Classification Shifting as an Earnings Management tool
Among EU and US firms
Name: Rudolf (Rolf) Hodel Student number: 10285482 Thesis supervisor: Reka Felleg Date: Jun 25, 2018
Word count: 15007
MSc Accountancy & Control, specialization Auditing and Control Faculty of Economics and Business, University of Amsterdam
Statement of Originality
This document is written by student Rudolf Hodel 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.
Abstract
This master thesis focusses on classification shifting as an earnings management tool and compares the degree to which EU firms reporting under IFRS and US firms reporting under US GAAP engage in this type of earnings management to opportunistically shift core earnings. I use a modified version of McVay (2006) her level and changes model in order to test whether current period non-recurring earnings are related to unexpected core earnings. I further test whether these core earnings are persistent in the following year. I find that both EU firms reporting under IFRS as well as US firms reporting under US GAAP engage in classification shifting to opportunistically core expenses to non-core expenses and non-core earnings to core earnings. Finally, I test whether classification shifting is more pervasive for EU firms than among US firms reporting. I find no evidence that this is the case, in fact I find significant evidence that EU firms engage in less classification shifting as compared to US firms. There results are valuable for a broad group of stakeholders, from analysts estimating future stock performance based on core earnings data to legislators aiming to curtail
Contents
1 Introduction ... 5
2 Theoretic framework ... 8
2.1 Agency Theory ... 8
2.2 Earnings management ... 8
2.2.1 Accrual-based earnings management ... 9
2.2.2 Real activity-based earnings management ... 10
2.2.3 Classification shifting ... 11
2.3 Reasons for earnings management ... 12
2.3.1 Earnings management to increase future performance ... 12
2.3.2 Earnings management and CEO compensation ... 13
2.4 Accounting Standards ... 15
2.4.1 IFRS ... 15
2.4.2 US GAAP ... 18
2.5 Hypothesis Development ... 18
3 Methodology ... 21
3.1 The Modified McVay Model ... 21
3.2 Comparing Classification Shifting ... 25
3.3 Sample selection and Descriptive statistics ... 25
4 Descriptive Statistics ... 28
4.1 Descriptive Statistics ... 28
5 Main Test Results... 32
5.1 Levels Model ... 32
5.2 Changes Model ... 33
5.3 Comparing Classification Shifting between the US and EU ... 37
6 Discussion and Conclusion ... 39
7 References ... 42
1 Introduction
Accrual-based earnings management and real earnings management are extensively researched earnings management tools in the accounting literature (Abernathy, Beyer, & Rapley, 2014). McVay (2006) argues the above tools are not the only way managers seek to alter the perceptions of stakeholders. She studies classification shifting as a tool to manage earnings. This manner of earnings management is different from conventional methods such as accrual-based earnings management and earnings management using real transactions, in that it does not change bottom line net profit. Instead, managers by shifting recurring costs to non-recurring costs and non-recurring revenues to recurring revenues, manage the amount of recurring profits they present in the income statement (McVay, 2006). Recurring profits are seen as sustainable as by definition they result from the normal operations of the company. If a manager opportunistically shifts revenues and expenses to increase the portion of profits that is reported as recurring, this can thus directly influence market opinion of the firms’ performance and value (Alfonso, Cheng, & Pan, 2015). Moreover, Haw, Ho, & Li (2011) suggest that classification shifting undermines the credibility of financial statements by misleading investors about the persistence of firms’ performance. This is further evidenced by Dichev, Graham, Harvey, and Rajgopal (2012) who based on extensive survey based research find that the most important characteristics of high quality earnings are that they are sustainable and do not include large non-recurring components. Hence, classification shifting directly impacts earnings quality and the ability of analysts to accurately analyze and value firm performance leading to overvaluations of firms opportunistically shifting recurring profits (Alfonso et al., 2015). Despite the importance of high quality reporting on sustainable earnings, the research field lacks systematic evidence on classification shifting in a European setting.
There are only a handful studies employing large sample based research on the subject and the large majority of studies on classification shifting only take into account United States firms reporting under United States generally accepted account standards, US GAAP, hereafter (Alfonso et al., 2015; Barua, Lin, & Sbaraglia, 2010; Fan, Barua, Cready, & Thomas, 2010; McVay, 2006). Even the research by Behn, Gotti, Herrmann, & Kang (2013) that aims to study classification in an international setting uses a sample of which 42 percent are US firms reporting under US GAAP and only 17 percent of the sample consists of EU firms reporting under IFRS. Furthermore, the study is aimed at examining the relation between internal governance and classification shifting and does not seek to compare the pervasiveness of classification shifting between US firms reporting under US GAAP and EU firms reporting under IFRS.
To my knowledge the only studies looking at classification shifting without using US firms in their sample are Haw et al. (2011) who study the practice in an East-Asian setting, and Athanasakou, Strong, & Walker (2008) and Zalata & Roberts (2016) who both look at classification shifting in the United Kingdom. Athanasakou et al. (2008) only find weak evidence for classification shifting for sub-sample 178 observations from large firms that just meet the earnings benchmark. Given of their main sample of 5,117 the sample for which they find classification shifting is only 3 percent. Zalata & Roberts (2016) however do find significant evidence for classification shifting for firms in the UK. The main difference between both studies is the period they cover. Athanasakou et al. (2008) use a sample of firms reporting under UK GAAP, while Zalata & Roberts (2016) use a sample of firms reporting under IFRS which became mandatory in 2005 for all European listed firms. In their work they reference to Athanasakou et al. (2008) stating that the main difference between their findings is likely the IFRS adopting by the UK in 2005. This would imply that the change from a rule-based to a more principle-based standard leaves more room for managers to opportunistically shift earnings. Furthermore, given the multiple studies (e.g. McVay, 2006; Fan et al., 2010) finding significant evidence for classification for US companies reporting under US GAAP, this could mean that classification shifting under EU firms reporting under IFRS is even more pronounced.
In order to test this, I use two different samples, one of US firms reporting under US GAAP and one of EU firms reporting under IFRS for the period of 2005 to 2016. The period of analysis spans from 2006 to 20161. The method I use to test for classification shifting is a
modified version of the two-stage multiple regression model introduced by McVay (2006). This method consists of two separate two-stage regression to identify classification shifting firms. The first two-stage regression is a levels model and tests for the level of unexpected core earnings that indicate classification shifting. The second two-stage regression is a changes model that tests whether the next-year unexpected change in core earnings is positive or negative. A negative change means that the prior year increase in core earnings was likely not based on real economic growth as it proved not sustainable and therefore was likely due to the firm using classification shifting to increase the core earnings.
Using these two models I first test whether EU firms reporting under IFRS and US firms reporting under US GAAP engage in classification shifting. Using the levels model, I find that both EU and US firms engage in classification shifting, however the evidence for US firms only
1
The EU implemented mandator IFRS adoption for firms listed at a regulated stock exchange in
2005. However, due to the inclusion of lagged variables in the McVay (2006) model I start my
period of analysis in 2006.
hold for a small percentage of the sample with negative non-recurring items. Using the changes model, I find that for both samples the non-recurring items are negatively and significantly related to the next-year unexpected change is negatively. Taken together this proves that both firms in the US and in the EU engage in classification shifting as an earnings management tool.
