THE EFFECT OF FINANCIAL LEVERAGE ON CLASSIFICATION SHIFTING

(1)

THE EFFECT OF FINANCIAL LEVERAGE ON

CLASSIFICATION SHIFTING

Sigrid Jansen

MSc Accountancy & Controlling, Faculty of Economics and Business Administrations, University of Groningen

Abstract

Classification shifting is an earnings management tool that refers to the deliberate choice to move core expenses to noncore special items within the income statement to increase core earnings. This study examines the effect of financial leverage on classification shifting in U.S. firms. High leverage has been considered to attract increased monitoring by outside creditors. Prior studies have shown a negative effect of leverage on accrual-based earnings management, while a positive effect of leverage on real earnings management. The preference for engaging in real earnings management might be due to the lower possibility of getting caught. This study examines whether highly leveraged firms, facing increased monitoring by creditors, still remain to engage in classification shifting. I established that the unexpected core earnings increase along with special items, indicating that managers shift core expenses to special items in order to increase core earnings. Besides, the results show that classification shifting is more prevalent among highly leveraged firms. The findings indicate that despite an increase in monitoring by creditors, firms engage in classification shifting.

Keywords: earnings management; classification shifting; financial leverage; special items.

Name: Sigrid Jansen Student number: S3541932 Thesis supervisor: C. K. Hoi Date: June 22, 2020

Word count: 7172

MSc Accountancy & Controlling

Faculty of Economics and Business Administrations University of Groningen

(2)

2

1. Introduction

An essential role of financial reports is to effectively communicate financial information to those outside the firm in a timely and credible manner (FASB, 1984). Earnings numbers are a major component of the annual reports that are used by outsiders to make decisions regarding the firm. Earnings reliability becomes questionable when managers have incentives to manipulate reported earnings numbers (Dechow & Skinner, 2000). A major issue to accounting research is the extent to which managers are allowed to manage reported earnings and to which extent it becomes fraud.

The use of earnings management to mask a firms’ true economic performance has been the focus of many papers. Literature has focused on two general earnings management tools: accrual-based earnings management and the manipulation of real economic activities. The paper of McVay (2006) examines a third potential earnings management tool, namely classification shifting. McVay (2006) was the first to determine empirical evidence that managers of U.S. firms shift core expenses to special items within the income statement in order to increase core earnings.

The purpose of this study is to examine the effect of financial leverage on classification shifting. High leverage has been considered to attract increased monitoring by outside creditors. Prior studies have shown a negative effect of leverage on accrual-based earnings management (e.g., Jelinek, 2007; Rodríguez-Pérez & van Hemmen, 2010), while a positive effect of leverage on real earnings management (e.g., Bartov, 1993; Anagnostopoulou and Tsekrekos, 2017). The preference for engaging in real earnings management might be due to the lower possibility of getting caught. Real earnings management is harder to detect and less likely to be scrutinized by outside parties. No study yet determined the relationship between leverage and classification shifting. I am interested in whether the findings by prior studies carry over to classification shifting. If highly leveraged firms, facing increased monitoring by creditors, still remain to engage in classification shifting, just like they still manipulate real activities. To further support and extend the literature on earnings management, I examine the effect of financial leverage on classification shifting. This leads to the following research question: What is the impact of

financial leverage on classification shifting in U.S. firms?

To examine the relation between financial leverage and classification shifting, data is obtained from the Compustat Capital IQ database (North America file) for the years 1988-2018. Classification shifting is measured using the regression model introduced by McVay (2006). If

(4)

4

firms engage in classification shifting, the unexpected core earnings are expected to increase along with special items.

This study provides evidence that the unexpected core earnings are positively associated with income-decreasing special items, indicating that managers shift core expenses to special items in order to increase core earnings, consistent with the findings by McVay (2006). Furthermore, the results show that classification shifting is more prevalent among highly leveraged firms. The findings indicate that despite an increase in monitoring by creditors, firms engage in classification shifting.

This study contributes to the existing literature in two ways. First, this study is the first to investigate the relationship between leverage and classification shifting. Second, high leverage has been considered to attract increased monitoring by outside creditors. Prior studies have shown a negative effect of leverage on accrual-based earnings management, but a positive impact of leverage on real earnings management. This study is the first that shows a positive association between leverage and classification shifting, indicating that classification shifting is more prevalent among highly leveraged firms. Therefore, this research further supports and extends the literature that argues that highly leveraged firms, facing increased monitoring by creditors, still engage in earnings management tools that are hard to detect and less likely to be scrutinized by outside parties. This research has implications for shareholders who strive to avoid investing in a firm with deceptive financial statements and auditors who are trying to assess a firm's potential for fraudulent behavior.

The paper is organized as follows. Section 2 provides a review of the related literature and develops the hypothesis. Section 3 describes the research design. Section 4 presents the empirical results. Finally, section 5 summarizes the findings, limitations, and further research topics.

(5)

5

2. The literature review

2.1 Earnings management

Agency theory explains the relationship between a principal (e.g., the shareholder) and an agent (e.g., the management). Agency costs arise when there is a conflict of interest between the principal and agent (Jensen & Meckling, 1976). Information asymmetry between the management and shareholders exists when managers possess private information about the firm that shareholders do not have (Richardson, 2000). Due to information asymmetry, shareholders might not have the necessary information on the firm's earnings to monitor the management, which provides the opportunity for managers to steer reported earnings. There are several different definitions of earnings management. According to Healy and Wahlen (1999), “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” (p. 368).

