EFFECTS OF OWNERSHIP STRUCTURE AND MONITORING
MECHANISMS ON EARNINGS QUALITY AND MARKET ASSESSMENT
by
RADZIAH ABDUL LATIFF
Thesis submitted in fulfillment of the requirements
for the degree of Doctor of Philosophy
April 2009
ACKNOWLEDGEMENTS
I am thankful for the study leave granted by Universiti Kebangsaan Malaysia together
with the financial support provided throughout my three and half years of study.
My sincere gratitude to my supervisor Associate Professor Dr Fauziah Md Taib for
agreeing to be my supervisor in the first place, amidst a trying time for her in many
aspects and her many academic commitments. Her intellectual inputs, patience and time
are highly appreciated.
I also appreciate the guidance of former Deputy Dean, Associate Professor Dr Zainal
Ariffin earlier on, and whose infectious enthusiasm is indeed inspiring. Special thanks
also to the existing Deputy Dean, Associate Professor Dr Zamri Ahmad whose openness
is indeed intellectually nurturing. The Dean, Associate Professor Dr Ishak Ismail has
also been a pillar of support for which I highly appreciate. I am indebted for the
valuable inputs of Dr Sofri Yahya and Associate Professor Datin Ruhani Hj Ali.
I am also grateful to other faculty members of both the School of Management,
Universiti Sains Malaysia and Faculty of Economics and Business, Universiti
Kebangsaan Malaysia who now and then provide guidance and pointers when I come to
some difficulties. I also take this opportunity to thank my friends, fellow Phd students
and colleagues whose encouragement, friendship and support I am truly blessed.
I dedicate this thesis to my family, especially my mother, Salasiah Abdul Hamid and my
late father, Abdul Latiff Muhammad for whom I am indebted for my very being.
ii
TABLE OF CONTENTS
Page ACKNOWLEDGEMENT iiTABLE OF CONTENTS iiiLIST OF TABLES viiLIST OF FIGURES xLIST OF APPENDICES xABSTRAK xiABSTRACT xiii CHAPTER 1 INTRODUCTION 1.1 Motivation of research
1
1.1.1 High standard of corporate governance (CG) and quality of information
3
1.1.2 Standard of corporate governance and quality of information in substance
4
1.1.3 Substantial shareholders 71.1.4 Earnings as a useful measure 101.1.5 A discerning market as enforcement agent 11
1.2. Problem statement 12
1.3. Research questions 14
1.4. Research objectives 15
1.5. Significance of study 15
1.5.1 Practical contributions 1.5.1.1 To the regulators 161.5.1.2 To market players 16
1.5.2 Methodical and theoretical contributions 17
1.6 Thesis outline
18
CHAPTER 2 LITERATURE REVIEW 2.0. Introduction
21
2.1. Earnings quality 22
2.1.1 Accounting based earnings attributes 232.1.2 Market based earnings attributes 262.1.3 The relevant earnings quality constructs 27
iii
2.2. Ownership structure, expropriation of non-controlling shareholders’ interest and earnings quality
2.2.1 Theoretical studies 292.2.2 Empirical studies 312.2.3 Consideration of the types of ultimate controlling party 352.2.4 Consideration of monitoring mechanisms – board structure,
substantial shareholders and audit committee 362.2.5 Justification for audit committee (AC) measurements 38
2.3. Earnings quality, information risks and market required return or assessment 2.3.1 Theoretical studies 402.3.2 Empirical studies 41
2.4. Market assessment or consequences of information quality 43
2.5. Endogeneity of Ownership Structure 46
2.6 Summary 46
CHAPTER 3 THEORETICAL FRAMEWORK AND HYPOTHESES DEVELOPMENT
3.0 Introduction
49
3.1 Relationship between ownership structure, monitoring mechanisms and earnings quality
50
3.2 Relationship between earnings quality and market assessment
53
3.3 Relationship between ownership structure, monitoring mechanisms and market assessment
55
3.4 Relationship that shows market assessment and the monitoring mechanisms could explain changes in ownership structure
58
CHAPTER 4 RESEARCH METHODOLOGY 4.0 Introduction
59
4.1. Population and sample 59
4.2 Variable definition and Measurement 60
4.2.1 Earnings quality constructs 4.2.1.1 Accruals quality 624.2.1.2 Time series properties- persistence and predictability 664.2.1.3 The non-discretionary determinants of earnings quality
68
iv
4.2.2 Ownership structure 694.2.2.1 Calculation of ownership structure variable 704.2.2.2 Type of ultimate controlling party 74
4.2.3 Audit committee characteristics 4.2.3.1 Independence 764.2.3.2 Competence 77
4.2.4 Substantial shareholders 774.2.5 Measures for the market assessment of earnings quality
4.2.5.1 Cost of equity 784.2.5.2 Market return 84
4.2.6 The validity of the cost of equity measures 85
4.3 Data Analysis
4.3.1 Theoretical/research framework 864.3.2 The simultaneity of equations 904.3.3 The choice of audit firms 93
CHAPTER 5 RESULTS 5.1 Data sources
94
5.2 Sample profile
94
5.3 General descriptive statistics
97
5.4 Descriptive statistics for ownership structure
102
5.5 Descriptive statistics for monitoring mechanisms – substantial shareholders’ voting rights and audit committee characteristics
104
5.6 Bivariate collinearity between EQ variables
105
5.7 Bivariate collinearity analysis of all variables in each sample
106
5.8 Construct validity of variables- COE/COEA , EQ and CFVR 5.8.1 Cost of equity 1125.8.2 Earnings quality 1125.8.3 Cash flow/voting rights 113
5.9. Multivariate analysis
5.9.1 Ownership structure and earnings quality 1155.9.2 Reestimation of equation 1 1245.9.3 Earnings quality and cost of equity 1315.9.5 Simultaneity test for equations 3 and 4 1445.9.6 Ownership structure, monitoring mechanisms and market
assessment
156
v
5.9.7 The relationship that examines whether market assessment and the
monitoring mechanisms could explain changes in ownership structure 156
5.9.8 Relationship between substantial shareholders’ voting rights and elements of ownership, monitoring mechanisms and cost of equity 159
5.9.9 Two stage least square of equations 3 and 4 1645.9.10 Comparisons of the ordinary least square results and two stage
least square results of equation 3 and 4 164
CHAPTER 6 DISCUSSION 6.0 Introduction- main findings 172
6.1 The ownership structure and earnings quality 6.1.1 Cash flow/voting rights and earnings quality 1746.1.2 Ultimate controlling party and earnings quality 1786.1.3 Monitoring mechanisms and earnings quality
6.1.3.1. Substantial shareholders voting rights and earnings quality 1816.1.3.2. Audit committee and earnings quality 183
6.2 Earnings quality and cost of equity
184
6.3 Cost of equity and market return 1856.4 The relationship between ownership structure and cost of equity
6.4.1 Cash flow/voting rights and cost of equity 1866.4.2 Ultimate controlling party and cost of equity 1876.4.3 Monitoring mechanisms and cost of equity
6.4.3.1. Substantial shareholders voting rights and cost of equity earnings quality 188
6.4.3.2. Audit committee and cost of equity earnings quality 1906.5 The relationship that examines whether market assessment and the
monitoring mechanisms could explain changes in ownership structure
190
6.6 The relationship between substantial shareholders’ voting rights and elements of ownership, monitoring mechanisms and cost of equity
191 CHAPTER 7 CONCLUSION 7.0 Introduction
193
7.1 Conclusion and contribution highlights
193
7.2 Implication
197
7.3 Limitations of study
198
vi
7.4 Future research 199
REFERENCES 201APPENDICES LIST OF TABLES
No. Title Page Table 1.1 ACP Industries Berhad- Analysis of Shareholdings as at
16 August 2004 7
Table 1.2 Glomac Berhad- Analysis of Shareholdings as at 30 June 2004
8
Table 1.3 Summary of motivation of study which leads to problem statement, research questions and objectives
20
Table 4.1 Variables brief description 60Table 4.2 MTD Capital Analysis of Shareholding 76Table 4.3 Calculation of cost of equity (COE) for ACP
INDUSTRIES 80
Table 5.1 Sample 95Table 5.2 Classification by industry 95Table 5.3 Sample size based on available data for the calculation
of earnings quality variables 96
Table 5.4 Breakdown of companies with various types of ownership
96
Table 5.5 Breakdown of companies with pyramidal(PYS) and non-pyramidal ownership(NON-PYS)
96
Table 5.6 Mean and dispersion of common variables in the three sample
98
Table 5.7 Descriptive Statistics - ABRES SAMPLE 99Table 5.8 Descriptive Statistics - ABSDATCA & ABSDATA
SAMPLE 100
Table 5.9 Descriptive Statistics for transformed ABSDATCA and ABSDATA i.e LABSCA and LABSTA
100
Table 5.10 Descriptive Statistics - PERS and PRED SAMPLE 101Table 5.11 Descriptive statistics of cash flow and voting rights and
ratio of cash flow to voting rights 103
Table 5.12 The descriptive statistics of substantial shareholders’ voting rights
104
Table 5.13 Pearson correlation coefficients between variables 106Table 5.14 Correlations - ABRES SAMPLE 109Table 5.15 Correlations- ABSDATCA and ABSDATA Sample 110Table 5.16 Correlations- PERS PRED Sample 111Table 5.17 Correlations between CF, VR and LGMV in full sample 114Table 5.18 Correlations between CF, VR and LGMV among PYS
companies 114
Table 5.19 Correlations between CF, VR and LGMV among NON-PYS companies
115
Table 5.20 Coefficients of equation 1 regression 118
Table 5.20 (a) Results of equation 1 regression for earnings quality 119
vii
ABRES
Table 5.20 (b) Results of equation 1 regression for earnings quality ABSDATA
120
Table 5.20 (c) Results of equation 1 regression for earnings quality ABSDATCA
121
Table 5.20 (d) Results of equation 1 regression for earnings quality PERS
122
Table 5.20 (e) Results of equation 1 regression for earnings quality PRED
123
Table 5.20 (f) Pyramidal companies in ABRES sample 126
Table 5.20 (g) Non-pyramidal companies in ABRES sample 127
Table 5.20 (h) Pyramidal companies in ABSDATA sample 127
Table 5.20 (i) Non-pyramidal companies in ABSDATA sample 128
Table 5.20 (j) Pyramidal companies in ABSDATCA sample 128
Table 5.20 (k) Non-pyramidal companies in ABSDATCA sample 129
Table 5.20 (l) Pyramidal companies in PERS sample 129
Table 5.20 (m) Non-pyramidal companies in PERS sample 130
Table 5.20 (n) Pyramidal companies in PRED sample 130
Table 5.20 (o) Non-pyramidal companies in PRED sample 131
Table 5.21 (I) Results of equation 2 regression using COE estimate 133
Table 5.21 (II) Results of equation 2 regression using COEA estimate 134
Table 5.21 (a) (i) Results of equation 2 regression - ABRES and COE estimate 135
Table 5.21 (a) (ii) Results of equation 2 regression - ABRES and COEA estimate 135
Table 5.21 (b) (i) Results of equation 2 regression- ABSDATA and COE estimate 136
Table 5.21 (b) (ii) Results of equation 2 regression- ABSDATA and COEA estimate 136
Table 5.21 (c) (i) Results of equation 2 regression - ABSDATCA and COE estimate 137
Table 5.21 (c) (ii) Results of equation 2 regression - ABSDATCA and COEA estimate 137
Table 5.21 (d) (i) Results of equation 2 regression - PERS and COE estimate 138
Table 5.21 (d) (ii) Results of equation 2 regression - PERS and COEA estimate
138
Table 5.21 (e) (i) Results of equation 2 regression - PRED and COE
viii
estimate 139
Table 5.21 (e) (ii) Results of equation 2 regression - PRED and COEA estimate 139
Table 5.22(a) Results of testing earnings quality (ABRES) and excess return
141
Table 5.22(b) Results of testing earnings quality (ABSDATCA) and excess return
141
Table 5.22(c) Results of testing earnings quality (ABSDATA) and excess return
142
Table 5.22(d) Results of testing earnings quality (PERS) and excess return
142
Table 5.22(e) Results of testing earnings quality (PRED) and excess return
143
Table 5.23 Pearson Correlation of cost of equity and return
143
Table 5.23 (a) ABRES sample- Results of estimating equation to test the exogenuity of CFVR and SSVR (with COE)
146
Table 5.23 (b) Testing for coefficients of ^CFVR and ^SSVR 146
Table 5.23 (c) ABRES sample- Results of estimating equation to test the exogenuity of CFVR and SSVR (with COEA)
147
Table 5.23 (d) Testing for coefficients of ^CFVR and ^SSVR
147
Table 5.24 (a) ABSDATA/ABSDATCA sample- Results of estimating equation to test the exogeneity of CFVR and SSVR (with COE)
148
Table 5.24 (b) Testing for coefficients of ^CFVR and ^SSVR 148Table 5.24 (c) ABSDATA/ABSDATCA sample- Results of estimating
equation to test the exogeneity of CFVR and SSVR (with COEA) 149
Table 5.24 (d) Testing for coefficients of ^CFVR and ^SSVR
149
Table 5.25 (a) PERS/PRED sample- Results of estimating equation to test the exogenuity of CFVR and SSVR (with COE)
150
Table 5.25 (b) Testing for coefficients of ^CFVR and ^SSVR
150
Table 5.25 (c) PERS/PRED sample- Results of estimating equation to test the exogenuity of CFVR and SSVR (with COEA) 151
Table 5.25 (d) Testing for coefficients of ^CFVR and ^SSVR
151
Table 5.26 (a) Results of equation 3 regression using COE estimate 154
Table 5.26(b) Results of equation 3 regression using COEA estimate
ix
155Table 5.27 (a) Results of equation 4 regression using COE estimates
157Table 5.27 (b) Results of equation 4 regression using COEA estimates
158
Table 5.28 (a) (i) ABRES Sample (with COE) 161Table 5.28 (a) (i) ABRES Sample (with COEA) 161Table 5.28 (b) (i) ABSDATA Sample (with COE) 162Table 5.28 (b) (ii) ABSDATA Sample (with COEA) 162Table 5.28 (c) (i) PERS/PRED Sample (with COE) 163Table 5.28 (c) (ii) PERS/PRED Sample (with COEA) 163Table 5.29 (a) ABRES Sample 166Table 5.29 (b) ABSDATA/ABSDATCA Sample 166Table 5.30(a) ABRES sample 167Table 5.30(b) ABSDATA/ABSDATCA sample 167Table 5.4 Summary of results 168Table 6.1 Percentage of companies whose controlling party is also
in an executive position
179
LIST OF FIGURES
No. Title Page Figure 1.1 Market-based regulatory environment 2Figure 3.1 Theoretical Framework 49
Figure 4.3.1 Theoretical/ research framework 86
LIST OF APPENDICES Appendix 1 Annual reports of APM Automotive Holdings Bhd Appendix 2 Annual reports of ACP Industries Berhad and other related
companies Appendix 3 List of companies in sample Appendix 4 SPSS Output (Equation 1) Appendix 5 Eviews output
(Equation 1 regression separating companies that are controlled by ultimate controlling party and those that are not)
Appendix 6 Eviews output (Equation 1 regression for family controlled companies separating those are family managed and those that are not)
x
KESAN STRUKTUR PEMILIKAN DAN MEKANISMA PENGAWASAN
TERHADAP KUALITI PEROLEHAN DAN PENAKSIRAN PASARAN
ABSTRAK
Tesis ini didorong oleh peralihan ke arah kawalan berdasarkan pasaran atau kawalan
kendiri bagi pasaran modal Malaysia. Di dalam persekitaran demikian, kualiti maklumat
adalah penting. Dengan berlatar belakangkan struktur pemilikan syarikat yang dikatakan
memburukkan konflik pengasingan pemilikan dan kawalan, dan yang berkemungkinan
menghadkan maklumat kepada pihak awam, tesis ini mengkaji sama ada struktur
pemilikan sedemikian membawa kepada kualiti perolehan yang rendah. Perolehan
adalah maklumat yang penting kepada pasaran. Dan sekiranya pasaran benar-benar
berkawalan kendiri, tesis ini mengkaji samaada pasaran menaksir kualiti perolehan dan
elemen tadbir urus; struktur pemilikan dan mekanisma pengawasan (jawatankuasa audit
dan pegangan pemegang saham utama). Pasaran menaksir elemen tersebut dengan
menghendaki pulangan tertentu, iaitu kos ekuiti, di mana elemen tersebut ditanggap
sebagai risiko maklumat. Kajian ini berdasarkan satu sampel syarikat tersenarai bagi
tahun perakaunan berakhir 2004. Ukuran kualiti perolehan yang digunakan ialah kualiti
akruan, akruan terpilih, keberterusan dan kebolehramalan. Kajian mendapati pegangan
pemegang saham utama iaitu satu mekanisma pasaran, berkait secara signifikan dengan
kualiti perolehan terpilih di mana ini bermakna yang pemegang saham utama adalah
mekanisma pengawasan yang penting. Ini berbeza dengan keputusan berhubung
mekanisma perundangan iaitu jawatankuasa audit. Tiada satu ciri jawatankuasa
juruaudit (kebebasan dan kecekapan) berkait secara signifikan dengan mana-mana
ukuran kualiti perolehan dan juga ciri tersebut tidak dinilai. Ini memberi implikasi
penting terhadap perbelanjaan sumber secara relatif terhadap mekanisma perundangan
dan mekanisma pasaran. Penemuan bahawa pegangan pemegang saham utama dinilai
xi
merupakan sumbangan penting kerana ianya memberi makna yang pegangan pemegang
saham utama meningkatkan aliran maklumat empunya kepada pihak awam dan
seterusnya mengurangkan risiko maklumat. Walau bagaimanapun kajian ini tidak
mendapat bukti yang mengaitkan hak aliran tunai/ mengundi dan jenis pihak mengawal
(keluarga, kerajaan, institusi, syarikat dan pengurusan) dengan kualiti perolehan dan kos
ekuiti. Kajian ini menyumbang bukti baru di Malaysia yang menunjukkan kualiti
perolehan mempengaruhi kos ekuiti. Keputusan berhubung kualiti akruan dan
keberterusan adalah konsisten dengan kedua-dua ukuran kos ekuiti. Akruan terpilih
adalah berkait secara signifikan dengan ukuran kos ekuiti. Implikasi penting ialah
syarikat mungkin mencapai objektif tertentu dengan melakukan aktiviti yang
menurunkan kualiti perolehan, tetapi syarikat terpaksa membayarnya dalam bentuk kos
ekuiti yang tinggi.
xii
EFFECTS OF OWNERSHIP STRUCTURE AND MONITORING
MECHANISMS ON EARNINGS QUALITY AND MARKET ASSESSMENT
ABSTRACT
This thesis is motivated by the move towards a market-based regulation or self-
regulation for the Malaysian capital market. In such environment the quality of
information is important. Against a background of companies’ ownership structure that
allegedly exacerbates the separation of ownership and control conflict, and that possibly
limits transparency of information to the public, this thesis examines if such ownership
structure leads to lower earnings quality. Earnings ia an important information to the
market. And if indeed the market is self-regulating this thesis examines if the market is
assessing earnings quality and the elements of governance; ownership structure and the
monitoring mechanisms (audit committee and substantial shareholding). The market
assesses these elements by requiring a certain return, the cost of equity, where
accordingly these elements are perceived to be an information risk. This study is based
on a sample of listed companies for the accounting year end 2004. The earnings quality
measures used are accrual quality, discretionary accruals, persistence and predictability.
It is found that substantial shareholding, a market mechanism, to be significantly
associated with the discretionary earnings quality which suggests the substantial
shareholders is an important monitoring mechanism. This is in contrast to the results of
a rule based mechanism, audit committee. None of the characteristics of audit
committee (independence and competence) is significantly associated with all measures
of earnings quality and neither are they priced. This has an important implication on the
relative spending of resources by regulators on market and rule based mechanisms. The
finding that substantial shareholding is priced is a significant contribution as it suggests
that substantial shareholding is a mechanism that increases proprietary information flow
xiii
to the public and hence reduces information risk. This study, however has not found any
evidence that relates cash flow /voting rights and the type of controlling party (family,
government, institution, company and management) with earnings quality and the cost
of equity. This study contributes new evidence in Malaysia that earnings quality
influences cost of equity. The results for accruals quality and predictability are
consistent across the two measures of cost of equity. Discretionary accruals are
significantly associated with one measure of cost of equity. An important implication is
that companies may achieve their objectives by engaging in activities that lead to lower
earnings quality, but they stand to pay a higher price in the form of higher cost of
equity.
xiv
CHAPTER 1
INTRODUCTION
1.1 Motivation of research
The Asian financial crisis has been claimed to be the wake-up call for corporate
governance (CG) reform in the Asian region. In response to the crisis, the Malaysian
regulators have taken a different approach to regulation by placing the responsibility of
valuing the companies in the hands of market players. Under the market based valuation
where the disclosure based regime operates, the regulator no longer assesses the merits
and worth of corporate proposal namely in security offerings and issuance (Securities
Commission 1999).
This market-based approach calls for the need for high quality disclosures/ financial
reporting and high standard of corporate governance (Securities Commission 1999,
Securities Commission Annual Report 2002). The role of regulators is to set standards
to meet this need. Figure 1.1 depicts this market based approach.
Whilst certain structural elements of CG such as corporate ownership and control are
the product of the socio-economic development and government policies, in this
reformed environment with high quality of information market players are expected to
be able to discern good governance in form and substance and make assessment
accordingly. In the words of Emeritus Professor Mohamed Ariff,
‘Good corporate governance is more than a check list of dos and don’ts.
It is essentially an infrastructure of built-in checks and balances. Good
1
governance is not confined to the top layer of the corporate hierarchy, as
governance and processes are intricately linked.’ (Emeritus Professor
Mohamed Ariff at http://www.mier.org.my/mierscan/ -‘Banking on
Corporate Governance’ 25 March 2005)
Regulators set standards
Corporate Governance mechanisms
Companies
High Quality Information Market
Players’ Assessment
Pricing
Figure 1.1 Market-based regulatory environment
The fact that market penalizes and rewards, or assesses companies for poor or good
governance, the market consequences of governance, is part of this checks and balances.
This research is motivated by this development into market based approach to
regulation which is although new but has been evolving.
2
1.1.1 High standard of corporate governance (CG) and quality of information
imperatives for effective market-based regulation
Substantial effort to improve reporting and CG practices is evidenced from the many
guidelines and rules for best practices established (for example Malaysian Code of
Corporate Governance (MCCG) (2000)), and laws enacted (Financial Reporting Act
1997, amendments to securities and companies law (2000)). A number of CG
mechanisms has long been adopted such as the rules governing independent directors
(1987) and audit committees (1993) (SC web page). These rules were subsequently
enhanced by the revamping of Exchange Listing Requirements in 2001. This major
revamp among others includes disclosure of the extent to which companies comply with
the MCCG.
Rules to protect investors and to promote transparency in ownership were also enhanced
to include for example rules regarding market manipulation, false and misleading
information, prohibition from hiding behind nominees (Securities Law) and one-share-
one-vote (s55 Companies act 1965).
In the area of financial reporting, both securities law and companies act have
incorporated requirements to comply with standards produced by Malaysian Accounting
Standards Board. Since August 1999, the exchange listing requirements provide for
companies to report financial information (which include income and cash flow
statements, balance sheet and explanatory notes) every quarter. The Bursa Malaysia
regularly investigates variances between these unaudited reported results and the year
end audited results.
3
These requirements together supposedly ensure that those who are in control of
companies act in the interest of all shareholders and that information that are made
available to the market actually reflect the economic performance of the companies. The
market players act accordingly through the pricing mechanism and thus ensure efficient
resource allocation. In other words, market players use disclosed information to assess
companies by requiring high return for high risk companies. Thus poor quality
information distort this risk assessment and market players may make wrong investment
decisions.
1.1.2 Standard of corporate governance and quality of information in substance
Mere compliance to disclosure requirements and corporate governance practices
guidelines, that is compliance in form, does not necessarily ensure that those who are in
control of corporate decisions do not in substance, subvert the intent and spirit of those
rules and guidelines. There are still practices that are not covered by the rules and
especially in financial reporting there is still room for managerial discretion. It has been
reported that each of the companies in the United States that was involved in fraudulent
scandals such as Enron, Tyco and Disney was in full compliance with the standards for
corporate governance related to the board of directors (BOD) and audit committee set
by the subsequently enacted Sarbanes Oxley Act (Pergola 2005). Most of them were
audited by one of the Big 4 (or 5 then) auditors.
Although not to the same scale as the scandals in the US market, the Malaysian market
is not short of improper practices even with the strengthening of the regulations
4
described above and with increased surveillance. The SC Annual Report 2003 for
example cited incidences of assets acquisitions and disposals at questionable prices. It
is also reported that companies create debts to offset contractual obligation. There are
also questionable transactions detected by the SC such as the acquisition for cash of a
private company that was subsequently disposed as a settlement of fictitious debt, the
use of money-lending licenses for what was purportedly in the ‘ordinary course of
business’, and the creation of a liability for a company that originated from private loan
arrangement between individuals. The findings reported in the SC annual report
emerged from targeted surveillance, where the SC focused on certain activities and
reporting standards. How widespread such activities among the listed companies is an
important empirical question. And of equal importance is the question of how pervasive
the practices that are although legal, not in pursuance of shareholders wealth
maximization.
The incidences of improper and unethical practices possibly signal a failure of the
various CG mechanisms to reduce information asymmetry and align interest of those in
control with other shareholders. Some of the mechanisms are to enforce independence
of BOD and audit committee. However independence in form does not necessarily
ensure independence in substance. This is so as independent directors could be
associated with the company or the chief executive officer in ways that are too subtle to
be captured by the rules and regulations.
Another possible reason which has not been given sufficient attention by researchers is
the ability to expropriate funds at relatively less cost to the perpetrator as the disparity
between cash flow rights (associated with ownership) and voting rights (associated with
5
control) of the perpetrator widens. This happens when concentration of control is
achieved through shareholding of multiple layers of companies, thus the term pyramidal
structure (this is explained in detail in Chapter 3).
To have the controlling votes is important from the perspective of corporate
governance. The owner has an influence over decisions such as dividend payments,
appointment of key management personnel, etc. Thus Mr. Zee needs relatively small
capital outlay to control PQR and could expropriate funds from the company as a
controlling shareholder with relatively small cash consequences as owner.
Samples of companies taken in past studies by Claessens, Djankov and Lang (2000) and
Fan and Wong (2002) indicate the existence of such control. Though s55 of Companies
Act 1965 requires one share to have one vote to prevent such disparity, in substance in
pyramidal structure such disparity exists.
This form of ownership concentration is not so apparent by cursory study of substantial
shareholders disclosure. Unlike in the US where companies are allowed to issue shares
that carry more than one vote, such disparity is transparent. Thus the shareholders who
hold the inferior shares with one vote each can discount the price of shares accordingly
knowing the voting power of the other class of shares (Francis, Schipper & Vincent
2005). However in Malaysia where it is not so transparent, therefore the non-controlling
shareholders not-knowing the existence of such control would not be able to do so,
hence the pricing mechanism for efficient resource allocation breaks down.
6
1.1.3 Substantial shareholders
A mechanism that may reduce moral hazard faced by non-controlling shareholders is
the existence of substantial shareholders i.e shareholders who own more than 5%.
According to Kaplan and Minton (1994), Pound (1988) and Shleifer and Vishney
(1986) a substantial shareholder has a role in controlling agency problems by actively
monitoring the controlling party who in a widely held company, is the management.
Similar role could be played by substantial shareholders in companies with concentrated
ownership.
In Malaysia it is common for companies to be held by a few substantial shareholders
with shareholdings far higher than the threshold 5%, instead of just one substantial
shareholder with the majority controlling rights, even though one may be with the
highest shareholding and the apparent controlling party,. The following extracts from
annual reports of ACP Industries Berhad and Glomac Berhad illustrate this type of
ownership concentration.
Table 1.1 ACP Industries Berhad- Analysis of Shareholdings as at 16 August 2004
Direct Interest
Indirect Interest Shareholders
Number of Shares
%
Number of Shares
%
Metacorp Berhad 38,734,790 29.02 MTD Capital Berhad 38,734,790 29.02 Lambang Simfoni Sdn Bhd 38,734,790 29.02 Employees Provident Fund Board 20,499,000 15.36
7
Metacorp Berhad, MTD Capital Berhad and Lambang Simfoni Sdn Bhd are companies
under the control of Dato’ Dr Nik Hussain Abdul Rahman and his family members. It is
fairly obvious that Dato’ Dr Nik Hussain and family are the controlling shareholder.
However, the Employees Provident Fund Board with shareholding of around 15% could
play a significant role in monitoring the controlling party actions.
Table 1.2 Glomac Berhad- Analysis of Shareholdings as at 30 June 2004
Direct Interest
Indirect Interest
Shareholders
Number of Shares
%
Number of Shares
%
Dato’ Mohamed Mansor Fateh Din
63,552,183 29.33
Datuk Fong Loong Tuck 47,404,490 21.88 Employees Provident Fund Board 14,456,590 6.67 ( Three other foreign companies with lesser shareholdings)
Similarly for Glomac Berhad, the substantial shareholder Datuk Fong may be able to
play a role in the checks and balance process assuming as apparent that Dato’ Mohamed
Mansor and Datuk Fong are not related.
In the West where aggressive takeovers bids are common, the existence of substantial
shareholders may control manager’s behavior as they have the ability to remove non-
performing managers by facilitating takeovers. Thus the effectiveness of the other
substantial shareholders hinges on whether such ability exists. According to Mak and Li
(2001) in economies such as Malaysia, hostile takeovers are rare and conflicts are
resolved through non-confrontational methods. Besides it would be easier to remove
non-performing managers than the other majority and controlling shareholders.
8
However the experience of KFC Holdings boardroom tussle in 2004/2005, where it was
reported that the move to remove the board executive chairman was initiated by a
substantial shareholder, indicated that perhaps the situation is changing (NST Business
Times, 20 May 2005, pg 1). Thus the effectiveness of the substantial shareholders in
monitoring the controlling party is an open question. The same arguments apply for
institutional investors such as the Kumpulan Wang Simpanan Pekerja (KWSP). The
fact that institutional investors played a major role in setting up the Minority
Shareholders Watchdog, suggests that institutional investors on their own may not serve
to be an effective monitoring mechanism.
Another factor which makes the effectiveness of substantial shareholders as a control
mechanism questionable in Malaysia is the potential alignment of interest between
substantial shareholders and the manager or the controlling party through kinship,
social or economic relationship (Lim 1981) and through political or governmental
affiliation (Gomez & Jomo 1999, Gomez 2002). Thus a substantial shareholder in a
company may share the same objectives as the controlling party resulting in a
potentially cohesive control to the detriment of other shareholders.
The role of substantial shareholders is an important and interesting area to consider as
the monitoring and governance effect if any is inherently non-legal or not imposed by
rules or standards unlike monitoring mechanisms such as audit committee and board of
directors. Thus the results would shed light on the relative effectiveness of a rule based
mechanism such as the audit committee and the board of directors and a market based
mechanism, such as a significant shareholding of a shareholder other than the
controlling shareholder. In the Malaysian capital market. in the context of the market
9
based approach to regulation as described earlier, the relevant regulator has a significant
role in setting standards and rules with regards to governance and quality of disclosure.
This does not preclude the development of mechanisms through market forces such as
the substantial shareholding. The substantial shareholder may play a role in aligning the
interest of the controlling party and other shareholders including other non-controlling
shareholders.
It is a conjecture at this stage that improper practices by the controlling party which
escape the law and other non-legal or market based mechanisms such as the substantial
shareholders, potentially undermine the credibility of reported results namely the
earnings figure. Even though financial accounting rules are extensive there is still room
for the exercise of judgment and discretion. Thus activities related to the expropriation
of non-controlling shareholders’ wealth, to management entrenchment and to
manipulation of accounts without legitimate underlying economic activity could be
hidden behind reported earnings numbers.
1.1.4 Earnings as a useful measure
It is expected that market players use a repertoire of measures and information from
various sources. However earnings is a summary measure widely used (Francis,
LaFond, Olsson & Schipper 2003, Liu, Nissim & Thomas 2002)). It reflects aggregate
effects of accounting policy choice made. A survey by Price Waterhouse (2000) found
that majority of chief executive officers believed price/earnings ratio was still relevant
for market valuation. In addition, anecdotal evidence indicates that market players
respond to earnings figure. It was reported that there were unusual share price
10
movements and dealings of Goh Bah Huat Berhad when it reported earnings of RM100
million in its 31 December 2004 quarterly results, which was later discovered to be
erroneous and turned out to be a loss of RM21 million. Apparently some market players
have acted on the erroneous profit. Thus for the proper functioning of the market, the
state of earnings quality in Malaysia and whether they reflect undesirable elements
described above are important empirical questions.
1.1.5 A discerning market as enforcement agent
In a market-based regulatory environment, market players play an important role in
enforcement through the pricing mechanism and thus ensure efficient resource
allocation. Investors must be able to reinforce proper conduct in companies by
rewarding or punishing appropriately, in technical terms by requiring higher rate of
return for high risk companies and vice versa. Certainly the rules and guidelines are
substantial enough to ensure high quality information to be available to the market for
investors to act accordingly. However for this approach to regulation to be effective, the
market players must be able to evaluate beyond the disclosed information.
For example in the case of Goh Ban Huat Bhd cited above, market responded to the
huge earnings announced without discerning the error impounded in the figure. It is not
that there has not been other indicators to doubt the figure such as past quarterly losses,
but the market players did not interpret cautiously the earnings figure. On the other hand
Mitton (2000) interpreted market reaction to the purchase of Renong shares by United
Engineers Malaysia at an inflated price as a penalty for bailing out the troubled parent
company. Market could see through the expropriation of other shareholders interest and
11
UEM share price dropped by 38% on the day the purchase was announced (NST 19
Nov 1997, pg 62).
Given the contradicting observations, it is therefore important to examine whether
market players in the Malaysian capital market appropriately prices, if at all they do,
earnings quality by requiring higher return from companies with lower earnings quality
and vice versa.
1.2 Problem Statement
Theoretical analyses (Berle & Means 1932, Jensen & Meckling 1976) have established
the moral hazard problems associated with information asymmetry when there is a
separation between ownership and control. In particular, the controlling party has an
incentive to expropriate company’s resources and to take actions that may be in
divergent to the interest of the other party, who have no access to information in order
to detect and monitor such practice. The separation of control and ownership is
particularly aggravated when the controlling party can further enhance control through
pyramid ownership structure, when there is concentration of ownership or with the
existence of shareholders that could exercise control by virtue of these shareholders
relationship with the controlling party.
Since rules and regulations, and other non-legal monitoring mechanism cannot firewall
completely improper practices as described in preceding paragraph, it is reasonable to
expect the higher the degree of separation of ownership and control, the greater the
likelihood of such improper practices. The improper practices are potentially manifested
12
in earnings which then result in low earnings quality. Unlike in a clinical and case by
case study, in an empirical study such improper practices are not easily observable.
