Corporate Tax Planning and Stock Returns. Shane Heitzman. Maria Ogneva. University of Southern California Marshall School of Business

Transcription

1 Corporate Tax Planning and Stock Returns Shane Heitzman Maria Ogneva University of Southern California Marshall School of Business October 30, 2015 Abstract This paper investigates the asset pricing implications of corporate tax planning. Tax planning influences real and reported activities to optimize expected tax liabilities. As a result, it changes the amounts, timing and variance of corporate cash flows. Basic intuition suggests that tax planning can significantly influence firm value by changing expected after-tax cash flows. It is less clear whether it can also affect expected returns. By exposing the firm to higher uncertainty associated with government actions, tax planning may increase the firm s non-diversifiable risk. Our results suggest that a tax planning-based risk premium exists but depends on firm size, the political party of the President, and the ideology of US Tax Court judges. These results have implications for assessing the importance of tax planning risks in the contexts of both capital markets and corporate governance and control. While the expected cash flows should remain the primary focus of evaluating the net present value of tax strategy, managers, boards and investors should also consider the impact of tax strategy on discount rates. Keywords: Tax Planning; Tax Avoidance; Risk; Asset Pricing We appreciate the comments and suggestions of Michelle Hanlon, Ed Maydew and participants at the 2015 Oxford Tax Symposium and the discussant Robert Ullman. This paper benefited from conversations with Robert Novy-Marx and Cliff Smith. shane.heitzman@marshall.usc.edu; maria.ogneva@marshall.usc.edu 0

2 Corporate Tax Planning and Stock Returns 1. Introduction Corporate tax planning investments generate cash flows for shareholders by reducing the share of asset returns paid to the tax authority. These risky cash flows must then be priced. Relative to other key corporate finance decisions capital structure and payout policy, for example we know surprisingly little about the capital market implications of tax planning. The purpose of this paper is twofold: to motivate the conditions under which tax planning investments affect investors risk exposure and to provide empirical evidence on the association between tax planning and investors expected returns, a first-order determinant of share values. The impact of corporate tax planning on security values depends on how tax planning affects after-tax cash flows and the discount rate. The numerator in a typical valuation model picks up the effect of tax planning on expected cash flows (cash tax savings less the costs of defending the position, accounting and audit costs, restructuring costs and agency costs). 1 In the denominator, the discount rate reflects investors exposure to risk, specifically risk that cannot be eliminated through diversification. Tax planning investment affects risk by increasing the variance of aftertax cash flows. However, it is not clear a priori how it should affect expected returns, if at all. Risk driven by randomness in the audit process for example should not require a risk premium. When tax planning risk is diversifiable, incremental cash flows becomes the only relevant metric for investor valuation and tax strategy purposes (ignoring agency considerations). But if this strategy exposes investors to systematic risk, investors should demand a risk premium. To the extent these premiums are more than negligible, they should be detectable in firms stock returns. 1 These net cash flow benefits clearly exist but can be challenging to document since the costs of tax planning primarily affect earnings before tax while the benefits of tax planning are difficult to infer given the limited disclosure required by financial reporting rules. See Hanlon and Heitzman (2010) for a review of the recent literature. 1

3 To motivate the potential for tax planning to affect investors risk exposure, consider the government s resemblance to a minority shareholder with a legal claim to a share of the firm s economic profits (Desai, Dyck and Zingales 2007). What makes this claim on firm value unique is that this sharing rule varies, both over time and in the cross-section, depending on how managers and the government behave. Corporate tax planning by managers typically results in transactions that reduce the share of corporate profits paid to the tax authority. This alters the firm s cost structure by reducing costs that have a relatively strong covariance with performance (by lowering the effective marginal tax rate) and increasing costs that have a significantly weaker covariance with performance (the fixed costs of tax planning). Thus, a leverage-like financial risk arises that increases the volatility of residual cash flows to shareholders and should be associated with higher betas. 2 A more appealing question is whether the firm s tax planning policies expose investors to greater risk from future changes in the government s tax policies. There is growing evidence to suggest that exposure to government policy risks are important determinants of stock returns (Boutchkova et al. 2012; Croce et al. 2012; Pastor and Veronesi 2013). Tax legislation, enforcement and judicial decisions combine to determine the government s claim on asset returns. And every taxpayer faces some risk that this claim the expected cash flows paid to government will change in the future. However, for such policy risk to affect expected stock returns, it must be that the market expects tax policy changes to disproportionately affect high tax planning firms, thereby creating a co-variation among their returns because of correlated exposure to the shift in policy. This expectation might depend on business cycles (such as greater 2 In other words, in good times, shareholders benefit from a low marginal tax rate and hence have the equivalent of levered cash flows, raising returns. In bad times, managers have limited flexibility to unwind the transactions implemented precisely to reduce the marginal tax rate in good times. If tax planning increases investors exposure to financial risk, it should be priced. 2

