Effect of adoption of IFRS on earnings management

(1)

Effect of Adoption of IFRS on Earnings Management 1

Master in International Finance

2016 - 2018

Master Thesis

Effect of Adoption of IFRS on Earnings Management

Student Name: Manika Saxena Student Number: 11152869 Supervisor: Dr. S.R. (Stefan) Arping

(2)

Abstract

Purpose

International Accounting Standards Board (IASB) has developed a high quality and universal accounting standards International Financial Reporting Standards (IFRS), with the aim of transparency in financial statements. This paper sets out to examine whether the adoption of IFRS has led to lower levels of earnings management.

Design/Methodology/Approach

For the purpose, I investigate all the countries of the European Union block and listed firms with a revenue of at least €1 million. The resulting dataset comprises 350 firms and nearly 5600 observations. The period chosen has been from 2001 - 2015, dividing them further into bins of 3: namely (2001 to 2005: pre-IFRS, 2006 to 2010: post-IFRS – the early adoption years and 2011 to 2015: post-IFRS – the later maturity years). To calculate the discretionary accruals as widely accepted in previous literature, Modified Jones Model has been used.

Findings

The results do show a decline in earnings management; however, it is a decreasing trend with time. Interestingly the major drops appear during 2001 (year of voluntary IFRS adoption) and 2005 (year of mandatory IFRS implementation). However, IFRS alone cannot be attributed as the sole reason for lower levels of discretionary. With this, my findings are in line with the existing debate whether the IFRS can solely be the reason for the decline in earnings management.

Keywords

Earnings Management, Accounting Standards, Discretionary Accruals, Earnings Distribution, IFRS, European Union.

(3)

Table of Tables and Figures

Table 4.1 Number of Firms by Country ... 31

Table 4.2 Descriptive Statistics of Financial Data ... 32

Table 5.1 Coefficients estimated by the Modified Jones Model (2001 – 2015) ... 34

Table 5.2 Determinants of Discretionary Accrual ... 38

Table 6.1 Comparison between the Means of Discretionary Accrual in sub-period .... 40

Figure 6.1 Year group Means of absolute discretionary accrual ... 41

(5)

1. Introduction

1.1 Introduction to the subject

A firm in profit is the major attractor for the stakeholders and potential investors in the company. It is the reflection of its financial health and the prospect of growth. A profitable enterprise attracts vested interests from the investors having robust profits, plan and promising future performance. Due to this incentive, many firms or say, certain individuals, might be inclined to inflate or overstate financial performance.

Earnings manipulation has been the subject of extensive research since the 1960s (Burgstahler and Dichev 1997, Copeland 1968, DeAngelo 1986; Dechow and Sloan 1991, Glaum et al. 2004, Healy and Kaplan 1985, Jeanjean and Stolowy 2008, Jones 1991, Kirchheimer 1968, Leuz et al. 2003, Stolowy and Breton 2004). And yet, there have been numerous examples were scandals keep emerging every now and then. World’s largest accounting scandals concerning big weights like that of Enron, Worldcom, Tyco, Waste Management, Healthsouth, Freddie Mac, AIG, Lehman Brothers, Bernie Madoff and Satyam together accounted for $213.8 billion of investors’ wealth1_.

On the other hand, efforts to standardize accounting standards and thus increasing transparency and reporting quality have been a continuous task. Efforts in this direction were previously led by International Accounting Standards Committee (IASC) and later the International Accounting Standards Board (IASB). They are responsible for publishing International Financial Reporting Standards (IFRS).

With this paper, I will set out to examine “Whether the IFRS adoption has affected the earning management over the years”

(6)

Both by identifying and discouraging such practices that affect earnings management and adjustment techniques help stakeholders make sound financial decisions.

In the existing literature different terms and definitions for earnings management exist. For the purpose of this study according to Wagenhofer and Ewert (2007), it is defined as:

“Earnings management is the figuration and adoption of financial statements, conducted mostly by executive staff, by means of corporate policy or accomplishing personal targets, within or without the limits of statuary regulations.”

Globalization and breaking of borders have driven financial markets towards enhancement in the degree of corporate comparability and transparency on consolidated financial reporting. Thus, during the last decades’ various measurements and different approaches have been developed, e.g. the use of discretional accruals or the distribution of income to measure earnings management.

Attempts to eliminate earnings manipulation is one step towards the correction of accounting standards. IFRS standards set out to publish a de-facto approach to meeting the increasing capital market demand aiming for better comparability and transparency of cross-border companies. It aims to become one uniform accounting rule worldwide because local standards insufficiently meet requirements of comparable financial statements. Since its inception, out of 47,874 domestic listed companies on the 93 securities exchanges in the world, over 27,000 use IFRS Standards. And of those domestic listed companies that do not use IFRS Standards, nearly 90 percent are listed in China, India, Japan, and the United States.

From a global coverage perspective, nearly 166 jurisdictions or representing 98.8% of world’s GDP, including G20 countries have agreed to require or allow adoption of IFRS or have established timelines for IFRS adoption. Of which, 144 jurisdictions (87 per

(7)

cent) require IFRS Standards for all or most domestic publicly listed companies and financial institutions. Of the remaining 22 jurisdictions that have not adopted2_:

● 12 jurisdictions permit, rather than require, IFRS Standards: Bermuda, Cayman

Islands, Guatemala, Honduras, Japan, Madagascar, Nicaragua, Panama, Paraguay, Suriname, Switzerland, Timor-Leste;

● One jurisdiction requires IFRS Standards for financial institutions but not listed companies: Uzbekistan;

● One jurisdiction is in process of adopting IFRS Standards in full: Thailand;

● One jurisdiction is in process of converging its national standards substantially

(but not entirely) with IFRS Standards: Indonesia; and

● Seven jurisdictions use national or regional standards: Bolivia, China, Egypt, India, Macao SAR, United States, and Vietnam.

Moreover, the US Securities and Exchange Commission (SEC) supports international compatibility that allows the presentation of financial statements in compliance with IFRS for foreign companies in the USA. The widespread acceptance of IFRS underlines their growing significance.

Of these, Europe accounts for 44 jurisdictions of which 43 require IFRS standards for all or most domestic publicly accountable entities and only one jurisdiction that permit or require the standards for at least some (but not all or most) domestic publicly accountable entities. In the European Union, IFRS was adopted mandatorily for publicly traded companies on the regulated market of EU Member States since their securities were admitted to trading on or after January 1 in the year 2005 (EC Regulation No. 1606/2002). European IFRS adoption embodies a crucial milestone in the financial reporting convergence process Armstrong et al. (2010). The expected benefit of implementing IFRS is to enhance comparability and transparency of financial information and thus increase the quality of financial reporting.

(8)

1.2 Research question and contribution to prior research

This study investigates through the central hypothesis:

“Whether the IFRS adoption has affected the earning management over the years”

Firstly, the impact of IFRS on quality reporting and adherence to the standards. Using discretionary accruals as a measure to evaluate earnings management.

