Executive Influence Over Tax Expense: The Interactive Role of Incentives and Opportunities

Transcription

1 Executive Influence Over Tax Expense: The Interactive Role of Incentives and Opportunities Erik L. Beardsley* University of Notre Dame Mehmet C. Kara Texas A&M University Connie D. Weaver Texas A&M University Preliminary Draft Please do not cite or circulate without permission *Corresponding author Erik Beardsley May 26, 2017 Acknowledgments: Beardsley gratefully acknowledges support from the Mendoza College of Business. Weaver gratefully acknowledges financial support provided by the KPMG professorship and the Mays Business School.

2 Executive Influence Over Tax Expense: The Interactive Role of Incentives and Opportunities Abstract: We examine managerial influence of tax expense in two related, but distinct settings: 1) earnings management using tax expense, and 2) long-run tax avoidance. Prior research has examined the effect of firm characteristics or managerial incentives in isolation, but has largely ignored the possibility of an interactive effect. In an earnings management setting, we predict and find that both managerial incentives and firm opportunities, as well as their interaction, are associated with a greater propensity for firms to manage earnings using tax expense. Additionally, we find an interactive effect between managerial incentives and opportunities in a tax avoidance setting, a result that is new to the literature. We contribute to the literature by providing insight into the effect of managerial incentives with varying levels of opportunities, which is important to consider in studies that examine the causes and consequences of executive influence over corporate policy decisions. Keywords: Tax expense; Tax avoidance; Earnings management; Executive incentives

3 I. INTRODUCTION Recent academic research documents that individual executives have incremental effects on firms level of tax avoidance (e.g., Dyreng et al. 2010; Chyz 2013), respond to incentives to avoid taxes (Rego and Wilson 2012; Armstrong et al. 2012; Gaertner 2014) and use the tax accounts as a final opportunity to manage earnings (Dhaliwal et al. 2004). These studies suggest a certain level of managerial influence over the tax expense reported in the financial statements. 1 However, an open question in the literature is the extent to which managers respond to incentives when they have varying levels of opportunity for tax planning or discretion that can lead to accounting manipulation. 2 In this study, we investigate this question by examining how the interaction between managerial incentives ( incentives ) and firm characteristics that create opportunities to influence taxes ( opportunities ) affect executives influence on tax expense. It is important to examine the interaction between managerial incentives and opportunities because managers could respond to incentives in different ways depending on opportunities, resulting in vastly different levels of influence on taxes. Although prior research acknowledges that certain firm characteristics are associated with opportunities for tax avoidance and manipulation (e.g., Collins et al. 1998; Rego 2003; Mills and Newberry 2004; Robinson et al. 2010), we are aware of no study that considers how incentives and opportunities interact to produce different tax outcomes. Our study seeks to fill this gap in the literature by providing insight into the conditions under which executives respond to incentives. Our study will therefore be informative to boards of directors that may be interested in the effect of executive incentives in 1 We define executive influence over the tax accounts as purposeful actions on the part of the manager that result in desired tax outcomes, whether by tax planning or accounting manipulations. These desired tax outcomes can include a lower tax liability (i.e., lower effective tax rate) or the manipulation of tax expense to meet an earnings target. 2 We acknowledge that all firms have some level of opportunity to influence taxes; however, certain firm characteristics provide greater opportunities to some firms compared to others. 1

4 firms with differing opportunity sets. Further, policy makers and regulators may be interested in our study because we examine different situations that lead to earnings management or tax avoidance. We examine executive influence in two settings. First, we examine executive influence on the propensity for firms to use tax expense as a tool to manage earnings. Managers can manage earnings by exercising discretion over tax reporting choices late in the year (Dhaliwal et al. 2004). Second, we examine executive influence on firms overall level of tax avoidance, which managers accomplish through a broad range of tax planning. We examine both earnings management using tax expense as well as overall tax avoidance because they represent two related, but distinct, ways that managers can influence tax expense. Using tax expense to manage earnings represents a manipulation in reported tax expense to meet an earnings target (e.g., Dhaliwal et al. 2004) and firms tax avoidance broadly represents the level of explicit tax liability the firm reports (e.g., Hanlon and Heitzman 2010). While manipulation and tax avoidance represent distinct dimensions of managerial influence over tax expense, our predictions are similar for both settings. 3 We expect that incentives influence the extent of managerial discretion over tax expense, consistent with managers enhanced incentive to take riskier tax positions (Armstrong et al. 2013). 4 We also expect that firm-specific opportunities are associated with the extent of managerial influence over tax expense, consistent with managers using opportunities to tax plan or take advantage of firm complexity to influence tax expense. Further, we expect that executives in firms with greater 3 However, we acknowledge that our results may differ between settings (i.e., tax avoidance and earnings management settings) because incentives or opportunities may be more dominant in one setting compared to the other. 4 To investigate or research question, we examine three managerial incentives: CEO portfolio vega, portfolio delta, and after-tax compensation incentives. See section IV for details. 2

