Do lenders price tax risk in the debt contract?

Transcription

1 Do lenders price tax risk in the debt contract? Zawadi Lemayian Olin Business School Washington University in St. Louis Gerald J. Lobo Bauer College of Business University of Houston Arpita Shroff Davies College of Business University of Houston-Downtown August 2017 Abstract We investigate the relation between tax risk and the cost of private debt. Tax risk can lead to significant negative outcomes, including penalties, interest and additional tax payments, which can increase the uncertainty of future cash flows and make it more costly for a firm to borrow. We find that borrowers with higher tax risk incur higher loan spreads than less (tax) risky borrowers. We also find that on average, lenders are more likely to require collateral, include more covenants, and issue smaller loans, when borrowers exhibit higher tax risk. JEL Classification: G31, G32, H26, M41 Keywords: Cost of bank loans, Tax risk We are grateful to Olin Business School (Washington University in St. Louis) and College of Business (University of Houston-Downtown) for financial support.

2 Do lenders price tax risk in the debt contract? Abstract We investigate the relation between tax risk and the cost of private debt. Tax risk can lead to significant negative outcomes, including penalties, interest and additional tax payments, which can increase the uncertainty of future cash flows and make it more costly for a firm to borrow. We find that borrowers with higher tax risk incur higher loan spreads than less (tax) risky borrowers. We also find that on average, lenders are more likely to require collateral, include more covenants, and issue smaller loans, when borrowers exhibit higher tax risk. JEL Classification: G31, G32, H26, M41 Keywords: Tax risk, Cost of bank loans

3 1. Introduction Uncertainty about taxation has become increasingly important for corporations in recent years. The growing concern about future cash flow risk due to future uncertainty in corporate taxes is attributed to tighter tax regulation, inconsistency of tax policies across jurisdictions, and the media hype surrounding the amount of taxes corporations pay. A 2013 Lloyd s survey indicates that company executives view tax risk as one of the top risks they face. The report further shows a significant jump in the ranking of tax uncertainty (Lloyd s Risk Index 2013). Such interest seems warranted since a firm s ability to effectively plan taxes while minimizing future tax risk can influence the decisions of its stakeholders, who include lenders, shareholders, customers and governments. In the same vein, in a 2014 Ernst and Young survey of 830 tax and finance executives, more than 80% of the respondents agreed or strongly agreed that tax risk and tax controversy will become more important for their firms in the next two years (Ernst and Young 2014). Consistent with this shift, the Dow Jones Sustainability Index now includes tax strategy as one of the criteria used to assess a corporation s sustainability risk (S&P Dow Jones Indices). The literature has linked tax risk to various costly outcomes, including greater dispersion of analyst forecasts, higher likelihood of settling with tax authorities, and the need to hold more cash (Blouin, Gleason, Mills, and Sikes 2010; Hanlon, Maydew, and Saavedra 2016). More recently, Hutchens, and Rego (2015) demonstrate that firms with higher tax risk face a higher cost of equity capital. However, the literature is largely silent on whether these costly outcomes extend to the debt market. We contribute to this research by examining the consequences of a firm s tax risk on external debt financing. We investigate how lenders price a borrower s tax risk in the interest rate and other non-price terms of the loan contract. This question is pivotal to our understanding of lenders perception of tax risk and the implications for borrowers. 1

4 Our study builds on prior research that examines which borrower characteristics influence lenders preferences in debt contracts (e.g., Bhojraj and Sengupta 2003; Anderson, Mansi, and Deeb 2004; Klock, Mansi, and Maxwell 2005; Chava, Livdan, and Purnandam 2009; Fields, Fraser, and Subrahmanyam 2012; Rahaman and Zaman 2013; Chen, Huang, Lobo and Wang 2017). This stream of research also finds evidence that borrowers with poor accounting quality based on annual report readability, accruals, and restatement measures are considered risky, and therefore incur higher borrowing costs (Ertugrul, Lei, Qiu, and Wan 2017; Bharath, Sunder, and Sunder 2008; Graham, Li, and Qiu 2008). While these results suggest a positive association between risk and borrowing costs, they do not take into account that firm risk can occur for at least two reasons. First, a borrower can be considered risky due to uncertainty in pre-tax cash flows. The borrower s inability to meet contractual debt obligations could be due to lack of pre-tax operating profit due to uncertainty regarding economic conditions, firm policies or firm operations. Second, uncertainty about the future tax liability due to prior and current tax positions can also influence a lender s assessment of a borrower s risk. In this study, we distinguish between these two sources of firm risk and focus on the lender s perception of tax risk. Given that lowering tax liability is one of the most used and often the last recourse for managers to increase the bottomline, we examine whether lenders perceive this as a firm risk and price the uncertainty in future tax liability in the debt contract. Our study is the first to examine the cost of tax risk to borrowing firms. A related study by Hasan, Hoi, Wu, and Zhang (2014) documents that firms with higher tax avoidance are more likely to incur higher loan spreads in their debt contracts. Tax risk and tax avoidance are distinct constructs. Tax avoidance is the reduction of explicit taxes (lower effective tax rates), while tax risk reflects a firm s (in) ability to sustain a given tax rate (regardless of the level of effective tax 2

