2.2.1 Exchange Offers

To investigate whether the tax benefits of debt increase firm value (Prediction 1), Masulis (1980) examines exchange offers made during the 1960s and 1970s. Because one security is issued and another is simultaneously retired in an exchange offer, Masulis argues that exchanges hold investment policy relatively constant and are primarily changes in capital structure. Masulis’s tax hypothesis is that leverage increasing (decreasing) exchange offers increase (decrease) firm value because they increase (decrease) tax deductions. Note that Masulis implicitly assumes that firms are underlevered and that the exchange announcement is unanticipated. For a company already at its optimum, a movement in either direction (i.e. increasing or decreasing debt) would decrease firm value.

Masulis (1980) finds evidence consistent with his predictions: leverage-increasing exchange offers increase equity value by 7.6%, and leverage-decreasing transactions decrease by 5.4%. Moreover, the exchange offers with the largest increases in tax deductions (debt-for-common and debt-for-preferred) have the largest positive stock price reactions (9.8% and 4.7%, respectively). Using a similar sample, Masulis (1983) regresses stock returns on the change in debt in exchange offers and finds a debt coefficient of approximately 0.40 (which is statistically indistinguishable from the top statutory corporate tax rate during that era). This is consistent with taxes increasing firm value as in Eqn (4) (and is also consistent with some alternative hypotheses discussed below), but it is surprising because such a large coefficient implies near-zero personal tax and other costs to debt and/or large nontax benefits to debt. That is, the debt coefficient in Masulis (1983) measures the average benefit of debt (averaged across firms and averaged over the incremental net benefit of each dollar of debt for a given firm) net of the costs. An average net benefit of 0.40 requires that the costs are much smaller than the benefits for most dollars of debt. For the post-exchange offer capital structure to satisfy the MB = MC equilibrium condition, and given that the average benefit is approximately equal to the maximum statutory tax rate, the benefit or cost curve (or both) must be very steeply sloped near their intersection.

Myers (1984) and Cornett and Travlos (1989) argue that Masulis’s (1980) hypothesis is problematic. If firms optimize, they should only adjust capital structure to move toward an optimal debt ratio, whether that involves increasing debt or equity. In other words, increasing debt will not always add to firm value, even if interest reduces tax liabilities. Graham, Hughson, and Zender (1999) point out that if a firm starts at its optimal capital structure, it will only perform an exchange offer if something moves the firm out of equilibrium. They derive conditions under which stock price-maximizing exchanges are unrelated to marginal tax rates because market reactions aggregate tax and nontax informational aspects of capital structure changes. Therefore, nontax reactions might explain Masulis’s (1980) results. As described next, several papers have found evidence of nontax factors affecting exchange offer market reactions. It is important to note that these post-Masulis papers do not prove that the tax interpretation is wrong—but they do offer reasonable alternative interpretations.

First, some papers find evidence of positive (negative) stock reactions to leverage-increasing (leverage-decreasing) events that are unrelated to tax deductions: Asquith and Mullins (1986), Masulis and Korwar (1986), and Mikkelson and Partch (1986) find negative stock price reactions to straight equity issuance, and Pinegar and Lease (1986) find positive stock price reactions to preferred-for-common exchanges. Second, Mikkelson and Partch (1986) and Eckbo (1986) report that straight debt issuance (without equity retirement) produces a stock price reaction that is indistinguishable from zero. Third, some papers find that exchange offers convey nontax information that affects security prices, perhaps due to asymmetric information problems along the lines suggested by Myers and Majluf (1984) or due to signaling (Leland and Pyle, 1977; Ross, 1977). For example, Shah (1994) correlates exchange offers with information about reduced future cash flows (for leverage-decreasing offers) and decreased risk (for leverage-increasing offers). Finally, Cornett and Travlos (1989) provide evidence that weakens Masulis’s (1983) conclusions. Cornett and Travlos regress stock returns around the exchange event on the change in debt and two variables that control for information effects (the ex-post change in inside ownership and ex-post abnormal earnings). They find that the coefficient on the change in debt variable is insignificant while the coefficients on the other variables are significant, which implies that the positive stock price reaction is related to positive information conveyed by the exchange.¹⁰ Cornett and Travlos conclude that equity-for-debt exchanges convey information about the future—but find no evidence of increased value due to tax benefits.

Two papers examine the exchange of traditional preferred stock for trust preferred stock, also known as monthly income preferred stock (MIPS). These two securities differ primarily in terms of their tax characteristics, so any market reaction should have minimal nontax explanations. MIPS interest is tax deductible for corporations (like debt interest), and preferred dividends are not. On the investor side, corporate investors enjoy a 70% dividends received deduction (DRD) for preferred dividends, but recipients of MIPS interest receive no parallel deduction.¹¹ When issuing MIPS to retire preferred, corporations gain the tax benefit of interest deductibility but experience two costs: underwriting costs and possibly an increased coupon due to the personal tax penalty (because investors are fully taxed on MIPS interest in contrast to corporate investors, who receive the DRD on preferred dividends). Engel, Erickson, and Maydew (1999) compare MIPS yields to preferred yields and conclude that the tax benefits of MIPS are approximately $0.28 per dollar of face value, net of costs. Irvine and Rosenfeld (2000) use abnormal announcement returns to estimate the value at $0.26. Given that MIPS and preferred are nearly identical in all legal and informational respects, these studies provide straightforward evidence of the positive effect of interest tax deductions at the corporate level on firm value, net of underwriting and personal tax costs. Given that the maximum tax benefit was 34% during the sample period, and average benefits are about $0.27 according to these studies, the results imply that the non-bankruptcy costs of debt financing (including personal tax costs) are only moderate. (Note that expected bankruptcy costs are similar between MIPS and traditional preferred, so do not affect the numbers mentioned above).

2.2.2 Cross-Sectional Regressions

Fama and French (1998; hereafter FF) attempt to estimate Eqn (4) and Prediction 1 directly, by regressing V_{with debt} on debt interest, dividends, and a proxy for V_{without debt}. They argue that a positive coefficient on interest would be evidence of positive tax benefits of debt. FF measure V_{with debt} as the excess of market value over book assets. They proxy V_{without debt} with a collection of control variables, including current earnings, assets, and R&D spending, as well as future changes in these same variables. (All the variables in the regression are deflated by assets.) If these control variables adequately proxy for V_{without debt}, the regression coefficient on interest will measure the net benefit of debt (which is hypothesized to be positive). A major difficulty with this approach is that if the control variables measure V_{without debt} with error, the regression coefficients can be biased. (Another difficulty is that if all firms choose optimal capital structures and operate at the maximum V_{with debt} point, then one could not detect a causal empirical relation between value and debt usage. To detect causal relations between debt and value, one would require that a friction requires some firms to operate suboptimally).