For the main analysis, whether EU firms engage in more classification shifting than US firms, I create a dummy variable that takes the value of 0 for US firms and 1 for EU firms. This means that the US sample acts as a benchmark for the EU sample. I then compute the mean centered non-recurring items, as this is the main predicting variable for the unexpected earnings. I then multiply the mean centered non-recurring earnings with the dummy and solve for equation 7 with as the dependent the non-expected core earnings. This yields the value of the corresponding coefficient which is positive and significant and thus means that EU firms reporting under IFRS engage in more classification shifting than US firms reporting under US GAAP.
This thesis contributes to the literature on classification shifting as it is to my knowledge the first study that examines whether classification shifting is more pronounced under EU firms reporting under IFRS than US firms reporting under US GAAP. Furthermore, the results are valuable to a broad range of stakeholders, from analysts trying to predict future earnings based on reported core earnings to legislators who aim to minimalize earnings management. Furthermore, the results contradict anecdotal evidence by Dichev et al. (2012) that rule-based approaches such as US GAAP lead to lower quality earnings due to the centralized and mechanical nature creating a box-checking mentality.
2 Theoretic framework
In this section agency theory is first introduced which is seen as the underlying rationale for managers to engage in earnings management. Furthermore, the two ‘traditional’ types of earnings management: accrual-based and real earnings management are described, followed by the more novel form of earnings management by classification shifting as described by McVay (2006). In the third section the International Financial Reporting Standards as adopted by the European Union are described. And lastly in the fourth section the dual role of analysts as a factor influencing earnings management is explained.
2.1 Agency Theory
Agency theory first introduced by Jensen & Meckling (1976) explains the relationship between a principal and an agent and the problem arising how to align their interests. Fundamentally the goals of the agent and principal are the same, to maximize their own value. This creates opposing interests when the principal for example the shareholder, wants to maximize the firm value while the agent, for example the management, wants to maximize his pay package (Eisenhardt, 1989). This relationship between the principal and the agent is characterized by the concept of information asymmetry, both parties hold different levels of information as the shareholder cannot practically monitor everything the management does. This allows management to use this information asymmetry to benefit themselves at the expense of the shareholders (Fama & Jensen, 1983). The management prepares the financial statements of the company, using their superior information as to the performance of the company compared to the shareholders they thus have room to increase their reported performance translating into higher pay-packages and / or retaining their jobs (Fama & Jensen, 1983). According to Agrawal & Mandelker (1987) and Haugen & Senbet (1981) this issue can be solved only once both parties share the same goals. This is often done by issuing stock-plans thus making the agent partly principal, as now he has become a shareholder as well.
2.2 Earnings management
In a survey conducted by Graham et al. (2005) eighty percent of the questioned CFO’s stated they would engage in a form of earnings management in order to meet analyst expectations. And fifty-five percent would even go as far as postponing a new project if this
meant the company would meet analyst expectations. Hence it is fair to state the act is widespread through the financial world and when results in misleading stakeholders and/or the misallocation of resources can have serious consequences for the firm.
There are several different definitions of earnings management. Healy & Wahlen (1999) describe earnings management as a situation when managers use judgement in their reporting and the structuring of transactions to alter financial reports to mislead some stakeholder about the underlying economic performance of the company or to influence contractual outcomes that depend on accounting numbers. This definition reflects the presence of information asymmetry and the ability for managers to make judgements without the shareholders being able to monitor accurately if these judgements give a true and fair view. More modern research often wields a more compact definition such as Zang (2012, p. 676) who describes it simply as “the purposeful altering of reported earnings in a particular direction”. In the literature the two most cited forms of earnings management are accrual-based earnings management (AEM, hereafter) and real activity-accrual-based earnings management (REM, hereafter). These two types are described in the next two subsections. In the third subsection the more novel method of classification shifting is explored.
2.2.1 Accrual-based earnings management
Managers in their reporting have a certain amount of freedom in how to best reflect the underlying economic performance in their financial reports. This on the one hand allows them to better inform shareholders and decrease the information asymmetry, however on the other hand Healy & Wahlen (1999) argue it also create the opportunity for managers to use this freedom to select reporting methods, estimates, and disclosures to steer the reported performance in a particular direction. There are many ways to steer the reported figures using AEM, but all have in common that they influence future periods. If you increase the income in the current reporting period by example given increasing the useful life of your machines with the sole reason to increase net income, then in the future periods these expenses will still occur. Thus in effect AEM shifts mainly the timing of expenses and borrows funds (if income increasing) from future periods (Healy & Wahlen, 1999). Another characteristic of AEM according to Zang (2012) is that contrary to REM it can and often is applied at the end of the reporting period to steer the earnings to a desired outcome. This in contrast to REM which occurs throughout the year.
2.2.2 Real activity-based earnings management
Different from AEM, this type of earnings management is not achieved by changing the accounting methods or estimates but by altering the real activities itself. Chi, Lisic, & Pevzner (2011) outline that an important characteristic of REM is that it is achieved by business decisions rather than accounting decisions. This means that instead of utilizing reporting methods, estimates, and disclosures to steer earnings, managers instead use real transactions to steer earnings. For example, instead of using aggressive revenue recognition which is an accounting decision to increase net profits a manager using REM would reduce research and development expenses to boost net profits which is a business decision. It is difficult to determine when such a business decision constitutes earnings management, given the manager can reasonably argue that it is a result of the normal course of business. Zang (2012, p.676) defines REM as follows:
“The purposeful action to alter reported earnings in a particular direction, which is achieved by changing the timing or structuring of an operation, investment, or financial transaction, and which has suboptimal business consequences.”
Key in this definition is that the decision results in suboptimal business consequences, thus business decisions that steer earnings in the short term in a desired way but have suboptimal business consequences are regarded as REM. Cohen & Zarowin (2010) argue that it is hard to prove the suboptimal performances are due to earnings management and not simply due to suboptimal decision making. This means they argue that it is hard to detect REM and prosecute this type of earnings management. Dechow & Skinner (2000); Fudenberg & Tirole, (1995); Graham et al. (2005); Healy & Wahlen, (1999) all find evidence that companies use methods such as: acceleration of sales, alterations in shipment schedules, and delaying or scrapping of R&D and maintenance expenditures as ways to engage in REM. All these methods differ from AEM in that they not solely affect the manner of reporting the firm’s performance, instead they affect the underlying transactions, and thus the actual cash flows the firm generates.