Thus earnings quality is a proxy to the likelihood of improper practices.
Given the considerable amount of effort and resources that have been spent on putting
in place rules and standards for good corporate governance, it is not only important to
examine if good governance characteristics are associated with high earnings quality
and vice versa, it is also important to examine if the capital market is pricing correctly
the companies based on the earnings quality.
Thus the purpose of this research is to examine the relationship between the extent of
separation of ownership and control in Malaysian listed companies, together with the
rule based and market based mechanisms, and earnings quality. Further, drawing from a
theoretical assertion that information risk is priced, this study will determine if the
capital market rewards or penalizes companies for the companies’ quality of earnings
through required return or cost of equity. In essence this is a departure from traditional
theory that only systematic risk is priced. Any idiosyncratic such as information risk
arises from each company unique circumstances or company specific and can therefore
be diversified away. As in previous researches, this study characterizes earnings quality
as information risk. Low earnings quality poses a risk as investors cannot rely on
earnings information to make investments decision and accordingly affects cost of
equity.
The characteristics associated with the separation of ownership and control poses
information risk as the controlling party is privy to more information. Drawing parallel
13
to the original work that characterizes information asymmetry between informed and
uninformed investors as information risk, it is here characterized that information
asymmetry between the controlling party and other investors as information risk. Thus
this study examines if capital market rewards or penalizes companies, assesses
companies with characteristics associated with separation of ownership and control. The
association between the monitoring mechanisms, audit committee and substantial
shareholders, with market assessment is also examined to see if these monitoring
mechanisms is priced and therefore perceived as effective in reducing the information
risk.
1.3 Research Questions
Against such background, this study seeks answers to the following questions:
a) What is the nature of the separation of control and ownership amongst
Malaysian listed companies?
b) What is the state of earnings quality amongst Malaysian companies?
c) Does the separation of ownership and control, in the presence of CG
mechanisms, and the alleged potential conflict between controlling and non-
controlling managers/shareholders manifest itself in earnings quality?
d) Do investors price accordingly the information risk poses by the quality of
earnings?
e) Does the separation of ownership, in the presence of CG mechanisms, affect
market assessment, i.e. is priced? Does the market assessment in turn affect
the separation of ownership and control?
14
1.4 Objectives of Research
The main objectives of the study are:
a) to determine the extent of separation of control and ownership by examining
ownership structures of Malaysian listed companies from simple structure
inducing manager-shareholder conflict to a more complicated pyramidal
structures that induces controlling- non-controlling shareholders conflict,
b) to determine whether the degree of separation of control and ownership in
the presence of CG mechanisms, has an influence over the quality of
earnings,
c) to examine whether investors penalize or reward accordingly companies for
low or high quality earnings, through the required return measure or cost of
equity measure, and
d) to examine whether the degree of separation of control and ownership in the
presence of CG mechanisms, affect the required return or cost of equity.
1.5 Significance of Study
This study, as described earlier, is motivated by the development in the Malaysian
capital market towards market based regulation where self-regulation by market players
is an expected feature together with the active involvement of regulators in terms of
setting rules and standards. Thus this study is therefore timely and contributes
significantly towards understanding of a self-regulation aspect of the market and that is
the market assessment of earnings quality and of the various governance mechanisms.
The contribution of this study is towards understanding of the market consequences of
15
earnings information and governance mechanisms which unlike the contribution from
many previous researches that examine the determinants of earnings quality which
include the governance mechanisms themselves. The following describes the
contribution of study from different aspects.
1.5.1 Practical contribution
1.5.1.1.To the regulators
Regulators’ investigation is ‘clinical’ and targeted at certain area. Since this is a study of
market behavior, thus the market wide effect of the conflict of interest between
controlling and non-controlling parties on earnings quality, and the market assessment
of it could be understood better. Market based study determines the significance of the
relationship and extent of the problem.
It is also important that regulators are informed that the public resources spent on rules
and regulations are effective and that the market perceived them as such. Otherwise it is
best left to market forces and more resources are spent to ensure the market forces are
working well.
1.5.1.2. To market players
In the market-based approach as earlier mentioned, market players play a major role in
the price discovery. But market players must use the information and must know how.
The SC and BM have stressed on the need for investors education. This research
increase awareness of the potential conflicts brought about by control achieved through
16
pyramidal structure, the way earning figures, in substance should be read and the
potential mispricing if low quality earnings is not read as such.
1.5.2 Methodological and theoretical contribution
1.5.2.1 By including the different corporate governance mechanisms and not just the
ownership structure, this study examines the relative significance of the different
corporate governance mechanisms. These different mechanisms is viewed here as rule
based or imposed by rules and regulations such as the audit committee and one that
emerges from market forces or market based such as substantial shareholding.
1.5.2.2. This study attempt to measure the information risks poses by the CG
mechanisms and ownership structure themselves. Whilst previous research characterize
information risks as the imprecision in the information as reflected by the quality of
earnings, this study attempt to characterize information risk as the relative amount of
information that is kept private and made public.
The study contributes to the understanding of whether the capital market penalizes or
rewards companies with certain ownership and earnings characteristics by requiring a
higher or lower rate of return. The required return or cost of equity is an important input
into financing and investment decisions.
1.5.2.3. Previous studies on ownership structure and other variables are carried out
mainly in well developed economies or at regional level. Findings that explain well
developed economies may not necessarily be applicable in a less developed economies.
17
Studies at a regional level such as Asia may fail to capture unique characteristics of
specific country that explain differences across companies in that country. This study
look at cross-company differences of ownership structure, corporate governance
mechanisms, earnings quality and cost of equity in Malaysia given the unique
characteristics of Malaysian business environment in which ownership is known to be
concentrated and where there is a suggestion that there is a weak market for corporate
control and the role of the other substantial shareholders is relatively less researched.
1.5.2.4. A significant contribution of this research, in the context of the theory
associating information risk and required return, is from the examination of the
substantial shareholding. the associations between substantial shareholding and each of
earnings quality and required return have never been examined. In this context the
results contribute towards the theory by establishing the substantial share holding role in
increasing the precision of information and the flow of information from the private to
public domain.
1.6 Thesis Outline
The introduction in this chapter is followed by Chapter 2. Chapter 2 reviews literature to
derive theoretical justifications for the relationships that are examined in the thesis and
to establish the extent of empirical studies that have already been carried out. This
chapter ends by laying out the theoretical and research framework , and the hypotheses
thus developed.
18
19
Chapter 3 lays out the relationships under study and develops hypotheses by drawing
from the literature on theoretical and empirical studies reviewed in Chapter 2.
Chapter 4 describes in detail the sample and variables. Where relevant examples are
given to illustrate how a particular variable is measured. The justification for a chosen
measure from alternatives of measures is also given. Finally the equations representing
the relationships examined are laid out.
Chapter 5 presents the results and brief analysis of the descriptive, bivariate and
multivariate analysis of each of the relationships examined. Each hypothesis presented
is tested. Where appropriate comparisons are made with findings of previous researches.
Chapter 6 discusses the results, provides explanation and where relevant justifications
for findings. The discussion draws Important similarities or differences in the findings
from previous researches.
Chapter 7 concludes, highlights significant contribution, provides implications of the
findings and discusses limitations of the research.. This chapter ends with some
direction for future research.
Table 1.3 Summary of motivation of study which leads to problem statement, research questions and objectives
MOTIVATION OF RESEARCH PROBLEM STATEMENT RESEARCH QUESTIONS RESEARCH OBJECTIVES 1. The move towards disclosure/market based regulatory environment – requires high quality of information and corporate governance – requires market to make assessment. 1.1.1 There has been many rules, regulations etc to improve CG and quality of information, 1.1.2 but in substance CG and information may not be reliable due to nature of ownership in Malaysia (pyramidal structure ,etc) that exacerbates the separation of ownership and control conflict. 1.1.3 The substantial shareholder’s ambiguous role, another feature of ownership needs to be examined 1.1.4. Market uses many information but earnings very useful thus a useful measure/ a proxy of information quality in the market based regulation 1.1.5. In the market based environment, also important that market prices accordingly information specifically earnings and elements of governance.
Theoretical analyses establish the relationship between ownership structure (separation of ownership and control) and improper practices even with rules, regulations, etc. Improper practices lead to poor earnings quality, thus need to examine whether ownership structure is associated with earnings quality, and whether market prices earnings quality. Ownership structure poses information risk in terms of the proportion of information in public/private domain. Thus research also examines if market prices ownership structure together with the monitoring mechanisms.
a) What is the nature of the separation of control and ownership amongst Malaysian listed companies?
b) What is the state of earnings quality amongst Malaysian companies?
c) Does the separation of ownership and control, in the presence of CG mechanisms, and the alleged potential conflict between controlling and non-controlling managers/shareholders manifest itself in earnings quality?
d) Do investors price accordingly the information risk poses by the quality of earnings?
e) Does the separation of ownership, in the presence of CG mechanisms, affect market assessment, i.e is priced? Does the market assessment in turn affect the separation of ownership and control?
a) To determine the extent of separation of control and ownership by examining ownership structures of Malaysian listed companies from simple structure inducing manager-shareholder conflict to a more complicated pyramidal structures that induces controlling- non-controlling shareholders conflict,
b) To determine whether the degree of separation of control and ownership in the presence of CG mechanisms, has an influence over the quality of earnings,
c) To examine whether investors penalize or reward accordingly companies for low or high quality earnings, through the required return measure or cost of equity measure, and
d) To examine whether the degree of separation of control and ownership in the presence of CG mechanisms, affect the required return or cost of equity.
20
CHAPTER 2
LITERATURE REVIEW
2.0 Introduction
This chapter reviews literature to derive theoretical justifications for the relationships
that are examined in the thesis and to establish the extent of empirical studies that have
already been carried out.
Part 2.1 first reviews literature on the nature and measure of earnings quality and
provides justification for the chosen measures in view of the relationships that are being
examined.
To link the ownership structure and the monitoring mechanisms that are examined to
earnings quality, part 2.2 described the literature that constitutes the body of knowledge
related to agency theory and the information asymmetry problems when there is a
separation of ownership and control. Empirical research is also reviewed to justify the
measures used in the separation of ownership and control construct. The measures are
the cash flow/ voting rights and the type of controlling party. This section also
establishes the theoretical justification for using two measures, independence and
competence, for the audit committee and the role of audit committee and substantial
shareholders in information asymmetry related problems.
Part 2.2 establishes the expectation of association between cash flow/ voting rights, the
type of controlling party, audit committee characteristics and substantial shareholder
with earnings quality.
21
Part 2.3 reviews literature that establishes the theory related to the pricing of
information risks. There are two dimensions of information risks. One is with regards to
the imprecision of information which provides a link between earnings quality and
required return by the market. The other is with regards to the relative amount of
information being made public or kept private by companies. This establishes the
expectation between ownership structure and the monitoring mechanisms being
examined with the required return. This section also reviews the empirical researches
that explore those theoretical links, and discusses briefly the endogeneity problem in
studies involving ownership structure.
2.1 Earnings quality
In previous research, such as in Francis, LaFond, Olsson and Schipper 2004, earnings
quality is associated with characteristics or attributes of earnings figure that are regarded
as favorable and desirable or otherwise. It encompasses more than earnings
management whether in good faith, such as in situation where well informed manager
manages earnings to signal to users, or in bad faith where manager manages earnings to
mislead users.
There is a number of attributes for which this concept of earnings quality is assessed.
Each attribute incorporates the respective perspective of the role of earnings in users’
decision making framework. Thus there is no single agreed upon measure of earnings
quality (Schipper & Vincent 2004). Francis et al 2004, based on past research,
categorized these attributes as accounting based and market based. The accounting
22
based attributes capture the uncertainty of future cash flows or earnings, whilst the
market based attributes capture the market that is investors’ perception of such
uncertainty. The following described these attributes.
2.1.1. Accounting based earnings attributes
Accrual quality
This attribute measures how close is earnings to cash flow. The underlying view is that
high quality earnings is one that is close to cash flow or low in accruals in general.
Unlike cash flow, the incidence and magnitude of accruals is subjected to management
discretionary accounting choice, therefore could be subjected to management
opportunistic action to mislead users, and is also influenced by a company’s individual
and industry characteristics.
In line with this view, two approaches in measuring accruals that lead to low quality
earnings could be identified from past researches. In the first approach, researches
identify the discretionary component of total accruals such as in Jones (1991) and
Dechow, Sloan and Sweeney (1995) studies, or they identify the discretionary
component of specific accruals such as bad debts in McNichols and Wilson (1988) In
both cases the residuals from the regression of total or specific accruals on variables
explaining the non-discretionary components of the respective accruals, are measures of
discretionary accruals. This measure of discretionary accruals is taken as measure of
earnings management that causes low quality earnings.
23
The second approach, taken in Dechow and Dichev (2002) and Francis et al (2004),
focus on the direct relationship between accruals and earnings without regards to the
discretionary/ manipulative and non-discretionary/ unintentional components, as
measure of earnings quality. This approach views the role of accruals as to ‘adjust the
recognition of cash flows over time so that the adjusted numbers (earnings) better
measure firm performance’ (Dechow & Dichev 2002). Thus the measure of earnings or
accrual quality is the residual from the regression of changes in working capital on last
period, current period and next period cash flows from operations.
Persistence
A desirable attribute of earnings is if it is permanent or recurring. Earnings is of high
quality if it is sustainable or in the term used in past research, persistent. This could be
interpreted as a source of earnings from a company’s core operations. Earnings that is
low in persistence could be interpreted as of low quality in the sense that a significant
part of a company’s earnings is generated from sources that is temporary or ‘managed’
and therefore not recurring.
Formally in econometrics term persistence is a measure of how current period earnings
shock (unexpected changes) is carried forward or persist in the future. Thus past studies
employ time series economics forecasting methods with varying assumptions regarding
the earnings process. For example Francis et al (2003) as in previous studies (Lev 1983,
Ali & Zarowin 1992) employ autoregressive model of order 1 (AR1) to estimate
persistence (slope coefficient estimate). This assumes that current period earnings
depend on previous period earnings plus an error term. Other studies as described in
24
Collins and Kothari (1989), use autoregressive integrated moving average (ARIMA)
model where earnings times series is regarded as nonstationary.
Predictability
As apparent from the term, predictability is the ability of existing earnings to predict
future earnings (Lipe 1990, Francis et al 2004). Earnings is of high quality if for
instance investors could use the information on current pattern such as increasing
earnings to predict future pattern, as a component of information that they use to
evaluate the company.
Lipe (1990) and subsequently Francis et al (2004) measure predictability as the variance
of the shocks of the time series earnings (the variance of error term from the forecast
model that measures persistence). As in measuring persistence it is assumed that
earnings process is a univariate time series. Earnings is high in predictability if the
variance is low.
The difference between predictability and persistence is that the former measures
average absolute magnitude of unexpected changes and the latter measures the
autocorrelation in earnings.
Smoothness
Smoothing of earnings is a form of earnings management whish could have a favorable
connotation. This is based on the view that management knowing the unfavorable
consequences of highly variable earnings and having knowledge about the company
25
future prospect, would smooth earnings. The resulting earnings figure would be more
representative of the stable component of the underlying economic event.
On the other hand Leuz, Nanda and Wysocki (2003) indicated that smoothed earnings is
of lower quality because there has been management intervention and therefore the
earnings figure does not reflect the company true economic performance. As in Francis
et al. (2004), Leuz et al. (2003) measures smoothness as the ratio of the standard
deviation of earnings to the standard deviation of cash flows. Both earnings and cash
flows scaled by beginning total assets. A high ratio indicates less smoothness or less
management intervention and therefore earnings is of high quality and vice versa.
Whilst predictability statistically refers to the autocorrelation in earnings where it is
expected that the more predictable the earnings the better the quality, smoothness refers
to the variance in earnings where as explained in preceding paragraph the perception of
smooth earnings is ambiguous.
2.1.2 Market based earnings attributes
As indicated earlier , market based attributes incorporate the market perception of
quality in earnings. Therefore these attributes measures the relationship between
earnings and market returns. Since the quality of earnings measure is imputed from
market assessment, therefore market efficiency assumption is implicit in this measure.
26
Value relevance
It is often interpreted as usefulness. Earnings is of high quality if it is useful in the sense
that it is able to explain variation in returns (Francis et al 2004).
Timeliness
Timeliness is the quality of earnings that is defined as the ability to incorporate
economic income (Ball, Kothari & Robin 2000, Francis et al 2004), where economic
income is the change in market value.
Conservatism
Again in reference to change in market value, conservatism is a measure of the extent to
which earnings incorporates economic losses, relative to economic gains (Basu 1997,
Ball et al 2000).
2.1.3. The relevant earnings quality constructs.
As stated in chapter 1, the objective of the research is to examine if the potential conflict
between controlling and non-controlling parties brought about by the separation of
ownership and control, manifest itself in earnings quality. The manifestation of the
conflict in earnings quality is allegedly through the controlling party manipulation of
earnings to hide expropriation of the company resources or simply to inflate earnings.
Thus earnings quality constructs that could capture manipulation are the accrual quality
construct and the time series properties constructs; persistence, predictability and
smoothness. The accruals quality construct capture manipulation through accruals
27
management, for example by insufficient provision for doubtful debts, obsolete
inventories, and recognition of future revenues all of which to inflate earnings without
corresponding increase in cash flows (Richardson 2003). The time series properties
constructs measures the manipulation through ‘shocks’ in the time series earnings as the
controlling party intervenes in the earnings process.
The objective of the research is also to study the market consequences of low or high
quality earnings by examining the relationship between cost of equity, a form of market
assessment, and earnings quality. As such market based earnings quality measures that
incorporates measures taken from the market such as prices and returns would not serve
this objective.
As described in Francis et al (2004) and Richardson (2003), earnings quality is
subjected to non-intentional or non-manipulative factors as much as it is subjected to the
controlling party’s discretion. As such this research will control for those non
manipulative factors namely a company’s size, sales variability, length of operating
cycles and capital intensity (Francis et al 2004), as far as the data is available to
compute those factors.
Even though all the accounting based earnings quality are estimated from accounting
data, they generally represent different construct of earnings quality and therefore it is
justified to examine each separately. Francis et al (2004) found little overlap between
the market based and accounting based earnings attribute and that even though they
found the accounting based attributes are correlated with each other, the correlations are
not sufficiently strong as to treat them as one construct.
28
Aboody , Hughes and Liu (2005) similarly found that the two accruals quality
constructs based on Jones (1991), as modified in Dechow et al (1995) and based on
Dechow and Dichev (2002) are capturing different information as they are not strongly
correlated.
However the correlations between earnings could be sample driven. Francis, Nanda and
Olsson (2008b) found significant correlation between accrual quality (based on Dechow
and Dichev 2002), earnings variability and abnormal accruals based on modified Jones
(1991). They have used a common factor to represent earnings quality.
2.2 Ownership structure, expropriation of non-controlling shareholders’
interest and earnings quality
2.2.1 Theoretical studies
Ownership structure of a company refers to the distribution of control and ownership in
the company. Control is the ability to affect decisions and for shareholders this is
represented by voting power. While ownership is the right to cash flows of the company
and is proportionate to shareholdings. In general, the separation of control and
ownership of companies results in information asymmetry and agency related problems
namely moral hazards, between those in control of and those who are not.
In early studies such as Berle and Means (1932) and Jensen and Meckling (1976), the
problem has always been characterized along the conflict of interest between a manager
who is in control, who may or may not own any shares, and shareholders who own the
29
company and bears the cash flow consequences of any action. Managers do not own
significantly any shares. However more recent studies characterize the conflict as
between the controlling shareholders (who could also be the manager), i.e shareholders
who have acquired sufficient number of shares to be able to affect decisions, and the
other or non-controlling shareholders (Shleifer & Vishny 1997).
Ownership could become separated from control through holdings of shares with
different voting power, or through holdings of shares in a pyramid structure. This latter
type of control is reported to be more common in East Asia, for example in Malaysia.
As described in earlier chapter, s55 of the Companies Act prohibit the issuing of shares
that depart from one share one vote.
Research that examines the market consequences of certain ownership structure is based
on the premise that the controlling party has an incentive to expropriate funds at the
expense of the non-controlling party. Not only theoretical analyses of Berle and Means
(1932) and Jensen and Meckling (1976) lend support to this premise, but also certain
accounts of the Asian financial crisis (Prowse 1998, Rajan & Zingales 1998) point to
ownership concentration among others as a contributing factor.
According to Jensen and Meckling (1976) it is generally impossible for a perfect
convergence of interest to happen between the utility maximizing principal/ non-
controlling shareholders and the agent/ controlling party. The non-controlling party
could take actions and incur costs to monitor agent, and controlling party could incur
bonding costs to ensure the non-controlling party is compensated when he takes action
that is not maximizing the welfare of the non-controlling party. Nevertheless, there will
30
be some divergence of interest which result in welfare loss to the non-controlling party
which Jensen and Meckling (1976) describes as ‘residual loss’. The controlling party
“will then bear only a fraction of the costs of any non-pecuniary benefits he takes out in
maximizing his own utility” (Jensen and Meckling (1976), page 312).
Along a similar line of arguments, Harris and Raviv (1988) and Grossman and Hart
(1988), analyze theoretically the separation of control and ownership problem through
the holdings of dual class of shares. They too conclude that such separation leads to
lower accountability and specifically lead to situations where the controlling party could
take actions to maximize his utility while bearing costs not in proportion to the
shareholdings.
2.2.2 Empirical studies
The results of the following two empirical studies are consistent with Jensen and
Meckling (1976) analysis that as the cash flow rights of controlling party increases,
there is more wealth maximizing benefits to the company as there would be less
expropriating tendency by the controlling party and less monitoring costs.
Claessens, Djankov, Fan and Lang (1998a), examine expropriation of non-controlling
shareholders’ wealth in the context of corporate diversification policy for 2000
companies in nine East Asian countries in the period between 1991 and 1996. They
found that diversification is associated with the disparity between cash flow and control
rights. Further, there is evidence that the larger the disparity the more the
diversification. This is proven true especially at higher level of control. The larger the
31
disparity the more incentive to expropriate as the link between the controlling
shareholders’ wealth and the company performance is weaker.
In a separate study, Claessens, Djankov, Fan and Lang (1998b), establish the existence
of expropriation by examining the association between each of cash flow and control
rights, and market value. The study is a cross sectional study of 2658 companies in East
Asia in 1996. The found negative association between control rights and market value,
and positive association between cash flow rights and market value. This is especially
so when cash flow rights are low and control rights are high, which they conclude,
suggest expropriation of non-controlling shareholders’ wealth. Further, they examine
the role of the type of ultimate controlling shareholder i.e whether it is family, financial
institution, corporations or state. They conclude that family control is an important
factor in the negative association between control rights and market value. However the
same could not be concluded for state control and widely held corporations.
A number of studies examine the effect of ownership structure with the possibility of
expropriation on earnings quality as perceived by the market i.e on market based
measure of earnings (Fan & Wong 2002, Jung & Kwon 2002, Francis, Schipper &
Vincent 2005). Fan and Wong (2002) using data of 977 companies in East Asia reported
that concentrated ownership and pyramidal structure which creates cash flow and voting
rights disparity are associated with low earnings informativeness as measured by the
earnings-return relation. This result, they explain, is consistent with the view that
earnings figure loses credibility to the market as there is a tendency for the controlling
party to ‘report accounting information for self-interested purposes’ (Fan & Wong
32
2002). Another explanation is that the controlling party may not disclose completely
information regarding the company activities.
Similar results are found in a study by Jung and Kwon (2002) on Korean companies.
They reported that consistent with Jensen and Meckling (1976) convergence of interest
prediction, earnings are more informative as the holdings of manager/owner increase as
controlling and non-controlling party’s interests are aligned. On the effectiveness of
external monitoring, they found institutional investors’ and blockholders’ holdings are
associated with earnings informativenesss. However , when they partition the sample
into chaebol and nonchaebol companies, where chaebol is a business group in Korea
owned and controlled by family, they found no significant relationship between
earnings informativeness and owner holdings for the chaebol companies. This evidence
support the opposing view of the convergent of interest theory , that is the controlling
party become entrenched (Morck, Shleifer & Vishny 1988).
In a study for US companies where the disparity between cash flow and voting rights is
achieved through holdings of dual class shares, Francis et al (2005) found earnings are
less informative relative to dividends for companies with holdings of dual class shares.
They concluded that the existence of dual class shares, in other words, the disparity
between cash and voting rights, reduces the credibility of earnings.
There has not been many studies on ownership structure and earnings quality measures
based on the times series properties. The bulk of accounting based earnings quality
research focus on earnings management. Warfield, Wild and Wild (1995) examine
managerial ownership and earnings informativeness (a market based measure) and
33
discretionary accounting accruals adjustments (an accounting based measure).
Informativeness is the degree of correlation of earnings and returns. The premise of
their research is that the separation of ownership and control, at low level of managerial
ownership, leads to contracts containing accounting based constraints being written to
limit expropriating behaviour of managers. This in turn leads to managers manipulation
of accounting numbers in their self interest which makes accounting numbers less
informative. Thus they hypothesize a positive relationship between managerial
ownership and earnings informativeness. However they recognize the endogeneity of
managerial ownership, where managerial ownership increases in response to earnings
being less informative. They found positive association between managerial ownership
and earnings informativeness. The correlation between earnings and returns is stronger
at higher level of managerial ownership. Secondly, they examine directly whether
managers manipulate accounting numbers which is represented by the discretionary
accruals. They predict an inverse relationship between discretionary accruals and
managerial ownership. The results confirm their prediction.
For a sample of Australian companies, Koh (2003) hypothesizes that income increasing
discretionary accruals vary with the level of institutional ownership in a non-linear way.
The results support this prediction where it is found that at a lower level of institutional
holdings, there is an incentive for managers to manage earnings upwards. In contrast at
a higher level of institutional holdings, there is a negative association between
discretionary accruals and institutional holdings. This shows that long term holdings by
institutional investors prove to be effective monitoring mechanisms.
34
Chung, Firth and Kim (2004) examine earnings management behaviour for companies
with different growth opportunities and availability of free cash flows. They found that
low growth companies with high free cash flow use income-increasing accruals to offset
negative earnings. The relevant findings for the proposed research is that they found that
institutional holdings as well as audit quality moderates the relationship found. This is
consistent with Koh (2003) findings of the effectiveness of institutional investors.
Chung, Ho and Kim (2004) find that although discretionary accruals for Japanese
companies are value relevant or useful in general , the value relevance is reduced for
cross held companies. They conclude that this findings is consistent with the view that
cross-business shareholdings increase managerial expropriation of funds which they
term as tunneling or managerial manipulation of accruals. They also found that foreign
shareholdings and bond financing enhance value relevance which prove that these are
effective monitoring mechanisms.
2.2.3 Consideration of the types of ultimate controlling party
Findings from various studies (Lim 1981,Claessens et al 2000) suggest the type of the
ultimate controlling party; manager, family, institution, government or politically
affiliated group may not only effect the propensity to expropriate, but also may create
less demand for transparency, thus effect the earnings quality.
Claessens et al (2000) found the separation of ownership and control more pronounced
in family controlled than state controlled. A Malaysian study, one of the earliest on
corporate ownership and control, is by Lim (1981). Although his study takes a socio-
35
economic perspective, his major findings are relevant to this study. His purposive
sample consists of 100 large companies listed on the Kuala Lumpur Stock Exchange, at
the time. He proved that share ownership is often concentrated in the hands of a few
institutions, ultimately family or in the hands of cliques or interest groups that share
social or economic relationship. Further, the concentration of ownership enables these
large shareholders to inflate more control than his portion of shares or voting power
would have allowed.
Concentration of ownership of companies in Malaysia could also be seen as an outcome
of economic policies to advance inter ethnic economic equality (Gomez & Jomo 1999).
As such government enterprises were expected and still are involved in businesses and
share ownership. Also, given the ethnic based policies, company ownership could also
be traced to political affiliation.
2.2.4 Consideration of monitoring mechanisms – board structure, substantial
shareholders and audit committee
Intuitively, no matter how compelling the arguments are, examining the effect of
ownership structure on earnings quality alone would not be complete. In statistical
terms there is a potential that the model tested would be misspecified. Agrawal and
Knoeber (1996) suggest, based on their study on company performance and
mechanisms to control agency problems, that any analysis based on any single
mechanism may be misleading.
36
As described above Koh (2003) and Chung, Firth and Kim (2004) found evidence of the
effectiveness of institutional holdings. The role of substantial shareholders can also be
seen from ‘information argument’ (Fan & Wong 2002). A controlling party would have
an advantage in terms of control of the flow of knowledge about the company
(proprietary knowledge). A controlling party could limit the information flow to
outsiders so as not to leak information to competitor. On the negative side this control
could be potentially harmful as the controlling party could hide any wrong doing. The
presence of others, such as another substantial shareholder, potentially increase the
sharing of this proprietary knowledge as the substantial shareholder would want more
information, be more informed, in order for him to make investment decisions. Other
non-controlling shareholders and prospective investors could benefit from this. There is
less opportunity for the controlling party to hide any expropriation and thus the
substantial shareholder ability to leak information to the public or other non-controlling
is a deterrent to expropriating behavior.
Peasnell, Pope and Young (2000) found evidence on board of directors role on earnings
management. Park and Shin (2004) whilst did not find significant relationship between
outside directors and abnormal accruals, found directors from financial intermediaries
reduced earnings management.
A study on a sample of Malaysian listed company found significant relationship
between CEO-Chairman duality and discretionary accruals (Mohd Saleh, Rahmat &
Mohd Iskandar 2004a). Confirming the interest alignment theory, they also found
negative relationship between managerial ownership and discretionary accruals. They
37
did not however found evidence of relationship between proportion of outside directors,
board of directors’s size and multiple directorship, and earnings management.
Klein (2002) found negative relationship between audit committee independence and
abnormal accruals, a proxy for earnings management. They also conclude that an
independent board is effective in monitoring earnings management behavior. Another
study in the US (Abbott, Parker & Peters 2004) found significant negative relationship
between each of audit committee independence and activity level, and the incidence of
financial restatement not involving fraud. The study also found evidence of significant
association between audit committee member of at least one with financial expertise and
the incidence of restatement.
On the other hand, a study using Malaysian data, Mohd Saleh et al (2004a) did not find
relationship between audit committee characteristics (frequency of meetings, size,
accounting knowledge and proportion of non-executive members) and earnings
management. However Mohd Saleh, Rahmat and Mohd Iskandar (2004b) found fully
independent audit committee members (as opposed to audit committee with varying
degree of independence), and the interaction between proportion of audit committee
members with accounting knowledge and the frequency of meetings, reduce earnings
management.
2.2.5 Justification for Audit committee (AC) measurements
There is no theoretical justification specifically for measures of effectives of audit
committee. Past researches use compliance to regulations such as the Code of Corporate
38
Governance in Malaysia (Mohd Saleh et al 2004a and 2004b, Abdul Rahman &
Mohamed Ali 2006) or recommendations by the relevant authority such as by the Blue
Ribbon Committee (Abbott et al 2004) as measure of effectiveness. These researches
then examine the required or recommended characteristics such as size, independence
and frequency of meetings and whatever dependent variable is in focus such as
earnings management.
This research, however draws theoretical justifications for effectiveness measurement of
audit committee as a monitoring mechanism, from the theoretical justification for
external audit quality, another monitoring mechanism. Briefly, based on the work of
DeAngelo (1981) and Watts and Zimmerman (1986), the likelihood that an auditor will
report a non-compliance depend on; (1) the likelihood he discovers the non-compliance,
and (2) the likelihood that upon discovering the non-compliance he reports it. The first
depends on the auditor’s competence, which has been translated as audit firm size, and
the second depends on the auditor’s independence, which has been measured by
whether the auditor provides non-audit service and length of tenure.
Drawing parallel to the function of audit committee, to monitor the controlling party’s
expropriating behavior, the audit committee must have the technical competence. For
the purpose of this research, competence of the audit committee is measured by the
qualification and experience of audit committee members. Competence is not related to
running the business such as in managing human resource, in marketing or in expanding
business. It is related to the ability to recognize earnings management practices or
expropriating behavior by controlling party.
39
An experimental study by Mc Daniel, Martin and Maines (2002) suggests that the
existence of experts (who are audit managers in their sample) directs the discussion and
evaluation of companies’ financial reporting towards issues that are important to
financial reporting quality.
The independence characteristic is measured as in Klein (2002), where the members are
regarded as independent if they are truly outsiders and therefore will report non-
compliance or irregularity upon discovering one.
2.3 Earnings quality, information risks and market required return or
assessment
2.3.1 Theoretical studies
As discussed earlier regulators have a role in reducing information asymmetry by
setting financial reporting standards so that management produces high quality
information that closely reflect the underlying economic events of the company. Using
the disclosed information investors then could make investment decisions which include
pricing of the company’s shares based on its performance. However if the quality of
information such as earnings is suspected then intuitively investors would want a higher
return on their investment.
In the Capital Asset Pricing Model (CAPM) the only risk that is priced is the systematic
risk of a company i.e the covariance of the company cash flow in relation to the market
portfolio. Other risks are idiosyncratic risks that are uncorrelated across companies.
Investors could diversify away such risks by holding a portfolio of large number of
40
shares. As such under this model the risk poses by the quality of information such as
accounting information is not priced or does not explain cross sectional differences of
returns.
However, Easley and O’Hara (2001) and Leuz and Verrechia (2005) establish a
theoretical link between information risk and companies cost of capital counter to
CAPM analysis. Easley and O’Hara (2001) show that the composition of public and
private information could influence a company’s cost of capital. Investors would want
higher return from companies that have more private and therefore less public
information. The high return reflect the risk that uninformed investors have to face by
holding shares of such companies. Thus information risk is a type of systematic risk that
is priced. Leuz and Verrechia (2005) arrived at similar conclusion using a different
approach. They demonstrate that higher information quality lower cost of capital
because information quality could actually affect a company cash flow and not just
perceived cash flow.
2.3.2 Empirical studies
A number of research explores empirically the link between information quality as
proxied by a number of measures, and cost of equity as most studies focus on usage of
information by equity investors. Botosan (1997) examines the relationship between
disclosure level and cost equity. She developed a voluntary disclosure index from
information in annual reports as proxy to disclosure level or quality. Estimates of cost of
equity are based on the valuation formula developed by Edwards and Bell (1961),
Ohlson (1995) and Feltham and Ohlson (1995) which states that market price of a
41
company’s share is equal to the sum of expected dividends discounted at the company’s
cost of equity. Botosan (1997) found a negative association between disclosure level
and cost of equity, after controlling for market risk (beta) and company’s size for
companies that attract a low analyst following. However no significant association was
found for companies that have high analyst following. The reason for this is that the
disclosure index may not capture fully the level of information provided to investors as
analysts play a significant role in disclosure. Botosan and Plumlee (2001) reexamine the
association between disclosure and cost of equity by segregating different forms of
disclosure quality i.e level and timely. Findings for relationship between disclosure
level and cost of equity confirm previous results. However a positive association was
found between timely disclosure and cost of equity which is contrary to theoretical
assertion. An explanation for this is that timely disclosure increases volatility of share
prices and hence cost of equity.
