Supervisory boards can make a difference

(1)

Supervisory boards can make a difference

A quantitative study of governance quality in

Dutch Higher Professional Education

MASTER THESIS

University of Groningen

Faculty of Business and Economics Business Administration

Specialisation: Organization and Management Control Thesis supervisor: dr. T.A. Marra

(2)

Master thesis P.I. Hagenbeek, S 1760912 University of Groningen December 2011 2

Preface

This thesis is written to conclude my master study Business Administration at the Faculty of Economics and Business of the University of Groningen. As a former lecturer and now manager in the sector of higher professional education in the Netherlands, the organization and quality of education is a subject that is very near my heart. For this thesis I chose the perspective of governance quality within the sector of higher professional education in the Netherlands.

In June 2008 the then minister of Education, Culture and Sciences, dr. R.A. Plasterk, introduced the scholarship for lecturers. This scholarship and the time and money granted by my employer, Stenden University of applied sciences, allowed me to successfully finish this study. For this, I wish to express my thanks.

I am very grateful for the support, the inspiration and the challenges I received from the lecturers at the University of Groningen during my education. Without exception, I experienced them as inspired, kind and intelligent professionals. My special thanks go to dr. T.A. Marra who supervised my thesis. He kept me on track and motivated me to go beyond of what I considered to be my frontiers.

Finally, my love goes to my men at home. For their unremitting patience, support and love.

Pauline Irene Hagenbeek S1760912

(3)

Table of content

1.0 Introduction 6

1.1 The theoretical framework 7

1.1.1 Educational governance 7

1.1.2 Primary stakeholders 7

1.1.3 The public cause 8

1.2 Moral hazard incidents 9

1.3 Governance defined 10

1.4 Governance in the HPE sector in the Netherlands 10

1.5 The role of supervisory boards 12

1.6 Problem statement 12

2.0 Accruals: a measure of good governance 14

2.1 Positive accounting theory: motives to engage in earnings management 14

2.2 Can IFHE’s engage in earnings management? 15

2.3 Accrual models 15

2.3.1 Healy model 16

2.3.2 Jones model and Modified Jones model 16

2.3.3 Teoh model 16

2.4 Conclusion 17

3.0 Hypothesis and research design 18

3.1 Hypotheses on board characteristics 18

3.1.1 Wrong monitor syndrome 18

3.1.2 Role ambiguity 19

3.1.3 Number of meetings 20

3.1.4 Board size 21

3.1.5 Busy boards 21

3.2 Data collection 22

3.3 Sample size and financial data 23

4.0 Results 25

4.1 Board composition & expertise 25

4.2 Independence of the board members and role ambiguity 25

4.3 Board meetings, Audit Committee and Remuneration Committee 25

4.4 Board size and organizational complexity 25

4.5 Number of directorships 27

4.6 Conclusions 27

5.0 Assessing the levels of accruals 28

5.1 NDcA and DcA 28

5.2 Correlation of accrual models 29

(4)

6.0 Hypothesis testing 31

6.1 Board expertise and DA levels 31

6.2 Role ambiguity and DA levels 31

6.3 Number of meetings and DA levels 31

6.4 Board size, organizational complexity and DA levels 33

6.5 Additional functions and DA levels 33

6.6 Hypotheses confirmed? 34

6.7 Jones’s models versus Teoh model 34

7.0 Additional analysis 35

7.1 Additional hypotheses tested 36

8.0 Conclusions 38

9.0 Discussion and recommendations for further research 39

Literature 41

Reports 42

Websites 43

(5)

Abstract

This thesis studies the relation between the quality of governance and characteristics of the supervisory boards in Higher Professional Education in the Netherlands. As a measure of governance quality the level of discretionary (current) accruals were used. Governance in the sector of higher professional education is placed in the theoretical framework of the agency theory. Strictly speaking the Minister of Education, Culture and Sciences nor students can be named as owners of an institute for higher education. In this study I show that the agency theory leaves ample room for denoting the Minister and the students as the principal, and the executive board as the agent. The supervisory board monitors on behalf of the principals.

(6)

1.0 Introduction

In the Law for Higher Education and Scientific Research (WHW) the government guarantees a well defined public interest for the entire higher education. The execution of higher education, of which higher professional education is a part, is delegated to institutes. The policy of the government is aimed at controlling ‘at arms lengths’. The government leaves it to the discretion of the executive boards of the institutes to set their own organizations and to distribute funds to organizational entities. A retreating government must be placed within the theory of New Public Management (NPM) (Verhoest, 2005: p. 236). Within the NPM theory the government no longer controls with the aid of input controls such as assigned budgets, rather it uses output and result controls, it uses financial instruments or introduces marketing activities. The basic thought of NPM is that output and result oriented control mechanisms ultimately lead to more efficient institutes for higher education; institutes that, to put it bluntly, ‘produce’ more graduates of a higher quality within a shorter time frame.

Within the Higher Professional Education in the Netherlands, the development of marketing and commercialization is encouraged, together with the tapping of so called third funds. The Higher Professional Education (from now on referred to as HPE) moves from the pure public domain to a mixture of the public and private domain. The size of private funding in the HPE-sector in 2009 was € 190 million, 6% of the total income (almost € 3 billion)1. A retreating government and the devel-opment of private initiatives within the HPE-sector make demands on management of and supervision on the educational organizations. The focus of the management changed from a single focus on the effective use of government funds to a focus on an effective and efficient use of funds. Not only ‘what’ is achieved, but also ‘how’ it is achieved, matters. This notion calls for an adapted governance system and in 2006 the HPE council2 published the governance code for the HPE sector. Its goal is to ‘make clear how institutes for higher education react to the higher demands that are made to the

management and supervision of the institutes’ (Branchecode governance, 2006: p. 4).

All institutes for higher education are organised similarly and have a two-tier board: an executive board and a supervisory board - in the law referred to as the supervisory board model. The structure and tasks of the executive board are laid down in article 10.2 of the WHW. The number of executive directors is three at most. The specifications with regard to the division of management and supervision are laid down in article 10.3.d: the supervisory board must be independent and its task is to supervise the board of directors in executing its tasks, to advice and to assist the board of directors. The minimum tasks of the supervisory board are included in appendix 1. As in many private organizations the ‘ownership’, management and supervision within an institute for higher education (from now on referred to as IFHE) are split.

1

In table 4, page 19 the financial data of the HPE-sector are summarized.

2

(7)

1.1 The theoretical framework

The agency theory describes the difficult relation between agent and principal whenever ownership and management are split. The agency theory is applicable to any private corporation where shareholders are the owners and the where the board of directors looks after the interests entrusted to them, i.e. the pursuit of value maximization on behalf of the shareholders. The chain of governance of an institute for higher education looks exactly like that of a private company, but there are two main differences. Contrary to private organizations, there is no party that owns the IFHE. The minister and the students are important stakeholder groups, but do not own the organization. And although the IFHE’s do make profit, their organizational goal is not to pursue profit maximization. Whether or not the agency theory is applicable to non profit organizations is subject of discussion. A literature review of this discussion is outlined in the next paragraphs.