5 opportunities to influence tax expense will respond to incentives to a greater extent than their counterparts; that is, we predict an interactive effect. Our earnings management results suggest that both managerial incentives and firmspecific opportunities are associated with an increased propensity to manage earnings using tax expense. Further, our results are consistent with an interactive effect, where managers are more likely to manage earnings using tax expense when they have both higher incentives and greater opportunities to influence tax expense. In addition, the tax avoidance results demonstrate an interactive effect, where managers with both the high incentive and high opportunity to avoid taxes are associated with higher levels of tax avoidance. We note, however, that our tax avoidance results hold when examining cash ETRs but not GAAP ETRs. This study makes several contributions to the literature. First, our study contributes to the literature that examines the effects of executive compensation on corporate tax outcomes (e.g., Desai and Dharmapala 2006; Armstrong et al. 2012; Rego and Wilson 2012). We demonstrate that the extent to which managers respond to compensation-based incentives depends on the level of firm characteristics that create opportunities. Thus, we contribute to research that examines the heterogeneity in corporate tax rates (Dyreng et al. 2008). Second, we contribute to the literature that examines earnings management using tax expense (Dhaliwal et al. 2004; Cook et al. 2008; Comprix et al. 2012; Christensen et al. 2015). Our study indicates that both managerial incentives and opportunities are significant determinants of discretion over the tax accounts to achieve earnings targets. Therefore, our study responds to the call for additional documentation of earnings management through accounting for income taxes and contributes to our understanding of why and how earnings management using tax expense is accomplished. (Graham et al. 2012, page 423). Finally, this study promotes a framework for studying the interaction between 3

6 incentives and opportunities, which is important to consider in studies that examine the causes and consequences of executive influence over corporate policy decisions. This paper proceeds as follows. Section II discusses background and related literature. Section III develops our hypotheses. Section IV discusses the research design. Section V discusses our sample selection and descriptive statistics. Section VI and VII present the results of our earnings management tests and tax avoidance tests, respectively, and section VIII concludes. II. BACKGROUND AND RELATED LITERATURE Earnings Management Using Tax Expense versus Tax Avoidance In our study, we examine earnings management using tax expense as well as corporate tax avoidance as two related, yet distinct, avenues for managerial influence. Prior studies have examined factors associated with corporate tax avoidance, an issue highlighted by Dyreng et al. (2008), who find large variation in ETRs among firms. We note that this line of literature examines tax avoidance broadly, defined as a reduction in the level of a firm s explicit tax liability (e.g., Hanlon and Heitzman 2010). Earnings management through tax accounts, however, is the purposeful reduction in tax expense that results in achieving an earnings target, such as analysts forecast (Dhaliwal et al. 2004). Although firms propensity to use tax expense to meet an earnings target and their level of avoidance may be related, there are important differences between these concepts. For example, a firm may have a lower tax rate this year compared to last due to a decrease in income from hightax jurisdictions relative to income from low-tax jurisdictions. Although this would potentially decrease the tax rate, it would not necessarily represent purposeful manipulation of tax expense to beat analysts expectations. However, the tax avoidance literature is related to our study to the extent that tax avoidance is a function of opportunities to avoid tax and intentional actions taken 4

7 by firm managers to reduce ETRs. In terms of the level of tax avoidance, several academic studies have focused on either 1) firm characteristics associated with tax avoidance, or 2) the association between managerial incentives and corporate tax policy. We discuss studies that have examined earnings management using tax expense and studies that have examine corporate tax avoidance in more detail below. Earnings Management Using Tax Expense One line of research examines the use of tax expense as a last chance to manage earnings. 5 Dhaliwal et al. (2004) find that firms decrease tax expense from the third to the fourth quarter as a way to meet or beat analysts forecasts. Extending this line of research, Comprix et al. (2012) find that quarterly ETRs are systematically biased upward, creating slack that is later used to manage earnings. Cook et al. (2008) find that tax fees paid to auditors are associated with greater reductions in ETR from the third to the fourth quarter, and Christensen et al. (2015) provide evidence that industry expert auditors constrain earnings management through tax accounts. In a concurrent study, Gleason et al. (2016) conclude that the presence of a material weakness in internal controls over the tax function makes earnings management using the tax accounts easier for firms to implement. Overall, these studies provide robust evidence that firms decrease tax rates late in the year in an attempt to meet or beat analysts forecasts, consistent with earnings management incentives; however, these studies do not consider the extent to which executive risk incentives, firm characteristics, or their interaction are associated with the manipulation of tax expense to meet earnings targets. 6 5 Some studies examine specific tax-related accounts such as valuation allowance (e.g., Miller and Skinner 1998; Bauman et al. 2001; Frank and Rego 2006; Schrand and Wong 2003), uncertain tax positions (e.g., Gupta et al. 2016; Cazier et al. 2015), or permanently reinvested foreign earnings (APB 23) (e.g., Krull 2004; Graham et al. 2010). 6 Prior research provides evidence of an association between earnings management and equity incentives (e.g., Cheng and Warfield 2005; McVay et al. 2006); however, we are interested in the effect of equity incentives and their 5