Francis et (2008b) found that voluntary disclosure as measured by a self-constructed
index of items in companies’ annual report has no distinct pricing effect. It is the
earnings quality measured by a common factor of three earnings attributes which is the
primary driver of cost of capital. In other words companies with high earnings quality
tend to voluntarily disclose more.
Francis et al (2004) examine the relationship between earnings attributes as proxy to
information quality and cost of equity. Earnings attributes are categorized as market
based and accounting based, each as described in earlier paragraphs. As a whole their
findings confirm previous results of negative relationship between earnings quality and
cost of equity. When considered individually the accounting based earnings attributes,
42
in particular accrual quality, have larger effect on cost of equity than market based
attributes.
Chen, Chen and Wei (2003) examine the effects of various corporate governance
mechanisms and disclosure level on the cost of equity. They found significant negative
association between corporate governance mechanisms and disclosure level, and cost of
equity. Their study was on Asia’s emerging markets which include 42 Malaysian listed
companies.
2.4 Market assessment or consequences of information quality
Market assessment or consequences of information quality refer to the effect of
information quality on expected or required return and valuation of shares. Studies that
examine market consequences of accounting information generally assume market is
efficient. In an efficient market hypothesis (EMH) the assumption of rational investors
implies that investors correctly use information in making their assessment of the value
of companies’ shares (Hand 1990, Tinic 1990). On the other hand the functional
fixation hypothesis (FFH) views investors as unsophisticated and are therefore unable to
‘unscramble the true cash flow implications of accounting data’ (Hand 1990). Hand
(1990) proposes a middle ground view and that is at any given time share prices are
determined by sophisticated investors and at other by unsophisticated investors. He
conjectured that the probability of the share price being determined by unsophisticated
investors is measured by the relative shareholdings of such investors in a company. He
finds evidence consistent with the proposed view and inconsistent with EMH.
43
This part of the proposed research is not about testing EMH, FFH or extended FFH. The
measurements that will be used as explained in chapter 3 are far from those used in
researches that test these hypotheses. However the findings from each hypothesis are
relevant in that they draw attention to the existence of sophisticated investors and
unsophisticated investors.
Even in sophisticated market, such as in the US, there has been research that although is
not testing either of EMH or FFH, tests the sophistication of market participants. The
results are rather mixed. Collins and DeAngelo (1990) test separately analyst and
market reactions to earnings management in the context of proxy contest for board
seats. They find that both categories of market participants reacted similarly. Contrary
to common belief, they find that despite indications of earnings management during the
proxy contest, analyst reaction similar to market reaction, is more intense in prior
periods.
Francis et al ( 2003) examine whether the market is influenced by the earnings quality
in their price reaction to increasing earnings, quarterly earnings that meet or exceed
analyst forecasts and smooth earnings. They find that for all three earnings pattern
market either does not reward or penalizes patterns achieved with low quality earnings.
In other words the market is able to discern low quality earnings.
Richardson (2003) tests whether short sellers use information regarding a measure of
earnings quality, accrual quality. He predicts that investors short sell shares that are
associated with high accruals as the performance of those shares has proven to
experience a decline. He concludes that market does not use earnings quality
44
information. Apart from short sellers not knowing the information content of accruals,
he attributed the failure of short sellers to cost and risk in shares with high accruals.
The calculations of the required rate of return by the market can be based on ex ante
measures drawn from analyst earnings forecasts and ex post measures based on realized
returns. The pricing of information quality research uses primarily the ex ante version of
required return or the cost of equity (Botosan 1997, Botosan & Plumlee 2001, Chen et
al 2003, Francis et al 2004, Francis et al 2008b). Francis et al (2004) in addition uses
portfolios of realized returns as sensitivity tests. Francis et al (2008b) uses average daily
realized returns, annual realized returns, capital asset pricing model excess return and
Fama and French (1993) excess return with size and book to market as explanatory
variable for excess return. Aboody et al (2005) uses primarily excess returns based on
Fama and French (1993) model.
The use of ex ante version of expected return as a primary measure arises from doubts
regarding realized return as a measure of expected return (Elton 1999). The bulk of
empirical asset pricing researches use realized returns as proxy to expected return as
discussed in Francis et al (2004), based on the belief that information surprises cancel
out on average and based on rational expectations. For practical reasons the use of
realized returns eliminate the need to make estimates as needed in the calculation of cost
of equity.
45
2.5 Endogeneity of ownership structure
Past research has dealt with the issue of endogeneity of ownership structure. However it
is more an empirical than a theoretical assertion. Demsetz in particular in various of his
work (Demsetz 1983, Demsetz 1985 and Demsetz & Villalonga 2001, Kapopoulus &
Lazaretou 2007) stressed the need to take into account the endogeneity of ownership to
avoid biasness in estimating relationships. Although his work is mainly in relationship
between ownership structure and performance, parallel arguments can be drawn in
investigating ownership with other variables in particular cost of capital. He argues
ownership changes too in response to market expectations citing examples such as
leveraged buy-out of non-management shares by management and cases where
management compensation in the form of stock options.
Along similar arguments Mak and Li (2001) conclude that models that do not consider
the endogeneity of ownership structure may be misspecified. In their study of the
determinants and interrelationships of corporate ownership and board structure
characteristics for a sample of Singapore listed companies they find significant
interrelationships among board and ownership characteristics. For example the
proportion of outside directors is negatively related to managerial ownership, board size
and government ownership.
2.6 Summary
Earnings quality is the characteristics of earnings figure that are regarded as favorable.
Francis et al (2004) examine both market based and accounting based measures of
46
earnings quality that have been separately examined in previous researches. In view of
the objective of this study, only accounting based measures are examined. They are
accrual quality, persistence, predictability and smoothness. In addition to Francis et al
(2004) measure of accrual quality, this study also examines Jones (1991) abnormal
accruals.
As with more recent studies, this study characterizes the separation of ownership and
control as between the controlling party (who may be the management) and the non-
controlling party. Control is represented by voting power whilst ownership is the rights
to cash flows. Early studies (Berle & Means 1932, Jensen & Meckling 1976) as well as
the more recent ones (Grossman & Hart (1988), Harris & Raviv 988, Shleifer & Vishny
1997) establish a potential for the divergence of interest between the controlling party
and the non-controlling party. Empirical research (Fan & Wong 2002, Jung & Kwon
2002, Francis, Schipper & Vincent 2005) leads to the potential link between ownership
structure and the monitoring mechanisms to earnings quality. But the bulk of the
research uses market based measures and abnormal accruals measure for earnings
quality. The theoretical justification for using two measures, independence and
competence, for the audit committee is drawn from similar justification for measures of
external audit.
Theoretical studies (Easly & O’Hara 2001, Leuz & Verrechia 2005) establish two
dimensions of information risks. One is with regards to the imprecision of information
which provides a link between earnings quality and the cost of equity, a measure of
required return by the market. The other is with regards to the relative amount of
information being made public or kept private by companies. This establishes the
47
expectation between ownership structure and the monitoring mechanisms being
examined with the cost of equity.
Studies that examine the relationship between information quality and cost of equity
generally supports the theory. Most studies use cost of equity as a measure of market
assessment or consequences.
48
CHAPTER 3
THEORETICAL FRAMEWORK AND HYPOTHESIS DEVELOPMENT
3.0 Introduction
This chapter outlines the relationships under study and develops the hypotheses by
drawing from theoretical and empirical studies reviewed in chapter 2. Figure 3.1
depicts the relationships examined. The relationships are:
1. between ownership structure, monitoring mechanisms and earnings quality,
2. between earnings quality and market assessment,
3. between ownership structure, monitoring mechanisms and market assessment, and
4. the relationship that shows market assessment and the monitoring mechanisms can
in turn explains ownership structure
EQ
MA OS
1 2
3
4
Monitoring elements i.e AC and SS
1
3
4
Figure 3.1 Theoretical Framework
EQ - Earnings quality MA - Market assessment OS - Ownership structure AC - Audit committee SS - Substantial shareholding
49
3.1 Relationship between ownership structure, monitoring mechanisms and
earnings quality
Theoretical studies (Jensen & Meckling 1976, Harris & Raviv 1988, Grossman & Hart
1988), assert that ownership structure that exacerbates the separation of ownership and
control leads to situation where the controlling party of a company takes actions to
maximize his utility while bearing costs not in proportion to his shareholdings.
Empirical studies by Claessens, Djankov, Fan and Lang (1998a, 1998b) found evidence
of expropriation in East Asian companies where there is a cash flow/ voting rights
disparity that is where the degree of separation of ownership and control is high.
These actions potentially affect earnings quality. Empirical studies (Fan & Wong 2002,
Jung & Kwon 2002, Francis, Schipper & Vincent 2005) have found negative
association between ownership structure, in particular the degree of separation of
ownership and control, and earnings quality. However thus far empirical studies use the
market based earnings quality. These earnings attribute measures the market perception
of uncertainty in future cash flow or earnings.
This study examines the relationship between ownership structure and accounting based
measure of earnings quality. The measures of accrual quality, persistence, predictability
and smoothness are established accounting based measures of earnings quality that
reflect different aspect of uncertainty in future cash flow or earnings. In other words the
expropriating behavior of the controlling party is expected to lead to poorer accounting
based earnings quality, because such behavior directly implicates accounting
information through direct manipulation of the accounting information or actual
50
behavior such as the asset acquisition at questionable price cited in the SC report
(2003).
Thus it is hypothesize that ownership structure, specifically the degree of separation of
ownership and control is related to earnings quality. The higher the degree of separation
of ownership and control, the lower the earnings quality.
Hypothesis 1: There is a negative relationship between ownership structure and
the earnings quality.
Several studies found that there are controlling mechanisms that could reduce the
expropriating behavior of the controlling party. These mechanisms are substantial
shareholders (Chung et al 2003,Koh 2003), board structure (Peasnell et al 2000, Park &
Shin 2004, Mohd Saleh et al 2004a) and audit committee (Klein 2002).
However Koh (2003) and Chung et al (2003) examined specific type of substantial
shareholdings that is the institutional shareholdings. In their hypothesis development
Chung et al (2003) argue that institutional shareholders with substantial shareholdings
are, by nature of their large shareholding, more likely to monitor managers as they are
not able to easily dispose off their investments. Similarly Koh (2003) argues that long
term institutional shareholders are likely to monitor managers’ accrual discretion.
This study differs from these two key studies in that in the Malaysian context as
observed and illustrated in Chapter 1, substantial shareholdings are not limited to those
of private institutions. They could be a state owned institution, or the state itself and
51
other unrelated individuals. Along similar arguments as those in the two studies, by
nature of their illiquid shareholdings and of their long term holdings, these other type
of substantial shareholders could impose more monitoring than other non-controlling
shareholders. However it is also an open question whether they are more likely to be
monitoring, as these substantial shareholders could effectively be ‘partners’ to the
controlling party and thus may share the same motives. In either case their existence
cannot be regarded as neutral and therefore even though a relationship with earnings
quality is hypothesized, the sign of the hypothesized relationship is not predicted.
Hypothesis 2: There is a relationship between substantial shareholding and
earnings quality.
The evidence on the effectiveness of board structure (proportion of outside directors,
size, CEO-Chairman duality, financial intermediaries) is rather mixed. Similarly the
evidence on the effectiveness of audit committee is rather weak, except for evidence in
Klein (2002).
This study focus on audit committee as in Malaysia members of audit committee are
almost always the members of board of directors and they are specifically assigned a
monitoring role. Unlike similar Malaysian study (Mohd Saleh et al 2004b) this study
improves the measurement of audit committee independence by using the method in
Klein (2002), where any member that is formerly affiliated to the company (such as a
former employee or consultant) is not taken to be independent even though they are
declared as such. Independent members are truly outsiders.
52
Thus it is hypothesize that the more independent and competent the audit committee the
higher the earnings quality.
Hypothesis 3: There is a positive relationship between audit committee
characteristics and earnings quality.
(That is,
Hypothesis 3a: There is a positive relationship between audit committee
independence and earnings quality.
Hypothesis 3b: There is a positive relationship between audit committee
competence and earnings quality.)
3.2 Relationship between earnings quality and market assessment
Studies on market assessment or consequences of accounting information (Botosan
1997, Botosan & Plumlee 2001, Chen et al 2003, Francis et al 2004, Aboody, Hughes &
Liu 2005) examine the effect of the quality of information on expected return or
valuation of shares.
Low quality information poses an information risk and that this risk is priced by the
market (Easley & O’Hara 2001 and Leuz & Verrechia 2005). A study that explores this
link finds different effects when earnings quality construct is measured differently
(Francis et al 2004).
Previous studies based on data in the US capital market have used both ex ante forms
(Botosan 1997, Botosan & Plumlee 2001, Chen et al 2003, Francis et al 2004, Francis et
53
al 2008b) and ex post forms (Francis et al 2004, Aboody et al 2005, Francis et al
2008b) of expected return. The results from using ex ante measure, henceforth referred
to as the cost of equity and the ex post measure, henceforth referred to as market return
generally provide evidence that information quality is priced, that is market players
require higher return from companies with lower quality of information. The results are
consistent across various versions of each measure even though there are doubts with
regards to the use of ex post realized return as proxy to expected return (Elton 1999).
Thus a negative relationship is hypothesized between earnings quality and expected
return.
Hypothesis 4 : There is a negative relationship between earnings quality and cost
of equity .
Hypothesis 5 : There is a negative relationship between earnings quality and
market return.
Given that the input for the cost of equity estimation is from analysts’ assessment of the
companies and analysts’ assessment is also part of the market information structure it is
hypothesized that the cost of equity is positively related to market return.
Hypothesis 6 : There is a positive relationship between market return and cost of
equity.
54
3.3 Relationship between ownership structure, monitoring mechanisms
and market assessment
The hypothesized relationship between earnings quality and market assessment is based
on the theory that earnings quality poses a dimension of information risk and that is
related to the precision of the information and thus subsequently affect market players
assessment of the future uncertainty in cash flows or earnings (Easley and O’Hara
(2001), Leuz and Verrechia (2005) studies as discussed in Francis et al (2004)).
Another dimension of information risk is related to the relative information of a
company that is private and public. Market players view companies whose information
is largely private pose higher risk and therefore market players would want a higher
return of these companies (Easley and O’Hara (2001), Leuz and Verrechia (2005)
studies as discussed in Francis et al (2004)).
There has not been any study conducted based on the US capital market that test
relationship between ownership structure or other governance mechanisms and expected
return. This is due to companies in the US and other developed markets having
dispersed rather than concentrated ownership. The theoretical analysis by Easley and
O’Hara (2001) on composition of private and public information is based on
composition of uninformed and informed investors.
Chen et al (2003) examined the relationship between level of corporate governance and
cost of equity. They hypothesized a negative relationship based on the arguments that
corporate governance mechanisms can potentially reduce the risk of expropriation by
55
the controlling party. This risk is non-diversifiable because it has a component that is
related to the market condition and thus is priced. Note that in information risk
argument, information risk is firm specific and therefore diversifiable.
In both arguments, that is whether risk associated with corporate governance
mechanisms is a form of information risk and therefore firm specific or is a form of risk
related to the market, there is a strong case for the association between corporate
governance mechanisms and cost of equity.
The ownership structure which in this study refers to the distribution of cash flow and
voting rights could be related to both arguments. As separation of ownership and
control increases, information asymmetry between controlling and non-controlling party
increases thus private information is relatively more than those made public. The
tendency to expropriate also increases. Thus it is hypothesized that there is a positive
relationship between separation of ownership and control, and market assessment.
Hypothesis 7: There is a positive relationship between ownership structure and
market assessment.
The monitoring mechanisms has a potential to reduce the risk of expropriation or
increase the flow of information to the public domain. Thus a negative relationship is
hypothesized between audit committee characteristics and market assessment.
As discussed in chapter 2, the presence of substantial shareholders, may force the
controlling party to release proprietary knowledge to the substantial shareholders in
56
order for the substantial shareholders to make decision. This inevitably release
information to other non-controlling party or the public. “The larger the set of informed
individuals, the larger the likelihood that proprietary information leaks to the
public……” (Fan and Wong 2002, page 408).
However, the potential role of substantial shareholding again could be an open question.
Thus a significant relationship is hypothesize in either a positive direction ( in case
where substantial is not a monitoring mechanism but rather a partner of the controlling
party) or negative direction (in cases of effective monitoring by substantial
shareholders).
Hypothesis 8: There is a relationship between substantial shareholder and
market assessment.
Hypothesis 9: There is a negative relationship between audit committee
characteristics and market assessment.
(That is,
Hypothesis 9a: There is a negative relationship between audit committee
independence and market assessment.
Hypothesis 9b: There is a negative relationship between audit committee
competence and market assessment.)
57
3.4 Relationship that shows market assessment and the monitoring mechanisms
could explain changes in ownership structure
Demsetz (1983), Demsetz (1985), and Demsetz and Villalonga (2001) examine
ownership structure and performance, and also examine the relationship where
ownership structure is assumed to be endogenous. The argument for the endogeneity of
ownership is that controlling shareholders are also likely to change their holdings in
response to performance. Mak and Li (2001) also found significant interrelationship
between board and ownership characteristics.
For the relationships examine in this study that involves ownership structure, it is
conceivable that ownership structure is endogenous. For example ownership structure
could change by the controlling party changing his shareholding in response to changes
in market assessment (increase/decrease expected return) and changes in monitoring
mechanisms (increase/decrease monitoring). As Claessens et al (2000) found that type
of owners (whether family, government or institution) could explain ownership in terms
of the level of separation of ownership and control. Their results shows that the level of
separation of ownership and control is more in family owned than state controlled.
Hypothesis 10 : There is a positive relationship between type of ownership and
ownership structure.
Hypothesis 11 : There is a negative relationship between market assessment and
ownership structure.
Hypothesis 12 : There is a positive relationship between the monitoring mechanisms
and ownership structure.
58
CHAPTER 4
RESEARCH METHODOLOGY
4.0 Introduction
The first section of this chapter explains the population and sample selection. The
following section describes in detail how the variables involved in the relationships
examined are defined and measured or estimated. Where alternative measures are used
in previous studies, justification is given for the ones used in this study. The last section
explains procedures for the data analysis.
4.1 Population and sample
The sample of companies is drawn from companies listed on Bursa Malaysia. However
the sample is limited by data availability, initially with regards to earnings forecast that
are needed to estimate cost of equity. IBES provides on Bloomberg services, earnings
forecasts for two years ahead. The earnings forecasts of years 2005 and 2006 are
obtained for 213 companies. These companies comprise the sample for this research.
Thus the relationships depicted in the model on page 52 are examined
contemporaneously for year 2004 because the earnings forecasts obtained for years
2005 and 2006 enable cost of equity to be estimated for year 2004 only.
The single year cross sectional study is justified by the fact that ownerships structure
does not change as assumed by Claessens et al (1998a, p 4).
59
The key studies (Claessens et al 1998a, Claessens et al 1998b, La Porta et al 1999)
referred to, examine ownership and control across a number of countries especially in
the Asian region. However Gul (2006) and Miller (2004) (as discussed in Gul (2006))
raises the need to focus on a certain country or specific region where variables
information are available easily. This study focuses on Malaysian companies for
practical and conceptual reasons. Detail information on controlling and substantial
shareholding may not be easily available for other countries for example in the Asia
region. It is not known the existence of significant substantial shareholders is common
among other countries to be included. Therefore it cannot be hypothesized as a variable
that explain cross sectional variation of earnings quality and is subsequently priced.
More importantly the development towards market based regulation is expressively
unique in Malaysia during the period under study.
4.2 Variable definition and measurement
To provide an overview of the variables involved , the following table provide a brief
description of them.
Table 4.1 Variables brief description
Variables Measurement Earnings Quality (EQ):
1. Accrual quality - Mapping cash flows 2. Discretionary current accruals 3. Discretionary total accruals
ABRES
ABSDATCA
ABSDATA
Absolute residual from the regression of changes in working capital and past, current and future cash flows (Dechow & Dichev 2002) Modified Jones (1991) model
60
4.Persistence
PERS Slope coefficient of earnings time series model. The time series model is the regression of earnings per share on lagged earnings per share.
5. Predictability PRED Absolute value of the residuals from the earnings time series model. As for persistence, the time series model is the regression of earnings per share on lagged earnings per share.
Audit committee (AC):
1. Independence 2. Competence
ACI
ACC
Proportion of members that are outsiders (those who are not affiliated in any way with the company other than being a director) Proportion of members that have accounting/finance knowledge (through experience or qualification)
Substantial shareholding (SS)
SSVR
The voting rights of the substantial shareholder who has the next highest voting rights after the controlling party
Ownership structure: 1. Cash flow rights 2. Voting rights 3.Cash flow voting/controlling rights disparity 2. Ultimate controlling party
CF
VR
CFVR
UCP
Ratio of cash flow to voting/controlling rights UCP is the shareholder that holds more than 20% of shares and the one with the highest shares. There are 5 categories: manager, institution, government, family and foreign company, and requires 4 dummy variable as follows. Foreign company category is the reference and UCPMn = 1, for managerial controlled, =0 otherwise. UCPInst= 1, for institutional controlled, =0 otherwise. UCPGov = 1, for government controlled, =0 otherwise. UCPFam =1, for family controole, =0 otherwise.
Market assessment: Cost of equity : Ex ante measure of expected return
COE
COEA
Residual income model Alternative estimation
61
Market return : ex post measure of expected return
AVMR
ER
Realized returns –average monthly return Fama and French (1993) three factor model excess return
Control variables: Size
LGMV Log market value
Growth
LBTMV Log book to market value
Risk
β Beta from Datastream
Capital Intensity
CAPINT Net book value of property, plant and equipment to total assets
Operating cycle
OC Log of the sum of a company’s days accounts receivable and days inventory
Standard deviation of revenue
STDREV
4.2.1 Earnings quality constructs
4.2.1.1 Accruals quality
There are two measures of accrual quality as operationalized in Aboody, Hughes and
Liu (2005) based on models developed by Jones (1991), Dechow et al (1995) and
Dechow and Dichev (2002) . The first is a direct attempt of measuring the discretionary
part that is the part that the controlling party may manipulate. Accrual quality is
measured as the amount of discretionary accruals (DA). Large DA is associated with
low quality. First non-discretionary accruals (NDA) is measured from a model
developed by Jones (1991) and subsequently modified by Dechow et al (1995). The
estimation is done in the following 3 steps (time t refers to year 2004).
62
1. A cross-sectional regression of total accruals on change in revenue and plant,
property and equipment (all scaled by lagged total assets) is run to obtain
estimates of coefficients a,b and c of each industry.
TAt /A t-1= a (1/A t-1) + b ΔREVt//A t-1 + cPPEt /A t-1+ δ t
Where,
TA - Total accruals
= Change in current assets – Change in current liabilities –
Change in cash + Change in short term debt – Depreciation
A t-1 -Total assets at t-1
ΔREVt -Change in revenues
PPEt -Plant,property and equipment at t
2. Then the estimated coefficients a,b and c is substituted into the following equation’s
coefficients (α,β and γ) to obtain non-discretionary accruals for each company.
NDAt/ A t-1= α(1/A t-1) +β ( ΔREVt - ΔRECt)/ A t-1 + γPPEt/ A t-1
Where,
NDAt - Non-discretionary accruals at time t
ΔRECt - Change in receivables
:
3. Then, the discretionary accruals, DA = TAt /A t-1 - NDAt/ A t-1
In the following analysis the absolute DA is taken and the acronym ABSDATA
(absolute discretionary total accruals) is assigned to the variable. The following analysis
will also use the current accrual variation of the above model as given below.
63
NCAt/ A t-1 = α(1/A t-1) +β ( ΔREVt - ΔRECt)/ A t-1
Where,
NCAt - Non-discretionary current accruals at time t
The coefficients α and β are estimated from coefficients a and b of the following cross-
sectional regression by industry :
TCAt /A t-1= a (1/A t-1) + b ΔREVt//A t-1 + δ t
Then discretionary current accruals (DCA),
DCAt = TCAt /A t-1 - NCAt/ A t-1
Where,
TCAt = Total current accruals
= Change in current assets – Change in current liabilities –
Change in cash + Change in short term debt
In the analysis the acronym ABSDATCA (absolute discretionary total current accrual)
is used. The Bursa Malaysia classification of industry is used for the cross sectional
regression. The larger the discretionary accruals whether based on total accruals
(ABSDATA) or total current accruals (ABSDATCA) the poorer the earnings quality.
The second measure of accrual quality is based on Dechow and Dichev (2002) model
which does not attempt to separate out the discretionary or manipulative component. It
simply measures quality as how well accruals map current cash flows to last and future
64
cash flows. It is the residual from the regression of changes in working capital of past,
current and future cash flow. This study will use the cross sectional version
operationalized in Aboody et al. (2005) as follows.
TCA j,t /Avasset j,t= a + b CFOt-1// Avasset j,t + c CFOt /Avasset j,t +
d CFOt+1 /Avasset j,t + δ t
Where,
CFO - cash flow
= net income before extraordinary item – TA (total accruals)
Avasset- average asset over t and t-1
The coefficients a,b,c and d will be applied to individual companies current, past and
future cash flows. The difference between the predicted and actual company’s total
current accrual is the residual used as a measure of earnings quality. The acronym
ABRES (absolute residual) is assigned to the variable. The larger the value of ABRES
the poorer the quality of earnings as the current accruals do not map well with current,
past and future cash flows. If the residual is small, this means that the total current
accruals is largely translated into cash flows.
Times series versus cross section versions of Jones 1991 and Dechow and Dichev
2002
In deriving the company specific measures for discretionary accruals (both current and
total) and for absolute residual of the cash flow mapping, this study uses the cross
sectional regressions to obtain the relevant parameters in the respective models as in
65
Aboody et al (2005). Francis et al (2004) uses the time series version to obtain the
absolute residuals.
Each approach has its own merits and demerits. The cross sectional version arguably
provides ‘noisy measure’ due to differences across companies in the same industry
(Francis et al 2004), however the measure would not be bias towards companies that
survive longer as would a measure from the time series version.
For the purpose of this research, on balance, the cross sectional approach is preferred as
the time series approach provides parameters that are a company’s own benchmark
measures. As mentioned earlier given that ownership structure is fairly stable
(Claessens et al 1998a) a company own benchmark would not be useful. In addition a
previous study, Mohd Salleh (2003), on Malaysian data indicates that the cross sectional
approach provides measures that produce significant results.
4.2.1.2 Times series properties
Persistence and predictability
The two measures , persistence and predictability, are estimated by assuming the
earnings figures follow a time series process. The Box-Jenkins method can be used to
determine the time series model which fit the earnings data such as whether the data fits
autoregressive process (AR) or autoregressive integrated moving average process
(ARIMA) (Gujarati 1995). However due to limited data, the time series model used is
the auto regressive of order one, as used in Francis et al (2003) and Ali and Zarowin
(1992). The model is as follows :
66
Ej,t = φ 0,j + φ 1,j E j,t-1 + ε j,t
Where,
Ej,t - earnings of firm j at time t
Persistence is the slope coefficient estimate φ 1,j. Earnings is of higher quality the higher
the value of φ 1,j . It is a measure of how current period earnings shock (unexpected
changes) is carried forward or persist in the future as described in section 2.1.1. As the
indication of high and low quality of earnings is opposite to the other measures of
earnings quality, in the analysis the values of φ 1,j are negated. Persistence is assigned
the acronym PERS.
Predictability is the absolute value of the residuals. Large value of predictability
indicates low quality. As described in section 2.1.1 predictability is the ability of
existing earnings to predict future earnings. An alternative measure of predictability is
the standard deviation of the residuals from 5, 8 or 10 firm- year regressions. This
measure not only requires earnings figure over longer period, but also is reported to be
strongly and negatively correlated with persistence. (Dechow & Dichev 2002, Francis
et al 2003). Thus this study uses the absolute value of the residuals as used in Aboody et
al (2005). Predictability is assigned the acronym PRED.
Francis et al (2003) regresses earnings on lagged earnings over a period of ten years. So
as not to reduce further the sample number, this study carries out the company specific
regression over eight year period as in Dechow and Dichev (2002).
67
Francis et al (2003) carries out the regression based on maximum likelihood method. As
discussed in Gujarati (1995) the maximum likelihood has stronger statistical properties
than an ordinary least square method. However, the problem with maximum likelihood
method is that the method cannot find a solution for certain data. The alternative method
is to use ordinary least square. Gujarati (1995) shows that the coefficient estimates
based on maximum likelihood and ordinary least square methods are statistically
equivalent in all cases. However the maximum likelihood standard error estimate
(∑µi2/n) is biased whilst the ordinary least square standard error estimate (∑µi
2/(n-2)) is
not. As can been seem from the formula they would be equivalent as n approaches ∞, in
other words for large samples. This study uses ordinary least square so as not to reduce
further the sample size.
4.2.1.3 The non-discretionary determinants of earnings quality
Francis et al (2004) describes earnings quality as being determined by ‘intrinsic (innate)
factors’ that are not within the discretion of management. These factors have also been
examined by previous research such as Dechow and Dichev (2002), Penman and Zhang
(2002) and Baginski, Lorek, Willinger and Branson (1999) as discussed in Francis et al
(2004).
In summary these factors that influence earnings quality are size, cash flow variability,
sales variability, length of operating cycle, capital intensity, and the absence and
intensity of intangibles. The cross sectional measures of earnings quality (ABRES,
ABSDATCA and ABSDATA) are tested with economic determinants of earning quality
namely size, capital intensity (CAPINT) and operating cycle (OC). While persistence,
68
PERS and predictability, PRED being time series attributes are tested with standard
deviation of revenue (STDREV).
Since one the objectives of this study is to examine whether the ownership structure of
companies and the alleged conflict between controlling and non-controlling parties
manifest itself in earnings quality, this study inevitably is concerned with the
discretionary control. Thus where appropriate the effects non-discretionary factors of
earnings need to be separated out or controlled.
4.2.2 Ownership structure
As described earlier ownership structure refers to the distribution of control (measured
by the voting rights) and ownership (measured by cash flow rights) or rights to
benefits/cash. This research, based on past theoretical analyses and empirical
researches, examines whether ownership structure is associated with expropriating
behavior or inappropriate practices by the controlling party, which then leads to poorer
earnings quality and higher required return cost of equity. A controlling party holds
more than 20% of shares. A controlling party can be an individual or a group of related
individuals. A group of individuals are related if they are of the same family or hold the
shares through a single common entity such as a company or a partnership. The
relationship between individuals is analyzed from disclosure of analysis of
shareholders’ in the financial reports.
For this purpose companies are divided between those with pyramidal structure (PYS)
and those without pyramidal structure (NPYS). For PYS both cash flow and voting
69
rights of companies are collected and for NPYS the cash flow and voting rights are
equal.
Further, for PYS the ownership structure measure is the cash flow to voting rights ratio.
The lower the ratio, the larger the disparity between cash flow and voting rights and the
wider is the separation between ownership and control thus the higher is the expectation
of expropriating behavior.
4.2.2.1 Calculation of the ownership structure variable
4.2.2.1.1 PYS companies
The calculation of cash flow and voting rights is based on the method used in Claessens
et. al. (2000), and in other researches (Fan & Wong 2002). Voting rights is taken as the
‘weakest link’ in the chain of voting rights. The main weakness in this method is that it
does not take into account the existence of other controlling shareholders. The inclusion
of the other substantial shareholder addresses this weakness.The following diagrams
illustrate examples on page 91 of Claessens et. al. (2000).
Example 1. A family owns 11% of company A. Company A in turns own 21% of
company B. What is the degree of separation of ownership and control of company B?
Example 2.
Family
11%
A
21%
B
Ownership, cash flow rights = 11% x 21% = 2% Control, voting rights = 11%, the ‘weakest” link.
70
Example 2. A family owns 11% of company A. Company A in turns own 21% of
company B. The same family owns 25% of company C, which also holds 7% of
company B. What is the degree of separation of ownership and control of company B?
Family 11% 25% A 21% C 7% B
= 4%
Voting rights = 11% + 7% = 18%
Cash flow rights = 11% x 21% + 25% x 7%
Taking the weakest link in the voting rights chain , means taking the minimum
disparity between cash flow and voting rights.
For the sampled companies in Malaysia the cash flow and voting rights chain will be
extracted and analyzed from the shareholder’s statistics pages of the annual report. It is
also necessary to use information on the company profile, such as structure of the whole
group of companies in which the company belongs, which is sourced from annual
reports or the official website of the company. The following shows such calculations
for APM Automotive Holdings Bhd and ACP Industries Berhad, both of which are
companies with IBES earnings forecasts. Appendices 1 and 2 are shareholder’s statistics
extracted from respective companies annual reports and other companies’ annual reports
within the group that are relevant.
71
APM Automotive Holdings Bhd
For APM, additional information is obtained from the website
http://mgv.mim.edu.my/Newspaper/0107/0107235.Htm. Tan Chong Consolidated Sdn
Bhd holds in total 22.667% (17.15 %+ 3.35% + 1.4899% + 0.6705%) of APM, directly
and through nominees. From the website, it is also established that Tan Chong wholly
own Parasand Ltd a foreign incorporated company.
It is also established that the Tan family controls Tan Chong Consolidated Sdn Bhd.
Tan Chong Consolidated Sdn Bhd
100%
22.667% Parasand Ltd 20.0248% APM
Cash flow rights = 22.667% + 100% (20.0248%) = 42.6918% Voting rights = 22.667% + 20.0248% = 42.6918%
ACP Industries Berhad
The structure of ACP could be understood by also looking at Metacorp Berhad and
MTD Capital Bhd shareholder’s statistics. The group could be traced to Dato’ Dr Nik
Hussain Abdul Rahman (NHAR), although there are other substantial shareholders.
NHAR and family members are substantial shareholders of Nikvest Sdn Bhd and
Alloy Consolidated Sdn Bhd.
72
100%
Lambang Simfoni Sdn Bhd
74.17%
Metacorp Berhad
29.02%
ACP Industries Berhad
Nikvest Sdn Bhd Alloy Consolidated Sdn Bhd
21.18% 15%
MTD Captial Berhad
Cash flow rights = (21.18%x74.17%x29.02 ) + ( 15%x74.17%x29.02) = 7.78% Voting rights = 21.18% + 15% = 36.18%
4.2.2.1.2 NPYS companies
NPYS companies will be analyzed into widely held or manager control, and not widely
held. Widely held is the situation where none of the shareholders have more than 20%
shareholdings. In other words no shareholder has gained effective control and therefore
control is in the hands of manager. For these manager controlled companies, the voting
rights equals the cash flow rights which is simply the percentage holdings of shares by
the manager if any. This is consistent with Jensen and Meckling (1976) analysis,
although they begin from 100% owner controlled situation without outside
shareholdings. The agency related problems begins as outside shareholdings exist. Thus
manager controlled situation is the agency problem at its worst.
