# The influence of firm maturation on tax-induced financing and investment decisions.

ABSTRACTEmpirical evidence that taxes influence firms' capital structures has been elusive. Scholes and Wolfson (1989) argue that refinancing costs that accumulate with age affect the time-series variation in firms' tax-induced financing and investment policies. Specifically, they predict that as capital structures gradually become more constrained over time, firms' financing decisions will become less sensitive to their marginal tax rates; and that as firms are increasingly impeded from adjusting their capital structures, they will resort to relying more on investment-related tax shields. This study examines panel data spanning firms' first nine public years that provides strong, robust evidence that the evolution in debt and asset tax shields is consistent with Scholes and Wolfson's (1989) predictions. In separate tests, age is measured from a firm's initial public offering and from its incorporation to consider whether the duration of their public and private existence, respectively, affects tax shield choice.

Keywords: taxes; capital structure; investment.

JEL Classification: C23, G31, G32, H25, L14.

INTRODUCTION

Most empirical capital structure studies do not find that taxes matter to firms' financing and investment policies (Graham 2001; Pittman and Klassen 2001). Scholes and Wolfson (1989) provide an explanation for this scarcity of evidence that the tax deductibility of interest persuades firms to increase their leverage. They argue that extant research obscures the influence of taxes by neglecting to consider refinancing costs that impede firms from reacting to changes in their tax status.

Scholes and Wolfson (1989) suggest that these refinancing costs gradually accumulate with age, which in turn affects the time-series variation in firms' tax-induced financing and investment decisions. (1) Specifically, they predict that as firms' capital structures become more constrained over time, their financing decisions will become less sensitive to their tax rates. In addition, they predict that firms will increasingly shift toward relying on investment-related tax shields as issuing and retiring securities becomes more difficult. This research tests these hypotheses with panel data that span firms' first nine public years. Strong, robust evidence on the evolution in firms' debt and asset tax shields is consistent with Scholes and Wolfson's (1989) predictions.

This study contributes to the literature by developing a research design that isolates the impact of changing adjustment costs on firms' tax-induced financing and investment policies. This paper also provides the first evidence on the time-series properties of firms' reactions to stable tax shields. Prior capital structure research examines either cross-sectional variation, or time-series variation related to certain macroeconomic conditions. (2) Finally, this study answers Slemrod and Shobe's (1990) call for empirical research that models the time-series behavior of responses to tax incentives.

This paper continues as follows. Section two further develops the motivation for the hypotheses tests, while Section three describes the research design and sample selection. Section four reports results for tests that measure a firm's age as the number of years that have passed since its initial public offering. Section five provides the sensitivity analysis. Section six examines whether firms' private ages at their IPO dates seriously affect the main results. Finally, the conclusions in Section seven discuss implications for future research.

HYPOTHESES DEVELOPMENT

Analytical research, such as Modigliani and Miller (1963) and DeAngelo and Masulis (1980), contributes testable hypotheses that specify relations among capital structure, tax rates, and asset tax shields. However, until recently evidence supporting these predictions has been elusive. Myers (1984, 588) comments on this situation in his presidential address to the American Finance Association:

I know of no study clearly demonstrating that a firm's tax status has predictable, material effects on its debt policy. I think the wait for such a study will be protracted.

Although MacKie-Mason (1990), Givoly et al. (1992), Graham (1996a), and Graham et al. (1998) find that taxes affect financing decisions, Parrino and Weisbach (1999) and Graham (2000) report that firms are substantially underlevered, given the magnitude of debt tax shields.

The tax-planning perspective adopted in Scholes and Wolfson (1992) stresses that managers should evaluate all costs and benefits of a proposed transaction, not just its tax aspects. They suggest that a firm would reject a strategy that reduces the present value of the firm's tax payments when the nontax costs that would be incurred exceed the potential tax savings. For example, Scholes and Wolfson (1989) argue that managers determining their firms' capital structures will increasingly encounter tension between taxes and refinancing costs that accumulate with age.

Pittman and Klassen (2001) review prior research that, on balance, implies that refinancing costs increase with age as firms' capital structures evolve over their early years. These changes include that firms, over time, replace private debt with public debt, which is more difficult to renegotiate; reduce their financing with securities that have flexible repayment provisions; and increase their number of debt issues, which can contain covenants that penalize or prohibit additional borrowing. Pittman and Klassen (2001) then examine Scholes and Wolfson's (1989) prediction about refinancing costs by estimating the time-series pattern in firms' rate of adjustment to their optimal capital structures over their first nine public years. Evidence from these tests, which measure the fraction of the distance between actual and target leverage that firms move in each year, suggests that the speed of adjustment gradually slows over time.

This paper extends Pittman and Klassen (2001) by testing Scholes and Wolfson's (1989) argument that refinancing costs that increase with age affect the time-series variation in firms' tax-shield choices. They predict that as capital structures gradually become more constrained over time, firms' financing decisions will become less sensitive to their tax status, which motivates the first hypothesis (stated in alternate form):

[H.sub.1]: The association between financial leverage and marginal tax rates will become less positive as firms age.

Scholes and Wolfson (1989) also predict that firms will shift more toward investment-related tax shields as refinancing costs begin to rise over time. This motivates the second hypothesis:

[H.sub.2]: The association between financial leverage and investment-related tax shields will become more negative as firms age.

The hypotheses assume that firm age is an adequate proxy for gradual changes in refinancing costs over time. This assumption is consistent with evidence provided in Pittman and Klassen (2001). Further, tests in Section five of the paper examine the time-series variation in some more direct proxies for refinancing costs to lend additional evidence on this fundamental assumption. To the extent that firm age is not an adequate proxy, or there are identification problems with this variable, the results should be interpreted in that light.

RESEARCH DESIGN AND SAMPLE SELECTION

This section develops the research design applied to estimate the time-series variation in firms' reliance on debt and investment tax shields. The empirical tests of [H.sub.1] and [H.sub.2] provide evidence on Scholes and Wolfson's (1989) predictions about the influence of refinancing costs on the evolution in tax-induced financing and investment decision's. The description of the dependent and explanatory variables emphasizes the specification of firms' leverage, marginal tax rates, and investment tax shields. The section concludes by summarizing the sample selection procedure and reporting descriptive statistics for the regression variables.

Regression Model

The predictions about firms' debt and investment tax shields and those motivated by other theories of capital structure are initially examined with the following model:

[LEVERAGE.sub.it] = [alpha] + [beta][X.sub.it] + [[epsilon].sub.it] (1)

where [alpha] is the intercept that is common to all firms,[X.sub.it], is a vector of explanatory variables, and [[epsilon].sub.it] is the error term. The testing begins with pooled OLS estimation to report results that are comparable to existing cross-sectional research on capital structure. However, the panel data tests that follow provide the main evidence on the hypotheses. The regression variables are specified as:

[LEVERAGE.sub.it] = Leverage is the book value of total long-term debt, convertible debt, and short-term debt deflated by one-year-lagged firm market value (the sum of market value of equity and book value of total debt).

[MTR.sub.i,t-1] = The lagged marginal tax rate is a trichotomous variable, which is equal to: (1) the top statutory corporate tax rate if taxable income is positive and the tax loss carryforwards (TLCF) balance is nonpositive; (2) one-half the top statutory rate if either taxable income is positive or TLCF is nonpositive; and (3) 0 otherwise.

[D.sub.it] Tax-exhausted firm dummy variable that indicates firms that have been assigned a 0 marginal tax rate according to the trichotomous variable conditions.

[SECURITY.sub.it] = The debt security proxy is the sum of depreciation and investment tax credits grossed-up by the statutory corporate tax rate. This amount is scaled by firm market value.

[IRTS.sub.it] = The investment-related tax shield variable is the product of the debt security proxy and the tax-exhausted firm indicator variable.

[Z-SCORE.sub.it] = Altman's z-score is book total assets divided by the sum of 1.2 times working capital, 1.4 times retained earnings, 3.3 times earnings before interest and taxes, and net sales.

[INDUSTRY-MEAN.sub.it] = Industry-mean leverage is the contemporaneous industry mean of the dependent variable, where industry-mean is the mean among other firms in the narrowest SIC code that includes at least five firms other than the firm under study.

[GROWTH.sub.it] = Growth options is the market value of common equity divided by its book value.

[CONVERTIBLES.sub.it] = Convertible securities is the sum of the book values of convertible debt and preferred shares; this amount is deflated by one-year-lagged firm market value.

[EARNINGS.sub.it] = Operating earnings is earnings before depreciation, interest, and taxes deflated by one-year-lagged firm market value.

[SIZE.sub.it] = Firm size is the natural logarithm of net sales.

[NEG.EQUTTY.sub.it] = The negative book equity dummy indicates if the book common equity is negative.

[ASSETS.sub.it] = Asset structure is property, plant, and equipment scaled by one-year-lagged total assets.

[YEAR.sub.t] = Year indicates the calendar year of the observation.

[AGE.sub.it] = Firm age is the number of years that have elapsed since the firm went public.

AGE*[MTR.sub.i,t-1] = This interaction is the product of the lagged marginal tax rate and firm age.

AGE*[LRTS.sub.it] = This interaction is the product of the investment-related tax shield and firm age.

Leverage

Leverage is defined as total book long-term, convertible, and short-term debt scaled by firm market value for several reasons. Stein (1992) and Mayers (1998) argue that convertible debt provides efficient financing when informational asymmetries are severe such as for the young, small firms that dominate this sample. In fact, Helwege and Liang (1996) and others find that firms frequently issue convertible bonds in their early public years. Essig (1991) and Titman and Wessels (1988) report that small firms rely more on convertible securities and short-term debt, respectively.

The predictions are tested on debt levels. MacKie-Mason (1990) and Shevlin (1999a) argue that examining financing changes is preferable to studying debt ratios, which represent the accumulation of past decisions that may obscure the effect that taxes have on firms' capital structures. This implies that tests using marginal tax rate proxies to explain firms' debt levels will suffer from low explanatory power. This study directly evaluates this issue by observing firms over time starting with their first post-IPO year, which ensures that they initially have no public financial histories. Scholes and Wolfson's (1989) prediction that refinancing costs increasingly constrain leverage with age suggests that firms will have both contemporaneous and lagged responses to tax incentives; the differenced data would identify only the contemporaneous response.

Marginal Tax Rates

There is no generally accepted proxy for a firm's marginal tax rate, defined as the present value of current and expected future taxes on an extra dollar of income earned today. Each study that detects the predicted positive relation between leverage and tax status specifies a different MTR measure (MacKie-Mason 1990; Givoly et al. 1992; Graham 1996a; Graham et al. 1998). Graham (1996b) evaluates how well ten of the most widely used MTR proxies predict perfect foresight marginal tax rates. He finds that his simulation algorithm produces the most accurate MTR measure currently available. Graham also concludes that those used by MacKie-Mason (1990) and Givoly et al. (1992) are mediocre proxies for firms' marginal tax rates.

