

buys was about the same as the mean number of sells. Individual investors’ mean

volume of stocks sold (17,571) was higher than their volume of stocks bought per

account (14,728). In contrast, the mean volume of stocks sold (236,544) by group

investors was lower than the volume of stocks bought (287,291) per account. The mean

value of stocks transacted by group investors (KES 15,118,756) was about 23 times

higher than that for individual investors (KES662,360), t(11982)=-7, p<0.001. The

mean KES value of sells for individual investors was higher than the mean KES value

of buys. This is in line with the statistics reported by Barber & Odean (2000) where

individual investors in the US made more sales than purchases. However, the KES

value of sells for group investors was lower than the KES value of their buys.

15

4.2 PGR/PLR for Individual and Group Investors

Table 3 reports the descriptive statistics for the final sample of 2,515 individual

investors and 3,711 group investors. As a result of the selection criteria for the

determination of disposition effect, the individual and group investors in the final

sample generally have a greater average number of distinct stocks traded than that for

the entire sample. Individual and group investors in the final sample also trade more

frequently and have higher values of stocks traded than the entire sample.

To ascertain whether there is a difference in disposition effect exhibited by group

investors compared to individual equity investors at the NSE, I determine disposition

effect at the aggregate level and also based on the average disposition effect at account

level. Table 4 presents the proportion of gains realised and the proportion of losses

realised by investor category.

Out of the sample of 2,515 individual and 3,711 group investors tested for disposition

effect, 53% of individual investors and 62% of group investors realised gains at a faster

rate than losses (PGR>PLR). Based on the aggregated gains and losses across

investors, individual investors were about 1.2 times more likely to realise gains than

losses (PGR/PLR ratio = 1.17) while group investors were about 2.3 times more likely

to realise gains than losses (PGR/PLR ratio = 2.29). When PGR and PLR are averaged

across accounts and over time, individual investors were about 2.3 times more likely

to realise gains than losses (PGR/PLR ratio = 2.325). Based on the aggregated gains

and losses across investors, group investors were about 2.3 times more likely to realise

gains than losses (PGR/PLR ratio = 2.29). However, based on the average PGR/PLR

ratio per account, group investors were about 3.2 times more likely to realise gains

than losses (PGR/PLR ratio = 3.24). This result is consistent with results from other

studies from various parts of the world (Barber et al., 2007; Bashall et al., 2018; Brown

et al., 2006; Chen et al., 2007; Odean, 1998).

The noted difference between the aggregated PGR/PLR ratio and the average

PGR/PLR ratio per account could be due to the fact that in the case of averaged

PGR/PLR ratios, I could only calculate the ratio if the investor had both realised gains

and realised losses. Out of the 2,515 individual and 3,711 group investors in the

sample, it was not possible to calculate the PGR/PLR ratio for 35% of the individual

and 32% of the group investors since they either had no realised gains or no realised

losses.

At the aggregate level, individuals do exhibit disposition effect where the difference

in PGR and PLR (0.09%) is reliably positive (p value=0.000). However, the difference

between the PGR and PLR averaged across investors is negative (-0.13%) but is not

statistically significant (p value =0.6735). The finding of a significant difference

between the aggregate and the individual level measures of disposition effect is similar

to findings by Dhar & Zhu (2006). One possible reason for this difference is that the

aggregate measure of disposition effect does not reflect the idiosyncratic differences

in individual PGR and PLR. Table 5 presents an analysis of the disposition effect

across deciles based on the frequency of trade. Disposition effect is greater for most

active individual traders in decile 10 (PGR-PLR=0.0047) than for the overall sample

(PGR-PLR= –0.0013). In line with Dhar & Zhu (2006), I find that the aggregate

measure of disposition effect assigns more weight to frequent traders who tend to

exhibit a greater disposition effect in the sample used in this study, thus increasing the

16

magnitude of disposition effect at the aggregate level. In contrast, group investors do

exhibit disposition effect regardless of whether I aggregate gains and losses across

investors and over time or average across investors. The difference in group investors’

PGR and PLR based on aggregate values across investors and over time is 0.608%

while the difference averaged across investors is 0.782%. These results are both

reliably positive (p value<0.000).

