University of Ghana  http://ugspace.ug.edu.gh
UNIVERSITY OF GHANA
OVERCONFIDENCE BIAS OF INVESTORS’ INVESTMENT DECISIONS ON
GHANA STOCK EXCHANGE
BY
RAPHAEL KURANCHIE-PONG
(10552007)
THIS THESIS IS SUBMITTED TO THE UNIVERSITY OF GHANA, LEGON IN
PARTIAL FULFILMENT OF THE REQUIREMENT FOR THE AWARD OF MPHIL
FINANCE DEGREE
JULY, 2017
University of Ghana  http://ugspace.ug.edu.gh
DECLARATION
I, Raphael Kuranchie-Pong, do hereby declare that this work is the result of my own research
and has not been presented by anyone for any academic award in this or any other university.
All references used in this work have been fully acknowledged.
I bear sole responsibility for any shortcomings.
---------------------------- ----------------------
RAPHAEL KURANCHIE-PONG (Date)
(10552007)
i
University of Ghana  http://ugspace.ug.edu.gh
CERTIFICATION
I hereby certify that this thesis was supervised in accordance with procedures laid down by
the University.
---------------------------- -------------------------
Dr. LORD MENSAH (Date)
(Supervisor)
------------------------------ -----------------------------
Prof. A. Q. Q.  ABOAGYE (Date)
(Supervisor)
ii
University of Ghana  http://ugspace.ug.edu.gh
DEDICATION
This work is dedicated to the Almighty God for His grace, my parents, Mr. John Kuranchie-
Pong and Mrs. Mary Kwantwi, to Brig. Gen. Adokpa and all my siblings for their love and
supports.
iii
University of Ghana  http://ugspace.ug.edu.gh
ACKNOWLEDGEMENT
I am indebted to my supervisors, Dr. Lord Menash and Prof. A. Q. Q. Aboagye for their
timeless dedication during the supervision of this work. I also acknowledge Mr. Charles
Kumi, Mr. Kwao Anooma, Miss. Lydia Kuranchie-Pong and all who have one way or the
other contributed to the success of this work.
iv
University of Ghana  http://ugspace.ug.edu.gh
ABSTRACT
The Efficient Market Hypothesis (EMH) postulates that all publicly available information on
any firm’s stock future performance is incorporated in the firm’s stock price such that the
actual price of the stock is equal to its intrinsic value. The EMH is argued to lead to “correct
prices” which creates the necessary environment for efficient allocation of economic
productive scarce resources. The documentation of empirical evidence of weak-form
inefficiency on most stock markets especially Africa, has received much attention by
regulators, researchers, investors and many other players in the financial industry. The Ghana
stock market, for instance, is documented in the literature to be weak-form inefficient.
In this study, the researcher turns to behavioural finance to test the rationality of the investors
on Ghana stock market since the rationality assumption is one of the pillars upon which
market efficiency is built. To test the rationality of the market participants, overconfidence
behavioural bias is used as a conditional rationality test proxy. The concentration on
overconfidence bias does not suggest that overconfidence is the only bias worth considering
but it is due to the fact that it is reported in the literature to be one of the robust psychological
behavioural biases.
Overconfidence is the act of having a mistaken assessment and believing in these assessments
too strongly. To test the overconfidence bias, a market-wide Vector Autoregressive (VAR)
model and its associated impulse response function is used to investigate the lead-lag
relationship between market returns and market trading volume. Also, granger-causality test
was performed to test the causality between returns and trading volumes. This study uses the
Ghana Stock Exchange All-Share index (GSE-ASI) and the Ghana Stock Exchange
Composite Index (GSE-CI) monthly returns data. In addition, the monthly trading volume,
v
University of Ghana  http://ugspace.ug.edu.gh
monthly returns volatility, interest rate and inflation data from 2000 to 2015 are used for the
study.
The findings reveal a significant impact of past market returns on current trading volumes.
The impulse response function analysis also showed that the response of trading volume to a
shock of market return residual remains highly positive and significant at lag one and two.
The granger causality reveals unidirectional causality running from returns to trading
volumes. The empirical findings finally demonstrate that market participants on GSE exhibit
conditional irrationality (overconfidence bias as proxy) in their investment decisions.
vi
University of Ghana  http://ugspace.ug.edu.gh
TABLE OF CONTENTS
Contents Page
DECLARATION ................................................................................................................... i
CERTIFICATION ................................................................................................................ ii
DEDICATION..................................................................................................................... iii
ACKNOWLEDGEMENT.................................................................................................... iv
ABSTRACT ......................................................................................................................... v
TABLE OF CONTENTS .....................................................................................................vii
LIST OF FIGURES .............................................................................................................xii
LIST OF TABLES ..............................................................................................................xiii
LIST OF ABBREVIATIONS .............................................................................................xiv
CHAPTER ONE ................................................................................................................... 1
INTRODUCTION ................................................................................................................ 1
1.1 Background of the Study......................................................................................... 1
1.2 Problem Statement .................................................................................................. 5
1.3 Research Objectives ................................................................................................ 9
1.4 Research Questions ................................................................................................. 9
vii
University of Ghana  http://ugspace.ug.edu.gh
1.5 Significance of the Study ...................................................................................... 10
1.6 Scope of the Study ................................................................................................ 12
1.7 Chapter Disposition .............................................................................................. 12
CHAPTER TWO ................................................................................................................ 13
LITERATURE REVIEW .................................................................................................... 13
2.1 Introduction .......................................................................................................... 13
2.2 Structure of Ghana Stock Exchange ...................................................................... 13
2.3 Behavioural Finance ............................................................................................. 14
2.3.1 The Foundations of Behavioural Finance ....................................................... 14
2.3.2 Behavioural Biases of Investment Decisions .................................................. 16
2.4 Theoretical Review ............................................................................................... 19
2.4.1 Overconfidence Bias ...................................................................................... 19
2.4.2 Overconfidence Bias in financial markets ...................................................... 21
2.4.2.1 Overconfidence bias theories in finance...................................................... 23
2.4.2.2 Implications of Overconfidence bias on financial markets .......................... 26
2.4.2.2.1 Overconfidence bias and Trading Volume............................................. 26
2.4.2.2.2 Overconfidence bias and Market returns................................................ 27
viii
University of Ghana  http://ugspace.ug.edu.gh
2.4.2.2.3 Overconfidence and differential reactions to private and public
information ........................................................................................................... 29
2.4.2.2.4 Overconfidence bias and excessive volatility......................................... 29
2.4.2.2.5 Causality relations between trading volume and stock returns ............... 30
2.5 Review of Previous Studies on Overconfidence bias in financial markets.............. 32
2.5.1 Evidence from Developed Markets ................................................................ 33
2.5.2 Evidence from Emerging Markets .................................................................. 44
CHAPTER THREE............................................................................................................. 46
RESEARCH METHODOLOGY ........................................................................................ 46
2.1 Introduction .......................................................................................................... 46
3.2 Data Sources ......................................................................................................... 46
3.3 Sample period ....................................................................................................... 47
3.4 Market Activities of Ghana Stock Exchange from 2000 – 2015............................. 48
3.5 Variables............................................................................................................... 51
3.5.1 Endogenous Variables.................................................................................... 52
3.5.1.1 Monthly Detrended Logged Trading Volume (MDLTVOL) ....................... 52
3.5.1.2 Thin trading and Monthly Market Return (MRET) ..................................... 55
ix
University of Ghana  http://ugspace.ug.edu.gh
3.5.2 Control (Exogenous) Variables ...................................................................... 56
3.5.2.1 Market Return Volatility (MSIG) ............................................................... 56
3.5.2.2 Inflation and Interest Rate (Macroeconomic Variables) .............................. 57
3.6 Empirical Methodology ........................................................................................ 58
3.6.1 The General Vector Autoregressive Model (VAR) ......................................... 58
3.6.2 Theoretical framework for the model ............................................................. 59
3.6.3 VAR Model specification for the study .......................................................... 60
3.6.3.1 Impulse Response Functions (IRF) ............................................................. 61
3.6.3.2 Optimal Lag Selection ................................................................................ 62
3.6.3.2.1 Cross-equation Restrictions (Likelihood Ratio (LR) Test) ..................... 63
3.6.3.2.2 Information Criterion ............................................................................ 63
3.6.4 Granger Causality Test................................................................................... 64
CHAPTER FOUR............................................................................................................... 66
RESULTS AND DISCUSSION .......................................................................................... 66
4.1 Introduction .......................................................................................................... 66
4.2 Descriptive Statistics............................................................................................. 66
4.3 VAR Estimation and Test Results ......................................................................... 69
x
University of Ghana  http://ugspace.ug.edu.gh
4.3.1 Optimal Lag Selection ................................................................................... 69
4.3.2 Market VAR Estimation Results .................................................................... 71
4.3.3 Impulse Response Functions (IRF)................................................................. 75
4.4 Granger Causality for market returns and trading volume results........................... 76
CHAPTER FIVE ................................................................................................................ 78
SUMMARY, CONCLUSION AND RECOMMENDATIONS ........................................... 78
5.1 Introduction .......................................................................................................... 78
5.2 Summary .............................................................................................................. 78
5.3 Conclusion............................................................................................................ 79
5.4 Recommendations................................................................................................. 80
REFERENCES ................................................................................................................... 82
APPENDICES .................................................................................................................... 92
xi
University of Ghana  http://ugspace.ug.edu.gh
LIST OF FIGURES
Figure 2.1: Foundations of Behavioural Finance (Source: Schindler, 2007) ......................... 15
Figure 3.1: A Graph of Monthly Returns on GSE from 2000 – 2015 ................................... 47
Figure 3.2: A Graph of Monthly Trading Volumes on GSE from 2000 – 2015 ............. Error!
Bookmark not defined.
Figure 3.3: A Graph of Yearly Market Capitalization of GSE from 2000 – 2015 ................. 49
Figure 3.4: A Graph of Yearly Market Trading Volumes of GSE from 2000 – 2015............ 50
Figure 3.5: A Graph of GSE Annual Returns from 2000 – 2015 .......................................... 51
Figure 3.6: A Graph of Detrended Logged Market Trading Volume by HP filter.......... Error!
Bookmark not defined.
Figure 4.1: Impulse Response Function (IRF) Graph of the Market-wide VAR model ........ 75
xii
University of Ghana  http://ugspace.ug.edu.gh
LIST OF TABLES
Table 2.1: Some of the Most Common Behavioural Biases ................................................. 18
Table 4.1: Descriptive Statistics Results .............................................................................. 67
Table 4.2: VAR Lag Order Selection Criteria Results.......................................................... 70
Table 4.3: Vector Autoregression (VAR) Estimation Results .............................................. 72
xiii
University of Ghana  http://ugspace.ug.edu.gh
LIST OF ABBREVIATIONS
BOG - Bank Of Ghana
EMH - Efficient Market Hypothesis
GSE - Ghana Stock Exchange
GSE-ASI - Ghana Stock Exchange All-Share Index
GSE-CI - Ghana Stock Exchange Composite Index
HP - Hodrick-Prescott (1997) Filter
MDLTVOL - Monthly Detrended Logged Trading Volume
MRET - Monthly Index Returns
MSIG - Monthly Volatility
VAR - Vector Autoregressive
xiv
University of Ghana  http://ugspace.ug.edu.gh
CHAPTER ONE
INTRODUCTION
1.1 Background of the Study
In the literature, the nexus between economic growth and financial markets are demonstrated
and established in both theoretical and empirical grounds (Beck & Levine, 2004). Financial
markets such as capital markets play a very significant role in the building of the financial
systems and development of economies. The 2008 global financial crisis which originated in
USA and spread globally is a typical example of the impacts of financial markets on the
world at large. As a result, financial markets have become one of the key areas of study by
researchers and practitioners with focus on both advanced and emerging markets.
The Efficient Market Hypothesis (EMH) is one of the fundamental building blocks of
standard finance for understanding financial market activities. A capital market is efficient if
asset prices observed on the market fully reflect all publicly available information. This
implies that in an efficient market, information processing is done rationally in such a way
that no systematic errors are done and significant information is not discarded. As a result,
financial asset prices are claimed to be equal to the intrinsic value of the asset since all
available public information are incorporated into the asset price during the price formation.
Therefore, asset prices are said to be “correct”. Jensen (1978) argues further that in financial
settings where information accessibility comes with a cost, the market is said to be efficient
when prices of securities reflect information to the level where the marginal benefits of
employing the information is not greater than the marginal cost of obtaining the information.
It is worth stating that the EMH is built on perfect market assumptions which include no
1
University of Ghana  http://ugspace.ug.edu.gh
transaction costs, information is costless, homogenous expectations of investors and
rationality of market participants.
According to Fama (1970), for an economy to derive optimal benefit from capital market in
the form of capital allocation, the capital market must be efficient. A firm performance is
measured by its profitability and to a large extent, its efficiency. The society votes in terms of
their purchasing power on the markets for a product to some extent determine the profitability
of the firm which also directly affects the performance of the firm. The performance of the
firm then influences the price of the firm’s security. It is worth indicating that a firm’s
security price would reflect its present and future performance if the market is efficient. This
implies that under efficient market, security prices are formed efficiently. Therefore,
investors identifying profitable future performance of the firm would invest in such a firm.
Thus, indirectly the society’s scarce resources are channelled to productive sectors of the
economy. This implies that the efficient allocation of resources in any economy to a large
extent depends on the efficiency of the capital markets. On the other hand, if security prices
are inefficiently formed, it would lead investments to inappropriate sectors of the economy
which would cause a severe social cost to the economy. Thus, there is an established link
between capital market efficiency and economic efficiency in resource allocation. According
to Fama (1970), an efficient market is a necessary condition for efficient allocation of scarce
economic resources in any economy. Fama (1970) states that:
The primary role of the capital market is allocation of ownership
of the economy’s capital stock. In general terms, the ideal is a
market in which prices provide accurate signals for resource
2
University of Ghana  http://ugspace.ug.edu.gh
allocation: that is, a market in which firms can make production-
investment decisions, and investors can choose among the
securities that represent ownership of firms’ activities under the
assumption that securities prices at any time “fully reflect” all
available information. A market in which prices always “fully
reflect” available information is called “efficient”.
(Fama, 1970, p. 383)
As a result, over the past six decades, researchers have focused on testing and explaining
efficiency of financial markets. In testing the market efficiency, some researchers test the
EMH prediction to ascertain whether: asset prices move as random walk over time is true;
information is rapidly incorporated into asset prices so that current available information
cannot be used to predict future excess returns; technical analysis does not yield excess
return; fund managers do not beat the market; and asset prices remain at levels consistent
with economic indicators. However, research over the years has shown that most financial
markets including emerging markets are inefficient and characterized by many anomalies
such as excessive volatility, high trading volumes, market bubbles and crashes, calendar,
fundamental and technical anomalies. Yakob, Beal & Delpachitra (2005) for instance
employed data from ten Asian markets and documented the existence of seasonality effects.
Similar results are also reported on the African stock markets which Ghana stock market is
no exception. For example, Appiah-Kusi & Menya (2003), Dewotor & Gborglah (2004),
Simons & Laryea (2006) and Mensah, Bokpin & Owusu-Antwi, 2016 reported market
inefficiency and calendar anomaly on some African stock markets.
3
University of Ghana  http://ugspace.ug.edu.gh
Notwithstanding the benefits of efficient market, the evidence of inefficiencies and anomalies
on some equity markets have caused researchers to dive into the possible reasons for such
happenings. Thus, a more convincing and extensive way of thinking and understanding the
behaviour of market participants, financial markets and the whole society at large became
intense. As a result, academic discussions shifted from standard finance theories to the
application of human psychology to finance to test the validity of the market participants
rationality assumption of efficient market hypothesis. This led to the birth of Behavioural
finance.
Behavioural finance researchers and practitioners argue that human psychology and
behaviour plays an active role in modelling, analysing and understanding financial markets
since both are inherently related. It is worth pointing out that behavioural finance emerged
partly from the doubts and qualms about the validity of the assumptions of the standard
finance theories. Several psychological findings reported in the psychology and finance
literature demonstrate that market participants do not always act in a rational manner but
rather are victims of cognitive and emotional biases. These biases lead to deviations from the
predictions of standard finance theories. Behavioural finance theories claim that if the
assumptions of rationality of economic agents are relaxed, many of the financial puzzles and
anomalies would be better understood. Some behavioural biases observed on equity markets
such as disposition effect and overconfidence bias are documented to explain some of the
inefficiencies and anomalies observed on the markets.
Overconfidence bias is cognitive psychological bias phenomenon occurring when market
participants are too confident and exaggerate their own skills and abilities in selecting
securities and portfolios as well as predicting securities future returns (Grinblatt & Keloharju,
4
University of Ghana  http://ugspace.ug.edu.gh
2009). Research has shown that overconfidence bias of market players is the reason for some
of the inefficiencies documented on the financial markets. Overconfidence bias of investors is
documented to cause underreaction and overreaction of stock price to public information
which leads to short and long run autocorrelation of returns. Overconfidence of market
participants is also documented to lead to excessive trading (Odean, 1998). Research have
also shown that overconfidence bias of traders on financial markets contribute to excessive
volatility observed in financial markets (Chuang & Lee, 2006). Akerlof & Shiller (2010) and
Authers (2010) argue that overconfidence of market participants is one of the major variables
that trigger beliefs change of market actors and offers the necessary environment for the
various market anomalies being observed on financial markets. Therefore, is it possible that
the documentation of inefficiency and anomalies on Ghana stock markets are as a result of
behavioural biases of the market participants?
1.2 Problem Statement
Efficiency of stock markets have been tested on most markets, both developed and emerging
markets and the discussion around it is still very vivid. Testing of efficiency of a market
provides vital information as to whether market security prices are always “right”. The Bulk
of these empirical testing of markets efficiency is conducted on developed markets and mixed
results are reported (Mobarek & Keasey, 2000 (cited in Dickson, 2015)). For instance,
Choudhry (1994) provided empirical evidence to support market efficiency whiles Lo &
MacKinlay (1988) also provided empirical evidence of non-random walk of security prices
and as a result reject the market efficiency hypothesis.
African equity markets are characterized by small market size and less liquidity as compared
to developed markets. A vast number of empirical studies have been conducted on African
5
University of Ghana  http://ugspace.ug.edu.gh
equity markets as well to test its efficiency. However, there are mixed empirical results on
African stock markets’ efficiency. Some of these empirical results reported weak-form
efficiency (Appiah-Kusi & Menya, 2003) and others have also reported weak-form
inefficiencies (Appiah-Kusi & Menya, 2003; Simons & Laryea, 2006). Some researchers
argue that the efficiency of African stock markets progressively change over time (Jefferis &
Smith, 2005). Also, quite a number of studies on African equity markets (Alagidede &
Panagiotidis, 2009; Aly, Mehdian & Perry, 2004) have reported the existence of calendar
anomalies. However, no empirical possible reasons have been provided for such phenomena
on equity markets in Africa.
