Executive Compensation, Tax Reporting Aggressiveness, and Future Firm Performance. Sonja Olhoft Rego University of Iowa

Transcription

1 Executive Compensation, Tax Reporting Aggressiveness, and Future Firm Performance Sonja Olhoft Rego University of Iowa Ryan Wilson * University of Iowa December 22, 2008 Abstract This study investigates the association between tax reporting aggressiveness and the level of CEO and CFO compensation, after controlling for standard economic determinants of such compensation. Across six different measures of tax aggressiveness, we consistently find a positive relation between aggressive tax reporting and the level of CEO and CFO compensation. We investigate whether this positive relation is due to optimal contracting or to managers extracting rents from firms. We find no evidence that aggressive tax strategies allow managers to extract rents from firms. In particular, we find no evidence that firms with aggressive tax reporting and weak corporate governance pay their CEOs and CFOs more than other firms. Moreover, we find no evidence that the positive relation between executive compensation and tax reporting aggressiveness leads to poor future firm performance. JEL Classifications: M41, M52, G34 Keywords: Tax aggressiveness, executive compensation, corporate governance. * We appreciate helpful comments from Dan Collins, Cristi Gleason, Leslie Hodder, Paul Hribar, Steve Kachelmeier, Sean McGuire, Rick Mergenthaler, Lillian Mills, Tom Omer, John Robinson, Terry Shevlin, Connie Weaver, and workshop participants at Indiana University, Texas A&M University, the University of Iowa, and the University of Texas.

2 Executive Compensation, Tax Reporting Aggressiveness, and Future Firm Performance 1. Introduction Prior research links executive compensation to the economic determinants of such compensation (e.g., Smith and Watts 1992; Core and Guay 1999), to governance characteristics (e.g., Core, Holthausen, and Larcker 1999; Ashbaugh-Skaife, Collins, and LaFond 2006), and to earnings management incentives (e.g., Erickson, Hanlon, and Maydew 2006; Armstrong, Jagolinzer, and Larcker 2008). While Core and Guay (1999) generally conclude that compensation contracts on average reflect incentive alignment, several recent studies suggest a positive relation between executive compensation and accounting manipulation (e.g., Cheng and Warfield 2005; Bergstresser and Philippon 2006). Despite a surge in both executive compensation and tax shelter activity during the 1990 s and early 2000 s, few studies have considered the link between executive compensation and tax reporting aggressiveness. We examine the association between executive compensation and tax aggressiveness and the implications of that association for future firm performance. For purposes of our study, we consider tax aggressiveness to include standard tax planning practices that do not necessarily violate income tax rules, as well as fraudulent tax avoidance, including tax shelter transactions considered abusive by the IRS and the Treasury Department. Firms may compensate managers for aggressive tax reporting for various reasons, the most obvious of which is the tax savings associated with tax planning activities. These tax savings often increase after-tax cash flows, book income, net assets, and more generally, financial slack attributes that typically have a positive impact on firm value. In addition, firms may also compensate managers for aggressive tax reporting to reward risk-taking behaviors, which managers may otherwise avoid (e.g., Rajgopal and Shevlin 2002). 1

3 Despite the considerable tax savings generated by corporate tax shelters, significant cross-sectional variation exists in the degree to which firms pursue aggressive tax reporting strategies. Shevlin (2007) calls for research examining why some firms enter into tax shelters while other do not. Investigating the link between executive compensation and tax aggressiveness provides insight into one potentially important determinant of observed crosssectional variation in tax shelter activity. In particular, managers may forego aggressive tax sheltering unless such activities have a net positive impact on their total compensation. Our study also provides insights into whether optimal contracting or opportunistic behavior by top executives accounts for the association between executive compensation and aggressive tax reporting. Although we expect firms to compensate managers for some degree of tax aggressiveness, we do not expect firms to compensate managers for tax aggressiveness that imposes excess costs on the firm and thus reduces shareholder wealth. Highly aggressive tax reporting can cause firms to incur additional costs, such as the costs of implementing the aggressive tax plan (e.g., promoter and attorney fees), the costs associated with IRS audits and subsequent litigation (e.g., accounting and legal fees), as well as reputational costs associated with the public disclosure of tax shelter activity (e.g., negative stock price implications as in Hanlon and Slemrod 2007). Moreover, to the extent that conformity exists between financial and tax reporting, reporting lower taxable income can generate financial reporting costs (i.e., lower financial income). In sum, we do not expect firms to compensate managers for tax aggressiveness if the costs exceed the benefits. We investigate the extent that firms compensate managers for aggressive tax planning using different measures of tax aggressiveness and we 2