To ascertain whether there is a statistically significant difference in disposition effect

exhibited by group investors compared to individual equity investors at the Nairobi

Securities Exchange, I test the statistical significance of the difference in means by

performing a two-sample t-test. The results as reported in table 4 indicate a significant

difference in PGR-PLR between individual and group investors t (4,354) = -2.664, p

value = 0.0078. Thus, group investors do exhibit a significantly greater disposition

effect than individual investors.

4.3 Logit regression analysis

To examine the influence of investor characteristics on disposition effect among

investors at the NSE, I perform logistic regression analysis of disposition effect on

several investor characteristic variables. Table 6 reports the results of the logistic

regression of disposition effect on individual investor characteristics. I estimate four

models where model 1 includes all variables. Overall, results indicate that the number

of distinct stocks traded by the investor, the number of trades placed by the investor

and the investors age have a significant influence on the disposition effect of the

individual investor.

I focus the discussion on model 4 which indicates that a one unit increase in the number

of distinct stocks traded increases the odds of the individual investor exhibiting

disposition effect by 3.6%, controlling for the number of trades, investor nationality,

gender, age and location of the investor. The odds of an investor exhibiting disposition

effect increase by 0.2% with one unit increase in number of trades placed by the

investor, controlling for investor nationality, gender, age and location of the investor

(statistically significant at p<0.01). This implies that the chances of an individual

investor exhibiting disposition effect increase with the frequency of trade and with

every additional unit of distinct stock traded by the investor in the sample period. This

finding is in contrast with findings from several previous studies where disposition

effect diminished with greater investor sophistication and experience (B. M. Barber et

al., 2007; Bashall et al., 2018; Brown et al., 2006; G. Chen et al., 2007; Hincapié-

Salazar & Agudelo, 2020). The odds of exhibiting disposition effect are lower for those

in the higher age categories compared to those in the lowest age category (below 30

years) controlling for the number of stocks traded, investor nationality, gender, age

and the location of the investor (all statistically significant at p<0.1). This finding is in

line with that by Lehenkari & Perttunen, (2004) and Bouteska & Regaieg (2018).

The investor nationality, gender and location of the investor do not have a significant

influence on the disposition effect exhibited by the individual investor. The finding of

no gender effect on disposition effect is in line with findings by Lehenkari & Perttunen

(2004) and Barber et al. (2007) but is in contrast to the findings by Breitmayer et al

(2019) who found that men had a lower disposition effect than women. The location

and nationality of the investor does not significantly influence disposition effect,

which is contrary to findings by Chen et al. (2007) and Breitmayer et al. (2019).

17

Table 7 reports the results of the logistic regression of disposition effect on several

group investor characteristics. The number of distinct stocks traded by the group, the

number of trades placed by the group and the gender composition of the groups have

a significant influence on the disposition effect of group investors. Model 2 shows that

a one unit increase in the number of distinct stocks traded increases the odds of the

group exhibiting disposition effect by 2.7%, controlling for the number of trades,

group size, gender composition, location composition and mean age category of the

group. The odds of a group exhibiting disposition effect increase by 0.1% with one

unit increase in number of trades placed by the group, controlling for number of stocks

traded, group size, gender composition, location composition and mean age category

of the group. This is consistent with the findings for the individual investors in this

study, implying that for both group and individual investors disposition effect

increases with the number of distinct stocks traded and trading frequency.

The odds of a group exhibiting disposition effect decrease by 13.4% with one unit

increase in the proportion of males trading the joint account, controlling for number of

distinct stocks traded, number of trades, group size, location composition and mean

age category of the group (statistically significant at p<0.01). This means that groups

that had a greater proportion of males trading stocks over the sample period were more

prone to disposition effect than those with lesser proportion of males which is contrary

to findings by Breitmayer et al. (2019). This finding is also in contrast with that for the

individual investors in this study where I found no gender effect on disposition effect.

The group size, location composition and mean age category of the group do not have

a significant influence on the disposition effect exhibited by the individual investor.

In summary, I find evidence that both group and individual investors at the NSE

exhibit disposition effect. However, group investors exhibit significantly greater

disposition effect than individual investors. The number of distinct stocks traded and

the frequency of trade, as proxies for investor sophistication and experience, are

important determinants of disposition effect for both group and individual investors.

