An Essay on the Effects of Taxation on

Agency decision

Ask Donna

What actually matters in this document.

Text

the Corporate Financial Policy

George Contos, Internal Revenue Service

he taxation of corporate profits in the United States

has been one of the most widely discussed issues

in the area of public finance. Corporate revenues

are currently subject to double taxation. Profits are taxed

first at the corporate level and then, when distributed

as dividends or when capital gains are realized, taxed

a second time at the individual level. The share of tax

revenues from corporate profits has been decreasing

steadily over the past four decades. In 1962, corporate

tax receipts accounted for 21 percent of all tax revenues,

but, by 2003, their share dropped to 7.5 percent.1 In 2003,

a proposal by the Bush Administration brought corporate

tax integration back to the front pages. The final legislation, the Jobs and Growth Tax Relief Reconciliation Act

of 2003, did not eliminate double taxation, but it did

reduce the taxation of corporate profits at the individual

level.2 Double taxation is still a reality; so, the discussion

for corporate integration is clearly not over.

In understanding why corporate taxation is such a

highly contested issue, critics argue that the current tax

system discourages business entities from organizing as

taxable corporations and encourages corporations to veer

from socially efficient decisions (Scholes et al. (2005),

p. 336). Those critics believe that the losses to the U.S.

economy caused by the current tax system far exceed the

gains from the revenues raised. They call for a neutral

tax system that does not enter into the decisionmaking

process of firms and does not distort economic efficiency.

Supporters of corporate taxation reply to those allegations by saying that corporations are distinct entities

and should be taxed separately from their shareholders;

that corporations should pay a fee, tax, for the special

privileges they enjoy; and that corporate taxation prevents the sheltering of individual income from taxation

(Rosen (2002), p. 399).

A large body of research has tested for the effects

of corporate taxation. Although the results of empirical

models vary significantly, all models agree that, to some

degree, corporate taxation affects a broad range of the

decisions made by taxable corporations. The magnitude

of those effects and their overall impact on the economy

are still under debate. Jane Gravelle (1995) divides the

debate on corporate taxation into three key issues. “First

who carries the burden of corporate tax--capital, labor,

or consumers, and does it play a role in a progressive

tax system? Second, how significant are the distortions

caused by the excess corporate tax? And third, how can

the revenues raised from corporate tax be replaced?”

This paper focuses on the second question and more

specifically on how the deductibility of interest affects

the capital structure of taxable corporations. I test the

hypothesis that taxable corporations have a tax incentive

to use debt financing versus equity financing because

interest paid is tax-deductible while dividends paid to

shareholders are not. Measuring the excess debt that

corporations carry due to the tax incentive is important

because the excessive use of debt may lead to financial

distress and even bankruptcy.

This paper extends the work of Gordon and Lee

(2001). They use an aggregate data time-series, Tax

Years 1950 to 1995, to test for the effects of corporate

taxation on the financial policy of firms of different sizes.

They found that taxes have a large effect on the use of

debt for the smallest and the largest firms. In this paper,

I first estimated the Gordon and Lee (G&L) model using the same aggregate Statistics of Income (SOI) data

but for a different time period, Tax Years 1993 to 2000,

and my findings were qualitatively similar to those of

G&L. Next, I introduced a confidential SOI firm-level

dataset for the 8-year period, and found an unexpected

negative relation between tax rates and debt. However,

using a marginal tax rate constructed from taxable income before the interest deduction and the panel dataset,

I found, as expected, a positive relation between tax

rates and debt. Finally, I divided my panel dataset into

small, intermediate, and large size firms, and I found a

positive relationship between tax rates and debt for all

three firm sizes.

- 95 -

Contos

 Corporate Taxation

Before discussing existing research on how taxes

affect the corporate capital structure, it is useful to

review how double taxation affects the decisionmaking process of firms. Business entities have a financial

incentive to organize as “C corporations,” where the

term C corporation comes from the subchapter of the

Tax Code defining their structure. Corporations are legal entities that can have multiple owners and separate

management. The ability to attract multiple investors

through the sale of shares or bonds gives corporations

broad access to capital and greater potential for growth.

The shares of corporations can be easily transferred to

other investors without disrupting the operations of

the companies. The owners of corporations also enjoy

limited liability since, in case of default, their liability

is limited to the amount they have invested. Because,

in the United States, corporate profits are subject to

double taxation, corporations in essence pay a fee for

the right to incorporate. Corporate revenues are taxed

first on the corporate level and then, when distributed

as dividends or when capital gains are realized, taxed

a second time on the individual level. Business entities

can avoid double taxation but in the process lose some

of the special privileges mentioned earlier, if they organize as passthrough entities. Passthrough entities, such

as sole proprietorships, partnerships, and subchapter

S corporations, avoid double taxation by passing all

profits and losses onto their shareholders (Brealey and

Myers, 2000).

debt to finance their investments, but empirical evidence shows that they use significant amounts of equity

capital.3 Why is this so? There can be significant nontax

costs involved with debt financing. These costs include

both the standard costs of borrowing and risks of financial distress that fixed liabilities imply. Firms fall into

financial distress when they have difficulty making their

debt payments. Extended periods of financial distress can

lead to bankruptcy. The higher the debt payment levels,

the higher the probability that the firm could fall into

financial distress. As the probability of distress increases

the risk for the firm’s debtor increases, so they demand

higher return for their investments. Consequently, the

value of debt tax shields decreases as these forms of

nontax costs increase.

The firm can finance its investments using equity

or debt. Equity is either cash available to the firm or

funds raised by issuing stock, primarily common stock.

Dividends paid to stockholders are not tax- deductible;

thus, dividends are paid from after-tax income. A firm

raises debt by borrowing from its shareholders, from

financial institutions, or from the public. All interest paid

by a corporation to its lenders is tax-deductible, thus

generating a tax shield. Clearly, there is a tax incentive

for a taxable corporation to use debt instead of equity.

So, double taxation directly affects the corporate capital

structure.

The value of tax shields also depends on the marginal

tax rate of the firm, and the availability of nondebt tax

shields4 and tax credits. The marginal tax rate is the tax

liability generated, today and in the future, by an additional dollar of income earned today. Estimating the

marginal tax rate is not straightforward because of the

uncertainty of future earnings, the carryback and the carryforward provisions of the tax law, and the alternative

minimum tax (AMT). Corporations can “carry back” and

“carry forward” operating losses and tax credits--meaning they can apply them to reduce tax liabilities incurred

in past or future years. As Graham (1996) explains,

the relationship among operating losses, marginal tax

rates, and the value of tax shields is not always obvious.

For example, tax shields have very low, if no, value to

corporations that expect operating losses in the future.

Such firms will have very low marginal tax rates because

they can use those net operating loss deductions (NOL’s)

in the future to refund any taxes paid today. Firms that

experienced losses in the past and expect moderate

profits in the future can also use NOL’s to reduce future

tax liabilities. However, if that same firm carries back

its current-year NOL and the NOL is less than or equal

to is past liabilities, then the marginal tax rate of any

additional income earned today will be equal to the

applicable statutory tax rate. From these examples, it is

easy to see that the NOL deduction makes estimating the

marginal tax rate of a corporation complex.

Since all interest paid is tax-deductible, one would

expect that taxable corporations would rely heavily on

The value of debt tax shields also depends on the

availability of nondebt tax shields4 and tax credits. As

- 96 -

An Essay on the Effects of Taxation on the Corporate Financial Policy

DeAngelo and Masulis (1980) explain, one can make the

case of a tax shield substitution effect since the availability of nondebt tax shields may crowd out debt tax

shields. Finally, it has been shown that the foreign tax

credit limitations do not just reduce the value of debt

tax shields, but actually influence U.S. multinationals

to decrease their domestic debts by substituting them

with equity financing.

In this paper, the corporate marginal tax rate proxies

are constructed by selecting the marginal statutory rate

that applies to the highest dollar of the current-year taxable income, or taxable income before interest deduction,

reported on the tax return. Such proxies have been used

successfully in earlier research and can be applied to both

the aggregate and firm-level datasets used. Upcoming

research by the author explores the effects of the NOL

deduction and the various tax credits on the corporate

capital structure.

