The influence of reporting frequency on accruals-based earnings management

(1)

The influence of reporting frequency on accruals-based earnings

management

Name: Rogier Muis

Student number: 11837241

Thesis supervisor: dr. D. Veenman Date: June 25th, 2018

Word count: 22,685

MSc Accountancy & Control, specialization Control

(2)

Statement of Originality

This document is written by Rogier Muis, 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)

1

Abstract

This thesis investigates the relation between a firm’s frequency of issuing financial reports and accruals-based earnings management (AEM). Recent literature shows multiple outcomes from higher reporting frequencies but has, to my knowledge, not researched the relation between a firm’s reporting frequency and AEM. Using a discretionary accruals model and an accruals quality model for estimation, I find no significant evidence for the hypothesis that quarterly reporters are more positively related to AEM than semi-annual reporters. Specifically, I show no evidence for AEM and myopia related costs for higher reporting frequencies. This thesis thereby contributes to the emerging literature on the costs and benefits of higher reporting frequencies and provides inputs for the societal debate on quarterly reporting.

(4)

2

2.2.1 Theoretical foundations ... 13 2.2.2 Empirical findings ... 14 2.3 Reporting frequency ... 17 2.3.1 Theoretical foundations ... 17 2.3.2 Empirical findings ... 18 2.4 Institutional background ... 21 3 Hypothesis development ... 23 4 Research design ... 25 4.1 Sample selection ... 25 4.2 Method ... 26 4.2.1 Accruals estimation ... 26 4.2.2 Regression variables ... 30 5 Results ... 34 6 Additional analyses ... 35

6.1 Voluntary and mandatory reporting ... 35

6.2 Country-level effects ... 37

6.3 Accruals quality... 40

7 Discussion ... 44

8 Conclusion ... 49

Appendix A: Variable definitions ... 52

(5)

3

1 Introduction

This thesis examines the influence of a firm’s financial reporting frequency on accruals-based earnings management (AEM). More precisely, through an analysis of companies’ financial statements and the number of earnings statements that these companies publish per year, I attempt to answer the following research question:

RQ: Does the reporting frequency influence accruals-based earnings management?

This research is important, because the literature has until now, to my knowledge, been unable to answer how reporting frequency affects accruals-based earnings management. The stream of literature on reporting frequency and on the frequency of other disclosures mainly documents the advantages of higher disclosure frequencies, while research on the disadvantages of reporting frequency is relatively scarce (Verdi 2012). This paper answers to calls of Verdi (2012), which are partly answered by Ernstberger, Link, Stich and Vögler (2017) and Kraft, Vashishtha and Venkatachalam (2018), to conduct more research on the costs related to increased reporting frequencies.

In the last decade of research, the literature has shifted its focus from AEM to real activities manipulation (RAM) (e.g.: Ernstberger et al. 2017; Roychowdhury 2006; Zang 2012). The article of Graham, Harvey and Rajgopal (2005) was a watershed moment in research on earnings management (Walker 2013). Where most research before focused on accruals-based earnings management because this was thought to be the dominant form of earnings management, Graham et al. (2005) surprisingly found that the majority of financial executives would take actions in the form of RAM rather than in the form of AEM to meet and beat earnings (MBE) targets. This finding incubated research into RAM (Walker 2013).

Although there is evidence that RAM is the more common form of earnings management, I deem it important to contribute with evidence of the relationship between a firm’s reporting frequency and AEM. Graham et al. (2005) namely note that survey answers on AEM might be more biased than answers on RAM, because RAM is perceived as more ethical and socially desirable than AEM (Bruns & Merchant 1990; Fischer & Rosenzweig 1995).

Although Graham et al. (2005) show that RAM is much more widespread than thought before, it is due to possible bias in AEM-related answers not precisely clear how prevalent AEM is. Multiple archival-based researchers from a variety of perspectives have found recent evidence of AEM among firms (e.g.: Ali & Zhang 2015; Gore, Pope & Singh 2007; Kalyta 2009). Furthermore, there is evidence that AEM carries costs. AEM may lead to market discounts

(6)

4

(Francis, LaFond, Olsson & Schipper 2005; Ogneva 2012), AEM may lead to lower future financial performance (Bhojraj, Hribar, Picconi & McInnis 2009) and AEM may lead to a decrease in market efficiency (Elliott & Hanna 1996). The subsequent detection of AEM may also carry costs (Desai, Hogan & Wilkins 2006; Lev, Ryan & Wu 2008; Palmrose, Richardson & Scholz 2004). For example, in the case of Enron, the detection of earnings management led to its bankruptcy. Since there are costs associated with AEM and AEM continues to be an adopted tool for earnings management, I deem research into AEM to be of relevance.

There is an ongoing societal discussion on the effects of quarterly reporting. In the EU, quarterly reporting was made mandatory in 2004, but after criticism, the requirement was reversed (EU 2004; EU 2013; Kay 2012). More recently, the Singapore Exchange announced it is seeking feedback on whether to maintain the current regime of mandatory quarterly reporting (SGX 2018). Last year, the Canadian Securities Administrators requested feedback on a proposal that included abolishing mandatory quarterly reporting (CSA 2017). Others, such as Mr. Pozen of MIT Sloan School of Management, advocate in defense of quarterly reporting, stating that otherwise investors use less informative data to base their decisions on (Pozen 2016). My analysis of the relationship between quarterly reporting and AEM aims to act as an input to this discussion.

The literature on managerial myopia shows that managers might feel psychological pressures to reach targets (Degeorge et al. 1999) and that managers might change their long-term investments to cater to their investor clientele (Polk & Spazienza 2009). The research on earnings management furthermore shows that managers engage in earnings management to reach investor targets and compensation targets (e.g.: Ali & Zhang 2015; Bartov et al. 2002; Bergstresser & Phillippon 2006; Bhorjaj et al. 2009; Daske et al. 2006; Gore et al. 2007; Healy 1985). I argue that managers that report quarterly face more compensation and investor targets than managers that report semi-annually. Since AEM is related to such targets, I subsequently argue that quarterly reporting might lead to more instances where managers might be persuaded to engage in earnings management. I hypothesize that quarterly reporting is more positively related to accruals-based earnings management than semi-annual reporting.

I perform my research in the 28 member states of the European Union (EU-28). Because of differences in member states’ legislation on the periodicity of reports, the EU-28 provides a setting in which variation in terms of reporting frequency is present. I develop an accruals estimation model following the modified Jones model of Dechow, Sloan and Sweeney (1995). Using a sample of 22,981 firm-year observations, I subsequently regress the reporting

(7)

5

frequency on the absolute measure of discretionary accruals. A higher value for absolute discretionary accruals indicates a higher level of accruals-based earnings management. The results of my analysis show that within the EU-28, there is no significant positive relationship between the reporting frequency and AEM. Rather, the results show that a higher reporting frequency has no relationship or a negative relationship with AEM. I thus find no evidence for the hypothesis that managers conduct more accruals-based earnings management when they report quarterly instead of semi-annually.