Disposition effect seems to increase with group and individual investors sophistication

and experience.

18

REFERENCES

Aduda, J., Odera, E. R., & Onwonga, M. (2012). The Behaviour and Financial

Performance of Individual Investors in the Trading Shares of Companies

Listed At the Nairobi Stock Exchange, Kenya. Journal of Finance and

Investment Analysis, 1(3), 33–60.

Ammann, M., Ising, A., & Kessler, S. (2012). Disposition effect and mutual fund

performance. Applied Financial Economics, 22(1), 1–19.

https://doi.org/10.1080/09603107.2011.595676

Bailey, W., Kumar, A., & Ng, D. (2011). Behavioral biases of mutual fund investors.

Journal of Financial Economics, 102(1), 1–27.

Barber, B. M., Heath, C., & Odean, T. (2003). Good Reasons Sell: Reason-Based

Choice among Group and Individual Investors in the Stock Market.

Management Science, 49(12), 1636–1652.

Barber, B. M., Lee, Y.-T., Liu, Y.-J., & Odean, T. (2007). Is the Aggregate Investor

Reluctant to Realise Losses? Evidence from Taiwan. European Financial

Management, 13(3), 423–447. https://doi.org/10.1111/j.1468-

036X.2007.00367.x

Barber, B., & Odean, T. (2000). Too many cooks spoil the profits: The performance

of investment clubs. Financial Analyst Journal, 56(1), 17–25.

Barberis, N., & Xiong, W. (2009). What Drives the Disposition Effect? An Analysis

of a Long-Standing Preference-Based Explanation. The Journal of Finance,

64(2), 751–784.

Bashall, J., Willows, G. D., & West, D. (2018). The Extent to Which Professional

Advice Can Reduce the Disposition Effect: An Emerging Market Study.

Journal of Emerging Market Finance, 17(2), 229–249.

Bisrat, A., Kostas, K., & Feng, L. (2012). Are there financial benefits to join

RoSCAs? Empirical evidence from Equb in Ethiopia. Procedia Economics

and Finance, 1, 229–238.

Bouteska, A., & Regaieg, B. (2018). Investor characteristics and the effect of

disposition bias on the Tunisian stock market. Borsa Istanbul Review, 18(4),

282–299.

Braga, R., & Fávero, L. P. L. (2017). Disposition Effect and Tolerance to Losses in

Stock Investment Decisions: An Experimental Study. Journal of Behavioral

Finance, 18(3), 271–280. https://doi.org/10.1080/15427560.2017.1308946

Breitmayer, B., Hasso, T., & Pelster, M. (2019). Culture and the disposition effect.

Economics Letters, 184, 108653.

Brown, P., Chappel, N., Da Silva Rosa, R., & Walter, T. (2006). The Reach of the

Disposition Effect: Large Sample Evidence Across Investor Classes*.

International Review of Finance, 6(1–2), 43–78.

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. https://doi.org/10.1002/bdm.561

Chen, T., & Sun, X. (2016). Shared information bias in group decision-making:

Based on hidden profile paradigm. Advances in Psychological Science, 24(1),

132–142.

Choe, H., & Eom, Y. (2009). The disposition effect and investment performance in

the futures market. Journal of Futures Markets, 29(6), 496–522.

https://doi.org/10.1002/fut.20398

19

Cici, G. (2012). The Prevalence of the Disposition Effect in Mutual Funds’ Trades.

The Journal of Financial and Quantitative Analysis, 47(4), 795–820.

Dacey, R., & Zielonka, P. (2008). A detailed prospect theory explanation of the

disposition effect. The Journal of Behavioral Finance, 9(1), 43–50.

Dhar, R., & Zhu, N. (2006). Up close and personal: Investor sophistication and the

disposition effect. Management Science, 52(5), 726–740.

Faulmüller, N., Kerschreiter, R., Mojzisch, A., & Schulz-Hardt, S. (2010). Beyond

group-level explanations for the failure of groups to solve hidden profiles:

The individual preference effect revisited. Group Processes & Intergroup

Relations, 13(5), 653–671.