Ghana stock market was judged as the best performing market in the world by the end of
2003 (Adjasi, Harvey & Agyapong, 2008). However, most empirical evidences on Ghana
Stock Exchange indicate inefficiencies and anomalies on the market. A vast number of
empirical studies to test GSE efficiency have been conducted over the years. Magnusson &
Wydick (2002) using eight African equity indices which included Ghana investigated their
efficiency by employing partial autocorrelation from 1989 to 1998. Appiah-Kusi & Menya
(2003) also used eleven African equity markets including Ghana to test the weak-form
efficiency in these markets by employing EGARCH-M model with thin trading correction.
Dewotor & Gborglah (2004) employing serial and cross-sectional correlation tests
investigated the efficiency of Ghana stock market using daily, monthly and yearly data. Also,
Ayentimi, Mensah & Naa-Idar (2013) also examined the weak-form market efficiency on the
Ghana stock market. The findings of all these empirical studies unanimously reported that the
Ghana stock market is weak-form inefficient. However, these studies do not provide any
empirical possible reasons for the inefficiency being observed on GSE.
6
University of Ghana  http://ugspace.ug.edu.gh
Also, studies by Alagidede & Panagiotidis (2009) and Mensah et al. (2016) have all indicated
the presence of calendar anomalies specifically day of the week effect on Ghana stock
market. Dickson (2015) applied Variable Moving Average (VMA) technical trading
strategies on the Ghana Stock Exchange Composite Index (GSE-CI) to ascertain whether it
would be profitable and outperform the buy-and-hold strategy. Dickson (2015) reported that
VMA generated positive returns and four out of the five VMA employed outperformed the
results from buy-and-hold rule. These empirical evidences also shows that the GSE market is
inefficient and they do not provide any empirical reasons for such phenomena.
In addition, there are many empirical studies demonstrating a relationship between
macroeconomic variables and Ghana stock market performance. For instance, Issahaku,
Ustarz & Domanban (2013) employed monthly data from 1995 to 2010 to examine the
presence of causality between macroeconomic variables and stock returns in Ghana. They
used Vector Error Correction (VECM) model and Granger causality for this study. They
reported short run and long run relationship between stock returns and macroeconomic
indicators. Other studies reporting a nexus between stock performance and macroeconomic
indicators on Ghanaian market include Antwi, Ebenezer & Zhao (2012), Kuwornu (2012) and
Kyereboah-Coleman & Agyire-Tettey (2008). The empirical evidence of nexus between
stock returns and macroeconomic indicators imply the existence of arbitrage profit
opportunities which is contrary to the inferences from EMH that stock returns could not be
predicted. Also, these studies also do not provide any empirical reasons for these market
inefficiencies documented on the Ghana stock market.
In order to improve the efficiency on Ghana stock market, GSE authorities introduced
automation trading platform system in 2011. Mensah, Adom & Pomaa-Berko (2014)
7
University of Ghana  http://ugspace.ug.edu.gh
examined the impact of the automation trading system on the GSE efficiency and also
reported that the Ghana equity market is still inefficient. Moreover, the rejection of the weak
form efficiency and calendar anomalies are argued to be as a result of the small size of the
market and considerable high transection cost which leads to inadequate market activities and
less liquidity. It is worth pointing out that these arguments are only persuasive rather than
empirical arguments. According to Appiah-Kusi & Menyah (2003) the market size of an
equity market cannot alone lead to market inefficiency. Thus, there are no empirical possible
reasons for the evidence of inefficiency and calendar anomalies on Ghana stock market in the
literature.
To fill this gap in the literature, this study switches to behavioural finance to test the
rationality of investors on Ghana stock market since rationality of investors is one of the
cornerstones upon which market efficiency is built. According to Barberis & Thaler (2003),
market agents are rational if arrival of new information is interpreted correctly and uniformly
by the market participants. In addition, market participants are said to be rational if they make
decisions and choices in an unbiased manner and form expectations of the future in the
manner described by Baye’s law (Barberis & Thaler, 2003). Overconfidence bias is used as a
proxy to test the rationality of market participants on Ghana stock market. It is important
emphasising that the concentration on overconfidence bias does not suggest that
overconfidence is the only bias worth considering but it is due to the fact that it is reported in
the literature to be one of the robust psychological widespread behavioural biases (DeBondt
& Thaler, 1995). To the best of the researcher’s knowledge, no study has attempted to
examine the rationality of investors on GSE either by employing any behavioural biases
phenomena or any other means. This study provides empirical possible reason for the
observation of Ghana stock market inefficiency.
8
University of Ghana  http://ugspace.ug.edu.gh
1.3 Research Objectives
The primary objective of this study is to investigate the rationality of GSE market participants
by testing the existence of overconfidence behavioural bias of market actors on GSE.
Overconfidence bias of investors is important since some researchers such as Akerlof &
Shiller (2010) and Authers (2010) argue that overconfidence of investors is one of the major
variables that trigger beliefs change of investors on financial markets and offers the necessary
environment for the various market anomalies. Thus, the rationality of market participants on
Ghana stock market is tested by investigating the existence of overconfidence bias on the
market using data from 2000 to 2015. In view of this, the objectives of the study are:
i. To investigate whether overconfidence bias exists on GSE market.
ii. To examine whether trading volume granger cause stock returns or/and stock
returns granger cause trading volume.
1.4 Research Questions
Following the objectives considered for this study which has been provided above, the
following research questions are posed;
i. What is the relationship between the current GSE trading volumes and past
GSE market returns?
ii. Is there a bidirectional granger-causality between trading volumes and market
returns on GSE?
9
University of Ghana  http://ugspace.ug.edu.gh
1.5 Significance of the Study
Assets value and quantity are considered to be the elementary pillars for market theories.
Therefore, a better understanding of market participants and aggregate markets behaviour
will make the understanding of financial markets easier and better. In the financial economic
literature, most studies have focused extensively on understanding and explaining the linkage
between assets returns, asset returns volatility and asset trading volumes. According to
Karpoff (1987), the empirical nexus between asset returns and the asset trading volumes is
imperative since such a relationship of empirical evidence aid a better understanding of the
market information flow theories. Thus, this study provides some evidences on the
relationship between equity price and trading volumes and as result gives an insight into the
information flow on GSE.
The primary goal for this study is to investigate the (conditional) rationality of GSE market
participants by testing the presence of overconfidence bias on Ghana stock market. This is
achieved by analysing the empirical lead-lag relations existing between GSE equities trading
volumes and equities returns. The evidence of current market trading volumes being
explained by past equities returns is an indication of existence of overconfidence bias in the
market. This study contributes to literature on behavioural finance on African stock markets
in several ways. Firstly, it is one of the few studies in Africa which have considered the
overconfidence behavioural bias on African stock markets. This is important because
behavioural biases help in explaining some of the inefficiencies and anomalies on stock
markets. To the best of the researcher’s knowledge, this is the first study that examines any
behavioural bias on GSE markets. Moreover, the study contributes and expands the debate on
overconfidence bias and also the causality relationship existing between asset returns and
asset trading volumes by presenting evidence from Ghana. The study therefore provides an
10
University of Ghana  http://ugspace.ug.edu.gh
insight into rationality of GSE market participants and as a result provides empirical possible
contributing reason for the inefficiency and anomalies on GSE market.
The documentation of empirical evidence of behavioural biases on financial markets is
important for financial markets participants especially for hedge funds, mutual and
investment funds managers in developing their trading or investment strategies. Their main
aim is to diversify risk in the process of allocating their capital. If financial markets
participants are rational and markets are efficient, fund managers would follow Markowitz
Portfolio Theory to buy well diversified portfolios in order to balance out the expected return
and expected risk. This study would provide investors the necessary knowledge in order to
choose the appropriate investment strategy based on the behaviour of the market in order to
maximize their investment wealth. For instance, market participants can adopt Doviak (2015)
strategies for applying behavioural finance to one’s practise or overconfidence bias portfolio
strategy suggested by Daniel & Titman (1999) which is based on market capitalization,
momentum and book-to-market ratios and reported in the literature to earn excess returns
than can be explained by the traditional finance framework and Fama & French (1993) three-
factor framework.
In recent times, the globalization and the interdependence of financial markets have resulted
to foreign investors having keen interest in foreign markets as a way of diversifying their
portfolio risk. According to Cooray & Wickremasinghe (2007), a well-researched market in
terms of its efficiency and behaviour attracts foreign investment which boosts domestic
savings. Therefore, this study would contribute to the research on GSE especially to the
behaviour of market participants on GSE.
11
University of Ghana  http://ugspace.ug.edu.gh
This study would also assist policy makers on Ghana Stock Exchange to appreciate the fact
that the inefficiency on the market could be due to behavioural biases other than the argument
of small market size and illiquidity of the market. The study encourages policy makers on
GSE to turn to behavioural biases as the possible cause of the anomalies on the market so that
the appropriate policy prescriptions such as introducing behavioural bias into the GSE
programme content and caution to investors can be provided to reduce these anomalies.
1.6 Scope of the Study
Most of the empirical research works relating to behavioural biases for that matter
overconfidence bias on stock markets are conducted in the developed countries using the
setting pertaining there with little being documented on African markets. An empirical study
in this area is highly required especially in Africa stock markets which are characterized with
less liquidity, opaque and volatile. Due to constraints of access to data, this study is limited to
one African stock markets namely, Ghana stock market, considering data from 2000 to 2015.
1.7 Chapter Disposition
The study is organized into five (5) main chapters. Chapter one is made up of background of
the study, problem statement, research objectives and questions, significance of the study and
the scope of the study as well as chapter disposition. Chapter two covers the relevant
literature review on the study mainly from related journal publications in the area. Chapter
three looks at the methodology or framework adopted for the study, providing relevant
information on data sources, and how the data was analysed. The Chapter four of this study
presents the results and discussion on the findings of the study. The Chapter five of this study
presents the summary, conclusions and recommendations grounded on the empirical
evidences obtained in the preceding chapter.
12
University of Ghana  http://ugspace.ug.edu.gh
CHAPTER TWO
LITERATURE REVIEW
2.1 Introduction
The discussion of the literature review in this chapter delves into the structure of Ghana Stock
Exchange, behavioural finance, investment behavioural biases, and overconfidence bias and
its associated theories. Empirical literature on overconfidence bias is then reviewed.
2.2 Structure of Ghana Stock Exchange
A Governing Council made up of nine members with two representatives from listed
companies, two representatives from licensed dealing members, three independent members
and two executive officers steer the affairs of GSE. This Governing Council has three main
committees, namely, the Listing committee, Finance committee and the Risk Management
and Surveillance committee which foster the smooth running of the exchange (GSE, 2016).
The prime role of the Listing Committee is to consider applications for listing whiles the
Finance Committee is charged with the responsibility on financial matters of the Exchange
such as budgets and remuneration (GSE, 2016). Also, the appraisal, recommendation and
approval of applications for membership are the duties of the Risk Management and
Surveillance committee. However, an ex-officio member is appointed as the Managing
Director of the Exchange who becomes the head of the management staff and has
responsibility of managing the day-to-day activities of GSE with the help of a Deputy
Managing Director (GSE, 2016).
GSE is a public company limited by guarantee with its membership grouped under Associate
members, Licensed Dealing Members (LDMs) and Government Securities Dealers (PDs)
13
University of Ghana  http://ugspace.ug.edu.gh
(GSE, 2016). The Associate members of the Exchange are individuals and legal corporate
bodies that provide the necessary support for the aims of the Exchange to be achieved. On the
other hand, the LDMs are legally registered members of the exchange who are qualified to
trade on the Exchange. The PDs are also legally corporate entities who have the authorization
from both GSE and Bank of Ghana to deal only in securities issued by the government.
2.3 Behavioural Finance
Behavioural finance is a new and evolving paradigm of finance that studies how sociological
and psychological factors of financial market participants influence market participants’
decisions and choices while buying or selling assets (Nofsinger, 2001). Thus, behavioural
finance focuses on the impact of psychological and sociological factors on how financial
market participants make decisions and choices and its successive impacts on the markets as
a whole. Behavioural finance is a multidisciplinary field of studies which employs
psychological and sociological factors to explain the thinking processes of individual market
participant, groups, and entities including the emotional patterns involved in their financial
decision-making process. Behavioural finance tries to answer the why, how and what
investors consider in their financial decisions and investing. Behavioural finance explains that
psychological, social, and emotional variables like cognitive ability and neural processes as
well as external environment variables such as crowd psychology and media influence are
very essential to the market participants’ financial decisions and the behaviour of the
aggregate financial markets.
2.3.1 The Foundations of Behavioural Finance
Behavioural finance discussions in the academic literature have taken several forms and
different perspectives. Numerous researchers, authors and practitioners have provided their
14
University of Ghana  http://ugspace.ug.edu.gh
respective viewpoints, interpretations and understanding of this new evolving field. As a
result, it is necessary to explain the cornerstones of behavioural finance which are
psychology, sociology and finance.
Psychology Sociology
Behavioural Finance
Finance
Figure 2.1: Foundations of Behavioural Finance (Source: Schindler, 2007, p. 18)
Psychology as a discipline involves the scientific study of human behaviour and their mental
processes. It also considers how human mental abilities and capabilities are influenced by its
physical, mental state and the outward environment. Sociology on the other hand, involves
the systematic study of the social behaviour of humans and the impact of group influence on
human behaviour. Therefore, the sociology aspect focuses and considers the impact of
humans’ social relationships on their way of life and behaviour. Finance is a discipline that
primarily focuses on the determination of value and allocation of capital, including the
acquiring, investing and managing resources.
Application of psychological studies in finance has revealed that economic agents
persistently display certain biases in the process of forming their belief system and as a result
affect their decision making. Sociology on one hand emphasizes that a lot of the financial
decisions taken by market participants’ result from social communication such as crowd
psychology rather than these decisions being taken in isolation as postulated by standard
15
University of Ghana  http://ugspace.ug.edu.gh
finance theories. Thus, Sociology therefore refutes the axiom that economic agents make
choices and decisions without external impacts. However, standard finance is still the
linchpin in the concepts of behavioural finance but with the behavioural traits of
psychological and sociological variables of humans as the primary catalyst in this new area of
study. As a result, in order to appreciate behavioural finance, one must familiarize himself
with the basic concept of psychology, sociology and finance disciplines in order to become
acquainted with the overall theories of behavioural finance.
2.3.2 Behavioural Biases of Investment Decisions
Financial investment decisions such as where to invest one’s funds, where to save and bank
are routine decisions that are usually taken by several players like individual market actors,
governments, institutional investors, fund managers and other players in the financial world.
For example, institutional investors and fund managers have to decide which kinds of assets
to hold, the percentage of funds to be allocate to the various portfolios and when to sell their
positions in the markets.
One of the central goals of behavioural finance is to draw the attention and deepen the
understanding of financial market players and researchers on how market agents design their
expectations about the future, behave towards uncertainty and risk, make decisions and
choices, what kinds of assets investors should hold and the best way to trade in the financial
markets (Oslen, 2008). For such purpose to be achieved, behaviourists argue that some
aspects of the standard finance axioms about economic agents must be relaxed. That is,
economic agents are not as smart as standard finance theory postulates and Kahneman puts it
in this way; “the mind is a system of jumps to conclusions” in his 2002 Nobel Prize Lecture.
Psychological empirical evidences indicate that the economic agents’ information processing
16
University of Ghana  http://ugspace.ug.edu.gh
ability is quite often weak. According to Newell & Simon (1972), when humans are
confronted with difficult problems, they tend to rely on psychological occurrences, long
memories and heuristics resulting in biases in dealing with such problems. Many studies have
reported that many of these biases are as a result of analytical processing and cognitive
limitations. These entire constraints limit human’s capability to make accurate judgments.
According to psychologists, the human brain is a receptive of considerable large volumes of
sensory signals whenever it faces a challenge that needs a solution. Psychologists claim that
whenever the human brain comes into contact with a new problem, the brain recognizes the
changes and as a result arranges the necessary signals in a more efficient way and in the
process of doing that, a considerable amount of the data or signal is discarded instantly.
Specific psychological compression system is employed by the human brain to filter the data
received and as a result, only a very small proportion of it is transformed into significant
information. Therefore, beliefs concerning the probability of the occurrence are then formed
based upon this information, even though possible respective payoffs of the event’s outcomes
would be uncertain.
However, according to Kahneman, Slovic & Tversky, (1982) the determination of such
probabilities and forecasting the outcome of uncertain problems are converted to simple
judgmental exercise which is made possible by the human brain associating with past
experiences that are stored in human’s long memory. Thus, market participants usually
forecast about the future and make choices when confronted with complex and uncertain
problems and their choices may not be optimal since their decisions and choices are based on
heuristic simplifications which lead to behavioural biases in the form of a systematic and
consistent error in judgment (Chen, Kim, Nofsinger & Rui, 2007).
17
University of Ghana  http://ugspace.ug.edu.gh
Behavioural biases have enormous impact on the way in which financial decision makers
acquire and process information for their financial decisions and choices. However, Binmore
(1999) argued that these biases can only be considered to have influential impacts on humans
if it persists in an environment where persons continually come in contact with the similar
decision problems. Empirical studies conducted from laboratory atmosphere to trading floors
have all indicated the presence of behavioural biases when trading. Practical examples of
such behaviours is what was witnessed with most market participants actively involved in
trading during the 1999-2000 internet bubble and 2007 US housing burst. Table 2.1 below
shows some of the major behavioural and psychological biases reported in the literature.
Table 2.1: Some of the Most Common Behavioural Biases
Ambiguity Aversion: Refers to the situation Disposition effect: The tendency of market
investors make preference for ambiguity-free participants to sell winning assets and rather
options even when they seem less likely to hold on to the ones that has fallen in value.
succeed and rather stick with the “known” to
avoid uncertainty and chaos. Optimism: Investors believe that bad
happenings occur to others; wishful thinking.
Anchoring: The tendency to hang on to
previous signals such as arbitrary pricing Overconfidence: The situation where
levels when making an investment decisions. investors overestimates their own skills and
abilities and also attribute failure to bad luck
Cognitive Dissonance: The state of having while success to their skills
inconsistent thoughts, beliefs and attitudes
towards an investment decision. Recency: Overemphasis is placed upon
events that are most recent.
Confirmation: Selective screening and
seeking out data to fit the existing position
whilst avoiding information that undermine
the positions.
Author’s Compilation (Sources: Samuelson & Zeckhauser, 1988; Pompian & Longo, 2005)
18
University of Ghana  http://ugspace.ug.edu.gh
2.4 Theoretical Review
2.4.1 Overconfidence Bias
Psychologists believe that most individuals are not well calibrated and as a result, place too
much belief in the precision of their judgements and knowledge. Empirical studies in
psychology on calibration of subjective probabilities indicate that individuals overestimate
the correctness of their judgements and knowledge. Koriat, Lichtenstein & Fischhoff (1980)
indicate that “an individual is well calibrated if, over the long run, for all answers to given
questions are assigned a given probability, the proportion correct equals the probability
assigned” (p. 108).
