Trade-off effect between accrual based earnings management and real earnings management after the adoption of mandatory IFRS : evidence from listed companies in 22 European countries

(1)

Trade-off effect between accrual based earnings

management and real earnings management after the

adoption of mandatory IFRS.

Evidence from listed companies in 22 European countries

Student: Willem Cheng

Student number: 10106162 Date of final version: 21st_{June, 2015} Word count: 16089

MSc Accountancy & Control, variant Accountancy Amsterdam Business School

Faculty of Economics and Business, University of Amsterdam Supervisor: dr. A. Sikalidis

(2)

2 | P a g e Statement of Originality

This document is written by student Willem Cheng, who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

3 | P a g e

Abstract

The objective of this research is to examine whether there is a trade-off between accrual-based earnings management and real earnings management after the adoption of mandatory IFRS. Main prior literature about earnings management and mandatory IFRS adoption focused only on accrual-based earnings management. Including real earnings management gives a better overall view of earnings management. Researches in Europe that included real earnings management in their research, have taken assumptions about the substitution effect found in the results from US settings.

In this research, I investigate if mandatory IFRS adoption has an effect on accrual-based earnings management and/or real earnings management. Prior literature reports different results about this subject. The trade-off effect between accrual-based and real earnings management after the mandatory implementation of IFRS is also investigated. This has not been researched before in the time frame of mandatory IFRS adoption, and also not with an European setting.

The sample of this research consist of European listed firms. To investigate if there is a trade-off effect, a sample have been taken of firms that are suspected of managing earnings. Firms that meet or just beat earnings benchmarks/forecast are suspected using earnings management. The results suggest that accrual-based earnings management is likely to decrease after mandatory IFRS adoption. However, there is no significant effect found on real earnings management after the mandatory IFRS adoption. The results of the trade-off effect between accrual-based earnings management and real earnings management suggest that there is a substitution effect after mandatory IFRS adoption.

(4)

4 | P a g e

1. Introduction

1.1 Background

In 2002 the European Parliament issued a regulation that all listed companies in the European Union (hereafter: EU) have to prepare consolidated financial statements according International Accounting Standards by 2005. All listed companies in the EU have to adopt mandatory International Financial Reporting Standards (hereafter: IFRS) for fiscal years starting after 1 January 2005. This is an important regulatory change for accounting in the EU. Companies are required to report in accordance with IFRS. The objective of implementing this change is to enhance the comparability of financial reporting, corporate transparency and increase the quality of financial reporting and therefore also the reported earnings (EC Regulation 1606/2002). Financial reporting quality should increase if IFRS limits management’s opportunistic discretion in determining accounting amounts. This would mean that implementing mandatory IFRS adoption would decrease earnings management practices as of 2005. IFRS should provide higher quality than the country-specific Generally Accepted Accounting Principles (GAAP) that was used before IFRS was mandatory adopted. However, certain issues may also have a negative impact on the quality of information.

In the past management have used earnings management to report better performances of the company. This lead to accounting scandals like Enron, WorldCom and Parmalat. This triggered the discussions about rules-based and principle-based systems in the United States (US). The US-GAAP is more rules-based compared to IFRS which is more principle-based. A change from rules-based system to principles-based system, managers and auditors often need to use their professional judgment to ensure that the financial statements reflect the economic components of a transactions (Wüstemann and Kierzek, 2005). Opportunistic behaviour by managers in financial reporting has been a concern for a long time. If ownership and management are not the same, the accounting function will have agency problems (Jensen and Meckling, 1976). The agency problem starts with the incomplete and asymmetric information if a principal (stakeholder) hires an agent (manager). The interest of a principal and agent are not aligned. Without effective controls, managers may act in their own interest. With flexible rules, there is a higher chance of earnings management practices. With the mandatory IFRS adoption in the EU, management has a new regulations of financial reporting.

(6)

6 | P a g e

There has been a lot of prior researches about earnings management. An often used definition of earnings management is from Schipper (1989, p. 92). He defined earnings management as follows:

“A purposeful intervention in the external financial reporting process, with the intent of

obtaining some private gain.”

Earnings management can be split in two forms, namely accrual-based earnings management and real earnings management. Past literature have focused more on accrual-based earnings management compared to real earnings management. This research is going to focus on both forms of earnings management.

According to Schipper (2005), the mandatory adoption of IFRS in the EU gives a setting that is better able to test the determinants and economic consequences of accounting quality. The reason here for is that mandatory IFRS adoption provides the same accounting standards for countries in the EU since 2005.

To research if there is a trade-off between accrual-based earnings management and real earnings management when IFRS was mandatory adopted, the research question to investigate this is as follows:

Does firms trade-off accrual based earnings management with real earnings management or vice versa after the adoption of mandatory IFRS?

1.2 Motivation

In the prior literature there has been a lot of research about the effect of the mandatory IFRS adoption on earnings management. The main prior researches about this topic have focused on accrual-based earnings management. Besides accrual-based earnings management, there is also real earnings management. Zang (2012) researched the trade-off between accrual-based earnings management and real activities manipulation in the US. Recently, there is a research done by Doukakis (2014) about the effect of mandatory IFRS adoption on earnings management in the EU. Both accrual-based and real earnings management activities have been taken into account in this research. According to this research there is no significant impact on earnings management after the mandatory adoption of IFRS. The results of this research differs from prior research about earnings management. Main prior research only took accrual-based earnings management in to account as overall earnings management. The

(7)

7 | P a g e

difference could also be the result of taking assumptions from the findings of researches that are done in the US. Doukakis (2014) took the results of Zang (2012) that there is a substitution effect between accrual-based and real earnings management.

In my research question I am going to research if the earnings management behaviour of managers have changed after the adoption of mandatory IFRS. Therefore I am going to research if managers switch from accrual-based earnings management to real earnings management or vice versa after the implementation of mandatory IFRS. This is in accordance with the papers of Cohen et al. (2008) and Zang (2012). Cohen et al. (2008) researched real and accrual-based earnings management in the pre- and post-SOX periods. Zang (2012) researched the trade-off effect between real and accrual-based earnings management in the US without accounting for the passage of SOX. In this research I am also going to combine these two researches to see if there is a trade-off between the two forms of earnings management after mandatory IFRS adoption in the EU. In contrast to the research of Zang (2012), I will not focus on the relative cost between these two earnings management methods. This research is going to focus on an EU setting instead of an US setting. The periods before and after the mandatory adoption of IFRS is taken into account. Prior literature suggest that after implementing IFRS, it should increase financial reporting quality by reducing management’s opportunistic discretion in determining accounting amounts. This research question could explain the results of Doukakis (2012) that there no significant impact on earnings management after adopting mandatory IFRS is found. If accrual-based earnings management have been replaced by real earnings management after the adoption of mandatory IFRS, it would mean that the adoption of IFRS did have an impact on earnings management. This research is important for countries that are still considering to adopt mandatory IFRS.