Feng, L., & Seasholes, M. S. (2005). Do investor sophistication and trading

experience eliminate behavioral biases in financial markets? Review of

Finance, 9(3), 305–351.

Firth, C. (2015). The disposition effect in the absence of taxes. Economics Letters,

136, 55–58. https://doi.org/10.1016/j.econlet.2015.08.025

Grinblatt, M., & Keloharju, M. (2001). What makes investors trade? The Journal of

Finance, 56(2), 589–616.

Guzzo, R. A., & Dickson, M. W. (1996). Teams in organizations: Recent research on

performance and effectiveness. Annual Review of Psychology, 47(1), 307–

338.

Hincapié-Salazar, J., & Agudelo, D. A. (2020). Is the disposition effect in bonds as

strong as in stocks? Evidence from an emerging market. Global Finance

Journal, 46, 100508.

Janis, I. L. (1972a). Victims of groupthink: A psychological study of foreign-policy

decisions and fiascoes. Houghton Mifflin.

Janis, I. L. (1972b). Victims of groupthink: A psychological study of foreign-policy

decisions and fiascoes.

Jr, N. D. C., Mineto, C., & Silva, S. D. (2008). Disposition effect and gender.

Applied Economics Letters, 15(6), 411–416.

https://doi.org/10.1080/13504850600706560

Kahneman, D., & Tversky, A. (1979). Prospect theory: An analysis of decision under

risk. Econometrica: Journal of the Econometric Society, 263–291.

Kedir, A. M., & Ibrahim, G. (2011). ROSCAs in Urban Ethiopia: Are the

characteristics of the institutions more important than those of members?

Journal of Development Studies, 47(7), 998–1016.

Kerr, N. L., & Tindale, R. S. (2011). Group-based forecasting?: A social

psychological analysis. International Journal of Forecasting, 27(1), 14–40.

Laughlin, P. R. (1999). Collective induction: Twelve postulates. Organizational

Behavior and Human Decision Processes, 80(1), 50–69.

Laughlin, P. R., Bonner, B. L., & Miner, A. G. (2002). Groups perform better than

the best individuals on letters-to-numbers problems. Organizational Behavior

and Human Decision Processes, 88(2), 605–620.

Laughlin, P. R., & Hollingshead, A. B. (1995). A theory of collective induction.

Organizational Behavior and Human Decision Processes, 61(1), 94–107.

Lehenkari, M., & Perttunen, J. (2004). Holding on to the Losers: Finnish Evidence.

Journal of Behavioral Finance, 5(2), 116–126.

https://doi.org/10.1207/s15427579jpfm0502_5

Locke, P. R., & Mann, S. C. (2005). Professional trader discipline and trade

disposition. Journal of Financial Economics, 76(2), 401–444.

20

Moscovici, S., & Zavalloni, M. (1969). The group as a polarizer of attitudes. Journal

of Personality and Social Psychology, 12(2), 125.

Odean, T. (1998). Are investors reluctant to realize their losses? The Journal of

Finance, 53(5), 1775–1798.

Prates, W. R., da Costa, N. C. A., & Dorow, A. (2017). Risk Aversion, the

Disposition Effect, and Group Decision Making: An Experimental Analysis.

Managerial and Decision Economics, 38(7), 1033–1045.

https://doi.org/10.1002/mde.2843

Rau, H. A. (2014). The disposition effect and loss aversion: Do gender differences

matter? Economics Letters, 123(1), 33–36.

Sanders, G. S., & Baron, R. S. (1977). Is social comparison irrelevant for producing

choice shifts? Journal of Experimental Social Psychology, 13(4), 303–314.

Schultze, T., Mojzisch, A., & Schulz-Hardt, S. (2012). Why groups perform better

than individuals at quantitative judgment tasks: Group-to-individual transfer

as an alternative to differential weighting. Organizational Behavior and

Human Decision Processes, 118(1), 24–36.

Seru, A., Shumway, T., & Stoffman, N. (2010). Learning by trading. The Review of

Financial Studies, 23(2), 705–739.

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.

Stasser, G., & Titus, W. (1985). Pooling of unshared information in group decision

making: Biased information sampling during discussion. Journal of

Personality and Social Psychology, 48(6), 1467.

Thaler, R. (1985). Mental accounting and consumer choice. Marketing Science, 4(3),

199–214.