FF perform a series of regressions on a broad cross-section of firms, using both level-form and first-difference specifications. In all cases, the coefficient on interest is either insignificant or negative. Fama and French interpret their results as being inconsistent with debt tax benefits having a first-order effect on firm value. Instead, they argue that interest provides information about earnings that is not otherwise captured by their controls for V_{without debt}. In other words, V_{without debt} is measured with error, which results in the interest coefficient picking up a negative valuation effect related to financial distress or some other cost.

Kemsley and Nissim (2002) attempt to circumvent this measurement problem. They perform a switch of variables, moving the earnings variable (which they assume proxies V_{without debt} with error) to the left-hand side of the regression and V_{with debt} to the right side. Therefore, their regression tests the relation V_{without debt} = V_{with debt} − coeff*D.

When Kemsley and Nissim regress earnings before interest and tax (EBIT) on V_{with debt} and debt, the debt coefficient is negative, which they interpret as evidence that debt contributes to firm value. The coefficient also changes through time in conjunction with changes in statutory tax rates. The Kemsley and Nissim analysis should be interpreted carefully. First, their regression specification can be interpreted as measuring the effect of debt on earnings, just as well as it can be interpreted as a switch-of-variables that fixes a measurement error problem in Fama and French (1998). Second, the debt coefficient has the correct sign for the full sample only in a nonlinear specification in which all the right-hand side variables are interacted with a crude measure of the discount rate. Finally, the coefficient that measures the net benefit of debt has an absolute value of 0.40. While consistent with Masulis (1983), such a large coefficient implies near-zero average debt costs including a near-zero effect of personal taxes.

2.2.3 Marginal Tax Benefit Functions

Graham (2000, 2001) uses a third approach to estimate the contribution of debt interest deductibility to firm value. He simulates interest deduction marginal benefit functions and integrates under them to estimate the tax-reducing value of a given amount of interest expense. For a given level of interest deductions, Graham essentially averages over possible states of the world (i.e. both taxable and nontaxable states) to determine a firm’s expected τ_C, which specifies the expected tax benefit of an incremental dollar of interest deduction. Because this approach focuses on incremental interest deductions, in the context of incremental decisions this approach largely avoids the endogeneity problem described below in Section 2.3.3.

Marginal tax benefits of debt decline as more debt is added because the probability increases with each incremental dollar of interest that it will not be fully valued in every state of the world. Using simulation methods (described more fully in Section 2.3.2) and various levels of interest deductions, Graham maps out firm-specific interest benefit functions analogous to the supply of debt curve in Panel B of Figure 2.

Using this approach, Graham (2000) estimates that the tax benefit of debt equals approximately 9–10% of firm value during 1980–1994 (ignoring all costs). van Binsbergen et al. (2010) update these estimates and find that the gross tax benefits of debt averaged about 9.0% from 1995–2009. The fact that these figures are less than the 14% estimated (at the beginning of Section 2) with the back of the envelope “τ_CD” calculation reflects the reduced value of interest deductions in some states of the world. When personal taxes are considered, the tax benefit of debt falls to 7–8% of firm value during 1980–1994 (i.e. this is Graham’s estimate of the “firm surplus” in Panel B of Figure 2), meaning that the personal tax penalty reduces corporate tax benefits by about one-third. These averages are instructive but they mask that for some firms, the tax benefits of debt are even larger. van Binsbergen et al. (2010) sort by tax benefits and estimate that for the top decile of firms, gross tax benefits average about 25% of firm value, and for these firms tax benefits net of all costs average about 10%.

Graham (2000; his Figure 2) also estimates the “money left on the table” that firms could obtain if they levered up to the point where their last dollar of interest deduction is valued at the full statutory tax rate (i.e. the “kink” in the tax benefit function, which is the point just before incremental tax benefits begin to decline).¹² The money left on the table calculations in Graham are partially updated in Table 1 of this paper. If all firms lever up to operate at the kink in their benefit functions, they could add 10.5% to firm value over the 1995–1999 period. This number can be interpreted (if many firms are underlevered) as a rough measure of the value loss due to conservative corporate debt policy, or (if most firms are optimally levered) as a lower bound for the difficult-to-measure costs of debt that would occur if a company were to lever up to its kink.

Table 1 Annual calculations of the mean benefits of debt and degree of debt conservatism. Before-financing MTR is the mean Graham (1996) simulated corporate marginal tax rate based on earnings before interest deductions, and after-financing MTR is the same based on earnings after interest deductions. Kink is the multiple by which interest payments could increase without a firm experiencing reduced marginal benefit on incremental deductions (i.e. the amount of interest at the point at which a firm’s marginal benefit function becomes downward sloping, divided by actual interest expense) as in Graham (2000). The tax benefit of debt is the reduction in corporate and state tax liabilities occurring because interest expense is tax deductible, expressed as a percentage of firm value. Money left on the table is the additional tax benefit that could be obtained, ignoring all costs, if firms with kink greater than one increased their interest deductions in proportion with kink.

This analysis initially yields two implications. First, the average unexploited tax benefits appear to be larger than the costs of debt that would occur if one were to lever up. Going back to Miller (1977) researchers have noted that the historic incidence of default is relatively low, meaning that the ex-ante probability of default is low. Interacting this low likelihood of default with ex-post measures of the cost of default (e.g. about 20% of firms value as in Andrade and Kaplan (1998)) implies expected costs of debt that on average appear to smaller than the forgone tax benefits. Almeida and Philippon (2007) point out the weakness of this logic. These authors highlight that distress most often occurs in bad states of the world, when the marginal utility of money is particularly high. Using credit default spreads to back out risk neutral probabilities that capture this effect, and using these risk neutral probabilities rather than historic incidence as their input into the expected cost of distress, Almeida and Philippon show that the expected cost of default approximately equals the “money left on the table” net of personal taxes estimate in Graham (2000), implying that firms on average may not be underlevered. This is an important point. One issue worth mentioning, perhaps, is that the Almeida and Philippon estimate of the personal tax costs, also used by Graham and others, is based on crude estimates. If this personal tax penalty happens to be overstated (e.g. if the marginal price-setter in the economy happens to be tax-free), then it is possible that the “underleverage” puzzle might not have been fully resolved by Almeida and Philippon.