8 Firm Characteristics and Corporate Tax Avoidance Another line of research closely related to our study examines determinants of corporate tax avoidance. Research in this area has found evidence that certain firm characteristics are associated with firms level of tax avoidance (e.g., lower ETRs). For example, studies have documented an association between tax avoidance and capital structure, asset mix, performance, extent of foreign operations, R&D intensity, and firm size (e.g., Graham 1996; Gupta and Newberry 1997; Collins et al. 1998; Graham 2000; Rego 2003; Wilson 2009; Lisowsky et al. 2013). These studies provide consistent evidence that certain firm characteristics provide increased opportunities for tax planning. Consistent with the notion that firm characteristics provide opportunities for tax planning, Mills et al. (1998) find that tax planning expenditures increase with foreign operations, capital intensity and the number of legal entities. Their results suggest that an additional $1 spent in tax planning results in a $4 reduction in tax liabilities on average. Relatedly, Robinson et al. (2010) find that firms are more likely to evaluate their tax departments as profit centers when firm characteristics provide greater tax planning opportunities, and this incentivizes the tax department to increase net income through lower ETRs. In a concurrent working paper, DeSimone et al. (2014) classify firms based on characteristics that likely make them more income mobile (i.e., can more easily exploit multi-jurisdictional income shifting) and find that income mobile firms have greater long-term tax avoidance than non-income mobile firms. Taken together, these studies provide support for the notion that certain firms have characteristics that allow for greater opportunities to avoid taxes. Managerial Incentives and Corporate Tax Avoidance interaction with opportunities on managerial influence over the tax accounts, a specific channel through which earnings management may occur. 6

9 Other tax avoidance research has investigated the association between managerial incentives and firms levels of tax avoidance. Dyreng et al. (2010) conclude that top executives have incremental effects on firms ETRs that cannot be explained by firm characteristics. Desai and Dharmapala (2006) find that increases in incentive compensation for the top five executives reduces the level of tax sheltering, and this result is primarily evident among firms with weak governance. In contrast, Rego and Wilson (2012) find that larger CEO and CFO equity risk incentives are associated with greater tax aggressiveness. Armstrong et al. (2012) conclude that tax directors are provided with incentives to reduce GAAP ETR, but not other tax attributes. Finally, prior work suggests that compensating business unit managers and CEOs on an after-tax basis is negatively associated with ETRs (Phillips 2003; Gaertner 2014). 7 Overall, this stream of literature suggests an association between different managerial incentives and firm-level tax outcomes. It is important to note, however, that although these studies examine the effect of managerial incentives on corporate tax avoidance, they typically include firm characteristics as controls in their analysis and do not consider the interactive effect between managerial incentives and firm characteristics. 8 We, on the other hand, argue that many of these firm characteristics create opportunities to influence the tax accounts, and will likely interact with managerial incentives to produce different levels of managerial influence on tax expense. The intuition is that managers employed by firms with greater opportunities to influence tax expense are more likely to respond to their incentives to do so. However, it is unclear whether managers at firms with 7 Phillips (2003) does not find an association between CEO incentives and tax avoidance; however, Gaertner (2014) demonstrates that the CEO result in Phillips (2003) is likely attributable to sample size and low statistical power. 8 For example, Armstrong et al. (2012) include managerial incentives as their variable of interest and control for foreign assets and complexity, among other variables. Likewise, Rego and Wilson (2012) control for firm characteristics such as size, R&D intensity, and capital expenditures. 7

10 lower levels of opportunity to influence tax expense will be willing or able to implement tax planning, or engage in manipulation in response to their incentives. It is also unclear if managers take advantage of the high levels of tax planning opportunities to influence tax expense if they do not have the incentives to do so. III. HYPOTHESES Our primary research question is how managerial incentives, opportunities, and their interaction affect managerial influence over tax expense. In order to discuss our hypotheses, we partition firms into four theoretical subsamples based on both their incentives and level of opportunity for tax planning as follows: 1) firms with high incentive and high opportunity, 2) firms with low incentive and high opportunity, 3) firms with high incentive and low opportunity, and 4) firms with low incentive and low opportunity. Figure 1 visually depicts the four subsamples of interest. 9 We detail our procedures for determining firms levels of incentives and opportunities in Section IV. [Insert Figure 1 here] Earnings Management Hypotheses The propensity for managers to use the tax accounts to manage earnings likely differs between these four groups. First, we predict that firms with higher opportunity (cells 1 and 2 in Figure 1) are more likely to purposefully manipulate tax expense than firms with lower opportunity (cells 3 and 4 in Figure 1). Although this prediction is supported by prior research that finds an association between firm characteristics and tax avoidance, the extent to which these characteristics extend to our setting is unclear, as we are interested in intentional manipulations of 9 We acknowledge that levels of incentives and opportunities are continuous, not discrete; however, we partition into four groups to facilitate our discussion and development of our hypotheses. In addition, firms with low opportunities likely have some level of opportunity for tax planning, just not the same level as other firms. 8

11 tax expense (i.e., purposeful actions that result in tax rate reductions) as opposed to the overall level of tax avoidance. Second, firms with higher incentives (cells 1 and 3 in Figure 1) are likely to manage earnings using tax expense to a greater degree than those with low incentives (cells 2 and 4 in Figure 1). Prior research supports this prediction by providing evidence that compensation incentives are associated with managerial influence over tax accounts (e.g., Gaertner 2014; Rego and Wilson 2012). In addition, prior research also suggests that equity incentives generally motivate managers to pursue riskier investment decisions (e.g., Core and Guay 1999; Rajgopal and Shevlin 2002; Coles et al. 2006; William and Rao 2006; Armstrong et al. 2013). To the extent that manipulating the tax accounts in order to achieve an earnings target reflects riskier activity, these studies provide support for our prediction. However, we are unaware of any study that specifically documents the association between managerial compensation incentives and earnings management using tax accounts. Therefore, our first (2-part) hypothesis, which complements and extends prior studies, is as follows: 10 H1a: Firms with high tax opportunity manage earnings using tax expense to a greater degree than firms with low tax opportunity. H1b: Firms with high managerial incentives manage earnings using tax expense to a greater degree than firms with low managerial incentives. Because managers are better able to respond to the incentives when firm-specific characteristics afford them greater opportunities to influence tax expense, we expect that firms with both high incentive and high opportunities (cell 1 in Figure 1) influence tax expense to a greater degree than firms with only high incentive, only high opportunity, or neither high 10 We note that prior studies have shown the association between tax avoidance (e.g., level of GAAP or Cash ETR) and opportunities and incentives. Our study differs because no study has shown how opportunities and incentives affect the manipulation of tax accounts (i.e., purposefully decreasing the ETR in order to reach an earnings target). 9

