The influence of a US military background on classification shifting practices:

(1)

The influence of a US military background on classification

shifting practices:

Study Program: MSc Accountancy Supervisor: S. Wang

Student: Fabian van der Schaaf – 3791327

Abstract

This paper examines the effect of a US military background on the application of classification shifting practices in the US. Classification shifting is a method of earning manipulation in which core expenses are shifted towards special items to trick investors as they focus mainly on core earnings. Recent research provided evidence that, based on the assumption that individual characteristics influence organizational outcomes, a military background leads to less fraudulent behavior. A military background is associated with core values of integrity, honesty, a high level of obedience for the law and respect for the profession and the corresponding values. These values are not in line with the application of classification shifting in which managers try to fool investors. However, the influence of a military identity and other personal characteristics on the application of classification shifting is yet unclear as classification shifting research emerged only quite recently after the research of McVay (2006). I expect that firms with a military CEO engage less in classification shifting practices as their military identity works as a constraint on earnings manipulation. Results confirm the expectation that CEOs with a military background engage less in classification shifting, even in high incentive situation. However, additional tests show that the time in which a CEO has served in the military influence the degree in which they adopt the military identity and, in turn, the degree in which they apply classification shifting. So, a military background does not necessarily work as a constraint on fraudulent behavior as results suggest that a military with less than ten service years engages in classification shifting.

(2)

Introduction

Recently the SEC announced in a press release that they fined Marvell Technology Group Ltd. for 5,5 million dollars for the engagement in earnings manipulation activities and thus violating the antifraud and reporting provisions of the federal security law (SEC, 2019). This sanction by the SEC is showing that that the manipulation of earnings to trick investors by presenting earnings which are not in line with the current economic situation of the organization is an ongoing problem and there is still a need for additional research for the determinants of earnings management despite all the previous, extensive research.

Until 2006 the focus of earnings management research was mainly on either accrual earnings management or real activities earnings management. The former focusing on a manager’s ability to influence earnings through the manipulation of accruals which leads to investors relying on the mangers’ judgement and discretion due to the subject character of accruals (Dechow & Dichev, 2002). The latter method focuses on the manipulation of real activities for instance to increase the production temporarily to be able to spread the overhead costs on a larger number of products and capitalize these costs in the inventory. Furthermore, firms can temporarily or artificial boost or decrease sales (Roychowdhury 2006). In 2006 McVay was the first to provide evidence of a third type of earnings management, namely, classification shifting. With this technique managers use their ability to shift core expenses vertically to special items which in turn will lead to a higher ‘core earnings’. This can be an important tool because investors and specifically analysts tend to focus on core earnings before the inclusion of special items (McVay 2006). As this third type of earnings management is only addressed quite recently, there has not been a lot of research into the different determinants and prohibitors of classification shifting such as the influence of personal characteristics.

To address the influence of personal characteristics it is first important how a person’s characteristics are shaped, Tafjel, Turner, Austin & Worchel (1979) found that an individuals’ their identity are shaped towards the identity of the group of which they think they belong. Wolfendale (2009) found that in the military, individuals have a clear distinction between insiders and outsiders and adopt the common identity of the group of ‘insiders’ they belong to. This common identity can be characterized by the following core values: (i) integrity; (ii) loyalty; (iii) honesty; (iv) discipline; and (v) a high level of moral duty towards their profession. Elder (1998) concluded that this common military identity is persistent when recruits leave the military and start working outside the military. As Hambrick & Mason (1984) provide evidence that personal characteristics and the background of a CEO influence the decisions they make. Therefore, it is expected that a military background will influence organizational outcomes as

(3)

serving in the military leads to a common identity and personal characteristics influence organisationa outcomes. This has been confirmed with the research of Koch & Wernicke (2018) and Benmelech & Frydman (2015) amongst others who concluded that CEOs with a military background behave more ethical and therefore are involvement in less fraudulent financial report and other fraudulent activities.

Thus, as classification shifting is a method to report earnings in a dishonest way to fool investors and act in a way which is perceived not integer because the firms achieve a high return through dishonest reporting. In contrast the military shapes the identity of a recruit towards a common military identity where integrity and honest are key. Therefore, it is expected that a CEO with a military background would be less likely to engage in classification shifting than a CEO without such a military background.

To test whether the hypothesized relationship is true I apply the method of McVay (2006) to determine the degree of unexpected core earnings and the delta unexpected core earnings in a firm year observation. The next step is to determine whether classification shifting is present in the full sample consisting of 5016 firm year observations from 2012 until 2018. When I regress unexpected core earnings on income decreasing special items, for the whole sample, it is expected that the coefficient of unexpected core earnings is positive while the coefficient of delta unexpected core earnings is negative as managers are expected to shift core expenses towards special items. So, when an increase of unexpected core earnings is positively related to special items it could be that this is due to the application of classification shifting. When this effect is indeed due to classification shifting a reversal in the next year is expected, as the core earnings are artificially inflated. When I include the variable of interest, Military CEO and control variables and generate interacting variables of all variables and income decreasing special items. The coefficient of Miltiary CEO * special items and unexpected core earnings is expected to be negative as I expect that CEOs with a military background do not shift core expenses towards special items.

When the full model is tested in which the influence of a CEO with a military background and a set of control variables is included in the model of McVay (2006). The coefficient of the variable of interest is negative and significant which indicates that a military CEO engages in less classification shifting practices which suggests that a military identity works as a constraint on earnings management due to their key values of integrity, honesty and loyalty. The results hold up when these CEOs with a military background, face high incentive situations such as; (i) a high portion (>5%) of income decreasing special items scaled by sales; (ii) a threshold which has just been met or beaten; or (iii) when these CEOs have low tenure

(4)

and thus are judged mainly on earnings by the labor market. Furthermore, when testing whether the time a CEO has served in the military influences the degree in which they apply classification shifting techniques I found that there is a significant difference between CEOs with more and less than ten service years relating to the degree in which they apply classification shifting. Results of the main test do not hold up for CEOs with less than 10 service years which indicates that it takes time to adopt to the military identity in such a manner that it will be beneficial in corporate life.

My study contributes to the literature in several ways. First of all, I add to the literature by building on the research of Koch-Bayram & Wernicke (2018), who found that a US military background can serve as an indicator of honest and integer reporting, in that I provide evidence that the time a CEO has served in the military influences the degree in which these characteristics have an influence in corporate life, and in particular, the degree in which they apply classification shifting practices. Furthermore, I contribute to classification shifting research as I show that managerial characteristics can have an influence on the degree in which CEOs apply classification shifting.

The remainder of this paper is as follows, firstly I am going to introduce the literature review and hypothesis after which I describe the data collection and the method I will apply. Finally, I am going the presents the results and additional tests before concluding.