Graham et al. (1998) report that existing tax status proxies are endogenous to financing decisions, which induces a spurious negative relation when leverage is regressed on contemporaneous MTR. As interest payments are deductible, a firm that finances with debt reduces its taxable income, which in turn may lower its marginal tax rate. Accordingly, they modify Graham's (1996a) algorithm to obtain marginal tax rates (derived from prefinancing taxable income) that are not affected by leverage decisions. Graham et al. (1998) rely on their before-financing simulated MTRs to provide the first evidence of a positive relation between debt levels and tax status. (3)

However, using Graham et al.'s (1998) MTRs in this study would prevent examining firms' until their fourth public year since the simulations require at least three years of historical data. Instead, a trichotomous variable that Shevlin (1990) advocates is used to measure MTR as:

1) the top statutory corporate tax rate if taxable income is positive and the amount of tax loss carryforwards (TLCF) is nonpositive;

2) one-half the top tax rate if either taxable income is positive or TLCF is nonpositive; and

3) zero if taxable income is nonpositive and TLCF is positive.

This trichotomous variable was the second-most accurate proxy in Graham's (1996b) evaluation of extant MTR proxies. The trichotomous measure is lagged one year since this explanatory variable should be the MTR prevailing at the time corporate decisions are made (Scholes et al. 1990). (4)

Investment-Related Tax Shields

DeAngelo and Masulis (1980) predict an inverse cross-sectional relation between leverage and investment-related tax shields. Until recent studies by MacKie-Mason (1990), Dhaliwal et al. (1992), and Trezevant (1992), empirical support for the tax substitution effect was scarce. Early research neglects to control for two important influences, the debt securability effect and the tax exhaustion hypothesis. Accounting depreciation, depletion, and investment tax credits are the tax-shield substitutes for debt in DeAngelo and Masulis's (1980) model. However, these tax deductions and credits are generated by capital assets that serve as loan collateral. Scott (1976) and others suggest that firms with substantial collateral may borrow at lower rates and reduce credit rationing. This debt securability effect may confound tests of the tax-shield substitution effect.

MacKie-Mason (1990) finds that the substitution effect applies to firms with a substantial probability of tax shield loss. A firm in this position is described as being tax exhausted, i.e., its tax shields exceed its taxable income such that a portion of its available deductions and credits are carried forward. The present value of tax shields is reduced when they can only be used in future years, if ever. Therefore, the sensitivity of the predicted negative relation between debt and investment tax shields will depend on firms' proximity to tax exhaustion. Firms nearing tax exhaustion will substitute because their income is not sufficient to cover all of their tax shields.

The sample is partitioned according to the firms' probabilities of losing the deductibility of tax shields to examine whether the substitution effect depends upon on firms' marginal tax rates. This research design integrates both tax-exhausted and nontax-exhausted firms in a single regression equation that slightly modifies the specifications in Dhaliwal et al. (1992) and Trezevant (1992). The following portion of the main regression model, Equation (1), addresses both the debt securability effect and the tax exhaustion hypothesis:

[LEVERAGE.sub.it] = [[beta].sub.2][[D.sub.it] + [[beta].sub.3][SECURITY.sub.it] + [[beta].sub.4][IRTS.sub.it] (2)

[[beta].sub.3] is predicted to be positive representing the debt securability effect, which is assumed to be constant across firms irrespective of their tax status. This control for debt securability enables the tax-shield substitution effect to be isolated in [[beta].sub.4], which is predicted to be negative. (5) Dhaliwal et al. (1992) explain that it is important to empirically validate the assumption that the debt securability effect is common to all firms. For the sample firms, tests indicate that there is no statistical difference in [[beta].sub.3] for firms partitioned based on their tax status. Tax-exhausted firms are defined as those assigned a zero marginal tax rate according to the trichotomous variable conditions. Section five explores the sensitivity of the results to this definition.

Other Determinants of Capital Structure

The explanatory variables summarized in Table 1 are included in the regressions to control for other determinants of capital structure. The choice and specification of these variables follows prior research, e.g., Mackie-Mason (1990), Rajan and Zingales (1995), and Graham et al. (1998).

Sample Selection

Table 2 summarizes the data screening imposed to obtain the longitudinal sample of 189 firms from the listing of the 3,426 SEC-registered initial public offerings occurring between 1978 and 1988 acquired from Security Data Corporation. This is the maximal period that provides nine consecutive years of data to examine the time-series pattern in firms' reliance on debt and asset tax shields. The selection procedure follows prior research; e.g., Mackie-Mason (1990) and Graham et al. (1998).

This study restricts the sample to firms that survive through their ninth public year to ensure that changes in sample composition over time do not spuriously induce the evidence reported. Further, letting firms enter and leave the sample during the nine years could admit confounding from events that caused these firms to be removed from Compustat. This paper is more concerned with the typical evolution in capital structures than the major discontinuous shifts that firms might experience around their merger, acquisition, bankruptcy, or liquidation (Shyam-Sunder and Myers 1999). However, requiring nine consecutive years of data contributes to the attrition in the sample, which contains only 6 percent of the original population. The econometric implications of potential survivorship bias are considered later. Finally, inspection of the data suggests that the calendar year (Table 2) and industry (Table 3) clustering in the sample resembles the clustering in the population.

Descriptive Statistics

The descriptive statistics reported in Table 4 indicate that there is substantial variation across all variables. The generally low cross-correlations provide support that other determinants of capital structure do not explain the observed tax effects. (6)

EMPIRICAL RESULTS

Cross-Sectional Estimation

Equation (1) is initially estimated using ordinary least squares under the assumption that the error term is independent of [X.sub.it] to generate results that are comparable to extant cross-sectional research. In this regression, the time-series variation in tax-induced financing and investment decisions is examined by interacting firm age with the proxies for debt and investment tax shields.

Table 5 reports in Column (1) the results from the pooled OLS regression for the balanced panel, using standard errors obtained with White's (1980) heteroscedasticity-consistent covariance matrix. The evidence is consistent with both hypotheses. (7) The prediction in [H.sub.1] that the influence firms' marginal tax rates have on their financing decisions diminishes with age is supported because:

1) the coefficient on the trichotomous MTR proxy is positive and statistically significant; and

2) the coefficient on the age-MTR interaction is negative and statistically significant.

The positive relation between firms' leverage and their tax status in the pooled tests implies that lagging the marginal tax rate proxy was sufficient to avoid the endogeneity bias that can arise in debt levels tests. The negative and statistically significant coefficient on the age-investment tax-shield interaction supports [H.sub.2] by providing evidence that tax-shield substitution increases over time.

Most control variables are statistically significant in the predicted directions. An important exception is the estimate for the investment tax shield, which is not negatively correlated with leverage, an issue that is reexamined in Section six. The only other results inconsistent with theory are the coefficients on Altman's (1968) z-score and firm size, which are both not significant.

The results reported in Column (1) of Table 5 may be artifacts of omitted variable bias since, for example, the industry-mean debt ratio variable might be a proxy for unobservable firm-specific effects. The fixed effects model discussed in the next section is employed to examine whether the cross-sectional evidence is attributable to mainly between-firm effects or within-firm dynamics.

Fixed Effects Estimation

Suppose that the unobservable error term in Equation (1), [[epsilon].sub.it] is more precisely described as:

[[epsilon].sub.it] = [[mu].sub.i] + [e.sub.it] (3)

where [[mu].sub.i] is a firm-specific component and the residual term, [e.sub.it], is a measurement error or other form of stochastic shock. This would imply that the treatment of the firm-specific effect will depend on the assumptions made about the relationships among [[mu].sub.i], [e.sub.it], and [X.sub.it]. The simplest assumption is that [[mu].sub.i], [e.sub.it], and [X.sub.it] are mutually orthogonal, which would enable Equation (1) to be estimated as a random effects model containing firm-specific heteroscedasticity.

However, it is more plausible to assume that [[mu].sub.i] absorbs differences in, for example, the quality of management, which would suggest that firms that survive are probably higher quality. This would imply that a positive correlation exists between [[mu].sub.i] and [X.sub.it], although [[mu].sub.i] could remain orthogonal to [e.sub.it]. The concern that firm-specific correlated omitted variables are present in the levels data can be remedied by specifying an intercept for each firm, [[alpha].sub.i]. This eliminates the time-invariant [[mu].sub.i] such that consistent coefficient estimates are obtained:

[LEVERAGE.sub.it] = [[alpha].sub.i] + [beta][X.sub.it] + [e.sub.it] (4)

The time-series patterns observed in the debt and investment tax shields for this fixed effects model, which preserves the time-series variation in leverage while accounting for individual firm heterogeneity, are similar to the cross-sectional evidence provided earlier. (8) Colunm (3) of Table 5 reports results from the fixed effects test that strongly support both [H.sub.1] and [H.sub.2]. (9)

An F-test indicates that the null hypothesis that the fixed firm effects are all equal is strongly rejected. In the presence of the firm effects, another F-test supports including a variable for the calendar year of each observation. (10) Accordingly, the balanced panel results presented in Column (3) of Table 5 are for a two-way fixed effects model with correction for unspecified heteroscedasticity.

Despite a Hausman (1978) test suggesting that the fixed effects model is the proper design choice, external validity is impaired because the coefficient estimates apply only to the firms in the sample. (11) This is potentially important since the sample contains at most 6 percent of the population of IPOs conducted between 1978 and 1988 (see Table 2). Although the selection procedure was certainly not random (e.g., firms were only included if they were immediately followed by Corn pustat), this study is interested in providing evidence to justify inferences about the entire population.

Similar balanced panel results (not tabulated) consistent with [H.sub.1], and [H.sub.2] are obtained when Equation (4) is reestimated with a random effects model. This is an important statistical matter since Hsiao (1986) explains that observing consistent estimates across alternative panel data estimation techniques indicates the absence of serious errors in variables problems. In addition, this fixed and random effects evidence for both H, and H2 is robust to the unbalanced sample.

SENSITIVITY AND OTHER TESTS

Time-Series Variation in Proxies for Refinancing Costs

Extensive prior research suggests that refinancing costs affect firms' capital structures; e.g., Gilson (1997), Shyam-Sunder and Myers (1999), and Pittman and Klassen (2001). However, except for Gilson's (1997) evidence on financially distressed firms, these studies do not specify proxies for refinancing costs. Instead, they assume that these costs are responsible for any delay observed in the adjustment of actual leverage to target leverage.

Dhaliwal and Graham (2001) and Pittman and Klassen (2001) identify limitations with relying on such target adjustment models to investigate whether firms' capital structures become more constrained over time. Shevlin (1999a) calls for empirical research designs that incorporate proxies for nontax costs in settings in which these are expected to impede firms from responding to tax incentives. The following tests examine the time-series pattern in five proxies for refinancing costs to provide additional evidence on the prediction that these costs gradually increase with age:

1) An individual creditor's motive to refuse to participate in a debt restructuring plan is stronger the smaller their claim since the firm's future financial condition is less likely to depend on whether they grant concessions (Grossman and Hart 1980; Gertner and Scharfstein 1991). The average size of private lenders' claims is specified by dividing the book value of total private long-term debt by the number of private long-term debt contracts (Gilson 1997).