4 examine chief executive officers (CEOs) separately from chief financial officers (CFOs), as CFOs have greater responsibility for the accounting function, relative to CEOs. Using compensation data obtained in a survey of corporate executives, Phillips (2003) finds that compensating division managers (but not CEOs) on an after-tax basis leads to greater tax planning effectiveness. Hanlon, Mills, and Slemrod (2005) find a positive association between the level of equity incentives from exercisable stock options and proposed IRS deficiencies. In contrast, Desai and Dharmapala (2006) find that increases in equity-based incentives actually lead to a reduction in the level of tax avoidance. Desai and Dharmapala conjecture that because tax shelters are intentionally designed to obscure the economic substance of a transaction they therefore provide a shield for managers to engage in opportunistic behavior. Consequently, they argue that increases in equity compensation reduce the level of tax sheltering by better aligning the incentives of managers with those of shareholders. Desai and Dharmapala argue this improved alignment is especially true in firms with poor corporate governance structures where it is easier for managers to engage in opportunistic behavior through the use of corporate tax shelters. While it is possible tax shelters could be used opportunistically by managers, they often generate significant tax savings that benefit shareholders. Examining a sample of 44 identified tax shelter cases, Graham and Tucker (2006) find the average tax shelter generates an annual tax deduction sufficient to shield income equal to approximately 9 percent of asset value. Also consistent with tax shelters benefiting shareholders, Wilson (2008) finds firms with strong governance structures that engage in tax shelter transactions exhibit positive abnormal stock returns during the tax shelter period. Because aggressive tax reporting can result in significant tax savings, but can also generate significant costs, including the potential for managerial 3

5 opportunism, it is not clear whether executive compensation contracts will be designed to reward managers for aggressive tax reporting. We directly test this association using a number of recently developed measures of tax reporting aggressiveness. Specifically, we utilize traditional effective tax rates, cash effective tax rates, total book-tax differences, permanent book-tax differences, discretionary book-tax differences, and predicted tax shelter firms. These six measures reflect varying degrees of aggressiveness, with the latter two likely reflecting the most aggressive tax strategies. Our initial tests document a strong, positive association between total CEO and CFO compensation and aggressive tax planning. This result is robust for all six measures of tax aggressiveness, as well as controls for the economic determinants of executive compensation. We infer that firms compensate managers for varying degrees of tax aggressiveness. We then investigate whether this positive relation between CEO and CFO compensation and tax aggressiveness is consistent with optimal contracting or with managers extracting rents from firms. First, we build on Desai and Dharmapala (2006) and examine whether governance structures moderate the positive relation between executive compensation and tax aggressiveness. If firms with weaker governance mechanisms exhibit a more positive relation between executive compensation and tax aggressiveness, then we would infer that managers at firms with weaker governance structures are able to extract rents from those firms via aggressive tax planning. On the other hand, if firms with stronger governance mechanisms exhibit a more positive relation between executive compensation and tax aggressiveness, then we would conclude that such managers are compensated for efficient tax sheltering choices. Contrary to stories of managerial rent extraction, our results suggest that corporate governance structures do 4

6 not consistently moderate the relation between executive compensation and tax reporting aggressiveness. Our second test of optimal contracting vs. managerial rent extraction considers the implications of the positive relation between executive compensation and tax aggressiveness for future firm performance. Core et al. (1999) and Ashbaugh-Skaife et al. (2006) show that measures of excess compensation are associated with negative future return on assets and negative future stock returns. 1 They interpret their results as consistent with managers at firms with greater agency problems receiving higher compensation that leads to worse future firm performance. We model our analyses after their empirical tests and calculate the portions of CEO and CFO compensation that are related to tax reporting aggressiveness, which we refer to as CEO and CFO tax-based compensation. We then regress future return on assets and future stock returns on CEO and CFO tax-based compensation and control variables. We find that taxbased compensation is not associated with inferior future firm performance. Thus, we find no evidence that aggressive tax strategies allow managers to extract rents from firms. Our study contributes to several streams of research, including studies that investigate whether executive compensation practices align managers incentives with those of shareholders, and studies that examine the economic outcomes associated with tax reporting aggressiveness. To our knowledge, this is the first study to link tax aggressiveness to executive compensation and future firm performance. Our results have implications for corporate compensation committees that determine that structure of executive compensation contracts, for regulators and corporate stakeholders that monitor corporate tax practices, as well as researchers interested in 1 Core et al. (1999) and Ashbaugh-Skaife et al. (2006) calculate excess compensation as the portion of total compensation that is predicted by board and ownership structure variables. Core et al. argue that if contracting is optimal there should be no association between CEO compensation and governance structures, i.e., CEO compensation should only be a function of economic determinants. 5

7 understanding the factors that impact corporate tax aggressiveness. Specifically, our results suggest tax reporting aggressiveness is an important determinant of total CEO (CFO) compensation and that this association does not appear to be a result of managerial opportunism. The remainder of this paper proceeds as follows. Section 2 describes prior studies that we build upon and develops hypotheses. Section 3 explains our research design, while Section 4 discusses sample selection and empirical results. Section 5 concludes. 2. Background and Hypothesis Development 2.1 Prior Research Our study builds on separate streams of prior research. One stream investigates the valuation implications of aggressive tax reporting, as well as the links between managerial compensation practices and tax reporting aggressiveness. The other stream examines the determinants of executive compensation. We discuss each of these streams below Is Tax Avoidance a Value-Maximizing Activity? Few studies directly investigate whether tax avoidance (aka tax planning or tax aggressiveness) is a value-maximizing activity for firms to undertake. Bankman (1999) argues that the dollar benefit of participating in a tax shelter far outweighs the cost, given the low probability of detection by the tax authorities. Graham and Tucker (2006) state that tax sheltering should affect stock prices since tax sheltering can increase financial slack, reduce expected bankruptcy costs, enhance credit quality, reduce the risk of covenant violation, and reduce the cost of debt (p. 566). They find that tax shelter firms credit ratings improve one notch relative to those of matched, control firms, consistent with lower costs of debt for their 6