5 rate) over a period of time. Tax risk considers the variation in effective tax rates over time, not the level. As Guenther, Matsunaga, and Williams (2016) point out, a firm can achieve a higher level of tax avoidance without using risky strategies. In addition, Dyreng, Hanlon, and Maydew (2008) show that firms are able to maintain stable tax rates over long periods of time. Thus, tax avoidance need not necessarily increase tax risk. 1 Tax risk broadly captures the possibility that a chosen action will lead to a tax outcome that is different from initial expectations (Neuman, Omer, and Schmidt 2016). This notion is rooted in economics and finance, where the term risk generally denotes variability of an outcome. Uncertainty about tax law and its application to company facts, as well as the likelihood of audit by tax authorities, which might result in unfavorable outcomes such as penalties and interest payments, can lead to tax risk (Shevlin, Urcan, and Vasvari 2013; Dyreng, Hanlon, and Maydew 2016). Given that the lender s primary concern is whether the borrower will generate sufficient cash flows to repay the loan on time, the lender is likely to be concerned about tax risk to the extent that it leads to additional cash outflow. Accordingly, a lender evaluating the viability of such a borrower will consider future cash flows from which interest and principal repayments are to be made as volatile and uncertain, and by extension consider the firm to be risky. The lender is likely 1 To illustrate, consider three firms effective tax rates in years 1, 2, and 3: Firm A (30%, 30%, and 30%); Firm B (15%, 15%, and 15%); Firm C (30%, 0%, and 15%). Using dispersion of effective tax rates as a measure of tax risk, Firm A and Firm B have tax risk equal to 0, while Firm C s tax risk equals 30. In contrast, with an average effective tax rate of 15%, both Firm B and Firm C would be classified as displaying higher tax avoidance than Firm A, whose effective tax rate is 30%. Studies on tax avoidance investigate how Firm B s and Firm C s contracts differ from Firm A s, while we distinguish Firm A s and Firm B s contracts from Firm C s. For example, Nashua Corporation, which is one of the firms included in our sample, reduced its effective tax rate from 114.6% in 1999 to 58.3% in Between 2001 and 2003, the firm s tax rate went from 12% to 39.5%, before increasing to 54.3%. During the same time frame, the firm s borrowing cost increased from 175 basis points in excess of LIBOR in 1999 to 275 basis points in

6 to price this risk in the debt contract, making a loan costlier for a borrower with substantial tax risk. Alternatively, if the lender believes that a firm s uncertain tax positions will be upheld in the future, the firm s tax risk might not affect debt-financing terms. This conjecture would also hold if the expected future costs (penalties, additional tax payments, interest) of these uncertain positions are considered immaterial by lenders. Additionally, if lenders believe that the risky positions will be upheld, they may ascribe a higher value and assign better credit metrics (lower credit risk) to the borrower, discounting perhaps that these positions can be challenged. To address our research question, we examine how the terms of lending contracts reflect lenders assessment of tax-related risk. We follow the methodology of Hutchens and Rego (2015) and measure tax risk using the volatility of cash effective tax rates over the past five years. 2 We use the interest component of the loan contract as our primary measure of the cost of debt. We also investigate the use of other debt contract provisions as compensation for lending to a borrower with high tax risk. We examine the likelihood of including collateral requirements, loan size, loan maturity, the likelihood of including performance pricing provisions, and the total number of covenants included in the debt contract. The use of these non-price terms is consistent with the idea that they are important in debt contracts (Melnik and Plaut 1986) and can be used in conjunction with the price of debt (Dennis, Nandy, and Sharpe 2000). We conduct our analysis in the private debt market. Our focus on private bank loans is motivated by the asymmetric nature of lenders payoffs, where their upside benefits are limited to the interest and principal amounts in comparison to the substantial downside risk they bear. Compared to public debt and equity issuances where there are numerous holders, private loans 2 Tax uncertainty can increase the tax risk of a firm due to the uncertainty that the tax positions taken by the company will be upheld in the future. 4

7 have greater downside risk as the lender is typically one bank or a syndicate consisting of a few banks. In addition, a private lender is likely to have greater access to the borrower s management, which gives the lender the ability to access information that potentially provides insight into a borrower s future risks related to tax strategies. In sum, private lenders are more likely than issuers of public debt or equity to price tax risk more efficiently in their loan contracts. Using a comprehensive sample of 11,721 loan facilities issued to U.S. public firms between 1995 and 2012, we test the link between debt contract terms and tax risk. We document higher loan spreads for borrowers with higher tax risk. Our results hold even after we control for tax avoidance and cash flow volatility, suggesting that creditors view tax risk negatively, which in turn leads to higher borrowing costs. We conduct additional tests on other non-price attributes included in loan contracts and find evidence that they are also systematically related to a firm s tax riskiness. In particular, we document that lenders are more likely to require collateral for borrowers with higher tax risk. Additionally, when tax risk is high, lenders are likely to issue smaller loans and impose more covenants. In sum, our results suggest that tax risk matters to lenders and provide insight into how they tailor the terms of debt contracts to fit borrowers risk profiles. Understanding lenders assessment of tax risk is important, given that firms rely significantly on public and private debt as sources of capital to fund operations. For instance, in the first quarter of 2015, the outstanding commercial and industrial loans in the U.S. were about $1.85 trillion, which indicates that debt is an integral component of a firm s capital structure (Forbes 2015). Our study adds to several streams of literature as follows. First, we contribute to debt pricing and covenant research. We provide empirical evidence that lenders price tax risk in the loan contracts. Prior studies have mainly focused on lenders perception of borrowers riskiness 5