 Prior Empirical Research

Modigliani and Miller (1963) were the first to introduce the idea that corporate taxation affects the capital

structure of firms. As Scholes et. al. (2005) discuss,

Modigliani and Miller showed that if the only imperfection of the capital markets is corporate taxation, the

deductibility of interest generates a debt tax shield that

increases the value of corporations. When comparing

debt and equity financing, Modigliani and Miller explain

that borrowing is beneficial to corporations because the

cost of debt, interest paid, is tax-deductible while the

cost of equity, dividends, is not. In a later paper, Miller

(1977) pointed out that, if one takes into account the tax

status of corporate investors, equity financing can be a

competitive alternative to debt financing. If the interest

earned by the debt holders is taxed at a higher rate than

the dividends paid to stockholders, then the corporation’s

tax incentive is the difference between the sum of the corporate tax rate plus the dividend rate, and the individual

tax rate of the bondholders. The work of Modigliani and

Miller was advanced by DeAngelo and Masulis (1980),

who introduced the idea of tax shield substitution. Firms

can substitute nondebt tax shields, like the depreciation

deduction, for debt tax shields. The work of DeAngelo

and Masulis is important because it led to a hypothesis

that can be empirically tested; firms with large amounts

of nondebt tax shields will have lower levels of debt

than firms with small amounts of nondebt tax shields

(Scholes et al. (2005) p. 344).

Since the works of Modigliani and Miller (1963)

and DeAngelo and Masulis (1980), a number of empirical studies have examined the impact taxes have on the

financial structure of corporations. As Ayers, Cloyd,

and Robinson (2001) explain, the capital structure

literature can be divided into two streams. The first

stream of works compares taxable corporations that

have different tax incentives, hypothesizing that firms

with greater tax incentives will have higher levels of

debt. The second stream of works compares taxable

corporations to passthrough entities that are not subject

to corporate taxation because, by law, they have to pass

all income to their shareholders. Their hypothesis is that

taxable corporations will have higher levels of debt than

passthrough entities.

The earlier articles of the first stream do cross-section analysis of taxable corporations but do not find

convincing evidence that taxation affects the financial

policy of firms (Bradley, Jarrell, and Kim, 1984; and

Gaver and Gaver, 1985). The more recent articles of the

first stream are more successful in finding evidence of a

significant positive relationship between debt financing

and marginal tax rates. These articles introduce several

improvements over earlier work: They examine incremental financing decisions instead of debt levels (MacKie-Mason (1990); Graham (1996); Gropp (1997)); they

develop better proxies for marginal tax rates (Graham

(1996); Graham, Lemmon, and Schallheim (1998)); they

use the ratio of interest expense to gross profit rather than

the debt-to-equity ratio as the dependent variable (Cloyd,

Limberg, and Robinson (1997); and they research the

debt policies of corporations of different sizes (Gordon

and Lee (1999)). Here, I briefly present an overview of

this work, focusing on the data, the marginal tax rate

proxies used, and their key findings.

Bradley, Jarrell, and Kim (1984) use data from 851

large firms to estimate a general equilibrium model.

Although they have multiyear data for each firm, in

order to avoid business cycle variations or different

- 97 -

Contos

adjustment periods, they calculate a 20-year average or

“permanent” leverage ratio for each firm. They examine how these ratios vary with the industry of the firm,

the volatility in the firm’s earnings, the availability of

nondebt tax shields, and the expenditures on research

and development and advertising. They do not find

concrete evidence that taxation affects the firm’s leverage ratios, but they find evidence that the leverage ratios

are strongly influenced by the firm’s industry. They also

find that firms with volatile earnings have lower levels of

debt, suggesting that the risk of bankruptcy has a negative effect on the amount a firm borrows. Finally, they

find that firms with higher levels of nondebt tax shields

borrow more, a finding that contradicts the findings of

the earlier literature. Bradley, Jarrell, and Kim offer as a

possible explanation for this last finding that firms with

large amounts of assets have more collateral and thus

can borrow more.

Graham (1996) follows MacKie-Mason’s incremental choice approach, using a simulated firm-specific

marginal tax rate as a proxy for the firm’s tax incentives.

The data used are a pooled cross-section of differenced

time series from about 10,000 Compustat firms from

1980 to 1992. Although he finds a strong positive relation between tax status and incremental debt policy, he is

puzzled by the low R-squared of about 5 percent that his

regressions produce. He states that "future researchers

should study why, given the strong tax incentives firms

have to issue debt, taxes do not explain a larger portion

of debt policy." Finally, he tests the effectiveness of the

tax status proxies used by earlier papers and finds that

only the net operating loss dummy variable is a reasonable proxy.5

The Gaver and Gaver (1985) article does not test

directly for the relationship between taxes and debt ratios

but rather tests the hypothesis that there is a systematic

relationship between the firm’s investment opportunity

set and its corporate policy decisions. Using longitudinal data from 237 new and 237 established firms, they

find evidence that growth firms have significantly lower

debt-to-equity ratios than established firms. This is an

interesting result that could explain the differences in

the debt levels across firms.

Gropp’s (1997) paper builds on the work done

by MacKie-Mason and Graham, but, instead of using proxies for expected marginal tax rates, he uses a

simple rational expectations approach to estimate the

expected effective corporate tax rates of firms. He finds

"that current average effective tax rates have substantial

predictive power for the estimation of expected corporate tax rates." Controlling for other theories of capital

structure choices, he finds that corporate taxation affects

the financial policy of firms using a balanced panel from

Compustat of 929 publicly traded manufacturing U. S.

firms from 1979 to 1991.

The MacKie-Mason (1990) article uses the Compustat data on large publicly traded companies to examine

the relationship between nondebt and debt tax shields

to measure the firm’s tax incentive, using a dummy

variable for the net operating loss deduction. Instead

of using the aggregate debt over total assets ratio as the

dependent variable, he uses the annual change in the

total debt levels scaled by the firm’s total assets. He finds

evidence of substantial tax effects on the choice between

issuing debt or equity; that firms with net operating loss

carry-forwards are much less likely to use debt; and

that the existence of investment tax credits reduces the

probability of debt issues only when the firm’s marginal

tax rate is near zero. His findings support a significant

relationship between corporate taxation and the financial

decisions of a firm.

Graham, Lemmon, and Schallheim (1998) is the

first paper to find a positive relationship between the

tax incentive and debt financing using debt levels.

They provide evidence that the corporate tax status is

endogenous to financing decisions, producing a spurious relationship between the debt ratio and the marginal

tax rate of the firm; in other words, the estimated effects

of tax status on the debt levels will be biased because

companies that have high levels of debt also have low

marginal tax rates. To solve this problem, they propose

a direct measure of the corporate marginal tax rate using taxable income before the interest deduction as a

measure of the firm profits. Using a balance panel from

Compustat of 18,193 observations from 1981 to 1992,

they find a positive relationship between tax rates and

the usage of debt.

- 98 -

An Essay on the Effects of Taxation on the Corporate Financial Policy

Gordon and Lee (2001) is the first paper to research

the debt policies of corporations of all sizes and to find

a positive relationship between debt levels and afterfinancing tax rates. They create a dataset from the

aggregate data on corporations published by SOI and

test for the effects of taxation by comparing the ratios

of debt-to-assets of firms in different asset size-classes.

Over the 46-year period covered by their data, the corporate tax rates varied significantly,6 giving them adequate

variation both across time and across firms for a difference-in-difference procedure. This procedure compares

the changes in the debt-to-assets ratios for small versus

large firms with the changes in the relative tax rates they

face. They find that taxes have a large effect on the use

of debt for the smallest and the largest firms. For intermediate-sized firms, they estimate a much lower effect,

but they provide indirect evidence that this finding is a

result of measurement error in the tax variable. Since

the SOI data are grouped in asset classes, they only have

information on the average rate of return for firms in

each asset class, taxable income divided by assets; so,

they calculate the average marginal tax rate for firms in

each asset class. Due to this limitation, "they are not able

to capture the effects of heterogeneity in rates of return

across firms on the expected marginal tax rate, arising

from the nonlinearity in the tax structure." The effects of

heterogeneity in rates of return are more important for

intermediate firms since their "taxable incomes are near

the point where tax rates change dramatically."

