Document Sample
BHS Powered By Docstoc
					                  Debt and Taxes: Evidence from REITs

                                  Michael J. Barclay

                                    Shane Heitzman

                                  Clifford W. Smith*

                               This version: April 1, 2009


Empirical evidence on the effect of corporate taxes on leverage decisions is limited, in
part because of the difficulty in obtaining an effective proxy for the firm’s tax benefit of
debt. In this paper, we compare the leverage of Real Estate Investment Trusts to that of
taxable real estate firms and other industrial companies. The tax hypothesis implies that
REITs – where the tax benefit of debt is zero – should have less debt than their taxable
counterparts. We find little evidence that taxes are an important determinant of leverage;
leverage ratios for non-taxable and taxable real estate companies are not reliably

*Simon School of Business, University of Rochester.
 We thank Dan Dhaliwal, Dave Gardner, Michelle Hanlon, Anzhela Knyazeva, Diana
Knyazeva, Rob Luken, Sanjog Misra, Minjae Song, Jerry Warner and Wei Yang for
comments and suggestions.
                  Debt and Taxes: Evidence from REITs

       The central question in corporate finance concerns the firm’s optimal capital
structure. Given the capital required to support a company’s activities, how can that
capital be divided between debt and equity to maximize firm value? And this
immediately leads to the next question: What are the important factors in determining this
optimal leverage for a given firm?
       Modigliani and Miller (1958) provide the fundamental contribution to our
understanding of corporate capital structure. They lay the foundation for a positive theory
of capital structure by developing implications of market equilibrium for an optimal
capital structure. Their analysis rigorously demonstrates that, given the firm’s investment
policy and facing no taxes or contracting costs, the firm’s choice of financing policy has
no effect on its current market value. This M&M proposition has focused the profession’s
attention on those factors that are potentially important in determining optimal leverage.
If corporate financing decisions do affect firm value in predictable ways, they do so for at
least one of the following reasons: (1) they affect taxes paid by issuers or investors, (2)
they affect the probability―and associated expected costs―of financial difficulty, (3)
they affect management’s incentives to follow the value-maximizing rule of undertaking
all positive NPV projects, or (4) they lower information costs by providing credible
signals to investors of management’s confidence about the firm’s future cash flows.
       In the development of our understanding of capital structure choices, taxes have
played a prominent role. But despite substantial effort devoted to understanding the
implications of taxes on leverage, little empirical validation has been produced. Myers
(1984 p. 588) concludes, ―I know of no study clearly demonstrating that a firm’s tax
status has predictable, material effects on its debt policy.‖ This lack of compelling
evidence is generally explained by some combination of considerations: (1) Our theories
lead to tests with limited power because they yield only qualitative predictions. (2) A
firm’s tax returns are confidential and thus we have poor proxies for its effective tax
status. (3) Many debt issues have long stated maturities and thus expected future tax rates
are important. (4) Many instruments for tax status are potentially correlated with other

firm characteristics (for instance, firms with net operating loss carryforwards frequently
are also experiencing financial distress). (5) The corporate tax benefit of leverage is
offset by the personal tax disadvantages of debt and these offsetting tax effects are
difficult to disentangle. (6) There is limited time-series variation in the structure of the
United States tax code.
         In this paper, we extend the literature on taxes and leverage by examining the
capital structures of a set of firms where problems with measuring the effective marginal
tax rate over the life of a debt instrument is straightforward—we analyze real estate
investment trusts.1 If a REIT distributes at least 90 percent of its earnings to shareholders
each year, it avoids entity-level taxation. We know that the effective marginal corporate
tax rate—both currently and over the life of a debt issue—is zero. Theory implies that
these firms should have a comparative disadvantage in issuing debt and therefore should
select a low level for target leverage. By focusing on this special set of firms, we also
minimize potential measurement-error problems with the tax proxies that other studies
have faced.2 Finally, our analysis of the tax effects of leverage exploits heretofore
neglected data and thus our results are largely independent of the available evidence.
Therefore, our analysis of REITs should provide valuable evidence on the importance of
corporate tax status in corporate leverage choices.
         We find that when compared to a broad sample of Compustat industrial firms,
REITS choose substantially higher leverage. This simple summary statistic is inconsistent
with the notion that the corporate tax benefit of debt is a fundamentally important
determinant of capital structure. When we compare leverage of REITS to that of taxable
real estate firms with similar investment opportunities, the difference generally is
insignificant. Therefore, even with our focus on a sample of firms for which we have
precise measures of tax benefits, results consistent with the tax hypothesis are quite

  Throughout the paper, we use the terms REIT and non-taxable real estate firm interchangeably.
  There are potentially important proxy effects in some of the instruments. For instance, investment tax
credits may proxy for the firm’s asset characteristics, as well as tax status. Moreover, there are potentially
important time-series aspects of the problem that generally have been ignored—for instance, forecasts of
future expected marginal tax rates over the life of a debt issue. But our results for REITs appear insensitive
to such measurement problems: REIT tax rates are zero.

Overview. In section 2, we review the theory of and evidence on the impact of taxes on
the firm’s capital structure. We describe our data in section 3. In section 4, we estimate
standard regressions with leverage as the dependent variable. Other factors that affect a
firm’s target leverage choice—its investment opportunity set, size, regulatory status, and
asset tangibility—are independent variables; we also add dummy variables for tax status.
We analyze the robustness of this benchmark regression by examining different time
periods, different measures of leverage, and different real estate asset types. Our analysis
suggests that differences in leverage between taxable and non-taxable real estate firms are
not reliably different. In section 5, we offer our conclusions.

2.1. Theory
         The basic theory is straightforward. A corporation’s taxable income is reduced
by interest payments, but not dividends. Thus, as Modigliani and Miller (1963) noted,
adding debt to a firm’s capital structure lowers its expected tax liability and thereby
increases its after-tax cashflow. If there were only corporate taxation and no individual
taxation of the returns from corporate securities, the value of a debt-financed company
would exceed that of an identical all-equity firm by an amount equal to the present value
of its interest tax shields.
         Miller (1977) argues this analysis potentially overstates the tax advantage of debt
by considering only corporate taxes. Many investors who receive interest income must
pay taxes on that income. But those same investors are taxed at lower rates when they
receive equity income. Thus, although raising leverage lowers the firm’s corporate taxes,
it also increases the aggregate taxes paid by its investors. Because investors care about
their after-tax returns, they require compensation for these increased personal taxes in the
form of higher yields on corporate debt.3

  Miller’s (1977) tax effects ultimately flow from an asymmetry in the tax treatment of ordinary income and
capital gains that sets up a compensating differential in pricing. But if prices are the transmission
mechanism, security prices depend on the tax status of the marginal investor. For non-corporate investors
in the pre-2004 period, the tax treatment of dividends and capital gains is symmetric for investments in
REITs and taxable real estate firms. Thus, any corporate tax advantage of debt for taxable real estate firms
should not be undone by personal tax considerations.

        The extent to which a company benefits from interest tax shields also depends on
whether it has other tax shields. For example, DeAngelo and Masulis (1980) argue that
holding all else equal, companies with more non-interest tax shields—such as investment
tax credits, foreign tax credits, and tax loss carryforwards—should have lower leverage to
reflect the reduced value of their interest tax shields.

2.2. Evidence
        A number of studies have attempted to identify the impact of taxes on financing
policy and firm value.
        Direct estimates of tax-shield value. Fama and French (1998) estimate the value
of the firm’s interest tax shields directly. They run regressions with the market value of
the firm’s assets as the dependent variable; independent variables include interest
expense and a set of control variables (earnings, assets, R&D spending) to proxy for the
value of an unlevered firm. To control for heteroskedasticity, they normalize all of these
variables by the book value of assets. They run a series of regressions using panel data
specified both in levels and first differences. Their estimates of the coefficient on the tax
variable is uniformly non-positive. Based on their evidence, they conclude that tax
benefits are an immaterial component of firm value.
        But the Fama and French regressions are difficult to interpret for at least two
reasons: First, there is important self-selection in leverage choices that underlies their
observations. If each firm has selected the level of leverage that maximizes its value,
their estimated coefficient does not indicate how much firm value would change if
leverage changes. Second, their regressions ultimately put the market-to-book ratio on
the left-hand side and the marginal tax rate times the debt-to-book asset ratio on the right-
hand side of their regression. Consider a firm where more of its market value reflects the
intangible assets associated with its future growth opportunities. Agency theory suggests
that such firms should use less debt. For such firms, market-to-book ratios tend to be
high. Stock analysts estimate and capitalize the value of these intangible assets;
accountants do not. Thus the Fama/French coefficient estimate includes a potential
simultaneous-equations bias.