Secondly, it aims to add to the already existing literature on the topic by measuring the overall impact of IFRS adoption in the elimination or reduction of the earnings management malpractice. The previous studies have been focused towards the earlier years of IFRS adoption. Whereas this report considers data from 2001 to 2015. Dividing them into three bins, namely (2001 to 2005: pre-IFRS, 2006 to 2010: post-IFRS – the early adoption years and 2011 to 2015: post-post-IFRS – the later maturity years).

When IFRS was first introduced back in 2001, many firms had gone through the uphill task of dealing with ambiguity, tardiness, and unclarity and thus the dilemma of what benefits does standardization bring. Given ample time since its first adoption, the companies by now have moved ahead in the maturity curve. It would be interesting to see that over these years what positive outcomes have universal standard adoption contributed towards the elimination or reduction of earnings management.

Finally, the primary data includes entire European Union which represents all national legal systems3_{- Civil Law (mostly EU mainland) and Common Law (mainly UK, Cyprus} and Ireland) and other flavors (French Civil Law, German Civil Law Scandinavian Civil Law). The major difference being shareholder and creditor protection and promoting financial development.4_{All listed companies within the EU block having more than €1} million revenue have been investigated using the widely accepted, in existing literature around this subject, Modified Jones Model for the analysis.

3_{Contrast between Common and Continental Legal Systems} 4_{International differences in Investor Protection}

(9)

1.3 Structure of the thesis

The thesis is divided into the following sections. Chapter 1: Introduction

Provides a brief on the history of IFRS and its current state and coverage. It introduces the practice of earnings management and the rationale behind it. In the latter part, it presents the topic and motivation of the research, hypothesis and methodology, data used in order to conduct the research. All of these aspects are discussed in details in relevant chapters.

Chapter 2: Literature Review

It underpins the main existing theories and literature that exists in the area of earnings management and their outcomes. The evidence is given out by the theories and the arguments either in favor of or against the motion.

Chapter 3: Methodology and Hypothesis

Chapter three presents the commonly existing econometric models, their pros, and cons and then selecting the model for my investigation. The chapter further develops the regression equation in multiple steps that would form the basis for my research. Chapter 4: Data and Descriptive Statistics

This section introduces the length and breadth of data being used in the research. It lists the parameters being used to select, reject and categorize the data. Gives a brief on why the data was selected and the rationale of bucketing them. It also enlists the steps and the source from where the data was gathered.

Chapter 5: Empirical Results

This summarizes the results of the regression model used selected during the study. Finally, it presents the main result.

(10)

Chapter 6: Robustness Checks

Additional statistical analysis and results to verify and cross-check the results and the approach is presented in this section.

Chapter 7: Conclusion and Discussion

The final conclusion is presented in the section. It summarizes the steps that lead to the results and their economic meaning in relation to the research topic. It analyzes if the findings are in accordance with the existing literature or not and its implications in general. It will then end by mentioning the limitations of this research, followed by some recommendations for future research topics.

(11)

2. Literature Review

Earnings management has been a subject of a plethora of literature and associated research since ages. The early years of researches and literature have been focusing on examining why, how and in what situations earnings management is pursued and what are the consequences of such a behavior (McNichols 2000, Szczesny 2007).

Much of the earlier studies focused their research on the motives, techniques, restrictions, and designs of earnings management. (Fields et al. 2001, Healy and Wahlen 1999, McNichols 2000) initially defined the structure of earnings management, presented implications for accounting criteria and regulation, and discussed earnings manipulation research designs.

(McNichols 2000, Szczesny 2007) points out that research designs on earnings management mostly use discretionary or specific accruals. This idea has been supported by various other researchers. On contrary, a minority of studies use the distributions of earnings (Degeorge et al. 1999, Burgstahler and Dichev 1997) or measures of income smoothing (Bouwman 2014, Cai et al. 2008) to examine earnings management.

Of the ones that focus on distribution of income smoothing to examine earnings management, many point towards the advantages and disadvantages of the distribution of income smoothing (Beaver et al. 2007, Daske et al. 2006, Dechow et al. 2003, Durtschi and Easton 2005, Glaum et al. 2004, Jacob and Jorgensen 2007, Jeanjean and Stolowy 2008).

According to the literature, managers adopt accounting discretion to maximize their compensation Healy and Palepu (2001) and such opportunistic behavior increase the firms’ agency costs Francis et al. (2005). It seems that the firms’ agency costs are directly correlated to the amount of discretion in the selection of the firms’ accounting procedures. However, the literature does also describe some cases in which granting

(12)

management with increased accounting discretion, proved to moderate the firms’ agency cost Adut et al. (2013). Here, predictable earnings can lower a firms’ information risk, which can also lower the firms’ costs Kravet and Shevlin (2010).

However, using accruals quality as the proxy for information risk, Francis et al. (2005) suggest the existence of some heterogeneity in managers’ behavior. While many managers use discretionary accruals to improve the reporting quality (decreasing information uncertainty), previous research on earnings management has also documented how managers, in some time periods, make accounting choices that reduce accruals quality (increasing information uncertainty) Francis et al. (2005). And information risk increases when investors are concerned with managers’ discretionary accounting choices Kravet and Shevlin (2010).

As previously discussed, adoption of one accounting standard across the world has set a stage to progress towards an unbiased and reasonable solution to harmonization and potential elimination of earnings management. IFRS is leading the way towards this goal. It has been adopted by EU and many other countries as their de-facto accounting and reporting standard since its inception in the year 2001.

In the EU, IFRS was mandatory for publicly traded companies on the regulated market of EU Member States from January 1 in the year 2005 (EC Regulation No. 1606/2002). European IFRS adoption marks an important step in the financial reporting Armstrong et al. (2010). The resulting benefit is to enhance comparability and transparency of financial information and thus increase the quality of financial reporting. Although some studies suggest (Barth et al. 2008, Daske 2008, Jeanjean et al. 2008, Callao et al. 2010) the economic ramifications of adopting IFRS, both confirming and rejecting the positive influence of it.

Mandatory adoption of IFRS in many European countries has brought a significant change in accounting procedures. Over the past years, there has been a continental shift from a rule-based approach to a principle-based one Callao et al. (2010). IASB

(13)

leans towards the principle-based approach. This provides managers’ as well as auditors’ professional judgment to ensure that a financial statement is economically sound and transparent. The interpretation of standards under the principle-based approach gives a room for providing easier standards enforcement and earnings manipulation. The rule-based approach, on the other hand, leaves little space for professional judgment Wüstermann et. al (2011).

Implementing IFRS as a universal accounting standard has a positive impact is supported by a considerable quantity of prior research studies. The result supports the idea that tightening accounting standards increase earnings quality. Armstrong et al. (2010) indicate that investors in European firms react positively and are aware of the net convergence benefits of IFRS adoption such as improvement in information quality, a decrease in information asymmetry, more rigorous enforcement of accounting standards, and convergence.