The previous paragraph focuses on whether “on average” firms appear to be underlevered or not. The second implication of the money left on the table analysis is that cross sectionally the evidence seems to imply that the firms that appear to be conservative in their use of debt are large, profitable firms (which would seem to face the lowest costs of debt).¹³ This is puzzling because one would think that to forgo large tax benefits, a firm should face high expected costs of debt. In general, these implications are hard for a trade-off model to explain. There seem to be several possible explanations for this puzzle: (1) these firms that use debt conservatively do in fact face high costs but those costs have not been properly measured; (2) the apparent forgone tax benefits are overstated; (3) some firms are underlevered; and/or (4), the static trade-off model does not adequately explain debt policy. See Graham and Leary (2011) for a detailed examination of these issues. Briefly, I mention a couple points here.

Regarding (1), Graham (2000), Lemmon and Zender (2001) and Minton and Wruck (2001) try to identify nontax costs that are large enough in a trade-off sense that perhaps debt-conservative firms are not in fact underlevered. Blouin, Core, and Guay (2010) make some additional progress in identifying costs faced by apparently underlevered firms, though Graham and Kim (2009) argue that more evidence is needed to explain conservative debt policy. Regarding (2), it is important to emphasize that Graham’s (2000) tax benefit functions are calculated using financial statement data. Consequently, income or deductions that are “off balance sheet” may not be reflected in the income statement and therefore might not be captured in these tax benefit functions. This could result in financial statement income being overstated and hence potential tax benefits from levering up also being overstated.¹⁴Graham and Tucker (2006) gather data from 44 tax shelter legal filings and find that the annual deduction due to shelters is huge, averaging 9% of asset value. Given that these large deductions are not reflected in financial statement taxable income, these 44 companies had much less taxable income than was reflected in their financial statements, and hence the value gained from incremental interest deductions (i.e. apparent “money left on the table”) would also be less than estimated (because less “real” taxable income would mean that additional deductions in the form of interest would be less valuable). Similarly, Graham, Lang, and Shackleford (2004) hand-collect deductions occurring due to executive stock option exercises and find that these deductions are sizable, reducing taxable income for many firms and consequently reducing the tax benefits of incremental interest deductions. In both of these cases, firms that have substantial deductions that are not reflected in their financial statement filings would have smaller “tax benefits of debt” than would be estimated based on Graham’s financial-statement-based tax benefit functions. Therefore, researchers who do not incorporate deductions from off-balance sheet activity such as tax shelters and stock options may incorrectly conclude that a company uses too few interest deductions. Likewise, expenses associated with off-balance sheet pensions may have been missed by researchers attempting to measure taxable income using financial statements. It appears, however, that even after these special deductions are considered, the incremental tax benefits that could be obtained by issuing additional debt still appear to be fairly large. (One piece of good news: starting in reporting periods that end after June 2005, stock option deductions are now expensed, and hence financial statement income captures the effects measured by Graham, Lang, and Shackleford. Similarly, all pension expenses must now be reflected in the financial statements, so again, taxable income based on financial statement data should not be significantly flawed due to pension expense.)

To sum up, a fair amount of research has found evidence consistent with tax benefits adding to firm value. However, some of this evidence is ambiguous because nontax explanations or econometric issues cloud interpretation. Additional research in three specific areas would be helpful. First, we need more market-based research along the lines of the MIPS exchanges, where tax effects are isolated from information and other factors and therefore the interpretation is fairly unambiguous. Second, additional cross-sectional regression research that investigates the market value of the tax benefits of debt would be helpful in terms of clarifying or confirming the interpretation of existing cross-sectional regression analysis. Finally, if the tax benefits of debt do in fact add to firm value, an important unanswered question is why some firms do not use more debt, especially large, profitable firms.¹⁵ We need to better understand whether this implies that some firms are not optimizing, or whether previous research has not adequately modeled costs and other influences.

2.3.1 Static Tax Rates

2.3.2 Dynamically Simulated Marginal Tax Rates

One of the problems that led to Myers’s capital structure puzzle is related to properly quantifying corporate tax rates and incentives. For example, many studies use static MTRs that ignore important dynamic features of the tax code related to net operating losses carryback and carryforwards, investment tax credits and other nondebt tax shields, and the alternative minimum tax. Static MTRs miss the fact that a company might be profitable today but expect to experience losses in the near future. This firm might erroneously be assigned a high current-period tax rate, even though its true economic tax rate is low.²⁰ Conversely, an unprofitable firm might have a large current economic marginal tax rate if it is expected to soon become and remain profitable (because extra income earned today increases taxes paid in the future: an extra dollar of income today reduces losses that could be carried forward to delay future tax payments, thereby increasing present value tax liabilities).

Shevlin (1987, 1990) uses simulation techniques to capture the dynamic features of the tax code related to net operating loss carrybacks and carryforwards. His approach assumes that taxable income follows a random walk with drift.²¹ The first step in simulating an MTR for a given firm-year, based on random-walk income forecasts, involves calculating the historic mean and variance of the change in taxable income for each firm. The second step uses this historic information to forecast future income for each firm. These forecasts can be generated with random draws from a normal distribution, with mean and variance equal to that gathered in the first step; therefore, many different forecasts of the future can be generated for each firm. The third step calculates the present value tax liability along each of the income paths generated in the second step, accounting for the tax-loss carryback and carryforward features of the tax code. The fourth step adds $1 to current-year income and recalculates the present value tax liability along each path. The incremental tax liability calculated in the fourth step, relative to that calculated in the third step, is the present value tax liability from earning an extra dollar today, in other words, the economic MTR. A separate marginal tax rate is calculated along each of the forecasted income paths to capture the different tax situations a firm might experience in different future scenarios. The idea is to mimic the different planning scenarios that a manager might consider. The fifth step averages across the MTRs from the different scenarios to calculate the expected economic marginal tax rate for a given firm-year. Note that these five steps produce the expected marginal tax rate for a single firm-year. The steps are replicated for each firm for each year, to produce a panel of firm-year MTRs. The marginal tax rates in this panel vary across firms and can also vary through time for a given firm. The end result is greater cross-sectional variation in corporate tax rates (and hence tax incentives) than implied by statutory rates.