       Dyreng and Graham (2007) attempt to circumvent these issues by examining
market reactions to a change in corporate tax law. In late 2006, the Canadian Minister of
Finance unexpectedly announced that tax-exempt Canadian Income Trusts would become
subject to corporate income tax beginning in 2011. The market value of these income
trusts fell by approximately 10% at the announcement. Consistent with the tax
hypothesis, the drop in firm value was smaller for firms with more debt in their capital
structure. Dyreng and Graham estimate that every dollar of debt contributes between
$0.16 and $0.39 to firm value. However, their estimates must be interpreted with care
since the market reaction to the announcement potentially reflects expectations about the
government’s economic policies as well as anticipated adjustments to the firm’s
financing, investment, and payout policies.
       Exchange offers. Masulis (1983) examines exchange offers in which one security
is issued and another simultaneously retired. He argues that these transactions hold
investment policy fixed and thus represent an ideal way to examine the implications of a
pure change in capital structure. He claims that the tax hypothesis implies that leverage-
increasing offers increase firm value because they increase corporate tax deductions. His
evidence appears supportive: leverage-increasing exchange offers increase the average
firm’s stock price by 10.1%; leverage-reducing offers decrease the price by 5.1%.
       However Smith (1986) argues that such offers provide little evidence on factors,
such as taxes, that determine a firm’s optimal capital structure. Rather, the price changes
reflect the market’s assessment of management’s private information about the firm and
its prospects. Because debt represents a more senior claim in bankruptcy, its value is less
sensitive to changes in a company’s prospects than is the value of common stock.
Suppose a manager wants to increase leverage through an exchange offer of debt for
equity and has non-public information that the firm’s future prospects look especially
strong. Making the offer immediately requires less debt to retire a given number of shares
than if the manager waits until this information is made public and the share price rises.
Conversely, if the manager’s private information suggests that future prospects look poor,
waiting to make the offer would require less debt to retire the same number of shares. In
an efficient market, sophisticated investors will revise their estimates of firm value if
management announces a transaction that potentially exploits their informational

advantage. Since announcing a leverage-increasing transaction is more likely when
future prospects are strong, the market’s reaction to the announcement tends to be
positive. Consequently the stock price reaction associated with the announcement
provides insights into the market’s assessment of manager’s private information, but little
useful information about the role of taxes in determining optimal capital structure.
        Reorganizations and spin-offs. Mehrotra, Mikkelson and Partch (2005) examine
the capital structures chosen by firms that announce spin-offs. In a spin-off, a substantial
fraction of the firm’s assets are transferred to a separate corporation. Spin-offs thus
provide a potentially valuable opportunity to investigate the influence of various asset
characteristics on the choice of capital structure. They focus on 98 spin-offs over the
period 1979 – 1997. They argue that because this new firm is created from whole cloth,
its capital structure should be free of historical path-dependent influences. Their results
suggest that capital structures primarily reflect three factors: asset tangibility, the level of
operating profits and the variability of operating profits. However they find no
significant impact of the firm’s tax circumstances on its leverage choice.
        Alderson and Betker (1995) examine the composition of capital structures for
firms that are reorganized under Chapter 11 of the United States bankruptcy code. These
firms’ capital structures also should be free of historical influences. They find that firms
which face high liquidation costs emerge with low leverage; their debt is more likely to
be public, unsecured, and include fewer restrictive covenants. In their analysis, they
consider the impact of net operating loss carryforwards, because non-debt tax shields
should be substitutes for interest tax shields. But they find no significant impact of NOLs
on leverage.
        Leverage regressions. Theoretical models of optimal capital structure predict that
companies with more taxable income and fewer non-debt tax shields should have higher
leverage ratios. Several studies run regressions where leverage is the dependent variable
and factors that affect leverage—like tax status—are used as dependent variables. But
the evidence from studies that examine the effect of non-debt tax shields (depreciation,
tax-loss carryforwards, and investment tax credits) on corporate leverage generally find
that companies with more non-debt tax shields appear to have, if anything, higher

leverage. (See Bradley, Jarrell and Kim, 1984; Titman and Wessels, 1988; Smith and
Watts, 1992; Barclay, Smith and Watts, 1995; Barclay and Smith, 2005.)
        In each of these studies, the measures used for tax status also proxy for firm
characteristics other than the company’s effective tax circumstances. For example,
companies with investment tax credits and high levels of depreciation tend to have more
tangible fixed assets. As such assets generally represent good collateral for debt, the non-
debt tax shields potentially serve not only as a proxy for the firm’s tax circumstances, but
also for low contracting costs associated with debt financing. Similarly, companies with
tax-loss carryforwards often experience financial distress. And because equity values
typically are lower in such circumstances, financial distress itself causes leverage ratios
to increase. Thus, it is not clear whether tax-loss carryforwards proxy for low tax
benefits of debt or for substantial financial distress—whether this regression coefficient
measures variation in target leverage or deviations from target leverage.
        Incremental financing decisions. Mackie-Mason (1990) focuses on incremental
financing choices—that is, changes in the amount of debt or equity—rather than on the
levels of debt and equity. He examines registered security offers by U.S. corporations
and finds that firms are more likely to issue debt if they face a high marginal tax rate and
to issue equity if they have a low tax rate. But lacking access to firms’ (confidential) tax
returns, he uses the existence of ITCs to proxy for a high tax rate and NOL carryforwards
to proxy for a low rate. Thus, his results potentially reflect the fact that the firms with
ITCs generate substantial free cashflow, regular additions of internal equity and thus,
when they sell public securities, they sell debt. Conversely, firms with NOL
carryforwards are financially distressed, have limited access to public debt markets, and
hence when they do access public capital markets, tend to sell equity to recapitalize the
        Tax status proxies. In an attempt to avoid the difficulties stemming from crude
proxy variables, Graham (1996) uses a sophisticated simulation method to provide a
more detailed measure of a company’s effective marginal tax rate. After simulating such
rates for thousands of companies, Graham (1999) reports a positive association between
corporate debt and tax rates. Graham (2000) estimates that for the average US company

during the 1980’s and the 1990’s, the tax benefits of debt accounted for 5-10% of firm
         But even this variable has potential problems. Graham takes great care in
estimating a firm’s effective marginal tax rate in a given year. Because of tax loss
carrybacks and carryforwards, today’s effective marginal tax rate depends on outcomes in
other years (you can carry back losses for two years and forward for twenty—recently
changed from three and fifteen). Moreover, if a firm is considering issuing 10 year debt,
both current as well as future effective marginal tax rates are relevant. And because of
the carrybacks and carryforwards, this set of effective tax rates exhibits substantial time-
series dependence. Finally, because he uses NOL carryforwards and other tax credits in
his simulations, his tax measures also are potentially correlated with non-tax
characteristics of the firm.

2.3. The real estate industry
         Others have examined the capital structure of REITs, but they focus on different
questions than we do. Howe and Shilling (1988) find that equity investors react positively
when a REIT issues debt. Jaffe (1991) argues that the value of non-taxable entities such
as REITs and partnerships is invariant to capital structure decisions. Feng, Ghosh, and
Sirmans (2006) argue that REIT capital structure can be explained by the pecking order
theory. None of these papers focuses on the tax status of REITs and its implications for
their capital structures. We focus on REITs as a setting to address the basic unresolved
question of whether or not entity-level taxation and the associated tax benefits of debt
affect leverage decisions.
         By focusing on real estate firms, we obtain a more meaningful measure of the
expected tax benefits of debt than prior studies. Once we identify the tax-status of the
entity (taxable or non-taxable), we are able to compare observed leverage choices across
firms. In this case, we have quite precise measures of expected tax benefits while
avoiding most problems that arise because the tax variables are correlated with financial
distress or other non-tax asset characteristics.


         To examine the effect of tax-status on capital structures, we first identify a sample
of public companies operating as REITs using the CRSP/Ziman Real Estate Data Series.
In addition, we also identify a set of comparable taxable real estate firms based on the
firm’s industry membership.4 Our data span the years 1984 (when segment data was first
available) to 2006 and includes 2,631 firm-year observations for non-taxable real estate
firms, and 810 firm-year observations for taxable real estate firms. The annual number of
observations for real estate firms is depicted in figure 1.
         Using regression analysis to control for factors known to affect firms’ leverage
choices, we compare the capital structures of non-taxable real estate firms with the capital
structures of all Compustat firms also covered by CRSP (Note that we exclude financial
firms, including mortgage REITs). We particularly focus on comparing the capital
structures of non-taxable real estate firms with a set of comparable taxable real estate
firms. Because taxable and non-taxable real estate firms are likely to have similar assets,
comparing their capital structures helps control for any omitted variable bias in our
regression analysis.5
         Leverage. Table 1 provides summary statistics on leverage as well as other firm
characteristics for REITS, taxable real estate companies, and industrial companies.