To avoid such problems, I introduced a confidential

firm-level dataset of taxable corporations of all sizes, for

Tax Years 1993 to 2000. This dataset allowed studying

the effects of taxation on firms of all sizes, while capturing the heterogeneity in rates of return across firms. I

found an unexpected negative relation between tax rates

and debt. However, using a marginal tax rate constructed

from taxable income before the interest deduction, I

found the expected positive relation between tax rates

and debt. Next, I took advantage of the panel aspects of

the microdataset; by using fixed effects models, I controlled for the unobserved firm-specific effects and found

again a positive relation between taxation and debt. Finally, I divided the panel dataset into small, intermediate,

and large size firms, and I found a positive relationship

between tax rates and debt for all three firm sizes.

 Empirical Research

The data sample

The data used for this study are the firm-level data

collected by SOI and published on an aggregate basis

in the annual Corporate Source Book.7 The data come

from the tax returns of domestic corporations and foreign

corporations with U.S. business activities.8 The firmlevel data are confidential, although SOI employees--like

my self--can conduct analyses of the data and share the

results with outsiders subject to disclosure review by the

Internal Revenue Service (IRS).

I began my analysis with Tax Year 1993 since it is

the first year that three new tax brackets, for returns with

taxable income greater than 10 million dollars, came into

effect. The three brackets were introduced by the Tax

Relief Act of 1993 and give my time series additional

variation across firms compared to earlier years. I ended

my analysis with Tax Year 2000 because it is the last full

year before the recession that started in March of 2001.9

Tax receipts in Tax Year 2001 decreased significantly;

so, including these data would complicate the analysis

of my findings.10 During the 1993 to 2000 time period,

the corporate tax schedule remained unchanged; so, the

dataset provides significant variation across firms but

limited variation across time.

To create the panel, I limited my sample to companies that filed tax returns under the same Employer Identification Number (EIN) and were selected by the SOI

sampling process every tax year from 1993 to 2000.11 To

confine the data to nonfinancial firms with appreciable

business operations, I excluded all financial returns

because they follow different tax rules: 1120F filers

because SOI does not collect balance sheet information

from them; part-year returns which have tax periods

of 6 months or less; and all returns with total assets of

$10,000 or less because such firms are too small to help

the explanatory power of the empirical model. After

these exclusions, the panel consisted of 10,552 firms.

Constructing a “true” balanced panel of corporations

is complicated by the need to account and adjust for

mergers, acquisitions, and other changes to the structure

- 99 -

Contos

of each corporation in the sample. Given the difficulty

of this undertaking, and of analyzing firms undergoing

major changes, I decided to exclude from the panel all

companies for which total assets increased by more than

tenfold in a single year and all companies for which total

assets decreased by more than 90 percent between 1999

and 2000. The first criterion eliminates from the panel

corporations that have merged with or acquired another

business entity. The second criterion eliminates from

the panel corporations that are in financial distress and

will be going out of business in the near future.12 A total

of 60 records were dropped for these reasons, leaving a

“final” panel of 10,492 firms.

Apart from the large number of observations, the

SOI data offer several advantages over the financial data

used in the prior literature. The data collected by SOI

are reported by firms to the IRS when financial (book)

data are reported by corporations to their shareholders.13

As George Plesko (2004) points out, "differences in accounting rules for book and tax reporting purposes can

lead to differences in the amount of income reported

to shareholders and to the IRS." Mills, Newberry, and

Trautman (2002) find that book-tax income differences

grew throughout the 1990’s so that tax rates estimated

from book income will be wrong.14

Financial and tax data may also differ when a

parent corporation reports with its subsidiaries. For

financial purposes, a parent company must include in

the consolidation all domestic and foreign subsidiaries

which it owns by 50 percent or more. Under tax rules,

however, domestic subsidiaries must be 80-percent or

more owned to be included in the parent’s tax return, and

foreign subsidiaries cannot be consolidated. Since the

Compustat dataset reports financial consolidations and

does not separate foreign and domestic income, taxable

income could be inflated. The amount of debt reported

by some companies in their tax returns could be inflated

because they do not eliminate intercompany payables

and receivables. Mills, Newberry, and Trautman (2002)

report anecdotal feedback of such reporting, but, since

the dependent and the control variables of the empirical

model are ratios, the effects should be minimal.

Finally, another reason financial and tax data may

differ is off-balance sheet financing. Firms in the 1990’s

used special purpose entities to keep debt outside their

consolidated financial statements. Mills and Newberry

(2004) find "that these financial reporting effects occurred primarily during 1994-1999." So the financial

statements of large firms for that period could underreport both interest expense and debt and inflate taxable income. I believe that, overall, the use of tax data

improves the accuracy of my empirical work.

 Summary Statistics

In order to be able to compare my results using the

firm-level data with G&L results based on aggregate

data, I first present summary information of all variables

from the G&L sample and the present sample. As shown

in Table 1, the summary statistics of the two samples

match very well. The mean total debt-to-assets ratio is

about four percentage points higher in the present sample

compared to that of G&L, reflecting greater long-term

borrowing over prior decades. Looking at the asset side

of their balance sheets, firms in the two samples own

comparable amounts of depreciable property and land,

but firms in the present sample have higher amounts of

intangible assets.15 Finally, although the ratio of accounts

receivable to assets dropped by a little bit more than 3

percentage points, cash holdings increased by about 2

percentage points. In comparing the mean marginal rates

of the two datasets, it is obvious that, in recent years,

corporations have faced significantly lower statutory

corporate tax rates: Companies in the 1950 to 1995

period faced higher tax scales with top statutory rates

as high as 52 percent, while those in the 1993 to 2000

period faced significantly lower tax scales that topped

at 39 percent. The mean marginal tax rate (mrt) has decreased from 37.6 percent to 26.5 percent.16 In contrast,

the average yearly individual tax rate on interest faced

by individual taxpayers (ifmr) in the same two periods

was much more stable, slipping from 24.5 to 22.3.17 It

is clear that firms in the 1993 to 2000 period have considerably lower tax incentive (dmr) than firms in the

1950 to 1995 period.18

 Empirical Findings and Sensitivity

Analysis

I begin my empirical analysis by regressing the present aggregate sample. The first equation of the Gordon

- 100 -

An Essay on the Effects of Taxation on the Corporate Financial Policy

where tdr is the debt over asset ratio for firms in asset

class s at year t, rassts are the inflation- adjusted total

assets of firms in asset class s at year t , log(rassts) is

the ith order polynomial function of logged rassts, dmr

is the tax incentive of firms in asset class s at year t,

X is a matrix of the composition of the assets of firms

in asset class s at year t, and d t are Tax Year dummies.

The main hypothesis is that the coefficient of the tax incentive is positive. For the asset composition variables,

I expect that firms with higher depreciable assets, land,

and intangibles asset ratios will have higher debt-toasset ratios when firms with higher cash balances and

trade notes and accounts receivable will have lower

debt-to-asset ratios. A complete listing of the variables

is included in the appendix.