This research is going to follow the following manner. Chapter 2 contains the literature review. This chapter start with the agency theory, and then provides a background of earnings management. This is followed by a brief overview of prior research about earnings management and the hypotheses development of this thesis. In chapter 3 I present the models that are used for this research, and also the research and data sample. The descriptive statistics and results are presented in chapter 4. Lastly the conclusion is given in chapter 5.

(8)

8 | P a g e

2. Literature Review

In this chapter I will first explain the agency theory and give the relationship with earnings management. Secondly I will explain what earnings management is by giving a definition of earnings management. Afterwards I will explain why management use earnings management. I am also going to explain two forms of earnings management and I will give an overview of prior researches about earnings management and mandatory IFRS adoption. Finally, the hypotheses development is given in the last section of this chapter.

2.1 Agency theory

Managers may have personal goals that are not aligned with the goals of the shareholders. This problem is a well-known theory that is discussed throughout the literature, namely the agency theory. Agency theory describes the structure of the economic exchange between two parties, namely (i) the principal and (ii) the agent. Jensen and Meckling (1976, p. 5) defined the agency theory as follows:

“We define an agency relationship as a contract under which one or more persons (the principal(s)) engage another person (the agent) to perform some service on their behalf which involves delegating some decision making authority to the agent. If both parties to the relationship are utility maximizers, there is good reason to believe that the agent will not always act in the best interests of the principal.”

This definition gives a clear understanding of the agency problem. Managers (agent) have been empowered by the shareholders (principal) by giving them decision making rights and compensating them for making decisions that will maximize the wealth of the shareholders (principal). Managers may have a different interest than the shareholders. Both managers and shareholders want to maximize their own utility (Heracleous & Lan, 2012). Agency theory predicts that managers try to maximize their own interest/wealth at the cost of the shareholders. They can do this by managing earnings.

Earnings management can been seen as an agency cost if managers provide financial reports that are not consistent with the economic state of the firm. Agency cost can be seen as when shareholders make non-optimal investment and operational decisions due to inaccurate financial information. Earnings management is a good example of an agency problem because earnings management can cause or worsen the agency cost (Davidson III et al., 2004, p. 268).

(9)

9 | P a g e

Warfield et al. (1995) also mentions the separation of ownership and control. Managers delegate decision rights to the shareholders. This can affect the amount of information regarding accounting of earnings and the decisions of managers regarding accounting choices.

2.2 Earnings management

2.2.1 What is earnings management

Earnings management is a well-researched subject in the literature. Prior literature provides many definitions of earnings management. The two most often used definition of earnings management from Schipper (1989) and Healy and Wahlen (1999) are used to explain earnings management. Schipper (1989, p. 92) gives the following definition of earnings management:

“A purposeful intervention in the external financial reporting process, with the intent of

obtaining some private gain.”

Healy and Wahlen (1999, p. 368) gives a more detailed 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.”

Schipper (1989) defines earnings management as only a private gain, whereas Healy and Wahlen (1999) refer to earnings management due to a more general reason. In this study, I see earnings management as a combination of these two definitions. Managers manage earnings to receive a certain bonus, and therefore they mislead some stakeholders about the performance of the company due to earnings management. Management also does not want the earnings to be too high. If management reduces the earnings for the current year, they can easier reach the bonus benchmark for next year.

2.2.2 Using earnings management

There is evidence found that suggest that managers have incentives to avoid reporting decreasing earnings. Several researches provide evidence that there is an incentive to maintain a consistent increase in earnings. Management avoid reporting earnings declines and loses in

(10)

10 | P a g e

the financial reporting by using estimates, judgements and assumptions to manage earnings. A reason to report a consistent increase in earnings is given by Barth et al. (1999). They found in their research that firms that report a consistent increase in earnings have a higher price-to-earnings multiples. This multiples will increase if price-to-earnings increases is maintained for a longer period. If this consistency in earnings increases is broken, the higher multiples will substantially be reduced or even disappear (Barth et al. 1999). DeAngelo et al. (1996) found similar findings in their research. They found that if a firm breaks the pattern of consistent earnings increases, the firm will experience a 14% negative abnormal stock return in the year when this happens.

Burgstahler and Dichev (1997) found in their research that there is a relative low amount of small decreases in earnings and small losses, and a relative high amount of small increases in earnings and small positive income. Two components of earnings is found that is used to manage earnings, namely (i) the cash flow from operations and (ii) changes in working capital. Burgstahler and Dichev (1997) gives two theories in their research that gives an explanation why managers engage in earnings management. (i) The first theory is that management try to avoid decreases when reporting earnings and losses to decrease the cost imposed on the firm in transactions with stakeholders. (ii) The second theory is based on prospect theory which postulates an aversion to absolute and relative losses (Burgstahler and Dichev, 1997, p. 124).

According to Dichey et al. (2013) the most important reasons for managers to engage in earnings management are to influence stock price, influence executive compensation and to avoid debt covenant violations. Managers have pressure to reach earnings benchmarks. Senior managers fear the negative effects on their careers if they report bad performance. Cheng and Warfield (2005) mentions that managers manage earnings to receive a bonus following a compensation plan. Managers increase earnings to reach a certain benchmark when he/she will receive a bonus, or decrease earnings to put less pressure on reaching earnings benchmark for the following years.

2.3 Two forms of earnings management

Discretion of managers in the adoption of accounting methods can be used to manage earnings. This research is going to discuss earnings management. Earnings management can be split into two forms, (i) accrual based earnings management and (ii) real earnings management.

(11)

11 | P a g e 2.3.1 Accrual-based earnings management

Accrual-based earnings management can be achieved by changing the accounting methods or estimates used when presenting a given transaction in the financial statement (Zang, 2012, p. 676). Managers deliberately bias accounting information by making certain accounting choices. This can be done by changing the depreciation method for fixed assets or changing the estimates of provision for doubtful accounts. This can bias reported earnings. Accrual-based earnings management does not directly affect cash flow. Accrual-Accrual-based earnings management is a situation where the total accruals are being managed by managers of a firm. Accrual-based earnings management occurs after the fiscal year-end.

Prior literature used discretionary accruals as a proxy for accrual-based earnings management. Discretionary accruals can be seen as the difference between actual accruals of a firm and the normal level of accruals. Dechow et al. (1995) investigated which model is the best to measure accrual-based earnings management. They looked at the following models: (i) the Healy Model, (ii) the DeAngelo Model, (iii) the Jones Model, (iv) the Modified Jones Model and (v) the Industry Model. In their research it is found that the modified Jones Model provides the most powerful test for earnings management. Cohen et al. (2008) and Zang (2012) used the modified Jones (1991) model to estimate the discretionary accruals. Therefore I am also going to use the modified Jones (1991) model to estimate accrual-based earnings management.

2.3.2 Real earnings management

Real earnings management are deviations from normal operational practices by managers to deliberately mislead at least some stakeholders into believing that financial reporting goals have been achieved by normal operations. These deviations from normal operational practices by influencing accounting information by the management does not necessarily increase firm value, even though managers achieve their reporting goals due to these deviations (Roychowdhury, 2006, p. 337). Real earnings management has to occur during the fiscal year.