Literature review and hypotheses

Military CEOs

The first theory which is important is the identification theory, Tajfel et al. (1979) state that interest in a group creates intergroup relations which causes identification with the group. Interest in a group does also create the so-called inter-group relations which further increases an individuals’ identification with the group to which they think they belong. Thus, if an individual identifies themselves as a member of a certain group, they are likely to be influenced and shaped towards the common group-identity. Wolfendale (2009) investigates these relations in a military setting and found that military personnel often have a clear distinction between insiders and outsiders, they clearly identify themselves as a member of the military group. Which leads to a high level of intergroup relationships through which the identity of the individual members is transformed towards a common military identity. Furthermore, Jackson, Thoemmes, Jonkmann, Lüdtke and Ulreich (2012) state that a military training changes the identity of a recruit towards the desired military identity.

(5)

So, the unique military setting increases the intergroup relations and shape an individuals’ identity towards a common identity. Next we need to determine what the important aspects of this military identity are. Integrity is one of the most important values of this military identity, which can be defined in one of the two following ways according to Oakley & Cocking (2001) – virtue ethics and professional roles - virtue ethics can assist in defining proper boundaries in the execution of their profession in which integrity can be about refusing to provide service or about practicing the profession in such a manner that it does not distinct from their personal values. Based on their research Wolfendale (2009) determined that the role of integrity of military personnel is guiding their behavior and serves as an ideal to which they can strive to. Besides integrity, loyalty and honor amongst others are important characteristics of the military profession. Furthermore, discipline is an important aspect of the military identity, it relates both to the punishment in case of violations and the definition of desired behavior. (Akerlof & Kranton, 2005). These characteristics have not changed a lot through the years while their environment has changed a lot through the years. Wolfendale (2009) concluded that the military has the moral duty in which they should refuse to provide service when it is in conflict with their personal values and guiding ideals of the profession. have a clear idea of how a member of their profession should behave, they should obey the rules of the organization and should put their service before themselves. All these aspects of the military identity are promoted in the training through for instance, imitation, boot camps and sometimes severe training exercises in which the individuals are shaped. Thus, not only is the identity shaped through training it also causes for an increase of social cohesion, which in turn will lead to an incremental adoption of the group identity (King, 2006; Franke, 1998).

The next question arising is whether the key values which represent the military identity are constant over the years and if these characteristics are persistent. After the cold war the army changed its focus from solely on conventional combat situations towards more non-conventional, non-combat situations. However, the military values such as dignity, integrity, morality and subordination remain the core of the military identity (Franke 1998). Elder (1986) states that in the period after the wars in which individuals experienced the war themselves or had relatives who experienced the period of war the military proved to be a turning point in their lives which turned their disadvantages into opportunities and eventually advantages. These transitions have such severe implications that these effects will be persistent up and until later in life (Elder, 1998). This is later confirmed with the research of Jackson, et al. (2012) who find that the military changes a man and that these effects are persistent when a recruit leaves the military.

(6)

After concluding that military recruits adopt a common military identity during their training and that these effects are persistent when a recruit leaves the military it is important to determine if and how these characteristic influences corporate decision making. Hambrick & Mason (1984) were the first to examine the possible influence of personal characteristics on organizational outcomes. They state that the organizational outcomes could be seen as the reflection of personal values of CEOs and other high influential executives, providing evidence that these values can have a significant impact on organization. Their upper echelon theory states that these values are shaped by the different demographic characteristics such as age, education of functional track and that these can be used to predict the strategical actions of the CEOs and eventually performance of their organizations. According to Hambrick & Finkelstein (1987) this influence will be even greater when there are little constraints on managerial discretion from the environment in which they operate. However, Betrand & Schoar (2003) found that a high level of heterogeneity across CEOs and other high influential executives had an influence on the different policies of their organizations while previous research concluded that this heterogeneity did not had a significant impact due the high level of constraints in this area. Furthermore, Borchet & Welch (2011) state that the professional background of top executives has a significant impact on the financial reporting choices. Finally, there is extensive research about the influence of CEO values on the different earnings manipulation decisions. (Plöckeringer, Aschauer, Hiebl & Roahtschek, 2016). Thus, the personal values of CEOs and other high influential executives have an influence on organizational outcomes with both a high and low level of constraints.

A new alley of research build upon the research of Jackson et al. (2012) who investigated the direct effect of military experience and specifically, the effects of military training on the development of personal traits which in turn can have an influence on organizational outcomes. As members of the military are expected to behave in line with the common identity and that the effect will be persistent after leaving the military it is likely that such a background will have an impact on corporate outcomes according to the upper echelon perspective. Research has shown that executives with a military background are less likely to engage in tax avoidance behavior (Law & Mills, 2017); are less likely to be involved in fraudulent financial reporting (Koch-Bayram & Wernicke, 2018) and behave more ethical so they are less likely to engage in fraudulent activities (Benmelech & Frydman, 2015).

(7)

Earnings manipulation

Healy & Wahlen (1999) state that financial reporting adds value if it enables financial statements to portray differences in both the financial performance and financial position over the reporting period. They define earnings manipulation as follows: “Earnings management occurs when managers use judgment in financial reporting and in structuring transactions to alter financial reports to either mislead some stakeholders about the underlying economic performance of the company or to influence contractual outcomes that depend on reported accounting numbers.” Thus, earnings management is about CEOs or other high influential managers using their managerial judgement and flexibility provided by the different accounting policies and/or standards to mislead stakeholders about the true economic performance of the organization.

There are different techniques which can be used for the manipulation of earnings, broadly the following three types can be identified: Real activities earnings management, accrual earnings management and classification shifting in which the first two already have been researched extensively while the latter one has been addressed relatively recent and therefore is quite unexplored.

(i) real activities earnings management (REM) - executives use techniques to temporarily boost earnings or decrease sales. Some of these techniques are to increase discounts to boost sales or to produce more good than can be sold to spread the overhead costs over more units and reduce the costs of goods sold and capitalize these costs (Roychowdhury 2006). However, as the power of the key executives which have to monitor the CEOs and their time to retirement increases the degree of real earnings manipulation behavior is likely to decrease according to Cheng, Lee & Shevlin (2016). In addition, the use of real activities-based earnings management leads to an increase of the firm’s risk premium which is incremental to the increase resulting from the other types of earnings management (Kim & Sohn, 2013).

(ii) accrual earnings management (AEM) - an accrual can be used to spread both costs and earnings over a longer period, however this requires some estimates and therefore are somewhat subjective. So, due to their subjective character accruals can also be used to deliberately manage earnings both downward and upwards (Dechow & Dichev, 2002). However, accrual quality can be seen as a proxy for earnings quality and a lower earnings quality is associated with a higher cost of equity capital. This suggests that investors do not value lower accrual quality which in turn can be considered a constraint on accrual earnings management (Kim & Qi, 2010);

(8)

(iii) The third type of earnings management which is relatively unexplored is classification shifting. When applying this type of earnings management managers shift core expenses such as, costs of goods sold vertically to other classifications such as special items which in turn will lead to higher core earnings. This can be an important tool because investors and specifically analysts tend to focus on core earnings before the inclusion of these special items (McVay 2006). It can be considered a relatively cost-effective type of earnings management because the application of classification shifting, in contrast to REM and AEM, does not have any direct effects on the following or previous accounting periods. Therefore, classification shifting might be an interesting type of earnings manipulation for CEOs and other high influential executives.