By demonstrating that quarterly reporting may not be related to an increase in AEM, I document no evidence for one potential cost of an increased reporting frequency. Moreover, since AEM is a technique related to managerial short-termism (Bhojraj et al. 2009), I indirectly find no evidence for the contention that quarterly reporting is related to an increase in short-termism. These findings might provide important support for advocates of quarterly reporting. In additional analyses, I control for effects that might distort the analysis. Following Butler, Kraft and Weiss (2007) and Fu, Kraft and Zhang (2012), I investigate for differences in the relationship for mandatory and voluntary quarterly reporters and document no different results. I also investigate the relationship between the reporting frequency and AEM on a per-country basis. I find significant variation between countries in the sign and magnitude of the coefficient of RFit on |DACCit|. The analysis provides indications that there might be other country-level

factors involved than which I control for.

I find that the accruals estimation model that I employ is inaccurate in some respects. Therefore, I investigate accruals with a more comprehensive model that captures accruals quality (Dechow & Dichev 2002; Francis, LaFond, Olsson & Schipper 2005). The results that I find also provide no evidence that supports the acceptance of my hypothesis.

I contribute to the emerging literature on the costs and benefits of higher reporting frequencies by showing no evidence for one potential cost of higher reporting frequencies (e.g.: Butler et al. 2007; Ernstberger et al. 2017; Fu et al. 2012; Kraft et al. 2018). Moreover, I indirectly find no evidence that quarterly reporting managers are more myopic than semi-annual reporting managers. I thereby contribute to the literature on managerial myopia (e.g.: Bhojraj et al. 2009; Dechow and Sloan 1991; Polk & Spazienza 2009).

An important caveat to this study is endogeneity. The analysis may be distorted by endogeneity effects, for which I find no direct solution that is econometrically correct (Lennox et al. 2012). It is important to highlight that the significant coefficients that I find for certain countries may

(8)

6

be caused by underlying variables. This means that any found relationship with AEM might not relate to the reporting frequency, but rather be explained by underlying factors that determine the reporting frequency. I discuss these issues and I emphasize the need for more research on suitable control variables for two-stage analyses in disclosure studies.

In section 2, I document the related literature. Drawing on agency theory, I discuss how managerial myopia, accruals-based earnings management and reporting frequencies are related. I also set out the European regulatory environment. In section 3, I develop a hypothesis following the findings of section 2. Section 4 outlines my sample selection and methodology. In section 5, I describe the results, which I test for robustness in section 6. In sections 7 I discuss the findings and in section 8 I provide conclusions in relation to the literature and suggestions for future research.

2 Related literature

2.1 Managerial myopia

In this thesis, I define managerial myopia as behavior that has short-term benefits for the manager but long-term firm value-decreasing effects.

2.1.1 Theoretical foundations

Agency theory suggests that firms are nexus of contracts and that managers, referred to as agents, might not always act in the best interests of shareholders, referred to as principals, because the interests of the agents and principals do not align (Jensen & Meckling 1976). For an understanding of agency theory, it is important to note two assumptions. Firstly, agents are expected to be effort-averse, meaning that agents do not like working for internal motivational reasons (Scott 2015). Secondly, agents are expected to maximize their own utility, meaning that the agents will attempt to maximize their own benefits regardless of interests of other parties (Scott 2015). Within this relationship between agents and principals, there is information asymmetry, meaning that principals might not have the same information as agents and that principals cannot observe the actions of agents (Jensen & Meckling 1976).

Because of this information asymmetry, investors do not precisely know what actions managers undertake. Because managers want to maximize their own utility, they might engage in actions which do not align with investor interests, which is known as moral hazard (Jensen & Meckling 1976). Scott (2015) models moral hazard in his book. Consider a one-period theoretical setting

(9)

7

where principal and agent have a basic contract that specifies that the agent will get a fixed amount of salary, irrespective of firm performance (Scott 2015). The agent will in this case not work hard, because the agent dislikes effort and will receive the salary no matter how hard he works (Scott 2015). Because he does not work hard, there is a smaller chance that the firm will do well and the principal, the investor, will therefore receive fewer benefits (Scott 2015). This divergence of interests between investors and managers relates to adverse selection. Adverse selection is the other type of information asymmetry (Jensen & Meckling 1976). Adverse selection means that investors expect managers not to act in their interests and because they have such expectations, they decrease their valuations (Jensen & Meckling 1976). The actions of managers thus come at a potential cost.

Agency theory highlights that contract specifications are the key to reducing information asymmetry (Hölmstrom 1979). Contracts can for example include clauses that give managers a share of the net income of the period, thereby better aligning the interests of investors and effort-averse managers (Scott 2015). If investors through contracting know that there are no moral hazards, the adverse selection problem will also be reduced.

Hölmstrom (1979) models that when principals design contracts, they should trade off the contractual agency costs and benefits. By making this careful trade-off, an optimal point can be reached (Hölmstrom 1979). For example, the costs of monitoring an agent have to be traded off against the benefits of preventing an agent from conducting value-decreasing actions. Agency theory relates to managerial myopia in multiple forms. For example, consider a setting where the manager receives a bonus if the period’s firm performance is high. The manager might work hard to attain this bonus, or under the effort-averse assumption the more likely option, the manager might manipulate the performance number so the he or she has to exert less effort (Scott 2015). Due to information asymmetry, the investor cannot observe the manipulation. The manager can manipulate this performance number by ‘borrowing’ at a reduced rate from long-term earnings, thereby decreasing total firm value. Such behavior is myopic, because the manager acts in the short-term interest at a long-term cost.

Moral hazards can also stem from investor behavior. For example, Stein (1988) models how raiders, which are aggressive investors looking to take over firms, may pressure managers to signal the firm being in a good state. Under the pressure of being taken over, managers will attempt to defend themselves by increasing firm value (Stein 1988). Subsequently, the manager will forego long-term profits for short-term profits to signal the firm being a good state and to

(10)

8

influence the share price (Stein 1988). Such signaling is myopic, as it means foregoing profits in later years and earning these profits at a reduced rate in the current period (Stein 1988). Following the same reasoning, Bebchuk and Stole (1993) model how in settings where investments decisions are not observable for investors, managers will signal the firm being currently in the good state by foregoing long-term investments.

Degeorge, Patel and Zeckhauser (1999) investigate managerial responses to internal and external targets. The authors provide backgrounds on the psychological causes for managerial myopia. Specifically, they research the reasons why managers might feel pressured to reach a target. Degeorge et al. (1999) list three psychological reasons.

First, in human thought processes, there is a clear demarcation between positive and negative numbers (Degeorge et al. 1999). The norm is to show positive numbers and meeting this norm is critical. Managers thus feel psychological pressures to conform to a norm by showing positive numbers (Degeorge et al. 1999). Secondly, the authors mention biases stemming from prospect theory as a reason for such behavior (Degeorge et al. 1999). Barberis (2013) fairly recently reviewed the prospect theory literature. Prospect theory states that people choose among risky options with the guidance of a reference point and with a bias for losses known as loss aversion (Barberis 2013). The reference point means that people do not evaluate their outcomes, for example the bonus awarded by reaching a quarterly target, absolutely but rather to how it will affect their current wealth level (Barberis 2013). Loss aversion means that people dislike losses more than equal gains (Barberis 2013). For example, a manager is much more sensitive to a 100$ loss than to a 100$ gain. Managerial thinking that follows the prospect theory leads managers to feel pressures to avoid losses and it might lead them to make myopic choices. Third, Degeorge et al. (1999) mention threshold effects. Threshold effects are that the prominence of certain targets in the managerial environment might make them feel more pressured to reach the targets (Degeorge et al. 1999).