One difficulty with simulated tax rates is that they require a time series of firm-specific data. Moreover, they are usually calculated using financial statement data, even though it would be preferable to use tax return data. With respect to the first problem, Graham (1996b) shows that an easy-to-calculate trichotomous variable (equal to the top statutory rate if a firm has neither negative taxable income nor net operating loss (NOL) carryforwards, equal to one-half the statutory rate if it has one but not the other, and equal to zero if it has both), is a reasonable replacement for the simulated rate. With respect to the tax return issue, Plesko (2003) compares financial-statement-based simulated rates for 586 firms to a static tax variable calculated using actual tax return data. He finds that simulated rates (based on financial statements) are highly correlated with tax variables based on tax return data. Plesko’s evidence implies that the simulated tax rates are a robust measure of corporate tax status. Graham and Mills (2008) also find that the simulated financial statement tax variable is most highly correlated with a dynamic tax variable based on tax return data. The latter authors also provide an equation (based on regression coefficients) that can be used to estimate MTRs when outright simulation is not possible.

Note that by construction the simulated tax rates capture the influence of profitability on the corporate marginal tax rate. Graham (1996a) extends the simulation approach to directly capture the effects of nondebt tax shields, investment tax credits, and the alternative minimum tax. Graham (1996b) demonstrates that simulated tax rates are the best commonly available proxy for the “true” marginal tax rate (when “true” is defined as the economic tax rate based on realized taxable income, rather than simulations of the future). Using the simulated corporate marginal tax rates, Graham (1996a) documents a positive relation between tax rates and changes in debt ratios (consistent with Prediction 2), as do Graham, Lemmon, and Schallheim (1998) and Graham (1999) for debt levels. Since that time, numerous other studies have also used simulated tax rates to document tax effects in debt decisions. (See, for example, Kunieda and Real (2010) for evidence that in Japan the change in debt is positively correlated with simulated MTRs. Their results imply that a 10 percentage point decline in MTRs would lead to an annual one percentage point reduction in the debt ratio). These results help to resolve Myers’s (1984) capital structure puzzle; when tax rates are properly measured, it is possible to link tax status with corporate debt policy.

2.3.3 Endogeneity of Corporate Tax Status

Even if measured by a precise technique, tax rates are endogenous to debt policy, which can have important effects on tax research. If a company issues debt, it reduces taxable income, which in turn can reduce its tax rate. In essence, the more of the left tail of the income distribution that is negative, the lower the expected MTR; and, each incremental dollar of interest deduction pushes more of the left tail into (or closer to) negative territory. The more debt issued, the greater the reduction in the expected marginal tax rate. Therefore, if one regresses debt ratios on marginal tax rates, the endogeneity of corporate tax status can impose a negative bias on the tax coefficient. This could explain the negative tax coefficient detected in some specifications (e.g. Barclay and Smith, 1995b; Hovakimian, Opler, and Titman, 2001). Note that endogeneity can affect all sorts of tax variables, including those based on NOLs, or those based on the average tax rate (i.e. taxes paid/taxable income).

There are two solutions to the endogeneity problem. MacKie-Mason (1990) proposed the first solution by looking at (0,1) debt versus equity issuance decisions (rather than the debt level) in his influential examination of 1747 issuances from 1977 to 1987. Debt levels (such as debt ratios) are the culmination of many historical decisions, which may obscure whether taxes influence current-period financing choices. Detecting tax effects in the incremental approach only requires that a firm make the appropriate debt-equity choice at the time of security issuance, given its current position, and not necessarily that the firm rebalance to its optimal debt-equity ratio with each issuance (as is implicit in many debt-level studies). This approach also only requires that the incremental tax rate, for the next dollars of debt, is measured correctly (and of course, this is what marginal tax rates are designed to do). To avoid the endogenous effect of debt decisions on the marginal tax rate, MacKie-Mason uses the lagged marginal tax rate to explain current-period financing choice.²² He finds a positive relation between debt issuance and tax rates. Graham (1996a) follows a similar approach and examines the relation between changes in the debt ratio and lagged simulated MTRs. He finds positive tax effects for a large sample of Compustat firms.²³

If taxes exert a positive influence on each incremental financing decision, the sum of these incremental decisions should show up in an analysis of current debt levels—if one could fix the endogenous negative effect on tax rates induced by cumulative debt usage.²⁴ The second approach to fixing the endogeneity problem is to measure tax rates “but for” financing decisions. Graham et al. (1998) measure tax rates before financing (i.e. based on income before interest is deducted). They find a positive relation between debt-to-value and (endogeneity-corrected) “but for” tax rates. (They also find a “spurious” negative correlation in an experiment that uses an endogenous after-financing tax rate.)

2.3.4 Time-Series and Small-Firm Evidence of Tax Effects

The empirical evidence described thus far confirms cross-sectionally that firms with high tax rates use more debt than those with low tax rates. Presumably, there should also be time-series tax effects. For example, if a firm starts public life with a low tax rate, one would expect increased debt usage if the tax rate increases as the firm matures. There is not much direct time-series evidence of tax effects. Most of the evidence is cross-sectional, which could be problematic because it is difficult to control for all possible effects that might be correlated with tax rates but not represent actual tax-driven effects.

Examining, as in the previous section, changes in debt answers the question “are incremental decisions affected by tax status?” An alternative approach is to ask: “if tax rates exogenously change, how will a firm alter debt usage?” The Tax Reform Act of 1986 greatly reduced corporate marginal tax rates (see Figure 1), which in isolation implies a reduction in the corporate use of debt. Givoly, Hahn, Ofer, and Sarig (1992) find that firms with high tax rates prior to tax reform (firms that therefore probably experienced the largest drop in their tax rate) reduce debt the most after tax reform. This finding is somewhat surprising because their corporate marginal tax rate suffers from the negative endogeneity bias described earlier. Moreover, personal taxes are not modeled directly, even though they fell by more than corporate tax rates after the 1986 tax reform.²⁵ In a paper that examines international evidence during the same time period, Rajan and Zingales (1995) provide weak international evidence that taxes affect debt decisions.

Gordon and Lee (2001) use aggregate data from US tax returns. They show that a 1000 basis point change in tax rates leads to a 360 basis point change in debt ratios. In the first stage of their analysis, van Binsbergen et al. (2010) regress interest expense (divided by total assets) decreases on variables capturing the reduction in tax incentives following the Tax Reform Act of 1986, which greatly reduced corporate income tax rates. A novel feature of their experimental design is that the reduction in tax rates was phased in over two years, depending on a given firm’s fiscal year-end. Thus, the authors have nearly a difference-in-difference specification, where firms similar in every way have different tax rates just because their fiscal year ends in a different month. (The experiment is not quite a diff-in-diff because the actual calendar months covered by firms with different fiscal year ends differ by a certain number of months.) They find that debt usage fell with corporate income tax rates. As mentioned above, Faccio and Xu (2011) also find evidence that increases (decreases) in tax rates lead to additional (less) debt being used, especially for firms that use less (more) debt than the median.