2) Similarly, this creditor holdout problem should be more serious for firms that have more publicly traded debt, which tends to be more widely held. There is also extensive theory and evidence that private lenders are more apt to renegotiate debt contracts and to selectively relax loan covenants. Pittman and Klassen (2001) review this literature along with research that suggests that public debt issues have longer maturities and contain legal restrictions that prevent renegotiation without the consent of each bondholder. These refinancing costs will be higher the higher the fraction of public bonds in firms' total long-term debt. (12)

3) Scholes and Wolfson (1992) explain that, although more costly than long-term debt, short-term debt affords firms the flexibility to frequently change their capital structures. The ratio of short-term debt to total debt measures this refinancing cost proxy.

4) Asquith et al. (1994) and others provide evidence consistent with Bolton and Scharfstein's (1996) theory that debt renegotiations become more complicated the larger the number of creditors. This refinancing cost is estimated as the number of long-term debt contracts outstanding, with each issue of public bonds counted as a single contract (Gilson 1997).

5) Mayers (1998) argues that young firms finance with securities, such as callable public bonds, that have flexible repayment provisions to lower costs on subsequent issues. This proxy is the fraction of public bond issues that are callable.

All data for these proxies was collected from Moody's industrial Manual and Transportation Manual. (13) This produced a sample of 96 firms having the required data for their first nine years after going public. However, only 17 of these firms are also present in the sample described in Table 2, which prevents including these more direct proxies for refinancing costs in the main hypotheses tests. This sample is split into firms' first three public years, second three public years, and third three public years to explore the evolution in firms' capital structure characteristics.

The evidence in Table 6 generally indicates that firms change in ways that are consistent with refinancing costs gradually rising over their early years. Specifically, results from Wilcoxon matched-pairs signed-rank tests suggest that over time the fraction of long-term debt in public bonds increases; the fraction of total debt that is short-term decreases; and the fraction of public bonds that are callable decreases. (14) The predicted increase in the number of long-term debt contracts is only found between firms' first and second three public years. However, there is no evidence to support the prediction that the average size of private long-term debt contracts decrease with age. In fact, the results suggest that average contract size increases between firms' second and third three public years.

Overall, similar to Pittman and Klassen's (2001) target adjustment model estimates, the evidence in this section implies that refinancing costs increase over time. Together, these results justify using firm age to proxy for cumulative refinancing costs in the tests of [H.sub.1] and [H.sub.2].

Selection Correction

The survivorship requirement imposed to compile the balanced panel almost certainly results in removing firms that became financially distressed during their first nine public years. These firms probably would be more concerned with financing matters other than taxes relative to the firms in the sample. (15) The primary tests examine a balanced panel to ensure that changes in sample composition over time do not drive the results. Still, the considerable attrition that occurred in the selection process could lead to biased estimates when standard statistical methods are applied to this sample.

Therefore, the Heckman (1979) two-step estimation procedure is used to generate consistent estimates in the presence of attrition. The first step estimates the parameters of a sample composition equation with the dependent variable having the value of 1 if the observation stays in the sample (survival), and 0 otherwise (attrition). The independent variables are the financial distress proxies, Altman's (1968) z-score and the negative book equity indicator, which are intended to explain attrition. (16) This equation is estimated for the full sample (9,322 firm-year observations) with maximum likelihood probit. The estimates are used to determine the conditional expectation of the error term of the regression model given the inclusion of an observation in the sample.

The second step involves OLS estimation of the regression model specified in the top panel of Table 5 with this conditional expectation entered as an additional explanatory variable. This ensures that consistent estimates are obtained for the balanced panel containing 1,701 firm-year observations. Column (5) of Table 5 reports that the evidence consistent with [H.sub.1] and [H.sub.2] is robust to this correction for potential sample selection bias.

Specification of Marginal Tax Rate

As the trichotomous MTR variable relies partially on Compustat's TLCF amounts, the relative inaccuracy of the TLCF data field reduces its validity (see Kinney and Swanson 1993). In the primary tests, firms with missing TLCF observations are discarded, while observations with an error in the reported amount are kept since they cannot be identified without examining firms' financial statements. An alternative approach is to replace missing TLCF observations with zeros under the assumption that firms that do not report these amounts do not have any tax loss carryforwards (e.g., Graham 1996b; Klassen 1997). In both the balanced and unbalanced panels, the evidence supporting [H.sub.1] and [H.sub.2] remains for both the fixed and random effects models with this re-specification.

Although the trichotomous variable performs well in Graham's (1996b) and Plesko's (1999) evaluations of conventional tax rate proxies, there are serious limitations with using it in this study. (17) Shevlin (1990) explains that the trichotomous MTR proxy might be particularly imprecise for young firms that are apt to have erratic earnings. Further, Teoh et al. (1998) and others find that firms practice earnings management surrounding their IPO dates, which may worsen the misclassification of MTRs using the trichotomous conditions.

These problems should not as adversely affect Graham et al.'s (1998) before-financing simulated marginal tax rates that are specifically designed to examine debt levels. However, replacing the trichotomous variable with these tax rates would prevent observing firms until their fourth public year since their simulation algorithm requires at least three years of historical data. (18) Nonetheless, this research design may be justified since the financing decisions made in the three years following an IPO may relate more to the transition to becoming a public company than to continuing capital structure choices (Pagano et al. 1998). In both the balanced and unbalanced samples, the fixed and random effects models provide evidence of the negative time trends predicted in [H.sub.1] and [H.sub.2] when using Graham et al.'s (1998) simulations to measure tax status. (19)

Specification of Investment-Related Tax Shield

In the primary tests, the firm's investment tax shield was estimated as the sum of its depreciation and grossed-up investment tax credits, scaled by firm market value. However, there may be substantial discrepancies between the depreciation and ITC amounts reported in the financial statements and those reported on tax returns. (20) Generally, accelerated depreciation is claimed for tax purposes and straight-line depreciation is expensed in the financial statements. ITCs can be claimed immediately for tax purposes, but might be deferred and amortized for accounting purposes.

Accordingly, the investment tax shield was also measured as the sum of the depreciation expense and ITCs reported in the firms' financial statements, plus (minus) the grossed-up increase (decrease) in deferred taxes. This assumes that all book-tax timing differences between accounting income and taxable income arise from differences in the calculation of depreciation and ITCs. The empirical results with this re-specification are very similar to the evidence in Section four.

The results are also robust to respecifying the investment tax shield to include rental expense, which could substitute for depreciation expense for firms that prefer to lease rather than buy capital assets (Smith and Wakeman 1985). Further, the results are virtually identical when the investment tax shield excludes ITCs. Graham and Smith's (1999) simulations indicate that ITCs hardly affect the convexity of the effective corporate tax function, which they interpret as contradicting DeAngelo and Masulis's (1980) argument that the lower present value of unused 1TCs will induce convexity. (21)

MacKie-Mason (1990) argues that only firms that are tax-exhausted will substitute tax shields. In the primary tests, tax-exhausted firms are identified as those that have been assigned a zero marginal tax rate according to the trichotomous variable conditions. The sensitivity of the results to the identification of tax-exhausted finns is evaluated by replacing the trichotomous proxy with Graham et al.'s (1998) simulated MTRs. Negative and statistically significant coefficients supporting H2 were found when the tax-exhausted firms were defined as those in the bottom 10 percent, 25 percent, or 50 percent of the simulated MTh distribution for firms of the same age. (22)

Cloyd et al. (1997) and Ayers et al. (2001) report evidence consistent with their argument that only finns with moderate marginal tax rates, which they describe as being "tax-sensitive," have sufficient incentive to substitute investment tax shields for debt tax shields.23 Their research designs differ from other studies which specify that only tax-exhausted firms will substitute their tax shields; e.g., MacKie-Mason (1990), Dhaliwal et al. (1992), and Trezevant (1992). Further, it is difficult to compare the tests on public firms in this study to Cloyd et al. (1997) and Ayers et al.'s (2001) examination of the substitution decisions of private firms, especially since they use a different dependent variable than most empirical research on capital structure (Graham 2001). Nonetheless, identifying firms with moderate marginal tax rates as those that will substitute provides evidence consistent with H2 when using Graham et al.'s (1998) simulations, but not when using the trichotomous variable to measure tax status.

Time-Series Variation in Covariance StructureThe Pearson correlation of these nine [MTR.sub.t-1] coefficients with firm age is -0.82, which is significant at the 1 percent level in a one-tailed test assuming independence. However, since this correlation ignores the precision of the coefficient estimates, a weighted linear regression is run with the weights determined by the standard errors of the coefficients; i.e., the least weight is assigned to the observations that are measured with the most error. The weighted least squares regression of the [MTR.sub.t-1] estimates from the nine cross-sectional random coefficients models on firm age also provides evidence at the 1 percent level of a negative pattern. Similarly, using this procedure to examine the time-series variation in investment tax-shield substitution produces evidence consistent with [H.sub.2].

It is plausible that the magnitude of the marginal tax rate coefficient may not become less positive with age as [H.sub.1] predicts, but rather may become statistically less significant as the covariance of leverage and tax rates increases with age. I thank an anonymous reviewer for pointing out this issue. Lang (1991) and Pittman and Klassen (2001) apply the following methodology to more explicifly consider the potential impact of changes to the covariance structure over time. Equation (1) is reestimated in separate cross-sectional regressions for each of firms' first nine public years with a random coefficients model. This estimation technique, which has been modified to correct for heteroscedasticity, allows for cross-sectional variation in the [MTR.sup.t-1] coefficients, each representing an average of individual firm coefficients. The results from OLS estimation are virtually identical.

The Pearson correlation of these nine [MTR.sub.t-1] coefficients with firm age is -0.82, which is significant at the 1 percent level in a one-tailed test assuming independence. However, since this correlation ignores the precision of the coefficient estimates, a weighted linear regression is run with the weights determined by the standard errors of the coefficients; i.e., the least weight is assigned to the observations that are measured with the most error. The weighted least squares regression of the [MTR.sub.t-1] estimates from the nine cross-sectional random coefficients models on firm age also provides evidence at the 1 percent level of a negative pattern. Similarly, using this procedure to examine the time-series variation in investment tax-shield substitution produces evidence consistent with [H.sub.2].

Other Tests

The evidence that the time-series variation in firms' financing and investment decisions is consistent with [H.sub.1] and [H.sub.2] persists when the numerator of the dependent variable is respecified from the sum of convertible, long-term and short-term debt to: (1) the sum of long-term and short-term debt; and (2) long-term debt only. The results also are not sensitive to scaling the dependent variable with lagged book value of equity or total assets, rather than lagged firm market Value. (24) Scaling by book equity or total assets is justified by the practice of many rating agencies and treasurers to rely on book leverage. Using book equity in the denominator may better isolate deliberate financing choices by avoiding impounding share price variances arising from extraneous conditions (Miller 1977).