Wang, L., Doucet, L., & Northcraft, G. (2006). Culture, Affect, and Social Influence

in Decision-Making Groups. In National Culture and Groups. Emerald

Group Publishing Limited.

Weber, M., & Camerer, C. F. (1998). The disposition effect in securities trading: An

experimental analysis. Journal of Economic Behavior & Organization, 33(2),

167–184.

Williams, H. M., & Meân, L. J. (2004). Measuring gender composition in work

groups: A comparison of existing methods. Organizational Research

Methods, 7(4), 456–474.

21

APPENDIX Table 1

Descriptive statistics for individual and group investors at the NSE

Variable Description

Disposition effect The difference between PGR and PLR coded as a dummy

variable that takes the value of 1 if positive and 0 otherwise.

Distinct Stocks traded Number of distinct stocks traded by the investor over the

sample period

Number of trades Natural logarithm of the total number of trades of the investor

over the sample period

Value of stocks traded Value of stocks traded by the investor over the sample period

Age category A categorical variable with five levels that indicate the

investors age category

Mean age category A categorical variable with five levels that indicate the groups

mean age category as determined by the mean age of the joint

account members.

Gender A dummy variable that takes the value of 1 if male and 0

otherwise.

Gender composition A variable that captures the gender mix of the joint account

investors determined as the proportion of males in the joint

account that traded (Williams & Meân, 2004)

Group size The number of persons trading through the same account as

determined by the unique dates of birth under the joint

account)

Location A categorical variable with three levels that indicates

investor’s location (Kenyan urban town, Kenyan rural town or

foreign town)

Location composition A variable that captures the location mix of the joint account

investors determined as either the proportion of joint account

investors that are from urban, rural, or foreign towns.

Investor nationality A categorical variable with three levels that indicate whether

the investor is a Kenyan, other East African (individuals from

Uganda, Tanzania, Rwanda, Burundi and South Sudan) or

foreign (individuals from the rest of the world)

22

Table 2

Descriptive statistics for individual and group investors at the NSE

Individual investors

(N=26,549)

Joint account investors

(N=11,922)

Welch’s t statistic

Number of distinct stocks traded per account

Minimum 1 1

Mean 2.13 5.23 t (15,185) = -68, p value =0.000

Maximum 49 51

Standard deviation 2.51 4.61

Number of trades per account

Minimum 1 1

Mean 6.61 25.33 t (12,716) = -18, p value <0.001

Maximum 1,997 8,265

Standard deviation 29.05 107.01

Mean number of buys per account 3.2 12.9

Mean number of sells per account 3.41 12.4

Mean volume of buys per account 14,728 287,291

Mean volume of sells per account 17,571 236,544

Mean value of buys per account (KES) 242,376 8,303,672

Mean value of sells per account (KES) 364,311 6,813,817

Value of trades per account (KES)

Minimum 5 70

Mean 662,360 15,118,756 t (11,982) = -7, p value <0.001

Maximum 2,077,010,883 18,718,773,980

Standard deviation 16,828,012 221,844,027

Investor age

Minimum 11 1

Mean 58.35 56.73 t (32,816) = 5, p value <0.001

Maximum 116 116

Standard deviation 31.73 21.34

Gender

Male 67.5% 51%

Female 29.8% 34.6%

Not specified 2.7% 14.4%

23

Location

Urban 84.6% 79%

Rural 4.4% 7.4%

Foreign town 1.2% 2.5%

Not specified 9.8% 11.1%

Investor type

Kenyan 98.6% 93.4%

Other East African 0.6% 1.7%

Rest of the world 0.8% 4.9%

Group size

Minimum 2

Mean 2.5

Maximum 14

Standard deviation 1.31

Notes: Individual investors represent accounts with only one individual transacting based on the date of birth listed under the account. Joint account

investors represent accounts with more than one individual transacting based on dates of birth listed under the investor account. The investor’s age

is calculated as at 01/01/2016 which is the starting date of the data set. Urban city is defined as a Kenyan city with a population of at least 2,000, a

rural city is a Kenyan city with a population of less than 2,000 as defined by the 2019 Kenya Population and Housing Census while a foreign city,

is defined as a city outside of Kenya. Group size is the number of investors trading through one account based on the dates of birth. Several investors

do not specify their gender, date of birth and location, and are thus categorised into the ‘not specified’ category. investors are classified into three

categories based on the nationality of the investor. Kenyan individual investors refer to investors with a Kenyan nationality, East African investors

are investors from other countries in the East African community that include Uganda, Tanzania, Rwanda, Burundi and South Sudan. Rest of the

world investors are from countries outside of Kenya and East Africa.