By studying capital structure decisions among newly formed firms, one might be able to avoid long-lasting effects of past financing decisions. For example, Baker and Wurgler (2002) show that today’s market-to-book ratio and debt-equity issuance decisions continue to affect the firm’s debt ratios for ten or more years. Esty, Qureshi, and Olson (2000) describe various start-up financing issues, including selecting a target debt ratio, as well as how market conditions and collateralization affect the sequence of initial financing choices.

Pittman and Klassen (2001) examine capital structure in the years following an initial public offering (IPO). They perform annual (i.e. years since IPO) cross-sectional regressions and find evidence that taxes have a positive effect on the use of debt in the early years of a firm’s public life—but this relation wanes as the firm ages. Pittman and Klassen attribute this waning to an increase in refinancing transaction costs as firms age. Note that their evidence is not time series in terms of firms altering capital structure as tax rates change through time, though they do link debt policy to firm age. Pittman and Klassen also find that firms use relatively more NDTS as they age.

Almost all capital structure papers study Compustat companies. Ayers, Cloyd, and Robinson (2001) instead examine small companies with less than 500 employees that participated in the 1993 Federal Reserve National Survey of Small Business Finances. A total of 2600 firms meet the Ayers et al. data requirements. The authors regress interest expense divided by pre-interest pre-NDTS income on various variables, including tax expense divided by pre-interest income. They find a positive coefficient on the tax variable in both their outside and inside debt regressions (i.e. interest owed to nonowners and owners, respectively). It is difficult to compare their results to Compusat-based research because Ayers et al. use a different dependent variable to most studies, and they delete firms with a negative value for the dependent variable (which raises statistical issues).

2.3.5 Economic Importance of Tax Effects on Capital Structure Decisions

The evidence summarized in Section 2.3 indicates that there is a statistically significant relation between corporate tax rates and debt usage. Though not emphasized in every paper, there are important policy implications related to the economic magnitude of the taxes/debt policy relation. For example, as mentioned in the introduction, some argue that “debt bias” (i.e. using ‘excess’ debt in response to tax incentives) causes firms to become distressed too often, and in turn causes or magnifies economic recessions.

Graham (1996a) states that the tax effects appear to be second-order important economically. His estimated tax coefficient implies that if a firm’s tax rate increased from 0% to 35%, the firm would increase debt usage by about 2.5 percentage points. In contrast, a recent IMF meta-study (de Mooij, 2011) finds somewhat stronger evidence that the tax incentive to use debt is economically important. They conclude that an increase in a given company’s tax rate would increase its debt to assets ratio by between 6 and 10 percentage points. The IMF report, however, does not deal with the possibility that publication bias (i.e. a paper is much more likely to be published when it contains a significant result) could bias upward the average tax coefficient if it is primarily based on published estimates.

In a recent paper, Graham et al. (2011) explicitly examine debt bias and related issues. They study capital structure effects on corporate distress during the Great Depression and also during the recent 2008 recession. Not surprisingly, they find that the probability of distress increases with the amount of debt a firm uses. More relevant to this review, they also document a statistically significant tax effect on debt usage during the Depression. They use a two-stage process and examine whether this “extra” debt that firms took on due to tax incentives pre-Depression increased the probability that a firm encountered distress during the Depression. They do not find significant results in the second stage; that is, the authors do not find evidence that the tax component of debt usage increased the incidence of distress during the Depression. Nor do they find evidence of tax effects on capital structure in the recent recession time-frame. Two caveats should be mentioned. First, the tax variable the authors use during the Depression is crude due to data availability. Second, data quality in general is not as good in the Depression era, so lack of significance in the second stage could be related to lack of power. Much more empirical evidence is needed to explore this important issue.

To summarize Section 2.3, once issues related to measuring debt policy and tax rates are addressed, researchers have supplied ample evidence in response to Myers’s (1984) challenge to show that corporate debt usage is positively affected by tax rates. These results are consistent with survey evidence that interest tax deductibility is an important factor affecting debt policy decisions (ranking below only maintaining financial flexibility, credit ratings, and earnings volatility), and is especially important for large industrial firms (Graham and Harvey, 2001). Notwithstanding these empirical results, Myers is still not entirely convinced—in Myers et al. (1998) he argues that tax incentives are of “third-order” importance in the hierarchy of corporate decisions. My own take is that corporate tax effects do matter, though the effects often appear to be second-order in magnitude. It would be helpful for future research to investigate whether the tax effects on debt versus equity choice are economically important, and if they are not, to determine why not.

Several other challenges remain. First, few of the papers cited above provide time-series evidence that firm-specific changes in tax status affect debt policy. It would be quite helpful to examine whether a firm changes its debt policy as it matures and presumably its tax status changes. Second, Fama and French (2001) point out that with few exceptions the panel data examinations do not use statistical techniques that account for cross-correlation in residuals, and therefore, many papers do not allow for proper determination of statistical significance for the tax coefficients. Therefore, it is not clear if all of the tax effects documented above are robustly significant. Finally, most papers ignore the tax cost of receiving interest income from the investor’s perspective (though some work has been summarized herein).

2.4.1 Market-Based Evidence on How Personal Taxes Affect Security Returns

The papers cited above, though consistent with personal taxes affecting corporate financing decisions in the manner suggested by Prediction 3, are not closely tied to market-based evidence about the tax characteristics of the marginal investor between debt and equity. Instead, these papers assume that dividend clienteles exist, and they also make assumptions about the personal tax characteristics of these clienteles based on a firm’s payout policy. For example, many of these papers implicitly assume that there is a certain marginal investor who owns both equity and debt and (to estimate τ_P) that this same investor sets prices between taxable and tax-free bonds. The truth is that we know very little about the identity or tax-status of the marginal investor(s) between any two sets of securities, and deducing this information is difficult.

For example, assume that munis yield 7%, Treasuries 10%, and equities 8% (and assume that this equity return has been adjusted to make it risk-equivalent to the risk of munis and Treasuries). In a Gordon/MacKie-Mason/Graham type of equilibrium, r_muni = r_Treasury(1 − τ_P) = r_equity(1 − τ_equity) = 7%, which implies that τ_P = 30% and τ_equity = 12.5%. However, things are rarely so simple. First, it is difficult to determine the risk-adjusted equity return.²⁹ Second, if there are frictions or transactions costs limiting arbitrage between pairs of markets (or if risk adjustments are not perfect), one could observe, say, munis yielding 7%, Treasuries 10%, and equities 12%. In this case, it is not clear which pair of securities should be used to deduce τ_P. If Treasuries and equities are used, the implicit τ_P could be negative. For example, assume that dividend payout is 15%, that τ_{effective cap gains} = 5%, and that τ_equity is modeled as a weighted average between dividends and retained earnings: τ_equity = 0.15(1 − τ_div) + 0.85(1 − τ_{effective cap gains}). To ensure that r_Treasury(1 − τ_P) = r_equity(1 − τ_equity), in this example τ_P = −30% (assuming that τ_div = τ_P). Clearly, market frictions drive relative returns in this example, so the usual approach cannot be used to deduce the personal tax characteristics of the marginal investor(s).