2.2.3 Classification shifting
A third form of earnings management is classification shifting. This form of earnings management entails the deliberate misclassification of items within the income statement McVay (2006). She argues that up until her publication the bulk of the literature focuses on either AEM, REM, or both. McVay (2006) identifies firms in the United States of America using classification shifting between core earnings and special (non-recurring) items in order to show stronger key figures. Alfonso et al. (2015) explain that these key figures are then used by analysts in calculating the future cash flows the firm will generate and consequently overvalue the firm. Thus, without changing the bottom line earnings figure, by shifting between the different line items companies can still manipulate stakeholders.
McVay (2006) argues that this form of earnings management is distinct from AEM and REM in three main aspects. Firstly, it does not change the GAAP earnings, instead it shifts individual components of the income statement that are meant to be informative to financial statement users. Secondly while all three forms of earnings management increase expected profitability, AEM and REM actually reduce future (or past) period earnings. Either by forgoing a profitable project, lending accruals from other periods, or selling at heavy discount near the end of the year. Under classification shifting the next periods earnings are left unchanged. The period in which the earnings are shifted does not impact any other period than the current period itself. Thirdly, Nelson, Elliott, Tarpley, & Gibbins (2002) find that because GAAP net income remains unchanged there exists limited scrutiny from auditors and some regulators making classification shifting an attractive tool for engaging in earnings management. Furthermore drawing on Cohen & Zarowin (2010) and Zang (2012) their findings on the substitutive relationship between AEM and REM based on their relative costs the findings by Nelson et al. (2002) can be supported if classification shifting has lower relative costs as compared to AEM and REM and therefore constitutes a more attractive earnings management tool. Abernathy et al. (2014) tests this by extending Zang (2012) her research and finds that firms who are constrained in using AEM and REM are more likely to engage in classification shifting.
Although the bottom line net income (GAAP net income) remains unaltered this does not mean that classification shifting does not have consequences for future periods. Alfonso et al. (2015) show that the market sees the core (recurring) earnings as a more reliable indicator of future profitability than normal earnings including non-recurring items. Thus, firms engaging in classification shifting are overvalued by the market. This means that classification shifting has negative consequences to users of financial statements as it is not easily detected and causes an inefficient allocation of resources. This is confirmed by Alfonso et al. (2015)
their findings showing firms that engage in classification shifting are associated with negative future profitability.
2.3 Reasons for earnings management
In a large scale survey and interview based research Graham et al. (2005) find that eighty percent of the corporate financial officers they interviewed admit they would engage in some form of earnings management in order to meet the earnings expectations. Furthermore, fifty five percent say they would even go as far as avoiding initiating a very positive net present value project if this means they meet the earnings expectations. Indicating that firms are willing to sacrifice real long-term economic value to meet short-term earnings expectations. Graham et al. (2005) their findings thus show that the practice of earnings management is widespread and that a significant amount of the surveyed CFO’s admits they would engage in earnings management to meet or beat earnings expectations. In this section two incentives for earnings management are described in more detail: earnings management to increase future firm performance and earnings management to personal compensation
2.3.1 Earnings management to increase future performance
Firstly, Bartov, Givoly, & Hayn (2002) study the effects of firms meeting earnings expectations and their future performance. In their research they find that meeting or marginally beating earnings expectations is a leading indicator for future performance and increases the value of the firm. Moreover Bartov et al. (2002) find that the use of earnings management as a way of meeting or beating earnings expectations only marginally affects this relationship. Hence, firms can stand to profit from engaging in earnings management. This is further evidenced by Kasznik & McNichols (2002) who find that firms meeting or beating earnings expectations show significantly higher annual returns than firms that do not meat or beat earnings expectations. They also find that it is important that firms consistently meet or beat earnings expectations as missing the earnings expectations creates uncertainty about the firms (future) profitability and lower the stock price. Hence this increases the incentive for firms to engage in earnings management as not to break a trend of meeting or beating the earnings expectations. Following these findings firms can engage in earnings management simply because it yields positive effects for the firm in higher future performance and current
stock prices. On the other hand Cohen, Dey, & Lys (2008) argue that since the introduction of SOX the costs of engaging in AEM have increased. Hence it is uncertain if the findings of Bartov et al. (2002) and Kasznik & McNichols (2002) still hold.
Secondly, engaging in REM can be potentially costly. For example REM by reducing R&D costs can result in significant lower future profitability and REM by not initiating a very positive NPV project as fifty five percent of respondents said they would do according to Graham et al. (2005) could result in high opportunity costs. The research field on whether REM specifically increases future performance is divided. Gunny (2010) for example argues that REM can have a positive effect on future earnings of a company, while Bhojraj, Hribar, Picconi, & McInnis (2009) find that firms engaging in REM have a worse performance in the next three years than firms that did not engage in REM and missed the earnings expectations. As for classification shifting there exists to my best knowledge no information regarding the future profitability of firms engaging in it. However McVay (2006) argues that it is less affected by increased auditor scrutiny and that it does not hold any real cost like REM does. Therefore, the argument can be made that this form of earnings management is a potentially powerful tool in meeting and beating earnings expectations. Evidence for this in a US setting is provided by Barua et al. (2010). Their findings show that classification shifting in US firms is more pronounced for those firms that just met or beat the earnings expectations.
Summarizing, one incentive for earnings management is to meet or beat earnings expectations and by doing so potentially increase future profitability of the firm. Evidence for this is provided by Bartov et al. (2002) and Kasznik & McNichols (2002) but given the introduction of SOX in 2002 it is uncertain if their results still hold for AEM. As for REM the research is divided (Bhojraj et al., 2009; Gunny, 2010). And finally, for classification shifting there is to my knowledge no specific evidence that it increases future profitability. However, Barua et al. (2010) do find that firms often employ this method if they just meet or beat earnings expectations. This in turn can lead to firms having better future performance (Bartov et al., 2002; Kasnik & McNichols, 2002).
2.3.2 Earnings management and CEO compensation
Apart from using earnings management to increase future firm performance, there also exists extant research on CEO’s and other top executives engaging in earnings management in order to maximize their own compensation (Efendi, Srivastava, & Swanson, 2007). CEOs
engaging in earnings management is a direct consequence of the agency problem. Eisenhardt (1989) states that aligning the interests of the agent and the principal is the best solution to mitigate agency costs. This is often accomplished by making the CEO a shareholder by (restricted) stock-(option)plans and thus aligning the CEO goal of maximizing compensation with the shareholder goal of maximizing shareholder value. For this reason stock(option) plans are widespread with Sanders & Hambrick (2007) showing that in the year 2001 fifty percent of the CEO compensation package consisted of stock(option) compensation. However, regarding the effectiveness in mitigating opportunistic behaviour the research field is divided (Armstrong, Jagolinzer, & Larcker, 2010).