A vast number of studies have documented that people are miscalibrated in their judgements.
Miscalibration is one form of overconfidence manifestation. Such overconfidence has been
documented in several professional environments which include psychologists, physicians
and nurses, engineers, attorneys, negotiators, entrepreneurs, managers, investment bankers,
and market professionals such as security analysts and economic forecasters.
Another strand of the literature on overconfidence in psychology also shows that people are
positively unrealistic in assessing themselves. According to Taylor & Brown (1988) (cited in
Odean, 1998), most people perceive themselves as better than the average people and also see
themselves better than others perceive them. Thus, most people perceive their skills and
abilities to be greater as compared to their peers. Hence, people tend to attribute their own
skills and abilities to their past positive results and therefore easily recall their successes more
than their failures. Researchers argue that when people forecast the outcome of a certain
event and the outcome of the event turns out to be what they forecasted, people tend to
overestimate their contributions in bringing it about. Taylor & Brown (1988) (quoted in
19
University of Ghana  http://ugspace.ug.edu.gh
Odean, 1998) indicated that such successive beliefs by individuals in their own abilities may
result to “higher motivation, greater persistence, more effective performance and ultimately,
greater success.” Researchers have documented that such beliefs can trigger biased
judgements which is referred to as self-attribution bias. Several psychological evidences
indicate that most individuals tend to place too much credit on their abilities and knowledge
for their past success and blame external forces for their failures (Dunning, 2005).
Many factors have been identified in psychology to influence the level of overconfidence
among people. Some of these factors include complexity of the task, information and base
rate. Many empirical studies in the psychology literature posit that people generally exhibit
overconfidence when dealing with moderate or hard difficulty tasks. Thus, overconfidence is
reported to be very predominant when tasks being done are difficult and less when the tasks
are easy when the difficulty level is measured by the number of people who correctly answer
the question (Gigerenzer, Hoffrage & Kleinbolting, 1991). Also, the psychology literature
posits that overconfidence bias is predominant in males than females.
Other studies also argue that the amounts of information people possess influence their
decision-making process and as result, their confidence level. Some psychologists argue that
the large amount of information influences people’s decision-making process and therefore,
increases their overconfidence level. However, in an experiment conducted by Peterson &
Pitz (1986), using baseball wins prediction, revealed that large amounts of information to
people rather decrease overconfidence level since it increases the level of accuracy. Thus,
there is a mixed result in the psychology literature as to whether large amounts of information
increases overconfidence bias.
20
University of Ghana  http://ugspace.ug.edu.gh
Also, research has shown that the rate at which people assess their own decisions or
behaviour to some extents depends on what majority of the people perceive as norms of the
community or group the people belong to. Comparing one’s belief or decisions with the
larger community or group’s believes that an individual is a member of shows whether the
individual is consistent or inconsistent with the common norm. Consistency of an
individual’s decisions or beliefs with the larger group increases the individual’s
overconfidence level because being consistent with the community norm seems to confirm
one’s confidence. The behaviour of an individual with regard to that of the population
expressed as a percentage is called base rate in the psychology literature.
2.4.2 Overconfidence Bias in financial markets
Black (1986) indicated that irrational noise market participant’s beat the trading equilibrium
postulates by standard finance theory. Black (1986) also argued that the expectations of noise
traders deepen the challenge of forming models on the workings and behaviour of financial
markets. Also, De Long, Shleifer, Summers & Waldmann (1990) also posits that market
participants employ their beliefs of future cash flows of a security when selecting their
securities and portfolios but not the risk associated with such returns. De Long et al. (1990)
further argued that rational market participants must not compete with irrational or noise
market participants since it’s risky and costly. As a result, many researchers using noise
traders’ characteristics have developed models and empirical tested such models and
prominent among these characteristics is the overconfidence bias.
Overconfidence bias is cognitive psychology bias which is adopted into finance to help
explain how decisions of market participants are biased. Overconfidence bias occurs where
market agents’ confidence judgements are reliably stronger and greater than the correctness
21
University of Ghana  http://ugspace.ug.edu.gh
of those judgements. Overconfidence relates to economic agents who are too confident and
exaggerate their own skills and capabilities in dealing with problems (Nofsinger, 2001).
Overconfidence bias is the phenomena observed when economic agents think, believe and
behave as if they are smarter and poses superior market information than what they actually
possess. Daniel & Hirshleifer (2015) state that “Overconfidence means having mistaken
valuations and believing in them too strongly” (p. 61). Overconfidence is the propensity of
economic agents to put unwarranted level of confidence in their skills and abilities (Grinblatt
& Keloharju, 2001).
It is reported in both the psychology and finance literature that overconfidence bias is the
most robust empirical cognitive bias with its evidence cutting across several professions.
DeBont & Thaler (1995) indicated that “perhaps the most robust finding in the psychology of
judgement is that people are overconfident.” The argument that overconfidence bias is one of
the most robust finding is also confirmed by Shiller (2000) who also state that “yet some
basic tendency towards overconfidence appears to be a robust human character trait: the bias
is definitely toward overconfidence rather than underconfidence” (p. 142). Overconfidence
bias has been observed among financial practitioners. For instance, Ben-David, Graham &
Harvey (2013) documented overconfidence bias among corporate financial officers whiles
Glaser, Langer & Weber (2013) also documented overconfidence bias among investment
bankers and professional traders.
One of the remarkable questions to ask is why might financial researchers expect market
participants in financial market to exhibit overconfidence? The primary answer for this
question is that overconfidence is a widespread psychological occurrence and is
characterised by a cluster of related effects. A trader in financial market’s basic aim is to
22
University of Ghana  http://ugspace.ug.edu.gh
select securities that would provide him superior returns in future than other securities, which
is a complex activity but it is in this activity most traders depict the overconfidence
characteristics. Nofsinger (2001) states that “overconfidence causes people to overestimate
their knowledge, underestimate risks, and exaggerate their ability to control events. Does
overconfidence occur in investment decision making? Security selection is a difficult task. It
is precisely this type of task at which people exhibit the greatest overconfidence.” (Nofsinger,
2001, p. 138)
Predictability of an event which has low probability of occurrence such as securities returns
is documented to be the tasks that market participants have the highest tendency to be
overconfident since experts tend to overweight their models and theories. The psychology
literature demonstrates that overconfidence is more prone in activities which require
judgment such as making diagnosis of sickness and tasks characterised with delayed
feedback such as predicting a market returns as against activities that offer immediate and
conclusive outcome feedback and mathematical tasks such as solving calculation problems
(Griffin & Tversky, 1992).
2.4.2.1 Overconfidence bias theories in finance
Financial economists over the years have been puzzled over the eagerness of economic
agents to be involved in active trading in financial securities which is regarded to be highly
competitive. Empirical studies have shown that overconfidence in market participants’
security analysis skills is a widespread and persistent behavioural phenomenon and as a result
financial economists have postulates theories on overconfidence bias in financial markets.
Overconfidence theories in the finance literature are still evolving. Researchers in finance
23
University of Ghana  http://ugspace.ug.edu.gh
have developed overconfidence theories and provided testable implications under two main
perspectives.
Odean (1998) postulates overconfidence of market participants as overestimation of their
information precision. Odean (1998) explains that overconfident market participants
overestimate the accuracy of their information and behave to have useful signals when in fact
they have no useful information. The researcher further explains that overconfident traders
can trigger markets underrection to information resulting to positive serially correlated
returns when they underweight new information and also cause markets overreaction leading
to negatively serially correlated returns when they overweight new information. Odean
(1998) explains that the level of this underreaction or overreaction effects on the markets
depends on the proportion of all market participants who underweight or overweight their
information. However, the researcher did not distinguish between public signals and private
information signals.
As a result, Daniel, Hireshleifer & Subrahmanyam (1998) presented overconfidence bias
theory which explains overconfidence as overestimation by market participants as to the
precision of their private information than the public available information. In the world of
finance, financial analysts and other market participants gather information which they
believe to be relevant for their trading activities through various avenues like rumours and
analyses of financial statements. According to Daniel et al. (1998), when an analyst places
much weight on his ability and skills to gather information, the analysts tend to undercut their
forecast error. Therefore, the analyst becomes overconfident about signals that he has come
up with and not the public available information. Daniel et al. (1998) argues that when
market participants receive confirming private signals, their overconfidence level increases
24
University of Ghana  http://ugspace.ug.edu.gh
but receivable of disconfirming private signals trigger their overconfidence level to decrease.
Accordingly, they define overconfident investors “as one who overestimates the precision of
his private information signals, but not of information signals publicly received by all”
(Daniel et al., 1998, p.1841). Researchers have shown that varying abilities of traders for
selecting stocks might arise from overweighting of their private signals than the signals
available for the general public. This theory postulates that stock prices on competitive
securities markets overreact to private information signals and underreact to public signals.
Therefore, several empirical studies by researchers have modelled investors’ overconfidence
as an overestimation of the precision of their private information (such studies include
Hirshleifer & Luo, 2001; and Scheinkman & Xiong, 2003).
Another theory of overconfidence bias in the finance literature is built on the self-attribution
bias psychology theory. Psychologists have empirically demonstrated that people tend to
attribute past success to their abilities and failures to external variables (Hastorf, Schneider &
Polifka, 1970 (cited in Gervais & Odean, 2001)). In the financial market context, self-
attribution bias leads to overconfidence bias when successful investors ascribe their past
success to their own skills and abilities of selecting securities and failures to bad luck. This
implies that overconfidence bias resulting from self-attribution is a dynamic phenomenon
since past outcomes of investors keep on changing overtime. Hirshleifer (2001, p. 1549)
states that “Overconfidence and biased self-attribution are static and dynamic counterparts.”
Gervais & Odean (2001) used the self-attribution psychological bias theory to develop a
theory that investors infer their ability from their past successes and failures which lead them
to become overconfident. The researchers postulate that when market participants become
successful in their trades, they tend to attribute too much credit to their own skills and as a
25
University of Ghana  http://ugspace.ug.edu.gh
result upgrade their beliefs about their abilities upwards which trigger market participants’
overconfidence. Thus, the theory assumes overconfidence is a dynamic phenomenon which
changes with investors’ successes and failures. As a result, the researcher argue that market
participants who have been trading in the market for a short period are characterised by
greater overconfidence compared to more experience market actors who have developed
better self-assessments. Gervais & Odean (2001) further argue that overconfidence bias of
investors resulting from their past positive trading returns causes investors to trade more in
subsequent periods. Also, several empirical studies have modelled investors’ overconfidence
as resulting from past market performance leading to excessive trading (such studies include
Statman, Thorley & Vorkink, 2006; and Chuang & Lee, 2006).
2.4.2.2 Implications of Overconfidence bias on financial markets
The theoretical explanations of overconfidence bias as discussed above have been
documented in the literature to have various implications on financial markets around the
globe. It is important indicating that most of these implications have been tested by
researchers on different markets.
2.4.2.2.1 Overconfidence bias and Trading Volume
Researchers over the years have argued that information is gradually introduced to the
security market through market participants trading (Benos, 1998). However, past studies do
not specify the kind of information impeded in trading volume as well as the content and
source of the information. As a result, a number of recent empirical studies postulate that
overconfidence of traders trigger the excessive trading volumes witnessed in the global
financial markets (Odean, 1998; Daniel et al., 1998; Gervais & Odean, 2001; Statman et al.,
2006). These studies explain that the relationship between overconfidence bias and trading
26
University of Ghana  http://ugspace.ug.edu.gh
volume exists predominately in the market since market participants who tend to be
overconfident overestimate their own security selecting, signals and valuation skills. As a
result, trade more on the average than the rational market participants. It was argued that
overconfidence arises from traders’ overestimation of their private information and as a result
triggers traders to hold different beliefs which also lead to overestimation of their belief
systems. Therefore, it is argued that strong opposing belief systems strongly held in the
markets are necessary and sufficient grounds for high trading volumes to be witnessed.
Overconfidence of market players is reported in the literature to explain excessive trading
volumes being observed on financial markets. DeBont & Thaler (1995) note that “the key
behavioral factor needed to understand the trading puzzle is overconfidence” (p. 393). Thus,
investors who exhibit overconfidence characteristics usually end up doing too much trading.
According to Shiller (2000), “overconfidence, however generated, appears to be a
fundamental factor promoting the high volume of trade we observe in speculative markets”
(Shiller, 2000, p. 144). It is reported that overconfident market actor trade excessively in such
a way that their transection cost surpasses their expected gains. The concept that
overconfidence bias of market participants causes them to trade excessively is tested by many
studies such as Statman et al. (2006) and Chuang & Lee (2006). These studies confirmed the
assertion that overconfident invertors trade too much.
2.4.2.2.2 Overconfidence bias and Market returns
Overconfidence bias and market return nexus has been documented in the behavioural
finance literature. In the literature, it is argued that market returns have an impact on the level
of investors’ overconfidence. Thus, investors draw their level of confidence from past market
results.
27
University of Ghana  http://ugspace.ug.edu.gh
According Gervias & Odean (2001), market participants who make profitable trades tend to
be more overconfident due to the better than average overconfidence resulting from self-
attribution bias. As a result, they developed a model to assess how market participants learn
about their abilities. They indicated that the overconfidence level of market participants
changes and can be attributed to their previous success and failure. For instance, in a bull
market, market actors easily make gains and as a result, tend to be more overconfident. They
associate the success to their own expertise ignoring the fact that the gains they are enjoying
are total market phenomena at that point in time. More specifically, the literature explains
that aggregate overconfidence is higher if market returns are positive. In the same way, the
overconfidence level decreases during a negative market returns as investors blame external
factors for their failure.
According to Glaser & Weber (2009), both previous market returns and also previous
portfolio returns have significant impacts on an investor’s behaviour but the past market
returns has superior impact on market participants overconfidence. Also, Statman et al.
(2006) and Chuang & Lee (2006) indicate that past market returns have strong and significant
impact on market participants’ overconfidence.
Daniel & Hirshleifer (2015) argued that overconfidence plays a vital role of understanding
return anomalies observed on financial markets which are normally attributed to market
participant’s limited attention. They claim that the assertion of limited market participants’
attention has enormous impacts on price if the market participants are overconfident and as a
matter of fact do not realise that the importance of the information they are neglecting.
28
University of Ghana  http://ugspace.ug.edu.gh
2.4.2.2.3 Overconfidence and differential reactions to private and public information
Overconfidence bias theory such as Daniel et al. (1998) postulates that overconfident market
participants overreact to their private information whiles underreacting to public available
information. Daniel et al. (1998) argues that overconfident investors place more weight on
their private information and as a result overreact to arrival of private information signals. On
the other hand, market participants place less weight on public information and as a result
underreact to public information. Therefore, it is argued that a shock of private and public
information drives stock returns and trading volume. Thus, researchers postulate that
overconfident investors put too much weight on their private signals which causes stock price
to overreact.
One strand of the literature explains that private information plays more significant role in
driving trading volume than public signals (Liorente, Michaely, Saar & Wang, 2002). On the
other hand, some researchers explain that market participants’ various interpretations and
analysis of public signals rather trigger trading volumes. However, the overconfidence theory
posits that overconfidence market participants overweight the precision of their private
information and underweight public signals and a result trade excessively (Odean, 1998;
Daniel et al., 1998). Chuang & Lee (2006) tested this overconfidence implication and
confirmed its validity.
2.4.2.2.4 Overconfidence bias and excessive volatility
Quite a number of empirical studies have documented that volatility of stock prices are more
volatile than an efficient market theory postulates (such studies includes Shiller, 1981). One
of the common explanations given to this puzzle is the habit formation model. The habit
formation model explains that investors’ consumption variations relative to their habit lead to
29
University of Ghana  http://ugspace.ug.edu.gh
variation in the risk aversion of investors and therefore leads to changes in price-to-dividend
ratios. This changes aid to decrease the gap between the dividend growth volatility and
returns volatility.
However, overconfidence bias is suggested as an important contributing factor responsible
for the excessive volatility observed on financial markets. For instance, Chuang & Lee (2006)
formulated a model to test whether excessive trading of overconfidence market participants
contribute to the documented excessive volatility. They reported that overconfident market
participants’ excessive trading behaviour contributes to the documented excessive volatility.
It is worth pointing out that, the overconfidence literature does not argue that overconfident
market participants excessive trading is the only source of excessive volatility observed on
financial markets.
2.4.2.2.5 Causality relations between trading volume and stock returns
The causal relationship between trading volume and stock returns have been documented in
the literature to explain several phenomena on financial markets. Hiemstra & Jones (1994)
postulate that stock returns and trading volume causal relationship provides an insight into
changes in stock prices in relation to movements in trading volume. In the literature, it is
argued that past trading volume cannot trigger current securities price in an efficient market.
However, causality running form securities prices to current trading volume is explained as
market participants basing on past securities prices behaviour to predict the future prices
behaviour of the securities (Brennan & Cao, 1997). The causality relation between trading
volume and returns have received much attention in recent times and Chen, Firth & Rui
(2001) state that “the important issue should be whether information about trading is useful in
improving forecasts of price changes and return volatility” (p. 155).
30
University of Ghana  http://ugspace.ug.edu.gh
Moreover, other explanation to the price – causality is the sequential arrival information
hypothesis postulated by Copeland (1976) and Jennings et al. (1981) later extended it. The
sequential arrival information theory suggests a positive bidirectional causal relationship
between absolute values of price changes and trading volume. Sequential arrival information
framework assumes that the release of new information into the market is not disseminated to
all market agents simultaneously but the new information is release to one agent at a time. As
a result, the final information equilibrium is reached only after sequences of transitional
equilibriums have been achieved. Consequently, lagged absolute returns may trigger current
trading volume and vice versa due to sequential flow of information. For instance, Hiemstra
& Jones (1994) claimed that sequential information flow leads to lagged trading volume to
trigger current absolute price changes and lagged absolute price changes having predictive
power for current volume.
Lakonishok & Smidth (1989) also demonstrated that current volume can have relation to past
stock price changes due to tax – and non – tax associated trading motivations. For that reason
the causal relation for tax – associated trading motivations is negative whiles non – tax
associated trading motivations are positive.
Furthermore, the price and volume causality relationship can be explained by the noise –
trader model proposed by De Long et al. (1990). The noise – trader model argues that the
market activities of noise traders which deviate from the basic economic fundamentals
usually leads to stock prices mispricing in the short run. According to De Long et al. (1990),
in the absence of transitory component, the price moves to its mean value in the long run.
Thus, in the noise trader model, positive causal nexus running from stock returns to trading
31
University of Ghana  http://ugspace.ug.edu.gh
volume follows the positive feedback trading strategy of noise traders who usually rely on
past price changes to make their decisions.