Consider for the moment the interest alignment theory, the lower the voting rights held
by the controlling manager the higher is the expectation of inappropriate behavior. Thus
73
this is consistent with the reading of cash flow to voting rights ratio of PYS companies
and expectation of inappropriate behavior by the controlling party of the PYS
companies.
For non-widely held companies, the cash flow/ voting rights of the shareholders with
the highest shareholdings will be documented. Even though there is no disparity
between cash flow and voting rights and considering the interest alignment theory, the
lower the voting rights held by the controlling shareholder the higher is the expectation
of inappropriate behavior. Thus this is consistent with the reading of cash flow to voting
rights ratio of PYS companies and widely held companies described earlier, and
expectation of inappropriate behavior by the controlling party of the PYS companies
and widely held companies.
However, as Mock et al (1988) found there is a possibility that as the controlling party
voting rights increases the controlling party becomes entrenched and the opposite effect
to the effect of interest alignment theory would be found. This is an empirical question.
As such for the sample under study, consideration is made first of whether the interest
alignment effect or the entrenchment effect is taking place.
4.2.2.2 Type of ultimate controlling party
The ultimate controlling party will be classified as managerial, institutional,
governmental, family or company. This measure captures an aspect of control that is
beyond percentage holdings of voting power. As described by Lim (1981), Claessens
74
(1998b),Gomez and Jomo (1999) and Gomez (2002), it appears that who controls could
also aggravate the separation of control and ownership conflict.
For a widely held company, the ultimate controlling party is the manager. An
institutional controlling party is a private sector institution such as a private trust fund
(Public Ittikal Fund). A government controlled company is a company whose ultimate
controlling party is a government agency (federal or state). Such agencies are given on
the website of Khazanah Nasional (http://www.khazanah.com.my/) and they include
such institutions as Kumpulan Wang Simpanan Pekerja, Permodalan Nasional Bhd
(and all the funds under it) and Lembaga Urusan Tabung Haji.
Family members as well as a group of connected individuals are classified as family.
For examples, from the website , Datuk Tan Kim Hor, a director of APM and an elder
founding family member. Therefore the ultimate controlling party of APM is classified
as family.
For some other companies tracing the ultimate controlling party could be more
complicated as shown in the following extract of analysis of shareholdings for MTD
Capital Bhd, as at 21 July 2004. The direct interest disclosure shows a company Nikvest
Sdn Bhd to be the controlling party. The detail analysis of deemed interest shows that
Nikvest Sdn Bhd and Alloy Consolidated Sdn Bhd are held by Dato’ Nik Hussain and
his family members. By virtue of their shareholdings in both these companies Ruslan
Sulaiman and Mohd Dom Ahmad, though appear not to have any family connections to
Dato’ Nik Hussain would be deemed to be related to him. The ultimate holding
company of MTD Capital is classified as family.
75
Table. 4.2 MTD Capital Analysis of Shareholding
Direct Interest
Indirect Interest
Shareholders
Number of Shares
%
Number of Shares
%
Nikvest Sdn Bhd 58,444,940 21.18 - - Alloy Consolidated Sdn Bhd 43,271,750 15.68 8,406,066 3.0511
Dato’ Dr Nik Hussain bin Abdul Rahman
71,104 .03 63,362,494 22.96 2
Nik Fauzi Dato’ Nik Hussein - - 58,444,940 21.18 3
Nik Faizul Dato’ Nik Hussein - - 58,444,940 21.18 3
Datin Nik Fuziah Dato’ Nik Hussein
- - 51,677,816 18.72 4
Ruslan Sulaiman - - 56,881,816 20.61 5
Mohd Dom Ahmad - - 56,881,816 20.61 5
Employees Provident Fund 25,156,500 9.11
1. Deemed interest via its wholly owned subsidiary Alloy Concrete Engineering 2. Deemed interest by virtue of his spouse shareholdings in MTD Capital and his children’s
shareholdings in Nikvest Sdn Bhd. 3. Deemed interest by virtue of their major shareholdings in Nikvest Sdn Bhd. 4. Deemed interest by virtue of her major shareholdings in Alloy Consolidated Sdn Bhd. 5. Deemed interest by virtue of their major shareholdings in Alloy Consolidated Sdn, Alloy
Concrete Engineering and other private companies. 4.2.3 Audit committee characteristics
4.2.3.1 Independence
This study uses measures of independence of audit committee as those used in Klein
2002. Audit committee independence is measured by whether the committee consists of
insiders, outsiders and affiliates. Insiders are current employees. Outsiders are those not
affiliated with the company other than being a director. Affiliates are former employees,
relatives of the CEO or board interlocks. The affiliate measure attempt to capture the
relationship which makes the outside committee member deemed non-independent.
76
4.2.3.2 Competence
Audit committee competence is the proportion of the members that has accounting or
finance knowledge background. By having accounting/ finance knowledge background
means either having qualification in accounting/finance related discipline or experience.
Therefore there is a reasonable expectation that the audit committee members are able
to detect irregularity in accounting information.
4.2.4 Substantial shareholders
Substantial shareholders are those with shareholdings of more than 5% and listed as
such in the analysis of shareholders’ statistics in the financial reports. Having identified
the ultimate controlling party, the shareholder with the next highest shareholding is
identified as the substantial shareholder. This substantial shareholder is therefore not
related to the ultimate controlling party and expected to have a monitoring role.
Referring to MTD Capital Bhd example earlier the substantial shareholder is the
Employees’ Provident Fund with voting rights of 9.11%. For this purpose Ruslan
Sulaiman and Mohd Dom Ahmad, though appear not to have any family connections
with Dato’ Nik Hussain are deemed to be related as they all have interest in various
companies. The truly unrelated is the Employees’ Provident Fund.
4.2.5 Measures for the market assessment of earnings quality
Market assessment refers to the effect of information risk on required return or the cost
of equity. The market requires a higher rate of return for equity that poses higher
77
information risk. The cost of equity is primarily used as the required return by the
market as the relationship is examined in the equity market. Thus a higher cost of equity
implies this group of market participants require a higher return for a lower earnings
quality which poses information risks.
Measures based on realized return is also used to examine if the general market
participants similarly priced information risks. And assuming rational expectation
theory required return by the market is actual realized return.
4.2.5.1. Cost of equity
Two alternatives cost of equity measures will be used. First, the cost of equity is
estimated as in Botosan (1997) and Chen et al (2003), based on a residual income model
(Gebhardt, Lee and Swaminathan 2002) using earnings forecasts by professional
analyst..
The cost of equity COE is the internal rate of return that equates the current value of a
company and the intrinsic value of the company. The current value is the current share
price. The intrinsic value is the current book value of equity and present value of future
abnormal earnings. The infinite form of the relationship between current and intrinsic
values is as follows :
78
79
∝ Pt = BBt + ∑ [ (Et (ROEt+i )– COE ) B t+i-1 ]
i=1 (1+COE) i
where,
Pt - Share price at time t
BBt - Book value at time t
ROEt+I - Return on equity at time t
COE - Cost of equity
E(--) - Expectation based on information at t
E(ROEt+i ) = Forecast earnings per share (FEPS t+i) / B t+i-1
FEPS - Forecasted earnings per share
As in Chen et al (2003), the finite version of the above model will be used to estimate
cost of equity, COE. In the finite version, an equilibrium ROE at time T, where T is a
terminal period needs to be estimated as the terminal ROE of the above summation
series. There is no appropriate method to determine T, Chen et al (2003) has chosen 6.
However they did estimate COE at various T from 4 to 12 and found that estimates of
COE are highly correlated. See table 4.3 for example of the calculation of the cost of
equity.
Table 4.3 Calculation of cost of equity (COE) for ACP INDUSTRIES
t+1 t+2 t+3 t+4 t+5 t+6 t+7 t+8 t+9 t+10 t+11 t+12 Div Payout ratio
0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3
FEPS
0.026 1 0.1411 0.76461 0.8433 5 0.916 0.978 1.026 1.055 1.061 1.041 0.993 0.915
Beginning Book value 2
2.79 2.8082 2.9069 3.4421 4.0328 4.6743 5.3592 6.0773 6.8155 7.5580 8.2865 8.9815
Return on Equity (ROE)
0.00931 3 0.050213 0.2633 0.245 4 0.2274 0.2094 0.1914 0.1744 0.1564 0.1384 0.1204 0.1024 Target ROE
COE 8 0.1376 0.1376 0.1376 0.1376 0.1376 0.1376 0.1376 0.1376 0.1376 0.1376 0.1376 0.1376 Book value
Streams residual income discounted at COE
Share Price
2.79 0-.3146 6 -0.1896 0.2476 0.2210 0.1897 0.1547 0.1169 0.0778 0.0384 0.00023 -0.03571 -0.06826 -0.49605 7 2.73
The publicly available data required are the dividend payout ratio (which is either a company’s historical three year median or the industry three year median),
short term growth rate (extrapolated from the two years IBES forecasted earnings), analysts forecasted earnings for two subsequent years (FEPS, which are IBES
forecasted earnings), the target return on equity (industry historical five-year median), book value (at balance sheet date) and share price at balance sheet date. To
operationalize the above formula the purpose of this estimation is to obtain a COE that when applied to the residual income of a company, that is the summation series
in the above formula, the sum of the current book value 2.79 and the discounted residual income of the company equals the current share price. Thus that is the
80
implied required return by the market.
The steps in estimating COE is :
1. FEPS for t+1 and t+2 are 0.026 and 0.141. The implied growth rate = (0.141-0.026)/0.026 = 4.42. thus
FEPS t+3 = (1+g) x FEPS t+2=( 4.42+1) x .141 = 0.7646
2. The book value per share for end of financial year 2004( or beginning of 2005) is 2.79. For subsequent years the book value (B) is estimated as given
below.
B t+1 = B t + EPS (1-Payout ratio)
for example for 2005 , B = 2.79 + 0.026(1-0.30) = 2.8082
3. The return of equity for years t+1, t+2 and t+3 is FEPS/ Beginning book value. For example ROE t+1 = 0.026/2.79 = 0.00931
4. The return on equity for years t+4 onwards are assumed to be approaching the industry target of 0.102. In other words this method assume that in the long
run each company’s return will approach the industry’s average. An interpolation is carried out for years between t+4 to the industry target . For example ROE
t+4 = 0.263 – (0.263-.102)/9 = 0.245. At this stage streams of book value and ROE estimates are completed.
5. FEPS t+4 onwards can now be completed by taking ROE x Book value estimates. For example FEPS t+4 = 0.245 x 3.4422 = 0.8433
6. For each of the 12 years, beginning from year t+1 (or 2005) the residual income is estimated as follows :
(ROE t+1 – COE) x (1/(1+COE) 1….12) x B t
So for t+1 , (0.00931 – COE)/ (1/(1+COE) 1) x 2.79
81
82
7. In this finite version of the model a terminal value is estimated as (ROE t – COE) x (1/((1+COE) 12 COE)) x B t
8. The value of COE, 0.1376, is obtained by trial and error, in such a way that makes the sum of the book value (2.79), the series of discounted residual
income and the last terminal value equals the share price 2.73.
The second alternative measure of cost of equity referred to as, COEA, is estimated
based on Ohlson and Juettner-Nauroth (2000) and operationalize in Chen et al (2003) as
follows.
Pt= FEPS t+1 + (FEPS t+2 –FEPS t+1 – COEAxFEPS t+1(1-POUT)) COEA COEA(COEA- g)
Which is then rearranged to find COEA,
COEA =A + ( A2 + FEPS t+1 (FEPS t+2 –FEPS t+1 - g ) ) 1/2
Pt FEPS t+1
Where, A = ½ ( g + POUT x FEPS t+1 ) Pt
Pt - Share price at time t
FEPS - Forecasted earnings per share
POUT - Dividend payout ratio
g - estimated long term growth
Example : ACP Industries
A = 0.5 (0.014 + 0.3 x 0.026 ) 2.73
= 0.00843
COEA = 0.00843 + (0.00843 2 + 0.026 ( 0.141 – 0.026 - 0.014)) 0.5
2.73 0.026 = 0.2135
This alternative measure, COEA, avoids using book value in the estimation whilst COE
estimation uses book value. Another important difference is that the long term growth in
83
COEA estimation is proxied by the inflation rate, as in Chen et al (2003), at 1.4% for
year 2004 and that is the company wide economic growth. Whilst in COE estimation
earnings grow towards the industry return on equity.
4.2.5.2 Market return
The general market cost of equity or required return is measured by realized returns.
Based on rational expectation theory (Copeland and Weston 1988, pages 346-350),
realized return is a proxy to expected return as assumed in Francis et. al 2004.
Realized return = Pt – P t-1 P t-1
Where, Pt is price at time t. This study uses the average monthly return (AVMR).
Alternatively this study also uses Fama and French (1993) excess return (ER) as used in
Aboody et al (2005) and Francis et al (2004).
ER = R i,t – R f,t = α +β(R m,t – R f,t) + γ Size + ∂ BTMV
R i,t – Return on company security at time t
R f,t – Risk free return at time t (12 month Treasury bills rate)
R m,t – Market return at time t (return on composite index)
Size – Market value
BTMV – Book to market value
In terms of the applicability of the Fama and French (1993) in the Malaysian context,
Abdul Rahim (2006) and Abdul Rahim and Mohd. Nor (2008) showed that the forecast
84
error between Fama and French (1993) model and two other models based on liquidity
are not significantly different. The liquidity models are based on the belief that liquidity
is an important factor in developing market that explains excess return. However at this
stage more work need to be done to develop the liquidity based models (Abdul Rahim
& Mohd. Nor 2008).
4.2.6 The validity of the cost of equity measures
There has been discussion in the literature regarding the validity of ex ante cost of
equity measures. Botosan and Plumlee (2004) assesses the construct validity of four
alternative measures of cost of equity. They found two form of cost of equity estimates;
one derived from dividend discount model and another from price-earnings-growth
relationship, that is the Easton model, to be valid in terms of capturing the cross-
sectional variation in COE. Nevertheless this study uses cost of equity estimates
described above as the Easton model may limit the sample. The residual income cost of
equity estimate is equivalent to the dividend discount model estimate (Gebhardt et al
2001) . Nevertheless, a validity test will be carried out as the result from Botosan and
Plumlee (2004) assessment is data specific. A valid measure of COE should have a
positive relationship with risk, β, as the higher the risk the higher the required return to
compensate equity holder for bearing more risk. COE should also have a negative
relationship with size (Botosan 1997). The smaller the size of the company the higher
the risk and the higher the required return.
85
4.3 Data analysis
4.3.1 Theoretical /research framework
EQ
MA OS
1 2
3
4
Monitoring elements i.e AC and SS
1
3
4
Figure 4.3.1 Theoretical/ research framework
Separation of control and ownership, monitoring mechanisms and earnings quality
Earnings quality 1,2,3,4,5 = α0 + αi CFVR + α2,3,4,5 UCP + α6 SSVR + α7 ACI + α8
ACC + α9SIZE + α10Controls for EQ + δ1,2,3,4,5
Equation 1
Refer to section 3.1 for detail explanation of how the relationship is derived. CFVR for
PYS companies is the ratio of cash flow to voting rights and for NPYS is the cash flow
rights. The coefficient αi is expected to be negative as the higher the CFVR, the lower
the separation of ownership and control, and thus the higher the quality of earnings. As
low measure of the quality of earnings is associated with high earnings quality, the
association is expected to be negative. The variable UCP are the four dummy variables,
thus if any type of ownership could explain earnings quality the coefficients α2,3,4,5 are
expected to be positive. As explained in section 3.1 as the substantial shareholder could
86
be potentially monitoring or colluding with the controlling party, the coefficient α6 is
expected to be either positive or negative. The coefficients for audit committee
independence and competence, α7 and α8 , are expected to be negative as the higher the
measure of each audit committee characteristics, the higher the earnings quality, the
lower the earnings quality measure.
Controls for earnings quality are capital intensity, operating cycle and standard
deviation of revenue. The reasons for these controls are discussed in section 4.2.1.3
Prediction :
Coefficients Sign αi Negative
α2,3,4,5 Positive α6 Positive/Negative α7 Negative α8 Negative α9 Negative α10 Negative for CAPINT,
and Positive for OC and STDREV
Earnings quality and cost of equity
COE = φ0 + φ1,2,3,4Earnings quality 1,2,3,4 + φ5 SIZE + φ6 β + φ7 BTMV + δ6
Equation 2
Prediction:
Coefficients Sign Φ1,2,3,4 Positive Φ5 Negative Φ6 Positive Φ7 Positive
Refer to section 3.2 for detail explanation of how the relationship is derived. The
coefficients of each earnings quality measure φ1,2,3,4 is expected to be positive as the low
earnings quality (high measure) is expected to be associated with high required return,
87
COE. As established in section 4.2.6, COE is negatively related to size, thus φ5 is
expected to be negative, and positively related to beta and book to market, and thus φ6
and φ7 are expected to be positive.
Relationship between ownership structure, monitoring mechanisms and cost of
equity
COE = μ0 + μ1 CFVR + μ 2,3,4,5 UCP + μ 6 SSVR + μ 7 ACI + μ 8 ACC +
μ 9SIZE + μ 10BTMV + μ 11β + δ7,8,9
Equation 3
Refer to section 3.3 for detail explanation of how the relationship is derived. The
direction of each of the coefficients for CFVR, UCP, SSVR, ACI and ACC are the same
for the coefficients of each of these variables in equation 1 for similar reasons.
Prediction:
Coefficients Sign μ1 Negative
μ 2,3,4,5 Positive μ 6 Positive/Negative μ 7 Negative μ 8 Negative μ 9 Negative μ 10 Positive μ 11 Positive
Relationship between ownership structure, monitoring mechanisms and cost of
equity
CFVR = η0 + η1COE + η2,3,4,5UCP + η6 SSVR + η7 ACI + η8 ACC +
η9SIZE + η10BTMV + δ10,11,12
Equation 4
88
Refer to section 3.4 for detail explanation of how the relationship is derived. The
coefficient η1 is expected to be negative as the required return, COE increases, the
controlling party may reduce the disparity between cash flow and voting rights. The
coefficients η2,3,4,5 are expected to be positive as UCP are the four dummies. Again as in
previous equations, it is expected that the substantial shareholders could either be
monitoring or colluding, thus η6 is expected to be positive or negative. The coefficient
for size η9 is expected to be positive as the larger the size of the company the lesser the
disparity between cash flow and voting rights, the higher the ratio. The converse applies
to book to market and CFVR.
Prediction:
Coefficients Sign η1 Negative
η2,3,4,5 Positive η6 Positive/Negative η7 Negative η8 Negative η9 Positive η10 Negative
Relationship between substantial shareholders’ voting rights and elements of
ownership, monitoring mechanisms and cost of equity
In addition, as an exploratory exercise, SSVR could be explained by the ownership
structure elements and other monitoring mechanisms, the following equation is also
tested:
SSVR = ω0 + ω 1COE + ω 2,3,4,5UCP + ω 6 CFVR + ω 7 ACI + ω 8 ACC +
ω 9SIZE + ω 10BTMV + δ10,11,12
89
4.3.2 The simultaneity of equations
Equations 1 and 2, a priori, are not regarded as simultaneous equations. The earnings
quality variables in equation 2 is an exogenous regressor, that is externally determined
or given. In statistical term the earnings quality variables in equation 2 is not expected
to be correlated with the error term.
However, the simultaneity of equations 3 and 4, a priori, cannot be disregarded as the
formulation of the equations are based on the expectation that the ownership structure
variable, CFVR, could explain the cost of equity and other variables, and be explained
by the latter as well. As discussed in chapter 2, drawing from Demsetz (1983), Demsetz
(1985), Demsetz and Villalonga (2001), and Mak and Li ( 2001) CFVR could be
expected to be endogenous.
By the same arguments, the substantial shareholders’ voting rights (SSVR) too could be
expected to be endogenous. The expected relationship between cost of equity and
substantial shareholders’ voting rights, where the cost of equity is the dependent
variable, as hypothesized above, is based on the expectation of the monitoring effect or
otherwise of the substantial shareholders. However as argued by previous researchers
and as with the ultimate controlling party, substantial shareholders could change their
holdings with changes in cost of equity or other ownership and other variables.
As discussed in Gujarati (1995), if simultaneity exists, the ordinary least square method
would not provide consistent and efficient estimates and thus erroneous inference might
be made. To handle simultaneity, 2 stage least square method would provide consistent
90
estimates. However if the 2 stage least square method is performed when in fact there
are no endogenous regressors, the estimates produce are consistent but not efficient.
Therefore, the two suspected variables are first tested for endogeneity.
Following procedures in Gujarati (1995), reduced form equations are obtained by
regressing each of suspected endogenous variables with other independent and
exogenous variables as follows:
Reduced form equation 1 :
CFVR = a0 + a1,2,3,4UCP + a5 ACI + a6 ACC + a7SIZE + a8BTMV + δ1
Reduced form equation 2 :
SSVR = b0 + b1,2,3,4UCP + b5 ACI + b6 ACC + b7SIZE + b8BTMV + δ2
From these two reduced form equations the predicted value of CFVR, ^CFVR, and
predicted value of SSVR, ^SSVR, are obtained. Then COE is regressed with the
original regressors as in equation 3 above and these two predicted values.
COE = μ0 + μ1 CFVR + μ 2,3,4,5 UCP + μ 6 SSVR + μ 7 ACI + μ 8 ACC +
μ 9SIZE + μ 10BTMV + c1 ^CFVR + c2 ^SSVR + δ7,8,9
If CFVR and SSVR are exogenous, c1 and c2 are zero. Therefore F test will test if c1 = c2
=0. If CFVR and SSVR are exogenous then ordinary least square could be used
otherwise the two stage least square method will be used to estimate equation 3 and 4.
91
The two stage least square method used is the method developed by Theil and Basmann
as described in Gujarati (1995). Reduced form equations are formed for all endogenous
variables to obtain their predicted values. Thus in addition to the above two equations, a
third for COE is needed to obtain ^COE:
Reduced form equation :
COE = c0 + c1,2,3,4UCP + c5 ACI + c6 ACC + c7SIZE + c8BTMV + δ3
Equations 3 and 4, are estimated in this second stage using ordinary least square using
(predicted values) ^CFVR, ^SSVR and ^COE as the regressors substituting for the
actual values.
Relationship between ownership structure and cost of equity
COE = μ0 + μ1 ^CFVR + μ 2,3,4,5 UCP + μ 6 ^SSVR + μ 7 ACI + μ 8 ACC +
μ 9SIZE + μ 10BTMV + μ 11β + δ7,8,9
Relationship between ownership structure and elements of ownership, monitoring
mechanisms and cost of equity
CFVR = η0 + η1^COE + η2,3,4,5UCP + η6 ^SSVR + η7 ACI + η8 ACC +
η9SIZE + η10BTMV + δ10,11,12
92
Substantial shareholders’ voting rights could be tested if it could be explained by the
ownership structure and other variables by estimating the following equation in the
same manner as testing CFVR.
SSVR = ω0 + ω 1^COE + ω 2,3,4,5UCP + ω 6 ^CFVR + ω 7 ACI + ω 8 ACC +
ω 9SIZE + ω 10BTMV + δ10,11,12
4.3.3The choice of audit firm
The audit firm may reflect the quality of audit and hence may influence the earnings
quality. This factor is controlled in this research as the listed companies in the sample
are overwhelmingly audited by one of the big four firms. Only 29 of the companies are
audited by the non-big audit firms.
93
CHAPTER 5
RESULTS
5.1 Data sources
Accounting data for the purpose of estimating earnings quality variables (ABRES,
ABSDATCA, ABSDATA, PERS and PRED), market value, book to market value,
prices and beta are obtained from Datastream data base for the financial year end 2004
or as at financial year end 2004 as applicable. For the purpose of estimating ABRES
accounting data for year 2003 and 2005 are also required. Estimated earnings per share
for years 2005 and 2006 required for calculating cost of equity are downloaded from
Bloomberg data base services in January 2005. Data on audit committee, voting rights
of substantial shareholders, cash flow and voting rights of ultimate controlling party are
collected from the annual reports for the financial year ending in 2004.
5.2 Sample profile
The sample comprises of companies with IBES’s estimated earnings per share for year
2005 and 2006. There are originally 213 companies from which 10 companies are
eliminated due to insufficient data available for the estimation of cost of equity. (Table
5.1). Refer to Appendix 3 for full list of companies. Table 5.2 provides the breakdown
of the sample into various industry classification.
94
Table 5.1 Sample Number of companies with estimated earnings forecasts 213 Eliminated due to change in accounting year end and insufficient data to calculate cost of equity
10
203 Table 5.2 Classification by industry Industry Construction 11 Consumer product 26 Finance 13 Hotel 3 Industrial product 55 Infrastructure project companies 5 Plantations 7 Properties 23 Technology 10 Trade and services 50 203
Since the data requirements varies for the calculation of the earnings quality variables,
the composition of companies are further reduced into three samples- 1) ABRES, 2)
ABSDATCA/ABSDATA and 3) PERS/PRED.
Companies in ABRES sample are those with available data to calculate accrual quality
based on Dechow and Dichev (2002). Companies in ABSDATCA/ABSDATA sample
are those with available data to calculate discretionary accruals, both current and total
accruals versions, based on modified Jones (1991) model. Those companies with
available data to estimate both persistence and predictability of earnings are in sample
termed PERS/PRED.
Table 5.3 shows the number of companies in each sample. There are 70 companies that
are common in all three samples.
95
Table 5.3 Sample size based on available data for the calculation of earnings quality variables Number (%)Share of
Market Capitalization
ABRES 141 31 ABSDATCA /ABSDATA 151 32 PERS /PRED 118 27
Table 5.4 provides a breakdown of companies in each sample into those with various
type of ultimate controlling party. Table 5.5 shows the proportion of companies that are
with ultimate controlling party having controlling rights through layers of companies
(pyramidal ownership) and those that are with ultimate controlling party having direct
controlling rights (non-pyramidal).
Table 5.4 Breakdown of companies with various types of ownership
Sample Managerial Institutional Governmental Family Company ABRES (141) 16
(11%) 2
(1%) 25
(19%) 96
((68%) 2
(1%) ABSDATCA /ABSDATA (151)
16 ((11%)
2 (1%)
24 (16%)
106 (70%)
3 (2%)
PERS/PRED (118)
12 (10%)
1 (0.8%)
24 (20)
79 (68%)
2 (1.2%)
Table 5.5 Breakdown of companies with pyramidal(PYS) and non-pyramidal
ownership(NON-PYS) Sample PYS NON-PYS ABRES (141) 40
(28%) 101
(72%) ABSDATCA /ABSDATA (151)
46 (30%)
105 (70%)
PERS/PRED (118)
47 (40%)
71 (60%)
96
5.3 General descriptive statistics
Tables 5.7, 5.8, 5.9 and 5.10 show the descriptive statistics for each variable in the three
samples. The values of each variable used in the regressions are as estimated or
calculated. Only for both the discretionary accruals variables (ABSDATCA and
ABSDATA) the logged form is used as the calculated form is highly skewed as shown
in table 5.8. As given in table 5.9 the skewness and kurtosis problems of the logged
form (LABSCA and LABSTA) is much lesser than the original form of the variable.
Even though some of the other variables are skewed and peaked, they are not
transformed as the transformed variables are not much improved and to prevent further
reduction of sample size. Besides as discussed in Tabachnik. and Fidell (2001)
transformation poses interpretation problem and not widely recommended. Further, with
the given sample sizes Central Limit Theorem is relied on to predict normality. Market
values (MV) and book to market (BTMV) are transformed (LGMV and LBTMV) as
MV are large and transformation of BTMV achieved univariate normality.
The table 5.6 shows mean and standard deviation of the common variables for the three
samples. Cursory examination shows that there are not much difference in the measures
of central tendency and dispersion in all three samples.
97
98
Table 5.6 Mean and dispersion of common variables in the three sample
ABRES ABSDATCA & ABSDATA PERS and PRED
Mean Std.
Deviation Mean Std. Deviation Mean Std.
Deviation CFVR 0.490 0.218 0.488 0.216 0.506 0.225 SSVR 0.087 0.077 0.091 0.080 0.087 0.084 LGMV 6.585 1.377 6.555 1.338 6.955 1.401 BETA 0.961 0.413 0.932 0.475 1.008 0.483 LBTMV -0.394 0.645 -0.406 0.641 -0.283 0.633 ACI 0.634 0.200 0.636 0.196 0.608 0.221 ACC 0.339 0.165 0.345 0.170 0.341 0.197 COEA 0.147 0.066 0.149 0.065 0.145 0.070 COE 0.099 0.034 0.099 0.033 0.094 0.037
The standard deviations of the earnings quality measures are found to be large relative
to the means (tables 5.7,5.8 and 5.10). The mean to standard deviation ratios for the
earnings quality are as follows:
ABRES 0.039/0.039 = 1
ABSDATCA 0.090/0.211 = 0.43
ABSDATA 0.094/.208 = 0.45
PRED 0.171/0.242 = 0.71
PERS -0.215/0.414 = -0.52
However similar pattern is found in previous studies. Francis et al (2004) reported
mean/dispersion ratios of 0.026/0.023 = 1.13 for ABRES, -0.482/0.368 = -1.31 for
PERS and 0.876/1.054= 0.83 for PRED. Francis et al (2006) reported mean/ dispersion
of 0.0465/0.0283= 1.64 for abnormal accrual. Similarly for studies on Malaysian data,
large standard deviations relative to means are reported. In Mohd Salleh et al (2007) a
mean to standard deviation of -0.013/0.148 = -0.088 is reported for discretionary
accruals. Mohd Salleh et al (2005) reported similar ratio of -0.007/0.166= 0.042.
Table 5.7 Descriptive Statistics - ABRES SAMPLE
N Minimum Maximum Mean Std.
Deviation Skewness Kurtosis
Statistic Std. Error Statistic
Std. Error
CFVR 141 0.006 1.000 0.490 0.218 0.515 0.204 2.522 0.155 0.406 0.382 SSVR 141 0.000 0.401 0.087 0.077 0.814 0.204 3.990 0.928 0.406 2.289 ABRES 141 0.001 0.209 0.039 0.039 1.682 0.204 8.241 3.095 0.406 7.633 LGMV 141 4.222 10.300 6.585 1.377 0.582 0.204 2.849 -0.221 0.406 -0.544 BETA 141 0.160 2.100 0.961 0.413 0.520 0.204 2.546 -0.029 0.406 -0.070 LBTMV 141 -3.013 1.393 -0.394 0.645 -0.559 0.204 -2.738 1.910 0.406 4.709 ACI 141 0.000 1.000 0.634 0.200 -1.036 0.204 -5.077 1.888 0.406 4.655 ACC 141 0.000 0.750 0.339 0.165 0.317 0.204 1.555 0.438 0.406 1.079 COEA 141 0.021 0.520 0.147 0.066 2.190 0.204 10.729 8.333 0.406 20.547 COE 141 0.014 0.263 0.099 0.034 1.150 0.204 5.632 4.635 0.406 11.430
99
Table 5.8 Descriptive Statistics - ABSDATCA & ABSDATA SAMPLE
N Minimum Maximum Mean Std.
Deviation Skewness Kurtosis
Statistic Std. Error Statistic
Std. Error
COE 151 0.014 0.263 0.099 0.033 1.173 0.197 5.945 4.838 0.392 12.333 COEA 151 0.021 0.520 0.149 0.065 2.087 0.197 10.570 7.794 0.392 19.865 ACI 151 0.000 1.000 0.636 0.196 -1.083 0.197 -5.486 2.051 0.392 5.229 ACC 151 0.000 0.750 0.345 0.170 0.321 0.197 1.626 0.323 0.392 0.823 CFVR 151 0.006 1.000 0.488 0.216 0.609 0.197 3.085 0.299 0.392 0.763 SSVR 151 0.000 0.401 0.091 0.080 0.749 0.197 3.793 0.429 0.392 1.093 LGMV 151 4.222 10.300 6.555 1.338 0.656 0.197 3.323 -0.040 0.392 -0.101 BETA 151 -1.460 2.100 0.932 0.475 -0.475 0.197 -2.406 3.558 0.392 9.069 LBTMV 151 -3.013 1.393 -0.406 0.641 -0.524 0.197 -2.656 1.762 0.392 4.492 ABSDATCA 151 0.000 2.364 0.090 0.211 8.863 0.197 44.899 92.330 0.392 235.343 ABSDATA 151 0.000 2.335 0.094 0.208 8.850 0.197 44.835 92.375 0.392 235.460
Table 5.9 Descriptive Statistics for transformed ABSDATCA and ABSDATA i.e LABSCA and LABSTA
N Minimum Maximum Mean Std. Deviation
Skewness Kurtosis
Statistic Std. Error
Statistic Std. Error
LABSCA 151 -8.957 0.860 -3.271 1.371 -0.512 0.197 -2.594 1.420 0.392 3.620 LABSTA 151 -9.053 0.848 -3.224 1.515 -1.152 0.197 -5.838 2.486 0.392 6.336
100
101
Table 5.10 Descriptive Statistics - PERS and PRED SAMPLE
N Minimum Maximum Mean Std.
Deviation Skewness Kurtosis
Statistic Std. Error Statistic
Std. Error
COE 118 0.014 0.263 0.094 0.037 1.168 0.223 5.244 4.095 0.442 9.267 COEA 118 0.021 0.520 0.145 0.070 2.181 0.223 9.793 7.687 0.442 17.395 ACI 118 0.000 1.000 0.608 0.221 -0.887 0.223 -3.983 0.928 0.442 2.100 ACC 118 0.000 1.000 0.341 0.197 0.784 0.223 3.522 1.371 0.442 3.103 CFVR 118 0.006 1.000 0.506 0.225 0.454 0.223 2.037 0.100 0.442 0.226 SSVR 118 0.000 0.401 0.087 0.084 0.856 0.223 3.843 0.492 0.442 1.113 PRED 118 0.007 1.429 0.171 0.242 3.321 0.223 14.912 12.487 0.442 28.259 PERS 118 -2.439 0.608 -0.215 0.414 -1.407 0.223 -6.319 6.143 0.442 13.902 BETA 118 -0.120 2.100 1.008 0.483 0.287 0.223 1.287 -0.433 0.442 -0.979 LGMV 118 4.222 10.456 6.955 1.401 0.304 0.223 1.367 -0.394 0.442 -0.892 LBTMV 118 -3.013 1.393 -0.283 0.633 -0.944 0.223 -4.241 3.316 0.442 7.505
5.4 Descriptive statistics for ownership structure
A majority of companies in all three samples examined has a family as the ultimate
controlling party (refer to table 5.4). The majority of around 68% to 70% is higher than
that found in Claessens et al (1998b) study in which the sample comprises of companies in
Malaysia and 7 other countries in East Asia . In the latter it was found to be around 46%.