Although in theory stock(option) compensation should mitigate opportunistic behaviour by managers, information asymmetry still exists and can be utilized to opportunistically steer earnings. Depending on the period the CEO stock or stock-options are restricted, the CEO can opportunistically steer earnings during his tenure to increase the value of his portfolio and cash them in once they have vested. For example Coffee Jr. (2003) states that top executive and CEO stock(option) compensation was a major reason why Enron, a former energy giant in the USA, went bankrupt after severe earnings management and fraud. Coffee Jr. (2003) particularly states that this was because the CEO and other top executives could cash in their stock(options) before the market knew of the what was going on. Fuller & Jensen (2002) share this view of the negative effects of stock(options). They argue that the increased use of stock(options) in compensation packages led to increased preservation and increasing of short-term stock prices. Furthermore they argue that it became a personal as well as damaging priority for many CEO’s and CFO’s to maintain high shares prices to serve their managerial egos and they were reluctant to break the trend of quarter-over-quarter earnings increases Fuller & Jensen (2002, p. 42).
The research on the relation between CEO stock-(option)compensation and earnings management is divided on whether there exists a positive relation between the two (Armstrong, Jagolinzer, & Larcker, 2010). Several studies such as Burns & Kedia (2006), Efendi et al. (2007), Bergstresser & Philippon (2006), and Johnson, Ryan, & Tian (2009) find evidence that stock(options) and earnings management in the form of (severe) restatements and abnormal accruals. However, other researchers such as Erickson, Hanlon, & Maydew (2006) and Armstrong et al. (2010) find no relation between stock(options) held by CEOs and earnings management. The overall notion however is that unrestricted stock and options, which can be exercised at any time, are related to restatements and abnormal accruals. This is in line with the opinions of the CFO’s in Graham et al. (2005) his study who see the way of compensation as an important contributing factor to myopic management behaviour such as earnings management.
2.4 Accounting Standards
According to Jensen & Meckling (1976) the divide between ownership and management creates an agency problem. Due to the information asymmetry between the shareholder and management the latter can engage in opportunistic behaviour to increase their personal wealth at the expense of the shareholder (Fama & Jensen, 1983). One way to decrease the information asymmetry is by heavily restricting the allowed accounting options available to managers and minimalize the management from using judgement in their financial reporting. This is the line of thought underlying US GAAP and other rule-based approaches. Another approach is a more flexible one under which managers are granted the opportunity to exercise judgement and choose from a greater array of accounting methods to best represent their firms’ performance (Jeanjean & Stolowy, 2006). This is the idea behind the principle-based approach characterizing the International Financial Reporting Standards (IFRS).
2.4.1 IFRS
The IFRS are a set of accounting and reporting guidelines issued by the issued by the International Accounting Standards Board (IASB). IFRS also include the International Accounting Standards (IAS) issued by the International Accounting Standards Committee (IASC) which were both replaced with IFRS and the IASB respectively. Thus, the current IFRS standards includes both the IAS and IFRS standards that together offer a global reporting framework. The aims of IFRS are to increase reporting quality and comparability by offering a broad set of standards that aid to converge the broad array of local generally accepted accounting standards (GAAPs) used around the globe. Apart from the EU-members other notable countries that have adopted IFRS are Australia, Russia, and Turkey as well as multiple East-Asian and West-Latin-American countries2. This shows that IFRS is a broadly supported
set of reporting guidelines which the ability of increasing reporting quality and comparability will only increase the more countries adopt the standards (Behn et al., 2013).
In 2005 IFRS was officially adopted by the European Union3 and every EU company
that has its debt or equity listed at a public exchange is required to comply with IFRS4. This
means that no longer every firm in the EU can report using the local GAAP in the country where they are active. The aim of IFRS adoption is that it leads to greater comparability between similar firms in different countries within the EU and overall increased reporting quality. Not all listed companies however need to comply with IFRS. Firms listed on very small indexes for example are still allowed to use their local GAAP for reporting purposes.
It is unclear if IFRS has succeeded in increasing reporting quality in the EU. On the one hand IFRS can plug holes in existing local GAAPS by providing guidance on issues that were not included in the local guidelines (Callao & Jarne, 2010). On the other hand Jeanjean & Stolowy (2006) find that first time adopters such as Australia and the United Kingdom did not show a decrease in earnings management. Moreover, they find that in France earnings management in fact increased after IFRS adoption. Callao & Jarne (2010) strengthen these findings by providing evidence that IFRS adoption increased earnings management for their sample of non-financial companies listed at 11 large EU stock markets. Both researches however focus on AEM. Daniel A Cohen et al. (2008) find that stricter regulation, in their case the Sarbanes Oxley Act, leads to decreased AEM but higher REM. Hence the above studies might fail to take into account the full effect of IFRS on earnings management as the increase in AEM might be offset by a decrease in REM. Zang (2012) argues that REM might be potentially even more harmful than AEM, due to it destroying real economic value. Hence, firms using AEM instead of REM might be beneficial for shareholders as although they are still misled it does not destroy real economic value.
When it comes to classification shifting the effects of IFRS are largely unknown. However, while Athanasakou et al. (2008) failed to find strong evidence for classification shifting in the UK prior to IFRS adoption. The study by Zalata & Roberts (2016) find significant evidence for firms engaging in classification shifting in the UK. In their study they also note that one of the possible reasons is the adoption of IFRS granting firms more room to engage in classification shifting.
Under IFRS the standard relevant for classifications shifting is IAS 1 “Presentation of the Financial Statements”. This standard gives guidance on the presentation of the statement of financial position (balance sheet), statement of profit or loss and other comprehensive
3
https://europa.eu/info/publications/regulation-application-international-accounting-standards-ias_en
4
https://www.ifrs.org/use-around-the-world/use-of-ifrs-standards-by-jurisdiction/european-union/#commitment
income (income statement and OCI), statement of changes in equity, cash flow statement, and the notes supporting the statements. For my study on classification shifting only the income statement is relevant. Although shifting revenues and expenses between the income statement and OCI is also a practiced form of earnings management, however decisions on recycling between the income statement and OCI is more of a timing issue and falls in the scope of AEM. IAS1 requires that the total income and expense that was recognized in a period should be in the income statement. However, the standard leaves room for judgement as how to present these income and expenses. Although the reporting of extraordinary or special items is not allowed5 unlike under US GAAP managers under IFRS can still make
decisions on what items to include in the recurring and non-recurring sections of the income statement. This allows them to steer the reported recurring earnings by including or excluding certain non-recurring profits or losses.
Furthermore, they are free to use different subtotals in the income statement. Although the aim of IFRS is that this freedom serves to allow managers to better reflect their firm performance and inform investors, this also creates the opportunity to use certain adjusted profit figures or cherry pick income measures that show the strongest results. For example, KLM-AirFrance6 in their 2017 annual financial report include among others: EBITDAR
(Earnings before Interest Tax Depreciation Amortization, and Rent), EBITDA (Earnings before Interest Tax Depreciation, and Amortization), Income from current operations and Income from operating activities. On one hand one can argue that this better informs stakeholders. However, the used earnings definitions can also be seen as trying to influence certain stakeholders. By showing EBITDAR they decrease the Total Expenses line presented above by 1’939 euro while the rent expenses should likely be included. Furthermore, Income from current operations and income from operating activities are presented as separate accounts the former being 910 and the latter -939. The corresponding note emphasizes that the other non-current income and expenses that make up this difference to be a one-time special event7.