Internationally recognized standards such as IFRS require a high degree of disclosure of financial information. Theoretically, higher disclosure requirements lead to a decline in information asymmetry for investors. The information asymmetry together with the estimation risk may be reduced by one uniform accounting approach because it helps investors to distinguish between lower and higher quality companies. Lower risk for investors as a consequence of more informed valuation in the equity markets, elimination of international differences in financial reporting, adjustment analysis decreasing the investors’ costs, expected increased market efficiency, and removing of barriers to the cross-border acquisition are major advantages of IFRS adoption for investors Ball (2006). The IFRS may have a positive impact on companies switching from local accounting legislation to internationally recognized standards Daske et al. (2006). Furthermore, accounting diversity could affect the level of cross-border investment Bradshaw et al. (2004).

Even voluntary IFRS adoption may be beneficial for investors Covrig et al. (2007), but on the other hand, Jeanjean et al. (2008) criticize the benefits of voluntary IFRS

(14)

adoption because it suffers from an adverse selection bias based on the assumption that only companies expecting advantages from IFRS adoption would make it. Daske et al. (2006) indicate that both firms which have voluntarily adopted internationally recognized accounting standards (IFRS or US GAAP) and those which are subject to mandatory (IFRS or US GAAP) adoption have increased their reporting quality significantly with the IFRS adoption. Those and similar findings tend to show a positive influence of one shared set of standards (IFRS in this case).

Despite the benefits of IFRS, especially the aim to unify accounting rules worldwide, the desired uniformity may remain an elusive achievement due to a lack of adequate changes in other accompanying institutions Ding et al. (2007). Countries with low demand for information from published financial reports might be reluctant to adopt any common accounting standards that emphasize value relevance. Those countries lead to use accounting practice producing financial data with low-value relevance.

Global accounting language plays merely a partial role in improving comparability of accounting information (Leuz 2003, Burgstahler et al. 2006). IFRS aim to become global accounting standards but the economic and politic environment which drives the incentives of preparers and enforcers remains local. Those political and economic factors such as the legal system, strength of the capital market and the enforcement mechanism Burgstahler et al. (2006) influencing managers’ and auditors’ incentives are underlying factors of reporting quality (Ball et al. 2003, Ball 2006). In addition, the legal environment – in particular, the jurisdiction – should be prepared to implement IFRS. The absence of effective control and infrastructure might be relevant drivers of IFRS failure Dao (2007). Cultural differences among countries make it more difficult to fulfill requirements on comparability. One of the key terms in IFRS, the definition of fair value, may be crucial Ball (2006) in understandability within cross-border companies.

Despite limitations borne by IFRS (Leuz 2003, Burgstahler et al. 2006), it does contribute in the reduction of costs associated, especially in case of cross-border

(15)

transactions and investments Covrig et al. (2007). Tendello et al. (2005) found no difference in earnings management behavior between companies reporting under IFRS and local GAAP for instance. The actual net benefit of voluntary and mandatory adoption of IFRS is still an open question. To answer that inquiry is not an easy task regarding past differences in literature findings. The issue itself provides fertile ground for future research. The accounting quality is largely a determinate of firms’ reporting incentives created by market forces and institutional factors rather than by accounting standards (Leuz 2003, Ball et al. 2003, Ball 2006). Companies’ reporting incentives are different and the strength of enforcement differs considerably across countries. Strong enforcement mechanisms are associated with lower earnings manipulation and higher disclosure quality Daske (2008); therefore, IFRS themselves may merely mean material changes in reporting quality.

On contrary, countries with a weaker legal system and enforcement mechanism have stronger earnings management Burgstahler et al. (2006) for both public and private firms; thus, a stronger effect of IFRS appears in countries with a stronger power to enforce the rules.

Leuz et al. (2003) find that consistent strong protection limits insiders’ ability to acquire private control benefits, reducing their incentives to mask firm performance and they support an endogenous link between corporate governance and reported earnings quality. Apparently, the global accounting debate concentrates too much on the standardization and too little on institutional factors and market forces.

Aubert et al. (2012) reported that level of earnings manipulation in European countries for the pre-IFRS period (1997 to 2003) and post-IFRS period (2006 to 2008), had a positive influence of IFRS. Those and similar studies summarize that benefits for investors from IFRS adoption exceed the expected costs. In contrast, Callao et al. (2010) gave evidence of increased earnings management after the IFRS adoption. Nevertheless, they showed the relationship between the accruals and the institutional variables as a negative one for both the periods, which implies that the level of

(16)

investor protection and legal enforcement used in the implementation of the standards helped to keep in check those manipulative practices. Disproving findings associated with IFRS benefits are also provided by Jeanjean et al. (2008), who examined 1146 firms in Australia, France, and the United Kingdom. They suggested that uniformity of accounting rules is not sufficient in itself to create a common business language.

Brüggemann et al. (2013) summarized empirical evidence on financial reporting effects and divided them into three categories: compliance and accounting choice studies, accounting properties studies, and value relevance studies. Among them, only one-third of the studies mentioned are consistent with the IFRS regulatory objectives. Also, findings from studies such as (Ahmed et al. 2012, Atwood et al. 2011, Lang et al. 2010) disprove consistency with the IFRS regulatory objectives.

It is therefore considered that earnings management is the management intervention in the production process and reporting of accounting information in order to obtain certain self-benefits. The manipulation may further comprise the actual manipulation, in the sense that the manipulation of the results may result from the choice of the right moment for making funding or investments decisions Schipper (1989).

Some other theories such as the positive accounting theory also support the explanation of the effect of the bonus plan, the existence of debt equity and political costs that motivates on earnings management measures and the theory of economic consequences that support the explanation regulation or policy relevance of accounting rules on earnings management. The main purpose is to explain the positive accounting approach and predicts the standard choice by management to analyze the costs and benefits of certain financial disclosures related to various individuals and resource allocation in the economy Scott (1997) define economic consequences as an accounting policy choice that can affect the value of the company.

(17)

Other researches also support the results of research on the differences in the level of earnings management before and after the IFRS adoption. Christensen et al. (2015) who found that there is a decrease in earnings management after the adoption of IFRS, especially for companies that implement them voluntarily. Similarly, (Barth et al. 2008, Lippens 2010, Jeanjean et al. 2008) show a difference in the level of earnings management before and after IFRS adoption.

IFRS, as a globally recognized body of standards, is expected to lower transactions costs for foreign users of financial statements. That is, foreign financial statement users already familiar with IFRS through their own adoption of the standards are expected to incur lower barriers in analyzing overseas financials prepared under IFRS, which in turn can result in benefits accruing to entities reporting under these standards.

(18)

3. Methodology and Hypothesis

The listed companies in the European Union have been applying IFRS for their consolidated statements. Over time, it is essential to measure if there is any convincing evidence that adopting IFRS had either a positive or negative influence in earning management? What substantial results in reporting can support this primary change?