8 sample of tax shelter firms. In addition, Wilson (2008) finds that the average tax shelter transaction generates federal income tax savings of $375.5 million. Using a large sample of U.S. firms, Desai and Dharmapala (2007) investigate whether corporate tax avoidance is valued by investors. They find a positive relation between their proxy for corporate tax avoidance (as measured by book-tax differences adjusted for accruals) and firm value (as proxied by Tobin s Q). They also find that firms with higher levels of institutional ownership (their proxy for strong corporate governance) exhibit a more positive relation between those two variables than firms with less institutional ownership. Hanlon and Slemrod (2007) investigate a similar research question but implement a different research design. They examine how the stock market evaluates news of corporate tax aggressiveness by investigating stock price reactions to the initial press mention that a firm was involved in a corporate tax shelter. The stock price reactions should reflect competing stock price components, including the positive impact of corporate tax planning and the negative impact of the direct costs associated with tax sheltering, as well as indirect reputational costs associated with the negative publicity. Hanlon and Slemrod find a small stock price decline associated with the initial tax shelter announcements; however, the decline is relatively small compared to other accounting transgressions. They also find the stock price decline is smaller for firms with stronger corporate governance structures. Based on a sample of firms identified ex post as having participated in tax shelter transactions, Wilson (2008) develops a model of tax shelter participation and then uses the model to identify a sample of predicted tax shelter firms. He relies on both samples to investigate whether tax sheltering increases shareholder wealth or allows managers to extract rents from the firm. Wilson finds that tax shelter firms with strong corporate governance exhibit positive 7

9 abnormal stock returns during the period of active tax sheltering, and these returns are significantly greater than the returns of tax shelter firms with poor corporate governance. Similar to Desai and Dharmapala (2007) and Hanlon and Slemrod (2007), the results in Wilson (2008) suggest an interaction effect between tax aggressiveness, strength of corporate governance, and their impact on firm value. In sum, prior research shows that tax sheltering can be a value-maximizing activity, but most likely for those firms with stronger corporate governance structures Linking Tax Aggressiveness to Compensation Practices Few studies have examined the relation between executive compensation practices and tax reporting aggressiveness. Shackelford, Slemrod, and Sallee (2007) develop a general model of how taxes affect the real and accounting decisions of corporations. They note that financial reports are a means by which managers convey their inside knowledge to external investors, reducing information asymmetries and thus lowering the cost of obtaining capital. In their model, there are at least two reasons why accounting information may be important to managers. First, many contracts (including compensation contracts) rely on the firm s accounting information. Second, if users of financial statements cannot distinguish between low earnings arising from poor profitability and low earnings arising from tax plans that increase cash flow at the expense of accounting earnings, then managers may be unwilling to minimize real taxes paid. This discussion highlights the fact that although tax aggressiveness can improve the cash flows of the firm, managers may not engage in such activity if it will reduce their expected compensation. In an empirical study that more directly investigates the link between tax aggressiveness and executive compensation, Phillips (2003) examines whether compensating managers based on 8

10 after-tax performance measures leads to lower effective tax rates, his proxy for tax planning effectiveness. Phillips notes that firms should use after-tax performance measures to compensate managers only if the expected benefits exceed the expected costs of doing so. Based on a sample of 209 surveyed corporate executives, Phillips concludes that compensating business unit managers, but not CEOs, on an after-tax basis leads to lower effective tax rates. Desai and Dharmapala (2006) examine how equity-based compensation incentives influence tax sheltering decisions. Because equity-based incentives should align managerial interests with those of shareholders, Desai and Dharmapala predict that such incentives should induce managers to reduce their diversion of rents and increase their tax sheltering activities. However, Desai and Dharmapala also conjecture that complex tax shelter transactions that are designed to obscure the economic substance of the transactions may also obscure a firm s financial reporting and increase the opportunities for managerial diversion. 2 They test their model across well-governed and weaker-governed firms and find that increases in incentive compensation reduce the level of tax sheltering and this negative effect is driven primarily by weaker-governed firms. Desai and Dharmapala conclude that incentive compensation better aligns managerial incentives with those of shareholders and reduces opportunistic tax sheltering. Our study extends Phillips (2003) and Desai and Dharmapala (2006) by directly investigating whether CEOs and CFOs are compensated for tax reporting aggressiveness and then linking these compensation practices to future firm performance. We consider various definitions of tax aggressiveness and examine their interactions with different corporate 2 Desai and Dharmapala (2006) cite Dynegy as an example of a company that used an opaque, tax-motivated transaction to mislead the capital markets. Although the authors acknowledge that no evidence of managerial diversion exists in the Dynegy example, they claim that Dynegy illustrates how misrepresentations destructive to shareholders can be facilitated by tax avoidance. (p. 157) 9