4 enforcement in bad times) or political cycles (such as greater enforcement in Democratic administrations), and can vary in the cross section if firms differ in their adaptability to changing tax rules. 3 Across all potential sources of risk and there could be others the question is the same: does the intensity of corporate tax planning affect expected stock returns? An empirical answer to this question requires proxies for expected returns and tax planning intensity. Following prior asset pricing research, we match historical accounting and tax information from year t 1 with monthly realized returns from July of year t through June of year t + 1. Our primary tax planning proxy is an effective tax rate constructed from accrual-based measures of tax expense and pretax income over a three year period. We adopt this measure over the alternatives for several reasons. First, regressions of tax planning proxies on tax planning determinants indicate that the variation in the effective tax rate is driven by factors with higher tax uncertainty international operations and mobile capital (R&D) and not by variation in capital intensity which affects the timing of the tax liability but not the uncertainty. Because our tax planning proxy is accrual-based, it is potentially more informative about long-run tax planning policies than a cash-based measure. Additionally, Armstrong, Blouin and Larcker (2012) find that tax director compensation is most highly correlated with accrual-based effective tax rates, suggesting it is a more informative measure about tax planning performance. Finally, we show that accrual-based tax rates explain future settlement with the tax authority in the predicted direction; other measures do not. 3 For example, a tax authority might respond to a negative demand shock with a more aggressive approach to firms determined to push the limits, perhaps due to revenue or public pressure (Bagchi 2015). Similarly, Congress might respond to a negative demand shock by increasing incentives for investment or loss carryback in a way that primarily benefits tax-abiding firms. There is some evidence that the government responds to economic shocks both in writing tax law (Romer and Romer 2010) and interpreting it (Brennan, Epstein and Staudt 2009). If more intensive tax planning exposes shareholders to policy-based systematic risk, it should be priced. 3

5 We adopt industry-level measures of tax planning intensity to mitigate measurement error driven by firm-specific short-run measures (Dyreng, Hanlon and Maydew 2008) and biased inferences caused by dropping firms with losses (Henry and Sansing 2014). 4 This approach assumes that the potential for exposure to a tax planning risk factor is present for all firms, even those with current losses, and that opportunities for tax planning are correlated with industryspecific asset and operating characteristics. The results in this study provide evidence of a tax planning-induced risk premium. The average equal-weighted excess stock return for the portfolio of firms in industries with the lowest accrual-based tax rates is 3.6% higher annually than the return on the portfolio of firms in industries with the highest tax rates. These results are supported by cross-sectional regressions of monthly returns on industry effective tax rates. 5 Tests based on tax planning sorts within size quintiles suggest that the tax planning-based risk premium observed with equal-weighted returns is stronger in smaller firms. The average value-weighted returns are significantly higher for the lowest effective tax rate portfolios than for the highest effective tax rate portfolios, but only among the smallest stocks. This is not surprising 4 Problems attributed to firm-specific measures of tax planning include: a) bias caused by dropping or eliminating firms with negative pretax income, b) the impact of idiosyncratic accounting shocks in the measurement of earnings, c) the need for a long time series of data to calculate long-run cash tax rate measures, d) the relatively short time series of data on unrecognized tax benefits, e) lack of observability due to incentives for opaque disclosures, f) correlations between tax proxies and shocks to economic performance. In robustness tests, we examine the sensitivity of the results to two firm-specific measures of tax planning: long run cash tax rates and the tax reserve for uncertain tax benefits. 5 Of course, one can argue that the results are influenced by the mispricing of cash flows from tax planning. Suppose investors systematically underweight the cash savings from tax planning. This will cause high tax planning firms to have higher future returns as future cash flow realizations surprise investors. This could happen, for example, if investors are misled by a manager s decision to over-reserve for expected settlements with the tax authority. If this mispricing explanation is valid, however, it implies that the returns to an investment strategy that is long in high-tax planning firms and short in low-tax planning firms will earn excess positive returns in years where tax policy is more likely to be pro-taxpayer and negative in years where tax policy is more likely to be pro-government. It is also possible this risk premium will vary over time. If governmental responses to tax planning are expected to be more severe during Democratic presidencies, the risk premium required to invest in high tax planning firms will increase. This runs counter to the argument that cash flows are somehow mispriced. Chi, Pincus and Teoh (2013) argue that firms with low tax liabilities are overpriced and generate low future earnings and returns, suggesting that firms with low tax rates should also have lower returns. Our results do not support this. 4