8 without distinguishing whether the risk is due to uncertainty about pre-tax cash flows or the tax rate applied to pre-tax cash flows (e.g., Ertugrul et al. 2017; Hasan, Hoi, Wu, and Zhang 2017; Deng, Willis, and Xu 2017; Bharath et al. 2008; Graham et al. 2008). 3 We find evidence of higher borrowing costs and more stringent debt contract terms for borrowers as tax risk increases. Specifically, we document that borrowers with higher tax risk incur higher loan spreads. We also find that lenders are more likely to require collateral, require more control on the firm via covenants, and issue smaller loans when borrowers exhibit higher tax risk. These results hold even after controlling for the volatility of pre-tax cash flows, suggesting that tax risk is an important additional consideration for lenders. Second, we contribute to the emerging literature on the economic consequences of tax risk. We demonstrate that lenders perceive the costs of risky tax strategies (uncertainty of future cash flows, opaque financial reporting due to the complexity of risky tax strategies) to outweigh the potential benefits (cash savings), which results in higher charges for borrowers. Recent studies have examined how equity markets perceive tax risk. Bauer and Klassen (2014) document a negative market reaction to the revelation of news about significant tax settlements. Hutchens and Rego s (2015) findings that higher tax risk is associated with higher cost of equity capital and greater dispersion of analysts forecasts provide evidence of economic consequences stemming from tax-related uncertainties. De Simone, Mills and Stomberg (2015) identify a set of firms with greater opportunities for less risky tax avoidance mechanisms and find that the market values these firms more than firms with fewer such opportunities. We add to this literature by highlighting the extent to which tax risk matters to lenders, and show how lenders use price and non-price attributes 3 A notable exception is a study by Hasan et al. (2014) that examines whether lenders penalize firms for tax avoidance. 6

9 of the loan contract to compensate for the borrower s tax risk. To the best of our knowledge, this research question has not been addressed in previous studies. Third, we contribute to the literature on the importance of borrowers tax practices by showing that uncertainty about the tax rate to be applied to pre-tax cash flows matters to lenders. Specifically, lenders with high tax risk incur higher borrowing costs and more stringent contract terms. This finding provides insight into lenders preferences. Several papers in this area focus on tax avoidance, which, as described earlier, is distinct from tax risk. The studies on tax avoidance find conflicting evidence on whether lenders penalize or reward tax avoidance. One group of studies finds lower loan spreads and fewer covenant restrictions for firms with higher tax avoidance (e.g., Kim, Li, and Li 2010; Lim, 2011; Lisowsky, Mescall, Novack, and Pittman 2011; Saavedra, 2013). In contrast, Shevlin et al. (2013) and Hasan et al. (2014) document higher borrowing costs and more covenant restrictions for firms with greater tax avoidance. A potential explanation for the mixed evidence in this area is that tax avoidance can engender both certain and uncertain strategies. Indeed, tax avoidance does not necessarily translate to firm risk if the lower tax rate is consistent and likely to be upheld. We focus on tax planning that is more likely to be uncertain, and this can potentially explain the mixed evidence in this area of research. When tax rates are high, firms pay more taxes and therefore have less cash available. This could negatively affect lenders because it increases the probability of contract violation. It may also communicate poor management because management is not planning taxes efficiently, and this inefficiency could also be priced by lenders. Alternatively, when tax rates are low, it suggests that management is aggressive in tax planning, which could impose penalties and loss of reputation on the firm. This potential imposition of penalties and loss of reputation could also negatively affect lenders. Thus, both aggressive and 7

10 conservative tax planning could negatively affect lenders. Any departure from a firm s optimal level of tax planning is costly to lenders, and our focus on tax risk engenders this notion. Moreover, new evidence suggests that tax risk and tax avoidance are distinct, and on average tax avoidance encompasses strategies that do not necessarily increase tax risk (Dyreng et al. 2016; Guenther, Wilson and Wu, 2016; Guenther et al. 2016). Confirming the contrast between tax risk and tax avoidance, Guenther et al. (2016) document significant differences between the characteristics of firms with high tax risk and high tax avoidance. Firms with high tax risk tend to be smaller firms with lower investments in intangible assets, lower tax benefits from stock options, and higher leverage than the average firm. Alternatively, firms with high tax avoidance utilize more tax shields from stock options, research and development, and accelerated depreciation than the average firm. The rest of the paper is organized as follows. In Section 2, we discuss the prior literature and develop our hypotheses. In Section 3, we describe the sample selection procedure and provide descriptive statistics. In Sections 4 and 5, we present the research design and results for the main tests. In Section 6, we discuss the results of additional analyses and robustness tests. Finally, we conclude in Section Prior Literature 2.1. Prior research on tax avoidance in the credit market Prior literature has explored the role of tax-related factors in the credit market. Many of the papers in this field focus on tax avoidance, measured as the reduction of explicit taxes (Hanlon and Heitzman 2010). For example, studies have examined whether tax avoiders have lower credit ratings (Crabtree and Maher 2009; Ayers, Laplante, and McGuire 2010) and more frequent and pronounced credit rating disagreements (Bonsall, Koharki, and Watson 2015). 8

