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 TABLE OF CONTENTS LIST OF TABLES LIST OF FIGURES LIST OF APPENDICES ABSTRAK ABSTRACT CHAPTER 1 INTRODUCTION 1.1 Motivation of research 1.1.1 1.1.2 1.1.3 1.1.4 1.1.5 High standard of corporate governance (CG) and quality of information Standard of corporate governance and quality of information in substance Substantial shareholders Earnings as a useful measure A discerning market as enforcement agent 1 3 4 7 10 11 12 14 15 15 ii iii vii x x xi xiii

1.2. Problem statement 1.3. Research questions 1.4. Research objectives 1.5. Significance of study 1.5.1 Practical contributions 1.5.1.1 To the regulators 1.5.1.2 To market players Methodical and theoretical contributions

1.5.2

16 16 17 18

1.6 Thesis outline

CHAPTER 2 LITERATURE REVIEW 2.0. Introduction 2.1. Earnings quality 2.1.1 Accounting based earnings attributes 2.1.2 Market based earnings attributes 2.1.3 The relevant earnings quality constructs 21 22 23 26 27

iii

2.2. Ownership structure, expropriation of non-controlling shareholders interest and earnings quality 2.2.1 Theoretical studies 2.2.2 Empirical studies 2.2.3 Consideration of the types of ultimate controlling party 2.2.4 Consideration of monitoring mechanisms board structure, substantial shareholders and audit committee 2.2.5 Justification for audit committee (AC) measurements 2.3. Earnings quality, information risks and market required return or assessment 2.3.1 Theoretical studies 2.3.2 Empirical studies 2.4. Market assessment or consequences of information quality 2.5. Endogeneity of Ownership Structure 2.6 Summary CHAPTER 3 THEORETICAL FRAMEWORK AND HYPOTHESES DEVELOPMENT 3.0 Introduction 3.1 Relationship between ownership structure, monitoring mechanisms and earnings quality 3.2 Relationship between earnings quality and market assessment 3.3 Relationship between ownership structure, monitoring mechanisms and market assessment 3.4 Relationship that shows market assessment and the monitoring mechanisms could explain changes in ownership structure CHAPTER 4 RESEARCH METHODOLOGY 4.0 Introduction 4.1. Population and sample 4.2 Variable definition and Measurement 4.2.1 Earnings quality constructs 4.2.1.1 Accruals quality 4.2.1.2 Time series properties- persistence and predictability 4.2.1.3 The non-discretionary determinants of earnings quality 59 59 60 49 50 29 31 35 36 38

40 41 43 46 46

53 55

58

62 66 68

iv

4.2.2 Ownership structure 4.2.2.1 Calculation of ownership structure variable 4.2.2.2 Type of ultimate controlling party 4.2.3 Audit committee characteristics 4.2.3.1 Independence 4.2.3.2 Competence 4.2.4 Substantial shareholders 4.2.5 Measures for the market assessment of earnings quality 4.2.5.1 Cost of equity 4.2.5.2 Market return 4.2.6 The validity of the cost of equity measures 4.3 Data Analysis 4.3.1 Theoretical/research framework 4.3.2 The simultaneity of equations 4.3.3 The choice of audit firms CHAPTER 5 RESULTS 5.1 Data sources 5.2 Sample profile 5.3 General descriptive statistics 5.4 Descriptive statistics for ownership structure 5.5 Descriptive statistics for monitoring mechanisms substantial shareholders voting rights and audit committee characteristics

69 70 74 76 77 77 78 84 85

86 90 93

94 94 97 102 104

5.6 Bivariate collinearity between EQ variables 5.7 Bivariate collinearity analysis of all variables in each sample 5.8 Construct validity of variables- COE/COEA , EQ and CFVR 5.8.1 Cost of equity 5.8.2 Earnings quality 5.8.3 Cash flow/voting rights 5.9. Multivariate analysis 5.9.1 Ownership structure and earnings quality 5.9.2 Reestimation of equation 1 5.9.3 Earnings quality and cost of equity 5.9.5 Simultaneity test for equations 3 and 4 5.9.6 Ownership structure, monitoring mechanisms and market assessment

105 106

112 112 113

115 124 131 144 156

5.9.7 The relationship that examines whether market assessment and the monitoring mechanisms could explain changes in ownership structure 5.9.8 Relationship between substantial shareholders voting rights and elements of ownership, monitoring mechanisms and cost of equity 5.9.9 Two stage least square of equations 3 and 4 5.9.10 Comparisons of the ordinary least square results and two stage least square results of equation 3 and 4 CHAPTER 6 DISCUSSION 6.0 Introduction- main findings 6.1 The ownership structure and earnings quality 6.1.1 Cash flow/voting rights and earnings quality 6.1.2 Ultimate controlling party and earnings quality 6.1.3 Monitoring mechanisms and earnings quality 6.1.3.1. Substantial shareholders voting rights and earnings quality 6.1.3.2. Audit committee and earnings quality 6.2 Earnings quality and cost of equity 6.3 Cost of equity and market return 6.4 The relationship between ownership structure and cost of equity 6.4.1 Cash flow/voting rights and cost of equity 6.4.2 Ultimate controlling party and cost of equity 6.4.3 Monitoring mechanisms and cost of equity 6.4.3.1. Substantial shareholders voting rights and cost of equity earnings quality 6.4.3.2. Audit committee and cost of equity earnings quality 6.5 The relationship that examines whether market assessment and the monitoring mechanisms could explain changes in ownership structure

156 159 164 164

172

174 178 181 183 184 185

186 187

188 190

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 7.1 Conclusion and contribution highlights 7.2 Implication 7.3 Limitations of study 193 193 197 198

vi

7.4 Future research REFERENCES APPENDICES LIST OF TABLES No. 1.1 1.2 1.3 4.1 4.2 4.3 5.1 5.2 5.3 5.4 5.5 5.6 5.7 5.8 5.9 5.10 5.11 5.12 5.13 5.14 5.15 5.16 5.17 5.18 5.19 5.20 5.20 (a) Title ACP Industries Berhad- Analysis of Shareholdings as at 16 August 2004 Glomac Berhad- Analysis of Shareholdings as at 30 June 2004 Summary of motivation of study which leads to problem statement, research questions and objectives Variables brief description MTD Capital Analysis of Shareholding Calculation of cost of equity (COE) for ACP INDUSTRIES Sample Classification by industry Sample size based on available data for the calculation of earnings quality variables Breakdown of companies with various types of ownership Breakdown of companies with pyramidal(PYS) and non-pyramidal ownership(NON-PYS) Mean and dispersion of common variables in the three sample Descriptive Statistics - ABRES SAMPLE Descriptive Statistics - ABSDATCA & ABSDATA SAMPLE Descriptive Statistics for transformed ABSDATCA and ABSDATA i.e LABSCA and LABSTA Descriptive Statistics - PERS and PRED SAMPLE Descriptive statistics of cash flow and voting rights and ratio of cash flow to voting rights The descriptive statistics of substantial shareholders voting rights Pearson correlation coefficients between variables Correlations - ABRES SAMPLE Correlations- ABSDATCA and ABSDATA Sample Correlations- PERS PRED Sample Correlations between CF, VR and LGMV in full sample Correlations between CF, VR and LGMV among PYS companies Correlations between CF, VR and LGMV among NONPYS companies Coefficients of equation 1 regression Results of equation 1 regression for earnings quality

199 201

Page 7 8 20 60 76 80 95 95 96 96 96 98 99 100 100 101 103 104 106 109 110 111 114 114 115 118 119

Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table

vii

ABRES Table 5.20 (b) Results of equation 1 regression for earnings quality ABSDATA Results of equation 1 regression for earnings quality ABSDATCA Results of equation 1 regression for earnings quality PERS Results of equation 1 regression for earnings quality PRED Pyramidal companies in ABRES sample Non-pyramidal companies in ABRES sample Pyramidal companies in ABSDATA sample Non-pyramidal companies in ABSDATA sample Pyramidal companies in ABSDATCA sample Non-pyramidal companies in ABSDATCA sample Pyramidal companies in PERS sample Non-pyramidal companies in PERS sample Pyramidal companies in PRED sample Non-pyramidal companies in PRED sample Results of equation 2 regression using COE estimate Results of equation 2 regression using COEA estimate Results of equation 2 regression - ABRES and COE estimate Results of equation 2 regression - ABRES and COEA estimate Results of equation 2 regression- ABSDATA and COE estimate Results of equation 2 regression- ABSDATA and COEA estimate Results of equation 2 regression - ABSDATCA and COE estimate Results of equation 2 regression - ABSDATCA and COEA estimate Results of equation 2 regression - PERS and estimate COE 138 138 120

Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table

5.20 (c) 5.20 (d) 5.20 (e) 5.20 (f) 5.20 (g) 5.20 (h) 5.20 (i) 5.20 (j) 5.20 (k) 5.20 (l) 5.20 (m) 5.20 (n) 5.20 (o) 5.21 (I) 5.21 (II) 5.21 (a) (i) 5.21 (a) (ii) 5.21 (b) (i) 5.21 (b) (ii) 5.21 (c) (i) 5.21 (c) (ii) 5.21 (d) (i) 5.21 (d) (ii) 5.21 (e) (i)

121 122 123 126 127 127 128 128 129 129 130 130 131 133 134 135 135 136 136 137 137

Results of equation 2 regression - PERS and COEA estimate Results of equation 2 regression - PRED and COE

viii

estimate Table 5.21 (e) (ii) Results of equation 2 regression - PRED and COEA estimate Results of testing earnings quality (ABRES) and excess return Results of testing earnings quality (ABSDATCA) and excess return Results of testing earnings quality (ABSDATA) and excess return Results of testing earnings quality (PERS) and excess return Results of testing earnings quality (PRED) and excess return Pearson Correlation of cost of equity and return ABRES sample- Results of estimating equation to test the exogenuity of CFVR and SSVR (with COE) Testing for coefficients of ^CFVR and ^SSVR ABRES sample- Results of estimating equation to test the exogenuity of CFVR and SSVR (with COEA) Testing for coefficients of ^CFVR and ^SSVR ABSDATA/ABSDATCA sample- Results of estimating equation to test the exogeneity of CFVR and SSVR (with COE) Testing for coefficients of ^CFVR and ^SSVR ABSDATA/ABSDATCA sample- Results of estimating equation to test the exogeneity of CFVR and SSVR (with COEA) Testing for coefficients of ^CFVR and ^SSVR PERS/PRED sample- Results of estimating equation to test the exogenuity of CFVR and SSVR (with COE) Testing for coefficients of ^CFVR and ^SSVR PERS/PRED sample- Results of estimating equation to test the exogenuity of CFVR and SSVR (with COEA) Testing for coefficients of ^CFVR and ^SSVR Results of equation 3 regression using COE estimate

139 139 141 141 142 142 143 143 146

Table Table Table Table Table Table Table

5.22(a) 5.22(b) 5.22(c) 5.22(d) 5.22(e) 5.23 5.23 (a)

Table Table

5.23 (b) 5.23 (c)

146 147

Table Table

5.23 (d) 5.24 (a)

147 148

Table Table

5.24 (b) 5.24 (c)

148

Table Table Table Table

5.24 (d) 5.25 (a) 5.25 (b) 5.25 (c)

149 149 150 150

Table Table Table

5.25 (d) 5.26 (a) 5.26(b)

151 151

154 Results of equation 3 regression using COEA estimate

ix

155 Table Table Table Table Table Table Table Table Table Table Table Table Table Table 5.27 (a) 5.27 (b) 5.28 (a) (i) 5.28 (a) (i) 5.28 (b) (i) 5.28 (b) (ii) 5.28 (c) (i) 5.28 (c) (ii) 5.29 (a) 5.29 (b) 5.30(a) 5.30(b) 5.4 6.1 Results of equation 4 regression using COE estimates Results of equation 4 regression using COEA estimates ABRES Sample (with COE) ABRES Sample (with COEA) ABSDATA Sample (with COE) ABSDATA Sample (with COEA) PERS/PRED Sample (with COE) PERS/PRED Sample (with COEA) ABRES Sample ABSDATA/ABSDATCA Sample ABRES sample ABSDATA/ABSDATCA sample Summary of results Percentage of companies whose controlling party is also in an executive position 157 158 161 161 162 162 163 163 166 166 167 167 168 179

LIST OF FIGURES No. 1.1 3.1 Title Market-based regulatory environment Theoretical Framework Page 2 49 86

Figure Figure Figure

4.3.1 Theoretical/ research framework

LIST OF APPENDICES

Appendix 1 Appendix 2 Appendix 3 Appendix 4 Appendix 5

Appendix 6

Annual reports of APM Automotive Holdings Bhd Annual reports of ACP Industries Berhad and other related companies List of companies in sample SPSS Output (Equation 1) Eviews output (Equation 1 regression separating companies that are controlled by ultimate controlling party and those that are not) Eviews output (Equation 1 regression for family controlled companies separating those are family managed and those that are not)

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 selfregulation 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 donts. It is essentially an infrastructure of built-in checks and balances. Good

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

High Quality Information

Market Players Assessment

Companies

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.