Table 1

Aggregate Data ¹

Sample Means and Standard Deviations of Variables

Variables

Corporate debt-asset

ratios

Total debt-to-assets

Short-term debt-toassets

Long-term debt-toassets

Tax rates

Marginal tax ratetaxable income

Marginal tax ratetaxable income plus

interest paid

Individual tax rate

Marginal tax rate

minus individual tax

rate

Corporate assets

Depreciable assets-toassets

Land-to-assets

Cash-to-assets

Intangible assets-toassets

Accounts receivable to-assets

Gordon & Lee

Present Study

1950 - 1995 ²

1993 – 2000 ³

Notation

Mean

Standard

Deviation

Mean

Standard

Deviation

Tdr

25.18

8.05

29.12

6.83

Sdr

9.45

4.07

10.33

3.22

Ldr

15.73

4.36

18.78

4.62

Mrt

37.57

13.15

26.48

9.74

Mrtint

37.97

12.81

27.80

9.86

Ifmr

24.49

2.36

22.26

1.00

Dmr

13.04

12.72

4.22

9.75

Dprr

20.79

6.32

21.17

7.09

Landr

Car

3.66

9.5

2.46

4.00

3.51

11.37

2.06

6.58

Intr

1.12

1.08

2.45

0.84

Arr

22.83

4.53

19.01

4.70

¹ Source: SOI Source Book, amounts are in dollars.

² From Gordon and Lee (1999)

³ Author’s tabulations

and Lee empirical model measures the effects of tax

incentive (dmr), nontax factors, firm unique characteristics, and the business environment on the firm’s total

debt-to-assets ratios.19 To simplify the model, G&L assume that all nontax factors that affect the corporate financial policy do not change over time or change in a way

that is uncorrelated with relative tax rates. To account

for those nontax factors, they use an "arbitrary function

that measures desired debt-to-assets ratios ignoring tax

incentives." In estimation, this arbitrary function is a seventh-order polynomial function of logged real assets.20

The unique characteristics of the firms in each asset class

are measured by the composition of the assets of those

firms. Finally, the business environment is captured by a

set of Tax Year dummies. Thus, the equation estimated is:

tdrst = ∑ α i log(rassts st ) i + β dmrst +

i =1

γ X st + ∑ δ t d t + ε st

t =1

(1)

Gordon & Lee use OLS to estimate the first equation, finding the effects of taxes on debt to be modest.

Because the marginal tax rate proxy is based on taxable

income, they are concerned with possible endogeneity

bias: a firm’s debt levels through the interest deduction

directly affect its taxable income. To correct this bias,

they construct an exogenous instrument, based on the

findings of Graham, Lemmon, and Schallheim (1998)

and re-estimate the model using Instrumental Variable

(IV). The instrument is the average tax rate faced by

all firms in each time period if the interest deduction is

added back to taxable income. Their IV coefficients are

not significantly different from their OLS, which G&L

attribute to high correlation of the instrument with the

marginal tax rate proxy.

The results of the OLS regressions for the present

and G&L samples are shown in Table 2. Like Gordon

and Lee, I find an unexpected negative relation between

tax rates and debt. I next controlled for the firms’ size

and asset composition by regressing the first equation,

resulting as expected in a positive tax coefficient. The

coefficients of the control variables, except for the ratio

of land-to-assets, had the expected signs and are significant at the 1-percent level. So, I found that the 1990’s

aggregate data produce the same results as the aggregate

data from 1950 to 1995.

G&L also estimate the effects on financial policy of

any factors that change over time. These factors are the

business cycle, the nominal interest rates, and the tax en-

- 101 -

Contos

incentive faced by corporations at year t, tb is the nominal interest rate measured by the 3-year Treasury bond

rate, dj is a business cycle proxy equal to the ratio of the

Dow Jones index over Gross Domestic Product, and d

 86 is a dummy capturing any omitted aspects of the

Tax Reform Act of 1986.

Table 2

Aggregate Data

Regression Results

Variables

Dmr

G&L

Present

G&L

Present

Tdr

Sdr

Ldr

-0.393**

(0.020)

-0.384

(0.065)

0.079**

(0.019)

1.853**

(0.355)

- 0.641**

(0.135)

-0.568**

(0.068)

0.085**

(0.009)

0.019**

(0.004)

-0.004**

(0.001)

0.002**

(0.00038)

0.320**

(0.058)

0.317

(0.254)

-0.437**

(0.087)

1.447**

(0.341)

0.078**

(0.038)

0.034**

(0.007)

-0.015**

(0.003)

-0.002**

(0.0002)

0.0006**

(0.0001)

-0.00003**

(0.000009)

0.127**

(0.027)

0.021**

(0.005)

-0.012**

(0.002)

-0.002**

(0.0002)

0.0005**

(0.00007)

-0.00002*

(0.00004)

-0.048**

(0.028)

0.013**

(0.005)

-0.003**

(0.002)

-0.0002**

(0.0001)

0.0002

(0.00007)

-0.0002

(0.00004)

0.663**

(0.122)

-1.271**

(0.307)

-0.223

(0.225)

0.578*

(0.409)

0.096**

(0.083)

-1.606**

(0.208)

-0.394**

(0.152)

0.251

(0.276)

0.567**

(0.092)

-0.335*

(0.231)

0.171

(0.169)

0.326*

(0.307)

Log(rassts)

Dprr

Landr

Car

Intr

Arr

Constant

Year

Dummies

Obs.

Adj Rsquared

25.572**

(1.289)

Yes

0.311

(0.018)

Yes

-0.027

(0.040)

20.992**

(2.187)

Yes

-0.823**

(0.166)

0.433**

(0.062)

Yes

-0.630**

(0.112)

0.370**

(0.042)

Yes

-0.193**

(0.124)

0.063**

(0.047)

Yes

434

0.433

0.246

434

0.972

0.98

0.974

0.988

Table 3 reports both the unexplained yearly variation reported by the G&L and the present samples. According to G&L, if the first equation fully accounts for

the effects of taxation on the corporate financial policy,

then the tax coefficient of the second equation should be

zero; they find that the tax coefficient is positive, large

in magnitude, and statistically significant. Because the

dependent variable of the second equation is measured

net of the estimated effects of taxes estimated in the

first equation, to get the complete effect of taxation,

they combine the two IV tax coefficients. They find that

large firms in the 1970’s would finance 9.2 percent of

their assets with debt relative to the smaller firms. Using

seven annual observations, my replication of the timeseries aggregate model showed no unexplained yearly

variation. So, for the present sample, the first equation

seems to capture the tax incentive in its entirety. This is

not totally unexpected since, in the 8 years of my time

series, both business cycle and the nominal interest rate

variables remained fairly constant when their sample

* and ** indicate significance levels at 5 percent and 1 percent. Standard errors in parenthesis.

Note: Following G&L, I stopped adding powers to the polynomial when the next higher power

Table 3

was statistically insignificant.

vironment. The dependent variable for the second equation is the coefficients of the time dummies estimated

on the first equation. Having already controlled for the

tax incentives, size of firm, and asset composition, the

coefficients of the time dummies capture the effects on

financial policy of these nontax factors. In addition, by

including in the second equation a yearly measure of

the tax incentive (dmr), G&L also test if they have adequately controlled for taxes on the first equation. If they

have done so, then the coefficient of the tax incentive

must be equal to zero. Thus, the equation estimated is:

(2)

Aggregate Data

Unexplained yearly variation

OLS Regression Results

Variables

Dmrt

Mrt

Ifmr

Dummy for

post 1986

Constant

Obs.

Adj. Rsquared

where δˆt are the coefficients of the Tax Year dummies

estimated by the first equation, dmr is the average tax

- 102 -

G&L

Present

0.264**

(0.094)

-0.232

(0.291)

0.504**

(0.148)

-4.546**

(1.485)

3.313**

(0.692)

0.191

(1.978)

0.84

0.001

(0.003)

0.015

(0.020)

-0.004

(0.044)

0.90

*and ** indicate significance levels at 5 percent

and 1 percent. Standard errors in parenthesis.

An Essay on the Effects of Taxation on the Corporate Financial Policy

period permits 37 annual observations and gains power

from a structural change in 1986, as well as several

economic cycle changes.

adjusted R-squared of the regression is 0.14 percent. So,

my model provides a better fit than earlier firm-level

studies but is still unexpectedly poor.

