The Incentives for Tax Planning

Christopher S. Armstrong
[email protected]

Jennifer L. Blouin
[email protected] The Wharton School University of Pennsylvania 1300 Steinberg-Dietrich Hall Philadelphia, PA 19104-6365

David F. Larcker
[email protected] Stanford University Graduate School of Business Rock Center for Corporate Governance Stanford, CA 94305-5015

Draft: March 8, 2011

This paper has benefitted from helpful comments provided by Michelle Hanlon (editor), Alan Jagolinzer, an anonymous referee, and the workshop participants at Boston College, Duke University, Temple University, Texas A&M, the University of Missouri, the University of Oregon, and the University of Washington. We also gratefully acknowledge the research support provided by Jim Feeney. Christopher Armstrong is grateful for financial support from the Dorinda and Mark Winkelman Distinguished Scholar Award.

Electronic copy available at: http://ssrn.com/abstract=1416422

The Incentives for Tax Planning Abstract: We use a proprietary data set with detailed executive compensation information to examine the relationship between the incentives of the tax director and GAAP and cash effective tax rates, the book-tax gap, and measures of tax aggressiveness. We find that the incentive compensation of the tax director exhibits a strong negative relationship with the GAAP effective tax rate, but little relationship with the other tax attributes. We interpret these results as indicating that tax directors are provided with incentives to reduce the level of tax expense reported in the financial statements. Keywords: tax director incentives; equity incentives; effective tax rate; book-tax difference JEL Classification: H25, M41, M52

2
Electronic copy available at: http://ssrn.com/abstract=1416422

1. Introduction This paper investigates the type of tax planning included in tax directors compensation contracts. Using detailed data on tax director incentive compensation, we examine whether the incentives provided to tax directors are associated with lower effective tax rates and/or a wider book-tax gap. We also examine whether tax director incentives are more strongly linked to measures of the impact of taxes on cash flows (i.e., cash effective tax rates and taxable income) or earnings (i.e., GAAP effective tax rates and pre-tax book income). Finally, we investigate whether tax director incentives are associated with proxies for aggressive tax planning. Our study complements and extends the recent literature that links tax planning with top executive incentive compensation (e.g., Desai and Dharmapala, 2006; Rego and Wilson, 2010) and executive/corporate culture (e.g., Frank et al., 2009; Dyreng et al., 2010).1 Although these studies show that both the incentives of top management and corporate culture have a significant impact on effective tax rates, it is not clear whether top management directly engages in tax planning or whether the documented tax effects are a byproduct of the investing, financing, and operating decisions made within the firm. For example, an observed relationship between top managements incentives and measures of tax planning may be an indirect outcome of the investment decision rather than from deliberate tax planning. To understand the link between performance measures and observed tax attributes, it is necessary to explicitly consider the incentive compensation of the tax director, who is the executive charged with overseeing the firms tax function. However, an alternative scenario is that all tax planning decisions are
This literature suggests that a firms culture, representing the shared beliefs within the firm regarding the optimal course of action, can lead to endemic aggressiveness in all of a firms business decisions, including the tax function (e.g., Cronqvist et al., 2007).
1

1
Electronic copy available at: http://ssrn.com/abstract=1416422

merely dictated by top management to the tax director, who has no ability to independently affect firm behavior. If tax planning is a role merely encompassed in top managements duties, then we might not observe an association between tax director incentives and measures of tax planning after controlling for the incentives of top management. To investigate the relative importance of various measures of tax planning in tax director incentive compensation contracts, we use a proprietary data set that includes detailed compensation information for many executives, including the members of the tax department, for a sizeable sample of mostly large U.S. firms. This unique data exhibits several advantages relative to that used in prior research. First, it allows us to directly measure the incentive compensation of the executives who are directly responsible for the firms tax function. Second, it provides us with information about the incentive compensation of other members of the management team. This within-firm benchmark allows us to identify the attributes of the compensation plan that are unique to tax directors, as opposed to the general compensation policy of the firm. We first examine the association between executive incentives and the GAAP and Cash effective tax rates (ETRs). Recent research presents evidence that variation in corporate tax planning is captured by ETRs (Dyreng et al., 2008; Robinson et al., 2010). We find a strong negative relationship between tax director incentives and the GAAP ETR and no relationship between tax director incentives and the Cash ETR. These results indicate that tax directors are provided with incentives to reduce the tax expense reported in the financial statements. Thus, our findings are consistent with the GAAP ETR being a

more informative measure of the tax directors actions and, accordingly, it is allocated more weight in the incentive compensation contract We next examine the book-tax gap, because this measure has recently received considerable attention from the U.S. Treasury, regulators, and academics. We find no evidence of a relationship between tax director incentive compensation and the spread between pre-tax book income and taxable income. Since the book-tax gap is defined as the difference between pre-tax book income and taxable income, there are econometric reasons to disaggregate this variable into its two components (i.e., pre-tax book income and taxable income). In addition, separating the book-tax gap allows us to look for evidence of conforming tax planning (i.e., where tax planning reduces both book and tax income). The results of this analysis provide no evidence of an association between the incentive compensation of tax directors and lower taxable income. Finally, we examine the relationship between tax director incentive compensation and two measures developed to explicitly capture tax aggressiveness: a modified version of Frank et al.s (2009) DTAX measure and Wilsons (2009) measure of tax sheltering. We find no evidence that tax directors incentive compensation is associated with either measure of tax aggressiveness. Collectively, our results indicate that tax directors are provided incentives to reduce the level of tax expense reported in the financial statements, but provide little support for the notion that tax directors are compensated on the basis of the book-tax gap, the cash effective tax rate, or various measures of tax aggressiveness. The remainder of the paper consists of the following seven sections. Section 2 provides a description of the possible roles of the tax director, and Section 3 reviews the

prior empirical literature on tax planning. Section 4 describes our proprietary sample and other data used in our analysis. We develop our variable measurement and methodological approach in Section 5. Our results are presented in Section 6, and sensitivity analyses are provided in Section 7. Conclusions and summary comments are provided in Section 8.

2. Role of the Tax Director Although tax directors are responsible for one of the firms largest outflows of cash and one of the largest expenses on the income statement, little is known about how these executives are compensated. Since tax directors are rarely among the top five highest paid executives, information about their annual compensation and equity holdings is not available in annual proxy (Form DEF 14A) filings. Therefore, researchers are typically unable to directly observe the parameters of tax director incentive pay. The tax director fills at least three roles. First, the tax director is responsible for compliance. Since multinational firms are typically required to file thousands of tax forms annually, it is not unreasonable to infer that compliance is the tax directors primary duty. Second, the tax director may serve as an advisor to the firms senior executives by providing expertise in minimizing the tax cost of the firms operating, financing, and investing activities. As an advisor, the tax director would be present when strategic investment decisions are made but may not be responsible for selecting the investments. Third, the tax director can be charged with actively pursuing tax planning opportunities by generating investment opportunities where the net present value of the project derives solely from tax benefits. We view this third role, termed the active

planning role, as relatively more tax aggressive than the other two advisory roles.2 It is difficult to empirically distinguish the advisor from the active planner, as both roles could yield lower tax obligations, and there can easily be an overlap in duties. In terms of selecting the compensation contract for the tax director, traditional agency theory suggests that compensation should be based on performance measures that are controllable by the agent (Holmstrom, 1979; Lambert, 2001). Therefore, if the tax directors primary responsibility is compliance, then the incentive component of pay should not be based on cash flow or earnings objectives (i.e., there are more direct measures of tax director effort, such as fines paid for noncompliance). On the other hand, the incentive compensation of a tax advisor and an active planner should be a function of the firms financial attributes, since these performance measures are, in part, controllable by these types of tax directors. Our characterization of the various roles of the tax director is consistent with the Tax Executive Institute, Inc.s (TEI) 2004-2005 Corporate Tax Department Survey, which provides information on several dimensions of the operations of firms tax departments.3 This document is the compilation of the responses of approximately 1,300 tax departments (of which 945 were at publicly traded firms), representing a 57% response rate, to a mid-2004 survey generated by a consulting firm. Although the survey

Clearly, the advisory role can also generate positive net present value for the firm. However, we view the advisor as effectively reducing the tax cost of an incremental investment opportunity (i.e., the net present value of the project includes non-tax components). A tax advisor does not typically undertake transactions that fail the business purpose test (i.e., there are non-tax benefits to the transaction). In contrast, an active planner will need to justify why there are benefits beyond the tax attributes of the transaction (see Gregory v. Helvering, 293 U.S. 465 (1935)). As Hanlon and Heitzman (2010) point out, there is no universally accepted definition of tax aggressiveness. So we are left to make an (albeit coarse) attempt at segregating tax planning from tax aggressiveness for the purposes of our analysis. 3 The Tax Executives Institute is a professional organization of roughly 5,000 business executives who are responsible for taxation matters. The TEI survey data is separate from the proprietary data used later in our analysis.

focuses on identifying the specific duties of the tax departments (e.g., time spent on foreign versus domestic compliance, types of entities responsible for, and nexus issues), it does provide descriptive information on the organization of the tax department and some performance metrics of tax department personnel. The senior tax executives of the public survey respondents typically hold the title of tax director (40%) or vice president (35%) (described in aggregate hereafter as tax directors), and those titles suggest that these individuals hold high level positions in their respective firms. Sixty-five percent of the tax directors report to the chief financial officer. The majority (60%) of public companies report that their tax function is centralized and globally integrated, so the tax director is responsible for taxes worldwide. In terms of specific tasks, the majority of the tax directors time is spent doing research (33%) and compliance (28%).4 The average tax director spends 9% of his or her time on the tax provision and 8% of his or her time on tax audits. In terms of the average tax departments budget, 60% is allocated toward compensation and 20% toward outside consultants. Finally, tax directors appear involved in the overall operations of the firm, as the survey reports that tax directors spend 9% of their time aiding in operational issues of the firm (including governance and Sarbanes-Oxley compliance). The TEI Survey also provides some descriptive information on the performance metrics of a firms top tax executive. Seventy-one percent of the surveys senior tax executives of public companies reported that lack of surprises was used by senior management to evaluate their performance. Since the effective tax rate (63%) is of slightly more importance than generating cash savings (57%) for performance evaluation, it appears that income statement surprises could potentially dominate cash flow surprises.
4

Note that time on research includes 8.1% of their time related to mergers and acquisitions.

Note that the cash savings is not necessarily solely attributable to taxes, as only 49% of respondents indicated that they were evaluated on the basis of cash taxes. Although anecdotal, our conversations with 12 tax directors suggest that their primary performance metric was economic value added, as measured by financial statement earnings. However, only 19% of the tax directors surveyed agreed. Finally, the TEI survey reports that only 30% of the incentive compensation of executives other than the tax director was based on after-tax measures. Our analysis is based on the maintained assumption that the tax director is not the primary executive charged with selecting the firms investment activity. Although tax directors are involved in transaction planning and investment location decisions, their role appears to be primarily as an advisor. This characterization is consistent with the TEI survey evidence documenting that the majority of the tax directors time involves tasks related to return compliance and research. In later tests, we attempt to separate out the advisory role from the active planner role by studying specific measures intended to proxy for aggressive tax planning. To the extent that there is overlap between the duties of tax directors and other senior executives, our analyses should yield similar associations between our measures of tax planning and the incentives of the other executives. As we discuss below, the statistical differences observed across executive positions is a key feature of our methodological approach.