3.1 Hypothesis Development

The premise of mainly studying European Unions’ adoption of IFRS was that collectively it’s a mature economic market. Also, the quality of IFRS standard is higher than that of the local standards(Leuz 2000 and 2003, Ashbaugh 2008, Barth 2007 and 2008). EU has been following the IFRS since 2005. For that reason, the assumption is that the implementation of IFRS will have a positive effect on the overall accounting quality over the years since its adoption.

Based on this premise, it is expected that the adoption of IFRS eventually improves the earnings quality and restricts management discretion thus providing a more consistent approach to accounting measurement Daske et al. (2008). The findings of Ewert and Wagenhofer (2005) are also on the same lines, tightening of accounting standards can reduce the level of earnings management and, therefore, improve the quality of earnings. Capkun et al. (2008) argue that earnings management during the transition period may reduce the earnings management options in the next reporting period.

Since the adoption of IFRS in EU in 2005, it has been more than a decade worth of reporting data and confusion diffusion or cool off period for early adopters. I will explore the current status of IFRS adoption impact and outcomes through the central question - “Whether the IFRS adoption has affected the earning management over the years”

(19)

3.2 Research Design

3.2.1 Available Research Models

As discussed previously in the introduction section, earnings quality is primarily measured by discretionary accruals. It is often used for measuring earnings quality in the literature (Klein 2002, Myers et al. 2003, Menon and Williams 2004, Ashbaugh-Skaife et al. 2008). Discretionary accruals reflect misguided earnings reporting by managerial manipulation Dechow et al. (2010).

For measuring discretionary accruals, the best way is to investigate first the total accruals. Then by using the chosen model from the already existing ones the total accruals are divided into discretionary and nondiscretionary components.

This study considered five available models5_{for the purpose. They are widely available} and extensively used in earnings management research and literature. Below, I enumerate all models to facilitate comparability and finally narrow on one for the purpose of my research.

The Healy Model

Healy (1985) Model is the first model developed in the literature and estimates that the systematic earnings management will exist in every period. It is very simple yet it is criticized as being quite insufficient in estimating discretionary accruals by researchers like Young (1999). Healy compared to other studies differs in this area by pointing out that earnings management is systematic in every period.

The DeAngelo Model

The DeAngelo Model is considered a special version of the Healy (1985) in consequence of the facts that it does not require any estimation periods and the

(20)

estimation period of the nondiscretionary accruals is limited by the previous year’s observations Dechow et al. (1995). Just like Healy, DeAngelo also accepts the fact that mathematically, discretionary accruals cannot be calculated alone.

The Jones Model

Jones (1991) has brought a model to the literature in which the model itself confirms the assumption that nondiscretionary accruals are not constant. Unlike the Healy (1985) and DeAngelo (1986) models – that contain the assumption that the average change in nondiscretionary accruals is constant and the change in total accruals stems from discretionary accruals – Jones (1991) added the change in sales and the gross amount of fixed assets to the model in order to control the effects of the changes that may occur in the nondiscretionary accruals as a result of the firm’s economic position. The Jones Model for nondiscretionary accruals in the event year is:

NDAt = α1 (1/At-1) + α2 (∆REVt ) + α3 (PPEt )

where,

∆REVt = revenue in year t less revenue in year t-1 scaled by total assets at t-1

PPEt = gross property plant and equipment in year t scaled by total assets at t-1

At-1 = total assets at t-1 α1, α2, α3 = firm-specific parameters

Estimate of the firm-specific parameters, α1, α2, and α3 are generated using the following model in the estimation period;

TAt = a1 (1/At-1) + a2 (∆REVt ) + a3 (PPEt )+ ν1

where,

TAt = Total accruals

a1, a2, a3 denote the Ordinary Least Square (OLS) estimates of α1, α2, α3 and TA are total accrual scaled by lagged total assets. The results show that the Jones model successfully revealed one-quarter of the variation in total accruals.

(21)

The Modified Jones Model

The Modified Jones Model removes the conjectured tendency of the Jones Model. It returns discretionary accruals with an error when discretion is around revenue recognition. The nondiscretionary accruals are estimated during the event period (i.e., during periods in which earnings management is hypothesized) as:

NDAt = α1 (1/At-1) + α2 (∆REVt - ∆RECt) + α3 (PPEt )

where,

∆REC = net receivables in year t minus net receivables in year t-1 scaled by total assets at t-1.

The estimates of α1, α2, α3 and nondiscretionary accruals during the estimation period (in which no systematic earnings management is hypothesized) are coming from the Jones Model.

The change in revenues is adjusted for the change receivables in the event period as per the Jones Model. The original Model implicitly assumes that discretion is not exercised over revenue in either the estimation period or the event period. Whereas, the modified version of the Jones Model implicitly assumes that all changes in credit sales are resulting from earnings management. This is based on the fact that it’s easier to manage earnings by exercising discretion over the recognition of revenue on cash sales rather than on credit sales.

The Industry Model

Another model considered is the Industry Model that is proposed by Dechow and Sloan (1991). The Industry Model relaxes the assumption that nondiscretionary accruals are constant over time. Instead of attempting to model the determinants of nondiscretionary accruals directly, the Industry Model assumes that the variation in the determinants of nondiscretionary accruals is common across firms in the same industry.

(22)

3.2.2 Choosing the Research Model

The above-discussed models have been proposed to segregate total accruals into discretionary accruals and nondiscretionary accruals. Discretionary accrual is used as a proxy to determine the extent of earning management. They are obtained by subtracting the non-discretionary accruals from total accrual. Non-discretionary accruals are estimated by using the regression model that regress total accruals on several explanatory variables. A critical drawback to the total accrual approach is that we cannot distinguish the discretionary component from non-discretionary components.

The Jones Model (1991) and Modified Jones Model Dechow et al. (1995) are the ones that are most frequently used in related literature around the relation between earnings management and accounting standards.

These models assume that since nondiscretionary accruals are constant from one period to another, the difference between current and prior year accruals is because of changes in discretionary accruals.This study investigates the difference between discretionary accrual over the different phases over the period from 2001 - 2015 of implementation of IFRS in the EU.

Therefore, the model needs to separate discretionary accruals from total accrual. The model out of these five that I will be using in this research is The Modified Jones Model. According to Dechow et al. (1995), a modified version of the model developed by Jones (1991) exhibits the most power in detecting earnings management. Also, Chan et al. (2004), indicate that the Modified Jones Model is a better model for detecting earnings management than the Jones Model.

(23)

3.2.3 Modified Jones Model, Variables and Final Regression Equation

Here, I describe the main steps to calculate the total accrual and estimating Modifies Jones Model. And finally calculating the discretionary accrual which is a proxy for earning management.

The calculations will be done using SPSS statistic program. The significance of the individual variables and the explaining power of the model will be discussed on the basis of model outcomes of SPSS.

The R² explains the dependency of dependent variables on independent variables. Tests will be conducted to determine the reliability of the model and the reliability of the outcomes.

Step 1: Calculate the Total Accrual

The first step is to calculate the total accrual (TACC).