11 Studies have also examined lenders perception of the riskiness of firms that engage in tax avoidance. For example, Kim et al. (2010) and Lisowsky et al. (2011) document a lower interest expense for U.S based tax avoiders. Lim (2011) confirms these results for a sample of Korean firms. Conversely, Shevlin et al. (2013) find that firms with a higher level of tax avoidance incur a higher cost of borrowing in the bond market. Saavedra (2013) also finds that unsuccessful tax avoiders face higher borrowing costs. The author posits that these firms may have been unsuccessful as a result of engaging in riskier tax avoidance, but does not directly test this. Hasan et al. (2014) also document that firms with higher tax avoidance have greater spreads in both private and public debt markets, more covenants, and greater reliance on bank loans than on bonds (public debt) as a means of financing. In sum, the empirical evidence on the cost of borrowing for tax avoiders is mixed. A possible explanation for the conflicting evidence is that tax avoidance is a summary measure that engenders both certain and uncertain strategies. Tax avoidance is a broad term for the spectrum of tax planning strategies to reduce taxes. Tax risk, on the other hand, captures tax planning strategies whose outcome is ambiguous (e.g., Mills 1998; Graham and Tucker 2006). As Dyreng et al. (2016) note, tax planning strategies differ greatly in terms of uncertainty of the ultimate tax savings. For example, two firms, A and B, can achieve a similar effective tax rate if firm A engages in an uncertain tax strategy, such as a tax shelter, and firm B engages in a certain tax strategy, such as investing in municipal bonds. One might expect a higher borrowing cost for firm A if the strategy could be disallowed in future (reversing initial tax savings), or if the complexity of engaging in a tax shelter is also associated with attributes that lenders penalize, such as an opaque information environment (Balakrishnan, Blouin, and Guay 2012; Kerr 2013). Conversely, due to the relative certainty of firm B s strategy, one might expect lower borrowing 9

12 costs. This simplified example highlights a potential problem with using a summary measure of tax avoidance: magnitudes of risk can vary substantially for similar levels of tax avoidance, and it is unclear to what extent lenders recognize and price it in borrowers loan terms. In light of these concerns, recent research has examined tax risk and its relation to tax avoidance. Dyreng et al. (2016) find that firms with high tax avoidance bear tax risk, but this relation is driven by firms with subsidiaries in tax havens and high levels of research and development expenditures, which proxy for intangible-related transfer pricing strategies. De Simone et al. (2015) document that firms that shift income to low-tax jurisdictions report lower effective tax rates and have higher tax risk (uncertain tax benefits). Also, the majority of tax avoidance may not be uncertain (Guenther et al. 2016) or associated with either future tax rate volatility or firm risk (Guenther et. al. 2016). In sum, evidence in the literature suggests that tax avoidance and tax risk are distinct, and on average tax avoidance encompasses strategies that do not increase tax risk. Recent research has examined the effects of tax risk, which we discuss in the next section Research on consequences of tax risk Prior research has investigated several consequences of tax risk, including the frequency of settling with tax authorities (Blouin et al. 2010), managing earnings (De Simone et al. 2014), and cash holdings (Hanlon et al. 2016; Ciconte, Donohoe, Lisowsky, and Mayberry 2016). One of the most important regulations in this area is FIN 48, which governs the disclosure of firms tax risks. FIN 48 was passed by the Financial Accounting Standards Board (FASB) and became effective in Prior literature examines the information content of this disclosure by &blobheader=application/pdf (Last accessed 07/09/2017). 10

13 examining the market reaction around this event. For example, Frischmann, Shevlin, and Wilson (2008) document significant market reaction to the initial disclosures made under the accounting regulation FIN 48. Koester (2011) also documents a positive market reaction to disclosure of taxrelated liabilities, whereas Bauer and Klassen (2014) find that the equity market views tax risk negatively when a significant tax settlement is revealed. In sum, future uncertainty in income taxes matters to stakeholders of the firm. Moreover, recent evidence links tax risk to overall firm risk. In particular, Hutchens and Rego (2015) find a positive relationship between tax risk and the implied cost of equity capital. Guenther et al. (2016) corroborate these findings by showing that the volatility of cash tax rates is associated with future stock volatility. Taken together, these studies provide evidence that tax risk impacts equity holders assessment of a firm. Our study contributes to the literature by examining whether lenders also price tax risk in the private debt market Hypothesis development Lenders are concerned about future cash flows because of the asymmetric nature of their stake in the firm. They bear the downside risk without the possibility of sharing in any upside potential. Accordingly, lenders use screening and monitoring mechanisms to ensure that borrowers repay them in a timely manner. Tax uncertainty can result from implementation of a risky tax strategy, underestimation of required reserves reported in the firm s disclosures, the likelihood that uncertain tax positions will be upheld by the authorities, and/or possible fines and additional tax payments that might be imposed on the firm for tax positions that are subsequently disallowed. A borrower with high tax risk presents a credit risk; the lender can be exposed to volatility and uncertainty if unexpected payments to tax authorities in the future have to be made from the stream of cash flows that the lender expected to be paid from. Moreover, the uncertainty about the future 11