There are three potential earnings management tools, namely real earnings management, accrual-based earnings management, and classification shifting. The manipulation of real economic activities is an action to alter reported earnings in a particular direction by changing the timing or structuring of an operation, investment, or financing transaction (Zang, 2012). Instead of using real transactions to steer earnings, accrual-based earnings management is achieved by changing the accounting methods, or estimates used when presenting transactions in the financial statements (Zang, 2012). On the other hand, classification shifting refers to the deliberate choice to move core expenses (defined as cost of goods sold and selling, general, and administrative expenses) to noncore special items within the income statement to increase core earnings (McVay, 2006).McVay (2006) was the first to determine empirical evidence that managers of U.S. firms shift core expenses to special items in order to increase core earnings. Classification shifting differs from accrual-based and real earnings management in several ways. First, unlike accrual-based and real earnings management, classification shifting does not change the generally accepted accounting principles (GAAP) earnings, it shifts individual components of the income statement that are meant to be informative for financial statement users (McVay, 2006; Barua, Lin & Sbaraglia, 2010). Secondly, classification shifting does not change future period earnings. Managers move items within the income statement in the current period but do not move items to future periods. Therefore, future period earnings are left unchanged. Although all three methods of earnings management affect the expectations

(6)

6

of future performance, only real earnings management and accrual-based earnings management reduce future (or past) period earnings (McVay, 2006). Finally, when earnings management activities are revealed, activities that violate GAAP generally impose a higher legal cost than activities within the bounds of GAAP (Jiambalvo, 1996). Earnings management activities that do not violate GAAP are less likely to be subject to litigation because it is legal if managers can provide reasonable economic justification for such activities. Figure 1 visualizes two methods of earnings management: accrual-based and real earnings management (Dechow & Skinner, 2000). The two rows indicate what is allowed under GAAP and what is considered fraud. This suggests that accrual-based earnings management is more likely to violate GAAP and therefore has a higher legal risk. Just like real earnings management, classification shifting is also less likely to violate GAAP because the accounting standards on the classification of core expenses can be subjective. Thus, real earnings management and classification shifting have a lower legal risk compared to accrual-based earnings management.

Figure 1 (Dechow & Skinner, 2000, p. 239)

Prior studies investigated whether the earnings management tools are used as substitutes or as complements (Cohen, Dey & Lys, 2008; Zang, 2012). The study by Cohen et al. (2008) investigated the prevalence of both accrual-based and real earnings management activities in

(7)

7

the period prior to the passage of SOX and the period after the passage of SOX. Firms are subject to greater scrutiny since the passage of SOX. They found evidence that after the passage of SOX, accrual-based earnings management declined significantly, while real earnings management increased significantly. Additional analyses examined pre-SOX and post-SOX accrual-based and real earnings management for a subset of firms that are more likely to have managed earnings. Consistent with their full sample results, the subset of firms that are more likely to have managed earnings used significantly less accrual-based earnings management activities and significantly higher real earnings management activities after the passage of SOX, compared to the period prior to the passage of SOX. They suggested that real earnings management increased due to the higher level of scrutiny of accounting practice after the passage of SOX. Accrual manipulation may cause accounting fraud, while the manipulation of real activities does not violate GAAP but sacrifice a firm’s future economic benefits. The research by Zang (2012) extended the literature by examining whether managers use accrual-based and real earnings management as substitutes in managing earnings. The study shows that firms trade off real earnings management versus accrual-based earnings management based on costliness and timing. First, when one activity is relatively more costly, firms engage in more of the other. Second, the manipulation of real economic activities must occur during the fiscal year. After the fiscal year-end, managers still have the opportunity to engage in accrual-based earnings management. The study hypothesized that firms facing greater scrutiny from regulators and auditors have a higher level of real earnings management. Managers may find it harder to convince high-quality auditors than low-quality auditors of their accounting estimates. Besides, managers may believe that accrual-based earnings management is more likely to be detected by auditors when regulators heighten scrutiny of accounting practices. The results indicate that managers are constrained to use accrual-based earnings management due to a higher level of scrutiny of accounting practice after the passage of SOX, and therefore use real earnings manipulation to a greater extent.

Auditors might be more effective in monitoring accrual-based earnings management than real earnings management because the former can be detected by examining a firm’s accounting policies while the latter is caused by managers’ real economic actions. Real activities manipulation is hard to distinguish from normal business activities. Besides, auditors might be more motivated to monitor accrual-based earnings management compared to real earnings management because it is more likely to violate GAAP.

Briefly summarized, in the case of increased monitoring by regulators and auditors, managers prefer real earnings management over accrual-based earnings management because

(8)

8

the manipulation of real activities is harder to detect and less likely to be scrutinized by outside parties. Besides, auditors might be more effective in monitoring accrual-based earnings management. Just like real earnings management, classification shifting is also hard to detect and less likely to be scrutinized by outside parties (e.g., Haw, Ho, & Li, 2011). The accounting standards on the classification of core expenses as special items can be subjective, and shifting core expenses to special items does not change bottom-line net income making this reporting manipulation less likely to be detected by external monitors.

2.2 Effect of financial leverage on earnings management

High leverage has been considered to attract increased monitoring by outside creditors who are interested in assessing and monitoring a firm’s credit risk profile. Increased control by creditors may reduce the incentive to engage in earnings management activities. Prior studies investigated the effect of leverage on accrual-based and real earnings management. A lot of studies related to accrual-based earnings management use leverage as a control variable when they estimate the discretionary accruals (e.g., DeFond & Park, 1997; Becker, DeFond, Jiambalvo & Subramanyam, 1998; Chung, Firth, & Kim, 2005; Lee, Lev & Yeo, 2007; Zhong, Gribbin & Zheng, 2007). For instance, DeFond & Park (1997) examined the effect of current relative pre managed earnings and expected future relative earnings on the behavior of discretionary accruals. They included leverage as a control variable in the Jones (1991) model to estimate discretionary accruals. Their results indicate that leverage is significantly negatively associated with discretionary accruals. The study by Becker et al. (1998) examined the relationship between audit quality and earnings management through discretionary accruals. They estimated the discretionary accruals using the cross-sectional Jones (1991) model. They expected leverage to be associated with discretionary accruals because highly leveraged firms try to avoid debt covenant violation by increasing discretionary accruals. They added leverage as a control variable, and the results indicate that leverage is negatively associated with discretionary accruals. According to Chung et al. (2005), leverage is negatively associated with income-increasing discretionary accruals. They investigated whether low-growth firms with high free cash flow have incentives to steer reported earnings by using income-increasing discretionary accruals. They used debt, defined as total debt divided by total assets, as a control variable in the cross-sectional Jones (1991) model to estimate discretionary accruals. The control variable debt indicates a negative sign, and the coefficient is significant at the .01 level. Similar to the prior studies, Lee et al. (2007) and Zhong et al. (2007) also added leverage as a control variable in the cross-sectional Jones (1991) model to estimate discretionary accruals.