TACCt = ∆CAt - ∆Cash - ∆CLt + ∆DCLt - DEPt ……….……… (Eq.1)

where,

TACCt = Total accrual in year t,

∆CAt = Change in current assets in year t,

∆Cash = Change in cash and cash equivalent in year t, ∆CLt = Change in current liabilities in year t,

∆DCLt = Debt in current liabilities in year t,

DEPt = Depreciation and amortization expenses in year t.

t = year index range from 2001 until 2015.

All the variables in the equation are known, so the total accrual can be calculated. The year index includes a range from 2001 - 2015. This includes the period before (2001 –

(24)

2005) and after (2006 – 2010) IFRS implementation considering the first 5 years as chaos. The latter period (2011 – 2015) will show if there is any effect of earnings management.

Step 2: Estimate the Modified Jones Model

The TACC calculated above in (Eq. 1) is used as the dependent variable in the second step. 𝑻𝑨𝑪𝑪_𝒕 𝑨_𝒕−𝟏

= 𝜶

𝟏 𝟏 𝑨_𝒕−𝟏

+ 𝜶

𝟐 ∆𝑹𝑬𝑽_𝒕−∆𝑹𝑬𝑪_𝒕 𝑨_𝒕−𝟏

+ 𝜶

𝟑 𝑷𝑷𝑬_𝒕 𝑨_𝒕−𝟏

+ 𝜺

……… (Eq. 2) where,

TACCt = Total accrual in year t,

∆REVt = revenue in year t minus revenues in year t-1,

∆RECt = Delta revenues in year t less delta net receivables in year t-1,

PPEt = Gross property plant and equipment in year t,

At-1 = Total assets in year t-1,

α1, α2, α3 = Regression coefficients.

The variable A, ∆REVt, ∆RECt, PPEt used above in (Eq. 2) is directly coming from the financial statements of the companies used in the sample. The regression coefficients α1, α2, α3 can be calculated using historical data for all the variables. These regression coefficients measure the degree to which the dependent variable changes as the independent variable changes. For example, when the α is 0.1 it means that when the dependent variable increases with 10, the dependent variable increases with 0.1 times 10. After that, we can estimate this model to estimate the discretionary accruals in step 3.

Step 3: Calculate the Non-Discretionary Accruals

The next step is calculating the nondiscretionary accruals, in order to determine the discretionary accruals later. Since discretionary accruals do not arise out of

(25)

operational activities, they are an indication and therefore a measure of earnings management. The following equation is used for calculating the nondiscretionary accruals: 𝑵𝑫𝑨𝑪𝑪_𝒕 𝑨_𝒕−𝟏

= 𝜶

𝟏 𝟏 𝑨_𝒕−𝟏

+ 𝜶

𝟐 ∆𝑹𝑬𝑽_𝒕−∆𝑹𝑬𝑪_𝒕 𝑨_𝒕−𝟏

+ 𝜶

𝟑 𝑷𝑷𝑬_𝒕 𝑨_𝒕−𝟏

……… (Eq.3) where,

NDACCt = Non-discretionary accrual divided by total assets in year t-1,

∆REVt = revenue in year t minus revenues in year t-1,

∆RECt = Delta revenues in year t less delta net receivables in year t-1,

PPEt = Gross property plant and equipment in year t,

At-1 = Total assets in year t-1,

Since the only unknown variable in the above equation is NDACC, the non-discretionary accrual can be calculated.

Step 4: Calculate the Discretionary Accruals

When non-discretionary accruals are calculated by the above model, the next step is to calculate the discretionary accrual. Basically, discretionary accrual is a component of total accrual (TACC) which is calculated in step 1. Total accrual consists of the components of non-discretionary (NDACC) + discretionary accrual (DACC).

The discretionary accruals will be calculated with the next equation (4)

DACCt = TACCt - NDACCt ………. (Eq.4)

Step 5: Final Regression

Having discussed all the steps in calculating discretionary accruals, the model below tests whether earnings management is a function of IFRS or not. The fifth step contains a regression analysis where I will not only include the implementation of IFRS as a dummy variable to see if it has an effect on earning management but also include the

(26)

multiple control variables in order to check if it can have an effect on the change in earnings management over the years.

I use the absolute value of discretionary accruals (i.e., absolute discretionary accruals) as a proxy to measure earnings management since accruals can be managed downwards as well as upwards to smooth reporting.

Estimating the following regression model:

DACCt = β0 + β1 (IFRS)t + β2 (Size)t + β3 (ROA)t + β4 (LogEBIT) + ε………..(Eq.5)

where,

DACCt = Absolute value of discretionary accrual in year t, scaled by lagged total assets estimated by Modified Jones Model,

IFRSt = Dummy variable, the accounting standard used, {compliance to IFRS=1, else=0},

Sizet = the company size based on the natural logarithm of total assets for that year t,

ROAt = Return on assets,

LogEBIT = Natural logarithm of the earnings before interest and tax, ε = an error term.

t = Year index range from 2001 - 2015. β0, β1, β2, β3, and β4 = Regression coefficients

As per Heemskerk and Van der Tas (2006), accruals can be both positive and negative, therefore the absolute value of the discretionary accruals should be taken. Therefore, ‘DACC’ represents the discretionary accruals scaled by the lagged assets for the firm in year t. The absolute value of discretionary accruals is calculated in (Eq. 4).

(27)

With regression, it is possible to determine what factors have a significant explaining power. In this case, it is investigated whether the introduction of IFRS and the other controlled variables has any influence on the level of earnings management.

Variable ‘standard’ represents a dummy variable that is given a 1 when the firm is using IFRS (voluntary and mandatory) and a 0 otherwise. According to (Ahmed et al. 2012, Doukakis 2014, Houqe et al. 2012), the firm size can be estimated by the natural logarithm of total assets in year t, hence the variable ‘size’ is calculated using the total assets for each firm.

(Kasznik 1999, Kothari et al. 2005) included the return of assets (ROA) of the current year as a controlled variable for the extreme performance that I have included as well in my equation. Past studies proved that the profitability of the companies is closely related to the return on assets (ROA). Hence, reporting a good ROA might be an incentive for managers to show the future profitability of the company Demirkan and Platt (2009). Rahman and Ali (2006), used ROA as a measurement for the performance of the companies by operating income or earnings before interest and tax (EBIT) divided by the total assets.

With this final step the regression coefficients β0, β1, β2, β3, and β4 are calculated. While calculating the coefficient, both the direction and the significance are of importance. The direction of the coefficient can be either positive or negative suggesting that there is a positive or negative association between the dependent variable and the level of the discretionary accruals respectively. The significance of the coefficients means whether the variables standard and size are positive or negative in association with the level of discretionary accruals.

This is the final outcome that is needed to answer the hypothesis. Since the data sets include discretionary accruals, standard, and size, year, return on assets and EBIT (Earnings before Interest and Taxes) for every company for every year, the regression coefficients can be calculated.