Different studies have examined the trade-off between different types of earnings management and especially between REM and AEM. Zang (2012) concludes that REM and AEM can be considered substitutes and that the earning manipulation type of choice is determined by its relative costs. Further research stated that the preference of management shifted from accrual-based earnings management towards real earnings management following the implementation of the SOX and IFRS regulations (Ipino & Parbonetti, 2017; Francis, Hasan & Li, 2016). In addition, Fan, Barua, Cready & Thomas (2010) find evidence that executives tend to use classification shifting when the use of accrual-based earnings management is constrained. However, due to the shift in regulation which resulted in an increase of REM practices it is also important to consider the trade-off between REM and classification shifting and the determinants of this trade-off. Abernathy, Beyer & Rapley (2014) provided evidence of the determinants of this trade-off. They state that when REM is constrained by; (i) a poor financial position; (i) a high level of monitoring by shareholders and especially institutional shareholders; and (iii) a relatively low market share organization will shift to the application of classification shifting as earnings management technique instead of applying REM. Furthermore, Kim & Sohn (2013) state that the application of REM results in an increase of the cost of capital and therefore can be seen as an additional cost for the application of REM which, according to Zang (2012), can be considered as a constraint of the use of REM which is expected to lead to an increase of the application of classification shifting.

This study focuses on classification shifting because it is a relatively new and unexplored type of earnings manipulation, which is often used when there are constraints on accrual- and real activities-based earnings management which have increased over the years.

(9)

Classification shifting

Classification shifting is first addressed by McVay (2006), who provides evidence that US listed firms shift core expenses to special items to increase the core earnings in an attempt to mislead stakeholders as they often focus on core earnings and pay none or less attention to the special items as these are not perceived persistent. Classification shifting does not affect the bottom-line earnings of an organization and the method does not directly affect the earnings of the current, previous or next period while AEM and REM do affect this. Fan et al. (2010) confirmed the findings of McVay (2006) and provided evidence that managers are more likely to shift core expenses to special items in the period closest to an earnings announcement than in any other period during the year. Haw, Ho and Li (2011) were the first to investigate this type of earnings management outside of the US by examining the degree of classification shifting within firms in eight East Asian economies. They found that firms within these economies shift core expenses to manage their core earnings. However, they state that the results can mainly be attributed to family-controlled divergent firms. Behn, Gotti, Hermann & Kang (2013) extended this research by examining classification shifting using a broader scope through the use of a sample of firms from 41 different countries. They found that the results are not restricted to US listed firms, but that classification shifting is being used globally and that the results are not limited to countries with a low level of investor protection. They provide evidence that classification shifting occurs both in countries with a high and a low level of investor protection. Furthermore, Malikov, Manson & Coakley (2018) opened up a new alley of research by providing evidence that UK listed firms use the shifting of non-operating revenue to operating-revenues to overstate the operating-operating-revenues. They found that an unexpected increase in operating revenues is associated with a decrease in non-operating revenues and a reversal in the following year. These findings provide evidence that UK listed firms shift revenues from non-operating to non-operating to mislead investors. The increase in this type of classification shifting in the 1995-2014 period can be contributed to the mandatory adoption of IFRS in the UK for listed firms.

Both the shifting of core expenses to special items and non-operating revenues to operating revenues can be considered earnings manipulation as both techniques aim to fool stakeholders by either understating expenses or overstating operating revenues. This can be an effective way to manipulate earnings as analysts, investors and other stakeholders tend to focus on core or operating earnings - in which significant expenses are excluded as they are perceived non-recurring – instead of GAAP earnings (Bradshaw & Sloan, 2002). If organizations who apply a classification shifting technique do not apply such a technique in the following year the

(10)

core earnings are not expected to be persistent. Therefore, the market should not be fooled by these non-persistent core-earnings if the market is fully efficient. However, Alfons, Cheng & Pan (2012) found that the perceived persistence of core-earnings for organizations who apply classification shifting techniques is significantly higher than the actual persistence. This is in line with the research of Doyle, Lundholm & Soliman (2003) who provided evidence that excluding costs from earnings lead to an abnormal return. Finally, Athanasakou, Strong & Walker (2011) find that firms who achieve their earnings forecast through the use of classification shifting techniques will achieve an abnormal market return over firms who do not meet their earnings forecast. However, the return will be lower for shifting organization than for firms who meet the forecast using honest and integer reporting techniques. These results provide evidence that the market is, at least to some degree, fooled by dishonest, non-integer and thus misleading reporting styles. Thus, the application of classification shifting techniques will lead to an abnormal increase in firm value and can be considered an effective technique to boost earnings and firm value in a dishonest way.

In the research of what drives or limits earnings manipulation behavior there has been an extensive research on the implications of different governance mechanisms such as board structure (Obigbemi, Omolehinwa, Mukoro, Ben-Caleb & Olusanmi, 2016) (Jamaludin, Sanusi & Kamaluddin, 2015) the ownership structure (Kazemian & Sanusi, 2015) different subsidiaries (Beuselink, Casciono, Deloof & Vanstraelen, 2019) (Bonacchi, Cipollini & Zarowin, 2018) and both board and audit committee independence (Klein, 2002) (Xie, Davidson & Dadalt, 2003). Furthermore, there has already been extensive research to the incentives of CEOs and other executives to engage in earnings management practices such as; the portion of share-based pay (Berstresser & Philippon 2006) the need to achieve earnings targets and benchmarks set by analysts (McVay, 2006) and personal characteristics such as CEO tenure (Ali & Zhang; 2015). However, this last area of research is relatively unexplored while the theoretical perspective of the upper echelon theorist states that organizational outcomes can be seen as a reflection of the values and cognitive bias of the organization’s CEOs and other high influential executives. So, the personal values and characteristics are an important determinant of the outcomes of an organization and financial reporting decisions (Hambrick & Mason, 1984).

(11)

During their training, recruits are, to a certain degree indoctrinated, to adapt the military identity and due to the effect of social cohesion and inter-group relations the degree of adoption will increase over time. In addition, Jackson et al. (2012) stated that this effect will persist when recruits leave the military. So, executives with a military background will still possess the characteristics of the military identity and are therefore expected to act with a high level of integrity, honesty, obedience regarding the law and put respect for their profession and acting in line with the expected values over the pursuing of self-interest. Following the upper echelon perspective these values have a significant impact on organizational outcomes. Therefore, when CEOs with a military background faces the decision whether to manage the earnings through the shifting of core expenses or non-operating revenues or not it is expected that the CEO will act in an integer and honest way with respect for the law and will not manipulate earnings through the application of classification shifting. Therefore, I have developed the following hypothesis:

H1: There is a negative relationship between the military background of a CEO and the degree of classification shifting through the shifting of core expenses.