I now turn my attention to the balanced panel of

firm-level microdata. I began by regressing the first

equation on the final panel using OLS. The results of

these regressions are reported in the first two columns of

Table 4. The tax coefficient is significant at the 1-percent

level but negative, and it stayed negative even after I

controlled for the size of the firm and asset composition. The asset composition variables had the expected

signs, and their magnitudes are consistent with my

expectations and were statistically significant. Firms

with higher depreciable or intangible asset ratios have

higher debt-to-asset ratios, and firms with higher levels

of cash at hand and accounts and trade notes receivable

have lower debt-to-asset ratios. Finally, the land coefficient was again negative but significantly lower. The

Still not satisfied with the goodness of fit of the liner

model, I estimated a log-linear model,21 and the OLS

regression results are shown in the two last columns

of Table 4. The adjusted R-squared of the log-linear

regression was higher than the linear model, while the

sum of square errors was lower, suggesting a better

fit. In particular, the adjusted R-squared was now 0.2

percent, considerably higher than the ones reported by

similar firm-level studies. The tax coefficient was again

negative, and the asset composition variables had the

expected signs.

Table 4

OLS Regression Results

Variables

Dmr

Log(rassts)

Tdr

Log(tdr)

-0.821**

(0.009)

-0.581**

(0.006)

-0.381**

(0.008)

-8.079**

(0.735)

0.960**

(0.093)

-0.055**

(0.006)

0.002**

(0.0001)

-0.00002**

(0.000002)

0.272**

(0.005)

-0.028**

(0.010)

-0.384**

(0.008)

0.363**

(0.020)

-0.087**

(0.006)

26.654**

(2.294)

Yes

-0.240**

(0.005)

-4.417**

(0.436)

0.532**

(0.055)

-0.031**

(0.003)

0.0009**

(0.0001)

-0.000005

(0.000001)

0.263**

(0.004)

-0.038**

(0.007)

-0.411**

(0.006)

0.304**

(0.014)

-0.098**

(0.005)

14.353

(1.362)

Yes

83,936

0.09

83,936

0.14

83,936

0.09

83,936

0.20

Dprr

Landr

Car

Intr

Arr

Constant

Year

Dummies

Obs.

R-squared

*and ** indicate significance levels at 5 percent and 1 percent. Standard

errors in parenthesis.

Note:

The final panel includes 10,492 nonfinancial companies that filed

tax returns under the same EIN and were selected by the SOI sampling

process every tax year from 1993 to 2000 and their total assets did not

increase by more than 10 times from one period to the next and did not file

final returns in Tax Year 2000. Following G&L, I stopped adding powers

to the polynomial when the next higher power was statistically insignificant.

I next took advantage of the panel aspects of my

dataset by using fixed effects.22 Fixed effects allow us

to isolate the unobserved firm-specific effects and get

a better measure of the true effects of taxation on the

financial policy of firms. By unobserved firm-specific

effects, I refer to all those firm-unique characteristics

that do not change from year to year and help shape the

firm’s financial policy and capital structure. As shown

in Table 5, the relationship between the tax incentive

and debt-to-asset ratios is again negative. The tax coefficient when total debt is the dependent variable was

–0.115, while the coefficients of the asset composition

variables have the expected signs and, except for the

ratio of land-to-assets, were statistically significant.

The tax coefficient was negative even when I divided

debt into short-term and long-term, –0.057 and –0.065,

respectively. The overall R-squared of the total, short,

and long-term debt regressions were 0.14 percent, 0.016

percent, and .2 percent, respectively.

To test whether the tax coefficients are driven by the

presence in my sample of a significant number of firms

with no taxable income, I regressed the first equation using two subsets of the final panel. In the first, the sample

was limited to 8,900 firms that had a positive marginal

tax rate for at least 1 year. Here again, the fixed effects

tax coefficient was negative and significant. Next, the

sample is further restricted to the 3,100 companies that

had a positive marginal tax rate every year; the coefficient remained negative and significant. Both datasets

produced the expected signs for all control variables,

- 103 -

Contos

instrument is the average tax rate faced by all firms in

each time period if the interest deduction is added back

to taxable income but the instrumental variable tax coefficient is again negative.

Table 5

Fixed Effects Regression Results

Variables

Log(dmr)

Log(rassts)

Log(dprr)

Log(landr)

Log(car)

Log(intr)

Log(arr)

Constant

Year

Dummies

Obs.

R-squared

Log(tdr)

Log(sdr)

Log(ldr)

-0.115**

(0.004)

-2.432**

(0.526)

0.285**

(0.067)

-0.016**

(0.004)

0.0005**

(0.0002)

-0.00001**

(0.000003)

0.267**

(0.007)

0.145**

(0.013)

-0.108**

(0.006)

0.310**

(0.015)

-0.058**

(0.007)

8.148**

(1.621)

Yes

-0.057**

(0.003)

-2.202**

(0.409)

0.242**

(0.052)

-0.013**

(0.003)

0.0003**

(0.0001)

-0.00001**

(0.000002)

0.034**

(0.005)

-0.005

(0.011)

-0.076**

(0.005)

-0.018**

(0.012)

-0.021**

(0.005)

7.850**

(1.260)

Yes

-0.065**

(0.003)

-0.503**

(0.474)

0.073**

(0.060)

-0.005*

(0.004)

0.0002*

(0.0002)

-0.000003*

(0.000002)

0.251**

(0.006)

0.154**

(0.012)

-0.038**

(0.005)

0.344**

(0.014)

-0.040**

(0.005)

1.249

(1.461)

Yes

83,936

0.14

83,936

0.014

83,936

0.20

Since the instrument does not seem to correct the

bias, I followed the example of Graham, Lemmon, and

Schallheim and generated a second marginal tax rate

proxy (mrtint) using taxable income before the interest deduction as a measure of the profits. I proceeded to

estimate the log-linear models using fixed effects. Table 6

reports the results of these regressions. The fixed effects

tax coefficients of all three regressions are positive and

significant at the 1-percent level. The tax coefficient,

for the total debt regression, was equal to 0.06. So, after

using a modified measure of revenue, one that includes

the interest deduction, I found a significant distortion

on the corporate financial policy caused by taxation. I

estimated that firms in the 39-percent tax bracket are

*and ** indicate significance levels at 5 percent and 1 percent.

Standard errors in parenthesis.

Note: The final panel includes 10,492 nonfinancial companies

Table 6

that filed tax returns under the same EIN and were selected by the

SOI sampling process every tax year from 1993 to 2000 and their

Fixed Effects Regression Results

total assets did not increase by more than 10 times from one period

to the next and did not file final returns in Tax Year 2000.

Variables

Log(dmrtint)

Following G&L, I stopped adding powers to the polynomial when

the next higher power was statistically insignificant.

Log(rassts)

and the same or higher overall R-squared as the final

panel did.23

Log(rassts)2

To test whether the negative tax coefficient related

to the companies with extreme observations, I excluded

from my sample firms that had total debt greater than

80 percent of total assets or firms that had any single

asset equal to or greater than total assets. After these

restrictions, my sample was reduced down to about 9,000

records. The tax coefficient was again negative and significant, with the rest of the control variables having the

expected signs. Excluding those extreme observations

reduced significantly the unobserved firm-specific error

and raised the overall R-squared to 0.2 percent.

Log(rassts)5

Since the negative relationship between taxes and

capital structure seemed to be independent of the dependent variable and the sample, I turned my attention to the

possibility of endogeneity bias between the dependent

variable and the main regressor.24To correct the possible bias, I constructed an exogenous instrument. The

- 104 -

Log(rassts)3

Log(rassts)4

Log(dprr)

Log(landr)

Log(car)

Log(intr)

Log(arr)

Constant

Year

Dummies

Obs.