(28)

4. Data and Descriptive Statistics

European Union (EU); a 28-member block with Austria, Belgium, Bulgaria, Croatia, Cyprus, Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Ireland, Italy, Latvia, Lithuania, Luxembourg, Malta, Netherlands, Poland, Portugal, Romania, Slovakia, Slovenia, Spain, Sweden, and the United Kingdom as its member states came into existence on November 1st_{1993 in Maastricht, The} Netherlands. It took nearly 150 years to come to realization when the idea was first germinated by Victor Hugo in International Peace Congress in 18496_{. After the initial} treaty, it took another nearly one decade to come to a single European currency system (in 2002) abolishing the national currencies. At the heart of the treaty was the idea to establish a single jurisdiction with economic and political partnership across the European states. A single market with standardized laws and ensuring free movement of people, goods, services, and capital.

Enacting on this founding preamble further, with respect to accounting, the EU established and brought into force some laws, known as Directives that all EU and EEA members must comply with. Some of those Directives address accounting issues. The most notable is Directive 2013/34/EU of the European Parliament and of the Council of 26 June 2013 on the annual financial statements, consolidated financial statements, and related reports7_{. Together with these Directives, each member state may enact} additional accounting laws and regulations that add to the requirements of the Directives, but they cannot override the requirements of the Directives. It must be noted that even though the Directive came into force in 2013, EU adopted IFRS standards and its mandatory adoption in 20058_{for all its member states.}

The initial adoption years created much of misinterpretation and havoc in transitioning from regional regulations to an international standard. From 2006 until

6_{Peace Congress Speech}

7_{European Commission - EU rules on financial information disclosed by companies} 8_{IFRS use around the world}

(29)

2010 a period I chose to call the early adoption, borrowing the term from Everett Rogers, a professor of communications from his theory on Diffusion of Innovations9 and recently popularized by the book ‘Crossing the Chasm’ by Geoffrey Moore10_.

My research is anchored on the following parameters and associated data sample:

Region: European Union (28-member block)

Rationale: This member block by the size of purchasing power parity is the 2nd_largest in the world with a GDP of nearly $21 trillion and representing close to 1/5th_{of the} global economy11_{. Euro is the 2}nd_{most traded currency}12_{and amongst the top 50} economies of the world, 15 of those are members of the EU13_{. It’s also a member of} the G20 group of economies that represent the world’s largest advanced and emerging economies, representing about two-thirds of the world’s population, 85 percent of global GDP and over 75 percent of global trade.

I have selected all the listed companies of the European Union with revenue of more than 1 million EUR.

Period: (2001 – 2015)

Rational: EU was one of the early adopters of the IFRS standards14_{and by now any} confusion it went through in the beginning has subsided with time and maturity. There are three main sub-division of the periods used in this study, that is:

1) Pre-IFRS: the period from 2001 - 2005;

2) Post-IFRS: the period from 2006 - 2010; includes the early years of IFRS adoption.

9_{Diffusion of Innovation - A Macroview} 10_{Crossing the Chasm}

11_{International Monetary Fund Report}

12_{Compositional Analysis of Foreign Currency Reserves} 13_{International Monetary Fund Report}

(30)

3) Post-IFRS: the period from 2011 - 2015; includes the mature years of IFRS adoption.

The idea to break the period into three bins is to notice and eliminate any discrepancies created during the early years of adoption.

Data Source:

There are two primary data sources used for this study: 1) Orbis

The Orbis data source is used to create a list of companies using the above-given criteria (region and period). Only industrial companies with revenue of more than €1 million were selected. These filters provided a list of 1,589 companies for years 2001 - 2015. Of all these companies’ private firms were excluded, resulting in a total of 518 listed firms.

2) DataStream/Worldscope

The financial data for the above-short-listed companies is extracted from DataStream/Worldscope data source.

Exclusions:

1) Observations where there were missing values for the selected variables (e.g. data missing in any period between 2001 and 2015).

2) Following (Daske et al. 2008, Houqe et al. 2012, Francis and Wang 2008), financial service firms have been excluded. Among these are banks, insurance firms, pension funds, private equity or other investment firms. Since they are regulated, they are likely to differ from other companies in their incentive to manage earnings.

3) Following Houqe et al. (2012) utility firms were also excluded, such as research institutes, foundations, public authorities and other governmental entities. 4) All firms using other accounting systems after 2005 were excluded.

(31)

After taking into all considerations of the exclusion factors, the original set of 518 companies reduced to a set of 350 companies.

Table 4.1 presents the research sample (numbers of companies and observations), consisting of 350 European companies (5,584 firm-year observations).

Table 4.1 Number of Firms by Country

This table presents for each country the total number of firms, number of observations, and percentage from 350 companies in this study. The selection is done using the Orbis database. Selection criteria are as follows:

1) EU countries having listed firms with more than ϵ1 million revenue, 2) All the firms have no missing data between 2000 – 2015,

3) Financial services and utility firms have been excluded,

4) Firms using local or another accounting system after 2005 have been excluded.

Country Number of Firms Number of observation Percentage of Firms/Observation Austria 8 128 2.29% Belgium 10 160 2.86% Czech Republic 2 32 0.57% Denmark 11 176 3.14% Finland 16 256 4.57% France 68 1,088 19.43% Germany 58 928 16.57% Greece 6 96 1.71% Hungary 2 32 0.57% Ireland 11 176 3.14% Italy 21 336 6.00% Luxembourg 5 64 1.43% Netherlands 18 288 5.14% Poland 3 48 0.86% Portugal 3 48 0.86% Slovakia 1 16 0.29% Spain 17 272 4.86% Sweden 22 352 6.29% United Kingdom 68 1,088 19.43% Total 3 5 0 5 ,5 8 4 1 0 0 .0 0 %

(32)

Table 4.2, presents the descriptive statistics (Means and Medians) of the sample set on the main variables used in the study: total assets, current assets, current liabilities, net sales or revenues, net receivable, cash or cash equivalent property, plant and equipment (PPE), depreciation and amortization, earnings before interest and tax (EBIT).

Table 4.2 Descriptive Statistics of Financial Data

This table presents the Mean and Median for all firms from the year 2001 – 2015. It includes all the observation for the main financial information used in the study. Financial statement data is obtained from the DataStream database for each year. The list of firms is obtained from the Orbis database and ISIN code is used to extract the financial information from the DataStream database. I used only those firms that have complete information for all the years from 2001 - 2015; thus excluding the firms with missing data. Only firms that have revenue of more than 1 million EUR have been selected.

Since the Means are higher than the Medians, this is an indication of some extreme values in the sample set. For the steps in the next section, the values of all the variables are scaled by lagged total assets to avoid complications with heteroscedasticity.

N Mean Median Std. Deviation

Total assets 5,584 32,760,419 6,452,700 187,877,325 Current assets 5,584 21,430,760 4,856,022 119,355,795 Current Liabilities 5,584 9,646,740 1,946,166 53,172,575 Revenue 5,584 54,275,660 11,230,225 388,247,747 Receivables 5,584 4,673,508 1,007,500 23,095,716 Cash 5,584 4,337,448 844,104 27,438,191 PPE 5,584 13,333,637 1,485,438 106,861,269

Depriciation and Amortization 5,584 1,853,619 235,700 14,324,959

(33)

5. Empirical Results

Using the Modified Jones Model and its associated steps as described in 3.2.3 to analyze the collected raw data, several interesting results were found. This section discusses the outcome of the analytical procedures to either accept or reject the proposed hypothesis.