Statman et al. (2006) argues that little past empirical studies have considered current volume
and lagged returns relationship in the literature. They further argue that these empirical
findings fail to recognize that overconfidence bias of market participants may also be the
reason for causality running from past returns to current volume. One of the implications of
overconfidence hypothesis is the relation between current trading volume and past stock
returns. A causality running from lagged market returns to current trading volume is
explained to be an evidence of overconfidence bias. Researchers argue that market
participants draw their source of overconfidence from previous market returns and as a result
trade more in subsequent periods. This argument is consistent with Daniel et al. (1998) and
Gervais & Odean (2001) overconfidence hypothesis that overconfidence is keenly promoted
in a bull market. Based on this argument, Chuang & Lee (2006) developed a bivariate
Granger causality tests framework to differentiate between the overconfidence bias and other
trading volume hypothesis. Based on their model, they concluded that granger-causality
running from lagged market returns to current trading volume is an implication of
overconfidence bias on financial markets.
2.5 Review of Previous Studies on Overconfidence bias in financial markets
Behavioural finance and for that matter overconfidence bias in finance is a new area
emerging in the literature and for that matter not much empirical evidences have been
documented on overconfidence bias. However, there are some evidences of overconfidence
bias from both advanced and developing markets.
32
University of Ghana  http://ugspace.ug.edu.gh
2.5.1 Evidence from Developed Markets
Odean (1998) argued that market participants are overconfident and that overconfidence bias
affects financial markets activities. The study indicated that the how is dependent on which
individuals are overconfident in the market and moreover how information in the market is
disseminated. The study examined markets where the market agents are price-taking traders,
a strategic-trading insider and risk-averse traders. As a result, the study employed three
different kinds of empirical models to achieve this multidimensional analysis of
overconfidence.
Based on the results of the analysis, Odean (1998) argued that when market participants are
price takers, insiders or market makers then their overconfidence would result to increase in
trading volumes. The study claimed that this finding is the most robust empirical evidence of
effect of market participants’ overconfidence. Thus, when overconfidence of price takers
increases, price takers measure their own signals greatly than they weight those of other price
takers in the process of forming their beliefs. Therefore, their subsequent beliefs become
more dispersed and a matter of fact more trading takes place. Also, when an insider investor
becomes overconfident, the insider investor believes and places more weight on the private
information in his possession and has more precision than the reality is and forms a posterior
expectation farther from optimal which leads the insider to trade more aggressively leading to
increase in expected volume. Odean (1998) indicated that overconfident market markers
usually have flatter supply curve and this motivates market makers who are price sensitive to
trade more. This implies that, with all the three settings of market participants,
overconfidence bias of market participants result in high volumes of financial securities being
traded.
33
University of Ghana  http://ugspace.ug.edu.gh
Empirically, Odean (1998) revealed that overconfident market participants can trigger
markets underreaction to information which would lead to returns of the market being
positive correlated. On the other hand, overconfident market participants can cause market to
overreact to information which basically leads to negative serially correlated returns.
Furthermore, the study argued that the level of the under reaction or over reaction rests on the
proportion of market participants who actually underestimate or overestimate their own
information. However, this result is consistent with the concept in the psychology literature
that individuals underestimate the difficulty when dealing with highly statistical and abstract
information and rather places more weight on salient and extreme information.
Odean (1998) empirically proved that traders expected utility reduces as a result of
overconfidence. For instance, in the world of price taking traders, there are no exploitation
because there is the absence of noise traders and hence the total market returns expected from
the market activities must be zero. The study argued that in settings where overconfidence
arises from traders’ overestimation of their private information triggers traders to hold
differing beliefs which also result in overestimation of their belief systems. Therefore, the
study argued that strong opposing belief systems strongly held in the markets are necessary
and sufficient grounds for high trading volumes to be witnessed, high returns volatility in the
market to occur, deviation from intrinsic securities prices to be witnessed and reduction of
participants expected utility. Thus, overconfidence of market participants would reduce their
expected utilities since it leads to non-optimal risk sharing and as a result overconfident
market actors are in a position to hold not well diversified portfolios.
The study further reported that overconfidence causes market depth to increase. Also, insider
traders who are overconfident improve financial assets quality price whereas overconfident
34
University of Ghana  http://ugspace.ug.edu.gh
investors who are price takers worsen it. The study also revealed that overconfidence has an
effect on volatility and this is also dependent on who is overconfident. This study actually
prepared the grounds for much more empirical studies on overconfidence to be conducted
and provided several testable implications of overconfidence.
Daniel et al. (1998) argued that some empirical studies reported in the literature indicate that
markets underreact to information whiles others report overreaction evidence but there was
no theory in the literature to incorporate these empirical evidences and further make
suggestions as to when market overreaction or underreaction would occur. Therefore, Daniel
et al. (1998) fill this gap in the literature by postulating that overconfidence of market
participants result from market participants’ overreaction to private information while
underreacting to available publicly information.
Their theory was founded on the grounds that a significant amount of errors made by market
participants is their misinterpretation of their own new private information. In their results,
they indicated that informed traders who are overconfident end up losing money.
Furthermore, the study reported that overconfidence of risk-averse traders with regards to
genuine information permits them to make most use of the information and as a result their
expected returns can exceed that of fully rational traders.
With their model having its foundation on traders’ overconfidence bias resulting from traders
becoming more confident about their private information, Daniel et al. (1998) argued that the
predictability of returns would be highly great and consistent for securities having the
greatest information asymmetries. Daniel et al. (1998) further demonstrated that the presence
of investors’ overconfidence in the markets may be the reason for the observed negative long-
35
University of Ghana  http://ugspace.ug.edu.gh
lag returns autocorrelations and excessive volatility observed in some financial markets.
Daniel et al. (1998) study provided breaking grounds for more studies on overconfidence bias
in finance to thrive on.
Odean (1999) argued that since there is no model that establishes the required and good level
of trading volume to be observed in financial markets, it becomes extremely challenging to
investigate whether trading volumes observed in financial markets are too high. The study
then argued that if aggregate market trading volume is high, then the trading volumes of some
group of market participants must be high. Therefore, the study employs discount brokerage
accounts class of investors and postulates that due to the overconfidence of the discount
brokerage investors, they trade too much. The study indicated that the use of discount
brokerage investors trading activities is very good for investigating overconfidence bias of
excessive trading since the discount brokerage is not characterised by complicated agency
problems.
Odean (1999) reported that some of the market participants trade too much in such a way that
on the average their overall trade gains are reduced. The study demonstrated that after
excluding market activities that may have been inspired by investors’ demands for liquidity,
tax loss selling, rebalancing of portfolio or changing to securities with lower risk, too much
trading still reduces returns. The findings of the study also revealed that overconfident
investors even trade more when the cost of trading is ignored and this actually reduces their
profits and as a result even claim that this empirical finding is more extreme compared with
what overconfidence theory predicts.
36
University of Ghana  http://ugspace.ug.edu.gh
Similarly, the study investigated the return behaviour of these investors before and after they
have made a security purchase or sale. Odean (1999) indicated that most of these investors
purchase financial securities that over the past six or more months have decreased or
increased in value. Also, the study reported that, on the average, these investors tend to sell
financial securities that have increased swiftly in recent weeks and claim that these observed
patterns are not easy to comprehend. The study suggested that the possible reasons for these
observed patterns may be due to the large number of financial assets available for investors’
choices or the influence of external sources such as financial media or may be due to the
disposition effect of the investors.
In addition, Barber & Odean (2001) indicated that psychological research has demonstrated
that men have the higher tendency to be overconfident compared to women especially in
male-dominated areas like finance. It is also documented in finance literature that investors
who exhibit overconfidence usually overestimate the precision of their private information
and as a result trade excessively. Inferring from these two statements, Barber & Odean (2001)
investigate the hypothesis that men are more overconfident compared to women and as a
result trade more and perform worse compared to women. In order to achieve this purpose,
they employed 35,000 households’ data on both men and women on their investment
behaviour towards common stocks from a large discount brokerage firm which spans from
1991 to 1997.
The results of the study provided a strong empirical evidence for the hypothesis that men
truly trade more compared to women and as a result decrease men returns more as compared
to women returns decrease. They indicated that these empirical evidences are more robust
and more evident between single men and women. They reported that on the average men
37
University of Ghana  http://ugspace.ug.edu.gh
tend to trade approximately 45% more than women and this behaviour results in decrease of
men’s net returns by approximately 2.7% annually compared to 1.7% for women. Barber &
Odean (2001) therefore concluded that overconfidence among men is higher than among
women.
Gervais & Odean (2001) also developed a multi-period market model which describes both
the steps and channel through which market participants get to know about their abilities and
how a bias in such instances lead to overconfidence. In their model, market participants are
assumed firstly to not be aware of their own trading abilities and that market participants
deduce their abilities from their previous trading successes and failures. That is, in the
process of market participants assessing their abilities and competence level, they end up
placing more credit of their successes to their own skills and as result becomes overconfident.
Gervais & Odean (2001) argue that investors’ overconfidence is a dynamic phenomenon and
therefore changes with trading successes and failures of traders. The study indicated
overconfidence bias is more evident in traders’ learning stages of their trading profession and
as traders mature in their career they develop a better assessment of their own abilities. One
of their major hypotheses was that they expect aggregate overconfidence to be higher
whenever the market experience gains, and lower after market losses. Therefore, greater
overconfidence results in greater trading volumes and market losses leads to lower trading
volume.
Gervais & Odean (2001) further argued that since traders become overconfident based on
their successes, overconfidence does not lead to traders becoming wealthy but rather the
processes involved in becoming wealthy make traders overconfident. Basing on their wealth,
traders who exhibit overconfidence bias are not in danger of exiting from the marketplace,
38
University of Ghana  http://ugspace.ug.edu.gh
however, as they become more experienced the trader’s level of wealth and confidence
diminishes. They further indicated that in a market where inexperienced market participants
are born always into the market and the experienced ones fade out, overconfidence would
persist in the markets and therefore overconfidence market participants can play a significant
part in the behaviour of the global financial markets.
The findings of the study indicated that most investors tend to take too much credit for their
own successes which leads to overconfidence. The results also reported that overconfident
traders trade too aggressively which leads to excessive trading volumes. The study revealed
that traders’ level of learning bias increases the trading volume and returns volatility
observed in financial markets. The study empirically documented that overconfident traders
exhibit suboptimal behaviour in their trading and as a result lowering expected future gains.
Put differently, a more experienced and successful market participant has lower expected
returns compared to a less successful trader. The study also showed that a high level of
overconfidence triggers greater trading volumes and that trading volume would be greater
after market gains. However, it is worth pointing out that unlike De Long, Shleifer, Summers
& Waldmann (1991) model in which bias traders survive by locking in greater profits,
Gervais & Odean (2001) model postulate a market in which the overconfident trader earns on
average lower profits.
Biais, Hilton, Mazurier & Pouget (2005) experimentally investigated the impacts of
psychological factors on market participants’ financial behaviour by focusing on
miscalibration and self-monitoring. Biais et al. (2005) explained that miscalibration as a form
of judgemental overconfidence resulting from participants placing more weight on the
precision of their information and self-monitoring as a form of investors’ attentiveness to its
39
University of Ghana  http://ugspace.ug.edu.gh
environment. Employing questionnaires and 245 respondents, they investigate these two
factors by studying the behaviour of these respondents under experimental financial market
settings. The study postulated two hypotheses. Firstly, Biais et al. (2005) anticipated
miscalibrated traders to underestimate the uncertainty of securities price in such a way that
they are exposed to the winner’s curse. Secondly, Biais et al. (2005) anticipated high self-
monitors to act tactically and achieve superior results.
The study reported that miscalibration judgemental overconfidence leads to low trading
performance and as a result investors’ overconfidence results to lower gains. On the other
hand, the study indicated that self-monitoring enhances trading results and high self-monitors
have high probability of escaping from the winners’ curse trap. Therefore, it is possible to
make an inference that this experimental market study employed the winner’s curse traps to
be the channel through which the impacts of miscalibration and self-monitoring can be
measured.
Consistent with the findings of Barber & Odean (2001), Biais et al. (2005) also reported that
men tend to trade more compared to women. However, while Barber & Odean (2001) employ
gender as proxy for overconfidence, they reported that there exists no correlation between
miscalibration judgemental overconfidence and gender. However, their results also revealed
that miscalibration overconfidence in men worsen their trading performance while it does not
influence the trading performance of women.
Statman et al. (2006) argued that recently investors overconfidence theories such as Gervias
& Odean (2001) paper provides the basic framework that calls for an empirical investigation
to ascertain the validity of their claim that some market participants become more
40
University of Ghana  http://ugspace.ug.edu.gh
overconfident resulting from observed positive portfolio returns and exhibit low level of
overconfidence following negative portfolio returns by employing both market-wide and
security-specific data. The study employed from 1962 to 2002 monthly observations time
series data on all NYSE/AMEX common stocks. Statman et al. (2006) adopted VAR model
and its associated impulse response functions to investigate the lead-lag relationship between
turnover and market returns.
One of their major findings is that, there exists a statistically significant evidence for market-
wide trading volume (used turnover as proxy for trading volume) to upsurge in periods
following high positive market returns after introducing volatility and portfolio rebalancing
as control variables. Statman et al. (2006) claim that the market-wide evidence documented
can also be attributed to the manifestation of disposition effect of investors. Therefore, in
order to substantiate their claim, they employ this VAR and impulse function models to
individual stocks with the intention of empirically differentiating between overconfidence
bias and disposition effect. The study also reported that a security turnover is significantly
explained by previous market returns after the introduction of previous security returns and
indicted this evidence buttresses the overconfidence bias hypothesis. However, Statman et al.
(2006) stated that their empirical evidence is more prominent in small-cap stocks.
The findings of the study indicated a lead-lag nexus between market-wide turnover and
market-wide returns which confirms Gervais & Odean (2001) theory on investors’
overconfidence. This finding indicates the economic importance of high positive or negative
returns on subsequent market trading volume as well as the behaviour of the market. Statman
et al. (2006) interpreted this lead-lag relationship as empirical evidence of overconfidence
bias hypothesis. Consistent with the financial economics literature, the study also
41
University of Ghana  http://ugspace.ug.edu.gh
documented a contemporaneous positive nexus between trading volume and returns
volatility. It is important to say this study provided the methodological grounds for many
studies on overconfidence bias to be conducted.
Chuang & Lee (2006) also conducted a study on the empirical implications of overconfidence
bias hypothesis. In the study, the authors formulated four main empirical frameworks in an
attempt to provide empirical evidences of the validity of the overconfidence hypothesis as
reported in the literature with specific attention to an aggregate behaviour of overconfident
investors. Firstly, Chuang & Lee (2006) hypothesize that investors who exhibit
overconfidence bias tend to overreact to their own private information and underreact to
public available for all information. Also, overconfidence bias of investors arising from high
past market returns triggers more aggressive trade by these overconfidence investors in
subsequent periods. Thirdly, Chuang & Lee (2006) examine whether overconfidence bias of
investors has significant effects on the observed market dynamic volatility. Lastly, they
hypothesize investors overconfidence would lead investors to underestimate risk and as a
result trade more in financial assets that are riskier.
In order to examine the shock of private and public information on securities prices, Chuang
& Lee (2006) employed a restricted bivariate moving average representation model. They
claim that “with an initial underreaction, stock prices overreact to private information
followed by a correction process, and underreact to public information reaching an
equilibrium response without a significant long-run reversal” (Chuang & Lee, 2006, p. 2503).
This evidence they documented provides an empirical support for their first framework of
overreaction to private information and underreaction to public information by overconfident
42
University of Ghana  http://ugspace.ug.edu.gh
investors. This evidence they documented also confirms the overconfidence bias arguments
of Daniel et al. (1998).
In order to investigate the dynamic relationship between stock returns and trading volume as
postulated by Gervais & Odean (2001) hypothesis, they employed granger causality tests if
causality runs from market returns to trading volume. The results of the study indicated that
there exists a significant causality running from returns to trading volume. This implies that
low (high) market returns of securities granger cause low (high) trading volume in the
market. They also indicated that this positive dynamic nexus documented between returns
and trading volume is robust even for sub-sample periods.
With respect to investors asymmetric trading behaviour the study provided the evidence that
in a bull market, investors trade more aggressively compared to a bear market. Employing
GARCH-type specifications, they concluded that market participants become overconfident
and as a result excessively trade. It is worth pointing out that this empirical finding provides
evidence to support Gervais & Odean (2001) overconfidence hypothesis.
In order to ascertain the validity of the claim that overconfidence investors’ excessive trading
contributes to the documented high market volatility, they decomposed the market trading
volume into two components. Firstly, the overconfidence component which is trading volume
relating to past returns and non-overconfidence components which is trading volume
unrelated to past returns. Chuang & Lee (2006) then examine the nexus between these two
components and volatility of stock returns. They found that the overconfidence component
positively accounted for conditional volatility and concluded that this finding is empirical
43
University of Ghana  http://ugspace.ug.edu.gh
evidence supporting the hypothesis that overconfidence market participants’ excessive
trading contributes to observed excessive volatility in the financial market.
Also, in order to test the validity of the hypothesis that overconfident market participants
underestimate their risk and trade more in riskier assets, using firm-specific risk and return
volatility as proxy for risk, they formulate two portfolios which are based on the degree of
each risk. The study then employs seemingly unrelated regression (SUR) to test this
hypothesis. The findings of the study indicated that stocks with higher risk had strong high-
volume impact which confirms the hypothesis that if market participants are overconfident,
they deal in riskier financial assets.
Overall, this study makes available intensive and extensive empirical evidences for the
manifestations of overconfidence bias in financial markets. This study has provided one of
the comprehensive empirical evidences of the numerous extrapolations of the overconfidence
bias hypothesis and demonstrated that the overconfidence hypothesis remains one of the best
possible viable explanations of many observed anomalous in financial markets.
2.5.2 Evidence from Emerging Markets
Zaiane & Abaoub (2009) following the work of Statman et al. (2006), they investigated the
presence of overconfidence bias in the Tunisian market employing the VAR model.
However, Zaiane & Abaoub (2009) found evidence for the overconfidence hypothesis but
indicated it is remarkably weak. That is, previous market returns influence current trading
activities of market participants in subsequent months. That is, the paper provides empirical
evidence from the Tunisian stock market on overconfidence bias of market participants’
44
University of Ghana  http://ugspace.ug.edu.gh
impacts on trading volume. Also, consistent with the literature, Zaiane & Abaoub (2009)
reported contemporaneous positive nexus between trading volume and returns volatility.
Muhammad & Khalid (2012) also tested two implications of overconfidence bias in Pakistan.
Thus, the study examined the nexus between overconfidence and trading volume as well as
overconfidence and volatility on Karachi Stock Exchange, Pakistan. With respect to the
nexus between trading volume and overconfidence bias, Muhammad & Khalid (2012)
followed Statman et al. (2006) paper; and with respect to trading volume and returns
volatility, Muhammad & Khalid (2012) also followed Chuang & Lee (2006) paper.