Dhaliwal et al. (1992), Cloyd et al. (1997), and Ayers et al. (2001) argue that interest expense deflated by earnings is a more appropriate measure of financial policy when examining tax shields. In addition, DeAngelo and Masulis (1980) and Dhaliwal and Graham (2001) stress the importance of controlling for profitability in empirical research on leverage decisions. Although Scholes and Wolfson's (1989) predictions specifically concern capital and asset structures, these relations should behave similarly when using interest expense scaled by taxable income before interest, depreciation, and grossed-up investment tax shields as the dependent variable (and correspondingly deflating the explanatory variables). The evidence consistent with the hypotheses is robust to this re-specification.

Notwithstanding DeAngelo and Masulis's (1980) exogenous specification of the investment decision, it may be more realistic that the choice of tax shields is jointly determined (Trezevant 1994); Cloyd et al. 1997; Ayers et al. 2001). Estimating a simultaneous equations model by two-stage least squares provides evidence consistent with the hypotheses. Further, a series of tests indicate that the results for [H.sub.1] and [H.sub.2] are not time-period or industry-specific, or driven by influential observations or outliers dominating the data. This evidence on the evolution in firms' debt and asset tax shields also remains for various diminishing nonlinear transformations of firm age and after removing firms that changed their year-ends during the nine-year panel (to synchronize the tests at one-year intervals). (25) Similarly, extending the time-series from nine to 12 or 15 years provides evidence supporting [H.sub.1] and [H.sub.2]. Finally, the results are not attributable to censorship bias from treating lever age as a continuous variable (12.5 percent of the 1,701 firm-year observations have no debt in their capital structures), or to firms' risk or investment opportunity sets changing over time. (26)

PRIVATE OPERATING HISTORY

This section considers whether the duration of firms' private operating histories affects the time-series variation in their tax-induced financing and investment decisions. The preceding tests provide strong, robust evidence consistent with the prediction that firms gradually shift from debt to investment tax shields in their early public years. The following tests reexamine this evidence after bisecting the sample into the youngest and oldest firms at their IPOs. Scholes and Wolfson's (1989) argument implies that the financing decisions of the older firms will be relatively less responsive to their tax rates since they already have extensive histories at the time that they go public. (27) The median private age of this sample is 18 years vs. four years for the sample of younger firms.

As predicted, the pooled OLS results presented in Column (1) of Table 7 indicate that the capital structure decisions of the older firms are not affected by their marginal tax rates; i.e., the estimated coefficient on this variable is not statistically significant. These firms are also observed to have shifted toward relying on the investment tax shield according to its negative and significant coefficient. This evidence is consistent with the prediction that firms begin to resort to other tax strategies when refinancing costs begin to impede financing changes.

Results from two-way fixed effects tests that examine the within-firm dynamics predicted in the hypotheses are reported in Column (2) of Table 7. This evidence suggests that these older firms continue to gradually reduce their reliance on debt tax shields during their first nine public years; i.e., the coefficient on the firm age-marginal tax rate interaction is negative and strongly significant, which is consistent with [H.sub.1].

However, there is no empirical support for [H.sub.2], which implies that the evolution in tax-shield choice is relatively advanced for these older firms. It appears that the transition in firms' tax-induced financing and investment policies predicted by Scholes and Wolfson (1989) is largely complete for this sample. For example, the cross-sectional results in Column (1) of Table 7 provide evidence that the older firms rely only on investment-related tax shields; their marginal tax rates do not affect their capital structures. Accordingly, the stability in the time-series pattern of tax-shield substitution, which is observed through the firm age-investment tax-shield interaction in column (2), is expected.

Almost completely opposite evidence is found for the younger firms, which are defined as those in the lower half of the private age distribution as measured from their incorporation dates to their IPO dates. Again consistent with the prediction that younger firms have more flexibility to adjust their capital structures in reaction to changing tax incentives, a positive and statistically significant coefficient is estimated for the MTR variable in this sample (see Column (3) of Table 7).

Also, the coefficient on the investment-related tax shield proxy reported in Column (3) of Table 7 is positive and significant, which contradicts the negative relation modeled in DeAngelo and Masulis (1980). This cross-sectional evidence supports the argument that these younger firms have capital structures that are not seriously affected by accumulating refinancing costs, so there is less urgency for them to substitute tax shields. The fixed effects results for the younger firms in Column (4) of Table 7 provide support for both [H.sub.1] and H2, which suggests that they have not progressed as far as the older firms in their transition from relying on debt tax shields toward investment tax shields. (28)

CONCLUSIONS

This study reports evidence supporting Scholes and Wolfson's (1989) prediction that the positive relation between leverage and marginal tax rates will subside with age as firms gradually shift toward investment tax shields. These findings are consistent with refinancing costs increasingly impeding firms' from adjusting their capital structures. However, these tests, which provide the paper's main evidence, deliberately ignore firms' private ages when they go public. The impact of firms' private histories is examined by reestimating the regressions after the splitting the sample into younger and older firms as of their IPO dates. These results suggest that the younger firms are at an earlier stage in their development toward relying more on investment tax shields and less on debt tax shields.

This research could be extended by testing the predictions in other empirical settings such as firms emerging from bankruptcy. Gilson (1997) provides evidence that the relatively low transactions costs incurred during Chapter 11 protection give financially distressed firms more flexibility to choose optimal capital structures. This may be a cleaner setting than recently public firms because bankrupt firms can essentially select a brand new capital structure before leaving Chapter 11 (Alderson and Betker 1995). In contrast, evidence in this study suggests that post-IPO firms have capital structures that are partially inherited from their pre-IPO histories. Similar to firms leaving bankruptcy protection, Dittmar (forthcoming) argues that subsidiaries divested in spin-offs can also adopt a brand new capital structure, which suggests that future research could test Scholes and Wolfson's (1989) predictions using these firms' early years.

Another project could investigate whether young firms react more quickly to changes to tax laws affecting their financing incentives. For example, this research could reevaluate studies such as Givoly et al. (1992) for evidence that younger firms are more responsive to tax regime shifts. Finally, the research design in this paper specifies depreciation, depletion, and ITCs as the nondebt tax shield that firms shift toward when refinancing costs eventually prevent changes to their leverage. This choice was partly motivated by the extensive theory and cross-sectional evidence on investment tax-shield substitution. However, future research could examine whether firms adjust other nondebt tax shields when their capital structures become seriously constrained.