2.1.2 Empirical findings

Degeorge et al. (1999) then analyze whether firms engage in earnings management to avoid losses or decreases relative to targets. Earnings management is by the authors interpreted as the strategic exercise of managerial discretion to inflate or deflate earnings and thereby includes real as well as accruals-based forms of earnings management (Degeorge et al. 1999). The authors use a methodology known as the suspicious threshold methodology. This methodology involves investigating firms that are suspiciously close above the earnings benchmark, with

(11)

9

suspicious referring to that these firms might have engaged in earnings management (Degeorge et al. 1999). Degeorge et al. (1999) find that firms indeed use earnings management to come above earnings thresholds. Moreover, they show that for firms which nearly miss the same-earnings-as-last-quarter benchmark, earnings are managed upward with decreased earnings performance in later years (Degeorge et al. 1999). Since the short-term earnings increasing behavior has long-term value-destroying effects, the managerial behavior can be categorized as myopic.

More recent evidence of managerial myopia is provided by Ali and Zhang (2015). They research how CEO tenure relates to earnings management. Specifically, they document how earnings management differs between the early and final years of a CEO’s tenure (Ali & Zhang 2015). Ali and Zhang (2015) hypothesize that CEOs manage earnings more during the early years because at that point in time, CEOs need to build up credibility with the markets to avoid dismissal and being perceived as having low ability. They perform regressions and indeed find evidence corresponding to the hypothesis (Ali & Zhang 2015). Finally, the authors find evidence of earnings decreases in later years following earnings management, which supports the contention as set out by Degeorge et al. (1999) that firms ‘borrow’ earnings from later years and which supports the contention that the manager behaves myopically to build up credibility with the market (Ali & Zhang 2015).

Oppositely, there is literature that documents managerial myopia in the final years of a CEO’s tenure. This is a phenomenon known as the horizon problem. One of the notable papers on the horizon problem is written by Dechow and Sloan (1991). The horizon problem means that CEOs manage earnings upwards in their final years at the cost of later earnings. Since the CEO resigns in that year or the years directly after, he does not experience the negative implications of the earnings management (Dechow & Sloan 1991). CEOs are especially hypothesized to behave as such when they retire afterwards, as otherwise the CEO might experience reputational damage (Kalyta 2009).

However, there is mixed support for the horizon problem theory. Dechow and Sloan (1991) find evidence for the horizon problem but also find that this problem can be mitigated by CEO stock ownership. A potential explanation for this mitigation effect is that CEOs with stock ownership will not vest all the shares in the year of resignation and hence would otherwise be affected by their own myopic actions. Wells (2002) finds little evidence of the horizon problem in an Australian setting. Cheng (2004) shows that compensation committees effectively mitigate the horizon problem. Kalyta (2009), on the other hand, documents that the horizon

(12)

10

problem effect is present in cases where retiring CEOs’ pensions are based on firm performance. The findings of Dechow and Sloan (1991), Cheng (2004) and Kalyta (2009) integrate with this explanation, namely that there is in fact a horizon problem, but mostly in cases of CEOs’ retirement depending on their final year earnings and not in firms where CEOs hold stock or where there is a compensation committee. The horizon problem literature thus documents cases of managerial myopia, but the myopic effect seems to be easily mitigatable. Another stream of literature has extended managerial myopia research by investigating potential causes for managerial myopia. This literature specifically focuses on investor-related causes. Abarbanell and Bernard (2000) note that investors themselves might be myopic, as investors might underprice firms that have long investment payback times and might overprice firms that have short investment payback times. Using a sample of American firms from 1978-93, Abarbanell and Bernard (2000) find no evidence of investors behaving myopically. Specifically, the authors do not find that a trading strategy focused on long-term investing firms provides the positive abnormal returns that it should, given the claim that the market systematically underprices such firms. Since they do not document myopic investor behavior, Abarbanell and Bernard (2000) doubt the explanation that myopic managerial behavior stems from myopic investor behavior.

Bushee (2001) studies the association between institutional ownership and myopic pricing behavior by investors. The sample contains two types of institutional firms, namely those operating in regulated industries, such as bank trusts, and ‘transient’ institutional firms with a short investment horizon (Bushee 2001). Bank trusts are hypothesized to behave myopically because they are highly pressured by shareholders. Bushee (2001) adopts the same database and methodology as Abarbanell and Bernard (2000) and finds that both types of institutional firms prefer firms that have short investment payback times. Additionally, Bushee (2001) researches the pricing of stocks with high percentages of the two types of institutional ownership. Bushee (2001) finds that the transient institutional investors systematically underprice the long-term earnings and systematically overprice the short-term earnings of firms.

Bartov, Givoly and Hayn (2002) research investor behavior in relation to earnings target beating. They find that firms that beat investors’ earnings targets have higher abnormal returns than firms that do not meet the earnings targets (Bartov et al. 2002). Specifically, this means that not only absolute performance is important for stock returns, as documented long ago by Ball and Brown (1968), but that firm performance relative to investor targets is also important.

(13)

11

They also find that these higher abnormal returns exist for firms that use AEM to come above the target. However, Bartov et al. (2002) do not find that the positive abnormal returns are the result of myopic investor behavior. They namely document that a positive earnings target deviation (i.e., earnings surprise) positively relates to future firm performance (Bartov et al. 2002). This means that the investors act rationally rather than myopically to the earnings surprise, even if there is AEM involved.

Indirect evidence for managerial myopia and the role of investors is provided by Graham et al. (2005). They survey a large sample of CFOs on the importance of meeting investor targets. Graham et al. (2005) find that 73.5 per cent of the CFOs agree or strongly agree with the statement that the analyst consensus forecast is important for the quarterly earnings report (Graham et al. 2005: 22). Furthermore, they find that almost 42 per cent of the CFOs would not pursue a positive-NPV project if that means missing the analyst consensus earnings by 10 cents (Graham et al. 2005: 39). Taken together, these findings imply that managers find quarterly targets and forecasts important and that they will behave myopically to ensure meeting these targets and forecasts. The findings also show the prominent role of investors in managerial decision-making.