(9)

9

Lee et al. (2007) examined the relationship between organizational structure and earnings management, while Zhong et al. (2007) examined the relationship between outside block holder ownership and discretionary accruals for NYSE-listed U.S. firms. They both concluded that leverage is significantly negatively associated with discretionary accruals.

Prior studies also examined the relationship between leverage and accrual-based earnings management explicitly. For example, Jelinek (2007) examined the effect of leverage increases on accrual-based earnings management. They hypothesized that firms that undergo leverage increases have lower earnings management than consistently high leveraged firms. They argue that leverage increases will reduce earnings management because debt payments reduce the free cash available for spending at managers’ opportunistic behavior (Jensen, 1986), and debt financing faces increased scrutiny of lenders. They tested two groups of firms: (1) firms that undergo large leverage increases and (2) a control group of consistently high leveraged firms. Their results indicate that firms with large leverage increases have lower accruals compared to consistently high leveraged firms. Their findings suggest that leverage increases reduce earnings management. The study by Rodríguez-Pérez & van Hemmen (2010) investigated the relationship among debt, diversification and discretionary accruals in a sample of Spanish firms during the period 1992 to 2002. The study hypothesized that the level of a firm’s leverage is negatively associated with income-increasing earnings management. They suggest that highly leveraged firms face increased monitoring by creditors and bankers, which restrains the use of positive discretionary accruals. Banks may incur in monitoring costs to assess the real quality of debtors. Thus, high leverage constrains opportunistic behavior. Besides, debt restricts the use of discretionary accruals because the firm is subject to financial commitments (Jensen, 1986). They used multiple regression models to test the relationship between leverage and discretionary accruals. Leverage is defined as total debt over total assets. All regression models show that leverage negatively affects discretionary accruals, consistent with their hypothesis.

On the other hand, other researchers have investigated the relationship between leverage and real earnings management. For instance, Bartov (1993) studied whether managers manipulate earnings through the timing of income recognition from the disposal of long-lived assets and investments (e.g., real earnings management). They hypothesized that there is a positive relationship between income from asset sales and leverage as captured by the debt-equity ratio. They suggest that managers try to minimize violation of accounting-based restrictions in debt agreements through earnings manipulation because violating debt covenants is costly. They took leverage as a proxy for the existence and closeness of accounting-based

(10)

10

constraints. The results of their regression analysis indicate that the coefficient of the debt-equity ratio is significant and positive, which supports their hypothesis. Thus, leverage is positively associated with real earnings management. The study by Anagnostopoulou and Tsekrekos (2017) extended the literature by investigating the impact of leverage on the trade-off between accrual-based and real earnings management. They used a sample of firms that are suspected of implementing earnings management to avoid reporting losses, a decrease in earnings from previous-year levels, or missing earnings targets set by ﬁnancial analysts. They suggest that leverage is likely to encourage the motivation to present existing and potential equity investors an improved picture of the firm, in the case when managers want to reduce the negative impact of leverage on stock market performance. Besides, they expect higher levels of leverage or changes in leverage to be accompanied by increased monitoring by outside parties, such as regulators, auditors, and debt and equity investors. Increased control may limit earnings management. Managers prefer real earnings management over accrual-based earnings management because real earnings management is less likely to be detected. Their research results indicate that high levels of leverage (or increases in leverage) positively and significantly affect real earnings management but with no significant effect on accrual-based earnings management. However, they obtained some evidence on the existence of upward accrual-based earnings management for firms experiencing very high debt changes, compared to very low debt changes. Furthermore, they analyzed whether the level of leverage combined with financial health plays any role in the level of accrual-based and real earnings management. The results indicate that financially healthy firms tend to engage in real earnings management when leverage increases, but this tendency is not confirmed for distressed ﬁrms, except the results on abnormal discretionary expenses. The tendency of financially distressed firms to engage in upward real earnings management as financial leverage increases is not observed for any other real earnings management proxy. Besides, the results indicate that financially healthy and high leveraged firms have a greater preference for real earnings management compared to firms with lower leverage.

Based on these prior studies, one would expect a negative relationship between leverage and accrual-based earnings management but a positive relationship between leverage and real earnings management. The preference for real earnings management might be due to the lower possibility of getting caught. Just like real earnings management, classification shifting is also hard to detect and less likely to be scrutinized by outside parties (e.g., Haw, Ho, & Li, 2011). Thus, I expect that leverage will also not mitigate classification shifting.

(11)

11

Lenders could be more interested in debt services rather than in accounting information. Therefore, the impact of increased monitoring by creditors on earnings management (e.g., classification shifting) might depend on the requirements within debt covenants. Most debt covenants are based on GAAP earnings (Li, 2010). Li (2010) investigated contractual definitions of net income and net worth and the cross-sectional variation in deﬁnitions of net income in private debt contracts. Thirty-six percent (36%) of the contracts define net income differently from the GAAP. The majority of the differences are related to transitory earnings. Twenty-two percent (22%) of the contracts exclude extraordinary items in calculating net income. The study by Li (2010) is focused on private debt contracts, while my study focusses on public debt contracts. I expect the findings by Li (2010) to be quite similar to public debt contracts.

When debt contracts define net income similar to GAAP, creditors may not be fooled by classification shifting because classification shifting does not affect GAAP earnings. Therefore, I expect creditors might not be motivated to detect classification shifting. However, when debt contracts define net income similar to Earnings before Interest, Taxes, Depreciation, and Amortization (EBITDA), creditors might be fooled by classification shifting because shifting core expenses to special items increases the reported EBITDA. In this case, creditors might be motivated to detect classification shifting. Debt contracts based on GAAP earnings might also reduce the incentive to engage in classification shifting because it will not help to avoid debt covenant violation. Debt contracts based on EBITDA might increase the motivation to engage in classification shifting to avoid debt covenant violation.