(12)

Research design

Sample and data collection

First all, I use firms listed in the Standard and Poor’s 1500 index, as this index is perceived representative for US listed firms as it comprises 600 lower-cap firms, 400 mid-size firms and 500 large-size firms. I use these US listed firms as military service is not prohibited in the US so there is heterogeneity between the military background of CEOs. Furthermore, US legislations make it possible to determine the demographic characteristics of CEOs which is needed to test the hypotheses. The data regarding the military background of these CEOs is hand collected. Secondly, after determining the geographic focus of the study the period which is being used should be determined. The research of Aberthy, et al. (2014) and Ipino * Parbonetti (2017) provided evidence of an increase in the use of REM techniques as AEM was constraint in the period following the implementation of the SOX regulations in 2002. Furthermore, Aberthy et al. (2014) find evidence that the application of classifications shifting techniques increased as the flexibility of accounting systems decreased. As the amount of regulations is increasing and thus the flexibility is decreasing the use of classification shifting is expected to increase over the years therefore it would be beneficial to use a sample which is relatively recent. Furthermore, as this study builds upon others on the paper of McVay (2006) data after 2006 is used. Finally, I aim to reduce the possible bias resulting from the financial crisis, therefore, the sample used in this study will consist of firm year observations from 2012. For these firms the financial data is extracted from the Compustat database, Firm year observations are excluded if; (1) the data which is needed for the equations below is missing; (2) the annual level of sales is below one million (McVay 2006); (3) firms use a firm year end which is different than December (McVay 2006). The final sample consists of 5076 firm year observations of which 5016 firm year observations can be used for the testing of the hypothesis.

Measuring classification shifting

To measure classification shifting I follow the method of McVay (2006) to estimate the expected level of core earnings and the expected change in core earnings1 _{and apply the following} equations for firm I within each industry and for each fiscal year:

1_{Another method which can be applied is the method of Malikov et al. (2018) which focuses on the shifting of} non-operating revenues, however the regulation regarding the recognition of sales can be perceived stricter than those of expenses. Furthermore, the data regarding other operating expenses is not readily available for the sample consisting of US firms.

(13)

(1)

𝐶𝐸_! = 𝛽_"+ 𝛽_#𝐶𝐸_$%#+ 𝛽_&𝐴𝑇𝑂_$+ 𝛽_'𝐴𝐶_$%#+ 𝛽₍∆𝑆𝐴𝐿𝐸_$+ 𝛽₎𝑁𝐸𝐺∆𝑆𝐴𝐿𝐸_$+ 𝜇_$

(2)

∆𝐶𝐸_! = 𝛾_"+ 𝛾_#𝐶𝐸_$%#+ 𝛾_& ∆𝐶𝐸_$%#+ 𝛾_'∆𝐴𝑇𝑂_$+ 𝛾₍𝐴𝐶_$%#+ 𝛾₎∆𝑆𝐴𝐿𝐸_$+ 𝛾_*𝑁𝐸𝐺∆𝑆𝐴𝐿𝐸_$

+ 𝜀_$

In the above equations core earnings (CE) are defined as the sales in year t minus the cost of goods sold (COGS) and selling-, general- and administrative expenses divided by sales to control for size. In both equation one and two the first variable is lagged core earnings as core earnings are expected to be persistent in a high degree (McVay, 2006; Fan et al. 2010; Fan, Thomas & Yu, 2019). Next, in the second equation the change in core earnings from t-2 to t-1 is included to control for the effect of mean reversion as this is generally present from year to year but in a higher degree when there is a large deviation from the mean according to Fama and French (2000). Mcvay (2006) states that following this evidence, the variable should be included as the effect of mean reversion is larger in the tail. The next variable is asset turnover ratio (ATO) which is being included because Nissim & Penman (2001) found that this ratio is closely related to profit margin and thus can predict future payoffs in a relatively high degree. McVay (2006) states that the profit margin can been seen as equal to her definition of core earnings. Furthermore, she states that organizations with large income-decreasing special items are expected to implement changes to their strategy and in the process influence their turnover, highlighting the importance to include ATO in equation one. In the second equation the change in ATO between t-1 and t0 is used as the variable of interest is the change in core earnings.

The next variable of interest is accruals, McVay (2006) uses both current period and lagged accruals. As current period’s accruals can be used to control for extreme performance as accruals can being used to manipulate earnings and these abnormal high accruals will reverse in the next period. Furthermore, lagged accruals can be used to explain future performance (Sloan 1996). However, the use of current period accruals can lead to a possible bias as current accruals include accrual special items which can create a relation between the dependent and independent variable which is only mechanical. Therefore, I do not include current accruals in above equations. Furthermore, both change in sales from t-1 to t0 and the negative change of sales are included in the equation. The positive delta sales is included as control because the fixed costs are likely to be constant while sales can fluctuate which will lead to a decrease of fixed costs per dollar of sales when sales grow. Furthermore, the negative change of sales is included because the negative effects of a decrease in activity is expected to be incremental to

(14)

the positive effect of an equal increase in activity according to Anderson, Banker and Janakiraman (2003). As the SGA costs increase with a higher percentage due to an increase of 1% in sales than these costs decrease due to a decrease of 1% in sales.

Finally, the error variable in both equations (𝜇_$) and (𝜀_$), the error variable (𝜇_$) is the estimated level of unexpected core earnings and (𝜀_$) is the unexpected change in core earnings. The unexpected level of core earnings and the change in the unexpected level are estimated for each industry group following the Fama and French classification and for each firm year.

The method described above to determine the level of unexpected core earnings and the change in unexpected core earnings is in line with the previous research of Behn et al. (2013) Haw et al. (2011), Fan et al. (2010) and Fan et al. (2019) who build on the research of McVay (2006).

Testing of the hypothesis

As Mcvay (2006) states the level of unexpected core earnings is expected to be related to special items as firms are expected to recognize income decreasing special items in the same period in which they overstate their core earnings. Thus, if managers apply classification shifting techniques, the core earnings are expected to be overstated by the increase of special items. She further states that this effect could also be due to poor firm performance, if the coefficient is negative the special items are due to poor firm performance. However, this does not provide a firm year measure of the degree in which firm apply classification shifting techniques and therefore an interaction variable is introduced to the regression which is coefficient b3 in equation (3) below. If the coefficient is positive the special items are due to opportunistic reporting strategies and thus classification shifting. So, we expect this coefficient of unexpected core earnings and military background multiplied with special items to be negative when the firm has a CEO with a military background. as I expect that the coefficient in the first model will be negative and therefore there will be no classification shifting practices the second model of the McVay (2006) is not needed as there is no reversal expected in the following year because there will be no shifting at all in the first place.