3. Prior Literature 3.1. Executive incentives and financial and tax reporting In contrast to the fairly extensive literature on incentives and financial misreporting, there is relatively little work linking executives incentives with tax

aggressiveness.5 Slemrod (2004) develops the idea that shareholders select the level of tax aggressiveness by linking tax manager compensation with effective tax rates or stock price. However, one serious limitation in this model is that shareholders cannot observe the compensation contract or know whether managers are engaging in legal tax planning or illegal tax evasion. Thus, inappropriate aggressive behavior by the tax director constitutes a hidden action, because tax returns are not disclosed to investors (see Crocker and Slemrod, 2005), and shareholders do not know whether to alter the executives compensation contract until after the firm is penalized. In addition, it is difficult to contract on tax evasion, since the behavior is illegal and therefore would render any contract that is a function of this outcome unenforceable by the courts (Chen and Chu, 2005). Ultimately, Slemrod (2004) suggests that corporate tax noncompliance (tax evasion) could be the result of the design of incentive compensation plans. For the purposes of our study, a shortcoming of the theoretical literature investigating the link between incentives and tax planning is that it fails to address the fact that large public firms are typically not assessed fines or penalties upon audit. Most settlements are made with the taxing authorities at a fraction of the dollar amount of the original assessment.6 Therefore, non-evasive but aggressive tax planning can be viewed

See Armstrong et al. (2010) for a review of the literature on executives incentives for financial misreporting. Much of the extant work focuses on whether equity incentives align managers interests with respect to financial reporting with those of shareholders, or whether equity incentives instead induce managers to manipulate accounting information for personal gain. More recently, a number of studies, including Feng et al. (2009) and Jiang et al. (2010), examine the relationship between the incentives of the CFO and earnings management. Jiang et al. (2010) argue that because CFOs primary responsibility is financial reporting CFO equity incentives should play a stronger role than those of the CEO in earnings management. This rationale for examining CFOs incentives is analogous to our reasoning for examining the relationship between the incentives of the tax director and the firms tax attributes. 6 For example, GlaxoSmithKline PLC (Glaxo) settled a transfer pricing audit with the IRS for $3.4 billion. According to Glaxos 20-F filings and the IRSs press release, the payment was made to cover back taxes and interest no penalties were assessed. Press reports suggest that Glaxo settled for about 60% of the original assessment. Further, Glaxo reported in its 20-F filing that the after-tax cost of the settlement was $3.1 billion, suggesting that interest was roughly $1 billion of the settlement (assuming a 30% tax rate).

as merely a borrowing arrangement with tax authorities. These models also do not consider the nature of the underlying tax planning (temporary versus permanent tax deductions) and the tax plannings impact on the financial statements. The reversal of a permanent tax deduction will have a much larger impact on a firms GAAP ETR than will a reversal of a temporary tax deduction. Hence, we need models that jointly consider both the cash flows and earnings implications of tax planning when considering optimal tax director incentives. Finally, note that both Crocker and Slemrod (2005) and Chen and Chu (2005) focus on incentives and tax evasion rather than on linking incentives to general tax director planning duties that affect either earnings or cash flows. Desai and Dharmapala (2006) develop a model linking equity-based compensation and aggressive tax planning. They argue that there are complementarities between tax sheltering and rent extraction that imply that incentives lead to more sheltering behavior in better governed firms. Although equity incentives can motivate managers to increase tax sheltering, this decision will also lead to fewer opportunities for managers to engage in rent extraction. Thus, the relationship between equity-based compensation and tax sheltering is theoretically ambiguous.7 There is also a growing related literature that investigates whether corporate culture and individual managers influence a firms level of tax aggressiveness. Frank et
Since the dispute covered the period from 1989 to 2005, a rough estimate of the pre-tax interest rate on the settlement is 4.5% (or 3.2%, after taxes), assuming an eight year interest accrual period. If we consider that the IRS effectively forgave 40% of the assessment, then the effective borrowing rate is actually negative, since the initial loan from the government was for $4 billion (= 2.4 / 0.6) and the final payment was only $3.4 billion. In addition, the IRS Data Book (see Tables 9a and 17 for 2008 and 2009) provides an analysis of the amount of penalties assessed. Overall, for incorporated business, penalties appear to constitute only 1.45% (0.56%) of the additional taxes due in 2008 (2009). In addition, 32% (40%) of the corporate accuracy-related penalties assessed are abated in 2008 (2009). 7 Desai and Dharmapala (2006) argue that any directional prediction for the association between tax planning and equity incentives is conditional on the firms governance structure. However, it is not clear how the tax director would be able to extract rents from the firm. As pointed out in footnote 6, even detected aggressive tax planning can be a positive net present value project that benefits shareholders.

al. (2009) argue that a positive relationship between aggressive financial and tax reporting is consistent with a generally aggressive corporate tone and culture. Ultimately, they find evidence of a positive relationship between earnings management and tax planning and that the market appears to reward this aggressiveness. However, they do not find meaningful evidence that the CEO and CFO are provided with incentives to undertake such behavior. Phillips (2003) investigates whether compensating managers on a pre- versus after-tax basis affects the firms reported tax expense. Using a proprietary dataset, he finds that compensating business unit managers on an after-tax basis is associated with lower firm GAAP effective tax rates. Consistent with the finding that equity risk-taking incentives motivate managers to undertake more aggressive tax planning, Rego and Wilson (2010) find a positive association between option vega and measures of corporate tax aggressiveness. Dyreng et al. (2010) investigate whether specific members of the top management team are associated with firms level of tax aggressiveness. Although they conclude that top management is associated with tax planning, it is not clear whether their results are attributable to top managements explicit setting of the tone at the top with regard to tax aggressiveness or making strategic decisions such as investment and financing policies that are highly correlated with a firms tax position. Finally, Robinson et al. (2010) attempt to measure tax manager incentives by determining whether the tax department is viewed as a profit center (i.e., a contributor to the bottom line). Although an indirect measure of managerial incentives, this is one of the few studies to consider the incentive compensation plan of the tax manager (as opposed to the CEO, CFO, or other senior executives). They find evidence that firms

10

with tax departments that are viewed as profit centers have lower GAAP ETRs, but they have only weak evidence that firms with tax departments that are viewed as profit centers have lower cash ETRs. Collectively, prior literature provides limited evidence that managerial incentives influence tax planning choices. However, there is no evidence related to the precise incentives of the tax director, who is the person most directly involved in the tax decisions of the firm. Thus, our subsequent empirical analysis focuses primarily on the compensation and incentives of the tax director and the impact of these incentives on firms various tax attributes. However, we are careful to control for the incentives of other high level executives who might also be expected to play a role in the firms tax planning, which allows us to isolate the unique effect of the tax director on the firms various tax attributes. 3.2. Book-tax gap Treasury (1999), Desai (2003), and Boynton et al. (2005) all document the growth of the spread between aggregate financial statement income and aggregate taxable income as reported by the IRS (the book-tax gap).8 Book income in excess of taxable income is consistent with manipulation of earnings reported to the capital markets, tax aggressiveness, or some combination of these two activities.9 The discretion available in GAAP provides managers with an opportunity to manage book earnings upward without necessarily affecting taxable income (e.g., Badertscher et al., 2009). Similarly, Mills (1998) and Desai (2003) suggest that the expanding book-tax gap appears to be consistent
Graham, Raedy and Shackelford (2010) use financial statement data to estimate the BTG from 1993 to 2008 and find that book income has exceeded estimated taxable income in all years except 2001 and 2008 (i.e., years with excessive losses). 9 Seidman (2010) conjectures that the growing book-tax gap could also be due to changes in accounting methods over time.
8

11

with aggressive tax planning. Phillips et al. (2003), Hanlon (2005), and Ayers et al. (2006) provide evidence that the divide between tax and financial reporting is partially attributable to earnings management. Finally, Frank et al. (2009) find evidence suggesting that firms that manage earnings also aggressively manage taxes. Prior literature has not reached a consensus about the source of the increase in the book-tax gap.10 For the purposes of our study, examining the link between executive incentives and the book-tax gap can provide new insights into whether firms appear to compensate managers in general, and tax directors in particular, in a manner that potentially induces a wider book-tax gap.

4. Sample Our sample is derived from the proprietary data files provided to us by a large human resources consulting firm. The data covers the fiscal years from 2002 to 2006 and is obtained by a detailed annual survey by the consulting firm. The proprietary data provides information on the major components of executive compensation for many executive positions, including the tax director. We retain all public firms for which there is complete compensation data for CEO, CFO, general counsel, and tax director, and where we can estimate all of the tax attributes and control variables. Our final sample consists of 423 unique firms and 1,162 firm-year observations that are nearly uniformly
There continue to be many unanswered questions about the book-tax gap. It is frequently argued that tax aggressiveness must have increased over the mid- to late-1990s, as reported corporate profits surged more than did reported taxable income. However, it is unclear why firms would begin to seek to lower their tax burden more during this period rather than during others. What changed during this time period? While the sophistication of capital markets has increased over time, yielding new sheltering opportunities, the extent of the increase in the book-tax gap seems to far exceed documented examples of product sheltering (see Graham and Tucker, 2006, and Wilson, 2009). Furthermore, many new sheltering products were created because Treasury had eliminated an original transaction (e.g., the elimination of Boss led to the creation of Son of Boss).
10

12

distributed across the sample period (244, 239, 222, 234, and 223 observations from 2002 to 2006, respectively). The industry composition of the sample relative to all firms on Compustat for the same sample period (i.e., 2002 to 2006) is provided in Table 1, Panel A. Relative to the full Compustat sample, our sample of firm years includes a greater proportion of firms from Food, Textiles, and Chemicals and a somewhat smaller proportion of firms from Financial Services, but is otherwise very similar. In Panel B of Table 1, we report the representation of our sample in both the population of Compustat and the S&P 500. Across all years, our sample represents approximately 30% of total assets and 33% of the total cash paid for taxes of firms in Compustat. In addition, although our sample includes roughly 50% of the firms in the S&P 500, it includes 65% of the total assets and 60% of the total cash paid for taxes. Thus, our sample captures a large fraction of the economic activity (and taxes paid) in the U.S. economy as a whole and S&P 500 in particular. Descriptive statistics for our sample are reported in Table 2. The mean (median) revenue and market capitalization for our sample is $17,817 ($8,211) million and $23,555 ($9,173) million, respectively. Our sample firms are profitable, reporting both mean and median return on assets of approximately 5%. Consistent with other compensation and tax planning studies, our sample firms are highly levered, with a mean (median) debt-to-assets ratio of 20% (18%). Finally, our sample firms report mean taxable income of $1,270 billion on pre-tax income of $1,838 million. Thus, our sample is composed of very large firms that (as we will see in Section 5) have sizeable tax burdens and, therefore, a relatively greater proclivity to engage a tax director for tax planning.

13

5. Variable Measurement and Methodological Approach 5.1. Executive incentives We have detailed data about the annual salary, annual bonus, annual restricted stock and option grants, and expected payouts from long-term incentive plans (e.g., stock options, restricted stock, performance units, and performance shares) for each executive in our sample. We use this information to calculate two complementary measures of incentive compensation. Our first measure is the total value of annual executive compensation, which is computed as the sum of all of the components of compensation, where restricted stock is valued using the market value at the date of the grant and stock options are valued using the Black-Scholes risk-neutral value.11 This approach provides one measure of an agents incentives, since total annual flow compensation should vary with performance measures that are used in the compensation contract (Lambert and Larcker, 1987). Our second measure is compensation mix, which is defined the ratio of variable compensation (i.e., bonus, stock options, restricted stock, performance units, and performance shares) to total compensation.12 Compensation mix is a measure of variable remuneration, or how much an executive shares in the actual performance of the firm. Compensation mix is commonly used in consulting practice as a measure of managerial
11

The parameters of the Black-Scholes formula are calculated as follows. Annualized volatility is calculated using continuously compounded monthly returns over the prior 36 months (with a minimum of 12 months of returns). The risk-free rate is calculated using the interpolated interest rate on a Treasury Note, with the same maturity (to the closest month) as the remaining life of the option, multiplied by 0.7 to account for the prevalence of early exercise. Dividend yield is calculated as the dividends paid over the past 12 months scaled by the stock price at the beginning of the month. This is essentially the same method described by Core and Guay (2002). 12 We also disaggregate this ratio into bonus mix and equity mix, defined as the ratio of the annual bonus to total compensation and the ratio of the value of the current years stock and option grants to total compensation, respectively. For the sake of brevity, we do not report results from using these variables instead of compensation mix, but we find that they are generally similar to those with compensation mix which suggests that the incentives are not uniquely attributable to either the bonus or equity grants.

14

incentives and is similar to the measure used in prior studies such as Ittner et al. (2003) and Erickson et al. (2006).13,14 Descriptive statistics reported in Table 2 reveal that the mean (median) total level compensation ranges from $13,662,000 ($9,839,090) for the CEO to $787,727 ($558,714) for the tax director. The high level of CEO compensation is expected, since our sample generally consists of very large firms. In terms of compensation mix, the mean (median) for the CEO is 0.85 (0.90) and 0.59 (0.63) for the tax director. The tax director compensation mix is primarily attributable to equity (45% = 59% less 14% of bonus) and suggests that the tax directors duties extend beyond compliance. If the tax directors work were simply compliance, which is readily observable, there should be little difficulty (and, hence, less need to use incentive-based compensation) in ascertaining whether compliance tasks were performed adequately. 5.2. Effective tax rates We begin by examining the GAAP and Cash effective tax rates. The GAAP effective tax rate is the ratio of total tax expense to pre-tax book income (TXT/PI). It is composed of a current and deferred component. Although the current tax expense is an

The incentives literature also uses the equity portfolio delta (e.g., Core and Guay, 1999) as a measure of executive incentives. This measure captures the change in the risk-neutral value of an executives firmspecific equity portfolio holdings for a 1% change in the price of the underlying stock. One limitation of our data set is that it does not provide information about the firm-specific equity holdings of the executives, and we are therefore precluded from calculating the equity portfolio delta for either the tax director or the other executives. We suspect that the equity ownership by the tax director is modest and, thus, that analyzing the current years compensation and compensation mix will be a reasonable proxy of managerial incentives. In addition, because we have information about the equity portfolio holdings of the CEO and CEO, we repeat all of our tabulated analyses using equity portfolio delta as the measure of the incentives of these two executives and we find that our primary results are essentially unchanged. 14 An alternative approach is to adopt a two-stage research design in which we first regress either the executives total compensation or variable compensation mix on the economic determinants of incentive compensation, and then use the residual from this regression in a second-stage model of the various tax attributes of interest (BTG, GAAP and Cash ETR, etc.). The use of this alternative measure produces results that are substantively equivalent to those reported in the tables.