11 governance structures. Importantly, our empirical tests are based on existing models of executive compensation, which we discuss below Determinants of Executive Compensation In a seminal study in finance, Smith and Watts (1992) examine the determinants of corporate policy decisions, including the level of executive compensation. Smith and Watts predict that a firm s investment opportunity set, industry regulation, size, and accounting return should explain the level of salary compensation. Their results indicate that larger firms with greater growth opportunities, less regulation, and higher accounting return have significantly higher salary compensation. The authors also find that firms with greater growth options have higher salary compensation and use stock-based incentive plans more often. Core, Holthausen, and Larcker (1999) extend Smith and Watts (1992) and model CEO compensation as a function of the economic determinants of CEO compensation, as well as board of director characteristics and ownership structure. Core et al. propose that the board and ownership structures observed in practice should induce optimal CEO contracting and firm performance. If firms engage in optimal contracting then the economic determinants of CEO compensation should completely describe the cross-sectional variation in CEO compensation. In this case, board and ownership structure variables should be unrelated to CEO compensation. Contrary to expectations, after controlling for standard economic determinants of CEO compensation, Core et al. find that board and ownership structure variables have a significant impact on the level of CEO compensation. Subsequent tests indicate a negative relation between the amount of excess compensation (calculated as the portion of total compensation predicted by the board and ownership structure variables) and future operating and stock market performance. Together, their results suggest that firms with weaker governance structures have greater agency 10

12 problems, that CEOs at firms with greater agency problems receive greater compensation, and that these firms exhibit worse future firm performance. We rely on the theory and compensation model underlying Core et al. (1999) to investigate our hypotheses, which we develop below. 2.2 Hypotheses Core et al. (1999) contend that the board and ownership structures observed in practice should induce optimal CEO contracting and firm performance. 3 In this case, managers should select the level of tax reporting aggressiveness that maximizes firm value. 4 If Core et al. s model for the equilibrium wage of the CEO is well-specified (i.e., if the economic determinants fully explain the level of CEO compensation), then tax aggressiveness should not predict the level of CEO compensation. That is, tax aggressiveness should be a noisy proxy for the variables that explain CEO compensation (e.g., after-tax return on assets). Thus, our first hypothesis (stated in the alternative) examines the basic relation between tax aggressiveness and the level of executive compensation: H1: Tax reporting aggressiveness is incrementally useful in predicting the amount of CEO and CFO compensation beyond standard economic determinants of such compensation. If hypothesis 1 is supported by our empirical results, then either: 1) managers are compensated for tax planning activities that merit a higher equilibrium wage (i.e., optimal contracting exists), or 2) there exist unresolved agency problems that allow managers to use 3 Their argument presumes that firms and managers engage in optimal contracting. Prior studies (e.g., Eaton and Rosen 1983 and Demsetz and Lehn 1985) assume that firms choose optimal compensation schemes that are based on unique firm characteristics and the characteristics of their executives. These studies contrast other studies in finance (e.g., Jensen and Murphy 1990) that conclude CEO compensation contracts in practice are less than optimal. 4 We do not claim that the CEO is directly involved in the tax planning process. Instead, we view the CEO as having a significant impact on corporate policies and decision-making, including tax planning. This view is consistent with results in Dyreng, Hanlon, and Maydew (2008) that suggest CEOs and CFOs influence a firm s level of tax aggressiveness. 11

13 aggressive tax strategies to extract rents from the firm (i.e., optimal contracting does not exist). If managers are compensated for tax planning activities that merit a higher equilibrium wage, then our proxies for tax aggressiveness reflect economic determinants of CEO and CFO compensation that are not captured by the variables included in the Core et al. (1999) model. For example, our measures of tax aggressiveness may reflect long-term savings associated with tax planning activities that extend over a multi-year period. Our measures of tax aggressiveness may also reflect the negative relation between aggressive tax sheltering and a firm s reliance on debt as a tax shield. The results in Graham and Tucker (2006) suggest that firms with more aggressive tax reporting are generally less levered, and thus likely have lower costs of debt, for which managers may receive higher compensation. On the other hand, if unresolved agency problems exist, then aggressive tax strategies may allow executives to obtain levels of compensation in excess of those implied by standard economic determinants. Desai and Dharmapala (2006) conjecture that complex tax shelter transactions may obscure the financial reporting of a firm (i.e., increase the information asymmetry between managers and shareholders) such that tax sheltering increases a manager s opportunity for the diversion of rents. For example, certain tax shelter transactions may generate financial reporting benefits that are difficult to identify in the financial statements but nonetheless generate bonus and/or stock-based compensation that the manager would not have otherwise earned. 5 The rent extraction argument in Desai and Dharmapala (2006) hinges on unresolved agency conflicts, such as those discussed in Core et al. (1999). If firms do not operate with 5 Enron is illustrative of this argument. Enron s numerous tax-motivated transactions not only saved the corporation millions of dollars in Federal income taxes but also enhanced the net income and net assets shown in its financial statements. Enron reported near-zero taxable income between 1996 and 2000 but billions of dollars of book income over the same reporting period (Seida 2003). This divergence between book and taxable income likely had a direct (and positive) impact on Enron s executives total compensation. 12