12 incentive nor opportunity (cells 2, 3, and 4 in Figure 1). We therefore state our second hypothesis as follows: H2: Firms with both high managerial incentives and high tax opportunity manage earnings using tax expense to a greater degree relative to firms that have low managerial incentives, low tax opportunity, or both low managerial incentives and tax opportunity. We make no formal prediction regarding the relative level of tax avoidance for high opportunity/low incentive firms (cell 2 in Figure 1) or low opportunity/high incentive firms (cell 3 in Figure 1) as we have no prior evidence or theory to guide us in making this prediction. Managers may respond to incentives by taking tax positions that reduce tax expense even with firm characteristics that do not provide high levels of opportunities to manipulate tax expense. If this is the case, it is possible that managerial incentives have a greater effect on earnings management through tax expense. However, these actions may increase risk because of the limited opportunities to tax plan. On the other hand, managers with low incentives but high opportunities could take advantage of the opportunities and decrease tax expense without adding significant risk. If this is the case, it is possible that opportunities have a greater effect on earnings management through tax expense. Tax Avoidance Hypotheses Because firms level of tax avoidance and the use of tax expense to manage earnings represent different aspects of executive influence over tax expense, we state our tax avoidance hypotheses separately from our earnings management hypotheses. As with our earnings management hypotheses, we expect that the level of tax avoidance likely differs between the four groups outlined in Figure 1. As previously discussed, prior research suggests that firms with greater opportunities (cells 1 and 2 in Figure 1) avoid more tax than firms with lower opportunities (cells 2 and 4 in Figure 1), and firms with higher incentives (cells 1 and 3 in Figure 10

13 1) avoid more tax than firms with lower incentives (cells 2 and 4 in Figure 1). Because prior literature provides evidence that supports these associations, we do not separately hypothesize them; however, we do replicate these associations in our tests. Next, we predict an interactive effect between incentives and opportunities on firms level of tax avoidance. Managers are likely able to respond to incentives more effectively when the firm has characteristics that create opportunities for increased tax planning, or otherwise influencing tax avoidance activity. We therefore expect that firms with both high incentive and high opportunity (cell 1 in Figure 1) avoid more tax than firms without both high incentives and high opportunities (cells 2, 3, and 4 in Figure 1). Accordingly, we state our third hypothesis as follows: H3: Firms with both high managerial incentives and high tax opportunity avoid more tax relative to firms that have low managerial incentives, low tax opportunity, or both low managerial incentives and tax opportunity. While prior research suggests that both managerial incentives and opportunities are associated with tax avoidance, we do not make a prediction as to whether firms with high opportunity/low incentives (cell 2 in Figure 1) or firms with low opportunity/high incentives (cell 3 in Figure 1) have greater levels of tax avoidance. Even when firm characteristics do not create high levels of opportunities for tax planning, managers may still respond to risk incentives by engaging in risky tax planning strategies to avoid tax burdens. Likewise, if firm characteristics provide opportunities to avoid tax through planning, the firm could use the opportunities to reduce its tax burden without the need to significantly incentivize the manager. IV. RESEARCH DESIGN Measuring Managerial Incentives and Opportunities 11

14 We draw on prior literature for three measures of managerial incentives. First, VEGA measures the sensitivity of the CEO s wealth to stock return volatility, also known as the manager s portfolio vega (see Core and Guay 1999; Rego and Wilson 2012). Prior studies demonstrate that the equity incentives captured by this measure motivate managers to pursue riskier investment and financing decisions (e.g., Core and Guay 1999; Rajgopal and Shevlin 2002; Coles et al. 2006; Williams and Rao 2006; Rego and Wilson 2012). Therefore, we use this measure to capture the extent to which managers are incentivized to assume greater risk by influencing tax expense (e.g., achieving an earnings target by manipulating the tax accounts expense, or by aggressively avoiding more tax). Second, DELTA measures the sensitivity of the CEO s wealth to stock price, also known as pay-for-performance sensitivity or delta. Pay-forperformance sensitivity provides an incentive to managers to engage in positive net present value projects; thus, we use this measure to capture the extent to which managers are incentivized to influence tax expense in such a way that provides positive net present value to the firm (e.g., by engaging in efficient tax planning strategies). 11 Third, following Gaertner (2014), AFTER_TAX is an indicator variable equal to one if the CEO s bonus is based on after-tax accounting measures, and zero otherwise. After-tax compensation provides an incentive for managers to engage in activities that increase after-tax performance, including tax planning and potentially exercise discretion to reduce tax expense and increase performance. There is no single widely-used proxy for firm opportunities to avoid tax, so based on prior research we construct an opportunity score that captures firm complexity and ability to influence 11 Risk incentives, or portfolio vega, capture the convexity of the relation between managers wealth and stock price, measured as the change in a manager s stock option portfolio for a given change in stock return volatility (e.g., Guay 1999). These incentives are different from the slope of the relation between manager s wealth and stock price, or portfolio delta, measured as the change in a manager s wealth for a given change in stock price; the slope effect motivates managers to undertake positive net present value projects, while the risk incentive effect motivates managers to undertake risky projects (Rego and Wilson 2012). 12