                                         [Insert Table 1 Here]

         We measure leverage as the ratio of the book value of total debt to the market
value of assets. Total debt is defined as long-term debt (Compustat data item 9) plus debt

  We use the firm’s primary segment industry code reported by Compustat because CRSP industry
definitions are frequently not updated. We classify firms as taxable real estate firms if they have the
following SIC codes: 6512 (Operators of nonresidential buildings); 6513 (Operators of apartment
buildings); 6514 (Operators of dwellings other than apartment buildings); 6515 (Operators of mobile home
sites); 6517 (Lessors of railroad property); 6519 (Lessors of real property); 7011 (Hotels and motels); and
NAICS codes: 531120 (Lessors of nonresidential buildings); 531110 (Lessors of residential buildings and
dwellings); 531190 (Lessors of other real estate property); and 721110 (Hotels (except casino) and motels).
When segment data is not available, we use the historical SIC code reported by Compustat (data item 324).
  We estimate our model employing data from all industrial and real estate firms. A priori, we have no
reason to believe that sensitivity of leverage to the independent variables should differ between industrial
and real estate firms, but such differences could lead to potentially biased estimates of the coefficient on
our real estate dummies. In untabulated tests, we estimate our regression model employing only data from
real estate firms. Our conclusions are unchanged.

in current liabilities (data item 34).6 The market value of assets is defined as the book
value of assets (data item 6) minus the book value of shareholders equity (data item 60)
plus the market value of common equity (data item 199 x data item 54).
        Panel A of Table 1 reveals that non-taxable real estate firms have high leverage.
Mean leverage is 42.5%; median leverage is similar. The average market leverage ratio
for taxable real estate firms is 39.6%, which is significantly smaller than the average for
REITs (p-value of difference < 0.01). Median leverage ratios are statistically
indistinguishable between taxable and non-taxable real estate firms. Average leverage for
industrial firms is 18.1% (panel C). In figure 2, we plot annual average leverage ratios
over the sample period for REITs, taxable real estate firms, and industrial companies. In
general, the differences between real estate firms and industrial companies appear
reasonably stable over time.
        The data in Table 1 and figure 2 suggest that real estate firms generally have
higher leverage than other industrial firms, and that the difference between REITs and
taxable real estate firms is small. Although this comparison does not control for other
factors that affect leverage, the fact that REITs have more than twice as much debt as the
typical U.S. industrial firm, and at least as much as taxable real estate firms, is not clearly
consistent with the proposition that interest tax shields represent an important benefit of
corporate debt. The similarity of the leverage ratios for the taxable and non-taxable real
estate companies suggests that the costs and benefits of debt are related to the nature of
the companies’ assets and investment opportunities. But these simple summary statistics
provide no indication that corporate taxes play an important role in the leverage decision.
        Comparing leverage of REITs with leverage of taxable real estate companies
controls for many firm characteristics that affect firms’ leverage choices. However, to
account for observable differences between these groups, we estimate leverage
regressions that explicitly control for characteristics that have been shown to be
correlated with firms’ leverage choices. We include the following variables in our

 If the COMPUSTAT figure for long-term debt includes the current portion of long-term debt, total debt is
calculated as data item 9 plus data item 34 plus data item 44.

         Investment Opportunities. We measure growth options in the firm’s investment
opportunity set using the firm’s market-to-book ratio (the market value of the firm
divided by the book value of assets). The average market-to-book ratio for the taxable
real estate firms in our sample is 1.28; the ratio for REITs is 1.18 (p-value of difference <
0.01). There is no reliable difference between the medians. Both groups have
significantly lower average (median) market-to-book ratios than industrial companies at
1.84 (1.38).
         Regulation. To control for the effects of regulation, we construct a dummy
variable that equals one for firms in regulated industries and zero otherwise. Regulated
industries in our sample include gas and electric utilities (4900 to 4939). Only 3.9% of
our observations are regulated firms.
         Firm size. We measure firm size as the natural log of sales (data item 12) in
constant 2000 dollars. Based on averages, REITs are smallest (with annual sales of $182
million) followed by the taxable real estate companies ($500 million) and industrial
companies ($1.48 billion). If size is measured by the average book value of total assets,
REIT size is $1.21 billion, taxable real estate firms are $0.98 billion and other industrial
firms are $1.75 billion.
         Profitability. Profitability is measured by the return on assets – operating income
before depreciation (data item 13) divided by total assets. Average ROA for REITs is
6.30% and is indistinguishable from that for taxable real estate firms (6.06%), but is
significantly less than for other industrial firms (7.52%, p-value of difference < 0.01).
Median ROA is lowest for REITs (6.45%), followed by taxable real estate firms (7.28%),
and industrial firms (11.35%).
         Tangibility. We measure the tangibility of the firm’s assets with the fixed-asset
ratio. The fixed-asset ratio is defined as the sum of net property, plant and equipment
(data item 8) and investments and advances – other (data item 32) divided by total
assets.7 As predicted by the theory, the fixed asset ratio generally enters leverage
regressions with a strong positive coefficient. In Table 1, we report that that the median

  We do not employ the traditional definition of tangible assets – net PPE / total assets – because Compustat
rarely reports a value for property, plant and equipment (data item 8) for REITs. Instead, Compustat
bundles investments in fixed assets with other investments under investments and advances – other (data
item 32). To make the definition of this variable comparable across taxable and non-taxable firms, we
define the fixed asset ratio as (data item 8 plus data item 32) / total assets.

fixed-asset ratio for industrial firms is 0.27. However, the median fixed-asset ratio is
significantly higher for REITs at 0.90 than it is for taxable real estate companies at 0.75
(p-value of difference < 0.01), even though these two groups of firms appear to hold
similar assets. (We examine this variable in more detail below.)
         Taxes. Our main focus in this paper is the effect of taxes. We proxy for tax
status using two dummy variables: one for REITs and one for taxable real estate firms;
industrial companies are the omitted category.8 If our control variables fully capture the
determinants of leverage, the tax hypothesis predicts a significant and negative
coefficient on the non-taxable real estate firm dummy variable since REITs receive no
corporate tax benefits of debt. Our regression may fail to capture all the non-tax
differences between real estate and industrial firms that affect leverage. However,
because REITs are similar to taxable real estate companies in most non-tax dimensions,
correlated omitted variables should affect both REIT and taxable real estate dummy
variables in similar ways. Thus, the tax hypothesis implies that the REIT dummy
coefficient should be significantly smaller than the taxable dummy because taxable real
estate firms benefit from the interest tax shield while REITs do not.
         Arguably, however, tax-status dummies also could proxy for non-tax incentives to
use debt and thus could affect our inferences about the role of taxes. REITs pay out a
larger fraction of their return to shareholders in the form of cash dividends. This implies
that the agency costs of free cash flow in REITs – and the derived demand for monitoring
by debt holders – are lower compared to taxable real estate firms. Because this potential
bias is in the same direction as the tax hypothesis, results supportive of the tax hypothesis
must be interpreted with caution.

  In untabulated tests, we estimate the marginal corporate tax rate of the taxable firms in our sample using:
1) the effective tax rate estimated as income tax expense (data item 16) / pretax income (data item 170); 2)
a trichotomous tax rate described in Shevlin (1990) and 3) simulated marginal tax rates before financing
costs (see Graham, Lemmon, and Schallheim 1998). The average estimated marginal tax rate for the
taxable real estate firms (industrial firms) is: 1) effective tax rate: 34.2% (35.2%); 2) trichotomous tax rate:
24.1% (24.9%); 3) simulated tax rate before financing: 30.4% (29.4%). Thus, the taxable firms in our
sample generally should value the tax deductibility of interest payments. We thank John Graham for
providing his simulated tax rates.