Furthermore, there are a number of variables which I need to control for. First of all, the presence or absence or a Big 4 auditor can have a significant influence on earnings manipulation practices. Haw et al. (2011) found that Big 4 auditors are highly alert of less noticeable forms of earnings manipulation practices in countries with a high level of investor protection. As the US is considered to have a high level of investors protection this is a variable which needs to

(15)

be controlled. I control for this by included the variable Big_4 which will be 0 if the firm is audited by a non-Big 4 firm and 1 otherwise. 2

To control for the growth opportunities the variable BTM is included in the regression which is the book to market ratio following the method of Fan et al. (2019). The fourth and last variable for which I need to control is the degree to which the managers ability to manage accruals is constraint as this is related to the use of the application of classification shifting techniques. Fan et al. (2011) found that the degree to which classification shifting is applied is related to the level of net operating assets (NOA), if NOA was above average the firm the application of classification shifting techniques. Therefore, the variable HiNoa is included which will be 1 is the net operating assets of the firm are above average and 0 if it is below average. Finally, I control for growth opportunities as McVay (2006) observed that firms which use classification shifting techniques to beat the analysts’ forecast are mainly firms with high growth opportunities.

Thus, to determine how the military background of a CEO affects the application of classification shifting of firms I include the control variables mentioned above and interacting variables for each variable in equation 3.

(3) UE_CE = 𝛽_"+ 𝛽_#𝑆𝐼 + 𝛽_&𝑀𝑖𝑙𝑖𝑡𝑎𝑟𝑦 𝐶𝐸𝑂 + 𝛽_'𝑆𝐼 ∗ 𝑀𝑖𝑙𝑖𝑡𝑎𝑟𝑦 𝐶𝐸𝑂 + 𝛽( 𝐵𝑖𝑔_4 + 𝛽)Big_4 ∗ SI + 𝛽*𝐵𝑇𝑀 + 𝛽,BTM ∗ SI + 𝛽-𝐻𝑖𝑁𝑜𝑎 + 𝛽.𝐻𝑖𝑁𝑜𝑎 ∗ 𝑆𝐼

2_{Another variable to control for external monitoring is the percentage of share held by institutional} shareholder, as the structure of a firm in its legal environment has an influence on the degree in which it uses classification shifting (haw et al. 2011). This control variable is excluded due to the fact that data for this variable could not be collected in time and Big4 is already used to control for external monitoring.

(16)

Results

Classification shifting

Before I can conduct the pre-test and can start with the testing of the hypothesis, I need to determine the level of unexpected core earnings and the change in the level of unexpected core earnings. As mentioned, the method of McVay (2006) will be applied to a sample of firm year observations from 2012 of the S&P 1500 index. The descriptive statistics of the variables used in equation (1) and (2) and the level of unexpected core earnings and change in unexpected core earnings are presented in table 1 below.3

Tabel 1. Descriptive statistics full sample

Variable Mean Median Std. Dev. 25% 75%

Core Earningst 0.1646 0.1501 0.1935 0.0892 0.2394 Core earningst-1 0.1663 0.1502 0.1820 0.0881 0.2393 DCore earningst-1 0.0003 0.0002 0.1742 -0.0363 0.0373 Asset turnover ratiot 2.6976 1.5912 3.8070 0.7126 3.0078 DAsset turnover ratiot -0.0533 -0.0309 2.9727 -0.7295 0.6285 Accrualst -0.1079 -0.0565 0.2099 -0.1236 -0.0186 %Dsales 0.5395 0.0000 2.1591 -0.3193 0.5279 Negative%Dsales -0.1828 0.0000 0.2599 -0.3193 0.0000 Unexpected Core Earningst 0.0020 -0.0017 0.1198 -0.0432 0.0468 DUnexpected Core Earningst+1 0.0013 -0.0013 0.1066 -0.0402 0.0412 Special items 0.0216 0.0022 0.0597 0.0000 0.0150 The full sample consisting of all the needed variables are 3946 firm years observations relating to unexpected core earnings and the level of change in unexpected core earnings. The mean of unexpected core earnings is 0,0004 while the mean of the change in unexpected expected core earnings is 0,0019 which is in line with the research of McVay (2006).

Next, to check if firms apply classification shifting, I am going to apply the main tests of McVay (2006) to the whole sample, and apply the following equations4_:

3_{The asset turnover ratio is defined as sales divided by net operating assets where NOA is calculated as the}

total equity minus sales plus total debt following the method of Zang (2012). Accruals are defined as income before extraordinary items less cash flows from operations. Negative delta sales is the change in sales between t-1 and t0 if the change in sales is negative, if the change is positive the value will be 0. The unexpected core earnings is the error variable of equation (1).

4_{UE_CE and 𝑈𝐸_∆𝐶𝐸}

(17)

(4) 𝑈𝐸_𝐶𝐸$= 𝑎"+ 𝑎#%𝑆𝐼 + 𝜀$

(5)𝑈𝐸_∆𝐶𝐸_$/#= 𝜂_"+ 𝜂_#%𝑆𝐼 + 𝜐_$/#

To determine whether firms apply classification shifting techniques using the above equations (4) and (5). In equation (4) the coefficient of 𝑎# is expected to be positive as an increase in

income decreasing special items is expected to be related to unexpected core earnings. In equation (5) the coefficient 𝜂_# is expected to be negative, as firms are not expected to shift core expenses in following year McVay (2006). The results of these equations are shown in table 2 below.56

Tabel 2. Test McVay (2006) existence Classifcation Shifting

(1) (2) (3) (4)

VARIABLES UE_CE DUE_CE UE_CE DUE_CE

%SI 0.00423 -0.00271 0.00313 -0.00301 (0.0282) (0.0251) (0.0283) (0.0252) Constant 0.00190 0.00136 0.00215 0.00143 (0.00179) (0.00159) (0.00181) (0.00161) Observations 5,076 5,076 5,016 5,016 R-squared 0.000 0.000 0.000 0.000

Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

As can be seen in table 2 both for the whole sample (1) and (2) and the test sample (3) and (4) the coefficients of special items and unexpected core earnings is positive, and the coefficient of special items and delta unexpected earnings is negative. However, these coefficients are not significant and therefore, based on the entire test sample, I cannot conclude with certainty that firms in my sample apply classification in general.

5_{%SI is the special items as defined in footnote 3. The coefficient of SI and unexpected core earnings is 𝑎}

" which

is expected to be positive. Furthermore, the coefficient of %SI and the change in unexpected core earnings is expected to be negative. When both these coefficients move in the expected direction and are significant it can be concluded that firm use classification shifting techniques.

6_{After including control variables 60 firm year observations are excluded from the sample due to missing data so} the final sample for the testing of the hypothesis consists of 5016 firm year observations.