14 optimal corporate governance mechanisms in place, then managers may be able to engage in tax sheltering that increases their personal wealth rather than shareholder wealth. If aggressive tax reporting is associated with suboptimal contracting, we would expect this association to be strongest in firms with weak corporate governance. This leads to our second hypothesis (also stated in the alternative): H2: Firms with weaker corporate governance structures and more aggressive tax reporting pay more compensation to their executives. Even if we do not find results that support hypotheses 2, there may still exist unresolved agency problems that our corporate governance proxies are unable to capture. So we further try to distinguish between the optimal contracting and rent extraction explanations for the positive association between tax aggressiveness and executive compensation by examining the relation between the portion of compensation predicted by tax aggressiveness ( tax-based compensation ) and future firm performance. Our tests are modeled after those in Core et al. (1999). Specifically, if the association between managerial compensation and tax aggressiveness reflects agency conflicts (e.g., sub-optimal governance structures that allow managerial rent diversion), we expect to find a negative relation between tax-based compensation and future firm performance. A negative relation would be consistent with managerial rent extraction that reduces future firm performance. We expect no association (or perhaps a positive association) between tax-based compensation and future firm performance if tax aggressiveness is an economic determinant of executive compensation, consistent with optimal contracting. Thus, our third hypothesis predicts (also stated in the alternative): H3: Tax-based compensation is negatively associated with measures of future firm performance. 13

15 3. Proxies for tax reporting aggressiveness Before we describe our research design we provide a brief discussion of our measures of tax reporting aggressiveness. Because of the difficulty in accurately measuring tax reporting aggressiveness, we conduct our tests using six alternative measures. With the exception of the final measure of tax aggressiveness (SHELTER), we rank each measure by year and two-digit SIC code. We then create an indicator variable that is set equal to one for the firms in the top (bottom for ETR and CASH_ETR) quintile of each respective tax aggressiveness measure, and zero for all other firms. This approach allows us to identify the most aggressive tax reporting firms according to each measure while controlling for industry effects that could cause differences in these measures unrelated to aggressive tax reporting strategies. 6 Each of the measures has limitations which we discuss below. However, observing consistent results across these measures supports the contention that our proxies capture a firm s tax reporting aggressiveness. 3.1 Effective Tax Rate (ETR) We begin with a measure of the traditional effective tax rate (ETR), calculated as total tax expense divided by pre-tax book income less special items. Numerous studies have used ETR as a measure of corporate tax burden and firms are also required to disclose their ETR in their financial statements. Zimmerman (1983) examines firm size and its association with effective tax rates. Gupta and Newberry (1992) examine the associations between ETR and firm size, capital structure, asset mix, and performance. Further, policy makers such as Citizens for Tax Justice (CTJ) have used the ETR measure to identify which U.S. corporations have relatively low total tax burdens. 6 Nonetheless, in untabulated analyses, we also perform all of our empirical tests using continuous measures of tax aggressiveness. Inferences from these tests are substantially the same as those based on our tabulated results. 14

16 Hanlon (2003) identifies a number of limitations of ETR as a measure of tax aggressiveness. First, total tax expense is the sum of current and deferred taxes. Deferred taxes represent taxes that will be paid (or refunded) in the future associated with temporary book-tax differences. Including deferred taxes in the calculation means that the ETR measure will not reflect aggressive tax planning that causes temporary book-tax differences. This limitation is important because some aggressive tax reporting strategies involve accelerating expenses or deferring revenue for tax purposes. Second, if firms are recording contingencies (i.e., tax cushion) associated with uncertain tax benefits generated by aggressive tax positions, then ETR will understate the firm s true level of tax aggressiveness. This limitation occurs because the firm may not record the benefit associated with uncertain tax positions in its financial statements in the same year that it is taking the aggressive tax position on its returns. As a result, the firm s tax expense in that year is a noisy indicator of how aggressive the firm was on its actual corporate tax return. Third, until recently, GAAP required firms to record the tax benefit associated with the deduction for employee stock options directly into equity rather than reducing the firm s current tax expense. 7 If a firm has significant stock option deductions then ETR overstates the firm s actual tax burden. 3.2 Cash Effective Tax Rate (CASH_ETR) Our second tax aggressiveness measure is the long-run cash effective tax rate (CASH_ETR) introduced by Dyreng et al. (2008). CASH_ETR is calculated as the ratio of the five-year sum of cash taxes paid to the five-year sum of pretax financial accounting income less special items. Dyreng et al. (2008) note this measure of tax aggressiveness has several advantages over the traditional ETR measure. First, CASH_ETR is not affected by changes in the firm s tax contingencies. Regardless of whether the firm records a benefit associated with an 7 See Hanlon and Shevlin (2002) for a discussion of the accounting for the tax benefit of employee stock options. 15

17 aggressive tax position in its financial statements, the reduced cash tax payments that result from the aggressive position will be reflected in a lower CASH_ETR. Second, to the extent firms aggressively accelerate expenses or defer income for tax purposes, these activities will be reflected in a lower CASH_ETR provided those timing differences do not reverse within the fiveyear period over which CASH_ETR is calculated. Finally, the CASH_ETR measure will be reduced by the tax benefit associated with employee stock options and therefore provides a better measure of the firm s true tax burden than the traditional ETR measure. Despite these advantages, CASH_ETR still contains some measurement error, as this measure does not control for nondiscretionary sources of book-tax differences (e.g., depreciable and amortizable assets) and is biased downward for those firms that consistently manage their pretax book-income upward over extended periods of time. Moreover, CASH_ETR is designed to capture a broad array of tax planning activities, which may or may not be considered aggressive in nature. 3.3 Book-tax Differences (BTDs) Our third proxy for tax aggressiveness is an estimate of the difference between pretax book income and taxable income (i.e., book-tax differences). There are a number of studies that suggest book-tax differences can be used as a signal of tax aggressiveness. Mills (1998) finds that proposed IRS audit adjustments are positively related to large positive book-tax differences. Desai (2003) posits that the growing difference between book and taxable income during the 1990 s was caused by increased levels of tax sheltering. In addition, Wilson (2008) finds that book-tax differences are positively associated with actual cases of tax sheltering. Despite evidence that large positive book-tax differences are associated with aggressive tax reporting, this measure also has limitations. Hanlon (2003) and Manzon and Plesko (2002) 16