Before analyzing the data, in order to identify and eliminate outliers an interquartile range is performed. This is to ensure that the values that are outside of the distribution which impacts the calculation and thus the value of the Mean and standard deviation are discarded. To deal with outliers, the method of winsorizing is used. Winsorization helps in transformation of statistics by limiting extreme values in the data to reduce the effect of invalid outliers.

SPSS program has been used to further transform the resulting data into a standard format. As discussed earlier in section 3.2.3, the following steps were performed:

Step 1: Calculate the Total Accrual

Using (Eq. 1), both for the period (2001 - 2005) and (2006 - 2015) the total accruals (TACC) are calculated. In order to make the results comparable, all outcomes are scaled by the total assets of the corresponding company.

Step 2: Estimate the Modified Jones Model

Using the total accrual (TACC), the regression coefficient α1, α2, α3 in the (Eq. 2) are calculated using the least square regression. Before a regression analysis can further be performed, the following assumptions must be satisfied:

1. All variables should be numeric. Since all the variables in (Eq. 2) are numeric, this assumption is met.

2. Independent variables do not measure the same so there is no multicollinearity. Since all the variables include specific parts of the financial

(34)

numbers of a company and no variable includes the same financials as another variable, this assumption is also met.

Therefore, two important assumptions for a multiple regression analysis are met for (Eq. 2). Thus, the regression analysis can be performed and the results are presented in Table 5.1.

Table 5.1 Coefficients estimated by the Modified Jones Model (2001 – 2015)

This table represents the coefficients and the regression summary of the Modified Jones Model. This provides the p-values of the coefficients, r-square, adjusted r-square and the result of Durbin-Watson (DW test) of autocorrelation. The dependent variable is total accruals (TA).

Coefficient α1 = 1/At-1; coefficient α2 = (∆REVt - ∆RECt) / At-1 , variation in the net revenue of company i

from time t-1 to time t, weighted by the total assets at the end of time t-1, minus the variation in accounts receivable of company i from time t-1 to time t, weighted by the total assets at the end of time t-1; coefficient α3 =PPEt/ At-1 , property plant and equipment of company i at the end of time t,

weighted by total assets at end of time t-1.

These coefficients are explained by the column ‘unstandardized coefficients B’, and their significance value explained by the column ‘p-value’; VIF = variance inflation factor.

*, **, ***, significant at the 10%, 5%, and 1% levels, respectively. The regression equation being tested is as follows:

The above Table 5.1 is the summary of the regression on the Modified Jones Model. In this summary, the multiple R² is important; it explains the relationship between the independent variables and the dependent variable. R² is 0.020 which means that approximately 2 percent of the total accruals, the dependent variable, is explained by

Standardized Coefficients

B Std. Error Beta Tolerance VIF

Constant -.050 .004 -11.342 .000 α1 27446.81 3368.08 .121 8.149 0.000** .855 1.169 α2 .005 .002 .043 2.905 0.040** .847 1.181 α3 .016 .008 .026 1.908 0.0564* .985 1.016 N 5242 R2 .020 R2 adjusted .020

Standard error of estiamate .130

F(36.191) .000 Durbin-Watson Statistic 1.891 Unstandardized Coefficients t p-value Collinearity Statistics 𝑇𝐴𝐶𝐶𝑡 𝐴𝑡 − 1⁄ _{= 𝛼1 (1 𝐴𝑡 − 1}⁄ _{) + 𝛼2 (∆𝑅𝐸𝑉𝑡 − ∆ 𝑅𝐸𝐶𝑡 𝐴𝑡 − 1}⁄ _{) + 𝛼3 (𝑃𝑃𝐸𝑡 𝐴𝑡 − 1}⁄ ) +ε

(35)

the combined effect of the three independent variables Assets, Revenue-Receivables and Property Plant Equipment.

To detect the presence or absence of autocorrelation, Durbin-Watson (DW test) was also included. The results of the DW test demonstrate the value of 1.891 (where DW rate between -2 to +2), so we can conclude that there is no autocorrelation.

The F-stat (36.191) which is the result of ANOVA in regression tells whether the independent variables are able to predict together the dependent variable. The F-stat displays whether this is the case or not.

The F-value is the part of the explained variance divided by the part of the unexplained variance of the model. The number of df of the regression is the same as the number of the independent variables, which is 3. The number of the degrees of freedom of the residual is the total number of cases minus the df of the regression minus 1. The total number of cases is 350 companies, for 15 historic years which is 5243. So, the number of df of the residual is 5243 – 3 - 1 = 5239.

Thus, this model as a whole has a significant explaining power, since the significance is 0.000. This significance is called the P-value and it means that if the hypothesis is true, the chance is 0.000 that the results are a coincidence. In other words, the smaller the P-value, the stronger the evidence is that there is an association between the independent and the dependent variable which is not coincidental. The P-value has to be compared with a specified level of significance in order to determine whether the data is statistically significant. For a result to be significant, the P-value of the test has to be smaller than the significance level. In the existing literature, the significance level that is mostly used is 5 percent. This means that the data gives evidence against the hypothesis so strong that it would happen no more than 5% of the time that the hypothesis is rejected when the hypothesis is true. Since there is no reason to assume that a 5% significance level is not applicable here, this level will be used for

(36)

determining the significance of the regression results. In this case, the P-value of 0.000 is smaller than 0.05 so the model is statistically significant.

It can be concluded that the model as a whole is said to be significant, the next step is to see which individual variables have a significant explaining power. The three independent variables of this model are the total assets, the revenue minus the receivables and the gross property, plant, and equipment scaled by the total assets.

The column ‘unstandardized coefficients B’ in the table represent the regression coefficient α1, α2, and α3 for all three variables, thus the dependent variable changes as a result of a change in the independent variable. As all the coefficient are positives it indicates all the three independent variables have the positive impact on the dependent variable. However, while the overall model was said to be significant, not all individual variables are. As shown in the table, PPEt/A(t-1) it is not significant since the value is 0.056, which is slightly higher than the 0.05, though the other two independent variables are significant.

Step 3: Calculate the Non-Discretionary Accruals

We can now use these regression coefficients and calculate the value of non-discretionary accruals using (Eq. 3).

Step 4: Calculate the Discretionary Accruals

After calculating the non-discretionary accrual (NDACC) we can use the (Eq. 4) to calculate finally the discretionary accruals (DACC).

(37)

From 3.2.3, for the final regression.