Polk and Spazienza (2009) investigate managerial myopia and investor behavior from a catering theory perspective. Specifically, they analyze how a firm’s long-term investments are related to a firm’s stock price (Polk & Spazienza 2009). Catering theory is the theory that in cases when a stock is already overpriced or underpriced, the manager will either cater to the price expectations by overinvesting or underinvesting (Polk & Spazienza 2009). In the case of overpricing (underpricing), the manager will overinvest (underinvest) because the market will overvalue (undervalue) the investments (Polk & Spazienza 2009). The authors find corresponding results and additionally show that the over- or underinvestment leads to decreased stock returns in later years (Polk & Spazienza 2009). Since the short-term pricing gains lead to long-term losses, the behavior is myopic. Both the investors and managers engage in this myopic behavior. The managers behave myopically by overinvesting (underinvesting) and the investors behave myopically by overpricing (underpricing) the overinvestment (underinvestment) (Polk & Spazienza 2009). The research of Polk and Spazienza (2009) indicates that myopic investor behavior might influence managers to invest myopically. In additional analyses, Polk and Spazienza (2009) show that transient investors positively moderate the effect between overpricing and overinvestment.

(14)

12

Cadman and Sunder (2014) provide the most direct link between investor behavior and managerial myopia. They research how influential short-term investors influence the horizon of managerial incentives. Cadman and Sunder (2014) hypothesize that short-term investors focus on short-term stock price increases and therefore tend to incentive firm management with short-term bonusses. Since these investors are influential, they can implement short-term bonus schemes in the firms. The authors indeed find evidence of influential short-term investors steering the incentive compensation design towards the short term (Cadman & Sunder 2014). Next to that, Cadman and Sunder (2014) document the negative effects of such behavior. They show that firms that have incentives with a longer horizon have higher long-term abnormal returns, but that the returns for the first half year are not affected by the incentive horizon (Cadman & Sunder 2014). This means that the short horizon incentives provided by the short-term investors leads managers to protect the share price in the first half year. In later years, the short-term protection actions have a negative impact on share price. The research thereby documents how short-term investors can create myopic managers. (Cadman & Sunder 2014) The findings of Bushee (2001), Cadman and Sunder (2014) and Polk and Spazienza (2009) contradict the results of Abarbanell and Bernard (2000) and Bartov et al. (2002). Bushee (2001), Cadman and Sunder (2014) and Polk and Spazienza (2009) find evidence of myopic investor behavior influencing myopic managerial behavior. On the other hand, Abarbanell and Bernard (2000) and Bartov et al. (2002) document rational investor behavior, thereby showing no evidence for the myopia relationship between investors and managers. The findings of both perspectives might be integrated by focusing on transient investors. While Bushee (2001) and Polk and Spazienza (2009) document the strongest evidence for this type of investor and Cadman and Sunder (2014) show similar results, Abarbanell and Bernard (2000) and Bartov et al. (2002) do not incorporate investor characteristics in their analyses. An analysis by Bartov et al. (2002) of for example transient institutional investors and future earnings pricing might have yielded different results.

Overall, the literature on managerial myopia shows that the behavior is prevalent in multiple settings, such as at the beginning of a manager’s tenure, at the end of a manager’s tenure and in settings where managers face targets (Ali & Zhang 2015; Dechow & Sloan 1991; Degeorge et al. 1999). There is also investor-related evidence on the causes of managerial myopia (Bushee 2001; Cadman & Sunder 2014; Polk & Spazienza 2009). While in the preceding sections I have set out the antecedents of managerial myopia as well as the concept itself, I have not documented the effects.

(15)

13

Earnings management may be an outcome of managerial myopia. Namely, when a manager has a short-term focus and is unable to reach short-term earnings targets, he might engage in earnings management. Earnings management may then aid in reaching the target but may carry long-term costs. I touched upon earnings management in the previous section, but in the next section I elaborate on the concept of earnings management, the motives for earnings management and its relationship with managerial myopia. I focus on accruals-based earnings management.

2.2 Accruals-based earnings management

Accruals-based earnings management (AEM) is one of two types of earnings management and is the management of earnings through accruals. Earnings are often partly cash-based and partly accruals-based, meaning that some of the earnings have been received as a cash flow and that some of the earnings have been accounted for through accruals to match the timing of expenses and revenues (Picker et al. 2016). Within for example IFRS and US GAAP, managers have discretion over the accounting for some of the accruals (Walker 2013). These are known as discretionary accruals (Walker 2013). Examples of discretionary accruals are inventory valuation, revenue recognition, amortization rates and bad debt provisions (Walker 2013). These discretionary accruals are the accruals through which managers are assumed to manage earnings, because these allow the discretion to adjust earnings upwards or downwards. A manager might for example change the provision for bad debt from 5 per cent to 3 per cent of total receivables, thereby decreasing the write-off on the income statement.

Earnings management also relates to agency theory. Managers may engage in earnings management for motives related to compensation (Scott 2015). Such earnings management may be value-decreasing and can therefore be a moral hazard for investors (Scott 2015). This compensation setting relates to the negative perspective of earnings management (Scott 2015). On the other hand, earnings management may also provide agency theory-related benefits. Earnings management may namely decrease information asymmetry (Walker 2013). If investors allow management to conduct earnings management, the managers might be able to signal the actual state of the firm to the investors, thereby decreasing the information gap between investors and managers (Walker 2013). This is known as the information perspective of earnings management (Walker 2013).

(16)

14

There is a large amount of empirical research that investigates AEM. The research on AEM is roughly divided by the underlying motives for engaging in earnings management. One stream of research documents findings on earnings management for compensation motives (e.g.: Healy 1985). A second stream of research focuses on earnings management for reaching investor earnings targets (e.g.: Skinner & Sloan 2002). A third stream of research on earnings management discusses the concept in relation to SEOs and IPOs (e.g.: Cohen & Zarowin 2010), while a fourth stream focuses on accruals-based earnings management for maintaining covenants (e.g.: Sweeney 1994). Separate from this, there is literature that documents the costs and benefits of AEM (e.g.: Sloan 1996).

The research that I conduct connects with the compensation-related AEM and investor target-related AEM streams. The frequency of reporting may namely influence the amount of investor targets and the amount of compensation targets, but the reporting frequency does not necessarily influence SEOs or covenants. I therefore focus on reviewing the first two streams of AEM literature. Finally, I review the literature on the costs and benefits of AEM.

One of the most notable papers in the literature on compensation-related AEM is the research of Healy. Healy (1985) researches how discretionary accruals relate to accounting-based managerial compensation incentives. In the compensation structures of the sample of Healy (1985), there are bogeys and caps. A bogey is a minimum target that needs to be achieved to earn a bonus. After the bogey, the bonus increases as a function of how much the target is exceeded. The cap is the maximum on the bonus increase. After the cap, further increases in performance on the measure provide no additional bonus.

Healy (1985) poses and finds that managers have incentives to upwardly manage earnings in the form of accruals when they are likely to come above the bogey. On the other hand, he finds that managers are likely to downwardly manage earnings in the form of accruals when they are not likely to reach the bogey, thereby deferring revenues to later periods. This is commonly known as ‘taking a bath’. Healy (1985) shows that managers engage in earnings management to derive additional compensation.

Next to the papers of Ali and Zhang (2015), Dechow and Sloan (1991) and Kalyta (2009), which I discussed earlier, Bergstresser and Phillippon (2006) present evidence of managing earnings for compensation purposes. The authors research how equity-based managerial compensation incentives relate to discretionary accruals (Bergstresser & Phillippon 2006).