13

15

estimate of the current taxes owed to the tax authorities, it does not include the tax benefit from stock options until the post-FAS123R period. Deferred tax expense captures the tax implications of differences between book and tax accrual accounting. Consistent with other studies (e.g., Dyreng et al., 2008), Table 2 reports that mean (median) GAAP ETR is about 30% (31%). A GAAP effective tax rate lower than the statutory tax rates implies that firms have income included in book income that will never be recorded in taxable income (e.g., municipal bond interest and permanently reinvested earnings).15 We model the ETR as a function of both economic variables that have been suggested by prior literature (see below) and the incentives of the firms key executives. We specify the model as follows.
ETRi,t = 0 + 1Incentivesi,t + 2Return-onAssetsi,t + 3Std Dev.ROAi,t + 4Log(Market Cap)i,t + 5Leveragei,t + 6Change in Goodwilli,t + 7NewInvestmenti,t + 8Foreign Assetsi,t + 9GeographicComplexityi,t + 10IndustryComplexityi,t + 11Proportion Tax Fees
i,t

+ 12TaxFeesi,t + YearIndicators + IndustryIndicators + i,t

(1)

where ETR is either the GAAP or Cash effective tax rate. The Cash ETR measures the current cash outflows for income taxes as a percentage of pre-tax book income ((TXPD + (TXBCO + TXBCOF)) / PI).16 Unlike the GAAP ETR, the Cash ETR reflects deferral of cash tax payments. Table 2 reveals that the mean (median) Cash ETR is 29% (23%),
Note that ETR differentials, such as tax credits, also drive down the effective tax rate as they effectively represent a difference of (1-)/ between book and taxable income. 16 Unlike Dyreng et al. (2008), our measure of the cash effective tax rate considers the impact of stock options and special items. First, we view the cash benefit of stock options as a de facto cash payment for income taxes. Note that the tax benefit of options reduces the cash outflow for income taxes without any corresponding adjustment to book income. Hence, in the pre-SFAS 123R period, a firm that chooses to compensate employees with equity rather than cash will report a lower Cash ETR, not because the firm has a lower cash payment for taxes but because its book income does not include any compensation expense. The expensing of stock options under FAS 123R alleviates much this measurement error, but since our sample spans SFAS 123R adoption, it is critical for us to account for this issue. Second, Dyreng et al.s exclusion of special items from the denominator leads to downwardly biased estimates of the cash effective tax rates as 73% of the special items reported in Compustat between 2002 and 2006 are income decreasing. In Table 2, we also report descriptive statistics on Cash ETR w/o option benefit, which excludes the tax benefit from stock options in the numerator. The mean and median values of this measure of the cash effective tax rate are slightly lower than the mean and median values of Cash ETR.
15

16

which suggests that our sample has accruals that increase pre-tax book income but not taxable income. Note that both the GAAP and Cash ETRs are censored to fall between zero and one.17 Incentives are the incentives of the CEO, tax director, general counsel, and CFO, respectively, measured as either total annual compensation or compensation mix (described in Section 5.1). In addition, all independent variables are winsorized at the 1% level to ameliorate the effect of influential observations on the results. Since both dependent variables are bound in the (closed) interval [0,1], we estimate the models using a doubly censored Tobit specification to eliminate bias that would otherwise result from OLS estimation.18 Finally, since prior studies have documented substantial variation in ETRs across industries and over time, we include year and industry indicator variables to capture the average effect of this variation. 5.2.1. Economic controls We include Return on Assets (net income over beginning total assets) to control for the underlying economic activity of the firm. We include the standard deviation of the previous five years ROA (Std Dev. ROA) to capture variability in the firms operations. Next we include a size proxy, the natural logarithm of the firms market capitalization (Log(Market Cap)). Prior studies evidence on the relationship between GAAP ETR and firm size is mixed. For example, Rego (2003), Zimmerman (1983), and Omer et al. (1993) document a negative relationship between firm size and GAAP ETR, a finding that

The fraction of GAAP (Cash) ETRs that are truncated at zero and one are 6.9% (6.7%) and 1.2% (2.7%), respectively. Since these two variables are censored both below and above, we used a doubly censored Tobit in our analysis of the ETRs. 18 All tabulated specifications were also estimated using ordinary least squares and produced almost identical results, which is not surprising given the relatively low fraction of observations that are censored at either end of the [0,1] interval. It can be shown that OLS estimates will be identical to those from a doubly censored Tobit when there are no censored observations. In addition, we report tabulated results from OLS estimation, since doing so facilitates the interpretation of the marginal effects of the coefficient estimates.

17

17

is consistent with the political cost hypothesis. However, Jacob (1996), Gupta and Newberry (1997), and Mills et al. (1998) do not find a statistically significant relationship. We therefore include firm size but do not predict the sign of the relationship. Leverage, defined as the ratio of the firms long-term debt to total assets (LT/AT), is included to capture the extent of the tax shield of debt. Ceteris paribus, the greater the firms tax shield of debt, the lower the need for incremental tax planning (see MackieMason, 1990). Change in Goodwill, defined as the annual increase in the firms goodwill scaled by beginning total assets (GDWL/Beginning AT), is included to capture merger and acquisition activity of the firm during the year. If goodwill decreases, then Change in Goodwill is set to zero. New Investment, defined in Richardson (2006) as the sum of research and development expense, capital expenditures, and acquisitions less the sum of sales of property and depreciation all scaled by total assets ((XRD + CAPX + ACQ SPPE DPC) / Avg. AT), is included to control for the firms investment activity, since investment often leads to book-tax differences because the tax and accounting rules are different (e.g., the investment tax credit, accelerated depreciation methods, and bonus depreciation).19 Rego (2003) finds that multinational firms with more extensive foreign operations have lower worldwide GAAP ETRs, a finding that she interprets as being consistent with economies of scale in tax planning. We therefore include the foreign assets of the company to control for differences in international planning opportunities. Foreign Assets are estimated using the methodology described in Oler et al. (2007), which uses the consolidated turnover ratio and foreign segment sales to infer foreign assets. We include

19

Notice that New Investment captures cash-based M&A activity, whereas the Change in Goodwill captures taxable and non-taxable M&A activity.

18

an estimate of the firms foreign asset base to capture differential tax rates that typically apply to firms foreign activities. Finally, since more complex organizations require more executive talent (which commands higher pay) and more opportunity to tax plan, we include two measures of the complexity of the organization: Geographic Complexity and Industry Complexity. Following Bushman et al. (2004), these measures are revenue-based HirfindahlHirschman indices that capture the within-firm geographic and industry segment concentration of the firm. Lower values of these measures suggest a relatively more complex organization, so we predict a negative association between these measures of complexity and the ETR consistent with more opportunity for tax planning. 5.2.2. External tax planning Firms may choose to use external providers for tax planning, which might weaken the link between tax director incentive compensation and measures of tax planning. We include two proxies for the extent of a firms use of tax consulting services. Tax Fees is the total tax fees paid to a firms external auditor scaled by total assets as reported by Audit Analytics.20 The greater a firms tax fees are relative to its size, the more likely the firm has contracted with its auditor for tax planning purposes. Proportion Tax Fees is that ratio of tax fees to total fees paid to the external auditor. The greater this ratio, the more emphasis a firm places on tax planning and reporting. Measures of external tax services could either be positively or negatively associated with measures of tax planning. If a firm faces a relatively high tax burden, then external tax consultants could be engaged to help the tax director. On the other hand, the firm may have a relatively low tax burden

20

Note that to the extent the firm utilizes non-auditor tax consulting services, our Tax Fees measure will underestimate the extent to which the firm obtains outside tax assistance.

19

because external tax consultants were engaged. Hence, we make no prediction on the direction of the association with measures of tax planning. 5.3. Book-tax gap The next tax attribute we examine is the book-tax gap (BTG) because, assuming the BTG reflects tax planning, we can use the BTG and its components to investigate both conforming and non-conforming tax planning. If, after controlling for firm performance, the tax directors compensation is associated with a smaller book-tax gap, then we can infer that the tax director is compensated for managing taxable income in a manner that avoids the scrutiny of the tax authorities (Mills 1998). On the other hand, as firms face taxes in multiple jurisdictions, we conjecture that the BTG may better capture tax planning by measuring activity that reduces pre-tax income.21 If tax directors have incentives to reduce taxable income even at the expense of reducing book income (i.e., conforming tax planning), then we should observe either a negative or no association with the BTG and a negative association with both taxable income and pre-tax book income. On the other hand, non-conforming tax planning may manifest in a positive association between the incentive compensation of the tax director and the BTG. We model the book-tax gap as a function of both economic variables that have been suggested by prior literature and the incentives of the firms key executives. We specify the model as follows.

21

For example, consider two firms, A and B, that both have a foreign affiliate that generates 50% of the consolidated entitys income (with the other 50% generated by the domestic parent). Suppose the tax rate in As affiliate is 40% and the rate in Bs affiliate is 20%. Further suppose that firm As tax director shelters 20% of its affiliates income. However, Bs tax director shelters only 10% of its affiliates income. It is unclear which firm has the better tax director. On an after-tax basis, Firm B has a lower GAAP ETR (assuming PRE designation under APB 23) and Cash ETR. Yet Firm As tax director has generated a larger book tax difference. This is the reason we consider the BTG tests as capturing pre-tax tax planning.

20

BTGi,t = 0 + 1Incentives i,t + 2 Operating Cash Flow i,t + 3Std Dev. ROAi,t + 4log(Market Capi,t) + 5Leveragei,t + 6Change in Goodwilli,t + 7New Investmenti,t + 8ForeignAssetsi,t + 9Geographic Complexityi,t + 10Industry Complexityi,t + 11Proportion Tax Fees i,t + 12TaxFeesi,t + Year Indicators + Industry Indicators + i,t (2)

where BTG is the book-tax gap measured globally and scaled by total assets at the beginning of the period. The BTG is measured as the difference between pre-tax income less income attributable to minority interest (Compustat PI MII) and taxable income (defined as current federal tax expense (TXFED) grossed up by the maximum federal statutory tax rate (i.e., 35%) plus pre-tax foreign income (PIFO) less the annual change in NOLs (NOL)) scaled (by total assets).22 Incentives are the incentives of the CEO, tax director, general counsel and CFO, respectively, measured as either the level of compensation or compensation mix. Similar to equation (1), it is necessary to specify the appropriate control variables for the BTG. We rely on the same controls as those included in equation (1) except that we replace Return on Assets, which is effectively a component of our measure of the BTG, with Cash Flow From Operations scaled by average assets to control for the underlying economic activity of the firm. All of the remaining variables are as defined above. 5.3.1 Components of the book-tax gap
We adjust for the minority interest on the assumption that it represents a greater than 20% ownership stake by the minority. We continue to find similar results if we do not remove the minority interest. In addition, we estimate foreign taxable income by using reported pretax income, which implicitly assumes a relatively high level of book-tax conformity on foreign jurisdictions. Most studies estimate taxable income by grossing up current foreign tax expense by the top U.S. statutory rate, which biases taxable income (BTG) downward (upward) since foreign statutory tax rates are typically lower than U.S. rates. Finally, we ignore the tax benefit of stock options (TXBCO + TXBCOF) in our estimate of the BTG because the benefit of options was a known difference between book and taxable income during the majority of our sample period. The tax benefit of options, however, is considered in our measures of taxable income in later analyses. We presume that employee stock option exercise activity is exogenous to tax planning (barring extreme instances such as the one described in Matsunaga, Shevlin, and Shores, 1992). Ideally, in our BTG analysis, we would eliminate all differences between book income and tax income unrelated to tax planning.
22

21

Table 2 shows that the mean book-tax gap is 5.0% of total assets for our sample.23 The positive value indicates that book income is greater than taxable income. Although the book-tax gap is an important tax construct, there are both economic and econometric reasons why the book-tax gap might be problematic as a dependent variable. The booktax gap is measured as the difference between pre-tax book income and estimated taxable income, and each of its two components might exhibit a different relationship with the independent variables included in equation (1). From an economic perspective, the booktax gap itself might not be the construct of interest because some executives influence the measure indirectly (e.g., the CEO's setting of the firms investment policy) and others directly (e.g., the smoothing earnings by the CFO). Since we are specifically interested in the direct links between tax planning and managerial incentives, we disaggregate the book-tax gap into its two components and jointly estimate both equations using seemingly unrelated regression (Zellner, 1962). Joint estimation of the two resulting equations also allows us to formally compare coefficients across the two equations. Disaggregating the book-tax gap into its components also allows us to investigate whether there is evidence of conforming tax planning. When a firm undertakes conforming tax planning, book income will be reduced because the firm seeks to lower taxable income. Since large book-tax differences may draw the attention of the tax authorities, firms may be willing to undertake conforming tax planning because it increases the likelihood that the firm will prevail under audit (Mills, 1998; Cloyd, Pratt, and Stock, 1996). 5.4. Specific measures of tax aggressiveness
23

Untabulated analysis reveals that the mean BTG for our sample is positive for all years during our sample period. We also find that the mean BTG is positive for the firms in the S&P500 during each year of our sample period.