TABLE 1 MOTIVATION FOR CONTROL VARIABLES Explanatory Variable Theory Altman's z-score Static trade-off theory that firms balance the tax subsidy provided by debt against the costs of Negative book equity indicator financial distress (described in Myers 1984) Industry membership Titman (1984), Myers and Majluf (1984), and Jensen (1989) Growth options Myers (1984), Myers and Majluf (1984), and Jensen and Meckling (1976) Asset structure Convertible securities Myers and Majluf (1984) Operating earnings Ross (1977), Myers and Majluf (1984), and Jensen (1986) Firm size Macroeconomic conditions Firm age Diamond (1989) Explanatory Variable Evidence Altman's z-score Altman (1968), MacKie-Mason (1990), Graham (1996a), and Graham et al. (1998) Negative book equity indicator Industry membership Marsh (1982), Bradley et al. (1984), and Titman and Wessels (1988) Growth options MacKie-Mason (1990) and Givoly et al. (1992) Asset structure Convertible securities Stein (1992) and Helwege and Liang (1996) Operating earnings MacKie-Mason (1990), Graham (1996a), and Graham et al. (1998) Firm size Warner (1977) and Ang et al. (1982) Macroeconomic conditions Givoly et al. (1992), Trezevant (1992), Marsh (1982), and Taggart (1985) Firm age Peterson and Rajan (1994) and Cloyd et al. (1997) This table lists prior research that motivates including the control variables in the regression equations presented in Tables 5 and 7. The empirical specifications of these explanatory variables are provided in Table 4. TABLE 2 SAMPLE SELECTION SUMMARY Calendar Year 1978 1979 1980 1981 1982 1983 Number of SEC registered IPOs 38 62 149 348 122 685 Number of firms not followed by Compustat since initial public offering (20) (22) (67) (125) (40) (246) Number of firms from utilities, financial, insurance, and real estate industries (3) (4) (11) (26) (7) (51) Number of firms that did not survive through their first nine years of public operation (2) (10) (27) (83) (36) (178) Number of firms with missing Compustat observations (10) (22) (35) (95) (35) (163) Number of firms in the sample 3 4 9 19 4 47 Calendar Year 1984 1985 1986 1987 1988 Number of SEC registered IPOs 357 355 728 415 167 Number of firms not followed by Compustat since initial public offering (129) (131) (290) (150) (47) Number of firms from utilities, financial, insurance, and real estate industries (24) (29) (53) (27) (12) Number of firms that did not survive through their first nine years of public operation (111) (96) (166) (100) (50) Number of firms with missing Compustat observations (80) (81) (181) (112) (50) Number of firms in the sample 13 18 38 26 8 Calendar Year Total Number of SEC registered IPOs 3,426 Number of firms not followed by Compustat since initial public offering (1,267) Number of firms from utilities, financial, insurance, and real estate industries (247) Number of firms that did not survive through their first nine years of public operation (859) Number of firms with missing Compustat observations (864) Number of firms in the sample 189 TABLE 3 INDUSTRY DISTRIBUTION OF SAMPLE Two-Digit SIC Code Number Industry Description Percentage 20 6 Food 3.2 27 7 Printing and publishing 3.7 28 8 Chemicals 4.2 35 29 Industrial and commercial machinery 15.3 36 24 Electronic and electrical equipment 12.7 38 8 Measuring, analyzing, and 4.2 controlling equipment 39 5 Miscellaneous manufacturing 2.6 industries 42 6 Freight transportation and 3.2 warehousing 50 7 Wholesale trade--durable goods 3.7 56 8 Apparel and accessory stores 4.2 58 5 Restaurants 2.6 59 5 Miscellaneous retail 2.6 73 16 Business services 8.5 87 6 Engineering, accounting, research, 3.2 management, and related services Subtotal 140 74.1 49 Industries with fewer than 5 firms 25.9 Total 189 100.0 TABLE 4 SUMMARY STATISTICS Panel A: Correlations Marginal Debt Altman's Variable Leverage Tax Rate Security IRTS z-score Leverage -0.158 0.353 0.177 0.358 (0.001) (0.001) (0.001) (0.001) Marginal Tax -0.204 -0.225 -0.427 -0.114 Rate (0.001) (0.001) (0.001) (0.001) Debt Security 0.130 -0.115 0.235 0.149 (0.001) (0.001) (0.001) (0.001) Investment 0.200 -0.361 0.143 0.083 Tax Shield (0.001) (0.001) (0.001) (0.001) Altman's 0.015 0.022 0.009 -0.005 z-score (0.537) (0.361) (0.699) (0.849) Industry-Mean 0.247 0.028 0.109 0.056 0.031 Debt (0.001) (0.241) (0.001) (0.020) (0.196) Growth Options -0.186 0.013 -0.074 -0.046 -0.033 (0.001) (0.603) (0.002) (0.057) (0.175) Convertible 0.222 -0.106 0.024 0.157 0.004 Securities (0.001) (0.001) (0.319) (0.001) (0.870) Operating -0.080 0.253 0.034 -0.327 0.019 Earnings (0.001) (0.001) (0.162) (0.001) (0.445) Firm Size 0.070 -0.076 0.191 0.137 0.013 (0.004) (0.001) (0.001) (0.001) (0.585) Negative Book 0.180 -0.109 0.012 0.134 0.003 Equity (0.001) (0.001) (0.613) (0.001) (0.913) Asset Structure 0.194 -0.120 0.275 0.268 0.027 (0.001) (0.001) (0.001) (0.001) (0.270) Firm Age 0.093 -0.269 0.059 0.122 -0.016 (0.001) (0.001) (0.015) (0.001) (0.505) Tax-Exhausted 0.211 -0.515 0.062 0.669 -0.061 Firm Indicator (0.001) (0.001) (0.011) 0.001 (0.012) Industry- Mean Growth Convert. Operating Firm Variable Debt Options Securities Earnings Size Leverage 0.247 -0.378 0.349 -0.017 0.178 (0.001) (0.001) (0.001) (0.495) (0.001) Marginal Tax 0.007 0.157 -0.174 0.205 -0.041 Rate (0.763) (0.001) (0.001) (0.001) (0.095) Debt Security 0.190 -0.375 -0.006 0.360 0.450 (0.001) (0.001) (0.807) (0.001) (0.001) Investment -0.023 -0.092 0.144 -0.440 -0.110 Tax Shield (0.346) (0.001) (0.001) (0.001) (0.001) Altman's -0.030 -0.216 0.214 -0.228 -0.280 z-score (0.209) (0.001) (0.001) (0.001) (0.001) Industry-Mean -0.161 0.016 0.168 0.134 Debt (0.001) (0.508) (0.001) (0.001) Growth Options -0.053 0.002 0.017 -0.343 (0.029) (0.937) (0.496) (0.001) Convertible 0.021 0.008 -0.014 -0.127 Securities (0.382) (0.753) (0.001) (0.001) Operating 0.086 -0.010 -0.444 0.496 Earnings (0.001) (0.700) (0.001) (0.001) Firm Size 0.070 -0.109 0.066 0.227 (0.004) (0.001) (0.006) (0.001) Negative Book 0.030 -0.333 0.209 -0.194 -0.019 Equity (0.219) (0.001) (0.001) (0.001) (0.430) Asset Structure 0.351 -0.033 -0.051 0.032 -0.027 (0.001) (0.168) (0.036) (0.189) (0.272) Firm Age -0.079 -0.061 0.058 -0.025 0.094 (0.001) (0.012) (0.017) (0.304) (0.001) Tax-Exhausted -0.0324 0.020 0.101 -0.444 -0.044 Firm Indicator (0.192) (0.407) (0.001) (0.001) (0.071) Neg. Asset Firm Tax- Variable Equity Structure Age Exhausted Leverage 0.134 0.258 0.060 0.167 (0.001) (0.001) (0.013) (0.001) Marginal Tax -0.099 -0.084 -0.347 -0.427 Rate (0.001) (0.001) (0.001) (0.001) Debt Security 0.015 0.647 0.161 0.109 (0.548) (0.001) (0.001) (0.001) Investment 0.128 0.106 0.065 0.996 Tax Shield (0.001) (0.001) (0.007) (0.001) Altman's 0.010 0.033 0.050 0.077 z-score (0.681) (0.179) (0.038) (0.002) Industry-Mean 0.043 0.311 -0.079 -0.036 Debt (0.076) (0.001) (0.001) (0.141) Growth Options -0.182 -0.113 -0.148 -0.076 (0.001) (0.001) (0.001) (0.002) Convertible 0.080 -0.077 0.078 0.145 Securities (0.001) (0.001) (0.001) (0.001) Operating -0.051 0.168 0.001 -0.451 Earnings (0.036) (0.001) (0.977) (0.001) Firm Size -0.054 0.068 0.125 -0.129 (0.025) (0.005) (0.001) (0.001) Negative Book 0.051 0.026 0.120 Equity (0.034) (0.284) (0.001) Asset Structure 0.061 0.118 0.085 (0.012) (0.001) (0.001) Firm Age 0.026 0.136 0.056 (0.284) (0.001) (0.001) Tax-Exhausted 0.120 0.107 0.056 Firm Indicator (0.001) (0.001) (0.022) Panel B: Continuous Variables Variable n Mean Std Dev Median Min Leverage 1,701 0.195 0.212 0.120 0 Marginal tax rate 1,701 0.302 0.143 0.340 0 Debt security 1,701 0.053 0.100 0.037 0.002 Investment tax shield 1,701 0.009 0.035 0 0 Altman's z-score 1,701 -1.153 82.79 0.404 -3263 Industry-mean leverage 1,701 0.255 0.117 0.243 0.017 Growth options 1,701 2.517 4.021 1.932 -76 Convertible securities 1,701 0.022 0.111 0 0 Operating earnings 1,701 0.121 0.147 0.124 -2.908 Firm size 1,701 1.424 1.550 1.042 0.007 Asset structure 1,701 0.525 0.369 0.434 0.027 Firm age 1,701 5.000 2.583 5.000 1.000 Variable Max Leverage 0.967 Marginal tax rate 0.480 Debt security 3.626 Investment tax shield 0.484 Altman's z-score 614 Industry-mean leverage 0.764 Growth options 69 Convertible securities 3.572 Operating earnings 0.787 Firm size 21 Asset structure 3.645 Firm age 9.000 Panel C: Discrete variables Variable n Percent Tax-exhausted firm indicator 234 13.76 Negative book equity indicator 19 1.11 This table presents summary statistics for the sample of 1,701 firm-year observations over the period 1978 to 1988 used in the hypotheses tests. Panel A provides correlations for the explanatory variables with Pearson correlations presented below the diagonal and Spearman correlations presented above the diagonal; related probability values are presented in parentheses. Panel B presents the distributional statistics for the Continuous variables and Panel C presents those for the discrete variables. Continuous variables are as follows (all dollar-denominated variables are stated in millions). Leverage is the book value of the sum of short-term, long-term, and convertible debt deflated by one-year-lagged firm market value (the sum of market value of equity and book value of total debt). The marginal tax rate is a trichotomous variable which is equal to: (1) the top statutory corporate tax rate if taxable income is positive and the tax loss carryforwards (TLCF) balance is nonpositive; (2) one-half the top statutory rate if either taxable income is positive or TLCF is nonpositive; and (3) zero if taxable income is nonpositive and TLCF is positive. The top statutory corporate tax rate is 48 percent before 1979, 46 percent from 1979 to 1986, 39.5 percent in 1987, 34 percent from 1988 to 1992, and 35 percent starting in 1993. Deferred tax expense grossed up by the top statutory tax rate is subtracted from the sum of income before extraordinary items, income tax expense, minority interest, and extraordinary ite ms and discontinued operations (grossed up by one minus the top statutory tax rate) to arrive at taxable income. The debt security proxy is the sum of financial statement depreciation and investment tax credits (missing observations are replaced by zeros) grossed-up by the corporate statutory tax rate. This amount is scaled by firm market value, which is defined as the sum of the book value of total debt and the market value of equity. The investment-related tax shield variable is the product of the debt securability proxy and the tax-exhausted firm indicator variable (see below). Altman's z-score is book total assets divided by the sum of 1.2 times working capital, 1.4 times retained earnings, 3.3 times earnings before interest and taxes, and net sales. Industry-mean leverage is the contemporaneous industry mean of the dependent variable, where industry-mean represents the mean among other firms in the narrowest SIC code that includes at least five firms other than the firm being examined. Growth options is the market value of common stock divided by the book value of common equity. Convertible securities is the sum of the book values of convertible debt and convertible preferred shares; this amount is deflated by one-year-lagged firm market value. Operating earnings is earnings before depreciation, interest, and taxes deflated by one-year-lagged firm market value. Firm size is the natural logarithm of net sales. Asset