Williams (2001) points out that when there are more than two assets, different pairs of assets can be arbitraged by different investors, so prices might reflect a mixture of tax characteristics. It is difficult to know which assets are directly benchmarked to each other by the marginal investor(s) and which are “indirectly arbitraged”, and it is even difficult to know whether capital gains or income tax rates are priced into security returns.

It would be helpful if future research could quantify the relative importance of personal taxes on security prices, with an eye toward feedback into capital structure decisions. One area in which a fair amount of research has been done along these lines involves determining the investor tax rate implicit between municipals and taxable government bonds. Poterba (1989) finds that the yield difference between high-grade one-year munis and government bonds approximates the top statutory personal tax rate, implying that the marginal investor between these two securities is a highly taxed individual. However, even this experiment is not without difficulty. First, returns on long-term munis and taxables imply a tax rate for the marginal investor that is approximately half that implied by the short-term securities. Chalmers (1998) shows that this holds even when the muni interest payments are prefunded by T-bonds held in defeasement (i.e. the muni interest is paid by earnings on a portfolio of government bonds purchased with the proceeds of the muni issuance), and therefore, risk differences between munis and T-bonds do not explain this conundrum. Green (1993) proposes that taxable bonds might not be “fully taxable” because a portion of their return can come from capital gains (especially for long-term bonds) and also because to some degree the interest income can be offset by investment interest deductions. Mankiw and Poterba (1996)suggest that munis might be benchmarked to equities by one clientele of investors and taxable bonds might be benchmarked to equities by another clientele. In this case, munis and taxables might not be directly benchmarked to each other, which could explain the unusual implicit tax rate that is sometimes observed between the two securities. Finally, recently Longstaff (2010) used an affine term-structure approach and estimated a high implicit marginal tax rate, putting us more or less back where we started.

As an example of trying to link the effects of personal taxes to capital structure issues, consider the implications from Engel et al. (1999) and Irvine and Rosenfeld (2000) about the personal tax penalty.³⁰ Assume that corporations are the marginal investors in preferred stock but not in debt.³¹ Given the similarity of the securities, in equilibrium, we expect their after-investor-tax returns to be equal, within transactions cost bounds: r_preferred(1 − τ_DRD) = r_MIPS(1 − τ_P). Plugging in r_preferred = 8.14% and r_MIPS = 8.37% from Engel et al.’s Table 3, and assuming that the marginal corporate investor is taxed at 35% so that τ_DRD = 10.5%, we can back out the personal tax rate associated with interest income: 0.0814(1 − 0.105) = 0.0837(1 − τ_P) implies that τ_P = 13%. If we ignore the 30 basis point “yield premium” on MIPS imputed by Engel et al. and use r_MIPS = 8.67%, τ_P = 16%.

Table 2 Tax incentive to use debt in a US multinational firm with foreign tax credits and allocable domestic interest. Assume that a US multinational firm currently returns $1 of pre-corporate-tax earnings to its marginal investor as domestic equity. The one-period model in this table shows the tax effect of instead paying the $1 as foreign interest (rightmost column in each panel) or as $1 of domestic interest (second-to-rightmost column). The model is adapted from Collins and Shackelford (1992) and assumes that all foreign income (Inc_For) is repatriated every year and that tax rules are the same worldwide, except that only the United States allocates interest. The model ignores the AMT, carrybacks and carryforwards, personal taxes, and allocable items other than interest. Because the real-world tax-code is dynamic (i.e. it allows for carrybacks and carryforwards), the one-period nature of this model might overstate (understate) the largest (smallest) tax benefits. Note that foreign losses (i.e. Inc_For − Int_For < 0) cannot be repatriated as losses back to the United States. FTC_allow is allowable foreign tax credit (sometimes referred to as FTC_limitation), FA is foreign assets net of foreign debt, WA is worldwide assets net of foreign debt, and FSI is foreign source income, which equals Inc_For − Int_For −  Int_US.

^∗In a multiperiod model, FTCs above the allowable amount could be carried back or accumulated and carried forward. For example, in the excess credit case with interest allocation (row 5), of unused FTCs accumulate per incremental dollar of domestic interest.

Table 3 Summary of predictions and empirical evidence for multinational capital structure

To the extent that results based on MIPS interest carry over to debt interest, finding τ_P = 16% for the marginal debt investor is intriguing. First, note that the mean after-financing corporate tax rate in the 1993–1999 sample period is approximately 18% (see Table 1), which is a rough estimate of the tax benefit of the last dollar of interest deduction (ignoring all costs). If we make Miller’s (1977) assumptions that τ_E = 0 and that all firms face the same 18% marginal benefit of debt, then τ_P should equal 18% (i.e. MC should equal MB), quite close to the τ_P = 16% MIPS estimate. As argued by Green and Hollifield (2003), it would only take fairly small costs of bankruptcy to equalize the costs and benefits of debt, creating an environment conducive to equilibrium with internal optimal debt ratios. However, τ_E is most likely not zero for the marginal investor in equities. (Green and Hollifield argue that deferral reduces effective τ_E to about half its statutory level.) Another issue is that the estimated MIPS costs and benefits are average, not marginal. Even if the marginal costs and benefits are equal in an equilibrium like that depicted in Figure 2a, there is a firm surplus/benefit to using debt. Therefore, even if personal tax costs are large enough at the margin to equal marginal benefits, there appear to be tax-driven preferred capital structures for some firms. Presumably, the incremental tax savings would be near $0.35 per dollar for high-tax-rate firms, while the personal tax cost is only half that amount. Only if the nontax costs of debt are large for these high-tax-rate firms could a full Miller equilibrium hold, in which the benefits of debt are zero for all firms in equilibrium.