Consistent with Statman et al. (2006) overconfidence hypothesis predictions, Muhammad &
Khalid (2012) report a significant evidence of past market returns relationship with current
trading volume after controlling for returns dispersion and returns volatility. However,
Muhammad & Khalid (2012) indicated that there is no significant positive contribution of
overconfidence to conditional returns volatility.
Metwally & Darwish (2015) employing monthly data from 2002 to 2012 also investigated the
evidence of overconfidence bias on Egypt stock markets by following Statman et al. (2006).
Metwally & Darwish (2015) also report a significant empirical evidence of the impacts of
past market returns on current trading volume (used turnover as proxy). Based on this
finding, they concluded that there is an empirical evidence of investors’ overconfidence bias
on Egypt stock market. Metwally & Darwish (2015) also indicated that the market states at a
particular point in time are strong and significantly affect the trading activities on the
Egyptian stock market especially in a bull market.
45
University of Ghana  http://ugspace.ug.edu.gh
CHAPTER THREE
RESEARCH METHODOLOGY
2.1 Introduction
In the preceding chapter, a comprehensive review of existing theory and the literature on
overconfidence was presented. This served as a guide to the researcher on the choice of the
estimation techniques to be employed in analysing the data. Explaining the methodology for
the study is the focus of this chapter. In this chapter, data sources, variables used in the study,
and empirical methodologies used for this study are discussed.
3.2 Data Sources
In order to achieve the objectives of this study, secondary data (monthly GSE-index value,
monthly trading volume, monthly consumer price index and 91-days treasury bill rate) was
obtained from the Ghana Stock Exchange and the Bank of Ghana for the analysis. The GSE –
index value and trading volume was obtained from Ghana Stock Exchange database and the
macroeconomic variables were also obtained from the Bank of Ghana (BOG) time series
database.
The data used for the analysis is monthly based observations, specifically within month daily
GSE – index value data is used to compute the monthly volatility. Approximately 3,440 daily
index observations of GSE – index value and 192 monthly observations for all the variables
are used for this study.
46
University of Ghana  http://ugspace.ug.edu.gh
3.3 Sample period
The operation of GSE began from 1990 but this study focuses on a sample period spanning
from January, 2000 to December, 2015 as a result of the inability of the researcher to obtain
the needed data on trading volumes for the early periods. In order to achieve reliable results,
the sample period from 2000 to 2015 and sub-sample financial crisis period (January, 2008 to
December, 2010) were considered. The financial crisis sub-sample period were considered
due to the volatility clusters (See Figure 3.1 and Figure 3.2) that manifests in monthly GSE –
index returns ( ) and monthly trading volume ( ) which may be due to the 2008
global financial crises. Thus, in order to remove the possible impacts of the 2008 global
financial crises from the model, a dummy variable, 08 that takes 1 for January, 2008 to
December, 2010 (otherwise, 0) is introduced.
Figure 3.1: A Graph of GSE-index Monthly Returns (MRET) on GSE from 2000 – 2015
MRET
.3
.2
.1
.0
-.1
-.2
-.3
-.4
00 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15
Source: Author’s own computation based on data from GSE database from 2000 – 2015
Figure 3.2: A Graph of Monthly Trading Volumes (MVOL) on GSE from 2000 – 2015
MVOL
350,000,000
300,000,000
250,000,000
200,000,000
150,000,000
100,000,000
50,000,000
0
00 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15
Source: Author’s own computation based on data from GSE database from 2000 – 2015
47
University of Ghana  http://ugspace.ug.edu.gh
3.4 Market Activities of Ghana Stock Exchange from 2000 – 2015
The GSE started operations with 11 listed companies and with three brokerage firms. Over
the years, many companies have been listed on the exchange even though the number of
listed companies is still recognized to be small with regards to other African equity markets
such as Nigeria, Kenya, Egypt, Morocco and South Africa. The size of listed equities on the
exchange grew from 11 to 29 listed equities as at 2009; 35 equities in 2012 were listed; and
as at December 2015, 38 equities were listed on the exchange. Listed companies are from the
financial sector, distribution sector, the food and beverage sector, ICT, insurance sector,
agriculture, education, manufacturing and mining sectors of the economy.
From its inception in 1990, GSE had only three (3) Licenced Dealing Members but as at the
end of 2015, the GSE could boast of twenty-one (21) Licensed Dealing Members. It is worth
pointing out that GSE operate on fully automation trading platform. GSE market opens each
working day from Monday to Friday with official market activities taking place from
10:00am to 15:00pm except holidays pronounced by the authorities of the exchange in
advance. The types of securities traded on GSE are “shares (preference or equities); Debt in
the form of corporate bonds (and notes), municipal bonds (and notes), government bonds
(and notes); and Close-end unit trusts and mutual funds” (GSE, 2016). GSE has also
established Ghana Alternative Market with the objective of assisting SMEs to gain access to
the capital market which had four listed SMEs as at December 2015. Moreover, the exchange
has also established Ghana Fixed Income Market with the primary aim of providing
dedicated market for fixed income securities.
At the end of 2015, fifteen (15) corporate bonds and ninety-seven (79) Government bonds
were traded on the Exchange (GSE, 2016). According to the Ghana Stock Exchange 25th
48
University of Ghana  http://ugspace.ug.edu.gh
Anniversary Public Lecture held on 11th January 2016, a total amount of GH¢ 2.1 billion has
been realised as an equity finance through the GSE. Prior to 2011, the GSE All-share index
was the main stock index of GSE with base index value of 10,000. In January 2011, GSE
replaced GSE-ASI by Ghana Stock Exchange Composite Index (GSE-CI) with base index
value of 1,000. However, some brokers have established their own index, for instance
Databank Stock Index (DSI) and CAL Brokers Limited (CBL) All-Share Index.
The equities market capitalisation of the GSE since 2000, has increased tremendously from
GH¢365.5 million in 2000 to a value of GH¢57,116.87 million as at December 2015.
However, total market capitalization went down by 11.24% from December 2014 value of
GH¢64,352.42 million to December, 2015 value of GH¢ 57,116.87 million. A summary
report on the yearly market capitalization from 2000 to 2015 is depicted in Figure 3.3.
Figure 3.3: A Graph of Yearly Market Capitalization of GSE from 2000 – 2015
Market Capitalization - Equities (GH¢ million)
70,000
60,000
50,000
40,000
30,000
20,000
10,000
0
00 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15
Source: GSE Annual report, 2016
49
University of Ghana  http://ugspace.ug.edu.gh
Yearly volumes traded have not been encouraging even though volumes traded have
increased significantly from 222,000 to 2,464,283,600 as at December 2015. Summary
reports on volume traded since the year 2000 is depicted in Figure 3.4.
Figure 3.4: A Graph of Yearly Market Trading Volumes of GSE from 2000 – 2015
Volume Traded - (000)
600,000
500,000
400,000
300,000
200,000
100,000
0
00 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15
Source: GSE Annual report, 2016
The Exchange has extremely performed1 5w5e%ll in9 1te%rms of 7c9a%pital gains. For instance, theExchange recorded annual returns of , and in 2003, 2004 and 2013
respectively. However, since 2000, the index has documented four negative annual returns
(Specifically in 2005, 2009, 2011 and 2015). It can be observed that the market’s worst
performance since 2000 was recorded in the year 2009 (−47%) which can be attributed to
the 2008 financial crises. The GSE – index annual returns since 2000 are depicted in Figure
3.5.
50
University of Ghana  http://ugspace.ug.edu.gh
Figure 3.5: A Graph of GSE – Index Annual Returns from 2000 – 2015
GSE Returns (2000-2015)
155%
91%
79%
58%
46%
31% 32%
24%
17%
11%
5% 5%
-3%
-12%
-30%
-47%
00 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15
Source: GSE Annual report, 2016
3.5 Variables
The primary objective of this study is to test the rationality of market participants on GSE by
investigating for an evidence of overconfidence behavioural bias on GSE. The selection of
the variables is influenced by previous studies on overconfidence bias, stock returns, trading
volumes, returns volatility and macroeconomic variables relationships. The variables used to
investigate the overconfidence bias include endogenous variables and exogenous variables
(control variables). Moreover, all the variables used in this study are converted into their
natural logarithm form for the analysis and this is adopted because the operation of taking a
log constitutes a non-linear transformation (Brooks, 2008).
According to Brooks (2008), the employment of non-stationary data can lead to unreliable
estimation and spurious correlation and regression. In order to check for the stationarity of
51
University of Ghana  http://ugspace.ug.edu.gh
the endogenous and control variables, Augmented Dickey-Fuller (ADF) test, Phillips-Perron
(PP) test and Kwiatkowski, Phillips, Schmidt, & Shin (KPSS) test are employed. All
variables is tested for stationarity at level and if not stationary then the first difference is
considered under assumption of intercept and trend, without trend but intercept and with
neither intercept nor trend.
3.5.1 Endogenous Variables
In this study, monthly market-wide detrended logged trading volume (proxy for trading
activities) and monthly GSE – index returns (proxy for market returns) are the endogenous
variable for the analysis. These variables are used due to Gervais & Odean (2001), Statman et
al. (2006) and Chuang & Lee (2006) empirical arguments that overconfidence bias arises
from previous markets returns and leads to market participants trading more or less in
successive periods.
3.5.1.1 Monthly Detrended Logged Trading Volume (MDLTVOL)
In the literature, several researchers have used different indicators in shares, value and
turnover ratios as proxy for trading volume. Among these are: Aggregate Share Volume
(Hiemstra & Jones, 1994); Individual Share Volume (Epps & Epps, 1976); Individual Dollar
Volume (Lakonishok & Vermaelen, 1986); Relative Individual Dollar Volume (Tkac, 1996),
Individual Turnover (Stickel & Verrecchia, 1994); Aggregate Turnover (Statman et al.,
2006); and Total Number of Trades (Conrad, Hameed & Niden, 1994).
In this study, logged aggregate share trading volume is used as an indicator for trading
activities since it is documented in the literature that logged trading volume removes the
influence of growth (Statman et al., 2006). The availability of only monthly data of trading
52
University of Ghana  http://ugspace.ug.edu.gh
volume in shares to the researcher also necessitate the use of trading volume in shares as
proxy for trading activities on Ghana stock market. The monthly log trading volume
( ) is non-stationary at level (see Appendix A) and so following Statman et al. (2006)
and Zaiane & Abaoub (2009), the log trading volume is detrened.
In addition to non-stationary of the log trading volume, several studies have provided
empirical evidence of strong linear and non-linear time trend in times series of trading
volume (Gallent, Rossi & Tauchen, 1992; Chen et al., 2001). According to Statman et al.
(2006), the linear time trend detrending methodologies seems not flexible enough for trading
volume time series so following Statman et al. (2006) and Zaiane & Abaoub (2009), Hodrick-
Prescott (1997) algorithm (HP) is employed to detrend the log trading volume. Statman et al.
(2006) explains that the HP detrending of log trading volume helps to eliminate the
correlation between the level of the trend and volatility around the trend.
The HP filter is a smoothing method that is widely employed by macroeconomists to obtain a
smooth estimate of the long-term component of a series. The HP filter was firstly introduced
by Hodrick and Prescott in their working paper on the post-war U.S business cycles in the
early 1980’s which was later published in 1997. The HP filter methodology of detrending
series is two-sided linear filter that smooth a series by minimizing the variance of the series.
For instance, the HP computes the smooth series monthly detrended logged trading volume
( ) of the monthly logged trading volume ( ) series by minimizing the
variance of around subject to a penalty that constrains the second
difference of . Thus, the HP filter chooses to minimize:
53
University of Ghana  http://ugspace.ug.edu.gh
( − )
+ ( − ) − ( − ) ( . )
Note: MDLTVOL = monthly detrended logged trading volume; MLTVOL = monthly logged trading volume
The penalty parameter controls the smoothness level of the series .  Following
Hodrick & Prescott (1997) and Zaiane & Abaoub (2009), the is set to be equal to 14,400. It
worth noting that the larger , the smoother the series . Thus, the detrending of
the trading volume is performed to obtain a stationary series (see Appendix A) and so
therefore the detrended logged trading volume is used for the analysis.
Figure 3.6 shows the original log trading volume series (blue line), the smooth trend (red
line) and the stationary series (green line). The stationary series referred as cyclical in the
figure is what is employed in the VAR model.
Figure 3.6: A Graph showing the Detrending of Logged Market Trading Volume
Hodrick-Prescott Filter (lambda=14400)
20
18
4
16
3
2 14
1
12
0
-1
-2
00 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15
Note: LMTVOL = Logged Market TrLaMdinTgV VOoLlume Trend Cycle
54
University of Ghana  http://ugspace.ug.edu.gh
3.5.1.2 Thin trading and Monthly Market Return (MRET)
It is reported by several studies (such as Lo & MacKinlay, 1990; Stoll & Whaley, 1990;
Miller et al., 1994) in the literature that thin (infrequent) trading may cause autocorrelation in
the return series. Thin trading is also indicated in the literature to be a common characteristic
of most emerging markets and GSE is no exception. The issue is that the absence of a price
change between any two moments may be explained as being caused by the absence of a
price reaction to new information rather than being due to infrequent trading. It is worth
stating that it is explained in the literature that thin trading is very evidence in high frequency
data such as hourly, daily and weekly.
However, some studies such as Darwish (2012) have indicated that low frequency financial
data such as monthly observation may overcome thin trading since low frequency data have
enough time frame for price change to be observed. As a result, the thin trading correction as
proposed by some researchers is not effected for this study due to the use of monthly data.
The monthly GSE – index return is obtained from the monthly closing value of GSE – index.
The study employs the Ghana Stock Exchange All-Share Index (GSE – ASI) monthly closing
values from January 2000 to December, 2010. However, after January 4, 2011, the GSE –
ASI is replaced with the Ghana Stock Exchange Composite Index (GSE – CI). These indexes
reflect the average price movements in all the equities listed on the stock market. The
employment of the index data presents the inborn advantage that the observations used are
not biased by any peculiar effects taking place within individual stocks (Mensah et al., 2016).
The monthly return is obtained by employing returns continuously compounded formulae:
Note: MRET = monthly GSE – index retu=rns at( time t;)P = monthly GS(E .– in)dex value at time t.
55
University of Ghana  http://ugspace.ug.edu.gh
From equation 3.2, lnis the closing GSE – index value at month , is the closing GSE –index at lag one and denotes the natural logarithm. The continuously compounded return is
adopted because of time addictive and irrelevant of the frequency of compounding (Brooks,
2008). Also, the monthly returns ( ) was found to be stationary (see Appendix A).
3.5.2 Control (Exogenous) Variables
Three main control variables are used in this study, namely, monthly market volatility,
inflation and interest rate. The theoretical explanations and the methodology used to measure
these variables are explained below. However, it is worth pointing out that monthly market
volatility was found to be stationary at level (see Appendix A) whiles inflation and interest
rate were not stationary at level but become stationary at first difference (see Appendix A).
3.5.2.1 Market Return Volatility (MSIG)
Return volatility is introduced into the VAR model as a control variable based on Karpoff’s
(1987) survey of studies on the contemporaneous nexus between trading volume and returns
volatility. Karpoff (1987) argued that market participants employ trading volume to forecast
future price movement of securities and as a result, a relationship between volume and price
movements offers “an insight into structure of financial markets” (Karpoff, 1987, p. 109).
Moreover, the Mixture of Distribution Hypothesis postulates that there exists positive nexus
between stock returns and volume since both jointly depend on the volume which reveals the
market level of information flow. Therefore, security prices variation can be explained by the
flow of information to the market which implies that a positive contemporaneous nexus exist
between volumes and return volatility. Poon & Granger (2003) argues that understanding the
nexus between trading volume and returns volatility helps in modelling returns distributions
56
University of Ghana  http://ugspace.ug.edu.gh
appropriately. Many empirical evidences have reported a relationship between trading
volume and returns volatility. For instance, Chen et al. (2001) indicated that trading volume
acts as a catalyst in reducing stock returns volatility on Chinese Stock Market.
The volatility measure employed in this study follows Statman et al. (2006) and Zaiane &
Abaoub (2009) computation of volatility which is based on French, Schwarz & Stambaugh
(1987). French, Schwarz & Stambaugh (1987) explains that, assuming “N” days for the
month, month t’s volatility with representing the th trading day returns in the month is
given by:
= + ( . )
3.5.2.2 Inflation and Interest Rate (Macroeconomic Variables)
A vast number of empirical studies in literature have documented a significant nexus existing
between returns and macroeconomic variables in both advanced countries and emerging
countries. Specifically, quite a number of empirical studies have found a significant nexus
between stock performance and macroeconomic on Ghana stock market (such as Kumornu &
Owusu-Nantwi, 2011; and Issahaku et al., 2013).
Based on these empirical evidence, inflation and interest rate variable is introduced into the
VAR model. Following Issahaku et al. (2013), the log of monthly consumer price index
( ) is used as proxy for inflation and the log of 91-day Treasury bill rate ( ) as proxy
for interest rate. However, (proxy for inflation) and (proxy for interest rate) were
found not to be stationary hence the first differences (∆ and ∆ ) of these variables
which were found to be stationary were used for the analysis (see Appendix A).
57
University of Ghana  http://ugspace.ug.edu.gh
3.6 Empirical Methodology
In this study, overconfidence bias is tested by employing the Vector Autoregression (VAR)
model and its associated impulse response function. This model is adopted from Statman et
al. (2006) empirical model they used to investigate overconfidence bias.
3.6.1 The General Vector Autoregressive Model (VAR)
The VAR model used in this study was adopted from Statman et al. (2006) paper which
studied overconfidence bias in the US. Thus, Statman et al. (2006) constructed a VAR model
to investigate the lead-lag relationship between market return and market trading volume.
The general form of VAR model employed by Statman et al. (2006) is given by:
= + + + ( . )
where is a × 1 vector of period endogenous observations, for instance, they used
turnover and return. is a vector of control variables observations (they used volatility and
portfolio rebalancing) at time and is × 1 residual vector. The regression coefficients,
estimate the time series nexus between endogenous variables and its lags, estimate the
nexus between endogenous variables and the control variables, where is the number of
lagged endogenous observations and is the number of lagged exogenous observations. is
the last or the nth lag observation of the endogenous variables considered and is the last lag
or nth lag observation of the exogenous variables considered.
58
University of Ghana  http://ugspace.ug.edu.gh
3.6.2 Theoretical framework for the model
According to Statman et al. (2006), the VAR model was employed based on Odean (1998)
and Gervais & Odean’s (2001) paper. Odean (1998) predicts that high total market returns
make some investors overconfident about the precision of their information and investment
strategies. Gervais & Odean (2001) develop a multi-period market model to assess how
market participants learn about their abilities. They also tested how self-attribution bias leads
to overconfidence bias. They indicated that overconfidence level of investor’s changes and
can be attributed to previous market gains and losses. For instance, in a bull market, market
participants easily make gains and as a result traders become more overconfident. They
associate the success to their own expertise ignoring the fact that the gains they are enjoying
are total market phenomena at that point in time. Thus, aggregate market returns have
influence on the overconfidence level of market participants.