(18)

Main results

After determining that the coefficients move in the expected direction, I apply regression (3) and (4) of which the descriptive statistics and the Pearson correlation matrix are presented below in table 37_{and 4.}

Table 3 Descriptive statistics hypothesis testing

Variable Mean Median Std. Dev. 25% 75%

Unexpected Core Earningst 0.0022 -0.0014 0.1202 - 0.0423 0.0471 Military CEO 0.0614 0.0000 0.2401 0.0000 0.0000 Special items 0.0217 0.0022 0.0599 0.0000 0.0151 Big4 0.8911 1.0000 0.3115 1.0000 1.0000 BTM 0.4396 0.2181 0.8371 0.0835 0.4769 HiNoa 0.4131 0.0000 0.4924 0.0000 1.0000

Tabel 4. Pearson correlation matrix

Variable A B C D E F UE CE A 1.0000 Military B -0.0214 1.0000 %SI C 0.0016 -0.0006 1.0000 Big4 D 0.0088 -0.0240 0.0303* _1.0000 BTM E 0.0458** _-0.0456** _0.0058 _-0.0609*** _1.0000 Hi Noa F 0.0312* _-0.0253 _0.0124 _0.0085 _-0.0364** _1.0000 *** p<0.01, ** p<0.05, * p<0.1

7_{The total sample which is used for the testing of the hypothesis consists of 5016 firm year observations.} Unexpected core earnings is the residuals of equation (1). Military CEO is a variable to indicate whether a CEO has a military background of which the value will be 0 if the CEO does not and 1 if the CEO has a military background. Special items are the income decreasing special items as percentage of sales. Big4 is the control variable which will have the value 0 if the firm is audited by a non-big4 and 1 if the firm is audited by a big4 auditor. BTM is the book to market value which is the book value of equity divided by market value. Finally, HiNoa is a control variable which will have the value 1 is the net operating assets are above industry average and 0 if the net operating assets are below industry average.

(19)

The results of testing the hypothesis are presented in table 5 in which the interacting variable between the total amount of special items scaled by sales and military CEO is of interest. The coefficient of this interacting variable is negative and significant -0.261 (t = -2.11) which indicates that a military background constrains the degree in which a CEO applies classification shifting practices. Firms with a CEOs who have a military background exhibit a negative relation between unexpected core earnings and special items. This indicates that when these firm have unexpected core earnings their income decreasing special items swill be lower. Thus, these firms do not shift core expenses towards income decreasing special items.

Table 5 the effect of a military background on classification shifting

(1) VARIABLES UE_CE %SI -0.2839*** (0.0897) Military CEO -0.0007 (0.00763)

Military CEO * %SI -0.2614**

(0.124) Big4 0.0109* (0.00578) Big4 * %SI 0.3242*** (0.0902) BTM 0.0102*** (0.00225) BTM * %SI -0.0559* (0.0290) Hi Noa 0.0008 (0.00375) Hi NOA * %SI 0.0233 (0.0567) Constant -0.0431

Industry fixed effects Year fixed effects

(0.0378) YES YES

Observations 5,016

R-squared 0.060

(20)

As McVay (2006) and Fan et al. (2019) state, firms are less likely to engage in classification shifting practices if they do not have a significant reported number of special items in in their usual business the first place. Therefore, I follow McVay (2006) and compare the full sample results in table 5 firms with a sample consisting of firm year observations with more than five percent income decreasing special items8_{as firms with a higher reported number of special} items are more likely to apply classification shifting.

Table 6 the effect of a military background on classification shifting when special items are >5% of sales (1) VARIABLES UE_CE %SI -0.0716 (0.192) Military CEO 0.0794* (0.0466)

Military CEO * %SI -0.7093***

(0.258) Big4 0.0428 (0.0418) Big4 * %SI 0.2241 (0.192) BTM 0.0479*** (0.0135) BTM * %SI -0.1645*** (0.0607) Hi Noa -0.0340 (0.0250) Hi Noa * %SI 0.1275 (0.124) Constant -0.1350

(0.167) YES YES

Observations 508

R-squared 0.146

8_{Income decreasing special items are scaled by sales to determine the percentual amount of income decreasing} special items.

(21)

Again, the interacting variable between percentual special items is of interest. As can be seen in table 6 the coefficient is smaller (more negative) -0.709 (t = -2.75) however, after testing, the magnitude of this coefficient is not significantly lower than de coefficient of military CEO * procent_si in table 5. This provides evidence that a military CEO does not engage in classification shifting even when the number of special items increases. This, in turn, suggests that when the military CEOs are in a high incentive situation such as a high number of special items, they still benefit from their military identity as it provides a resistance against earnings management practices.

Additional tests

Just meeting or beating targets

Targets or threshold are mainly important because of two following reasons according to Degeorge, Patel and Zeckhauser (1999; (i) stakeholders who invest in the firm are concerned with the performance of the firm, value targets and therefore managers value them as well; (ii) managers can benefit from beating or meeting targets for their personal gains. Furthermore they distinguish three types of targets and their relative performance; (1) the first and most important target is to report positive earnings; (2) the second and second most important target is to achieve at least persistent earnings; (3) the last and least important of the three is the target based on expectations of analysts. In their research they show that a relatively large number of firms report small profits compared to reporting small losses as the market is known to punish firm which report small losses. Furthermore, they provide evidence that a relatively large number report a small increase in earnings per share and report a shaved distribution around zero change. Finally, they report that more firms perform above the analyst’ expectation or exactly meet them than report earnings below their expectations. These findings indicate that firms often apply earnings management techniques to just meet or beat any of the three targets. I am going to test whether the degree in which military CEOs engage in classification shifting increase when they face a high incentive situation to engage in earnings management practices. I am going to classify firms into two categories; (i) firm who have just meet or beat the first two targets Degeorge et al. (1999) identify; (ii) firm who do not have just achieved these targets.

Next, it is important to define the range which will be used to determine whether a firm has just met or beaten its target. Different papers apply different ranges, for instance McVay (2006), Fan et al (2019) and Degeorge et al. (1999) all use earnings per share and use the range between 0.00 and respectively 0.01, 0.02 and 0.05 earnings per share. When applying these methods, the sample size of the just met group will be relatively small and the number of

(22)

military CEOs will be even smaller. Therefore, I applied method of Jeanjean & Stolowy (2008) who define a small profit and thus a small exceedance of the threshold as income before extraordinary items scaled by lagged total assets and use a range between 0.00 to 0.01. I run regression (3) again on these two subsamples, the results are presented in table 7 below. Table 7 just meeting or beating targets

(1) (2)

VARIABLES UE_CE UE_CE

%SI -0.3618*** 1.151***

(0.0959) (0.347)

Military CEO 0.0003 -0.0153

(0.00893) (0.0155)