R-squared

Log(tdr)

Log(sdr)

Log(ldr)

0.058**

(0.006)

-1.831**

(0.530)

0.213**

(0.067)

-0.012**

(0.004)

0.0003**

(0.0001)

-0.000003

(0.000002)

0.274**

(0.007)

0.156**

(0.014)

-0.130**

(0.006)

0.320**

(0.016)

-0.069**

(0.007)

6.269**

(1.633)

Yes

0.014**

(0.004)

-1.974**

(0.410)

0.215**

(0.052)

-0.011**

(0.003)

0.0003**

(0.00009)

-0.000003

(0.000002)

0.038**

(0.005)

0.010*

(0.011)

-0.086**

(0.005)

-0.013**

(0.012)

-0.027**

(0.005)

7.139**

(1.265)

Yes

0.049**

(0.005)

-0.344*

(0.116)

-0.032*

(0.011)

0.001*

(0.0004)

-0.00002*

(0.000007)

83,936

0.13

83,936

0.01

83,936

0.20

0.256**

(0.006)

0.160**

(0.013)

-0.051**

(0.005)

0.350**

(0.014)

-0.047**

(0.006)

-0.043

(1.467)

Yes

*and ** indicate significance levels at 5 percent and 1 percent.

Standard errors in parenthesis.

Note: The final panel includes 10,492 nonfinancial companies

that filed tax returns under the same EIN and were selected by the

SOI sampling process every tax year from 1993 to 2000 and their

total assets did not increase by more than 10 times from one period

to the next and did not file final returns in Tax Year 2000.

Following G&L, I stopped adding powers to the polynomial when

the next higher power was statistically insignificant.

An Essay on the Effects of Taxation on the Corporate Financial Policy

forecasted to finance 1.5 percent more of their assets

with debt than firms in the 15-percent tax bracket. Firms

in the top tax bracket, large firms, are forecasted to

finance 1.2 percent more of their assets with debt than

small firms. The coefficients of the asset composition

variables have the expected signs and are significant at

the 1-percent level.

Dividing debt into short-term and long-term also

produces very interesting results. The tax coefficient

of the long-term debt regression is greater than the tax

coefficient of the short-term regression, 0.049 compared

to 0.013. These coefficients are drastically different from

the aggregate data coefficients presented in Table 2.

The coefficients of the asset composition variables for

both the short-term and long-term regressions have the

expected signs and are statistically significant, except

for the land and intangible assets coefficients of the

short-term regression that are statistically insignificant.25

Firms with higher depreciable assets have higher longterm debt-to-assets ratios compared to their short-term

debt ratios. Firms with higher ratios of cash-to-assets

have higher short-term debt-to-assets ratios compared

to their long-term debt ratios.

impact on the tax liability of the larger firms. If interest

paid was not tax-deductible, then the 7 cents of additional

tax liability for large firms would have been 10 cents.

These findings are not surprising, since large firms hold

more debt, but they give us a measure of the importance

of the interest deduction as a tax shield.

The fixed effects regression results of the log-linear

model for separate asset-sized classes are reported in

Table 7. The dependent variable for the fixed effects

regression is the marginal tax rate based on taxable

income before the interest deduction (mrtint).27 The

estimated tax coefficients are: 0.057 for small firms,

0.055 for intermediate firms, and 0.085 for large firms.

So, I found evidence of a positive relationship between

taxation and corporate debt for all three types of firms.

Contrary to the G&L findings, taxes had the largest effect on the use of debt for the largest firms, and the tax

effect for intermediate firms is comparable to the tax

effect for small firms. The coefficients of the majority of

the control variables had the anticipated sign and were

statistically significant.

To get a better understanding of the effects of taxation on the financial policy of firms of different size, I

divide my sample into small, intermediate, and large

firms.26 Small firms have lower debt-to-asset ratios than

the rest of the firms, 26 percent of total assets compared

to 31 percent for intermediate and large firms. The majority of that debt for all three categories is long-term debt,

but, for small firms, long-term debt is a lower percentage

of total debt. Large firms have the highest combined ratio

of depreciable and intangible assets, with intermediate

firms being a close second. The amount of cash firms

hold is inversely related to their sizes. Firms in the lowest asset class hold more than one fifth of their assets

in cash, while firms in the highest asset class hold only

about 6 percent of their assets in cash. The progressiveness of the tax system is evident in both marginal tax rate

proxies. The average marginal tax rates, for both proxies,

increase as the asset classes rise. An additional dollar of

taxable income increases the tax liability of large firms

by more than 7 cents, 22.7 percent, whereas an additional

dollar of taxable income increases that of small firms by

15.8 percent. The interest paid deduction has the highest

- 105 -

Table 7

Fixed Effects Regression Results

$1 under

$10,000,000

Log(tdr)

Variables

Log(drtint)

Log(rassts)

Log(dprr)

Log(landr)

Log(car)

Log(intr)

Log(arr)

Constant

Obs.

R-squared

0.057**

(0.007)

-0.422**

(0.101)

0.029**

(0.007)

-0.0006**

(0.0002)

0.292**

(0.008)

0.156**

(0.016)

-0.134**

(0.007)

0.378**

(0.024)

-0.095**

(0.008)

2.113

(0.444)

54,024

0.17

$10,000,000

under

$100,000,000

or more

Log(tdr)

0.055**

(0.014)

-2.807**

(0.514)

0.158**

(0.031)

-0.003**

(0.0006)

0.268**

(0.013)

0.192**

(0.031)

-0.108**

(0.012)

0.307**

(0.026)

0.037**

(0.014)

16.564**

(2.883)

21,360

0.09

0.085**

(0.036)

-0.826**

(0.159)

0.042**

(0.009)

-0.0007**

(0.0001)

0.144**

(0.021)

0.118**

(0.058)

-0.190**

(0.022)

0.232**

(0.027)

-0.050**

(0.021)

-5.479**

(1.008)

8,552

0.10

*and ** indicate significance levels at 5 percent and 1 percent. Standard

errors in parenthesis.

Note: Following G&L, I stopped adding powers to the polynomial when

the next higher power was statistically insignificant.

Contos

Next, I divided debt into short-term and long-term,

and I re-estimated the model. All tax coefficients were

positive and statistically significant. The effect of taxation on the long-term debt of small firms was large when

the effect on short-term debt was very small. The opposite was true for large firms, where the effect of taxation

on short-term debt was approximately two times the

effect on long-term debt. Finally, the effects of taxation

on short-term and long-term debt for intermediate firms

were approximately the same. I believe that these finding can be supported by intuition. Although small firms

have relatively less long-term debt than intermediate and

large firms, this debt doubles as debt tax shield. Large

firms have more mature capital structures; they follow

debt target level for their long-term borrowing and use

short-term borrowing to create tax shields as needed.

Summarizing my findings, I found evidence of a positive

relationship between corporate taxation and the total debt

ratios of small, intermediate, and large firms.

When using a firm-level dataset, and after isolating

the unobserved firm-specific effects and using a modified measure of revenue, my findings suggest that there

is a positive relationship between taxation and the use

of corporate debt. Contrary to the G&L findings, taxes

have the largest effect on the use of debt for the largest

firms and a positive effect on the use of debt for intermediate firms.

Appendix

Definitions of Variables and Expected

Signs

Dependent Variables

Tdr Ratio of total debt to total assets. Measures total

debt as a percentage of total assets. Total debt

is equal to the sum of mortgages, notes, bonds

payable (Form 1120, page 4 balance sheet, lines

17 and 20).

 Conclusion

Past empirical research on the effects of taxation on

corporate financial policy has been limited, due to lack

of data, to large publicly-traded firms or small closelyheld partnerships. The more recent studies of the capital

structure literature find a positive relationship between

taxation and the debt levels of those firms. The only

work that looks at the entire corporate population is a

study by Gordon and Lee. They utilized an aggregate

time-series dataset from 1950 to 1995 to find evidence

that taxation increases the use of debt. In this study, I

used the SOI aggregate and microdata files to research

the effects of taxation on the corporate financial policy

from Tax Years 1993 to 2000.

When using the aggregate dataset, my findings suggest that taxation in the 1990’s still affected the financial

policy of firms but to a somewhat lesser extent. I found

that large firms in the 1990’s finance 1.4 percent more of

their assets with debt relative to the smaller firms. That

it is a significant decrease compared to the 9.2 percent

estimated by G&L. I believe that this decrease is in its

entirety due to the lower tax rates faced by all firms and

by the reduction in the gap between the tax rates faced

by small versus large firms.