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-shareone-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.

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

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

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.

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

Shareholders Metacorp Berhad MTD Capital Berhad Lambang Simfoni Sdn Bhd Employees Provident Fund Board

Direct Interest Number of Shares % 38,734,790 29.02

Indirect Interest Number of Shares % 38,734,790 38,734,790 29.02 29.02

20,499,000

15.36

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

Shareholders Dato Mohamed Mansor Fateh Din 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)

Direct Interest Number of Shares % 63,552,183 29.33

Indirect Interest Number of Shares %

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 managers behavior as they have the ability to remove nonperforming 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.

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

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 companys 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 noncontrolling 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

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.

19

Table 1.3 Summary of motivation of study which leads to problem statement, research questions and objectives
MOTIVATION OF RESEARCH 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 shareholders 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. PROBLEM STATEMENT 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) RESEARCH QUESTIONS 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) RESEARCH OBJECTIVES 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 controllingnon-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 companys 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 companys core operations. Earnings that is low in persistence could be interpreted as of low quality in the sense that a significant part of a companys 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 partys discretion. As such this research will control for those non manipulative factors namely a companys 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/ noncontrolling 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 partys 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 directorss 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 earnings management. whatever dependent variable is in focus such as

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 auditors competence, which has been translated as audit firm size, and the second depends on the auditors 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 partys 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 noncompliance 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 companys 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 OHara (2001) and Leuz and Verrechia (2005) establish a theoretical link between information risk and companies cost of capital counter to CAPM analysis. Easley and OHara (2001) show that the composition of public and private information could influence a companys 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

companys share is equal to the sum of expected dividends discounted at the companys cost of equity. Botosan (1997) found a negative association between disclosure level and cost of equity, after controlling for market risk (beta) and companys 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 Asias 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 noncontrolling 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 & OHara 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

Monitoring elements i.e AC and SS

EQ
1 3 2

OS
4

MA

Figure 3.1 Theoretical Framework EQ MA OS AC SS Earnings quality Market assessment Ownership structure Audit committee 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 & OHara 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 OHara (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 OHara (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 OHara (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 Earnings Quality (EQ): 1. Accrual quality Mapping cash flows 2. Discretionary current accruals 3. Discretionary total accruals ABRES Measurement Absolute residual from the regression of changes in working capital and past, current and future cash flows (Dechow & Dichev 2002)

ABSDATCA Modified Jones (1991) model ABSDATA

60

4.Persistence 5. Predictability

PERS PRED

Slope coefficient of earnings time series model. The time series model is the regression of earnings per share on lagged earnings per share. 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. 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)

Audit committee (AC): 1. Independence 2. Competence ACI ACC

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 CF 2. Voting rights 3.Cash flow voting/controlling rights disparity 2. Ultimate controlling party VR CFVR Ratio of cash flow to voting/controlling rights

UCP

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. Residual income model Alternative estimation

Market assessment: Cost of equity : COE Ex ante measure of expected return COEA

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 Log market value Log book to market value Beta from Datastream Net book value of property, plant and equipment to total assets Log of the sum of a companys days accounts receivable and days inventory

Size Growth Risk

Control variables: LGMV LBTMV CAPINT OC STDREV

Capital Intensity Operating cycle Standard deviation of revenue

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 REVt//A t-1 + cPPEt /A t-1+ 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

REVt -Change in revenues PPEt -Plant,property and equipment at t

2. Then the estimated coefficients a,b and c is substituted into the following equations coefficients (, and ) to obtain non-discretionary accruals for each company. NDAt/ A t-1= Where, NDAt RECt : 3. Then, the discretionary accruals, DA = TAt /A t-1 - NDAt/ A t-1 Non-discretionary accruals at time t Change in receivables (1/A t-1) + ( REVt - RECt)/ A t-1 + PPEt/ 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 Where, NCAt -

(1/A t-1) + ( REVt - RECt)/ A t-1

Non-discretionary current accruals at time t

The coefficients and are estimated from coefficients a and b of the following crosssectional regression by industry :

TCAt /A t-1= a (1/A t-1) + b REVt//A t-1 + 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 + t Where, CFO cash flow

= net income before extraordinary item TA (total accruals) Avassetaverage 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 companys 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 companys 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 = 0,j + 1,j E j,t-1 + j,t

Where, Ej,t - earnings of firm j at time t

Persistence is the slope coefficient estimate 1,j. Earnings is of higher quality the higher the value of
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 the acronym PERS.
1,j

are negated. Persistence is assigned

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 (i2/n) is biased whilst the ordinary least square standard error estimate (i2/(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?
Family 11% A 21%

Ownership, cash flow rights = 11% x 21% = 2% Control, voting rights = 11%, the weakest link.

Example 2.
B

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% A 21% 7% B C 25%

Cash flow rights = 11% x 21% + 25% x 7% = 4% Voting rights = 11% + 7% = 18%

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 shareholders 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 shareholders 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. Cash flow rights = 22.667% + 100% (20.0248%) = 42.6918% Voting rights = 22.667% + 20.0248% = 42.6918%

Tan Chong Consolidated Sdn Bhd 100% 22.667% Parasand Ltd 20.0248% APM

ACP Industries Berhad The structure of ACP could be understood by also looking at Metacorp Berhad and MTD Capital Bhd shareholders 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

Nikvest Sdn Bhd 21.18% MTD Captial Berhad 100% Lambang Simfoni Sdn Bhd 74.17% Metacorp Berhad 29.02% ACP Industries Berhad

Alloy Consolidated Sdn Bhd 15% 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 Number of Shares % 58,444,940 21.18 43,271,750 15.68 71,104 .03 25,156,500 9.11 Indirect Interest Number of Shares % 8,406,066 3.0511 63,362,494 22.96 2 58,444,940 58,444,940 51,677,816 56,881,816 56,881,816 21.18 3 21.18 3 18.72 4 20.61 5 20.61 5

Shareholders Nikvest Sdn Bhd Alloy Consolidated Sdn Bhd Dato Dr Nik Hussain bin Abdul Rahman Nik Fauzi Dato Nik Hussein Nik Faizul Dato Nik Hussein Datin Nik Fuziah Dato Nik Hussein Ruslan Sulaiman Mohd Dom Ahmad Employees Provident Fund

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 childrens 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

Pt

Bt + [ (Et (ROEt+i ) COE ) B t+i-1 ] i=1 (1+COE) i

B

where, Pt Bt
B

= -

Share price at time t Book value at time t Return on equity at time t Cost of equity Expectation based on information at t Forecast earnings per share (FEPS t+i) / B t+i-1 Forecasted earnings per share

ROEt+I COE E(--) E(ROEt+i ) FEPS

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.

79

Table 4.3 Calculation of cost of equity (COE) for ACP INDUSTRIES

t+1 0.3 t+2 0.3 t+3 0.3 t+4 0.3 t+5 0.3 t+6 0.3 t+7 0.3 t+8 0.3 t+9 0.3 t+10 0.3 t+11 0.3 t+12 0.3

Div Payout ratio FEPS

0.026 1 2.79

0.1411 2.8082

0.76461 2.9069

0.8433 5 3.4421

0.916 4.0328

0.978 4.6743

1.026 5.3592

1.055 6.0773

1.061 6.8155

1.041 7.5580

0.993 8.2865

0.915 8.9815

Beginning Book value 2 Return on Equity (ROE) COE 8 Book value 2.79

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

0.1376

0.1376

0.1376

0.1376

0.1376 0.1897

0.1376 0.1547

0.1376 0.1169

0.1376 0.0778

0.1376 0.0384

0.1376 0.00023

0.1376 -0.03571

0.1376 -0.06826 -0.49605 7 Share Price 2.73

Streams residual income discounted at COE 0-.3146 6 -0.1896 0.2476 0.2210

The publicly available data required are the dividend payout ratio (which is either a companys 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 for example for 2005 , B = B t + EPS (1-Payout ratio) = 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 companys return will approach the industrys 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

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.

82

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 Pt FEPS t+1 Where, A = ( g + POUT x FEPS t+1 Pt Pt FEPS POUT g Share price at time t )

- g ) ) 1/2

Forecasted earnings per share Dividend payout ratio 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 = +(R m,t R f,t) + 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 crosssectional 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, , 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 Monitoring elements i.e AC and SS

EQ
3 3 2

OS
4

MA

Figure 4.3.1 Theoretical/ research framework

Separation of control and ownership, monitoring mechanisms and earnings quality

Earnings quality

1,2,3,4,5

= 0 + i CFVR + 2,3,4,5 UCP + 6 SSVR + 7 ACI + 8 ACC + 9SIZE + 10Controls for EQ + 1,2,3,4,5 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 i 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 2,3,4,5 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 6 is expected to be either positive or negative. The coefficients for audit committee independence and competence, 7 and 8 , 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 i 2,3,4,5 6 7 8 9 10 Sign Negative Positive Positive/Negative Negative Negative Negative Negative for CAPINT, and Positive for OC and STDREV

Earnings quality and cost of equity

COE

= 0 + 1,2,3,4Earnings quality 1,2,3,4 + 5 SIZE + 6 + 7 BTMV + 6 Equation 2

Prediction: Coefficients 1,2,3,4 5 6 7 Sign Positive Negative Positive Positive

Refer to section 3.2 for detail explanation of how the relationship is derived. The coefficients of each earnings quality measure 1,2,3,4 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 5 is expected to be negative, and positively related to beta and book to market, and thus 6 and 7 are expected to be positive.

