tax rates and corporate decision making
TRANSCRIPT
Columbia Law School Davis Polk & Wardwell Tax Policy
Colloquium
“TAX RATES AND CORPORATE DECISION MAKING”
Michelle Hanlon
(Co-written with John Graham,Terry Shevlin, & Nemit Shroff)
Tuesday, November 18, 4:20-6:10 pm
Jerome Greene Hall, Room 304
Colloquium schedule and papers are available at:
http://web.law.columbia.edu/tax-policy
Tax Rates and Corporate Decision Making
John R. Graham
Duke University [email protected]
Michelle Hanlon
Massachusetts Institute of Technology [email protected]
Terry Shevlin
University of California, Irvine [email protected]
Nemit Shroff
Massachusetts Institute of Technology [email protected]
November 2014
ABSTRACT
It has long been suspected that managers use short-cuts (e.g., heuristics) to make many decisions and that their decisions are affected by behavioral biases such as a tendency to overly rely on ‘salient’ or vivid metrics/information. We document that managers do indeed rely on heuristics when they incorporate taxes into decision-making, and that tax rate salience affects their decision-making. Further, we document, for the first time in a corporate setting, that these behavioral biases lead to suboptimal decisions, and we provide estimates of the economic magnitude of the loss in firm value as a result of these biases. For example, we find that many firms employ the average tax rate paid on their income (i.e., the GAAP effective tax rate [ETR]) rather than the marginal tax rate (MTR) to evaluate incremental decisions, and that using the GAAP ETR for decision-making leads the typical firm to experience a deadweight loss of nearly $10 million for making the incorrect capital structure decisions. We also find that firms that employ the GAAP ETR for investment decision-making are less responsive to their growth opportunities and have smaller acquisition announcement returns than those using the MTR, leading to a loss in firm value of more than $20 million from suboptimal acquisitions. We appreciate helpful comments from Mary Barth, Lil Mills (discussant), Richard Sansing, seminar participants at the Baruch College, Columbia Law School, INSEAD conference, MaTax Conference in Germany, Stanford Summer Camp, University of California, Davis, University of Tennessee, and Virginia Tech University. Thanks to the following people for helpful comments on the development of survey questions: Jennifer Blouin, James Chyz, Merle Erickson, Ken Klassen, Ed Maydew, Peter Merrill, Lil Mills, Sonja Rego, Richard Sansing, Stephanie Sikes, Joel Slemrod, and Ryan Wilson. We also appreciate the support of PricewaterhouseCoopers (especially Peter Merrill) and the Tax Executives Institute (especially Tim McCormally) in asking firms to participate and in reviewing the survey document. Finally, each author is grateful for the financial support of the Fuqua School of Business, Paton Accounting Fund at the University of Michigan, MIT Junior Faculty Research Assistance Program, and the Paul Merage School of Business at the University of California-Irvine. All errors are our own.
1
1. Introduction
Taxes represent a significant cost for most profitable corporations and, thus, are an
important input into many corporate decisions (see Shackelford and Shevlin (2001), Graham
(2003), and Hanlon and Heitzman (2010) for reviews). According to theory, the marginal tax rate
(MTR), defined as the present value of additional taxes paid on an additional dollar of income
earned today (Scholes, Wolfson, Erickson, Hanlon, Maydew, and Shevlin, 2014), is the
appropriate rate to use to evaluate incremental corporate decisions (MacKie-Mason, 1990;
Graham, 1996a; Graham, 1996b; Brealey, Myers, and Allen, 2014; and Scholes et al., 2014).
Consistent with this theory, prior research finds that the MTR is correlated with firms’ capital
structure decisions (e.g., Graham, 1996a; Heider and Ljungqvist, 2014). However, prior research
also finds that firms often have conservative debt policies given their MTRs (e.g., Graham, 2000;
Strebulaev and Yang, 2013), raising the question of whether managers do indeed incorporate the
MTR in incremental leverage decisions or whether the MTR is simply correlated with the tax
rate they use.1
In this paper, we use a survey to directly ask tax executives, for the first time, which tax
rate their firms use when making corporate financing and investment decisions. Survey data are
particularly useful to address this question because the rate employed cannot be directly
observed. Indeed, archival-based tests designed to identify which tax rates managers incorporate
into decision making are necessarily joint tests of what measure of taxes (if any) is used to make
the decision and whether the researchers’ empirical proxy of the rate is correct.2 We combine the
survey data with Compustat data to explore the determinants of managers’ tax rate choices and
the economic consequences of deviating from the theoretically preferred tax rate.
1 Fama and French (1998) caution that cross-sectional studies are vulnerable to endogeneity biases as firms’ tax status may correlate with omitted variables. 2 Survey data are, of course, subject to many of their own concerns, which we acknowledge and attempt to mitigate. Our goal in this paper is to complement the evidence in prior empirical-archival studies using survey evidence.
2
We directly ask tax executives of both public and private firms which tax rates their
company employs in the following decisions: (i) mergers and acquisitions (M&A), (ii) capital
structure, (iii) capital investment, (iv) purchase versus lease, (v) cost of capital computations, (vi)
where to locate new facilities, and (vii) compensation. We find that fewer than 13% of our
sample firms say that they use the MTR in any of these decision making contexts. Rather, survey
responses indicate that most firms use either the U.S. statutory tax rate (STR) or the ‘average’ tax
rate computed in financial statements (the GAAP effective tax rate [GAAP ETR], defined as
income tax expense scaled by pretax book income, both financial accounting numbers) for
evaluating incremental decisions.3 For example, 30% (26%) of the respondents indicate that they
use the GAAP ETR (STR) for making capital structure decisions.
Although these findings are seemingly inconsistent with corporate finance theory, they
are potentially consistent with the theories and evidence in psychology and behavioral
economics, which have focused primarily on consumers/individuals in non-corporate settings
(DellaVigna, 2009). For example, prior research finds that (i) individuals (and managers in some
cases) use simple heuristics in many decision-making contexts rather than more complex (and
fully rational) approaches (e.g., Tversky and Kahneman, 1974; Simon, 1979; Gabaix, Laibson,
Moloche, and Weinberg, 2006) and (ii) individuals are affected by tax rate salience (e.g., de
Bartolome, 1995; Finkelstein, 2009; Chetty, Looney, and Kroft, 2009). Computing the MTR is
complicated due to unique features of the tax code (e.g., the treatment of net operating losses,
alternative minimum tax, etc.) coupled with the need to forecast taxable income many years into
the future. It is precisely in such complicated situations that managers are likely to make
decisions based on heuristics such as the STR, which is well-known to all firms and readily
available. Similarly, for publicly traded companies, the GAAP ETR is easily computed using
financial statement data, and is indeed the focus of much managerial attention (Healy, 1985; 3 GAAP stands for Generally Accepted Accounting Principles.
3
Degeorge, Patel, and Zeckhauser, 1999; Graham, Harvey, and Rajgopal, 2005; Graham, Hanlon,
Shevlin, and Shroff, 2014), which potentially leads to a salience bias. That said, prior research in
psychology argues that profit-maximizing firms are less likely to deviate from fully rational
decision-making because relative to individuals, firms have more experience and are disciplined
by forces such as competition (DellaVigna, 2009; DellaVigna and Pollett, 2013).
We explore each of these issues in detail. We find that the difference between the STR
and the MTR is less than two percentage points for the majority of firms that use the STR as the
tax rate input for decision making.4 This suggests that when a firm uses the STR, in many cases
the STR closely approximates the MTR, consistent with the STR being a simple heuristic
employed by managers. We also find that public firms (relative to private firms) and firms with
high analyst following (relative to firms with low analyst following) are more likely to use the
GAAP ETR as the tax rate in their decisions, consistent with the idea that capital market pressure
increases the salience of the GAAP ETR and thus increases its usage in decision-making.
Finally, we find that larger firms (conditional on being public), high R&D intensity firms, and
high institutional ownership firms are more (less) likely to use the MTR (ETR) for decision
making, consistent with sophisticated firms and firms with greater external monitoring (i.e.,
institutional ownership) more correctly incorporating taxes into their decision making.5
While the MTR is the theoretically preferred tax rate to evaluate the tax impact of
incremental decisions, and the STR is equivalent to the MTR for highly profitable firms without
NOLs and thus qualifies as an efficient heuristic, there is very little justification (at least in
theory) for using an average rate such as the GAAP ETR as the tax rate input for decision-
4 We estimate MTRs using the approaches developed by Graham (1996a) and Blouin, Core, and Guay (2010). 5 Our results suggest that analyst following induces inefficiencies caused by capital market pressures whereas institutional investors serve to monitor managers and reduce their focus on short-term earnings. Such opposing effects are consistent with the recent research; specifically, He and Tian (2013) find that analyst coverage leads managers to focus on near-term earnings and thus hinders innovation. In contrast, Aghion, Van Reenen, and Zingales (2013) find that institutional investors serve to promote innovation and mitigate a short-term focus.
4
making (see Brealey et al., 2014, p.449). As a result, we next explore whether there are negative
economic consequences of using the GAAP ETR in corporate decisions. To conduct this test we
focus on firms that say they use the GAAP ETR for decision-making. We gather GAAP ETR
and estimated MTR measures for these firms over time and associate the GAAP ETR vs. MTR
difference with outcomes of corporate decisions. In this analysis we focus on capital structure,
M&A, and capital investment outcomes because they are important, often-studied corporate
decisions, and because we have empirical proxies from prior research to evaluate the efficiency
of these decisions. Our prediction is that there is a positive association between the amount of
estimated error in the rate employed (i.e., the magnitude of the difference between the GAAP
ETR and the estimated MTR) and the inefficiency of the outcome.6
We find that firms using GAAP ETRs as the tax rate input for their capital structure
decisions adopt a sub-optimal debt policy when their GAAP ETR differs from their estimated
MTR. To assess the dollar cost of these inefficiencies, we trace firm-specific marginal cost of
debt and marginal benefit of debt curves using the approach in van Binsbergen, Graham, and
Yang (2010). The intersection of these cost and benefit curves provides an estimate of the
optimal leverage ratio; the area between the benefit and cost curves, integrated over the distance
from chosen to optimal leverage, provides an estimate of the cost of being under- or over-levered
(see Figure 2). We find that a one percentage point increase in the difference between the MTR
and GAAP ETR leads to a 1.75 percentage point increase in the distance between the optimal
and actual leverage ratios. This increase in the distance between the optimal and actual leverage
ratios leads to a deadweight cost of $1.28 million (or 0.034% of assets), on average. The average
6 An important benefit of our research design is that it exploits time-series differences between the ETR and estimated MTR of the same firm, thereby holding firms’ survey responses constant. As a result, our design mitigates potential response biases induced via the survey instrument. In untabulated analyses, we use an alternative research design that compares the investing/financing behavior of the group of firms that say they use the GAAP ETR with the group of firms that say they use the MTR/STR for decision-making. Consistent with our prediction and the results using our primary research design described above, we find that firms using the GAAP ETR for decision-making are relatively inefficient compared to firms that use the MTR/STR for decision-making (see Section 6.5.2).
5
firm in our sample that responded that their company uses the GAAP ETR has a GAAP ETR that
is 7.7 percentage points away from their MTR, suggesting that the average firm has a deadweight
cost of using the incorrect tax rate of $9.86 million or 0.26% of assets.
We also test investment outcomes. We find that firms that use the GAAP ETR as the tax
rate input for investment decisions are less responsive to their investment opportunities when the
GAAP ETR is different than their MTR. In addition, these firms have acquisition announcement
returns that decrease in the difference between the firm’s MTR and GAAP ETR. For example, a
one percentage point increase in the difference between the MTR and GAAP ETR (for firms
using ETRs for decision making) reduces the average acquisition announcement return by 0.08–
0.10 percentage points, representing a 2.03–2.64% lower average announcement return for the
acquirer. In dollar terms, our coefficients imply that the acquirer experiences a $21–28 million
drop in market capitalization from using the GAAP ETR for M&A decisions.
Our primary results are robust to controlling for a number of variables associated with
firms’ debt and investment policy including firm fixed effects. In addition, we conduct a
falsification test where we examine outcome efficiency for firms that state they use the MTR for
decision making. This test serves to mitigate concerns that our results are spurious, that
measurement error in the MTR proxy (e.g., when firms have foreign income in low-tax
jurisdictions and/or other non-debt tax shields such as employee stock options) is inducing our
results, and other concerns. The intuition is that if firms use the MTR as the tax rate input for
their decisions, then the difference between the MTR and GAAP ETR should be uncorrelated
with their capital structure, investment, and acquisition outcomes. This is exactly what we find.
To summarize, despite the importance of understanding taxes and corporate decision
making, there is little empirical evidence on the extent to which company management actually
use the MTR as the input to capture tax effects when making corporate decisions. We contribute
to the literature by asking managers directly what tax rate they use when making corporate
6
decisions. We then examine potential determinants and investigate whether there are observable
negative outcomes of using the GAAP ETR, a theoretically incorrect rate. Our paper provides
further evidence that there is an association between taxes and corporate decision-making, such
as leverage and investment; associations that have drawn skepticism at times (see Hanlon and
Heitzman (2010) for a review). In addition, our paper goes further to show and quantify the
negative economic effects from managers using the GAAP ETR in their decisions. If our
conjecture is true that the decision to use the GAAP ETR is due to salience, our estimates of such
costs could be interpreted as estimates of the costs of such biases.