The next higher type of ultimate controlling party is government which is between 16% to
20%. In Claessens et al (1998b) study government owned corporation is found to be 7%.
Consistent with the findings in La Porta et al (1999) that companies all over the world are
not widely held, the samples in this study consist of only around 11% widely held
companies (ultimate controlling party classified as manager).
The cash flow, voting/control rights and ratio of voting/control rights are found to be higher
than those reported in Claessens (1998b) study. For the sample of 238 Malaysian
companies included in their report the mean of cash flow rights is 24%, voting/control
rights is 28% and the mean ratio of cash flow to voting/controlling rights is 85%. However
the means for cash flow and voting rights for the samples in this study are found to be
higher as given in the following table, whilst the ratio of cash flow to voting rights is lower
in this study than in Claessen (1998b).
102
Table 5.11 Descriptive statistics of cash flow and voting rights and ratio of cash flow to voting rights
ABRES ABSDATCA PERS/PRED CF VR CFVR CF VR CFVR CF VR CFVR Mean 0.41 0.54 0.49 0.40 0.53 0.49 0.39 0.53 0.51 Minimum 0.01 0.20 0.01 0.01 0.20 0.01 0.01 0.20 0.01 Maximum 0.89 1.00 1.00 0.89 1.00 1.00 0.89 1.00 1.00 Std Dev 0.19 0.23 0.22 0.18 0.22 0.22 0.19 0.22 0.23 Percentiles 25 0.28 0.38 0.32 0.28 0.36 0.32 0.24 0.37 0.34 50 0.41 0.50 0.50 0.41 0.48 0.48 0.40 0.48 0.51 75 0.54 0.61 0.60 0.53 0.60 0.60 0.54 0.60 0.60
Claessens et al (1998b)’s sample is taken in 1996, the year before the Asian financial crisis.
Thus for the year under study, the disparity indicated by the cash flow/voting rights ratio
appears to be higher than Claessens et al (1998b) study and at a higher level of control as
indicated by the mean voting rights.
Another point to note is that sample in Claessens et al studies (1998 b, 2000) is based on
availability of ownership structure data on Worldscope database which largely comprises of
large companies. The share of the total market capitalization of the Malaysian companies in
the sample is 74% (Claessens 2000). The share of market capitalization of the companies in
the samples in this study is between 27%-32%. Since it is based on availability of IBES
analysts’ earnings forecasts it comprises of companies of interest to analysts which are not
just large companies but also newly listed and not as large.
103
5.5 Descriptive statistics for the monitoring mechanisms
- substantial shareholders’ voting rights and audit committee characteristics
In both Claessens (1998(b)) and (2000) studies, they acknowledge but ignore the possible
existence of another controlling party and the effect of a substantial shareholder, the
shareholder with the next highest voting rights. These are shareholders who have
significant shareholding (above 3%) but may or may not have enough controlling rights. In
all the samples about 40 companies are without substantial shareholders. From the table
5.12 about 25% of the companies in the samples are with substantial shareholders with
voting rights above 13% for ABRES sample and above 15% for ABSDATCA/ABSDATA
and PERS/PRED samples.
Table 5.12 The descriptive statistics of substantial shareholders’ voting rights
ABRES ABSDATCA/ ABSDATA PERS/PRED
Mean 0.087 0.09 0.09 Median 0.081 0.08 0.08 Std. Deviation 0.077 0.08 0.08 Minimum 0.000 0.00 0.00 Maximum 0.401 0.40 0.40 Percentiles: 25 0.000 0.00 0.00 50 0.081 0.08 0.08 75 0.137 0.15 0.15
The mean of the audit committee independence is just over 60% in each of the three
samples (tables 5.7, 5.8 and 5.10) which just exceeds the majority required by the
Malaysian Code of Corporate governance. A lower percentage at just over 33% in each
sample is reported for the mean audit committee competence. It suggests that companies
104
are just fulfilling the requirement. In fact there even those without independent or
competent members.
5.6 Bivariate collinearity between earnings quality variables
The different earnings quality measures (except for ABSDATA and ABSDATCA)
represent different dimensions of earnings attributes. The values of the earnings quality
variables are to be read as large for poor earnings quality. For example large ABRES means
that current accruals are not well translated to cash flows (past, current and future) and
therefore earnings is of low quality. Large ABSDATCA and ABSDATA (discretionary or
abnormal accruals) are similarly read. Whilst large PRED, absolute residual of earnings
time series model, means earnings of low quality because current earnings does not predict
future earnings well, large PERS, coefficient of earnings time series model, means earnings
is highly persistent or earnings is sustainable, thus earnings of high quality. To be
consistent with the way other earnings measures are read, values for persistence are negated
such that large values means earnings of poor quality. The following table 5.13 provides
Pearson correlation coefficients between the different EQ measures.
105
Table 5.13 Pearson correlation coefficients between variables
ABRES ABSDATCA ABSDATA PRED PERS ABRES 1.00 ABSDATCA 0.22 1.00 (0.01)*** ABSDATA 0.19 0.99 1.00 (0.02)** (0.00)*** PRED 0.24 0.08 0.07 1.00 (0.02)** (0.43) (0.45) PERS -0.18 0.13 0.13 0.04 1.00 (0.07)* (0.19) (0.17) (0.69) The figures in italics and parenthesis are the probabilities. *** Probabilities that are significant at 1% level ** Probabilities that are significant at 5% level * Probabilities that are significant at 10% level
As in Francis et al (2004) each earnings quality measure is treated as distinct dimension of
earnings quality. The Pearson correlation coefficients between the earnings quality
variables are found to be between 21% to 48%. The rather low correlation (except for
between ABSDATA and ABSDATCA) is consistent with the view that they are distinct.
On the other hand Francis et al (2008b) (with different sample from Francis et al 2004)
reported correlation coefficients between earnings measures of between 50%-91%. Thus
they have used a common factor of earnings quality in their study.
5.7 Bivariate collinearity analysis of all variables in each sample
Tables 5.14, 5.15 and 5.16 show the Pearson correlation coefficients and the associated
significance level of all metric variables.
106
It is expected that the poorer the earnings quality, the larger would be the cost of equity, as
market penalizes poor earnings quality. The correlation is expected to be positive. The
measure cost of equity, COE, is only significantly correlated in the positive direction with
predictability, PRED (Table 5.16) and not with other earnings quality measures. The
alternative measure of cost of equity COEA however is significantly and positively
correlated with all earnings quality variables except, persistence, PERS.
The relationship between substantial shareholders’ voting rights (SSVR) and cost of equity
is not predicted given the mixed results from prior research. The role of a block holder or a
substantial shareholder could be either way in mitigating the effects of separation of
ownership and control. In all the samples SSVR is not significantly correlated with COE in
any way. However SSVR is consistently and significantly correlated in a negative direction
with COEA in all samples. The correlation is weakest for the PERS PRED sample i.e at
10% significant level.
Both measures of cost of equity are not significantly correlated with the other independent
variables; cash flow/voting rights, CFVR, audit committee competence, AC and audit
committee independence, ACI. As expected both measures are significantly correlated with
size, LGMV in a negative direction and with book to market, LBTMV in positive direction.
However, neither COE nor COEA are correlated with beta.
The correlation between the regressors appear to be trivial at this stage for example in
ABRES sample, between ACC and LBTMV, in ABSDATCA sample between ACI and
107
108
LABSCA/LABSTA. In the multivariate analysis multicollinearity is checked for the level
of severity in the problem.
It should be noted that substantial shareholders’ voting rights,SSVR and cash flow/ voting
rights, CFVR are consistently and significantly correlated in all samples in negative way.
This suggests interestingly that companies with a lower CFVR and therefore with a higher
expectation of problems associated with separation of ownership and control such as
expropriation by the ultimate controlling party, have a substantial shareholder with a higher
voting rights. On the other hand, companies with a higher CFVR, in other words with a
smaller disparity between cash flow and voting rights and therefore are expected to face
less separation of ownership and control related problems, have substantial shareholder
with a lower voting rights.
Table 5.14 Correlations - ABRES SAMPLE
COE COEA ACI ACC SSVR CFVR ABRES LGMV LBTMV BETA CAPINT OC
COE 1.000 COEA 0.722 1.000 (0.000)*** ACI 0.105 0.045 1.000 (0.215) (0.597) ACC 0.122 0.082 0.043 1.000 (0.148) (0.335) (0.609) SSVR -0.086 -0.225 -0.018 0.122 1.000 (0.308) (0.007)*** (0.836) (0.149) CFVR 0.102 0.068 -0.141 -0.083 -0.214 1.000 (0.227) (0.424) (0.094)* (0.328) (0.011)** ABRES 0.085 0.199 0.110 -0.077 -0.140 -0.087 1.000 (0.315) (0.018)** (0.192) (0.367) (0.098)* (0.305) LGMV -0.370 -0.330 -0.089 -0.123 -0.070 0.143 -0.203 1.000 (0.000)*** (0.000)*** (0.294) (0.147) (0.406) (0.090)* (0.016)** LBTMV 0.567 0.289 0.001 0.157 0.106 0.125 -0.188 -0.269 1.000 (0.000)*** (0.001)*** (0.990) (0.064)* (0.213) (0.140) (0.025)** (0.001)* BETA 0.116 0.053 0.028 0.127 -0.036 -0.068 -0.011 -0.113 0.130 1.000 (0.170) (0.535) (0.745) (0.134) (0.675) (0.421) (0.897) (0.182) (0.123) CAPINT 0.080 0.070 -0.104 -0.041 -0.024 0.038 -0.158 0.097 0.169 -0.128 1.000 (0.348) (0.410) (0.218) (0.627) (0.778) (0.658) (0.062)* (0.253) (0.045)** (0.130) OC 0.096 0.126 0.236 0.087 -0.015 -0.049 0.124 -0.286 0.141 0.043 -0.255 1.000 (0.258) (0.137) (0.005)*** (0.303) (0.857) (0.566) (0.144) (0.001)*** (0.096)* (0.613) (0.002)*** The figures in italics and parenthesis are the probabilities. *** Probabilities that are significant at 1% level ** Probabilities that are significant at 5% level * Probabilities that are significant at 10% level.
109
Table 5.15 Correlations- ABSDATCA and ABSDATA Sample
COE COEA ACI ACC SSVR CFVR ABSD ATCA
ABSD ATA
LG MV
LBT MV BETA
CAP INT OC
COE 1.000 COEA 0.714 1.000 (0.000)*** ACI 0.095 0.045 1.000 (0.246) (0.581) ACC 0.116 0.067 0.048 1.000 (0.156) (0.414) (0.561) SSVR -0.091 -0.197 -0.023 0.136 1.000 (0.264) (0.015)** (0.779) (0.095)* CFVR 0.082 0.026 -0.132 -0.094 -0.205 1.000 (0.314) (0.752) (0.106) (0.253) (0.012)** ABSDATCA 0.029 0.185 0.178 0.001 -0.127 -0.119 1.000 (0.720) (0.023)** (0.028)** (0.988) (0.119) (0.145) ABSDATA 0.011 0.181 0.229 -0.065 -0.094 -0.032 0.732 1.000 (0.898) (0.026)** (0.005)*** (0.427) (0.249) (0.695) (0.000)*** LGMV -0.364 -0.336 -0.094 -0.133 -0.083 0.197 -0.284 -0.287 1.000 (0.000) (0.000) (0.249) (0.105) (0.312) (0.015)** (0.000)*** (0.000)*** LBTMV 0.598 0.318 0.007 0.179 0.129 0.083 -0.150 -0.129 -0.282 1.000 (0.000)*** (0.000)*** (0.931) (0.028)** (0.115) (0.311) (0.066)* (0.115) (0.000)*** BETA 0.075 0.026 0.015 0.077 -0.060 0.007 -0.007 0.089 -0.034 0.136 1.000 (0.359) (0.751) (0.852) (0.349) (0.466) (0.933) (0.930) (0.278) (0.674) (0.096)* CAPINT 0.062 0.048 -0.096 -0.071 -0.072 0.080 -0.195 -0.216 0.117 0.166 -0.074 1.000 (0.449) (0.554) (0.240) (0.383) (0.378) (0.329) (0.017)** (0.008)*** (0.152) (0.041)** (0.365) OC 0.096 0.156 0.240 0.111 0.044 -0.104 0.367 0.293 -0.332 0.190 0.015 -0.248 1.000 (0.239) (0.055)* (0.003)*** (0.173) (0.590) (0.205) (0.000)*** (0.000)*** (0.000)*** (0.019)** (0.858) (0.002)*** The figures in italics and parenthesis are the probabilities. *** Probabilities that are significant at 1% level ** Probabilities that are significant at 5% level * Probabilities that are significant at 10% level.
110
111
Table 5.16 Correlations- PERS PRED Sample
COE COEA ACI ACC SSVR CFVR NEGPERS PRED LGMV LBTMV BETA STDREV
COE 1.000 COEA 0.690 1.000 (0.000)*** ACI 0.093 0.108 1.000 (0.314) (0.244) ACC 0.028 0.072 0.128 1.000 (0.764) (0.439) (0.168) SSVR -0.050 -0.151 0.081 0.027 1.000 (0.594) (0.100)* (0.381) (0.771) CFVR 0.114 0.030 -0.242 -0.133 -0.204 1.000 (0.218) (0.748) (0.008)*** (0.152) (0.027)** NEGPERS -0.035 -0.059 -0.008 0.037 -0.127 0.079 1.000 (0.709) (0.528) (0.934) (0.693) (0.172) (0.394) PRED 0.236 0.274 0.009 0.039 -0.012 0.058 0.059 1.000 (0.010)*** (0.003)*** (0.927) (0.678) (0.897) (0.533) (0.525) LGMV -0.392 -0.355 -0.119 -0.134 -0.102 0.067 -0.092 -0.033 1.000 (0.000)*** (0.000)*** (0.198) (0.149) (0.271) (0.471) (0.320) (0.722) LBTMV 0.602 0.348 0.043 0.150 0.113 0.105 0.011 0.022 -0.383 1.000 (0.000)*** (0.000)*** (0.640) (0.105) (0.222) (0.260) (0.909) (0.813) (0.000)*** BETA 0.062 0.049 0.072 0.218 -0.111 -0.044 0.161 0.235 -0.112 0.172 1.000 (0.505) (0.598) (0.439) (0.018)** (0.232) (0.638) (0.082)** (0.010)*** (0.226) (0.063)* STDREV 0.148 -0.037 -0.113 -0.123 0.072 -0.049 0.038 0.165 -0.028 -0.039 -0.080 1.000 (0.146) (0.714) (0.268) (0.227) (0.484) (0.632) (0.710) (0.100)* (0.783) (0.700) (0.435)
The figures in italics and parenthesis are the probabilities. *** Probabilities that are significant at 1% level ** Probabilities that are significant at 5% level
* Probabilities that are significant at 10% level.
5.8 Construct validity of variables - COE/COEA, EQ and CFVR
5.8.1 Cost of Equity - COE/COEA
Cost of equity reflects the pricing of risks. Therefore it should be related to the usual risks
factors other than those information related risks that are directly being tested in this study.
Those risks factors are market value, book to market and beta. Both the cost of equity
measures as discussed above are correlated with market value and book to market.
However they are not related to beta. As discussed in Chen et al (2003) there have been
studies conducted in emerging markets that found stock returns not to be related with beta.
Therefore both the cost of equity measures are considered to be reasonably valid.
To recapitulate COE is estimated as in Botosan (1997) and Chen et al (2003), based on a
residual income model (Gebhardt et al 2002) using earnings forecasts by professional
analyst (IBES). COE is the internal rate of return that equates the current value of a
company and the intrinsic value of the company. The current value is the current share
price. COEA is estimated based on Ohlson and Juettner-Nauroth (2000) as operationalized
in Chen et al (2003).
5.8.2 Earnings Quality
For the earnings quality variables to be valid they must be related to some of their
economic determinants as demonstrated in Francis et al (2008b). As described in section
4.2.1.3, the cross sectional measures of earnings quality (ABRES, ABSDATCA and
ABSDATA) are tested with economic determinants of earning quality namely size, capital
112
intensity (CAPINT) and operating cycle (OC). While persistence, PERS and predictability,
PRED being time series attributes are tested with standard deviation of revenue (STDREV).
As shown in the related tables, ABRES is significantly related to size and CAPINT, but not
with OC. ABSDATCA and ABSDATA are significantly related to all its economic
determinants. PERS are not related to size or STDREV while PRED is significantly related
STDREV. Therefore except for PERS, the earnings quality variables are valid
representation of earnings quality.
5.8.3 Cash Flow/Voting Rights
The cash flow and voting/controlling rights and subsequent ratio for the pyramidal structure
companies are calculated as in Claessens et al (2000). The higher the ratio, the closer the
value of cash flow to controlling rights, the lesser the tendency to expropriate. For the
companies without pyramid ownership structure, the cash flow rights is the proxy for
expropriating behavior. The higher the cash flow rights the lesser the tendency to
expropriate. The justification for the use of cash flow for non-pyramidal companies is that
there is evidence in Malaysia that the interest of the controlling party is aligned with other
non-controlling shareholders with increase in cash flow rights. In a study by Mohd Saleh et
al (2004b) on Malaysian listed companies found negative relationship between managerial
ownership and discretionary accruals as proxy to earnings management. A positive
relationship between managerial ownership and discretionary accruals would prove the
competing theory of managerial entrenchment and invalidate the use of cash flow rights in
the same manner as ratio of cash flow to voting rights.
113
Claessens (1998b) found significant positive effect of cash flow rights to market valuation.
The study measures market valuation as the ratio of actual market capitalization to an
imputed market value. Drawing parallel to Claessens (1998b) findings, a positive
relationship between cash flow rights and market value suggests positive relationship
between cash flow rights and interest alignment. The following tables show significant
positive relationship between market value and cash flow rights in full samples ABRES and
ABSDATA.
Table 5.17 Correlations between CF, VR and LGMV in full sample .
ABRES ABSDATA PERS LGMV LGMV LGMV
CF 0.16 0.16 0.05 (0.06)* (0.05)** 0.63
VR 0.07 -0.11 -0.12 0.40 0.18 0.19
The figures in italics and parenthesis are the probabilities. *** Probabilities that are significant at 1% level ** Probabilities that are significant at 5% level * Probabilities that are significant at 10% level.
Table 5.18 Correlations between CF, VR and LGMV among PYS companies
ABRES ABSDATA PERS LGMV LGMV LGMV
CF -0.07 -0.12 -0.08 0.66 0.47 0.58
VR 0.21 -0.02 0.05 0.19 0.91 0.72
The figures in italics and parenthesis are the probabilities. *** Probabilities that are significant at 1% level ** Probabilities that are significant at 5% level * Probabilities that are significant at 10% level.
114
Table 5.19 Correlations between CF, VR and LGMV among NON-PYS companies
ABRES ABSDATA PERS LGMV LGMV LGMV
CF 0.35 0.36 0.27 0.00 0.00 0.02
The figures in italics and parenthesis are the probabilities. *** Probabilities that are significant at 1% level ** Probabilities that are significant at 5% level * Probabilities that are significant at 10% level.
5.9 Multivariate analysis
5.9.1 Ownership structure and earnings quality
Equation 1:
Earnings quality 1,2,3,4,5 = α0 + αi CFVR + α2,3,4,5 UCP + α6 SSVR +
α 7 ACI + α8 ACC + α9SIZE +
α10 Controls for EQ + δ1,2,3,4,5
The results are reported in summary in table 5.20 and in detail in tables 5.20 (a), (b), (c) (d)
and (e).
This equation is estimated for all companies in each sample. This is so as there is a similar
expectation that as the cash flow/voting rights ratio for the pyramidal companies and cash
flow (or voting rights) for the non-pyramidal companies increase, there is a lesser
expectation of expropriating behavior (refer to the discussion in 5.8.3).
115
The R squared for the three regressions where each of ABRES, ABSDATCA and
ABSDATA is in turn the dependent variable are significant. The variance inflation index
are all below 10 (refer to Appendix 4 for SPSS output ) which shows there is no problem of
multicollinearity. Thus the significant correlations between regressors as reported earlier
(section 5.7) could be regarded as trivial. The problem of heteroskedasticity is dealt with by
using White’s adjusted standard error where necessary.
Hypothesis 1 : There is a negative relationship between ownership structure and earnings
quality
The main measure of ownership structure is CFVR. The type of ultimate controlling party
is added to test if the type of controlling party exacerbates and therefore explains earnings
quality. Since it is a dummy for each category of ultimate controlling party the predicted
sign is positive. As shown in Table 5.20 below the coefficients of both variables are not
significant. Therefore hypothesis 1 is not supported.
Hypothesis 2 : There is a relationship between substantial shareholders voting rights
(SSVR) and earnings quality.
The coefficients for SSVR in the cases where ABRES, ABSDATCA and PERS are the
dependent variables for earnings quality are significant and negative. Whilst the
coefficients for the other earnings quality dependent variables are not significant. It should
be noted however that the sign is consistently negative. Therefore there is some evidence to
support hypothesis 2.
116
Hypothesis 3 : There is a relationship between audit committee characteristics and earnings
quality .
Coefficients of ACI and AC are all not significant. There is no support for the hypothesis
in all cases.
117
Table 5.20 Coefficients of equation 1 regression
Earnings quality 1,2,3,4,5 = α0 + αi CFVR + α2,3,4,5 UCP + α6 SSVR + α 7 ACI + α8 ACC +
α9SIZE + α10 Controls for EQ + δ1,2,3,4,5
Predicted sign ABRES ABSDATA ABSDATCA PERS PRED
(Constant)
UCPMn +ve -0.01 0.28 -0.36 -0.18 -0.04
UCPInst +ve 0.01 -0.81 -0.20 -0.23 -0.09
UCPGov +ve -0.01 -0.42 -1.18 -0.05 0.03
UCPFam +ve 0.00 -0.13 -0.59 -0.06 0.03
SSVR ? -
0.09*** -2.01 -3.41*** -0.77** -0.28
ACI -ve 0.01 1.11* 0.43 0.06 0.04
ACC -ve -0.02 -0.81 -0.15 0.11 0.02
CFVR -ve -0.01 0.52 -0.21 -0.00 0.08
LGMV -ve -0.004* -0.24** -0.14* -0.05 0.00
CAPINT -ve -0.03** -1.07** -0.64
OC +ve 0.87* 1.47***
STDREV +ve 0.17 0.29*
R2 0.12 0.21 0.26 0.06 0.08 F 1.82 3.40 4.37 0.58 0.76
Sig 0.06* 0.00*** 0.00*** 0.82 0.67
N 141 141 151 151 118
*** Coefficients that are significant at 1% level ** Coefficients that are significant at 5% level * Coefficients that are significant at 10% level.
118
Table 5.20 (a) Results of equation 1 regression for earnings quality ABRES
ABRES = α0 + αi CFVR + α2,3,4,5 UCP + α6 SSVR +α 7 ACI + α8 ACC + α9SIZE + α10
Controls for EQ + δ
Dependent Variable: ABRES Included observations: 141 after adjustments White Heteroskedasticity-Consistent Standard Errors & Covariance
Variable Coefficient Std. Error t-Statistic Prob.
C 0.092640 0.024308 3.811065 0.0002 UCPMN -0.005935 0.013195 -0.449809 0.6536
UCPINST 0.010566 0.036525 0.289275 0.7728 UCPGOV -0.011214 0.009439 -1.188127 0.2369 UCPFAM 0.002469 0.008806 0.280388 0.7796
SSVR -0.091919 0.036736 -2.502114 0.0136 ACI 0.012500 0.014962 0.835433 0.4050 ACC -0.020365 0.021794 -0.934421 0.3518
CFVR -0.014970 0.015212 -0.984110 0.3269 LGMV -0.004254 0.002306 -1.844967 0.0673
CAPINT -0.025294 0.012671 -1.996184 0.0480
R-squared 0.122298 F-statistic 1.811405 Adjusted R-squared 0.054782 Prob(F-statistic) 0.064580 S.E. of regression 0.037535
119
Table 5.20 (b) Results of equation 1 regression for earnings quality ABSDATA
ABSDATA = α0 + αi CFVR + α2,3,4,5 UCP + α6 SSVR +α 7 ACI + α8 ACC + α9SIZE +
α10 Controls for EQ + δ
Dependent Variable: ABSDATA Included observations: 151 after adjustments White Heteroskedasticity-Consistent Standard Errors & Covariance
Variable Coefficient Std. Error t-Statistic Prob.
C -3.498777 1.591336 -2.198641 0.0296 UCPMN 0.286607 0.497123 0.576532 0.5652
UCPINST -0.809175 0.578915 -1.397745 0.1644 UCPGOV -0.417655 0.583196 -0.716149 0.4751 UCPFAM -0.128597 0.475471 -0.270462 0.7872
ACC -0.812367 0.744025 -1.091855 0.2768 ACI 1.108497 0.632959 1.751294 0.0821
CFVR 0.517643 0.578760 0.894400 0.3727 SSVR -2.014903 1.339638 -1.504065 0.1348
LGMV -0.240589 0.116881 -2.058414 0.0414 CAPINT -1.069421 0.513841 -2.081229 0.0392
OC 0.875636 0.478115 1.831433 0.0692
R-squared 0.212088 F-statistic 3.401431 Adjusted R-squared 0.149736 Prob(F-statistic) 0.000338
120
Table 5.20 (c) Results of equation 1 regression for earnings quality ABSDATCA
ABSDATCA = α0 + αi CFVR + α2,3,4,5 UCP + α6 SSVR +α 7 ACI + α8 ACC + α9SIZE +
α10 Controls for EQ + δ
Dependent Variable: ABSDATCA Included observations: 151 after adjustments White Heteroskedasticity-Consistent Standard Errors & Covariance
Variable Coefficient Std. Error t-Statistic Prob.
C -4.525943 1.449884 -3.121589 0.0022 UCPMN -0.363396 0.378125 -0.961048 0.3382
UCPINST -0.203948 0.411217 -0.495961 0.6207 UCPGOV -1.184706 0.447421 -2.647857 0.0090 UCPFAM -0.591863 0.277602 -2.132055 0.0348
ACC -0.148402 0.601677 -0.246647 0.8055 ACI 0.431023 0.414867 1.038942 0.3006
CFVR -0.210126 0.507543 -0.414005 0.6795 SSVR -3.406535 1.234491 -2.759465 0.0066 LGMV -0.141408 0.086504 -1.634699 0.1044
CAPINT -0.640600 0.493847 -1.297161 0.1967 OC 1.468395 0.450099 3.262385 0.0014
R-squared 0.257012 F-statistic 4.371135 Adjusted R-squared 0.198215 Prob(F-statistic) 0.000013
121
Table 5.20 (d) Results of equation 1 regression for earnings quality PERS
PERS= α0 + αi CFVR + α2,3,4,5 UCP + α6 SSVR +α 7 ACI + α8 ACC + α9SIZE + α10
Controls for EQ + δ
Dependent Variable: PERS Included observations: 98 after adjustments White Heteroskedasticity-Consistent Standard Errors & Covariance
Variable Coefficient Std. Error t-Statistic Prob.
C 0.169770 0.396223 0.428470 0.6694 UCPMN -0.178708 0.179605 -0.995007 0.3225
UCPINST -0.233684 0.155306 -1.504669 0.1360 UCPGOV -0.047423 0.160449 -0.295564 0.7683 UCPFAM -0.059351 0.142050 -0.417818 0.6771
SSVR -0.769400 0.394764 -1.949013 0.0545 ACI 0.065818 0.211390 0.311356 0.7563 ACC 0.114777 0.173784 0.660458 0.5107
CFVR -0.002842 0.198705 -0.014302 0.9886 LGMV -0.048040 0.034693 -1.384740 0.1697
STDREV 0.168939 0.293628 0.575351 0.5665
R-squared 0.062728 F-statistic 0.582255 Adjusted R-squared -0.045005 Prob(F-statistic) 0.824347
122
Table 5.20 (e) Results of equation 1 regression for earnings quality PRED
PRED = α0 + αi CFVR + α2,3,4,5 UCP + α6 SSVR +α 7 ACI + α8 ACC + α9SIZE + α10
Controls for EQ + δ
Dependent Variable: PRED Included observations: 98 after adjustments White Heteroskedasticity-Consistent Standard Errors & Covariance
Variable Coefficient Std. Error t-Statistic Prob.
C 0.014874 0.145001 0.102579 0.9185 UCPMN -0.042280 0.070063 -0.603450 0.5478
UCPINST -0.088591 0.064926 -1.364503 0.1759 UCPGOV 0.029800 0.064131 0.464675 0.6433 UCPFAM 0.028086 0.060076 0.467499 0.6413
SSVR -0.278021 0.252621 -1.100547 0.2741 ACI 0.043652 0.065219 0.669321 0.5051 ACC 0.016816 0.061846 0.271910 0.7863
CFVR 0.083134 0.104572 0.794995 0.4288 LGMV 0.004746 0.010494 0.452208 0.6522
STDREV 0.292701 0.157739 1.855606 0.0669
R-squared 0.080348 F-statistic 0.760102 Adjusted R-squared -0.025359 Prob(F-statistic) 0.666274
123
5.9.2 Reestimation of Equation 1- Ownership Structure and Earnings Quality,
separating the pyramidal and non-pyramidal companies (Tables 5.20(f)-(o))
As explained in Chapter 4, for PYS companies the lower the cash flow to voting rights ratio
(CFVR) and that is the higher the disparity between cash flow and voting rights, the higher
the expectation to expropriate as there is a greater incentive to do so. For the NPYS, the
ownership structure measure is simply the cash flow which equals the voting rights and this
measure is similarly read as CFVR and that is the higher the measure the less incentive to
expropriate as the interest between controlling party and the company become aligned. That
is the lower the cash flow rights the higher the expectation to expropriate and the lower the
earnings quality. This treatment is justified in section 5.8.3. as that expectation is generally
reasonable, even though Mock et al.(1988) reported an entrenchment effect where the
controlling party may become entrenched at high level of control.
Given that there is a possibility of such entrenchment effect taking place for NPYS,
Equation 1 is run separating PYS and NPYS companies for each sample. The results are
given in tables 5.20(f)-(o).
The regressions’ R squared for all the pyramidal samples are not significant except for the
regression run for earnings quality measure ABSDATA that is weakly significant at 10%
level. This is due to the small number of companies relative to the number of independent
variables.
124
Similar to the regression runs for the combined pyramidal and non-pyramidal sample, the R
squared for the regressions of PERS and PRED non-pyramidal samples are not significant.
R squared for the regression of non-pyramidal samples ABRES, ABSDATA and
ABSDATCA are significant. In terms of the individual coefficients, the results are similar
to those from the regressions of the combined samples.
The ownership variables CFVR (pyramidal) and CF (for non-pyramidal) are highly
insignificant except for the regression of PRED non-pyramidal sample. In the PRED non-
pyramidal sample the coefficient is positive, that is opposite to the one predicted. This
suggests that the ultimate holding party is entrenched as increasing control is associated
with poorer quality of earnings. It is also interesting to note that in this instance the
substantial shareholders’ voting rights is not significantly related to earnings quality.
Thus, except for the PRED non-pyramidal sample, similar conclusion can be drawn as that
for the regression of the combined samples, that hypothesis 1 is largely not supported.
Again similar to results for the combined samples, the coefficients of SSVR for ABRES
and ABSDATCA are negative and significant, although weakly for ABSDATCA. The
coefficients for the pyramidal ABRES, ABSDATA and ABSDATCA are negative. Thus
there is some support for hypothesis 2.
125
In the regressions of ABRES non-pyramidal sample and PRED pyramidal sample the
coefficients for audit committee competence are weakly supported.
Table 5.20 (f) Pyramidal companies in ABRES sample
Dependent Variable: ABRES Included observations: 42 after adjustments
Variable Coefficient Std. Error t-Statistic Prob.
C 0.079044 0.064963 1.216751 0.2329 UCPMN -0.022383 0.060500 -0.369973 0.7139
UCPINST 0.006926 0.049669 0.139447 0.8900 UCPGOV -0.002034 0.044137 -0.046088 0.9635 UCPFAM 0.011548 0.040638 0.284181 0.7782
SSVR -0.064072 0.077022 -0.831865 0.4118 ACI 0.013304 0.028337 0.469492 0.6420 ACC 0.051293 0.040165 1.277040 0.2111
CFVR -0.014715 0.025607 -0.574631 0.5697 LGMV -0.008127 0.004859 -1.672604 0.1045
CAPINT -0.012946 0.033361 -0.388065 0.7006
R-squared 0.189137 F-statistic 0.723088 Adjusted R-squared -0.072432 Prob(F-statistic) 0.696922
126
Table 5.20 (g) Non-pyramidal companies in ABRES sample Dependent Variable: ABRES Included observations: 102
Variable Coefficient Std. Error t-Statistic Prob.
C 0.086939 0.035751 2.431803 0.0170 UCPMN -0.001822 0.016780 -0.108561 0.9138
UCPGOV -0.018284 0.015139 -1.207809 0.2302 UCPFAM 0.004055 0.012398 0.327052 0.7444
SSVR -0.128991 0.055416 -2.327692 0.0221 ACI 0.006698 0.022100 0.303090 0.7625 ACC -0.043743 0.023624 -1.851645 0.0673 CF -0.011682 0.030630 -0.381403 0.7038
LGMV -0.001483 0.003423 -0.433368 0.6658 CAPINT -0.023827 0.017775 -1.340504 0.1834
R-squared 0.173867 F-statistic 2.151351 Adjusted R-squared 0.093049 Prob(F-statistic) 0.032669
Table 5.20 (h) Pyramidal companies in ABSDATA sample
Dependent Variable: ABSDATA Included observations: 40 after adjustments
Variable Coefficient Std. Error t-Statistic Prob.
C -4.107776 2.340185 -1.755321 0.0898 UCPMN 1.626625 2.307689 0.704872 0.4865
UCPINST 2.418191 1.874137 1.290295 0.2071 UCPGOV 2.738160 1.682786 1.627159 0.1145 UCPFAM 3.517478 1.531328 2.297011 0.0290
SSVR -2.005748 2.974324 -0.674354 0.5054 ACI 1.616366 1.116992 1.447071 0.1586 ACC -0.342052 1.522174 -0.224713 0.8238
CFVR -0.196153 0.943685 -0.207858 0.8368 LGMV -0.343239 0.183410 -1.871431 0.0714
CAPINT -2.100986 1.324875 -1.585799 0.1236
R-squared 0.414124 F-statistic 2.049854 Adjusted R-squared 0.212098 Prob(F-statistic) 0.064330
127
Table 5.20 (i) Non-pyramidal companies in ABSDATA sample
Dependent Variable:ABSDATA Included observations: 118 after adjustments
Variable Coefficient Std. Error t-Statistic Prob.