Sdr Ratio of short-term to total assets. Measures

short-term debt as a percentage of total assets.

Short- term debt is equal to the sum of mortgages, notes, bonds payable in less than 1 year

(Form 1120, page 4 balance sheet, line 17).

Ldr Ratio of long-term to total assets. Measures

long-term debt as a percentage of total assets.

Long-term debt is equal to the sum of mortgages,

notes, bonds payable in 1 year or more (Form

1120, page 4 balance sheet, line 20).

Tax Variables

- 106 -

Dmr Equal to mrt minus ifmr. Measures the tax

incentive the firm has to use debt. (+)

Mrt Proxy for marginal rate using taxable income.

The rate is set equal to the marginal statutory

rate that applies to the highest dollar of taxable

income (Form 1120, page 1, line 30). The rate

is set to zero when taxable income is zero. (+)

An Essay on the Effects of Taxation on the Corporate Financial Policy

assets (amortizable only) less accumulated

amortization (Form 1120, page 4 balance

sheet, lines 13 a (c) and b (c)). (+)

Dmrtint Equal to mrtint minus ifmr. Measures the

tax incentive the firm has to use debt. (+)

Mrtint

Ifmr

Proxy for marginal rate using taxable income before the interest deduction. The rate

is set equal to the marginal statutory rate

that applies to the highest dollar of taxable

income before interest deduction (Form

1120, page 1, lines 30 and 18). The rate

is set to zero when taxable income before

interest deduction is zero. (+)

Proxy for yearly individual tax rate on interest income multiplied by the fraction of

household assets held outside of pensions

and life insurance. The yearly rate is the

weighted average marginal tax rate reported

in the SOI individual returns publication. (-)

Yearly Variables

Ydmr Yearly average of dmr.

Total assts (Form 1120, page 4 balance

sheet, line 15d) deflated by CPI. Real total

assets.

Dprr

Ratio of net depreciable assets to total assets. Net depreciable assets are equal to

buildings and other depreciable assets less

accumulated depreciation (Form 1120, page

4 balance sheet, lines 10 a (c) and b (c)). (+)

Landr

Ratio of cash to total assets (Form 1120,

page 4 balance sheet, line 1(d)). (-)

Arr

Ratio of trade notes and accounts receivable

to total assets. Trade notes and accounts

receivable are equal to trade notes and accounts receivable less allowance for bad

debts (Form 1120, page 4 balance sheet,

lines 2 a (c) and b (c)). (-)

Intr

Three-year Treasury Bill rate. Proxy for nominal interest rate.

Average Dow Jones index deflated by GDP.

Proxy for the business cycle.

Source: Congressional Budget Office Web site;

Table 3 Revenues by Major Source, 1962-2003.

Beginning in 2003, the maximum tax rates on

qualified dividends have been lowered to 15

percent from 39.6 percent. For sales and other

dispositions of property after May 5, 2003, the

maximum tax rates on net capital gains have been

lowered to 15 percent from 20 percent.

Although the ratios fluctuate from year to year,

firms relay primarily on internal generated cash

(retained earning plus depreciation) to finance new

investments. Industry averages show that the ratio

can range from 40 percent to 85 percent (Brealey

and Myers, 2000).

The most widely used nondebt tax shields in Tax

Year 2000 were: depreciation, compensation of

officers, employee benefit programs, advertising,

and contributions to pensions and profit-sharing

plans.

In a later paper (1996), he adds two more acceptable marginal tax rate proxies, a trichotomous

variable and the statutory marginal tax rate.

The top corporate tax rate for that time period

ranged from a high of 52 percent, from 1952 to

1963, to a low of 34 percent, from 1988 to 1992.

Ratio of land to total assets. Land is equal

to land net of any amortization (Form 1120,

page 4 balance sheet, line 12). (+)

Car

Proxy personal marginal tax rate.

 Endnotes

Control Variables

Rassts

Imr

Ratio of intangible assets to total assets.

Intangible assets are equal to intangible

- 107 -

Contos

The data are aggregated based on the end-of-year

total assets reported in the balance sheet by each

firm. For the studies used by Gordon and Lee, the

number of asset classes ranged between ten and

fourteen. For my dataset, there are eleven asset

classes. The breakdown of the asset classes is: (1

under 0.1m), (0.1m under 0.25m), (0.25m under

0.5m), (0.5m under 1m), (1m under 5m), (5m

under 10m), (10m under 25m), (25m under 50m),

(50m under 100m), (100m under 250m), (250m

or more), and (zero assets). The last asset class

groups returns that had no ending assets, and was

not used in my analysis.

The term domestic corporation refers to companies incorporated in the United States but does

not necessarily imply that all their activities are

domestic. For foreign corporations engaged in

trade or business in the United States, only income

that was considered effectively connected with the

conduct of a trade or business in the United States

was included in the statistics.

The Business Cycle Dating Committee of the National Bureau of Economic Research, November

26, 2001, reports that the longest expansion in

the NBER chronology reached its peak in March

of 2001.

Tax receipts are total income tax after credits

reported on Table 1 of the Corporate Income Tax

Returns Publication..

The sample selection process is set up in such a

manner that any firms selected into the sample in

a given year will be selected again the next year,

providing that the firm files a return using the

same employer identification number (EIN) in

the two years and that it falls into a stratum with

the same or higher sampling rate. Note that a firm

will usually change its EIN when it merges with

another firm. For more detailed explanation of the

sampling process, see Section 3 of the Corporate

Income Tax Returns Publication.

Financial reporting usually follows the generally

accepted accounting principles (GAAP) rules issued by the Financial Accounting Standards Board

(FASB).

The use of book data is an issue for all prior literature, Auerbach and Poterba (1987) review pre

TRA86 data and they report that the differences

between the tax and book amounts reported by

firms can be significant.

The intangible assets number maybe inflated by

the Internet bubble.

My findings are in line with the historical marginal tax

rates reported at the Tax Policy Center’s Web site.

Proxy for yearly individual tax rate multiplied

by the fraction of household assets held outside

of pensions and life insurance. The yearly rate is

the weighted average marginal tax rate reported

in the SOI individual returns publication.

I set the tax incentive as the simple difference

between the corporate marginal tax rate and the

individual tax rate on interest income. Other literature is investigating the tradeoff and how the

individual tax rate differences (dividends versus

interest versus capital gain rates) are affecting

capital structure, but this issue is beyond the scope

of this paper.

The total debt is the sum of mortgages, notes bonds

payable in less than 1 year and mortgages, notes

bonds payable in 1 year or more.

This is the only variable deflated using the Consumer Price Index (CPI); the rest of variables are

in current dollars.

To estimate the model, following the work of Gentry (1994), I transformed all dependent, tax, and

control variables by adding one to all observations.

I did so because those variables have observations

that are equal to zero. I also tried another model

with the log of the total debt ratio as the dependent variable, but the log-liner model consistently

produced the highest adjusted R-squared.

Such firms have unusually large amounts of debt

and no taxable income.

- 108 -

An Essay on the Effects of Taxation on the Corporate Financial Policy

Originally, I thought that, due to the large number

of observations in our panel, random effects may

be the better choice than fixed effects, but the

Hausman test rejected the random coefficients as

inconsistent.

Because for these regressions I dropped observations based on the magnitude of the dependent

variable, these results may be spuriously induced.

I also allowed for the possibility of dynamics of

adjustment of the debt-over-asset ratio by including in the right-hand side of the empirical model

a one-period lag of the ratios and estimating the

model using the method of Arellano and Bond.

The one-period lag coefficient was both positive

and significant with the tax incentive still having

a negative effect, but I found that the instrument

variables, dmr and dprr, were correlated to some

set of residuals and are not acceptable, and the

model failed the Sargan test of overidentifying

restrictions.

The time dummy coefficients for these regressions were statistically insignificant; so, I did not

estimate the second equation.