22

The final tax attribute that we examine is tax aggressiveness. As discussed above, there is no generally accepted definition of this construct, and it is thus difficult to measure. Nevertheless, we investigate two recently developed measures of tax aggressiveness: Frank et al.s (2009) DTAX and Wilsons (2009) SHELTER. Frank et al. (2009 p. 468) define aggressive tax reporting as a downward manipulation of taxable income through tax planning that may or may not be considered fraudulent tax evasion. Anecdotal evidence suggests that for public companies, the ideal tax planning transactions are those that create permanent differences.24 By creating permanent differences, the firm reduces cash taxes paid without decreasing financial statement income. To investigate whether incentives are associated with permanent differences and ETR differentials (hereafter perm diffs for expositional simplicity), we begin by computing a modified version of Frank et al.s (2009) DTAX measure of discretionary perm diffs. Frank et al. (2009) compute DTAX as the residuals from a regression of an estimate of perm diffs on measures of intangible assets, income of unconsolidated subsidiaries, minority interest, state tax burdens, changes in NOLs, and lagged perm diffs. We modify this computation by including Oler et al.s (2007) measure of foreign assets (Foreign Assets) to control for the existence of multinational operations. By including Foreign Assets in the first stage, we attempt to control for ETR differentials that result from ordinary overseas operations. Without this modification, DTAX would

24

The 1999 Treasury White Paper on corporate tax shelters noted the following: There is a current trend among public companies to treat corporate in-house tax departments as profit centers that strive to keep the corporations effective tax rate low and in line with that of competitors. Accordingly, in most recent corporate tax shelters involving public companies, the financial accounting treatment of the shelter item has been inconsistent with the claimed federal income tax treatment. See McGill and Outslay (2004) for further discussion of the prevalence of permanent difference creation in tax shelter arrangements.

23

suggest that firms with extensive foreign operations or foreign operations in low tax jurisdictions are always more aggressive tax planners.25 Table 2 reports that the mean (median) DTAX is 7.5% (0.30%) of total assets. Positive levels of DTAX indicate that our sample firms have discretionary perm diffs that reduce taxable income relative to the population of Compustat firms that are used to estimate DTAX. We investigate whether executives have incentives to engage in creating permanent differences by estimating the following model:
DTAXi,t = 0 + 1Incentivesi,t + 2Return-on-Assetsi,t + 3Std Dev. ROAi,t + 4log(Market Capi,t)+ 5Leveragei,t + 6Change in Goodwilli + 7NewInvestmenti,t + 8Foreign Assetsi,t + 9 GeographicComplexityi,t + 10IndustryComplexityi,t + 11Proportion Tax Fees i,t + 12TaxFeesi,t + i,t (3)

where the variables are as defined above. Finally, we investigate whether tax director incentives are associated with the predicted presence of tax shelters. We estimate Wilsons (2009) SHELTER measure which was developed using a sample of approximately 60 firms identified in court documents as having participated in illegal tax shelters. We use the following model to determine whether tax director incentives are associated with a high probability of sheltering behavior.
SHELTERi,t = 0 + 1Incentivesi,t + 2Return-on-Assetsi,t + 3Std Dev. ROAi,t + 4log(Market Capi,t)+ 5Leveragei,t + 6Change in Goodwilli + 7NewInvestmenti,t + 8Foreign Assetsi,t + 9 GeographicComplexityi,t + 10IndustryComplexityi,t + 11Proportion Tax Fees i,t + 12TaxFeesi,t + i,t
25

(4)

Note that Frank et al.s measure, DTAX, implicitly assumes that any variation in perm diffs that does not result from the known sources of perm diffs included in the first stage (i.e., change in NOL, minority interest income, state income taxes, intangibles, and unconsolidated subsidiaries) are the result of tax planning. We conjecture that foreign operations are another source of perm diffs that are not necessarily discretionary tax planning and are therefore not necessarily indicative of aggressive behavior. Accordingly, we modify Frank et al.s DTAX measure by also including foreign assets in the first-stage regression. Nevertheless, all of our reported results are robust to excluding foreign assets and measuring DTAX exactly as described by Frank et al.

24

6. Results 6.1. GAAP ETR The results of estimating equation (1) where the GAAP ETR is the dependent variable are presented in Table 3. We first consider the role of the level of compensation, which is presented in columns (1) and (2). In the first model, where all of the positions are simultaneously considered, we find a strong negative relationship between tax director compensation and the GAAP ETR, but no significant relationship with the other three executives. This strong negative relationship continues to hold when we estimate the model considering only the tax executive.26 When we measure incentives as compensation mix in columns (3) and (4), we again find a strong negative relationship between tax director incentives and the GAAP ETR when all executives are considered together. We continue to find a negative relationship when only the tax directors incentives are included in the analysis (i.e. column 4). Overall, the evidence presented in Table 3 provides strong support for the notion that tax directors, but not the CEO, general counsel, and CFO have incentives to manage the GAAP ETR. We also find a significant relationship between the GAAP ETR and certain control variables included in the specification. In particular, Return on Assets is significantly positive which indicates that more profitable firms have higher GAAP ETRs. Our proxy for the operating risk, Std. Dev. ROA, is negative and significant, consistent with convexity in the tax function (i.e., losses can reduce future earnings). Foreign Assets is positive and is generally significant, suggesting that firms with greater multinational
26

When we include the incentives of each of the other executives (CEO, CFO, GC) alone in the model, they are never significant.

25

activities have higher GAAP ETRs. Although we expected foreign activity to give rise to lower tax liabilities, our results are consistent with foreign investment being correlated with future profitability. Geographical Complexity is positive and significant, suggesting that firms with more dispersed geographic operations have lower effective tax rates. Consistent with Mills, Erickson, and Maydew (1998), we also we find that a higher ProportionTaxFees is associated relatively lower GAAP ETRs, which is consistent with the presence of more sophisticated tax planning. Interestingly, these results suggest that tax directors are compensated on the basis of after-tax financial reporting measures rather than pre-tax measures. In particular, since we find no association between taxable income and tax director incentives, the negative relationship with the GAAP ETR is consistent with tax directors undertaking of tax planning involving permanent differences and ETR differentials. Thus, our results corroborate Robinson et al.s (2010) finding that firms provide incentives to the tax department to mitigate the impact of taxes on bottom-line net income. 6.2. Cash ETR The results of estimating equation (1) with the Cash ETR are presented in Table 3, columns (5) through (8), and parallel our analysis of the GAAP ETR. When we consider the level of compensation as the measure of incentives in columns (5) and (6), we fail to find a significant relationship between the incentives of any of the executives and the Cash ETR. When compensation mix is used as the measure of incentives in columns (7) and (8), we again find no relationship between executive incentives and the Cash ETR in any of the specifications.27

27

When we use the leading three-year average Cash ETR, Cash ETR (3 year), as the dependent variables, we find no association with our various measures of incentives. Note that this analysis reduces our sample

26

We document several significant relationships between the cash effective tax rate and the control variables. Return on Assets is significantly positive which indicates that more profitable firms have higher cash effective tax rates. Std ROA is negative and significant, consistent with the finding that a greater variation in a firms earnings results in a lower Cash ETR because of the convexity in the statutory rate structure. Change in Goodwill is positively related to the Cash ETR, a finding that suggests that M&A activity leads to lower pre-tax book income with no corresponding reduction to cash taxes paid (e.g., in-process R&D). Finally, we find that, consistent with the presence of incremental tax incentives, New Investment exhibits a negative and significant relationship with Cash ETR. Overall, the lack of a relationship between tax director incentives and Cash ETR is in sharp contrast to the strong negative relationship for the GAAP ETR presented in Table 3.28 Collectively, these results suggest that tax directors have incentives to reduce the GAAP ETR. One agency-theoretic explanation for this result is that the contracting weight on a performance measure is a function of its signal-to-noise ratio (Holmstrom, 1979; Lambert and Larcker, 1987; Banker and Datar, 1989; Lambert, 2001). Accordingly, an increase in either the sensitivity of the performance measure to the agents actions (i.e., the strength of the signal about the agents actions), or a decrease in its conditional variance (i.e., the noise with which it reflects the agents actions) will, ceteris paribus, increase the relative weight placed on the performance measure in the
to 772 firm years, since we are unable to estimate Cash ETR (3 year) for 2006. The negative association between GAAP ETR and tax director incentives continues to hold in this reduced sample. 28 To alleviate concerns that our GAAP ETR results are driven solely by tax accounting accruals, we have collected data on firms valuation allowance and cushion activity for a random subsample of our firms. When we include the change in the valuation allowance and/or cushion, we continue to find a significant negative (no) association between the tax directors compensation (both levels and mix) and the GAAP (Cash) ETR.

27

optimal compensation contract. From a signal perspective, the GAAP ETR is likely to be more sensitive to the tax directors actions than the Cash ETR is, because, unlike the Cash ETR, it contains an accrual component that estimates the future effects of actions in the current period.29 From a noise perspective, the Cash ETR of our sample firms exhibits more variation than does the GAAP ETR (Table 2 reports that the standard deviation of Cash (GAAP) ETR is 26.2% (14.8%)). This is likely attributable to the discrete nature of items such as option exercises and tax audit cash outflows, which is undesirable from a contracting perspective.30 Thus, our results are consistent with the GAAP ETR being a more informative measure of the tax directors actions, and, accordingly, it is allocated more weight in the incentive compensation contract. Our results could also be construed as consistent with firms potentially forgoing cash savings to gain financial reporting benefits (e.g., Engel et al., 1999; Erickson et al., 2004). Finally, our results also corroborate Robinson et al.s (2010) finding of little to no association between the profit center motive of the firms tax department and Cash ETR. 6.3. Book-tax gap and its components The results of the book-tax gap analysis are presented in Table 4. We find no relationship between the book-tax gap and either the level of compensation or the compensation mix of any of the executives. In addition, we find no relationship between the incentives of the tax director and either the PBI or TI component of the BTG.

In a multiperiod contracting setting, a number of authors such as Dutta and Reichelstein (2003) show that accruals are valuable for contracting because they provide a current (albeit noisy) estimate of the future cash flows generated by the agents actions in the current period. Accordingly, accruals are useful for overcoming agency problems that result from differences in discount rates between the principal and agent. 30 This idea is a manifestation of the controllability principal from agency theory. Lambert (2001, p. 23) defines a performance measure as controllable if the agents actions influence the probability distribution of that variable. To the extent that a performance measure that is not controllable by the agent is used in the compensation contract, it imposes additional risk on the agent for which he or she must be compensated.