Relationship between ownership structure, monitoring mechanisms and cost of equity COE = 0 + 1 CFVR + 2,3,4,5 UCP + 6 SSVR + 7 ACI + 8 ACC + 9SIZE + 10BTMV + 11 + 7,8,9 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 1 2,3,4,5 6 7 8 9 10 11 Sign Negative Positive Positive/Negative Negative Negative Negative Positive Positive

Relationship between ownership structure, monitoring mechanisms and cost of equity

CFVR

0 + 1COE + 2,3,4,5UCP + 6 SSVR + 7 ACI + 8 ACC +

9SIZE + 10BTMV + 10,11,12 Equation 4 88

Refer to section 3.4 for detail explanation of how the relationship is derived. The coefficient 1 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 2,3,4,5 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 6 is expected to be positive or negative. The coefficient for size 9 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 1 2,3,4,5 6 7 8 9 10 Sign Negative Positive Positive/Negative Negative Negative Positive Negative

Relationship between substantial shareholders voting rights and ownership, monitoring mechanisms and cost of equity

elements of

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 =

0 + 1COE + 2,3,4,5UCP + 6 CFVR + 7 ACI + 8 ACC + 9SIZE + 10BTMV + 10,11,12

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 + 1

Reduced form equation 2 : SSVR = b0 + b1,2,3,4UCP + b5 ACI + b6 ACC + b7SIZE + b8BTMV + 2

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

0 + 1 CFVR + 2,3,4,5 UCP + 6 SSVR + 7 ACI + 8 ACC + 9SIZE + 10BTMV + c1 ^CFVR + c2 ^SSVR + 7,8,9

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 + 3

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

0 + 1 ^CFVR + 2,3,4,5 UCP + 6 ^SSVR + 7 ACI + 8 ACC + 9SIZE + 10BTMV + 11 + 7,8,9

Relationship between ownership structure and elements of ownership, monitoring mechanisms and cost of equity

CFVR

0 + 1^COE + 2,3,4,5UCP + 6 ^SSVR + 7 ACI + 8 ACC + 9SIZE + 10BTMV + 10,11,12

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

0 + 1^COE + 2,3,4,5UCP + 6 ^CFVR + 7 ACI + 8 ACC + 9SIZE + 10BTMV + 10,11,12

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 IBESs 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 Eliminated due to change in accounting year end and insufficient data to calculate cost of equity 213 10 203

Table 5.2 Classification by industry

Industry Construction Consumer product Finance Hotel Industrial product Infrastructure project companies Plantations Properties Technology Trade and services 11 26 13 3 55 5 7 23 10 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 ABRES ABSDATCA /ABSDATA PERS /PRED 141 151 118 (%)Share of Market Capitalization 31 32 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 ABRES (141) ABSDATCA /ABSDATA (151) PERS/PRED (118) Managerial 16 (11%) 16 ((11%) 12 (10%) Institutional 2 (1%) 2 (1%) 1 (0.8%) Governmental 25 (19%) 24 (16%) 24 (20) Family 96 ((68%) 106 (70%) 79 (68%) Company 2 (1%) 3 (2%) 2 (1.2%)

Table 5.5 Breakdown of companies with pyramidal(PYS) and non-pyramidal ownership(NON-PYS)

Sample ABRES (141) ABSDATCA /ABSDATA (151) PERS/PRED (118) PYS 40 (28%) 46 (30%) 47 (40%) NON-PYS 101 (72%) 105 (70%) 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

Table 5.6 Mean and dispersion of common variables in the three sample
ABRES Mean 0.490 0.087 6.585 0.961 -0.394 0.634 0.339 0.147 0.099 Std. Deviation 0.218 0.077 1.377 0.413 0.645 0.200 0.165 0.066 0.034 ABSDATCA & ABSDATA Mean 0.488 0.091 6.555 0.932 -0.406 0.636 0.345 0.149 0.099 Std. Deviation 0.216 0.080 1.338 0.475 0.641 0.196 0.170 0.065 0.033 PERS and PRED Std. Mean Deviation 0.506 0.225 0.087 0.084 6.955 1.401 1.008 0.483 -0.283 0.633 0.608 0.221 0.341 0.197 0.145 0.070 0.094 0.037

CFVR SSVR LGMV BETA LBTMV ACI ACC COEA COE

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 0.43 0.45 0.71 -0.52

ABSDATCA 0.090/0.211 ABSDATA PRED PERS 0.094/.208 0.171/0.242

-0.215/0.414 =

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.

98

Table 5.7 Descriptive Statistics - ABRES SAMPLE

N Minimum Maximum Mean Std. Deviation Skewness Statistic 0.515 0.814 1.682 0.582 0.520 -0.559 -1.036 0.317 2.190 1.150 Std. Error 0.204 0.204 0.204 0.204 0.204 0.204 0.204 0.204 0.204 0.204 Kurtosis Statistic 0.155 0.928 3.095 -0.221 -0.029 1.910 1.888 0.438 8.333 4.635 Std. Error 0.406 0.406 0.406 0.406 0.406 0.406 0.406 0.406 0.406 0.406

CFVR SSVR ABRES LGMV BETA LBTMV ACI ACC COEA COE

141 141 141 141 141 141 141 141 141 141

0.006 0.000 0.001 4.222 0.160 -3.013 0.000 0.000 0.021 0.014

1.000 0.401 0.209 10.300 2.100 1.393 1.000 0.750 0.520 0.263

0.490 0.087 0.039 6.585 0.961 -0.394 0.634 0.339 0.147 0.099

0.218 0.077 0.039 1.377 0.413 0.645 0.200 0.165 0.066 0.034

2.522 3.990 8.241 2.849 2.546 -2.738 -5.077 1.555 10.729 5.632

0.382 2.289 7.633 -0.544 -0.070 4.709 4.655 1.079 20.547 11.430

99

Table 5.8 Descriptive Statistics - ABSDATCA & ABSDATA SAMPLE

N Minimum Maximum Mean Std. Deviation Skewness Statistic 1.173 2.087 -1.083 0.321 0.609 0.749 0.656 -0.475 -0.524 8.863 8.850 Std. Error 0.197 0.197 0.197 0.197 0.197 0.197 0.197 0.197 0.197 0.197 0.197 Kurtosis Statistic 4.838 7.794 2.051 0.323 0.299 0.429 -0.040 3.558 1.762 92.330 92.375 Std. Error 0.392 0.392 0.392 0.392 0.392 0.392 0.392 0.392 0.392 0.392 0.392

COE COEA ACI ACC CFVR SSVR LGMV BETA LBTMV ABSDATCA ABSDATA

151 151 151 151 151 151 151 151 151 151 151

0.014 0.021 0.000 0.000 0.006 0.000 4.222 -1.460 -3.013 0.000 0.000

0.263 0.520 1.000 0.750 1.000 0.401 10.300 2.100 1.393 2.364 2.335

0.099 0.149 0.636 0.345 0.488 0.091 6.555 0.932 -0.406 0.090 0.094

0.033 0.065 0.196 0.170 0.216 0.080 1.338 0.475 0.641 0.211 0.208

5.945 10.570 -5.486 1.626 3.085 3.793 3.323 -2.406 -2.656 44.899 44.835

12.333 19.865 5.229 0.823 0.763 1.093 -0.101 9.069 4.492 235.343 235.460

Table 5.9 Descriptive Statistics for transformed ABSDATCA and ABSDATA i.e LABSCA and LABSTA
N Minimum Maximum Mean Std. Deviation 1.371 1.515 Skewness Statistic LABSCA LABSTA 151 151 -8.957 -9.053 0.860 0.848 -3.271 -3.224 -0.512 -1.152 Std. Error 0.197 0.197 Kurtosis Statistic -2.594 -5.838 1.420 2.486 Std. Error 0.392 0.392

3.620 6.336

100

Table 5.10 Descriptive Statistics - PERS and PRED SAMPLE

N Minimum Maximum Mean Std. Deviation Skewness Statistic 1.168 2.181 -0.887 0.784 0.454 0.856 3.321 -1.407 0.287 0.304 -0.944 Std. Error 0.223 0.223 0.223 0.223 0.223 0.223 0.223 0.223 0.223 0.223 0.223 Kurtosis Statistic 4.095 7.687 0.928 1.371 0.100 0.492 12.487 6.143 -0.433 -0.394 3.316 Std. Error 0.442 0.442 0.442 0.442 0.442 0.442 0.442 0.442 0.442 0.442 0.442

COE COEA ACI ACC CFVR SSVR PRED PERS BETA LGMV LBTMV

118 118 118 118 118 118 118 118 118 118 118

0.014 0.021 0.000 0.000 0.006 0.000 0.007 -2.439 -0.120 4.222 -3.013

0.263 0.520 1.000 1.000 1.000 0.401 1.429 0.608 2.100 10.456 1.393

0.094 0.145 0.608 0.341 0.506 0.087 0.171 -0.215 1.008 6.955 -0.283

0.037 0.070 0.221 0.197 0.225 0.084 0.242 0.414 0.483 1.401 0.633

5.244 9.793 -3.983 3.522 2.037 3.843 14.912 -6.319 1.287 1.367 -4.241

9.267 17.395 2.100 3.103 0.226 1.113 28.259 13.902 -0.979 -0.892 7.505

101

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 CF VR 0.41 0.54 0.01 0.20 0.89 1.00 0.19 0.23 0.28 0.38 0.41 0.50 0.54 0.61 ABSDATCA CF VR 0.40 0.53 0.01 0.20 0.89 1.00 0.18 0.22 0.28 0.36 0.41 0.48 0.53 0.60 PERS/PRED CF VR 0.39 0.53 0.01 0.20 0.89 1.00 0.19 0.22 0.24 0.37 0.40 0.48 0.54 0.60

Mean Minimum Maximum Std Dev Percentiles

25 50 75

CFVR 0.49 0.01 1.00 0.22 0.32 0.50 0.60

CFVR 0.49 0.01 1.00 0.22 0.32 0.48 0.60

CFVR 0.51 0.01 1.00 0.23 0.34 0.51 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 0.087 0.081 0.077 0.000 0.401 0.000 0.081 0.137 ABSDATCA/ ABSDATA 0.09 0.08 0.08 0.00 0.40 0.00 0.08 0.15 PERS/PRED 0.09 0.08 0.08 0.00 0.40 0.00 0.08 0.15

Mean Median Std. Deviation Minimum Maximum Percentiles: 25 50 75

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 ABSDATCA ABSDATA 1.00 0.22 1.00 (0.01)*** 0.19 0.99 1.00 (0.02)** (0.00)*** 0.24 0.08 0.07 (0.43) (0.45) (0.02)** -0.18 0.13 0.13 (0.19) (0.17) (0.07)* PRED PERS

1.00 0.04 (0.69) 1.00

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

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.

108

Table 5.14 Correlations - ABRES SAMPLE

COE COE COEA ACI ACC SSVR CFVR ABRES LGMV LBTMV BETA CAPINT OC 1.000 0.722 (0.000)*** 0.105 (0.215) 0.122 (0.148) -0.086 (0.308) 0.102 (0.227) 0.085 (0.315) -0.370 (0.000)*** 0.567 (0.000)*** 0.116 (0.170) 0.080 (0.348) 0.096 0.045 (0.597) 0.082 (0.335) -0.225 (0.007)*** 0.068 (0.424) 0.199 (0.018)** -0.330 (0.000)*** 0.289 (0.001)*** 0.053 (0.535) 0.070 (0.410) 0.126 0.043 (0.609) -0.018 (0.836) -0.141 (0.094)* 0.110 (0.192) -0.089 (0.294) 0.001 (0.990) 0.028 (0.745) -0.104 (0.218) 0.236 0.122 (0.149) -0.083 (0.328) -0.077 (0.367) -0.123 (0.147) 0.157 (0.064)* 0.127 (0.134) -0.041 (0.627) 0.087 (0.303) -0.214 (0.011)** -0.140 (0.098)* -0.070 (0.406) 0.106 (0.213) -0.036 (0.675) -0.024 (0.778) -0.015 (0.857) -0.087 (0.305) 0.143 (0.090)* 0.125 (0.140) -0.068 (0.421) 0.038 (0.658) -0.049 (0.566) -0.203 (0.016)** -0.188 (0.025)** -0.011 (0.897) -0.158 (0.062)* 0.124 (0.144) -0.269 (0.001)* -0.113 (0.182) 0.097 (0.253) -0.286 (0.001)*** 0.130 (0.123) 0.169 (0.045)** 0.141 (0.096)* -0.128 (0.130) 0.043 (0.613) -0.255 (0.002)*** 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 COEA ACI ACC SSVR CFVR ABRES LGMV LBTMV BETA CAPINT OC