Fan, Thomas & Yu (2019) investigated whether managers are more likely to engage in classification shifting when loan contracts that contain debt covenants are close to violating their EBITDA-related covenants. Their results show that managers are more likely to engage in classification shifting when the firm is close to a violation of at least one EBITDA-related covenant. Besides, they found little evidence of classification shifting when EBITDA-related covenants are not close to violation or firms with non-EBITDA-related covenants close to a violation.

2.3 Hypothesis development

Based on prior research, one would expect a negative relationship between leverage and accrual-based earnings management but a positive relationship between leverage and real earnings management. I expect that highly leveraged firms, facing increased scrutiny by outside creditors, do not eliminate earnings management activities altogether, but will change

(12)

12

managers’ preference for different earnings management devices. The preference for real earnings management might be due to the lower possibility of getting caught. Just like real earnings management, classification shifting is also hard to detect and less likely to be scrutinized by outside parties (e.g., Haw, Ho, & Li, 2011). Thus, I expect that leverage will also not mitigate classification shifting.

Furthermore, although higher debt results in increased monitoring, creditors might care less about classification shifting when debt covenants are based on GAAP earnings. However, according to Li (2010), debt covenants may be more specific than GAAP. When debt covenants are based on EBITDA, creditors might be more motivated to detect classification shifting.

Overall, I expect that leverage will not mitigate classification shifting. All this leads to the following hypothesis:

(13)

13

3. Research design

3.1 McVay model

To examine whether U.S. firms engage in classification shifting, I follow the regression model introduced by McVay (2006). If firms engage in classification shifting, it is expected that they have a higher than expected level of core earnings in year t and a lower than expected change in core earnings in year t+1. The unexpected core earnings in year t are expected to be increasing with special items in year t, and the unexpected change in core earnings from year t to t+1 is expected to be declining in special items in year t (McVay, 2006). McVay (2006) developed a model of expected core earnings in levels (to examine year t) and in changes (to examine year t+1). This study uses the levels model of McVay (2006).

Levels model:

(1) CEₜ = β₀ + β₁CE t-1 + β₂ATOₜ + β₃ACCRUALSt-1 + β₄ACCRUALSₜ + β₅ΔSALESₜ +

β₆NEG_ΔSALESₜ + εₜ

Each of the variables is described below in table 1.

The levels model includes the variable lagged core earnings (CEt-1) to capture core

earnings persistence over time. The variable asset turnover (ATOₜ) is included to control for the inverse relationship between asset turnover and profit margin. Including the asset turnover ratio is also important because firms that have large income-decreasing special items are likely to make changes to their operating strategy, possibly altering the firms' mix of margin and turnover (McVay, 2006). The variable prior-year operating accruals (ACCRUALSt-1) is added to capture

the information content of past accruals. According to Sloan (1996), holding earnings constant, accrual levels are an explanatory variable for future performance. Current year accruals (ACCRUALSₜ) are included to control for extreme performance. DeAngelo, DeAngelo, and Skinner (1994) find that extreme performance is correlated with changes in accrual levels. Controlling for accruals allows for a stronger prediction of abnormal core earnings associated with classification shifting. Sales growth (ΔSALESₜ) is included as an explanatory variable. The relationship between changes in sales and core earnings is not expected to be constant due to the impact of sales growth on fixed costs. Finally, the variable negative sales (NEG_ΔSALESₜ) is added to allow for different slopes for sales increase and decrease.

(14)

14

By using the Compustat database for all U.S. firms, the first equation is regressed to obtain the coefficients β0, β1, β2, β3, β4, β5, and β6. The coefficients are plugged into the equation to calculate the expected core earnings [E(CE)ₜ]. Equations 1 is run cross-sectionally for each industry-year and excluding firm i.

After that, the unexpected core earnings (UE_CEₜ) are calculated by subtracting the estimated core earnings [E(CE)ₜ] from the reported core earnings (CEₜ), as seen in equations 2. The estimated core earnings are derived from equation 1.

Levels model:

(2) UE_CEₜ = CEₜ - E(CE)ₜ

3.2 Hypothesis testing

The next step is to determine the association between the unexpected core earnings and special items. To test my hypothesis, I include the interaction of non-recurring items with financial leverage (LEV). The regression takes the following form:

Levels model:

(3) UE_CEₜ = αₒ + α₁%SIt + α₂LEV + α3%SIt × LEV + α4SIZ + α5CFO + α6ROA +

α7MBV

Similar to Barua, Lin & Sbaraglia (2010) and Zalata & Roberts (2016), to control for performance the following control variables are added to the model: firm size (SIZ), cash flow from operations (CFO), return on assets (ROA), and market to book value (MBV). Barua et al. (2010) compared firm characteristics between firm-year observations with and without discontinued operations. They observed significant differences in the variables: firm size, cash flow from operations, return on assets, and book to market value.

To test the hypothesis, the variables of interest are %SI and %SI × LEV. Hypothesis 1 predicts α1 to be significantly positive to confirm that firms engage in classification shifting.

Besides, hypothesis 1 predicts α3 to be significantly positive to confirm that financial leverage

(15)

15

Table 1 Variable definitions

Variable Definition

CEₜ Core Earnings (before Special Items and Depreciation), calculated as (Sales - Cost of Goods Sold - Selling, General, and Administrative Expenses) / Sales. Cost of Goods Sold and Selling, General, and Administrative Expenses exclude Depreciation and Amortization, as determined by Compustat.

ATOₜ Asset Turnover Ratio, defined as Salesₜ/((NOAₜ + NOAₜ-1)/ 2), where Net

Operating Assets (NOA), is equal to the difference between Operating Assets and Operating Liabilities. Operating Assets = Total Assets - Cash - Short-Term Investments. Operating liabilities = Total Assets - Total Debt - Book Value of Common and Preferred equity - Minority Interests. The average net operating assets are required to be positive.

ACCRUALSₜ Operating Accruals, calculated as (Net Income before Extraordinary Items - Cash From Operations)/Sales.

ΔSALESₜ Percent Change in Sales, calculated as (Salesₜ - Salesₜ-1)/Salesₜ-1.