1.1.1 Educational governance

Strikwerda is an important contributor to the discussion of governance in the educational sector in the Netherlands. He argues that looking at educational governance from a perspective of the market mechanism and corporate governance is too limited. He developed principles for educational governance in which he states that ‘educational governance must find its own way which cannot be

copied from corporate governance’ (Strikwerda, 2005: p. 21). He states as well that if and when the supervisory board model is defined by law, the principles of the code Tabaksblat can be applied to these supervisory boards, which implies that the code Tabaksblat could be applied to public organizations. An important notion of Strikwerda cannot be left unmentioned in the discussion of corporate government principles in the educational sector and this refers to the difference between

Wertrationalität and Zweckrationalität. ‘The question we must ask is whether an institute can be

governed along the same principles as an organization that operates under the mechanism of the market economy. The answer is no: institutes concentrate on thick values like truth and ethics (Wertrationalität) Production organizations concentrate on the thin values efficiency and effectiveness (Zweckrationalität)’ (Strikwerda, 2005: p. 10). This implies that whenever choices have to be made, a hierarchy of values exists. In the discussion about governance in the HPE sector, this hierarchy of values is not always made explicit nor is it clear and explicit which values prevail. New Public Management assumes the prevalence of thin values.

1.1.2 Primary stakeholders

Speckbacher (2008) treats the issue from a stakeholders’ point of view. He argues that the financial view of corporate governance has been fruitful but reduces the governance problem to the agency conflict between owners and managers, which he describes as a ‘somewhat simplistic view of

(8)

Master thesis P.I. Hagenbeek, S 1760912 University of Groningen December 2011 8 absent, contractual governance is applicable to IFHE’s. The rights and duties of the minister, of students and of teaching staff are indeed covered in the WHW, the Law for Higher Education and Scientific Research.

He continues with the notion that as long as parties operate on a quid pro quo basis (an equal exchange basis) no substantial governance problem arises. However, in complex economic relationships where it is not possible to define exactly what the contribution of each party is (the so-called contractual incompleteness (Speckbacher, 2008: p. 301)), substantial governance problems do arise. In his discussion of governance mechanisms he reasons that ‘control tasks should be fulfilled by

representatives of primary stakeholders, while management tasks are fulfilled by professional managers’ (Speckbacher, 2008: p. 311) This implies for an IFHE that the supervisory board should act on behalf of the Minister and students. The issue of the execution of control tasks by a one tier or a two tier board is left undecided but whichever system is chosen ‘primary stakeholders are the

principals and they delegate decision and control rights to boards and managers’ (Speckbacher, 2008: p. 311). This means that the principal-agent approach is applicable when studying the governance within the HPE sector.

1.1.3 The public cause

Hoek (2007) describes the problem between the government and organizations that execute government policies. If one hands over the execution of a public goal to a third party – and that is the situation that has been defined for the HPE sector in the WHW – than a contract must be made with that party. Rules and regulations must ensure that the executing party is accountable and supervision on the execution is necessary. He concludes that new public management (NPM) leads to an increased need of public governance (Hoek, 2007: p. 95). Hoek notes that there are two different principles when it comes to the content and arrangement of public governance.

The first principle is derived from the private sector and is based on the idea that governance principles from the private sector can be applied to the domain where public goals are realized. This line of thinking is in line with the principle of NPM where output and result controls are used to steer and control the execution of government policy. Public governance can than be studied from the point of view of the agency theory where the government is the principal and the executing organization, the institute for higher education, is the agent. Information asymmetry and goal conflicts exist and the agent will pursue his private goal and will not always act in the best interest of the principal. In other words: moral hazard incidents do occur. Governance mechanisms are put in place to reduce infor-mation asymmetry and to align the goals of both the agent and the principal.

(9)

Master thesis P.I. Hagenbeek, S 1760912 University of Groningen December 2011 9 of the market mechanism is important and new external monitoring parties arise. An example is the Dutch telecom market. This was privatized in the ’90-ties of the last century and since 2006 the state is no longer a shareholder. The monitoring party is the OPTA which is a non governmental public body (NPB).

Analogous to these approaches, the role of the government in public governance can be regarded in different ways. One approach has as a central notion that the government has the primacy in public governance: the government is the centre of power and only she guarantees the ‘public cause’ She is the principal and uses a series of instruments (laws and regulations, supervision, disclosure and accountability) to steer the behavior of the agent. This theorem is leading in the Netherlands (Hoek, 2007: p. 107)

1.2 Moral hazard incidents

Although the discussion on the applicability of the agency theory has different perspectives, it gives enough arguments to use the agency theory as a framework to study the relation between a government at arms length and the management of an institute for higher education that acts on behalf of this government. The central issue within the agency theory is how the principal can control the agent in a situation of information asymmetry and goal conflict. Information asymmetry gives the opportunity and the goal conflict provides the motive for moral hazard and adverse selection incidents. These incidents do occur within the HPE-sector. In 2004 the Commissie Schutte investigated possible fraudulent behaviour of Dutch institutes for higher education with regard to the defrayment of (international) students. The report contained 17 conclusions and recommendations about the ways of defrayment, about the central policy of the Ministry of Education, Culture and Sciences, about procedures and internals controls applied by institutes, about the tension between public responsibility and private initiatives undertaken by IFHE’s and about the sensitivity to fraud of the defrayment system. The commissie Schutte concluded in its report that ‘onderwijsland is geen fraudeland3’ (Schutte, 2004: p. 86). There is no evidence of systematic fraudulent behaviour, yet moral hazard incidents do occur regularly.

Since 2004 new incidents have occurred: Saxion Hogescholen unlawfully granted Master of Arts titles to master students, and during the summer of 2010 Hogeschool INHolland made the headlines with ‘alternative graduation courses’ for those students who exceeded the nominal duration of study of 4 years. The Commissie Leers conducted an initial investigation in 2010 and subsequently the

Onderwijs-inspectie4 investigated 15 educations of 10 different institutes of higher education5 during the winter and spring of 2011. The Education Inspection concluded that there is insufficient discipline within the HPE sector to abide by legal regulations that should grant the graduation level. Furthermore

3

‘The world of education is not by definition a fraudulent world’.

4

Education Inspection

5

(10)

Master thesis P.I. Hagenbeek, S 1760912 University of Groningen December 2011 10 it marked the quality of four educations as very weak and that of four other educations as ‘disturbing’. As a third point the inspection concluded that there are gaps in the system of external quality assurance; more specifically there are insufficient means to detect weak educations in time. The time in between two accreditations leaves room for ‘quality risks and quality failure’ (Steur, 2011: p. 5). If and when moral hazard incidents occur, the quality of governance within a sector or institute is at stake.