Thus, by using several income measures the company highlights higher persistent earnings in the form of income from current operations.
5
IFRS, IAS 1, paragraph 87
6
https://www.klm.com/corporate/en/publications/2017_Annual_Report.html
Page 96 Income Statement
2.4.2 US GAAP
Under US GAAP unlike IFRS firms do not have much discretion in dealing with non-recurring items. All non-recurring items need to be separately classified, presented, and disclosed based as either special or extraordinary items. A special item is either unusual or infrequent in nature or occurrence. And in case that an item is both unusual and infrequent it is classified as an extraordinary.
In 2015 the FASB (Financial Accounting Standards Board) issued new guidance that eliminates the concept of extraordinary events and transactions and aims to simplify the income statement presentation. This is a move towards a more aligned standard with IAS 1 discussed above, which also prohibits firms from presenting and disclosing extraordinary items.
2.5 Hypothesis Development
After McVay (2006) introduced classification shifting in the academic field a handful of other studies have taken an interest in this novel form of earnings management. From these studies the majority focuses on US firms reporting in compliance with US GAAP (McVay, 2006; Fan, Barua, Cready, & Thomas, 2010; Abernathy et al., 2014; Barua et al., 2010). These studies examine the use of special or discontinued items in the case of Barua et al. (2010) and find evidence consistent with these firms opportunistically misclassifying earnings to increase core profits. These studies are extended by Behn et al. (2013) who study classification shifting in a more global context by using a sample consisting of firms from forty countries.
Contrary to the above studies Athanasakou et al. (2008) who study classification shifting in the UK only find weak evidence of companies engaging in clasisifaction shifting. From their origional sample of 5,117 firm they only find evidence of classification shifting for a small sub-sample of about 3 percent of their entire sample. Zalata & Roberts (2016) who also study classification shifting in a UK setting however find significant evidence for their entire sample. One of the possible reasons they cite for this is the change of the UK from UK GAAP for IFRS for publicly listed firms.
IFRS and more specifically IAS 1 “Presentation of Financial Statements” differs from US GAAP in how it deals with non-recurring items. For example the special items used for
classification shifting in McVay (2006) her study are not allowed to be reported under IFRS. Instead under IFRS management is required to make a clear distinction between what constitute core- and non-core earnings. Because according to Alfonso et al. (2015) the effect of classification shifting depends on the manner analysts perceive reported recurring earnings, the manner in which these are reported due to the accounting standards can influence effectiveness and thus attractiveness of engaging in classification shifting.
Furthermore, apart from different standards there exist also different enforcement regimes and levels of investor protection between the US and the EU. In the US regulations are structured according to common law, this results to a higher level of investor protection and significantly higher fines than under codified law that (apart from the United Kingdom) forms the basis of the legal underpinnings of the EU countries (Athanasakou et al., 2008). This can mean that the costs of engaging management when discovered are lower in EU countries. This all taken together means that EU firms face lower legal costs while having more room to engage in classification shifting. Therefore, I expect European firms reporting under IFRS to engage in more classification shifting. In order to test this hypothesis, I will first test whether both US and EU firms engage in classification shifting, and then analyze whether EU firms do so to a greater degree. This yields the following hypotheses:
H1a: European Union firms reporting under IFRS engage in classification shifting H1b: US firms reporting under US GAAP engage in classification shifting
H2: European Union firms reporting under IFRS are more likely to engage in classification shifting than US firms reporting under US GAAP.
3 Methodology
The aim of this thesis is to examine whether EU firms use classification shifting as an earnings management tool to a higher degree than US firms. I test this by using two samples, one consisting of US firms reporting under US GAAP and one consisting of EU firms that report under IFRS. For both samples I use a modified version of the McVay model for determine the degree of classification shifting in both the US and EU. This model introduced by McVay in 2006 is used by every study looking at classification shifting in both US and international settings (Abernathy et al, 2014; Behn et al., 2013). The remainder of this section is structured as follows: in the next paragraph I first introduce McVay her model to tests for classification shifting and how it is modified to be applicable in an international setting. Thereafter, I explain how I test whether EU firms significantly shift more earnings than US firms. Lastly, I go into detail on the sample and data collection process.
3.1 The Modified McVay Model
To test whether EU firms engage in classification shifting I adopt a modified version of the 2-stage OLS regression model introduced by McVay (2006). With this model I first estimate the level of core earnings (CE) based on prior year core earnings, asset turnover (ATO), current and prior year accruals (ACCRUALS), the change in sales between the current and last year in percentage points (ΔSALES) and finally an indicator for negative sales equaling 0 if sales are positive and the sales change if sales are negative (NEG_ΔSALES). Furthermore, by using the reported core earnings as the dependent variable, I can compute the coefficients and use these to calculate the expected core earnings. To compute the reported core earnings I adopt Behn et al. (2013) their measure to calculate the core earnings. In their research they take a global approach to earnings management by classification shifting offering a definition of core earnings that can be used to calculate in both IFRS and US GAAP settings8. Behn et
al. (2013) derive core earnings as total sales deducted by operating expenses and scaled by sales. Using this definition allows me to compute the core earnings in a comparable manner for both the sample of US and EU firms. The equation for the first stage is given in equation 1
8
Although Behn et al. (2013) study classification shifting in an global setting they focus on the role
of internal governance on classification shifting instead of on the pervasiveness of classification
shifting in the EU versus the US.
below. In the following paragraphs the inclusion of each variable is explained in more detail. All variables used in this thesis are defined in appendix A.
First-Stage: Levels Model
(1) 𝐸(𝐶𝐸)𝑡 = 𝛽0+ 𝛽1𝐶𝐸𝑡−1+ 𝛽2𝐴𝑇𝑂𝑡+ 𝛽3𝐴𝐶𝐶𝑅𝑈𝐴𝐿𝑆𝑡−1+ 𝛽4𝐴𝐶𝐶𝑅𝑈𝐴𝐿𝑆𝑡+ 𝛽5∆𝑆𝐴𝐿𝐸𝑆𝑡
+ 𝛽6𝑁𝐸𝐺_∆𝑆𝐴𝐿𝐸𝑆𝑡+ 𝑒𝑡
Firstly, I include last year core earnings as they are assumed to be persistent over time9. The asset turnover ratio (ATO) is included because research has shown that it is
inversely related to the profit margin10 and the core earnings used in this model closely parallel
profit margin (Nissim & Penman, 2001; McVay, 2006). Furthermore, ATO captures the effects of firms chancing their operating strategy leading to a different asset turnover ratio and affecting core earnings. Lagged accruals (ACCRUALSt-1) are added as Sloan (1996) finds
that the accrual component of current earnings helps predict the future earnings. This effect is found to be even stronger when there is no AEM. Furthermore, current year accruals (ACCRUALSt) are added to control for extreme performance as DeAngelo, DeAngelo, &
Skinner (1994) find that extreme performance is highly correlated to changes in current year accruals. The change in sales (∆SALES) is included as McVay (2006) states that fixed costs as a percentage of sales decreases when sales increase. Therefore, the relationship between core earnings and the change in sales is expected not to be constant. The growth in sales variable acts as an explanatory variable for this effect. Also, a dummy variable for an increase or decrease in sales is added (NEG_ ∆SALES). This variable is equal to 0 if there is a positive change in sales and is equal to the change in sales (∆SALES) when sales are negative. The reason for adding this variable is to allow for different slopes for decreasing and increasing sales. This is done because Anderson, Banker, & Janakiraman (2003) find costs to be sticky relative to changes in sales. They document that per 1 percentage point of a sales increase selling, general, and administrative costs rise by 0.55 percent while for a percentage point decrease in sale these costs only decrease by 0.35 percent.