structure is total property, plant, and equipment scaled by one-year-lagged total assets. Firm age is the number of years that have elapsed since the firm's initial public offering. Age*MTR is the product of firm age and the marginal tax rate. Age*Investment-related tax shield is the product of firm age and the investment-related tax shield variable. Indicator variables are as follows. The year dummy variables indicate the calendar year of the observation. A tax-exhausted firm dummy variable indicates those firms that have been assigned a zero marginal tax rate according to the above trichotomous variable conditions. The negative book equity dummy indicates if the book value of common equity is negative. TABLE 5 POOLED OLS, FIXED EFFECTS, AND SELECTION-CORRECTED REGRESSION RESULTS Pooled OLS regression [LEVERAGE.sub.it] = [alpha] + [[gamma].sub.t] + [[beta].sub.1] [MTR.sub.i,t-1] + [[beta].sub.2][D.sub.it] + [[beta].sub.3] [SECURITY.sub.it] + [[beta].sub.4][IRTS.sub.it] + [[beta].sub.5] [Z-SCORE.sub.it] + [[beta].sub.6][INDUSTRY-MEAN.sub.it] + [[beta].sub.7][GROWTH.sub.it] + [[beta].sub.8][CONVERT.sub.it] + [[beta].sub.9][EARNINGS.sub.it] + [[beta].sub.10][SIZE.sub.it] + [[beta].sub.11][NEG.EQUITY.sub.it] + [[beta].sub.12][ASSETS.sub.it] + [[beta].sub.13][AGE.sub.it] + [[beta].sub.14][AGE.sub.it] * [MTR.sub.i,t-1] + [[beta].sub.15][AGE.sub.it]*[IRTS.sub.it] + [[epsilon].sub.it] Fixed effects regression [LEVERAGE.sub.it] = [[alpha].sub.i] + [[gamma].sub.t] + [[beta].sub.1][MTR.sub.i,t-1] + [[beta].sub.2][D.sub.it] + [[beta].sub.3][SECURITY.sub.it] + [[beta].sub.4][IRTS.sub.it] + [[beta].sub.5][Z-SCORE.sub.it] + [[beta].sub.6][INDUSTRY-MEAN.sub.it] + [[beta].sub.7][GROWTH.sub.it] + [[beta].sub.8][CONVERT.sub.it] + [[beta].sub.9][EARNINGS.sub.it] + [[beta].sub.10][SIZE.sub.it] + [[beta].sub.11][NEG.EQUITY.sub.it] + [[beta].sub.12][ASSETS.sub.it] + [[beta].sub.13][AGE.sub.it] * [MTR.sub.i,t-1] + [[beta].sub.15][AGE.sub.it] * [IRTS.sub.it] + [[epsilon].sub.it] Pooled OLS Balanced Unbalanced Variable Prediction 1 2 Intercept ? -0.038 -0.097 *** Marginal tax rate + 0.149 ** 0.134 *** Tax-exhausted indicator ? 0.124 *** 0.072 *** Debt security + 0.105 ** 0.051 Investment-related tax shield - 0.284 0.898 *** Altman's z-score - 0.000 0.000 Industry-mean leverage + 0.379 *** 0.669 *** Growth options - -0.008 *** 0.000 Convertible securities + 0.447 *** 0.371 *** Operating earnings + 0.149 *** 0.097 *** Firm size + -0.000 0.010 *** Negative book equity indicator + 0.101 ** 0.171 *** Asset structure + 0.037 *** 0.010 ** Firm age + 0.022 *** 0.021 *** [lambda] (inverse Mills ratio) ? Age*MTR ([H.sub.1]) - -0.059 *** -0.051 *** Age*IRTS ([H.sub.2]) - -0.077 * -0.156 *** Adjusted [R.sup.2] 0.219 0.268 F-statistic 31.54 [degrees] 90.38 [degrees] Number of observations 1,701 5,613 Fixed Effects Balanced Unbalanced Variable 3 4 Intercept 0.136 *** 0.062 *** Marginal tax rate 0.016 0.016 Tax-exhausted indicator 0.090 *** 0.061 *** Debt security 0.042 0.089 *** Investment-related tax shield 0.580 * 0.813 *** Altman's z-score 0.001 0.000 Industry-mean leverage 0.061 0.350 *** Growth options -0.003 *** 0.000 Convertible securities 0.232 *** 0.286 *** Operating earnings -0.018 -0.051 *** Firm size -0.003 -0.001 Negative book equity indicator 0.121 *** -0.120 *** Asset structure 0.033 * 0.004 Firm age 0.012 *** 0.015 *** [lambda] (inverse Mills ratio) Age*MTR ([H.sub.1]) -0.032 *** -0.025 *** Age*IRTS ([H.sub.2]) -0.125 *** -0.139 *** Adjusted [R.sup.2] 0.681 0.683 F-statistic 17.33 [degrees] 15.17 [degrees] Number of observations 1,701 5,613 Selection- Corrected OLS Variable 5 Intercept 0.392 Marginal tax rate 0.129 ** Tax-exhausted indicator 0.118 *** Debt security 0.102 ** Investment-related tax shield 0.380 Altman's z-score -0.001 Industry-mean leverage 0.373 *** Growth options -0.008 *** Convertible securities 0.431 *** Operating earnings 0.162 *** Firm size -0.002 Negative book equity indicator 0.372 Asset structure 0.042 *** Firm age 0.022 *** [lambda] (inverse Mills ratio) -0.289 Age*MTR ([H.sub.1]) -0.057 *** Age*IRTS ([H.sub.2]) -0.087 * Adjusted [R.sup.2] 0.212 F-statistic 29.56 [degrees] Number of observations 1,701 This table presents regression results for the levels leverage models for both balanced and unbalanced panels using OLS and two-way fixed effects estimation for the firms' first through ninth years of public operation. Also, the results for the Heckman (1979) two-stage estimation procedure are reported in the selection-corrected column. The balanced sample discards the entire time-series of firms if any missing observations are encountered in the nine years; the unbalanced panel discards only the firm-year when missing observations are encountered. The dependent variable and the explanatory variables are defined in Table 4. All regressions include unreported calander year dummy variables. Regression equation F-tests significant at less than 0.001 are identified by a [degrees] superscript. In this table, the subscripts i and t identify firms and time, respectively. The superscript asterisks indicate explanatory variable coefficient significance at p-values less than 0.10 (*), 0.05 (**), and 0.01 (***) in one-tailed tests where directional predictions are made and two-tailed tests otherwise. TABLE 6 TESTS OF THE TIME-SERIES VARIATION IN REFINANCING COST PROXIES Variable Prediction Years 1-3 Years 4-6 Years 7-9 Average size of private - 21.45 22.86 28.85 ** long-term debt contracts Fraction of long-term + 0.078 0.168 *** 0.200 *** debt in public bonds Fraction of total debt - 0.142 0.114 ** 0.087 * in short- term debt Number of long-term + 3.57 4.23 *** 4.21 debt contracts Fraction of public bonds - 0.895 0.848 * 0.747 *** that are callable This table presents results for tests that examine differences in mean capital structure characteristics between firms' first three public years (years 1 to 3) and their second three public years (years 4 to 6) and between their second three pubic years and their third three public years (years 7 to 9). Variable definitions follow with those that are dollar-denominated stated in millions. The average size of private long-term debt contracts is the book value of total private long-term debt divided by the number of private long-term debt contracts outstanding. The fraction of long-term debt in public bonds is the book value of public long-term debt divided by book value of total long-term debt. The fraction of total debt in short-term debt is the book value of short-term debt divided by the sum of the book value of short-term and long-term debt. The number of long-term debt contracts counts each issue of public bonds as a single contract. The fraction of public bonds that are callable is the number of callable public bond issues outstanding divided by the total number of callable and noncallable public bond issues outstanding. All test variables represent the three-year means corresponding to firms' first three public years, second three public years in Columns 1, 2, and 3, respectively. The differences in these variables are evaluated with the Wilcoxon matched-pairs signed-rank test. The superscript asterisks indicate p-values less than 0.10 (*), 0.05 (**), and 0.01 (***) in one-tailed tests for the directional predictions when the value this period is significantly different from the corresponding, value in the preceding period. The sample consists of 96 firms that went public between 1978 and 1988. TABLE 7 POOLED OLS AND FIXED EFFECTS REGRESSION RESULTS - BALANCED PANEL PARTITIONED BY PRIVATE AGE AT IPO DATE Pooled OLS regression [LEVERAGE.sub.it] = [alpha] + [[gamma].sub.t] + [[beta].sub.1] [MTR.sub.i,t-1] + [[beta].sub.2][D.sub.it] + [[beta].sub.3] [SECURITY.sub.it] + [[beta].sub.4][IRTS.sub.it] + [[beta].sub.5] [Z-SCORE.sub.it] + [[beta].sub.6]INDUSTRY-[MEAN.sub.it] + [[beta].sub.7][GROWTH.sub.it] + [[beta].sub.8][CONVERT.sub.it] + [[beta].sub.9][EARNINGS.sub.it] + [[beta].sub.10][SIZE.sub.it] + [[beta].sub.11][NEG.EQUITY.sub.it] + [[beta].sub.12][ASSETS.sub.it] + [[beta].sub.13][AGE.sub.it] + [[beta].sub.14][AGE.sub.it] * [MTR.sub.i,t-1] + [[beta].sub.15][AGE.sub.it]*[IRTS.sub.it] + [[epsilon].sub.it] Fixed effects regression [LEVERAGE.sub.it] = [[alpha].sub.1] + [[gamma].sub.t] + [[beta].sub.1][MTR.sub.i,t-1] + [[beta].sub.2][D.sub.it] + [[beta].sub.3][SECURITY.sub.it] + [[beta].sub.4][IRTS.sub.it] + [[beta].sub.5][Z-SCORE.sub.it] + [[beta].sub.6][INDUSTRY-MEAN.sub.it] + [[beta].sub.7][GROWTH.sub.it] + [[beta].sub.8][CONVERT.sub.it] + [[beta].sub.9][EARNINGS.sub.it] + [[beta].sub.15][AGE.sub.it] * [IRTS.sub.it] + [[epsilon].sub.it] Older Firms Pooled OLS Fixed Effects Variable Prediction 1 2 Intercept ? 0.026 0.197 *** Marginal tax rate + 0.058 0.072 Tax-exhausted ? 0.174 *** 0.129 *** indicator Debt security + 0.977 *** 1.628 *** Investment-related tax - -3.362 *** -2.2l5 *** shield Altman's z-score - 0.039 *** 0.019 *** Industry-mean leverage + 0.428 *** 0.044 Growth options - -0.011 *** -0.007 *** Convertible securities + 1.187 *** 0.779 *** Operating earnings + -0.355 *** -0.405 *** Firm size + 0.020 *** -0.019 *** Negative book equity + 0.004 -0.008 indicator Asset structure + -0.004 -0.066 *** Firm age + 0.016 *** 0.013 *** Age*MTR ([H.sub.1]) - -0.045 *** -0.040 *** Age*IRTS ([H.sub.2]) - 0.172 0.101 Adjusted [R.sup.2] 0.366 0.743 F-statistic 34.57[degrees] 20.36[degrees] Number of observations 873 873 Younger Firms Pooled OLS Fixed Effects Variable 3 4 Intercept -0.041 0.091 ** Marginal tax rate 0.189 *** 0.057 Tax-exhausted 0.090 *** 0.064 *** indicator Debt security 0.072 * 0.011 Investment-related tax 1.276 ** 1.167 *** shield Altman's z-score 0.000 0.000 Industry-mean leverage 0.338 *** 0.075 Growth options -0.004 *** -0.006 Convertible securities 0.326 *** 0.212 *** Operating earnings 0.244 *** 0.072 ** Firm size -0.009 *** -0.006 Negative book equity 0.205 *** 0.148 *** indicator Asset structure 0.033 ** 0.036 * Firm age 0.022 *** 0.011 *** Age*MTR ([H.sub.1]) -0.066 *** -0.035 *** Age*IRTS ([H.sub.2]) -0.l42 ** -0.157 *** Adjusted [R.sup.2] 0.227 0.676 F-statistic 17.21[degrees] 14.80[degrees] Number of observations 828 828 This table presents regression results for the levels leverage models for older and younger firms using ordinary least squares and two-way fixed effects estimation for the firms' first through ninth years of public operation. The balanced sample discards the entire time-series of firms if any missing observations are encountered in the nine years. Older (younger) firms are defined as those for which the duration of their private history, which is measured as the number of years that have elapsed between their incorporation and their initial public offering, is above (below) the median for the sample. The dependent variable and the explanatory variables are defined in Table 4. All regressions include unreported calendar year dummy variables. Regression equation F-tests significant at less than 0.001 are identified by a [degrees] superscript. In this table, the subscripts i and t identify firms and time, respectively. The superscript asterisks indicate explanatory variable coefficient significance at p-values less than 0.10 (*), 0.05 (**),and 0.01 (***) in one-tailed tests where directional predictions are made and two-tailed tests otherwise.