15 tax expense. To the extent that firm complexity provides managers greater discretion over tax planning, these characteristics allow greater tax planning opportunities. 12 We identify seven firmspecific characteristics to develop the composite opportunity score: firm size, number of business segments, number of geographic segments, percent sales from foreign operations, R&D expenditures, intangibles, and capital expenditures. While there may be additional firm characteristics that create opportunities for tax planning, we select these seven because the literature commonly identifies these as characteristics that allow firms additional opportunities for tax planning. We select size (SIZE) because larger firms are more likely to have complex operations and are able to take advantage of economies of scale with their tax planning (Omer et al. 1993; Rego 2003; De Simone et al. 2014). We select number business and geographic segments (NUM_BUSSEGMENTS and NUM_GEOSEGMENTS) because firm complexity increases with number of segments (Robinson et al. 2010; De Simone et al. 2015) and allows for additional managerial discretion through income shifting, transfer pricing, and other tax planning. We select percent of sales from foreign operations (FI) because the rate differential between the U.S. and foreign countries allows firms to engage in multi-jurisdictional income shifting through transfer pricing and other tax planning (Klassen et al. 1993; Collins et al. 1998; De Simone et al. 2014). We select R&D expenditures (RND) and intangible assets (INTANG) because the tax credits associated with R&D, significant managerial discretion as to what qualifies for R&D tax credits, and the ability to shift income through placement of intangible assets creates tax-related complexity (Robinson et al. 2010; Bratten et al. 2016). We select capital expenditures (CAPEX) 12 We note that De Simone et al. (2014) take a similar approach to identify firms with greater income mobility that can achieve tax avoidance with less risk. Although related, our measure is intended to capture general firm complexity that creates tax planning opportunities rather than income mobility which allows for international income shifting, one form of tax expense manipulation. 13

16 because capital intensity could indicate manufacturing capabilities, which contribute to opportunities for tax planning (Gupta and Newberry 1997). Because it is unclear whether an equal weighting of each characteristic is appropriate to provide a comprehensive opportunity score, we perform a principal component analysis using these seven characteristics to construct an opportunity score (OP_SCORE). See Appendix B for details. Earnings Management Multivariate Model Dhaliwal et al. (2004) find that firms decrease their ETR from third to fourth quarter to a greater extent when they would have missed analysts forecast using the third quarter ETR. Consistent with prior research (e.g., Degeorge et al. 1999; Skinner and Sloan 2002; Habib and Hansen 2008; Gleason and Mills 2008), we are not only interested in actions that appear consistent with earnings management (i.e., decreases in ETR from the third to the fourth quarter as in Dhaliwal et al. 2004), but, more importantly, we are interested in firms that are actually successful in meeting or beating analysts forecast as a result of earnings management. For example, a firm may decrease its ETR from third to fourth quarter, but unless that decrease enables the firm to meet or beat its target, this action does not result in successful earnings management. 13 Thus, our multivariate model to examine earnings management through tax expense is a variation on the one developed in Dhaliwal et al. (2004). Specifically, we examine the propensity for firms to successfully meet or beat analysts forecasts using tax expense by estimating the following logistic regression: _, (1) where MBWTAX is an indicator variable equal to 1 if the firm would have missed the forecast using its ETR in the third quarter, but ended up meeting or beating the forecast using its ETR in 13 See also Dhaliwal et al. (2004), page 445, for additional discussion regarding how ETR decreases may or may not actually result in actually meeting or beating the target. 14

17 the fourth quarter, and zero otherwise. In order to isolate firms that could potentially use tax expense to manage earnings, we estimate Equation (1) using a sample of firms that would have missed the forecast using their third quarter ETR. Despite the reduction in sample size due to this restriction, we consider this a more powerful test of our research question because it provides direct evidence regarding firms that were successful in meeting or beating the threshold by using a decrease in tax expense. OP_SCORE is the opportunity score based on the principal component analysis, and INCENTIVE is our managerial incentive variable (either VEGA, DELTA, or AFTER_TAX), defined above. Our controls follow Dhaliwal et al. (2004) and include INDUCED_CHG_ETR to control for changes in ETR due to unexpected earnings, TAXOWED to control for misestimation of tax owed in prior quarters, ETR3 to control for the amount by which the firm could decrease the ETR as well as for any underlying mean reversion, and ACCRUALS to control for accruals management. Specific variable definitions are included in Appendix A. We include year and industry fixed effects to control for macroeconomic effects causing systematic misestimation of ETRs and cluster standard errors by firm (Petersen 2009; Gow et al. 2010). A positive coefficient for β1 (β2) would be consistent with H1a (H1b) and would indicate a greater propensity to manage earnings using tax expense when the firm has greater opportunities (incentives). To test H2, we require an interaction between managerial incentives and opportunities. That is, we examine how the association between managerial incentives and earnings management using tax expense varies with opportunities. Therefore, we modify Equation (1) by adding an interaction between OP_SCORE and INCENTIVE as follows: _ _ x (2) 15