(0.258) (0.137) (0.005)*** 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 COE COEA ACI ACC SSVR CFVR ABSDATCA ABSDATA LGMV LBTMV BETA CAPINT OC 1.000 0.714 (0.000)*** 0.095 (0.246) 0.116 (0.156) -0.091 (0.264) 0.082 (0.314) 0.029 (0.720) 0.011 (0.898) -0.364 (0.000) 0.598 (0.000)*** 0.075 (0.359) 0.062 (0.449) 0.096 0.045 (0.581) 0.067 (0.414) -0.197 (0.015)** 0.026 (0.752) 0.185 (0.023)** 0.181 (0.026)** -0.336 (0.000) 0.318 (0.000)*** 0.026 (0.751) 0.048 (0.554) 0.156 0.048 (0.561) -0.023 (0.779) -0.132 (0.106) 0.178 (0.028)** 0.229 (0.005)*** -0.094 (0.249) 0.007 (0.931) 0.015 (0.852) -0.096 (0.240) 0.240 0.136 (0.095)* -0.094 (0.253) 0.001 (0.988) -0.065 (0.427) -0.133 (0.105) 0.179 (0.028)** 0.077 (0.349) -0.071 (0.383) 0.111 (0.173) -0.205 (0.012)** -0.127 (0.119) -0.094 (0.249) -0.083 (0.312) 0.129 (0.115) -0.060 (0.466) -0.072 (0.378) 0.044 (0.590) -0.119 (0.145) -0.032 (0.695) 0.197 (0.015)** 0.083 (0.311) 0.007 (0.933) 0.080 (0.329) -0.104 (0.205) 0.732 (0.000)*** -0.284 (0.000)*** -0.150 (0.066)* -0.007 (0.930) -0.195 (0.017)** 0.367 (0.000)*** -0.287 (0.000)*** -0.129 (0.115) 0.089 (0.278) -0.216 (0.008)*** 0.293 (0.000)*** -0.282 (0.000)*** -0.034 (0.674) 0.117 (0.152) -0.332 (0.000)*** 0.136 (0.096)* 0.166 (0.041)** 0.190 (0.019)** -0.074 (0.365) 0.015 (0.858) -0.248 (0.002)*** 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 COEA ACI ACC SSVR CFVR ABSD ATCA ABSD ATA LG MV LBT MV BETA CAP INT OC

(0.239) (0.055)* (0.003)*** 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

Table 5.16 Correlations- PERS PRED Sample

COE COE COEA ACI ACC SSVR CFVR NEGPERS PRED LGMV LBTMV BETA STDREV 1.000 0.690 (0.000)*** 0.093 (0.314) 0.028 (0.764) -0.050 (0.594) 0.114 (0.218) -0.035 (0.709) 0.236 (0.010)*** -0.392 (0.000)*** 0.602 (0.000)*** 0.062 (0.505) 0.148 0.108 (0.244) 0.072 (0.439) -0.151 (0.100)* 0.030 (0.748) -0.059 (0.528) 0.274 (0.003)*** -0.355 (0.000)*** 0.348 (0.000)*** 0.049 (0.598) -0.037 0.128 (0.168) 0.081 (0.381) -0.242 (0.008)*** -0.008 (0.934) 0.009 (0.927) -0.119 (0.198) 0.043 (0.640) 0.072 (0.439) -0.113 0.027 (0.771) -0.133 (0.152) 0.037 (0.693) 0.039 (0.678) -0.134 (0.149) 0.150 (0.105) 0.218 (0.018)** -0.123 (0.227) -0.204 (0.027)** -0.127 (0.172) -0.012 (0.897) -0.102 (0.271) 0.113 (0.222) -0.111 (0.232) 0.072 (0.484) 0.079 (0.394) 0.058 (0.533) 0.067 (0.471) 0.105 (0.260) -0.044 (0.638) -0.049 (0.632) 0.059 (0.525) -0.092 (0.320) 0.011 (0.909) 0.161 (0.082)** 0.038 (0.710) -0.033 (0.722) 0.022 (0.813) 0.235 (0.010)*** 0.165 (0.100)* -0.383 (0.000)*** -0.112 (0.226) -0.028 (0.783) 0.172 (0.063)* -0.039 (0.700) -0.080 (0.435) 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 COEA ACI ACC SSVR CFVR NEGPERS PRED LGMV LBTMV BETA STDREV

(0.146) (0.714) (0.268) 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.

111

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 LGMV CF VR 0.16 (0.06)* 0.07 0.40 ABSDATA LGMV 0.16 (0.05)** -0.11 0.18 PERS LGMV 0.05 0.63 -0.12 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 LGMV -0.07 0.66 0.21 0.19 ABSDATA LGMV -0.12 0.47 -0.02 0.91 PERS LGMV -0.08 0.58 0.05 0.72

CF VR

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 LGMV 0.35 0.00 ABSDATA LGMV 0.36 0.00 PERS LGMV 0.27 0.02

CF

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 = 0 + i CFVR + 2,3,4,5 UCP + 6 SSVR + 7 ACI + 8 ACC + 9SIZE + 10 Controls for EQ + 1,2,3,4,5 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 Whites 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 = 0 + i CFVR + 2,3,4,5 UCP + 6 SSVR + 7 ACI + 8 ACC + 9SIZE + 10 Controls for EQ + 1,2,3,4,5
Predicted sign (Constant) UCPMn UCPInst UCPGov UCPFam SSVR ACI ACC CFVR LGMV CAPINT OC STDREV R2 F Sig N +ve +ve +ve +ve ? -ve -ve -ve -ve -ve +ve +ve 0.12 1.82 0.06* 141 0.21 3.40 0.00*** 141 0.26 4.37 0.00*** 151 -0.01 0.01 -0.01 0.00 0.09*** 0.01 -0.02 -0.01 -0.004* -0.03** 0.28 -0.81 -0.42 -0.13 -2.01 1.11* -0.81 0.52 -0.24** -1.07** 0.87* -0.36 -0.20 -1.18 -0.59 -0.18 -0.23 -0.05 -0.06 -0.04 -0.09 0.03 0.03 -0.28 0.04 0.02 0.08 0.00

ABRES

ABSDATA

ABSDATCA

PERS

PRED

-3.41*** 0.43 -0.15 -0.21 -0.14* -0.64 1.47***

-0.77** 0.06 0.11 -0.00 -0.05

0.17 0.06 0.58 0.82 151

0.29* 0.08 0.76 0.67 118

*** 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 = 0 + i CFVR + 2,3,4,5 UCP + 6 SSVR + 7 ACI + 8 ACC + 9SIZE + 10 Controls for EQ +
Dependent Variable: ABRES Included observations: 141 after adjustments White Heteroskedasticity-Consistent Standard Errors & Covariance Variable C UCPMN UCPINST UCPGOV UCPFAM SSVR ACI ACC CFVR LGMV CAPINT R-squared Adjusted R-squared S.E. of regression Coefficient 0.092640 -0.005935 0.010566 -0.011214 0.002469 -0.091919 0.012500 -0.020365 -0.014970 -0.004254 -0.025294 0.122298 0.054782 0.037535 Std. Error 0.024308 0.013195 0.036525 0.009439 0.008806 0.036736 0.014962 0.021794 0.015212 0.002306 0.012671 F-statistic Prob(F-statistic) t-Statistic 3.811065 -0.449809 0.289275 -1.188127 0.280388 -2.502114 0.835433 -0.934421 -0.984110 -1.844967 -1.996184 Prob. 0.0002 0.6536 0.7728 0.2369 0.7796 0.0136 0.4050 0.3518 0.3269 0.0673 0.0480 1.811405 0.064580

119

Table 5.20 (b) Results of equation 1 regression for earnings quality ABSDATA

ABSDATA = 0 + i CFVR + 2,3,4,5 UCP + 6 SSVR + 7 ACI + 8 ACC + 9SIZE + 10 Controls for EQ +

Dependent Variable: ABSDATA Included observations: 151 after adjustments White Heteroskedasticity-Consistent Standard Errors & Covariance Variable C UCPMN UCPINST UCPGOV UCPFAM ACC ACI CFVR SSVR LGMV CAPINT OC R-squared Adjusted R-squared Coefficient -3.498777 0.286607 -0.809175 -0.417655 -0.128597 -0.812367 1.108497 0.517643 -2.014903 -0.240589 -1.069421 0.875636 0.212088 0.149736 Std. Error 1.591336 0.497123 0.578915 0.583196 0.475471 0.744025 0.632959 0.578760 1.339638 0.116881 0.513841 0.478115 F-statistic Prob(F-statistic) t-Statistic -2.198641 0.576532 -1.397745 -0.716149 -0.270462 -1.091855 1.751294 0.894400 -1.504065 -2.058414 -2.081229 1.831433 Prob. 0.0296 0.5652 0.1644 0.4751 0.7872 0.2768 0.0821 0.3727 0.1348 0.0414 0.0392 0.0692 3.401431 0.000338

120

Table 5.20 (c) Results of equation 1 regression for earnings quality ABSDATCA

ABSDATCA = 0 + i CFVR + 2,3,4,5 UCP + 6 SSVR + 7 ACI + 8 ACC + 9SIZE + 10 Controls for EQ +

Dependent Variable: ABSDATCA Included observations: 151 after adjustments White Heteroskedasticity-Consistent Standard Errors & Covariance Variable C UCPMN UCPINST UCPGOV UCPFAM ACC ACI CFVR SSVR LGMV CAPINT OC R-squared Adjusted R-squared Coefficient -4.525943 -0.363396 -0.203948 -1.184706 -0.591863 -0.148402 0.431023 -0.210126 -3.406535 -0.141408 -0.640600 1.468395 0.257012 0.198215 Std. Error 1.449884 0.378125 0.411217 0.447421 0.277602 0.601677 0.414867 0.507543 1.234491 0.086504 0.493847 0.450099 F-statistic Prob(F-statistic) t-Statistic -3.121589 -0.961048 -0.495961 -2.647857 -2.132055 -0.246647 1.038942 -0.414005 -2.759465 -1.634699 -1.297161 3.262385 Prob. 0.0022 0.3382 0.6207 0.0090 0.0348 0.8055 0.3006 0.6795 0.0066 0.1044 0.1967 0.0014 4.371135 0.000013

121

Table 5.20 (d) Results of equation 1 regression for earnings quality PERS

PERS= 0 + i CFVR + 2,3,4,5 UCP + 6 SSVR + 7 ACI + 8 ACC + 9SIZE + 10 Controls for EQ +

Dependent Variable: PERS Included observations: 98 after adjustments White Heteroskedasticity-Consistent Standard Errors & Covariance Variable C UCPMN UCPINST UCPGOV UCPFAM SSVR ACI ACC CFVR LGMV STDREV R-squared Adjusted R-squared Coefficient 0.169770 -0.178708 -0.233684 -0.047423 -0.059351 -0.769400 0.065818 0.114777 -0.002842 -0.048040 0.168939 0.062728 -0.045005 Std. Error 0.396223 0.179605 0.155306 0.160449 0.142050 0.394764 0.211390 0.173784 0.198705 0.034693 0.293628 F-statistic Prob(F-statistic) t-Statistic 0.428470 -0.995007 -1.504669 -0.295564 -0.417818 -1.949013 0.311356 0.660458 -0.014302 -1.384740 0.575351 Prob. 0.6694 0.3225 0.1360 0.7683 0.6771 0.0545 0.7563 0.5107 0.9886 0.1697 0.5665 0.582255 0.824347

122

Table 5.20 (e) Results of equation 1 regression for earnings quality PRED

PRED = 0 + i CFVR + 2,3,4,5 UCP + 6 SSVR + 7 ACI + 8 ACC + 9SIZE + 10 Controls for EQ +

Dependent Variable: PRED Included observations: 98 after adjustments White Heteroskedasticity-Consistent Standard Errors & Covariance Variable C UCPMN UCPINST UCPGOV UCPFAM SSVR ACI ACC CFVR LGMV STDREV R-squared Adjusted R-squared Coefficient 0.014874 -0.042280 -0.088591 0.029800 0.028086 -0.278021 0.043652 0.016816 0.083134 0.004746 0.292701 0.080348 -0.025359 Std. Error 0.145001 0.070063 0.064926 0.064131 0.060076 0.252621 0.065219 0.061846 0.104572 0.010494 0.157739 F-statistic Prob(F-statistic) t-Statistic 0.102579 -0.603450 -1.364503 0.464675 0.467499 -1.100547 0.669321 0.271910 0.794995 0.452208 1.855606 Prob. 0.9185 0.5478 0.1759 0.6433 0.6413 0.2741 0.5051 0.7863 0.4288 0.6522 0.0669 0.760102 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 nonpyramidal 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 C UCPMN UCPINST UCPGOV UCPFAM SSVR ACI ACC CFVR LGMV CAPINT R-squared Adjusted R-squared Coefficient 0.079044 -0.022383 0.006926 -0.002034 0.011548 -0.064072 0.013304 0.051293 -0.014715 -0.008127 -0.012946 0.189137 -0.072432 Std. Error 0.064963 0.060500 0.049669 0.044137 0.040638 0.077022 0.028337 0.040165 0.025607 0.004859 0.033361 F-statistic Prob(F-statistic) t-Statistic 1.216751 -0.369973 0.139447 -0.046088 0.284181 -0.831865 0.469492 1.277040 -0.574631 -1.672604 -0.388065 Prob. 0.2329 0.7139 0.8900 0.9635 0.7782 0.4118 0.6420 0.2111 0.5697 0.1045 0.7006 0.723088 0.696922