6 if small firms find it costlier to hedge against future shifts in tax policy or protect their interest through political influence, or if the exposure to policy changes at large firms is lower because such firms are already under continuous audit by the tax authority. We also examine whether the tax planning premium is associated with time-varying factors that capture increased uncertainty in government reactions to corporate tax planning. We find evidence that a tax planning-based risk premium is stronger during Democratic presidencies, and holds for both equal- and value-weighted returns suggesting it is less sensitive to size effects. This is consistent with Bagchi (2015) who finds that IRS enforcement efforts increase when a Democrat is in office and with Belo, Gala and Li (2013) who find that stock returns are higher during Democratic presidencies for firms that benefit more from government spending. There is also some evidence that the ideology of the US Tax Court, measured by the political party of each judge s appointing President, affects returns. In periods with more Democrat-appointed judges on the Tax Court (19 judges each serving 15-year terms), stock returns are higher for high tax planning firms. The results for both the President and Tax Court are consistent with an ex ante premium to tax planning in periods of heighted risk of an adverse policy change. In contrast, we find evidence that high tax planning firms generate higher returns during periods of low IRS enforcement measured by corporate return audits. One explanation for this is that news about low realized audit rates causes investors to revalue expected cash flows from prior tax planning. This study contributes to the growing literature on the economic consequences of tax planning and the specific properties of firms tax policies. An implicit assumption in this literature is that greater tax planning has an effect on cash flow risk orthogonal to operating and investing risks. Our paper provides two key innovations. First, we focus on the non-diversifiable risk of tax planning, which is the only risk that should affects firms cost of capital in an efficient 5

7 market. Second, we consider a direct link between tax planning and firms systematic risk that results from the correlated shocks due to government s tax policy and tax enforcement changes. This is motivated by the recent work on the pricing of policy uncertainty and stands in contrast to existing studies that argue tax planning affects stock returns through corporate transparency. That approach relies on strong assumptions about the link between tax planning and corporate transparency as well as the pricing of transparency. Our research also contributes to a growing body of evidence on the importance of economic policy risk by focusing on a key element of fiscal policy: taxation and tax enforcement. Remarks by the IRS, law firms, governance advisors and accounting firms point to the increasing importance of managing tax risk. 6 The evidence presented here on shareholders demand for a premium to invest in firms with more intensive tax planning contributes to our understanding of the pricing of the firm s securities, and by extension, the types of company policies that will maximize firm value. Moreover, understanding the valuation consequences of tax planning appears crucial for the efficient design of incentives and organizational structures that exploit value-enhancing opportunities for coordinated tax planning. Thus, our findings are relevant to studies on the role of agency conflicts in tax planning and the broader discussion of tax risk management occurring in the boardroom. While managers and boards should continue to focus on the numerator expected incremental cash flows when evaluating tax planning opportunities, our results suggest that denominator effects can be an important consideration depending on the attributes of the firm, economic conditions and the political climate. 7 6 IRS News Release IR ; The Role of Executives, Counsel and Boards of Directors in Tax Risk Oversight Skadden, May 18, 2010; Bridging the divide; Highlights from the 2014 Tax risk and controversy survey Ernst and Young Of course, if the t-statistic cutoffs for statistical significance are increased following the suggestions of Harvey, Liu and Zhu (2014], the strength of our inferences must be moderated. 6

8 2. Prior literature 2.1. Valuation consequences of tax planning Economic intuition tells us that managers, properly incentivized, should adopt tax planning strategies that maximize firm value. 8 As firms have unique optimal investment and financing choices, the efficient tax strategy will also be unique, leading to variation in tax planning over time and across firms. If managers implement tax strategies that maximize firm value on average, there should be no association between measures of tax planning and equity values. However, if tax planning is associated with unresolved agency problems such predictions can change. Desai and Dharmapala (2006) link aggressive tax planning to corporate opacity and argue that the value of tax planning to shareholders is conditional on corporate governance. 9 In support of this, Desai and Dharmapala (2009) provide evidence that market-to-book, their proxy for value, is increasing in tax avoidance but only among firms with more institutional ownership. 10 This interpretation relies on the argument that tax planning has a directional effect on firm value and that the direction is conditional on external monitoring. Of course, agency conflicts can cause managers to choose too much or little tax planning and thus any deviation from optimal will reduce value. Consistent with this, Armstrong et al. (2014) provide evidence suggesting that greater monitoring appears to reign in both overly-conservative and overly- 8 Agency conflicts complicate the matter if managerial preferences for tax strategy diverge from shareholder preferences (Crocker and Slemrod 2005; Desai and Dharmapala 2006; Rego and Wilson 2012). 9 If opacity is required to hide aggressive tax positions from the tax authority, monitoring of managers also becomes more difficult, reducing the manager s cost to extract rents from shareholders (Desai and Dharmapala 2006). In this world, more aggressive tax planning signals an increase in agency costs and self-serving managerial behavior, but can be mitigated by greater monitoring. Kim, Li and Zhang (2011) provide evidence that firms that avoid more taxes are also more likely to experience a future stock price crash, but this likelihood apparently falls with stronger monitoring by analysts and institutions. 10 In related studies, Koester (2011) finds that the liability reserve for uncertain tax benefits is positively associated with share values, suggesting that investors view aggressive tax planning activity as value-increasing. Koester, Lim and Vigeland (2014) show that the valuation of this tax reserve falls when the internal controls have a tax-related material weakness. 7