1.3 Governance defined

Governance or corporate governance has many definitions. In this thesis governance is studied within the field of higher professional education in the Netherlands. The institutes for higher education are Non Departmental Public Bodies (NDPB’s)6, Therefore I choose to work with a description made for the field of public organizations and specifically the one made by Hoek (Hoek, 2007: p. XV) “Governance is a set of rules and manners for administration, supervision and accountability. These

rules and manners must ensure that organizations and persons, to the best of their abilities, look after the interests entrusted to them.” Mechanisms of governance can be divided in internal and external mechanisms. Internal mechanisms comprise the ownership structure, internal controls, board characteristics, and remuneration variables. External mechanisms are codes and laws and to a lesser extent the monitoring role of financial institutes.

1.4 Governance in the HPE sector in the Netherlands

As the principal in the chain of governance, the government has various means at hand to reduce information asymmetry and goal conflict. It can put in force monitoring systems that measure and evaluate the performance and skills of the agent. Monitoring in the HPE sector takes place via the supervisory boards of each institute for higher education and via the Education Inspection that can initiate its own investigation next to the investigation of the annual reports. Although, as sketched in the introduction, the government wishes to operate at arms’ lengths, it still uses behaviour-based contracts. The WHW and subsequently the ‘Directive annual reports in Education’7 gives very detailed instructions with regard to drafting and publishing the financial reports of the institute for higher education. Lastly, the government uses risk transferring financial systems to transfer the risk from the principal to the agent. Within the HPE-sector this is done by means of paying a graduation fee to the institutes. The more the student exceeds the nominal duration of study of 4 years, the lower this fee is. From a financial point of view the institute for higher education gains when its students graduate as quickly as possible.

In 2005 the Netherlands Court of Audit (NCA) published an overview of instruments to promote broad - internal, vertical and horizontal - accountability of NDPB’s in the Netherlands. The NCA

6

Netherlands Court of Audit 2005: p71

7

(11)

Master thesis P.I. Hagenbeek, S 1760912 University of Groningen December 2011 11 recommended that the government should develop a framework on the supervision and monitoring of NDPB’s. Now in 2011, the WHW and the sector code HBO 2006 are in force and the Education Inspection acts as the prime monitoring body. The NCA is in favour of the usage of performance contracts, at the same time it recognizes the risk that performance contracts may lead to “potential

perverse effects”. IFHE’s are partly financed on the basis of input, namely number of students, and partly financed on the basis of output or performance. The output-based financing has three components: number of graduates, number of drop-outs and speed of graduation. IFHE’s are ‘rewarded’ for every student that graduates after four or five years and are ‘fined’ for those students who graduate after a study period of 6 or more years. IFHE’s are also fined for drop-outs and the fine gets higher with every additional year of study. (The fine for a dropout in the 3rd year is € 4.500,-; the fine for a dropout in the 7th year is € 32.100,-8.)

The NCA recommended9 the use of third party audits of financial statements by an accountant. For the HPE sector, the content of financial statement is regulated in the “Richtlijn Jaarverslag Onderwijs

2008”. The external accountant is appointed by the supervisory boards. According to the NCA the issuance of annual reports by NDPB’s and the quality of these reports has increased over the years to 2005. The vast majority of IFHE’s issued an annual report in 2008 and 2009. The instrument of external reviews and certification for the HPE sector is executed by the NVAO10. Its scope and responsibilities are laid down in the law (WHW H5a art. 5.1.1 – 51.16). An example of the instrument of voluntary benchmarking in the HPE sector is the national student survey11. The HPE council is one of the initiating parties. As a last item the NCA refers to more or less formal ways of communication with stakeholders. For IFHE’s the communication with internal stakeholders is formalized in the WHW and organized on three levels. The institutes’ co-determination council

(medezeggen-schapsraad) with whom the supervisory board meets at least twice a year; the co-determination council of a faculty or other organizational unit (the so-called deelraad) and the educational committee (de opleidingscommissie) which is organized per education and consists of students and teaching staff. The educational committee consults with the management of the education.

In 2006 the HPE council published the sector code governance for the HPE sector. Following the code Tabaksblat the sector code contains general principles on governance that are further defined into specific mechanisms and controls. The institute for higher education should ‘comply or explain’ in their annual report as to why and how they follow the governance code. The code contains four sections. The first two cover principles about the executive board and the supervisory board. The third section contains principles about the audit of financial reporting, the role and position of internal

8

Pricelevel 2007. An example calculation is added in appendix 3

9

The NCA report comprised 3.200 NDPB’s for some of which apparently audits of financial statements by accountants are not mandatory.

10

Nederlands Vlaamse Accreditatie Organisatie

11

(12)

Master thesis P.I. Hagenbeek, S 1760912 University of Groningen December 2011 12 control and the role of the external accountant. The last section covers principles with regard to the horizontal dialogue, i.e. the dialogue with relevant internal and external stakeholders.

1.5 The role of supervisory boards

In 2005 Hoek, van Montfort and Vermeer published the article: “Enhancing Public Accountability in

the Netherlands” in which they criticize the functioning of supervisory boards of non departmental public bodies. They state that members of a supervisory board must be able to act independently, yet at the same time, members can represent a specific stakeholder group or their election can be brought up by a stakeholder group. E.g., a co-determination council of an institute for higher education has the right to nominate a candidate for the supervisory board.

The supervisory board of a corporation serves the interest of the shareholders. Such a clear framework is missing for supervisory boards in the HPE-sector. The sector code governance states: “in executing

its task the supervisory board is lead by the interest of the institute en weighs all relevant interests from all stakeholders of the institute..” In principle the minister has the authority to monitor the work of the supervisory boards. She can put the supervisory boards in a position in which they can play an important role in the system of vertical supervision. Hoek, van Montfort and Vermeer (2005) conclude that this does not happen systematically. In 2005 the Education Inspection published its report of its investigation into the functioning of supervisory boards in the HPE sector. The inspection concluded that ‘...when the government retreats, other instruments and actors for example supervisory boards

must be put into position....’ She concluded as well that the HPE-sector ‘lacks a clear guarantee for

the quality, the continuity and the functioning of the internal supervision’. Specifically the inspection discussed the aspects related to the supervisory board: composition, appointment, authority, tasks, roles and procedures and accountability. Precisely this internal supervision could be an important tool in managing the governance quality of institutes for higher education.