29

28

In terms of the control variables, the positive association between BTG and Operating Cash Flow suggests that more profitable firms have a greater wedge between book and taxable income. When the dependent variable is pre-tax book (taxable) income, in all models, the estimated coefficient on cash flows of approximately 0.84 (0.59) indicates that pre-tax book (taxable) income is, on average, roughly 16% (41%) lower than cash flows from operations. The positive relationship between log(Market Capitalization) and measures of firm performance (PBI and TI) is consistent with profitable firms being more valuable. In addition, the negative association between Industry Complexity and the BTG suggests that firms with more variation in industry composition have greater book-tax gaps. However, consistent with earnings management, the association appears to stem from the fact that these firms have relatively higher pretax book income. Overall, the results of the taxable income specification support the notion that there is little cross-sectional variation in the executives incentives to reduce taxable income, at least as we can estimate this measure. As our sample is composed of large public firms, it is likely that these firms are averse to tax planning opportunities that reduce both book and tax income. 6.4. Measures of tax aggressiveness Our final set of results relates to tax aggressiveness and is presented in Table 5. Using the level of compensation as our measure of incentives, we find no relationship between executive incentives and Frank et al.s (2009) measure of discretionary permanent differences/ETR differentials. We also fail to find a correlation between any

29

of the measures of executive incentives and Wilsons (2009) measure of sheltering behavior (SHELTER).31 We document several significant relationships between the measures of tax aggressiveness and the control variables. Interestingly, we find that DTAX is positively associated with Change in Goodwill, a finding that is consistent with increases in permanent differences around M&A activity. The negative association between DTAX and New Investment suggests that firms with relatively more new depreciation deductions have smaller discretionary permanent differences. Finally, Proportion Tax Fees is significantly positively related to DTAX, a finding that suggests that hiring the firms auditor for tax services complements aggressive tax planning. Return on Assets exhibits a positive and significant relationship with SHELTER which indicates that more profitable firms have greater predicted probability of participating in tax shelters. We also find that SHELTER is positively related to the size of the firm which is consistent with the large fixed costs of entering into shelter transactions. The negative association between geographic complexity and SHELTER suggests that firms with more foreign affiliates have a higher probability of being involved in a tax shelter. However, overall, we find little evidence that the either the CEO, tax director, general counsel, or CFO are provided incentives to undertake aggressive tax planning as measured by DTAX and SHELTER.32

31

We obtain similar results when we use Compensation Mix as our measure of incentives. We also obtain similar results when we include the five variables from the first-stage DTAX specification that are not included in equation (3), namely, the ratio of (i) intangible assets to total assets, (ii) earnings of subsidiaries, (iii) minority interest income, (iv) state taxes, and (v) change in NOL. 32 Note that there are limitations of the DTAX and SHELTER measures. First, DTAX measures only incremental permanent differences and ETR differentials (perm diffs), so, ceteris paribus, a tax director that has replaced one perm diff with another (perhaps because of a change in the tax law) will have zero DTAX. Second, much of the variation in DTAX could stem from changes in foreign operations. Although we include foreign assets in our model of DTAX, we still are not able to control for the tax rates of the

30

7. Sensitivity Analyses 7.1. Governance characteristics Prior literature such as Core, Holthausen, and Larcker (1999) has shown that various governance characteristics are related to firms incentive compensation practices. In addition, recent research such as Desai and Dharmapala (2006) suggests a link between firms governance characteristics and the extent to which they engage in aggressive tax planning.33 We therefore repeat the analyses above, including a broad array of governance variables that have been used in prior research. In particular, prior studies linking corporate governance to compensation practices have examined three broad categories of governance constructs: (i) board of director characteristics; (ii) Board structure; and (iii) state antitakeover laws. We obtain board of director data from the Equilar analysis of proxy statements and use six variables to capture board characteristics: (i) the number of directors, (ii) the fraction of inside directors, (iii) the fraction of directors who are over 69 years old, (iv) the fraction of the board that is busy (which is measured as serving on at least two additional boards), (v) an indicator that equals one if the lead director is classified as affiliated, and zero otherwise, and (vi) the fraction of directors classified as outsiders who were appointed after the current CEOs term began.

jurisdictions where the assets are located. Third, variation in DTAX could be attributable to changes in the tax contingency that is related to past tax planning instead of current tax planning. Regarding SHELTER, the majority of the variation in the measure results from firm size. Since our sample consists almost entirely of large firms, there is almost no variation in this measure within our sample (Table 2 reports that the interquartile range is from 0.946 to 0.994). Since our tests are cross-sectional, the lack of variation in these measures across our sample suggests that this measure produces a low-powered test. 33 In particular, Desai and Dharmapala (2006) argue that good governance increases the link between equity compensation and aggressive tax planning.

31

Next, we use data from FactSet SharkRepellent to measure board structure similar to Gompers et al. (2003), Daines and Klausner (2001), Bebchuk and Cohen (2005), and Faleye (2007). First, we include a set of indicator variables to capture whether the companys board members are elected annually or are elected to staggered, multiyear terms.34 We also include an indicator for whether the firm has multiple classes of shares with unequal voting rights. Finally, similar to Bebchuk et al. (2002), Cheng et al. (2004), and Wahal et al. (1995), we include variables for state antitakeover laws. Their results suggest that the introduction of stronger antitakeover legislation leads to greater managerial entrenchment. Our results (untabulated) are robust to the inclusion of governance metrics. Specifically, we continue to find a significant negative relationship between tax director incentives and the GAAP ETR and no relationship between tax director incentives and BTG, Cash ETR, DTAX, and SHELTER. 7.2. Reversing the regression Prior studies that investigate cross-sectional variation in firms tax planning model executive compensation as a function of the firms tax attributes (e.g., Rego and Wilson, 2010; Desai and Dharmapala, 2006). These studies find that higher levels of compensation are associated with a higher level of tax aggressiveness. In addition, this specification allows researchers to simultaneously evaluate multiple measures of tax planning. We therefore estimate regressions of both total compensation and compensation mix of the CEO, tax director, general counsel, and CFO on a variety of economic controls as well as GAAP ETR and Cash ETR. The results of the total

34

Activist shareholders argue that staggered terms impede shareholders monitoring of the Board by making it more difficult for them to alter the boards composition over a short time period.

32

compensation regressions are reported in Table 6, and they reveal that GAAP ETR continues to exhibit a strong negative relationship with the level of tax director compensation.35 Interestingly, and consistent with our signal-to-noise interpretation of the various performance measures, when we replace Cash ETR with Three-year lagged Cash ETR, we find a negative (but insignificant) relationship with the level of CEO and tax director compensation. This finding suggests that the signal becomes less noisy as the window over which Cash ETR is measured lengthens. 7.3. Matched-pair research design There are two important limitations of the traditional linear regression approach used in the preceding tests of the relationship between executive incentives and firm tax attributes. First, this approach relies on a linear functional form linking the outcome variable of interest (i.e., tax attributes) with both the independent variable of interest (i.e., executive compensation) and the other control variables, or covariates (e.g., firm size, operating performance, and tax professional fees). To the extent that this linearity assumption is violated, the model is misspecified and can produce biased coefficient estimates. Second, to the extent that there is endogenous matching of executives with firms (and compensation contracts) on the basis of some unobserved characteristics (e.g., executive risk aversion or talent), this gives rise to traditional correlated omitted variable problems and this will bias the parameter estimates obtained from a linear model. To mitigate these econometric concerns, we also use a propensity score matchedpair research design (Rosenbaum and Rubin, 1983; Rosenbaum, 2002; Armstrong et al., 2009), in which we form matched pairs of firm years that are similar along all of their

35

The coefficient on Cash ETR and GAAP ETR are statistically different for both the tax director (p = 0.027, two-tail) and the general counsel (p = 0.071, two-tail).

33

(observable) economic characteristics, but most dissimilar in terms of their executives incentives.36 After matching on these variables, any difference in outcome of interest (e.g., BTG, GAAP ETR, Cash ETR, SHELTER, and DTAX) can be more appropriately attributed to differences in the level of the executives incentives rather than to differences in the other variables, regardless of the underlying structural form. We also assess the sensitivity of our results to unobserved correlated omitted variables by determining the magnitude of the correlated omitted variable bias that is necessary to cause any statistically significant differences between matched pairs to become insignificant.37 Although this approach does not resolve the endogeneity problem per se, the computation enables us to provide some insight into whether our results are robust to endogenous matching. A formal test of the difference in GAAP ETR between matched pairs is presented in Panels A and B of Table 7, respectively. We observe that both higher levels of tax director compensation and tax director compensation mix exhibit a strong negative relationship with the GAAP ETR (Wilcoxon p-values of 0.0001 and 0.0275, respectively, and t-statistic p-values of 0.0005 and 0.0266, respectively). Further, both the mean and median differences in the GAAP ETR of -0.029 and -0.018 between the firms with a
Our propensity score matched-pair research design requires a model for the conditional probability of receiving a given level or mix of compensation given observable features of the contracting environment. Consistent with many prior studies, we assume that the choice of the level and mix of compensation are a function of the economic and tax planning characteristics discussed in the Sections 5.2.1 and 5.2.2. We also required matched pairs to be from the same year and industry. The matched sample was constructed using a nonbipartite matching algorithm suggested by Derigs (1988), which is an optimal algorithm in the sense that it considers the potential distances between other pairs when forming a particular matched pair. The matching was done without replacement. 37 An alternative way of characterizing the correlated omitted variable problem in our setting is that there is the potential for endogenous matching (or sorting) of executives and companies. If, for example, executives who are relatively more risk-tolerant decide to work for companies that offer more risky, incentive-based compensation packages (because these are the executives who are better able to bear the associated risk), any observed relationship between executive incentives and certain tax characteristics (e.g., tax aggressiveness) could be a result of differences in executive risk-tolerance rather than differences in the observed incentives.
36

34

relatively high and relatively low level of tax director compensation, respectively, and 0.019 and -0.011 for tax director compensation mix, respectively, are economically significant. To provide a benchmark of the magnitude of our GAAP ETR results, we also provide the propensity score matched-pair analysis for Cash ETR.38 The Cash ETR results in Panel B show that none of the incentives variables exhibits a significant relationship with the Cash ETR. These results are consistent with the linear regressions results presented above and suggest that the relationship is robust to possible misspecification of the functional form linking the incentive and control variables with the tax outcomes. Similar to all observational studies, our results are susceptible to hidden bias if there are correlated omitted variables that are not balanced across the two categories. Rosenbaum (2002, 2007) develops a bounding approach for assessing the sensitivity of the matched-pair results to hidden bias.39 In untabulated sensitivity analysis, we find that the results remain statistically significant to the existence of a correlated omitted variable that shifts the assignment of executives to the treatment group from a 50%/50% probability of being assigned to the high- and low-incentives categories to a

Note that a similar propensity score matched-pair analysis of BTG, SHELTER, and DTAX provides no evidence that any of these tax attributes exhibits a significant relationship with executive incentive compensation. 39 In our context, hidden bias exists if two executive (denoted i and j) have the same observed (economic and tax planning) covariates, but different probabilities (denoted ) of having a particular level of incentives. The odds that each executive was assigned to the high- and low-incentives category are denoted i/(1-i) and j/(1-j), respectively. If the odds ratio (following Rosenbaum (2002), denoted by ), does not equal one, then the two executives have an unequal probability of being assigned to a category, and hidden bias exists. Rosenbaum (2002) shows that relaxing the assumption that = 1 allows for a computation of the amount of hidden bias (or, the strength of a correlated omitted variable) that is needed to alter any significant inferences. Smaller values of indicate statistically significant results that are more sensitive to hidden bias.

38

35

66.5%/33.5% probability assignment.40 Thus, our results are robust to substantive correlated omitted variables, and this provides some justification for drawing causal inferences regarding the relationship between tax director incentives and GAAP ETRs.

8. Conclusion Although tax directors are responsible for one of the firms largest outflows of cash and one of the largest expenses on the income statement, almost nothing is known about how these executives are compensated. To the best of our knowledge, this is the first paper to directly study the link between the incentives of tax directors and measures of the extent of their firms tax planning. Overall, our analyses of the book-tax gap, its components, the cash effective tax rate, and alternative measures of tax aggressiveness provide little evidence that our sample of large, publicly traded firms explicitly incentivize their tax function to undertake measures to lower the firms contemporaneous cash tax burden. Instead, consistent with tax directors receiving incentives to reduce the level of tax expense reported in the financial statements, we find that tax executives compensation is negatively associated with the GAAP ETR. An agency theoretic interpretation of this result is that the GAAP ETR is relatively controllable by the tax director and can be measured with sufficient precision so as to make it valuable for

In other words, the statistical tests conducted in Table 7 are based on the assumption that, conditional on the variables in the propensity score model, executives have a 50% probability of being assigned to both the treatment and control groups (i.e., the high- and low-incentives groups). If instead, because of some correlated omitted variable (e.g., executive risk tolerance), executives with a higher observed level of incentive compensation were actually assigned to the treatment group 66.5% of the time and assigned to the control group only 33.5% of the time (or = 2), a statistically significant difference in the GAAP ETR between the two groups would still obtain. Since no strict benchmark exists to determine whether a given is large or small, the designation is inherently subjective and depends on beliefs as to the degree of endogenous selection on unobservable factors. Given the extensive controls included in the propensity score matching, we believe that it is unlikely that a correlated omitted variable exists that will produce a > 2.

40

36

contracting. In contrast, the firms other tax attributes are either not sufficiently controllable by the tax director and/or are too noisy to be valuable for contracting. As with all empirical observational studies, there are several caveats regarding our results. First, our sample consists primarily of large profitable U.S. domiciled, multinational firms. Accordingly, our results may not generalize to smaller firms. Second, the firms in our sample are known to have (highly paid) tax directors. Clearly, the presence of a tax director is prima facie evidence of these firms undertaking some activity incremental to the duties of the top management team. Therefore, our results do not imply that the CEO or CFO is never responsible for the firms tax planning function, but only that this link is attenuated in the presence of a tax director. Third, like almost all prior incentives research, we are unable to directly observe the tax directors compensation contracts. Therefore, we do know with certainty the precise mechanism(s) through which his or her incentives are generated which, in turn, makes causality difficult to establish. Fourth, we only investigate the monetary incentives provided to tax directors. Tax directors (and other executives) employment might be contingent on some other measure of tax planning (e.g., Cash ETR), but the nature of our data inhibits our ability to study termination as an incentive mechanism. Finally, we cannot directly observe tax planning, so our measures of tax planning undoubtedly include measurement error.