9 aggressive tax avoidance. Given incomplete models of optimal tax planning, an inherent complexity and opacity of corporate tax reporting, and infrequent natural experiments, it has been difficult to draw concrete inferences about the relation between tax planning and value. In contrast to price-level studies, event studies offer the potential for more powerful identification of tax-based effects. Hanlon and Slemrod (2009) examine market reactions to allegations the firm is involved in a tax shelter. If this very aggressive form of tax planning is value destroying, and the public allegation informs market participants about this activity (as well as managers preference for it), the reaction should be negative. While they find negative average market reactions across the entire tax shelter sample, firms that appear to be less intensive tax planners have less negative market reactions, implying a positive market response to news of aggressive tax avoidance among firms perceived as not aggressive enough beforehand. Gallemore, Maydew and Thornock (2014) show that the average price declines at the announcement of the allegations are only temporary, consistent with tax shelter participation having minimal negative long-run effects on managerial reputation Tax planning and risk The notion of tax planning is broad and is often described with more colorful language depending on the context: avoidance, aggressiveness, sheltering or evasion, for example. It 11 Examining market reactions to corporate inversion announcements, Desai and Hines (2002) and Cloyd, Mills and Weaver (2003) find mixed evidence that the market responds systematically to announcements to relocate offshore. Part of the difficulty in interpreting market reactions in the inversion case arise because of investor level tax effects. In studies related to changes in accounting rules, Frischmann, Shevlin and Wilson (2008) examine the market s response to new uncertain tax benefit disclosure requirements under FIN48. They find little evidence that a firm s tax planning activity predicts stock returns around key FASB pronouncement dates related to FIN48, suggesting that the new accounting rule has little expected effect on the costs or benefits of tax planning. When the event date is centered on the firm s initial disclosure of the tax reserve, stock returns are positive and increasing in the reserve, suggesting that investors are, on average, positively surprised by the level of tax planning. Robinson and Schmidt (2013) find that the market s positive reaction to the initial disclosure of uncertain tax benefits is significantly weaker when the firm s disclosure quality is high, suggesting the value of aggressive tax planning is mitigated when firms are more transparent (and thus more likely to have a weaker position when challenged). 8

10 requires intentional actions that involve structuring the organization, its investments, transactions and reporting to exploit tax-based opportunities to increase shareholder wealth. These opportunities can arise from investments that are tax favored, transactions that exploit ambiguity in the tax law to reduce marginal tax costs, and discretion over the location, timing and characterization of reported income and deductions in the tax returns. There should be little disagreement that tax planning increases the riskiness of cash flows to investors: holding pretax cash flows constant, any reduction in the marginal tax rate will increase the volatility of after-tax cash flows. 12 However, whether tax planning leads to higher expected returns depends on how it affects the covariance of cash flows with performance and whether it changes the firm s exposure to tax policy shocks. The idea that tax planning has implications for the risk profile of the firm is gaining traction. Practitioner literature in tax management and corporate governance point to the increasing prominence of tax risk considerations in corporate decision making and control (Levin et al. 2006; Larsen 2011; Ernst & Young 2014). Neubig and Sangha (2004) describe a number of taxbased sources of risk that managers and directors should consider, some which apply to all firms and other that apply only to firms that exploit ambiguity in the tax law. Though the distinction is imperfect, one can think of tax risk as exposure to things in the manager s control (tax compliance activities, the location and type of investment and financial policies) as well as things out of his control (political ideologies of the President and Congress, IRS budgets and enforcement priorities, the macroeconomy, etc.). The IRS is becoming more outspoken on the role of corporate governance structures in addressing tax risk, ostensibly to influence corporate tax planning by appealing directly to those responsible for tax decision making and control. 12 Even if perfectly legal and certain, a reduction in the government s claim to firm assets increases the variance of returns to shareholders. 9

11 Recently, a number of studies address the basic link between tax planning and risk relying on an indirect corporate transparency argument. They argue that if aggressive tax planning is associated with greater asymmetric information a la Desai and Dharmapala (2006), and if asymmetric information is priced, then greater tax planning will increase the cost of equity capital. In contrast, one can appeal to the narrative in Gallemore and Labro (2015) who show that firms engaged in more intensive tax planning actually require high quality information systems to actually identify and support their tax planning strategies. Henry (2014) proposes a discount rate explanation for the link between stock returns and tax expense surprises (Hanlon, Laplante and Shevlin 2005; Thomas and Zhang 2014), and argues that tax surprises proxy for priced information risk arising from earnings manipulation and tax planning-driven opacity (Francis et al. 2005; Desai and Dharmapala 2006). 13 In debt markets, Shevlin, Urcan and Vasvari (2013) provide evidence that more aggressive tax avoidance leads to higher bond yields by reducing transparency. On the other hand, Goh et al. (2014) follow Gallemore and Labro and predict that greater tax avoidance should decrease the cost of equity capital by enhancing information quality. Using implied measures of expected returns and book-tax differences, and effective tax rates to measure tax planning, they document a negative association between tax avoidance and expected returns. This results in Shevlin et al. (2013) and Goh et al. (2014) are intriguing as one would 13 Henry (2014) uses variance decomposition techniques to link tax surprises to contemporaneous changes in stock prices and concludes that changes in tax expense are correlated with changes in priced risk. The results in Henry (2014) are difficult to interpret. Tax surprises appear negatively related to total news, although the effect is larger for the portion of news attributed to changes in the discount rate. However, when the regressions are estimated on firms with positive tax surprises only, the conditional association with discount rate news is only one-quarter the magnitude of the association with cash flow news. Henry partitions the sample on the sign of tax surprise to obtain a sort on firms more or less likely to have used tax expense to manage earnings, however, the sort on tax surprise is also a sort on profitability and growth: firms with increasing tax expense are also the same firms with expanding taxable profits. 10