(17)

15

They find that firms with more incentivized CEOs, meaning CEOs whose wealth is more sensitive to stock price, which is known as delta, have higher levels of AEM (Bergstresser & Phillippon 2006). Chava and Purnanandam (2010) show similar results and add that managers whose wealth is more sensitive to stock price volatility, known as vega, engage less in AEM because AEM is an income-smoothing measure and hence leads to less stock price volatility. On the other hand, there is research that documents no relationship between equity compensation and AEM (Walker 2013). Armstrong, Jagolinzer and Larcker (2010) employ a comprehensive matching analysis and find indicative evidence that supports the contrary, namely that equity incentives decrease AEM-related accounting irregularities. Erickson, Hanlon and Maydew (2006) show that equity incentives do not lead to more fraud. Walker (2013) integrates the findings of Erickson et al. (2006) and Armstrong et al. (2010) with other literature. The author outlines that it might be possible that equity incentives might indeed have AEM-increasing effects, but only for within-GAAP practices, and that equity incentives have no effect on accounting alterations that fall outside of GAAP (Walker 2013).

The literature that describes accruals-based earnings management in relation to investor targets is known as the meeting and beating earnings (MBE) benchmarks literature. This literature stresses the pressures on managers to meet and beat investor earnings benchmarks. The pressure on managers is evident from the paper of Skinner and Sloan (2002). They find that for growing firms, not meeting the investor benchmark is penalized (Skinner & Sloan 2002). The authors model that this difference in negative returns between growing stocks and mature stock is around -4 per cent (Skinner & Sloan 2002: 292). Such economically significant penalties may lead to managerial myopic behavior because managers might become loss-averse. The paper of Bartov et al. (2002), which I discussed earlier, is also an example of MBE research. They find a positive relationship between MBE benchmarks and abnormal returns, even when the benchmarks are reached through AEM (Bartov et al. 2002). This finding shows a clear incentive for management to engage in AEM. Similar findings have been documented by Gore et al. (2007), who find that firms manage earnings to MBE benchmarks, by Daske, Gebhardt and McLeay (2006) who find AEM used for income smoothing relative to targets for European firms, and by Das, Kim and Patro (2011). There is thus widespread evidence in literature for managers meeting and beating earnings benchmarks through AEM.

Regarding the cost and benefits of AEM, there are various papers in the MBE stream that not only research MBE behavior but also the potential consequences of the behavior. Bhorjaj et al.

(18)

16

(2009) show that there are benefits for the first three years to use AEM to MBE instead of missing the benchmark without AEM, but that these benefits reverse after these three years. After the three years, the firms that missed the benchmark because they did not apply AEM show higher performance. The findings of Bhorjaj et al. (2009) are consistent with the managerial myopia contention of earnings management. On the other hand, Bartov et al. (2002) show that there are long-run benefits to using AEM for MBE benchmarks. Das et al. (2011) also find that there are costs to earnings management, but that these are outweighed by the benefits. This seeming contradiction may be explained by time window differences. Whereas Bartov et al. (2002) research the effects for up to three years and Das et al. (2011) for a duration of a maximum of two quarters, Bhorjaj et al. (2009) show that the performance reversal starts from three years.

There is indicative evidence that AEM for compensation motives also carries costs. Elliott and Hanna (1996) research the concept of taking a bath and its effect on the earnings response coefficient (ERC). The ERC indicates how strongly investors react to new earnings (Elliott & Hanna 1996). They find that managers who often incur special write-downs, have a lower earnings response coefficient (Elliott & Hanna 1996). Such a lower ERC might translate to higher agency costs because of information asymmetry.

Another type of cost of AEM is the cost of detection. When managers are found to have made excessive use of AEM, financial market authorities might order an earnings restatement. Firm announcements of such pending earnings restatements generally trigger negative market reactions. Palmrose et al. (2004), for example, find negative abnormal stock returns of about -9 per cent for the two days surrounding the announcement (Palmrose et al. 2004: 5-9). Desai et al. (2006) find that managers who are ordered to restate earnings have lower career prospects and a higher chance of dismissal. Lev et al. (2008) find that earnings restatements significantly increase the probability of litigation, and especially so if they change the momentum of the earnings pattern.

An example of a paper that has documented positive earnings management effects is the research of Bowen, Rajogpal and Venkatachalam (2008). The authors associate managerial discretion with future firm performance. Managerial discretion is often interpreted as allowing managers to behave opportunistically (Bowen et al. 2008). However, following the information perspective from agency theory, managerial discretion might also enable managers to signal private information to investors (Scott 2015). The authors find that managerial discretion is

(19)

17

positively linked to future firm performance, which confirms the information perspective hypothesis (Bowen et al. 2008).

In an experimental research setting, Koonce and Lipe (2010) research income smoothing. Income smoothing is a form of AEM done by managers to make the earnings pattern less volatile. Koonce and Lipe (2010) find that the participants give higher valuations to firms that have a smoother earnings pattern.

Managers thus have multiple motives to engage in accruals-based earnings management. The reviewed motives come in the form of attaining a market premium for stock when beating benchmarks habitually or in the form of gaining additional compensation when beating a target. There is also evidence of AEM being myopic. Degeorge et al. (1999), amongst others, show this for compensation motives. Bhorjaj et al. (2009) document long-term costs for MBE behavior. Other costs of AEM are detection costs and lower earnings responses (Desai et al. 2006; Elliot & Hanna 1996; Lev et al. 2008; Palmrose et al. 2004). Finally, there are some authors who note benefits to AEM, such as Bowen et al. (2008).

The findings on AEM for both compensation motives and MBE motives can be related to a firm’s reporting frequency. The reporting frequency influences the frequency of investor earnings targets because the reports provide information for target setting. For the same reason, a firm’s reporting frequency may also influence the frequency of compensation targets. In the next section, I outline the literature on reporting frequencies and its connections with AEM and myopia.

2.3 Reporting frequency

The concept of reporting frequency relates to the theoretical debate on disclosures. A higher reporting frequency is logically associated with a higher level of disclosure, ceteris paribus. The costs and benefits of disclosures have been researched from multiple perspectives. Beyer, Cohen, Lys and Walther (2010) review the literature on voluntary and mandatory disclosures. The research into voluntary disclosures is related to a theory known as the unraveling result (Beyer et al. 2010). The unraveling result is the theory that specifies under which conditions a manager will voluntarily disclose information (Beyer et al. 2010). These conditions create a perfect market and in that perspective the theory is similar to the theory of Modigliani and Miller (1958;1961). If the conditions hold, the argument of the theory is as follows. Investors

(20)

18

make rational inferences. From these rational inferences, investors know that if managers do not disclose information, the information is probably bad. Hence, if managers opt to non-disclose, investors estimate a lower price for the firm. Managers, knowing how investors will behave, will disclose information to differentiate themselves from firms that have worse information. (Beyer et al. 2010)

One of the key conditions for this line of reasoning to hold is that there are no costs for managers to disclose information (Beyer et al. 2010). In real world situations, this condition is unlikely to uphold, as there are both direct costs to disclosures, such as the costs of making a report, as well as indirect costs, such as proprietary costs (Beyer et al. 2010).