1.6 Problem statement

(13)

Master thesis P.I. Hagenbeek, S 1760912 University of Groningen December 2011 13 In this thesis I focus on the role of the supervisory board as governance mechanism within the HPE sector with the leading problem statement :

(14)

2.0 Accruals: a measure of good governance

An extensive line of research exists on the relation between corporate governance mechanisms and accruals. Kent et al. (2010) build on prior research of Dechow et al. (1995), Klein (2002), Davidson et al. 2005 and Koh et al. (2007) in researching ‘whether and how they (accruals) can be migitated by

sound governance’ (Kent, 2010: p. 172). Kents’ analysis shows that sound governance is generally associated with a higher quality of both discretionary accruals and innate (non discretionary) accruals. Larcker et al. (Larcker, 2007: p. 984) also make the assumption that indices of good governance relate negatively to abnormal accruals. Accruals are generally defined as the difference between net profit and cash flow from operations (Healy, 1985 and Scott: 2011, p 424). Accruals can be split in discretionary and non-discretionary accruals. The latter are the accruals that cannot be influenced by the manager. Discretionary accruals are adjustments to cash flows selected by the manager and by managing the size of discretionary accruals, the management can influence the size of the net profit of a company. This is called earnings management. Earnings management in itself is not per se fraudulent behaviour but there is only a fine line between earnings management and earnings mismanagement.

Scott (2011) defines earnings management as ‘the choice of accounting policies, or real actions,

affecting earnings so as to achieve some specific reported earnings objective’. Earnings can be affected negatively: ‘taking a bath’ or ‘income minimization’. Income minimization may be done by firms under political scrutiny. Proof of that is given by Jones (1991). She reports that ‘.. managers

make income-decreasing accruals during import relief investigations. Discretionary accruals are more income-decreasing during the year the ITC completed its investigation (year 0) than would otherwise be expected.’ (Jones, 1991: p. 223). Earnings can affected positively as well e.g. for bonus purposes or to avoid violation of a debt covenant. Reasons for income smoothing or the avoidance of income volatility are relatively stable bonuses, the prevention of debt covenant violation, reduction of the risk of being fired or to signal a persistent earning power to the capital market.

2.1 Positive accounting theory: motives to engage in earnings management

(15)

Master thesis P.I. Hagenbeek, S 1760912 University of Groningen December 2011 15 High or excessive profits can indicate to the minister that either the budgets provided by the minister were too high or that the IFHE did not spend enough of the budget for the education of the students. There is no reason to assume that IFHE’s are not subject to debt covenants. Neither is there a reason to assume that persistent negative financial results will be without consequences for the members of the executive board. Earnings management for financial bonus purposes does not seem to be a valid reason. In 2009 five out of the 31 institutes in the survey applied a moderate form of bonus based compensation for the executive board. Two of those five reported that the bonus was handed back to the organization. In two other cases the bonus was so low it seems to have a ritual rather than a financial incentive function (6% and less than 0.7% of the fixed salary respectively). Implicit bonuses like status or (political) power were not taken into account in this study. On average the HPE sector reported an EBIT of € 950.000,- in 2008 and € 1.800.000,- in 2009. This equates to a return on sales of 1% and 2% respectively.

2.2 Can IFHE’s engage in earnings management?

The ‘Directive annual report in Education’ gives freedom of choice in accounting policies and depreciation. Assets may be valued at cost or at current value and may be depreciated straight line or parallel to value decrease. Once chosen the depreciation method must be used consistently. Depreciation expenses are generally considered to be non-discretionary. An increase in accounts receivable is discretionary when caused by e.g. more relaxed credit terms. They are considered non-discretionary when the change is caused by an increase in business. As IFHE’s belong to the service industry, deliberate changes in inventory such as ‘channel stuffing’ or ‘manufacturing for stock only’ are hardly applicable. Real actions that manage cash flows and real earnings, and that are available to an IFHE are: advertising expenditures, R&D, maintenance, timing of purchases and disposals of capital assets. It is fair to conclude that any IFHE has the possibility to engage in earnings manage-ement although this may be to a lesser extent than in private corporations,

2.3 Accrual models

Kent et al. (2010) measure accrual quality by the use of the Dechow Dichey model (DD), and the McNichols model. In the DD model: ∆WCt= β0 + β1 * CFOt-1 + β2 * CFOt +β3* CFOt+1+ εt the standard deviation of the regression residual εt is the measure of accruals quality: the higher the standard deviation, the higher the accruals estimation error and the lower the accruals quality. The McNichols model is extended with change in revenue (∆ REV) and property, plant and equipment: ∆WCt= β0 + β1 * CFOt-1 + β2 * CFOt +β3* CFOt+1+ β4* ∆REV t + β5* PPEt + εt

(16)

Master thesis P.I. Hagenbeek, S 1760912 University of Groningen December 2011 16 detect earnings management. In this study they conclude that ‘..all models, except the DeAngelo

model, are successful in discriminating between firms that manage earnings and firms that do not.’ (Bartov et al, 2001: p. 450). Another approach is taken by Teoh et al. (1998 and 1998b). They focus on current accruals only because current accruals are easier to manipulate (Xie et al., 2003: p.301). Underneath I will discuss the Healy and Jones models as well as the Teoh model.

2.3.1 Healy model

The Healy model is a time series model and reads:

NDAt = 1/nΣt(TA t/ A t-1), where n represents the number of years from the estimation period, TA stands for Total Accruals and A is Total Assets. For this thesis the balance sheet data from 2007, 2008 and 2009 are available. So the equation would become:

NDA (Non Discretionary Accruals)2009= ½ [(TA 2009/ A 2008) +(TA 2008/ A 2007)] DA (Discretionary Accruals)2009 = (TA 2009/ A 2008 )– NDA2009

In his study of 1984 Healy used data from 16 consecutive years. Bartov et al. (2001) used data from 17 consecutive years. In this study, I can use data from three consecutive years. Therefore I conclude that I have insufficient data to effectively use the Healy model. This was confirmed in the early analyses where the outcomes of the analyses with the Healy model were not in line with the Jones, Modified Jones and Teoh models.

2.3.2 Jones model and Modified Jones model

The Jones model and the modified Jones model include the change in revenue (∆REV) and the fixed tangible assets (PPE). Both are scaled by lagged total assets (At-1). The equation for non discretionary accruals is: NDAt = α1 ( 1/ A t-1) + α2 (∆REV t / A t-1 ) + α3(PPE t / A t-1 ). The coefficients α1, α2 and α3 are calculated with the equation: TAt/At-1= α1 ( 1/ A t-1) + α2 (∆REV t / A t-1 ) + α3(PPE t / A t-1 ) + ε. Subsequently the discretionary accruals in yeart are calculated as follows: DAt = TAt/At-1 - NDAt = ε. The modified model corrects the change in revenue with the change in accounts receivable in order to establish the realised increased revenues rather than the reported.

The modified Jones model reads NDAt = α1 ( 1/ A t-1) + α2 ((∆REV t - ∆AR t ) / A t-1 ) + α3(PPE t / A t-1 ). The coefficients α1, α2 and α3 in this model are taken from the original Jones model. The data availability is this study is such that both Jones models are cross sectional models.