Once the expected core earnings are calculated the unexpected core earnings can be derived by deducting the estimated core earnings from the reported core earnings as seen in equation 2 below. The unexpected core earnings (UE_CE) are then used in the second stage as the dependent variable as seen below in equation 3. This enables the estimation of the coefficient of the recurring items (NRI). This coefficient explains to what degree the
9
Highly significant correlation between CE and CE lagged of 0,816 visible in the correlation table.
10Significant and negative correlation can be seen between ATO and CE in the correlation table.
recurring items drive unexpected core earnings. I expect NRI to be positive and significant when firms engage in classification shifting, as the difference between reported and estimated core earnings is then explained by an increase in non-recurring items.
(2) 𝑈𝐸_𝐶𝐸𝑡 = 𝐶𝐸𝑖,𝑡− 𝐸(𝐶𝐸)𝑖,𝑡
Second-Stage: Levels Model
(3) 𝑈𝐸_𝐶𝐸𝑡 = 𝛽0+ 𝛽1𝑁𝑅𝐼𝑡+ 𝛽2𝑆𝐼𝑍𝐸 + 𝛽3𝐿𝐸𝑉 + 𝛽4𝐶𝐹𝑂 + 𝛽5𝑅𝑂𝐴 + 𝑒𝑡
The use of non-recurring items differs from McVay (2006) her original model in which she only examines special items reported under US GAAP. However, special items are not allowed to be reported in the income statement under IFRS. Therefore, instead of using special items I look at classification shifting using non-recurring items. Under IFRS companies have to make a clear distinction between recurring and non-recurring earnings and this leaves the opportunity to opportunistically shift income and expenses. Zalata & Roberts (2016) used a similar approach in their study examining the relationship between internal corporate governance and classification shifting in the UK, a country reporting under IFRS as adopted by the EU. In line with Zalata & Roberts (2016) I calculate the non-recurring items as the difference between reported core earnings and bottom line net income (net income – core earnings). This means that if firms are using classification shifting to increase their core earnings the non-recurring items are negative and decrease the bottom-line net income. Another difference with McVay is that she did not include any control variables arguing that the control variables included in the first stage suffice. Based on Barua et al. (2010) and Behn et al. (2013). The control variables aim to capture the effect of the size (SIZE), capital structure as measured in debt to equity ratio (LEV) and firm performance by looking at both the cash flow from operations and the return on assets. Finally, like Behn et al. (2013) I scale all main variables that are not fractions themselves by sales. This also means that I exclude firms with less that 1 million in sales from the sample, as dividing by earnings between 0 and 1 for scaling greatly inflates the scaled variable and distorts the sample. The precise calculation of all variables is shown in appendix A.
The above discussed model to detect classification shifting only detects unexpected core earnings and the degree to which these are driven by non-recurring items. However, as McVay (2006) notes positive unexpected earnings do not necessarily indicate that a firm engaged in classification shifting. Positive unexpected core earnings can also result from
better than expected performance for example due to successful restructuring of a company and increasing long term performance (McVay, 2006). To differentiate between real economic drivers of positive unexpected core earnings and firms engaging in classification shifting to achieve positive unexpected core earnings McVay extends the first 2-stage levels model with a 2-stage changes model. In the first-stage the change in core earnings for the current year is estimated using equation 4.
First-stage: Changes model
(4) ∆𝐶𝐸𝑡 = 𝛽0+ 𝛽1𝐶𝐸𝑡−1+ 𝛽2∆𝐶𝐸𝑡−1+ 𝛽3∆𝐴𝑇𝑂𝑡+ 𝛽4𝐴𝐶𝐶𝑅𝑈𝐴𝐿𝑆𝑡−1+ 𝛽5𝐴𝐶𝐶𝑅𝑈𝐴𝐿𝑆𝑡
+ 𝛽6∆𝑆𝐴𝐿𝐸𝑆𝑡+ 𝛽7𝑁𝐸𝐺∆𝑆𝐴𝐿𝐸𝑆𝑡+ 𝑒𝑡
The variables used to calculate the change in core earnings are similar to the variables used in the levels model. However, there are three major change. Firstly, the changes model includes the lagged change in core earnings (ΔCEt-1) which measures the difference between
the core earnings from year t-2 and t-1. This variable is added because in case of classification shifting core earnings are expected not to be sustainable. Therefore, positive prior year changes in core earnings are expected to have a negative effect on the current year change in core earnings. Secondly, the change in asset turnover (ΔATOt) is included instead of the
asset turnover (ATOt) that is included in the levels model. The change in asset turnover
controls for a change in firm performance that can explain the change in core earnings. A positive change in asset turnover is expected to increase core earnings. The remaining independent variables: accruals both current and lagged, as well as the change in sales and negative change in sales are the same as in the stage levels model. Just as with the first-stage levels model, the first-first-stage changes model is used to derive the expected change in core earnings, which can then be used to calculate the unexpected change in core earnings using equation 5 shown below.
(5) 𝑈𝐸_∆𝐶𝐸𝑡= ∆𝐶𝐸𝑖,𝑡− 𝐸(∆𝐶𝐸)𝑖,𝑡
With the unexpected change in core earnings the coefficients of the second-stage of the changes model can be estimated using equation 6 shown below. The second-stage changes model looks at the relation between current year non-recurring items and the change in core earnings in the following year. Therefore, if a firm uses classification shifting to increase core earnings resulting in positive unexpected core earnings in the year t, it is expected that this level of unexpected core earnings is not sustainable in the year t+1 resulting in a negative change in core earnings in that year. The reason for this is that the current year unexpected
core earnings were not based on an actual improvement in real economic performance, and therefore not sustainable. Just as with the second-stage of the levels model McVay (2006) does not include control variables arguing that the first-stage model already controls for other effects. Based on Barua et al. (2010) and Behn et al. (2013) I include control variables for the size, capital structure, and performance of the firm as factor besides non-recurring items that drive the change in core earnings.