NEG_ΔSALESₜ Percent Change in Sales (ΔSALESₜ) if ΔSALESₜ is less than 0, and 0 otherwise.

%SI Income-Decreasing Special Items as a Percentage of Sales, calculated as (Special Itemst x -1)/Salest when Special Items are income-decreasing,

and 0 otherwise.

LEV Leverage is deﬁned as the sum of short- and long-term debt, scaled by total assets.

SIZ Firm size is calculated as the log of total assets at the end of the ﬁscal year.

CFO Cash flow from operations divided by lagged total assets.

ROA Return on assets deﬁned as income before extraordinary items, scaled by total assets.

MBV Market-to-book ratio. Market value is calculated by multiplying the closing price at ﬁscal year-end by the number shares outstanding.

3.3 Sample selection

The sample of U.S. firms is derived from the Compustat Capital IQ database (North America file) for the years 1988-2018. The sample begins in 1988 because, as of this year, Compustat reported a newly required Cash From Operations, which allows calculating Accruals as Earnings less Cash From Operations as prescribed by Hribar and Collins (2002). The newly reported Cash From Operations is essential because Accruals measured using the balance-sheet approach causes errors, mainly for firms that have had Merger and Acquisition (M&A) activities (Hribar and Collins, 2002). Because the variables require one year of lagged data, the actual years examined are 1989-2018.

(16)

16

I deleted firms that have less than $1 million in sales to avoid the creation of outliers caused by scaling variables by sales. Also, to ensure that years are comparable, I eliminated all firms that had a fiscal-year-end change from t-1 to t or from t to t+1. Besides, I require at least 15 observations per industry per fiscal year to estimate the expected core earnings. Industries are classified following Fama and French (1997). Finally, all variables are winsorized by year at the 1st and 99th percentiles.

Consistent with Elliott and Hanna (1997) and McVay (2006), I assigned a zero to the variable Special Items if that data item is missing. Also, the variable Extraordinary Items and Discontinued Operations (Statement of Cash Flows), which is used to calculate Accruals, is assigned zero if that data item is missing.

(17)

17

4. Empirical results

4.1 Descriptive statistics

Table 2 below reports the descriptive statistics (mean, median 25th and 75th percentile values and standard deviation) for the main variables used to determine the association between financial leverage and classification shifting. The mean (median) core earnings scaled by sales for all U.S. firms is 0.0031 (0.1217). The mean (median) of income-decreasing special items is 2.36 percent (0.00 percent) of sales. Consistent with other studies in the field (e.g., Cohen et al., 2008; Anagnostopoulou & Tsekrekos, 2017), the leverage median is 20.89 percent.

Table 2 Descriptive statistics

Variable Mean Median

Standard Deviation 25% 75% CEt 0.0031 0.1217 0.7529 0.0338 0.2542 UE_CEt -0.0061 -0.1041 0.5873 -0.2239 -0.0106 CEt-1 0.0377 0.1256 0.6392 0.0409 0.2570 ATOₜ 1.9506 1.4583 6.0789 0.4624 2.8882 ACCRUALSt-1 -0.1407 -0.0613 0.3859 -0.1591 -0.0086 ACCRUALSt -0.1573 -0.0634 0.4483 -0.1670 -0.0092 ΔSALESₜ 17.34% 7.63% 0.4803 -2.78% 23.07% NEGΔSALESₜ -3.39% 0.00% 0.0956 0.00% 0.00% %SIt 2.36% 0.00% 0.0956 0.00% 0.3269% LEV 0.2734 0.2089 0.2829 0.0470 0.4010 SIZ 5.6628 5.6668 2.5071 3.8485 7.3913 CFO 0.0385 0.0641 0.2065 0.0062 0.1268 ROA -0.0532 0.0163 0.2946 -0.0310 0.0594 MBV 2.5883 1.6583 4.8888 0.9578 3.0072 SI 14.4953 0.00 65.4561 0.00 0.55

Table 3 distinguishes firms with and without large income-decreasing special items. Income-decreasing special items are defined large if the income-decreasing special items are above 5 percent of sales. The mean core earnings are significantly lower for firms with large income-decreasing special items compared to firms without large income-decreasing special items (-0.3734 versus 0.0482), which is consistent with the results of prior studies showing that

(18)

18

firms reporting large special items tend to be poor performers (DeAngelo et al. 1994; McVay, 2006). According to McVay (2006), firms with large income-decreasing special items have a greater opportunity to engage in classification shifting. Table 3 indicates that firms with large income-decreasing special items have significantly higher financial leverage than those firms with small income-decreasing special items. Besides, accruals are significantly more negative for firms with large special items.

Table 3 Descriptive statistics subgroups

Firms without Income-Decreasing Special Items ≥ 5% of sales

Firms with Income-Decreasing Special Items ≥ 5% of sales

P-value for statistical difference between firms with and

without Special Items ≥ 5% of sales

Variable Mean Median Mean Median t-test

Wilcoxon Rank Sum Test CEt 0.0482 0.1289 -0.3734 0.0318 0.0000 0.0000 UE_CEt -0.0506 -0.1121 0.3444 0.0158 0.0000 0.0000 CEt-1 0.0728 0.1303 -0.2539 0.0753 0.0000 0.0000 ATOₜ 1.9978 1.5033 1.5532 1.1475 0.0000 0.0000 ACCRUALSt-1 -0.1212 -0.0576 -0.2965 -0.1098 0.0000 0.0000 ACCRUALSt -0.1104 -0.0536 -0.5324 -0.2524 0.0000 0.0000 ΔSALESₜ 17.41% 8.05% 16.72% 2.01% 0.0519 0.0000 NEGΔSALESₜ -2.86% 0.00% -8.94% 0.00% 0.0000 0.0000 %SIt 0.33% 0.00% 23.88% 13.31% 0.0000 0.0000 LEV 0.2665 0.2056 0.3321 0.2428 0.0000 0.0000 SIZ 5.7215 5.7357 5.1651 4.9938 0.0000 0.0000 CFO 0.0504 0.0697 -0.0567 0.0120 0.0000 0.0000 ROA -0.0176 0.0232 -0.3532 -0.1563 0.0000 0.0000 MBV 2.6062 1.6819 2.4452 1.3969 0.0000 0.0000 SI 7.9951 0.00 83.2827 13.471 0.0000 0.0000

Table 4 reports the Pearson/Spearman correlations among the main variables used to test my hypothesis. The Pearson (Spearman) correlations are below (above) the diagonal. All variables are winsorized at 1st and 99th percentiles. The results indicate a high correlation between core earnings and lagged core earnings, consistent with the theory that core earnings

(19)

19

(20)

20

are persistent over time. Special items as a percentage of sales are negatively correlated with core earnings and positively correlated with unexpected core earnings, consistent with the findings by McVay (2006) that unexpected core earnings are expected to be increasing with special items in year t when managers engage in classification shifting. The Pearson correlation shows that financial leverage is positively correlated with unexpected core earnings.

