Interest rates, taxes and corporate financial policies.

INTRODUCTION

There has been a substantial literature examining both theoretically and empirically the effects of the tax structure on corporate use of debt vs. equity finance. According to the theoretical literature, debt finance should be encouraged to the degree that more taxes are saved on corporate interest deductions than are owed on the resulting interest income.

Most all past papers on taxes and corporate borrowing, though, ignore an important element in the theory. According to the theory, the size of the net tax savings per dollar of corporate debt is proportional to nominal interest rates. Everything else equal, therefore, the size of the tax incentives affecting use of corporate debt should be high in years such as the early 1980s in the U.S. when nominal interest rates were high (around 14 percent), and should be very low in the years immediately after 2000 when at least short-term nominal interest rates were around one percent.

If this variation in interest rates were statistically independent of the variation in tax rates, then prior estimates could still be unbiased, even if highly dependent on the sample period. However, we find below that years when corporate tax rates were high relative to personal tax rates, encouraging use of corporate debt, also tended to be years in which nominal interest rates were low, weakening the estimated effects of any tax incentives. The past neglect of variation in interest rates when estimating the sensitivity of use of corporate debt to tax incentives may be one explanation why this past literature has not found much effect of taxes on use of debt.

For similar reasons, the term structure of interest rates can affect a firm's choice of the maturity of its debt structure. This point is developed formally in Gordon (1982) and Brick and Ravid (1985). (1) In particular, to the degree that the long-term interest rate is higher than the short-term rate, the tax savings from use of long-term debt increase relative to the tax savings from an equivalent amount of short-term debt. These incentives are stronger the larger the tax differential.

There are a few past papers that attempted to take into account the interaction of taxes and market interest rates on use of corporate debt. For example, Gordon (1982) tested these predictions using aggregate time-series data for the U.S. While the point estimates were consistent with the theory, (2) standard errors were high based on 25 observations. There have also been a few attempts to test the effects of the term structure of interest rates on the maturity structure of debt using the largely cross-sectional data from Compustat (see Barclay and Smith (1995), Stohs and Mauer (1996), Guedes and Opler (1996), Harwood and Manzon (2000) and Newberry and Novack (1999)). Here the estimated effect of tax incentives is commonly statistically insignificant or the wrong sign, given the theory. These studies using Compustat data suffer from the problem that the proxies for the corporate tax rate can be picking up important non-tax effects on behavior.

The aim of this paper is to reexamine the combined effects of interest rates and taxes on corporate debt and debt maturity, using the U.S. Statistics of Income (SOI) Corporate Income Tax Returns, building on the identification strategies developed in Gordon and Lee (2001).

As reported below, we find that estimated tax effects are large and significant statistically, in contrast to the results in most previous papers. In particular, reducing the tax rate on corporate income by ten percentage points is forecast to reduce the fraction of capital financed with debt by three percentage points, given average interest rates. (3) The results also show important effects of the level of nominal interest rates on the overall use of debt: given average tax incentives, we estimate that an increase in market interest rates by five percentage points should increase the fraction of capital financed with debt by 5.4 percentage points. In addition, we report evidence that the term structure of interest rates has smaller but statistically significant effects on the maturity structure of corporate debt.

The rest of the paper is organized as follows. The second section briefly examines the hypotheses and develops the empirical strategy. The data are described in the third section, and regression results are reported in the fourth section. The paper concludes with a brief summary and discussion.

THEORY AND SPECIFICATION

The theoretical forecasts for the combined effects of interest rates and taxes on a corporation's choice of an overall debt level, and of the term structure on the choice of long-term vs. short-term debt, have been laid out in the past (see, e.g., Gordon (1982) and Brick and Ravid (1985)). In order to motivate the particular empirical specification we use, however, it is helpful to summarize explicitly the nature of the theoretical argument.

Consider the financial decisions made by a corporation. Denote its total debt by D, its long-term debt by [D.sub.L], and its short-term debt by [D.sub.s]. Denote the current short-term nominal interest rate by [r.sub.s] and the current long-term nominal interest rate by [r.sub.L]. Owners of long-term debt also receive an ex-post capital gain (loss if negative) of [??] in a given time period.