126

Table 5.20 (g) Non-pyramidal companies in ABRES sample

Dependent Variable: ABRES Included observations: 102 Variable C UCPMN UCPGOV UCPFAM SSVR ACI ACC CF LGMV CAPINT R-squared Adjusted R-squared Coefficient 0.086939 -0.001822 -0.018284 0.004055 -0.128991 0.006698 -0.043743 -0.011682 -0.001483 -0.023827 0.173867 0.093049 Std. Error 0.035751 0.016780 0.015139 0.012398 0.055416 0.022100 0.023624 0.030630 0.003423 0.017775 F-statistic Prob(F-statistic) t-Statistic 2.431803 -0.108561 -1.207809 0.327052 -2.327692 0.303090 -1.851645 -0.381403 -0.433368 -1.340504 Prob. 0.0170 0.9138 0.2302 0.7444 0.0221 0.7625 0.0673 0.7038 0.6658 0.1834 2.151351 0.032669

Table 5.20 (h) Pyramidal companies in ABSDATA sample

Dependent Variable: ABSDATA Included observations: 40 after adjustments Variable C UCPMN UCPINST UCPGOV UCPFAM SSVR ACI ACC CFVR LGMV CAPINT R-squared Adjusted R-squared Coefficient -4.107776 1.626625 2.418191 2.738160 3.517478 -2.005748 1.616366 -0.342052 -0.196153 -0.343239 -2.100986 0.414124 0.212098 Std. Error 2.340185 2.307689 1.874137 1.682786 1.531328 2.974324 1.116992 1.522174 0.943685 0.183410 1.324875 F-statistic Prob(F-statistic) t-Statistic -1.755321 0.704872 1.290295 1.627159 2.297011 -0.674354 1.447071 -0.224713 -0.207858 -1.871431 -1.585799 Prob. 0.0898 0.4865 0.2071 0.1145 0.0290 0.5054 0.1586 0.8238 0.8368 0.0714 0.1236 2.049854 0.064330

127

Table 5.20 (i) Non-pyramidal companies in ABSDATA sample

Dependent Variable:ABSDATA Included observations: 118 after adjustments Variable C UCPMN UCPGOV UCPFAM SSVR ACI ACC CF LGMV CAPINT OC R-squared Adjusted R-squared Coefficient -4.131862 0.457482 -0.903557 -0.113961 -2.212612 0.111077 -0.946022 1.173596 -0.151666 -0.824332 1.103235 0.207683 0.133634 Std. Error 1.606964 0.612346 0.511233 0.445352 1.664680 0.754123 0.747393 1.063377 0.115207 0.631639 0.525159 F-statistic Prob(F-statistic) t-Statistic -2.571223 0.747098 -1.767409 -0.255890 -1.329152 0.147292 -1.265763 1.103649 -1.316464 -1.305067 2.100764 Prob. 0.0115 0.4566 0.0800 0.7985 0.1866 0.8832 0.2083 0.2722 0.1908 0.1947 0.0380 2.804691 0.003999

Table 5.20 (j) Pyramidal companies in ABSDATCA sample

Dependent Variable: ABSDATCA Included observations: 40 after adjustments Variable C UCPMN UCPINST UCPGOV UCPFAM SSVR ACI ACC CFVR LGMV CAPINT R-squared Adjusted R-squared Coefficient -1.751461 -3.239049 -0.483200 -2.040747 -0.560491 -3.774609 2.048086 1.257858 -0.879102 -0.208104 -0.712102 0.279964 0.031675 Std. Error 2.315959 2.283800 1.854736 1.665366 1.515476 2.943533 1.105428 1.506416 0.933916 0.181511 1.311160 F-statistic Prob(F-statistic) t-Statistic -0.756257 -1.418272 -0.260522 -1.225404 -0.369845 -1.282339 1.852753 0.835001 -0.941307 -1.146508 -0.543109 Prob. 0.4556 0.1668 0.7963 0.2303 0.7142 0.2099 0.0741 0.4105 0.3543 0.2610 0.5912 1.127574 0.376272

128

Table 5.20 (k) Non-pyramidal companies in ABSDATCA sample

Dependent Variable: ABSDATCA Included observations: 118 after adjustments Variable C UCPMN UCPGOV UCPFAM SSVR ACI ACC CF LGMV CAPINT OC R-squared Adjusted R-squared Coefficient -3.349700 0.028147 -1.011089 -0.534950 -2.541033 -0.298327 -0.538576 0.940903 -0.260736 -1.288789 1.379634 0.303481 0.238386 Std. Error 1.442890 0.549824 0.459035 0.399881 1.494713 0.677125 0.671083 0.954804 0.103444 0.567148 0.471539 F-statistic Prob(F-statistic) t-Statistic -2.321522 0.051193 -2.202641 -1.337772 -1.700015 -0.440578 -0.802547 0.985440 -2.520547 -2.272405 2.925811 Prob. 0.0222 0.9593 0.0298 0.1838 0.0920 0.6604 0.4240 0.3266 0.0132 0.0251 0.0042 4.662111 0.000016

Table 5.20 (l) Pyramidal companies in PERS sample

Dependent Variable: PERS Included observations: 43 Variable C UCPMN UCPINST UCPGOV UCPFAM SSVR ACI ACC CFVR LGMV STDREV R-squared Adjusted R-squared Coefficient 0.974650 0.066860 -0.245729 -0.403055 -0.354819 0.676202 -0.197221 -0.014887 0.109101 -0.065547 0.053343 0.152454 -0.112404 Std. Error 0.705859 0.532511 0.510575 0.390832 0.371806 0.758610 0.250528 0.303912 0.239404 0.058129 0.417408 F-statistic Prob(F-statistic) t-Statistic 1.380800 0.125556 -0.481279 -1.031274 -0.954312 0.891369 -0.787221 -0.048986 0.455720 -1.127627 0.127795 Prob. 0.1769 0.9009 0.6336 0.3101 0.3471 0.3794 0.4369 0.9612 0.6517 0.2679 0.8991 0.575608 0.821341

129

Table 5.20 (m) Non-pyramidal companies in PERS sample

Dependent Variable: PERS Included observations: 64 after adjustments Variable C UCPMN UCPGOV UCPFAM SSVR ACI ACC CF LGMV STDREV R-squared Adjusted R-squared Coefficient -0.622320 0.192981 -0.040326 0.100582 0.452263 0.124942 -0.568724 0.180638 0.116904 0.043469 0.174149 0.036507 Std. Error 0.523918 0.283230 0.206996 0.195749 0.705313 0.291516 0.342848 0.466489 0.044158 0.599514 F-statistic Prob(F-statistic) t-Statistic -1.187820 0.681360 -0.194814 0.513833 0.641223 0.428593 -1.658824 0.387229 2.647388 0.072507 Prob. 0.2401 0.4986 0.8463 0.6095 0.5241 0.6699 0.1029 0.7001 0.0106 0.9425 1.265233 0.277031

Table 5.20 (n) Pyramidal companies in PRED sample

Dependent Variable: PRED Included observations: 43 Variable C UCPMN UCPINST UCPGOV UCPFAM SSVR ACI ACC CFVR LGMV STDREV R-squared Adjusted R-squared Coefficient 0.320356 -0.007202 0.108869 0.179332 0.153775 0.013404 -0.020527 -0.216313 -0.078818 -0.025063 -0.000270 0.141828 -0.126350 Std. Error 0.285882 0.215674 0.206789 0.158292 0.150586 0.307247 0.101467 0.123088 0.096962 0.023543 0.169056 F-statistic Prob(F-statistic) t-Statistic 1.120590 -0.033393 0.526471 1.132920 1.021177 0.043627 -0.202307 -1.757377 -0.812879 -1.064554 -0.001596 Prob. 0.2708 0.9736 0.6022 0.2657 0.3148 0.9655 0.8410 0.0884 0.4223 0.2950 0.9987 0.528858 0.856793

130

Table 5.20 (o) Non-pyramidal companies in PRED sample

Dependent Variable: PRED Included observations: 64 after adjustments Variable C UCPMN UCPGOV UCPFAM SSVR ACI ACC CF LGMV STDREV R-squared Adjusted R-squared Coefficient -0.261046 -0.007916 -0.008722 0.035839 -0.136762 0.085467 0.192479 0.394237 0.002328 0.693636 0.217214 0.086750 Std. Error 0.219802 0.118825 0.086843 0.082124 0.295904 0.122302 0.143837 0.195709 0.018526 0.251518 F-statistic Prob(F-statistic) t-Statistic -1.187640 -0.066617 -0.100430 0.436401 -0.462183 0.698820 1.338175 2.014399 0.125683 2.757799 Prob. 0.2402 0.9471 0.9204 0.6643 0.6458 0.4877 0.1864 0.0490 0.9004 0.0079 1.664934 0.120544

5.9.3 Earnings quality and cost of equity Equation 2: COE = 0 + i,2,3,4Earnings quality 1,2,3,4 + 5 SIZE + 6 + 7 BTMV + 6

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 Whites 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 (Constant) t Sig. EQ t Sig. LGMV t Sig. BETA t Sig. LBTMV t Sig. R2 F Sig N -ve +ve ABRES 10.61 0.000 0.13** 2.22 0.03 -0.004*** -3.22 0.00 0.002 0.35 0.73 0.03*** 6.39 0.000 0.39 21.99 0.00 141 ABSDATA 14.36 0.00 0.00 0.38 0.70 -0.01*** -3.58 0.00 -0.00 -0.12 0.90 0.03*** 6.63 0.00 0.40 24.34 0.00 151 ABSDATCA 12.51 0.00 0.00 0.89 0.38 -0.004*** -3.38 0.00 -0.00 -0.10 0.92 0.03*** 6.98 0.00 0.40 24.60 0.00 151 PERS 9.87 0.00 0.00 0.94 0.35 -0.01*** -2.97 0.00 -0.00 -0.58 0.57 0.03*** 6.79 0.00 0.40 18.64 0.00 118 PRED 9.17 0.00 0.04** 1.88 0.06 -0.00*** -2.76 0.01 -0.01 -1.33 0.18 0.03*** 7.19 0.00 0.45 23.20 0.00 118

+ve

+ve

*** 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 (Constant) t Sig. EQ t Sig. LGMV t Sig. BETA t Sig. LBTMV t Sig. R2 F Sig N +ve ABRES 6.95 0.00 0.35** 2.10 0.04 -0.01*** -2.95 0.00 -0.00 -0.06 0.96 0.03*** 3.30 0.00 0.19 7.96 0.00 141 ABSDATA 9.77 0.00 0.01** 2.09 0.04 -0.01*** -2.89 0.00 -0.00 -0.44 0.66 0.03*** 3.69 0.00 0.19 8.52 0.00 151 ABSDATCA 9.98 0.00 0.01*** 2.53 0.01 -0.01*** -2.94 0.00 -0.00 -0.25 0.80 0.03*** 3.78 0.00 0.19 8.67 0.00 151 PERS 7.20 0.00 -0.01 -0.97 0.33 -0.01*** -3.17 0.00 -0.00 -0.11 0.91 0.03*** 3.41 0.00 0.19 6.46 0.00 118 PRED -6.90 0.00 0.31** 1.97 0.05 -0.07*** -2.60 0.01 -0.03 -0.38 0.71 0.18*** 2.88 0.00 0.20 7.14 0.00 118