18 identify firm specific characteristics that generate book-tax differences, but are not necessarily reflective of aggressive tax reporting. For example, firms with large capital expenditures could have significant book-tax differences associated with depreciation, but these differences would not be reflective of aggressive tax strategies. In addition, Phillips, Pincus, and Rego (2003) and Hanlon (2005) demonstrate how temporary book-tax differences can reflect earnings management activities. To the extent that earnings management and innate firm characteristics unrelated to aggressive tax reporting are the primary determinants of book-tax differences, booktax differences will be a noisy proxy for tax aggressiveness. Book-tax differences (BTD) are computed as pre-tax book income less an estimate of taxable income. We calculate taxable income by grossing up the sum of current federal tax expense and current foreign tax expense and subtracting the change in net operating loss carryforward. If the current federal tax expense is missing, total current tax expense is calculated by subtracting deferred tax expense, state tax expense, and other income taxes from total tax expense. We measure book income as pre-tax book income less minority interest. 3.4 Permanent BTDs (PERM_BTD) Many aggressive tax reporting strategies result in permanent book-tax differences. The majority of the cases of tax sheltering examined by Wilson (2008) resulted in permanent booktax differences. Further, the U.S. Congress Joint Committee on Taxation (1999), Weisbach (2002), and Shevlin (2002) describe the ideal tax shelter as creating permanent, rather than temporary, book-tax differences. Consequently, we also include permanent book-tax differences (PERM_BTD) as a measure of tax aggressiveness in this study. The primary advantage of this measure is its exclusion of temporary differences that can reflect earnings management activity and/or book-tax rule differences unrelated to tax planning. We calculate PERM_BTD as total 17

19 book-tax differences less deferred tax expense grossed-up by the applicable federal statutory tax rate. 3.5 Discretionary Permanent BTDs (DTAX) Our fifth proxy for tax reporting aggressiveness is a measure of discretionary permanent book-tax differences. Frank, Lynch, and Rego (2008) calculate discretionary permanent differences (DTAX) by regressing permanent book-tax differences on nondiscretionary items unrelated to tax planning that are known to cause permanent differences. The nondiscretionary items include intangible assets and state tax expense, among others. The residual from this regression is then used as a proxy for tax reporting aggressiveness. See Appendix A for a complete description of the model used to calculate the DTAX variable. Similar to total book-tax differences in Wilson (2008), Frank et al. (2008) find that their measure of discretionary permanent differences is associated with actual cases of tax sheltering. 3.6 Predicted Tax Shelter Firms Our final measure of tax aggressiveness is based on a model developed by Wilson (2008) to identify active tax shelter participants. The model of tax sheltering was developed using a sample of identified tax shelter participants and is composed of variables predicted to be associated with tax shelter activity. We use the model of tax sheltering to rank firm-year observations into quintiles based on the probability the firm is engaged in tax shelter activity in year t. Firm-year observations ranked in the top quintile are then designated as predicted tax shelter firms (SHELTER). See Appendix A for a complete description of the tax shelter model. In summary, we utilize six different measures of tax reporting aggressiveness because no single measure perfectly captures tax aggressiveness. Moreover, each measure captures varying degrees of tax aggressiveness. The proxies discussed earlier (ETR, CASH_ETR, and 18

20 BTD) likely reflect less aggressive tax planning than the proxies discussed later (PERM_BTD, DTAX, and SHELTER). Thus, to the extent we obtain similar results across these measures of tax aggressiveness we can be confident that our results are highly robust. 4. Research Design 4.1 Tests Linking Executive Compensation and Tax Aggressiveness Hypothesis 1 predicts that tax reporting aggressiveness is associated with the level of executive compensation. We base our tests of hypothesis 1 on Core et al. (1999) and Core, Guay, and Larcker (2008), which model total compensation as a function of the economic determinants of such compensation. In particular, we regress the natural log (Log) of total CEO (CFO) compensation on proxies for CEO and CFO experience, firm size, growth opportunities, stock and accounting returns, and the standard deviation of stock and accounting returns in the prior five years. Recall that we examine CEO compensation separately from CFO compensation, as CFOs have greater responsibility for the accounting function, relative to CEOs. Log(TOTAL COMP t ) = β 0 + β 1 Log(TENURE t ) + β 2 Log(SALES t-1 ) + β 3 S&P500 t + β 4 BTM t-1 + β 5 RETt + β 6 RET t-1 + β 7 ROA t + β 8 ROA t-1 + β 9 σ(ret) t-1 + β 10 σ(roa) t-1 + β 11 LOSS t + β 12 OCF t + β 13 MNC t + β 14 TA t + β 15 YEAR + β 16 INDUS + ε τ (1) where TOTAL COMP is a measure of CEO (CFO) salary, bonus, long-term incentive plan payouts, the value of restricted stock grants, the value of options grants during the year, and any other annual pay; TENURE is the logarithm of the CEO s (CFO s) tenure at the firm; SALES is the logarithm of total sales; S&P500 is 1 if the firm is in the S&P 500, and 0 otherwise; BTM is the firm s book-to-market ratio; RET is the firm s stock return; ROA is income before 19