Second-stage: Changes model
(6) 𝑈𝐸_∆𝐶𝐸𝑡+1= 𝛽0+ 𝛽1𝑁𝑅𝐼𝑡+ 𝛽2𝑆𝐼𝑍𝐸 + 𝛽3𝐿𝐸𝑉 + 𝛽4𝐶𝐹𝑂 + 𝛽5𝑅𝑂𝐴 + 𝑒𝑡
3.2 Comparing Classification Shifting
To test whether EU firms engage in more classification shifting than US firms I adopt the modified McVay model and add in a dummy variable called EU_DUMMY which takes the form of 0 for US firms and 1 for EU firms. Furthermore, I add the mean-centered non-recurring items (NRI_MC) as well as the four control variables for firm size, capital structure and performance. I run three regressions for: total NRI, Positive NRI, and Negative NRI. I expect classification shifting to be most pronounced for positive non-recurring items, as these items increase the operating income. The coefficient of interest if 𝛽3, if this coefficient is positive and
significant then that means that the EU sample has a positive effect on the level of unexpected core earning compared to the reference group. Meaning that firms in the EU engage in more classification shifting than firms in the US. I test this model with the predicted values for UE_CEt calculated during the testing for classification shifting on the separate samples of EU
and US firms.
(7) 𝑈𝐸_𝐶𝐸𝑡 = 𝛽0+ 𝛽1𝑁𝑅𝐼𝑡+ 𝛽2𝐸𝑈𝑑𝑢𝑚𝑚𝑦 + 𝛽3𝑁𝑅𝐼_𝑀𝐶 ∗ 𝐸𝑈𝑑𝑢𝑚𝑚𝑦 + 𝛽4𝑆𝐼𝑍𝐸 + 𝛽5𝐿𝐸𝑉
+ 𝛽6𝐶𝐹𝑂 + 𝛽7𝑅𝑂𝐴 + 𝑒𝑡
For the main test I collect my data from the Compustat Capital IQ databases. My sample of EU firms is derived from the Global file, and the sample of US firms is derived from the North-American file. I exclude financial service firms in line with prior literature and use the accounting standard (ACCTSTD) mnemonic to filter out companies that do not report in compliance with US GAAP for the US sample and IFRS for the EU sample. Secondly, using the stock exchange code (EXCHG) mnemonic I remove firms not listed on major exchanges and firms listed on foreign exchanges that were not filtered out with the country data selection. The reason is that IFRS is only required for listed firms in the EU that are listed on a regulated exchange. And foreign exchanges may require different compliance and reporting requirements that are outside the scope of this study. Furthermore, I remove firms that have less than one million in revenue as I use this metric to scale all other variables that are not fractions or percentages themselves in line with (Behn et al., 2013). Including sales below 1 million would greatly inflate those variables. Lastly, I remove firms that lack the necessary data to compute the variables necessary to test for my main hypothesis. The final sample consists of 2,627 observations for the US sample and 16,215 observations for the EU sample. Because I use multiple linear regression which is a technique sensitive to outliers I winsorize all data at the 1st and 99th percentile for my research, this is in line with prior research (Behn et al., 2013;
Zalata & Roberts, 2016). Data from countries that use a different currency than the euro is translated to euros based on the foreign exchange rate at the end of the fiscal reporting period using historic exchange rate obtained from the European Central Bank.11
11
https://www.ecb.europa.eu/stats/policy_and_exchange_rates/euro_reference_exchange_rates
/html/index.en.html
Although country-level effects can influence the pervasiveness in classification shifting between different EU countries. I focus in my research solely on the difference between the pervasiveness of classification shifting between US firms reporting under US GAAP and EU firms reporting under IFRS. The countries included in the EU sample are shown in table 2 below.
4 Descriptive Statistics
In this chapter I first describe the results rom the descriptive statistics and Pearson correlation table both presented in table 3 and 4 respectively. Then in the second section I test whether companies in both the US and EU sample show evidence of classification shifting using both the modified two-stage levels and changes models from McVay (2006). Then in the third part I compare the two samples to examine whether classification shifting is more pervasive among EU firms.
4.1 Descriptive Statistics
Table 3 below shows the descriptive statistics for the main variables split up in two panels. Panel A shows the descriptive statistics for the United States sample and panel B for the EU sample. I included the 1st and 99th percentile to show the minimum and maximum values that the data is winsorized at for every main variable14. For the US sample the mean
core earnings (CE) of 0,098 are close to the mean core earnings of 0,070 that McVay (2006) finds. Furthermore, the unexpected core earnings have a (rounded) mean of zero, this is in line with other research showing unexpected core earnings to be zero or near zero. For example, Zalata & Roberts (2016) studying classification shifting in the United Kingdom find a mean value for unexpected core earnings of 0,004 and McVay (2006) finds a mean of 0,001.
The change in future unexpected core earnings (UE_CEt+1) is -0,018 for the US sample, consistent with the idea that unexpected changes in core earnings are not sustainable and could indicate classification shifting. For the EU sample the future unexpected core earnings are 0,003, thus marginally positive. This does however not imply that there exists no classification shifting, a few large firms showing strong consistent increases in unexpected future core earnings can increase the mean to a positive value. This is further evidenced by McVay (2006) who finds significant evidence for classification shifting in an US setting, even though the mean of the future change in unexpected earnings is 0,001, thus positive, in her research. Furthermore, both the median value of the future change in unexpected core earnings in the EU sample and in the paper by McVay is negative, adding to the evidence that a few strong performing large firms increase the mean value to a positive one.
Lastly, non-recurring items (NRI) in both models is positive. As the non-recurring items are calculated as reported core earnings deducted by bottom line net income and scaled by sales this means that the average of non-recurring items is income decreasing. The average value of non-recurring items is higher than Zalata & Roberts (2016) who finds an average value of 0,061. Furthermore, the average NRI for the US sample is twice as large as the
average for the EU sample, 0,274 versus 0,125 respectively. This could indicate that US firms have larger non-recurring items and thus engage in more classification shifting. However as described above, the non-recurring earnings can be part of real-economic value generation and are only considered classification shifting if they lead to unsustainable unexpected core earnings. Given the unexpected future change in core earnings (UE_CEt+1) is positive it is
unlikely that the higher value of positive non-recurring items is linked to more earnings management in the US. The control variables show similar values to those of Behn et al. (2013) and Barua et al. (2010) on which they are based. With both current and lagged accruals (ACCRUALS) having small negative values.
In table 4 the Pearson correlation tables for both the US and EU sample are presented. From this table it is apparent the current year core earnings (CEt) is largely predicted by the
last year core earnings (CEt-1) showing that core earnings are persistent. Secondly, both
non-recurring earnings (NRI) and the positive non-non-recurring earnings (Positive NRI) have a negative relation with core earnings while negative non-recurring items (Negative NRI) have a positive relation with current year core earnings. This is in line with the finding of McVay whether positive special items have a negative relation with core earnings. The same effect exists between the negative- and positive NRI and the unexpected core earnings, in both samples negative NRI has a positive effect on unexpected core earning, while positive non-recurring items have a negative relation with unexpected core earnings. Lastly, the unexpected core earnings (UE_CEt) in year t have a negative relationship with the future
change in unexpected earnings (UE_ΔCEt+1), evidencing that unexpected core earnings are
generally not sustainable and can possibly be attributed to classification shifting. This is in line with Athanasakou et al. (2008) their research that finds a negative correlation between both variables as well.