        We estimate our benchmark regression of market leverage on the control
variables, as well as the non-taxable and taxable real estate company dummy variables,
for the period 1987 through 2006. Following Fama and MacBeth (1973) and Fama and
French (2002), we estimate these regressions with a single cross-section each year, and
report the average of the slope coefficients from these cross-sectional regressions:

Leverage  1.78  4.57          3.24 Regulation  1.03 ln(Sales)
          (1.24) (24.57) Book (6.59)              (22.61)

             11.39 ROA  17.55 Fixed-asset  14.12 REIT  14.90 Taxable RE
              (8.27)     (22.38) ratio       (20.25) dummy (22.20) dummy

        The t-statistics (in parentheses) are calculated using the time-series standard error
of the average slope coefficients. The average adjusted R2 is 29.32%. Our primary
interest is in capital structure choices for non-taxable real estate firms.9 The REIT
dummy variable is positive and statistically significant. After controlling for the other
determinants of leverage in this regression, the implied leverage for a non-taxable real
estate firm is 14.12 percentage points greater than the implied leverage for an otherwise
similar industrial company (t-stat = 20.25). This result is inconsistent with the tax
hypothesis and suggests that the corporate tax benefits of debt are not a first-order
determinant of leverage.
        Because the coefficient on non-taxable real estate firms could be capturing the
effect of unmeasured attributes of real estate firms, we include a second dummy for
taxable real estate firms. The coefficient for taxable real estate firms of 14.90 (t-stat =
22.20) is larger than the coefficient on the REIT dummy, but the difference between the
coefficients is insignificant – the one-tail p-value is 0.166.10 These results suggest that

  If we estimate our model on all firms but exclude real estate firm dummy variables, the control variables
in the regressions have the same sign and approximate magnitude as reported elsewhere. On average, firms
tend to have higher target leverage if they are larger, less profitable, have fewer growth options, and are
   Throughout the study, we focus on one-tail p-values because the tax hypothesis unambiguously predicts
that leverage should be lower in non-taxable firms.

real estate firms have attributes that are not captured by the other variables in the model.
However, this comparison offers little support for the tax hypothesis.
         We next investigate the robustness of this result to alternative specifications. First,
we examine a potential bias in the tangibility variable and its impact on our results.
Second, we consider the effect of regulatory regimes which affect the economic
characteristics of REITs. Third, we partition the real estate firms into groups based on the
types of property in which they invest and investigate the robustness of the results across
property categories. Fourth, we investigate whether our results are sensitive to alternative
measures of market leverage. Finally, we analyze the data for firms that switch their tax
status to examine the endogeneity of tax status and leverage.
         Asset tangibility. Our benchmark regression includes the fixed-asset ratio; its
estimated coefficient is significant as expected (coeff. = 17.55, t-stat = 22.38). However,
the reported fixed-asset ratios for non-taxable and taxable real estate companies reflect
potentially important different tax incentives and financial reporting rules. To the extent
that our measured fixed-asset ratios are lower for taxable firms as a result of different
reporting incentives, our estimates of the taxable real estate dummy coefficient is
         When a firm acquires operations, they allocate the acquisition price among the
acquired assets. This allocation determines the size and timing of future tax deductions;
thus, taxable firms have incentives to allocate the purchase price to those assets which
generate larger deductions.11 For instance, land is not depreciable, buildings are
depreciable over periods exceeding 27.5 years, but other assts (including intangibles) can
be depreciated over periods ranging from 3 to 15 years. Therefore, taxable real estate
companies face an incentive to over-allocate the purchase price to this third category and
under-allocate to land. REITs do not face these same incentives. Because intangible
assets are excluded from the definition of fixed assets, this tax-based incentive potentially
leads to systematically lower reported fixed-asset ratios for taxable real estate firms.
  The tax incentives for purchase price allocations are limited to transactions which result in stepping up
the tax basis of the acquired assets. The value of increasing the tax basis of an asset is due to the larger
depreciation deductions that can be taken. Acquisitions of free-standing C corporations as well as tax-free
acquisitions are usually structured with no step-up in basis. The seller’s historical tax basis in the assets is
used by (or carried over to) the buyer, even though accounting values are written up to fair market value.
Acquisitions of assets and acquisitions of subsidiaries or flow-through entities in which a §338(h)(10)
election is made result in the seller taking a tax basis of the assets equal to their fair market value.

Moreover, the accounting value of fixed assets will decline faster for taxable firms if
more of the property cost is allocated to equipment and buildings, as opposed to land.
This leads to systematically lower measured fixed-asset ratios for taxable firms.12
Unfortunately, the effect of these forces is difficult to identify and there is limited
research on the extent to which firms trade off tax and financial reporting considerations
when making purchase price allocations.13
         Modern REITs are frequently structured as an umbrella partnership (UPREIT). In
an UPREIT, the REIT creates an operating partnership (OP). The REIT can acquire
assets by exchanging OP units (which are convertible into REIT share) for property. The
principal advantage of this structure is that the exchange of property for OP units is a
non-taxable event under partnership tax law. The OP unit-holder defers the tax until the
units are converted into shares of the REIT. On its consolidated financial statements, the
REIT records the initial acquisition at the assets’ fair market value. When the OP units
are subsequently converted into shares in the REIT, the REIT records the conversion at
the shares’ fair market value. The difference between the market value of the shares
issued and the book value of the units exchanged is allocated to all the REIT’s existing
assets based on their relative fair market values. If a REIT’s share value appreciates faster
than the book value of the partnership unit (as generally occurred over our sample
period), the revaluation at the exchange of OP units for stock increases the fixed-asset
ratio. Such revaluation is less likely to happen in taxable firms.14

   In untabulated analyses, we estimate the relative bias in the fixed-asset ratio for taxable firms that would
be necessary to induce the dummy coefficient differences between taxable and non-taxable real estate firms
as reported in column (1) of Table 2. To do this, we add increments of 0.001 to the fixed asset ratio for
taxable and industrial firms until the estimated dummy coefficients on the taxable real estate firms and
REITs are equal. The estimated bias in the fixed-asset ratio, under the assumption that the true dummy
coefficients are equal, is 0.042. In later tests that focus on the results in different time periods and using
various measures of leverage, the inferences are generally the same whether or not we include the fixed-
asset ratio in the model.
   The tax rules for allocating purchase prices to assets are similar to but separate from those for financial
reporting. When the tax basis of an acquired asset is being stepped-up to fair market value, the IRS
frequently requests valuation analyses performed by independent parties, and are likely to question tax
allocations that are at odds with financial-reporting allocations. Therefore, the financial-reporting
allocations on which we rely to calculate fixed-asset ratios likely embed tax considerations, and we expect
that managers choose allocations that maximize the net present value of future cash flows.
   Another potential bias arises when a REIT acquires property from a taxable seller. Depreciation recapture
rules for depreciable personal property (IRC §1245) treat some or all of any gain as ordinary income to the
extent of previous depreciation. Any remaining gain is capital, and generally taxed at favorable rates. To
the extent the seller has property for which depreciation was taken (for example fixtures, furnishings, and
equipment), there is an incentive to under-allocate the purchase price to that property and over-allocate to

         If the true fixed-asset ratio is independent of firm type, then excluding the
tangibility variable from the regression leaves us with consistent estimates of our dummy
variable coefficients. Omitting the fixed-asset ratio from the model, our estimated
dummy variables suggest that REITs have higher target leverage than their taxable
counterparts. In untabulated results for the 1987 – 2006 period, the estimated coefficient
on the REIT dummy (13.54, t-stat = 11.30) is larger than that of the taxable real estate
firm dummy (11.64, t-stat = 8.15).
         Leverage across REIT regulatory regimes. The modern REIT structure was
created by Congress in 1960 to facilitate investment in large-scale, income-producing real
estate. (For additional detail on REIT institutional background, see Fass, Shaff, and Zief;
2007.) Prior to1986, REITs were generally restricted to hold portfolios of passive
investments; REITs owned the properties, but were required to outsource their operation
and management to third parties. As part of the Tax Reform Act of 1986, Congress
significantly altered the real estate investment landscape by relaxing the restrictions on a
REIT’s operations and permitting most REITs to both own and operate their properties
by providing customary services. REITs were limited on how much prohibited income
they could receive. In December 1999, as part of the REIT Modernization Act, Congress
again substantially relaxed the rules on REIT operations. The primary change in 1999
was that REITs could now engage in otherwise prohibited activities via the use of taxable
REIT subsidiaries. Thus, the REIT Modernization Act further closed the gap between
asset portfolios of REITs and taxable real estate firms.15 The implication of these
developments is that the most appropriate comparisons between REITs and taxable real
estate firms focus on time periods when regulatory constraints on REITs are low.
         In Table 2, we present the results for regressions estimated over various time
periods. In each case, the full benchmark model is estimated. For brevity, we report only
the coefficients on the real estate dummy variables. None of the untabulated coefficients

buildings and land to avoid depreciation recapture. REITs have a comparative advantage in agreeing to
such an allocation. (There are also depreciation recapture provisions for buildings (IRC §1250), but these
rarely apply.)
   For example, a hotel REIT is now permitted to receive revenues from room charges, food and beverage
sales, as well as sales from ancillary services like golf courses and spas, through a taxable REIT subsidiary
without disqualifying its tax-exempt status (subject to certain restrictions, see IRC §856-7). Even with these
liberalizations, lodging and health care REITs must outsource the day-to-day management of the facility to
an independent contractor (§856(l)(3)).

change materially across our various specifications. In the first column, we summarize
the results for our benchmark case (1987 – 2006) for comparative purposes; Column (2)
reports results from the period 1984 – 2006; Column (3), from the period 2000 – 2006.