Our paper also contributes to the growing body of evidence in psychology and behavioral
economics that agents often deviate from decision rules predicted by standard decision-making
models that assume agents to be fully rational and optimize perfectly. This literature has
predominantly focused on decision-making by individuals in non-corporate settings. DellaVigna
(2009) discusses that the intuition for focusing on individuals rather than firms is that, “firms can
specialize, hire consultants, and obtain feedback from capital markets. Firms are also subject to
competition…therefore, firms are less likely to be affected by biases (except for principle–agent
problems), and we expect them to be close to profit maximization.” (p. 361) However, Camerer
and Malmendier (2007) discuss that even managers could make mistakes or inefficient decisions
that markets do not fully correct when the decisions are infrequent and/or lacking clear feedback
(e.g., acquisitions). They go on to discuss that managers are more likely to make inefficient or
suboptimal decisions when they do not specialize in making those decisions, which is likely the
case when it comes to incorporating the effect of taxes into decisions. We contribute to this
literature by showing that biases that affect individuals also affect managers, as predicted by
Camerer and Malmendier (2007), with respect to their choice of tax inputs for decision-making.7
7 Camerer and Malmendier (2007) summarize research documenting that managers are affected by biases other than those documented here, such as overconfidence and corporate socialism, which also affect corporate decisions.
7
Finally, our findings shed light on the effect of tax policy on corporate behavior.
Governments use tax policy to provide incentives and disincentives for certain actions (e.g.,
investment) and thus the degree to which taxes actually impact corporate decisions determines
whether such policies are effective. Standard economic models assume that firms respond to
changes in their MTR by making the right connection between their income and tax schedule
(Mirrlees, 1971; Atkinson and Stiglitz, 1976; Diamond, 1998). As a result, the effects of tax
changes predicted by standard economic models are likely to be different when managers base
decisions on an average tax rate such as the GAAP ETR rather than the MTR. Our results
provide fresh insights that potentially help clarify the implications of tax policy changes for
corporate behavior.8
The paper proceeds as follows. Section 2 describes our survey methodology and sample.
Section 3 provides descriptive statistics about our sample firms and discusses tests of non-
response bias. Section 4 presents the responses to the survey questions. Sections 5 and 6 examine
the determinants and consequences of managers’ survey responses, and Section 7 concludes.
2. Survey methodology and sample9
We developed an initial survey instrument to ask about taxes in the context of key
corporate decisions. We solicited feedback from several academic researchers, Tax Executives
Institute (TEI) and PricewaterhouseCoopers (PwC) on the survey content and design.10 Survey
Sciences Group (SSG), a survey research consulting firm, assisted with the survey formatting 8 Prior literature has documented the use of average rates by individuals. We discuss this literature below. 9 Our survey has four parts and the data from different parts of this survey are used in Graham, Hanlon, and Shevlin (2010, 2011), and Graham, Hanlon, Shevlin, and Shroff (2014). As a result, the discussion in this section is similar to that in those papers. However, we note that the research questions addressed in these papers are very different than the research question of the current paper. Specifically, Graham et al. (2010) focus on the 2004 American Jobs Creation Act and repatriation decisions in response to that Act. Graham et al. (2011) explore questions concerning the location, reinvestment, and repatriation of foreign earnings. In particular, they examine the effect of an accounting rule, APB 23, on these decisions. Finally, Graham et al. (2014) examine the effects of reputational and financial accounting concerns on tax planning decisions. 10 TEI is an association founded in 1944. Its members are executives responsible for the tax affairs of U.S. and foreign businesses. The member companies are from a wide range of industries.
8
and programmed an online version. We had two executives beta test the survey and we made
revisions based on their suggestions. The final survey contained 64 questions, most with
subparts. The paper version of the survey was 12 pages long. There were many branching
questions and, as a result, many firms were directed to answer only a portion of the questions.
The paper version of the survey is available upon request.
An initial email invitation was sent on August 9, 2007 to the highest ranking tax
executive who is a member of Tax Executives Institute (TEI) at 2,794 firms (thus, only one
invitation was sent to each company); three of these were returned as undeliverable. We also sent
a letter via two-day express mail to fifteen companies for which we did not have email addresses.
Thus, a total of 2,806 companies received invitations to complete the survey. SSG sent three
email reminders throughout August and September. For those who had still not responded, we
then sent a paper version of the survey, along with instructions of how to complete the survey
online, in September and October. We closed the online survey on November 9, 2007.
A total of 804 firms accessed the survey. Sixty of these companies entered no more than
two responses and thus we delete them from our sample, leaving 744 usable responses. The
response rate for our survey is 26%, which compares favorably to many prior survey studies.11
We eliminate 11 firms that indicate they are not subject to the U.S. corporate income tax (i.e.,
businesses not taxed at the entity level, such as S corporations and other flow-through entities).
We also eliminate 29 companies that indicate that they did not file a corporate income tax
return—Form 1120 (under the assumption that these companies are also not taxed at the entity
level). We restrict the sample further by eliminating firms that are subsidiaries of foreign parents
since their corporate decisions and tax planning incentives are likely to be affected by the tax
rules in the parent’s home country. Finally, we lose 95 firms that did not respond to the section
11 For example, Trahan and Gitman (1995), Slemrod and Blumenthal (1996), Graham and Harvey (2001), Slemrod and Venkatesh (2002), Brav et al. (2005), and Graham et al. (2005) report response rates between 10.4% and 21.8%.
9
concerning tax rates and decision-making, the subject of our study. This leaves 500 remaining
firms on which we conduct our analyses. The sample size varies across the individual decisions
(e.g., M&A, capital structure) because of occasional missing responses.
There are caveats and limitations to survey research. First, firms that decide to answer the
survey may systematically differ from those that do not answer the survey. We address this
concern by comparing our survey respondents to the average Compustat firm to get a sense of
the characteristics of our sample firms relative to the typical sample of firms studied in the extant
literature.12 In addition, we also compare firms that responded to the survey with firms that did
not respond. These data are tabulated and discussed below (see Section 3).
Another concern with survey based research is that it is plausible that survey respondents
do not tell the complete truth in their responses. In addition, we may not have asked the
questions clearly, the respondent may not have understood some questions, or perhaps the
respondent just answered questions randomly. There is no way to completely eliminate these
possibilities; however, we attempted to mitigate these concerns by having academics,
practitioners, and a set of beta firms carefully review the survey before it was distributed. We
also employed a professional survey consulting firm to assist in programming the survey online
and to provide advice on how to best ask the questions. Finally, most of our inferences are based
on associations between our survey data and Compustat data, which helps mitigate concerns
related to biases in the survey data.
Finally, an observation worth noting is that our survey participants are corporate tax
executives. It is plausible that tax executives are not directly involved in corporate decision
making and thus are not the ideal survey population for questions concerning investment and/or
financing decisions. This concern is less relevant because our research question is about the
12 Unlike much research based on surveys, we know the identities of the firms that responded (and that did not respond), allowing tests for potential non-response bias.
10
manner in which tax rates are incorporated into decision making. Even if tax executives are not
closely involved in these corporate decisions, they likely provide the tax rates and any associated
data that other managers use as inputs.
3. Descriptive statistics and non-response bias test
We obtain demographic information from both survey data and, for publicly traded firms
additional data from Compustat. Table 1 presents the descriptive statistics of our sample firms.
Survey responses indicate that 78% of our sample firms are publicly listed. The average firm in
our sample has $7.8 billion in assets (Assets), with the average public (private) firm having $9.2
($2.5) billion in assets (untabulated). Survey responses also indicate that the average firm in our
sample has 19% of its assets in foreign locations (Foreign Assets) and has a GAAP ETR of 31%.
Finally, 46.3% of our respondents have a net operating loss carryforward in the U.S. (US NOL).
Additional data from Compustat indicate that the average public firm in our sample has
$5.8 billion in sales (Sales), a market capitalization (MVE) of $8.5 billion and earns a 5.8%
return on assets (i.e., net income scaled by assets; ROA). Tobin’s Q, sales and asset growth (Sales
Growth; Asset Growth) for the average public firm in our sample are 1.97, 14.1% and 13.6%,
respectively. Approximately 45% of the public firms in our sample invest in R&D and the
average R&D Intensity is 2.4%, where R&D Intensity is R&D expense scaled by assets. The
average public firm in our sample is followed by ten analysts (Analyst Following) and has 53.6%
institutional ownership (Institutional Ownership). In terms of tax rate proxies, we find that the
average public firm in our sample has a marginal tax rate before accounting for the interest on
debt (MTR) of 30.9% based on Graham’s methodology and 33.1% based on Blouin, Core, and
Guay’s (2010; BCG henceforth).13 Similarly, the average after-interest marginal tax rates (MTR
13 The two methodologies differ in their approach for forecasting future taxable income. Graham’s MTR methodology assumes that the level of future taxable income follows a random walk with drift. In contrast, the BCG methodology is a nonparametric approach that employs bins of firms grouped by their profitability and size, and then traces the taxable income of these firms into the future.
11
A.I.) are 21.4% and 31.6%, respectively, and the average GAAP ETR is 30.9%.14 The average
unsigned difference between the GAAP ETR and MTR (|MTR – GAAP ETR|) is 13.2% (9.5%)
using Graham’s (BCG’s) MTR methodology and that between the GAAP ETR and MTR A.I.
(|MTR A.I. – GAAP ETR|) is 16.7% (10.1%). All variable are defined in the Appendix.
Table 2 compares firms that responded to the survey with other firms. The analysis in this
table is restricted to public firms only because the data on firm characteristics come from
Compustat. We present descriptive statistics of (i) the average Compustat firm, (ii) the average
firm we sent the survey to, (iii) the average firm that responded to the survey, and (iv) the
average firm that did not respond to the survey. Our average surveyed firm is larger than the
average Compustat firm (measured by Assets, MVE, and Sales) and different than Compustat
firms along most dimensions such as liquidity/profitability (Cash and ROA), growth (MB, Sales
Growth, and Asset Growth), investment intensity, and tax rates (GAAP ETR and MTR). To
further examine the source of the differences between our survey firms and Compustat firms, we
match each survey firm with Compustat firms based on size (i.e., Assets, MVE, and Sales) and
re-examine the differences in firm characteristics. We find that the average survey firm is
statistically indistinguishable from the average Compustat firm along the other dimensions once
we control for size (untabulated). Specifically, we find that the survey firms and size-matched
Compustat firms have similar Leverage, MB, ROA, Asset Growth, Sales Growth, investment
intensity, NOL, 3-Yr Cash ETR, and MTR. Therefore, once one controls for size (which we do in
our analysis below), the survey sample appears representative of Compustat.
Next, we compare respondents to non-respondents (with Compustat data) to test for non-
response bias. We find that the average respondent firm is statistically no different than the
14 The before interest MTR means computing taxable income prior to the deduction for interest expense (i.e., before financing). The after interest MTR means computing taxable income including the interest deduction. See Graham, Lemmon, and Schallheim (1998) for a discussion of when to use before- and after-interest MTRs.
12
average non-respondent firm in terms of size (i.e., Assets, MVE and Sales), Leverage, Cash,
growth (MB, Asset Growth, Sales Growth), investment intensity, acquisition intensity, R&D
intensity, NOL, GAAP ETR, 3-Yr Cash ETR, and MTR. The only statistical difference is that the
respondent firms have, on average, a higher ROA than non-respondent firms. We cannot think of
any biases that arise for our tests because of this difference. Overall, there is little evidence of
non-response bias in our sample, but we recognize that it is always a possibility with survey data.
4. Which tax rates do managers use to incorporate taxes into their decisions?
Finance theory suggests that managers (should) use the MTR to evaluate incremental
decisions because it is the rate paid on an incremental dollar of income earned today. For
example, Graham (1996a, p. 42) states that, “Financial theory is clear that the marginal tax rate
is relevant when analyzing incremental financing [and investing] choices.” In a similar vein, the
popular corporate finance text, Brealey, Myers, and Allen (2014, p. 449), explicitly prescribes
that managers should “Always use the marginal corporate tax rate, not the average rate.” Despite
the widely held belief that managers (should) evaluate incremental corporate decisions using the
MTR, there is little direct evidence that managers indeed do so.
In our survey, we ask the corporate tax executives ‘What is the primary tax rate your
company uses to incorporate taxes into each of the following forecasts or decision making
processes?’ The survey respondent is allowed to choose from the following options (or, write in
an answer if the options given are not sufficient): (i) U.S. statutory tax rate, (ii) GAAP effective
tax rate, (iii) jurisdiction-specific statutory tax rate, (iv) jurisdiction-specific effective tax rate,
(v) marginal tax rate, and (vi) other. The question then listed the following decision contexts for
the respondent to indicate the primary tax rate used in that setting: (i) mergers and acquisitions,
(ii) capital structure (debt versus equity), (iii) investment decisions (property, equip., etc.), (iv)
decision to purchase versus lease (property, equip., etc.), (v) weighted average cost of capital,
13
(vi) where to locate new facilities, and (vii) compensation decisions. The respondent could
choose one tax rate for each decision context. Because we were not certain that tax executives
would hold the same definition of MTRs that exists in academia, we defined in parenthesis the
MTR as “an estimation of the change in the present value of taxes from earning a marginal dollar
of income, where the present value computation takes into account net operating loss carrybacks
and carryforwards.” We acknowledge that defining the MTR and no other rate may induce
response biases. Particularly, it is plausible that by defining MTRs, we heightened respondents’
awareness to MTRs over the other tax rates, thereby increasing the number of executives
choosing the MTR.15 Therefore, our results might possibly represent an upper bound estimate of
the number of firms using the MTR as their tax rate input for decision making.16
Figure 1 and Table 3, Panel A present the survey responses to the above question. The
most popular tax rate managers claim to incorporate into their decision making is the GAAP
ETR and the second most popular is the STR. For example, with respect to capital structure
decisions, survey responses indicate that 30% of firms use the GAAP ETR, 26% use the U.S.