Also, Gervais & Odean (2001) indicated that overconfidence bias of market participants leads
them to trade more. They explained that although market returns is market wide, market
participants mistakenly ascribe gains to their own skills to select portfolios and stocks. Based
on this empirical evidence, Statman et al. (2006) postulated that if the lag terms of market
return are significant in explaining current market trading volume then that is evidence of
overconfidence bias. Therefore, Statman et al. (2006) constructed a VAR model to
investigate the lead-lag relationship between market returns and market trading volumes.
The VAR model used by Statman et al. (2006) to measure overconfidence is valid since it
rests on two major implications of overconfidence documented on financial markets by
several researchers. Thus, the VAR model used employs one of the variables reported to be a
source of overconfidence (market returns) and also one variable reported to be the
59
University of Ghana  http://ugspace.ug.edu.gh
predominant effect of overconfidence of investors on financial markets (excessive trading).
Also, the VAR model is suitable for this analysis since it does not put any restrictions on any
of the two main variables of interest (trading volume and returns) as to which of the variables
are exogenous or endogenous but rather all variables are considered endogenous. The VAR
model also permits for a covariance structure to exist in the residual vector that captures the
contemporaneous correlation between control variables which helps to have a better result
than other structural models when control variables are introduced into the model.
3.6.3 VAR Model specification for the study
It is worth pointing out that the adopted VAR model and its associate impulse response
function is employed to address the research objective one, which is investigating an
evidence of overconfidence bias on GSE. As indicated above, the VAR model employed for
this study is adopted from Statman et al. (2006) and modified with the introduction of
macroeconomic variables. Vector Autoregressive model (VAR) is a system regression model
constructed to investigate the lead-lag relationships among variables and Sims (1980)
popularised this model in the econometrics literature as the univariate autoregressive model
generalisation (Brooks, 2008). Typically, a VAR is used to estimate several equations
simultaneously without indicating which of these variables are endogenous and exogenous.
Based on (t3h.e5 )general VAR model (equation 3.4), a market version of the general model inEquation which contains two endogenous variables, market monthly detrended trading
volume and market return, and three control variables, market volatility, inflation and interest
rate is constructed as:
60
University of Ghana  http://ugspace.ug.edu.gh
= +
+ + + , , ( . )
where is the monthly detrended logged trading volume at month ; is the
mΔLoCnPthIly GSE – index return at month ; is the market returns volatility at month ;is0 t8he proxy for inflation at month , is the proxy for0 i8nterest rate at monthand is the 2008 financial crisis dummy variable. The is considered due to
volatility clusters (See Figure 3.1 and Figure 3.2) that manifest in monthly GSE – index
returns ( ) and monthly trading volume ( ) which may be a result of the 2008
global financial crises. Thus, in order to remove the possible impacts of the 2008 global
financial crises from the model, a dummy variable, 08 that takes 1 for January, 2008 to
December, 2010 (otherwise, 0) is introduced.
3.6.3.1 Impulse Response Functions (IRF)
The VAR model indicates which of the variables used in the model have statistically
significant impacts on each of the variables future values in the system. However, the overall
impact of the endogenous and the control variables lags are not fully captured by the VAR
model. Therefore, several researchers employ the impulse response function which uses all
the coefficient estimates of VAR to trace the complete impact of a residual shock (one
sample standard deviation) on the endogenous variables is performed to ascertain the full
impacts of all the variables used in the VAR.
61
University of Ghana  http://ugspace.ug.edu.gh
Impulse response function traces the impacts of one standard deviation shock to the residual
of the endogenous variables on the endogenous variables’ current and future values thro
the dynamic structure of the VAR model (Statman et al., 2006). For instance, Equation (3u.g5h)
have: two endogenous variables, namely, market trading volume and market index returns;
and also three control variables, namely, market returns volatility, inflation and interest rate.
For instance, a shock of one standard deviation to any of the endogenous variable residual,
say , would instantaneously vary the trading volume current value ( )
and consequently affect the future values of . Moreover, this shock would also
influence the market returns ( ) through the coefficient due to the fact that the
lagged values are in both equations.
The primary purpose of this study is to investigate market evidence of overconfidence bias
hypothesis on GSE employing lead-lag relationship between market returns and trading
volume and as a result a one-standard deviation shock is applied to the endogenous variables
residual to ascertain the full impact of the variables in the VAR model. Thus, employing all
the VAR estimates and its dynamic structure, the responds of market detrended logged
trading volume and market returns to the endogenous variables residual shock is graphed.
Therefore, the impulse response function gives a simple picture as to how endogenous
variables respond to these residual shocks over time.
3.6.3.2 Optimal Lag Selection
Deciding on the appropriate lag order for autocorrelation functions is one of the crucial
decisions to make to ensure a valid estimation results. However, financial theory and for that
matter overconfidence bias theories do not specify the suitable lag length that a researcher
should adopt for a VAR model and its associated dynamic analysis. The model used in this
62
University of Ghana  http://ugspace.ug.edu.gh
study is a VAR system and as a result an optimal lag is necessary for valid estimation results.
Therefore, two main approaches are employed to obtain the optimal lag for the VAR model.
Thus, the cross-equation restrictions (Likelihood Ratio (LR) test) and information criteria
approach.
3.6.3.2.1 Cross-equation Restrictions (Likelihood Ratio (LR) Test)
In order to employ the LR test to determine the optimal lag structure, a sequential
modification is performed on the LR. Beginning from the possible highest lag, a jointly zero
hypothesis on the estimated coefficients of equation (3.6) is conducted:
= ( − ) ∑ − ∑ ( . )
Where ∑ refers to the determinant of the variance-covariance matrix r∑esulting from theresiduals of the restricted VAR model (say lag 4) estimation. Also, refers to the
variance-covariance matrix determinant resulting from the unrestricted VAR model (say lag
12) estimation. The refers to the sample size whiles the refers to the number of
parameters per equation. ( − ) in Equation (3.6) is Sims (1980) modification for small
sample size. The test statistics has distribution with the number of restrictions as the
degrees of freedom. To achieve the optimal lag, this test “compare the modified LR statistics
to the 5% critical values starting from the maximum lag, and decreasing the lag one at time
until the first rejection is obtained. The alternative lag order from the first rejected test is
selected as optimal lag” (Eviews 9, User’s Guide).
3.6.3.2.2 Information Criterion
Another way to choose the optimal lag is to select the model order that minimizes
information criterion value. The information criterion is performed by fitting VAR (p)
63
University of Ghana  http://ugspace.ug.edu.gh
models with different lag length on the data and the lag length (p) that provides the minimum
value of the information criterion is selected.  The models used for the information criterion
selection of optimal lags include Akaike Information Criterion (AIC), Schwarz-Bayesian
(SC) and Hannan-Quinn Criterion (HQ).
= − +
= − + ( ) ( . )
= − + ( )
where ℓ = − (1 + ln(2 ) + ln( / )), refers to the number of estimated parameters,
is the sample size. According to Brooks (2008), SC embodies much stiffer penalty term than
AC while HC is somewhere in between.
3.6.4 Granger Causality Test
The granger causality model is used to address the research objective two of examining
whether trading volume has impact on stock returns or stock returns have impact on trading
volume or both exist on GSE. In this study, a granger causality test is conducted alongside the
VAR model estimation to detect the causality between returns and trading volume on GSE as
well as to validate the results of VAR model. The adoption of the granger causality is partly
motivated by Hiemstra & Jones (1994) argument that the trading volume and returns causal
nexus can help to explain stock price movement and changes in trading volume. It partly also
rests on Chuang & Lee (2006) argument that overconfidence causes a positive causal relation
between stock returns and trading volume.
64
University of Ghana  http://ugspace.ug.edu.gh
Granger (1969) proposes an approach to answer the question whether a series say causes
another series say or whether causes . Thus Granger (1969) approach basically seek
to know whether the current values of ′ is explained by past values of . Therefore, isclaim to Granger-cause if lagged explain current ′ and as a result helps in the
predi(c3ti.o9n) of . Specifically, a bivariate Granger causality test of the form equation (3.8)and is adopted to empirically test the causal relation between trading volume and
returns on GSE.
= + + + ( . )
= + + + ( . )
Where is the market trading volume at month , is the market-wide return
at month , The coefficients, and estimate the time series impacts between returns
and trading volume. is the number of lagged observations and is the last or the nth lag
observation considered. and are constant whiles
Based on the model above, a joint hypothesis of the form (3.10)and are the error terms.is tested and the F-statistics
(Wald Statistics) is reported: = = ( . )
The null hypothes(is3 .f8o)r this test is that does not granger-cause withregard to (e3q.u9a)tion and also does not granger-cause with respect toequation . The decision criterion in this test is that, if the corresponding p-value of the
F-statistics test is less than 1% or 5% or 10% significance level, the null is rejected.
65
University of Ghana  http://ugspace.ug.edu.gh
CHAPTER FOUR
RESULTS AND DISCUSSION
4.1 Introduction
The aim of this chapter is to present and discuss results obtained using the data and
methodologies explained in the previous chapter. In this chapter, descriptive statistics, the
market VAR estimation and results and the Granger Causality results are presented and
discussed.
4.2 Descriptive Statistics
The summary statistics on monthly market-wide variables used for this study is presented in
Table 4.1. Thus, the descriptive statistics on monthly trading volume ( ), monthly
G∆SE – index return ( ), market volatility ( ), inflation (∆ ) and interest rate( ) are presented in Table 4.1. The Table 4.1 shows the descriptive statistics on the
above variables for the full sample period (January, 2000 – December, 2015) and the
financial crisis sub-sample period (January, 2008 – December, 2010).
The (and monthly raw trading volume (unreported)) for the full sample period
indicated positive skewness. This suggests that the distribution of the trading volume variable
is skewed to right and leptokurtic. 1T6h,e4 mean shares for the full sample period is 64,250market–wide monthly raw trading volume in(unreported). The descriptive on
indicates kurtosis greater than three for the full sample. The normality hypothesis is rejected
by the Jarque-Bera statistics for trading volume over the full sample period at 5%
significance level. This empirical evidence of non-normality of trading volume is consistent
with Jlassia, Naouib & Mansourc (2013) descriptive results of 27 economies (11 advanced
66
University of Ghana  http://ugspace.ug.edu.gh
markets, 4 Latin Amercian markets, 7 Asian markets and 5 emerging markets) using data
from 2000 to 2012.
Table 4.1: Descriptive Statistics Results
Full sample periodM (2D0L00T VtoO 2L015) – 19M2 RObEsT MSIG ∆LCPI ∆LTBR
Mean --00..001742949522 00..000195561777 00..000001330099 00..001052774842 -00..00002Median 3.283529 0.201080 0.020403 1.256102 0.282030
00306
Maximum 49
Minimum -2.04
Std. Dev. 00..85352
1276143 -00..0352831124 -3.63E-05 -1.383777 -0.71157725 03..800426756938 -03..12633
8658841 -00..036083593269
Skewness
Kurtosis 3.869594 9.926166 21.93949 63.497877070 07..239152892346
Jarque-Bera 015.0.0106690 3Probability 509 09.090.90707000 30.304030.105008 029.0601060.0260 01.5050.80700300
Financial crisis sub--s0a.1m0p3le1 1p8eriod – dumMean -0.238697 00.0
0083 0(821008 to 2001.000)2–53562 Obs 0.011031 0.00
Median 3.283529 0..011704905949 00..000210249043 00..001341541313 00..0210
40000190
Maximum
Minimum -11..08986298 -0.323124 1.46E-08 -0.015222 -0.17
74036553
Std. Dev.
Skewness 0.9559670219 -01..048285479165 20..090033792969 -00..051033016452 00..507854806075
Kurtosis 4.278534 6.810323 12.48229 2.257077 3.956476
Jarque-Bera 07..907188252166 30.30.9007030820 10.8050.40600070 20..334096831122 30..1356487Probability 86702
Note: Obs = Observations; MDLTV∆OL= monthly detrended log trading volume; MRET∆= monthly GSE-indexreturns; MSIG= market volatility; LCPI = inflation (first difference of log CPI), and LTBR = interest rate
(first difference of log 91-day T.Bill rate)
The full sample mean for the12m×on1th.6ly=G1S9E.2–%index returns ( ) isannualised average return is . The kurtosis value of 9.19.266%1;6 6and a simplefor the full
67
University of Ghana  http://ugspace.ug.edu.gh
sample period is greater than three which indicates that GSE – index returns have significant
leptokurtosis. The market returns are negatively skewed indicating that the distribution has a
long left tail. Moreover, the standard deviation that measures the GSE stock index’s return
volatility is high compared to mean return. Likewise, the Jarque-Bera test result rejects the
normality hypothesis of monthly returns at 1% significant level. The high kurtosis value
observed follows the descriptive summary findings on emerging economics of Jlassia et al.
(2013). This empirical finding of non-normality of GSE – index monthly returns is consistent
with Jlassia et al. (2013), the conventional theory (Fama, 1965, 1970) and it is also consistent
with descriptive empirical evidence of Appiah-Kusi & Menyah (2003) when they tested the
weak-form market efficiency of eleven African stock markets including Ghana.
Additionally, Table 4.1 shows the summary statistics on the control variables, namely;
volatility ( ), inflation (∆ ), and interest rate (∆ ). The results of the control
variables over the full sample period show a positive mean return with the exception of
interest rate. In terms of skewness, and ∆ both exhibit return distributions which
are positively skewed over the full sample period. ∆ on the other hand exhibits a
negative skewness over the full sample which means that it has a long left tail. The null
hypothesis of normality is rejected for these variables over the full sample period at 5% level
using the Jarque-Bera statistics. The negative skewness of inflation and non-normality of the
macroeconomic variables follows the descriptive statistics of Kuwornu & Owusu-Nantwi
(2011) when they study the relationship between macroeconomic variables (using consumer
price index (proxy for inflation), crude oil price, exchange rate and 91-day Treasury bill rate
(as a proxy for interest rate)) and market returns on the GSE.
68
University of Ghana  http://ugspace.ug.edu.gh
T26he,6 2m5e,9an1 6raw trading volume (unreported) for the financial crisis sub-sample period is. The kurtosis value for the is greater than three and thus, the trading
activity variables have significant leptokurtosis. The maximum trading volume over the full
sample period is observed in the sub-financial crisis period. The normality hypothesis of
is rejected by the Jarque-Bera statistics at 5% significant level over the financial
crisis sub-sample period.
The mean 3re.6tu%rn of the financial crisis period is approximately 0.3% and a simple annualizedreturn is . Over the financial crisis sub-sample period, the market returns is also
negatively-skewed and the kurtosis values are greater than three and thus, the index has
higher peaks or leptokurtic relative to the normal distribution. The minimum return (-
0.323124) for the full sample period is observed in the sub-financial crisis period. Financial
crisis sub-sample period returns volatility is higher as compared to the standard deviation for
the whole sample period. This implies that volatility on GSE is high during financial crisis
periods. Moreover, the normality hypothesis for the market-wide monthly returns in the sub-
sample period is rejected at 1% significant level which means the return distribution does not
follow normal distribution. These empirical evidences of minimum return and maximum
trading volume being observed in this financial crisis sub-sample also motivated the dummy
variable introduction. The summary statistics on the control variables over the financial crisis
sub-sample period are also reported in Table 4.1.
4.3 VAR Estimation and Test Results
4.3.1 Optimal Lag Selection
The determination of optimal lag length is important for a valid VAR model results. The
overconfidence bias theories do not specify the optimal lag to use for the VAR model.
69
University of Ghana  http://ugspace.ug.edu.gh
Therefore the optimal lag order for the VAR analysis for this study is obtained following the
LR and information criteria test. Eviews computation of the likelihood ratio and different
version of information criteria until the 12th lags is depicted in Table 4.2. Table 4.2 indicates
that both Likelihood Ratio (LR) and Akaike Information Criterion (AIC) reveals that the third
lag is the optimal lag for the endogenous variables. Using the three endogenous variables, the
model is developed by adding the lags of the control variables and then compares the
information criteria of these models. However, it turns out that the optimal lag length for the
model with both the endogenous and control variables is also three lags.
TaLbagle 4.2: VLAogRL Lag OrdeLrR Selection C0 51. FP
riEteria ResulAtIsC SC HQ
21 8843.
03
.277473
40933 58.N6A 0.00264456 2.6449716864 00..00001199273 -
-00.260282 0.249 -0..555709835260 -00..00095
2113670* --00..0565177 -0.3248684
57151*
43 9915..6445064910 162.6.25725* 0.001916* -0.583694* 0.134080 -0.292
859239
56 9987..2206040973 31..04
2648
81720060
1 0.001922 -0.581248 0.208304 -0.2610
08 00..000021907431 --00..552525796383 00.305396 -0.2066
3161
78 91903.9.67672657 62..818203559293 00..000022109077 --00..449963396223 00.
.450170936318 --00..144.583299 -0.005869
333718
61
9 01110
110078..66027461 61..4955110.0248 2.2567925
9876 00..00002111 -0.492627 0.0022128578 --00..442598
96
6851
1
90 0
0..676515440770.8623346
-00..003267016939
* i1n2dicates l1a11.9633 3.044710 0.002315 -0.406365 0.957405 0
0..104946732273
ALFPRICE: :s:  AeFqkinuaaiekln ept
g o
 irianeldf omSC: Schwarz infoi
rcodtideorinf is eeedrler LocRtre tde bsty s tthaeti sctriicte (reiaocnh test at 5% level)
HQ: Hannan-Quirnrmmna aitntiiofoonnr mc crraiittteieorrinioo ncnriterion
70
University of Ghana  http://ugspace.ug.edu.gh
4.3.2 Market VAR Estimation Results
The results of market VAR system which is used to address the research objective one
estimation is reported in Table 4.3. The market VAR system is made up of two endogenous
variables: monthly market-wide trading volume ( ) and monthly GSE – index
return ( ); and three control variables: market volatility ( ), the first difference of
inflation (∆ ) and the first difference of interest rate (∆ ). Table 4.3 is arranged into
columns for endogenous variables and rows for the lag terms of both endogenous and control
variables. For each of the variables, estimated coefficient, standard error, test statistics and p
– value results are reported. In order to verify if the variable estimated coefficient is
significantly dif1fe0r%ent from zero, the p-values are employed. Thus, variables with reported p –value less than are considered as significant in this study. Also, variables with reported
p – value less than 5% or018% are also considered as highly significant in this study. It isworth stating that the refers to the 2008 global financial crisis sub-sample period and
as indicated in the Table 4.3, has no significant statistical impact on the VAR system.