24

Table 3

Descriptive statistics for final sample of individual and group investors

Entire sample final sample

mean sd mean sd

Panel A: Individual investors

Number of observations 26,549 2,515

Number of distinct stocks traded per account 2.13 2.51 5.275 4.862

Value of trades (KES) 662,360 16,828,012 3,180,000 24,400,000

Number of trades 6.61 29.05 32.37 83.78

Age 58.35 31.73 38.61 18.88

Panel B: Group investors

Number of observations 11,922 3,711

Number of distinct stocks traded per account 5.23 4.61 8.28 5.84

Value of trades (KES) 15,118,756 221,844,027 37,874,153 378,599,316

Number of trades 25.33 107.01 56.81 185.63

Age 56.73 21.34 50 21.35

Notes:

Individual investors represent accounts with only one individual transacting based on the date of birth listed under the account.

The number of distinct stocks traded per account is the number of unique stocks traded by the investor during the 1-year sample

period of January 2016 to December 2016. The investor’s age is calculated as at 01/01/2016 which is the starting date of the data

set.

25

Table 4

Proportion of gains realised (PGR) and proportion of losses realised (PLR) by investor type

Individual investors

(N=2,515)

Group investors

(N= 3,711)

Aggregate

Average per

account Aggregate Average per account

Paper gains 808,249 321.37 868,000 245.61

Paper losses 1,007,928 400.77 2,121,615 582.86

Realised gains 5,112 2.03 9,460 3.61

Realised losses 5,461 2.17 10,033 3.97

% gains realised (PGR) = RG/(PG + RG) 0.63% 1.86% 1.078% 1.449%

% losses realised (PLR) = RL/(PL + RL) 0.54% 1.99% 0.47% 0.677%

% of investors where PGR>PLR 53% 62%

PGR/PLR ratio 1.17 2.325 2.29 3.24

PGR – PLR 0.0009 –0.0013 0.00608 0.00782

t statistic1 7.8945 -0.4495 3.852 6.8599

p value 0.000 0.6735 0.00006 0.000

Accept/Reject Ho @5% level Reject Accept Reject Reject

Notes: Notes: RG = realized gain; RL = realized loss; PG = paper gain; PL= paper loss. Individual investors represent accounts with only one individual

transacting based on the date of birth listed under the account. Group investors represent accounts with more than one individual transacting based

on dates of birth listed under the investor account. The aggregate values are determined by summing the paper gains, paper losses, realised gains and

realised losses across accounts and then calculating the proportion of gains realised (PGR) and proportion of losses realised (PLR) based on the

aggregate values. The averaged values per account are calculated individually for each investor and then averaged across investors. There is a

significant difference in PGR-PLR between individual and group investors t (4,354) = -2.664, p value = 0.0078.

1The t-statistic at aggregate level is calculated following a similar procedure as used by Odean (1998) as follows:

𝑡 =𝑃𝐺𝑅−𝑃𝐿𝑅

𝜎(𝑃𝐺𝑅−𝑃𝐿𝑅) , where 𝜎(𝑃𝐺𝑅−𝑃𝐿𝑅) = √

𝑃𝐺𝑅(1−𝑃𝐺𝑅)

𝑁𝑅𝐺+𝑁𝑃𝐺+

𝑃𝐿𝑅(1−𝑃𝐿𝑅)

𝑁𝑅𝐿+𝑁𝑃𝐿 where NRG, NPG, NRL, and NPL are the number of realized gains, paper

gains, realized losses, and paper losses.