C -4.131862 1.606964 -2.571223 0.0115 UCPMN 0.457482 0.612346 0.747098 0.4566
UCPGOV -0.903557 0.511233 -1.767409 0.0800 UCPFAM -0.113961 0.445352 -0.255890 0.7985
SSVR -2.212612 1.664680 -1.329152 0.1866 ACI 0.111077 0.754123 0.147292 0.8832 ACC -0.946022 0.747393 -1.265763 0.2083 CF 1.173596 1.063377 1.103649 0.2722
LGMV -0.151666 0.115207 -1.316464 0.1908 CAPINT -0.824332 0.631639 -1.305067 0.1947
OC 1.103235 0.525159 2.100764 0.0380
R-squared 0.207683 F-statistic 2.804691 Adjusted R-squared 0.133634 Prob(F-statistic) 0.003999
Table 5.20 (j) Pyramidal companies in ABSDATCA sample
Dependent Variable: ABSDATCA Included observations: 40 after adjustments
Variable Coefficient Std. Error t-Statistic Prob.
C -1.751461 2.315959 -0.756257 0.4556 UCPMN -3.239049 2.283800 -1.418272 0.1668
UCPINST -0.483200 1.854736 -0.260522 0.7963 UCPGOV -2.040747 1.665366 -1.225404 0.2303 UCPFAM -0.560491 1.515476 -0.369845 0.7142
SSVR -3.774609 2.943533 -1.282339 0.2099 ACI 2.048086 1.105428 1.852753 0.0741 ACC 1.257858 1.506416 0.835001 0.4105
CFVR -0.879102 0.933916 -0.941307 0.3543 LGMV -0.208104 0.181511 -1.146508 0.2610
CAPINT -0.712102 1.311160 -0.543109 0.5912
R-squared 0.279964 F-statistic 1.127574 Adjusted R-squared 0.031675 Prob(F-statistic) 0.376272
128
Table 5.20 (k) Non-pyramidal companies in ABSDATCA sample
Dependent Variable: ABSDATCA Included observations: 118 after adjustments
Variable Coefficient Std. Error t-Statistic Prob.
C -3.349700 1.442890 -2.321522 0.0222 UCPMN 0.028147 0.549824 0.051193 0.9593
UCPGOV -1.011089 0.459035 -2.202641 0.0298 UCPFAM -0.534950 0.399881 -1.337772 0.1838
SSVR -2.541033 1.494713 -1.700015 0.0920 ACI -0.298327 0.677125 -0.440578 0.6604 ACC -0.538576 0.671083 -0.802547 0.4240 CF 0.940903 0.954804 0.985440 0.3266
LGMV -0.260736 0.103444 -2.520547 0.0132 CAPINT -1.288789 0.567148 -2.272405 0.0251
OC 1.379634 0.471539 2.925811 0.0042
R-squared 0.303481 F-statistic 4.662111 Adjusted R-squared 0.238386 Prob(F-statistic) 0.000016
Table 5.20 (l) Pyramidal companies in PERS sample
Dependent Variable: PERS Included observations: 43
Variable Coefficient Std. Error t-Statistic Prob.
C 0.974650 0.705859 1.380800 0.1769 UCPMN 0.066860 0.532511 0.125556 0.9009
UCPINST -0.245729 0.510575 -0.481279 0.6336 UCPGOV -0.403055 0.390832 -1.031274 0.3101 UCPFAM -0.354819 0.371806 -0.954312 0.3471
SSVR 0.676202 0.758610 0.891369 0.3794 ACI -0.197221 0.250528 -0.787221 0.4369 ACC -0.014887 0.303912 -0.048986 0.9612
CFVR 0.109101 0.239404 0.455720 0.6517 LGMV -0.065547 0.058129 -1.127627 0.2679
STDREV 0.053343 0.417408 0.127795 0.8991
R-squared 0.152454 F-statistic 0.575608 Adjusted R-squared -0.112404 Prob(F-statistic) 0.821341
129
Table 5.20 (m) Non-pyramidal companies in PERS sample
Dependent Variable: PERS Included observations: 64 after adjustments
Variable Coefficient Std. Error t-Statistic Prob.
C -0.622320 0.523918 -1.187820 0.2401 UCPMN 0.192981 0.283230 0.681360 0.4986
UCPGOV -0.040326 0.206996 -0.194814 0.8463 UCPFAM 0.100582 0.195749 0.513833 0.6095
SSVR 0.452263 0.705313 0.641223 0.5241 ACI 0.124942 0.291516 0.428593 0.6699 ACC -0.568724 0.342848 -1.658824 0.1029 CF 0.180638 0.466489 0.387229 0.7001
LGMV 0.116904 0.044158 2.647388 0.0106 STDREV 0.043469 0.599514 0.072507 0.9425
R-squared 0.174149 F-statistic 1.265233 Adjusted R-squared 0.036507 Prob(F-statistic) 0.277031
Table 5.20 (n) Pyramidal companies in PRED sample
Dependent Variable: PRED Included observations: 43
Variable Coefficient Std. Error t-Statistic Prob.
C 0.320356 0.285882 1.120590 0.2708 UCPMN -0.007202 0.215674 -0.033393 0.9736
UCPINST 0.108869 0.206789 0.526471 0.6022 UCPGOV 0.179332 0.158292 1.132920 0.2657 UCPFAM 0.153775 0.150586 1.021177 0.3148
SSVR 0.013404 0.307247 0.043627 0.9655 ACI -0.020527 0.101467 -0.202307 0.8410 ACC -0.216313 0.123088 -1.757377 0.0884 CFVR -0.078818 0.096962 -0.812879 0.4223 LGMV -0.025063 0.023543 -1.064554 0.2950
STDREV -0.000270 0.169056 -0.001596 0.9987
R-squared 0.141828 F-statistic 0.528858 Adjusted R-squared -0.126350 Prob(F-statistic) 0.856793
130
Table 5.20 (o) Non-pyramidal companies in PRED sample
Dependent Variable: PRED Included observations: 64 after adjustments
Variable Coefficient Std. Error t-Statistic Prob.
C -0.261046 0.219802 -1.187640 0.2402 UCPMN -0.007916 0.118825 -0.066617 0.9471
UCPGOV -0.008722 0.086843 -0.100430 0.9204 UCPFAM 0.035839 0.082124 0.436401 0.6643
SSVR -0.136762 0.295904 -0.462183 0.6458 ACI 0.085467 0.122302 0.698820 0.4877 ACC 0.192479 0.143837 1.338175 0.1864 CF 0.394237 0.195709 2.014399 0.0490
LGMV 0.002328 0.018526 0.125683 0.9004 STDREV 0.693636 0.251518 2.757799 0.0079
R-squared 0.217214 F-statistic 1.664934 Adjusted R-squared 0.086750 Prob(F-statistic) 0.120544
5.9.3 Earnings quality and cost of equity
Equation 2:
COE = φ0 + φi,2,3,4Earnings quality 1,2,3,4 + φ5 SIZE + φ6 β + φ7 BTMV + δ6
The results are reported in summary in tables 5.21 (I) and 5.21 (II) and in details in tables
5.21 (a) (i) &(ii), Table 5.21 (b)(i) & (ii), Tables 5.21 (c) (i) &(ii) Tables 5.21 (d) (i) &(ii)
and Tables 5.21 (e) (i) &(ii)).
The R squared statistics for all the regressions are significant. Again the problem with
heteroskedasticity are dealt with by using the White’s adjusted standard error where
necessary. The VIF are all below 10.
131
Hypothesis 4 : There is a negative relationship between earnings quality and cost of equity.
Since large earnings quality measures represent poor quality and vice versa the predicted
coefficient is positive. Tables 5.21(a)(i), 5.21(b)(i), 5.21(c)(i), 5.21(d)(i) and 5.21(e)(i)
shows results using COE and Tables 5.21(a)(ii), 5.21(b)(ii), 5.21(c)(ii),5.21(d)(ii) and
5.21(e)(ii) shows results using the alternative measure COEA.
In the regressions using COE, ABRES and PRED have positive significant coefficients.
ABSDATCA , ABSDATA and PERS do not explain variability in COE. In the regression
using COEA, coefficients for all EQ variables except PERS are positive and significant.
Thus hypothesis 4 is largely supported.
132
Table 5.21 (I)Results of equation 2 regression using COE estimate Predicted sign ABRES ABSDATA ABSDATCA PERS PRED
(Constant)
t 10.61 14.36 12.51 9.87 9.17 Sig. 0.000 0.00 0.00 0.00 0.00
EQ +ve 0.13** 0.00 0.00 0.00 0.04**
t 2.22 0.38 0.89 0.94 1.88 Sig. 0.03 0.38 0.35 0.06 0.70
LGMV -ve -0.004*** -0.01*** -0.004*** -0.01*** -0.00***
t -3.22 -3.58 -3.38 -2.97 -2.76 Sig. 0.00 0.00 0.00 0.00 0.01
BETA +ve 0.002 -0.00 -0.00 -0.00 -0.01
t 0.35 -0.12 -0.10 -0.58 -1.33 Sig. 0.73 0.90 0.92 0.57 0.18
LBTMV +ve 0.03*** 0.03*** 0.03*** 0.03*** 0.03***
t 6.39 6.63 6.98 6.79 7.19 Sig. 0.000 0.00 0.00 0.00 0.00
R2 0.39 0.40 0.40 0.40 0.45 F 21.99 24.34 24.60 18.64 23.20
Sig 0.00 0.00 0.00 0.00 0.00
N 141 151 151 118 118
*** Coefficients that are significant at 1% level ** Coefficients that are significant at 5% level * Coefficients that are significant at 10% level.
133
Table 5.21 (II) Results of equation 2 regression using COEA estimate
Predicted sign ABRES ABSDATA ABSDATCA PERS PRED
(Constant)
t 6.95 9.77 9.98 7.20 -6.90 Sig. 0.00 0.00 0.00 0.00 0.00
EQ +ve 0.35** 0.01** 0.01*** -0.01 0.31**
t 2.10 2.09 2.53 -0.97 1.97 Sig. 0.04 0.04 0.01 0.33 0.05
LGMV -ve -0.01*** -0.01*** -0.01*** -0.01*** -0.07***
t -2.95 -2.89 -2.94 -3.17 -2.60 Sig. 0.00 0.00 0.00 0.00 0.01
BETA +ve -0.00 -0.00 -0.00 -0.00 -0.03
t -0.06 -0.44 -0.25 -0.11 -0.38 Sig. 0.96 0.66 0.80 0.91 0.71
LBTMV +ve 0.03*** 0.03*** 0.03*** 0.03*** 0.18***
t 3.30 3.69 3.78 3.41 2.88 Sig. 0.00 0.00 0.00 0.00 0.00
R2 0.19 0.19 0.19 0.19 0.20 F 7.96 8.52 8.67 6.46 7.14
Sig 0.00 0.00 0.00 0.00 0.00
N 141 151 151 118 118
*** Coefficients that are significant at 1% level ** Coefficients that are significant at 5% level * Coefficients that are significant at 10% level.
134
Table 5.21 (a) (i) Results of equation 2 regression - ABRES and COE estimate COE = φ0 + φ ABRES + φ5 SIZE + φ6 β + φ7 BTMV + δ6
Dependent Variable: COE Included observations: 141 after adjustments White Heteroskedasticity-Consistent Standard Errors & Covariance
Variable Coefficient Std. Error t-Statistic Prob.
C 0.133755 0.012606 10.61061 0.0000 ABRES 0.129050 0.058032 2.223794 0.0278 LGMV -0.004673 0.001453 -3.216660 0.0016 BETA 0.002087 0.005964 0.349988 0.7269
LBTMV 0.028085 0.004398 6.386051 0.0000
R-squared 0.392715 F-statistic 21.98693 Adjusted R-squared 0.374854 Prob(F-statistic) 0.000000
Table 5.21 (a) (ii) Results of equation 2 regression - ABRES and COEA estimate COEa = φ0 + φ ABRES + φ5 SIZE + φ6 β + φ7 BTMV + δ6
Dependent Variable: COEA Included observations: 141 after adjustments White Heteroskedasticity-Consistent Standard Errors & Covariance
Variable Coefficient Std. Error t-Statistic Prob.
C 0.212813 0.030630 6.947958 0.0000 ABRES 0.351431 0.167227 2.101520 0.0374 LGMV -0.010327 0.003501 -2.949929 0.0037 BETA -0.000742 0.013408 -0.055376 0.9559
LBTMV 0.027584 0.008361 3.299339 0.0012
R-squared 0.189766 F-statistic 7.963163 Adjusted R-squared 0.165935 Prob(F-statistic) 0.000008
135
Table 5.21 (b) (i) Results of equation 2 regression- ABSDATA and COE estimate COE = φ0 + φ ABSDATA + φ5 SIZE + φ6 β + φ7 BTMV + δ6
Dependent Variable: COE Method: Least Squares Included observations: 151 after adjustments White Heteroskedasticity-Consistent Standard Errors & Covariance
Variable Coefficient Std. Error t-Statistic Prob.
C 0.145026 0.010097 14.36344 0.0000 ABSDATA 0.000490 0.001290 0.379850 0.7046
LGMV -0.005024 0.001403 -3.580372 0.0005 BETA -0.000542 0.004353 -0.124409 0.9012
LBTMV 0.027900 0.004210 6.626383 0.0000
R-squared 0.400131 F-statistic 24.34664 Adjusted R-squared 0.383696 Prob(F-statistic) 0.000000
Table 5.21 (b) (ii) Results of equation 2 regression- ABSDATA and COEA estimate COEA = φ0 + φ ABSDATA + φ5 SIZE + φ6 β + φ7 BTMV + δ6
Dependent Variable: COEA Included observations: 151 after adjustments White Heteroskedasticity-Consistent Standard Errors & Covariance
Variable Coefficient Std. Error t-Statistic Prob.
C 0.254799 0.026082 9.769311 0.0000 ABSDATA 0.006929 0.003318 2.088405 0.0385
LGMV -0.010297 0.003566 -2.887237 0.0045 BETA -0.004700 0.010744 -0.437454 0.6624
LBTMV 0.028993 0.007851 3.693010 0.0003
R-squared 0.189323 F-statistic 8.524073 Adjusted R-squared 0.167112 Prob(F-statistic) 0.000003
136
Table 5.21 (c) (i) Results of equation 2 regression - ABSDATCA and COE estimate COE = φ0 + φ ABSDATCA + φ5 SIZE + φ6 β + φ7 BTMV + δ6
Dependent Variable: COE Included observations: 151 after adjustments White Heteroskedasticity-Consistent Standard Errors & Covariance
Variable Coefficient Std. Error t-Statistic Prob.
C 0.146053 0.010066 14.50995 0.0000 ABSDATCA 0.001391 0.001569 0.886659 0.3767
LGMV -0.004714 0.001393 -3.383814 0.0009 BETA -0.000429 0.004307 -0.099676 0.9207
LBTMV 0.028368 0.004061 6.984845 0.0000
R-squared 0.402602 F-statistic 24.59826 Adjusted R-squared 0.386235 Prob(F-statistic) 0.000000
Table 5.21 (c) (ii) Results of equation 2 regression - ABSDATCA and COEA estimate COEA = φ0 + φ ABSDATCA + φ5 SIZE + φ6 β + φ7 BTMV + δ6
Dependent Variable: COEA Included observations: 151 after adjustments White Heteroskedasticity-Consistent Standard Errors & Covariance
Variable Coefficient Std. Error t-Statistic Prob.
C 0.255970 0.025642 9.982527 0.0000 ABSDATCA 0.008082 0.003193 2.531434 0.0124
LGMV -0.010123 0.003437 -2.944965 0.0038 BETA -0.002620 0.010544 -0.248500 0.8041
LBTMV 0.029373 0.007775 3.777826 0.0002
R-squared 0.191941 F-statistic 8.669945 Adjusted R-squared 0.169802 Prob(F-statistic) 0.000003
137
Table 5.21 (d) (i) Results of equation 2 regression - PERS and COE estimate COE = φ0 + φ PERS + φ5 SIZE + φ6 β + φ7 BTMV + δ6
Dependent Variable: COE Included observations: 118 after adjustments White Heteroskedasticity-Consistent Standard Errors & Covariance
Variable Coefficient Std. Error t-Statistic Prob.
C 0.141047 0.014283 9.874852 0.0000 PERS 0.004583 0.004872 0.940544 0.3489
LGMV -0.005154 0.001737 -2.967324 0.0037 BETA -0.003302 0.005732 -0.576000 0.5658
LBTMV 0.030963 0.004563 6.785149 0.0000
R-squared 0.397487 F-statistic 18.63692 Adjusted R-squared 0.376159 Prob(F-statistic) 0.000000
Table 5.21 (d) (ii) Results of equation 2 regression - PERS and COEA estimate COEA = φ0 + φ PERS + φ5 SIZE + φ6 β + φ7 BTMV + δ6
Dependent Variable: COEA Included observations: 118 after adjustments White Heteroskedasticity-Consistent Standard Errors & Covariance
Variable Coefficient Std. Error t-Statistic Prob.
C 0.244983 0.034038 7.197300 0.0000 PERS 0.014323 0.014756 0.970664 0.3338
LGMV -0.013442 0.004247 -3.165286 0.0020 BETA -0.001469 0.013414 -0.109486 0.9130
LBTMV 0.027470 0.008051 3.412025 0.0009
R-squared 0.186177 F-statistic 6.462690 Adjusted R-squared 0.157369 Prob(F-statistic) 0.000101
138
Table 5.21 (e) (i) Results of equation 2 regression - PRED and COE estimate COE = φ0 + φ PRED + φ5 SIZE + φ6 β + φ7 BTMV + δ6
Dependent Variable: COE Included observations: 118 after adjustments White Heteroskedasticity-Consistent Standard Errors & Covariance
Variable Coefficient Std. Error t-Statistic Prob.
C 0.139314 0.015192 9.169980 0.0000 PRED 0.037015 0.019686 1.880247 0.0626 LGMV -0.004924 0.001783 -2.762287 0.0067 BETA -0.008333 0.006250 -1.333316 0.1851
LBTMV 0.031473 0.004375 7.193619 0.0000
R-squared 0.450902 F-statistic 23.19800 Adjusted R-squared 0.431465 Prob(F-statistic) 0.000000
Table 5.21 (e) (ii) Results of equation 2 regression - PRED and COEA estimate COEA = φ0 + φ PRED + φ5 SIZE + φ6 β + φ7 BTMV + δ6
Dependent Variable: LCOEA Included observations: 118 after adjustments
Variable Coefficient Std. Error t-Statistic Prob.
C -1.481669 0.214931 -6.893707 0.0000 PRED 0.308715 0.156397 1.973924 0.0508 LGMV -0.073811 0.028409 -2.598152 0.0106
LBTMV 0.182519 0.063394 2.879116 0.0048 BETA -0.029862 0.079500 -0.375621 0.7079
R-squared 0.201711 F-statistic 7.138209 Adjusted R-squared 0.173453 Prob(F-statistic) 0.000037
139
Hypothesis 5 : There is a negative relationship between earnings quality and market return.
When COE is substituted with average monthly return (and year end return) none of the EQ
variables has significant coefficient. In contrast Francis et al (2008b) earnings quality is
significantly related to all measures of capital; cost of equity, average daily return,
annualized return, CAPM excess return and Fama and French (1993) 3 factor excess return.
In addition, the following regression using excess return based on Fama and French (1993)
three factor model as used in Francis et al (2004) :
R i,t – R f,t = α +β (R m,t – R f,t) + γ Size f,t + ∂ BTMV f,t + θEQ f,t + ε f,t
Results are given in tables 5.22(a)-(e). In all the regressions only the coefficient excess
market return is significant. Francis et al (2004) find that the various earnings attributes
examined to be significantly and negatively related to not only the cost of equity measures
but also the realized returns measure.Thus there is no support for hypothesis 5.
140
Table 5.22(a) Results of testing earnings quality (ABRES) and excess return
R i,t – R f,t = α +β (R m,t – R f,t) + γ Size + ∂ BTMV + θABRES + ε
Dependent Variable: R i,t – R f,t White Heteroskedasticity-Consistent Standard Errors & Covariance
Variable Coefficient Std. Error t-Statistic Prob.
C -0.108864 0.139256 -0.781757 0.4356 (R m,t – R f,t) 3.511877 1.327909 2.644667 0.0091
LGMV -0.009587 0.020209 -0.474367 0.6360 LBTMV -0.073929 0.049777 -1.485191 0.1397 ABRES 0.357854 0.670017 0.534096 0.5941
R-squared 0.207669 F-statistic 9.370014 Adjusted R-squared 0.185506 Prob(F-statistic) 0.000001
Table 5.22(b) Results of testing earnings quality (ABSDATCA) and excess return
R i,t – R f,t = α +β (R m,t – R f,t) + γ Size + ∂ BTMV + θ ABSDATCA + ε
Dependent Variable: R i,t – R f,t White Heteroskedasticity-Consistent Standard Errors & Covariance
Variable Coefficient Std. Error t-Statistic Prob.
C -0.056494 0.121957 -0.463227 0.6439 (R m,t – R f,t) 3.453557 1.309422 2.637468 0.0092
LGMV -0.004491 0.022839 -0.196653 0.8444 LBTMV -0.062109 0.049976 -1.242779 0.2159
ABSDATCA 0.022153 0.030079 0.736504 0.4626
R-squared 0.199576 F-statistic 9.474829 Adjusted R-squared 0.178512 Prob(F-statistic) 0.000001
141
Table 5.22(c) Results of testing earnings quality (ABSDATA) and excess return
R i,t – R f,t = α +β (R m,t – R f,t) + γ Size + ∂ BTMV + θ ABSDATA + ε
Dependent Variable: R i,t – R f,t White Heteroskedasticity-Consistent Standard Errors & Covariance
Variable Coefficient Std. Error t-Statistic Prob.
C -0.062973 0.119938 -0.525049 0.6003 (R m,t – R f,t) 3.535445 1.362388 2.595035 0.0104
LGMV -0.004271 0.023913 -0.178599 0.8585 LBTMV -0.062712 0.050863 -1.232965 0.2195
ABSDATA 0.021989 0.028202 0.779689 0.4368
R-squared 0.200271 F-statistic 9.516088 Adjusted R-squared 0.179225 Prob(F-statistic) 0.000001
Table 5.22(d) Results of testing earnings quality (PERS) and excess return
R i,t – R f,t = α +β (R m,t – R f,t) + γ Size + ∂ BTMV + θPERS + ε
Dependent Variable: R i,t – R f,t White Heteroskedasticity-Consistent Standard Errors & Covariance
Variable Coefficient Std. Error t-Statistic Prob.
C -0.060824 0.130289 -0.466840 0.6415 (R m,t – R f,t) 3.878367 1.530388 2.534237 0.0126
LGMV -0.011102 0.018521 -0.599464 0.5500 LBTMV -0.027595 0.044521 -0.619831 0.5366
PERS -0.006507 0.091027 -0.071489 0.9431
R-squared 0.238020 F-statistic 9.136813 Adjusted R-squared 0.211969 Prob(F-statistic) 0.000002
142
Table 5.22(e) Results of testing earnings quality (PRED) and excess return
R i,t – R f,t = α +β (R m,t – R f,t) + γ Size + ∂ BTMV + θPRED + ε
Dependent Variable: R i,t – R f,t White Heteroskedasticity-Consistent Standard Errors & Covariance
Variable Coefficient Std. Error t-Statistic Prob.
C -0.063713 0.134166 -0.474887 0.6358 (R m,t – R f,t) 3.872478 1.480569 2.615534 0.0101
LGMV -0.011379 0.019804 -0.574596 0.5667 LBTMV -0.028069 0.043772 -0.641258 0.5226 PRED 0.020400 0.112599 0.181176 0.8565
R-squared 0.238095 F-statistic 9.140629 Adjusted R-squared 0.212047 Prob(F-statistic) 0.000002
H ypothesis 6 : There is a positive relationship between market return and COE/COEA .
For the purpose of testing this hypothesis a Pearson correlation test is conducted on both
COE measures and average monthly return of the companies for the ABRES sample. There
is significant correlation between COE and returns.
Table 5.23 Pearson Correlation of cost of equity and return COE COEA AVMR YER COE 1 .722 .015 -.008 .000 .864 .922 COEA .722 1 .014 -.012 .000 .866 .891 AVMR .015 .014 1 .975 .864 .866 .000 YER -.008 -.012 .975 1 .922 .891 .000
143
5.9.5 Simultaneity test for equations 3 and 4
As described in 4.3.2, a simultaneity test need to be conducted first because if simultaneity
exists ordinary least square estimates would not be consistent and efficient. The reduced
form equations to estimate ^CFVR and ^SSVR are as follows.
Reduced form equation 1 :
CFVR = a0 + a1,2,3,4UCP + a5 ACI + a6 ACC + a7SIZE + a8BTMV + δ1
Reduced form equation 2 :
SSVR = b0 + b1,2,3,4UCP + b5 ACI + b6 ACC + b7SIZE + b8BTMV + δ2
The predicted values ^CFVR, ^SSVR and ^COE are obtained. The following equation is
run to test for the exogeneity of CFVR and SSVR.
COE = μ0 + μ1 CFVR + μ 2,3,4,5 UCP + μ 6 SSVR + μ 7 ACI + μ 8 ACC +
μ 9SIZE + μ 10BTMV + c1 ^CFVR + c2 ^SSVR + δ7,8,9
The results are given in table 5.23(a) ABRES sample (with COE), 5.23(c) ABRES sample
(with COEA), 5.24(a) ABSDATA/ABSDATCA sample (with COE),5.24(c)
ABSDATA/ABSDATCA sample (with COEA), 5.25(a) PERS/PRED sample (COE) and
5.25(c) PERS/PRED sample (with COEA). .
144
If CFVR and SSVR are exogenous, c1 and c2 are zero. Therefore F test will test if c1 = c2
=0. The results are given in table 5.23(b),(d). 5.24(b),(d) and 5.25(b),(d). As can be seen
from the tables at 5% level of confidence, the F-statistic is not significant, which means that
c1 and c2 are not significantly different from zero and therefore the variables CFVR and
SSVR are exogenous. Therefore equations 3 and 4 are not simultaneous equations. The
analysis proceeds to estimate equations 3 and 4 using ordinary least square.
However throughout this chapter thus far, significance level of 10% has been treated as
weak level significance. Therefore the two stage least square estimation of equations 3 and
4 of ABRES and ABSDATA/ABSDATCA samples (with COE) are performed to compare
the results with the ordinary least square method.
145
Table 5.23 (a) ABRES sample- Results of estimating equation to test the exogenuity of CFVR and SSVR (with COE) Dependent Variable: COE Included observations: 141 after adjustments
Variable Coefficient Std. Error t-Statistic Prob.
C 0.117788 0.018123 6.499316 0.0000 LGMV -0.003618 0.001880 -1.924232 0.0565
LBTMV 0.032740 0.004563 7.174682 0.0000 BETA -0.001414 0.005520 -0.256070 0.7983 CFVR 0.014285 0.011858 1.204647 0.2305 SSVR -0.055998 0.030384 -1.843008 0.0676
^CFVR -0.006543 0.003307 -1.978745 0.0500 ^SSVR -0.006919 0.003488 -1.983528 0.0494
UCPINST 0.018665 0.021681 0.860898 0.3909 UCPFAM 0.014092 0.004992 2.822951 0.0055
ACI 0.003991 0.011890 0.335698 0.7376 ACC 0.016535 0.015098 1.095202 0.2755
R-squared 0.452174 F-statistic 9.679643 Adjusted R-squared 0.405460 Prob(F-statistic) 0.000000
Table 5.23 (b) Testing for coefficients of ^CFVR and ^SSVR
Wald Test:
Test Statistic Value df Probability
F-statistic 2.678529 (2, 129) 0.0725Chi-square 5.357059 2 0.0687
Null Hypothesis Summary:
Normalized Restriction (= 0) Value Std. Err.
C(7) -0.006543 0.003307C(8) -0.006919 0.003488
Restrictions are linear in coefficients.
146
Table 5.23 (c) ABRES sample- Results of estimating equation to test the exogenuity of CFVR and SSVR (with COEA)
Dependent Variable: COEA Included observations: 141 after adjustments
Variable Coefficient Std. Error t-Statistic Prob.
C 0.232324 0.041737 5.566332 0.0000 LGMV -0.010651 0.004331 -2.459531 0.0152
LBTMV 0.028066 0.010509 2.670591 0.0085 BETA -0.006169 0.012713 -0.485225 0.6283 CFVR 0.017223 0.027309 0.630671 0.5294 SSVR -0.235330 0.069974 -3.363123 0.0010
^CFVR -0.007192 0.007616 -0.944384 0.3467 ^SSVR -0.001030 0.008034 -0.128215 0.8982
UCPINST -0.004079 0.049930 -0.081692 0.9350 UCPFAM 0.017634 0.011496 1.533863 0.1275
ACI -0.004792 0.027382 -0.174990 0.8614 ACC 0.017359 0.034770 0.499258 0.6184
R-squared 0.246630 F-statistic 3.839137 Adjusted R-squared 0.182389 Prob(F-statistic) 0.000086
Table 5.23 (d) Testing for coefficients of ^CFVR and ^SSVR
Wald Test: Equation: EQEXOCOEA
Test Statistic Value df Probability
F-statistic 0.507749 (2, 129) 0.6030Chi-square 1.015497 2 0.6018
Null Hypothesis Summary:
Normalized Restriction (= 0) Value Std. Err.
C(7) -0.007192 0.007616C(8) -0.001030 0.008034
Restrictions are linear in coefficients.
147
Table 5.24 (a) ABSDATA/ABSDATCA sample- Results of estimating equation to test the exogenuity of CFVR and SSVR (with COE)
Dependent Variable: COE White Heteroskedasticity-Consistent Standard Errors & Covariance
Variable Coefficient Std. Error t-Statistic Prob.
C 0.132124 0.014244 9.275499 0.0000 UCPMN 0.016691 0.009771 1.708206 0.0898
UCPINST 0.014993 0.008976 1.670342 0.0971 UCPFAM 0.016901 0.004050 4.173364 0.0001
ACC 0.015666 0.011719 1.336793 0.1835 ACI 0.006167 0.010215 0.603702 0.5470
LGMV -0.003889 0.001473 -2.641088 0.0092 LBTMV 0.029885 0.003427 8.720063 0.0000 BETA -0.004622 0.004525 -1.021412 0.3088 CFVR 0.013343 0.013684 0.975131 0.3312 SSVR -0.067063 0.031652 -2.118766 0.0359
^SSVR -0.153670 0.090459 -1.698778 0.0916
R-squared 0.470226 F-statistic 11.21599 Adjusted R-squared 0.428301 Prob(F-statistic) 0.000000
Table 5.24 (b) Testing for coefficients of ^CFVR and ^SSVR Probability to test null hypothesis c2=0 is 0.0916. The ^CFVR has been automatically dropped.
148
Table 5.24 (c) ABSDATA/ABSDATCA sample- Results of estimating equation to test the exogenuity of CFVR and SSVR (with COEA)
Dependent Variable: COEA Included observations: 151 after adjustments
Variable Coefficient Std. Error t-Statistic Prob.
C 0.228413 0.048826 4.678079 0.0000 UCPMN 0.013170 0.022028 0.597872 0.5509
UCPINST -0.008619 0.049446 -0.174306 0.8619 UCPFAM 0.021214 0.014378 1.475507 0.1423
ACC 0.007020 0.034761 0.201944 0.8403 ACI -0.000385 0.025651 -0.014991 0.9881
LGMV -0.011243 0.004159 -2.702931 0.0077 LBTMV 0.026379 0.009113 2.894711 0.0044 BETA -0.004844 0.011357 -0.426577 0.6703 CFVR 0.007368 0.026056 0.282775 0.7778 SSVR -0.219707 0.063788 -3.444321 0.0008 ^SSVR 0.095329 0.315595 0.302061 0.7631
R-squared 0.252465 Mean dependent var 0.148808 Adjusted R-squared 0.193308 S.D. dependent var 0.065392 S.E. of regression 0.058732 Akaike info criterion -2.755513 Sum squared resid 0.479481 Schwarz criterion -2.515729 Log likelihood 220.0412 F-statistic 4.267684 Durbin-Watson stat 2.050598 Prob(F-statistic) 0.000018
Table 5.24 (d) Testing for coefficients of ^CFVR and ^SSVR Probability to test null hypothesis c2=0 is 0.76. The ^CFVR has been automatically dropped.
149
Table 5.25 (a) PERS/PRED sample- Results of estimating equation to test the exogenuity of CFVR and SSVR (with COE)
Dependent Variable: COE Included observations: 118 after adjustments
Variable Coefficient Std. Error t-Statistic Prob.
C 0.149486 0.042026 3.556943 0.0006 LGMV -0.005266 0.002192 -2.402582 0.0180
LBTMV 0.033224 0.006192 5.365775 0.0000 BETA -0.010618 0.008540 -1.243310 0.2165 CFVR 0.004950 0.013895 0.356221 0.7224 SSVR -0.060397 0.033229 -1.817602 0.0719
PREDCFVR 0.005964 0.035030 0.170265 0.8651 PREDSSVR -0.174906 0.228285 -0.766171 0.4453 UCPINST 0.010957 0.029123 0.376237 0.7075 UCPFAM 0.013907 0.006579 2.113900 0.0369
ACI 0.017496 0.014245 1.228184 0.2221 ACC -0.009263 0.014502 -0.638781 0.5243
R-squared 0.450696 F-statistic 7.906510 Adjusted R-squared 0.393693 Prob(F-statistic) 0.000000
Table 5.25 (b) Testing for coefficients of ^CFVR and ^SSVR Wald Test: Equation: Untitled
Test Statistic Value df Probability
F-statistic 0.420436 (2, 106) 0.6579Chi-square 0.840873 2 0.6568
Null Hypothesis Summary:
Normalized Restriction (= 0) Value Std. Err.
C(7) 0.005964 0.035030C(8) -0.174906 0.228285
Restrictions are linear in coefficients.
150
Table 5.25 (c) PERS/PRED sample- Results of estimating equation to test the exogenuity of CFVR and SSVR (with COEA) Dependent Variable: COEA Included observations: 118 after adjustments
Variable Coefficient Std. Error t-Statistic Prob.
C 0.235433 0.093475 2.518660 0.0133 LGMV -0.011567 0.004875 -2.372805 0.0195
LBTMV 0.029061 0.013772 2.110147 0.0372 BETA -0.008995 0.018995 -0.473541 0.6368 CFVR -0.000210 0.030906 -0.006789 0.9946 SSVR -0.200100 0.073907 -2.707447 0.0079
PREDCFVR -0.016334 0.077914 -0.209647 0.8343 PREDSSVR 0.107186 0.507753 0.211099 0.8332 UCPINST -0.007141 0.064776 -0.110239 0.9124 UCPFAM 0.020702 0.014632 1.414837 0.1600
ACI 0.018468 0.031684 0.582887 0.5612 ACC -0.000183 0.032255 -0.005689 0.9955
R-squared 0.255154 F-statistic 3.301019 Adjusted R-squared 0.177858 Prob(F-statistic) 0.000633
Table 5.25 (d) Testing for coefficients of ^CFVR and ^SSVR Wald Test: Equation: EQEXOCOEA
Test Statistic Value df Probability
F-statistic 0.071934 (2, 106) 0.9306Chi-square 0.143869 2 0.9306
Null Hypothesis Summary:
Normalized Restriction (= 0) Value Std. Err.