Step 5: Final Regression

DACCt=β0 + β1 (IFRS)t + β2 (Size)t + β3 (ROA)t + β4 (LogEBIT) + ε………..(Eq.5)

where,

DACCt = Absolute value of discretionary accrual in year t, scaled by lagged total assets estimated by Modified Jones Model,

IFRSt = Dummy variable, the accounting standard used, {compliance to IFRS=1, else=0},

Sizet = company size based on the natural logarithm of total assets for year t,

ROAt = Return on assets,

LogEBIT = Natural logarithm of the earnings before interest and tax, ε = an error term.

t = Year index range from 2001 - 2015. β0, β1, β2, β3, and β4 = Regression coefficients

For this regression equation, the same assumptions as explained before also need to be satisfied.

1. Since the variable IFRS is a categorical variable hence it has to be restated to 1 and 0 in order to fulfill assumption 1.

2. Since the size of a company does not change as a result of a new accounting standard, it is obvious that these two variables do not measure the same thing. Thus, the second assumption, that there is no multicollinearity, is also satisfied.

Therefore, two important assumptions for a multiple regression analysis are met for (Eq. 5). Thus, the regression analysis can be performed and the results are presented in Table 5.2.

(38)

Table 5.2 Determinants of Discretionary Accrual

This table represents the coefficients and the regression summary of the equation (5) used in this study to find the effect of IFRS and other control variable to discretionary accrual. It provides the p-values of the coefficients, r-square, adjusted r-square and the result of Durbin-Watson (DW test) for autocorrelation. The dependent variable is the absolute value of discretionary accruals (absDACC). The effect of independent variables IFRS, Size, ROA and LogEBIT are described by the values coefficient β1, β2, β3, and β4 respectively by the column ‘unstandardized coefficients B’ and their significance is explained by the column ‘p-value’; VIF = variance inflation factor. *, **, ***: significant at the 10%, 5%, and 1% levels, respectively.

The regression equation being tested is as follows:

DACCt = β0 + β1 (IFRS)t + β2 (Size)t + β3 (ROA)t + β4 (LogEBIT) + ε

The R² for the above-presented regression model is only 0.021 which means that approximately 2 percent of the discretionary accruals is explained by the independent variables accounting standard under which the financial report is made, company size, ROA, and EBIT. Hence, the explaining power of this model is almost zero. There were no significant outliers in the data that can skew the results. Therefore, this cannot explain the low value of R².

The next step is to determine the significance of the whole model. Here, the F-value (26.672) is the number of the explained variance divided by the number of the unexplained variance. The significance of the model is 0.000. This further illustrates that the model as has significant explaining power since it is smaller than 0.05.

The last step is to calculate the significance of the individual independent variables that are also presented in Table 5.2. Here, the constant variable has a value of 0.130,

Standardized

Coefficients Collinearity Statistics

B Std. Error Beta Tolerance VIF

(Constant) .130 .019 6.893 .000 IFRS -.024 .004 -.098 -6.780 .000 .969 1.032 Size .011 .003 .152 3.469 .001 .105 9.522 ROA .003 .000 .133 6.599 .000 .492 2.033 LogEBIT -.034 .007 -.218 -4.863 .000 .100 10.006 N 4875 R2 .021 R2 adjusted .021

Standard error of estiamate .10867

F(26.672) .000

Durbin-Watson Statistic 2.0103

Unstandardized Coefficients

(39)

regression coefficient of IFRS -0.024, company size is 0.011, ROA is 0.003 and LogEBIT is -0.34.

The independent variable IFRS is a qualitative variable because before 2005 companies were using the different accounting standards and after 2005 only IFRS. Since this study is the effect of IFRS so we recode this variable into two categories; code IFRS as 1 for years 2005 onwards and before 2005 were given a code of 0.

The regression coefficient for the accounting standard, when the dependent variable is discretionary accruals, is -0.24. Since it is a negative coefficient, IFRS is associated with a lower level of discretionary accruals. Besides the coefficient is very low, it is significant since the corresponding P-value is 0.000, which is smaller than 0.05.

The other independent variables are numeric so they don’t have to be restated. All the variables used are significant as the p-value for all the independent variable is less than 0.05.

Therefore, the hypothesis "Whether the IFRS adoption has affected the earning management over the years." is therefore not rejected. It can be concluded that the IFRS has a significant effect on earnings management based on described independent variables.

(40)

6. Robustness Checks

6.1 Comparing Means using ANOVA

To test the results first I divided the years into two main groups’ pre-IFRS and post-IFRS. The pre-IFRS period is from 2001 - 2005 and post- IFRS period is from 2006 – 2015. The post-IFRS is further divided into two subgroups, the transition period (2006 – 2010) as well as the mature years after the adoption (2011 – 2015).

The following Table 6.1 presents the comparison of the three groups using the ANOVA test. Group (1) is pre-IFRS i.e. 2001 - 2005, group (2) is the early year of adoption of IFRS i.e. 2006 - 2010 and group (3) is the mature years of adoption i.e. 2011 - 2015.

Table 6.1 Comparison between the Means of Discretionary Accrual in sub-period

This table represents the comparison of Means of absolute discretionary accrual (absDACC) between the sub-period of group 1 (2001 - 2005), group 2 (2006 - 2010), and group 3 (2011 - 2015). I used the ANOVA test to compare the Means of these three groups, and ANOVA post-hoc test of Games-Howell for multiple comparisons between the groups. The upper part of the table represents the results of the Games-Howell test, and the lower part represents the results of the ANOVA test. The column ‘Mean Difference’ shows the difference of Means in corresponding comparison groups.

*, **, ***: significant at the 10%, 5%, and 1% levels, respectively.

The lower part of Table 6.1 represents the results of the ANOVA test, where the p-value of the F-Stat (39.802) is less than 0.05. This means the ‘Means’ between the groups are not equal. The Mean (0.1313) of discretionary accrual for the group (1), is higher than the Means of the other two groups (2) and (3). Also, we notice that the

2001-2005 gp(1) 2011-2015 gp(3) .03355** .00388 .000 2006-2010 gp(2) 2001-2005 gp(1) -.02057** .00411 .000 2011-2015 gp(3) 2006-2010 gp(2) -.01297** .00334 .000 2001-2005 (Mean) .1313 2006-2010 (Mean) .1107 2011-2015 (Mean) .0978 F (39.802) .0000 N 5243

Year groups Mean

Effect of adoption of IFRS on earnings management

Master in International Finance

2016 - 2018

Master Thesis

Effect of Adoption of IFRS on Earnings Management

Abstract

Table of Contents

Table of Tables and Figures

1. Introduction

1.1 Introduction to the subject

1.2 Research question and contribution to prior research

1.3 Structure of the thesis

2. Literature Review

3. Methodology and Hypothesis

3.1 Hypothesis Development

3.2 Research Design

3.2.1 Available Research Models

3.2.2 Choosing the Research Model

3.2.3 Modified Jones Model, Variables and Final Regression Equation

= 𝜶

+ 𝜶

+ 𝜶

+ 𝜺

= 𝜶

+ 𝜶

+ 𝜶

4. Data and Descriptive Statistics

5. Empirical Results

6. Robustness Checks

6.1 Comparing Means using ANOVA