2.3.3 Teoh model

The Teoh model uses three equations to split current accruals in year t12: CAt / A t-1= α0 ( 1/ A t-1) + α1 (∆REV t / A t-1 ) + ε (1)

The coefficient α0 and α1 are used to calculate the non discretionary accruals for a specific firm: NDcAt = α0 ( 1/ A t-1) + α1 ((∆REV t - ∆AR t ) / A t-1 )) (2)

12

(17)

Master thesis P.I. Hagenbeek, S 1760912 University of Groningen December 2011 17 The discretionary accruals then become:

DcAt = CAt / A t-1 - NDcAt (3)

Where CA stands for current accruals. ∆REV t is the change in revenues from year t-1 to year t, ∆AR t is the change is accounts receivable from year t-1 to year t, A t-1 represents Total Assets, DcA are discretionary current accruals and NDcA are the non-discretionary current accruals in year t-1. Teoh et al. concentrate on short term – current - accruals. It is fair to assume that the yearly changes of PPE and thus the changes in depreciation on PPE are relatively small in the HPE sector. Therefore, the change in the level of total accruals would be mainly caused by short term or current accruals. Data on the level of accruals confirm this: the mean total accruals dropped from 2008 tot 2009 with just over € 250.000. Over the same period the mean current accruals dropped with € 600.000,- (see page 24). Xie et al. (2003) use the Teoh model in their paper in which they study the relation between board characteristics and earnings management. They repeat the analysis using Jones’s long term model and reports that the results are qualitatively unchanged but with lower statistical significance.

2.4 Conclusion

Although the majority of the literature is based on surveys in the for profit sector, earnings management is not necessarily limited to the for profit sector only. There is no reason why earnings management would not take place in the non profit sector such as the HPE sector in the Netherlands. Therefore also in the HPE sector accruals can be used as a measure of governance quality. In this study I use the Jones, Modified Jones and Teoh model to assess the size of discretionary and non-discretionary accruals in the HPE sector so that the research model becomes:

(18)

3.0 Hypothesis and research design

In order to answer the problem statement ‘how do characteristics of the supervisory board relate to

the governance quality within an institute for higher education in the Netherlands and which are the areas that are open to improvement?’ research hypotheses were drawn up.

Literature on corporate governance in the private sector often relates governance mechanisms to company performance. I.e. it does not qualify the governance, it rather links governance dimensions to company performance. Based on the direction of the correlation a governance mechanism either contributes to company performance or not. As a measure for the latter, Tobin’s Q is often used. An example of such a study is the one by Larcker et al. (2007). Company performance as such is a rather useless measure in the HPE sector. An IFHE has no market value of equity. And in essence the desired profit level in a given year is nil. Some IFHE’s actually state this as their profit objective in their annual report. While research on governance mechanisms in Anglo-Saxon, private companies is available in abundance, little data are available on governance mechanisms and governance quality in IFHE’s in the Netherlands. For the lack of a better measure the starting point for this study is the literature on boards of directors and on supervisory boards in the private domain which can give clues about relevant characteristics of boards in the HPE-sector. Earnings management is measured by the level of discretionary current accruals. Low levels of D(c)A indicate the absence of earnings management and are therefore a sign of good governance. As stated in chapter two, this is based on literature of Kent et al (2010) and Larcker et al (2007).

3.1 Hypotheses on board characteristics

3.1.1 Wrong monitor syndrome

The literature (Tirole, 2006: p. 336) gives various reasons why monitoring through incumbents has its limitations. Tirole mentions the ‘wrong monitor syndrome’: he argues that is not always possible to determine ex ante whether or not a candidate possesses the knowledge and skills needed to be a suitable member of a supervisory board. Therefore it is not possible to determine whether or not a monitor – or in this thesis, a member of a supervisory board – possesses the knowledge and skills to supervise the board of directors of an IFHE. In 2005 The Education Inspection found as members of a supervisory board of an IFHE ‘... a common denominator of financial, business administrative and

public administrative expertise.’ (Inspectie van het onderwijs, 2005: p. 8). The WHW prescribes that the supervisory board approves the budget, the annual accounts, the annual report and the institutional plan (art. 10.3d); but it should also supervise the design of the system of quality management and in general terms supervise the effective and efficient expense of public funds. Principle III, item 3.2 of the sector code 2006 states that at least one board member is a financial expert.

(19)

Master thesis P.I. Hagenbeek, S 1760912 University of Groningen December 2011 19 The presence of sector specific knowledge and skills in the board is emphasized by the findings of Hau and Thum (Hau et al., 2009: p. 741). The presence of financial and management expertise in the board was directly related to the size of losses suffered by German banks during the financial crisis of 2008 and 2009.

Hence, in order to execute its job properly the supervisory board of an IFHE should encompass members with a financial and/or accountancy background and educational or administrative educa-tional background and public administrative background. This leads to my first hypothesis.

H1: to the extent that the supervisory board consists of members with financial,

educational and public administrative expertise, the governance quality is better.

The WHW (art. 10.3d lid 4) and the sector code (principle III.2) require the supervisory board members to be independent. One of the members may be nominated by the codetermination council. To measure independence I check whether or not supervisory board members explicitly state that they represent a stakeholder group. If this is not the case, independence is assumed.

3.1.2 Role ambiguity

A supervisory board of an institute for higher education supervises the executive board of that institute on behalf of a third party. There are several opinions on who this third party is or should be. The WHW and the sector code do not give any information on this issue. The WHW only states that the supervisory board accounts for her tasks in the annual report of the institute for higher education (WHW, art. 10 3d. lid 2i). The sector code HPE clearly mentions that the supervisory board must focus on the interests of the institute for higher education. Unlike John Carver who states in his policy governance® model that the supervisory board represents the owner, and must serve his or her interests and not the interests of the organization. He states that in the case of an institute for higher education, the supervisory board represents the minister and it should serve her interest: ‘governing is

ownership one step down, not management one step up’ (Carver, 2007: p.1035).

In 2005 the Education Inspection criticizes the absence of a clearly defined party on behalf of whom the supervisory board executes its tasks and the inspection concludes: ‘.. the accountability is not

concrete and in the absence of a clear assessment framework there is only limited insight into how the control is implemented." Members of the supervisory board in turn state that the feedback by the minister on their account in the annual report is very limited, which causes the account to be ritual rather than a fundamental contribution to good governance (Onderwijsinspectie, 2005: p. 31)

(20)

Master thesis P.I. Hagenbeek, S 1760912 University of Groningen December 2011 20 ambiguity is negatively related to the extent to which board members can focus their time and attention to the desired activities or are to able evaluate and adapt their own work properly13.

Because there is not a clear party to whom the supervisory board is accountable, role ambiguity exist which leads to the second hypothesis:

H2: To the extent role ambiguity exists, the governance quality is worse.

The extent of role ambiguity is measured by whether or not supervision on behalf of the minister or another party is mentioned in the annual report.