26

Table 5

Investors disposition effect based on number of trades placed over the period (1: the least active traders, 10: the most active traders)

Decile groups 1 2 3 4 5 6 7 8 9 10

Panel A: Individual investors

No. of observations 327 246 220 287 223 253 229 236 246 248

Mean no. of trades 3.09 5.46 7.49 9.97 12.89 16.89 22.26 30.39 46.38 176.25

RG 168 163 174 233 251 313 347 407 660 2396

PG 27,043 35,728 42,235 66,799 58,357 86,948 84,333 103,611 126,189 177,006

RL 227 185 181 348 254 383 384 456 662 2381

PL 41,428 40,415 44,987 77,543 64,977 98,821 97,575 124,935 145,776 271,471

PGR 0.0062 0.0045 0.0041 0.0035 0.0043 0.0036 0.0041 0.0039 0.0052 0.0134

PLR 0.0054 0.0046 0.004 0.0045 0.0039 0.0039 0.0039 0.0036 0.0045 0.0087

PGR–PLR 0.0007 0 0.0001 -0.001 0.0004 -0.0003 0.0002 0.0003 0.0007 0.0047

PGR/PLR 1.133 0.997 1.024 0.778 1.1 0.929 1.045 1.076 1.151 1.536

Panel B: Group investors

No. of observations 464 336 376 342 407 327 350 367 375 367

Mean no. of trades 5.95 9.99 13.49 17.43 22.41 28.67 37.13 50.69 80.28 309.15

RG 46 58 120 143 273 313 487 726 1227 3833

PG 4,234 5,224 14,177 14,116 26,423 34,253 49,925 73,352 107,525 150,470

RL 41 52 121 127 256 314 559 894 1,420 4,012

PL 8,795 13,144 28,667 30,711 61,649 79,962 117,256 175,721 273,465 382,913

PGR 0.0305 0.0273 0.0204 0.0175 0.0254 0.0118 0.023 0.016 0.0114 0.0215

PLR 0.0161 0.0168 0.0058 0.0073 0.0042 0.0045 0.0051 0.0062 0.0058 0.0118

PGR–PLR 0.0144 0.0105 0.0146 0.0102 0.0212 0.0073 0.0178 0.0097 0.0056 0.0097

PGR/PLR 2.1186 2.6176 4.9968 1.9323 4.7966 2.3961 4.688 2.6222 2.735 3.3246

Notes:

Individual investors represent accounts with only one individual transacting based on the date of birth listed under the account. Group investors

represent accounts with more than one individual transacting based on dates of birth listed under the investor account.

27

Table 6

Influence of investor characteristics on disposition effect among individual investors at the NSE

(1) (2) (3) (4)

DE DE DE DE

Number of distinct stocks traded 1.037*** 1.049*** 1.036***

(0.013) (0.011) (0.013)

Number of trades 1.001* 1.003*** 1.002*

(.001) (0.001) (0.001)

Value of trades 1

(0)

Investor nationality

Rest of the world investor 0.712 0.708 0.724 0.714

(0.729) (0.727) (0.722) (0.73)

Kenyan investor 0.591 0.581 0.605 0.592

(0.538) (0.533) (0.533) (0.539)

Gender 0.897 0.907 0.894 0.897

(0.089) (0.09) (0.089) (0.089)

Age category

30-44 years 0.986 0.988 0.997 0.987

(0.106) (0.106) (0.107) (0.106)

45-59 years 0.788* 0.796* 0.813* 0.789*

(0.098) (0.098) (0.1) (0.098)

60-74 years 0.682** 0.686** 0.718** 0.683**

(0.103) (0.103) (0.107) (0.103)

75 years and above 0.536** 0.552* 0.56* 0.54**

(0.168) (0.168) (0.174) (0.167)

Location

Rural town 0.391 0.381 0.399 0.391

(0.263) (0.258) (0.266) (0.263)

Urban town 0.931 0.92 0.961 0.932

(0.563) (0.56) (0.574) (0.563)

Constant 1.99 1.992 2.123 1.986

(2.189) (2.204) (2.275) (2.184)

Observations 2342 2342 2342 2342

28

Pseudo R2 0.015 0.014 0.012 0.015

Robust standard errors are in parentheses *** p<.01, ** p<.05, * p<.1

Notes:

For the logistic regressions the dependent variable is the disposition effect (binary variable that takes the value of 1 if PGR-PLR is positive

and 0 otherwise. The independent variables include the number of stocks traded by each investor, number of trades that each investor has

executed, value of trades executed by the investor, investor nationality (Other East African, rest of the world or Kenyan), gender of the

investor (dummy variable that takes the value of 1 if investor is male and 0 otherwise), the investor’s age category and the location of the

investor (foreign, rural or urban town. Other East African investor, below 30 years, and foreign town are baselines for their respective groups.