14 payments may lead the firm to hold additional cash (in anticipation of future payments) and thereby may constrain the firm s investment options, as well as introduce agency problems. Hanlon et al. (2016) document greater cash holdings for firms with uncertain tax positions, possibly to satisfy future cash payments. Guenther et al. (2016) also find that tax uncertainty is associated with future stock volatility, confirming that it contributes to overall firm risk. Accordingly, we expect lenders to demand greater compensation from borrowers with higher tax risk. Prior studies show that lenders evaluate a borrower s risk profile and demand greater compensation from riskier borrowers. Graham et al. (2008) find that loans issued after a firm restates its earnings have higher spreads, shorter maturities, greater likelihood of being secured, and more covenant restrictions. The relation is stronger if the restatement was fraud-related. Bharath et al. (2008) also document higher loan spreads for borrowers with high information risk. Wittenberg-Moerman (2008) finds a similar relation in the secondary debt market. We also know that tax positions are challenged, which leads to unsuccessful tax avoidance outcomes (Saavedra 2013). Therefore, we predict that borrowing costs increase with tax risk. Our first hypothesis, stated in alternative form, is: H1: Borrowing costs are increasing in a borrower s tax risk. There are several reasons we might not observe the predicted relationship. First, if lenders believe that a firm s uncertain tax positions will be upheld in the future, a firm s tax risk might not affect debt pricing. This conjecture would also hold if the costs incurred (penalties, additional tax payments, interest) when uncertain positions likely to be challenged in future are considered immaterial by lenders. Additionally, if lenders believe that the firm s tax positions will be upheld, they may ascribe a higher value to the company and therefore assign a better credit metric and 12

15 lower credit risk to the company. These arguments imply a negative relationship between the cost of debt and tax risk. Our next hypothesis explores the non-price terms lenders can use in addition to the explicit costs (loan spreads) borrowers are charged. Our premise is that these additional debt contract terms can be tailored to a borrower s risk profile. Hypothesis 2 predicts how various non-price loan terms relate to tax risk. Lenders can secure their future payments by including the requirement of collateral in the debt contract for firms with higher tax risk. Different theoretical perspectives have been offered regarding the types of borrowers that are more likely to be required to provide collateral. One view is that collateral is more likely to be pledged by low-risk borrowers as a signal of their creditworthiness (Boot and Thakor 1994), because low risk implies a lower likelihood of losing pledged collateral. An alternative view is that lenders demand more collateral from riskier borrowers because of the higher credit risk they present (Rajan and Winton 1995; Manove and Padilla 1999; Manove, Padilla, and Pagano 2001; Ertugrul et al. 2017). Requiring riskier borrowers to pledge collateral mitigates moral hazard. Empirically, the evidence from research in this area supports the view that lenders demand greater collateral from riskier borrowers (Berger and Udell 1990, 1992; Booth 1992). In turn, borrowers who pledge collateral obtain more credit and cheaper borrowing costs (Berger and Udell 1995; Jimenez, Salas, and Saurina 2006). This relationship is magnified for collateral that can be redeployed (Benmelech and Bergman 2009). Conversely, borrowers who cannot offer collateral, or whose collateral value decreases, face negative outcomes. For example, Gan (2007) finds that firms obtain smaller amounts of credit and are less likely to sustain a relationship with lenders following a decrease in the value of collateral. Thus, we expect lenders 13

16 to tighten collateral requirements to mitigate risks associated with borrowers prior tax positions. Stated in alternative form, our hypothesis is: H2a: The likelihood of a collateral requirement is increasing in a borrower s tax risk. Hypothesis 2b examines loan size. In a lending relationship, the size of the loan can affect the lender s exposure. In general, because of a higher probability of default, riskier borrowers tend to get smaller loans. Bharath, Dahiya, Saunders, and Srinivasan (2011) find that borrowers with greater information asymmetry get smaller loans, which suggests that lenders ration access to credit for risky borrowers. In part, this is because of the lower information asymmetry for larger loans as a result of the falling fixed costs to obtain information about the firm (Jones, Lang, and Nigro 2005). Further, Wittenberg-Moerman (2008) points out that there is a higher amount and quality of information for larger debt issues. Thus, for riskier borrowers, lenders prefer to offer smaller loans. A firm seeking access to more credit would obtain less favorable rates as a result of the increased uncertainty about its ability to generate cash flows to repay the loan. As a result, we expect the amount of credit available to a borrower to decrease as tax risk increases. This yields our next hypothesis: H2b: The size of the loan is decreasing in a borrower s tax risk. Hypothesis 2c focuses on the use of performance pricing provisions to impose greater discipline on borrowers. Performance pricing provisions link the cost of borrowing to a borrower s performance by letting the cost of debt vary with a performance metric such as credit rating or profitability. By including performance pricing provisions in a debt contract, the borrower benefits from reduced interest rates (when performance improves) and avoids renegotiation costs (when performance improves or declines). The lender benefits from reduction in the probability of default (when the borrower s performance improves) and increased interest (when the borrower s 14