21 extraordinary items divided by average total assets; and σ(ret) [σ(roa)] is the standard deviation of annual stock returns (return on assets) over the five prior years. Core et al. (2008) base their model on prior research that examines the economic determinants of executive compensation, including Smith and Watts (1992), Core et al. (1999), and Murphy (1999). These studies generally find that total CEO compensation is positively related to firm size (as measured by total sales) and current and prior year stock returns, and negatively related to the standard deviation of ROA and the book-to-market ratio. Core et al. (2008) also show that firms in the S&P 500 pay more total compensation to their CEOs. We expand the model developed by Core et al. (2008) to include additional variables that could be associated with both executive compensation and our measures of tax aggressiveness, including any current year losses, multinational activity, and net operating cash flows. In particular, LOSS is 1 if the firm reports negative income before extraordinary items, and 0 otherwise; OCF is operating cash flow divided by average total assets; and MNC is 1 if the firm reports income from foreign operations, and 0 otherwise. We also estimate equation (1) with fixed-effects for year (YEAR) and 2-digit SIC codes (INDUS). The focus of our analysis is on our measures of tax aggressiveness, TA. Our first hypothesis predicts that managerial compensation is positively associated with tax reporting aggressiveness (i.e., β 14 is significantly greater than 0). Our data comes from three sources. We obtain CEO and CFO compensation data from the Execucomp database. We use CRSP data to calculate stock returns, while all other data are obtained from the Compustat database. For a firm-year observation to be included in our sample it must have all the data necessary to calculate the variables described in equation (1). The sample for our first set of tests consists of 18,827 CEO-year observations from 1992 through 20

22 2006 that were successfully matched to CRSP and Compustat firm-year data. All variables are winsorized at the 1 st and 99 th percentiles. Panel A of Table 1 provides descriptive statistics for the variables used in our first set of tests. We tabulate continuous data for our measures of tax aggressiveness in Table 1, but include the quintile indicator variables for each measure of tax aggressiveness in all subsequent tables. The median sample CASH_ETR is 27 percent, compared to a median ETR of 33 percent. The median firm is profitable and has positive book-tax differences, BTD. We calculate the SHELTER variable using all Compustat firm-years from 1992 through 2006 that have the requisite data and then match these firm-years to the CEO (CFO) data on ExecuComp. As a result, the SHELTER indicator variable is set equal to one for more than 20 percent of our sample firm-years. Similarly, we calculate the DTAX variable using all Compustat firm-years from 1992 through 2006 that have the requisite data and then match these firm-years to the ExecuComp observations. As a result, the mean DTAX is not equal to zero despite being the residual from annual cross-sectional regressions. 8 Panel B presents the Pearson and Spearman correlation coefficients for each of the six tax aggressiveness measures. Consistent with expectations, each of the measures is significantly correlated and in the anticipated direction. INSERT TABLE 1 HERE We examine the basic relation between CEO compensation and tax reporting aggressiveness in Figure 1, which contains a graph of average (logged) total CEO compensation across firms ranked based on two measures of tax aggressiveness, ETR and CASH_ETR. In particular, all firm-year observations for CEO compensation are placed into 20 groups based on 8 See Appendix A for a complete discussion of the calculation of the SHELTER and DTAX variables. 21

23 separate rankings of ETR and CASH_ETR. 9 Figure 1 shows that total CEO compensation generally increases as the level of tax aggressiveness increases (i.e., as CASH ETR and ETR decrease). However, for groups 18, 19, and 20, which contain the highest levels of tax aggressiveness, this trend reverses and total compensation declines as tax aggressiveness continues to increase. This result suggests that while compensation committees may generally reward CEOs for more aggressive tax reporting, there appears to be a point at which this is no longer the case. This decrease in CEO compensation for the highest levels of tax aggressiveness is consistent with extreme levels of tax aggressiveness being viewed as value-destroying. At such high levels, the costs of aggressive tax planning may outweigh the benefits and, as a result, executives are actually punished in the form of lower compensation for engaging in overly aggressive strategies. INSERT FIGURE 1 HERE Table 2 presents the results of our estimation of equation (1) for the CEO compensation sample using each of the alternative measures of tax aggressiveness. Consistent with prior research we find that total compensation is positively related to CEO tenure (Log(TENURE)), firm size (Log(SALES)), S&P 500 membership, current and lagged annual stock returns (RET t, RET t-1 ), and the standard deviation of stock returns σ(ret) t-1, and negatively related to the standard deviation of return on assets (σ(roa) t-1,) and the book-to-market ratio (BTM). Further, we observe a significant positive relation between the existence of foreign operations (MNC) and executive compensation. This finding is consistent with executives receiving greater compensation for managing more complex, international operations. We also observe a significant positive association between total compensation and tax aggressiveness for each of 9 We remove loss firms from this analysis prior to ranking the firms into the 20 groups because loss firms are likely to have very low ETR measures. However, the results are generally consistent with those in Figure 1 when the loss firms are included. 22