3.1.3 Number of meetings

An extensive research string exists on corporate governance mechanisms, board characteristics and their relation with company performance. Larcker et al. (2007) reported on 14 different dimensions of corporate governance, among others the number of meetings. They reported an average number of board meetings, of audit committee meetings and of the remunerations committee meetings of 7,31, 6,33 and 3,92 respectively (Larcker: 2007, p. 970). Koh finds an average of 11,15 board meetings and 2,85 AC meetings respectively (Koh: 2007, p. 315). Koh includes the number of meetings of the audit committee in assessing audit-related governance mechanisms. Hoitasch et al. (2009) discuss the number of board meetings and audit committee meetings. More meetings can be a sign of a more vigilant board. But an increase in the number of meeting can also be the result of incidents, i.e. governance incidents occurred and as a result of that the board decides to increase its number of meetings.

The law states that the supervisory board meets at least two times per year with the codetermination council of the IFHE. The sector code requires one meeting with the executive board, in which the remuneration is determined, and at least one meeting with the external accountant. The number of meetings of an audit committee or a remuneration committee is not mentioned in the WHW nor in the sector code. For the supervisory board nine different tasks are defined in the WHW. The literature is indecisive; the code, the law and common sense lead to a minimum of four meetings per year. The above leads to the hypotheses:

H 3a: The number of board meetings is at least four per year

H 3b: The number of board meetings, AC meetings and RC meetings is related to governance

quality.

13

(21)

3.1.4 Board size

Research into board size shows contradictory results. Yermack (1996) reports a positive correlation between a small board and financial performance. The rationale is that small boards are more vigilant in their supervision than big boards. Adams et al. argue that based on their ‘bargaining based model’ a positive relation should exist between the board size and the financial performance. (Adams, Hermalin & Weisbach, 2010: p. 73). The underlying rationale is that good CEO’s know that small boards are more vigilant so they will bargain to increase the size of their boards. The size of a supervisory board is prescribed neither in the governance code nor in the WHW. The Education Inspection (2005: p. 19) acknowledges a lack of clarity around the issue of board size. She speaks of an optimum of seven members but also acknowledges that the board size should be related to organizational complexity. To avoid a deadlock in votes, the number of member should be odd.

Rather than assessing the actual size, Coles et al. (2008) display the relation between board size and complexity of the company. Complex companies profit from big boards. The rationale is that complex companies have increased advising requirements which are provided by big boards rather than by small boards. Company complexity is measured by company size, extent of diversification and the D/E ratio. IFHE’s in the Netherlands differ in size and complexity. Based on Coles’ measure of organizational complexity14 I define the complexity of an IFHE as: ln(revenues)+ #domestic entities+ #entities abroad+ #commercial entities + D/E ratio. The numbers of entities are a proxy for diversification. The fourth hypothesis reads:

H4: the size of the supervisory board of an IFHE is positively related to the complexity of an

IFHE.

3.1.5 Busy boards

Fich and Shivdasani researched the relationship between busy boards and company performance. They found that a company with board members that hold more directorships (>3 directorships) has a ‘..lower market to book value, has weaker profitability and lower CEO turnover sensitivity to firm

performance (Fich and Shivdasani, 2006: p. 689). This finding is in line with the ‘busyness view’ that says that busy directors lack time and attention to perform their job adequately. Other scholars reason that the number of directorships held by a member of the supervisory board signals quality and claim that busy directors make up better boards. The fact that they seat in different boards adds to their monitoring experience and skills. The latter is referred to as the ‘reputational view’. Both the busyness view and the reputational view are described by Hooghiemstra and Marra (2011). The sector code nor the Dutch governance code mention a limitation to the number of additional functions. They do require that additional functions of members of the supervisory board are published. The contradicting

14

(22)

Master thesis P.I. Hagenbeek, S 1760912 University of Groningen December 2011 22 results in the literature on governance quality in relation to board busyness are the base for the fifth hypothesis:

H5: board busyness is related to governance quality.

Busyness can be defined in different ways. Either by the mean number of additional directorships per board member, or by calculating the fraction of busy and non-busy members per board. Both ways will be taken into account. The table underneath sums up the hypotheses and the expected signs.

Board characteristics Expected sign

1. board expertise (-) 2. role ambiguity (+) 3. number of meetings (+/-) 4. board size and organization

complexity

(+)

5. board busyness (+/-)

Table 1 Hypotheses

3.2 Data collection

The hypotheses are tested using data from annual reports of 2007, 2008 and 2009 of 31 IFHE’s in the Netherlands. All data are hand collected. The 2009 annual reports were downloaded from the HBO council website. Usually the annual accounts contain year-on-year data so a lot of 2008 financial data could be obtained from the 2009 reports. When these were not available I downloaded the 2008 reports from the site of the specific IFHE. If digital information was not available I asked the IFHE to provide me with a hard copy the 2008 and/or 2007 annual report. All IFHE’s provided their cooperation. One IFHE did not release the 2008 annual report. For that institute, the necessary financial data for 2008 were approximated by comparing the end 2007 balance (= opening balance 2008) to the 2009 opening balance (= closing balance 2008). Annual reports from 2006 and earlier are hardly available – that is to say: in digital form.

(23)

Increase Decrease Example HS Utrecht15

EBIT + 6.037.000 ∆ Provisions + - + 2.154.000 ∆ Inventory - + - 71.000 ∆ Accounts receivable - + - 2.538.000 ∆ Accounts payable + - + 2.052.000 Depreciation + + 15.120.000

Results interest paid -/ received +

- 2.000.000 Cash flow from

operations

+ 20.754.000

Table 2 Example cash flow calculation

This calculation is in accordance with the calculation prescribed by the ministry of education16. It is mandatory that a cash flow statement is included in the financial statements so one wonders why eleven institutes do not include this in their annual reports. An explanation could be that the cash flow statement was included in the report submitted to the minister but was excluded from the reports published for broader stakeholder groups.

Subsequently total accruals were calculated using the equation from Scott (2011): total accruals = cash flow from operations +/-net profit. Following Xie et al. (2003) current accruals are defined as: ∆ Inventory +/- ∆ Accounts receivable +/- ∆ Accounts payable. Total assets and plant, property and equipment data were taken from the balance sheet. The D/E ratio was calculated with the use of data from the balance sheet: Debt = long term debt + short term debt + provisions and Equity = equity + general and specific reserves. In order to obtain the data on board characteristics all annual reports were systematically hand searched with the aid of a list of keywords. The list is included in appendix 6. If necessary, data were supplemented with public information from websites. This concerns specifically data on the expertise and additional functions of the supervisory board members. Data on the average years of study are available from the HBO council.

3.3 Sample size and financial data

In 2009 the Netherlands counted 39 institutes for higher education (see appendix 7); of eight institutes data were not available, which results in a net sample of 31 institutes.

Gross sample size 39

Included in consolidated data mother organization

2

In process of merger 1

No financial statement or annual report available or downloadable

3 Insufficient detailed information in financial

statement

2

Net sample size 31

Table 3 Gross and net sample size

15

Financial statement of HS Utrecht in appendix 5

16

(24)

Master thesis P.I. Hagenbeek, S 1760912 University of Groningen December 2011 24 The table underneath displays the main financial data from the sample.