17 creditworthiness declines prior to default). Consequently, lenders can use this provision to mitigate agency problems. Including this provision allows the lender to be compensated for small increases in a borrower s risk. Accordingly, we predict that lenders use of performance pricing provisions increases with a borrower s tax risk. H2c: The likelihood of inclusion of a performance pricing provision is increasing in a borrower s tax risk. Hypothesis 2d focuses on debt maturity. Theoretical work posits that the maturity of debt can be used as a mechanism to address agency cost problems when lending to risky borrowers (Flannery 1986; Diamond 1991). Several theoretical models provide a framework linking loan maturity to borrower risk. One stream of research predicts longer debt maturity for risky borrowers. Flannery (1986) models a firm with positive net present value projects spanning two periods. These projects can be financed using long-term debt maturing in two periods or shortterm debt maturing in one period, with the option to refinance at the end of the first period. Ex ante, the future refinancing rate is unknown. In this model, managers are more informed than the lenders regarding project quality. However, creditors can observe performance at the end of the first period. The model also assumes the existence of transaction costs, which are high enough to prevent pooling of bad and good borrowers. Consequently, good firms end up using short-term debt to finance their projects; they roll over at favorable rates at the end of the first period. In contrast, bad firms use long-term debt; they accept a higher rate of borrowing to avoid transacting at the end of the first period. Thus, this model predicts that loan maturity increases with borrower risk. 5 5 Diamond (1991) builds on this model by introducing a third category of risk, medium risk. Firms compare the benefits of using short-term debt and possibly renegotiating favorable future rates to the risk of liquidation, which would make it impossible to roll over debt. This model predicts that low-risk and high-risk firms borrow short-term, while medium-risk firms borrow long-term. In line with this conjecture, several studies find that debt with shorter 15

18 On the other hand, short-term loans force the lender and the borrower to renegotiate the loan terms, so riskier borrowers may not be granted long-term loans (Ertugrul et al. 2017; Graham et al. 2008). In particular, Ertugrul et al. (2017) document that borrowers with low readability of annual reports have shorter term loans. In view of the disagreement between the empirical evidence and the predictions of certain models, we do not make a directional prediction on the relation between loan duration and borrower s tax risk and, instead, present our hypothesis in the null form as follows: H2d: Loan maturity is not related to a borrower s tax risk. Our final hypothesis focuses on covenant usage. Covenants can be used to restrict managerial actions that can reduce debt and firm value (Jensen and Meckling 1976; Myers 1977; Smith and Warner 1979). Covenants can be beneficial to a lender if the borrower is risky, because covenant violation signals a borrower s decline in creditworthiness and can trigger the lender to exercise greater control rights. Covenants also incentivize lenders to continuously monitor the borrower (Rajan and Winton 1995). Indeed, empirical evidence supports the conjecture that the use of covenants is positively related to the borrower s risk (Demiroglu and James, 2010). In turn, several benefits accrue from their inclusion. Himmelberg and Morgan (1995) and Bradley and Roberts (2011) find that covenant use reduces borrowing costs and improves access to credit. Taken together, these findings suggest that covenant usage can limit a lender s exposure to tax risk. Our final hypothesis is: H2e: The intensity of covenant usage is increasing in a borrower s tax risk. maturity reduces managerial incentives to engage in risky activities (Leland and Toft, 1996; Stulz, 2000). Barclay and Smith (1995) argue that the shorter duration forces borrowers frequent disclosure and renegotiation. This in turn affords the lender greater flexibility to monitor the borrower (Rajan and Winton, 1995). 16

19 3. Main Variables, Sample, and Descriptive Statistics 3.1. Variables of Interest We follow Hutchens and Rego (2015) and Guenther et al. (2016) and measure tax risk, CETRVOL, as the standard deviation of annual cash effective tax rates (computed as the ratio of cash taxes paid to pretax income adjusted for special items) in the five-year period immediately preceding loan issuance. We use the natural log of this variable because of its skewed nature. CETRVOL reflects the volatility in the amount of cash paid for taxes resulting from the reversal of earlier positions, as well as related penalties and/or interest charges. A firm can exhibit high cash tax volatility if it pays significantly different rates of taxes in successive years, which would suggest that it has had little success in effectively applying tax strategies. In our multivariate models (described later), we control for tax avoidance, CASH_ETR, calculated as the cash taxes paid divided by the pretax income adjusted for special items in the year prior to loan issuance. Following prior literature we truncate this variable so that its value ranges between 0 and 1, and then multiply it by -1, so that an increasing value represents higher tax avoidance. If firms that avoid taxes are considered risky, lenders can demand higher costs when lending to borrowers that are aggressive in their tax strategies. Alternatively, the cash savings from successful tax avoidance can reduce the risk of default, leading to reduced borrowing costs Sample selection We use the Thomson-Reuters LPC DealScan database available on WRDS to access data on bank loans. In general, DealScan offers comprehensive coverage of the U.S. loan market (Chava and Roberts, 2008). The data are organized by packages (deals), which represent a contract 17

20 between the lender and the borrower. DealScan provides detailed information on loans including issue date, size, coupon rate, coupon frequency, maturity, collateral requirements, performance pricing provisions, covenant types, syndicate information, credit rating, purpose, and identity of the lending banks, as well as borrower-specific information. The basic unit of private borrowing is a lending facility. A firm can obtain multiple facilities with the same loan package in a contract year, and loan terms can differ across facilities. As a result, we treat each facility-year as a distinct observation. We match the same firm-year information to multiple facility-year observations if a firm obtains multiple facilities in one year. We first identify 29,104 facilities on Dealscan from 1995 to We require firms to have sufficient data to calculate different loan terms (Sufi 2007), deleting those with missing information about loan spread as well as those with spreads of less than zero. We then delete observations with lenders whose country of syndication is not the U.S. To be included in our merged dataset, observations must have non-missing assets. Similar to Hutchens and Rego (2015), we also exclude observations from the following industries: utilities, real estate investments, and financial institutions (Fama-French industry codes 31, 44, 45, 46 and 47), as their borrowing incentives differ from those of the remaining Compustat sample. In addition, firms in these industries face regulatory constraints on the permissible firm risk, which in turn might make their loan pricing terms different from those of companies without such limitations. We also delete firms with negative book value and firms with assets lower than one million dollars. Finally, we retain only observations with the required data for our empirical tests. We winsorize all continuous 6 We use the link file provided by Chava and Roberts (2008) to match the firms from DealScan with the Annual Compustat file from WRDS. The link file data are available only until year 2012, which limits our sample facilities to year