Regulation on disclosures, such as the mandatory 10-Q statement in the U.S., provides evidence against the workings of the unraveling result theory (Beyer et al. 2010). Namely, if managers know that investors will devalue their stock if they withhold any information, managers would be inclined to disclose all information. In such a world, mandatory disclosures would not be necessary.

Research has proposed multiple reasons why mandatory disclosures are enacted (Beyer et al. 2010). The key underlying argument is that without regulation, firms might make suboptimal disclosure choices from a societal point of view (Beyer et al. 2010). For example, regulation may increase the comparability of the earnings statements, which may lead to increases in the efficiency and accuracy of analyst forecasts (Beyer et al. 2010). However, the research on mandatory regulation has been unable unify the perspectives into one theory, as most of the benefits also bear equally high costs and hence finally provide no reason for mandatory disclosures (Beyer et al. 2010).

The research on the effects of reporting frequencies is still developing. Evidence of this is that most of the research on the topic has been written in recent years (e.g.: Ernstberger et al. 2017; Fu et al. 2012; Kraft et al. 2018). Two streams have developed within the empirical literature on reporting frequency. One stream documents benefits to high reporting frequencies, while the other stream shows costs associated with high reporting frequencies. Another strand of literature researches the influence of the frequency of other types of disclosures.

Butler et al. (2007) examine the hypothesis that a quarterly reporting frequency should lead to benefits in terms of higher earnings timeliness with annual earnings. Earnings timeliness is the amount of time it takes for accounting information to be reflected in the stock price (Butler et

(21)

19

al. 2007). With additional disclosures during the year to predict the annual earnings at the end of the year, analysts should be able to make better forecasts and be able to adjust their valuations more swiftly following the annual earnings disclosure (Butler et al. 2007). Butler et al. (2007) find no evidence of higher earnings timeliness for mandatory quarterly reporters relative to mandatory semi-annual reporters. However, the authors do find that for voluntary quarterly reporters, there is indeed higher earnings timeliness (Butler et al. 2007). They explain this finding from a managerial rationality perspective: managers voluntarily increase the reporting frequency if this provides benefits (Butler et al. 2007).

Fu et al. (2012) also approach reporting frequency from a perspective of benefits. The authors use the same setting as Butler et al. (2007). One of the benefits, according to theory, is that an increased reporting frequency might increase disclosures and thereby reduce the information asymmetry between managers and investors (Lundholm 1991). Another hypothesized benefit of a higher reporting frequency is a lower cost of equity, because a higher reporting frequency reduces adverse selection problems (Fu et al. 2012). It might reduce such adverse selection problems because the investor knows more about the actual state of the firm. If the investor knows more about the actual state, he will not decrease his valuation as under normal adverse selection scenarios. Fu et al. (2012) show that regardless of the reporting frequency being mandatory or voluntarily, a higher frequency indeed reduces information asymmetry and cost of equity. Both Butler et al. (2007) and Fu et al. (2012) thus show that there are economic reasons to report quarterly.

Ernstberger et al. (2017) examine the potential costs of higher reporting frequencies. They use a sample of European firms that had to switch to quarterly reporting after the EU’s Transparency Directive was implemented in 2004 (Ernstberger et al. 2017). They split their sample in firms that included earnings information in their quarterly reports and in firms that did not include such information (Ernstberger et al. 2017). For the firms that did include such information, they find that the adoption of quarterly reporting led to increases in RAM (Ernstberger et al. 2017). The authors also analyze what the impact is of an increase in the reporting frequency on operating performance (Ernstberger et al. 2017). They find that firms with an increased reporting frequency report higher operating performance in the first year, but that the operating performance is lower than expected in up to six years after (Ernstberger et al. 2017: 52).

(22)

20

The findings of Ernstberger et al. (2017) correspond to the survey answers to the study of Graham et al. (2005), who find that managers are willing to conduct RAM to reach targets. When there are more targets per year due to more earnings reports being released, the level of real activities manipulation should logically increase. The findings of Ernstberger et al. (2017) also provide evidence that an increase in the reporting frequency and the subsequent increase in RAM is related to myopia.

Kraft et al. (2018) research the impact of reporting frequencies on managerial investment decisions. They conduct their research with a sample of American firms over a period from 1950 to 1970, a time period in which the mandatory reporting frequency steadily increased (Kraft et al. 2018). The authors find that an increase in reporting frequency is related to a decrease in investments (Kraft et al. 2018). The decrease that Kraft et al. (2018) present is statistically and also economically significant, namely 1.2 per cent of capital expenditures (Kraft et al. 2018: 260). In additional analyses, Kraft et al. (2018) attempt to document what caused the decline in investments. They have two alternative hypotheses. The managerial myopia hypothesis links the underinvestment to managers responding to capital market pressures, while the managerial discipline hypothesis links the underinvestment to managers spending less capital on excess investments (Kraft et al. 2018). The authors find that more frequent reporters have in the long-run reduced growth and productivity, which confirms the managerial myopia hypothesis (Kraft et al. 2018).

Houston, Lev and Tucker (2010) investigate voluntary disclosures. Some firms opt to include forecasts for the next quarter in their quarterly reports, which is known as giving earnings guidance. The authors conduct their research to provide inputs to the debate whether the practice of giving earnings guidance is beneficial or costly (Houston et al. 2010). Houston et al. (2010) compare firms that stop issuing earnings guidance with firms that maintain the disclosure behavior. They find no evidence of the stoppers acting less myopically, as proxied by R&D and capital expenditure increases (Houston et al. 2010). R&D expenditures and capital expenditures are both forms of long-term investment. Hence, myopic behavior would be characterized by underinvestment on these items. Rather, Houston et al. (2010) find that the stopping of earnings guidance leads to less overall firm disclosures, greater analyst earnings forecast errors, and decreases in analyst following.

Call, Chen, Miao and Tong (2014) also study earnings guidance reports and provide inputs to the debate. Specifically, Call et al. (2014) research how the disclosure of earnings guidance

(23)

21

reports relates to accruals-based earnings management. The authors find that firms that issue guidance engage less in accruals-based earnings management than firms that do not issue earnings guidance (Call et al. 2014). This contrasts to the societal expectation that earnings guidance increases earnings but not cash flows (CFA Institute 2006).

Related to this, is the method of disclosing voluntary information. Brochet, Loumioti and Serafeim (2015) study the association between short-termism of managers in conference calls to investors and earnings management. They create a proxy to determine through the wording of the conference disclosures the level of short-termism of the manager (Brochet et al. 2015). To ensure the correctness of the proxy, the authors compare their proxy with proxies for short-termism used by other researchers and find that there is significant association between the constructs (Brochet et al. 2015). They find that there is a positive association between short-term wordings and RAM and AEM (Brochet et al. 2015).