We examine first the incentives faced by the "marginal" shareholder whose preferences determine the pricing of corporate equity. (4) Assume that this "marginal" shareholder has a personal tax rate of m, while the marginal tax rate on corporate income (including any subsequent personal taxes for this shareholder) is denoted by [tau]. By construction, this marginal investor is just indifferent between investing another dollar in the firm's equity and investing it instead in either long-term or short-term bonds. In particular, the investor's marginal indifference between long-term and short-term bonds implies that

[1] [(1--m)[r.sub.L] + [(1- [t.sub.g])[c.sub.g] = [(1 - m)[r.sub.s],

where [C.sub.g] denotes the equilibrium certainty--equivalent (pretax) income generated by the random capital gain on long-term debt, while [t.sub.g] is the capital--gains tax rate on this random return.

We assume that the objective of the firm's managers when making financial decisions is to maximize the value of the firm. The specific financial choices we focus on are the fractions of the firm's capital to finance with long-term and short-term debt, denoted by [d.sub.L] and [d.sub.s] respectively. When choosing the optimal use of long-term and short-term debt finance, the firm trades off the resulting tax savings/dissavings with any non-tax costs arising from a use of debt different from that which would be chosen ignoring tax considerations. Let these non-tax costs be denoted by C([d.sub.L],[d.sub.s])K, where K denotes the firm's capital stock. By making this function proportional to K, we assume that financial choices will be the same regardless of the scale of the firm, though we relax this assumption in the empirical work below. We also assume that these non-tax costs are a convex function of both [d.sub.L] and [d.sub.s]. In particular, assume that [C.sub.LL] > 0 and [C.sub.ss] > 0. (5) Intuitively, firms face pressures to match the time pattern of their income and financial liabilities, to avoid having to come up with the funds to repay debt at a date when they are cash constrained and may be forced to sell non-liquid assets at a deep discount. Short-term debt would then be ideal to handle seasonal variation in expenses vs. revenue, whereas long-term debt is preferable in financing longer-lived assets. In addition, assume that the cost of adding more of one maturity of debt is higher the more debt the firm has of the other maturity: [C.sub.LS] > 0, with [C.sub.LS][C.sub.SL] < [C.sub.SS][C.sub.LL]. If all that matters is total debt, due, for example, to the risk of bankruptcy when total debt exceeds firm value, then [C.sub.LS] = [C.sub.LL] = [C.sub.SS]. (6)

The value of the firm will adjust to leave this marginal investor indifferent at the margin between investing further in equity and short-term bonds. (7) The certainty-equivalent nominal return from equity should, therefore, equal the return the investor could have received instead from investing the same funds in short-term bonds:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

where the numerator of the left-hand side is the after-tax income from equity, and the denominator V is the initial market value of equity. Here, [bar.f] (K)(1 - [rho]) is the certainty-equivalent pretax income to the firm for any given capital stock K, where [bar.f](K) is the expected return and [rho] captures the equilibrium marginal cost of risk bearing. The inflationary increase in the value of the firm's capital and the certainty-equivalent loss from bearing the capital-gains risk on long-term debt are not part of the corporate income tax base, so are taxed only at the shareholder's capital-gains tax rate, [t.sub.g]. Note that by definition V = qK(1 - [d.sub.L]-[d.sub.s]), where q equals the ratio of the market value to the book value of equity.

At the financial policies that maximize firm value, [partial derivative]q/[partial derivative][d.sub.L] = [partial derivative]q/[partial derivative][d.sub.s] = 0. (8) Using equation [1] to simplify, the resulting first-order conditions are:

[2a] ([tau] - m)[r.sub.L]/(1 - [tau]) = [C.sub.L]([d.sub.s], [d.sub.L])

and

[2b] ([tau] - m)[r.sub.s]/(1 - [tau]) = [C.sub.s]([d.sub.s], [d.sub.L]).

We then solve these two equations for the optimal values of [d.sub.s] and [d.sub.L], finding in general that