Submitted: January 2001

Accepted: April 2002

(1.) For expositional convenience, the term "refinancing costs" refers to all contracting costs other than the direct or indirect reorganization costs incurred by bankrupt firms. For example, firms' capital structures may change over their early years in ways that make renegotiating with creditors more difficult while restricting additional borrowing.

(2.) Besides studies that examine whether tax regime shifts affect financing decisions (e.g., Givoly et al. 1992; Trezevant 1992), empirical research has considered a variety of other macroeconomic factors such as recent capital market performance (Marsh 1982) and the availability of competing securities (Taggart 1985). Conversely, testing in this study suppresses confounding from changes in macroeconomic conditions.

(3.) Other research that finds a positive relation between debt policy and tax rates study incremental financing decisions (MacKie-Mason 1990; Givoly et al. 1992; Graham 1996a), which avoids the endogeneity problem.

(4.) The issue of whether lagging is sufficient to suppress endogeneity bias is examined in Section four. The sensitivity of the results to specifying MTR with the trichotomous variable is considered in Section five.

(5.) The regression also includes two control variables intended to remove any remaining confounding influences on the tests of the tax-shield substitution hypothesis. First, the contemporaneous industry mean of the dependent variable controls for industry-specific factors (Dammon and Senbet 1988). Second, the ratio of property, plant, and equipment to total book assets controls for differences in asset structures. However, Harris and Raviv (1991) and Dhaliwal et al. (1992) explain that this variable may absorb lingering variation from the debt securability effect that is not captured by specifying ITCs and depreciation expense as the proxy for investment-related tax shields.

(6.) In particular, the low correlation between firm age and operating earnings suggests that time-series variation in profitability does not drive any evidence on the hypotheses (Dhaliwal and Graham 2001). However, to address potential multi-collinearity, explanatory variables with absolute cross-correlations with the tax proxies exceeding 0.15 were removed from the equations one at a time. None of these re-specifications qualitatively affect the results.

(7.) This research design emphasizes results from tests on the balanced panel that is compiled by discarding the entire time-series of firms if any missing observations occur in the nine years. However, results are also provided for the unbalanced sample that is compiled by discarding only firm-years that have missing observations. Although the larger unbalanced panel increases statistical efficiency and may be less susceptible to survivorship bias, there is concern that evidence for this sample could be caused by changes in its composition over time.

(8.) The coefficients on many control variables tend to approach zero or are estimated imprecisely in the fixed effects regression, which removes cross-sectional variation in the data. However, the theories motivating including these potential determinants of capital structure are more amenable to explaining differences in financing across firms. For these cross-sectional predictions, only the results that are reported in Column (1) of Table 5 are valid.

(9.) Since the variable that tests [H.sub.2] is a three-way interaction, i.e., the product of the debt securability, the tax-exhausted firm indicator, and age proxies, it may be important to estimate the regression with all component two-way interactions as well as the main variables to avoid biasing the coefficients on the interaction term of interest.

(10.) The potential influence of changing macroeconomic conditions is partially controlled in this study by aligning the data in event time. Observing firms as they age starting with the year that they went public, rather than across calendar years, spreads the fixed time effects. Still, this only imperfectly suppresses the impact of prevailing conditions since the sample firms are not equally distributed across the IPO years; e.g., only three firms in the sample underwent an IPO in 1978, while 47 firms had their IPO in 1983 (see Table 2).

(11.) "Hausman (1978) relies on the fact that the random effects model is only a valid alternative to the fixed effects model when the firm-specific and time-specific effects are uncorrelated with the explanatory variables. His test statistic exploits the notion that in the absence of correlation both models are consistent, but the fixed effects model is inefficient. In the presence of correlation, the fixed effects model is consistent, but the random effects model is inconsistent since it does not allow for the variable coefficients.

(12.) Replacing this proxy with the fraction of firms with public debt outstanding does not qualitatively alter the results reported in this section. Extant research suggests that the presence of public bonds in firms' capital structures undermines renegotiations with private lenders (Pittman and Klassen 2001). For example, banks seldom offer concessions to firms with noninvestment grade public debt; young firms typically have speculative grade issues.

(13.) There are other potential proxies for refinancing costs such as whether the firm's debt covenants provide flexibility, e.g., allowing the firm to postpone principal or interest payments or to extend the maturity of its debt, etc. However, neither Compustat nor the various Moody's manuals report this level of detail on firms' capital structures.

(14.) The Wilcoxon matched-pairs signed-rank test is used to address censoring in some of the proxies; all results are robust to t-tests of the paired observations that assume the differences are normally distributed.

(15.) For example, the sample probably excludes, at least until the later years of the time-series, distressed firms that are acquisition targets. Zweibel (1996) argues that managers may increase leverage when experiencing takeover threats as a defensive strategy to signal its commitment to value-increasing restructuring. For investment policy, fixed production capacity can be difficult to reverse when firms become financially distressed (Jensen 1993).

(16.) The panel data examined in this study was selected by conditioning on survival, which suggests that these financial distress proxies are adequate predictors of attrition. In fact, descriptive statistics (not tabulated) for these proxies indicate that there are time-series changes within firms in the probability of surviving through the nine years, which implies that selection cannot be dismissed as a random occurrence. The fixed effects regression in Table 5 that models within-firm variation in financing decisions removes any time-invariant sample selection bias.

(17.) Although Shevlin (1999b) criticizes his methodology, Plesko (1999) finds that the trichotomous variable usually performs better than both Graham's (1966a) after-financing MTR and Graham et al.'s (1998) before-financing MTR in estimating the firm-specific MTR derived from corporate tax return data. The evidence consistent with [H.sub.1] and [H.sub.2] is robust to reestimating Equation (1) in first-differences with Graham's (1996a) simulated rates representing firms' marginal tax rates.

(18.) Jain and Kini (1994) provide evidence that a significant decline in firms' operating performance occurs in the three years following IPO, which implies that pre-IPO earnings cannot be reliably used as the seed data for the Graham (1996a) and Graham et al. (1998) simulated marginal tax rate algorithms.

(19.) In addition, respecifying tax status with Graham et al.'s (1998) simulated tax rates increases the balanced panel from 189 to 438 firms, which provides additional evidence that survivorship bias does not drive the results.

(20.) This may be a particularly important issue since Manzon and Plesko (2001) report that the book-tax spread increased between 1988 and 1999, a period that coincides with the later years of this study.

(21.) In fact, there are other reasons to exclude LTCs from the investment-related tax shield proxy. First, the Tax Reform Act of 1986 suspended the accumulation of investment tax credits, although firms were permitted to carry forward any unused amounts for up to 15 years. Most sample firms are examined in years surrounding this change. Second, Scholes et al. (1990) explain that firms reduce their tax burdens by investing in tax shelters such as ITCs. Accordingly, tax credit carryforwards may be considered a mixed signal that may indicate a high level of taxable income that the company is attempting to shelter, rather than a lack of taxable income. MacKie-Mason (1990) argues that only firms with low taxable incomes will substitute their tax shields.

(22.) There is no substantive evidence or theory on the proper assignment of firms in the sample based on their probabilities of losing the deductibility of tax shields. Any delineation could be considered arbitrary, which implies that the tests should be re-run on several ad hoc cut-offs (Dhaliwal et al. 1992).

(23.) Cloyd et al. (1997) and Ayers et al. (2001) maintain that neither finns with low marginal tax rates ("taxexhausted" firms) nor those with high marginal tax rates ("tax-insatiable" firms) will substitute since the amount of their investment-related tax shields will hardly, if at all, affect the tax subsidy on debt for these firms.

(24.) The empirical design in this paper follows extant research by deflating (where appropriate) the regression variables by one-year-lagged firm market value. However, this will only generate consistent coefficient estimates if either the variables are measured at short intervals or growth is absent. The data satisfy neither of these conditions: the variables are measured at annual intervals; and the firms in the balanced panel grow somewhat during their early public years with the mean (median) annual sales increase amounting to 6 percent (0 percent). Since this could particularly bias results for firms with higher growth rates, the fixed and random effects models were reestimated after excluding observations with more than 13 percent sales annual growth (the upper quartile). These robustness tests, which follow prior research that sorts firms according to such within-sample characteristics when examining this issue (e.g., Kaplan and Zingales 1997), provide evidence consistent with [H.sub.1] and [H.sub.2]

(25.) As the influence of an additional year of public existence may decline with age, nonlinear transformations of firm age (the natural logarithm of one plus age, second-order logs, square roots, and reciprocals) were specified to provide the data flexibility. [R.sup.2] was virtually identical in each case, suggesting that the data do not distinguish between a linear specification and the nonlinear specifications.

(26.) Rajan and Zingales (1998) and Helwege and Liang (1996) report descriptive statistics consistent with there being a life-cycle pattern in corporate financing, with firms more dependent on external capital in their early years to finance their lucrative initial investment projects. This implies that firms' demand for external financing subsiding with age may spuriously induce the evidence in this paper; i.e., as firms' investment needs stabilize, so should their capital structures. Similarly, Diamond (1989) argues that lenders perceive firms to become less risky with age.

(27.) Pittman and Klassen (2001) find some evidence that firms with longer private histories at their IPO dates have capital structures that are more constrained by refinancing costs. Similarly, estimates (not tabulated) of the time-series variation in the impact of lagged leverage (Gilson 1997) and profitability (Dittmar 2002) on firms' financing decisions also provide indirect evidence that these older firms have more difficulty adjusting leverage.

(28.) Some firms were founded more than 100 years before going public, implying that mean private age could not be reliably used as the measure of central tendency in this section. The results are not sensitive to replacing age with its natural logarithm. Data on firms' private ages were mainly collected from Ward's Business Directory of U.S. Private and Public Companies and a database that Professor Jay Ritter of the University of Florida maintains.

REFERENCES

Alderson, M., and B. Betker, 1995. Liquidition costs and capital structure. Journal of Financial Economics 39 (1): 45-69.

Altman, E. I. 1968. Financial ratios, discriminant analysis, and the prediction of corporate bankruptcy. Journal of Finance 23 (4): 589-609.

Ang, J. S., J. H. Chua, and J. J. McConnell. 1982. The administrative costs of corporate bankruptcy. A note. Journal of Finance 37 (1): 219-226.

Asquith, P., R. Gertner, and D. Scharfstein. 1994. Anatomy of financial distress: An examination of junk bond issuers. Quarterly Journal of Economics 109 (3): 625-658.

Ayers, B. C., C. B. Cloyd, and J. R. Robinson. 2001. The influence of income taxes on the use of inside and outside debt by small businesses. National Tax Journal 54 (1): 27-55.

Bolton, P., and D. S. Scharfstein. 1996. Optimal debt structure and the number of creditors. Journal of Political Economy 104 (1): 1-25.

Bradley, M., G. Jarrell, and E. Kim. 1984. On the existence of an optimal capital structure: Theory and evidence. Journal of Finance 39 (3): 857-877.

Cloyd, C. B., S. T. Limberg, and J. R. Robinson. 1997. The impact of federal taxes on the use of debt by closely held corporations. National Tax Journal (2): 261-277.

Dammon, R. M., and L. W. Senbet. 1988. The effects of taxes and depreciation on corporate investment and leverage. Journal of Finance 43 (2): 357-373.

DeAngelo, H., and R. W. Masulis. 1980. Optimal capital structure under corporate and personal taxation. Journal of Financial Economics 8 (1): 5-29.