STR, 15% use the jurisdiction specific STR, 15% use the jurisdiction specific ETR, and 12% use
the MTR. Averaging across all decision contexts, the survey responses suggest the following
pattern of tax rate use: 25.8% use GAAP ETRs, 23.1% use STRs, 19.6% use jurisdiction specific
STRs, 17.0% use jurisdiction specific ETRs, 11.2% use MTRs, and 3.2% use some other rate. 15 It is also plausible that defining the rate made it seem complicated and the executive did not choose it as a result. This alternative seems unlikely if the respondents truly use MTRs for decision making. Further, given the education, experience, and expertise of the respondents, it is unlikely that the definition of MTRs would be too complicated. 16 The definition of MTR provided in our survey follows from the Scholes-Wolfson framework (see Scholes et al., 2014). It is plausible that managers use a tax rate that incorporates the Scholes-Wolfson intuition without explicitly computing MTRs per that definition (e.g., a company might have separate tax rates for different economic scenarios – high, medium, and low – and these scenarios might include the effect of NOLs, etc.). If so, the number of firms saying they use the MTR might be understated. We created a provision for such an outcome in our survey by allowing respondents to choose the “other” option and by providing space for them to write-in an answer. Depending on the decision, 11 to 20 firms listed a response in this “other” space. From these we see that one firm indicated that it uses an “Employees marginal tax rate average” for its compensation decisions, and other firms stated they use “Business plan ETR,” “A standard 41%,” “Effective cash tax rate,” and “cash tax rate.” No firm suggested using something that we could interpret as equivalent or similar to the MTR in concept.
14
These results may seem surprising because very few firms use the MTR. However, the
STR closely approximates the MTR for firms that generate high taxable income and foresee
themselves continuing to do so. Indeed, in untabulated analyses, we find that over 80% of our
sample firms that indicate they use the STR for capital structure decisions have a simulated MTR
that is within two percentage points of the STR. Thus, for the majority of firms that say they use
the STR, we find that the STR closely approximates the estimated MTR and thus, this choice is
unlikely to induce costly errors for these firms. To the extent the STR and the jurisdiction
specific STR closely approximate the MTR of firms, our survey results suggest that over 50% of
our survey respondents use tax rates that are consistent with the predictions of standard finance
theory (e.g., Graham, 2003; Brealey et al., 2014; Scholes et al., 2014).
Another noteworthy observation is that when making a location specific decision, the
majority of managers answered that they use jurisdiction specific rates (statutory and effective)
as the tax rate input for decision making. For example, over 54% of respondents indicated they
use a jurisdiction specific rate for the location of new facilities. In contrast, only 30% and 26% of
respondents say that they use a jurisdiction specific rate for capital structure decisions and
weighted average cost of capital computations, respectively. This change in responses across the
factors is consistent with expectations and suggests that the respondents carefully considered the
questions and varied their answers when appropriate (i.e., the respondents did not provide the
same response for each factor without thought).
The most puzzling finding in the table is likely that many firms claim to use a form of
ETR as the tax rate input in all decision making contexts examined. Specifically, in our sample
of firms and policies examined, approximately 17% to 34% of respondents (or 40% to 47% when
including use of a jurisdiction specific ETR) indicate that their firm uses the GAAP ETR as the
tax rate input for decision making. This result is puzzling because the GAAP ETR is an average
15
tax rate based on financial accounting data and as such has very little theoretical basis to use as
the tax rate input. An average tax rate (such as the GAAP ETR) is unlikely to be the rate that is
paid on the company’s marginal transaction but instead is a (GAAP accounting) estimate of the
average tax rate paid on all transactions combined. We discuss potential reasons for this choice
in the next section (Section 5) and test whether this choice is associated with inefficient decision-
making in Section 6.
Table 3, Panel B presents correlations of tax rate choices across the different decision
contexts. As expected, we find that the tax rate choices are highly correlated across the different
decision contexts presented to them. The correlation coefficients range from a low of 0.66 to a
high of 0.93. The highest correlation is between the tax rates used in the purchase vs. lease
decision and capital investment decisions and the lowest correlations are between the tax rates
used in the compensation and the other decisions (correlation ranging from 0.66 to 0.77). Given
the high correlations, for the remainder of our analyses, we focus on firms’ capital structure,
capital investment, and acquisition decisions for the sake of brevity; these are important
decisions, and the literature provides measures to evaluate the outcomes of these decisions.
Further, in our analyses of the determinants of managers’ tax rate choice (in the next section), we
combine firms indicating that they use the jurisdiction specific STR (ETR) for decision making
with those using the U.S. STR (GAAP ETR) in the interest of parsimony.
5. Determinants of managers’ tax rate choice
To better understand the factors affecting tax rate choices, we examine the association
between managers’ survey responses and firm characteristics chosen largely based on research in
psychology and behavioral economics. We posit that managers’ tax rate choices are affected by
the following: (i) the complexity of calculating the tax rate and the availability of simple
16
heuristics that (are equal to or) approximate the expected tax costs, (ii) the salience of different
tax rates to managers, (iii) the likely availability of tax planning opportunities, and (iv) the
existence of governance mechanisms to monitor managers. We recognize that the factors we
examine do not comprise a complete list of all the factors that affect managerial tax rate choices;
we view our analyses in this section as an initial step to understand managers’ tax rate choices.
Heuristics: Since Simon’s (1955) early work, research in behavioral economics
recognizes that managerial decision-making often falls short of the purely rational model (see
Barberis and Thaler (2003), Camerer and Malmendier (2007), and DellaVigna (2009) for
reviews of the literature). Research trying to explain deviations from rational decision-making
mostly focuses on biases and heuristics to explain the deviations (Tversky and Kahneman, 1974),
where biases and heuristics are decision rules, cognitive mechanisms, and subjective opinions
people use to assist in making decisions. Kahneman et al. (1982) indicate that managers rely on
biases and heuristics when the decision requires complex calculations and when there is high
uncertainty. They also indicate that the application of heuristics often yields acceptable solutions
that are not too far from a fully rational solution.
Computing a firm’s MTR requires information about the firm’s taxable income well into
the future, the applicable statutory tax rate, the presence of NOL carryforwards, the number of
years before NOLs will be utilized, and whether the firm will have to pay any alternative
minimum tax. As a result, MTRs are very complicated to calculate for many firms; it is precisely
in such a circumstance that managers are likely to rely on heuristics for decision making
(Kahneman et al., 1982; Gabaix et al., 2006). Thus, when viewed through the lens of behavioral
economics theories, it is perhaps not surprising that such few companies use the MTR for
decisions making. Some evidence consistent with the use of a heuristic is found in our data as
discussed above; the majority of our firms that indicate they use the STR have a simulated MTR
that equals the STR and over 80% have simulated MTRs that are within 2% of the STR
17
(untabulated). To further examine this heuristic hypothesis, we compare the tax rate used by
firms with significant foreign operations with that of firms without significant foreign operations.
Our prediction is that firms with more foreign operations are less likely to use MTRs for decision
making and more likely to use heuristics that approximate firms’ MTRs (e.g., a jurisdiction-
specific STR). Our intuition is that foreign operations increase the complexity and uncertainty of
computing MTR estimates (because of differences in the likelihood and timing of foreign
earnings repatriations, and cross-country differences in the tax rates/schedules, transfer pricing
rules, tax policy uncertainty, etc.), which increases the likelihood that managers use a heuristic
(Kahneman et al., 1982).
Salience: Prior research finds evidence that individuals often use average tax rates instead
of marginal tax rates when faced with marginal investment choices. For example, de Bartolome
(1995) investigates which tax rates individuals use when making marginal economic decisions.
He finds that many people use the average tax rate “as if” it were the marginal tax rate. He
investigates the cause via a controlled experiment and concludes that they do this because of
salience; if the tax table is changed to stress the marginal tax rate, the marginal tax rate is used.
Similarly, Finkelstein (2009), Chetty (2009), and Chetty et al. (2009) find evidence that tax
salience affects behavior at the individual level. Most prior research on the effects of tax salience
concerns decisions at the individual level rather than decisions by corporate managers. If
managers are more sophisticated decision-makers than individuals, a salience bias might not
affect managerial decision-making (DellaVigna, 2009). However, Faulhaber and Baumol (1988)
find that most companies base their pricing decisions on average costs, not marginal costs (at
least in the 1970s), suggesting that corporate entities are at times no more sophisticated than
individuals. Camerer and Malmendier (2007) discuss some reasons why behavioral biases are
unlikely to be eliminated in a corporate setting or remedied by organization design and
18
governance choices. Thus, the salience explanation is plausible in our setting because the GAAP
ETR and STR are very salient whereas the MTR is not.
To examine the salience hypothesis, we compare the tax rates used by public and private
firms. Our prediction is that public firms are most likely to use the GAAP ETR because it is
more salient for them. Prior research finds that top management of public firms view GAAP
earnings as the most important performance metric of a firm (Graham et al., 2005) and that tax
executives are often incrementally compensated for lowering GAAP ETRs and often view the
GAAP ETR as a more important metric than cash taxes paid (Armstrong, Blouin, and Larcker,
2012; Graham et al., 2014). The focus on GAAP earnings and by extension GAAP ETRs is
likely to make the ETR significantly more salient to the decision-makers in public firms. In
contrast, prior research finds that private firms are significantly less focused on GAAP based
numbers than public firms (Graham et al., 2005; Graham et al., 2014), making them less likely to
use the GAAP ETR as their tax rate input for decision making. We also use additional variables
such as analyst following as a proxy for capital market pressure (He and Tian, 2013), and a
survey question on whether managers are more concerned about the GAAP ETR or cash taxes
paid to examine whether salience is a potential determinant of the tax rate choice.17
Tax planning opportunities: We conjecture that firms with greater tax planning
opportunities are more likely to be tax savvy and hire well-trained tax personnel because they
can derive greater benefits from the tax planning. Further, tax savvy firms are more likely to
17 Another potential source of salience is that managers are often taught to use the GAAP ETR or STR because those rates are salient to professors. For example, when forecasting earnings in financial statement analysis classes the GAAP ETR is (and should be) used and this is perhaps carried over to the executives’ (i.e., former students’) marginal decision-making in other contexts. Similarly, most managerial accounting textbooks just assume a tax rate of 35 or 40%. While some corporate finance texts mention the MTR and discuss the general effects of losses, at least for the capital structure decision, they do not generally go into detail about how to compute the MTR. For example, in Horngren, Datar, Foster, Rajan, and Ittner's (2009) text, taxes are provided for in a discounted cash flow analysis by applying a 40% tax rate stating “the income tax rate is 40 percent of operating income each year” (p. 744). Brealey et al. (2014) use a flat 35% in their computation but include a footnote that states “Always use the marginal corporate tax rate, not the average rate” (p. 449).
19
accurately compute MTR estimates and use the MTR (or STR when appropriate) rather than the
GAAP ETR for decision-making. To test this conjecture, we proxy for a firm’s tax planning
opportunity set using firm size and R&D intensity. Our intuition for these proxies is as follows.
First, larger firms benefit from economies of scale of tax planning, and thus are more likely to
invest in well-trained/sophisticated tax personnel (Rego, 2003). Specifically, large firms (by
definition) engage in more business activities than small firms, which potentially allows them to
avoid taxes through intercompany transactions, tax-advantaged leasing and financing
arrangements, and the use of flow-through entities such as partnerships and real estate
investment trusts among other things. Consistent with this argument, Mills, Erickson, and
Maydew (1998) find that larger firms have lower average costs of tax planning. And second, we
use R&D intensity as a proxy for firms’ tax planning opportunities based on the intuition that
R&D generates intellectual capital that can be transferred to low tax jurisdictions to save taxes
(Grubert and Slemrod, 1998). Further, R&D investments generate tax credits that can be offset
against current and future income. Thus, all else equal, high R&D firms are likely to have more
tax planning opportunities than low R&D firms.
Monitoring and competitive pressures: Prior research finds that external monitoring
mechanisms such as institutional investors and product market competition serve to discipline
managers and curb agency problems, thereby making the firm more efficient. For example, prior
research finds that institutional ownership is associated with more efficient outcomes such as
greater R&D and innovation (Bushee, 1998; Aghion, Van Reenen, and Zingales, 2013), better
acquisitions and post-acquisition performance (Chen, Harford, and Li, 2007), greater pay-for-
performance sensitivity of executive compensation (Hartzell and Starks, 2003), more accurate
disclosure (Chung, Firth, and Kim, 2002; Ajinkya, Bhojraj, and Sengupta, 2005; Shroff, Sun,
White, and Zhang, 2014), and greater firm value and overall performance (McConnell and
20
Servaes, 1990; Cornett, Marcus, Saunders, and Tehranian, 2007). These papers suggest that
institutional investors monitor managers and thus reduce agency problems.
Similarly, Shleifer and Vishny (1997, p. 738) argue that “product market competition is
probably the most powerful force towards economic efficiency in the world.” The intuition is
that competition makes it costlier for firms to forego long-run value since such actions would
increase the likelihood of bankruptcy. As a result, competition increases managerial focus on
firm value and reduces their incentives and ability to slack (or to lead a “quiet life”). A number
of studies find evidence consistent with this intuition in a variety of settings (see e.g., Nickell,
1996; Bernard, Jensen, and Schott, 2006; Giroud and Mueller, 2010; Shroff et al., 2014). Based
on the above, we conjecture that firms with greater institutional ownership and firms operating in
more competitive environments are more likely to use MTRs (or STRs when appropriate) for
decision-making and/or are less likely to use ETRs for decision-making.