37

References Armstrong, C.S., A.D. Jagolinzer, and D.F. Larcker, 2009. Chief executive officer equity incentives and accounting irregularities. Journal of Accounting Research 48, 225-271. Armstrong, C.S., W.R. Guay, and J.P. Weber, 2010. The role of information and financial reporting in corporate governance and contracting. Journal of Accounting and Economics 50, 179-234. Ayers, B., J. Jiang, and P. Yeung. 2006. Discretionary accruals and earnings management: An analysis of pseudo earnings management. The Accounting Review 81(3), 617-652. Badertscher, B., J. Phillips, M. Pincus, and S. Rego, 2009. Earnings management strategies and the trade-off between tax benefits and detection risk: To conform or not to conform? The Accounting Review 84(1), 63-97. Banker, R.D., and S.M. Datar, 1989. Sensitivity, precision, and linear aggregation of signals for performance evaluation. Journal of Accounting Research 27, 21-39. Bebchuk, L.A., and A. Cohen, 2005. The costs of entrenched boards. Journal of Financial Economics 78, 409-433. Bebchuk, L.A., J.C. Coates IV, and G. Subramanian, 2002. The powerful antitakeover force of staggered boards: theory, evidence, and policy. Stanford Law Review 54, 887-951. Boynton, C., P. DeFilippes, and E. Legel, 2005. Prelude to Schedule M-3: Schedule M-1 corporate book-tax difference data 1990-2003. Proceedings of the Annual Conference on Taxation. National Tax Association: 131-137. Bushman, R., Q. Chen, E. Engel, and A. Smith, 2004. Financial accounting information, organizational complexity, and corporate governance systems. Journal of Accounting and Economics 37: 167-201. Chen, K., and C. Chu, 2005. Internal control versus external manipulation: a model of corporate income tax evasion. RAND Journal of Economics 36(1): 151-164. Cheng, S., V. Nagar, and M. Rajan, 2004. Identifying control motives in managerial ownership: Evidence from antitakeover legislation. Review of Financial Studies 18, 637-672. Cloyd, B., J. Pratt, and T. Stock, 1996. The use of financial accounting choice to support aggressive tax positions: Public and private firms. Journal of Accounting Research 34(1): 23-43. Core, J., R. Holthausen, and D. Larcker, 1999. Corporate governance, chief executive officer compensation, and firm performance. Journal of Financial Economics 51, 371-406. Core, J.E., and W.R. Guay, 1999. The use of equity incentives to manage optimal equity incentive levels. Journal of Accounting and Economics 28, 151-184. Core, J.E., and W.R. Guay, 2002. Estimating the value of employee stock option portfolios and their sensitivities to price and volatility. Journal of Accounting Research 40, 613-630.

38

Crocker, K., and J. Slemrod, 2005. Corporate tax evasion with agency costs. Journal of Public Economics 89, 1593-1610. Cronqvist, H., A. Low, and M. Nilsson, 2007. Does corporate culture matter for firm policies? Working Paper, The Ohio State University. Daines, R., and M. Klausner, 2001. Do IPO charters maximize firm value? Antitakeover protection in IPOs. Journal of Law, Economics, and Organization 17, 83-120. Derigs, U., 1988. Solving non-bipartite matching problems via shortest path techniques. Annals of Operations Research 13, 225-261. Desai, M. 2003. The divergence between book and tax income. Poterba, J.M. (ed.), Tax Policy and the Economy, Volume 17, Cambridge, Mass.: NBER and MIT Press, 169206. Desai, M., and D. Dharmapala. 2006. Corporate tax avoidance and high-powered incentives. Journal of Financial Economics 79, 145-179. Dutta, S., and S. Reichelstein, 2003. Leading indicator variables, performance measurement, and long-term versus short-term contracts. Journal of Accounting Research 41, 837-866. Dyreng, S., M. Hanlon, and E. Maydew, 2008. Long-run corporate tax avoidance. The Accounting Review 83(1), 61-82. Dyreng, S., M. Hanlon, and E. Maydew, 2010. The effects of managers on corporate tax avoidance. The Accounting Review 85(4), 1163-1189. Engel, E., M. Erickson, and E. Maydew, 1999. Debt-to-equity hybrid securities. Journal of Accounting Research 37(2), 249-274. Erickson, M., M. Hanlon, and E. Maydew, 2004. How much will firms pay for earnings that do not exist? Evidence of taxes paid on allegedly fraudulent earnings. The Accounting Review 79(2), 387-408. Erickson, M., M. Hanlon, and E. Maydew, 2006. Is there a link between executive equity incentives and accounting fraud? Journal of Accounting Research 44(1), 113-143. Faleye, O., 2007. Classified boards, firm value, and managerial entrenchment. Journal of Financial Economics 83, 501-529. Feng, M., W. Ge, S. Luo, and T. Shevlin, 2009. Why do CFOs become involved in material accounting manipulations? Working Paper, University of Washington. Frank, M.M., L. Lynch, and S. Rego, 2009. Tax reporting aggressiveness and its relation to aggressive financial reporting. The Accounting Review (84)2, 467-496. Gompers, P., J. Ishii, and A. Metrick, 2003. Corporate governance and equity prices. Quarterly Journal of Economics 118, 107-155. Graham, J., J. Raedy and D. Shackelford, 2010. Research in accounting for income taxes. Working Paper, University of North Carolina. Graham, J., and A. Tucker, 2006. Tax shelters and corporate debt policy. Journal of Financial Economics 81, 563-594.

39

Gupta, S., and K. Newberry, 1997. Determinants of the variability in corporate effective tax rates: evidence from longitudinal data. Journal of Accounting and Public Policy 16, 1-34. Hanlon, M., 2005. The persistence and pricing of earnings, accruals, and cash flows when firms have large book-tax differences. The Accounting Review 80(1), 137-166. Hanlon, M., and S. Heitzman, 2010. A review of tax research. Journal of Accounting and Economics 50, 127-178. Holmstrom, B., 1979. Moral hazard and observability. Bell Journal of Economics 10(1), 74-91. Ittner, C., R. Lambert, and D. Larcker, 2003. The structure and performance consequences of equity grants to employees of new economy firms. Journal of Accounting and Economics 34(1-3), 89-127. Internal Revenue Service, 2009. Internal Revenue Service Data Book, 2008. Publication 55B, Washington, D.C. Internal Revenue Service. 2010. Internal Revenue Service Data Book, 2009. Publication 55B, Washington, D.C. Jacob, J. 1996. Taxes and transfer pricing: Income shifting and the volume of intrafirm transfers. Journal of Accounting Research 34(2), 301-311. Jiang, J., K. Petroni, and I. Wang, 2010. CFOs and CEOs: Who have the most influence on earnings management? Journal of Financial Economics 96, 513-526. Lambert, R.A., 2001. Contracting theory and accounting. Journal of Accounting and Economics 32, 3-87. Lambert, R.A., and D.F. Larcker, 1987. An analysis of the use of accounting and market measures of performance in executive compensation contracts. Journal of Accounting Research 25, 85-125. Mackie-Mason, J. 1990. Do taxes affect corporate financing decisions? Journal of Finance 45(5), 1471-1493. McGill, G., and E. Outslay, 2004. Lost in translation: Detecting tax shelter activity in financial statements. Tax Notes 96(8), 1125-1146. Mills, L. 1998. Book-tax differences and Internal Revenue Service adjustments. Journal of Accounting Research 36(2), 343-356. Mills, L., M. Erickson, and E. Maydew, 1998. Investments in tax planning. Journal of the American Taxation Association 20, 1-20. Oler, M., T. Shevlin, and R. Wilson, 2007. Examining investor expectations concerning tax saving on repatriations of foreign earnings under the American Jobs Creation Act of 2004. Journal of the American Taxation Association 29(2), 25-55. Omer, T.Z., Molloy, K.H., Ziebart, D.A., 1993. An investigation of the firm size/effective tax rate relationship in the 1980s. Journal of Accounting, Auditing and Finance 8(2), 167-182. Phillips, J., 2003. Corporate tax-planning effectiveness: The role of compensation-based incentives. The Accounting Review 78(3), 847-874.

40

Phillips, J., M. Pincus, and S. Rego, 2003. Earnings management: New evidence based on deferred tax expense. The Accounting Review 78(2), 491-521. Prendergast, C., 2002. The tenuous trade-off between risk and incentives. The Journal of Political Economy 110(5), 1071-1073. Rego, S., and R. Wilson, 2010. Executive compensation, equity risk incentives, and corporate tax aggressiveness. Working Paper, University of Iowa. Rego, S.A., 2003. Tax avoidance activities of U.S. multinational corporations. Contemporary Accounting Research 20, 805-833. Richardson, S., 2006. Over-investment of free cash flow. Review of Accounting Studies 11(2-3), 159-189. Robinson, J.R., S.A. Sikes, and C.D. Weaver, 2010. The impact of evaluating the tax function as a profit center on effective tax rates. The Accounting Review 85(3), 10351064. Rosenbaum, P., and D. Rubin, 1983. The central role of the propensity score in observational studies for causal effects. Biometrika 70, 41-55. Rosenbaum, P.R., 2007. Sensitivity analysis for m-estimates, tests, and confidence intervals in matched observational studies. Biometrics 63, 456-464. Rosenbaum, P.R., 2002. Observational Studies, 2nd ed. Springer Series in Statistics, Berlin. Seidman, J., 2010. Interpreting the book-tax income gap as earnings management or tax sheltering, Working Paper, University of Texas. Slemrod, J., 2004. The economics of corporate tax selfishness. National Tax Journal 57(4), 877-899. Tax Executives Institute, Inc. 2004-2005 Corporate Tax Department Survey. Tax Executives Institute, Inc., Washington D.C. U.S. Department of the Treasury. 1999. The problem of corporate tax shelters: discussion, analysis and legislative proposals. U.S. Government Press, Washington, D.C. Wahal, S., K.W. Wiles, and M. Zenner, 1995. Who opts out of state antitakeover provisions? The case of Pennsylvanias SB 1310. Financial Management 24, 22-39. Wilson, R., 2009. An examination of corporate tax shelter participants. The Accounting Review 84(3), 969-999. Zellner, A., 1962. Estimators for seemingly unrelated regression equations: some finite sample results. Journal of the American Statistical Association 58, 977-992. Zimmerman, J., 1983. Taxes and firm size. Journal of Accounting and Economics 5, 119149.

41

Appendix A: Variable Definitions

Variable Total CEO Compensation, Total Tax Director Compensation, Total GC Compensation, and Total CFO Compensation Definition Total annual compensation of the CEO, Tax Director, General Counsel, and CFO, respectively, where total annual compensation is defined as the sum of salary, bonus, the market value of target long-term incentive payouts, perquisites, the market value of restricted stock, and the risk-neutral (Black-Scholes) value of stock options received during the year (with a 70% discount applied to the time to maturity). Ratio of variable compensation (i.e., bonus, restricted stock, and stock options) to total compensation. Ratio of the annual bonus to total compensation. Difference between pre-tax book income and taxable income. Book-Tax Gap scaled by average Total Assets. The GAAP effective tax rate for the year defined as total income-tax expense scaled by pre-tax income. Current effective tax rate defined as total income tax expense less deferred income tax expense scaled by pre-tax income. Deferred effective tax rate defined as deferred income tax expense scaled by pre-tax income. The cash effective tax rate for the year defined as the sum of total income taxes paid and the tax benefit of stock options scaled by pre-tax income. Source and Data Codes Proprietary compensation consultant data

CEO Compensation Mix, Tax Director Compensation Mix, GC Compensation Mix, and CFO Compensation Mix CEO Bonus Mix, Tax Director Bonus Mix, GC Compensation Mix, and CFO Bonus Mix Book-Tax Gap (BTG)

Proprietary compensation consultant data

Proprietary compensation consultant data Compustat PI less TXFED/0.35 less PIFO plus TLCF Compustat BTG divided by average AT Compustat TXT / PI

Scaled Book-Tax Gap GAAP ETR

Current ETR

Compustat ((TXT TXDI) / PI)

Deferred ETR

Compustat (TXDI / PI)

Cash ETR

Compustat ((TXPD + (TXBCO + TXBCOF)) / PI)

42

Cash ETR (w/o Option Benefit) Cash ETR (3 Year) Pre-Tax Book Income Taxable Income

Cash effective tax rate excluding the tax benefit of stock options. Three-year Cash ETRs. Pre-tax income less minority interest. Current federal tax expense grossed up by the maximum federal statutory tax rate plus foreign pre-tax income less the change in the firms net operating loss carryforward. Taxable income less the tax benefit of stock options grossed up by 35%. Operating cash flows. The tax benefit of stock options not included in income.

Compustat TXPD/ PI 3 years leading Compustat PI MII Compustat TXFED/0.35 + PIFO - TLCF

Taxable Income (Less Option Benefit) Cash Flow From Operations Tax Benefit of Stock Options

TXFED/0.35 (TXBCO + TXBCOF)/0.35 + PIFO - TLCF Compustat OANCF Hand collection from 2002 to 2004, Compustat TXBCO + TXBCOF 2005 and 2006 Compustat

DTAX

Modified discretionary permanent differences and ETR differentials as defined in Frank et al. (2009) and modified with the inclusion of estimated foreign assets. Probability that the firm engages in a tax shelter as defined in Wilson (2009).