C(7) -0.016334 0.077914C(8) 0.107186 0.507753
151
5.9.6 Ownership structure, monitoring mechanisms and market assessment
Equation 3:
COE = μ0 + μ1 CFVR + μ 2,3,4,5 UCP + μ 6 SSVR + μ 7 ACI + μ 8 ACC + μ
9SIZE + μ 10BTMV + δ7,8,9
The estimation using ordinary least square method is carried out in the absence of
simultaneity. Summary results are given in table 5.26 (a) with COE estimate and table 5.26
(b) with COEA estimates. All the R squared are significant. Again VIF shows there is no
problem with multi-collinearity and where necessary heteroskedasticity is dealt with by
using White’s standard error.
Hypothesis 7 : There is a relationship between ownership structure and COE
Coeficients of UCPMn and UCPFAM are both positive and significant in the ABRES and
ABSDATCA/ ABSDATA samples. This means with reference to UCPCom, UCPMN and
UCPFAM explain variability of COE. However in the regression using COEA, none of the
coefficients of UCP’s is significant. Therefore there is some mixed support for hypothesis
7.
Hypothesis 8 : There is a relationship between substantial shareholder voting rights and
COE.
152
Substantial shareholders’ voting rights, SSVR consistently shows a negative and significant
relationship with cost of equity using both measures of cost of equity in all samples. Thus
there is a strong support for hypothesis 8.
Hypothesis 9 : There is a relationship between audit committee characteristics and COE.
The coefficients of ACI and ACC are not significant in regressions using both COE and
COEA in all samples examined. There is no support at all for hypothesis 9.
153
Table 5.26 (a) Results of equation 3 regression using COE estimate
Predicted sign ABRES ABSDATA/ABSDATCA PERS/PRED
(Constant)
UCPMn +ve 0.221** 0.180* 0.022
UCPInst +ve 0.046 0.032 0.045
UCPGov +ve 0.165* 0.111 0.110
UCPFam +ve 0.354*** 0.286*** 0.226
SSVR ? -0.128* -0.164*** -0.139*
ACI -ve 0.067 0.053 0.076
ACC -ve 0.051 0.031 -0.075
LGMV -ve -0.190*** -0.179** -0.208**
LBTMV +ve 0.469*** 0.519*** 0.506***
BETA +ve -0.017 -0.039 -0.081
CFVR -ve 0.093 0.088 0.030
R2 0.405 0.428 0.394 F 9.680 11.216 7.907
Sig 0.000 0.000 0.000 N 141 151 118
*** Coefficients that are significant at 1% level ** Coefficients that are significant at 5% level * Coefficients that are significant at 10% level.
154
Table 5.26(b) Results of equation 3 regression using COEA estimate
Predicted sign ABRES ABSDATA/ABSDATCA PERS/PRED
(Constant)
UCPMn +ve 0.099 0.055 0.009
UCPInst +ve 0.005 -0.009 -0.015
UCPGov +ve -0.013 -0.035 -0.041
UCPFam +ve 0.166 0.144 0.134
SSVR ? -0.274*** -0.270*** -0.241***
ACI -ve -0.002 -0.006 0.074
ACC -ve 0.069 0.034 0.014
LGMV -ve -0.228*** -0.223*** -0.227**
LBTMV +ve 0.233*** 0.278*** 0.280***
BETA +ve -0.039 -0.044 -0.082
CFVR -ve 0.057 0.024 -0.001
R2 0.182 0.193 0.178 F 3.839 4.268 3.301
Sig 0.000 0.000 0.001
N 141 151 128
*** Coefficients that are significant at 1% level ** Coefficients that are significant at 5% level * Coefficients that are significant at 10% level.
155
5.9.7 The relationship that examines whether market assessment and the monitoring
mechanisms could explain changes in ownership structure
Equation 4:
CFVR = η0 + η1COE + η2UCP + η3 SSVR + η4 ACI + η5 ACC + η6SIZE +
η7BTMV + δ10,11,12
Hypothesis 10 : There is a relationship between type of ownership and ownership structure.
The coefficients of UCPMn are significant in all regressions and in the negative direction.
Hypothesis 11 : There is relationship between market assessment and ownership structure.
The coefficients of cost of equity for both measures and in all samples are not significant.
Therefore hypothesis 11 is not supported.
Hypothesis 12 : There is a relationship between the monitoring mechanisms and ownership
structure.
There is a weak support at 10% for relationship between audit committee independence and
CFVR for sample PERS/PRED. There is a weak significant relationship between audit
committee competence in all regressions except in PERS/PRED sample. The direction in
all regression is negative.There is a significant relationship between SSVR and CFVR in a
negative direction in all regressions.
156
Table 5.27 (a) Results of equation 4 regression using COE estimates
Predicted
sign ABRES ABSDATA/ABSDATCA PERS/PRED (Constant)
UCPMn +ve -
0.472*** -0.449*** -0.413***
UCPInst +ve -0.064 -0.056 0.001
UCPGov +ve 0.063 0.053 0.092
UCPFam +ve -0.195 -0.170 -0.143
ACI -ve -0.064 -0.055 -0.144*
ACC -ve -0.133* -0.128* -0.134
LGMV +ve 0.045 0.109 -0.048
LBTMV -ve 0.127 0.113 0.125
COE -ve 0.120 0.116 0.038
SSVR ? -0.165** -0.157** -0.179**
R2 0.238 0.226 0.220 F 5.383 5.376 4.309
Sig 0.000 0.000 0.000
N 141 151 118 *** Coefficients that are significant at 1% level ** Coefficients that are significant at 5% level * Coefficients that are significant at 10% level.
157
Table 5.27 (b) Results of equation 4 regression using COEA estimates
Predicted
sign ABRES ABSDATA/ABSDATCA PERS/PRED (Constant)
UCPMn +ve -0.455*** -0.434*** -0.414***
UCPInst +ve -0.060 -0.052 0.003
UCPGov +ve 0.084 0.067 0.096
UCPFam +ve -0.163 -0.142 -0.135
ACI -ve -0.056 -0.049 -0.141*
ACC -ve -0.131* -0.126* -0.138
LGMV +ve 0.035 0.094 -0.056
LBTMV -ve 0.172** 0.168** 0.144
COEA -ve 0.055 0.021 -0.001
SSVR ? -0.167** -0.172** -0.185**
R2 0.232 0.218 0.220 F 5.234 5.192 4.292
Sig 0.000 0.000 0.000
N 141 151 118 *** Coefficients that are significant at 1% level ** Coefficients that are significant at 5% level * Coefficients that are significant at 10% level.
158
5.9.8 Relationship between substantial shareholders’ voting rights and elements of
ownership, monitoring mechanisms and cost of equity
In addition, to test whether SSVR could be explained by the ownership structure elements
and other monitoring mechanisms, the following is run.
SSVR = ω0 + ω 1COE + ω 2,3,4,5UCP + ω 6 CFVR + ω 7 ACI + ω 8 ACC + ω
9SIZE + ω 10BTMV + δ10,11,12
The results are in table 5.28 (a) (i) – 5.28 (c)(ii). The R squared for all samples except for
sample PERS/PRED (using COE), are significant although they are not high. The rather
low R squared is expected. This suggests that there are other determinants of the
shareholdings of these shareholders.
In all the regressions the coefficients of COE, CFVR and LBTMV are consistently
significant. The coefficient of COE is negative throughout the samples which suggests that
high substantial shareholders’ shareholdings is associated with low cost of equity.
The coefficient of CFVR is negative, which also suggests that high substantial
shareholders’ shareholdings is associated with low cash flow/voting rights. In other words
substantial shareholders’ voting rights are high in companies where the separation of
ownership and control related problems are expected to be high. On the other hand, high
159
substantial shareholders voting rights is associated with high book to market as indicated by
the positive coefficient.
160
Table 5.28 (a) (i) ABRES Sample (with COE) Dependent Variable: SSVR Included observations: 141 after adjustments
Variable Coefficient Std. Error t-Statistic Prob.
C 0.184785 0.058855 3.139658 0.0021 COE -0.456015 0.247953 -1.839121 0.0682
UCPMN -0.029491 0.030132 -0.978732 0.3295 UCPINST 0.037359 0.056814 0.657562 0.5120 UCPGOV -0.049904 0.026733 -1.866770 0.0642 UCPFAM -0.014510 0.022515 -0.644461 0.5204
CFVR -0.070698 0.033451 -2.113508 0.0365 LGMV 0.000808 0.005438 0.148566 0.8821
LBTMV 0.031144 0.012110 2.571699 0.0112 ACI -0.011434 0.032257 -0.354467 0.7236 ACC 0.044703 0.039598 1.128928 0.2610
R-squared 0.141326 F-statistic 2.139626 Adjusted R-squared 0.075274 Prob(F-statistic) 0.025648
Table 5.28 (a) (i) ABRES Sample (with COEA) Dependent Variable: SSVR Included observations: 141 after adjustments
Variable Coefficient Std. Error t-Statistic Prob.
C 0.202368 0.055273 3.661256 0.0004 COEA -0.340682 0.101607 -3.352922 0.0010
UCPMN -0.031433 0.028791 -1.091740 0.2770 UCPINST 0.030478 0.055148 0.552652 0.5815 UCPGOV -0.054951 0.025613 -2.145414 0.0338 UCPFAM -0.016861 0.021180 -0.796063 0.4274
CFVR -0.066991 0.032398 -2.067745 0.0406 LGMV -0.000943 0.005289 -0.178295 0.8588
LBTMV 0.026946 0.010533 2.558153 0.0117 ACI -0.015802 0.031209 -0.506336 0.6135 ACC 0.046723 0.038471 1.214488 0.2268
R-squared 0.189109 F-statistic 3.031747 Adjusted R-squared 0.126733 Prob(F-statistic) 0.001760
161
Table 5.28 (b) (i) ABSDATA Sample (with COE) Dependent Variable: SSVR Included observations: 151 after adjustments
Variable Coefficient Std. Error t-Statistic Prob.
C 0.204849 0.060073 3.409991 0.0008 COE -0.627132 0.255102 -2.458355 0.0152
UCPMN -0.030916 0.030951 -0.998884 0.3196 UCPINST 0.032253 0.058739 0.549089 0.5838 UCPGOV -0.055653 0.027814 -2.000938 0.0473 UCPFAM -0.010504 0.023041 -0.455883 0.6492
CFVR -0.068305 0.033511 -2.038258 0.0434 LGMV 0.001274 0.005627 0.226488 0.8212
LBTMV 0.039877 0.012619 3.160128 0.0019 ACI -0.014488 0.032860 -0.440889 0.6600 ACC 0.046809 0.038409 1.218699 0.2250
R-squared 0.159224 F-statistic 2.651287 Adjusted R-squared 0.099169 Prob(F-statistic) 0.005424
Table 5.28 (b) (ii) ABSDATA Sample (with COEA)
Dependent Variable: SSVR Included observations: 151 after adjustments White Heteroskedasticity-Consistent Standard Errors & Covariance
Variable Coefficient Std. Error t-Statistic Prob.
C 0.211024 0.055620 3.794064 0.0002 COEA -0.355192 0.091690 -3.873835 0.0002
UCPMN -0.037411 0.024377 -1.534703 0.1271 UCPINST 0.023679 0.050848 0.465683 0.6422 UCPGOV -0.062329 0.024463 -2.547841 0.0119 UCPFAM -0.015027 0.019289 -0.779048 0.4373
CFVR -0.071253 0.033912 -2.101100 0.0374 LGMV 4.54E-07 0.005668 8.02E-05 0.9999
LBTMV 0.032377 0.011867 2.728254 0.0072 ACI -0.020034 0.036056 -0.555624 0.5794 ACC 0.045942 0.034497 1.331769 0.1851
R-squared 0.190623 F-statistic 3.297256 Adjusted R-squared 0.132810 Prob(F-statistic) 0.000727
162
Table 5.28 (c) (i) PERS/PRED Sample (with COE) Dependent Variable: SSVR Included observations: 118 after adjustments
Variable Coefficient Std. Error t-Statistic Prob.
C 0.228923 0.072237 3.169061 0.0020 COE -0.467521 0.276300 -1.692076 0.0935
UCPMN -0.059670 0.038345 -1.556120 0.1226 UCPINST -0.034430 0.088128 -0.390681 0.6968 UCPGOV -0.051099 0.035770 -1.428537 0.1560 UCPFAM -0.025882 0.029970 -0.863584 0.3897
CFVR -0.083112 0.039472 -2.105581 0.0376 LGMV -0.002592 0.006749 -0.384033 0.7017
LBTMV 0.038995 0.016663 2.340195 0.0211 ACI 0.014123 0.036439 0.387574 0.6991 ACC -0.011985 0.040935 -0.292769 0.7703
R-squared 0.117753 F-statistic 1.428127 Adjusted R-squared 0.035300 Prob(F-statistic) 0.177755
Table 5.28 (c) (ii) PERS/PRED Sample (with COEA) Dependent Variable: SSVR Included observations: 118 after adjustments
Variable Coefficient Std. Error t-Statistic Prob.
C 0.234071 0.067287 3.478706 0.0007 COEA -0.313487 0.120056 -2.611162 0.0103
UCPMN -0.058497 0.037676 -1.552604 0.1235 UCPINST -0.045201 0.086437 -0.522937 0.6021 UCPGOV -0.056764 0.034976 -1.622942 0.1075 UCPFAM -0.026958 0.029204 -0.923093 0.3580
CFVR -0.082536 0.038740 -2.130537 0.0354 LGMV -0.003599 0.006607 -0.544819 0.5870
LBTMV 0.034062 0.014884 2.288445 0.0241 ACI 0.015249 0.035747 0.426586 0.6705 ACC -0.004207 0.039969 -0.105260 0.9164
R-squared 0.148410 F-statistic 1.864736 Adjusted R-squared 0.068822 Prob(F-statistic) 0.058116
163
5.9.9 Two stage least square of equations 3 and 4 for ABRES and
ABSDATA/ABSDATCA samples
The reduced form equation for COE, the only other endogenous variable is estimated first
to obtain ^CFVR. And the second stage is performed by regressing equation 3 and 4 using
^CFVR, ^SSVR and ^COE on the right hand side.
COE = μ0 + μ1 ^CFVR + μ 2,3,4,5 UCP + μ 6 ^SSVR + μ 7 ACI + μ 8 ACC +
μ 9SIZE + μ 10BTMV + μ 11β + δ7,8,9
The results are in table 5.29(a) ABRES sample and table 5.29(b) for
ABSDATA/ABSDATCA sample.
CFVR = η0 + η1^COE + η2,3,4,5UCP + η6 ^SSVR + η7 ACI + η8 ACC +
η9SIZE + η10BTMV + δ10,11,12
The results are in table 5.30(a) ABRES sample and table 5.30(b) for
ABSDATA/ABSDATCA sample.
5.9.10 Comparison between the ordinary least square results and two stage least
square results of equations 3 and 4 for ABRES and ABSDATA/ABSDATCA
samples
The results of estimating equation 3 under the two methods are similar. Under both
methods, the coefficients of size (LGMV) , book to market (LBTMV), substantial
shareholders’ voting rights (SSVR) and ultimate controlling party family (UCPFAM) are
164
significant. Under the ordinary least square method, the coefficient of ultimate controlling
party management (UCPMN) is also significant but not under the two stage least square.
In the estimation of equation 4, under the two stage least square method none of the
coefficients is significant, whilst estimation under the ordinary least square method, the
coefficient of substantial shareholders’ voting rights is significant.
165
Two stage least square regression of
COE = μ0 + μ1 ^CFVR + μ 2,3,4,5 UCP + μ 6 ^SSVR + μ 7 ACI + μ 8 ACC +
μ 9SIZE + μ 10BTMV + μ 11β + δ7,8,9
Table 5.29 (a) ABRES Sample Dependent Variable: COE Included observations: 141 after adjustments
Variable Coefficient Std. Error t-Statistic Prob.
C 0.111894 0.017903 6.249931 0.0000 LGMV -0.003877 0.001854 -2.091129 0.0384
LBTMV 0.029297 0.003879 7.551818 0.0000 BETA -0.000974 0.005597 -0.174099 0.8621
^SSVR -0.006824 0.003158 -2.160952 0.0325 UCPMN 0.014460 0.009069 1.594538 0.1132
UCPINST 0.022019 0.022494 0.978865 0.3295 UCPFAM 0.017922 0.006126 2.925417 0.0041
ACI 0.007248 0.011496 0.630511 0.5295 ACC 0.021960 0.015941 1.377554 0.1707
R-squared 0.426696 F-statistic 10.83332 Adjusted R-squared 0.387308 Prob(F-statistic) 0.000000
Table 5.29 (b) ABSDATA/ABSDATCA Sample Dependent Variable: COE Included observations: 151 after adjustments
Variable Coefficient Std. Error t-Statistic Prob.
C 0.140970 0.019784 7.125452 0.0000 LGMV -0.003676 0.001793 -2.050316 0.0422
LBTMV 0.030808 0.003870 7.961765 0.0000 BETA -0.004468 0.004918 -0.908526 0.3652
^SSVR -0.234542 0.133846 -1.752324 0.0819 UCPMN 0.012441 0.008840 1.407437 0.1615
UCPINST 0.013939 0.021402 0.651317 0.5159 UCPFAM 0.015899 0.006171 2.576656 0.0110
ACI 0.005320 0.011089 0.479784 0.6321 ACC 0.013954 0.014989 0.930980 0.3535
R-squared 0.435368 F-statistic 12.08001 Adjusted R-squared 0.399327 Prob(F-statistic) 0.000000
(^CFVR and UCPGOV have been automatically dropped)
166
Two stage least square regression of CFVR = η0 + η1^COE + η2,3,4,5UCP + η6 ^SSVR + η7 ACI + η8 ACC +
η9SIZE + η10BTMV + δ10,11,12
Table 5.30(a) ABRES sample Dependent Variable: CFVR Included observations: 141 after adjustments
Variable Coefficient Std. Error t-Statistic Prob.
C 0.659967 0.128311 5.143474 0.0000 LGMV 0.005764 0.013692 0.420968 0.6745
LBTMV 0.075271 0.028405 2.649901 0.0090 PREDSSVR -0.028532 0.023139 -1.233069 0.2197
UCPMN -0.322826 0.065730 -4.911361 0.0000 UCPINST -0.075879 0.165968 -0.457191 0.6483 UCPFAM -0.086149 0.044820 -1.922111 0.0567
ACI -0.071339 0.084848 -0.840784 0.4020 ACC -0.124718 0.115976 -1.075381 0.2842
R-squared 0.254124 F-statistic 5.621643 Adjusted R-squared 0.208920 Prob(F-statistic) 0.000004
Table 5.30(b) ABSDATA/ABSDATCA sample Dependent Variable: CFVR Included observations: 151 after adjustments
Variable Coefficient Std. Error t-Statistic Prob.
C 0.353963 0.175945 2.011779 0.0461 PREDCOE 3.829805 2.669489 1.434659 0.1536 UCPMN -0.403098 0.128100 -3.146754 0.0020
UCPINST -0.045916 0.178926 -0.256621 0.7978 UCPGOV -0.156878 0.264480 -0.593159 0.5540 UCPFAM -0.174957 0.124667 -1.403398 0.1627
PREDSSVR -2.423920 3.017809 -0.803205 0.4232 ACI -0.124357 0.112463 -1.105756 0.2707 ACC -0.037661 0.207832 -0.181208 0.8565
LGMV 0.037754 0.026198 1.441102 0.1518
R-squared 0.244085 F-statistic 5.058775 Adjusted R-squared 0.195835 Prob(F-statistic) 0.000006
167
Summary of results
The following table provides a summary of results of hypothesis testing.
Table 5.4 Summary of results
Hypothesis Hypothesis statements Results
1
There is a negative relationship between earnings quality and ownership structure
Not supported for all earnings quality measures (ABRES, ABSDATA, ABSDATCA, PERS and PRED) and both ownership structure measures (CFVR and UCP)
2
There is a relationship between earnings quality and substantial shareholders
Supported for between earnings quality measures ABRES ,ABSDATCA, PERS and SSVR. Not supported for between ABSDATA, PRED, and SSVR
3
There is a relationship between earnings quality and audit committee
Supported (weakly) for between earnings quality measure ABSDATA and ACI. Not supported for between ABRES, ABSDATCA, PERS and PRED, and ACI. Not supported for between ABRES, ABSDATA, ABSDATCA,PERS and PRED, and ACC.
4
There is a positive relationship between cost of equity and earnings quality
Supported for between COE and earnings quality measures ABRES and PRED. Not supported for between COE and earnings quality measures ABSDATA, ABSDATCA and PERS. Supported for between alternative market assessment measure COEA and earnings quality measures ABRES, ABSDATA, ABSDATCA and PRED. Not supported for between alternative market assessment measure COEA and earnings quality measure PERS.
168
Table 5.4 Continued Hypothesis Hypothesis statements Results
5
There is a negative relationship between market return and earnings quality
Not supported for both measures of market assessment AVMR and YER
6
There is a positive relationship between ex post market return -AVMR and ex ante market assessment
Not supported for both measures of market assessment AVMR and YER, and both measures of analyst assessment COE and COEA
7
There is a relationship between cost of equity and ownership structure
ABRES sample : Supported for between COE and UCPFam. Not supported for between COE and all other UCP. Not supported for between COE and CFVR Not supported for between COEA and CFVR, COEA and all UCP ABSDATA/ABSDATCA sample : Supported for between COE and UCPFam. Not supported for between COE and all other UCP. Not supported for between COE and CFVR. Not supported for between COEA and CFVR, COEA and all UCP PERS/PRED sample : Not supported for between COE and any of SOC measures CFVR and UCP Not supported for between COEA and CFVR, COEA and all UCP
169
Table 5.4 Continued Hypothesis Hypothesis statements Results
8
There is a relationship between cost of equity and substantial shareholders
ABRES sample : Supported for between COE and SSVR (negative relationship) Supported for between COEA and SSVR (negative relationship) ABSDATA/ABSDATCA sample : Supported for between COE and SSVR (negative relationship) Supported for between COEA and SSVR (negative relationship) PERS/PRED sample : Supported(weak) for between COE and SSVR (negative relationship) Supported for between COEA and SSVR (negative relationship)
9
There is a relationship between cost of equity and audit committee
Not supported in all samples and measures of analyst assessment and audit committee
10
There is a relationship between type of ownership and cash flow/ voting rights disparity.
Not supported
11
There is a reverse relationship between cash flow/ voting rights disparity and cost of equity.
Not supported.
170
Table 5.4 Continued Hypothesis Hypothesis statements Results
12
There is a reverse relationship between cash flow/ voting rights disparity and the monitoring mechanisms.
ABRES sample : Supported for between CFVR and SSVR (negative relationship) Not supported for between CFVR and ACI (negative relationship) Weak support for between CFVR and ACC (negative relationship) ABSDATA/ABSDATCA sample : Supported for between CFVR and SSVR (negative relationship) Not supported for between CFVR and ACI (negative relationship) Weak support for between CFVR and ACC (negative relationship) PERS/PRED sample : Supported(weak) for between CFVR and SSVR (negative relationship) Weak support for between CFVR and ACI (negative relationship) Not supported for between CFVR and ACC (negative relationship)
171
CHAPTER 6
DISCUSSION
6.0 Introduction
A caveat
A caveat is in order for the small sample size. As described in section 5.2 the sample
size is reduced a number of times by the lack of information to estimate the variables.
Even though the final sample size is sufficient for the statistical analyses employed, the
statistical inference made is somewhat limited for generalization. The results shed some
light on the relationships studied and provide some evidence to support the hypotheses
tested. Future research that includes more companies and years is needed to provide a
robust set of results.
Main findings
This study has not found a significant relationship between ownership structure and
earnings quality. As such the existence of expropriation or otherwise by the controlling
party is not associated with the ownership structure measures; the cash flow rights for
non-pyramidal structure companies or the cash flow to voting rights ratio for pyramidal
structure companies and the type of ultimate controlling party.
This insignificant result augurs well with the significant negative association between
substantial shareholders voting rights and earnings quality in majority of the samples
examined. This suggests an effective monitoring role played by substantial
shareholders. However similar evidence is not found for the association between audit
committee characteristics; independence and competence, and earnings quality. Only in
172
one sample, a weak significant association is found between audit committee
independence and abnormal total accruals as hypothesized.
The hypothesis that there is a cost of equity effect of earnings quality is largely
supported. However the results are not consistent across both estimates of cost of equity
and the different earnings quality measures. The evidence is consistent and strongest for
between accrual quality (ABRES) and predictability (PRED) and both measures of cost
of equity. Discretionary current and total accruals is significantly related to the
alternative measure of cost of equity (COEA). Persistence (PERS) is not significantly
associated with either measures of cost of equity.
There is some support that the type of ownership, in particular where the type of
controlling party is family/related individuals, is perceived as information risk and is
priced. Since there is no support for the hypothesis that the type of ultimate controlling
party influences earnings quality, therefore the significant relationship between type of
controlling party and cost of equity is not driven by the effect of earnings quality on
cost of equity. Market prices family type of controlling party because market perceives
family controlled companies to have relatively more information that is private than
public hence these companies pose an information risk.
However the hypothesized relationship between cash flow/voting rights and cost of
equity is not supported. This is consistent with the insignificant relationship between
cash flow/voting rights and earnings quality. Since there is no evidence of expropriation
associated with cash flow/voting rights judging by the insignificant relationship
between cash flow /voting rights and earnings quality, cash flow/voting rights is not an
173
information risk and therefore priced. By the same token, there is strong and consistent
evidence that substantial shareholding is priced as there is evidence of effective
monitoring role. Neither of the audit committee characteristics are found to be priced by
the market.
There is no evidence that cash flow/voting rights changes in response to changes in cost
of equity. However the results suggest that cash flow/voting rights changes in response
to changes in monitoring mechanisms. The evidence is consistent for changes in
substantial shareholdings in all the three samples. The results for the audit committee
characteristics are rather mixed. The negative association suggest that increase
monitoring is associated with lowering of cash flow/ voting rights disparity and control.
6.1 The ownership structure and earnings quality
6.1.1 Cash flow/ controlling rights and earnings quality
The expected relationship between ownership structure, cash flow/voting rights (CFVR)
and earnings quality is negative given that high CFVR represents low expectation of
expropriating behavior and that high earnings quality measures represents poor quality.
The expected relationship applies to both companies with pyramidal ownership
structure and those with non-pyramidal ownership structure. For pyramidal structure
high CFVR means the cash flow rights is closer to voting rights thus there is no
incentive to expropriate because the loss suffered is proportionate to the control. Unlike
where CFVR is low the loss suffered is disproportionately low. For non-pyramidal
companies, the higher the cash flow rights which equals the voting rights the more
174
aligned the interest of the controlling party to the interest of the company and therefore
the interest of other non-controlling parties.
For all samples the coefficients of CFVR are highly insignificant. In fact only for
sample ABRES the coefficient is negative. The coefficients for the other samples are
positive. The results show that the disparity between cash flow and voting rights for
pyramidal companies, and levels of cash flow rights for non-pyramidal companies, do
not explain earnings quality. In other words, there is no association between high (low)
CFVR and less (more) expropriation or manipulation of earnings. A possible
explanation to this is that cash flow rights do not provide an incentive to expropriate or
manipulate earnings for this sample because as further explained below, the companies
in this sample are with ultimate controlling party with a higher level of control rights
than those in previous research (Claessens 1998(b), Fan & Wong 2002) and therefore
the cash flow rights is not an incentive to expropriate.
The comparison between Claessens (1998b) sample and the samples under study
provides an insight why this is so. Table 4.11indicates that even the minimum voting
rights found in the samples in this study is quite close to the mean of 28% in Claessens
(1998b) study. Seventy five percent of companies in the samples have ultimate
controlling party with voting rights above 37%. Similarly the mean cash flow rights in
all samples in this study are almost twice that reported in Claessens (1998b). 75% of
the companies have ultimate controlling party with cash flow rights above 28%.
Similar pattern is observed in Fan and Wong (2002) whose study includes 177
Malaysian companies. The mean voting rights is 31% whilst the mean cash flow rights
175
is reported to be 26%. As in Claessens et al (1998b) study the reported mean CFVR is
85%. Further both Claessens et al (1998b) and Fan and Wong (2002) capped the voting
rights at 50%. They stopped analyzing the voting rights of the ultimate controlling party
once the voting rights breaches 50%. So the maximum voting rights for the companies
in the sample is 50%.
Francis et al (2005) reported significant association between disparity of cash flow and
voting rights, and informativeness of dividends and earnings for US listed companies. It
is well known that capital markets in Europe and the US consist of companies with
diffused ownership. Thus the controlling party ownership rights is likely to be at lower
level.
The other noteworthy difference is that both Fan and Wong (2002) and Francis et al
(2005) use market based of earnings quality measures. Thus they examine association
between market perception as embedded in the measure of informativeness and cash
flow/controlling rights disparity. Thus what they are measuring is the credibility of
earnings figures in the face of cash flow/controlling rights disparity. In Fan and Wong
(2002) words - “This does not always mean that there is an outright earnings
manipulation to cover up possible earnings effect of wealth extraction”. The accounting
based earnings quality measures as used in this study are measuring earnings
manipulation after controlling for other economic condition. Francis et al (2004)
reported low correlation between market based and accounting based earnings quality
measures.
176
Certainly the off setting effect of increasing cash flow rights and increasing voting
rights is complex and merit more research. Previous research such as Claessens (1998b)
also found that at higher level of control the tendency to expropriate is higher. However
the insignificant results in this study is consistent with this finding and the prevailing
theory and prediction regarding the association between cash flow/voting rights and
earnings quality. It suggests that at higher level of control the ultimate controlling party
with varying degree of disparity between cash flow and voting rights would expropriate.
However the cash flow rights are not the incentive and thus the lack of significant
association.
Or the other possible explanation is that the ultimate controlling parties do not
expropriate as at higher level of control with the even higher level of cash flow rights,
the ultimate controlling parties of pyramidal and non-pyramidal companies may find the
cash flow loss is too much to expropriate.
Consider this hypothetical example of pyramidal companies. Suppose there are two
companies, A and B. The controlling party of A has cash flow rights of 8% and voting
rights of 40%, thus CFVR ratio of 20%. The controlling party of B has cash flow rights
of 20% and controlling rights of 100%, thus CFVR of also 20%. The controlling party
of company A would share a loss of 80,000 for a loss of 1 million in company A, whilst
a loss of 1 million in company B, the controlling party’s share in the loss is 200,000.
For non-pyramidal companies the lack of significant results does not prove that there is
an interest alignment between the ultimate controlling party and other shareholders. It
simply suggests that expropriation behavior is not found or not associated with levels of
177
cash flow rights. The coefficient of CFVR is positive and significant in a separate
regression run for non-pyramidal PRED sample. This suggests that the ultimate
controlling party is entrenched as increasing control is associated with poorer quality of
earnings. It is also interesting to note that the substantial shareholders’ voting rights is
not significantly related to earnings quality.
To prove interest alignment or otherwise entrenchment requires further research. It
requires measurement of not only expropriating behavior, but also measurement of
value maximizing behavior. In a way the positive relationship between cash flow and
market value as described in section 5.8.3 , suggests that perhaps there is an interest
alignment. Of course this is far from proving it as market value is a crude measure.
6.1.2 Ultimate controlling party and earnings quality
The composition of ultimate controlling party in this study reflects generally what is
found in previous studies and that is there is a high proportion of family as the ultimate
controlling party. Thus there is an expectation of the alleged expropriating behavior.
However none of the coefficients in equation 1 regressions is significant. When the
regressions are run separately for pyramidal and non-pyramidal companies similar
results are obtained. Thus there is no support for the hypothesis that the type of
controlling party is associated with high or low earnings quality.
Again this could be attributable to the composition of companies in the sample, where it
is dominated by those with ultimate controlling party with high controlling votes.
Another plausible explanation is that the controlling party is far removed from the
companies operations and delegate the running of the companies to professional
178
managers. The executive managers are the ones who possibly would know the company
enough to expropriate or manipulate accounting figures.
A recent accounting scandal is a good example. The company involved, Transmile Bhd
is ultimately controlled by the Kuok family. The alleged perpetrators are the former
chief executive officer, who is the founder but not the ultimate controlling party even
though he has a substantial shareholding, together with a former chief financial officer
and an executive director.
The following Table 6.1 provides the breakdown of companies in the three samples
(ABRES, ABSDATA,ABSDATCA, PERS/PRED) whose ultimate controlling party is
also in an executive position for example the controlling party is the chief executive
officer or managing director. Although majority of the ultimate controlling party is in an
executive position but the proportion of about 55% is not overwhelming. The proportion
of family controlled companies with a family member in an executive position is higher
than similar proportion for all types of ultimate controlling parties.
Table 6.1 Percentage of companies whose controlling party is also in an executive
position ABRES ABSDATA/
ABSDATCA PERS/PRED
Percentage of companies whose ultimate controlling parties is also in an executive position
55%
55%
54%
Percentage of companies whose ultimate controlling parties is a family (Table 5.4)
68%
70%
68%
Percentage of companies from those whose ultimate controlling parties is a family and the family member is in an executive position
73%
86%
79%
179
Equation 1 is rerun replacing the ultimate controlling party variable (UCP) with a
variable MNGT ( where 1 is for companies whose ultimate controlling party is in
executive position and 0 otherwise). The results (refer to Appendix 5) are largely the
same as those described in section 5.9 (Table 5.20). When the same relationship is rerun
for family controlled companies similar results are obtained (refer to Appendix 6). Thus
there is no conclusive evidence regarding the type of controlling party having an
influence on earnings quality whether they are in executive position or not.
Even though the ultimate controlling party is in an executive position, there may not be
an association with expropriating behavior for two reasons. One, that there is no cash
flow incentive, as discussed earlier, and two, they may not have control of information
in the company, sufficient to manipulate it in cases where there are chief financial
officer who are more technically competent in accounting matters and may have the
incentive through accounting measure performance based compensation.
The complexity of the type of controlling party vis-a-vis the issue of proprietary control
of information merits a separate study. The sample size of this existing study cannot
sufficiently accommodate such variations. The sample size of this study has been
limited by data availability of other variables such as the cost of equity and earnings
quality which future research need not be limited by, by focusing away from those
variables.
Thus future research needs to consider the existence of the ‘empire building’ managers.