18 This specification allows us to test H2 by examining the interaction of managerial risk incentives and firm opportunities (i.e., β3). H2 predicts a positive coefficient estimate for β3, consistent with an interactive effect of managerial incentives and opportunities on earnings management using tax expense. Tax Avoidance Multivariate Model To examine the effect of managerial incentives and opportunities on tax avoidance, we start with the following OLS regression: _, (3) where ETR is forward-looking 3-year GAAP ETR (LR_ETR_3) or 3-year Cash ETR (LR_CETR_3), following Dyreng et al. (2008). 3-year GAAP and 3-year Cash ETR serve as our proxies of tax avoidance, where lower ETRs indicate greater levels of tax avoidance. INCENTIVE is VEGA, DELTA, or AFTER_TAX and OP_SCORE is the opportunity score as defined above. For our control variables, we follow recent tax avoidance research (e.g., Kubick et al. 2015) to control for other firm-specific factors known to affect effective tax rates but not captured by our opportunity score. Specific variable definitions are included in Appendix A. We also include year and industry fixed effects and cluster standard errors by firm (Petersen 2009; Gow et al. 2010). Although we do not hypothesize the association, a negative coefficient on β1 (β2) would be consistent with prior research and would suggest that firms avoid more tax when they have greater managerial incentives or more firm-specific characteristics that provide opportunities to avoid tax. To test H3, we require an interaction between managerial incentives and opportunities. Therefore, we modify Equation (3) by adding an interaction between OP_SCORE and INCENTIVE as follows: 16

19 _ _ x (4) This specification allows us to test H3 by examining the interaction of managerial incentives and firm opportunities (i.e., β3). H3 predicts that β3 will be positive, consistent with an interactive effect of managerial incentives and opportunities on tax avoidance. V. SAMPLE SELECTION AND DESCRIPTIVE STATISTICS Sample Our sample consists of observations at the intersection of Compustat, I/B/E/S, and Execucomp databases for the years 2003 to We obtain equity incentive data (DELTA and VEGA) from Execucomp, after-tax compensation data from IncentiveLab, analyst forecast data from I/B/E/S, and firm financial data used to calculate ETRs and control variables from Compustat. We begin our sample after 2002 due to different audit regulations and reporting requirements following SOX. Consistent with prior research, we remove observations with negative pretax income or negative ETRs, and observations with assets less than $10 million. We also trim outliers at the 1 st and 99 th percentiles based on ETR3, ETR4_ETR3, and INDUCED_CHG_ETR. Finally, we remove observations with missing data for the variables used in our regression. These restrictions leave us with 14,783 observations for our tax avoidance sample. Following prior research (e.g., Dhaliwal et al. 2004) we further restrict our sample for our earnings management tests. From our tax avoidance subsample, we remove observations with an analyst forecast error greater than five cents (i.e., we only keep observations with actual earnings per share within five cents of the consensus analyst forecast). We also remove observations that have earnings based on third quarter ETR that is greater than five cents away from the analyst 17

20 consensus forecast. These observations are likely too far from the forecast to reasonably manipulate ETR from the third to the fourth quarter to meet or beat the forecast. Finally, because we are interested in examining the behavior of firms that may use tax expense to manage earnings, we remove observations that would have beaten the analyst consensus forecast without a change in their third quarter ETR. These criteria result in a final sample of 870 observations for our earnings management sample. Table 1 provides details regarding our sample selection procedures. [Insert Table 1 here] Descriptive Statistics Table 2, Panel A presents descriptive statistics for our earnings management sample. Consistent with prior research (Dhaliwal et al. 2004; Comprix et al. 2012), the mean and median ETR4_ETR3 is negative, suggesting that firms decrease their ETR from the third to the fourth quarter on average. We further note that descriptive statistics for MISS_AMOUNT, MISS, INDUCED_CHG_ETR, TAXOWED, ETR3, ETR4, and ACCRUALS are consistent with Christensen et al. 2015, who use a similar time period for their analysis. The descriptive statistics for VEGA, representing managerial risk incentives, and DELTA, representing pay-forperformance sensitivity, are consistent with Rego and Wilson (2012) and suggest significant variation in managerial incentives within our sample. OP_SCORE has an average of by design, standard deviation of 0.815, and inter-quartile range of to 0.403, suggesting substantial variation in our sample of firm characteristics that create opportunities to influence the tax accounts. Table 2, Panel B presents descriptive statistics for our tax avoidance sample. In spite of differing sample sizes and sample selection procedures, the descriptive statistics for our tax 18