We present univariate (Table 4) and multivariate (Table 5) tests that examine the above
conjectures. In the multivariate tests, we present results from nine regressions where the
dependent variable is an indicator variable for the manager’s tax rate choice in each decision
context and the independent variables include all firm characteristics examined in Table 4 as well
as some additional characteristics (not tabulated in Table 4 for brevity). The sample in Panel A
of Tables 4 and 5 is comprised of both public and private firms and thus the partitioning (and
independent) variables are limited to firm characteristics obtained from the survey. The sample
in Panel B in these tables is comprised of just public firms and uses data from both the survey
and Compustat. Below, we discuss the results in both tables by grouping them as being
consistent or inconsistent with the conjectures put forward in the discussion above.
Our first conjecture is that firms are more likely to rely on simple heuristics (i.e., STR)
for decision making when the purely rational decision rule (i.e., MTR) is complex and uncertain.
Consistent with conjecture, we find that firms with a large proportion of their assets in foreign
21
locations are significantly less likely to use the MTR and more likely to use the STR (in this case
often a jurisdiction specific STR) as their tax rate input for decision making (see Tables 4 and 5,
Panels A and B).18 This result is consistent with the hypothesis in Kahneman et al. (1982).
Consistent with our second conjecture that tax rate salience affects managers’ tax rate
choice, we find that private firms are significantly more likely to use STRs in their decision
making process, whereas public firms are significantly more likely to use the GAAP ETR in their
decision making process (Panel A in Tables 4 and 5). To further explore the salience explanation
for this result, we also partition our sample firms based on a survey question: “Which metric is
more important to top management in your company?” The possible answers to the question are
(i) GAAP ETR, (ii) cash taxes paid, or (iii) both are equally important. Both Panels in Table 4
show that when the top management view the GAAP ETR as more important than cash tax taxes
paid, their firm is more likely to use the GAAP ETR for decision making (however this
association is weaker in the multivariate tests in Table 5). This result, although somewhat weak,
is consistent with a salience based explanation for why managers use the GAAP ETR as their tax
rate input for decision making. Finally, Table 5, Panel B shows that firms with high analyst
following are more likely to use the GAAP ETR and less likely to use the MTR/STR for
decision-making. To the extent financial analysts increase capital market pressure (He and Tian,
2013) and thus increase managerial focus on financial accounting earnings, this result is also
consistent with a salience explanation leading firms to use the GAAP ETR for decision-making.
Another noteworthy observation in Panel B in Tables 4 and 5 is that when the magnitude
of the differences between the MTR and GAAP ETR (|MTR – GAAP ETR|) is large, firms are
more likely to use the GAAP ETR as their tax rate input. Kahneman et al. (1982) suggest that
18 In untabulated analyses, we proxy for foreign operations using foreign income rather than foreign assets and find similar results. We use foreign assets in the tabulated analyses because we obtain these data for both public and private companies from our survey instrument, whereas data on foreign income is obtained from Compustat and thus is available only for public companies.
22
managers are more likely to rely on heuristics when the heuristic approximates the “fully rational
solution” and serves as an efficient approach to reach the rational solution. Our result that
managers are more likely to use the GAAP ETR when it is farther away from the MTR suggests
that firms do not treat the GAAP ETR as an easy-to-compute heuristic that is to be used as it
converges with the MTR. Rather, it appears that the decision to use the GAAP ETR is a result of
other factors/biases (such as salience) but not heuristics.
Our third conjecture is that firms with greater tax planning opportunities (proxied by firm
size and R&D intensity) are more likely to be tax savvy and thus are more likely to use the MTR
(or STR when appropriate) and less likely to use the GAAP ETR for decision-making. Tables 4
and 5 show that conditional on being public, larger firms and high R&D intensity firms are more
likely to use the MTR and/or STR for decision-making. In addition, larger firms and high R&D
intensity firms are less likely to use the GAAP ETR for decision making.19 We interpret these
results as suggesting that firms with greater tax planning opportunities being more tax savvy and
thus more (less) likely to use MTRs or STRs (GAAP ETRs) for decision making.
Our final prediction is that firms with greater external monitoring (i.e., institutional
ownership and competitive pressure) are more likely to use the MTR and less likely to use the
ETR for decision making. Table 5, Panel B shows that firms with high institutional ownership
are significantly more likely to use the MTR for capital structure and M&A decisions, and less
likely to use the GAAP ETR for M&A decisions. This finding is consistent with such investors
monitoring managers, thereby providing them additional incentives to maximize firm value.20
19 In Table 4, Panel A we find that firm size is unrelated to managers’ tax rate choice. This is perhaps because the univariate relation between firm size and tax rate choice is confounded by the effect of ownership on tax rate choice (i.e., public firms are larger than private firms and are more likely to use the GAAP ETR due to salience; but conditional on being public, larger firms are less likely to use the GAAP ETR because such firms have more tax planning opportunities and thus are more likely to be tax savvy). 20 In untabulated analyses, we decompose institutional ownership using Bushee’s (1998) classification. Bushee (1998) classifies institutional investors into three groups based on their trading patters: “quasi-indexers” are
23
Finally, we find little evidence of competition affecting a firm’s tax rate choice. For example,
based on our primary measure of competition, the Hoberg and Phillips’s (2013) text-based
industry concentration index, we find no statistically significant evidence that competition affects
managers’ tax rate choice. Moreover, in untabulated tests, we use a number of additional
competition proxies (e.g., the Census based concentration measure from Ali, Klasa, and Yeung
(2009), the 10-K disclosure measure from Li, Lundholm, and Minnis (2013), and the traditional
Herfindhal index) and find no association between these proxies and managers’ tax rate choice.
In Table 5, Panel B, we also examine the association between tax rate choices and a
number of additional firm characteristics (these are untabulated in Table 4 for brevity). We find
little evidence that our proxy for a firm’s tax complexity, ROA, leverage, NOLs, and intangible
intensity affects a manager’s tax rate choice. The lack of evidence could either be because of
limitations in our empirical proxies or because these factors truly do not affect tax rate choices.
Overall, the results in Tables 4 and 5 are consistent with (i) managers’ relying on
heuristics, such as the STR, for decision-making when it is likely to approximate the MTR and
when computing the MTR is more complex, (ii) a salience effect leading managers to use the
GAAP ETR for decision-making, and (iii) larger firms, firms with more R&D, and firms with
high institutional investor ownership using (not using) the MTR (GAAP ETR) for decision-
making. However, we caveat that the findings in this section are somewhat exploratory and
based on fairly rough proxies for our constructs of interest. The discussion above should be
interpreted in light of these factors. institutions that follow an indexing strategy; “dedicated” institutions have concentrated holdings and do not trade much; and “transient” institutions are diversified investors that trade often in and out from individual stocks. Bushee (1998) finds that transient investors induce myopic behavior but the other two do not. We find that only ownership by “quasi-indexers” is associated with managers’ tax rate choice and “dedicated” and “transient” investor ownership is not significantly associated with managers’ tax rate choice. To the extent quasi-indexers follow an indexing strategy and thus are unable to sell their shares, they have stronger incentives than the other types of institutions to monitor management to ensure it is acting in the long-term interests of the firm.
24
6. Economic consequences of tax rate choices
In this section, we examine whether using the GAAP ETR as the tax rate input leads to
inefficient corporate decisions. One approach to test for such inefficiencies (that we adopt in our
additional analyses section below) is to directly compare the outcomes of corporate decisions of
companies that say they use the GAAP ETR for decision-making with those of companies that
say they use the MTR or STR (henceforth, the cross-sectional approach). However, testing the
direct relation between managers’ tax rate choices and the outcomes of their corporate decisions
has two important weaknesses. First, it is plausible that the manner in which we present or phrase
our survey questions (e.g., defining MTRs) leads to response biases that are systematically
correlated with the manner in which managers make corporate decisions. For example, by
defining the MTR in our survey, we made it more salient to managers and plausibly influenced
some managers to respond that they use the MTR for decision making. It is plausible this bias
towards salient factors also influences the same managers when making corporate decisions.
Thus, if managers’ response biases in the survey are indicative of their biases when making other
corporate decisions, we cannot draw reliable inferences about the consequences of managers’ tax
rate choices from observing an association between survey responses and corporate outcomes.
Second, it is also plausible that responses to the survey questions are reflective of certain
firm characteristics that are correlated with corporate decisions. For example, the availability of
tax planning opportunities is likely to affect managers’ survey responses and is also likely to
affect corporate decisions. The existence of such factors that are correlated with both managers’
tax rate choices and their corporate decisions potentially exposes inferences based on the cross-
sectional approach to an omitted variable bias. Finally, notwithstanding the above challenges, we
note that empirically identifying inefficient corporate decisions is difficult and the proxies used
in prior research to measure inefficiencies are noisy and prone to measurement error. If
25
measurement error in these proxies is correlated with managers’ survey responses, any inference
drawn from observing associations between these proxies is problematic (if the survey responses
are uncorrelated with measurement error in our proxies, the coefficients are biased toward zero).
We choose a research design to mitigate the above concerns. Specifically, we restrict our
analyses in this section to just those firms that use the GAAP ETR as the tax rate input for
decision making. We then exploit time series variation in the difference between a firm’s MTR
and its GAAP ETRs to draw our inferences. In other words, rather than comparing firms that
provide different responses to our survey question, we compare the same firm at different points
in time.21 We predict that firms using the GAAP ETR as their tax rate input for decision making
are likely to make better decisions when their GAAP ETR is close to their MTR. However, as
the difference between the MTR and GAAP ETR becomes larger, these firms’ decisions are
likely to become more inefficient. The benefits of our chosen research design are two-fold: First,
and perhaps foremost, inferences based on the above research design are unaffected by
(potential) response biases induced by our survey. Second, because we only exploit variation in
the difference between a firm’s MTR and its GAAP ETR, our inferences are less likely to be
affected by any unaccounted for relation between firms’ tax rate choices and their characteristics.
We merge survey data for firms that say they use the GAAP ETR as the tax rate input
with Compustat data from 1997 to 2006, thereby creating a ten year panel. We measure GAAP
ETRs for each firm-year as total income-tax expense scaled by pre-tax book income and we
measure MTRs for each firm-year using the approaches developed by (i) Shevlin (1987, 1990)
and Graham (1996a) and (ii) Blouin et al. (2010). We recognize that both MTR proxies have
measurement error and devise a falsification test to mitigate this (and other) concern(s).
21 An implicit assumption with this research design is that the rate firms say they use in the survey is the rate they use in the years we include in the sample (i.e., that this choice is time invariant during our sample period). To mitigate concerns that this assumption affects are inferences (and for the sake of completeness), we verify the robustness of our inferences to cross-sectional testing approach described above in Section 6.5.2.
26
6.1. Capital structure consequences
We begin by examining whether firms that use the GAAP ETR as the tax rate input for
capital structure decisions end up using debt either conservatively or aggressively. If a firm’s
GAAP ETR is higher (lower) than its MTR and the GAAP ETR is used for decision making,
then such a firm is likely to overestimate (underestimate) the tax benefits of debt and have a
relatively aggressive (conservative) debt policy. We use the approaches in Graham (2000) and
van Binsbergen et al. (2010) to measure how aggressively firms use debt and how far firms are
from their optimal leverage ratios. The approach in Graham (2000) uses information in a firm’s
marginal tax benefit of debt function to evaluate the extent to which the firm’s capital structure
policy is aggressive or conservative. Whereas, the approach in van Binsbergen et al. (2010)
simulates both the marginal cost and tax benefit of debt functions to identify a firm’s optimal
capital structure. Although the van Binsbergen et al. (2010) method provides a more
comprehensive estimate of the economic cost of not being at the optimal leverage ratio, the
estimation procedure requires more assumptions than the approach in Graham (2000). Thus, we
conduct both tests below.
The first test approach is based on Graham’s (2000) “debt kink” calculation. Graham
(2000) constructs a firm’s tax benefit function using a series of hypothetical marginal tax rates,22
with each rate corresponding to a specific level of interest deductions. The benefit function is
initially flat for small interest deductions but eventually becomes downward sloping as interest
deductions increase (see Figure 2). This result occurs because interest deductions reduce taxable
income, which decreases the probability that a firm will be fully taxable in all current and future
states, which in turn reduces the tax benefit from the incremental deductions. Graham (2000)
22 The MTR measures the tax savings benefit of being able to deduct interest. For example, a firm facing a 35% MTR will save $0.35 in taxes if it is able to deduct $1 in interest from taxable income.
27
quantifies how aggressively a firm uses debt with the “kink” in the firm’s tax benefit function
(i.e., the point where marginal benefits begin to decline and thus the function begins to slope
downward) (see Figure 2). Specifically, the “kink” is defined as the ratio of the amount of
interest required to make the tax benefit function slope downward to actual interest expense. If
kink is less than one, a firm operates on the downward-sloping part of its tax benefit function and
is considered to use debt aggressively because it expects reduced tax benefits on the last portion
of its interest deductions. If kink is greater than one, a firm could increase interest expense and
expect full benefit on these incremental deductions; such a firm can be thought of as using debt
conservatively. Debt conservatism increases with kink. We compute the tax benefit function and
the “debt kink” using the procedures from both Graham (1996a) and Blouin et al. (2010).