                                    [Insert Table 2 Here]

       We first extend the beginning of the sample period back to 1984, prior to the
liberalization of REIT regulation. Reported in column (2), the difference in leverage
between taxable and non-taxable real estate firms is statistically significant (p-value =
0.02) and suggests that taxes affect observed capital structure in ways consistent with the
tax hypothesis. However, this result obtains only when including data from the 1984 –
1986 period. The difference in leverage is insignificant after 1986 (Column 1) and after
1999 estimated leverage for non-taxable real estate firms exceeds leverage for taxable
real estate firms (column 3). Moreover, if we omit our asset tangibility variable, the
estimated dummies uniformly suggest that REITs have higher leverage than taxable real
estate firms
       Overall, our results appear sensitive to the time period chosen. This evidence
suggests that the magnitude of the difference between REITs and taxable real estate firms
is smaller during periods where the REITs’ economic activities are more similar to those
of their taxable counterparts. But in general, leverage for taxable real estate firms is not
reliably higher than for non-taxable real estate firms. This again is inconsistent with the
hypothesis that corporate tax status is an important determinant of corporate leverage
       Investor-level taxes. Miller (1977) argues that the tax disadvantage of debt at the
investor level offsets the tax benefits of debt at the corporate level. Although REITs face
a disadvantage in issuing debt from a corporate tax perspective, it could be the case that
they have a comparative advantage in issuing debt given investor-level taxes. Since
investor-level taxation of income from investments in debt securities is the same for

claims issued by taxable and non-taxable firms, a tax-based preference for debt by REITs
could be driven by relatively higher investor-level taxes on equity.16
         At the statutory level, taxation of income from investments in regular corporations
and REITs was generally similar prior to 2003.17 Individuals paid ordinary income tax on
both interest income and dividend payments and were subject to capital gains taxation. In
May of 2003, the Jobs and Growth Tax Relief Reconciliation Act reduced the top
individual tax rate on qualified ordinary dividends to 15%. This change in personal
taxation reduced investor-level taxes on equity investments in taxable corporations, but
not in REITs. This affects the relative advantage of debt versus equity and thus potential
capital structure decisions following the 2003 change in the tax law. In column (4) of
Table 2, we report the results over the period 1987 – 2003, and find that taxable firms
appear to have more leverage than non-taxable firms, but this difference is only
marginally significant (p-value = 0.099). However, if we exclude the asset tangibility
variable, the difference is insignificant.
         Although the statutory treatment of taxable and non-taxable firm income for non-
corporate investors was generally similar before 2003, it also could be the case that the
marginal investors for taxable and non-taxable real estate firms differ. In some studies
(see, Ayers, Lefanowicz, and Robinson, 2003), institutional holdings is used to proxy for
the likelihood that the marginal investor is not an individual.18 Based on institutional
holdings from 2002, we find no significant difference in the fraction of institutional

   With respect to dividend policy, REITs are required to pay out at least 90% of their taxable income as
dividends to maintain tax-exempt status. Because operating cash flow exceeds taxable income (because of
depreciation deductions) many REITs choose to pay out much more. If the REIT distributes the excess
cash, it is often in the form of a non-taxable return of capital which avoids additional immediate taxation
(although it does reduce stock basis and increases future taxation at long-term capital gains tax rates).
Weber and Li (2007) find, on average, that 23% of a REIT’s dividend is a non-taxable return of capital.
Taxable corporations often cannot take advantage of this because the tax status of dividends is determined
in reference to undistributed earnings accumulated over the life of the corporation.
   One exception is the treatment of preferred stock and common stock dividends received by taxable
corporations. To mitigate the effect of multiple layers of taxation on corporate distributions, corporations
normally can deduct 70% of dividends received by another corporation (Prior to 1987, the benchmark was
80%, and before 1986, it was 85%). Since there is generally no corporate tax on a REIT’s income,
corporations cannot take a dividends-received deduction for investments in REIT stock (IRC §243(d)(3)).
   One group of potentially important capital market participants are security dealers; they are well
capitalized and extremely active in the market. Because of their access to trading facilities, they have a
comparative advantage in arbitrage. Moreover, the tax code treats their dividends and capital gains
symmetrically. Furthermore, pension plans (like 401K or 403B plans) tax individuals at withdrawal – again
treating interest income, dividend income and capital gains symmetrically. Thus, if the marginal investor is
a security dealer or a pension plan, then limited pricing results should be expected.

ownership between taxable and non-taxable real estate firms (46.2% for taxable real
estate firms compared to 51.2% for REITs, p-value of difference = 0.57). Ultimately, as
Graham points out, ―we know very little about the identity or tax-status of the marginal
investor(s) between any two sets of securities, and deducing this information is difficult.‖
(2006, p. 603).19
         Real estate property types. The optimal leverage ratio potentially varies across the
type of property owned by the real estate firms. If taxable and non-taxable firms
concentrate in different types of property investments, then it would be inappropriate to
attribute differences in the taxable and non-taxable coefficients to taxes alone. To
examine this possibility, we classify both taxable and non-taxable real estate firms into
four categories based on the type of property owned: non-residential property; residential
property; hotels; and other. For REITs, we identify property type based on the property-
type code reported in the CRSP/Ziman database. For non-REITs, we identify the property
type based on the industry code of the firms’ primary segment.20
         In Table 3, we report the results from re-estimating our benchmark model with the
addition of property-type indicators. For brevity, we report only the coefficients on the
real estate dummy variables (untabulated coefficients again are insensitive to this

                                          [Insert Table 3 Here]

   Miller’s (1977) expression for the tax benefit of debt with personal taxes can be used to show that the
price-setting investor’s tax rate on equity income would have to be significantly higher in a REIT to
completely offset the effect of the loss of tax deductibility of interest payments on the net tax benefit of
debt. Therefore, we expect the likelihood that differences in equity investor-level taxation drive the results
to be small.
   For REITs, we classify firms with primary investments in ―industrial/office‖, ―retail‖, and ―diversified‖
as non-residential; ―residential‖ as residential; ―lodging/resorts‖ as hotel; ―unknown‖, ―unclassified‖,
―health care‖, and ―self storage‖ as other. We exclude mortgage REITS. For non-REITs, we classify firms
with a primary segment SIC code of 6512 (Operators of Nonresidential Buildings) or NAICS code of
531120 (Lessors of Nonresidential Buildings) as non-residential; SIC codes of 6513 (Operators of
Apartment Buildings) or 6514 (Operators of Dwellings Other Than Apartment Buildings) or NAICS code
of 531110 (Lessors of Residential Buildings and Dwellings) as residential; SIC code 7011 (Hotels and
Motels) or NAICS code 721110 (Hotels (except Casino Hotels) and Motels) as hotel; and SIC codes 6515
(Operators of Residential Mobile Home Sites), 6517 (Lessors of Railroad Property), 6519 (Lessors of Real
Property, NEC) or NAICS code 531190 (Lessors of Other Real Estate Property) as other.

        For both taxable and non-taxable real estate firms, residential property firms have
the highest leverage, followed by hotel firms, non-residential firms, and firms in the other
property category. The fact that the ordering of the coefficient magnitudes is the same for
both taxable and non-taxable firms suggests that observed leverage choices are correlated
with property type. Comparing the estimated coefficients in column (1), only taxable
residential firms have significantly higher leverage than REITs (p-value of difference <
0.001). Leverage ratios for taxable and non-taxable non-residential, hotel, and other real
estate firms are not significantly different from each other at the 10% level using a one-
tail test. As we focus on alternative time periods in the remaining columns, the
conclusions are the same. If we exclude the fixed-asset ratio from the model, our
inferences are unchanged.
        Alternative measures of leverage. We next investigate the robustness of the
results to alternative definitions of leverage. REITs hold substantially less cash than their
taxable counterparts. The summary statistics in Table 1 indicate that cash represents
approximately 4.59% of the average REIT’s assets in market value terms. This is
significantly less than the average cash holdings of both taxable real estate firms
(12.61%) and industrial firms (10.85%). The tax hypothesis would suggest that taxable
firms have incentives to avoid holding cash for the same reasons that they would use
leverage – the interest income from their cash holdings would be taxed.21 Because firms
can use cash to reduce outstanding debt, an alternative approach is to measure leverage
on a net-of-cash basis.