2009 n = 31 ∆ 2008 n = 31 Mean revenues € 95.499.392 +6% € 90.283.270 Median € 53.008.000 +6% € 49.959.000 % Government funding 70% 69% % Tuition fees 18% 17,5% % Commissioned activities 6% 7% % other (overig) 6% 6,5% Mean EBIT € 1.814.598 +90% € 954.511

Mean cash flow from operations

€ 8.279.364 +8% € 7.674.031

Mean total accruals € 6.464.766 -4% € 6.719.519 Mean current accruals € 1.334.700 -31% € 1.945.550 Mean total assets € 93.932.894 +10% € 85.498.531

Mean PPE € 68.624.015 +7% € 64.340.782

(25)

4.0 Results

In this chapter I present and evaluate the descriptive statistics of the board characteristics.

4.1 Board composition & expertise

Contrary to expectations almost half of the boards do not have members with educational expertise, and one third of the boards lack specific knowledge in public administration. Almost a quarter of the boards function without a member with an accountancy or financial background. The latter violates principle III, item 3.2 of the sector code which states that at least one member is a financial expert.

4.2 Independence of the board members and role ambiguity

These results are in accordance with principle III item 2 of the sector code which requires supervisory board members to be independent. It remains unclear on behalf of whom the supervisory board conducts its supervision. Only one supervisory board mentions in her report on behalf of whom she executes her task. Hypotheses three predicts a negative effect of role ambiguity (the absence of a clear party to whom you are responsible) on governance quality.

4.3 Board meetings, Audit Committee and Remuneration Committee

The number of board meetings must be published. In four cases this was not done. An audit or remuneration committee (AC resp. RC) are not mandatory. The sector code leaves it to the discretion of the supervisory board to establish those committees. However, if they do exist, they are obliged to mention the number of meetings of the committees in the annual report (sector code, principle III.5). This principle is violated in most cases. In 10 cases (50%) no meetings of the audit committee were reported; in 13 cases no meetings of the remuneration committee were reported (68%). The number of reported meetings (5,2 meetings) is below the numbers found by Larcker et al. and Koh. An explanation of this difference could lie in the fact that both Larcker and Koh measured the number of meetings of a one-tier board whereas I measure the number of meetings of the supervisory board only. The average number of meetings is above the predicted minimum of four. In one case, only three meetings were reported. Audit and remuneration committees are fairly new in the HPE sector – the sector code and the possibility of installing an AC and RC only exist since 2006. This relative novelty could be an explanation for the low frequency of meeting and for the sloppy way of reporting the frequency of meetings in the annual report.

4.4 Board size and organizational complexity

(26)

Table 5 Board descriptives - 1 -

N = 31 %

Expertise Dummy FinAC 24 77%

Dummy EDU 16 52%

Dummy PUB 21 68%

Independence #not repr stakeholder 31 100%

#repr. stakeholder 1 3.2%

Ambiguity Superv. minister 0 0%

Superv. other 1 3.2%

Superv. not ment. 30 96.8%

Meetings Supervisory board meetings reported 27 87%

Dummy AC 20 65%

Dummy RC 19 61%

Reported meetings audit committee (n=20)

10 50%

Reported meetings remuneration committee (n=19)

6 32%

Additional Dummyrepfunction17 24 77%

functions #Dummy av.p.member 10 40%

Dummy fraction busy 15 62,5%

#memberFin/AC: number of members with a financial or accounting background. Measured using titles, main function or additional function or c.v.; Dummy FinAC: dummy variable with value 1 if board has at least one FIN or AC member; 0 if not; #memberEDU: number of members with experience in education. Measured using main function or additional function or c.v. Dummy EDU: dummy variable with value 1 if board has at least one EDU member; 0 if not; #memberPUB: number of members with experience in public administration. Measured using main function or additional function or c.v.; DummyPUB: dummy variable with value 1 if board has at least one public administration member; 0 if not; #FinAC+#EDU+#PUB: number of members with experience in either FinAC of EDU or PUB. #memberSB: number of members of the supervisory board; #repr stakeholder: mentions representation of stakeholder group; #not repr. stakeholder: no mention of representation of stakeholder group; Superv. minister, Superv. other, Superv. not ment.: board mentions on behalf of whom they supervise; # meetings year: number of SB meetings per year; AC # meeting: number of audit committee meetings per year; DummyAC: dummy variable with value 1 if board has an AC, 0 if not; RC#meeting: number of remuneration committee meetings per year; DummyRC: dummy variable with value 1 if board has an RC, 0 if not; #SB+AC+RC: number of SBmeetings + ACmeetings + RCmeetings per year; #memb addfunc: number of members with an additional function; #total addfunc board: total additional functions per board; #add. per member: average number of additional functions per board member; Dummyrepfunction: dummy variable with value 1 if additional functions are reported in annual report or on website, value 0 if not; Frac memb nonbusy: number of members in the

board that hold 3 or less additional functions; Frac memb busy: number of members in the board that hold more than 3 additional functions. #dummy av. p. member: dummy variable with value 1 if members hold on average 4,9 or more additional functions, value 0 if less than 4,9. Dummy fraction busy: dummy variable with value 1 if the majority of the members hold more than three additional functions.

Table 6 Board descriptives – 2 -

Mean Highest Lowest SD Median

Expertise # memberSB 5.90 9 2 1.56 6

# memberFinAC 1.03 3 0 .80 1

# memberEDU 0.87 5 0 1.15 1

# memberPUB 1.1 3 0 .978 1

# memberOTH 2.87 5 0 1.408 3

Meetings # meetings year (n = 27) 5.2 8 3 1.30 5

AC#meeting (n = 10) 3.5 5 2 1.99 3.5

RC#meeting (n=6) 1.6 4 1 1.02 1

Complexity Organizational complexity 23.35 37.64 16.96 5.58 23.41 Additional functions #memb addfunc 5.38 9 2 1.40 5

#total addfunc 26.13 59 5 13.89 26.5

#add. per member 4.92 13.5 1 2.64 4.63

Frac. memb nonbusy 2.83 7 0 2,04 2.50

Frac. memb busy 3.25 7 0 2.152 3.50

17

(27)

Master thesis P.I. Hagenbeek, S 1760912 University of Groningen December 2011 27 Statistical analyses on the fit between organizational complexity and number of board members gave no meaningful results. The correlation between the two variables is weak (ρ = .2) and not significant. For a further analyses the data were split into ‘small boards’ (≤ 5) and ‘large boards’ (≥ 6) and ‘less complex organizations’ (<23,42) and ‘more complex organizations (≥ 23,42). Again, this did not reveal any relation between the two variables. The cut of point in both cases was the median value.