21 avoidance sample are consistent with those produced by prior studies utilizing similar variables. Specifically, the means, medians and standard deviations for controls unique to our tax avoidance sample (i.e. INTANG, RND, FI, PPE, ROA, MTB, LEV, FCF, NOL, CHG_NOL) are close to those generated by Kubick et al. (2015). 14 [Insert Table 2 here] VI. EARNINGS MANAGEMENT RESULTS Earnings Management Univariate Analysis Table 3 Panel A presents the proportion of firms with MBWTAX =1 for the four subsamples of firms outlined in Figure 1. This represents the proportion of firms that would have missed analysts forecast using 3 rd quarter ETR, but the reduction in their ETR from third to fourth quarter allowed them to meet or beat the analyst forecast. We separately present the results for all three incentive variables (VEGA, DELTA, and AFTER_TAX) to split the sample on high/low incentives. We present univariate tests of differences in these proportions across subsamples in Table 3 Panel B. Our results indicate that firms with high opportunity have a higher propensity to meet or beat the analyst consensus forecast with a decrease in tax rate than firms with low opportunity, and firms with high incentives have a higher propensity to meet or beat the analyst consensus forecast with a decrease in tax rate than firms with low incentives (p-values < 0.01). These results are consistent with H1a and H1b, respectively. We also find that firms with both high opportunity and high incentive (HOppHInc = 1) have a higher proportion of firms that meet or beat the analyst consensus forecast with a decrease in tax rate than firms with high opportunity and low incentive 14 We note a number of variables significantly differ between our tax avoidance sample and earnings management sample. This is due to the greater restrictions on the earnings management sample (i.e., being close to the analyst forecast). 19

22 (HOppLInc = 1) and firms with low opportunity and high incentive (LOppHInc = 1) (p-values < 0.01). These results are consistent with H2 and suggest an interactive effect between incentives and opportunities. [Insert Table 3 here] Earnings Management Multivariate Results Table 4 presents the results of estimating Equation (1) using our earnings management sample to test H1a and H1b. In Models I, II, and III we include DELTA, VEGA, and AFTER_TAX, respectively, as our managerial incentive variables, as well as OP_SCORE to proxy for firm-level opportunities to influence tax expense. In Model I (II) we find that the coefficient on DELTA (VEGA) and OP_SCORE are both positive and significant (p-values < 0.01), consistent with H1a and H1b. In Model III the coefficient on AFTER_TAX is not significant at conventional levels. 15 Overall, these results suggest that when firms have greater opportunities to influence tax expense, they are more likely to use tax expense as a means of managing earnings to meet or beat analysts forecasts. Furthermore, when managers have incentives to engage in more risky activity (i.e., they have higher vega) or they have incentives tied to stock price (i.e., they have higher delta), they are also more likely to use tax expense as a means of managing earnings. 16 [Insert Table 4 here] To test H2, we estimate Equation (2) and present the results in Table 5. In Model I, II, and III, we use DELTA, VEGA, and AFTER_TAX, respectively, to determine levels of incentive. These tests allow us to examine whether opportunities interact with the different incentives. In all 15 We note that our sample size is significantly reduced for this analysis due to data limitations from IncentiveLab, and this could be one reason for not finding significant results in Model III. 16 Gleason and Mills (2008) find that the market discounts the reward for meeting or beating analysts forecast when using a decrease in ETR; however, if managers believe ex ante there is a stock price reward for this behavior they might engage in earnings management using tax expense to gain their expected reward. 20

23 three models, we find the coefficient on the interaction between incentives and opportunities is positive (p-values < 0.10, 0.10, and 0.05, respectively), suggesting that firms with high opportunity respond to their incentives to a greater extent than firms with lower opportunities. The coefficients on the interactions suggest that a standard deviation increase in opportunities results in a 21.4%, 16.6%, and 82.5% increase in the likelihood of meeting or beating analysts forecasts when the CEO has high DELTA, VEGA, or AFTER_TAX incentives, respectively. 17 Thus, the interaction between incentives and opportunities is economically meaningful in an earnings management context. Overall, these results are consistent with H2 and provide evidence in support of an interactive effect of incentives and opportunities on manager s influence of tax expense in an earnings management context. [Insert Table 5 here] VII. TAX AVOIDANCE RESULTS Tax Avoidance Univariate Analysis Table 6 Panel A presents the mean 3-year GAAP ETRs (LR_ETR_3) for the four subsamples of firms outlined in Figure 1. We present univariate tests of differences in 3-year GAAP ETR across subsamples in Table 6 Panel B. As expected, we find that firms with high opportunity have lower GAAP ETRs than firms with low opportunity, and that firms with high incentive have lower GAAP ETRs than firms with low incentives. We also find that firms with both high opportunity and high incentive (HOppHInc = 1) have lower GAAP ETRs than firms with high opportunity and low incentive (HOppLInc = 1) and firms with low opportunity and high incentive (LOppHInc = 1) (p-values < 0.01). These results are consistent with H3 and suggest an interactive effect between incentives and opportunities on GAAP ETR. 17 We note these estimates may be sensitive to our specific sample, in particular because sample sizes are not very large. 21