                                         [Insert Table 4 Here]

        In Table 4, we re-estimate our benchmark regression using four alternative
measures of leverage. In panel A, we report the results for the tax-status variables only.
When we subtract cash holdings from total debt (col. 1), the coefficient on the non-
taxable dummy is 15.24 compared to the coefficient on the taxable dummy of 13.24. This
difference is of the opposite sign as implied by the tax hypothesis. If we also adjust the

  Furthermore, Riddick and Whited (2007) argue that a firm’s cash holding is related to external financing
costs, investment opportunities, and corporate taxes. While we do not test the hypotheses implied by these
studies, they do suggest that it is important to consider how cash holdings affect the results.

denominator of our leverage measure for cash, as reported in column (2), the difference is
insignificant (p-value = 0.229). These findings offer no support for the tax hypothesis.22
         The traditional measure of leverage also ignores other fixed claims, such as
preferred stock. To the extent the composition of fixed claims varies across taxable and
non-taxable real estate entities, we potentially draw incorrect inferences by focusing on
debt alone. In the remaining two columns, we address the issue by including preferred
stock in the measure of leverage. In general, our inferences are unchanged. If we add
preferred stock to the numerator in column (3), and we adjust the measure of fixed claims
for cash holdings in column (4), the leverage difference is insignificant. In panel B, we
focus on the differences across property types. The significant difference in leverage
between taxable and non-taxable residential real estate firms disappears in column (4);
our other inferences are unchanged. (Excluding the fixed-asset ratio from the model does
not affect our findings)
         Selection and tax-status. Our empirical analysis assumes that observed leverage
choices and tax status are not jointly determined by some unobserved factor that is
omitted from the model. Prior literature on the determinants of tax-status is limited.
Damadoran, John, and Liu (1997) document the stated reasons why real estate firms
switch among taxable corporation, REIT, and partnership forms. Unsurprisingly, firms
switch to non-taxable status primarily for the tax benefits. They report that firms
switching away from the REIT structure cite the desire to have more flexibility in
managing assets, making investment decisions, and engaging in prohibited activities. But
their sample of firms that switch status covers only the period from 1966 through 1989.
Their data generally predates our post-1986 sample period during which the tax code
imposed fewer restrictions on REITs’ management and investment policies.

  An alternative explanation for differences in cash holdings is related to foreign operations. Foley,
Hartzell, Titman and Twite (2007) argue that firms with foreign operations have tax-based incentives to
hold more cash in foreign subsidiaries. REITs historically do not invest in foreign property. In our sample,
2.13% of REITs report foreign income, and of those, foreign income averages 0.27% of sales. 10.37% of
taxable real estate firms report foreign income. Of those, foreign income averages 2.97% of sales. As a
result, variation in cash balances across REIT and taxable firms, and hence our cash-adjusted leverage
ratios, simply could be reflecting variation in foreign operations. To check for this, we re-estimate the
regressions in columns (1) and (2) after excluding firms in which foreign income is positive. In untabulated
results, our findings are qualitatively similar.

       If firms choose their tax status (REIT vs. taxable C corporation) our statistical
methods may fail to control effectively for the differences across these two sets of firms.
To evaluate the impact of this potential selection bias, we examine leverage decisions
surrounding a firm’s switch between REIT and non-REIT status. We examine both
market leverage as well as the residuals from our benchmark regression (re-estimated
excluding the tax-status dummies). Using residuals should control for simultaneous
changes in the determinants of leverage that also are affected by the change in tax status.
Among firms with sufficient data that switch to REIT status during the sample period (12
firms), we find that leverage increases by 0.21% from the year before to the year after the
switch; this increase is statistically insignificant (two-tailed p-value = 0.98). Analyzing
residuals indicates that abnormal leverage falls by 2.95%; this decrease also is
insignificant (p-value = 0.47). Among firms that switched away from REIT status (7
firms), we find a decrease in leverage from the year before to the year after the switch;
again, this difference is insignificant. Overall, the evidence from the sample of firms
switching tax-status during the sample period indicates no material selection bias and
provides no support for the tax hypothesis.

       At least since Modigliani and Miller (1963), corporate finance theory has
recognized that a potentially important benefit of leverage is the value of the corporate
interest tax shield. Over the last four decades, a number of papers have employed an
array of empirical strategies to identify the impact of this tax benefit. However to date,
supportive empirical evidence is limited. Potential reasons are: (1) our proxies for the tax
benefits of debt are poor and potentially correlated with other firm characteristics (for
instance, tax loss carryforwards capture other attributes of the firm, such as the expected
costs of financial distress) (2) the corporate tax benefit of leverage is offset by the
personal tax disadvantages of leverage and these offsetting tax effects are difficult to
        We believe that our empirical methods largely avoid both of these problems. We
identify REITs as a set of firms where we know that the corporate tax benefit of debt is
zero. The tax hypothesis thus suggests that these firms should have less debt than their

taxable counterparts. Moreover, prior to 2004 the tax code generally treated investments
in REITs and taxable real estate firms symmetrically, hence any differential corporate tax
advantages would not be offset by differential personal tax disadvantages. We test the tax
hypothesis by comparing the leverage choices of these non-taxable REITs with the
leverage choices of taxable real estate firms and other industrial companies.
          Our empirical analysis provides little evidence that taxes are an important
determinant of leverage decisions. REITs have leverage that is significantly higher than
comparable industrial firms and is not significantly lower than that of taxable real estate
companies, which have similar assets but quite different tax circumstances. This result is
surprising in light of the facts that industrial companies are generally fully taxable and
that REITs receive no benefit from the interest tax shield. Our results indicate that the
costs and benefits of debt are closely related to the nature of the firm’s assets and
investment opportunities and that the tax deductibility of debt plays a secondary role, at
best, in the firm’s leverage decision. Our results, which are summarized in Table 5,
appear to be quite robust. They generally obtain across different sub-periods, using
different leverage measures, and disaggregating across various asset types.23 Of the
limited evidence we do obtain to support the tax hypothesis, it is concentrated in the
results that control for asset tangibility in panel A. But even these results are problematic
since different tax incentives and accounting rules across taxable and non-taxable real
estate firms create potentially biased estimates of asset tangibility. When we exclude the
fixed-asset ratio from the empirical model in panel B, there is no reliable evidence that
leverage differs between taxable and non-taxable real estate firms.
         Therefore, despite compelling logic that taxes matter in corporate financing policy
decisions, the supporting evidence is far from overwhelming. Our evidence when
combined with the scant extant evidence of material tax effects suggests that taxes are
simply less important in driving corporate leverage decisions than our intuition has

  The most supportive evidence uses data from residential real estate firms – leverage employed by taxable
real estate firms in this sector is higher than that for REITs. But there is no a priori reason to expect that
residential real estate provides a particularly powerful test of the tax hypothesis.

Alderson, M.J. and B.L. Betker, 1995, ―Liquidation Costs and Capital Structure,‖ Journal of
       Financial Economics, 39, 45-69.

Ayers, B.C., C.E. Lefanowicz, and J.R. Robinson, 2003, ―Shareholder Taxes in Acquisition
       Premiums: The Effect of Capital Gains Taxation,‖ Journal of Finance, 58, 2783-

Barclay, M., C.W. Smith and R.L. Watts, 1995, ―The Determinants of Corporate Leverage
       and Dividend Policies,‖ Journal of Applied Corporate Finance, 7, 4-19.

Barclay, M., and C.W. Smith, 2005, ―The Capital Structure Puzzle: The Evidence Revisited,‖
       Journal of Applied Corporate Finance, 17, 8-17.

Bradley, M., G. Jarrell, and E.H. Kim, 1984, ―On the Existence of an Optimal Capital
       Structure: Theory and Evidence,‖ Journal of Finance, 39, 857-878.

Damadoran, A., K. John and C.H. Liu, 1997, ―The Determinants of Organizational Form
     Changes: Evidence and Implications from Real Estate,‖ Journal of Financial
     Economics, 45, 169-192.

DeAngelo, H. and R.W. Masulis, 1980, ―Optimal Capital Structure under Corporate and
      Personal Taxation,‖ Journal of Financial Economics, 8, 3-29.

Dyreng, S., and J. Graham, 2007, ―How Much do Tax Shields Contribute to Firm Value? A
       Natural Experiment in the Canadian Income Trust Market,‖ Working paper, Duke

Fama, E.F. and K.R. French, 1998, ―Taxes, Financing Decisions, and Firm Value,‖ Journal
       of Finance, 53, 819-843.