21 variables with values at the top and bottom 1% to limit the influence of extreme values. Our final sample consists of 11,721 facilities for 7,612 firm years Descriptive statistics Table 2 Panel A reports descriptive statistics for the variables used in our regressions. As indicated in Panel A, the average (median) rate of interest charged on the facilities is about 118 (150) basis points over LIBOR. 7 Also, the mean (median) number of general covenants in our sample facilities is 9.47 (11). Almost 99% of our sample facilities have seniority over other borrowings of the company, and a majority of the loans (91%) are syndicated (more than one lender). About half of the debt contracts in the sample include a performance pricing covenant. Similarly, about half of the sample facilities require collateral, and about 60% of the observations are revolver lines of credit. These loan characteristics are similar to those reported in prior research (Graham et al. 2008; Hasan et al. 2014). The mean (median) cash tax rate paid by the sample firms is 21.5% (22.3%), which is significantly lower than the statutory corporate rate of 35%. 8 This indicates that the majority of firms engage in tax planning. This is not surprising as a frequent allegation is that corporations do not seem to pay their fair share of taxes. Finally, the firms in our sample are relatively large, with an average of $968.4 million in assets. Overall, these statistics are in line with those of previous studies (Kim et al. 2010; Balakrishnan et al. 2012; Hasan et al. 2014). We report the industry distribution of the firms in our sample in Table 1 Panel C. The industries with the highest number of facilities are Business Services, Retail, and Petroleum and Gas (1,055, 912, and 894, respectively), while those with the fewest observations are Precious 7 Note that the values reflected in the table are natural logs of the interest rate of the facilities. 8 The values in the table reflect a cash effective tax rate multiplied by -1, so that avoidance is increasing in CASH_ETR. 19

22 Materials, Cigarettes, and Coal (15, 28, and 29, respectively). Overall, the table reveals that the loans are not concentrated in particular industries, as there is substantial variation in the number of firms from different industries. We provide correlations between the variables we use in our tests in Table 2 Panel B. Our tax risk measure (CETRVOL) is significantly and positively correlated with loan spreads, which suggests initial support for the prediction that loans are costlier for borrowers with substantial tax risk. We observe a negative correlation between CETRVOL and CASH_ETR, confirming some of the evidence that tax avoidance does not necessarily lead to tax risk. Finally, we also find evidence of significant correlations between loan spreads and loan- and firm-specific variables. In sum, the univariate analyses do offer some support for the predicted association between tax risk and borrowing cost. Next, we check whether these preliminary results hold in multivariate tests in which we account for systematic interactions between variables. 4. Borrowing costs and tax risk 4.1. Research design for H1: Relation between borrowing costs and tax risk Our first hypothesis predicts higher borrowing costs for firms with higher tax-related risk. Testing this prediction requires a proxy for borrowing costs. The construct we use is SPREAD, the natural log of loan spread. Loan spread is measured as the excess basis points over LIBOR or the LIBOR equivalent for each dollar drawn down (all-in drawn) for each facility. The spread includes both the fixed spread charge over LIBOR and other fees incurred. We use the following ordinary least squares (OLS) regression model that relates the cost of borrowing (SPREAD) to our measure of tax risk (CETRVOL) and several firm-specific and loanspecific controls: SPREAD = α + β1 CETRVOL + β2 CONTROLS + β3 INDUSTRY FE 20

23 +β4 YEAR FE + ε (1) In this regression, the main coefficient of interest is β 1. A positive and significant estimate would confirm our prediction that lenders view tax risk negatively. We include several firm-level controls that previous studies find to impact the cost of borrowing. We control for the firm s tax avoidance in the year preceding loan issuance using the firm s cash effective tax rate, CASH_ETR. By controlling for a firm s effective tax rate, which can be viewed as certain, we are better able to isolate the role of risky tax strategies, which are more uncertain. Findings in recent studies also suggest that lenders take into account a firm s tax avoidance (Shevlin et al. 2013; Hasan et al. 2014). Since a lender s main concern is the borrower s ability to service and repay the loan, we control for the volatility of pre-tax cash flows using CFVOL, computed as the standard deviation of cash flows from the statement of cash flows adjusted for taxes paid for the period from t-4 to t. By including CFVOL, we control for uncertainty arising from factors such as economic conditions, firm policies, and firm operations that are unrelated to tax risk. We also control for firm size because larger firms are generally more stable and likely to be more successful at obtaining better financing terms. We compute SIZE as the natural log of total assets. We calculate LEVERAGE as the ratio of total liabilities to total assets. More highly levered firms might incur higher borrowing costs to the extent that substantial amounts of debt increase default probability. We include BTM, the firm s book-to-market ratio, as a proxy for the firm s growth opportunities. Lenders may penalize a firm with high growth opportunities if the presence of growth opportunities encourages risky choices or increases the likelihood of financial distress. ROA, measured as the ratio of income before extraordinary items to total assets, captures a firm s profitability. Profitable firms are less likely to default. CASH is calculated as the ratio of 21