Real earnings management can be examined in the following ways: though increasing earnings by reducing the cost of goods sold by overproducing inventory, cutting discretionary expenditures, including R&D, advertising, and selling, general, and administrative (SG&A) expenditures (Roychowdhury, 2006; Cohen et al., 2008, Zang, 2012). Cohen et al. (2008) and Zang (2012) measure these real earnings manipulation activities using the method of Roychowdhury (2006) by using (i) abnormal levels of cash flow from operations, (ii) abnormal levels of discretionary expenses and (iii) abnormal level of production costs.

(12)

12 | P a g e

Cohen and Zarowin (2010), Bruns and Merchant (1990) and Graham et al. (2005) mentions two reasons that managers manage earnings through real earnings management activities than through based earnings management. (i) They mention that accrual-based earnings management is more likely to attract auditor or regulatory scrutiny than real decisions. (ii) Besides this, they mention that only relying on accrual-based management is risky. As mentioned before, accrual-based earnings management can only occur after the fiscal year-end. If manipulating accruals is not enough to reach the desired amount, real earnings management cannot be used anymore. As mentioned above, real earnings management has to occur during the fiscal year. Cohen et al. (2008) mentions that real earnings management is a safer way to manage earnings because it is harder to detect, even though that it is more costly.

2.4 Prior research about earnings management and mandatory IFRS adoption.

Zang (2012) focused in her study on the trade-off between real activities manipulation and accrual-based earnings management. Zang (2012) did not research earnings management with regard to the mandatory IFRS adoption period or the passage of SOX. This study researched the difference by using variables that explains the costs of real and accrual earnings management. She took a sample of 6.500 firm-years observations that are suspected of earnings management from the period 1987 till 2008. Like prior literature about accrual-based earnings management, discretionary accruals is used as a proxy to estimate the amount of accrual-based earnings management. To estimate real activities manipulation the cross-sectional model that Roychowdhury (2006) used in her paper is also used here. The real activities manipulation includes overproducing inventory and cutting discretionary expenditures like R&D and SG&A expenditures. Her research was according to Zang (2012) the first study to identify the cost for real earnings management and examine the trade-off effect on real and accrual-based earnings management. In her setting, the two earnings management methods is traded-off by managers. It is found that when accrual-based earnings management is constrained, firms use more real activities manipulation. However, when the use of real activities manipulation it is more costly than accrual-based earnings management, firms will use more accrual-based earnings management.

Callao and Jarne (2010) has researched the effect of IFRS on earnings management. The objective of this paper is to see whether after the adoption of mandatory IFRS in the EU have affected earnings management. Unlike the paper of Zang (2012), Callao and Jarne (2010)

(13)

13 | P a g e

does not take real earnings management in to account. Discretionary accruals has been taken as a proxy for earnings management. To see if the discretionary accounting practices have changed, the amount of discretionary accruals is estimated before and after the mandatory IFRS adoption. The discretionary accruals that have been broken down into current and long-term adjustments. This paper used the model of Larcker and Richardson (2004) to get the discretionary accruals. This model comes from the modified Jones (1991) model proposed by Dechow et al. (1995). The sample for this paper consist of 1.408 non-financial firms that are listed on the stock markets in the 11 EU member states. The data is collected from the years 2003 to 2006. Only one year after the mandatory IFRS adoption is taken into account in this research design. The results of this paper shows that the discretionary accruals have increased in the periods after the mandatory IFRS adoption. This implies that earnings management has increased after mandatory IFRS adoption in Europe.

Jeanjean and Stolowy (2008) researched whether mandatory adoption of IFRS had an impact on earnings management. They focused on three countries, namely Australia, France and the UK. Research design for earnings management can be defined in three categories, namely (i) those that use discretionary accruals, (ii) those that use specific accruals, and (iii) those that study statistical properties of earnings to identify thresholds. Jeanjean and Stolowy (2008) used the research design that analyse the distribution of earnings. They research whether managers manage earnings to avoid losses after the mandatory IFRS adoption have decreased. A sample of 1.146 firms have been used in this research. The data is collected from the periods 2002 till 2006. Also in this paper, only one year after the implementation of mandatory IFRS have been taken into account. The findings of Jeanjean and Stolowy (2008) suggest that mandatory IFRS adoption did not lead to a decline of earnings management. It even lead to an increase of earnings management in France.

Zeghal, Chrourou and Sellami (2011) researched the effects of mandatory IFRS adoption on earnings management by French firms only. Discretionary accruals are also taken as a proxy for earnings management. This study also does not take real earnings management into account. The model of Kothari et al. (2005) is used to calculate the discretionary accruals. The sample consist of 851 French-listed firms. The period of this sample is taken from the years 2003 till 2006. It is found in this research that discretionary accruals has decreased after the adoption of mandatory IFRS in French. The result suggest that the adoption of mandatory IFRS have decreased the amount of earnings management for the French firms. This is a

(14)

14 | P a g e

different result than the research of Callao and Jarne (2010) and Jeanjean and Stolowy (2008), were the amount of discretionary accruals have increased after the adoption of mandatory IFRS.

The paper of Doukakis (2014) also researched the effect of mandatory IFRS adoption on earnings management. This paper examines both the real and accrual-based earnings management activities. A sample of 15.206 observations was used from firms that mandatory adopted IFRS in the periods between 2000 and 2010. This study includes a control sample of voluntary IFRS adopters. It is expected that implementing IFRS in 2005 would raise the quality of financial reporting. Financial reporting quality should increase if it limits management’s opportunistic discretion in determining accounting amounts. Based on this, the initial thought of implementing IFRS would decrease earnings management practices. The accrual-based earnings management is calculated using discretionary accruals as proxy. The modified Jones model (1991) is used for this. As the study of Zang (2012), Doukakis (2014) also used the model of Roychowdhury (2006) to estimate real earnings management. This research suggest that there is no significant impact on the amount of accrual and real earnings management after mandatory implementing IFRS.

2.5 Hypotheses development

Main prior studies (Jeanjean & Stolowy, 2008; Callao & Jarne, 2010; Zeghal et al., 2011) about earnings management and mandatory IFRS adoption examine only accrual-based earnings management. Including real earning management, like studies of Cohen et al. (2009), Zang (2012) and Doukakis (2014), give a better view of overall earnings management.

Overall, the findings about mandatory IFRS adoption and earnings management give different results. There is a positive or no effect found with regard to earnings management after the mandatory implementation of IFRS. Callao and Jarne (2010) and Jeanjean and Stolowy (2008) found in their studies that earnings management have increased after the mandatory adoption of IFRS. These studies only took accrual-based earnings management into account. In contrast, Doukakis (2014) also real earnings management into account. His research found that there is no significant impact on earnings management after the mandatory adoption of IFRS. The results of Zeghal, Chrourou and Sellami (2011) suggest that the adoption of mandatory IFRS have decreased the amount of earnings management. But this research contains French firms only.