24 the tax aggressiveness measures. These findings support hypothesis 1, which predicts an association between executive compensation and tax aggressiveness, even after controlling for standard economic determinants of executive compensation. Thus, our measures of tax aggressiveness do not simply reflect tax savings, since our measures of tax aggressiveness are significant after controlling for current (ROA, OCF) and anticipated (RET) firm performance. We explore the cause of this association when we test hypotheses 2 and 3 below. INSERT TABLE 2 HERE Table 3 presents the results for estimating equation (1) using the CFO compensation sample, which is composed of 13,590 CFO-year observations that were successfully matched to Compustat and CRSP firm-year data. Consistent with the results regarding CEO compensation, we find that CFO total compensation is positively associated with CFO tenure, firm size, S&P 500 membership, stock returns, the standard deviation of stock returns, and the existence of foreign operations, and negatively related to the standard deviation of return on assets and growth opportunities. CFO compensation is significantly associated with five of the six measures of tax aggressiveness, with the only exception being DTAX. Similar to the results for CEOs, the results in Table 3 suggest that CFOs are also rewarded for engaging in aggressive tax reporting. INSERT TABLE 3 HERE 4.2 Tax-Related Compensation and Corporate Governance To test our second hypothesis, we modify equation (1) to include a variable that indicates the presence of weak corporate governance (WEAK_GOV), as well as that term s interaction with tax aggressiveness (TA WEAK_GOV): 23

25 Log(TOTAL COMP t ) = β 0 + β 1 Log(TENURE t ) + β 2 Log(SALES t-1 ) + β 3 S&P500 t + β 4 BTM t-1 + β 5 RET t + β 6 RET t-1 + β 7 ROA t + β 8 ROA t-1 + β 9 σ(ret) t-1 + β 10 σ(roa) t-1 + β 11 LOSS t + β 12 OCF t + β 13 MNC t + β 14 WEAK_GOV t + β 15 TA t + β 16 TA t WEAK_GOV t + β 17 YEAR + β 18 INDUS + ε τ (2) where WEAK_GOV is 1 for firm-years with weak governance (as defined below), and 0 otherwise. Our second hypothesis predicts that if managers engage in aggressive tax reporting to extract rents from the firm (i.e., if tax aggressiveness is related to excess compensation), then β 16 should be positive, consistent with managers at firms with aggressive tax reporting and weak corporate governance receiving more compensation than managers at firms with aggressive tax reporting and stronger corporate governance. Because of the difficulty in measuring the quality of a firm s governance structure we estimate equation (2) using two alternative measures of corporate governance. We start with the Gompers et al. (2003) index of shareholder rights, the corporate governance metric employed by Desai and Dharmapala (2006) and Wilson (2008). 10 This measure is constructed based on 24 different provisions that can be classified into five categories tactics for delaying hostile bidders, voting rights, director and officer protection, other takeover defenses, and state laws. Each of these categories represents potential determinants of a firm s takeover vulnerability. The Gompers et al. (2003) index ranges from 0-24, where low scores indicate a lower degree of insulation for managers from hostile takeovers and therefore a higher quality of corporate governance. We expect that well-insulated managers (i.e., those at firms with higher scores and lower quality corporate governance) will have more opportunities to engage in tax shelters for personal gain. GSCORE is an indicator variable set equal to one for firms with a Gompers et al. 10 Both of these studies consider interactions between tax reporting aggressiveness and corporate governance strength. 24

26 (2003) index score above the sample median (and thus indicates weaker governance), and set equal to zero for all other observations. Our second proxy for corporate governance measures the number of independent blockholders with at least 5 percent stock ownership. In their study of the impact of corporate tax avoidance on firm value, Desai and Dharmapala (2007) also rely on blockholder data as a proxy for the strength of corporate governance. Jensen (1993) contends that blockholders play an important role in a firm s governance structure because they are independent and have a financial interest that gives them incentives to monitor management. Because blockholders control significant amounts of stock they can use their voting power to pressure management if they believe managers are behaving opportunistically. BLOCK is an indicator variable set equal to one for firm-years where the number of independent blockholders owning greater than 5 percent of the stock of a particular firm is below the sample median (and thus indicates weaker governance). BLOCK is set equal to zero for all other observations. Blockholder data was obtained from the Wharton Research Data Services website (WRDS) and is only available for the years Thus, when we match the blockholder data to our CEO-year observations, our sample size is reduced to 5,147 observations. Table 4 presents the results of estimating equation (2) for the CEO sample, including two of our six measures of tax aggressiveness. 11 We first estimate equation (2) using the CASH_ETR measure and find that it continues to be significantly positively associated with total compensation. In contrast, its interaction with GSCORE (CASH_ETR GSCORE) is negative and significant. This result does not support our second hypothesis. In fact, it is consistent with 11 We tabulate results for just two of the six measures of tax aggressiveness to minimize the length of our manuscript. Consistent with the results reported in Table 4, the interaction terms between corporate governance and the remaining four measures of tax aggressiveness are not significant in the predicted direction for any of the four measures. 25