According to Table 4.3, it is documented that the monthly trading volume is not
autocorrelated. Thus, the first, second and third lagged market-wide monthly trading volume
estimated coefficients are insignificant even though the first lag is positive. It indicates that
the monthly market-wide trading volume is not significantly affected by its own behaviour in
past months. Using the monthly raw trading volume as a robust check, the results
(unreported) indicated that monthly raw trading volume is not autocorrelated. This implies
that market participants cannot use past trading volume to predict future trading volume on
GSE. This empirical evidence is consistent with Zaiane & Abaoub (2009) results on Tunisian
71
University of Ghana  http://ugspace.ug.edu.gh
Table 4.3: Vector Autoregression (VAR) Estimation Results
−00.1.01754529528 00..0
.
07788619509 −10.
−4 − . − −
0.045713 0.077905 0. .9
7484305511 00..31445038 −00.0.00072198528 00..000044650881 −10.5.66471798003 00..51127149
3−.428.019202158∗0∗∗ 11..532061026348 −21.6
52846873777 00..05057981 00.5.04083776892 0.004564∗∗∗ 0.077700 70..802628672503 00..04007009
−10.3.03522572
.392490 0.1647 −0.1
012 01..301864066965 −10.0.61222448706 00..35312410 00.1.096
9587298∗∗ 0.087939 −2.220819 0.0270
−21.06351 26.59476 −0.792017 0.4289 −1.70468524
41∗∗ 00..007074916004 21..535183734735 00..01811812
−7.283127 28.01752 0.259949 0.7951 −0.12284
358 11..654581047274 −−10..012742084492 00..92640274
−56
3.25.579685589 27.44409 −1.193914 0.2334 2.336669 1.607832 1.453304 0.10.01464∗5∗ 026.4.0303520612 −20.1.04376655964 00..90732085 −10.0.01270779695 01..502423849462 −00.6.858956670 0.50
47991
−00.1.38106182453 00..442233743524 −00.4.72457377414 00..64750518 −−00..003023527613 00..002244880258 −−01..219449502
36
666 00
0..3.818
765
0.042251 0.402121 0.105071 0.9164 −0.002433 0.023559 −0.103292 0.919
56
−1.817571 1.07 7
487
−11.2.7422134098
∗
6 11..11768
2189 −1.6951
3620744 −11.0.46636387
95 0.091
416 00..21488
04 00..004102949915 00.062815 0.198846 0.8425
0.942202 1.095373 0.860165 0.390343 −−00..007801423973 0
..006688745464 −01.5.098958 0.54960.064173 −1.25318918837 00..22199176
No08= Steta: n**d*a,*
−
r* 
0a.06d enrdr 
8*7 in76or; td-iscat
0e.17128tatss  =si gtnei
5fican−t 0a.t4 10%15, 258%  and0 1.608%8 3level re−sp0e.0ct0i7ve1l7y7; Coe0f.f0 =10 c0o3e5fficie−n0t;.7 S1t5. e2r1r2or 0.4750
∆ st statistics. Where: MDLTVOL= monthly detrended log trading volume;MRET= monthly GSE-index returns; MSIG= market volatility; ∆LCPI = inflation (first difference of log CPI),and LTBR = interest rate (first difference of log 91-day T.Bill rate)
72
University of Ghana  http://ugspace.ug.edu.gh
stock market when they also used detrended trading volume and returns to examine
overconfidence bias in Tunisian market. However, this result contradicts with Statman et al.
(2006) findings from advanced market. Statman et al. (2006) used share turnover as proxy for
trading volume and they reported that turnover is autocorrelated, with a highly significant
first lagged coefficient of 0.284.
More importantly, the first lag of market returns is found to have highly significant positive
impacts on market trading volume. The significant impact of the market return on the market
trading volume only exists in the first lag whiles the second and third lag of market return are
n3o.t4 8s1ta2t2is5tical significant. The highly significant posi1ti%ve first lagged market return( )  impact on the market-wide trading volume at confidence level demonstrates
the overconfidence hypothesis that high market returns triggers overconfidence of investors
which makes them trade more in subsequent periods. The empirical evidence of positive
highly significant nexus between current market trading volume and first lag market returns
is the major findings of this study. This empirical finding indicates that GSE – index past
market returns triggers overconfidence of market participants on GSE which makes them
trade more in subsequent periods. This empirical result is consistent and also confirms the
overconfidence hypothesis results of empirical studies reported in the finance literature by
scholars such as Statman et al. (2006), Chuang & Lee (2006), Zaiane & Abaoub (2009) and
Metwally & Darwish (2015). Statman et al. (2006) for instance employed monthly shares
turnover, returns, market volatility and dispersion reported highly significant first lag
coefficient (0.819) and second lag coefficient (0.433). Statman et al. (2006) explained this
finding as an evidence of overconfidence on US stock markets. Also, Zaiane & Abaoub
(2009) following Statman et al. (2006) examined overconfidence bias on Tunisian stock
market reported only one significant coefficient (the fifth lag) of monthly returns impacts on
73
University of Ghana  http://ugspace.ug.edu.gh
trading volume. They interpreted this result as evidence of overconfidence but indicated it is
remarkably weak.
According to Table 4.3, the exhibits autocorrelation which implies that market return
prediction anomaly exists on GSE. Thus, estimated coefficients for first, second and third
lags of are highly significant. This means that past monthly market returns influence
the behaviour of current period’s market returns. Specifically, market returns are significantly
positively influenced by its first lag. This finding is also consistent with Mensah et al. (2016)
evidence of the GSE – index returns exhibiting a significant autocorrelation. The influence of
past market returns on the current market returns imply that the investors on GSE can use
past market returns to predict future returns. This result is inconsistent with the most result
from advanced markets. For instance, Statman et al. (2006) found no autocorrelation on
monthly market returns in US.
Considering the contemporaneous nexus between monthly return volatility and monthly
trading volume, results in Table 4.3 indicate that volatility has no significant influence on
trading volume at contemporaneous level. This finding implies that the GSE market has
liquidity challenges since return volatility has no contemporaneous effect on trading volume.
This empirical result is inconsistent with the findings of Karpoff (1987), Statman et al. (2006)
and Zaiane & Abaoub (2009) on the same subject matter. However, return volatility has
positive highly significant impact on trading volume at lag three.
With respect to the relationship between macroeconomic variables and trading volume,
interest rate has negative significant impacts on the trading volume at contemporaneous level.
According to Adjasi et al. (2008), the negative statistical significant inflation estimates
74
University of Ghana  http://ugspace.ug.edu.gh
implies that under a good-looking Treasury bill rates, stock analyst forecast investors being
rational may shift much of the funds at their disposal from stocks to treasury bills and as a
result affect stock market trading activities.
Moreover, in order to ensure the coefficients reported above are significant, a VAR Residual
Portmanteau Tests for Autocorrelations is performed and the null of no residual
autocorrelations up to lag 30 cannot be rejected (see Appendix B).
4.3.3 Impulse Response Functions (IRF)
Impulse Response Function which uses all the coefficient estimates of the VAR to trace the
complete impact of a residual shock (one sample standard deviation) on the endogenous
variables is performed. Figure 4.1 consist of IRF graph employing all the VAR estimates
depicted in Table 4.3. The vertical axis measures the percentage increase in endogenous
variable with respect to a shock. The blue line in the graph outlines the response of the
endogenous variable to a shock whilst the red line is the two standard error band. The
impulse function becomes zero in the long term since the monthly market-wide trading
volume is detrended (Statman et al., 2006).
Figure 4.1: Impulse Response Function (IRF) Graph of the Market-wide VAR model
Response to Cholesky One S.D. Innovations ± 2 S.E.
Response of MDLTVOL to MRET
1.0
0.8
0.6
0.4
0.2
0.0
-N0o.t2e: MDT1VOL = m2onthly 3market-4wide tra5ding vol6ume; MR7ET = m8onthly G9SE-inde1x0 return1s1 12
75
University of Ghana  http://ugspace.ug.edu.gh
Figure 4.1 depicts the response of monthly trading volume to a residual shock of monthly
market returns. The Figure 4.1 shows that the response by market trading volume to a
residual shock of returns exists until the seventh period. Precisely, in lag one and two, the
response is evidently large and positive. The shock of market return residual brings 0.12%
increase in market trading volume at first lag and 0.19% increase at second lag. Statman et
al. (2006) interpret this evidence as market returns influence on overconfidence level of
investors which trigger changes in their subsequent trading activities. This confirms the key
finding of this study established in the VAR model results. However, there are few
behavioural or rational expectations models that are consistent with the positive lead – lag
relationship between returns and volume such as portfolio rebalancing and heterogeneous
interpretation of informational (Sequential arrival of information). However, the magnitude
and persistence of the lead – lag relationship between returns and trading volume is so strong
that the empirical are interesting independent of one’s interpretation (Statman et al., 2006).
For the other graphs of the impulse response function, see Appendix C.
4.4 Granger Causality for market returns and trading volume results
The granger causality test is conducted to address the second research objective. Granger
causality test is conducted to detect the causality between returns and trading volume on
GSE. However, this test is not conducted for such only purpose but also is conducted
alongside the VAR model estimation in this study to check the validity of the results obtained
from the VAR model. Employing 3 lags of the endogenous variables and 189 observations,
the granger causality tests results is obtained (see Appendix D).
The null hypothesis that monthly trading volume does not granger-cause monthly market
returns cannot be rejected since it has p-value more than .05 (F − score = 1.20932). Thus,
76
University of Ghana  http://ugspace.ug.edu.gh
the granger test reveals that trading volume has no significant impact on market returns. On
the other hand, the second hypothesis that mFa−rksect orreetu=rns3 .d2o21n6o6t granger cause tradingvolume is rejected at 5% level of significance ( ). Thus, the rejection of
the second null hypothesis implies that monthly returns granger-cause monthly trading
volume. This shows that the monthly returns have impact on monthly trading volume. This
implies that there exists unidirectional causality from monthly market returns to monthly
trading volume on GSE. This result is in conformity with the VAR findings.
The unidirectional causality running from returns to trading volume is consistent with Tariq
& Ullah (2013) findings of unidirectional causality running from market returns to trading
volume on Pakistan stock markets. However, it is inconsistent with Metwally & Darwish
(2013) who found bi-directional causality from returns and trading volume (using turnover as
proxy) on Egypt stock markets.
The primary goal of this study is to test the rationality of market participants by investigating
the evidence of overconfidence bias on Ghana stock market. Quite a number of interesting
empirical evidence emerged from the data analysis. In general, the results above indicate that
there is a significant association between past returns and current trading volumes on GSE.
This empirical finding is confirmed by the VAR estimation, the impulse response function
and the granger causality test. There is unidirectional causality running from market returns
to market trading volume on GSE. It is also observed that market returns are autocorrelated
which means investors can use past returns to predict future market returns. However, no
significant contemporaneous nexus between trading volume and volatility on GSE is found as
the literature postulates. Conversely, a negative significant contemporaneous relationship is
recorded for trading volume and interest rate.
77
University of Ghana  http://ugspace.ug.edu.gh
CHAPTER FIVE
SUMMARY, CONCLUSION AND RECOMMENDATIONS
5.1 Introduction
The study examined the rationality of market participants on GSE by using overconfidence
bias of market players on Ghana stock market as a proxy. The aim of this chapter is to
provide a summary of the previous chapters as well as to make recommendations to policy
makers and future researchers.
5.2 Summary
The basic goal of this study is the testing of the rationality of investors on GSE. The goal was
necessitated by vast empirical evidences of market inefficiencies and anomalies documented
on GSE which deny GSE from the efficient allocation of resource benefits attributed to
efficient capital market in the literature. Overconfidence bias as a behavioural bias is
employed to test the rationality of market participants on GSE. The overconfidence bias of
market participants have been investigated by numerous financial economists. According to
Gervias & Odean (2001), due to the self-attribution bias, the overconfidence level of market
agents is positively related to market return. Also, Odean (1998) reveals that overconfidence
of market participants leads market participants to trade more. Based on these prior empirical
studies, the lead-lag nexus between trading volume and returns is the framework adopted for
this study following Statman et al. (2006).
Following Statman et al. (2006), a market-wide VAR model with monthly market trading
volume and market returns as endogenous variables and volatility, interest rate and inflation
78
University of Ghana  http://ugspace.ug.edu.gh
as control variables is employed to examine the lead-lag between trading volume and returns
on Ghana stock market. The study employed data from 2000 to 2015.
The empirical result indicates that previous market return is positively related to current
market trading volume. A further analysis to ascertain the duration of the relationship is
conducted by employing an impulse response function. The res
a shock to market returns residual result to approximately 0u.l1ts2 from t0h.e1 I9RF revealed thatand rise in market
trading volume at lag one and two respectively. However, the magnitude and persistence of
the lead – lag relationship between returns and trading volume is so strong that the empirical
are interesting independent of one’s interpretation (Statman et al., 2006). As a result, this
finding is an empirical evidence of overconfidence bias on GSE. This is consistent with
overconfidence bias as explained by Statman et al. (2006). However, this overconfidence
effect in the GSE market does not persist over a longer period as in the U.S. market. The
granger causality test result revealed unidirectional causality from returns to trading volume.
This also confirms the overconfidence bias since returns granger cause trading volume.
However, the positive contemporaneous nexus between trading volume and returns volatility
documented in the literature, in line with Mixture of Distribution Hypothesis, is not evident
on the Ghana stock market. Market returns is found to be autocorrelated but trading volume
was not found to be autocorrelated. A cotemporaneous relationship between trading volume
and interest rate is documented on GSE.
5.3 Conclusion
Consistent to expectation, the VAR model results indicate that past market return is related to
current trading volume. The impulse response functions also show higher response of market
79
University of Ghana  http://ugspace.ug.edu.gh
trading volume to a shock of market returns residual. This implies that market participants on
the Ghana stock market exhibit characteristics of overconfidence. Therefore, the market
players on GSE exhibit conditional irrationality in the form of overconfidence bias. Also, a
unidirectional causality from market returns to trading volume was documented. The
researcher additionally introduced macroeconomic variables (inflation and interest rate) as a
control variable in the VAR model, to capture the returns and macroeconomics variable
relationship and the results is economically meaningful and as well as statistically significant
for interest rate influencing trading volume in Ghana.
5.4 Recommendations
Based on these empirical results, the GSE regulators must appreciate the fact that the
inefficiencies in the market may be as result of behavioural biases of market participants such
as the overconfidence bias found in this study. The GSE regulators should consider
behavioural biases as one of the possible cause of market inefficiency on GSE than arguing
with the size of the market. The GSE regulators should come out with training programmes
for market participants on behavioural biases and warn investors on existence of behavioural
biases on the market so that they would not “confuse brains with a bull market” as the old
Wall Street adage put it. Also, Ghana Stock Exchange Commission should incorporate
behavioural finance in the programmes they run on the GSE to educate potential market
participants on the presence of conditional irrational traders on the market.
Additionally, financial advisors, investment banks and brokers in Ghana should educate their
clients about evidence of overconfidence behavioural bias on the exchange. They should
educate them about the hazard of having behavioural biases such as overconfidence on their
long run returns in the markets. If possible, they should conduct a study on their client to
80
University of Ghana  http://ugspace.ug.edu.gh
know which kind of behavioural biases they are exposed to so that they can advise
accordingly.
Market participants should also note that whenever there is a positive market returns, the
subsequent period trading volume will be high. Therefore, market participants can take
advantage when designing their trading strategy in the GSE market where liquidity is a
problem. Moreover, market participants should also not “confuse brains with a bull market”
so that they can avoid this overconfidence bias which is reported in the literature to lead to
low returns in the long run. Market participants should also understand that conditional
irrational traders exist on the market and can cause deviations of stock prices.
Further studies should look at the overconfidence bias using individual market participant’s
data on Ghana stock market. Future empirical research can also distinguish between
individual and institutional investors to test the evidence of overconfidence bias for these
classes of investors. Similarly, it is recommended that other studies also consider other
behavioural biases documented in the literature such as disposition effect on GSE.
81
University of Ghana  http://ugspace.ug.edu.gh
REFERENCES
Adjasi, C. K., Harvey, S. K., & Agyapong, D. (2008). Effect of exchange rate volatility on
the Ghana Stock Exchange. African Jouurnal of Accounting, Economic, Finance and
Baking Research, 3(3), 28-47.
Akerlof, G. A., & Shiller, R. J. (2010). Animal spirits: How human psychology drives the
economy, and why it matters for global capitalism. Princeton University Press.
Alagidede, P., & Panagiotidis, T. (2009). Calendar anomalies in the Ghana stock
exchange. Journal of Emerging Market Finance, 8(1), 1-23.
Aly, H. Y., Mehdian, S. M., & Perry, M. J. (2004). An analysis of day of the week effects in
the Egyptian stock market. International Journal of Business, 9(3), 301-308.
Antwi, S., Ebenezer, A. M. F. E., & Zhao, X. (2012). The effect of macroeconomic variables
on stock prices in emerging stock market: Empirical evidence from
Ghana. International Journal of Social Sciences Tomorrow, 1(10), 1-10.
Appiah-Kusi, J., & Menyah, K. (2003). Return predictability in African stock
markets. Review of financial economics, 12(3), 247-270.
Authers, J. (2010). The fearful rise of markets: Global bubbles, synchronized meltdowns, and
how to prevent them in the future. Pearson Education.
Ayentimi, D. T., Mensah, A. E., & Naa-Idar, F. (2013). Stock Market Efficiency of Ghana
Stock Exchange: An Objective Analysis. International Journal of Management,
Economics and Social Sciences, 2(2), 54-75.
Barber, B. M., & Odean, T. (2001). Boys will be boys: Gender, overconfidence, and common
stock investment. Quarterly Journal of Economics, 141(1), 261-292.
82
University of Ghana  http://ugspace.ug.edu.gh
Barberis, N., & Thaler, R. H. 2003. A Survey of Behavioural Finance, in Constantinides,
G.M, Harris, M. and Stulz, R. eds., Handbook of the Economics of Finance. North
Holland, Amsterdam.
Beck, T., & Levine, R. (2004). Stock markets, banks, and growth: Panel evidence. Journal of
Banking & Finance, 28(3), 423-442.
Ben-David, I., Graham, J. R., & Harvey, C. R. (2013). Managerial Miscalibration. Quarterly
Journal of Economics, 128(4), 1547-1584.
Benos, A. V. (1998). Aggressiveness and survival of overconfident traders. Journal of
Financial Markets, 1(3), 353-383.
Biais, B., Hilton, D., Mazurier, K., & Pouget, S. (2005). Judgemental overconfidence, self-
monitoring, and trading performance in an experimental financial market. The Review
of Economic Studies, 72(2), 287-312.
Binmore, K. (1999). Why experiment in economics? The Economic Journal, 109(453), 16-
24.
Black, F. (1986). Noise. The Journal of Finance, 41(3), 528-543.
Brennan, M. J., & Cao, H. H. (1997). International portfolio investment flows. The Journal of
Finance, 52(5), 1851-1880.