(15)

15 | P a g e

As can be seen from the aggregate results of prior literatures done in the EU, the finding concerning earnings management gives different results. Doukakis (2014) make assumptions in his paper, based on finding from the US. One assumption is the substitution effect between accrual-based and real earnings management that Zang (2012) found. Before looking if this substitution assumption, that have been found in the US, also holds in an EU setting or if it is applicable in a period of implementing new regulations, I first look at the effect of mandatory IFRS adoption on earnings management. Therefore, our first hypothesis is about the effect of implementing mandatory IFRS on earnings management. The hypothesis is split in to two parts, accrual-based earnings management and real earnings management. In line with the results of Doukakis (2014), I expect that there is no impact on both accrual-based and real earnings management after mandatory IFRS adoption in the EU. These hypotheses are based on the finding of Doukakis (2014) and are as follows:

Hypothesis 1a: There is no significant impact on the amount of accrual-based earnings

management after the mandatory implementation of IFRS in Europe.

Hypothesis 1b: There is no significant impact on the amount of real earnings

management after the mandatory implementation of IFRS in Europe.

As before mentioned, Doukakis (2014) made an assumption based on the prior research to use the substitution argument that Zang (2012) found. According to Zang (2012), accrual-based earnings management and real earnings management can act like substitutes. Like discussed before, real earnings management can take place independently of accrual-based earnings management (Doukakis, 2014, p. 556). The research of Zang (2012) is conducted in the US. By using data from the EU could lead to different results. Both accrual-based and real earnings management are costly activities. Consistent with the research of Zang (2012), where she investigated if there is a trade-off between accrual-based and real earnings management, my research question also investigates whether managers trade-off accrual-based earnings management with real earnings management. This research sample will only consist of EU listed firms and looks at the period before and after mandatory IFRS adoption. I am going to look if the amount of accrual-based or real earnings management changes and if there is a negative relationship between them, when mandatory IFRS is adopted. Consistence with the findings from Zang (2012) in the US, I expect a negative relationship between accrual-based and real earnings management, after the mandatory

(16)

16 | P a g e

adoption of IFRS in the EU. Based on the results of Zang (2012), the hypotheses are as follows:

Hypothesis 2a: The amount of accrual-based earnings management is negatively

related to the amount of real earnings management after the mandatory IFRS adoption in Europe.

Hypothesis 2b: The amount of real earnings management is negatively related to the

amount of accrual-based earnings management after the mandatory IFRS adoption in Europe.

(17)

17 | P a g e

3. Data and Research methodology

In this chapter I give an overview how I conduct my research. First I explain the research methodology, how accrual-based earnings management and real earnings management is measured. Then I explain how the regression model is used to measure the trade-off between accrual-based and real earnings management. Secondly, I describe how the data sample is chosen.

3.1 Measuring accrual-based earnings management

Accrual-based earnings management have been researched a lot in prior literature. As mentioned before, many different models have been examined to measure accrual-based earnings management. Dechow et al. (1995) concluded that the modified Jones Model provides the most powerful test for earnings management. Many prior literature used this model to measure accrual-based earnings management. For this, I also use the modified Jones model. The modified Jones model is used to get the discretionary accrual variable. Discretionary accruals are calculated by using total accruals and non-discretionary accruals.

Total accruals are calculated as follows:

Where = total accruals, calculated as firm i’s net income minus cash flows from operations in year t taken from the statement of cash flows

= total assets for firm i in year t - 1;

= change in sales for firm i during the year t, so Salesit – Salesit-1 ;

= gross value of property, plant and equipment for firm i in year t.

Non-discretionary accruals are calculated as follows:

Where = non-discretionary accruals for firm i in year t;

= change in accounts receivable for firm i during the year t, so .

(18)

18 | P a g e

As before mentioned, discretionary accruals (3) are calculated by the difference between the total accruals (1) and the non-discretionary accruals (2).

Where = discretionary accruals for firm i in year t.

Firms can manage earnings through either income-increasing or income-decreasing accruals (Warfield et al., 1995). Therefore, consistent with prior literature (Cohen et al. (2008), Doukakis (2012), Kim et al. (2012) Zang (2012)), the absolute value of discretionary accruals (ABS_DA) is used for our main analyses to take both income-increasing and income decreasing earnings management into account. The absolute value if discretionary accruals will be used as the dependent variable in two of the regression models. The higher the value of absolute discretionary accruals indicates more earnings management through accruals. The results of income-increasing (Positive_DA) and income-decreasing accruals (Negative_DA) are also shown in the results for a better overview of discretionary accruals.

3.2 Measuring real earnings management

In accordance with the research of Zang (2012), real earnings management activities will be examined following Roychowdhury (2006). In the paper of Roychowdhury (2006), three proxies are mentioned to estimate real earnings management. Real earnings management can be measured through (i) abnormal cash flow from operations (AB_CFO), (ii) cost of production (AB_PROD) and (iii) discretionary expenses (AB_DISX). Total real earnings management is computed using these standardized variables.

The following methods can be used to manage real earnings through (i) cash flow from operations, (ii) increased production and (iii) discretionary expenses (Roychowdhury (2006) and Cohen et al. (2008)):

Managing (i) cash flow from operations can be done by increasing sales through price discounts or providing more lenient credit terms. This will increase sales temporarily, but will disappear when the firm apply the old prices again. The increase in sales will boost the earnings of that period, but will result in lower cash flows for the current period due to more lenient credit terms.

By (ii) increasing production, the cost of goods sold can be lowered. By producing more units, managers can spread the cost of fixed costs over more products. The fixed cost

(19)

19 | P a g e

per unit will be lower, and thus the cost per unit will decline. Producing more units can also lead to more costs. An example is the increased holding costs of these more products.

By (iii) decreasing discretionary expenses, earnings of that period can increase. Examples of discretionary expenses are advertising expense, research and development, and SG&A expenditures. It could also lead to higher cash flows in that period.

To estimate real earnings management, AB_CFO, AB_PROD and AB_DISX are calculated. To get the abnormal values, the difference between the actual values and the normal values are calculated. Normal values of cash flows from operations, production costs and discretionary expenditures will be estimated according to the models used by Roychowdhury (2006), Cohen et al. (2008) and Zang (2012).

The normal values are calculated as follows:

Where = the normal level of cash flows from operations from firm i in year t.

Where = the sum of the cost of goods sold from firm i in year t and the change in inventory from year t - 1 to year t;

∆Salesit-1 = change in sales for firm i during the year t – 1, so Salesit-1 – Salesit-2

Where = the discretionary expenses for firm i in year t, is defined as selling, general and administrative expenses.

Other variables are as previously defined.

To understand the effect of total real earnings management, the meanings of the three standardized variables have to be understood. The higher the positive value of AB_CFO indicates less earnings management. The higher the positive value of AB_PROD indicates more earnings management. The higher the negative value of AB_DISX indicates more earnings management.