Military CEO * %SI -0.2652** 0.7057

(0.130) (0.889) Big4 0.0130** 0.0127 (0.00660) (0.0121) Big4 * %SI 0.410*** -0.9085*** (0.0970) (0.329) BTM 0.0124*** -0.0073 (0.00248) (0.00572) BTM * %SI -0.0655** -0.1193 (0.0310) (0.108) Hi Noa 0.0011 0.0010 (0.00432) (0.00770) Hi Noa * %SI 0.0174 -0.1262 (0.0599) (0.257) Constant -0.0425 -0.0536

(0.0437) YES YES (0.0701) YES YES Observations 4,070 946 R-squared 0.062 0.136

In the first column of table 7 are the results of regression (3) when firms do not have just meet or beaten the thresholds. The coefficient of military CEO * procent_si is of interest. In this column the coefficient is significantly negative -0.265 (t = -2.04) which is as expected. In the second column of table 7 are the results of the regression (3) for firms who do have just beat or met the threshold as defined by Jeanjean & Stolowy (2008), the coefficient of military CEO * procent_si is positive 0.706 although not significant, this could be an indication that firms who have just beat or met the threshold have applied classification shifting to achieve this. However,

(23)

additional tests show that there is no reversal in the following year, this suggests that the positive coefficient is due to ‘normal’ economic consequences which were not expected and thus do military CEOs not apply classification shifting even when they have just met certain thresholds.

CEO Tenure

In this section I am going to test whether the tenure of military CEOs influences the degree in which they apply classification shifting techniques. Graham, Harvey & Rajgopal (2005) found that CEOs take the effects on their reputation into account when they have to make decisions. In addition, Zhang (2009) found that managers use more aggressive accounting techniques in the beginning of their period as CEO as they try to build their reputation and the CEO labor market focusses mainly on the performance of their firms the early days of a CEOs focuses. Ali & Zang (2015) build on this research and found that earnings management and the application of earnings management techniques is higher in the early of the CEO’s tenure. So, it is expected that CEOs engage in more classification shifting in the early years of their tenure as they are trying to build their reputation and the labor market tends to rely on firm performance. To test whether the expected relationship still holds up for military CEOs who are in their early of their corporate career, I am going to following the definition of Ali and Zang (2015) who define the early years as the first three firm years the CEO works for the firm. Once the early days and therefore low tenure CEOs are defined, I am going to run regression (3) again on the sub sample of low and high tenure CEOs.

The results of the test are present in table 8 below. In the first column of table 8 are the results of regression (3) relating to CEOs with a maximum of three years tenure. The coefficient of military CEO * procent_si of these is positive 0.482 (t = 0.87). Although not significant, this could be an indication that these firm could apply classification shifting practices through the shifting of core expenses to special items. However, further tests indicate that there is no reversal in the following year. This suggests that the positive coefficient between Military CEO * procent_si and unexpected core earnings in year t, for low tenure CEOs, is due to unexpected benefits caused by restructuring in the early tenure of military CEOs (McVay 2006). In the second column of table 8 are the results for CEOs with at least 4 years of tenure, for this subsample the coefficient Military CEO * procent_si is both negative and significant -0.2948 (t = -2.27) which indicates that military CEOs in their later tenure do not shift core expenses towards special items to inflate core earnings

Table 8 classification shifting when CEOs have high versus low tenure

(24)

%SI -0.2792* -0.2948***

(0.168) (0.108)

Military CEO -0.0202 0.0047

(0.0159) (0.00906)

Military CEO * %SI 0.4818 -0.2971**

(0.553) (0.131) Big4 0.0234* 0.0080 (0.0122) (0.00667) Big4 * %SI 0.2461 0.3619*** (0.176) (0.107) BTM 0.0040 0.0127*** (0.00371) (0.00286) BTM * %SI 0.0861 -0.0988*** (0.0616) (0.0336) Hi Noa 0.0050 1.11e-05 (0.00712) (0.00447) Hi Noa * %SI 0.1012 -0.0077 (0.106) (0.0683) Constant -0.0748 -0.0407

(0.114) YES YES (0.0406) YES YES Observations 1,346 3,670 R-squared 0.067 0.071

It could be that the results above are driven by the fact that newly appointed CEOs often apply a technique called taking a bath which would lead to similar results. Taking a bath relates to the reporting of large losses by including additional income decreasing items and the underreporting in the current when the firm makes a small loss with the purpose of reporting a profit in the following year (Kirschenheiter & Melumad, 2001). Previous research of Masters-Stout, Costigan & Lovata (2008) shows that newly appointed CEOs apply this technique as they are able to contribute this loss to their predecessor. To rule out the possibility that the positive coefficient Military CEO * Procent_si of the low tenure subsample is due to the fact that low tenure military CEOs take a big bath in their early tenure I run regression (3) again on the subsample of low tenure CEOs in which I exclude the year in which there was a CEO turnover. The results of this additional test are presented in table 9 below.

Again, the variable Military CEO * procent_si is of interest, this time the firm year observations which are the year in which there was a CEO turnover are excluded from the

(25)

sample. Again, although not significant, the variable is positive 0.2179 (t=0.38) which could be an indication that these firms do apply classification shifting. However, additional tests show that there is no reversal in the following year and therefore we can attribute the positive coefficient for these lower tenure CEOs to an unexpected restructuring benefit. Thus, the results of table 9 are not driven by CEOs who take a big bath in the first year of their tenure.

Table 9 classification shifting for low tenure firms excluding year of turnover

(1)

VARIABLES UE_CE

%SI -0.2531

(0.171)

Military CEO Low -0.0202

(0.0169)

Military CEO * %SI 0.2179

(0.577) Big4 0.0266** (0.0130) Big 4 * %SI 0.2054 (0.180) BTM 0.0059 (0.00416) BTM * %SI 0.0508 (0.0667) Hi Noa 0.0092 (0.00768) Hi Noa * %SI 0.0638 (0.118) Constant -0.0163

(0.0386) YES YES

Observations 1,170

R-squared 0.073

(26)

Time served in the military

Another variable which may influence the degree to which military CEOs apply classification shifting is the time they have served in the military as Elder (1986) pointed out that the early entrants in the military experienced a stronger influence of their service. Duffy (2006) builds on this research by addressing the similarities between leaders in the army and corporate CEOs and how a military background would benefit a corporate career. He found that in the military there is a lot of emphasis on working together as a team, planning and performing under high levels of pressure. In addition, the military is the place in which an individual gets a leadership role at a relatively young age. The theoretical perspective was confirmed by a CEO with a military background who stated that the military is an excellent starting point for a corporate career. Furthermore, he emphasizes that the first ten years of one’s military experiences would benefit him or her the most in a future corporate career. Therefore, to examine the influence of the time served in the military I am going to redefine the military CEO variable and qualify a military CEO as a CEO who has a least ten service years. Once the military background variable is redefined, I am going to run regression (3) again.