5 Main Test Results
The main objective in this thesis is to examine whether EU firms reporting under IFRS engage in more classification shifting than US firms reporting under US GAAP. To test this, I first use the two different two-stage models McVay introduces, the levels and changes models, to determine whether there is evidence of classification shifting in both samples and support or reject hp. The first step is to test the level of unexpected core earnings in year t (UE_CEt)
driven by the non-recurring items in year t (NRIt).
5.1 Levels Model
To calculate the unexpected core earnings, I first use the first-stage regression given in equation 1 to estimate the expected core earnings. In line with McVay (2006) and Behn et al. (2013) who estimate the expected core earnings by industry and year, I firstly aimed to estimate the expected core earnings by country and year. However due to a lack of data the resulting regressions showed high levels of multicollinearity12. Therefore, I run the regression
on country level instead of country and year. The results of these first regressions are shown in panel A of table 4, where the reported numbers represent the coefficients and the significance is market as *,**,or *** for p-values smaller or equal than 0.1, 0.5, and 0.01 respectively. As expected the estimated current year core earnings are largely driven by the lagged core earnings. This also translates in a relatively high adjusted R2 as the persistence
in core earnings means that the prior year core earnings can accurately predict the current year core earnings. The coefficients of all included variables except for the asset turnover (ATO) are equal to their predicted sign. However, the asset turnover does show a negative effect when non-recurring items are positive, and thus income increasing. Hence, in the case of firms using classification shifting to increase core earnings, the affect from the asset turnover is as expected. Furthermore, the Durbin-Watson statistic is not significantly different that 2 therefore the results are not subject to autocorrelation.
(1) 𝐸(𝐶𝐸)𝑡 = 𝛽0+ 𝛽1𝐶𝐸𝑡−1+ 𝛽2𝐴𝑇𝑂𝑡+ 𝛽3𝐴𝐶𝐶𝑅𝑈𝐴𝐿𝑆𝑡−1+ 𝛽4𝐴𝐶𝐶𝑅𝑈𝐴𝐿𝑆𝑡+ 𝛽5∆𝑆𝐴𝐿𝐸𝑆𝑡
+ 𝛽6𝑁𝐸𝐺_∆𝑆𝐴𝐿𝐸𝑆𝑡+ 𝑒𝑡
12
Running regressions per country and year resulted in VIF levels far exceeding the commonly
accepted threshold of 4 leading to outcomes that are not robust.
With the expected core earnings derived from equation 1 I calculate the unexpected core earnings using equation 2, as reported core earnings deducted by the expected core earnings
(2) 𝑈𝐸_𝐶𝐸𝑡 = 𝐶𝐸𝑖,𝑡− 𝐸(𝐶𝐸)𝑖,𝑡
Second-Stage: Levels Model
(3) 𝑈𝐸_𝐶𝐸𝑡 = 𝛽0+ 𝛽1𝑁𝑅𝐼𝑡+ 𝛽2𝑆𝐼𝑍𝐸 + 𝛽3𝐿𝐸𝑉 + 𝛽4𝐶𝐹𝑂 + 𝛽5𝑅𝑂𝐴 + 𝑒𝑡
Then with the derived unexpected core earnings I calculate the coefficients of the second-stage levels model, the results are shown in panel B of table 4 presented below. Despite the model being significant there is no significant effect of non-recurring items (NRI) in the US sample and a significant non-effect in the EU sample for total non-recurring items. As for positive non-recurring items, and thus income increasing non-recurring items, the relationship with the level of unexpected core earnings is insignificant. Lastly, for negative NRI I find both a significant and positive effect on the level of unexpected earnings. Furthermore, the included control variables included to control for firm size, performance and capital structure are equal to their predicted signs yet only significant for the European sample.
5.2 Changes Model
The second step in determining whether firms engage in classification shifting is determining whether the unexpected core earnings are due to one-off events hinting at classification shifting or whether they are sustainable and based on real-economic circumstances. To do this I use the two-stage multiple linear regression changes model by McVay (2006) given in equations 4 to 6 below. Just as with the two-stage levels model I first
estimate the change in core earnings using equation 4. Then I deduct the estimated change in core earnings from the reported change in core earnings to derive the unexpected change in core earnings for the year t using equation 5. Then because the aim is to predict the future change in core earnings I regress equation 6 using the next year change in core earnings as the dependent variable (UE_ΔCEt+1). The equations used to estimate the change in core
earnings and the unexpected change in core earnings are given below by equation 4 and 5. The results are presented in table 5.
First-stage: Changes model
(4) ∆𝐶𝐸𝑡 = 𝛽0+ 𝛽1𝐶𝐸𝑡−1+ 𝛽2∆𝐶𝐸𝑡−1+ 𝛽3∆𝐴𝑇𝑂𝑡+ 𝛽4𝐴𝐶𝐶𝑅𝑈𝐴𝐿𝑆𝑡−1+ 𝛽5𝐴𝐶𝐶𝑅𝑈𝐴𝐿𝑆𝑡
+ 𝛽6∆𝑆𝐴𝐿𝐸𝑆𝑡+ 𝛽7𝑁𝐸𝐺∆𝑆𝐴𝐿𝐸𝑆𝑡+ 𝑒𝑡
(5) 𝑈𝐸_∆𝐶𝐸𝑡= ∆𝐶𝐸𝑖,𝑡− 𝐸(∆𝐶𝐸)𝑖,𝑡
Second-stage: Changes model
(6) 𝑈𝐸_∆𝐶𝐸𝑡+1= 𝛽0+ 𝛽1𝑁𝑅𝐼𝑡+ 𝛽2𝑆𝐼𝑍𝐸 + 𝛽3𝐿𝐸𝑉 + 𝛽4𝐶𝐹𝑂 + 𝛽5𝑅𝑂𝐴 + 𝑒𝑡
In panel A of table 5, both lagged core earnings (CEt-1) as well as the change in lagged
core earnings (ΔCEt-1) have a negative relation with the current ear core earnings (CEt). This
negative relation is explained by the fact that previous year positive core earnings and positive changes in core earnings are on average not sustainable and thus decrease the current year change in core earnings (McVay, 2006). The change in asset turnover (ΔATOt) is also
negatively related to the current year change in core earnings. This is different than I expected, as a positive change in asset turnover should be a signal of improved firm efficiency, thus result in higher core earnings. The variables are all in line with their predicted sign.
Panel B of table 5 shows the results of the regression on the unexpected change in future core earnings. The coefficient of interest in this model is the one of the non-recurring items (NRI). A negative coefficient means that positive, thus income increasing, current year non-recurring income is not sustainable in future years, and hence and indicator that firms