I compute real earnings management in two ways. For the hypotheses 1a and 1b, I follow the paper of Kim et al. (2012). They computed total real earnings management

(20)

20 | P a g e

(RM_PROXY) as AB_CFO – AB_PROD + AB_DISX. According to this measure, when the proxy for real earnings management (RM_PROXY) decreases, firms engage in more aggressive earnings management through real activities. For the hypotheses 2a and 2b, the papers of Cohen et al. (2008), Zang (2012) and Doukakis (2014) are used to compute the total amount of real earnings management (RM_PROXY1). Firms can manipulate earnings upwards through reductions in research and development. This results in lower values of abnormal. This is due to fact that realized values of research and development will be lower than the normal values of research and development. Therefore AB_CFO and AB_DISX are first multiplied by -1 (AB_CFO1 and AB_DISX1). By multiplying with -1, the higher amount of

AB_CFO1 and AB_DISX1 will indicate that firms are more likely to be engaging in real sales

manipulation through price discounts and cutting discretionary expenses (Doukakis, 2012, p. 559). So in aggregate, the higher the amount of the sum of these three standardized variables (AB_CFO1 + AB_PROD + AB_DISX1 = RM_PROXY1) will indicate more earnings management through real activities. The proxy for total real earnings management will be used as a dependent variable in the regression models. The three individual variables can have different implications for earnings management that the total real earnings management. Therefor the results of the three individual variables are also reported.

3.3 Used regression models specification

For the hypotheses 1a and 1b the models 7 and 8 are used. The variable of interest in these models is POST2005. This is a indicator variable that indicates when an observation is from the years 2005 or later. The indicator variable POST2005 is given an “1” if the observation year is 2005 or later. This indicates when IFRS is mandatory adopted in the EU. If the observation year is from the year 2004 or before, the indicator variable POST2005 is given a “0”. A positive significant coefficient for POST2005 in model 7 would indicate that earnings management through accruals have increased after mandatory adoption of IFRS. A negative significant coefficient for POST2005 in model 8 would indicate that earnings management have increased through real earnings management after mandatory adoption of IFRS. Total amount of real earnings management (RM_PROXY) is included in model 7, and the absolute value of discretionary accruals (ABS_DA) is included in model 8 to control for the effect of these earnings management on the models. This is also done in the models of Kim et al. (2012) and Doukakis (2012).

(21)

21 | P a g e

For the hypotheses 2a and 2b the models 9 and 10 are used. The variable of interest in these models are RM_PROXY1xPOST2005 for model 9 and ABS_DAxPOST2005 for model 10.

RM_PROXY1xPOST2005 and ABS_DAxPOST2005 are interaction terms. The interaction

term in model 9 is generated by multiplying the total real earnings management (RM_PROXY1) with the indicator variable POST2005. The effect of this interaction variable is that model 9 only takes the total real earnings management from the post-mandatory IFRS adoption period into account. The interaction term in model 10 is generated by multiplying

ABS_DA with POST2005. This result in only taking the absolute discretionary accruals

(ABS_DA) from the post-mandatory IFRS adoption period as a control variable in model 10.

ABS_DA is the absolute value of discretionary accruals, where discretionary accruals are

computed through the cross-sectional modified Jones model. RM_PROXY and RM_PROXY1 are a proxy for the total real earnings management. Prior literature show that earnings management is affected by different factors. Prior literature mentions factors as the size of the firm, market to book value, profitability, type of the auditor and financial leverage. The control variable Size is the natural logarithm of the total assets of a firm in year t. It controls for the size of an observation. According to Watts and Zimmerman (1990), the variable Size can indicate for political attention. Larger firms are expected to prefer income-decreasing earnings management, because it is expected that for larger firms potential government scrutiny increases. MB is the market-to-book equity ratio, measured as the market value of equity over the book value of equity. The MB gives the growth opportunities and controls for the impact it has on the growth in earnings management. Managers can use income increasing earnings management when growth slows down (Summers and Sweeney, 1998). MB is calculated by common shares outstanding multiplied by the market price end of year scaled by the difference between total assets and total liabilities. ROE is the return on equity, and is calculated as net income before extra items/preferred dividends divide by last year’s total shareholders’ equity. This controls for the relationship between profitability and earnings

(22)

22 | P a g e

management. BIG4 is an indicator variable that indicates when a firm hires one of the Big 4 audit firms. Prior literature (Francis et al., 1999) suggests that when a Big 4 audit firm is hired, accrual-based earnings management will be reduced. The indicator variable is given a “1” if the audit is done by a Big 4 audit firm. “0” is given if the audit is done by a non-Big 4 audit firm. Leverage is calculated by the ratio of total liabilities in year t to last year’s total assets. Prior literature (DeFond & Jiambalvo, 1994) found evidence that highly leveraged firms have incentives to use income-increasing earnings management to avoid violation of debt covenant. The variable INSTIOWNER is for the institutional ownership. This controls for the role of ownership structure on earnings management. This is measured as the percentage of closely held shares. Following prior studies (Koh (2003), Aubert & Grudnitski (2012), Doukakis (2012)), income increasing earnings management are negatively associated with high institutional ownership levels. This suggest that institutional ownership can limit aggressive earnings management of managers. This control variable is only used in models 9 and 10, because by using it in models 7 and 8 would result in to many missing observations. Dummy variable SUSPECTED_EM is used in models 7 and 8. Observation that are suspected in managing earnings have been given an “1”, if not a “0” have been given. The dummy variable INDUSTRY is included to control for industry effects on earnings management. This variable indicate if the firm is in a certain industry. Two-digit SIC code is used for this variable. If the firm is in a certain industry, the dummy variable have been given an “1”, if not a “0” have been given. (For a summary of the variable definitions, see Appendix A)

3.4 Data Sample

The data for this research is collected from countries in the European Union. The countries consist of Austria, Belgium, Cyprus, Czech Republic, Denmark, Finland, France, Germany, Greece, Hungary, Ireland, Italy, Lithuania, Luxembourg, Norway, Poland, Portugal, Spain, Sweden, Switzerland, The Netherlands and United Kingdom. All the necessary data are collected from Datastream and Compustat databases for the years 1998 till 2012. Datastream is namely used because this database has a large base of European data. Compustat is used in case when data is not available from Datastream. Missing values are deleted from the sample.

The sample consist of European publicly listed firms that have adopted mandatory IFRS in the year 2005. The range from this research ranges from 2000 till 2012. For hypotheses 1a and 1b a sample is used that consist of 51.252 listed firm-year observations. The sample consist of 22 European countries. In Panel A of Table 1 gives a summary of the amount of observations per country. The country with the most observations in this sample is

(23)

23 | P a g e

the United Kingdom (27.72 percent), followed by Germany (13.11 percent) and France (12.23 percent). Periods after the mandatory IFRS adoption in 2005 (post-IFRS) have been compared with the periods before the mandatory IFRS is adopted (pre-IFRS). For calculating discretionary accruals and total real earnings management for this sample, at least 10 companies are needed per industry. This is needed to compare a company with its industry.