Future research needs also to consider whether the ultimate controlling party traced is in
180
a position to control the information flow from the company. If they are not they cannot
be expected to expropriate or otherwise, or they do not pose an information risk.
Even though there are family members who are in executive positions, future research
needs to analyze the existence of others who have proprietary information of the
company such as a chief financial officer who has incentive to manipulate earnings.
Finally the lack of significant association between the type of ultimate controlling party
and earnings quality could be due to the effective presence of a substantial shareholder.
As found, substantial shareholders’ voting rights, SSVR, is fairly consistent in showing
negative association with earnings quality and cost of equity variables as discussed
further below.
6.1.3 Monitoring mechanisms and earnings quality
6.1.3.1 Substantial shareholder voting rights and earnings quality
The results from regression of equation 1 show strong negative relationship between
substantial shareholder voting rights and earnings quality measures of accrual quality
(ABRES) and absolute discretionary current accruals (ABSDATCA). It suggests that
the higher the degree of substantial shareholders’ presence in the companies the lower
the measures of accrual quality and discretionary accrual or the higher the quality of
earnings. Similar results are obtained when equation 1 is rerun separating the companies
into those whose controlling party is in executive position and those who are not
(Appendix 5) and separating family owned companies into those that are family
181
managed and those that are not. In the reruns the significance of substantial
shareholders’ voting rights coefficient is at 10% level.
Since these two variables are direct measure of accounting manipulation, it also
suggests that substantial shareholders’ presence is associated with lower accounting
manipulation. This is generally consistent with previous research. However previous
researches examine the shareholding of specific type of substantial shareholder such as
institution (Jung and Kwon 2002, Koh 2003 and Chung et al 2004,) and foreign (Chung
et al 2004).
The insignificant coefficients of substantial shareholders’ voting rights in the
regressions using PERS and PRED suggest that the presence of a substantial
shareholder does not affect earnings persistence and predictability. One possible
explanation to this is that PERS and PRED are time series properties. Even though
manipulation and expropriation would influence these properties, other influence
namely variability of revenue may be overriding. Variability of revenue is something
not within the control of substantial shareholder to the extent that it is economically
related.
The results from separate regressions of pyramidal and non-pyramidal sample of
companies provide similar conclusion. Of the three significant regression equations for
the non-pyramidal samples, two estimated equations are with coefficients of substantial
shareholders’ voting rights that are negatively significant.
182
6.1.3.2 Audit committee and earnings quality
The results from table 5.20 show no support for the association between audit
committee independence and earnings quality variables. This is not consistent with
Mohd Salleh et al (2007) findings, where a significant relationship is found although
Mohd Salleh et al (2007) found significant relationship with fully independent audit
committee. Similarly no association is found between audit committee competence with
measures of earnings quality. As with Mohd Salleh et al (2005) no significant
association is found between audit committee competence as measured by possession of
accounting knowledge, with discretionary accruals. Mohd Salleh et al (2005) found
though the interaction between frequency of committee meeting and competence could
explain discretionary accruals. Mohd Salleh (2005) did not find significant relationship
between other characteristics of audit committee such as frequency of meetings, size,
accounting knowledge and proportion of non-executive members).
This is in contrast to research done in the US market for example Klein (2002) and
Abbott et al (2005) that find significant association between audit committee
independence and earnings management. Kinney, Palmrose and Scholz (2004) find a
significant association between the non-audit services fees (which renders external
auditor less independent) and restatements (a proxy for reporting quality).
The descriptive statistics from Table 5.8 shows that on average companies just meet the
requirement of independence and competence. The relatively small standard deviations
possibly indicate that there is no sufficient variability to explain variability in earnings
quality.
183
The insignificant results from this research and other Malaysian study (Mohd Salleh et
al 2007) possibly show that compliance in form of requirements is not sufficient to
achieve the desired effects.
6.2 Earnings quality and cost of equity
Earnings quality variables proxy information risk in terms of the risk associated with the
reliability of information. As discussed in Francis et al (2004), in general reliable
earnings information assists investors to make investment decisions and without
reliable information (as proxied by earnings quality measures) or perceived reliable
information (as proxied by market based earnings quality measures), investors require a
higher return.
The results from Table 5.21 (a)(i) and (ii) for accrual quality (ABRES) is consistent
with this theory. For both measures of cost of equity, high value of earnings quality
measure(low earnings quality) is significantly associated with high COE/COEA (high
required return). Investors demand high return for companies with current accruals that
maps poorly with previous, current and future period cash flow. Large value of ABRES
means a large proportion of accrual are not translated to cash flows or cannot be
explained by cash flows in the preceding, current and future period. This implies also
that a large proportion of earnings are not translated to cash flows and therefore the
earnings figure is to be suspected.
184
Similarly for predictability (the absolute residuals from regression of earnings on past
earnings), which provides measure of how well earnings information can be used to
predict future earnings, the regression results show significant association with
COE/COEA. This is evident of not only the theory but also the reported importance of
earnings figure by practitioners namely analysts. However persistence which is derived
from the same procedure and is the coefficient of lagged earnings, is not similarly
priced. Persistence supposedly measure the proportion of earnings that is recurring is
not significantly associated with both COE measures. It is to be noted that PERS is also
not correlated with the established economic determinants of earnings (section 5.8.2).
This brings into question the construct validity of PERS.
Both the absolute discretionary accruals which are proxies of earnings management are
significantly related to COEA measures but not to COE. Hence this is fair evidence of
abnormal discretionary accruals being priced.
This study therefore provides fairly strong evidence of the cost of equity effect of
earnings quality. It is also consistent in general with Francis et al (2004) and Francis et
al (2006).
6.3 Cost of equity and market return
When the cost of equity measures are replaced by market return measures, average
monthly return and year end return (lagged three months), in an attempt to differentiate
assessment by market (ex ante and ex post), none of earnings quality measures explain
significantly market returns. Neither are the return measures significantly correlated
185
with the cost of equity measures. The same results are obtained when excess returns
measure is used.
This is in contrast with studies in the US (Francis et al 2004, Aboody et al 2005, Francis
et al 2008b) where measures using both ex post (average daily and annual returns,
CAPM and (Fama & French 1993) excess returns) and ex ante versions (cost of equity)
produce consistent results and that is earnings quality is priced. This besides the fact
that there has been doubt about realized return being a good proxy for expected return
(Elton 1999).
A possible explanation is that realized returns for Malaysian companies are influenced
by many information surprises that render it a poor measure of expected return. Elton
(1999) discussed this possibility in examining other measures of expected returns.
6.4 The relationship between ownership structure and cost of equity
6.4.1 Cash flow/ controlling rights and cost of equity
Whilst earnings quality poses risk in terms of reliability/precision of information,
ownership structure poses risk in terms of amount of information that is private. The
more private the information the higher the required return. This is the essence of the
theoretical studies of Easley and O’Hara (2004) and Leuz and Verrecchia (2004) as
discussed in Francis et al (2004). There has not been any studies in the US that test
ownership structure and the cost of equity. This is so because in the US ownership is
diffused and if disparity of cash flow and voting rights exists they are through the
existence of dual class shares and at low level of control. Also the existence of dual
186
class shares requires disclosure and therefore is transparent. Fan and Wong (2002)
examine ownership structure and earnings informativeness on the premise that
concentrated ownership through pyramidal ownership structure inhibits information to
the public.
The regression results for equation 3 provide no support for the hypothesis that
ownership structure is priced. The coefficients for CFVR are not significant in all
samples using both measures of cost of equity. Market perceived CFVR as irrelevant in
required return. The lack of any effect on cost of equity may be due to mitigating factor
in particular the presence of substantial shareholder as discussed below. Briefly in
relation to the theory, with the mitigating factor the amount of private information is not
sufficiently large as to pose a risk.
6.4.2 Ultimate controlling party and cost of equity
There is a weak support for the hypothesis that the type of ultimate controlling party
has an effect on cost of equity. In reference to company ultimate controlling party, the
family ultimate controlling party is highly significant in explaining cost of equity in
samples ABRES and ABSDATA/ABSDATCA for when measure COE is used.
Similarly the managerial ultimate controlling party moderately proved to be significant.
By comparing the coefficients and associated probabilities family ultimate controlling
party is more significant. None of the other types of ultimate controlling party is
significant. In the regression using COEA measure, however, none of all the types of
ultimate controlling party is significant. The significant results, albeit weak, are
187
consistent with other studies such as Claessens (1998 b) that found negative relationship
between family control and market valuation.
Another point to note is that as earlier discussed type of ultimate controlling party is not
significantly related to earnings quality. However it is found that managerial and family
ultimate controlling party are significantly related to COE and are therefore priced.
This suggests that even though family and managerial ultimate controlling party are not
associated with earnings quality and the implied earnings manipulation and
expropriation, these two types of ultimate controlling party are perceived to be
relatively more associated with information risk related to the amount of private
information. The companies with family and management as ultimate controlling party
are perceived to have more private information.
6.4.3 Monitoring mechanisms and cost of equity
6.4.3.1 Substantial shareholder voting rights (SSVR) and cost of equity
As with earnings quality, the coefficients of SSVR in the regression with cost of equity
are consistently and negatively significant. When COEA measure is used the
coefficients for all samples are significant at 1% level. In the regression using COE, the
coefficient of SSVR in the ABSDATCA/ABSDATA sample is significant at 5% level,
while for the other two samples at a relative weak 10% level.
The effectiveness of the monitoring role of a substantial holder is ambiguous and
therefore an empirical question. The significant negative relationship with earnings
quality suggests an effective monitoring role. The negative relationship with cost of
188
equity suggests that a lower percentage holding of substantial shareholder poses
information risk. This is consistent with the theory in general, as an increase in voting
rights afford the substantial shareholder more bargaining power for inclusion in the
decision making process such as being a member of the board of directors. Therefore
this increase the chance of more information flow to the public, information which
otherwise would be in the proprietary control of the ultimate controlling party.
The significant negative relationship found between substantial shareholders voting
rights and cost of equity also contributes towards the ‘information argument’ (Fan &
Wong 2002). That is the presence of others other than the controlling party increase the
likelihood that proprietary knowledge of the company is shared to the others and
decrease the likelihood that it is concentrated to certain individual which leads to
opacity of information. The wider the set of informed individuals the greater the
likelihood that information ‘leaks’ to the public and thus reduce the company’s
information risk.
Previous researches (Jung and Kwon 2002, Koh 2003 and Chung et al 2004) associate
substantial shareholder’s voting rights with earnings quality measure as described
earlier. The cost of equity effect, that is the information risk effect, has never been
examined.
189
6.4.3.2 Audit committee and cost of equity
As with earnings quality, audit committee independence and competence could not
explain variability in cost of equity. It suggests that audit committee characteristics are
perceived as having no impact of the amount of information made public.
6.5 The relationship that examines whether market assessment and the monitoring
mechanisms could explain changes in ownership structure
This aspect of the study is exploratory. Past researchers (Demsetz 1983, Demsetz 1985,
Demsetz and Villalonga 2001, and Mak and Li 2001) examined the issue of whether
performance in turn affect ownership as an empirical question and not a question
grounded on theory. Demsetz in particular argues that ownership changes too in
response to market expectations.
The coefficients of COE/COEA in all regressions are not significant which means that
the cost of equity does not explain CFVR. This means the ultimate controlling party do
not change in response to changes in market assessment. This augurs well with the
insignificant findings above that suggests investors see as irrelevant the disparity
between cash flow and voting rights in the pricing of risk.
There is support in all regressions (except a weak one for PERS/PRED sample) for the
hypothesis that variability of SSVR explains variability in CFVR in a negative direction
that is there is a significant negative relationship. This means the disparity between cash
flow and voting rights narrows (high CFVR) in the face of lower SSVR which in turn
190
would be penalized with higher cost of equity. Note that the substantial shareholder’s
share holding is priced (as described earlier that higher SSVR is associated with lower
COE) and that there is no significant relationship between CFVR and cost of equity
either way. Thus the significant relationship between SSVR and CFVR is not to do with
the consequential effects of SSVR on cost of equity. It may have to do with the control
of company. In the face of increasing SSVR, hence increasing bargaining power for
involvement in decision making, the ultimate controlling party may relinquish his cash
flow rights.
There is no evidence to support the hypothesis that audit committee competence has a
positive relationship with earnings quality. Neither there is evidence to support that that
measure is priced. However in samples ABRES and ABSDATA/ABSDATCA
regression there is a weak support at 10% for the hypothesis that variation in ACC
explains variation in CFVR. Again as with the substantial shareholder voting rights, this
may have to do with the control over decision regarding the company’s accounting
choices.
6.6 The relationship between substantial shareholders’ voting rights and elements
of ownership, monitoring mechanisms and cost of equity
The rather low R squared is expected as there are conceivably other factors that
determine the shareholdings of these shareholders. As any other shareholders, other than
performance of companies, a substantial shareholder would also consider for example
the relative risk/ return relationship in his portfolio of investments. This cannot be
captured by the estimated relationship.
191
However the limited purpose of the estimation and that is to examine whether the
substantial shareholders voting rights could be explained by the stated variables, is
served. In all the regressions the coefficients of COE, CFVR and LBTMV are
consistently significant. The coefficient of COE is negative throughout the samples
which suggests that high substantial shareholders’ shareholdings is associated with low
cost of equity.
The coefficient of CFVR is negative, which also suggests that high substantial
shareholders’ shareholdings is associated with low cash flow/voting rights. In other
words substantial shareholders’ voting rights are high in companies where the
separation of ownership and control related problems are expected to be high.
On the other hand, high substantial shareholders voting rights is associated with high
book to market as indicated by the positive coefficient.
192
CHAPTER 7
CONCLUSION
7.0 Introduction
This thesis is motivated by the desire of the relevant authority to move towards a
market-based regulation for the Malaysian capital market, which inevitably means self-
regulation by the market players, and where the quality of information to the market is
important. Against a background of alleged ownership structure of companies that
exacerbates the separation of ownership and control conflict, and possibly limit
transparency of information to the public, this thesis set out to examine if such
ownership structure leads to lower earnings quality. And if indeed the market is self-
regulating this thesis examines if the market players are assessing the earnings quality
and the elements of governance namely the ownership structure and the monitoring
mechanisms that, one, is imposed by the relevant authority (audit committee) and
another that emerges from market forces (substantial shareholding). The market
assesses these elements by requiring a higher or lower return where accordingly these
elements lead to the companies being of higher or lower information risk.
7.1 Conclusion and contribution highlights
The results from this study show that the ownership structure of Malaysian listed
companies is indeed characterized by high concentration of control or voting rights, and
dominated by family or individuals as the ultimate controlling party. These elements
point towards a tendency to expropriate, to manipulate earnings and to limit information
flow to the public. This study, however, has not found any evidence that relates cash
193
flow /voting rights ratio with earnings quality the proxy for expropriation and
manipulation of earnings. This suggests that whether there is or there is not such
behavior, it is not related to cash flow/voting rights.
Whilst this may also suggests that controlling party’s behavior might have been aligned
with other shareholders, it is yet to be proven as there is a need to prove the controlling
party is value maximizing. The conclusion remains that for this sample of companies
with ultimate controlling party holding high level of control, there is no more
expropriation or manipulation of earnings at higher level of disparity between cash flow
and voting rights or higher level of control, as there are at lower level. Thus there is
lack of significant association.
An also important finding is that investors do not perceive such ownership structure as
information risk and therefore priced. The importance of this finding lies in the
consistency with the finding that there is no significant association between cash
flow/voting rights and earnings quality.
The consistency in the two findings is an important contribution to the theory and the
body of knowledge in ownership structure and the pricing of information risks. Whilst
past researches examine ownership structure and earnings quality, and separately
examine earnings quality and cost of equity, none has examined ownership structure as
a source of information risk. Although the results are not in the affirmative, it confirms
the theory, that is what is perceived as a source of risk is priced.
194
There is evidence that the type of controlling party namely management and family is
priced, whilst they are not significantly associated with earnings quality. Again this
finding albeit weak contributes to theory. As these types of ownership are concerned
with individuals’ interest directly as compared to others in the study such as
government, institution and foreign company, they are perceived to have more private
information thus pose a risk.
There is a fairly strong association between the cost of equity and earnings quality.
There is nothing new in this finding except that this study examines companies in an
emerging market. Therefore even in an emerging market investors are sophisticated and
do price earnings quality with low earnings quality being perceived as information risks.
In relation to the ownership structure, investors do not however perceive ownership
structure in particular the cash flow and voting rights disparity as the primary driver of
earnings quality.
An important finding and therefore contribution is the significance of predictability as a
dimension of earnings quality in required return. Whilst there have been many studies
on Malaysian companies that examine abnormal accruals in relation to many variables
such as board characteristics, managerial ownership, etc. there is none to date on
predictability, a dimension that standard setters have long expounded as an important
attribute of accounting information. The finding that predictability is priced has an
important implication to Malaysian standard setters and preparers of accounts as
elaborated below. Also the finding on abnormal accruals and the significant results
while consistent with other studies add to the body of knowledge as there has never
been any Malaysian studies that prove that abnormal accruals are priced. The
195
consequence of this is that the preparer may gain in manipulation of accounts but they
stand to lose in terms of higher required return by investors.
This study has consistently shown that substantial shareholders’ voting rights is
significantly associated with high earnings quality especially the discretionary ones
which suggest the presence of substantial shareholder is an important monitoring
mechanism. In contrast to the results shown by the rules based mechanism in particular
audit committee, where none of the characteristics of audit committee is significantly
associated with earnings quality and neither is it priced. The finding that substantial
shareholders’ shareholding is priced is a new and significant contribution not only in the
Malaysian context but also elsewhere. Thus not only substantial shareholder is an
important mechanism, it is also perceived as such by the market.
The amount of shares that substantial shareholders buy in the company depends on a
number of factors which is beyond this research. However it is not conceivable that it is
related to the cash flow/voting rights of the controlling party, even though statistically
they are correlated. Given that in the reverse relationship test, substantial shareholding
can affect the cash flow/voting rights of the controlling party, it is an important
contribution in the sense that it confirms the belief that ownership changes in response
to changes in market, even though as earlier reported ownership does not response to
market assessment or changes in cost of equity. The controlling party can change his
shareholding in the face of changing market expectation. Market expects with increase
in substantial shareholding there will be an increase in monitoring and increase
information to the market.
196
Note that this is an evidence from cross-sectional variation. It is not in contradiction
with earlier report that ownership is fairly stable over time (Claessens et al 1998a,
Section 4.1). To reconcile, variation of shareholdings (controlling, substantial) are not
seen overtime.
7.2 Implication
1. Earnings quality is relevant for investors in that they require a premium to
compensate for the risk they take that is associated with unreliable earnings figure or
earnings figure that suggests there has been some manipulation. This concurs well with
anecdotal evidence on the use of earnings figure for market valuation and as found by
Price Waterhouse survey despite earnings figure being historical.
The implication for preparers is that manipulation of earnings figure may achieve their
goal to cover misrepresentation or omission, but this behavior backfires when they stand
to pay higher price in the form of higher cost of equity.
Given that predictability of earnings is priced, this quality is not an artificial construct
that glossed statement of accounting objectives and has relevant in the market. An
implication for standard setter is that in selecting a standard from a choice of treatment
careful consideration need to be made on the impact of such treatment on the earnings
figure, that is whether they will make earnings figure more predictable or otherwise.
This gives weight, for example to the argument against fair valuation that makes
earnings more volatile and possibly less predictable.
197
2. Setting up extensive rules and regulations to improve corporate governance drains
resources of the regulators and society at large. If it is proven that company complies
only in form and not in substance, and that market mechanisms would be more
effective, then resources should be diverted to ensuring conducive conditions exist in
the market for the mechanisms to thrive.
7. 3 Limitations of study
1. This study is limited by the availability of estimated earnings per share. To be able to
examine more years require historical estimates which are far too expensive. The
limited number of data limits simultaneous testing of relationship, which would provide
richer analysis. However the main research questions are answered.
2. This research has looked at one dimension of the ultimate controlling party and that is
type as established in past research. However the ability and potential to expropriate
may be explained by another layer of control and that is control obtained by possession
of knowledge regarding the operations. This is achieved by direct involvement with
operation or close relationship with those in direct involvement. This research has not
differentiate the ultimate controlling party who are and are not in executive position.
3. This research has not separated out the substantial shareholders who are in actual fact
a partner of the ultimate controlling party and the substantial shareholders who are
really an outsider. Obviously the presence of the former may impair its monitoring role.
198
7.4. Future research
1. Certainly the complex effects of the ultimate controlling party holding varying
degree of cash flow rights at varying levels of cash flow rights held and the ultimate
controlling party holding varying degree of voting rights at different levels of control is
an important and exciting area of research. Future research should also examine the
combined effects of such ownership rights.
In future examination of the effects of ownership structure, separate measurement
should be made for the tendency to expropriate and for the tendency to maximize
wealth. The reason being one behavior does not necessarily exclude the other. It may
also be necessary to draw from theories predicting human threshold for tolerance for
loss and gain at different level of cash or wealth possession in general.
2. There is still a wider scope of research into the ability and potential to expropriate or
otherwise by the controlling party. The voting rights certainly afford the ability,
however this may be enhanced by direct involvement in company’s operations or
impede by non-involvement or the existence of other parties not necessarily with the
controlling rights, such as the ‘empire building’ manager or a ‘significant other’
substantial shareholder. This research has provided a modest evidence regarding the
association of substantial shareholder rights and earnings quality and cost of equity, but
more can be done in terms of explicating the relationship between this ‘significant
other’ party and the controlling party and the consequences.
199
3. From observation the presence of this ‘partnering’ ownership, that is, an ultimate
controlling party and a ‘significant other” shareholder appear to be common. Again how
common and what are the consequence should be a subject of an empirical question.
Also it would be useful to examine the effectiveness of a substantial shareholder’s
presence at different levels of share holdings to take into account the investment time
horizon of the substantial shareholder.
4. The mixed results from this research and previous research point to a need to assess
the relative importance and effectiveness of a rule based and a market imposed
monitoring mechanisms.
200
REFERENCES
Abbott, L.J., Parker, S. and Peters, G.F. (2004). Audit committee characteristics and restatements. Auditing: A Journal of Practice and Theory. 23(1),p.69-87 Abdul-Rahim, R (2006). Model penetapan harga asset : Peranan kecairan dalam konteks model Fama-Frence. Phd Thesis. Universiti Kebangsaan Malaysia. Abdul-Rahim, R and Mohd. Nor, A.H.S. (2008). Evaluating the accuracy of Fama-French model versus liquidity based three factor models in forecasting portfolio returns. Labuan Bulletin of International Business and Finance. 6,p.77-107. Abdul Rahman, R and Mohamed Ali, F.H. (2006). Board, audit committee, culture and earnings management: Malaysian evidence. Managerial Auditing Journal. 21(7), p. 783-804 Aboody D., Hughes J. and Liu J. (2005) Earnings quality, insider trading and cost of capital. Journal of Accounting Research. 43(5),p.651-673 Agrawal A. and Knoeber C. R. (1996). Firm performance and mechanisms to control agency related problems between managers and shareholders. Journal of Financial and Quantitative Analysis.31(3),p.377-397. Ali, A. and Zarowin, P. (1992) Permanent versus transitory components of annual earnings and estimation error in earnings response coefficients. Journal of Accounting and Economics.15, p.249–264 Ball, R., Kothari, S.P., and Robin, A. (2000). The effect of institutional factors on properties of accounting earnings. Journal of Accounting and Economics.29, p.1-51 Basu, S. (1997). The conservatism principle and the asymmetric timeliness of earnings. Journal of Accounting and Economics.24, p.3-37 Baginski, S., Lorek, K., Willinger G.L. and Branson, B..(1999). The relationship between economic characteristics and alternative annual earnings persistence measures. The Accounting Review .74,p.104-120 Berle, A., and Means, G., (1932). The modern corporation and private property. New York:Harcourt, Brace & World. Botosan, C.A. (1997). Disclosure level and the cost of equity capital. The Accounting Review.72(3),p.323-349
201
Botosan,C.A. and Plumlee,M.A. (2001).A Re-examination of disclosure level and the expected cost of equity capital. Journal of Accounting Research.40(1),p.21-40. Botosan,C.A. and Plumlee,M.A. (2004). Assessing the construct validity of alternatives proxies for expected cost of equity capital. Working paper Bushman, R., Chen, Q., Engel, E. and Smith, A. (2004). Financial accounting information, organizational complexity and corporate governance systems. Journal of Accounting and Economics.37,p.167-201. Chen K.C.W.,Chen Z. and Wei K.C.J.(2003). Disclosure, corporate governance and the cost of equity : Evidence from Asia’s emerging markets. Working Paper. Chung R., Firth M. and Kim J. B. (2004). Earnings management, surplus free cash flow and external monitoring. Journal of Business Research.58(6),p.766-776 Chung R., Ho S. and Kim J. B.(2004). Ownership structure and the pricing of discretionary accruals in Japan. Journal of International Accounting, Auditing and Taxation.13,p.1-20. Claessens,S., Djankov, S., and Lang,L.H.P.(2000). The separation of ownership and control in East Asian corporations. Journal of Financial Economics.58, p.81-112 Claessens,S., Djankov, S., Fan, J.P.H. and Lang,L.H.P. (1998a). Corporate diversification in East Asia: The role of ultimate ownership and group affiliation. World Bank Working Paper. Claessens,S., Djankov, S., Fan, J.P.H. and Lang,L.H.P.(1998b). Expropriation of Minority Shareholders : Evidence from East Asia. World Bank Working Paper. Companies Act 1965 Collins D. W. and DeAngelo L.(1990). Accounting information and corporate governance. Market and analyst reactions to earnings of firms engaged in proxy contest. Journal of Accounting and Economics.13(3),p.213-247. Collins D.W., and Kothari, S.P. (1989) An analysis of intertemporal and cross-sectional determinants of earnings response coefficients. Journal of Accounting and Economics.11, p. 143–181 Copeland T. E. and Weston J. F.(1988). Financial Theory and Corporate Policy.United States of America:Addison-Wesley Publishing Company. DeAngelo (1981). Auditor size and audit quality. Journal of Accounting and Economics. 3,p.183-199
202
Dechow,P.M. and Dichev,I.D. (2002). The quality of accruals and earnings: The role of accruals estimation errors. The Accounting Review.77,p.35-59. Dechow,P.M, Sloan R.G. and Sweeney, A.P. (1995). Detecting earnings management. The Accounting Review.70(2),p.193-225. Demsetz, H.(1983). The structure of ownership and the theory of the firm. Journal of Law and Economics. 26,p.375-390. Demsetz, H. (1985). The structure of corporate ownership : causes and consequences. Journal of Political Economy. 93,p.1155-1177. Demsetz, H. and Villalonga, B. (2001).Ownership structure and corporate performance. Journal of Corporate Finance.7,p.209-233. Easly, D. and O’Hara, M. (2001). Information and the cost of capital. Cornel University Working Papers. Easton P.D. (2004). PE ratios, PEG ratios, and estimating the implied expected rate of return on equity capital. The Accounting Review. 79(1),p.73-95. Edwards E. and Bell P. (1961). The theory and measurement of business income.Berkeley, CA:University of California Press Elton, E.J. (1999). Expected returns, realized returns and asset pricing test. Journal of Finance.4,p.1199-1220 Fan, J.P.H and Wong, T.J., (2002).Corporate ownership structure and the informativeness of accounting earnings in East Asia. Journal of Accounting and Economics.33,p.401-425. Feltham G. and Ohlson J. (1995). Valuation and clean surplus accounting for operating and financial activities. Contemporary Accounting Research. Spring,p.689-731 Francis,J., LaFond,R., Olsson, P. and Schipper, K. (2003). Earnings quality and the pricing effects of earnings patterns. Working Paper. April. Francis,J., LaFond,R., Olsson, P. and Schipper, K. (2004). Costs of equity and earnings attribute. The Accounting Review.79(4),p.967-1010. Francis, J., D. J. Nanda, and P. Olsson (2008b), ‘Voluntary disclosure, earnings quality, and cost of capital’. Journal of Accounting Research.46,p 53-99
203
Francis,J., Schipper, K. and Vincent,L.(2005). Earnings and dividend informativeness when cash flow rights are separated from voting rights. Journal of Accounting and Economics.39,p.329-360 Gebhardt, W.R., Lee, C.M. and Swaminathan,B.(2001).Toward an implied cost of capital.Journal of Accounting Research.39,p.135-175. Gomez and Jomo (1999). Malaysia’s political economy: politics, patronage and profits. United Kingdom:Cambridge University Press,. Gomez, E.T. (2002). Political business in Malaysia: party factionalism, corporate development, and economic crisis in Political business in Asia (Gomez, E.T. ed), London :Routledge,. Grossman,S. and Hart, O.(1988). One share/one vote and the market for corporate control. Journal of Financial Economics.20,p.175-202 Gujarati, D.N. (1995). Basic econometrics. McGraw-Hill International Editions. Hand J. R. M. (1990). A test of the extended functional fixation hypothesis. The Accounting Review. 65(4), p.740-763 Harris, M. and Raviv, A.(1988). Corporate governance: voting rights and majority rules. Journal of Financial Economics.20,p.203-235 Hermalin, B. and Weisbach, M.,(2003). The determinants of board composition. RAND Journal of Economics.19,p.589-606. Jensen, M.C. and Mekling, W.H. (1976).Theory of the firm: managerial behavior, agency costs and ownership structure. Journal of Financial Economic.3,p.305-360 Jones J. (1991). Earnings management during import relief investigations. Journal of Accounting Research.29,p.193-228. Jung, K. and Kwon, S. Y., (2002). Ownership structure and earnings informativeness: evidence from Korea. The International Journal of Accounting.37, p.301-325. Kaplan, S. N. and Minton, B. A.(1994). Outside activity in Japanese companies: determinants and managerial implications. Journal of Financial Economics.36, p.225-258. Kapopoulos, P. and Lazaretou, S. (2007). Corporate ownership structure and firm performance : evidence from Greek firms. Corporate Governance.15(2),p. 144-158.
204
Khazanah Nasional official homepage [Online ]. Available on World Wide Web: (http://www.khazanah.com.my/) Kinney, W.R., Palmrose, Z. and Scholz, Z. (2004). Auditor independence, non-audit services and restatements: Was the US government right? Journal of Accounting Research. 42,p.561 Klein A. (2002). Audit committee, board of director characteristics and earnings management. Journal of Accounting and Economics. 33,p.375-400. Koh, P.S. (2003). On the association between institutional ownership and aggressive corporate earnings management in Australia. The British Accounting Review. 35,p.105-128. Leuz, C. and Verrechia, R.E. (2005). Firm’s capital allocation choices, information quality, and the cost of capital. Working Paper Leuz, C.,Nanda, D. and Wysocki, P. D.(2003). Earnings management and investor protection: an international comparison. Journal of Financial Economics.69,p.505-527. Lev, B. (1983) Some economics determinants of time series properties of earnings. Journal of Accounting and Economics.5,p.31-48 Lim Mah Hui (1981). Ownership and control of the one hundred largest corporations in Malaysia. Kuala Lumpur:Oxford University Press. Lipe R. (1990). The relation between stock returns and accounting earnings given alternative information.The Accounting Review.65(1),p.49 Liu J., Nissim D. and Thomas J. (2002). Equity valuation using multiples. Journal of Accounting Research. 40(1),p.135-172. Mak , Y. T. and Li, Y. (2001). Determinants of corporate ownership and board structure: evidence from Singapore. Journal of Corporate Finance. 7,p.235-256 McDaniel, L., Martin, R.D., and Maines, L.A. (2002). Evaluating financial reporting quality: The effects of financial expertise vs financial literacy. The Accounting Review. 77,p.139-167 McNichols M. and Wilson P.(1988). Evidence of earnings management from the provision for bad debts. Journal of Accounting Research.26,Supplement. Mock, R., Shleifer, A. and Vishny, R., (1988). Management ownership and market valuation: an empirical analysis. Journal of Financial Economics.20,p.293-315
205
Mohamed Ariff (2005). Banking on Corporate Governance [Online ]. Available on World Wide Web: http://www.mier.org.my/mierscan/ Mohd Saleh, N., Rahmat,M.M. and Mohd Iskandar,T. (2005). Earnings management and board characteristics: Evidence from Malaysia. Jurnal Pengurusan. 24,p.77-103 Mohd Saleh, N., Rahmat,M.M. and Mohd Iskandar,T. (2007). Audit committee characteristics and earnings management: evidence from Malaysia. Asian Review of Accounting.15(2),p.147-163 . Mohd Saleh, N (2003) Accounting policy choice by firms undergoing debt renegotiation. Phd Thesis, La Trobe University. Ohlson, J. A., (1995). Earnings, book values and dividends in equity valuation. Contemporary Accounting Research.11(2),p. 661-687. Pergola, T.M.(2005). Management Entrenchment : Can it negate the effectiveness of recently legislated governance reform? Journal of American Academy of Business. 6(2),p.177- 183 Penman,S. and Zhang, X-J.(2002). Accounting Conservatism, the quality of earnings and stock returns.The Accounting Review. 77,p.237-264 Pound, J. (1988). Proxy contest and the efficiency of shareholder oversight. Journal of Financial Economics. p.293-315. Pricewaterhouse Coopers (2000). Inside the mind of the CEO. Asia Pacific : A survey for the year 2000. World Economic Forum, Asia Pacific Economic Summit 2000, Melbourne, Australia. Prowse S.(1998). Corporate Governance: Emerging issues and lessons from East Asia. World Bank. Rajan R. and Zingales L (1998). Which capitalism? Lessons from the East Asian crisis. Journal of Applied Corporate Finance Richardson, S. 2003. Earnings quality and short sellers. Accounting Horizons. Supplement. Pg 49. Schipper K. and Vincent L. (2003). Earnings Quality. Accounting Horizons.17,p.97-110. Securities Commission (1999).Disclosure based Regulation- What directors need to know [Online].Available on World Wide Web: (http://www.sc.com.my /ENG/HTML/resources/inhouse/dbrbi.pdf).
206
Securities Commission Annual Report (2002) [Online ]. Available on World Wide Web: (http://www.sc.com.my/) Shleifer, A. and Vishny, R., (1986). Large shareholders and corporate control. Journal of Political Economy. 94,p.461-488. Shleifer, A. and Vishny, R., (1997). A survey of corporate governance. Journal of Finance. 52,p.737-783 Tabachnick. B. G. and Fidell L. S.(2001). Using Multivariate Statistics. Allyn and Bacon Tinic S. M. (1990). A perspective on the stock market’s fixation on accounting numbers. The Accounting Review. 65(4),p.781-796 Warfield T. D., Wild J. J., and Wild K. L.(1995). Managerial ownership, accounting choices and informativeness of earnings. Journal of Accounting and Economics. 20,p.61-91. Watts, R. L. and Zimmerman, J.L.(1986). Positive Accounting Theory. New Jersey, United States of America:Prentice-Hall.
207