Dhaliwal, D., R. Trezevant, and S. Wang. 1992. Taxes, investment-related tax shields and capital structure. The Journal of the American Taxation Association 14 (1): 1-21.

-----, and J. R. Graham. 2001. Discussion of: The influence of firm maturation on the time-series variation to optimal capital structure. The Journal of the American Tax Association 23 (Supplement): 95-98.

Diamond, D. W. 1989. Reputation acquisition in debt markets. Journal of Political Economy 97 (4): 828-862.

Dittmar, A. K. Capital structure in corporate spin-offs. Journal of Business (forthcoming).

Essig, S. 1991. Convertible securities and capital structure determinants. Ph.D. dissertation, University of Chicago.

Gertner, R., and D. Scharfstein. 1991. A theory of workouts and the effects of reorganization law. Journal of Finance 46 (4): 1189-1222.

Gilson, S. C. 1997. Transactions costs and capital structure choice: Evidence from financially distressed firms. Journal of Finance 52 (1): 161-196.

Givoly, D., C. Hayn, A. H. Ofer, and O. Sarig. 1992. Taxes and capital structure: Evidence from firms' response to the Tax Reform Act of 1986. The Review of Financial Studies 5 (2): 331-355.

Graham, J. R. 1996a. Debt and the marginal tax rate. Journal of Financial Economics 41 (1): 41-73.

-----. 1996b. Proxies for the corporate marginal tax rate. Journal of Financial Economics 42 (2): 187-221.

-----. M. L. Lemmon, and J. S. Schallheim. 1998. Debt, leases, taxes, and the endogeneity of corporate tax status. Journal of Finance 53 (1): 131-161.

-----, and C. W. Smith. 1999. Tax incentives to hedge. Journal of Finance 54 (6): 2241-2262.

-----. 2000. How big are the tax benefits of debt? Journal of Finance 55 (5): 1901-1941.

-----. 2001. Taxes and corporate finance: A review. Working paper, Duke University.

Grossman, S., and O. Hart. 1980. Takeover bids, the free-rider problem and the theory of the corporation. Bell Journal of Economics 11 (Spring): 42-64.

Harris, M., and A. Raviv. 1991. The theory of capital structure. Journal of Finance 46 (1): 297-355.

Hausman, J. A. 1978. Specification tests in econometrics. Econometrica 46 (6): 1251-1271.

Heckman, J. J. 1979. Sample selection bias as specification error. Econometrica 47 (1): 153-162.

Helwege, J., and N. Liang. 1996. Is there a pecking order? Evidence from a panel of IPO firms. Journal of Financial Economics 40 (3): 429-458.

Hsiao, C. 1986. Analysis of Panel Data. Cambridge, U.K.: Cambridge University Press.

Jain, B. A., and O. Kini. 1994. The post-issue operating performance of IPO firms. Journal of Finance 49 (5): 1699-1726.

Jensen, M. C., and M. H. Meckling. 1976. Theory of the firm: Managerial behavior, agency costs, and ownership structure. Journal of Financial Economics 3 (2): 305-360.

-----. 1986. Agency costs of free cash flow, corporate finance, and takeovers. American Economic Review 76 (2): 323-329.

-----. 1989. Eclipse of the public corporation. Harvard Business Review (September-October): 61-74.

-----. 1993. The moden industrial revolution and the challenge of internal control systems. Journal of Finance 48 (3): 831-880.

Kaplan, S. N., and L. Zingales. 1997. Do investment-cash flow sensitivities provide useful measures of financing constraints? Quarterly Journal of Economics 112 (1): 169-215.

Kinney, M. R., and E. P. Swanson. 1993. The accuracy and adequacy of tax data in Compustat. The Journal of the American Tax Association 15 (1): 121-135.

Lang, M. 1991. Time-varying stock price response to earnings induced by uncertainty about the time-series process of earnings. Journal of Accounting Research 29 (2): 229-257.

MacKie-Mason, J. K. 1990. Do taxes affect corporate financing decisions? Journal of Finance 45 (5): 1471-1493.

Manzon, G. B., Jr., and G. A. Plesko. 2001. The relation between financial and tax reporting measures of income. Working paper, Massachusetts Institute of Technology.

Marsh, P. 1982. The choice between equity and debt: An empirical study. Journal of Finance 37 (1): 121-144.

Mayers, D. 1998. Why firms issue convertible bonds: The matching of financial and real investment options. Journal of Financial Economics 47 (1): 83-102.

Miller, M. H. 1977. Debt and taxes. Journal of Finance 32 (2): 26 1-275.

Modigliani, F., and M. Miller. 1963. Corporate income taxes and the cost of capital: A correction. American Economic Review 53 (3): 433-443.

Myers, S. C. 1984. The capital structure puzzle. Journal of Finance 39 (3): 575-592.

_____, and N. S. Majluf. 1984. Corporate financing and investment decisions when firms have information that investors do not have. Journal of Financial Economics 3 (2): 187-221.

Pagano, M., F. Panetta, and L. Zingales. 1998. Why do companies go public? An empirical analysis. Journal of Finance 53 (1): 27-65.

Parrino, R., and M. Weisbach. 1999. Measuring investment distortions arising from stockholder-bondholder conflicts. Journal of Financial Economics 52 (1): 3-42.

Petersen, M. A., and R. G. Rajan. 1994. The benefits of lending relationships: Evidence from small business data. Journal of Finance 49 (1): 3-37.

Pittman, J. A., and K. J. Klassen. 2001. The influence of firm maturation on the time-series variation to optimal capital structure. The Journal of the American Taxation Association 23 (Supplement): 70-94.

Plesko, G. 1999. An evaluation of alternative measures of corporate tax rates. Working paper, Massachusetts Institute of Technology.

Rajan, R., and L. Zingales. 1995. What do we know about capital structure: Evidence from G-7 countries. Journal of Finance 50 (5): 1421-1460.

_____, and _____. 1998. Financial dependence and growth. American Economic Review 88 (3): 559-586.

Ross, S. 1977. The determination of financial structure: The incentive signalling approach. Bell Journal of Economics 8 (Spring): 23-40.

Scholes, M. S., and M. A. Wolfson. 1989. Issues in the theory of optimal capital structure. In Frontiers of Modern Finance, edited by S. Bhattacharya, and G. Constantinides. New York, NY: Rowman & Littlefield.

_____, G. P. Wilson, and M. A. Wolfson. 1990. Tax planning, regulatory capital planning and financial reporting strategy for commercial banks. The Review of Financial Studies 3 (4): 625-650.

_____, and M. A. Wolfson. 1992. Taxes and Business Strategy: A Planning Approach. Englewood Cliffs, NJ: Prentice Hall.

Scott, J. H. 1976. A theory of optimal capital structure. Bell Journal of Economics 7 (Spring): 33-54.

Shyam-Sunder, L., and S. C. Myers. 1999. Testing static tradeoff against pecking order models of capital structure. Journal of Financial Economics 51(2): 219-244.

Shevlin, T. 1990. Estimating corporate marginal tax rates with asymmetric tax treatment of gains and losses. The Journal of the American Taxation Association 11 (2): 51-67.

_____. 1999a. Research in taxation. Accounting Horizons 13 (4): 427-441.

_____. 1999b. A critique of Plesko's "An evaluation of alternative measures of corporate tax rates." Working paper, University of Washington.

Slemrod, J., and W. Shobe. 1990. The use of panel data in the analysis of the behavioral response to taxation. In Simulation Models in Tax and Transfer Policy: Proceedings of an International Symposium, edited by J. K. Brunner, and H.-G. Petersen. Frankfurt, Germany: Campus Verlag.

Smith, C. W., and L. M. Wakeman. 1985. Determinants of corporate leasing policy. Journal of Finance 40 (3): 895-908.

Stein, J. C. 1992. Convertible bonds as backdoor equity financing. Journal of Financial Economics 32 (1): 3-22.

Taggart, R. A. 1985. Secular patterns in the financing of U.S. corporations. Corporate Capital Structures in the U.S. Washington, D.C.: National Bureau of Economic Research.

Teoh, S. H., I. Welch, and T.J. Wong. 1998. Earnings management and the long-run market performance of initial public offerings. Journal of Finance 53 (6): 1935-1975.

Titman, S. 1984. The effect of capital structure on a firm's liquidation decision. Journal of Financial Economics 13 (1): 137-151.

_____, and R. Wessels. 1988. The determinants of capital structure choice. Journal of Finance 43 (1): 1-19. Trezevant, R. 1992. Debt financing and tax status: Test of the substitution effect and the tax exhaustion hypothesis using firms' responses to ERTA of 1981. Journal of Finance 47 (4): 1557-1568.

_____. 1994. How did firms adjust their tax-deductible activities in response to the Economic Tax Act of 1981? National Tax Journal 67 (2): 253-271.

Warner, J. B. 1977. Bankruptcy costs: Some evidence. Journal of Finance 32 (2): 33-54.

White, H. 1980. A heteroscedasticity-consistent covariance matrix and a direct test heteroscedasticity. Econometrica 48 (4): 817-838.

Zweibel, J. 1996. Dynamic capital structure under management entrenchment. American Economic Review 86 (5): 1197-1215.

Jeffrey A. Pittman is an Assistant Professor at the Memorial University of Newfoundland.

This paper is based on a chapter from my Ph.D. thesis completed at the University of Waterloo. I am indebted to my thesis committee, Alan V. S. Douglas, Ranjini Sivakumar, Tony Wirjanto, and especially my supervisor, Kenneth J. Klassen, for their guidance. I also thank Dan S. Dhaliwal for kindly consenting to be the external examiner for my thesis. This research has also benefited from comments from Frances L. Ayres (the editor), two anonymous reviewers, Anthony A. Atkinson, Sati P. Bandyopadhyay, Kate Bewley, Phelim Boyle, Shane Dikoli, Leonard G. Eckel, Glenn Feltham, Steve Fortin, Alan Macnaughton, Amin Mawani, Thomas Matthews, Susan A. McCracken, Gordon D. Richardson, Jay Ritter, Daniel B. Thornton, and participants at the 1999 AAA Annual Meeting, the 1998 CAAA Conference, and seminars at Memorial University of Newfoundland and the University of Waterloo. The financial support of the Deloitte & Touche Center for Taxation Education and Research, the Chartered Accountants' Education Foundation of Newfoundla nd, and the Society of Management Accountants of Canada is gratefully acknowledged. Sarah C. Mavrinac and John R. Graham generously provided IPO data and simulated marginal tax-rate data, respectively.

Printer friendly Cite/link Email Feedback | |

Author: | Pittman, Jeffrey A. |
---|---|

Publication: | Journal of the American Taxation Association |

Geographic Code: | 1USA |

Date: | Sep 22, 2002 |

Words: | 15211 |

Previous Article: | The impact of firm characteristics on book-tax-conforming and book-tax-difference audit adjustments. |

Next Article: | The impact of embedded intelligent agents on tax-reporting decisions. |

Topics: |