In sum, the implicit personal tax costs estimated here suggest that at the margin the tax costs and tax benefits might be of similar magnitude. However, they do not explain cross-sectionally why some inframarginal firms (with large tax benefits of interest) do not use more debt. (More details on this issue are presented in Section 2.4) One other area in which there has been a fair amount of success in deducing marginal investor tax characteristics—though not unambiguously so—is related to ex-day dividend returns. This discussion is deferred to Section 4, which explores how taxes affect corporate dividend policy.

In the most general sense, any research that shows that personal tax rates affect security returns sheds light on Miller’s (1977) claims. Using the CAPM-with-taxes specification, Auerbach (1983) finds evidence that tax-related preferences result in clienteles of investors that purchase stocks based on firm-specific dividend-price ratios. Constantinides (1983) and Dammon, Spatt, and Zhang (2001) investigate how favorable capital gains taxation affects investment and consumption choices. Seida and Wempe (2000) show that individual investors accelerated recognizing capital gains (and delayed losses) in anticipation of the increase in capital gains tax rates associated with the 1986 tax act. For a review of articles related to how personal taxation affects the timing and value of asset sales and purchases, see Poterba (2001).

2.5.1 Leasing

The discussion thus far has considered the debt versus equity choice; however, it can be extended to leasing. In certain circumstances, a high-tax-rate firm can have a tax incentive to borrow to purchase an asset, even if it allows another firm to lease and use the asset. With true leases (as defined by the IRS), the lessor purchases an asset and deducts depreciation and (if it borrows to buy) interest from taxable income. The lessee, in turn, obtains use of the asset but cannot deduct interest or depreciation. The depreciation effect therefore encourages low-tax-rate firms to lease assets from high-tax-rate lessors. This occurs because the lessee effectively “sells” the depreciation (and associated tax deduction) to the lessor, who values it more highly (assuming that the lessee has a lower tax rate than the lessor). This incentive for low-tax-rate firms to lease is magnified when depreciation is accelerated, relative to straight-line depreciation. Furthermore, the alternative minimum tax (AMT) system can provide an additional incentive for a lessee to lease, in order to remove some depreciation from its books in an attempt to stay out of AMT status.

There are other tax effects that can reinforce or offset the incentive for low-tax-rate firms to lease. Lessors with relatively large tax rates receive a relatively large tax benefit of debt, which provides an additional incentive (to borrow to) buy an asset and lease it to the lessee. Moreover, tax incentives provided by investment tax credits (which have existed at various times) associated with asset purchases are also relatively beneficial to high-tax-rate lessors. In contrast, the relatively high taxes that the lessor must pay on lease income provide a tax disincentive for firms with high tax rates to be lessors (and similarly the relatively small tax benefit that a low-tax-rate firm obtains from deducting lease expense works against the incentive for low-tax-rate firms to lease rather than buy). The traditional argument is that low-tax-rate firms have a tax incentive to lease from high-tax-rate lessors, though this implication is only true for some combinations of tax rules (e.g. depreciation rules, range of corporate tax rates, existence of investment tax credits, or AMT) and leasing arrangements (e.g. structure of lease payments). See Smith and Wakeman (1985) and Eisfeldt and Rampini (2009) for details on how nontax effects can also influence the leasing decision.

Prediction 5: All else equal, the traditional argument is that low-tax-rate firms should lease assets from high-tax-rate lessors, though this implication is conditional on specifics of the tax code and leasing contract.

There are several complications associated with investigating whether firms lease in response to tax incentives. First, because leasing expense is tax deductible, leasing endogenously reduces a lessee’s effective tax rate, which can bias an experiment in favor of detecting tax effects. Similarly, lessor tax rates could be endogenously increased from the effects of lease income. Second, financial statement definitions of leasing are not one-to-one with IRS definitions, making it difficult to use Compustat data to test Prediction 5.³³ Using endogenously affected tax variables, Barclay et al. (1995b) and Sharpe and Nguyen (1995) find that low-tax-rate firms use relatively many capital leases. However, capital leases do not meet the IRS definition of true leases. (Instead, they are likely a mixture of true leases and conditional sales contracts, the latter of which are treated like debt, so that the lessee deducts interest and depreciation, and one might expect to find a positive relation between the conditional sales contract in capital leasing and taxes.) Therefore, the documented negative relation between capital leases and taxes is hard to interpret because it might be spurious.

Graham et al. (1998) address the first issue (endogeneity) by measuring tax incentives “but-for financing decisions”, that is, calculating tax rates using income before debt interest and the implicit interest portion of lease payments are deducted. They address the second issue by focusing on operating leases, which are defined in a manner similar to the IRS definition of true leases. Graham et al. (1998) find that the use of operating leases is negatively related to before-financing tax rates, consistent with Prediction 5, and that capital leases are unrelated to before-financing tax rates. Graham et al. also show that erroneously using an after-financing tax rate would double the magnitude of the negative tax coefficient for operating leases, and spuriously assign a negative tax coefficient to capital lease usage.

Eades and Marston (2001) find that lessors tend to be high-tax-rate firms (consistent with Prediction 5). Finally, O’Malley (1996) finds no evidence that firms systematically lease in response to tax incentives imposed by the AMT. We need research investigating whether the tax benefit of leasing adds to firm value. The jury is still out on whether debt and leasing are substitutes for the lessee (as they might be in the sense considered by DeAngelo and Masulis (1980), because both lead to tax deductions).

2.5.2 Pensions

Black (1980) argues that pension plans and the overall company are a single economic entity that should have an integrated financing and investment strategy. Due to interest tax deductions, the cost of corporate borrowing is the after-tax cost of debt. Because they are tax-free entities, defined benefit pension plans (DBs) earn the before-tax rate of interest on bondholdings. Therefore, Black suggests that DBs should increase (decrease) bond (equity) holdings, while the rest of the firm should do the reverse. This action should not increase firm risk because the increase in corporate debt offerings is offset by the increase in bonds held in the pension plan. (Interestingly, though, Shivdasani and Stefanascu (2010) find that US debt ratios are 35% larger when pension assets and liabilities are incorporated into measures of capital structure.) In an MM (1963) world, the net effect is that the company earns τ_C times the amount of bonds held, as in Eqn (4). Tepper (1981) argues that there can be a tax advantage to the strategy of corporate borrowing and DBs investing in bonds, even in a Miller (1977) world. In this case, the benefit occurs when the DB is an inframarginal investor in bonds, thereby earning the “extra” return necessary to compensate individual investors for the personal tax penalty associated with interest income (i.e. DBs capture some of the investor surplus depicted in Figure 2). The Tepper incentive for DBs to hold bonds increases with the difference between personal tax rates on interest and equity income.