24 cash and short-term investments to total assets. We expect firms with greater cash holdings to incur lower borrowing costs, since cash can signal ability to make future interest and principal payments. TANGIBILITY, which we calculate as total property plant and equipment divided by total assets, captures the availability and redeployability of collateral. Finally, ZSCORE measures the borrower s bankruptcy risk. Lenders can ration credit or make it more costly to access for borrowers with a heightened risk of going bankrupt. We also include several loan-specific controls in our model. TOTCOV is the natural log of (1 + total number of covenants for each package). Covenants restrict managerial actions. Their inclusion in debt contracts increases the monitoring benefits to lenders, since covenant violations provide the lender with greater control rights. MATURITY is measured as the natural log of the number of months to maturity. We expect the lender s perception of a borrower s riskiness to influence loan duration. LOANSIZE is the log of the total loan facility. On one hand, larger loans can increase the risk of default, leading to greater borrowing costs. Conversely, larger loans may introduce economies of scale from a lender s perspective and thus reduce a borrower s cost. PP is an indicator variable that equals one for loans with performance pricing provisions and zero otherwise. COLLATERAL is an indicator variable that equals one if the facility has a collateral requirement and zero otherwise. REVOLVER equals one if the facility is a revolving credit facility and zero otherwise. SYNDICATE equals one if the lenders are in a syndicated arrangement, and zero if they are not. 9 Finally, our specification also includes year and industry fixed effects to account for any year- and industry-specific trends Results for test of H1: Relation between borrowing costs and tax risk 9 Although majority of our facilities are syndicated, we include this control variable for completeness. 22

25 In Table 3, we present estimation results for Equation (1), which examines the relation between the cost of borrowing and tax risk. We find a significant and positive coefficient on CETRVOL (β1= 0.350, t-stat = 5.87), consistent with higher borrowing costs for borrowers with higher tax risk, after controlling for loan and firm characteristics. In terms of economic significance, a 1% increase in the volatility of cash effective tax rates raises loan spreads by 0.35%, which translates to an increase of about 4.46 basis points (average spread (118 basis points) * β1 (0.35)* standard deviation of CETRVOL (0.108)). These costs are economically significant and imply that tax risk leads to increased cost of debt. 10 We find that firms with higher tax avoidance face greater borrowing costs, which indicates that tax risk is priced by lenders even after controlling for a borrower s tax avoidance. This finding is important because it suggests that lenders consider cash tax savings separately from tax risk, and the significance of tax risk in debt contracts is incremental to lenders consideration of the cash tax benefits (from certain and uncertain tax positions). This result corroborates the findings of prior research that tax risk and tax avoidance are distinct (Guenther et al. 2016). Further, we find a positive and significant coefficient on CFVOL (β1= 1.113, t-stat = 9.58), which indicates that lenders consider pre-tax cash flow volatility as an important metric to assess a borrower s repayment ability. A 1% increase in the volatility of pre-tax cash flows raises loan spreads by 1.113%, which translates to an increase of about 7.75 basis points. The effect of cash flow uncertainty is comparable to the effect of tax rate uncertainty (4.46 basis point increase). Our finding that both CETRVOL and CFVOL are significant in this specification indicates that both forms of risk (uncertainty of pre-tax cash flows and the tax rate to be applied to the cash flows) 10 For comparison, Bharath et al. (2008) find that a 1% increase in the standard deviation of accounting quality is associated with ~6 basis points reduction. Similarly, Hasan et al. (2014) show that a 1% increase in the standard deviation of tax avoidance results in ~5 basis points increase. 23

26 matter to lenders and are priced in the debt contract. Among the remaining control variables, we find that firms with greater cash holdings, higher leverage, and higher bankruptcy risk incur higher spreads on bank loans. In addition, spreads are higher for loans with more covenants, longer maturity, and collateral requirements. The cost of borrowing is lower for firms with more assets, loans that are smaller in size, loans that include performance pricing provisions, revolver loans, and loans with seniority. Overall, our results confirm our first hypothesis that higher tax risk is associated with increased borrowing costs. 5. Non-price terms and tax risk 5.1. Research design for H2: Relation between non-price contract terms and tax risk Our next test examines whether non-price terms are related to tax risk, based on the premise that they are important in debt contracts and can be used in conjunction with the price of debt to make total borrowing costs (implicit and explicit) either more or less costly (Melnik and Plaut 1986; Dennis et al. 2000). To test this prediction, we use several variables to capture the non-price loan contract terms and regress each of them on a baseline model that includes tax risk and controls for firm attributes and loan characteristics. NON-PRICE TERMS = α + β1 CETRVOL + β2 CONTROLS + β3 INDUSTRY FE + β4 YEAR FE + ε (2) The dependent variable, NON-PRICE TERMS, represents the non-price terms included in the loan contract. We use COLLATERAL, an indicator variable that equals one if the facility has a collateral requirement, and zero otherwise. We expect lenders to require borrowers to pledge collateral when their tax risk is high, because lenders can claim the collateral to compensate them for their losses in the event of default (Berger and Udel 1995). LOANSIZE is an indicator variable that equals one for loans above the median size, and zero for those below. If a borrower is risky, 24