12 expect tax-driven information quality problems to affect debt and equity in similar ways. 14 Guenther, Matsunaga and Williams (2013) argue that the volatility of a firm s cash tax payments is more indicative of tax risk than the level of tax payments used in prior studies, and show that the variance of a cash tax rate is positively correlated with future stock return volatility while the level of the effective tax rate is not. Rego and Wilson (2012) argue that tax planning increases risk and that risk-averse managers must be incentivized to undertake such risk. Consistent with that, they document that CEOs with stronger wealth sensitivity to risk (portfolio vega) appear to implement more risky tax strategies. However, they are silent on the link between tax planning and priced risk. This paper differs from existing work on several key dimensions. First, we provide a simple descriptive theoretical framework to motivate the existence and pricing of a tax planning risk factor. This contrasts with prior studies that either ignore the conditions under which tax planning should lead to higher expected returns or rely on tenuous links between tax planning, information quality and the pricing of information risk. Second, we employ standard empirical asset pricing methodologies to test whether tax planning is associated with expected returns. 15 Concurrent studies that measure expected returns using implied cost of capital measures suffer from bias caused by correlations between tax planning proxies and earnings shocks, exclusion of firms without analyst coverage and exclusion of firms with losses or negative expected earnings. Third, we employ industry-level accrual-based proxies for tax planning. We use correlations between tax planning proxies and firm s exposure to future conflicts with the tax authority to 14 Hutchens and Rego (2013) also examine the association between a number of firm-specific tax proxies and the discount rate implied by the relation between prices and analyst earnings forecasts. They find little evidence linking tax planning proxies to expected returns. 15 See Core, Guay and Verdi (2008) and Barth, Konchitchki and Landsman (2013) for examples from the accounting literature. 11

13 identify the proxy most consistent with our research question: accrual-based effective tax rates. Moreover, we rely on the logical assumption that tax planning incentives and opportunities are driven primarily by industry-based competition in product markets, organizational structure and investment opportunities. Because expected returns are based on forward looking assumptions, this allows us to maximize sample size and reduce bias induced by researchers typical selection on historical profitability. 3. Theoretical development 3.1. Tax planning and financial risk There are at least two potential channels through which tax planning should affect expected returns: financial risk and tax policy risk. It is well known that financial leverage from debt financing should be priced. Investors also appear to demand a risk premium to invest in firms with higher operating leverage, that is, firms whose operating cost structure is comprised of relatively greater fixed costs that have a lower covariance with the market (Lev, 1974; Novy- Marx, 2011). In a similar vein, tax planning can exhibit leverage-like characteristics depending on how it alters the total tax cost function. To illustrate, consider the taxable firm. Notwithstanding the asymmetry in the taxation of profits and losses embedded in the tax code, a firm s tax costs absent planning will be highly sensitive to profitability. When the firm engages in tax planning, it incurs new costs that to a first approximation do not vary with economic performance they are fixed (Mills, Erickson and Maydew 1998). These costs include the costs of identifying tax planning transactions, direct legal and accounting costs to structure the transactions, indirect costs when tax planning activities affect the distribution of information within the firm, legal costs and opportunity costs of managerial time if transactions are challenged by tax authorities, and lower pretax returns on 12