-ve

+ve

+ve

*** 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 = 0 + ABRES + 5 SIZE + 6 + 7 BTMV + 6

Dependent Variable: COE Included observations: 141 after adjustments White Heteroskedasticity-Consistent Standard Errors & Covariance Variable C ABRES LGMV BETA LBTMV R-squared Adjusted R-squared Coefficient 0.133755 0.129050 -0.004673 0.002087 0.028085 0.392715 0.374854 Std. Error 0.012606 0.058032 0.001453 0.005964 0.004398 F-statistic Prob(F-statistic) t-Statistic 10.61061 2.223794 -3.216660 0.349988 6.386051 Prob. 0.0000 0.0278 0.0016 0.7269 0.0000 21.98693 0.000000

Table 5.21 (a) (ii) Results of equation 2 regression - ABRES and COEA estimate COEa = 0 + ABRES + 5 SIZE + 6 + 7 BTMV + 6

Dependent Variable: COEA Included observations: 141 after adjustments White Heteroskedasticity-Consistent Standard Errors & Covariance Variable C ABRES LGMV BETA LBTMV R-squared Adjusted R-squared Coefficient 0.212813 0.351431 -0.010327 -0.000742 0.027584 0.189766 0.165935 Std. Error 0.030630 0.167227 0.003501 0.013408 0.008361 F-statistic Prob(F-statistic) t-Statistic 6.947958 2.101520 -2.949929 -0.055376 3.299339 Prob. 0.0000 0.0374 0.0037 0.9559 0.0012 7.963163 0.000008

135

Table 5.21 (b) (i) Results of equation 2 regression- ABSDATA and COE estimate COE = 0 + ABSDATA + 5 SIZE + 6 + 7 BTMV + 6

Dependent Variable: COE Method: Least Squares Included observations: 151 after adjustments White Heteroskedasticity-Consistent Standard Errors & Covariance Variable C ABSDATA LGMV BETA LBTMV R-squared Adjusted R-squared Coefficient 0.145026 0.000490 -0.005024 -0.000542 0.027900 0.400131 0.383696 Std. Error 0.010097 0.001290 0.001403 0.004353 0.004210 F-statistic Prob(F-statistic) t-Statistic 14.36344 0.379850 -3.580372 -0.124409 6.626383 Prob. 0.0000 0.7046 0.0005 0.9012 0.0000 24.34664 0.000000

Table 5.21 (b) (ii) Results of equation 2 regression- ABSDATA and COEA estimate COEA = 0 + ABSDATA + 5 SIZE + 6 + 7 BTMV + 6
Dependent Variable: COEA Included observations: 151 after adjustments White Heteroskedasticity-Consistent Standard Errors & Covariance Variable C ABSDATA LGMV BETA LBTMV R-squared Adjusted R-squared Coefficient 0.254799 0.006929 -0.010297 -0.004700 0.028993 0.189323 0.167112 Std. Error 0.026082 0.003318 0.003566 0.010744 0.007851 F-statistic Prob(F-statistic) t-Statistic 9.769311 2.088405 -2.887237 -0.437454 3.693010 Prob. 0.0000 0.0385 0.0045 0.6624 0.0003 8.524073 0.000003

136

Table 5.21 (c) (i) Results of equation 2 regression - ABSDATCA and COE estimate COE = 0 + ABSDATCA + 5 SIZE + 6 + 7 BTMV + 6

Dependent Variable: COE Included observations: 151 after adjustments White Heteroskedasticity-Consistent Standard Errors & Covariance Variable C ABSDATCA LGMV BETA LBTMV R-squared Adjusted R-squared Coefficient 0.146053 0.001391 -0.004714 -0.000429 0.028368 0.402602 0.386235 Std. Error 0.010066 0.001569 0.001393 0.004307 0.004061 F-statistic Prob(F-statistic) t-Statistic 14.50995 0.886659 -3.383814 -0.099676 6.984845 Prob. 0.0000 0.3767 0.0009 0.9207 0.0000 24.59826 0.000000

Table 5.21 (c) (ii) Results of equation 2 regression - ABSDATCA and COEA estimate COEA = 0 + ABSDATCA + 5 SIZE + 6 + 7 BTMV + 6

Dependent Variable: COEA Included observations: 151 after adjustments White Heteroskedasticity-Consistent Standard Errors & Covariance Variable C ABSDATCA LGMV BETA LBTMV R-squared Adjusted R-squared Coefficient 0.255970 0.008082 -0.010123 -0.002620 0.029373 0.191941 0.169802 Std. Error 0.025642 0.003193 0.003437 0.010544 0.007775 F-statistic Prob(F-statistic) t-Statistic 9.982527 2.531434 -2.944965 -0.248500 3.777826 Prob. 0.0000 0.0124 0.0038 0.8041 0.0002 8.669945 0.000003

137

Table 5.21 (d) (i) Results of equation 2 regression - PERS and COE estimate COE = 0 + PERS + 5 SIZE + 6 + 7 BTMV + 6

Dependent Variable: COE Included observations: 118 after adjustments White Heteroskedasticity-Consistent Standard Errors & Covariance Variable C PERS LGMV BETA LBTMV R-squared Adjusted R-squared Coefficient 0.141047 0.004583 -0.005154 -0.003302 0.030963 0.397487 0.376159 Std. Error 0.014283 0.004872 0.001737 0.005732 0.004563 F-statistic Prob(F-statistic) t-Statistic 9.874852 0.940544 -2.967324 -0.576000 6.785149 Prob. 0.0000 0.3489 0.0037 0.5658 0.0000 18.63692 0.000000

Table 5.21 (d) (ii) Results of equation 2 regression - PERS and COEA estimate COEA = 0 + PERS + 5 SIZE + 6 + 7 BTMV + 6

Dependent Variable: COEA Included observations: 118 after adjustments White Heteroskedasticity-Consistent Standard Errors & Covariance Variable C PERS LGMV BETA LBTMV R-squared Adjusted R-squared Coefficient 0.244983 0.014323 -0.013442 -0.001469 0.027470 0.186177 0.157369 Std. Error 0.034038 0.014756 0.004247 0.013414 0.008051 F-statistic Prob(F-statistic) t-Statistic 7.197300 0.970664 -3.165286 -0.109486 3.412025 Prob. 0.0000 0.3338 0.0020 0.9130 0.0009 6.462690 0.000101

138

Table 5.21 (e) (i) Results of equation 2 regression - PRED and COE estimate COE = 0 + PRED + 5 SIZE + 6 + 7 BTMV + 6

Dependent Variable: COE Included observations: 118 after adjustments White Heteroskedasticity-Consistent Standard Errors & Covariance Variable C PRED LGMV BETA LBTMV R-squared Adjusted R-squared Coefficient 0.139314 0.037015 -0.004924 -0.008333 0.031473 0.450902 0.431465 Std. Error 0.015192 0.019686 0.001783 0.006250 0.004375 F-statistic Prob(F-statistic) t-Statistic 9.169980 1.880247 -2.762287 -1.333316 7.193619 Prob. 0.0000 0.0626 0.0067 0.1851 0.0000 23.19800 0.000000

Table 5.21 (e) (ii) Results of equation 2 regression - PRED and COEA estimate COEA = 0 + PRED + 5 SIZE + 6 + 7 BTMV + 6

Dependent Variable: LCOEA Included observations: 118 after adjustments Variable C PRED LGMV LBTMV BETA R-squared Adjusted R-squared Coefficient -1.481669 0.308715 -0.073811 0.182519 -0.029862 0.201711 0.173453 Std. Error 0.214931 0.156397 0.028409 0.063394 0.079500 F-statistic Prob(F-statistic) t-Statistic -6.893707 1.973924 -2.598152 2.879116 -0.375621 Prob. 0.0000 0.0508 0.0106 0.0048 0.7079 7.138209 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 = + (R m,t R f,t) + Size f,t + BTMV f,t + EQ f,t + 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 = + (R m,t R f,t) + Size + BTMV + ABRES +

Dependent Variable: R i,t R f,t White Heteroskedasticity-Consistent Standard Errors & Covariance Variable C (R m,t R f,t) LGMV LBTMV ABRES R-squared Adjusted R-squared Coefficient -0.108864 3.511877 -0.009587 -0.073929 0.357854 0.207669 0.185506 Std. Error 0.139256 1.327909 0.020209 0.049777 0.670017 F-statistic Prob(F-statistic) t-Statistic -0.781757 2.644667 -0.474367 -1.485191 0.534096 Prob. 0.4356 0.0091 0.6360 0.1397 0.5941 9.370014 0.000001

Table 5.22(b) Results of testing earnings quality (ABSDATCA) and excess return R i,t R f,t = + (R m,t R f,t) + Size + BTMV + ABSDATCA +
Dependent Variable: R i,t R f,t White Heteroskedasticity-Consistent Standard Errors & Covariance Variable C (R m,t R f,t) LGMV LBTMV ABSDATCA R-squared Adjusted R-squared Coefficient -0.056494 3.453557 -0.004491 -0.062109 0.022153 0.199576 0.178512 Std. Error 0.121957 1.309422 0.022839 0.049976 0.030079 F-statistic Prob(F-statistic) t-Statistic -0.463227 2.637468 -0.196653 -1.242779 0.736504 Prob. 0.6439 0.0092 0.8444 0.2159 0.4626 9.474829 0.000001

141

Table 5.22(c) Results of testing earnings quality (ABSDATA) and excess return R i,t R f,t = + (R m,t R f,t) + Size + BTMV + ABSDATA +

Dependent Variable: R i,t R f,t White Heteroskedasticity-Consistent Standard Errors & Covariance Variable C (R m,t R f,t) LGMV LBTMV ABSDATA R-squared Adjusted R-squared Coefficient -0.062973 3.535445 -0.004271 -0.062712 0.021989 0.200271 0.179225 Std. Error 0.119938 1.362388 0.023913 0.050863 0.028202 F-statistic Prob(F-statistic) t-Statistic -0.525049 2.595035 -0.178599 -1.232965 0.779689 Prob. 0.6003 0.0104 0.8585 0.2195 0.4368 9.516088 0.000001

Table 5.22(d) Results of testing earnings quality (PERS) and excess return R i,t R f,t = + (R m,t R f,t) + Size + BTMV + PERS +

Dependent Variable: R i,t R f,t White Heteroskedasticity-Consistent Standard Errors & Covariance Variable C (R m,t R f,t) LGMV LBTMV PERS R-squared Adjusted R-squared Coefficient -0.060824 3.878367 -0.011102 -0.027595 -0.006507 0.238020 0.211969 Std. Error 0.130289 1.530388 0.018521 0.044521 0.091027 F-statistic Prob(F-statistic) t-Statistic -0.466840 2.534237 -0.599464 -0.619831 -0.071489 Prob. 0.6415 0.0126 0.5500 0.5366 0.9431 9.136813 0.000002

142

Table 5.22(e) Results of testing earnings quality (PRED) and excess return R i,t R f,t = + (R m,t R f,t) + Size + BTMV + PRED +

Dependent Variable: R i,t R f,t White Heteroskedasticity-Consistent Standard Errors & Covariance Variable C (R m,t R f,t) LGMV LBTMV PRED R-squared Adjusted R-squared Coefficient -0.063713 3.872478 -0.011379 -0.028069 0.020400 0.238095 0.212047 Std. Error 0.134166 1.480569 0.019804 0.043772 0.112599 F-statistic Prob(F-statistic) t-Statistic -0.474887 2.615534 -0.574596 -0.641258 0.181176 Prob. 0.6358 0.0101 0.5667 0.5226 0.8565 9.140629 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 COE COEA AVMR YER 1 .722 .000 .015 .864 -.008 .922 COEA .722 .000 1 .014 .866 -.012 .891 AVMR .015 .864 .014 .866 1 .975 .000 YER -.008 .922 -.012 .891 .975 .000 1

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 + 1

Reduced form equation 2 : SSVR = b0 + b1,2,3,4UCP + b5 ACI + b6 ACC + b7SIZE + b8BTMV + 2

The predicted values ^CFVR, ^SSVR and ^COE are obtained. The following equation is run to test for the exogeneity of CFVR and SSVR.