4.5 Number of directorships

Although it is mandatory to publish the number and character of additional functions18 (sector code principle III.1.3), seven organizations (23%) fail to do so. For this purpose I checked the annual reports as well as information on the website. On average 5.38 directors per board hold one or more additional functions. And on average they hold 4.92 additional functions. This is more, but not significantly more, than the optimum number of 3 or less, reported by Fich and Shivdasani. Per board the average number of additional functions is just over 26. Board busyness is measured in two ways. Based on the average of 4.92 additional functions board were labelled as busy or not (Dummy av. p. member). Based on this measure 10 out of the 24 boards are busy. The second measure was the number of busy supervisors versus de number of non-busy supervisors per board (Frac memb busy). At this measure a board was labelled busy if the majority of its members holds more than three additional functions. According to this measure 15 boards are busy (62,5%).

4.6 Conclusions

These analysis show that quite a few aspects of the supervisory board appear to function well in reality. All IFHE’s report on the number of members; all boards have independent members; the vast majority (87%) publishes the number of board meetings. On average a board meets once every two and a half months. Also the majority reports on the number and nature of the additional functions of the supervisory board members. Almost all members have additional functions. The busiest board counts 9 members with in total 59 reported additional functions.

Most boards lack one or more experts, which violates the provisions in both the law and the sector code so there is definitively room for improvement when it comes to the financial, accountancy and educational expertise in de supervisory board. The reporting on the AC and RC meetings is downright sloppy. Contrary to the literature the size of supervisory boards of IFHE’s is not related to organizational complexity.

18

(28)

5.0 Assessing the levels of accruals

The most simple and straightforward approach to assess whether or not the height of accruals is ‘normal’, is to assume that accruals are non-discretionary: not amendable by a manager. If this would be the case, the accruals in year t would be the same as in year t-1(ceteris paribus) Any deviation would be a sign of earnings management. When both variables are scaled by lagged total assets there is no correlation (ρ = 0.036) which in itself is a signal for further study. The value of the case with the highest ratio was winsorized to the nearest value as the original value led to a strong negative correlation (see appendix 9).

5.1 NDcA and DcA

With the use of the equations of Jones and Teoh, the level of ND(c)A and D(c)A for the IFHE’s were assessed. The Jones-model reads as follows:

NDAt = .633 ( 1/ A t-1) + -.055(∆REV t / A t-1 ) + .082(PPE t / A t-1 ) DAt = TAt/At-1 - NDAt = ε

The modified Jones model reads:

NDAt = .633( 1/ A t-1) + -.055((∆REV t - ∆AR t)/ A t-1 ) + .082(PPE t / A t-1 ). The coefficients α1, α2 and α3 in this model are taken from the original Jones model. Details are given in appendix 10.

The data for the Teoh-model are:

NDCAt = .582 ( 1/ A t-1) + -.223 ((∆REV t - ∆AR t ) / A t-1 )) The discretionary accruals then become:

DCAt = CAt / A t-1 - NDCAt , α0 and α1 are calculated with the equation CAt / A t-1= α0 ( 1/ A t-1) + α1 (∆REV t / A t-1 ) + ε

Details are given in appendix 11.

The details of the models show that Teoh is stronger is its explanation of the level of discretionary (current) accruals. Its R2 is .545 as opposed to an R2 of .376 of the Jones model.

With the aid of the above equations, the level of discretionary (current) and non discretionary (current) accruals can be assessed. Because the level of accruals is subject of this research and not the sign (positive or negative) I use absolute data for discretionary (current) accruals. Again the outlier values (Iselinge) in all three calculations were winsorized. (see appendix 12 and 13)

(29)

DA-Master thesis P.I. Hagenbeek, S 1760912 University of Groningen December 2011 29 levels were compared to data from studies by Jones and Teoh. The DA levels found in the HPE sector are not uncommon (Jones:1991, p 216; Healy: 1985, p 98; Xie et al.: 2003, p 302).

Table 7 Descriptives of accrual data

Mean Highest Lowest SD

JonesABS DA09wins .0515 .1900 .0043 .0504 ModJonesABS DA09wins .0513 .1900 .0049 .0503 TeohABS DcA09wins .0459 .1700 .0022 .0402 Jones NDA 09 .0576 .09 .02 .0166 ModJones NDA 09 .0572 .09 .00 .0189 Teoh NDcA 09 -.0117 .012 -.049 .0118 LnJonesABS DA09wins -3.424 -1.687 -5.460 .9692 LnModJonesABS DA09wins -3.414 -1.660 -5.327 .9910 LnTeohABS DA09wins -3.440 -1.771 -6.114 .9367

For the remainder of this study I will continue to use the three models. Jones and modified Jones, because they are most detailed and refined and proved to be the ‘best’ detector of earnings management in earlier studies (Bartov et al, 2001). The Teoh model is used because it concentrates on short term – current - accruals. All analyses were initially carried out with the absolute data and later repeated with log transformed data to double check the initial outcomes. In almost all cases the sign, the strength and the significance of the correlation was equal. Only when the log transformed data give deviating results I will present those results separately.

5.2 Correlation of accrual models

In order to assess whether or not the three accrual models give congruent results, the correlation coefficients were calculated (see table 8). As expected Jones and Modified Jones show a perfect correlation. The correlation of both the Jones models and the Teoh model is very strong and significant at the 1% level. The correlation between the Jones models and the Teoh model are not perfect which follows logically from the financial data on page 24 which show that percentage wise the year-on-year changes of total and current accruals vary.

Table 8 Correlation of accrual models, absolute data Log transformed data JonesABS DA09 ModJones ABS DA09 TeohABS DcA09 LnJones ABS DA 09 LnMod Jones ABS DA 09 LnTeoh ABS DcA 09 JonesABS DA09 1 1.000** .723** LnJones ABS DA09 1 .995** .580** ModJones ABS DA09 1 .725** LnMod Jones ABS DA09 1 .589** TeohABS DcA09 1 LnTeoh ABS DcA09 1

(30)

5.3 Assessing homoscedasticity

In the regression models of Jones and Teoh all data are scaled by total assets to reduce heteroscedasticity (Jones, 1991: p. 212). If a model suffers from heteroscedasticity it means that the variance of the error variable or σ2ε, is not constant i.e. it fluctuates with changes in another, unknown variable.

Figure 1. Example showing heteroscedasticity; source Keller 2005, p 616.

Heteroscedasticity is detected by plotting the predicted value of y (ŷ) against the value of ε. The results are given underneath. Both the Jones’s - and the Teoh models do not display signs of heteroscedasticity. Regression Standardized Predicted Value 4 2 0 -2 R e g re s s io n S ta n d a rd iz e d R e s id u a l 2 1 0 -1 -2 -3

Dependent Variable: Tacc09/Tass08 Modified Jones model - heteroscedasticity

R Sq Linear = 1,11E-16 Regression Standardized Predicted Value 5 4 3 2 1 0 -1 R e g re s s io n S ta n d a rd iz e d R e s id u a l 2 0 -2 -4

Teoh - heteroscedasticity Dependent Variable: CurrentAcc09/Tass08

R Sq Linear = 0