Prediction 6: Defined benefit pension plans have an incentive to hold bonds (equity) that increases (decreases) in the corporate tax rate, while the rest of the firm has the reverse incentive.

Myers (2001) finds evidence consistent with the Black (1980) case: she reports that DB bondholdings increase with a simulated corporate marginal tax rate. She does not find evidence consistent with the Tepper argument. In a less direct test of the same incentives because it focuses on the size of pension contributions rather than pension capital structure, Thomas (1988) finds time-series evidence that firms decrease DB contributions when their tax rate is falling, and cross-sectional evidence that high-tax firms have larger DB funding levels.

Clinch and Shibano (1996) study pension reversions, which occur when a firm terminates an overfunded pension, settles its liabilities, and reverts the excess assets to the firm, all in one year. The reverted assets are taxable in the reversion year. Clinch and Shibano state that firms with the largest tax benefit of reverting do so, and also that firms’ time-reversion decisions occur in years with particularly large tax benefits. One nice aspect of the Clinch and Shibano experiment is that their tax variable equals the tax consequence of reverting relative to the tax consequence associated with the next best alternative (e.g. amortizing the excess assets over several years) rather than a “do nothing” alternative.³⁴

2.5.3 Debt Maturity

In the spirit of Modigliani and Miller (1958), Lewis (1990) derives an irrelevance null hypothesis for debt maturity. If corporate taxes are the only market imperfection, Lewis shows that the optimal firm-specific debt policy (i.e. optimal level of promised interest payments) can be achieved by various combinations of short- and long-term debt. This implies that firm value is unaffected by debt maturity structure and that capital market imperfections beyond corporate taxes, like costs to restructuring debt or underinvestment, are needed for debt maturity to matter. (Incomplete contracts can also lead to debt maturity mattering.)

Rather than modeling the simultaneous choice of debt level and maturity structure as in Lewis (1990), Brick and Ravid (1985) assume that firms choose debt level before debt maturity. If the expectations theory of interest rates holds, firms pay the same present value of interest in the long run regardless of debt maturity. However, issuing long-term debt accelerates interest payments, thus maximizing the present value of the interest tax shield. Brick and Ravid (1985) use this logic to argue that debt maturity should increase with the slope in the yield curve.

Prediction 7: Debt maturity increases in the slope in the yield curve.

Most empirical evidence does not support this prediction. Barclay and Smith (1995a) and Stohs and Mauer (1996) include a stand-alone yield curve variable that is either insignificant or has the wrong sign. Guedes and Opler (1996) maintain that the slope of the yield curve should only affect firms with a positive tax rate, and therefore the yield curve variable should interact with the corporate marginal tax rate. Neither Guedes and Opler (using a crude measure of the corporate tax rate) nor Harwood and Manzon (1998) (using a simulated corporate tax rate) find a significant coefficient on the yield curve variable. The one exception is Newberry and Novack (1999), who use a dummy variable equal to one during 1992 and 1993 (when the term premium was relatively high) and equal to zero for all other years 1987–1995. Newberry and Novack find a positive coefficient on the yield curve dummy in their public debt regression but not in their private debt analysis.

Kane, Marcus, and McDonald (1985) determine optimal debt maturity in a model that trades off corporate tax benefits with personal tax, bankruptcy, and flotation costs. The implications of their model are that debt maturity decreases with the corporate MTR and increases with the personal tax rate: long maturity implies less frequent recapitalization and relatively low transactions costs, so long-term debt can be desirable even if the net tax benefit is low. Maturity also decreases with the volatility of firm value because volatile firms are more likely to restructure debt.

Prediction 8: Debt maturity decreases with the corporate MTR and the volatility of firm value and increases with the personal tax rate.

Stohs and Mauer (1996) find the following support for Prediction 8: volatile firms generally use shorter term debt. The evidence relating to the tax-rate prediction is weaker. Stohs and Mauer report that debt maturity decreases with corporate tax rates—but their MTR variable is very crude (equal to income tax expense divided by pretax income when this ratio is between 0 and 1, and equal to 0 otherwise). Opler and Guedes (1996) find a negative coefficient on a tax expense divided by assets variable but the wrong sign on an NOL-based tax variable. Finally, Harwood and Manzon (1998) and Newberry and Novack (1999) point to a positive relation between a simulated tax rate variable and debt maturity, opposite the Kane et al. prediction.³⁵ A positive coefficient makes sense if large simulated MTRs identify firms that use long-term debt because they are relatively likely to be able to deduct interest in current and future periods.

Finally, debt maturity can affect the tax-timing option for firms to opportunely retire debt (e.g. Emery, Lewellen, and Mauer, 1988). If the corporate tax function is convex, the expected present value tax benefit of short-term debt declines with interest rate volatility, while the tax deductions with long-term debt are fixed. Therefore, long-term debt is preferred when interest rates are volatile. Long-term debt also increases the value of the timing option for investors to tax-trade securities (Kim, Maurer, and Stohs, 1995) because option value increases with security maturity and long-term bond prices are more sensitive to changes in interest rates.

Prediction 9: Debt maturity increases with interest rate volatility.

Kim et al. (1995) find that debt maturity increases with interest rate volatility, but Guedes and Opler (1996) do not. Nor do Guedes and Opler find significance for a second variable that interacts interest rate volatility with a corporate MTR variable.

In an analysis conducted for this chapter, I perform a more direct test on the hypothesis that uncertainty about future tax-paying status reduces the use of long-term debt. I use the standard deviation of the simulated marginal tax rate to measure uncertainty about tax-paying status, with the standard deviation calculated across the simulated scenarios for any given firm-year. I do not find any relation between debt maturity and uncertainty about tax-paying status.

The evidence linking tax incentives to debt maturity is mixed. One factor that makes it difficult to draw general conclusions is that debt maturity is defined differently in various papers. Barclay et al. (1995a) use a dependent variable measuring the portion of outstanding debt that matures in four or more years; Guedes and Opler (1996) use the log of the term to maturity for new debt issues; Stohs and Mauer (1996) use the book value weighted average of the maturity of a firm’s outstanding debt; Newberry and Novack (1999) use the same for new issues; and Harwood and Manzon (1998) use the portion of outstanding debt that is long term. Another issue that might affect inferences about tax variables is the apparently nonlinear relation between debt maturity and nontax influences (Guedes and Opler, 1996). Unless the nonlinearity of the overall specification is properly controlled, it might adversely affect the ability to detect tax effects. Finally, the yield curve was never inverted during the periods studied by most of these papers, so the tests of Brick and Ravid (1985) focus on the steepness of the yield curve rather than on the sign.