Brooks, C. (2008). Introductory econometrics for finance. Cambridge University Press.
Chen, G. M., Firth, M., & Rui, O. M. (2001). The dynamic relation between stock returns,
trading volume, and volatility. Financial Review, 36(3), 153-174.
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.
83
University of Ghana  http://ugspace.ug.edu.gh
Choudhry T. (1994). Stochastic Trends and Stock Prices: An International Inquiry. Applied
Financial Economics, 4(6), 383-390.
Chuang, W. I., & Lee, B. S. (2006). An empirical evaluation of the overconfidence
hypothesis. Journal of Banking & Finance, 30(9), 2489-2515.
Conrad, J. S., Hameed, A., & Niden, C. (1994). Volume and autocovariances in short‐horizon
individual security returns. The Journal of Finance, 49(4), 1305-1329.
Cooray, A., & Wickremasinghe, G. (2007). The efficiency of emerging stock markets:
Empirical evidence from the South Asian region. The Journal of Developing
Areas, 41(1), 171-183.
Daniel, K. D., & Hirshleifer, D. (2015). Overconfident Investors, Predictable Returns, and
Excessive Trading. Journal of Economic Perspectives, 29(4), 61-88.
Daniel, K. D., & Titman, S. (1999). Market Efficiency in an Irrational World. Financial
Analysts’ Journal, 55(6), 28-40.
Daniel, K. D., Hirshleifer, D., & Subrahmanyam, A. (1998). Investor Psychology and
Security Market Under- and Over-Reaction. The Journal of Finance, 53(6), 1839-1885.
Darwish, M. J. (2012). Testing the Contemporaneous and Causal Relationship between
Trading Volume and Return in the Palestine Exchange. International Journal of
Economics and Finance, 4(4), 182-192.
De Long, B. J., Shleifer, A., Summers, L. H., & Waldmann, R. J. (1990). Noise Trader Risk
in Financial Markets. Journal of Political Economy, 98(4), 703-737.
De Long, J. B., Shleifer, A., Summers, L., & Waldmann, R. J. (1991). The survival of noise
traders in financial markets. Journal of Business, 64(1), 1–20.
84
University of Ghana  http://ugspace.ug.edu.gh
DeBondt, W. F., & Thaler, R. H. (1995). Financial decision-making in markets and firms: A
behavioral perspective. Handbooks in operations research and management
science, 9(13), 385-410.
Dewotor, F. S., & Gborglah J. (2004). The Information Content of Stock Prices in Ghana,
Databank Research Paper.
Dickson, G. K. (2015). Assessing the performance of Active and Passive trading on the
Ghana Stock Exchange, unpublished thesis paper, University of Ghana.
Doviak, P. (2015). How to incorporate behavioural finance theory into a planning practice.
Journal of Finance Planning, 28(3), 30-32.
Dunning, D. (2005). Self-insight: Roadblocks and detours on path to knowing thyself. New
York: Psychology Press.
Epps, T. W., & Epps, M. L. (1976). The stochastic dependence of security price changes and
transaction volumes: Implications for the mixture – of – distributions
hypothesis. Econometrica: Journal of the Econometric Society, 44(2), 305-321.
EViews 9 User’s Guide, 2015, Quantitative Micro Software, USA.
Fama, E. F. (1965). The behavior of stock-market prices. The Journal of Business, 38(1), 34-
105.
Fama, E. F. (1970). Efficient capital markets: a review of theory and empirical work. Journal
of Finance, 25(2), 383−417.
Fama, E. F., & French, K. R. (1993). Common risk factors in the returns on stocks and
bonds. Journal of Financial Economics, 33(1), 3-56.
French, K. R., Schwert, G. W. and Stambaugh, R. F. (1987). Expected Stock Returns and
Volatility. Journal of Financial Economics 19(1), 5-29.
85
University of Ghana  http://ugspace.ug.edu.gh
Gallant, A. R., Rossi, P. E., & Tauchen, G. (1992). Stock prices and volume. Review of
Financial Studies, 5(2), 199-242.
Gervais, S., & Odean, T. (2001). Learning to be overconfident. Review of Financial
Studies, 14(1), 1-27.
Ghana Stock Exchange (2016). GSE Anniversary Handout.
Gigerenzer, G., Hoffrage, U., & Kleinbolting, H. (1991). Probabilistic Mental Models: A
Brunswikian theory of confidence. Psychological Review, 98(1), 506-528.
Glaser, M., & Weber, M. (2009). Which past returns affect trading volume?. Journal of
Financial Markets, 12(1), 1-31.
Glaser, M., Langer, T., & Weber, M. (2013). True overconfidence in interval estimates:
Evidence based on a new measure of miscalibration. Journal of Behavioral Decision
Making, 26(5), 405-417.
Granger, C. W. J. (1969). Investigating Causal Relations by Econometric Models and Cross-
Spectral Methods. Econometrica, 37(1), 424–438.
Griffin, D., & Tversky, A. (1992). The weighing of evidence and the determinants of
confidence. Cognitive Psychology, 24(3), 411-435.
Grinblatt, M., & Keloharju, M. (2001). What makes investors trade?. The Journal of
Finance, 56(2), 589-616.
Hastorf, A. H., Schneider, D. J., & Polefka, J. (1970). Person Perception, Addison-Wesley,
Reading, Mass.
Hiemstra, C., & J‐ones, J. D. (1994). Testing for linear and nonlinear Granger causality in thestock price volume relation. The Journal of Finance, 49(5), 1639-1664.
86
University of Ghana  http://ugspace.ug.edu.gh
Hirshleifer, D. (2001). Investor psychology and asset pricing. The Journal of Finance, 56(4),
1533-1597.
Hirshleifer, D., & Luo, G. Y. (2001). On the survival of overconfident traders in a
competitive securities market. Journal of Financial Markets, 4(1), 73-84.
Hodrick, R. J., & Prescott, E. C. (1997). Postwar U.S. Business Cycles: An Empirical
Investigation. Journal of Money, Credit, and Banking, 29(1), 65-91.
Issahaku, H., Ustarz, Y., & Domanban, P. B. (2013). Macroeconomic Variables and Stock
Market Returns in Ghana: Any Causal Link?. Asian Economic and Financial Review,
3(38), 1044-1062.
Jefferis, K., & Smith, G. (2005). The changing efficiency of African stock markets. South
African Journal of Economics, 73(1), 54-67.
Jensen, M. C. (1978). Some anomalous evidence regarding market efficiency. Journal of
Financial Economics, 6(2/3), 95-101.
Jlassia, M., Naouib, K., & Mansour, W. (2013). Overconfidence behaviour and dynamic
market volatility: evidence from international data. Procedia Economics and Finance,
13(1), 128-142.
Kahneman, D., Slovic, P., & Tversky, A. (1982). Judgment under uncertainty: Heuristics and
Biases. Cambridge, UK: Cambridge University Press.
Karpoff, J. M. (1987). The relation between price changes and trading volume: A
survey. Journal of Financial and Quantitative Analysis, 22(1), 109-126.
Koriat, A., Lichtenstein, S., & Fischhoff, B. (1980). Reasons for confidence. Journal of
Experimental Psychology: Human learning and memory, 6(2), 107-118.
87
University of Ghana  http://ugspace.ug.edu.gh
Kuwornu, J. K. (2012). Effect of macroeconomic variables on the ghanaian stock market
returns: A co-integration analysis. Agris on-line Papers in Economics and
Informatics, 4(2), 1-15.
Kuwornu, J. K. M., & Owusu - Nantwi, V. (2011). Macroeconomic variables and stock
market returns: Full information maximum likelihood estimation. Research Journal in
Finance and Accounting, 2(4): 49-63.
Kyereboah-Coleman, A., & Agyire-Tettey, K. F. (2008). Impact of macroeconomic
indicators on stock market performance: The case of the Ghana Stock Exchange. The
Journal of Risk Finance, 9(4), 365-378.
Lakonishok, J., & Vermaelen, T. (1986). Tax-induced trading around ex-dividend
days. Journal of Financial Economics, 16(3), 287-319.
Liorente, G., Michaely, R., Saar, G., & Wang, J. (2002). Dynamic volume-return relation of
individual stocks. Review of Financial Studies, 15(9), 1005-1047.
Lo, A. W., & MacKinlay, A. C. (1988). Stock market prices do not follow random walks:
Evidence from a simple specification test. Review of financial studies, 1(1), 41-66.
Lo, A. W., & MacKinlay, A. C. (1990). Data – snooping in tests of financial asset pricing
models, The Review of Financial Studies, 3(1), 431 – 467.
Magnusson, M., & Wydick, B. (2002). How efficient are Africa's emerging stock
markets? Journal of Development Studies, 38(4), 141-156.
Mensah, L., Bokpin, G. A., & Owusu-Antwi, G. (2016). Time your investment on the Ghana
Stock Exchange (GSE). African Journal of Economic and Management Studies, 7(2),
256-267.
Mensah, T. J., Adom, P. H., & Pomaa-Berko, M. (2014). Does Automation Improve Stock
Market Efficiency in Ghana? African Review of Economics and Finance, 6(1) 69–101.
88
University of Ghana  http://ugspace.ug.edu.gh
Metwally, A. H., & Darwish, O. (2015). Evidence of Overconfidence Bias in the Egyptian
stock market in different states. The Business and Management Review, 6(4), 178-198.
Miller M. H., Muthuswamy, J. & Whaley, R. E. (1994). Mean reversion of Standard & Poor’s
500 Index basic changes: Arbitrage – induced or statistical illusion? The Journal of
Finance 49(2), 479 – 513.
Mobarek, A., & Keasey, K. (2000). Weak-Form Market Efficiency of an Emerging Market:
Evidence from Dhaka Stock Market of Bangladesh, Paper Presented at ENBS
Conference held on Oslo, 2000, 1-3.
Muhammad, F. S., & Khalid, R. (2012). Overconfidence Bias, Trading Volume and Returns
Volatility: Evidence from Pakistan. World Applied Sciences Journal, 18(12), 1737-
1748.
Newell, A., & Simon, H. A. (1972). Human problem solving (Vol. 104, No. 9). Englewood
Cliffs, NJ: Prentice-Hall.
Nofsinger, J. R. (2001). Investment madness: how psychology affects your investing – and
what to do about it. USA: Pearson Education.
Odean, T. (1998). Volume, volatility, price, and profit when all traders are above
average. The Journal of Finance, 53(6), 1887-1934.
Odean, T. (1999). Do investors trade too much? American Economic Review 89(5),
1279−1298.
Olsen, R. A. (2008). Cognitive dissonance: The problem facing behavioral finance. Journal
of Behavioral Finance, 9(1), 1–4.
Peterson, D. K., & Pitz, G. F. (1986). Effects of amount of information on predictions of
uncertain quantities. Acta Psychological, 61(1), 229-241.
89
University of Ghana  http://ugspace.ug.edu.gh
Pompian, M. M., & Longo, J. M. (2005). Incorporating Behavioral Finance into Your
Practice. Journal of Financial Planning, 18(3), 58-63.
Poon, S. H., & Granger, C. W. (2003). Forecasting volatility in financial markets: A
review. Journal of Economic Literature, 41(2), 478-539.
Samuelson, W., & Zeckhauser, R. (1988). Status quo bias in decision making. Journal of risk
and uncertainty, 1(1), 7-59.
Scheinkman, J. A., & Xiong, W. (2003). Overconfidence and Speculative Bubbles. Journal of
Financial Economics, 111(6), 1183-1219.
Schindler, M. (2007). Rumors in financial markets: Insights into behavioral finance (Vol.
413). John Wiley & Sons.
Shiller, R. J. (2000). Measuring bubble expectations and investor confidence. The Journal of
Psychology and Financial Markets, 1(1), 49-60.
Shiller, R., J. (1981). Do stock prices move too much to be justified by subsequent changes in
dividends?. American Economic Review, 71(2), 421-498.
Simons, D., & Laryea, S. A. (2006). The Efficiency of Selected African Stock
Markets. Finance India, 20(2), 553-571.
Sims, C. (1980). Macroeconomics and Reality. Econometrica, 48(1), 1-48.
Statman, M., Thorley, S., & Vorkink, K. (2006). Investor overconfidence and Trading
Volume. Review of Financial Studies, 19(4), 1531-1565.
Stickel, S. E., & Verrecchia, R. E. (1994). Evidence that trading volume sustains stock price
changes. Financial Analysts Journal, 50(6), 57-67.
Stoll, H. R. & Whaley, R. E. (1990). The dynamics of stock index and stock index futures
returns. Journal of financial and Quantitative Analysis, 25(1), 441 – 468.
90
University of Ghana  http://ugspace.ug.edu.gh
Tariq, B., & Ullah, N. (2013). Investor Overconfidence and Stock Returns: Evidence from
Pakistan. Journal of Business and Management, 8(1), 77-84.
Taylor, S. E., & Brown, J. D. (1988). Illusion and well-being: a social psychological
perspective on mental health. Psychological Bulletin, 103(2), 193-210.
Tkac, P. (1996). A Trading Volume Benchmark: Theory and Evidence. Working paper,
Department of Finance and Business Economics, University of Notre Dame.
Yakob, N. A., Beal, D., & Delpachitra, S. (2005). Seasonality in the Asia Pacific stock
markets. Journal of Asset Management, 6(4), 298-318.
Zaiane , S., & Abaoub, E. (2009). Investor Overconfidence And Trading Volume: The Case
Of An Emergent Market. International Review of Business Research Papers, 5(2),
213-222.
91
University of Ghana  http://ugspace.ug.edu.gh
APPENDICES
Appendix A: Unit root test for stationarity
ADF PP KPSS
Time series Intercept Trend and None Intercept Trend and None Intercept Trend and
intercept intercept intercept
MLTVOL −−312.8.3479436166∗∗ −11.09054∗∗MDLTVOL −4.011922∗∗ −−172.9.0486280389∗∗ −−
01.22.8186429323 −−83.780846∗∗ −812
.745951∗∗ −
.1.9153210437∗∗ −−1
11.342
82.1.91008
36∗∗ 0
27429∗∗ −−172
.5.11304.9676264333
4 01..043019064223 00.0.13461∗∗ 0.150388∗∗ 0.05110704
522∗∗
MRET −9.4 ∗∗ ∗∗ ∗∗ ∗∗ ∗∗ ∗∗ ∗∗ 0∗∗MSIG
LCPI −−128.2
4758260979∗∗ −−9.88646 −11
.4
8.7
72
.9796
49604∗∗ −8.01909 −108.3
6296
.99093606
1
79
∗∗ −
−−
9
12
.4.33517684219.631977
∗∗ −−92.4−19.0
5167440107∗∗ −80..348169853930∗∗ 00..181899697736∗∗ 00.0.37288575088∗∗
ΔLCPI −2 ∗∗LTBR −8.1.20301442491 −−81.1.8263767367
∗∗ −0.571958∗∗ −1.987544∗∗ −1.8.6103037545∗∗ −−109.5.5327315023∗∗ 00..245002 −8.11 4
849907∗∗ 00.0.23788830179∗∗
ΔLTBR ∗∗ ∗∗
Lag length Note: *1*4*,** and * indi1ca4tes significan1
54774∗∗ −8.361481∗∗ −8.380459∗∗ −8.376260∗∗ 0.121375∗∗ 0.043171∗∗
F(1u9ll9e2r )( 1T9e9st7) Test; PP = Phillips-Perrt oant  1(1%9,8 58%)  T aensdt; 1K0P%S Sle =ve Kl wreisaptkeoctwivsekliy, PAhDilFli p=s,A Sucghmmeindtte, Ad nDdi cSkheiyn-
92
University of Ghana  http://ugspace.ug.edu.gh
Appendix B: Residual Portmanteau VAR tests for Autocorrelations
VAR Residual Portmanteau Tests for Autocorrelations
Null Hypothesis: no residual autocorrelations up to lag h
Lags Q-Stat Prob. Adj Q-Stat Prob. df
1 0.192324 NA* 0.193364 NA* NA*
2 1.378357 NA* 1.392288 NA* NA*
3 2.712487 NA* 2.748289 NA* NA*
4 7.596342 0.1075 7.739482 0.1016 4
5 9.894491 0.2725 10.10112 0.2580 8
6 10.59409 0.5640 10.82403 0.5441 12
7 13.72015 0.6196 14.07235 0.5933 16
8 25.75831 0.1739 26.65154 0.1454 20
9 25.90585 0.3580 26.80659 0.3135 24
10 28.00334 0.4643 29.02325 0.4113 28
11 32.78800 0.4282 34.10867 0.3665 32
12 36.41691 0.4493 37.98784 0.3789 36
13 38.57358 0.5345 40.30657 0.4567 40
14 46.94196 0.3528 49.35610 0.2677 44
15 50.68471 0.3681 53.42716 0.2737 48
16 56.25098 0.3188 59.51731 0.2209 52
17 58.08464 0.3984 61.53542 0.2846 56
18 60.35956 0.4627 64.05408 0.3363 60
19 68.42211 0.3297 73.03393 0.2055 64
20 70.08058 0.4076 74.89221 0.2647 68
21 79.82521 0.2467 85.87708 0.1263 72
22 82.48333 0.2859 88.89178 0.1480 76
23 84.08806 0.3556 90.72294 0.1936 80
24 84.38716 0.4676 91.06635 0.2804 84
25 86.29376 0.5315 93.26900 0.3301 88
26 91.26422 0.5021 99.04716 0.2892 92
27 94.68894 0.5187 103.0534 0.2929 96
28 98.92050 0.5117 108.0349 0.2741 100
29 99.80070 0.5982 109.0777 0.3473 104
30 102.5998 0.6286 112.4150 0.3663 108
*The test is valid only for lags larger than the VAR lag order.
df is degrees of freedom for (approximate) chi-square distribution
*df and Prob. may not be valid for models with exogenous variables
93
University of Ghana  http://ugspace.ug.edu.gh
Appendix C: Impulse Response Function of the VAR model
Response to Cholesky One S.D. Innovations ± 2 S.E.
Panel A
Response of MDLTVOL to MDLTVOL
1.0
0.8
0.6
0.4
0.2
0.0
-0.2
1 2 3 4 5 6 7 8 9 10 11 12
Panel B Panel C
Response of MRET to MDLTVOL Response of MRET to MRET
.06 .06
.05 .05
.04 .04
.03 .03
.02 .02
.01 .01
.00 .00
-.01 -.01
1 2 3 4 5 6 7 8 9 10 11 12 1 2 3 4 5 6 7 8 9 10 11 12
94
University of Ghana  http://ugspace.ug.edu.gh
Appendix D: Pairwise Granger Causality tests for Returns and Volume (Lags: 3)
Null Hypothesis: Obs F-Statistic Prob.
MMRDETTV OdLo edso neos tn Gorta Gnrgaenrg Cear uCsaeu MseD MTVROETL 189 31..2202196362 00..03207389
95