14 tax-favored investment structures. With positive demand shocks, shareholders benefit from the lower marginal tax rate created by tax planning, generating the equivalent of leveraged cash flows and raising returns. In bad times, managers have limited flexibility to unwind the investments, transactions and policies implemented precisely to reduce the marginal tax rate in good times. By reducing the covariance of tax costs with market performance, tax planning that induces leverage-like behavior in the firm s cost structure increases shareholders exposure to financial risk that should affect financial risk Tax planning and policy risk A second channel runs through exposure to tax policy risk. Tax policy is a cornerstone of the government s economic policies. Uncertainty about the government s policies can affect stock returns in the cross-section if a) investors expect tax policies to change, and b) the expected policy response has a disproportionate effect on firms that engage in more intensive tax planning. A growing literature in financial economics addresses the link between policy uncertainty and stock returns. In Pastor and Veronesi (2013), policy uncertainty matters because future policy choices can affect firm cash flows independent of the broader economic conditions and because of uncertainty about the impact of current policies on cash flows. Moreover, exposure to policy risk is magnified in weak economic conditions if policies are more likely to change during those times. Using the economic policy uncertainty index constructed by Baker, Bloom and Davis (2012) an instrument constructed from news coverage of policy uncertainty, expiration of tax provisions, and macroeconomic forecast disagreement they find consistent evidence that periods of greater policy uncertainty are associated with stronger correlations in stock returns, 16 Of course, if tax planning does not affect cost structure in this way, it may not affect investors exposure to systematic risk and thus will not be associated with returns through this channel. 13

15 higher risk premia, and risk premia that increase in weak economic conditions. In the cross section, Belo, Gala and Li (2013) show that firms in industries exposed to government expenditure policy (those in industries where more of the output is purchased by the government) command risk premiums in periods when a Democrat holds the presidency. The risks that derive from tax policy uncertainty result from at least three sources: tax legislation, tax enforcement and judicial ideology. Analyzing several decades of tax legislation, Romer and Romer (2010) conclude that tax legislation appears motivated by attempts to encourage and stimulate long-term growth. In some periods, tax policy is countercyclical increasing tax burdens to slow down expansions while providing tax breaks to stimulate investment and hiring in bad times. Only before the 1980s are tax changes spending driven. Uncertainty about future tax legislation is a key component of the economic policy uncertainty index of Baker et al. (2012). However, the extent to which uncertainty about tax legislation matters more for investors in high tax planning firms is not obvious. If investors expect tax legislation will directly target high tax planning firms (perhaps because the ability of these firms to effectively manage taxes downward is a response to controversial tax rules or incentives implemented to favor certain industries), then high tax planning firms should command a premium. If not, the impact on expected returns is not clear. Tax enforcement offers a more direct link between tax planning and policy uncertainty. The government s enforcement policies affect firms by changing the expected payoffs to tax planning. These policies are likely a function of political ideologies and macroeconomic conditions. For example, Bagchi (2015) finds that the political party of the President appears to influence how IRS enforcement resources are allocated, specifically the number of returns that get audited during the year. When a Democrat occupies the White House, more corporate returns 14

16 are audited. The number of corporate returns audited is also increasing in the federal deficit and decreasing in GDP growth, consistent with revenue pressures driving variation in enforcement efforts. Of course, the agency charged with enforcing tax laws sets tax policy other ways as well: evidence from the Treasury Department s anti-inversion guidance suggests targeted responses to firms pursuing mergers with low-tax foreign partners (IRS Notice ), transactions viewed politically as having primarily a tax avoidance motive. Finally, an interesting question is whether uncertainty about judicial perceptions of tax planning, and their impact on tax policy, also matter. This is relevant because controversial tax strategies with high potential payoffs are those where the wedge between the firm s and the tax authority s interpretations is large. Like enforcement, this offers a direct link between tax planning choices and policy risk. Before a tax position is adopted, the firm forecasts expected outcomes if challenged to determine expected cash flows. Ex post, when a dispute with the tax authority arises, the firm and IRS assess the optimal settlement strategy, both forming expectations of an uncertain judicial response. In both cases, knowing how the court is likely to rule will impact the decision to adopt the tax plan or fight the challenge. The potential for a systematic risk factor enters in if the court s attitude toward tax planning shifts over time due to economic or political forces. Brennan, Epstein and Staudt (2009) find that Supreme Court tax decisions vary predictably, becoming pro-government in periods of depression. Staudt, Epstein, and Wiedenbeck (2006) find that Supreme Court ideology (liberal vs. conservative) affects the likelihood of a pro-government ruling. More liberal courts are more likely to rule against corporate taxpayers, but are no different in their rulings against individual taxpayers. In this paper, we consider time series variation in the composition of the US Tax Court, the venue where most challenges to corporate tax planning are decided. 15

17 The collective evidence above suggests that economic policy uncertainty matters to investors, and that tax enforcement and judicial ideology are particularly relevant tax policy instruments that affect the government s claim to firm cash flows through their targeted impact on firms engaged in more tax planning (and hence more likely subject to enforcement or to end up in court). Not only do tax laws, enforcement efforts and judicial beliefs change over time, they can also be explained by the political ideology of the President and macroeconomic conditions. We consider a variety of instruments designed to capture time-varying risk that the government s tax policies shift against firms that engage in more aggressive tax planning: including the political party of the President, the intensity of corporate tax enforcement, and the composition of the Tax Court Caveats Despite these arguments, the theoretical and empirical progress on understanding tax planning cost behavior and fiscal policy motivations is still early and incomplete, making room for a credible (and interesting) null. The basic proposition that more intensive tax planning increases the variance of after-tax cash flows and hence firm risk is unsurprising, but whether investors require compensation for such risk is not obvious ex ante. If tax planning risks are driven by enforcement outcomes that are largely idiosyncratic and diversifiable, there should be no obvious risk premium. Even if a policy-based risk premium does exist, it may be too small to detect or may switch signs over time. Moreover, if tax planning does not change the firm s cost structure in a way that leads to increased financial risk for example, if tax planning investments are reversible in bad times or managers have other ways to hedge the downside risk of these positions a risk premium should not exist. We also lack a robust and well-accepted theory on the definition and measurement of tax planning risk, in part due to its correlation with business 16