The research on reporting frequencies and disclosures shows both benefits as costs to quarterly disclosures. Earnings guidance literature mainly documents benefits of quarterly guidance, such as a lower cost of equity and more precise analyst estimates. The reporting frequency literature shows both benefits as costs related to higher reporting frequencies. The research question of this thesis is most closely related to the reporting frequency literature. This because the voluntary reports, such as earnings guidance statements, are subject to much different levels of regulation than quarterly statements. Earnings guidance statements also contain more forward-looking information than quarterly earnings reports. To my knowledge, no research has been conducted on the relationship between reporting frequency and AEM.

The next section sets out the regulatory environment in the EU-28 on reporting frequencies. Since there are different effects documented for voluntary disclosures and mandatory disclosures, I find it important to discuss the disclosure regimes.

2.4 Institutional background

In 2004, the European Parliament and Council issued the Transparency Directive 2004/109/EC. One of the main objectives of the Directive was the harmonization of reporting requirements (EU 2004). The Directive required, amongst others, that all companies include interim management statements (IMSs) for the periods between the semiannual reports (EU 2004). The directive was subsequently transposed to member state regulations between 2007-2009. From then on, companies that were previously reporting semiannually had to report quarterly.

(24)

22

The requirements of these IMSs are found in Article 6 of the Directive. It outlines that the statements should include “an explanation of material events and transactions that have taken

place during the relevant period and their impact on the financial position of the issuer and its controlled undertakings, and a general description of the financial position and performance of the issuer and its controlled undertakings during the relevant period” (EU 2004: 390/46).

The directive did thereby not include any obligation for companies to report quantitative earnings information. It also allowed stock exchanges and member states to impose stricter rules regarding the amount of yearly reports and the information in the reports (Link 2012). In Austria and Germany, for example, the prime markets additionally required the quarterly reports to include earnings information (Link 2012). Other countries, such as Bulgaria, Latvia and Poland, additionally required that earnings information be included in quarterly reports for all listed firms (Link 2012). The option to impose stricter requirements resulted in that the harmonization objective of the Directive was not achieved (Link 2012). Different countries and different stock exchanges imposed different rules.

In 2010, Mazars published a report on behalf of the European Union that surveyed stakeholders on the impact of the Directive. Stakeholders on the investment-side were generally satisfied with the new IMS standard, with 82% of them answering that the interim management statements were useful and 69% of them responding that the statements aided in market efficiency (Mazars 2010: 9). Some stakeholders on the issue-side reported that the statements created difficulties (Mazars 2010). This complaint mainly came from SMEs, as for these firms the new requirements were especially costly (Mazars 2010). The Directive was thus generally well-received, but the assessment clarified that there was some room for improvement in terms of requirements for SMEs.

Given the positive results of the Directive, it was surprising that in 2013 the EU decided to abolish the requirement of IMSs altogether through Directive 2013/50/EU, which was subsequently enacted in national legislations between 2014 and 2016. The directive stipulated that “member states should not be allowed to impose in their national legislation the

requirement to publish periodic financial information on a more frequent basis than annual financial reports and half-yearly financial reports” (EU 2013: 294/13). The EU justified the

directive that through the abolishment the EU intended to encourage long-term investment and to reduce the costs for SMEs (EU 2013). The move to abolish quarterly reporting is less surprising than the Mazars report would suggest, when the societal circumstances are considered.

(25)

23

In the period between 2004 and 2013 the financial crisis and euro crisis occurred, which both severely damaged the European economies. The financial crisis is partly attributed to investor short-termism (Dallas 2012). The large-scale investments in derivatives and other securities paid off short-term benefits, but had long-term value destroying effects because they did not align with the underlying risks of the economy (Dallas 2012). Within national discussions, there was support for abandoning quarterly reporting. In the UK in 2012, for example, a review was undertaken on the effects of quarterly reporting. The review’s final report recommended abolishing mandatory quarterly reporting (Kay 2012). The amendments made by the Directive were in line with the trend of Dallas (2012) and the Kay Review (2012).

Table 1 provides an overview on the reporting frequency regimes changes within the EU-28. It shows the national reporting regimes in 2010 and it also shows the years of the transpositions of the 2013/50/EU Directive. Most countries enacted the directive that abolished mandatory quarterly reporting in 2015. For the member states that already had a voluntary reporting regime, the Directive had no effect. For the member states that did have a mandatory quarterly reporting requirement, the Directive led to quarterly reporting becoming voluntary. The directive thereby reduced the mandatory reporting frequency from quarterly reports to semi-annual reports in Austria (for prime markets), Bulgaria, Croatia, Estonia, Finland, Germany (for prime standard), Greece, Hungary, Latvia, Lithuania, Poland, Portugal, Slovakia and Sweden.

3 Hypothesis development

Related literature suggests that managers may behave myopically for various motives. Dechow and Sloan (1991) and Kalyta (2009) indicate that managers are short-term focused to attain additional compensation. Degeorge et al. (1999), Polk and Spazienza (2009) and Cadman and Sunder (2014), amongst others, document myopic managerial behavior stemming from investor pressures.

In the earnings management literature, there is evidence of compensation motives affecting managers (e.g.: Ali & Zhang 2015; Bergstresser & Phillippon 2006; Healy 1985) and investor pressures affecting managers (e.g.: Bartov et al. 2002; Bhorjaj et al. 2009; Daske et al. 2006; Gore et al. 2007).

Specifically, the earnings management literature documents that investor pressures and compensation targets can lead managers to conduct accruals-based earnings management.

(26)

24

Research on how the reporting frequency of earnings information affects managerial myopia is scarce and still developing, with some major topics currently not yet addressed. One of these topics is accruals-based earnings management. While research documents associations between higher reporting frequencies and RAM (e.g.: Ernstberger et al. 2017) and higher reporting frequencies and underinvestment (e.g.: Kraft et al. 2018), there is, to my knowledge, no research conducted on AEM.

A higher reporting frequency may lead to more earnings management, because logically, firms that face earnings targets that they find necessary to meet more often, have more instances where they might be persuaded to engage in earnings management. This leads to the following hypothesis:

H1: Firms that issue quarterly reports have higher accruals-based earnings management relative to firms that report semi-annually.

TABLE 1 REPORTING REGIMES Country/Stock Market In 2010 2013/50/EU in effect Country/Stock Market In 2010 2013/50/EU in effect

Austria – P.M. Vol. 2015 Ireland Vol. 2017

Austria – N.P.M. Mand. 2015 Italy Vol. 2016

Belgium Vol. 2016 Latvia Mand. 2016

Bulgaria Mand. 2016 Lithuania Mand. 2015

Croatia Mand. 2015 Luxembourg Vol. 2016

Cyprus Vol. 2017 Malta Vol. 2016

Czech Republic Vol. 2016 Netherlands Vol. 2016

Denmark Vol. 2015 Poland Mand. 2016

Estonia Mand. 2015 Portugal Mand. 2016

Finland Mand. 2015 Romania Vol. 2017

France Vol. 2015 Slovakia Mand. 2016

Germany – G.S. Vol. 2015 Slovenia Vol. 2015

Germany – P.S. Mand. 2015 Spain Vol. 2015

Greece Mand. 2016 Sweden Mand. 2016

Hungary Mand. 2015 United Kingdom Vol. 2015

This table summarizes the reporting regimes within the EU-28 and the adjustments to these regimes. “Vol." indicates that the disclosure of earnings information is voluntary, while "Mand." indicates that such disclosures are mandatory. The column “In 2010” provides information on the reporting regime at the start of my sample. The column "2013/50/EU in effect" shows when subsequent EU legislation was enacted that abolished all quarterly reporting requirements. “N.P.M.” and P.M.” stand for “Non-Prime Market” and “Prime Market” respectively, “G.S.” and “P.S” stand for “General Standard” and “Prime Standard” respectively. Information on EU-15 countries has partly been compiled using information from Ernstberger et al. (2017). I compile other information from the Eur-Lex database, which documents national transpositions. For robustness, I check the observations for any noticeable changes per country in the distribution of semi-annual and quarterly reporters.