The second test approach is based on van Binsbergen et al. (2010), the authors of which
estimate company-specific optimal capital structure as occurring at the intersection of the
marginal benefit and marginal cost of debt functions. These authors simulate the marginal
benefit of debt function using the approach described above (Graham, 2000). The authors use
shifts in the benefit function to observe a series of optimal cost/benefit intersection points, and
infer the marginal cost of debt function by statistically “connecting the dots” provided by the
intersection points.23 Once we identify a firm’s ‘optimal’ or ‘equilibrium’ leverage ratio at the
intersection of the marginal benefit and cost curves, we then normalize this equilibrium leverage
ratio using the observed leverage ratio to construct an Equilibrium Factor. An Equilibrium
Factor of one implies that the firm is at equilibrium and its optimal leverage ratio is 100% of its
actual. An Equilibrium Factor of 1.2 implies that the firm’s optimal leverage ratio is 20% larger
than its actual leverage ratio (the firm is underlevered). Similarly, an Equilibrium Factor of 0.8,
23 An important assumption in the van Binsbergen et al. (2010) approach is that the cost function remains constant as the benefit function shifts. The authors accomplish this by including control variables designed to hold the cost function fixed and by observing exogenous variation in benefit functions induced by tax regime changes.
28
implies that the firm’s optimal leverage ratio is 20% smaller than its actual leverage ratio (the
firm is overlevered).
To test our prediction that differences between the MTR and GAAP ETR lead to debt
policies that are either too aggressive or too conservative, we estimate the following regression:
, , ′ , (1)
where Debt Kink is our measure of debt policy conservatism and Equilibrium Factor is the
optimal capital structure normalized by the actual capital structure such that larger (smaller)
values of Equilibrium Factor imply that the firm is underlevered (overlevered). The signed
difference between firms’ before-interest MTR and their GAAP ETRs (MTR – GAAP ETR) is the
primary independent variable of interest.
We use before-interest MTRs in this analysis because the dependent variables, Debt Kink
and Equilibrium Factor, are based on the cumulative financial policy of all financing decisions
rather than just the marginal financing decision. Before-interest MTRs remove the effect of past
financing decisions (that are still part of existing capital structure) and are appropriate to use
when the dependent variable is a stock measure of capital structure (like the debt kink and
optimal leverage ratio).24 In the regression above, i indexes firms, t indexes years, and are
industry and year fixed effects, X is a vector of control variables that includes a number of
nontax factors that affect debt policy, following Graham (2000) or other factors affecting optimal
capital structure (van Binsbergen et al., 2010). is predicted to be positive: when MTR – GAAP
ETR is large, the MTR is larger than the GAAP ETR, leading firms that use the GAAP ETR to
underestimate the tax benefit of debt, and thus leading such firms to use debt too conservatively
as captured by a higher kink/equilibrium factor. Conversely, when MTR – GAAP ETR is small,
24 In contrast, after-interest MTRs are based on income after interest is deducted and are the appropriate MTRs in most other settings, including those in which the dependent variable measures future incremental investing and financing choices (see Graham, Lemmon, and Schallheim, 1998; Graham and Mills, 2008).
29
the MTR is smaller than the GAAP ETR, leading firms that use the GAAP ETR to overestimate
the tax benefit of debt, and thus leading such firms to use debt too aggressively as captured by a
smaller kink/equilibrium factor. We cluster standard errors by firm.
Table 6, Panel A presents the descriptive statistics for the variables used in the Debt Kink
regression and Table 6, Panel B presents the regression results when Debt Kink is the dependent
variable. For brevity, we do not discuss the descriptive statistics. In Panel B, we find that the
coefficient for MTR – GAAP ETR is positive and statistically significant at the 5% level using
both MTR proxies, Graham (1996a) and Blouin et al. (2010). These coefficients suggest that as
the distance between a firm’s MTR and GAAP ETR widens such that the firm’s GAAP ETR is
higher (lower) than its MTR, the firm adopts a relatively aggressive (conservative) debt policy,
which is consistent with our prediction.
Table 7, Panel A presents the descriptive statistics for the variables used in the
Equilibrium Factor regression. In addition to presenting the regression results, we tabulate the
mean and median Equilibrium Factor of the firms in each quartile of MTR – GAAP ETR (Table
7, Panel B). This panel not only shows us the relation between MTR – GAAP ETR and
Equilibrium Factor but also tells us if the mean/median firm across the distribution of MTR –
GAAP ETR is under- or over-levered (since an Equilibrium Factor smaller (larger) than one
implies that the firm is overlevered (underlevered)). Table 7, Panel B shows that there is a near
monotonic increase in Equilibrium Factor as we move from the first to the fourth quartile of
MTR – GAAP ETR. Further, the table shows that the median firm in the 1st quartile of the MTR –
GAAP ETR distribution is overlevered but the median firm in the remaining three quartiles is
underlevered. The near monotonic relation between MTR – GAAP ETR and Equilibrium Factor
combined with the observation that Equilibrium Factor is greater than one in the top three
quartiles of MTR – GAAP ETR tells us that firms using the GAAP ETR for capital structure
30
decisions become more underlevered as MTR becomes larger than the GAAP ETR, consistent
with our prediction.25
Table 7, Panel C presents the results from a regression of Equilibrium Factor on MTR –
GAAP ETR. Consistent with the univariate results in Panel B, we find that the coefficient for
MTR – GAAP ETR is positive and statistically significant at the 5% level. The coefficient
estimate for MTR – GAAP ETR in the multivariate regression is 1.75, which tells us that a one
percentage point increase in the difference between the MTR and GAAP ETR leads to a 1.75
percentage point increase in Equilibrium Factor. Consider a hypothetical firm that uses the
GAAP ETR for decision making and is currently at its optimal leverage ratio with its GAAP
ETR equal to its MTR. Let’s say the GAAP ETR for this firm decreases from 35% to 34% while
the MTR is remains constant. In this case, our regression coefficient predicts that the firm’s
actual leverage ratio will move from the optimal to being only 98.25% of the optimal (the firm
would be underlevered as a result).
To get a better sense of the magnitude of the loss in firm value as a result of using the
GAAP ETR instead of MTR, in Panel D we change the dependent variable in the above
regression to the total deadweight cost of being under- or over-levered, where Total Deadweight
Loss is measured as the area between the cost and benefit curves when a firm has more/less debt
than recommended by our model (see Figure 2). We use the absolute value of the difference
between MTRs and ETRs in this regression because the dependent variable combines the loss
from being underlevered for some firms with the loss from being overlevered for other firms.26
25 That most firms are underlevered is also consistent with prior research (van Binsbergen et al., 2010; Korteweg, 2010) and the observation that the cost of being overlevered is asymmetrically higher than the cost of being underlevered (Leary and Roberts, 2005; van Binsbergen et al., 2010). 26 However, in untabulated analyses, we find that the signed difference between the MTR and ETR is positively (negatively) associated with the deadweight cost of being overlevered (underlevered), which is consistent with our expectation. Interestingly, we find that the coefficient for MTR – GAAP ETR is larger in magnitude when the dependent variable is the deadweight cost from being overlevered. This result is consistent with the asymmetrically higher loss of overleverage relative to underleverage (Leary and Roberts, 2005).
31
Total Deadweight Loss is reported as a percentage of book value in perpetuity; for example, a
loss of 5% would occur if the annual loss was 0.5% and the discount rate was 0.10.27 The
average Total Deadweight Loss for our sample firms is 1.1% (from Table 7, Panel A), which is
consistent with these firms being off equilibrium (underlevered), on average.
In Table 7, Panel D we find that the relation between |MTR – GAAP ETR| and Total
Deadweight Loss is positive and significant (as expected). The coefficient for |MTR – GAAP
ETR| suggests that a one percentage point increase in the difference between MTRs and ETRs
leads to a 3.8 percentage point increase in the Total Deadweight Loss. In dollar terms, this result
implies that a one percentage point increase difference between MTRs and ETRs leads to $1.28
million increase in deadweight cost.28 Since the average difference between the MTR and ETR
(for firm that responded that their company uses the GAAP ETR) is 7.7 percentage points, our
results suggest that the average firm experiences a deadweight loss of $9.86 million or 0.26% of
assets for making the incorrect capital structure decision. Collectively, the results in Tables 6 and
7 suggest that using the GAAP ETR for decision-making reduces firm value.
6.2. Capital investment consequences
We next examine whether firms that use the GAAP ETR as the tax rate input for capital
investment decisions make inefficient capital investment decisions. The intuition is that
differences between the MTR and GAAP ETR can lead firms to incorrectly forecast the after-tax
cash flows from their current and potential investments. Incorrect cash flow forecasts are likely
to create biases in firms’ evaluation of their existing investments’ NPV as well as that of their
investment opportunities, leading to too little investment in high NPV projects and/or too much
27 As indicated in the Appendix, we obtain data on Equilibrium Factor and Total Deadweight Cost from van Binsbergen et al. (2010). Also note that these authors use the Moody’s average corporate bond yield as the discount rate for all firms in a given year. 28 The dollar magnitude is computed by multiplying the average the book value of assets ($3.8 billion) and the change in the Total Deadweight Loss (1.138%-1.10%) from a one percentage point increase in |MTR – GAAP ETR|.
32
investment in low (or negative) NPV projects, and thus leading to a weaker relation between
investment and investment opportunities (i.e., inefficient decisions). For example, if a firm’s
GAAP ETR is higher (lower) than its MTR and the GAAP ETR is used for decision making,
then such a firm is likely to underestimate (overestimate) the after-tax cash flows from investing
and thus underinvest (overinvest). To test this prediction, we examine the association between
the difference between the MTR and GAAP ETR and responsiveness of investment to
investment opportunities, which prior research interprets as an outcome of more efficient
investment behavior (Hubbard, 1998; Bekaert, Harvey, Lundblad, and Siegel, 2007; Badertscher,
Shroff, and White, 2013; Asker, Farre-Mensa, and Ljungqvist, 2014). Specifically, we estimate
the following equation using OLS in the tradition of Q-theory of investment (Fazzari et al. 1988):
, 1_ , 1 2
| . . | , 1
3| . . | _ , 1 , (2)
where CAPEX is capital expenditures scaled by lagged total assets, INV_OPP is our proxy for
investment opportunities (discussed below), |MTR A.I. – GAAP ETR| is the unsigned difference
between firms’ simulated after-interest MTRs and their GAAP ETRs, CFO measures a firm’s
cash flows from operations scaled by total assets, and ( ) are industry (year) fixed effects.
Investment efficiency is estimated by the coefficient for the INV_OPP variable, β1, which is
predicted to be positive. The coefficient of interest in the above equation is β3, which captures
the incremental sensitivity of investment to investment opportunities for firm-years with larger
differences between the MTR and ETR. As any under- or over-investment represents investment
inefficiency, β3 is predicted to be negative.
We use the absolute value of the difference between MTR A.I.s and GAAP ETRs in this
analysis because both over- and under-investment reduces the sensitivity of a firms’ investment
to its investment opportunities. For example, firms that increase investment when their growth
33
opportunities are declining (overinvestment firms) are likely to have lower investment-growth
opportunity sensitivities, and firms that decrease investment when their growth opportunities are
increasing (underinvestment firms) are also likely to have lower investment-growth opportunity
sensitivities. As a result, we use the absolute value of the difference between MTRs and ETRs in
this analysis. Also note that we use an after-interest estimate of MTR in the above regression
(rather than a before interest estimate as we did for our analyses of capital structure decisions)
because the cash flows from the marginal investment decision are subject to the firm’s marginal
tax rate after deducting interest. In other words, capital investment decisions are incremental
decisions and should be analyzed using the tax rate applicable to incremental decisions, which is
the after-interest MTR (Graham et al., 1998; Graham and Mills, 2008).
We proxy for investment opportunities using Tobin’s Q and Sales Growth following a
long list of prior studies (Wurgler, 2000; Whited, 2006; Bloom, Bond, and Van Reenen, 2007;
Badertscher et al., 2013; Shroff et al., 2014). While these proxies for investment opportunities
are measured with error (as in prior research), our inferences are only affected to the extent the
measurement error in these proxies is correlated with time-series changes in the differences
between the MTRs and ETRs of firms. We have no reason to expect such a correlation, however,
our falsification test described in Section 6.4 mitigates such potential concerns.29
Table 8, Panel A presents the descriptive statistics for the variables used in the regression
described above, and Panel B presents the regression results. Consistent with our prediction, we
find that the coefficient for |MTR A.I. – GAAP ETR| × INV_OPP is negative and statistically
significant in all four regressions presented in the table. These coefficients suggest that as the gap
between a firm’s MTR and GAAP ETR widens, firms that use the GAAP ETR as the tax rate
29 A significant limitation of using Tobin’s Q as a proxy for investment opportunities is that the numerator in Q includes the market value of equity. To the extent investors anticipate and identify firms that use the GAAP ETR as their tax rate for decision making and incorporate this information into the firm’s stock price, our inferences could be confounded by an endogeneity bias. However, this concern does not affect the Sales Growth proxy.
34
input for decision making become less responsive to their growth opportunities. This result can
be interpreted as indicating that a firm using the GAAP ETR as its tax rate input is less efficient
in its investment decision-making when the GAAP ETR differs from its MTR. The table also
shows that the coefficients for INV_OPP and CFO are positive and statistically significant in all
our regressions, consistent with our expectations and prior research.