18 risk and the interdependence of tax strategies. Ultimately, we adopt accrual-based effective tax rates because they perform the best at predicting future settlements with the tax authority. With that in mind, the growing importance of tax planning risk to regulators and practitioners suggests a clear demand for evidence on the apparent economic consequences of tax planning. 4. Empirical Methodology 4.1. Sample and variable definitions The sample consists of all U.S. firms traded on NYSE, AMEX or Nasdaq with at least two years of Compustat coverage and monthly returns available from CRSP between July of 1988 and December The sample begins in 1988 to ensure consistency in the definitions of tax expense. Stock returns from July of year t through June of year t + 1 are matched to accounting and industry information for the fiscal year ending in year t 1. Market returns are based on the value-weighted market portfolio. Industry classifications are obtained from Compustat. Tax planning measures Following recent research, we rely on effective tax rates to measure the tax planning intensity. Our setting requires a measure of exposure to future government actions that affect payoffs to tax planning policies. Prior research provides little guidance on which proxies for tax planning capture such risk best. To select the most appropriate measure, we examine correlations between three commonly used candidate proxies (effective tax rates, cash tax rates and book-tax differences) and subsequent adverse outcomes associated with the firm s tax policy (tax settlements). We measure tax planning proxies over a three-year window ending in year t, and look for settlements over the following three-year period as reported in mandated accounting disclosures that begin in We summarize our findings in Table 1. 17

19 We find that firms with low effective tax rates (high tax planning) have significantly higher rates of settlements with the tax authority. Surprisingly, firms with high cash tax rates (low tax planning) also have the highest rate of settlements even after controlling for size and profitability, suggesting that cash tax rates do a poor job picking up exposure to the types of tax planning that are subject to ex post challenges by the tax authority. We also consider book-tax differences calculated as the spread between book income and grossed-up current tax expense divided by assets and find that it behaves similarly to cash tax rates. Our primary tax planning measure, the effective tax rate, is defined at the industry level. We do so to mitigate measurement error in firm-specific measures and to accommodate the inclusion of firms with negative earnings or unreasonable tax rate proxies these firms are typically dropped from studies using effective tax rates proxies. The industry effective tax rate is the median three-year effective tax rate of firms with positive pretax income before special items in the firm s industry. It is measured as the sum of total tax expense over the three year period ending in year t 1 divided by the sum of pretax income before special items over the same period. 17 The cash tax rate is defined similarly except that cash taxes paid replaces total tax expense in the numerator. Industries are defined using the 3-digit SIC codes. Firm characteristics In the cross-sectional asset pricing regressions, we control for the common characteristics associated with firm risk, including market capitalization measured at the end of June in year t, book-to-market measured as the book value of equity (shareholders equity less book value of preferred stock plus deferred taxes) divided by the market value of equity at the end of the fiscal year, and leverage measured as total debt divided by the market value of assets (total assets less 17 The results are robust to using pretax income without adjusting for special items. 18

20 book value of equity plus market value of equity). We also include additional controls associated with incentives and opportunities for tax planning, including net property plant and equipment (PPE), balance sheet intangibles, and R&D and advertising expenditures (Dyreng, Hanlon and Maydew 2008). We control for financial performance with EBT (earnings before taxes and special items, scaled by assets). We also control for foreign operations, an important driver of tax planning opportunities, by including an indicator variable equal to one when the firm reports foreign pretax earnings. Because the existence of NOL carryforwards mitigates the value of tax planning, we include an indicator equal to one when the firm has positive net operating loss carryforwards at the end of the year Descriptive statistics and determinants of tax planning Table 2, Panel A reports average characteristics for portfolios sorted into quintiles each year based on effective tax rates. Firms are assigned to portfolios at the end of each June based on their tax planning intensity measured in the prior year. Assets and market value are adjusted for inflation. Panel A suggests key differences between high and low tax planning firms along dimensions such as growth opportunities, R&D, fixed assets and foreign operations. To explore these associations more formally, and to compare the drivers of effective tax rates with those of cash tax rates, we regress the effective tax rates on firm characteristics and compare these results to regressions using cash-based tax rates. The regressions are estimated annually using one observation per firm. The table reports averages of the time-series of coefficient estimates and their t-statistics. In the first regression, high tax planning firms (those with low effective tax rates] have higher book-to-market ratios, lower leverage, greater R&D and less advertising. They are also more likely to have foreign income or NOL carryforwards and have greater past return volatility. 19