24 [Insert Table 6 here] Table 6 Panel C presents the mean 3-year Cash ETRs (LR_CETR_3) for the four subsamples of firms outlined in Figure 1. We present univariate tests of differences in 3-year Cash ETR across subsamples in Table 6 Panel D. As expected, we find that firms with high opportunity have lower Cash ETRs than firms with low opportunity. We also generally find across our incentive variables that firms with both high opportunity and high incentive (HOppHInc = 1) have lower Cash ETRs than firms with high opportunity and low incentive (HOppLInc = 1) and firms with low opportunity and high incentive (LOppHInc = 1). These results are consistent with H3 and suggest an interactive effect between incentives and opportunities on Cash ETR. Tax Avoidance Multivariate Results Table 7 presents the results of estimating Equation (4) using our tax avoidance sample. In Models I through III, LR_ETR_3 is the dependent variable. In Models I, II, and III we do not find strong evidence of an interactive effect between incentives and opportunities on GAAP ETRs (all p-values > 0.10). We do, however, find the coefficient on OP_SCORE is negative and significant across all specifications (p-values < 0.01), and the coefficients on VEGA and AFTER_TAX are negative and significant (p-values < 0.01), consistent with Rego and Wilson (2012) and Gaertner (2014). In Models IV through VI, LR_CETR_3 is the dependent variable. In Model IV and V, the interactions between DELTA (VEGA) and OP_SCORE are negative and significant (p-values < 0.10 and 0.05, respectively), consistent with H3. The coefficient on the interaction between AFTER_TAX and OP_SCORE in model VI is insignificant. In sum, the results in Table 7 provide evidence of an interactive effect between managerial incentives and opportunities on firms cash 22

25 ETR, but we do not find strong evidence of an interactive effect between incentives and opportunities on GAAP ETRs. [Insert Table 7 here] VII. SUMMARY AND CONCLUSION This paper examines the extent to which managerial incentives, firm characteristics that create opportunities to influence tax accounts, and the interaction between these factors are associated with managerial influence over tax expense. We examine managerial influence of tax expense in two related, but distinct settings: 1) earnings management using tax expense, and 2) long-run tax avoidance. While prior research has examined the effect of individual executives (e.g., Dyreng et al. 2010; Chyz 2013), managerial risk incentives (e.g., Rego and Wilson 2012), or firm-specific characteristics on firms level of tax avoidance, no study has investigated the role of these factors on earnings management using tax expense (Dhaliwal et al. 2004). Further, no study has examined how the interaction between incentives and opportunities could influence the extent to which managers influence tax expense. We assert that the extent to which managers influence tax expense, whether through earnings management using the tax accounts or tax avoidance, depends on both their incentives and opportunities to do so. We predict and find that both managerial incentives and firm opportunities are associated with a greater propensity for firms to manage earnings using tax expense. Further, we find strong evidence for an interactive effect between incentives and opportunities; that is, firms with both high incentives and high opportunities manage earnings using tax expense to a greater degree than firms with either high incentives or opportunities, but not both. Consistent with prior research, we also find that both managerial incentives and firm opportunities are associated with a greater tax 23

26 avoidance. Our results also suggest an interactive effect between incentives and opportunities on cash ETRs, a result that is new to the literature. This study contributes to the literature in at least three ways. First, we demonstrate situations when managers purposefully reduce tax expense, which contributes to the literature investigating the determinants of tax avoidance. Second, we show that managers respond to compensation-based incentives differently based on varying sets of opportunities, which contributes to the literature that examines executive influence over tax accounts (e.g., Dyreng et al. 2010) and the role of managerial incentives (e.g., Rego and Wilson 2012). Finally, our paper responds to the call in Graham et al. (2012) to provide additional documentation of earnings management through tax expense by examining not only the role of managerial incentives and opportunities, but more importantly the interaction of these factors. 24

27 APPENDIX A Variable Definitions (Alphabetical Order) ACCRUALS AFTER_TAX CHG_NOL DELTA ETR4_ETR3 ETR3 FCF FI HOppHInc HOppLInc INDUCED_CHG_ETR INTANG LEV LOppHInc LR_CETR_3 LR_ETR_3 MBWTAX Total accruals scaled by pre-tax book income, as specified by Dhaliwal et al. (2004). Indicator variable equal to one if the CEO equal to one if the CEO s bonus is based on after-tax accounting measures, and zero otherwise. After-tax bonus compensation data are from IncentiveLab database. The change in firm i s net operating loss scaled by lagged total assets. CEO pay-for-performance sensitivity, which captures the slope of the relationship between manager s wealth and stock price. Calculated as the change in the manager s wealth for a given change in stock price. The difference between ETR4 and ETR3. Total income taxes through three quarters divided by pretax income through three quarters. Operating cash flows minus capital expenditures scaled by lagged total assets. Pretax foreign income scaled by lagged total assets. An indicator variable equal to 1 if the firm has both high incentives and high opportunities, and zero otherwise (i.e., cell 1 in Figure 1). An indicator variable equal to 1 if the firm has high opportunities but low incentives, and zero otherwise (i.e., cell 2 in Figure 1). The change in total tax expense due to the difference between actual earnings per share and the consensus forecast divided by pretax income. Calculated as the statutory tax rate (0.35 for our sample) minus ETRQ3 times the difference between actual earnings per share and the consensus analyst forecast times common shares used to calculated EPS divided by 1 minus the statutory tax rate (0.35) all divided by pretax income. Intangible assets for year t scaled by lagged total assets. Total long-term debt scaled by lagged total assets. An indicator variable equal to 1 if the firm has low opportunities but high incentives, and zero otherwise (i.e., cell 3 in Figure 1). The sum of total cash taxes paid for the current year and next two years divided by the sum of pretax book income less special items for the current year and the next two years. The sum of total tax expense for the current year and next two years divided by the sum of pretax book income less special items for the current year and the next two years. An indicator variable equal to 1 for firms that would have missed analysts forecasts using their third quarter ETR, but ended up 25