Fama, E.F. and K.R. French, 2002, ―Testing Tradeoff and Pecking Order Predictions about
       Dividends and Debt,‖ Review of Financial Studies 15, 1-37.

Fama E.F, and J. MacBeth, 1973, ―Risk, Return, and Equilibrium: Empirical Tests,‖ Journal
      of Political Economy, 81, 607-636.

Fass, P.M., M.E. Shaff, and D.B. Zief, 2007, Real Estate Investment Trusts Handbook.
        Thomson West.

Feng, Z., C. Ghosh, and C.F. Sirmans, 2006, ―On the Capital Structure of Real Estate
       Investment Trusts (REITS),‖ Working paper, University of Connecticut.

Foley, F.C., J.C. Hartzell, S. Titman, and G. Twite, 2007, ―Why do Firms Hold So Much
        Cash? A Tax-Based Explanation,‖ Journal of Financial Economics 86, 579-607.

Graham, J.R., 1996, ―Debt and the Marginal Tax Rate,‖ Journal of Financial Economics 41,

Graham, J.R., 1999, ―Do Personal Taxes Affect Corporate Financing Decisions?‖ Journal of
      Public Economics 73, 147-185

Graham, J.R., 2000, ―How Big Are the Tax Benefits of Debt?‖ Journal of Finance 55, 1901-

Graham, J.R., 2006, ―A Review of Taxes and Corporate Finance,‖ Foundations and Trends
      in Finance 1, 573–691.

Graham, J. R., M. Lemmon, and J. Schallheim, 1998, ―Debt, Leases, Taxes, and the
      Endogeneity of Corporate Tax Status,‖ Journal of Finance 53, 131–162.

Howe, J.S. and J.D. Shilling, 1988, ―Capital Structure Theory and REIT Security Offerings,‖
       Journal of Finance, 43, 983-993.

Jaffe, J.F., 1991, ―Taxes and the Capital Structure of Partnerships, REIT’s, and Related
         Entities,‖ Journal of Finance, 46, 401-407.

MacKie-Mason, J. K., 1990, ―Do Taxes Affect Corporate Financing Decisions?‖ Journal of
      Finance 45, 1471–1493.

Masulis, R.W., 1983, ―The Impact of Capital Structure Change on Firm Value: Some
       Estimates,‖ Journal of Finance, 38, 107-126.

Mehrotra, V, W. Mikkelson and M. Partch, 2005, ―Do Managers Have Capital Structure
       Targets? Evidence from Corporate Spinoffs,‖ Journal of Applied Corporate Finance,
       17, 18-25.

Miller, M.H., 1977, ―Debt and Taxes,‖ Journal of Finance, 32, 261-275.

Modigliani, F., and M.H. Miller, 1958, ―The Cost of Capital, Corporation Finance and the
       Theory of Investment,‖ American Economic Review, 48, 261-297.

Modigliani, F., and M.H. Miller, 1963, ―Corporate Income Taxes and the Cost of Capital: A
       Correction,‖ American Economic Review, 53, 315-335.

Myers, S. C., 1984, ―The Capital Structure Puzzle,‖ Journal of Finance, 39, 575–592.

Riddick, L.A., and T.M. Whited, 2007, ―The Corporate Propensity to Save,‖ Working paper,
       University of Wisconsin.

Shevlin, T., 1990, ―Estimating Corporate Marginal Tax Rates with Asymmetric Tax
       Treatment of Gains and Losses,‖ The Journal of the American Taxation Association
       12, 51–67.

Smith, C.W., 1986, ―Investment Banking and the Capital Acquisition Process,‖ Journal of
       Financial Economics, 15, 3-29.

Smith, C.W. and R.L. Watts, 1992, ―The Investment Opportunity Set and Corporate
       Financing, Dividend and Compensation Policies,‖ Journal of Financial Economics,
       32, 263-292.

Titman, S. and R. Wessels, 1988, ―The Determinants of Capital Structure Choice,‖ Journal of
       Finance, 43, 1-19.

Weber, D., and O.Z. Li, 2007, ―The Ex-Day Pricing of Dividends for REITs,‖ Working
      paper, University of Arizona.

                                            Figure 1

          Annual num ber of observations for taxable and non-taxable real
                                  estate firm s



























                             Non-taxable Real Estate          Taxable Real Estate

                                            Figure 2

         Annual m ean m arket leverage for taxable real-estate firm s, non-
               taxable real estate firm s, and other industrial firm s























                 Non-taxable Real Estate            Taxable Real Estate            Other Industrial

                                                      Table 1
                                                Descriptive Statistics
Market leverage is total debt (Compustat data items 9 + 34) / market value of assets (total assets (item 6) – book
value of equity (item 60) + market value of equity (item 199 x item 54)). Market-to-book is the ratio of market
value of assets / Book value of total assets (item 6). Sales (item 12) and Assets (item 6) are inflation-adjusted.
ROA is EBITDA (item 13) / Book value of total assets. Fixed-asset ratio is equal to (property, plant, and
equipment (item 8) + other investments (item 32)) / Book value of total assets. Cash is equal to cash and cash
equivalents (item 1) / Market value of assets.

Panel A: REITs (N = 2,631)
                                  Mean               Std. Dev         25%              50%              75%
Market Leverage (%)                 42.46               19.98         30.08             41.86            54.38
Market-to-Book                       1.18                0.37          0.97              1.14             1.32
Sales ($M)                         181.72              373.55         23.30             70.03           174.98
Assets ($M)                      1,209.62            2,345.33        156.74            462.50         1,174.77
ROA (%)                              6.30                5.17          4.71              6.45             8.25
Fixed-asset ratio                    0.84                0.20          0.82              0.90             0.95
Cash (%)                             4.59               10.15          0.51              1.46             4.27

Panel B: Taxable real estate firms (N = 810)
                                  Mean               Std. Dev         25%              50%              75%
Market Leverage (%)                 39.61*              25.42            19.18          40.62            59.47
Market-to-Book                       1.28*               0.77             0.94           1.10             1.39
Sales ($M)                         500.17*           1,567.33             8.82          42.27*          256.81
Assets ($M)                        977.94            2,076.55            48.31         144.60*          799.82
ROA (%)                              6.06                8.84             2.58           7.28*           10.79
Fixed-asset ratio                    0.65*               0.27             0.49           0.75*            0.87
Cash (%)                            12.61*              21.56             1.88           4.76            11.49

Panel C: Industrial firms (N = 83,045)
                                  Mean               Std. Dev         25%              50%              75%
                                            *                                                   *
Market Leverage (%)                 18.13               17.93             2.22          13.50            29.04
Market-to-Book                       1.84*               1.32             1.05           1.38*            2.09
Sales ($M)                       1,482.22*           7,079.65            33.06         150.00*          678.41
Assets ($M)                      1,750.41*          10,536.31            36.69         145.20*          666.91
ROA (%)                              7.52*              18.84             3.90          11.35*           17.14
Fixed-asset ratio                    0.33*               0.24             0.13           0.27*            0.48
Cash (%)                            10.85*              15.27             1.59           5.27            13.79

 Indicates difference between sample mean (median) of non-taxable real estate firms significant at the 0.01 level
using a two-tailed t-test (Z-test).

                                         Table 2
        Market leverage for real estate firms during various regulatory regimes
Column (1) includes the benchmark period of 1987 - 2006. Column (2) includes the period 1984 - 2006.
Column (3) includes the period after the REIT Modernization Act. Column (4) includes the period after the
Tax Reform Act but before the Jobs and Growth Tax Relief Reconciliation Act. In each column, the full
model is estimated, and only the coefficient estimates on real estate firms are reported. The dependent
variable is total debt divided by the market value of assets at the end of year. Reported coefficients are the
average annual estimated coefficients, and the t-statistic is based on the distribution of the annual
coefficient estimates.
                                   1987 – 2006        1984 – 2006        2000 – 2006        1987 – 2003
                                       (1)                (2)                (3)                (4)

All Real Estate Firms:
 REITs                                14.12              11.88              14.88              13.94
                                     (20.25)             (8.60)            (36.21)            (17.19)
 Taxable                              14.90              14.26              14.22              15.07
                                     (22.20)            (20.33)            (20.94)            (19.95)
 p-value of difference                 0.166              0.020             >0.500              0.099
Ave. Adj. R2                          29.32%             29.15%             31.91%             28.78%

REITs are identified using the CRSP/Ziman real estate database. Taxable real estate firm dummies are
based on the SIC or NAICS code of the firm’s primary segment. The p-value of the difference in
coefficients is based on the annual differences using a one-tail test of the hypothesis that taxable firm
leverage coefficient exceeds the non-taxable leverage coefficient. Untabulated independent variables are
defined in Table 1.