21 The results paint an intuitive picture: firms with high growth opportunities (low book-tomarket) have yet to generate taxable profits to shield while firms with high leverage have higher effective tax rates when the tax benefits of debt (and thus marginal tax rates) are higher. Capital intensity reduces current tax payments under the long-standing accelerated deduction regime and affects cash tax rates, but not effective tax rates that include an accrual for the timing difference. Firms with R&D not only benefit from tax credits, the assets created from that investment are more mobile and generate income streams that can be shifted to lower-taxed jurisdictions. Firms with foreign income have greater opportunities through income shifting, and a negative association between loss carryforwards and effective tax rates can be driven by the fact that firms with loss carryforward have lower marginal tax rates by definition. Before continuing, we briefly consider the explanatory power of industry membership in determining firm-level tax planning, a key consideration of our use of industry proxies for firmlevel tax planning risks. Panel C of Table 2 reports incremental adjusted R-squared values from regressions of firm-level effective tax rates on year dummies, industry dummies and firm controls. Industry effects increase explanatory power by 120% for effective tax rates (from 4.09% to 9.00%) and 197% for cash tax rates (from 3.55% to 10.54%), both relative to year effects alone. Once year and industry effects are included, firm-specific controls contribute just 11.2% to effective tax rate variation (from 9% to 10.01%) and 27.6% to cash tax rate variation (from 10.54% to 13.45%). Industry membership explains tax planning proxies relatively well and avoids the problems inherent in using firm-specific measures described earlier Cross-sectional regressions of returns on tax planning To test whether tax planning intensity is able to predict returns, we first estimate Fama- MacBeth regressions of monthly stock returns on industry effective tax rates and firm 20

22 characteristics (e.g. Fama and French 1992). Table 3 reports average coefficients from the monthly cross-sectional regressions and the t-statistics corresponding to the time-series distribution of these coefficients. Effective tax rates have a statistically significant association with returns. The coefficients on the effective tax rate reported in the first three columns are statistically significant at conventional levels. In column (2), reducing the effective tax rate by ten percentage points (from 35% to 25%) is associated with a 3.74% higher stock return (annualized) after controlling for book-to-market, size and leverage (0.312 coeff. x 0.1 ETR change x 12 months), and 2.24% when the full complement of control variables are included (col. 3) Evidence from portfolio returns The cross-sectional tests suggest that tax planning intensity can explain future stock returns. In this section, we examine returns on portfolios formed on the basis of effective tax rates. Stocks are assigned to quintile portfolios at the end of June in year t based on the tax planning measure from year t 1. This timing ensures that public information about tax planning activity is impounded in price prior to the measurement of stock returns. The portfolios are re-balanced monthly to incorporate stock delistings. When a stock delists, we include the delisting return in the portfolio return calculation. Missing delisting returns are substituted with average delisting returns from Shumway (1997) and Shumway and Warther (1999). Table 4 reports average returns on tax planning portfolios. We estimate returns in excess of the risk-free rate, as well as alphas intercepts from the time-series regressions of excess returns on risk factors. The one-factor model includes excess return on the market, the three-factor model includes the market, size and book-to-market factors, and the four-factor model includes the market, size, book-to-market and momentum factors. All factor series are obtained from 21

23 Kenneth French s website. The last column contains returns on the hedge portfolio that is long (short) in the quintile of stocks with low (high) effective tax rates. The portfolio returns in Table 4 are almost monotonically decreasing along effective tax rate portfolios. The excess returns, three- and four-factor alphas on hedge portfolios based on effective tax rates are statistically significant. The average alpha from the three-factor model amounts to 37 basis points per month (4.4% annually) (t = 2.55), while the average alpha from the four-factor model amounts to 30 basis points per month (3.6% annually) (t = 2.01). These results suggest limited evidence of a tax planning-based risk premium. Of course, if the t-statistic cutoff for statistical significance were closer to 3.0 to account for the long history of risk factor tests as suggested by Harvey, Liu and Zhu (2014), these results would effectively disappear, suggesting no link between tax planning intensity and priced risk. In the following section, we consider the role of firm size, and time-series variation in tax planning premiums Sensitivity tests Time-series variation in risk premiums. The tax planning-induced systematic risk may be stronger in some periods than in others. For example, there is a significant increase in the proportion of IRS resources devoted to criminal investigations and in the number of corporate tax audits during the Democratic presidential administrations (Bagchi 2015). If Democratic administrations are perceived by the market as being more determined to curtail tax planning opportunities (through legislated tax policy, appointments to the Treasury Department, enforcement budgets and mandates and so on), and political interest in tax law depends on broader economic conditions (Romer and Romer 2006; Pastor and Veronesi 2013), then firms with more intensive tax planning should be exposed to more policy risk during those administrations. 22