SHELTER

Compustat Tax Shelter = - 4.86 + 5.20*BookTax Differences + 4.08*Discretionary Accruals - 1.41*Leverage + 0.76*Size + 3.51*ROA + 1.72*ForeignIncome + 2.42*R&D Compustat (CSHO * PRCC_F) / (AT LT) Compustat CSHO * PRCC_F Compustat SALE Compustat NIt / ATt-1

Market-to-Book Ratio Market Capitalization

Market capitalization over the book value of Total Shareholders Equity. The number of common shares outstanding multiplied by the price per share at the fiscal-year end. Total net sales. Net income (or loss) scaled by beginning total assets.

Revenue Return on Assets

43

Prior Years Return Std. Dev. ROA Prior Years Volatility Leverage Change in Goodwill New Investment

Cumulative return on the firms stock price over the previous year. Standard deviation of Return on Assets over the previous five fiscal years. The volatility of the firms monthly stock returns over the previous year. Long-term debt over total assets. The annual change in goodwill if greater than 0; otherwise 0. The annual investment as described in Richardson (2007) and defined as Research and Development expense plus Capital Expenditures plus Acquisitions minus Sale of Property minus Depreciation all scaled by average Total Assets. Estimated foreign assets using the methodology described in Oler et al. (2007) over Total Assets. Estimated as the revenue-based Hirfindahl-Hirschman indices calculated as the sum of the squares of each geographic segments sales as a percentage of the total firm sales as described in Bushman et al. (2004). Estimated as the revenue-based Hirfindahl-Hirschman indices calculated as the sum of the squares of each industry segments sales as a percentage of the total firm sales as described in Bushman et al. (2004). Total annual audit fees (in $millions) paid by the firm. Total annual tax fees (in $millions) paid by the firm over total assets. Total annual tax fees (in $millions) paid by the firm over audit fees.

CRSP RET Std dev(Compustat NIt / ATt-1) CRSP RET Compustat LT / AT Compustat (GDWL/ Avg AT) Compustat ((XRD + CAPX + ACQ SPPE DPC) / Avg. AT)

Foreign Assets

Compustat

Geographic Complexity

Compustat

Industry Complexity

Compustat

Audit Fees Tax Fees Proportion Tax Fees

Audit Analytics Audit Analytics Audit Analytics

All compensation and control variables are measured as of time t except where specified. Dependent variables (ETRs, taxable income, BTG, etc.) are computed at time t+1.

44

Table 1 Panel A: Industry Composition of Sample

Industry 1. Mining and Construction 2. Food 3. Textiles, Printing, and Publishing 4. Chemicals 5. Pharmaceuticals 6. Extractive Industries 7. Durable Manufacturers 8. Computers 9. Transportation 10. Utilities 11. Retail 12. Financial Institutions 13. Insurance and Real Estate 14. Services 15. Other Total Proprietary Sample Number Frequency 23 1.98% 82 7.06% 74 6.37% 80 6.88% 57 4.91% 36 3.10% 277 23.84% 28 2.41% 43 3.70% 72 6.20% 137 11.79% 174 14.97% 59 5.08% 2 0.17% 18 1.55% 1,162 100% Compustat Number Frequency 1,263 2.75% 859 1.87% 1,568 3.42% 955 2.08% 2,719 5.93% 1,686 3.67% 12,518 27.28% 2,338 5.09% 2,907 6.33% 1,797 3.92% 3,595 7.83% 6,806 14.83% 2,096 4.57% 3,862 8.42% 920 2.00% 45,889 100.00%

This table presents the industry classification of the proprietary sample and the Compustat database over the 2002 to 2006 period. The number and frequency of the 1,162 sample firm-year observations and the Compustat database from 2002 to 2006 are reported according to their industry classification following Barth et al. (1998).

Panel B: Proportion of Sample Tax Expense and Cash Paid for Taxes Ratio of Proprietary Sample to Compustat Tax Expense
36.05% 32.55% 28.35% 28.93% 28.45%

Ratio of Proprietary Sample in S&P500 to Total S&P500 # of Firms

50.80% 50.80% 51.20% 50.60% 49.20%

Assets 2002 2003 2004 2005 2006

31.07% 30.32% 29.03% 28.84% 28.09%

Cash Paid for Taxes

34.61% 32.09% 32.94% 36.27% 33.86%

Assets
64.78% 64.64% 64.19% 62.51% 60.83%

Tax Expense
64.67% 61.35% 58.61% 57.43% 55.26%

Cash Paid for Taxes

62.96% 62.55% 60.72% 58.26% 56.28%

Columns (1) to (3) of this table present the ratio of total assets (TA), total tax expense (TXT), and total cash paid for taxes (TXPD) for the proprietary sample to the total Compustat population of firms with data available, respectively. Columns (4) to (7) present the percentage of the S&P 500 firms in the proprietary sample, and the ratio of total assets (TA), total tax expense (TXT), and total cash paid for taxes (TXPD) for the S&P 500 firms in the proprietary sample to all S&P 500 firms.

45

Table 2 Descriptive Statistics

Mean
Compensation Variables Total CEO Compensation Total Tax Director Compensation Total GC Compensation Total CFO Compensation CEO Compensation Mix Tax Director Compensation Mix GC Compensation Mix CFO Compensation Mix CEO Bonus Mix Tax Director Bonus Mix GC Bonus Mix CFO Bonus Mix Tax Variables Book-Tax Gap Scaled Book-Tax Gap GAAP ETR Cash ETR Cash ETR (w/o Option Benefit) Cash ETR (3 Year) Pre-Tax Book Income Taxable Income (Less Option Benefit) Taxable Income Cash Flow From Operations Tax Benefit of Stock Options DTAX SHELTER Financial Variables Market-to-Book Ratio Market Capitalization Revenue Return on Assets Prior Years Return Std. ROA Prior Years Volatility Leverage Additional Control Variables Change in Goodwill New Investment Foreign Assets Geographic Complexity Industry Complexity Audit Fees Tax Fees Proportion Tax Fees 13,662,000 787,727 2,114,037 1,178,060 0.848 0.593 0.718 0.682 0.133 0.138 0.138 0.139 274 0.050 0.297 0.287 0.283 0.331 1,838 1,270 1,397 2,283 44.50 0.102 0.941 3.38 23,555 17,817 0.052 0.093 0.159 0.008 0.202 0.027 0.038 0.268 0.730 0.657 7,154 0.159 0.183

Median
9,839,090 558,714 1,355,193 888,292 0.902 0.629 0.762 0.722 0.105 0.128 0.119 0.127 67 0.016 0.313 0.233 0.230 0.267 683 446 470 796 0.00 0.010 0.981 2.42 9,173 8,211 0.049 0.128 0.128 0.004 0.184 0.001 0.020 0.166 0.853 0.773 4,140 0.067 0.137

Standard Deviation
13,722,377 695,740 2,200,675 980,002 0.161 0.205 0.183 0.181 0.120 0.090 0.103 0.098 2,153 0.340 0.148 0.262 0.262 0.274 3,730 3,178 2,984 4,126 374.27 0.478 0.114 3.39 42,365 31,817 0.066 0.329 0.144 0.014 0.132 0.072 0.065 0.313 0.298 0.368 11,289 0.221 0.164

25th Percentile
4,670,010 340,909 758,726 550,343 0.823 0.483 0.644 0.604 0.056 0.078 0.069 0.075 -41 0.000 0.246 0.123 0.118 0.179 187 96 104 271 0.00 -0.015 0.946 1.62 3,060 3,183 0.016 -0.021 0.065 0.002 0.104 0.000 0.001 0.000 0.462 0.314 1,963 0.013 0.050

75th Percentile
17,645,995 974,043 2,642,015 1,445,396 0.939 0.747 0.848 0.811 0.182 0.189 0.188 0.190 358 0.043 0.358 0.338 0.333 0.356 1,791 1,387 1,419 2,140 11.90 0.075 0.994 3.79 21,329 17,854 0.090 0.257 0.203 0.008 0.285 0.014 0.055 0.458 1.000 1.000 8,127 0.222 0.271

This table presents descriptive statistics for the sample of 985 firm-year observations for which we have all of the variables in the table. All variables are as defined in Appendix A.

46

Table 3 Determinants of the GAAP and Cash Effective Tax Rates

GAAP ETR Compensation Level
Intercept CEO Incentives Tax Director Incentives General Counsel Incentives CFO Incentives Return on Assets Std. Dev. ROA Log(Market Cap) Leverage Change in Goodwill New Investment Foreign Assets Geographic Complexity Industry Complexity Proportion Tax Fees Tax Fees Nobs Adjusted R2 (1) 0.411*** (3.09) -0.001 (-0.09) -0.028*** (-2.70) 0.001 (0.12) 0.008 (0.87) 0.343*** (3.67) -0.089** (-2.53) 0.008 (1.20) -0.018 (-0.34) -0.027 (-0.49) 0.074 (0.82) 0.037* (1.83) 0.095*** (3.68) -0.005 (-0.30) -0.042 (-1.34) 0.012 (0.27) 985 12.7% (2) 0.461*** (4.26)

Cash ETR Compensation Level

(5) 0.260 (1.54) -0.014 (-1.55) -0.023 (-1.36) 0.011 (1.12) 0.022 (1.43) 0.264** (2.21) -0.118*** (-3.14) 0.003 (0.37) -0.022 (-0.32) 0.196* (1.70) -0.362*** (-3.24) 0.018 (0.76) 0.044 (1.34) 0.006 (0.30) -0.044 (-1.09) 0.062 (0.95) 985 9.4% (6) 0.354** (2.33)

Compensation Mix
(3) 0.205*** (2.90) 0.010 (0.21) -0.075** (-2.20) -0.008 (-0.17) 0.019 (0.42) 0.371*** (3.95) -0.091** (-2.56) 0.005 (0.81) -0.021 (-0.41) -0.025 (-0.45) 0.081 (0.90) 0.037* (1.86) 0.095*** (3.64) 0.000 (0.00) -0.052* (-1.66) 0.022 (0.49) 985 12.3% (4) 0.213*** (3.17)

Compensation Mix
(7) 0.234*** (2.76) -0.107 (-1.59) -0.040 (-0.62) 0.016 (0.23) 0.118 (1.62) 0.267** (2.23) -0.116*** (-3.07) -0.001 (-0.12) -0.024 (-0.34) 0.190 (1.64) -0.354*** (-3.14) 0.017 (0.71) 0.040 (1.23) 0.009 (0.43) -0.046 (-1.14) 0.063 (0.97) 985 9.2% (8) 0.204*** (2.65)

-0.024*** (-2.78)

-0.064* (-1.95)

-0.015 (-1.11)

-0.010 (-0.21)

0.347*** (3.72) -0.09** (-2.52) 0.009 (1.44) -0.017 (-0.32) -0.026 (-0.47) 0.074 (0.83) 0.036* (1.85) 0.093*** (3.58) -0.004 (-0.23) -0.044 (-1.41) 0.014 (0.31) 985 12.6%

0.374*** (4.05) -0.092** (-2.57) 0.005 (0.89) -0.021 (-0.41) -0.025 (-0.45) 0.082 (0.93) 0.037* (1.86) 0.094*** (3.63) 0.000 (0.03) -0.053* (-1.67) 0.022 (0.49) 985 12.3%

0.272** (2.26) -0.113*** (-3.03) 0.004 (0.50) -0.019 (-0.28) 0.199* (1.73) -0.362*** (-3.23) 0.019 (0.83) 0.041 (1.22) 0.008 (0.37) -0.050 (-1.21) 0.064 (0.98) 985 8.9%

0.289** (2.42) -0.115*** (-3.07) -0.001 (-0.08) -0.021 (-0.30) 0.201* (1.74) -0.361*** (-3.26) 0.019 (0.81) 0.042 (1.26) 0.010 (0.48) -0.055 (-1.36) 0.069 (1.06) 985 8.7%

This table presents the estimates of equation (1) where the dependent variable is either the GAAP ETR or the Cash ETR. GAAP ETR is the financial effective tax rate for the year defined as total income tax expense scaled by pre-tax income (Compustat TXT / PI). Cash ETR is the cash effective tax rate for the year defined as the sum of total income taxes paid and the tax benefit of stock options scaled by pre-tax income ((TXPD + (TXBCO + TXBCOF)) / PI). Incentives for the CEO, tax director, general counsel, and CFO are measured as either Compensation Level (i.e., the natural logarithm of total annual compensation) or Compensation Mix (i.e., the ratio of variable compensation to total compensation). The remaining variables are as defined in Appendix A. Year and industry indicators are included but unreported. tstatistics calculated using robust standard errors (based on two-way clustering by firm and year) are reported in parentheses below the coefficient estimates. *, **, and *** denote statistical significance (twosided) at the 10%, 5%, and 1% levels, respectively.