4.2 Regression results

Table 5 represents the regression results for the model of expected core earnings in levels (Equation 1). The overall adjusted R2 is 73,58%. The one-tailed p-values show that all

Table 5 Model of Expected Core Earnings (levels)

Dependent variable: CEt Independent variable Predicted sign Mean coefficient (one-tailed p-value) Median coefficient Percent significant (p-value ≤ 0.10, one-tailed test) Percent with sign in the predicted direction Intercept 0.0315 (0.0000) CEt-1 + 0.8178 (0.0000) 0.8223 100.00% 100.00% ATOₜ + 0.0012 (0.000013) 0.0005 28.26% 63,04% ACCRUALSt-1 - -0.1822 (0.0000) -0.1616 89.13% 95,65% ACCRUALSt + 0.2505 (0.0000) 0.2024 95.65% 100.00% ΔSALESₜ + 0.1645 (0.0000) 0.1289 82.61% 95.65% NEGΔSALESₜ + 0.4209 (0.0000) 0.2988 93.48% 97.83% Adjusted R2 73.58%

There are 151,335 ﬁrm-year observations for the period 1989 to 2018. The regression is estimated cross-sectionally by industry and year. All variables are winsorized at 1st and 99th percentiles. The percent significant column indicates how many industry regressions have significant variables. The last column shows the percentage of regressions in which the variables conform to the predicted sign direction.

CEₜ = β₀ + β₁CEt-1 + β₂ATOₜ + β₃ACCRUALSt-1 + β₄ACCRUALSₜ + β₅ΔSALESₜ +

(21)

21

coefficients are statistically significant. The results indicate that the variable lagged core earnings (CEt-1) is a strong predictor of core earnings (CEt), with a mean coefficient of 81.78%

and a median coefficient of 82.23%.

The variable asset turnover (ATOₜ) is significantly positively associated with core earnings. However, only 28 percent of the industry regressions are significant, which indicates a relatively weak association between asset turnover and core earnings. According to Soliman (2004), the weak relationship between profit margin and asset turnover is driven by industry associations. Besides, the positive direction of the variable asset turnover is not similar to McVay’s results. However, table 5 indicates that 63 percent of the industry regressions show a negative association between asset turnover and core earnings, consistent with McVay (2006).

The variable lagged accruals (ACCRUALSt-1) has a negative coefficient of -0.16, and

the variable current accruals (ACCRUALSt) has a positive coefficient of 0.20, consistent with

the theory that higher levels of accruals have lower earnings persistence. Consistent with McVay (2006), the coefficient for negative sales growth (NEGΔSALESₜ) is significantly larger than the coefficient for sales growth (ΔSALESₜ). According to Anderson, Banker & Janakiraman (2003), the slope coefficient on sales growth is larger for firms experiencing a decline in sales.

Equation 1 is used to determine the expected core earnings by multiplying the regression coefficients by the actual values for each firm. After that, the unexpected core earnings are determined by subtracting the expected core earnings from the reported core earnings (equation 2). The unexpected core earnings are used in equation 3 as the dependent variable to determine if firms engage in classification shifting.

Before investigating whether leverage affects classification shifting, I started with a basic regression to check whether there is a positive relationship between unexpected core earnings and special items (table 6). Table 6 considers three samples, (1) all Compustat firms, (2) firms with non-zero decreasing special items, and (3) firms with large income-decreasing special items of at least 5 percent of sales. As expected, special items are positively associated with unexpected core earnings (α₁ = 0.1097). Besides, the adjusted R2 increases as the sample narrowed down to firms with a greater opportunity to engage in classification shifting (41.56% versus 44.83%), consistent with McVay (2006).

Table 7 summarizes the regression results of equation 3 to examine whether financial leverage affects the level of classification shifting. Just as table 6, table 7 also indicates that special items are positively associated with unexpected core earnings (α₁ = 0.0513). One

(22)

22

Table 6 Regression of Unexpected Core Earnings on Special Items.

Dependent variable: UE_CEt

Independent variable Predicted sign All Compustat firms Non-Zero Income-Decreasing Special Items Income-Decreasing Special Items ≥ 5% of sales Intercept 0.0675 (19.68)*** 0.0261 (4.27)*** -0.0299 (-1.62) %SIt + 0.1097 (46.42)*** 0.2753 (71.00)*** 0.2949 (43.03)*** SIZ - -0.0818 (-36.76)*** -0.0685 (-19.25)*** -0.0436 (-6.06)*** CFO - -0.2759 (-102.86)*** -0.3019 (-75.74)*** -0.3991 (-55.59)*** ROA - -0.3466 (-113.77)*** -0.2199 (-45.02)*** -0.1311 (-14.98)*** BMV + 0.0535 (25.79)*** 0.0360 (11.51)*** 0.0218 (3.69)*** Adjusted R2 41.56% 46.69% 44.83%

The full sample consists of 135,964 observations. T-statistics are shown in parentheses. ***, **, and * indicate statistical signiﬁcance at the 1%, 5%, and 10% levels (two-tailed),

respectively. All variables are winsorized at 1st and 99th percentiles. UE_CEₜ = αₒ + α₁%SIt + α2SIZ + α3CFO + α4ROA + α5MBV

standard deviation of increase in special items increases the unexpected core earnings (scaled by sales) by 49 basis points. Determined by multiplying the coefficient from table 7 (0.0513) by the standard deviation of special items from table 2 (0.0956). As expected, the coefficient of the variable special items increases as the sample narrows down to firms with a greater opportunity to engage in classification shifting (0.0513 versus 0.0864), consistent with McVay (2006). One standard deviation of increase in special items increases the unexpected core earnings (scaled by sales) by 83 basis points for firms with large income-decreasing special items of at least 5 percent of sales.