I decided against using the thirteen SOI asset

classes because their breakouts were too detailed.

My breakouts, based on yearend total assets are:

small firms, less than $10,000,000; intermediate

firms, $10,000,000 less than $100,000,000; and

large firms, $100,000,000 or more.

In order to retain the panel aspects of my datasets

and because firms over the eight years time-series

moved in and out of asset classes I assigned to all

eight observation of each firm the same asset class

based on the firms’ 1996 year-end total assets.

 References

Ayers, Benjamin C.; Cloyd, C. Bryan; and Robinson,

John R. (2001), “The Influence of Income Taxes

on the Use of Inside and Outside Debt By Small

Businesses,” National Tax Journal, Volume 54,

Number 1, pp. 27-55.

Auerbach, A. (1985), “Real Determinants of Corporate Leverage,” in Corporate Capital Structures in

the United States, edited by B. Friedman, University of Chicago Press, Chicago, pp. 125-143.

Auerbach, A. and Poterba, J. (1987), “Tax Loss Carryforwards and Corporate Tax Incentives,” in The

Effects of Taxation on Capital Accumulation, edited by Martin S. Feldstein, University of Chicago

Press, Chicago, pp. 305-337.

Bradley, M.; Jarrell, G.; and Kim, E. (1984), “On the

Existence of an Optimal Capital Structure: Theory

and Evidence,” Journal of Finance, Volume 39,

pp. 857-878.

Brealey, Richard A. and Myers, Stewart C. (2000),

“Principles of Corporate Finance, Sixth Edition,”

McGraw-Hill Irwin.

Carroll, Robert; Hassett, Kevin; Mackie III, James B.,

“The Effect of Dividend Tax Relief on Investment

Incentives,” National Tax Journal, Volume 56,

Number 3, pp. 629-651.

Cloyd, C. Bryan; Limberg, Steven T.; Robinson, John

R. (1997), The Impact of Federal Taxes on the

Debt-Equity Structure of Closely-Held Corporations, National Tax Journal, Volume 50, Number

2, pp. 261-277.

DeAngelo, Harry and Masulis, Ronald W. (1980),

“Optimal Capital Structure Under Corporate and

Personal Taxation,” Journal of Financial Economics, Volume 8, Number 1, pp. 3-29.

Feldstein, M. (1995), “The Effect of Marginal Tax

Rates on Taxable Income: A Panel Study of the

1986 Tax Reform Act,” The Journal of Political

Economy, Volume 103, Number 3, pp. 551-572.

Gaver, Jennifer J. and Gaver, Kenneth M. (1993),

“Additional Evidence on the Association between

the Investment Opportunity Set and Corporate

Financing, Dividend, and Compensation Policies,”

Journal of Accounting and Economics, Volume

16, Number 1/2/3, pp. 125-160.

- 109 -

Contos

Gentry, William M. (1994), “Taxes, Financial Decisions and Organizational Form,” Journal of Public

Economics, Volume 53, Number 2, pp. 223-244.

Gordon, Roger H., and Young Lee (2001), “Do Taxes

Affect Corporate Debt Policy? Evidence from

U.S. Corporate Tax Return Data,” Journal of Public Economics, Volume 82, pp. 195-224.

cal Society. Series B (Methological), Volume 54,

Number 3, pp. 761-771.

Hinkins, Susan; Jones, Homer; and Scheuren, Fritz

(1988), “Design Modifications for the SOI Corporate Sample: Balancing Multiple Objectives,”

1988 Proceedings of the Survey Research Methods Section, American Statistical Association.

Internal Revenue Service, Statistics of Income, Corporate Source Book, Washington DC, various issues.

Graham, John R. (1996), “Debt and the Marginal Tax

Rate,” Journal of Financial Economics, Volume

41, Number 1, pp. 41-73.

Graham, John R. (1996), “Proxies for the Corporate

Marginal Tax Rate,” Journal of Financial Economics, Volume 42, Number 1, pp. 187-221.

Internal Revenue Service, Statistics of Income, Corporation Income Tax Returns, Washington DC,

various issues.

Graham, John R. (1998), “Do Personal Taxes affect

Corporate Financing Decisions?”

Internal Revenue Service, Statistics of Income, Individual Income Tax Returns, Washington DC,

various issues.

Graham, John R. (2000), “How Big are the Tax Benefits of Debt,” Journal of Financial Economics,

Volume 55, Number 5, pp. 1901-1941.

MacKie-Mason, Jeffrey K. (1990), “Do Taxes Affect

Corporate Financing Decisions?,” The Journal of

Finance, Volume 45, Number 5, pp. 1471-1493.

Graham, John R.; Lemmon, Michael L.; Schallheim,

James S. (1998), “Debt, Leases, Taxes, and the

Endogeneity of Corporate Tax Status,” The Journal of Finance, Volume 53, Number 1, pp. 131-162.

MacKie-Mason, Jeffrey K. and Gordon, Roger H.

(1997), “How Much do Taxes Discourage Incorporation,” The Journal of Finance, Volume 52,

Number 2, pp. 477-505.

Gravelle, Jane and Kotlikoff, L. (1989), “The Incidence and Efficiency Costs of Corporate Taxation

When Corporate and Noncorporate Firms Produce

the Same Good,” The Journal of Political Economy, Volume 97, Number 4, pp. 749-780.

Miller, Merton H. (1977), “Debt and Taxes,’’ Journal

of Finance, Volume 32, pp. 261-275.

Gravelle, Jane (1995), “The Corporate Income Tax:

Economic Issues and Policy Options,” The National Tax Journal, Volume 48, Number 2, pp. 267-277.

Gropp, Reint E. (1997), “The Effect of Expected

Effective Corporate Tax Rates on Incremental Financing Decisions,” International Monetary Fund

Staff Paper, Volume 44, Number 4, pp. 485-509.

Hadi, Ali S. (1992), “Identifying Multiple Outliers in

Multivariate Data,” Journal of the Royal Statisti-

Mills, Lillian F.; Newberry, Kaye J.; Trautman, William B. (2002), “Trends in Book-Tax Income and

Balance Sheet Differences,” Tax Notes, August

19, pp. 1109-1124.

Mills, Lillian F. and Newberry, Kaye J. (2004),

“Firms’ Off-Balance Sheet Financing: Evidence

from Their Book-Tax Reporting Differences,”

working paper.

Modigliani, Franco and Miller, Merton H. (1963),

“Corporate Income Taxes and the Cost of Capital: A Correction,” American Economic Review,

Volume 53, Number 3, pp. 433-443.

- 110 -

An Essay on the Effects of Taxation on the Corporate Financial Policy

Newberry, Kaye J. (1998), “Foreign Tax Credit Limitations and Capital Structure Decisions,” Journal

of Accounting Research Volume 36, Number 1,

pp. 157-166.

Plesko, George A. (2004), “Reconciling Corporate

Book and Tax Net Income, Tax Years 1995-2001,”

Data Release

Omer, Thomas C. and Terrando, William (1999), “The

Effect of Risk and Tax Differences on Corporate and Limited Partnership Capital Structure,”

National Tax Journal, Volume 52, Number 4, pp.

699-715.

Plesko, George A. (2002), “Reconciling Corporate

Book and Tax Net Income, Tax Years 1996-1998,”

SOI Bulletin (Spring 2002), pp.1-16.

Rosen, Harvey S. (2002), “Public Finance, Sixth Edition,” McGraw-Hill Irwin.

Scholes, Myron S. and Wolfson, Mark A. (1992),

“Taxes and Business Strategy,” Prentice-Hall,

Englewood Cliffs, NJ.

Scholes, Myron S.; Wolfson, Mark A.; Erickson,

Merle; Maydew, Edward L.; and Shevlin, Terry

(2005), “Taxes and Business Strategy, A Planning

Approach,” Prentice-Hall, Upper Saddle River, NJ.

- 111 -

This is a copy of a public record, reproduced as it was published. It is not legal advice, and it may not be the version a court would rely on. Check the official source before you cite it.