6.3. M&A consequences
Finally, we examine whether firms that use the GAAP ETR as the tax rate input for
M&A decisions are more prone to making value-decreasing acquisitions. Acquisitions are among
the largest and most readily observable forms of corporate investment and thus an important
corporate decision with a significant effect on firm value. Typically, acquisition decisions require
rigorous due diligence that includes evaluating the target firm’s value as a stand-alone business
and the value of potential synergies gained from combining businesses. Both processes require
firms to forecast after-tax cash flows. To the extent that firms use the GAAP ETR to measure the
tax impact of the acquisition decisions, they are likely to under- or over-estimate the value of the
acquisition. Underestimating the expected value created from an acquisition is likely to result in
the firm simply not engaging in the acquisition or being outbid by competing bidders. However,
we are unable to empirically measure such lost opportunities. On the other hand, conditional on
completing an acquisition, companies using the GAAP ETR as their tax rate input for decision
making are more likely to have either overbid for the target and/or more likely to have made
other errors in forecasting the value of the acquisition (e.g., the value of the target’s NOLs).
Thus, companies that use the GAAP ETR for decision making are more likely to make value
destroying acquisitions, especially when the GAAP ETR is different from the MTR.30
30 We recognize that in some cases firms engage in very large acquisitions that result in a combined entity that is fundamentally different than the acquirer (e.g., in the case of corporate inversions). For such acquisition, it is
35
Following a long list of prior studies, we measure whether acquisitions are value
enhancing or decreasing by examining the market response to their acquisition announcements
(see e.g., Harford, 1999; Fuller, Netter, and Stegemoller, 2002; Moeller, Schlingemann, and
Stulz, 2004; Masulis, Wang, and Xie, 2007; Goodman, Neamtiu, Shroff, and White, 2014). We
estimate the following OLS regression to test our prediction:
| . . | , ′ , (3)
where is the five–day [-2,2] cumulative abnormal return around the acquisition
announcement date. We measure abnormal returns as the firm’s return minus the return of the
CRSP value weighted index. |MTR A.I. – GAAP ETR| is the unsigned difference between an
acquirer’s simulated MTR after interest deductions and its GAAP ETR. X is a vector of control
variables that includes a number of other determinants of acquirer returns following Harford
(1999), Masulis et al. (2007), and Goodman et al. (2014). The regression also includes year and
industry fixed effects, and the standard errors are clustered at the firm-level. The coefficient of
interest, , captures the incremental acquisition announcement return for M&A deals in firm-
years with larger differences between the MTR and ETR, and is predicted to be negative.
Table 9, Panel A presents the descriptive statistics for the variables used in the above
regression and Table 9, Panel B presents the regression results. Consistent with our prediction,
we find that the coefficient for |MTR A.I. – GAAP ETR| is negative and significant using both
MTR proxies (i.e., Graham/Shevlin and BCG). These coefficients suggest that as the gap
between a firm’s MTR and GAAP ETR widens, firms that use the GAAP ETR as the tax rate
input for decision making complete acquisitions that are (relatively) value decreasing for
conceivable that the pre-acquisition MTR of the acquirer is not the theoretically appropriate tax rate to use because such acquisitions are not really “marginal” decisions. However, we note that the average (75th percentile) acquisition in our sample represents only 5.8% (7.4%) of the acquirers’ pre-acquisition total assets. Thus, the acquisitions in our sample of firm-years appear to be marginal decisions that should (in theory) be evaluated using the MTR.
36
shareholders.31 This result can be interpreted as indicating that a firm using the GAAP ETR as its
tax rate input is less efficient in its investment decision-making when the GAAP ETR differs
from its MTR. In terms of economic magnitude, our coefficients imply that a one percentage
point increase in the difference between the MTR and ETR reduces the acquisition
announcement return by 0.08 to 0.10 percentage points (depending on the MTR proxy). Since
the average (unsigned) stock return to an acquisition announcement is 3.90%, our results suggest
that using the GAAP ETR for M&A decisions reduces the average acquisition announcement
return by 2.03 to 2.64% per one percentage point increase in the difference between the MTR
and GAAP ETR. Since the market value of the average acquirer is $27 billion (untabulated), in
dollar terms, our coefficients imply that an acquisition announcement by firms using the GAAP
ETR for decision making lowers acquirer market values by $21–$28 million, on average.32
6.4. Falsification tests
Thus far, we focus on firms that use the GAAP ETR as their tax rate input and examine
whether these firms make inefficient corporate decisions when their ETRs differ from their
MTRs. Although this analysis offers a number of advantages, potential concerns about
measurement error and correlated omitted variables remain. We devise a falsification test using
firms that say they use the MTR for decision making. The intuition is that if firms use the MTR
as the tax rate input for their decisions, then the difference between the MTR and GAAP ETR
should be uncorrelated with their capital structure, investment, and acquisition outcomes. If the
difference between the MTR and GAAP ETR captures confounding factors that have a direct 31 Note that although we use acquisition announcement returns to proxy for acquisition quality, we do not assume that investors know which tax rates managers use for decision-making. Our analysis only assumes that investors recognize the expected value creation from an acquisition and can distinguish between value-increasing and value-decreasing deals without necessarily knowing the reason why managers engage in value-decreasing deals. 32 The dollar magnitude of the decrease in acquirer market value is computed by multiplying the coefficient estimate for |MTR A.I. – GAAP ETR| (i.e., -0.103% or -0.079%) with the average acquirer market value (i.e., $27 billion).
37
relation with corporate decisions, then we would (spuriously) find that this difference is related
to the corporate outcomes even for firms that use the MTR for decision making.
Table 10 presents the results of the falsification analyses. Specifically, Panel B (C, D)
presents the results for our analyses of the Debt Kink (Capital Expenditures, Acquisition
Announcement Returns) and Panel A presents the descriptive statistics for all the variables. The
regressions in Table 10 have the same structure that we used in our earlier analyses but now are
estimated using the firms that say they use the MTR for decision-making. Consistent with our
expectations, we find that the coefficient for MTR – GAAP ETR in Panel B, |MTR A.I. – GAAP
ETR| × INV_OPP in Panel C, and |MTR A.I. – GAAP ETR| in Panel D are all statistically
insignificant. These results indicate that the difference between a firm’s MTR and ETR does not
affect capital structure and investment policy for firms that use the MTR as their tax rate input
for decision-making (consistent with our prediction).33 These results help mitigate concerns that
the inferences drawn from Tables 6, 8 and 9 are confounded by spurious correlations.
6.5. Untabulated additional analyses and robustness tests
6.5.1. Measurement error in the MTR proxies
Simulated MTRs have measurement error when firms have non-debt tax shields such as
foreign income in low tax jurisdictions and employee stock options, among other things. If
measurement error in the MTR proxy is correlated with the corporate outcomes we examine,
then our results could be spurious. To at least partially address this concern, we partition our data
into two groups based on whether firms have foreign income. Since there is more measurement
error in the MTR proxies for firms with foreign income, we predict that our results from Tables 6
33 In untabulated analyses, we repeat these tests for firms that use the STR for decision-making and find that MTR – GAAP ETR is not associated with any of the corporate outcomes for firms that say they use the STR for decision making (which is consistent with our prediction).
38
to 9 will be weaker (stronger) for firms with (without) foreign income. Consistent with this
prediction, we find that all our results continue to hold in the sample of domestic firms (with no
foreign income) and the results become weaker in the sample of firms with foreign income
(untabulated). This result is consistent with our expectations and provides additional evidence
suggesting that our results are unlikely to be explained by measurement error in the MTR proxy.
Notwithstanding the above test, we note that the falsification test discussed in Section
6.4. also helps mitigate the concern that measurement error in our MTR proxies explain our
results. Specifically, if measurement error in MTR induces a relation between MTR – GAAP ETR
and the corporate outcomes we examine, then we should observe this relation for both firms that
use the GAAP ETR for decision-making as well as firms that use the MTR for decision-making.
However, as discussed above, we do not observe an association between MTR – GAAP ETR and
our proxies for corporate decision-making efficiency for firms that use the MTR for decision-
making.
6.5.2. Cross-sectional comparison of firms using different tax rates for decision-making
So far our analyses of the consequences of firms’ tax rate choice have focused on
companies that use the GAAP ETR for decision-making, and we used time-series variation in
difference between MTRs (the theoretically preferred rate) and GAAP ETRs (the rate chosen by
managers) to test our predictions. In this section, we examine whether our inferences are robust
to using an alternative research design that compares the outcomes of corporate decisions of
companies that use the GAAP ETR for decision-making with those of companies that use the
MTR or STR. Essentially, this analysis investigates whether the cross-sectional variation in
corporate decision-making efficiency is associated with cross-sectional variation in managers’
tax rate choice (controlling for prior determinants of decision-making efficiency). We also
39
exploit variation in the difference between a firm’s MTR and GAAP ETR to increase the power
of our tests. Specifically, we estimate regressions of the following structure:
1
2| |
3| |
The coefficient of interest in the above regression is , which captures the difference in
decision-making efficiency of firms that use the GAAP ETR for decision-making and whose
ETR differs from its MTR compared to firms that use the MTR or STR for decision-making. We
find that companies that use the GAAP ETR for decision-making have sub-optimal debt levels,
are less responsive to their investment opportunities, and make poorer acquisitions when the
difference between the MTR and ETR is large, relative to companies that use the MTR/STR for
decision-making (untabulated). These tests provide additional support for our main inference.
7. Conclusion
We survey corporate tax executives to examine the manner in which companies
incorporate taxes into their decision-making. We directly ask tax executives what tax rate their
companies use in decision-making and we merge their responses with Compustat data. We find,
surprisingly, that many firms use the GAAP effective tax rate (ETR) as their tax rate input for
decision making and few firms use the marginal tax rate (MTR). We provide some conjectures
and descriptive results in an attempt to explain the survey responses. We find that the difference
between the statutory tax rate (STR) and MTR is less than two percentage points for the majority
of the firms using the STR as the tax rate input for decision making. This result is consistent with
managers relying on simple heuristics that approximate theoretically correct constructs in their
decision making process. We also find that companies that are more focused on external
reporting (e.g., public firms and firms with high analyst following) are significantly more likely
to use the GAAP ETR as the tax rate in their decisions. In contrast, firms less focused on external
40
reporting (e.g., private firms) are significantly more likely to use STRs as the tax rate in their
decisions. These results suggest that managers have a salience bias with respect to the tax rate
used for decision-making. Finally, we find that larger firms (conditional on being public), firms
with high R&D intensity and high institutional ownership are more likely to use the MTR for
decision making, suggesting that firms with greater tax planning opportunities and more external
monitoring effectively incorporate taxes into their decision making.
We then go on to examine the magnitude of adverse economic consequences for firms
that use the GAAP ETR as their tax rate input for decision making. We find that firms using
GAAP ETRs as the tax rate input for their capital structure decisions adopt an aggressive
(conservative) debt policy when their GAAP ETR is greater (less) than their MTR. In terms of
investment outcomes, we find that the firms that use the GAAP ETR as the tax rate input for
investment decisions are less responsive to their investment opportunities and have lower
acquisition announcement returns when the difference between their firm’s MTR and GAAP
ETR is large. These results suggest that the tax rate firms use to incorporate taxes into their
business decisions has important economic consequences. Specifically, the use of GAAP ETRs
instead of the theoretically suggested MTR as the tax input for decision making leads to
inefficient corporate decisions. We estimate that the capital structure inefficiency alone costs the
firm $10 million (on average) in forgone value.
Our paper answers a call for research from Shackelford and Shevlin (2001) to find out
what rate managers actually use in decision-making settings. Our results run contrary to standard
economic theory and suggest that managers behave in an inefficient manner (i.e., use an average
tax rate for making incremental decisions), leading to a loss in firm value. As Camerer and
Malmendier (2007) explain, these inefficiencies likely persist because managers do not
specialize in taxation and because there isn’t a clear feedback mechanism that helps managers
learn from their past mistakes related to the use of improper tax rates for decision-making.
41
References Aghion, P., Van Reenen, J., and Zingales, L. 2013. Innovation and Institutional Ownership.
American Economic Review 103, 277-304. Ajinkya, B., Bhojraj, S., and Sengupta, P. 2005. The association between outside directors,
institutional investors, and the properties of management. Journal of Accounting Research 43, 343-376.
Ali A., Klasa, S., and Yeung, E., 2009, The Limitations of Industry Concentration Measures
Constructed with Compustat Data: Implications for Finance Research. Review of Financial Studies 22: 3839 – 3872.
Altman, E., 1968. Financial ratios, discriminant analysis and the prediction of corporate
bankruptcy. Journal of Finance 23,589–609. Armstrong, C., Blouin, J. and Larcker, D. 2012. The incentives for tax planning. Journal of
Accounting and Economics 53: 391-411. Asker, J., Farre-Mensa, J., Ljungqvist, A., 2014. Corporate Investment and Stock Market Listing:
A Puzzle? Review of Financial Studies, forthcoming. Atkinson, A. B. and Stiglitz, J.E. 1976. The Design of Tax Structure: Direct versus Indirect
Taxation. Journal of Public Economics, 6(1–2): 55–75. Badertscher, B., Shroff, N. and White, H. 2013. Externalities of public firm presence: Evidence
from private firms’ investment decisions. Journal of Financial Economics 109 (3): 682–706. Barberis, N. and Thaler, R.H. 2003. A survey of behavioral finance. Handbook of the Economics
of Finance (ed. G. Constantinides, M. Harris, and R. Stulz). Amsterdam: North-Holland. Bernard, A., Jensen, B. and Schott, P. 2006. Trade Costs, Firms and Productivity. Journal of
Monetary Economics 53, 917-937. Bekaert, G., Harvey, C., Lundblad, C., and Siegel, S. 2007. Global growth opportunities and
market integration. Journal of Finance 62: 1081–1137. Bloom, N., Bond, S., Van Reenen, J., 2007. Uncertainty and investment dynamics. Review of
Economic Studies 74, 391–415. Blouin, J., Core J., and Guay, W. 2010. Have the tax benefits of debt been overestimated.