To examine whether financial leverage affects the level of classification shifting, I focus on the interaction between %SI and LEV. Table 7 shows that the coefficient of %SI × LEV is positive (α3 = 0.0309) and significant at 1%, supporting hypothesis 1. Thus, financial leverage

(23)

23

is more prevalent among highly leveraged firms. Thus, hypothesis 1 can be accepted.

Table 7 Analysis impact Financial Leverage on Classification Shifting.

Independent variable Predicted sign All Compustat firms Non-Zero Income-Decreasing Special Items Income-Decreasing Special Items ≥ 5% of sales Intercept 0.1111 (31.48)*** 0.1011 (15.88)*** 0.1430 (7.49)*** %SIt + 0.0513 (24.36)*** 0.0841 (25.82)*** 0.0864 (14.01)*** LEV - -0.1113 (-49.77)*** -0.1392 (-37.46)*** -0.2251 (-26.22)*** %SIt × LEV + 0.0309 (12.91)*** 0.1227 (29.08)*** 0.1944 (21.27)*** SIZ - -0.0629 (-28.24)*** -0.0458 (-12.50)*** -0.0072 (-0.97) CFO - -0.2645 (-98.82)*** -0.2993 (-73.26)*** -0.3989 (-54.14)*** ROA - -0.4188 (-139.39)*** -0.3494 (-71.59)*** -0.2608 (-29.57)*** BMV + 0.0426 (20.47)*** 0.0309 (9.63)*** 0.0083 (1.35) Adjusted R2 _42.02% _44.35% _42.17% Number of observations 135,964 54,737 15,907

The full sample consists of 135,964 observations. T-statistics are shown in parentheses.

***, **, and * indicate statistical signiﬁcance at the 1%, 5%, and 10% levels (two-tailed), respectively. All variables are winsorized at 1st_{and 99}th_percentiles.

UE_CEₜ = αₒ + α₁%SIt + α₂LEV + α3%SIt × LEV + α4SIZ + α5CFO + α6ROA + α7MBV

4.3 Sensitivity analyses

I conducted a sensitivity analysis by using other definitions of the independent variables LEV and %SIt × LEV. The main analysis, following prior studies, defined leverage as the sum

of short- and long-term debt scaled by total assets. I chose an alternative definition of leverage, the sum of short- and long-term debt scaled by total equity, for the sensitivity analysis. Debt

(24)

24

divided by equity is also a widely used definition of leverage. Table 8 indicates that the results for all variables are quite similar.

Table 8 Sensitivity analysis

Independent

variable Predicted sign

All Compustat firms – Debt/Assets

All Compustat firms – Debt/Equity Intercept 0.1111 (31.48)*** 0.0536 (15.51)*** %SIt + 0.0513 (24.36)*** 0.0548 (26.15)*** LEV - -0.1113 (-49.77)*** -0.0965 (-38.06)*** %SIt × LEV + 0.0309 (12.91)*** 0.0352 (15.19)*** SIZ - -0.0629 (-28.24)*** -0.0629 (-28.01)*** CFO - -0.2645 (-98.82)*** -0.2667 (-100.06)*** ROA - -0.4188 (-139.39)*** -0.3972 (-144.30)*** BMV + 0.0426 (20.47)*** 0.0890 (38.07)*** Adjusted R2 42.02% 41.59% Number of observations 135,964 135,964

The full sample consists of 135,964 observations. T-statistics are shown in parentheses.

***, **, and * indicate statistical signiﬁcance at the 1%, 5%, and 10% levels (two-tailed), respectively. All variables are winsorized at 1st and 99th percentiles.

(25)

25

5. Discussion and conclusion

5.1 Findings

This study examines the effect of financial leverage on classification shifting. High leverage has been considered to attract increased monitoring by outside creditors. I argued that highly leveraged firms, facing increased scrutiny by outside creditors, do not eliminate earnings management activities altogether, but will change managers’ preference for different earnings management devices. Based on prior research, one would expect a negative relationship between leverage and accrual-based earnings management (e.g., Jelinek, 2007; Rodríguez-Pérez & van Hemmen, 2010), but a positive relationship between leverage and real earnings management (e.g., Bartov, 1993; Anagnostopoulou and Tsekrekos, 2017). The preference for real earnings management might be due to the lower possibility of getting caught. Just like real earnings management, classification shifting is also hard to detect and less likely to be scrutinized by outside parties.

To examine the relation between financial leverage and classification shifting, data is obtained from the Compustat Capital IQ database (North America file) for the years 1988-2018. Classification shifting is measured using the regression model introduced by McVay (2006). If firms engage in classification shifting, the unexpected core earnings are expected to increase along with special items.

I established that the unexpected core earnings increase along with special items in year t, indicating that managers shift core expenses to special items in order to increase core earnings, consistent with the findings by McVay (2006). Besides, the results indicate that the likelihood to engage in classification shifting increases when firms have a greater opportunity to engage in classification shifting. Furthermore, the results show a significant positive association between leverage and classification shifting, indicating that classification shifting is more prevalent among highly leveraged firms. Thus, hypothesis 1 can be accepted. The findings indicate that despite an increase in monitoring by creditors, firms engage in classification shifting. The results are consistent with the literature that argues that highly leveraged firms, facing increased monitoring by creditors, still engage in earnings management tools that are hard to detect and less likely to be scrutinized by outside parties.