47

Table 4 Determinants of the Book-Tax Gap and Components

BTG
Intercept CEO Incentives Tax Director Incentives General Counsel Incentives CFO Incentives Operating Cash Flow Std. Dev. ROA Log(Market Cap) Leverage Change in Goodwill New Investment Foreign Assets Geographic Complexity Industry Complexity Proportion Tax Fees Tax Fees Nobs Adjusted R2
0.183* (1.78) -0.001 (-0.17) -0.007 (-1.08) 0.003 (0.85) -0.004 (-0.74) 0.155* (1.85) 0.020 (0.56) 0.002 (0.70) -0.014 (-0.36) 0.065 (1.21) -0.062 (-0.90) 0.000 (0.00) 0.024 (1.10) -0.068* (-1.96) 0.000 (0.02) 0.016 (1.12)

Compensation Level PBI TI

-0.110* (-1.74) -0.004 (-1.36) -0.006 (-1.46) 0.003 (1.11) 0.005 (1.42) 0.836*** (11.99) 0.012 (0.42) 0.014*** (3.96) -0.016 (-0.64) -0.042 (-1.18) 0.091 (1.51) 0.020 (1.56) 0.036*** (2.73) -0.040* (-1.89) 0.031** (2.03) -0.004 (-0.50) -0.14** (-2.00) -0.003 (-1.34) 0.002 (0.45) 0.001 (0.44) 0.004 (1.08) 0.591*** (11.22) -0.002 (-0.11) 0.007*** (3.05) -0.037* (-1.77) -0.067* (-1.72) 0.089* (1.81) 0.011 (0.93) 0.016 (0.96) 0.010 (0.44) 0.026* (1.84) -0.003 (-0.36)

Diff
0.240 (0.63) 0.020 (0.90) 3.340 (0.07) 0.390 (0.53) 0.100 (0.76) 17.250 (0.00) 0.370 (0.54) 3.910 (0.05) 0.840 (0.36) 0.650 (0.42) 0.000 (0.97) 0.890 (0.35) 1.770 (0.18) 6.550 (0.01) 0.170 (0.68) 0.030 (0.86)

BTG
0.100 (1.39) -0.045 (-0.94) -0.033 (-1.31) 0.055* (1.81) 0.008 (0.24) 0.153* (1.80) 0.019 (0.54) 0.001 (0.19) -0.016 (-0.44) 0.063 (1.15) -0.061 (-0.88) -0.002 (-0.12) 0.024 (1.13) -0.069* (-1.91) 0.002 (0.08) 0.016 (1.16)

Compensation Mix PBI TI

-0.117*** (-2.86) -0.033** (-1.98) -0.001 (-0.06) 0.013 (0.81) 0.033* (1.91) 0.839*** (12.37) 0.012 (0.43) 0.012*** (4.23) -0.015 (-0.62) -0.044 (-1.22) 0.089 (1.46) 0.019 (1.63) 0.036** (2.49) -0.041* (-1.83) 0.031** (2.18) -0.004 (-0.46) -0.091** (-2.19) -0.013 (-0.62) 0.027 (1.62) -0.018 (-1.04) 0.019 (1.11) 0.593*** (11.16) -0.001 (-0.05) 0.007*** (3.04) -0.035* (-1.69) -0.068* (-1.73) 0.088* (1.76) 0.011 (1.00) 0.017 (0.96) 0.010 (0.38) 0.025* (1.73) -0.003 (-0.37)

Diff
0.470 (0.49) 0.970 (0.32) 3.910 (0.05) 3.700 (0.05) 0.730 (0.39) 17.450 (0.00) 0.350 (0.56) 3.540 (0.06) 0.790 (0.37) 0.620 (0.43) 0.000 (0.98) 0.800 (0.37) 1.710 (0.19) 6.600 (0.01) 0.240 (0.62) 0.010 (0.92)

985 7.6%

985 60.8%

985 50.1%

985 8.0%

985 60.8%

985 50.3%

This table presents the estimates of equation (2) where the dependent variable in the first column is the Book-Tax Gap (BTG). The second and third columns present results when the Book-Tax Gap is disaggregated into its two components, Pre-Tax Book Income (PBI) and Taxable Income (TI), and the equations are jointly estimated using Seemingly Unrelated Regression. The fourth column, Diff, presents the results of a chi-square test of the difference between the estimated coefficients in the PBI and TI models. Incentives for the CEO, tax director, general counsel, and CFO are measured as either the natural logarithm of total compensation or compensation mix. The remaining variables are as defined in Appendix A. Year and industry indicators are included but unreported. t-statistics calculated using robust standard errors (based on two-way clustering by firm and year) are reported in parentheses below the coefficient estimates. *, **, and *** denote statistical significance (two-sided) at the 10%, 5%, and 1% levels, respectively.

48

Table 5 Measures of Tax Aggressiveness (DTAX and Shelter)

DTAX
Intercept CEO Incentives Tax Director Incentives General Counsel Incentives CFO Incentives Return on Assets Std. Dev. ROA Log(Market Cap) Leverage Change in Goodwill New Investment Foreign Assets Geographic Complexity Industry Complexity Proportion Tax Fees Tax Fees Nobs Adjusted R2 0.435 (0.74) -0.013 (-0.38) -0.079 (-0.83) -0.048 (-1.08) 0.079 (0.82) -0.006 (-0.01) -0.317 (-0.38) 0.008 (0.23) -0.047 (-0.18) 1.483** (2.23) -1.119* (-1.86) 0.032 (0.23) 0.006 (0.04) -0.019 (-0.15) 0.586** (2.35) -0.270 (-0.92) 985 10.4% 0.579 (0.84)

Shelter
0.523*** (4.53) 0.001 (0.20) 0.003 (0.37) 0.005 (1.04) -0.002 (-0.32) 0.183* (1.87) -0.221 (-1.64) 0.033*** (6.08) -0.073* (-1.79) 0.025 (0.54) -0.038 (-0.58) 0.019 (1.07) -0.060** (-2.38) -0.006 (-0.43) -0.051 (-1.57) 0.064 (1.59) 985 41.9% 0.542*** (5.68)

-0.072 (-1.03)

0.005 (0.72)

0.018 (0.04) -0.383 (-0.46) 0.008 (0.21) -0.031 (-0.12) 1.489** (2.25) -1.136* (-1.89) 0.031 (0.22) 0.003 (0.02) -0.020 (-0.15) 0.570** (2.36) -0.274 (-0.94) 985 10.1%

0.184* (1.88) -0.216 (-1.64) 0.033*** (6.32) -0.074* (-1.83) 0.025 (0.53) -0.035 (-0.53) 0.019 (1.08) -0.060** (-2.43) -0.005 (-0.39) -0.051 (-1.57) 0.065 (1.62) 985 41.9%

This table presents the estimates of equations (3) and (4) where the dependent variable is either DTAX (i.e., excess permanent differences and ETR differentials as defined in Frank et al., 2009 and modified with the inclusion of estimated foreign assets) or Shelter (i.e., Wilsons (2009) measure of the probability that the firm engages in a tax shelter). Incentives for the CEO, tax director, general counsel, and CFO are measured as the level of total compensation. The remaining variables are as defined in Appendix A. Year and industry indicators are included but unreported. t-statistics calculated using robust standard errors (based on two-way clustering by firm and year) are reported in parentheses below the coefficient estimates. *, **, and *** denote statistical significance (two-sided) at the 10%, 5%, and 1% levels, respectively.

49

Table 6 Reverse Regression - Compensation Level

CEO
Intercept Return on Assets Std. Dev. ROA Log(Market Cap) Leverage Change in Goodwill New Investment Foreign Assets Geographic Complexity Industry Complexity Proportion Tax Fees Tax Fees GAAP ETR Cash ETR 12.617*** (26.57) 2.291*** (4.04) -1.771*** (-2.66) 0.381*** (10.99) -0.180 (-0.65) 0.027 (0.06) 0.533 (0.92) -0.046 (-0.33) -0.197 (-1.10) 0.184 (1.63) -0.375 (-1.59) 0.778** (2.32) -0.138 (-0.54) 0.044 (0.25) 0.008 0.927 985 36.1%

Tax
10.015*** (28.74) 1.829*** (4.53) 0.317 (0.60) 0.350*** (12.55) 0.018 (0.08) -0.045 (-0.17) 0.231 (0.60) 0.050 (0.47) -0.056 (-0.39) -0.048 (-0.56) 0.125 (0.64) 0.093 (0.36) -0.363*** (-2.63) 0.123 (0.98) 4.90** 0.027 985 44.9%

GC
10.85*** (24.79) 2.104*** (4.10) 0.955 (1.39) 0.294*** (7.73) -0.073 (-0.26) 0.029 (0.07) 0.557 (1.10) 0.015 (0.13) -0.263 (-1.43) 0.134 (1.22) -0.239 (-1.00) 0.551* (1.70) -0.140 (-0.70) 0.262* (1.71) 3.27* 0.071 985 31.7%

CFO
11.041*** (38.60) 2.041*** (5.21) -0.374 (-0.65) 0.287*** (12.55) 0.066 (0.33) 0.069 (0.24) 0.261 (0.70) -0.005 (-0.05) -0.212* (-1.76) 0.120 (1.54) -0.315* (-1.77) 0.36* (1.65) -0.087 (-0.64) 0.155 (1.33) 0.64 0.426 985 43.1%

GAAP ETR vs. Cash ETR (F-stat) p-value Nobs Adjusted R2

This table presents the estimates of regressing CEO, tax director, general counsel, and CFO incentive compensation on economic determinants of incentive compensation and the GAAP ETR and Cash ETR. The dependent variables are measured as the natural logarithm of the level of total annual compensation. The remaining variables are as defined in Appendix A. Year and industry indicators are included but unreported. t-statistics calculated using robust standard errors (based on two-way clustering by firm and year) are reported in parentheses below the coefficient estimates. The F-statistic and associated p-value from the test of the equality of the estimated coefficients on GAAP ETR and Cash ETR are reported. *, **, and *** denote statistical significance (two-sided) at the 10%, 5%, and 1% levels, respectively.

50

Table 7 Propensity Score Matched-Pair Analysis Panel A: Differences in GAAP ETR

Treatment
CEO Compensation Tax Director Compensation General Counsel Compensation CFO Compensation CEO Comp Mix Tax Director Comp Mix General Counsel Comp Mix CFO Comp Mix

Wilcoxon
0.557 -4.050*** 0.321 -0.890 -0.477 -2.204** 0.275 -1.389

p-value
0.577 0.000 0.749 0.373 0.633 0.028 0.784 0.165

t-statistic
-0.446 -3.519*** 0.202 -0.230 -1.224 -2.223** 0.125 -0.734

p-value
0.655 0.000 0.840 0.818 0.221 0.027 0.901 0.463

Mean Difference in GAAP ETR

-0.003 -0.029 0.001 -0.002 -0.009 -0.019 0.001 -0.005

Median Difference in GAAP ETR

0.005 -0.018 0.000 -0.006 0.000 -0.011 0.000 -0.008

Panel B: Differences in Cash ETR

Treatment
CEO Compensation Tax Director Compensation General Counsel Compensation CFO Compensation CEO Comp Mix Tax Director Comp Mix General Counsel Comp Mix CFO Comp Mix

Wilcoxon
0.062 0.679 -0.233 0.920 -1.381 0.110 0.592 0.989

p-value
0.951 0.497 0.816 0.357 0.167 0.912 0.554 0.323

t-statistic
0.402 -0.004 -0.712 1.567 -1.079 0.180 0.687 0.793

p-value
0.688 0.997 0.477 0.118 0.281 0.857 0.492 0.428

Mean Difference in Cash ETR

0.005 0.000 -0.009 0.020 -0.014 0.002 0.009 0.010

Median Difference in Cash ETR

-0.003 0.015 0.000 0.000 -0.009 0.000 0.000 0.011

This table presents the results of the difference in GAAP ETR (Panel A) and Cash ETR (Panel B) between the matched pairs of firms for different treatments which are either the level or mix of compensation for the CEO, tax director, general counsel, and CFO. The first column presents the treatment for which we modeled the conditional probability of having a certain level of the treatment (conditional on economic and tax planning characteristics) and then matched firms with the most similar conditional probability of treatment but the largest difference in the observed level of treatment. The second and third columns present a Wilcoxon statistic of the rank-sum difference in the median ETR between the matched pairs of firms and the corresponding p-value (two-sided). The fourth and fifth columns present a t-statistic for a test of the difference in the mean ETR between the matched pairs and the associated p-value (two-sided). The sixth and seventh columns present the mean and median difference in GAAP ETR (Panel A) and Cash ETR (Panel B) between the matched pairs. *, **, and *** denote statistical significance (two-sided) at the 10%, 5%, and 1% levels, respectively.

51