COE

0 + 1 CFVR + 2,3,4,5 UCP + 6 SSVR + 7 ACI + 8 ACC + 9SIZE + 10BTMV + c1 ^CFVR + c2 ^SSVR + 7,8,9

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 C LGMV LBTMV BETA CFVR SSVR ^CFVR Coefficient 0.117788 -0.003618 0.032740 -0.001414 0.014285 -0.055998 -0.006543 -0.006919 0.018665 0.014092 0.003991 0.016535 0.452174 0.405460 Std. Error 0.018123 0.001880 0.004563 0.005520 0.011858 0.030384 0.003307 0.003488 0.021681 0.004992 0.011890 0.015098 F-statistic Prob(F-statistic) t-Statistic 6.499316 -1.924232 7.174682 -0.256070 1.204647 -1.843008 -1.978745 -1.983528 0.860898 2.822951 0.335698 1.095202 Prob. 0.0000 0.0565 0.0000 0.7983 0.2305 0.0676 0.0500 0.0494 0.3909 0.0055 0.7376 0.2755 9.679643 0.000000

^SSVR
UCPINST UCPFAM ACI ACC R-squared Adjusted R-squared

Table 5.23 (b) Testing for coefficients of ^CFVR and ^SSVR

Wald Test:

Test Statistic F-statistic Chi-square

Value 2.678529 5.357059

df (2, 129) 2

Probability 0.0725 0.0687

Null Hypothesis Summary: Normalized Restriction (= 0) C(7) C(8) Restrictions are linear in coefficients. Value -0.006543 -0.006919 Std. Err. 0.003307 0.003488

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 C LGMV LBTMV BETA CFVR SSVR ^CFVR ^SSVR UCPINST UCPFAM ACI ACC R-squared Adjusted R-squared Coefficient 0.232324 -0.010651 0.028066 -0.006169 0.017223 -0.235330 -0.007192 -0.001030 -0.004079 0.017634 -0.004792 0.017359 0.246630 0.182389 Std. Error 0.041737 0.004331 0.010509 0.012713 0.027309 0.069974 0.007616 0.008034 0.049930 0.011496 0.027382 0.034770 F-statistic Prob(F-statistic) t-Statistic 5.566332 -2.459531 2.670591 -0.485225 0.630671 -3.363123 -0.944384 -0.128215 -0.081692 1.533863 -0.174990 0.499258 Prob. 0.0000 0.0152 0.0085 0.6283 0.5294 0.0010 0.3467 0.8982 0.9350 0.1275 0.8614 0.6184 3.839137 0.000086

Table 5.23 (d) Testing for coefficients of ^CFVR and ^SSVR

Wald Test: Equation: EQEXOCOEA Test Statistic F-statistic Chi-square Value 0.507749 1.015497 df (2, 129) 2 Probability 0.6030 0.6018

Null Hypothesis Summary: Normalized Restriction (= 0) C(7) C(8) Restrictions are linear in coefficients. Value -0.007192 -0.001030 Std. Err. 0.007616 0.008034

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 C UCPMN UCPINST UCPFAM ACC ACI LGMV LBTMV BETA CFVR SSVR Coefficient 0.132124 0.016691 0.014993 0.016901 0.015666 0.006167 -0.003889 0.029885 -0.004622 0.013343 -0.067063 -0.153670 0.470226 0.428301 Std. Error 0.014244 0.009771 0.008976 0.004050 0.011719 0.010215 0.001473 0.003427 0.004525 0.013684 0.031652 0.090459 F-statistic Prob(F-statistic) t-Statistic 9.275499 1.708206 1.670342 4.173364 1.336793 0.603702 -2.641088 8.720063 -1.021412 0.975131 -2.118766 -1.698778 Prob. 0.0000 0.0898 0.0971 0.0001 0.1835 0.5470 0.0092 0.0000 0.3088 0.3312 0.0359 0.0916 11.21599 0.000000

^SSVR
R-squared Adjusted R-squared

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 C UCPMN UCPINST UCPFAM ACC ACI LGMV LBTMV BETA CFVR SSVR ^SSVR R-squared Adjusted R-squared S.E. of regression Sum squared resid Log likelihood Durbin-Watson stat Coefficient 0.228413 0.013170 -0.008619 0.021214 0.007020 -0.000385 -0.011243 0.026379 -0.004844 0.007368 -0.219707 0.095329 0.252465 0.193308 0.058732 0.479481 220.0412 2.050598 Std. Error 0.048826 0.022028 0.049446 0.014378 0.034761 0.025651 0.004159 0.009113 0.011357 0.026056 0.063788 0.315595 t-Statistic 4.678079 0.597872 -0.174306 1.475507 0.201944 -0.014991 -2.702931 2.894711 -0.426577 0.282775 -3.444321 0.302061 Prob. 0.0000 0.5509 0.8619 0.1423 0.8403 0.9881 0.0077 0.0044 0.6703 0.7778 0.0008 0.7631 0.148808 0.065392 -2.755513 -2.515729 4.267684 0.000018

Mean dependent var S.D. dependent var Akaike info criterion Schwarz criterion F-statistic Prob(F-statistic)

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 C LGMV LBTMV BETA CFVR SSVR PREDCFVR PREDSSVR UCPINST UCPFAM ACI ACC R-squared Adjusted R-squared Coefficient 0.149486 -0.005266 0.033224 -0.010618 0.004950 -0.060397 0.005964 -0.174906 0.010957 0.013907 0.017496 -0.009263 0.450696 0.393693 Std. Error 0.042026 0.002192 0.006192 0.008540 0.013895 0.033229 0.035030 0.228285 0.029123 0.006579 0.014245 0.014502 F-statistic Prob(F-statistic) t-Statistic 3.556943 -2.402582 5.365775 -1.243310 0.356221 -1.817602 0.170265 -0.766171 0.376237 2.113900 1.228184 -0.638781 Prob. 0.0006 0.0180 0.0000 0.2165 0.7224 0.0719 0.8651 0.4453 0.7075 0.0369 0.2221 0.5243 7.906510 0.000000

Table 5.25 (b) Testing for coefficients of ^CFVR and ^SSVR

Wald Test: Equation: Untitled Test Statistic F-statistic Chi-square Value 0.420436 0.840873 df (2, 106) 2 Probability 0.6579 0.6568

Null Hypothesis Summary: Normalized Restriction (= 0) C(7) C(8) Restrictions are linear in coefficients. Value 0.005964 -0.174906 Std. Err. 0.035030 0.228285

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 C LGMV LBTMV BETA CFVR SSVR PREDCFVR PREDSSVR UCPINST UCPFAM ACI ACC R-squared Adjusted R-squared Coefficient 0.235433 -0.011567 0.029061 -0.008995 -0.000210 -0.200100 -0.016334 0.107186 -0.007141 0.020702 0.018468 -0.000183 0.255154 0.177858 Std. Error 0.093475 0.004875 0.013772 0.018995 0.030906 0.073907 0.077914 0.507753 0.064776 0.014632 0.031684 0.032255 F-statistic Prob(F-statistic) t-Statistic 2.518660 -2.372805 2.110147 -0.473541 -0.006789 -2.707447 -0.209647 0.211099 -0.110239 1.414837 0.582887 -0.005689 Prob. 0.0133 0.0195 0.0372 0.6368 0.9946 0.0079 0.8343 0.8332 0.9124 0.1600 0.5612 0.9955 3.301019 0.000633

Table 5.25 (d) Testing for coefficients of ^CFVR and ^SSVR

Wald Test: Equation: EQEXOCOEA Test Statistic F-statistic Chi-square Value 0.071934 0.143869 df (2, 106) 2 Probability 0.9306 0.9306

Null Hypothesis Summary: Normalized Restriction (= 0) C(7) C(8) Value -0.016334 0.107186 Std. Err. 0.077914 0.507753

151

5.9.6 Ownership structure, monitoring mechanisms and market assessment Equation 3: COE =
9SIZE

0 + 1 CFVR + 2,3,4,5 UCP + 6 SSVR + 7 ACI + 8 ACC + + 10BTMV + 7,8,9

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 Whites 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 UCPs 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 (Constant) UCPMn UCPInst UCPGov UCPFam SSVR ACI ACC LGMV LBTMV BETA CFVR R2 F Sig N +ve +ve +ve +ve ? -ve -ve -ve +ve +ve -ve 0.221** 0.046 0.165* 0.354*** -0.128* 0.067 0.051 -0.190*** 0.469*** -0.017 0.093 0.405 9.680 0.000 141 0.180* 0.032 0.111 0.286*** -0.164*** 0.053 0.031 -0.179** 0.519*** -0.039 0.088 0.428 11.216 0.000 151 0.022 0.045 0.110 0.226 -0.139* 0.076 -0.075 -0.208** 0.506*** -0.081 0.030 0.394 7.907 0.000 118 ABRES ABSDATA/ABSDATCA PERS/PRED

*** 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 (Constant) UCPMn UCPInst UCPGov UCPFam SSVR ACI ACC LGMV LBTMV BETA CFVR R2 F Sig N +ve +ve +ve +ve ? -ve -ve -ve +ve +ve -ve 0.099 0.005 -0.013 0.166 -0.274*** -0.002 0.069 -0.228*** 0.233*** -0.039 0.057 0.182 3.839 0.000 141
*** Coefficients that are significant at 1% level ** Coefficients that are significant at 5% level * Coefficients that are significant at 10% level.

ABRES

ABSDATA/ABSDATCA

PERS/PRED

0.055 -0.009 -0.035 0.144 -0.270*** -0.006 0.034 -0.223*** 0.278*** -0.044 0.024 0.193 4.268 0.000 151

0.009 -0.015 -0.041 0.134 -0.241*** 0.074 0.014 -0.227** 0.280*** -0.082 -0.001 0.178 3.301 0.001 128

155

5.9.7 The relationship that examines whether market assessment and the monitoring mechanisms could explain changes in ownership structure

Equation 4: CFVR = 0 + 1COE + 2UCP + 3 SSVR + 4 ACI + 5 ACC + 6SIZE + 7BTMV + 10,11,12

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 (Constant) 0.472*** -0.064 0.063 -0.195 -0.064 -0.133* 0.045 0.127 0.120 -0.165** 0.238 5.383 0.000

ABRES

ABSDATA/ABSDATCA

PERS/PRED

UCPMn UCPInst UCPGov UCPFam ACI ACC LGMV LBTMV COE SSVR R2 F Sig

+ve +ve +ve +ve -ve -ve +ve -ve -ve ?