Journal of Financial Economics 98, 195-213. Brav, A., Graham, J., Harvey, C., and Michaely, R. 2005. Payout policy in the 21st century.
Journal of Financial Economics 77: 483-527.
42
Brealey, R., Myers, S., and Allen, F. 2014, Principles of Corporate Finance (11th ed.), McGraw-Hill.
Bushee, B. J. 1998. The Influence of Institutional Investors on Myopic R&D Investment
Behavior. The Accounting Review 73 (3), 305–33. Camerer, C., and U. Malmendier, 2007. Behavioral Economics of Organizations. In: P. Diamond
and H. Vartiainen (eds.), Behavioral Economics and Its Applications, Princeton University Press.
Chen, X., Harford, J., and Li, K. 2007. Monitoring : which institutions matters? Journal of
Financial Economics 86:279-305. Chetty, R. 2009. The simple economics of salience and taxation. Unpublished working paper,
Harvard University. Chetty, R., Looney, A., and Kroft, K. 2009. Salience and taxation: Theory and evidence.
American Economic Review, Vol. 99, No. 4 (September), pages 1145-1177. Chung, R.M., Firth, M., and Kim, J.B., 2002. Institutional monitoring and opportunistic earnings
management. Journal of Corporate Finance 8, pp. 29-48.
Cornett, M., Marcus, A., Saunders, A., Tehranian, H., 2007. The impact of institutional ownership on corporate performance. Journal of Banking and Finance 31, 1771-1794.
de Bartolome, C. A. M. 1995. Which Tax Rate Do People Use: Average or Marginal? Journal of
Public Economics, 56(1): 79–96. Degeorge, F., Patel, J., Zeckhauser, R., 1999. Earnings manipulation to exceed thresholds.
Journal of Business 72, 1–33. DellaVigna, S., 2009. Psychology and Economics: Evidence from the Field. Journal of
Economic Literature, 47(2): 315–72. DellaVigna, S., and J. Pollet. 2013. Capital Budgeting versus Market Timing: An Evaluation
using Demographics. Journal of Finance, 68(1): 237–270.
Diamond, P. A., 1998. Optimal income taxation: An example with a u-shaped pattern of optimal marginal tax rates. American Economic Review 88, 83-95.
Fama, E. F., and French, K.R. 1998. Taxes, financing decisions, and firm value. Journal of Finance 53, 819-843.
Faulhaber, G. R. and Baumol, W.J. 1988. Economists as innovators: Practical products of
theoretical research. Journal of Economic Perspectives 26, 577-600.
43
Fazzari, S., Hubbard, R.G., and Petersen, B. 1988. Financing constraints and corporate investment. Brookings Papers on Economic Activity, 141-195.
Finkelstein, A. 2009. E-ZTAX: Tax Salience and Tax Rates. Quarterly Journal of Economics,
124(3), 969-1010. Fuller, K., Netter, J. and Stegemoller, M. 2002. What do returns to acquiring firms tell us?
Evidence from firms that make many acquisitions. Journal of Finance 57 (4): 1763–1794. Gabaix, Xavier, David Laibson, Guillermo Moloche, and Stephen Weinberg. 2006. Costly
Information Acquisition: Experimental Analysis of a Boundedly Rational Model. American Economic Review, 96(4): 1043–68.
Giroud, X., and Mueller, H.M. 2010, Does corporate governance matter in competitive
industries? Journal of Financial Economics 95, 312–331. Goodman, T., Neamtiu, M., Shroff, N., White, W., 2014. Management Forecast Quality and
Capital Investment Decisions. The Accounting Review 89, 331-365. Graham, J.R., 1996a, Debt and the marginal tax rate, Journal of Financial Economics 41, 41-73. Graham, J.R., 1996b, Proxies for the corporate marginal tax rate, Journal of Financial
Economics 42, 187-221. Graham, J.R., 2000. How big are the tax benefits of debt? Journal of Finance 55, 1901-1941. Graham, J.R., 2003. Taxes and corporate finance: A review. Review of Financial Studies 16,
1074–1128. Graham, J.R., and Harvey. C., 2001. The theory and practice of corporate finance: evidence from
the field. Journal of Financial Economics 60: 187-243. Graham, J.R., Harvey, C. and Rajgopal, S. 2005. The economic implications of corporate
financial reporting. Journal of Accounting and Economics 40: 3-73. Graham, J.R., Hanlon, M., and Shevlin. T. 2010. Barriers to mobility: the lockout effect of U.S.
taxation of worldwide corporate profits. National Tax Journal 63: 1111-1144. Graham, J.R., Hanlon, M., and Shevlin. T. 2011. Real effects of accounting rules: evidence from
multinational firms’ investment location and profit repatriation decisions. Journal of Accounting Research 49: 137-185.
Graham, J.R., Hanlon, M., Shevlin, T., and Shroff, N. 2014. Incentives for Tax Planning and
Avoidance: Evidence from the Field. The Accounting Review 89, 991-1023.
44
Graham, J., Lemmon, M., Schallheim, J., 1998. Debt, leases, taxes, and the endogeneity of corporate tax status. Journal of Finance 53, 131–162.
Graham, J.R., and Mills, L. 2008. Simulating marginal tax rates using tax return data, Journal of
Accounting and Economics 46, 366-388. Grubert, H. and Slemrod, J. 1998. The effect of taxes on investment shifting and income shifting
to Puerto Rico. The Review of Economics and Statistics 80: 365-373. Hadlock, J.H. and Pierce, J. R., 2010. New evidence on measuring financial constraints: Moving
beyond the KZ index. Review of Financial Studies, 23(5): 1909-1940. Hanlon, M. and Heitzman, S. 2010. A review of tax research. Journal of Accounting and
Economics 50: 127-178. Harford, J., 1999. Corporate cash reserves and acquisitions. Journal of Finance 52, 1969-1997. Hartzell, J., Starks, L., 2003. Institutional investors and executive compensation. Journal of
Finance 58, 2351–2374. He, J. and X. Tian, 2013. The Dark Side of Analyst Coverage: The Case of Innovation. Journal
of Financial Economics 109 (3): 856–878. Healy, P. M. 1985. The effect of bonus schemes on accounting decisions. Journal of Accounting
and Economics 7 (1-3): 85–107. Heider, F., and Ljungqvist, A. 2014, As certain as debt and taxes: Estimating the tax sensitivity
of leverage from state tax changes. Journal of Financial Economics, forthcoming. Hoberg, G., Phillips, G., 2013. Text-based network industries and endogenous product
differentiation. Unpublished working paper. Horngren, C. T., Datar, S.M., Foster, G., Rajan, M.D., Ittner, C. 2009. Cost Accounting: A
Managerial Emphasis, Prentice Hall, Upper Saddle River, NJ, 07458. Hubbard, R.G., 1998. Capital-market imperfections and investment. Journal of Economic
Literature, 193-225. Kahneman, D., Slovic, P., and Tversky, A. 1982. Judgment under Uncertainty: Heuristics and
Biases. New York: Cambridge University Press. Korteweg, A. 2010, The net benefits to leverage, Journal of Finance 65, 2137–2170. Leary, M. T. and Roberts, M.R. 2005, Do firms rebalance their capital structures? Journal of
Finance 60, 2575–2619.
45
Li, F., Lundholm, R., Minnis, M., 2013. A new measure of competition based on 10-K filings. Journal of Accounting Research 51, 399-436.
MacKie-Mason, J.K., 1990, Do taxes affect capital structure decisions?, Journal of Finance 45,
1471-1493. Masulis, R., Wang, C., Xie, F., 2007. Corporate governance and acquirer returns. Journal of
Finance 112, 1851-1889. McConnell, J., Servaes, H., 1990. Additional evidence on equity ownership and corporate value.
Journal of Financial Economics 27, 595–613. Mills, L., Erickson, M., Maydew, E., 1998. Investments in tax planning. Journal of American
Taxation Association 20, 1–20. Mirrlees, James A. 1971. An Exploration in the Theory of Optimum Income Taxation. The
Review of Economic Studies, 38(2): 175–208. Moeller, S., Schlingemann, F., and Stulz, R., 2004. Firm size and the gains from acquisitions.
Journal of Financial Economics 73, 201-228. Nickell, S. 1996. Competition and Corporate Performance. Journal of Political Economy 104,
724-746. Rego, S., 2003. Tax avoidance activities of U.S. multinational corporations. Contemporary
Accounting Research 20, 805–833. Scholes, M., Wolfson, M., Erickson, M., Hanlon, M., Maydew, E., Shevlin, T., 2014. Taxes and
Business Strategy: A Planning Approach. 5th edition, Prentice-Hall, Upper Saddle River, NJ. Shackelford, D., and Shevlin, T. 2001. Empirical research in accounting. Journal of Accounting
and Economics 31: 321-387. Shevlin, T. 1987, Taxes and off-balance sheet financing: Research and development limited
partnerships, The Accounting Review 62, 480–509. Shevlin, T. 1990, Estimating corporate marginal tax rates with asymmetric tax treatment of gains
and losses, Journal of the American Taxation Association 12, 51–67. Shleifer A. and Vishny, R.W. 1997. A survey of corporate governance. Journal of Finance 52:
737-784. Shroff, N., Sun, A., White, H., and Zhang, W. 2014. Management forecasts and seasoned equity
offerings. Unpublished working paper, MIT.
46
Shroff, N., Verdi, R., and Yu, G., 2014. Information environment and the investment decisions of multinational corporations. The Accounting Review 89, 759-790.
Simon, H. A., 1955. A behavioral model of rational choice. Quarterly Journal of Economics 69:
99-118. Simon, H. A., 1979. Information processing models of cognition. Annual Review of Psychology
30: 363-396. Slemrod, J. and Blumenthal, M. 1996. The income tax compliance cost of big business. Public
Finance Quarterly (October) 24: 411-438. Slemrod, J. and Venkatesh, V. 2002. The income tax compliance cost of large and midsize
businesses. A Report to the IRS LMSB Division. Strebulaev, I.A., Yang, B., 2013. The mystery of zero-leverage firms. Journal of Financial
Economics 109, 1–23. Trahan, E.A. and Gitman, L.J. 1995. Bridging the theory-practice gap in corporate finance: a
survey of chief financial officers. Quarterly Review of Economics and Finance 35 (Spring): 73-87.
Tversky, A. and Kahneman, D. 1974. Judgment under uncertainty: heuristics and biases. Science,
185:1124-1131. van Binsbergen, J. H., Graham, J.R. and Yang, J. 2010. The cost of debt, Journal of Finance, 65,
2089-2136. Whited, T. M. 2006. External finance constraints and the intertemporal pattern of intermittent
investment. Journal of Financial Economics 81, 467-502. Wurgler, J. 2000. Financial markets and the allocation of capital. Journal of Financial
Economics 58, 187-214.
47
APPENDIX Variable Definitions
This table provides a detailed description of the procedure used to compute each variable used in our analyses. Our data are obtained through (i) survey questions, (ii) Compustat, (iii) I/B/E/S, (iv) Thomson Reuters, (v) John Graham’s website (JG Website), (vi) the Hoberg-Phillips Data Library (H-P Website) (vii) SDC Platinum, and (viii) courtesy of Jie Yang. The cross-sectional variables are constructed using data from year 2006. All continuous variables are winsorized at 1% and 99% of the distribution and all dollar amounts are in millions. The variables are listed in alphabetical order.
Variable Definition
3-Yr Cash ETR
3-Yr Cash ETR is computed as the ratio of the sum of the Cash ETR numerator for the preceding three years and sum of Cash ETR denominator for the preceding three years. Where Cash ETR is the cash effective tax rate defined as the sum of total tax paid (data TXPD) divided by pretax income (data PI).
Acquisition Announcement Return
Five-day cumulative abnormal return for the acquirer around the announcement date of an acquisition. Abnormal returns are calculated using the Fama-French three-factor daily market model. The market-model parameters are estimated using daily returns data from the year prior to the acquisition. The announcement date is the event date disclosed in SDC.
Acquisition Expenditure
Acquisition expenditures (data AQC) scaled by lag total assets (data AT). Missing acquisition expense data are coded as zero.
Advertising Intensity The ratio of advertising expense (data XAD) scaled by total assets (data AT). Missing advertising expense data are coded as zero.
Analyst Following The number of analyst following a firm in I/B/E/S. We assume that analyst following is zero for public firms not covered by I/B/E/S.
Assets Worldwide assets (in millions) and corresponds with Compustat data item AT. Asset Growth Changes in total assets scaled by lag total assets (data AT).
BGY Fin. Constraints Indicator
Indicator variable for firm-years in the bottom two terciles of all of the following variables: long-term debt issuance (data DLTIS), long-term debt reduction (data DLTR), equity issuance (data SSTK), and equity reduction (data PRSTKC).
Cash Acquisition An indicator equal to one if the acquisition was paid for in cash.
Cash ETR Cash ETR is the cash effective tax rate defined as the sum of total tax paid (data TXPD) divided by pretax income (data PI).
Capital Expenditures Capital expenditures (data CAPX) scaled by lag total assets (data AT). CFO Cash flow from operations (data OANCF) scaled by lag total assets (data AT).
Credit Rating Dummy Indicator variable equal to one if the firm has an S&P credit rating (non-missing Compustat data item SPLTICRM).
Debt Kink Ratio of the amount of interest required to make the tax rate function slope downward to actual interest expense. The tax rate function is computed using the MTRs from Graham (1996a) or Blouin, Core, and Guay (2010). See Graham (2000) for more details.