In column 1 the logistic regression includes all independent variables. Regression in Column 2 excludes number of trades and the value of

trades, column 3 excludes number of stocks traded and value of trades while column 4 excludes the value of trades placed as independent

variables.

29

Table 7

Influence of investor characteristics on disposition effect among group investors at the NSE

(1) (2) (3) (4)

DE DE DE DE

Number of stocks traded 1.025*** 1.027*** 1.027*** 1.027***

(.008) (.008) (.008) (.008)

Number of trades 1.001** 1.001* 1.001* 1.001*

(.001) (0) (0) (0)

Value of trades 1

(0)

Group size 1.024 1.023 1.023 1.023

(.023) (.023) (.023) (.023)

Gender composition .863* .866* .867* .867*

(.073) (.073) (.073) (.073)

Location composition

Proportion urban 1.29 1.292

(.258) (.258)

Proportion rural .726

(.199)

Proportion foreign .817

(.283)

Mean age category

30-44 years .897 .895 .895 .897

(.101) (.101) (.101) (.101)

45-59 years .9 .895 .898 .897

(.118) (.117) (.117) (.117)

60-74 years 1.004 .996 .988 .987

(.147) (.146) (.144) (.144)

Constant 1.015 1.011 1.299** 1.289**

(.229) (.227) (.16) (.159)

Observations 3711 3711 3711 3711

Pseudo R2 .01 .01 .01 .01

Robust standard errors are in parentheses

*** p<0.01, ** p<0.05, * p<0.1 Notes: For the logistic regressions the dependent variable is the disposition effect (binary variable that takes the value of 1 if

PGR-PLR is positive and 0 otherwise. The independent variables include the number of stocks traded by each investor, number

of trades that each investor has executed, the value of trades executed by the investor, gender composition (proportion of males

in the group account), group size (the number of persons trading under the same account), the proportions of investors in a group

30

that are located in an urban town, rural town or foreign town, the group’s mean age category as determined by the mean age of

the group members. Below 30 years is the baselines for the mean age category. In column 1 the logit regression includes all

independent variables except proportion of group members in rural and foreign towns. Regression in Column 2 excludes the value of trades and the proportion of group members in rural and foreign towns, column 3 excludes value of trades and the

proportion of group members in urban and foreign towns while column 4 excludes the value of trades and the proportion of group

members in urban and rural towns as independent variables.

31

Figure 1: Data Clean-up and Classification Process

Where:

• LI – Local individual which means Kenyan individual

• EI – Individual from the East African Community, that is, Tanzania, Uganda, Rwanda,

Burundi and South Sudan

• FI – Foreign individuals are from the rest of the world

• LC – Local companies refers to Kenyan corporates

• EC– Corporates from the East African Community, that is, Tanzania, Uganda, Rwanda,

Burundi and South Sudan.

• FC – Foreign corporates are from the rest of the world

• JR – Junior investors are minors

• BR – brokers

• Gender of investor – Can be F(female), M(male) or N (not applicable)

• DOB – This is date of birth of the investor or date of incorporation of the company

All Accounts

41,630

Individuals & Corporates

(LI,EI,FI,LC,EC,FC)

41,625

Joint Accounts

(Have multiple DOB)

10,548

Individual Accounts

(Have one DOB and no blanks)

22,367

Individual Investors

21,109

Corporate Investors

1,258

Unknown

Have no DOB (2,078) or 1 DOB and no DOB (6,632)

Individual Accounts

(Have one gender)

7,323

Individual investors

5,440

Corporate investors

1,883

Joint Accounts

(Have multiple gender)

1,387

Others

(BR,JR)

5

Individual Investors

(26,549)

Group Investors

(11,935) Dropped Accounts

(1,263)

32

Figure 2: Predictive probability of disposition effect across age categories for individual

investors