Hypotheses 2a and 2b uses a different sample. This sample is collected from listed European firms that are suspected to have managed earnings in the pre-IFRS and post-IFRS periods. For the second hypotheses, firms that manage earnings is of interest. A sample is taken that only consist of firms that are suspected of earnings management. This increases the power of the results about the trade-off of various forms of earnings management. This sample consist of 1.656 firm-years observations. Panel B of Table 1 shows a summary of the amount of observations per country. The country with the most observations is the United Kingdom (17.93 percent), followed by France (15.1 percent) and Germany (13.77 percent).

Following the researches of Bartov et al. (2002), Roychowdhury (2006), Reichelt and Wang (2010) and Zang (2012), firm-years for this sample are classified as firms that manage earnings when they meet or just beat earnings benchmarks/forecast. It is more likely that earnings management occur when firms just beat or meet important earnings benchmarks. This can be measured by looking at the forecast errors. Forecast errors are calculated by the comparing actual data with expected/forecasted data.

Roychowdhury (2006) defined firm-years that are beating or meeting the zero benchmarks as firm-years with earnings before extraordinary items over lagged assets when forecast errors are between 0 and 0.005. Zang (2012) measure firm-years that beating/meeting last-year earnings as the change in basic earnings per share (EPS), excluding extraordinary items from last year, is between 0 and 0.02. Reichelt and Wang (2010) and Chi et al. (2011) used forecast errors between 0 and 0.01. Consistence with prior literature, I take firm-years that are meeting or just beating earnings benchmarks by comparing the actual EPS with the expected EPS. Therefore I take a forecast error of between 0 and 0.02. If a firm-year observation have a forecast error within this range, this firm-year observation have been included in this sample. If it is not within this range, observations have not been taken into the sample. Firms that beat the analyst forecast by a larger amount are more likely to have done this due their performance than by managing earnings.

Managers receives incentives from shareholders to reach a certain earnings benchmark to receive a bonus. According to Barth et al. (1999), managers will try to avoid reporting decreasing earnings and try to maintain a consistent small increase in earnings. The likelihood

(24)

24 | P a g e

of managers engaging in earnings management will increase. Therefore I also include firms-year observations in this sample when earnings did not increase compared to the previous year, and firm-year with a relative small increase in earnings before interest and taxes compared to the previous year. A relative small increase is defined by an increase of 2 percent or smaller per year.

For calculating discretionary accruals and total real earnings management for this sample, at least 3 companies are needed per industry. This amount is lower than the first sample, because this sample is much smaller. By using a higher amount of companies needed in an industry would limit this sample too much.

Financial institutions have been excluded from both samples. Financial institutions like banks, insurance and investment companies have been marked with a SIC code between 6000 and 6999.

Table 1

Sample composition by country Country

Panel A:

Firm-years sample for H1 Number of Observations

Panel B:

Firm-years sample for H2

% Number of Observations % Austria (OE) 703 1.37 14 0.85 Belgium (BG) 967 1.89 34 2.05 Cyprus (CP) 362 0.71 7 0.42 Czech Republic (CZ) 153 0.30 1 0.06 Denmark (DK) 1336 2.61 26 1.57 Finland (FN) 1393 2.72 113 6.82 France (FR) 6266 12.23 250 15.10 Germany (BD) 6717 13.11 228 13.77 Greece (GR) 2242 4.37 57 3.44 Hungary (HN) 278 0.54 4 0.24 Ireland (IR) 660 1.29 74 4.47 Italy (IT) 2161 4.22 174 10.51 Lithuania (LN) 177 0.35 6 0.36 Luxembourg (LX) 276 0.54 9 0.54 Norway (NW) 1699 3.31 29 1.75 Poland (PO) 2631 5.13 37 2.23 Portugal (PT) 602 1.17 46 2.78 Spain (ES) 888 1.73 66 3.99 Sweden (SD) 3841 7.49 70 4.23 Switzerland (SW) 2124 4.14 57 3.44 The Netherlands (NL) 1567 3.06 57 3.44

United Kingdom (UK) 14209 27.72 297 17.93

(25)

25 | P a g e

4. Results and analysis

This chapter contains the results and analyses of the regression models that are used to tests the hypotheses of this research. For the two different samples, two descriptive statistics is given. After the descriptive statistics the related regression analyses are discussed. Lastly, a summary is given of the results of the analyses.

4.1 Effect of mandatory IFRS adoption on earnings management 4.1.1 Descriptive statistics

The indicator variable POST2005, is used to identify if firms are in the post-mandatory IFRS adoption period. This indicator variable POST2005 takes a value of “1” if the observation is in the year 2005 or in a year after 2005, and “0” if the observation is in the year 2004 or before.

Table 2 presents the sample distribution by the two-digit SIC code industry. The most heavily represented industry with the most observations is Manufacturing (SIC code 20 - 39, with 43.37 percent of the total sample of Hypothesis 1). Followed by the industry Services (SIC code 70 - 89) with 26.25 percent and Transportation & Public Utilities (SIC code 40 - 49) with 9.93 percent.

Table 2

Sample Description: Distribution of Firm-year Observations by Industry for H1

Two digit Number of % of Cumulative

Industry SIC Observations Sample Percent

Agriculture, Forestry, Fishing 01 - 09 383 0.75 0.75

Mining 10 - 14 3034 5.92 6.67

Construction 15 - 17 1954 3.81 10.48

Manufacturing 20 - 39 22229 43.37 53.85

Transportation & Public Utilities 40 - 49 5088 9.93 63.78

Wholesale Trade 50 - 51 2236 4.36 68.14

Retail Trade 52 - 59 2874 5.61 73.75

Services 70 - 89 13454 26.25 100.00

Total 51252 100

Table 3 presents descriptive statistics and Pearson correlations for the hypotheses 1a and 1b (for variable definitions, see Appendix A). To make sure that the influence of the few extreme observations is reduced, winsorization is used. Therefore, all the continuous variables are winsorized at the first and ninety-ninth percentiles of their distributions. Panel A from Table 3 shows that the ABS_DA has a mean value of 0.097. The mean values of AB_CFO, AB_PROD,

(26)

26 | P a g e

the firms in our sample are audited by a Big 4 audit firm. Only 4.1 percent of the firms in our sample are suspected in managing earnings. For ABS_DA and AB_CFO the initial observations of 51.252 is used. For AB_PROD, AB_DISX and RM_PROXY the amount of observations is lower. Panel B presents a closer look of the mean of the dependent variables with the dummy variable POST2005. The mean value of RM_PROXY from post-2005 suggesting that, on average, firms engage more in real earnings management compared to pre-2005. The value of RM_PROXY decreases, as firms engage in more aggressive earnings management through real activities. The amount of accrual-based earnings management is almost the same.

Panel C presents the Pearson correlations coefficients for the dependent and control variables within the sample for hypothesis 1. This table indicates that the absolute value of discretionary accruals are significantly and negatively correlated with POST2005. Whereas for RM_PROXY there is a small positive correlation, but not significant (p-value of 0.3206). In the correlation table, you can also observe that ABS_DA is significantly negative correlated with Size and BIG4, which indicates that a large firm or firm that is audited by a Big 4 audit firm, discretionary accruals will decrease. There is also a smaller significant negative correlation between RM_PROXY and Size, but no significant correlation between RM_PROXY and BIG4 (p-value = 0.1103).