                                        Table 3
    Market leverage for real estate firms classified by property type during various
                                  regulatory regimes
Column (1) includes the benchmark sample period of 1987 - 2006. Column (2) includes the period 1984 –
2006. Column (3) includes the period after the REIT Modernization Act. Column (4) includes the period after
the Tax Reform Act but before the Jobs and Growth Tax Relief Reconciliation Act. In each column, the full
model is estimated, and only the coefficient estimates on real estate firms are reported. The dependent variable
is total debt divided by the market value of assets at the end of year. Reported coefficients are the average
annual estimated coefficients, and the t-statistic is based on the distribution of the annual coefficient estimates
                                  1987 – 2006          1984 – 2006           2000 – 2006           1987 – 2003
                                      (1)                  (2)                   (3)                   (4)

 REIT                               13.54                11.21                 15.99                 13.07
                                   (11.30)               (6.74)               (35.25)                (9.48)
 Taxable                            11.64                10.40                  9.33                 11.44
                                    (8.15)               (7.29)                (4.13)                (6.97)
 p-value of difference              >0.500               >0.500                >0.500                >0.500
 REIT                               19.20                17.43                 18.34                 18.78
                                   (13.39)              (11.00)               (13.91)               (11.23)
 Taxable                            30.06                29.04                 33.94                 29.79
                                   (20.81)              (20.41)               (22.56)               (17.58)
 p-value of difference              <0.001               <0.001                <0.001                <0.001
 REIT                               15.80                12.79                 15.94                 15.96
                                    (8.05)               (5.29)               (21.36)                (6.89)
 Taxable                            15.94                15.87                 14.06                 17.03
                                   (12.94)              (14.68)                (7.43)               (14.06)
 p-value of difference               0.472                0.104                >0.500                 0.319
 REIT                               8.88                  6.87                   5.94                 9.82
                                   (5.54)                (3.82)                 (2.87)               (5.48)
 Taxable                            0.64                  0.94                -10.26                  2.21
                                   (0.27)                (0.42)                (-3.31)               (0.86)
 p-value of difference             >0.500                >0.500                >0.500                >0.500
Ave. Adj. R2                       29.67%                29.48%                32.34%                29.09%
REITs are identified using the CRSP/Ziman real estate database. Taxable real estate firm dummies are based
on the SIC or NAICS code of the firm’s primary segment. Property type classification is described in the text.
The p-value of the difference in coefficients is based on the annual differences using a one-tail test of the
hypothesis that taxable firm leverage coefficient exceeds the non-taxable leverage coefficient. Untabulated
independent variables are defined in Table 1.

                                     Table 4
   Alternative measures of market leverage for real estate firms from 1987 – 2006
In each column, the full model is estimated, and only the coefficient estimates on real estate firms are
reported. The alternative measures of leverage in the dependent variable takes are: (debt – cash) / MVA;
(debt – cash) / (MVA – cash); (debt + preferred stock) / MVE; and (debt + preferred stock – cash) / MVA.
Reported coefficients are the average annual estimated coefficients, and the t-statistic is based on the
distribution of the annual coefficient estimates.
                                                                           Debt +            Debt + Pref.
                                    Debt – Cash      Debt – Cash           Pref. Stk.          Stk. – Cash
                                       MVA           MVA – Cash             MVA                  MVA
                                         (1)              (2)                (3)                  (4)

Panel A: Tax-status Indicators

Real Estate Firms
 REIT                                15.24              15.48             14.78              15.97
                                    (20.04)            (21.04)           (16.26)            (15.72)
 Taxable                             13.24              16.32             14.06              12.57
                                    (15.10)            (16.66)           (20.64)            (14.58)
 p-value of difference               >0.500              0.229            >0.500             >0.500
Ave. Adj. R2                         28.28%             24.61%            26.10%             25.63%

Panel B: Property type indicators

 REIT                                14.50              14.63             14.22             15.28
                                    (11.23)            (10.91)            (9.74)            (9.60)
 Taxable                              9.13              16.15             10.55              8.36
                                     (6.13)             (9.87)            (6.89)            (5.28)
 p-value of difference               >0.500              0.218            >0.500            >0.500
 REIT                                20.14              21.36             19.49              20.49
                                    (14.38)            (13.09)           (16.87)            (17.28)
 Taxable                             23.24              24.78             29.07              22.42
                                    (14.15)            (15.38)           (21.00)            (13.35)
 p-value of difference                0.038              0.018            <0.001              0.113
 REIT                                16.99              17.04             16.76              17.94
                                     (8.31)             (8.04)            (7.99)             (8.43)
 Taxable                             15.42              16.32             15.17              14.69
                                    (10.37)            (11.16)           (12.81)            (10.38)
 p-value of difference               >0.500             >0.500            >0.500             >0.500
 REIT                                10.95             10.45               9.44             11.58
                                     (6.71)            (6.35)             (6.49)            (7.87)
 Taxable                              2.45              1.28              -0.18              1.90
                                     (0.96)            (0.51)            (-0.06)            (0.65)
 p-value of difference               >0.500            >0.500            >0.500             >0.500
Ave. Adj. R2                         28.61%            26.29%            26.37%             25.89%

Non-taxable real estate firms are identified using the CRSP/Ziman real estate database. Taxable real estate
firm dummies are based on the SIC or NAICS code of the firm’s primary segment and are disaggregated
by property type. The p-value of the difference in coefficients is based on the annual differences using a
one-tail test of the hypothesis that taxable firm leverage coefficient exceeds the non-taxable leverage
coefficient. Untabulated independent variables are defined in Table 1.

                                   Table 5
Summary of leverage comparisons between taxable and non-taxable real estate firm
Summary of results from comparing the annual estimated dummy coefficients for taxable and non-taxable
real estate firms over various regimes, asset ownership types, and leverage definitions after including or
excluding the fixed asset ratio from the benchmark regression.

                        Panel A: Fixed asset ratio included          Panel B: Fixed asset ratio excluded

                                   Leverage measure                           Leverage measure
  Property      Time
    type       period     1       2        3        4       5          1       2       3       4         5

                 1        ∙        ∙       ∙        ∙       ∙          ∙       ∙       ∙       ∙         ∙
                 2        ●        ∙       ●        ∙       ∙          ∙       ∙       ∙       ∙         ∙
                 3        ∙        ∙       ∙        ∙       ∙          ∙       ∙       ∙       ∙         ∙
                 4        ●        ∙       ●        ∙       ∙          ∙       ∙       ∙       ∙         ∙
                 1        ∙        ∙       ∙        ∙       ∙          ∙       ∙       ∙       ∙         ∙
                 2        ∙        ∙       ●        ∙       ∙          ∙       ∙       ∙       ∙         ∙
                 3        ∙        ∙       ∙        ∙       ∙          ∙       ∙       ∙       ∙         ∙
                 4        ∙        ∙       ●        ∙       ∙          ∙       ∙       ∙       ∙         ∙

                 1        ●       ●        ●       ●        ∙         ●        ∙       ∙      ●          ∙

                 2        ●       ●        ●       ●       ●          ●        ∙       ●      ●          ∙
                 3        ●       ●        ●       ●        ●         ●        ∙       ∙      ●          ∙
                 4        ●       ●        ●       ●        ●         ●        ∙       ∙      ●          ∙

                 1        ∙        ∙       ∙        ∙       ∙          ∙       ∙       ∙       ∙         ∙
                 2        ∙        ∙       ∙        ∙       ∙          ∙       ∙       ∙       ∙         ∙
                 3        ∙        ∙       ∙        ∙       ∙          ∙       ∙       ∙       ∙         ∙
                 4        ∙        ∙       ∙        ∙       ∙          ∙       ∙       ∙       ∙         ∙
                 1        ∙        ∙       ∙        ∙       ∙          ∙       ∙       ∙       ∙         ∙
                 2        ∙        ∙       ∙        ∙       ∙          ∙       ∙       ∙       ∙         ∙
                 3        ∙        ∙       ∙        ∙       ∙          ∙       ∙       ∙       ∙         ∙
                 4        ∙        ∙       ∙        ∙       ∙          ∙       ∙       ∙       ∙         ∙
  Significance level                       Leverage definition                             Time period
     ●          ≤1%               1     Debt / Assets                                  1     1987 – 2006
     ●          ≤5%               2     (Debt – Cash) / Assets                         2     1984 – 2006
      ●        ≤10%               3     (Debt – Cash) / (Assets – Cash)                3     2000 – 2006
      ∙        >10%               4     (Debt + Preferred Stock) / Assets              4     1987 – 2003
                                  5     (Debt + Preferred Stock – Cash) / Assets


Shared By: