The impact of the Combined Code on
Corporate Governance on earnings
management for UK listed companies
Student number: 10002273
MSc Accountancy & Control, track Accountancy
Amsterdam Business School
Faculty of Economics and Business, University of Amsterdam
Supervisor: dr. M. Lubberink
C. Baars | Abstract 2
Title: The impact of the Combined Code on Corporate Governance on earnings
management for U.K. listed companies.
Author: Cynthia Baars
Student number: 10002273
Place and date: Krommenie, June 2012
Course: MSc Accounting & Control, track Accountancy
Amsterdam Business School
Faculty of Economics and Business, University of Amsterdam
Year: 2011/2012, semester 2
C. Baars | Abstract 3
This research examines the impact of the Combined Code on corporate governance of 2003 on earnings management in the U.K. I examine the change in the use of earnings management by managers after the implementation of the Combined Code of 2003. Especially, I look at the accrual-based earnings management and real-accrual-based earnings management. For my research I use non-financial companies of the FTSE 350 for the years 1998 till 2010. I divided my sample period in a pre-Code (1998-2003) and post-pre-Code period (2004-2010). I use the modified-Jones model and a model which was used by Roychowdhury (2006) for the calculations of accrual-based and real-based earnings management.
I expect that due to the increased scrutiny of the Combined Code of 2003, accrual-based earnings management will decrease and real-based earnings management will increase, because real-based earnings management is harder to detect.
I find no evidence to support my expectations, which suggest that the Combined Code of 2003 has still enough flexibility to conduct earnings management and it failed to reduce earnings management in a detectable way. A possible explanation for this is that the Combined Code has a principle-based nature. Furthermore, I find evidence that accrual-based earnings management decreased over my total sample period.
C. Baars | Abstract 4
Table of contentsAbstract ... 3 1. Introduction ... 5 1.1 Background ... 5 1.2 Research question ... 7
1.3 Motivation of this research ... 7
1.4 Contribution of this research ... 8
2. Literature review and hypothesis ... 9
2.1 Earnings management ... 9
2.1.1 Why do managers conduct earnings management? ... 10
2.2 Earnings management in the U.K. (comparing with the U.S.) ... 11
2.3 Corporate Governance Code ... 12
2.3.1 History of the U.K. Corporate Governance Code ... 12
2.3.2 Effects of the different reports/codes on earnings management ... 14
2.3.3 Sarbanes-Oxley act of 2002 ... 15
2.4 Hypotheses development ... 17
3. Research methodology ... 18
3.1 Metrics for earnings management ... 18
3.1.1 Accruals model of Jones ... 18
3.1.2 Model for real-based earnings management ... 19
3.2 Regression model ... 20
3.2.1 Control variables ... 21
4. Results ... 22
4.1. Descriptive statistics ... 22
4.2 Regression results... 30
4.2.1 Robust regression results ... 32
5. Sensitive analyses and conclusion ... 33
5.1 Sensitive analyses ... 33
5.2 Conclusion and limitations ... 38
C. Baars | 1. Introduction 5
Earnings management is a widely investigated topic. There is substantial evidence that managers are engage in earnings management (Roychowdhury, 2006; and Graham et al., 2005). Managers have many incentives to boost their earnings to achieve a better result. After some major scandals around 1990, the Cadbury Report was published in 1992 in the U.K. This is a set of principle of corporate governance. Later, after some scandals in the U.S., the Sarbanes-Oxley act (SOX) was published in 2002 in the U.S. and the Combined Code on corporate governance of 2003 was published in the U.K. Cohen et al. (2008) researched earnings management in the U.S. in the period before the SOX was implemented and in the period after the SOX was implemented. Evidence in the U.K. is limited. I investigate the use of earnings management in the period before the passage of the
Combined Code of 2003 and the period after this passage in the U.K. More specially, I look at the use of accrual-based earnings management and real-based earnings management.
I find evidence that accrual-based earnings management has a declining pattern over my sample period (1998-2010). In my robust test, I find no evidence that the Combined Code of 2003 changed the use of earnings management in the U.K. This finding will make regulators be aware that when the scrutiny increases it does not eliminate earnings management.
The remainder of the paper proceeds as follows. Section 1.1 till 1.4 reviews the background, research question, motivation, and contribution of this research. Section 2 reviews relevant prior literature and provides the development of the hypotheses. Section 3 describes the research methodology. Section 4 presents the results. Finally, Section 5 presents the sensitive analyses and conclusions of my research.
There is essential evidence that companies are engaged in earnings management (Roychowdhury, 2006; and Graham et al., 2005). Earnings management is important for managers, because meeting or exceeding benchmarks is very important. Graham et al. (2005) found that CFOs appear as keen in the post-Sarbanes-Oxley act of 2002 (SOX) period as the pre-SOX period to meet earnings
benchmarks. They want especially meet or beat the analyst forecasts. Executives belief that this will build credibility with the market and helps to maintain or increase their company’s stock price. The market believes that companies can “find the money” to meet or beat their targets. As mentioned by Graham et al. (2005):
C. Baars | 1. Introduction 6 “Not being able to find one or two cents to hit the target might be interpreted as evidence of hidden problems at the firm”.
Furthermore, a missing target can indicate that a company is managed poorly. They also found supported survey evidence that companies are engaging in real activities in manipulating their reports. They found that executives have a strong preference for smooth earnings, a major part (78%) of the surveyed executives would give up economic value in exchange for smooth earnings. Smooth earnings are perceived as less risky by investors.
Roychowdhury (2006) found evidence that companies are trying to avoid losses by offering price discounts to temporarily increase sales, engaging in overproduction to lower costs of goods sold, and reducing discretionary expenditures aggressively to improve margins.
Zang (2012) researched how managers trade-off real-based earnings management and accrual-based earnings management. Zang found several constrains of real-based earnings
management. These constrains are the company’s competitive status in the industry, their financial health, the scrutiny from institutional investors, and the immediate tax consequences of the manipulation by real-based earnings management. The constrains of accrual-based earnings management are the presence of high-quality auditors, heightened scrutiny of accounting practice after the passage of the SOX, and company’s accounting flexibility. Hence, when there is a higher level of scrutiny of accounting practice in the post-SOX period, and when limited accounting flexibility is possible, companies use more real-based earnings management and less accrual-based earnings management. However, the evidence of Zang for the trade-off decisions is not depended on a specific period, such as Cohen et al. (2008) did in their study.
Cohen et al. (2008) did their research around the passage of the SOX. They found that the level of real earnings management declined prior to SOX and increased significantly after the passage of the SOX of U.S. listed companies. Further did Cohen et al. found that the accrual-based earnings management activities were high in the period immediately preceding the SOX and declined after the passage of the SOX. The SOX-act of 2002 is a response to some major corporate failures, examples are WorldCom and Enron. The SOX-act is only for U.S. listed companies and their subsidiaries in other countries.
All the above studies have investigated U.S. companies. Peasnell et al. (2000) researched earnings management in the U.K. They researched the extent of accrual management to meet earnings target in the pre- and post-Cadbury period. The Cadbury Report was published in 1992 and is a corporate governance code for the U.K. A corporate governance code is a set of principles of good corporate governance. Peasnell et al. (2000) found evidence of accrual management to meet earnings targets in both periods. This suggests that the Cadbury report have failed to completely to
C. Baars | 1. Introduction 7 eliminate earnings management activities. Further did Peasnell et al. (2000) found that the increasing use of audit committees during their sample period does not appear capable of explaining the observed structural break in the association between abnormal accruals and board compositions.
1.2 Research question
To extend the papers of Peasnell et al. (2000) and Cohen et al. (2008), I want to research which impact the Combined Code on corporate governance had on earnings management for U.K.
companies. Hence, my research question is: “In which extent does the Combined Code on corporate governance of 2003 influence the use of earnings management in the U.K. listed companies?”
I will look at the use of real-based and accrual-based earnings management in the U.K. I predict that after the Combined Code of 2003, companies use more real-based earnings
management and before the Combined Code companies use more accrual-based earnings management.
1.3 Motivation of this research
The most studies have been done on U.S. companies, examples are Cohen et al. (2008),
Roychowdhury (2006) and Graham et al. (2005). Academic literature is limited for U.K. cases. A study which has been done on U.K. companies was done by Peasnell et al. (2000).
My research will be an extension of prior research, for example the papers of Peasnell et al. (2000) and Cohen et al. (2008). The paper of Cohen (2008) is only relevant for U.S. companies and their subsidiaries and it not relevant for the most companies in the U.K. This research will be more relevant for companies, regulators, auditors, investors, stakeholders, etc., which are located in the U.K.
It is interesting to look at the U.K. instead of the U.S. for several reasons. First of all, the U.K. is a pioneer in the field of corporate governance regulations. The Cadbury Report, published in 1992, has influenced the development of many corporate governance codes globally and has been adopted as a benchmark by several countries (Mallin, 2004; and Arcot and Bruno, 2007). The Code is a
response to some corporate governance failures in the UK, examples of these are Maxwell
Communications, Polly Peck, Bank of Credit Commerce International (BCCI), and Coloroll. In 2003, the Combined Code on corporate governance was published. This Code is an improved version of a couple of reports (included the Cadbury Report), this will be explained later in the section of the literature review. The Combined Code was aiming to improve and strength the existing Combined Code of 1998.
C. Baars | 1. Introduction 8 of both countries is different. The Combined Code of the U.K. is more flexible than the SOX-act of the U.S. One of the differences between the Combined Code and the SOX-act is that the Combined Code is based on principles and the SOX-act on rules. The rule-based approach of the SOX is a list of detailed rules that companies must be follow. The principle-based approach of the Combined Code allows more judgement from the managers because there is not one solution for all companies. Besides that the Combined Code is more flexible, managers in the U.K. have different
incentives for earnings management than managers in the U.S. (Brown and Higgins, 2001). Conyon et al. (2011) and Fernandes et al. (2011) found both that the compensation in the U.S. is higher than in the U.K. This suggests that managers in the U.K. conduct less earnings management relative to the U.S. Further companies in the U.K. have a less widely dispersed ownership than the U.S., which leads to less agency problems.
1.4 Contribution of this research
This research will be relevant for companies, regulators, auditors, investors, stakeholders, etc., which are located in the U.K. It will give these parties awareness about the use of earnings management by managers in the U.K. This can help different parties with, for example, to understand the financial statement better and for better decision making.
This research will be interesting for regulators. As also mentioned by Zang (2012) it will make regulators be aware that when the scrutiny increased, by for example the Combined Code, it does not eliminate earnings management activities, but it may only change managers’ strategy of using earnings management
Further, real earnings management is more costly and is harder to detect (Graham, 2005). Nelson et al. (2002) found that managers are more likely to attempt earnings management, and auditors are less likely to adjust earnings management attempts in certain circumstances. My research will make auditors more aware of the extent of the use of real earnings management and the use of accruals.
For investors this research is also interesting. Real earnings management can be costly for investors (Zang, 2012). Accruals are more strongly associated with stock returns and evidence suggests that accruals play an important role in improving the ability of earnings to reflect company performance (Dechow, 1994). The use of more real earnings management might reduce future earnings, for example, by the reduction of Research and Development expenses (R&D) (Peasnell et al., 2000). The reduction of R&D can lead to lower future earnings.
C. Baars | 2. Literature review and hypothesis 9
2. Literature review and hypothesis
In this section I will give a deeper understanding about earnings management and Corporate Governance codes.
2.1 Earnings management
Earnings management is a broadly investigated topic. Scott (2003) described earnings management as a choice of accounting policies or actions affecting earnings, which is decided by a manager. The manager wants to achieve some specific reported objective. Healy and Wahlen (1999) give the following definition of earnings management:
“Earnings management occurs when managers use judgment in financial reporting and in structuring transactions to alter financial reports to either mislead some stakeholders about the underlying economic performance of the company or to influence contractual outcomes that depend on reported accounting numbers”.
Managers are using earnings management to maximize company’s interest or to maximize own interest. This causes an agency problem. The shareholders (principle) hire a management (agent) to act for him or her. However, both have different interests and this leads to conflicts (Jensen, 2005 and Leuz et al., 2003). There exist an information asymmetry between both parties and earnings management increase this. This might lead that investors make wrong decisions, because the information which they have is misleading. Hence, the information is in the manager’s best interest. However, earnings management may reduce this information asymmetry. Managers have more information about the company and the future of the company, and can help investors to see this by using earnings management. A solution to reduce the agency problem is a good corporate
The positive accounting theory (Watts and Zimmerman, 1986) predicts choices of accounting policies and response to new accounting standard by the company’s manager. It predicts that managers act out of self-interest and that managers are willing to sacrifice long-term gain for short-term gain. Long-short-term gain is in the best interest of the company and short-short-term gain is in the best interest of the manager (Scott, 2011).
Badertscher (2011) distinguished three types of earnings management, these are accruals-based earnings management, real-accruals-based earnings management and non-GAAP earnings
management. In this research I only focus on accrual-based and real-based earnings management. Accrual-based earnings management means changing the numbers without changing cash flows and is not in contradiction with the national GAAP. Examples of accruals are changing accounting policy
C. Baars | 2. Literature review and hypothesis 10 or increasing net accounts receivable (Scott, 2003,). Several studies have been done on accruals. Examples are Dechow (1994), Jackson et al. (2009), Badertscher (2011) and Graham et al. (2005). Real-based earnings management are real actions which are taken to manage earnings. Real action can be sales discounts, changes in production levels, changes in R&D and changes in other
discretionary expenditures (Roychowdhury, 2006). Graham et al. (2005) found evidence of real-based earnings management. They found that 80 percent of the CFOs would decrease R&D, advertising and maintenance expenditures, in order to deliver earnings. Further, they found that 55 percent of the CFOs would postpone new projects.
2.1.1 Why do managers conduct earnings management?
There are several reasons why managers conduct earnings management. First of all, managers can have contractual motivations, for example to manage his cash bonus or to manage debt covenants (Scott, 2011). Managers are compensated on earnings in several ways. These are both explicitly and implicitly. Examples of explicitly are salary, bonus and stock options, and examples of implicitly are job security and reputation. Managers with high equity incentives are more likely to report earnings that meet or beat analysts’ forecast (Cheng and Warfield, 2005).
A second motivation for earnings management is to lower political ‘heat’. The last motivation is to meet earnings expectations. Graham et al. (2005) found that earnings are an important
benchmark for managers and they found that managers are willing to make real decisions to meet the benchmark. They found four reasons why managers want to meet or beat the benchmark. First, to the build credibility with the capital market. Second, to maintain or increase stock price. Third, they want to improve the external reputation of the management. The last reason is to covey future growth prospects.
Graham et al. (2005) also found that missing a target by one or two cents might be
interpreted as evidence of hidden problems and leads to significant negative effects on share price and reputations. Managers try to meet expectations by avoiding losses and to smooth earnings (Leuz et al., 2003).
Managers want to avoid losses (Burgstahler and Dichev, 1997), because losses reduce the company value and managers believe that they may be fired when they reported losses (Bamber et al., 2010). Second, managers want to smooth earnings.
‘Smoothing’ earnings means reduce the variability of the reported earnings. Less volatile earnings are perceived as less risky by investors, which will lead to a higher company value (Graham et al., 2005).
C. Baars | 2. Literature review and hypothesis 11 psychological effects of low book values of assets (Jackson et al., 2009). Scott (2011) came with four patterns of earnings management. These are: taking a bath, income minimization, income
maximization, and income smoothing.
2.2 Earnings management in the U.K. (comparing with the U.S.)
There have been several studies done about earnings management. However, the most studies have been done in the U.S. Some studies are comparing the U.S. with other countries.
Brown and Higgins (2001) compared earnings management in the U.S. with 12 other countries. They found that managers in the U.S. are more likely to manage earnings surprises than other countries. The cause is that the corporate governance and the legal environments are different in the U.S. than other countries. Another reason why managers in the U.S. manage more earnings is because there is a rapid increase in stock and option compensation in the U.S., further is the U.S. more short-term oriented than other countries. Companies in the U.S. have the most widely dispersed ownership, which leads to more agency problems. Ownership in the U.K. is also broadly dispersed, but less than the U.S. Hence, managers in the U.K. are using relatively more earnings management than other non-U.S. managers.
Fernandes et al. (2011) looked at CEO compensation of the U.S. and other countries. They found that the mean CEO compensation in 2006 was $ 5.5 million in the U.S. and $ 2.9 million in the U.K. The median CEO compensation in 2006 was $ 3.3 million in the U.S. and $ 1.7 million in the U.K. As you can see in the U.K. managers are receiving less compensations than in the U.S. This suggests that managers in the U.K. have less incentives in earnings management than the U.S. However, Fernandes et al. also looked at the risk-adjusted CEO pay. This is the amount of compensation after all non-tradable options granted to an undiversified risk-averse executive, and have been replaced with equivalent riskless cash compensation the executive would exchange for the options. The risk-adjusted CEO compensation is for the U.K. ($ 2.3 million) higher than the U.S. ($ 2 million).
Conyon et al. (2011) also found that U.S. managers have a higher compensation. This is why they have much more stock and option incentives than U.K managers. Further they found that the risk-adjusted compensations is not much different in the U.S. than the U.K. Conyon et al. (2011) explain this as differences in pay between both countries.
Leuz et al. (2003) found less earnings management in countries with developed equity markets, dispersed ownership structures, strong investor rights, and legal enforcement. They found that earnings are smoother in Continental Europe and Asia than in Anglo-American countries. The authors also looked at the magnitude of accruals comparing with the magnitude of the operating cash flows. They found that the magnitude of company’s accruals is small in the U.K. and the U.S.
C. Baars | 2. Literature review and hypothesis 12 compared to Austria, Germany and South Korea. Further they found that European and Asian companies have a greater degree of loss avoidance than Anglo-American companies.
This suggests that managers in the U.S. have more incentives for earnings management than managers in the U.K.
2.3 Corporate Governance Code
The corporate governance is a system of directing and controlling a company. The corporate governance code is a set of principles of good corporate governance. Each country has a different corporate governance code, however some Codes are influenced by others or is the Code of one country a benchmark for another.
2.3.1 History of the U.K. Corporate Governance Code
The U.K. is a pioneer in the field of corporate governance regulations. After some major failures and scandals around 1990, examples are Maxwell Communications, Polly Peck, BCCI and Coloroll, Sir Adrian Cadbury investigated the British corporate governance system and published the Cadbury Report in 1992. An important part of this report was the Code of Best Practice.
The Cadbury Report is a voluntary code and is characterised by shareholder pressure for adoption. They choice for a voluntary code instead of a statutory code, because the principles that one size does not fit all in matter of corporate governance. Each company has a different
environment and circumstances and that is why each company has to choose their optimal corporate governance structure. The Cadbury Report is a guideline for the listed companies. Some major recommendations of the Cadbury Report of 1992 are:
The separation of the role of the chairman from the executive functions of the CEO;
Companies must have three committees: the audit, nomination and remuneration committees, each with a specific function and clearly defined terms of reference;
The board should have at least three non-executive directors within the board with a monitoring function (Arcot and Bruno, 2007; and Cadbury Report, 1992).
The Cadbury Report has influenced the development of many corporate governance codes globally and has been adopted as a benchmark by several countries (Mallin, 2004; and Arcot and Bruno, 2007).
After the Cadbury Report there were published several different/newer reports. In 1995 came the Greenbury Report. This report was mainly focused on remuneration committees and
C. Baars | 2. Literature review and hypothesis 13 remuneration package. Sir Ronald Greenbury recommended that the progress of the corporate governance must be reviewed every three years.
In 1998 the Hampel Committee came with a revised corporate governance system. Hampel emphasized the role that institutional investors could play in corporate governance (Mallin, 2011). The Cadbury, Greenbury and Hampel reports came together in one Code, namely the Combined Code of 1998. This Code comprised the recommendations of all these separate reports. Arcot and Bruno (2007) mentioned that companies often believed that the Cadbury and Greenbury report was too prescriptive, without paying attention to the explanation part. The main focus of this new Combined Code was on the quality of information that the Code should convey, essentially the explanations provided. The Combined Code is a principle-based standards and involved judgements of managers. It operates on ‘comply or explain’, which means that good governance cannot be constrained by statutory regulations because this tends towards a ‘one-size-fits-all’ solution. Nelson et al. (2002) found that managers trend to adjust their decisions based on the latitude given by the standards.
In response of the public shock following the scandals in the U.S. (for example Enron and WorldCom), both Derek Higgs and Robert Smith published a report on corporate governance in 2003. Even though there were no scandals or failures in the U.K., however the public was scared of failure of the U.K corporate governance system. Higgs’ report was focus on the role and effectiveness of non-executive directors. Smiths’ report was a response on the collapse of Arthur Anderson and Enron and focused on the independence of auditors. Those two reports became together in one report in the revised Combined Code of 2003. The Combined Code was aiming to improve and strength the existing Combined Code of 1998. The role of boards and the board sub-committees is the starting point of good corporate governance (Mallin, 2011). The Combined Code of 2003 focus on the degree of power in a company. It wants to prevent that all the power (or too much power) is in the hand of one person. That is the reason why the Combined Code recommends a separate role of chair and CEO. Other recommends of the Combined Code of 2003 are:
Establishment of an audit committee, which act as a link between external auditors and audit committee;
Establishment of a remuneration committee, which set the remuneration of executive directors;
There should be a formal and transparent nominations process for nominating new directors.
C. Baars | 2. Literature review and hypothesis 14 and Combined Code of 2003). The revised Combined Code of 2003 applied for companies which reported on or after the first of November 2003.
Where the U.S. came with a prescriptive legislation code, Derek Higgs strongly believed in the non-prescriptive approach. However, the presence of a higher number and detailed provisions was seen as a movement towards a more prescriptive legislation.
In 2006 the Combined Code of 2003 was revised. There were three main changes; these are (Mallin, 2011):
The company chair is allowed to serve on the remuneration committee, but is not allowed to chair this committee;
Provide a ‘vote withheld’ option on proxy appointment forms to enable a shareholder to indicate that they wish to withhold their vote;
Recommend companies to publish details of proxies lodged at general meetings where votes were taken on a show of hands, on their website.
Two years after the revised Combined Code of 2006, the revised Combined Code of 2008 was published. This Code had two changes. The first one was to remove the restriction on an individual chairing more than one FTSE 100 company. Second, listed companies outside the FTSE 350 are allowed the company chair to sit on the audit committee where he or she was considered independent on appointment (Mallin, 2011).
In 2010 was a new revised Combined Code published. This Code was a response on the financial crisis and came into force for companies with financial years beginning on or after 29 June 2010.
2.3.2 Effects of the different reports/codes on earnings management
Every report on corporate governance in the U.K. in the past 20 years emphasise different issues, although all issues of the different reports are included in the Combined Code of 2003. In 1998, the Hampel Committee emphasized the role that institutional investors could play in corporate
governance (Mallin, 2011). The Hampel report is included in the Combined Code of 1998 and 2003. The Combined Code 2003 and earlier corporate governance reports emphasize the audit committee, remuneration committee, and the nomination committee. Another important issue in the Combined Code of 2003 are the presence non-execute directors in the boards. In 2006 it was allowed for a company chair to serve on the remuneration committee, however not as the chair of this committee (Mallin, 2011). All of these issues may have effects on earnings management. There are several researches done on some of these issues.
C. Baars | 2. Literature review and hypothesis 15 The Hampel report emphasise the instructional investors. Koh (2007) did research on the association between institutional investor type and the company’s earnings management strategy. Koh found evidence that support the view that long-term institutional investors constrain accruals management among companies that manage earnings to meet/beat benchmarks. This suggests long-term institutional investors can mitigate aggressive earnings management among these firms. However, this does not apply for transient institutional investors.
Peasnell et al. (2000) did research on the relation between earnings management in the pre- and post-Cadbury period and also looked at non-execute directors. Peasnell et al. found no evidence that non-execute directors in the board would reduce earnings management in the post-Cadbury period in the U.K.
Klein (1998) found evidence that audit committees are less in dependent if the CEO is on the nominating committee. This suggests a positive relation between CEO being on the board’s
nominating committee or remuneration committee and earnings management. In another research, Klein (2002) found evidence in the U.S. that the audit committee reduces earnings management, only when the committee has less than a majority of independence directors, and hence not a 100% audit committee independence. Other evidence Klein found is that earnings management is
negatively related to the CEO’s shareholdings and it is negatively related to whether a large outside shareholder sits on the board’s audit committee.
After the implementation of the revised Combined Code of 2006 the company chairman is allowed to serve on the remuneration committee, although the company chair is not allowed to be the chairman of this committee. Klein (2002) found evidence that earnings management increases when the CEO sits on the board’s remuneration committee.
2.3.3 Sarbanes-Oxley act of 2002
The Sarbanes-Oxley Act (SOX) of 2002 was a response of some major scandals in the U.S. The SOX was implemented to improve the transparency, timeliness, and quality of financial reporting. The SOX is not only for companies and auditors in the U.S., but also for foreign companies which are listed in the U.S. or companies which are a subsidiary of an U.S. company. The SOX contains 11 titles that describe guidance for financial reporting, some of these are mandatory and some are
requirements. One of these titles is the Public Company Accounting Oversight Board. This board is an important part of the SOX. This board has the power to set auditing standards and also to inspect auditors. Further, there are penalties for fraud and for obstructing an investigation (Merchant and van der Stede, 2007). Some important provision of SOX include the following (Larcker and Tayan, 2011):
C. Baars | 2. Literature review and hypothesis 16
The requirement that the CEO and CFO certify financial results, with misrepresentations subject to criminal penalties;
An attestation by executives and auditors to the sufficiency of internal controls;
Independence of the audit committee of the board of directors;
A limitation of types of non-audit work an auditor can perform for a company;
A ban on most personal loans to executives or directors.
The two important sections of the SOX are Section 302 and Section 404. Section 302
mandates a set of internal procedures designed to ensure accurate financial disclosure. The CFO and CEO must certify that they are responsible for establishing and maintain internal controls and have designed such internal controls to ensure that material information relating to the company is made known to the CFO and CEO by others within the company. Further, it is required that an external auditor issues an opinion on whether effective internal control over financial reporting was maintained in all material respects by management.
Section 404 of SOX requires public companies to report management’s assessment of the effectiveness of internal control over financial reporting. This evaluation is integrated with the audit of the financial statements. The evaluation increases user confidence about future financial
reporting, because an effective internal control reduces likelihood of future misstatements in financial statements. Most companies in the U.S. use the Committee of Sponsoring Organizations of the Treadway Commission (COSO) framework for their internal control (Arens et al. 2011). This is a framework that can guide the company in the risk-management processes. In the U.K. the most companies use the Turnbull report which is similar to COSO (Larcker and Tayan, 2011).
Different than the U.K. Corporate Governance Code, the SOX is a prescriptive legislation. The Combined Code is more flexible than the SOX, because of its ‘comply or explain’ approach. Although, since some changes in the SOX in 2007, the SOX became also more flexible. The changes of 2007 included removing the requirement for the external auditor to assess management’s process for assessing the system of internal control, and the revising of the definitions of significant deficiency and material weakness. The new standard aims to encourage companies and auditors to move away from the checklist approach to Sarbanes-Oxley audits, where auditors have to delve into each financial element in fine detail. The Securities and Exchange Commission (SEC), which is the most important regulatory body in the U.S. and it emphasis on the protection of security holder rights and the prevention of corporate fraud (Larcker and Tayan, 2011), is recommending in the new standard a risk-based approach that will encourage executives, auditors, directors, and audit committee
C. Baars | 2. Literature review and hypothesis 17 to transfer some of the responsibility for auditing corporate governance to a company’s own management and internal auditing teams, counting on them to determine and check what is really important (EDN, 2007). This might lead to more judgement which might create space to conduct earnings management.
2.4 Hypotheses development
To give a better understanding of the financial statement to several parties, like investors, auditors, regulators, etc., it is important to show the pattern of the use of earnings management in the U.K. I am interesting whether the Combined Code of 2003 helped in reducing earnings management, this was found by Cohen et al. (2008) for U.S. companies after the implementation of the SOX. Further, I want to know what type of earnings management managers use more in the U.K.
I will focus on two types of earnings management, namely accrual-based earnings management and real-based earnings management. I will look at the discretionary accruals and changes in real activities/decisions. Consistent with Zang (2012), I predict that managers use accrual-based earnings management and real-accrual-based earnings management as substitutes to achieve the desired earnings targets. To see whether the use of earnings management changed after the passage of the Combined Code of 2003, I will look at the use of earnings management in the period before the and after the passage of the Combined Code.
Based on the paper of Zang (2012) I expected that when scrutiny increased, managers are using more real-based earnings management because this is harder to detect. I will look whether companies replaced accruals-based earnings management with real-based earnings management. This leads to my fist hypothesis:
Hypothesis 1: The level of real earnings management increased in period after the passage of the Combined Code of 2003.
To detect real-based earnings management, I will look at the patterns in cash flow from operations and production costs.
Further, I want to find out whether the Combined Code of 2003 is ‘better’ than the Cadbury Report of 1992. Peasnell et al. (2000) found accrual-based earnings management in both pre-Cadbury and post-pre-Cadbury period. Which suggest that the pre-Cadbury report have failed to completely eliminate earnings management. I do not expect that the Combined Code eliminate earnings management, however as mentioned above, when the scrutiny is higher I expect less accrual-based earnings management and more real-based earnings management in the post-Code period. This leads to my second hypothesis:
C. Baars | 3. Research methodology 18
Hypothesis 2: The level of accrual-based earnings management decreased in the period after the passage of the Combined Code of 2003.
3. Research methodology
As I mentioned before, there are three types of earnings management, namely accrual-based earnings management, real-based earnings management and non-GAAP earnings management (Badertscher, 2011). I only focus on the accrual and real-based earnings management in my research. I collect my data from Datastream and I use annual data. I will compare the period before the Combined Code on corporate governance of 2003 with the period after the Combined Code of 2003. The sample of the period before the Combined Code of 2003 will be from 1998 till 2003 (pre-Code), and the sample of the period after the Combined Code of 2003 will be from 2004 till 2010 (post-Code). This code is for listed UK companies, hence I will look at the FTSE 350 UK non-financial companies with available data. The reason why I use the FTSE 350 is that some provisions do not apply to companies below the FTSE 350. Further, I exclude financial companies (SIC 6400-6999) because the overall regulatory environment is significantly different from non-financial companies.
3.1 Metrics for earnings management
To examine in which extent the Combined Code on corporate governance of 2003 influence the use of earnings management in the U.K, I will start examine both accruals and real-based earnings management. With the results of the calculations of both accruals and real-based earnings management I can make my regression, see section 3.2.
For the examination of the accrual, I will use the modified cross-sectional Jones model (Jones, 1991; Dechow, 1995; and Cohen et al., 2008). For the examination of the real earnings management, I will use a model which is developed by Dechow et al. (1998) and is also used by Roychowdhury (2006) and Zang (2012). All my variables, in the below explained formula’s, will be winsorized to the 1st and 99th percentiles of their distributions (expect for 1/Ait-1, because here there is a minimum of
zero). I do this to make my tests more robust. Further, I use a two-way cluster-robust procedure (Cameron et al., 2011) for my calculations of accruals and real-based earnings management. I cluster the years and industries in my sample.
3.1.1 Accruals model of Jones
To calculate the accruals, I will look at the two-digit SIC code to classified industries. This is necessary to estimate the model for every industry (Cohen et al., 2008). I will require at least 8 companies per
C. Baars | 3. Research methodology 19 industry, so I can compare an individual company with its industry and that I can reflect the impact on earnings management from industry wide economic conditions during the year. The formula for total accruals (TACC) is as follows:
where i stands for company and t stands for year; TACCit = EBXIit – CFOit, where EBXI stands for earnings before extraordinary items and CFO stands for cash flow from operations; Ait-1 is total assets
from the previous year (t-1); ∆REVit is the change in revenue from previous year to preceding year;
PPEit stands for the gross value of property, plant and equipment; Є is the residual.
The coefficients estimates from the above (1) equation are used to estimate the company-specific normal accruals (Cohen et al., 2008). The formula for the normal accruals (NA) is as follows:
where ∆ARit is the change in accounts receivable from previous year to preceding year. To calculate the discretionary accruals (DA) I will look at the difference between the total accruals and the normal accruals, thus:
3.1.2 Model for real-based earnings management
To calculate real earnings management I will look at a model which is used by Roychowdhury (2006) and Zang (2012). I will look at the abnormal levels of cash flow from operation (CFO), as well as at production costs. Also here I will look at the two-digit SIC code to classified industries. I will require at least 8 companies per industry. To calculate the abnormal level of CFO I use the following formula:
The abnormal CFO is the actual CFO minus the normal level of CFO. The normal level of CFO is calculated using the estimated coefficient from the above (3) formula.
For the calculation of the production costs I will look at the cost of goods sold (COGS) and the change in inventory during the year. The COGS and change in inventory are calculated as follows:
C. Baars | 3. Research methodology 20
Productions costs (Prodit) is the sum of COGS and change in inventory. The normal level of the production costs is calculated as following formula, with using the coefficients from formula 5 and 6:
The abnormal level of the production costs is calculated as the actual production cost minus the normal level of production costs.
3.2 Regression model
To summarize time trends of the accrual-based and real-based earnings management proxies, I will regress each of the variables on a time trend variable. The reason to use this model is to describe the variables because many of the variables exhibit significant time trends (non-stationary), rendering traditional summary statistics uninformative (Cohen et al., 2008). I use a regression model to capture the variations resulted from the time specific effects of the passage of the Combined Code of 2003. The regression estimation provides the time specific effects in the sense that earnings management may be different or may be changed across time with regarding to the passage of the Combined Code. Similar regression models are used by Peasnell et al. (2000) and Cohen et al. (2008) to
investigate the impact of the Cadbury Report of 1992 in the U.K. and the SOX-act of 2002 in the U.S. on earnings management across time. To examine my hypotheses, I use the following regression model:
where Depj represent the various earnings management metrics. These metrics are discretionary accruals, based CFO manipulation, based production cost manipulation and the sum of real-based manipulations (earnings management). Time is a trend variable equal to the difference between the current year and 1998. PostCode is a dummy variable, which is equal to 1 if the year is the range of 2004-2010. Post2006 is a dummy variable, which is equal to 1 if the year is equal to 2007 and 2008. Post2008is a dummy variable which is equal to 1 if the year is equal to 2009 and 2010. The sum of Controls is the sum of all my control variables, these are size, change in number of employees, change in sales, financial distressed, Big 4, and leverage. These control variables will be explained in the next section.
C. Baars | 3. Research methodology 21 among the residuals within year clusters in my regression model (Petersen, 2009).
As mentioned earlier, I will compare the period before Combined Code on corporate
governance of 2003 with the period after the Combined Code of 2003. This is why I will use a dummy for the Code period. However, there were two revised Combined Codes published in the post-Code period of my sample. To make sure that my findings are not influenced by these revised post-Codes, I will use two extra dummies, for the years 2007-2008 and 2009-2010.
I expect that the coefficient of regression for the discretionary accruals will be negative and the coefficient for real-based earnings management will be positive in the post-Code period. Which suggest that management/companies reduce the use of accrual-based earnings management in the post-Code period and increase the use of real-based earnings management in the same period. I do not expect that the revised Codes of 2006 and 2008 influence my results, because there were not significant changes made into these Codes. Thus, the coefficients of these periods I expect to be insignificant.
3.2.1 Control variables
To check whether my results are not influenced by other variables, I will use the following control variables: Size, ∆ Employees, ∆ Sales, Fin distress, Big 4, and LEV.
To check whether Size of the companies is influences my results, I will look at the total assets of the companies. Bigger companies are more professional and that is why I expect less earnings management for those companies. Further, I expect less earnings management for those companies who are growing. Therefore I have two control variables ∆ Employees and ∆ Sales. I will look at the change of the number of employees and at the change of the sales.
I expect that companies which are not doing well will use more earnings management to boost their earnings, this is consistent with evidence from Burgstahler and Dichev (1997). I will look at financially distressed companies. Consistent with prior literature (Jaggi and Lee, 2002 and Jaggi and Sun, 2006), I consider a financially distressed company if the company has losses for two consecutive years and/or the company has a cash flow from operations which is negative for two consecutive years. Fin distressed is equal to 1 when a company satisfies with the above mentioned requirements.
Big 4 is equal to 1 if the external auditor of the companies is one of the big 4, which is Deloitte, Ernst&Young, KPMG, or PwC. Becket et al. (1998) found evidence that Big 6 auditors are generally more effective in deterring earnings management than others.
My last control variable is LEV. LEV is the financial leverage and it is measured as the ratio of total debt to total assets. Companies with a high leverage ratio may have incentives to adjust
C. Baars | 4. Results 22 earnings upward to avoid debt-covenant violation. Although, those companies might be under a close scrutiny of lenders and are less able to do this.
4.1. Descriptive statistics
My final sample consists 130 industry-year observations from differ industries. Table 1B gives an oversight of the different industries which are included in my sample. The industry with the largest observations is the industry “Advertising and market research”, with 19 companies in this industry. The mean (median) is 4 (2) companies per industry. I required in my sample of companies that in each industrial group must have at least eight observations in each year, this requirement is for the comparison of an individual company with its industry and for the reflection of the impact on earnings management from industry wide economic conditions during the year. With this requirement the maximum number of observations for each industry is still 19, and the mean (median) is 9.7 (8). First, I provide the descriptive statistics per industry, then I provide the
descriptive statistics of my total sample, and lastly, I will look at the descriptive statistics of the pre- and post-Code period. Further, I provide the model parameters and the correlations between the variables.
Table 1A provides an overview of the medians of the main variables for each industry of my sample. Table 2 provides a summary of statistics over time for my total sample. Panel A of T1 gives an oversight of the statistics of the main variables over the whole sample period. Panel B gives also an oversight of the statistics of the main variables, however, here I divided my sample in two period, namely pre-Code period and post-Code period. This gives a good oversight of the differences
between the two periods.
In Table 1A the median of the ROE is the lowest for the industry “Manufacture of textiles” with 12.17 per cent and the industry with the highest ROE is “Waste collection, treatment and disposal activities”, with 17.40 per cent. The median (mean) ROE for the total sample (Table 2 Panel A) is 15.04 (30.64) per cent. A ROE between 15 and 20 per cent is generally considered to be good. The high ROE for the industry “Waste collection, treatment and disposal activities” is due to their low shareholders’ equity. The median of the ROE increased significantly in the post-Code period
Descriptive statistics per industry (median)
Mfg food Mfg textiles Mfg leather Mfg
chemicals Mfg machinery Sewerage Wast collection Land transport Water transport Adv. & research TA 1,669,614 741,849 1,031,000 2,178,000 597,848 2,305,850 453,561 6,766,950 532,350 568,600 LEV 0.1917 0.1174 0.1258 0.2695 0.1487 0.2315 0.2611 0.4343 0.2179 0.1643 ROA 0.0713 0.0679 0.0515 0.0625 0.0744 0.0418 0.0619 0.0382 0.0706 0.0701 ROE 0.1266 0.1217 0.1637 0.1666 0.1673 0.1354 0.1740 0.1497 0.1670 0.1417 ACCRUALS -0.0551 -0.0580 0.0236 -0.0445 -0.0536 -0.0445 -0.0444 -0.0408 -0.0291 -0.0601 DA 0.0536 0.0546 0.0045 0.0385 0.0340 0.0239 0.0270 0.0603 0.0188 0.0179 PDA 0.0536 0.0546 0.0047 0.0385 0.0340 0.0239 0.0270 0.0603 0.0188 0.0179 NDA - - -0.0004 - - - - Ab_CFO 0.0171 0.0219 -0.0771 0.0061 0.0086 -0.0221 -0.0053 -0.0226 -0.0174 0.0222 Ab_prod 0.0194 0.0510 0.1368 -0.0121 -0.0868 0.0387 -0.0318 0.0383 -0.0234 -0.0839 REM_proxy 0.0550 0.0720 0.0740 -0.0166 -0.0735 0.0332 -0.0559 0.0152 -0.0438 -0.0760 N 96 96 143 119 96 88 96 96 96 225
This table represents the medians of several variables. Where Mfg food represents industry “Manufacture of food products”; Mfg textiles represents industry “Manufacture of textiles”; Mfg leather represents industry “Manufacture of leather and related products”; Mfg chemicals represents industry “Manufacture of chemicals and chemical products"; Mfg machinery represents industry “Manufacture of machinery and equipment n.e.c.”; Sewerage represents industry “Sewerage”; Waste collection represents industry “Waste collection, treatment and disposal activities, materials recovery”; Land transport represents industry “Land transport and transport via pipelines”;Water transport represents industry “Water transport”; Adv. & research represents industry “Advertising and market research”; TA represents total assets; LEV represents the financial leverage, which is measures as the ratio of total debt to total assets; ROA represents return on assets, calculated by net income after preferred dividends divided by total assets; ROE represents return on equity, calculated by net income after preferred dividends divided by shareholders’ equity; ACCRUALS represents (EBXI-/-CFO)/total assets from the previous year, where EBXI represents earnings before extraordinary items; DA represents discretionary accruals computed by using the Modified Jones Model; PDA represents the positive value of discretionary accruals; NDA represents the negative value of discretionary accruals; Ab_CFO represents the abnormal level of real earnings management using cash flow from operations; Ab_prod represents the abnormal level of real earnings management using production costs; REM_proxy represents the sum of the two real earnings management proxies, i.e., Ab_CFO and Ab_prod.
TABLE 1B SIC codes
SIC-code Description of the code Number of companies per
10 Manufacture of food products 8
13 Manufacture of textiles 8
15 Manufacture of leather and related products 12
20 Manufacture of chemicals and chemical products 10
28 Manufacture of machinery and equipment n.e.c. 8
37 Sewerage 8
38 Waste collection, treatment and disposal activities;
49 Land transport and transport via pipelines 8
50 Water transport 8
73 Advertising and market research 19
TABLE 2 Descriptive statistics Panel A: total sample
Variable Mean Standard deviation Median N 25th Percentile 75th Percentile TA 5,177,931 15,000,000 895,196 1151 353,832 3,463,200 LEV 0.2264 0.1975 0.2003 1151 0.0881 0.3231 ROA 0.0613 0.0864 0.0619 1151 0.0315 0.0975 ROE 0.3064 5.8976 0.1504 1151 0.0744 0.2182 ACCRUALS -0.0485 0.0752 -0.0436 1151 -0.0809 -0.0094 DA* 0.0350 0.0255 0.0299 1151 0.0154 0.0519 PDA* 0.0352 0.0254 0.0303 1145 0.0155 0.0520 NDA* -0.0004 0.0002 -0.0004 6 -0.0004 -0.0003 ABS_DA 0.0350 0.0255 0.0299 1151 0.0154 0.0519 Ab_CFO 0.0014 0.0857 -0.0068 1151 -0.0491 0.0392 Ab_prod -0.0140 0.2366 0.0174 1151 -0.0902 0.0995 REM_proxy -0.0126 0.2101 0.0119 1151 -0.0850 0.0857
Panel B: Differences between the pre-Code and post-Code period
Variable Mean Median Mean Median Means
(p-value) Medians ( p-value) TA 3,617,374 655,092 6,294,275 1,067,780 2,676,901 (0.0023) 412,688 (0.0000) LEV 0.2223 0.2051 0.2294 0.2002 0.0071 (0.7812) -0.0049 (0.6410) ROA 0.0508 0.0602 0.0688 0.0659 0.0180 (0.0006) 0.0057 (0.0455) ROE 0.4872 0.1377 0.1771 0.1610 -0.3101 (0.3830) 0.0233 (0.0000) ACCRUALS -0.0516 -0.0476 -0.0463 -0.0409 0.0053 (0.2260) 0.0067 (0.1829)
C. Baars | 4. Results 25 DA* 0.0388 0.0335 0.0324 0.0272 -0.0064 (0.0000) -0.0063 (0.0000) PDA* 0.0388 0.0335 0.0327 0.0277 -0.0061 (0.0001) -0.0058 (0.0000) NDA* - - -0.0004 -0.0003 - - ABS_DA 0.0388 0.0335 0.0324 0.0272 -0.0064 (0.0000) -0.0063 (0.0000) Ab_CFO -0.0076 -0.0084 0.0078 -0.0056 -0.0002 (0.0021) 0.0028 (0.0318) Ab_prod 0.0018 0.0312 -0.0253 0.0053 -0.0271 (0.0550) -0.0259 (0.0005) REM_proxy -0.0058 0.0211 -0.0174 0.0050 -0.0116 (0.3537) -0.0161 (0.0134)
Descriptive data of a sample of U.K. companies over years 1998 to 2010. TA represents total assets; LEV represents the financial leverage, which is measures as the ratio of total debt to total assets; ROA represents return on assets, calculated by net income after preferred dividends divided by total assets; ROE represents return on equity, calculated by net income after preferred dividends divided by shareholders’ equity; ACCRUALS represents (EBXI-/-CFO)/total assets from the previous year, where EBXI represents earnings before extraordinary items; DA represents discretionary accruals computed by using the Modified Jones Model; PDA represents the positive value of discretionary accruals;
NDA represents the negative value of discretionary accruals; ABS_DA represents the absolute value of discretionary accruals; Ab_CFO represents the abnormal level of real earnings management using cash flow from operations;
Ab_prod represents the abnormal level of real earnings management using production costs; REM_proxy represents the sum of the two real earnings management proxies, i.e., Ab_CFO and Ab_prod; Pre-Code represents the years 1998-2003; Post-Code represents the years 2004-2010.
* DA is subdivided in two variables, namely PDA and NDA, for this reason are the amounts of observations for these two variables different. NDA has 0 observations in the pre-Code and 6 observations in the post-Code. PDA has 480 observations in the pre-Code and 665 observations in the post-Code.
The median (mean) of the ROA of the total sample is 6.19 (6.13) per cent. The highest median of the ROA is for the industry “Manufacture of machinery and equipment n.e.c., with 7.44 per cent and the lowest is 3.82 per cent for industry “Land transport and transport via pipelines”. Financial leverage is measured as the ratio of total debt to total assets. The difference between the highest leverage industry and the lowest is significant at the 1% level. The median of the total sample in Table 2 Panel 1 is 20.03 per cent. The highest median leverage is for the industry “Land transport and transport via pipelines” with 43.43 per cent, where the lowest is 11.74 per cent for the industry “Manufacture of textiles”. This might suggest that the industry “Land transport and transport via pipelines” has more incentives to adjust their earnings upward to avoid debt-covenant violation. Although, as mentioned in section 3.2.1, it is also possible that the companies in this industry are not adjusting their earnings because they are under a close scrutiny of lenders. Table 1A suggests that accruals-based earnings management and real-based earnings management are substitutes of each other for some industries. A good example is the industry “Manufacture of leather and related products”. This industry has the highest median for real-based earnings management, namely 0.0740 where the lowest median is -0.0760 and the median of the total sample is 0.0119, and it has the lowest median for discretionary accruals comparing with other industries, namely 0.0045 where the highest median is 0.0603 and the median of the total sample is 0.0299.
C. Baars | 4. Results 26 The same table suggests that abnormal cash flow from operations and abnormal production costs are also substitutes of each other. In the same industry as before, “Manufacture of leather and related products” has the lowest value for abnormal cash flow from operations (-0.0771, where the median of the total sample is -0.0068) and has the highest value of abnormal production costs (0.1368, where the median of the total sample is 0.0174). Another good example is the industry “Advertising and market research”, which has the lowest value for real-based earnings management. This industry has the highest value for the abnormal level of cash flow from operations (0.0222) and has the second lowest value for the abnormal level of production costs (-0.0839). An explanation for the low abnormal level of production costs for this industry might be because industry “Advertising and market research” is a service company and not a manufacturing company. Service companies have no/less production costs and that is why it is hard to conduct earnings management with production costs.
Table 2 provides a summary of statistics over time. Panel A gives an oversight of the statistics of the main variables over the whole sample period, and Panel B gives an oversight of the statistics of the main variables, where my sample is divided in two period, namely pre-Code period and post-Code period.
The median ROA, in Panel B T2, increased significantly at the 5% level in the post-Code period compared with the pre-Code period and the median ROE increased significantly in the post-Code period. Both increased because this is leverage driven, the median of shareholders’ equity also increased significantly. Further, there were no cases of negative discretionary accruals in the pre-Code. In the post-Code period there were 6 cases of negative accruals. The median and mean of discretionary accruals are statistically significantly lower in the post-Code period compared with the pre-Code period. The median and mean of total of real-based earnings management is higher in the post-Code period compared with the pre-Code period, however the mean is not significant (p-value of 0.3537), and the median is significant at the 5% level.
Figure 1 provides a graphical illustration of the development of the median, which is less sensitive for outliners, of the discretionary accruals and the total real-based earnings management over the total sample period. The figure shows the decrease of the discretionary accruals over time. In the first two year (1999-2001) the value of discretionary accruals drops the most. Furthermore, the figure shows a quite stable value for the period 2001-2005. In this period, year 2003, the Combined Code was implemented, and that is why I expect a decrease.
Figure 1 Panel B provides an oversight of the development of the median of real-based earnings management. Notable is the difference in pattern of discretionary accruals and real-based earnings management. Where discretionary accruals (Panel A of Figure 1) have a declining pattern,
C. Baars | 4. Results 27
Development of the median of earnings management over the total sample period (1998-2010)
Panel A: Development of discretionary accruals over time
Panel B: Development of the median of total real-based earnings management over time
and real-based earnings management (Panel B of Figure 1) has a volatile pattern.
Real-based earnings management is the sum of the abnormal level of cash flow from operations and the abnormal level of production costs. Figure 2 provides the development of the median of both measurements of real-based earnings management. In this figure the abnormal level of cash flow from operations has a volatile and increasing pattern. The abnormal level of production cost has a decreasing pattern. Notable is the year of 2009, which has a high value of abnormal production cost. .0 2 .0 2 5 .0 3 .0 3 5 .0 4 D iscre ti o n a ry a ccru a ls 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 Year -. 0 2 0 .0 2 .0 4 T o ta l re a l-b a se d e a rn in g s ma n a g e me n t 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 Year
C. Baars | 4. Results 28
Development of the median of abnormal level of cash flow from operations and production costs
Table 3 gives a brief summary of the regressions for the calculations of accruals and real-based earnings management. It reports the coefficients, p- values, and the adjusted R2. For these regressions I winsorized variables to the 1st and 99th percentiles of their distributions (except for 1 / TAit-1, because assets cannot be negative). I use a two-way cluster-robust procedure (Cameron et al., 2011) for my calculations of accruals and real-based earnings management. The regressions are based on industry-year observations.
For the calculation of accruals I use the modified-Jones model, which has three variables. Table 3 shows that two variables have a coefficient which is significant negative in this model, and the coefficient for the change of revenue divided by total assets of the previous year (∆REVit / TAit-1) is insignificantly different from zero. An insignificant coefficient means that this variable is not an important factor in predicting accruals. The other two variables, inverse assets (1 / TAit-1) and PPE divided by total assets (PPEit / TAit-1), have a more important factor in predicting accruals, both they have a negative coefficient of -996.48 and -0.0529, respectively. These negative coefficients suggest that these variables decrease with 996.48 for inverse assets and 0.0529 for PPEit / TAit-1 by the amount of change of total accruals, expecting that the other variables are held constant. This also counts for PPEit / TAit-1 when I look at inverse assets. Thus, inverse assets decrease with 996.48 by the amount of change of accruals (acc), expecting that all other variables (including PPEit / TAit-1) are held constant. Further, the adjusted R2 in the regression for acc is 10.55%, meaning that the modified-Jones model has a quite low explanatory power. Only 10.55% of the variance of the accruals is accounted for by the model by its variables.
Table 3, also reports the regressions of real-based earnings management. In the regression for cash flow from operations (CFO_lagA) is only the variable sales divided by total assets of the previous year (Salesit / TAit-1) insignificant different from zero, meaning that this variable is less important than the other variables in predicting cash flow from operations. The other variables are significant and have a positive relationship with the CFO_lagA. The explanatory power of the model
-. 0 3 -. 0 2 -. 0 1 0 .0 1 Ab n o rma l le ve l o f ca sh f lo w f ro m o p e ra ti o n s 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 Year -. 0 2 0 .0 2 .0 4 .0 6 Ab n o rma l le ve l o f p ro d u ct io n co st s 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 Year
C. Baars | 4. Results 29
Model parameters estimations of the normal levels of accruals, cash flow from operations and productions costs
Acc CFO_lagA Prod_lagA
Intercept -0.0117 (0.380) 0.0989 (0.000) -0.1687 (0.000) 1 / TAit-1 -996.48 (0.007) 1,835.80 (0.062) -9,069.82 (0.029) ∆REVit/ TAit-1 -0.0061 (0.585) 0.0989 (0.000) -0.2428 (-0.016) PPEit / TAit-1 -0.0529 (0.000) Salesit / TAit-1 0.0056 (0.692) 0.8553 (0.000)
∆Salesit-1/ TAit-1 -0.0517
(%) 10.55 11.46 88.09
This table provides the coefficients and (p-values). The regressions are estimated for every industry for every year. The regressions are estimated at the 2-digit SIC industry level, were in each industrial group there must be at least 8 companies in each year. Acc represents total accruals divided by total assets from the previous year by using the Modified Jones Model; CFO_lagA represents cash flow from operations divided by total assets from the previous year; Prod_lagA
represents productions costs (sum of cost of goods sold and change in inventories) divided by total assets from
previous year;TAit-1 represents total assets from the previous year (t-1); ∆REVit is the change in revenue from
previous year to preceding year; PPEit stands for the gross value of property, plant and equipment.
is the coefficient of determination adjusted for degrees of freedom, in %.
of CFO_lagA is low. The average adjusted R2 for the regressions of real-based earnings management across industry-years is 11.46% for cash flow from operations, and the adjusted R2 for production costs is 88.09%. An adjusted R2 of 88.09% means that 88.09% of the variance of the production costs is accounted for by the model by the variables. The production costs are calculated by two models, namely one model looks at the change in inventories and the other model looks at COGS. The model of COGS has a quite high explanatory power, with an adjusted R2 of 86.74%.
Table 4 provides the Pearson’s correlations between various variables. The correlation of Pearson between abnormal real-based earnings management (REM_proxy) and abnormal discretionary accruals (DA) is 5%, a positive correlation suggest that managers are using both accruals-based earnings management and real-based earnings management (Zang, 2012). However, this correlation is insignificant. The high significant correlation between abnormal production costs and abnormal real-based earnings management (93%) is because the proxy of abnormal real-based earnings management is the sum of abnormal production cost and abnormal cash flow from operations.
The correlation between abnormal accruals and abnormal cash flow from operations is positive and significant (30%), this is inconsistent with prior studies. Both Roychowdhury (2006) and Cohen (2008) expected a negative correlation between abnormal accruals and abnormal cash flow