-0.449*** -0.056 0.053 -0.170 -0.055 -0.128* 0.109 0.113 0.116 -0.157** 0.226 5.376 0.000 151

-0.413*** 0.001 0.092 -0.143 -0.144* -0.134 -0.048 0.125 0.038 -0.179** 0.220 4.309 0.000 118

N 141 *** 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 (Constant) UCPMn UCPInst UCPGov UCPFam ACI ACC LGMV LBTMV COEA SSVR R2 F Sig N +ve +ve +ve +ve -ve -ve +ve -ve -ve ? -0.455*** -0.060 0.084 -0.163 -0.056 -0.131* 0.035 0.172** 0.055 -0.167** 0.232 5.234 0.000 141 -0.434*** -0.052 0.067 -0.142 -0.049 -0.126* 0.094 0.168** 0.021 -0.172** 0.218 5.192 0.000 151 -0.414*** 0.003 0.096 -0.135 -0.141* -0.138 -0.056 0.144 -0.001 -0.185** 0.220 4.292 0.000 118 ABRES ABSDATA/ABSDATCA PERS/PRED

*** 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 = 0 + 1COE + 2,3,4,5UCP + 6 CFVR + 7 ACI + 8 ACC +
9SIZE

+ 10BTMV + 10,11,12

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 shareholdings of these shareholders. the

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 C COE UCPMN UCPINST UCPGOV UCPFAM CFVR LGMV LBTMV ACI ACC R-squared Adjusted R-squared Coefficient 0.184785 -0.456015 -0.029491 0.037359 -0.049904 -0.014510 -0.070698 0.000808 0.031144 -0.011434 0.044703 0.141326 0.075274 Std. Error 0.058855 0.247953 0.030132 0.056814 0.026733 0.022515 0.033451 0.005438 0.012110 0.032257 0.039598 F-statistic Prob(F-statistic) t-Statistic 3.139658 -1.839121 -0.978732 0.657562 -1.866770 -0.644461 -2.113508 0.148566 2.571699 -0.354467 1.128928 Prob. 0.0021 0.0682 0.3295 0.5120 0.0642 0.5204 0.0365 0.8821 0.0112 0.7236 0.2610 2.139626 0.025648

Table 5.28 (a) (i) ABRES Sample (with COEA)

Dependent Variable: SSVR Included observations: 141 after adjustments Variable C COEA UCPMN UCPINST UCPGOV UCPFAM CFVR LGMV LBTMV ACI ACC R-squared Adjusted R-squared Coefficient 0.202368 -0.340682 -0.031433 0.030478 -0.054951 -0.016861 -0.066991 -0.000943 0.026946 -0.015802 0.046723 0.189109 0.126733 Std. Error 0.055273 0.101607 0.028791 0.055148 0.025613 0.021180 0.032398 0.005289 0.010533 0.031209 0.038471 F-statistic Prob(F-statistic) t-Statistic 3.661256 -3.352922 -1.091740 0.552652 -2.145414 -0.796063 -2.067745 -0.178295 2.558153 -0.506336 1.214488 Prob. 0.0004 0.0010 0.2770 0.5815 0.0338 0.4274 0.0406 0.8588 0.0117 0.6135 0.2268 3.031747 0.001760

161

Table 5.28 (b) (i) ABSDATA Sample (with COE)

Dependent Variable: SSVR Included observations: 151 after adjustments Variable C COE UCPMN UCPINST UCPGOV UCPFAM CFVR LGMV LBTMV ACI ACC R-squared Adjusted R-squared Coefficient 0.204849 -0.627132 -0.030916 0.032253 -0.055653 -0.010504 -0.068305 0.001274 0.039877 -0.014488 0.046809 0.159224 0.099169 Std. Error 0.060073 0.255102 0.030951 0.058739 0.027814 0.023041 0.033511 0.005627 0.012619 0.032860 0.038409 F-statistic Prob(F-statistic) t-Statistic 3.409991 -2.458355 -0.998884 0.549089 -2.000938 -0.455883 -2.038258 0.226488 3.160128 -0.440889 1.218699 Prob. 0.0008 0.0152 0.3196 0.5838 0.0473 0.6492 0.0434 0.8212 0.0019 0.6600 0.2250 2.651287 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 C COEA UCPMN UCPINST UCPGOV UCPFAM CFVR LGMV LBTMV ACI ACC R-squared Adjusted R-squared Coefficient 0.211024 -0.355192 -0.037411 0.023679 -0.062329 -0.015027 -0.071253 4.54E-07 0.032377 -0.020034 0.045942 0.190623 0.132810 Std. Error 0.055620 0.091690 0.024377 0.050848 0.024463 0.019289 0.033912 0.005668 0.011867 0.036056 0.034497 F-statistic Prob(F-statistic) t-Statistic 3.794064 -3.873835 -1.534703 0.465683 -2.547841 -0.779048 -2.101100 8.02E-05 2.728254 -0.555624 1.331769 Prob. 0.0002 0.0002 0.1271 0.6422 0.0119 0.4373 0.0374 0.9999 0.0072 0.5794 0.1851 3.297256 0.000727

162

Table 5.28 (c) (i) PERS/PRED Sample (with COE)

Dependent Variable: SSVR Included observations: 118 after adjustments Variable C COE UCPMN UCPINST UCPGOV UCPFAM CFVR LGMV LBTMV ACI ACC R-squared Adjusted R-squared Coefficient 0.228923 -0.467521 -0.059670 -0.034430 -0.051099 -0.025882 -0.083112 -0.002592 0.038995 0.014123 -0.011985 0.117753 0.035300 Std. Error 0.072237 0.276300 0.038345 0.088128 0.035770 0.029970 0.039472 0.006749 0.016663 0.036439 0.040935 F-statistic Prob(F-statistic) t-Statistic 3.169061 -1.692076 -1.556120 -0.390681 -1.428537 -0.863584 -2.105581 -0.384033 2.340195 0.387574 -0.292769 Prob. 0.0020 0.0935 0.1226 0.6968 0.1560 0.3897 0.0376 0.7017 0.0211 0.6991 0.7703 1.428127 0.177755

Table 5.28 (c) (ii) PERS/PRED Sample (with COEA)

Dependent Variable: SSVR Included observations: 118 after adjustments Variable C COEA UCPMN UCPINST UCPGOV UCPFAM CFVR LGMV LBTMV ACI ACC R-squared Adjusted R-squared Coefficient 0.234071 -0.313487 -0.058497 -0.045201 -0.056764 -0.026958 -0.082536 -0.003599 0.034062 0.015249 -0.004207 0.148410 0.068822 Std. Error 0.067287 0.120056 0.037676 0.086437 0.034976 0.029204 0.038740 0.006607 0.014884 0.035747 0.039969 F-statistic Prob(F-statistic) t-Statistic 3.478706 -2.611162 -1.552604 -0.522937 -1.622942 -0.923093 -2.130537 -0.544819 2.288445 0.426586 -0.105260 Prob. 0.0007 0.0103 0.1235 0.6021 0.1075 0.3580 0.0354 0.5870 0.0241 0.6705 0.9164 1.864736 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

0 + 1 ^CFVR + 2,3,4,5 UCP + 6 ^SSVR + 7 ACI + 8 ACC + 9SIZE + 10BTMV + 11 + 7,8,9

The

results

are

in

table

5.29(a)

ABRES

sample

and

table

5.29(b)

for

ABSDATA/ABSDATCA sample. CFVR = 0 + 1^COE + 2,3,4,5UCP + 6 ^SSVR + 7 ACI + 8 ACC +

9SIZE + 10BTMV + 10,11,12

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 = 0 + 1 ^CFVR + 2,3,4,5 UCP + 6 ^SSVR + 7 ACI + 8 ACC + 9SIZE + 10BTMV + 11 + 7,8,9 Table 5.29 (a) ABRES Sample
Dependent Variable: COE Included observations: 141 after adjustments Variable C LGMV LBTMV BETA ^SSVR UCPMN UCPINST UCPFAM ACI ACC R-squared Adjusted R-squared Coefficient 0.111894 -0.003877 0.029297 -0.000974 -0.006824 0.014460 0.022019 0.017922 0.007248 0.021960 0.426696 0.387308 Std. Error 0.017903 0.001854 0.003879 0.005597 0.003158 0.009069 0.022494 0.006126 0.011496 0.015941 F-statistic Prob(F-statistic) t-Statistic 6.249931 -2.091129 7.551818 -0.174099 -2.160952 1.594538 0.978865 2.925417 0.630511 1.377554 Prob. 0.0000 0.0384 0.0000 0.8621 0.0325 0.1132 0.3295 0.0041 0.5295 0.1707 10.83332 0.000000

Table 5.29 (b) ABSDATA/ABSDATCA Sample

Dependent Variable: COE Included observations: 151 after adjustments Variable C LGMV LBTMV BETA ^SSVR UCPMN UCPINST UCPFAM ACI ACC R-squared Adjusted R-squared Coefficient 0.140970 -0.003676 0.030808 -0.004468 -0.234542 0.012441 0.013939 0.015899 0.005320 0.013954 0.435368 0.399327 Std. Error 0.019784 0.001793 0.003870 0.004918 0.133846 0.008840 0.021402 0.006171 0.011089 0.014989 F-statistic Prob(F-statistic) t-Statistic 7.125452 -2.050316 7.961765 -0.908526 -1.752324 1.407437 0.651317 2.576656 0.479784 0.930980 Prob. 0.0000 0.0422 0.0000 0.3652 0.0819 0.1615 0.5159 0.0110 0.6321 0.3535 12.08001 0.000000

(^CFVR and UCPGOV have been automatically dropped)

166

Two stage least square regression of CFVR = 0 + 1^COE + 2,3,4,5UCP + 6 ^SSVR + 7 ACI + 8 ACC + 9SIZE + 10BTMV + 10,11,12 Table 5.30(a) ABRES sample
Dependent Variable: CFVR Included observations: 141 after adjustments Variable C LGMV LBTMV PREDSSVR UCPMN UCPINST UCPFAM ACI ACC R-squared Adjusted R-squared Coefficient 0.659967 0.005764 0.075271 -0.028532 -0.322826 -0.075879 -0.086149 -0.071339 -0.124718 0.254124 0.208920 Std. Error 0.128311 0.013692 0.028405 0.023139 0.065730 0.165968 0.044820 0.084848 0.115976 F-statistic Prob(F-statistic) t-Statistic 5.143474 0.420968 2.649901 -1.233069 -4.911361 -0.457191 -1.922111 -0.840784 -1.075381 Prob. 0.0000 0.6745 0.0090 0.2197 0.0000 0.6483 0.0567 0.4020 0.2842 5.621643 0.000004

Table 5.30(b) ABSDATA/ABSDATCA sample

Dependent Variable: CFVR Included observations: 151 after adjustments Variable C PREDCOE UCPMN UCPINST UCPGOV UCPFAM PREDSSVR ACI ACC LGMV R-squared Adjusted R-squared Coefficient 0.353963 3.829805 -0.403098 -0.045916 -0.156878 -0.174957 -2.423920 -0.124357 -0.037661 0.037754 0.244085 0.195835 Std. Error 0.175945 2.669489 0.128100 0.178926 0.264480 0.124667 3.017809 0.112463 0.207832 0.026198 F-statistic Prob(F-statistic) t-Statistic 2.011779 1.434659 -3.146754 -0.256621 -0.593159 -1.403398 -0.803205 -1.105756 -0.181208 1.441102 Prob. 0.0461 0.1536 0.0020 0.7978 0.5540 0.1627 0.4232 0.2707 0.8565 0.1518 5.058775 0.000006

167

Summary of results The following table provides a summary of results of hypothesis testing. Table 5.4 Summary of results Hypothesis 1 Hypothesis statements There is a negative relationship between earnings quality and ownership structure Results Not supported for all earnings quality measures (ABRES, ABSDATA, ABSDATCA, PERS and PRED) and both ownership structure measures (CFVR and UCP)

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 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. 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.

There is a relationship between earnings quality and audit committee

There is a positive relationship between cost of equity and earnings quality

168

Table 5.4 Continued Hypothesis Hypothesis statements 5

Results

There is a negative Not supported for both measures of market relationship between assessment AVMR and YER market return and earnings quality There is a positive relationship between ex post market return AVMR and ex ante market assessment There is a relationship between cost of equity and ownership structure Not supported for both measures of market assessment AVMR and YER, and both measures of analyst assessment COE and COEA

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 8 There is a relationship between cost of equity and substantial shareholders

Results 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)

There is a relationship between cost of equity and audit committee There is a relationship between type of ownership and cash flow/ voting rights disparity.

Not supported in all samples and measures of analyst assessment and audit committee

10

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 12 There is a reverse relationship between cash flow/ voting rights disparity and the monitoring mechanisms.

Results 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 partys 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 OHara (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 companys information risk.

Previous researches (Jung and Kwon 2002, Koh 2003 and Chung et al 2004) associate substantial shareholders 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 shareholders 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 companys 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 selfregulation 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 selfregulating 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 partys 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 companys 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 shareholders 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 FamaFrench 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.249264 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 Asias 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 crosssectional determinants of earnings response coefficients. Journal of Accounting and Economics.11, p. 143181 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 OHara, 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.689731 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). Malaysias 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.225258. 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.105128. Leuz, C. and Verrechia, R.E. (2005). Firms 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.97110. 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 markets 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