Figures 1, 2, 3 and 4 in Appendix B provide graphical illustrations of the medians of the dependent variables. These graphs provide the median of every year of this sample.

Table 3, Panel A

Descriptive Statistics of Selected Variables for H1

N Mean Median Std. Dev. 25th Prc. 75th Prc. Dependent Variables ABS_DA 51252 0.097 0.052 0.149 0.022 0.112 DA 51252 -0.012 -0.007 0.178 -0.060 0.043 AB_CFO 51252 0.025 0.042 0.159 0.010 0.075 AB_PROD 49401 0.869 0.738 0.724 0.406 1.143 AB_DISX 50452 0.433 0.213 14.958 0.114 0.362 RM_PROXY 48727 -0.415 -0.434 15.204 -0.777 -0.192 Control Variables POST2005 51252 0.704 1.000 0.457 0.000 1.000 Size 51252 12.069 11.900 2.391 10.449 13.582 MB 51252 2.718 1.494 84.249 0.833 2.692 ROE 51247 0.124 0.071 42.557 -0.084 0.183 BIG4 51252 0.636 1.000 0.481 0.000 1.000 Leverage 51252 0.813 0.569 10.042 0.381 0.748 SUSPECTED_EM 51252 0.041 0.000 0.199 0.000 0.000

(27)

27 | P a g e SIC_Agriculture 51252 0.007 0.000 0.086 0.000 0.000 SIC_Mining 51252 0.059 0.000 0.236 0.000 0.000 SIC_Construction 51252 0.038 0.000 0.192 0.000 0.000 SIC_Manufacturing 51252 0.434 0.000 0.496 0.000 1.000 SIC_Transportation 51252 0.099 0.000 0.299 0.000 0.000 SIC_Wholesale 51252 0.044 0.000 0.204 0.000 0.000 SIC_Retail 51252 0.056 0.000 0.230 0.000 0.000 SIC_Services 51252 0.263 0.000 0.440 0.000 1.000

Variables are defined in Appendix A.

Table 3, Pabel B

Descriptive Statistics of Dependent Variables means pre- and post-2005 for H1

Period ABS_DA AB_CFO AB_PROD AB_DISX RM_PROXY

Pre-2005 0.099 0.027 0.939 0.399 -0.522

Post-2005 0.096 0.024 0.841 0.447 -0.371

Total 0.097 0.025 0.869 0.433 -0.415

Variables are defined in Appendix A. Table 3, Panel C

Correlations among the dependent and control variables for H1

1 2 3 4 5 6 1 ABS_DA 1 2 DA -0.1977*** 1 3 AB_CFO -0.3411*** 0.0513*** 1 4 AB_PROD 0.0708*** 0.1058*** 0.2042*** 1 5 AB_DISX 0.037*** 0.0058 -0.0418*** 0.0202*** 1 6 RM_PROXY 0.0288*** 0.0057 -0.0422*** -0.0243*** 0.999*** 1 7 POST2005 -0.01** 0.004 -0.0095** -0.0616*** 0.0015 0.0045 8 Size -0.2967*** 0.0322*** 0.3524*** -0.0092** -0.0336*** -0.0284*** 9 MB -0.005 -0.0329*** -0.0088** -0.0027 0.0005 -0.0002 10 ROE 0.011** -0.0145*** -0.0018* 0.0003 0.0001 0 11 BIG4 -0.156*** -0.0065 0.188*** 0.0241*** -0.0085* -0.0072 12 Leverage 0.0843*** 0.007 -0.0786*** 0.0809*** 0.0077* 0.0033 13 SUSPECTED_EM -0.0383*** 0.0185*** 0.0256*** -0.0057 -0.0015 -0.0009 14 SIC_Agriculture 0.0058 0.0178*** 0.0083* -0.0152*** -0.0014 -0.0005 15 SIC_Mining 0.0802*** -0.0157*** -0.1197*** -0.1748*** 0.0029 0.0097** 16 SIC_Construction 0.003 0.0127*** -0.0097** 0.0501*** -0.0038 -0.0063 17 SIC_Manufacturing -0.0933*** 0.0237*** 0.0604*** -0.038*** -0.008* -0.0055 18 SIC_Transportation -0.0154*** 0.0081* 0.0268*** -0.0911*** -0.0033 0.0011 19 SIC_Wholesale -0.0184*** 0.0076* 0.012*** 0.1801*** -0.0025 -0.011** 20 SIC_Retail -0.0046 -0.0004 0.0624*** 0.1492*** 0.0243*** 0.0187*** 21 SIC_Services 0.0811*** -0.0362*** -0.0576*** 0.0153*** 0.0007 -0.001

* Indicate a statistical significance at the ≤ 0.10 levels ** Indicate a statistical significance at the ≤ 0.05 levels *** Indicate a statistical significance at the < 0.01 levels Variables are defined in Appendix A.

(28)

28 | P a g e Panel C, Correlations among the dependent and control variables for H1 (continued)

7 8 9 10 11 12 7 POST2005 1 8 Size 0.0023 1 9 MB -0.0012 -0.0028 1 10 ROE 0.0004 0.0004 0.0019 1 11 BIG4 -0.0525*** 0.4581*** 0.0042 -0.0011 1 12 Leverage -0.004 -0.0044 -0.0001 0.0071 -0.011** 1 13 SUSPECTED_EM 0.0061 0.058*** -0.0034 0 0.044*** -0.0017 14 SIC_Agriculture 0.0176*** 0.0068 -0.0013 -0.0002 0.0015 -0.0015 15 SIC_Mining 0.0668*** -0.063*** 0.0027 -0.0027 -0.0473*** -0.0048 16 SIC_Construction 0.0194*** 0.0747*** -0.0012 -0.0003 0.0007 -0.0009 17 SIC_Manufacturing -0.0301*** 0.098*** 0.0013 -0.0016 0.0688*** -0.0097** 18 SIC_Transportation 0.0177*** 0.1964*** -0.0034 -0.0021 0.0739*** 0.0179*** 19 SIC_Wholesale -0.0183*** -0.0011 -0.0169*** -0.0006 -0.0374*** -0.0027 20 SIC_Retail -0.0255*** 0.0352*** -0.0022 -0.0004 -0.0036 -0.0032 21 SIC_Services -0.0041 -0.2617*** 0.0091** 0.0053 -0.0837*** 0.005

Panel C, Correlations among the dependent and control variables for H1 (continued)

13 14 15 16 17 18 13 SUSPECTED_EM 1 14 SIC_Agriculture 0.0081* 1 15 SIC_Mining 0.0077* -0.0218*** 1 16 SIC_Construction -0.0086* -0.0173*** -0.0499*** 1 17 SIC_Manufacturing -0.0069 -0.0759*** -0.2195*** -0.1742*** 1 18 SIC_Transportation 0.0293*** -0.0288*** -0.0833*** -0.0661*** -0.2905*** 1 19 SIC_Wholesale 0.0002 -0.0185*** -0.0536*** -0.0425*** -0.1869*** -0.0709*** 20 SIC_Retail 0.0013 -0.0211*** -0.0611*** -0.0485*** -0.2133*** -0.0809*** 21 SIC_Services -0.0149*** -0.0518*** -0.1497*** -0.1188*** -0.5221*** -0.1981***

Panel C, Correlations among the dependent and control variables for H1 (continued)

19 20 21

19 SIC_Wholesale 1

20 SIC_Retail -0.0521*** 1

21 SIC_Services -0.1274*** -0.1454*** 1

4.1.2 Results of Accrual-based Earnings Management and Post-2005 period.

Panel A of Table 4 presents the results of the regression analysis of model 7 for hypothesis 1a. This regression analysis looks at discretionary accruals. The R-squared is 12.04%. R-squared tells us how much of the total variance of absolute discretionary accruals (ABS_DA) is

(29)

29 | P a g e

explained by this regression model. The Adj R-squared is roughly the same amount (12.02%). Adj R-squared takes into account the number of variables in the model. If a variable is added to the model, that does not affect the dependent variable, Adj. R-squared will decrease. The Adj. R-squared gives us more information about the regression model. This regression sees if there is any evidence of relationship between the control variables and dependent variable, controlled for the other control variables. POST2005 is the variable of interest. The coefficient of this variable is negative (-0.005) and significant (p-value < 0.01). This indicates that in the period after mandatory IFRS adoption (post-2005) there is, on average, a decrease of the value ABS_DA by the amount of 0.005 compared to the period before mandatory IFRS adoption (pre-2005). This is not consistent with hypothesis 1a, which hypothesize that there is no significant relationship between accrual-based earnings management after mandatory IFRS implementation. The results indicates that firms are less likely to manage earnings through accruals after mandatory IFRS adoption. This result is consistent with the finding of Zéghal et al. (2011).

Table 4, Panel A

Regression results of Accrual-Based Earnings Management after mandatory IFRS adoption

Source SS df MS Number of observations 48723

Model 98.0850159 15 6.53900106 F( 15, 48707) 444.57 Residual 716.406527 48707 0.01470849 Prob > F 0 Total 814.491543 48722 0.01671712 R-squard 0.1204 Adj R-squard 0.1202 Root MSE 0.12128

ABS_DA Coef. Std. Err. t-test P > | t | [95% Conf. Interval]

POST2005 -0.0050046 0.0012183 -4.11*** 0.000 -0.0073925 -0.0026167 RM_PROXY 0.00016 0.0000362 4.43*** 0.000 0.0000892 0.0002309 Size -0.0163832 0.0002725 -60.12*** 0.000 -0.0169174 -0.0158491 MB 0.0000403 0.00000917 4.4*** 0.000 0.0000224 0.0000583 ROE 0.0000546 0.0000134 4.06*** 0.000 0.0000282 0.0000809 BIG4 -0.0058946 0.001295 -4.55*** 0.000 -0.0084329 -0.0033563 Leverage 0.0018317 0.0000832 22.02*** 0.000 0.0016686 0.0019947 SUSPECTED_EM -0.0150954 0.0027352 -5.52*** 0.000 -0.0204565 -0.0097343 SIC_Agriculture -0.0029066 0.0071015 -0.41 0.682 -0.0168256 0.0110125 SIC_Mining 0.0212282 0.0036479 5.82*** 0.000 0.0140782 0.0283782 SIC_Manufacturing -0.0292792 0.0029331 -9.98*** 0.000 -0.0350281 -0.0235303 SIC_Transportation -0.0016898 0.0033123 -0.51 0.610 -0.0081819 0.0048023 SIC_Wholesale -0.0326313 0.0038336 -8.51*** 0.000 -0.0401453 -0.0251174 SIC_Retail -0.0218595 0.0037128 -5.89*** 0.000 -0.0291366 -0.0145824 SIC_Services -0.0169669 0.0030475 -5.57*** 0.000 -0.0229401 -0.0109937 _cons 0.3170198 0.0043979 72.08*** 0.000 0.3083998 0.3256398

*** Indicate a statistical significance at the < 0.01 levels Variables are defined in Appendix A.

(30)

30 | P a g e

Panel A Table 4 also shows a significant and positive coefficient for the variable

RM_PROXY. This means if firms decreases real earnings management activities, firms are

more likely to manage earnings through accrual-based earnings management. The coefficient of Size is significant and negative. This indicates that lager firms are less likely to manage earning management through accruals. MB and ROE are both significant and have positive coefficients. This indicates that if firms are more likely to manage earnings through accruals if they have more growth opportunities and are more profitable. The indicator variable BIG4 is also significant and negative. This indicates if a firm is audited by a Big 4 audit firm, they are less likely to manage earnings through accruals. The coefficient of Leverage is also significant and positive. This means that highly leveraged firms are more likely to manage earnings through accrual-based earnings management. The indicator variable

SUSPECTED_EM is significant and negative. This indicates that firms that meet or just beat

(31)

31 | P a g e Table 4, Panel B

Regression results of Positive_DA and Negative_DA variables

Positive_DA Coefficient (t-stat) Negative_DA Coefficient (t-stat) POST2005 -0.0032 0.0065 (-1.79)* (3.97)*** RM_PROXY 0.0021 -0.0000 (14.88)*** (-0.52) Size -0.0152 0.0172 (-37.12)*** (47.25)*** MB 0.0001 -0.0000 (1.57) (-4.13)*** ROE -0.0002 -0.0001 (-1.34) (-4.02)*** BIG4 -0.0076 0.0046 (-4.04)*** (2.59)** Leverage 0.0033 -0.0014 (19.35)*** (-14.28)*** SUSPECTED_EM -0.0061 0.0230 (-1.56) (6.08)*** SIC_Agriculture -0.0089 -0.0341 (-0.97) (-3.09)*** SIC_Mining 0.0010 0.0033 (0.18) (0.29) SIC_Manufacturing -0.0368 0.0266 (-8.58)*** (2.51)** SIC_Transportation -0.0107 -0.0020 (-2.21)** (-0.18) SIC_Wholesale -0.0360 0.0322 (-6.39)*** (2.9)*** SIC_Retail -0.0292 0.0190 (-5.44)*** (1.71)* SIC_Services -0.0308 0.0098 (-6.89)*** (0.92) _cons 0.3058 -0.3278 (47)*** (-28.75)*** R-squared 0.1163 0.1374 Adj R-squared 0.1157 0.1369 Number of observations 22074 26649

Note: variable SIC_Construction is omitted because of collinearity.

Panel B presents the results of regression analyses of Positive_DA and Negative_DA. The coefficients and t-stat are reported with the indication if it is significant.

Trade-off effect between accrual based earnings management and real earnings management after the adoption of mandatory IFRS : evidence from listed companies in 22 European countries