This study contributes to the existing literature in two ways. First, this study is the first to investigate the relationship between leverage and classification shifting. Second, high leverage has been considered to attract increased monitoring by outside creditors. Prior studies have shown a negative effect of leverage on accrual-based earnings management, but a positive

(26)

26

impact of leverage on real earnings management. This study is the first that shows a positive association between leverage and classification shifting, indicating that classification shifting is more prevalent among highly leveraged firms. Therefore, this research further supports and extends the literature that argues that highly leveraged firms, facing increased monitoring by creditors, still engage in earnings management tools that are hard to detect and less likely to be scrutinized by outside parties. This research has implications for shareholders who strive to avoid investing in a firm with deceptive financial statements and auditors who are trying to assess a firm's potential for fraudulent behavior.

5.2 Limitations and further research

The study is subject to limitations. In particular, although I present a sensitivity analysis indicating that the results are robust, classification shifting and leverage are endogenous. Future research could address this concern by examining a robust model that corrects for endogenous variables. The method of instrumental variable approach, considering endogeneity, will lead to a more robust model.

(27)

27

References

Anagnostopoulou, S. C., & Tsekrekos, A. E. (2017). The effect of ﬁnancial leverage on real and accrual-based earnings management. Accounting and Business Research, 47(2), 191-236.

Anderson, M., Banker, R., & Janakiraman, S. (2003). Are selling, general, and administrative costs "sticky"? Journal of Accounting Research, 41(1), 47-63.

Bartov, E. (1993). The timing of assets sales and earnings manipulation. Accounting

Review, 68(4), 840–855.

Barua, A., Lin, S., & Sbaraglia, A. S. (2010). Earnings management using discontinued operations. The Accounting Review, 85, 1485-1509.

Becker, C. L., DeFond, L. M., Jiambalvo, J., & Subramanyam, K. R. (1998). The effect of audit quality on earnings management. Contemporary Accounting Research, 15(1), 1–24.

Chung, R., Firth, M., & Kim, J. B. (2005). Earnings management, surplus free cashﬂow, and external monitoring. Journal of Business Research, 58(6), 766–776.

Cohen, D., Dey, A., & Lys, T. (2008). Real and accrual-based earnings management in the pre- and post- Sarbanes Oxley period. Accounting Review, 83(3), 757–787.

DeAngelo, H., DeAngelo, L., & Skinner, D. (1994). Accounting choice in troubled companies. Journal of Accounting and Economics, 17(1), 113-143.

Dechow, P. M., & Skinner, D. J. (2000). Earnings management: Reconciling the views of accounting academics, practitioners, and regulators. Accounting Horizons, 14(2), 235-250.

DeFond, M., & Park, C. (1997). Smoothing income in anticipation of future earnings.

Journal of Accounting and Economics, 23(2), 115–139.

Elliott, J. A., & Hanna, J. D. (1997). Repeated accounting write-offs and the information content of earnings. Journal of Accounting Research, 34, 135–156.

Fama, E. F., & French, K. R. (1997). Industry costs of equity. Journal of Financial

(28)

28

Fan, Y., Thomas, W. B., & Yu, X. (2019). The impact of financial covenants in private loan contracts on classification shifting. Management Science, 65(8), 3637–3653.

Financial Accounting Standards Board (FASB). (1984). Statement of Financial Accounting Concepts No. 5.

Haw, I. M., Ho, S. S., & Li, A. Y. (2011). Corporate governance and earnings management by classiﬁcation shifting. Contemporary Accounting Research, 28(2), 517–553.

Healy, P., & Wahlen, J. (1999). A Review of the Earnings Management Literature and Its Implications for Standard Setting. Accounting Horizons, 13(4), 365–383.

Hribar, P., & Collins, D. W. (2002). Errors in estimating accruals: implications for empirical research. Journal of Accounting Research, 40, 105–134.

Jelinek, K. (2007). The Effect of Leverage Increases on Earnings Management. Journal

of Business & Economic Studies, 13(2), 24-46.

Jensen, M. C. (1986). Agency costs of free cash flow, corporate finance, and takeovers.

The American Economic Review, 76(2), 323–329.

Jensen, M. C., & Meckling, W. H. (1976). Theory of the firm: Managerial behavior, agency costs and ownership structure. Journal of Financial Economics, 3(4), 305–360.

Jiambalvo, J. (1996). Discussion of “causes and consequences of earnings manipulation: an analysis of firms subject to enforcement actions by the sec.” Contemporary Accounting

Research, 13(1), 37–47.

Jones, J. (1991). Earnings management during import relief investigations. Journal of

Accounting Research, 29(2), 193–228.

Lee, K.W., Lev, B., & Yeo, G. (2007). Organizational structure and earnings management. Journal of Accounting, Auditing and Finance, 22(2), 293–331.

Li, n. (2010). Negotiated measurement rules in debt contracts. Journal of Accounting

Research, 48(5), 1103–1144.

McVay, S. E. (2006). Earnings management using classiﬁcation shifting: An examination of core earnings and special items. The Accounting Review, 81(3), 501–531.

(29)

29

Richardson, V. (2000). Information asymmetry and earnings management: Some evidence. Review of Quantitative Finance and Accounting, 15(4), 325-347.

Rodríguez-Pérez, G., & van Hemmen, S. (2010). Debt, diversiﬁcation and earnings management. Journal of Accounting and Public Policy, 29(2), 138–159.

Sloan, R. (1996). Do stock prices fully reflect information in accruals and cash flows about future earnings? The Accounting Review, 71(3), 289-315.

Soliman, M. (2004). Using industry-adjusted DuPont analysis to predict future profitability. Working paper, Stanford University

Zalata, A., & Roberts, C. (2016). Internal corporate governance and classification shifting practices: an analysis of u.k. corporate behavior. Journal of Accounting, Auditing and

Finance, 31(1), 51–78.

Zang, A. Y. (2012). Evidence on the trade-off between real activities manipulation and accrual-based earnings management. Accounting Review, 87(2), 675–703.

Zhong, K., Gribbin, D., & Zheng, X. (2007). The effect of monitoring by outside blockholders on earnings management. Quarterly Journal of Business and Economics, 46(1), 37–60.

THE EFFECT OF FINANCIAL LEVERAGE ON CLASSIFICATION SHIFTING