(27)

25

Although I find the strongest indications in the literature for a positive relationship between reporting frequency and AEM, it could be that there is no relationship between the two concepts. Fu et al. (2012) find that quarterly reporters have lower information asymmetry than semi-annual reporters. Lower information asymmetry might increase the likelihood of AEM being detected. It may then be that the increase in AEM detection risk offsets the increase in AEM incentives for higher reporting frequencies.

4 Research design

4.1 Sample selection

I use the European Union over the years 2010-2017 as my research setting. In this time period, quarterly reporting was both mandatory and voluntary, depending on the country and year. Since there is variation in the frequency of earnings reporting in the European Union, the EU provides the possibility to test my hypotheses. Next to that, the EU provides a common market with common regulation, which controls for some of the country-effects that might affect my models. However, as documented by Leuz et al. (2003), within the EU-28 there are differing levels of enforcement against earnings management. Therefore, in my analysis, I control for the country-level of enforcement. In later analyses, I control for differences between mandatory quarterly reporters and voluntary quarterly reporters.

Information on reporting frequencies and on equity market values is collected using the Datastream/Worldscope databases, while the majority of the other information is retrieved from Compustat. I make use of data from the years 2010 to 2017. I use 2005-2009 data to create lagged variables. I have selected 2010 as base year to partially negate the effects of the financial crisis (Ernstberger et al. 2017). However, by employing 2009 data for the majority of the lagged variables and 2005-2009 data for volatility variables, the effects of the financial crisis may still be marginally impacting my data.

Table 2 panel A provides an overview of my sample selection methodology. The initial sample of EU-28 firm-years in COMPUSTAT consists of 48,603 unique observations. The tools that the university offers in combination with Datastream/Worldscope constrain the maximum amount of observations. Most methods for data extraction from Datastream/Worldscope lead to only the largest firms in terms of market capitalization being included. To mitigate such bias, I include only active firms. Active firms are firms that are listed in the year 2017 and that are not liquidated, delisted or merged between 2010 and 2017. While the measure is not unbiased,

(28)

26

as smaller stock are more probable to become delisted due to costs associated with listings, it is a method to obtain a relatively random sample of European firms from the Datastream/Worldscope databases. From this sample, I delete firms that are cross-listed to prevent different reporting regimes affecting my sample, firms that have missing data on SIC industry classification and firms with negative total assets. Next, I remove financial services firms because such firms use different reporting methodologies that do not fit well with earnings management detection models. I delete observations from 2009 after creating lagged variables. I subsequently remove observations that have missing observations for reporting frequency. I also delete observations with missing values for determining accruals. Finally, I delete observations with missing data for the control variables. This selection methodology leads to a sample of 22,981 firm-years.

Panel B presents the distribution of my sample across the EU-28 countries. There are large differences in the mean reporting frequency of countries. From the sample of firms from Estonia, Croatia and Latvia, only firms that report quarterly are in the final sample. Nearly all Bulgarian, Lithuanian, Portuguese and Swedish firms also report quarterly. This corresponds to the overview provided in table 1, which shows that these countries all had mandatory quarterly reporting regimes. The noticeable exceptions are Greece and to some extent Finland and Hungary, where the percentage of quarterly reporters decreased relatively rapidly since the recent abolishment of mandatory quarterly reporting.

4.2 Method

4.2.1 Accruals estimation

Most accruals models are specified either following the Jones model or the modified Jones model. These models differentiate between total accruals, non-discretionary accruals and discretionary accruals. In these models, total accruals is the amount of accruals calculated from the balance sheet or cash flow statement (Jones 1991). The accruals estimated on the other side of the equation are known as non-discretionary accruals, because these accruals are ‘normal’ given the level of operations (Jones 1991). The models then estimate discretionary accruals as the difference between total accruals and non-discretionary accruals (Jones 1991).

The changes in revenues and the amount of property plant and equipment (PPE) influence the level of total accruals but also the level of non-discretionary accruals (Jones 1991). Total accruals is positively influenced by revenues.

(29)

27

TABLE 2 SAMPLE SELECTION

Panel A: Sample selection methodology

Firm-years

European firms in Compustat (2009-2017) 48,603 - Negative assets or no asset data -303 - Cross-listed firms -1,953 - Missing industry classification data (SICH) -126 - Financial services firms (SIC 6000-6999) -8,496 - Dropping year 2009 observations -3,691 - Missing accruals data -4,104 - Missing reporting frequency data -3,822 - Missing control variables data -2,620

- Not allocated to accrual industry-year groups -507

Final sample 22,981

Panel B: Accrual estimation sample distribution by country Country No. Obs. Mean reporting

frequency

Country No. Obs. Mean reporting frequency Austria 266 0.90 Estonia 88 1.00 Belgium 507 0.20 Finland 730 0.93 Bulgaria 235 0.98 France 3,045 0.07 Croatia 344 1.00 Germany 2,584 0.73 Cyprus 211 0.24 Greece 1,043 0.83

Czech Republic 34 0.85 Hungary 63 0.94

Denmark 604 0.76 _Ireland 201 0.08

Country No. Obs. Mean reporting frequency

Country No. Obs. Mean reporting frequency Italy 1,131 0.79 Portugal 269 0.96 Latvia 131 1.00 Romania 232 0.71 Lithuania 173 0.97 Slovakia 17 0.76 Luxembourg 138 0.69 Slovenia 128 0.77 Malta 57 0.04 Spain 705 0.66 Netherlands 509 0.34 Sweden 2,200 0.98

Poland 2,211 0.99 United Kingdom 5,125 0.03

Panel A presents the sample selection process for the sample that combines COMPUSTAT and Datastream/Worldscope databases. Panel B shows the sample distribution per country and the mean reporting frequency per country. Appendix A provides information on the reporting frequency variable. A reporting frequency of 1 equals quarterly reporting, while a reporting frequency of 0 equals semi-annual reporting. I categorize a firm as quarterly reporting when it includes net income in its quarterly reports.

As revenues increase, net income increases, but a portion of these sales is not converted to cash in the accounting period. This portion of sales is recognized as an accrual. Hence, when revenues increase, total accruals increase. The increase in revenues is recognized as a non-discretionary accrual because the effect occurs naturally, without managers influencing it discretionally (Jones 1991).

The influence of reporting frequency on accruals-based earnings management