In the first column of table 9 are the results of regression (3) in which the military CEO variable is redefined as a CEO who has at least 10 years of service for the whole test sample of 5016 firm year observations. The first variable of interest is Military CEO high * procent_si, the coefficient is significantly negative -0.4772 (t = -2.97)

In the second column of table 9 are the results of regression (3) in which the military CEO has les 10 service years. The variable of interest is Military CEO low * procent_si. Interestingly, the variable of interest is positive 0.0513 (t = 0.27). Furthermore, further test show that the coefficient of the column of delta unexpected core earnings and Military CEO low * procent_si is negative -0.192 (t = -0.99), although not significant this is an indication that these firms apply classification shifting techniques.

When comparing the coefficient of Military CEO high * procent_si and Military CEO low * procent_si the coefficient of military CEO high is smaller (more negative) than the coefficient of military CEO low and the difference between both is significant. These results indicate that the amount of service years of the CEO influences the degree in which they engage in classification shifting as CEOs with at least ten service years have a strong resistance and do not apply classification shifting practices. Furthermore, it suggests that when a CEO has a low number of service years (less than ten years) they will not or at least less adapt to the military identity and thus their military background does not influence, or at least influences them less, relating to organizational decisions whether or not to apply classification shifting techniques.

(27)

Table 10 classification shifting time served in the military

(1) (2)

%SI -0.2786*** -0.2877***

(0.0897) (0.0897) Military CEO high 0.0033

(0.0101) Military CEO high * %SI -0.4772***

(0.161)

Military CEO low -0.0056

(0.0112)

Military CEO Low * %SI 0.0513

(0.187) Big4 0.0109* 0.0110* (0.00578) (0.00578) Big4 * %SI 0.3236*** 0.3100*** (0.0901) (0.0902) BTM 0.0102*** 0.0101*** (0.00224) (0.00225) BTM * %SI -0.0570** -0.0511* (0.0290) (0.0290) Hi Noa 0.0009 0.0008 (0.00375) (0.00375) Hi Noa * %SI 0.0163 (0.0567) (0.0568) 0.0207 Constant -0.0432 -0.0429

Industry Fixed effects Year Fixed effects

(0.0378) YES YES (0.0378) YES YES Observations 5,016 5,016 R-squared 0.061 0.059

Conclusion

In this study I examine the effects of a Military background on the degree in which CEOs apply classification shifting. I follow the definition of McVay (2006) who defines classification shifting as a technique in which core expenses are shifted towards special items in an attempt to artificial inflate core earnings as this is the main focus of investors. CEOs who have served in the US army are expected to act with a high level of integrity, obedience regarding the law and honesty as these are key values of a common military identity. Thus, as classification shifting is a way to mislead investors and therefore is not in line with the core

(28)

values of a military identity, I expect that CEOs with a military background have an aversion against such reporting.

Consistent with my expectation I find that when a CEO has served in the US military, they are unlikely to engage in classification shifting. Further tests reveal that; (i) the degree in which military CEOs apply classification shifting techniques decreases, although the difference is not significant, when a firm has more than 5% special items, scaled by sales. Firms with more than 5% special items are better able to shift items undetected as they report a certain amount special items in their regular operations; (ii) when firms have just met or exceeded a threshold income decreasing special items are positively related to unexpected core earnings, however, there is no reversal in the following year so military CEOs do not apply classification shifting even when they report small earnings. Small earnings or increase in earnings could be an incentive as the capital market punishes firms if they do not achieve these thresholds; (iii) military CEOs do not engage in classification shifting even in their low tenure. In their low tenure the incentives are higher as the labor market rates their performance mainly on earnings; (iv) It takes time adopt the common military identity as results show that CEOs who have served more than ten years do not apply classification shifting, while results suggest that CEOs with less than 10 service years do apply classification shifting. Thus, a military background of a CEO works as a constraint on classification shifting, even in high incentive situations. However, it is important to consider the time which a CEO has served as results show that there is a significant difference between less than and more than ten service years.

My study also has some limitations. First, as the sample consists of US listed firms and the military background relates to the US military, results are not generalizable across different countries. Secondly, the model used to determine core earnings is not perfect which possibly influence the outcomes (McVay 2006).

There are also possibilities for future research. First, as McVay (2006) points out, the model to determine the expected core earnings is imperfects but it still is a respected method in research. Future research could focus on improving the method to determine the unexpected core earnings. Secondly, As I documented that the amount of service years has an influence on the degree in which managers apply classification shifting future research could extend this research and examine the impact of the amount of service years in more detail and in different settings. Finally, as my study is one of the first to examine the influence of managerial characteristics on classification shifting, future research could examine the effect of other managerial characteristics on the application classification shifting as it is a method of earnings management which is quite new and therefore relatively unexplored.

(29)

References

Abernathy, J. L., Beyer, B., & Rapley, E. T. (2014). Earnings management constraints and classification shifting. Journal of Business Finance & Accounting, 41(5-6), 600-626.

Akerlof, G. A., & Kranton, R. E. (2005). Identity and the Economics of Organizations. Journal of Economic perspectives, 19(1), 9-32.

Ali, A., & Zhang, W. (2015). CEO tenure and earnings management. Journal of Accounting and Economics, 59(1), 60-79.

Anderson, M. C., Banker, R. D., & Janakiraman, S. N. (2003). Are selling, general, and administrative costs “sticky”? Journal of accounting research, 41(1), 47-63.

Alfonso, E., Cheng, C. A., & Pan, S. (2015). Income classification shifting and mispricing of core earnings. Journal of Accounting, Auditing & Finance, 0148558X15571738.

Ali, A., & Zhang, W. (2015). CEO tenure and earnings management. Journal of Accounting and Economics, 59(1), 60-79.

Athanasakou, V., Strong, N. C., & Walker, M. (2011). The market reward for achieving analyst earnings expectations: Does managing expectations or earnings matter? Journal of Business Finance & Accounting, 38(1‐2), 58-94.

Behn, B. K., Gotti, G., Herrmann, D., & Kang, T. (2013). Classification shifting in an international setting: Investor protection and financial analysts monitoring. Journal of International Accounting Research, 12(2), 27-50.

Benmelech, E., & Frydman, C. (2015). Military ceos. Journal of Financial Economics, 117(1), 43-59.

Bergstresser, D., & Philippon, T. (2006). CEO incentives and earnings management. Journal of financial economics, 80(3), 511-529.

Bertrand, M., & Schoar, A. (2003). Managing with style: The effect of managers on firm policies. The Quarterly journal of economics, 118(4), 1169-1208.

Beuselinck, C., Cascino, S., Deloof, M., & Vanstraelen, A. (2019). Earnings management within multinational corporations. The Accounting Review, 94(4), 45-76.

Bonacchi, M., Cipollini, F., & Zarowin, P. (2018). Parents’ use of subsidiaries to “push down” earnings management: Evidence from Italy. Contemporary Accounting Research, 35(3), 1332-1362.