Diversifying Acquisition
An indicator equal to one if the target’s one-digit SIC code differs from that of the acquirer’s.
Dividend Dummy Indicator variable equal to one if the firm pays a dividend in year in the fiscal year before the acquisition.
Equilibrium Factor
Equilibrium Factor is defined as the intersection of the marginal benefit of debt and marginal cost of debt curves. The marginal benefit curve is computed following Graham (2000) and the marginal cost curve is computed following van Binsbergen, Graham, and Yang (2010). If Equilibrium Factor = 1, the firm’s equilibrium leverage is 100% of its actual (i.e., firm is at equilibrium). If Equilibrium Factor = 1.2, the firm should have 120% of its actual leverage (i.e., firm is underlevered). If Equilibrium Factor = 0.8, the firm should have 80% of its actual leverage (firm is overlevered). We thank Jie Yang for sharing this variable with us.
Firm Size The book value of worldwide assets (data AT). Foreign Acquisition Indicator variable that takes a value of one if the target is a foreign company.
48
Foreign Assets Proportion of assets owned in foreign locations. Foreign Income Foreign pre-tax income (data PIFO) divided by total assets (data AT).
GAAP ETR GAAP ETR is the GAAP effective tax rate defined as total tax expense (data TXT) divided by pretax accounting income (data PI). The GAAP ETR is set to missing of PI is less than or equal to zero.
GAAP ETR Importance
This variable is computed using the survey responses to the question “Which metric is more important to the top management at your company?” The respondents could pick one of the following answers: (i) GAAP ETR, (ii) Cash Taxes Paid, or (iii) Both are equally important. GAAP ETR Importance is an indicator variable that takes on the value of one for firms indicating that the GAAP ETR is the most important metric to top management or both cash taxes and the GAAP ETR are equally important to top management.
HP Fin. Constraints Indicator
Indicator variable for firms in the top tercile of the yearly size-age index developed by Hadlock and Pierce (2010). The size-age index is (−0.737 × Size) + (0.043 × Size2) − (0.040 × Age), where size equals the log of inflation-adjusted book assets, and age is the number of years the firm is listed with a non-missing stock price on Compustat. In calculating this index, size is winsorized at $4.5 billion, and age is winsorized at 37 years.
Intangible Intensity The ratio of intangible assets (data INTAN) scaled by total assets (data AT). Missing data are coded as zero.
Institutional Ownership (%)
The percentage of the firm’s equity held by institutional investors in year t. Calculated from data provided in the Thomson-Reuter’s Institutional Holdings (13F) Database. Set equal to zero if the data are missing.
Leverage Ratio of long-term debt (data DLTT) plus the debt included in current liabilities (data DLC) to total assets (data AT).
MB Market-to-book ratio (MVE/data CEQ).
MTR MTR is the simulated marginal tax rate (before financing) obtained from John Graham’s website (https://faculty.fuqua.duke.edu/~jgraham/taxform.html) or from WRDS for the Blouin, Core, and Guay (2010) procedure.
MTR After Interest (MTR A.I.)
MTR A.I. is the simulated marginal tax rate (after financing) obtained from John Graham’s website (https://faculty.fuqua.duke.edu/~jgraham/taxform.html) or from WRDS for the Blouin, Core, and Guay (2010) procedure.
MVE MVE is the market value of equity (data PRCC_F multiplied by data CSHO).
NOL An indicator variable that equals one if the firm has a positive tax loss carry-forward (TLCF) on Compustat.
Non-Distress Indicator
Indicator variable for firm-years in the top tercile of the z-score distribution (following van Binsbergen, Graham, and Yang (2010)). Z-score is a bankruptcy score that is computed as follows: 1.2 × working capital + 1.4 × retained earnings + 3.3 × earnings before interest and taxes + 1.0 × sales.
No. of Bidders Indicator variable that equals one if there is more than one bidder for the target firm.
Organization Tax Complexity
Sum of (i) the number entities in the consolidated IRS Form 1120, (ii) the number of IRS Form 1120s filed, (iii) the number of IRS Form 5471s filed, and (iv) the number of flow through entities in which the firm has ownership interest.
PPE Book value of Net Property, Plant, and Equipment (data PPENT) scaled by lag total assets (data AT).
Pre-Acquisition Return Run Up
Abnormal stock returns for the period (-250, -12) prior to the acquisition announcement date, as in Harford (1999). Abnormal returns are calculated using the Fama-French three-factor daily market model. Market-model parameters are estimated over the period (-370,-253).
Public An indicator variable that takes on the value of one for publicly traded firms. Public Target An indicator equal to one if the target is identified as public company in SDC Platinum.
R&D Intensity Ratio of research and development expense (data XRD) scaled by total assets (data AT). Missing R&D data are coded as zero.
49
Relative Value of Target
The value of the acquisition scaled by the acquirer’s market value of equity.
ROA ROA is return-on-assets defined as net income (data NI) divided by total assets (data AT). Sales Worldwide net sales and corresponds to Compustat data item SALE. Sales Growth Changes in sales scaled by lag sales (data SALE).
Text-Based HHI
This is an industry concentration measure computed using the Herfindahl-Hirschmann sum of squared market shares formulation (HHI). These HHI data are computed by grouping firms based on their product similarity in their 10-K filings. Specifically, the measure of product similarity is obtained based on text-based analysis of firm 10-K product descriptions. See Hoberg and Phillips (2013) for a detailed description of the procedure used to compute this measure of industry concentration.
Tobin’s Q The market value of worldwide assets, which is computed as the market value of equity plus book value of debt scaled by the book value of assets [(data PRCC_F × data CSHO + data DLTT + data DLC) / (data AT)].
Total Deadweight Loss
Total Deadweight Loss is the loss in firm value from being either underlevered or overlevered (van Binsbergen, Graham, and Yang, 2010). It is measured as the area between the cost and benefit curves when a firm has more/less debt than recommended by our model. Total Deadweight Loss is reported as a percentage of book value in perpetuity; for example, a loss of 5% would occur if the annual loss was 0.5% and the discount rate was 0.10. We use the Moody’s average corporate bond yield as the discount rate for all firms in a given year. Figure 2 provides a graphical description of this variable. We thank Jie Yang for sharing this variable with us.
US NOL US NOL is an indicator variable that equals one if the firm has a US net operating loss carryforward (from survey responses).
Z-Score Z Score is the bankruptcy score from Altman (1968) as modified by MacKie-Mason (1990). In terms of Compustat data items it equals: 1.2 × [ACT - LCT]/AT + 1.4 × RE/AT + 3.3 × EBIT/AT + 1.0 × SALE/AT.
Survey R
Notes: This incorporate tto choose frorate, (ii) GAA(v) marginal indicate the p(iii) investmweighted aveprovides the separately for
Responses Des
figure presentsaxes into each
om the followinAP effective tatax rate, and (
primary tax ratent decisions erage cost of percentage of r public and pr
cribing the Tax
s the response of the following options (or,ax rate, (iii) jur(vi) other. Thete used in that (property, equcapital, (vi) wfirms using ea
rivate firms. Th
x Rates used byDecisions Ex
s to the surveying forecasts owrite in an an
risdiction-spece question thendecision: (i) muip., etc), (iv)where to locatach tax rate avhe correspondin
50
Figure 1 y Firms for Coxamined in the
y question “W
or decision maknswer if the optcific statutory tn listed a numb
mergers and acqdecision to p
te new facilitiveraged across ng data in table
orporate Decise Survey
What is the primking processestions given aretax rate, (iv) ju
mber of decisionquisitions, (ii) purchase versuies, and (vii) all seven decie form are in T
sion Making Av
mary tax rate s?” The surveye not sufficienturisdiction-spens and settingscapital structurus lease (propcompensation ision contexts. Table 3, Panel A
veraged Across
your companyy respondent ist): (i) U.S. statuecific effective s for the respore (debt versus
perty, equip., edecisions. ThThe results ar
A.
s all
y uses to s allowed utory tax tax rate,
ondent to s equity), etc.), (v) he figure re shown
Notes: This equilibrium underleveredx*, in the casdepicted by tin which the
Measu
figure shows tlevel of debt,
d is depicted bye in which the the shaded areaactual level of
ure of Total De
the marginal bx*, that occu
y the shaded aractual level of
a between the Mf debt is greater
eadweight Cos
benefit of debturs where the rea between thf debt is less thMC and MB cur than the equil
51
Figure 2 st from Being O
t curve, MB(x)marginal coste MC and MBhan the equiliburves from the librium level o
Overlevered an
), the marginalt and benefit curves from th
brium level of equilibrium, x
of debt.
nd Underlevere
l cost of debt curves intersehe actual debt,debt. The cost *, to the observ
ed
curve, MC(x),ect. The cost , x0, to the equof been overle
ved debt, x1, in
, and the of being
uilibrium, evered is
n the case
52
Table 1 Descriptive Statistics
Notes: The above data are obtained through (i) survey questions, (ii) Compustat, (iii) I/B/E/S, (iv) Thomson Reuters, (v) John Graham’s website (JG Website: https://faculty.fuqua.duke.edu/~jgraham/taxform.html) or (iv) the Hoberg-Phillips Data Library (H-P Website: http://alex2.umd.edu/industrydata/industryconcen.htm). All continuous variables are winsorized at 1% and 99% of the distribution and all dollar amounts are in millions. All variables are defined in the Appendix.
Variable Source N Mean SD P25 P50 P75
Public Survey 500 0.778 0.416 1.000 1.000 1.000
Assets (in millions) Survey 493 7,804 19,639 473 1,313 4,895
MVE (in millions) Compustat 354 8,518 20,311 703 1,936 6,164
Sales (in millions) Compustat 356 5,772 12,130 575 1,474 5,260
Tobin's Q Compustat 354 1.970 1.031 1.271 1.594 2.359
ROA Compustat 356 0.058 0.072 0.024 0.059 0.096
Foreign Assets (% of assets) Survey 488 0.187 0.216 0.000 0.100 0.305
Organization Tax Complexity Survey 488 104.553 388.393 17.000 36.000 78.500
Leverage Compustat 356 0.212 0.196 0.048 0.174 0.320
Sales Growth Compustat 353 0.152 0.239 0.039 0.098 0.193
Asset Growth Compustat 354 0.143 0.275 0.003 0.071 0.189
US NOL Survey 478 0.462 0.499 0.000 0.000 1.000
R&D Intensity Compustat 356 0.024 0.044 0.000 0.000 0.032
Intangible Intensity Compustat 351 0.188 0.182 0.030 0.137 0.307
Analyst Following I/B/E/S 356 9.798 8.253 3.000 8.000 15.000
Institutional Ownership (%) Thomson Reuters 357 0.536 0.402 0.000 0.688 0.869
Text-Based HHI H-P Website 340 0.213 0.232 0.059 0.121 0.262
GAAP ETR (survey response) Survey 398 0.309 0.165 0.273 0.340 0.377
GAAP ETR Compustat 311 0.313 0.154 0.278 0.336 0.375
MTR JG Website 203 0.309 0.099 0.350 0.350 0.350
MTR After Interest (MTR A.I.) JG Website 234 0.214 0.156 0.029 0.347 0.350
MTR WRDS-Blouin/Core/Guay 334 0.331 0.044 0.337 0.345 0.350
MTR After Interest (MTR A.I.) WRDS-Blouin/Core/Guay 334 0.316 0.060 0.317 0.341 0.348
|MTR - GAAP ETR| JG Website & Compustat 187 0.132 0.289 0.024 0.050 0.118
|MTR A.I. - GAAP ETR| JG Website & Compustat 210 0.167 0.164 0.029 0.096 0.306
|MTR - GAAP ETR| Blouin/Core/Guay & Comp 291 0.095 0.154 0.023 0.044 0.099
|MTR A.I. - GAAP ETR| Blouin/Core/Guay & Comp 291 0.101 0.156 0.026 0.048 0.107
3-Yr Cash ETR Compustat 297 0.287 0.236 0.171 0.254 0.356
53
Table 2 Non-Response Bias Test
Notes: All dollar amounts are in millions. All Compustat variables are measured in the fiscal year ending in 2006 and are winsorized at 1% and 99% of the distribution. Column (1) consists of all the firms on Compustat except for firms with a negative book value, firms whose names indicate they are limited partnerships, and firms incorporated outside the United States. Column (2) includes all the firms that were sent a survey (as described in Section 2 in the paper) that we could match to and retrieve the data from Compustat. Column (3) includes the survey responders with data available on Compustat. Column (4) consists of the group of firms that are on Compustat and that we sent a survey to but did not receive a response. All variables are defined in the Appendix.
N Mean N Mean N Mean N Mean 1 vs. 2 1 vs. 3 2 vs. 3 3 vs. 4 Assets 5,938 3,580.2 1356 9,075.2 414 10,173.8 942 8,592.3 0.00 0.00 0.45 0.28MVE 5,443 2,369.1 1288 7,206.3 396 8,283.1 892 6,728.2 0.00 0.00 0.33 0.15Sales 5,911 1,719.4 1342 5,049.7 400 5,717.8 942 4,766.1 0.00 0.00 0.29 0.13Leverage 5,921 0.19 1341 0.22 400 0.21 941 0.22 0.00 0.21 0.25 0.12Cash 5,937 0.20 1337 0.14 396 0.15 941 0.13 0.00 0.00 0.30 0.16MB 5,443 4.35 1279 3.30 387 3.32 892 3.30 0.00 0.00 0.95 0.93ROA 5,911 -0.04 1338 0.05 396 0.07 942 0.04 0.00 0.00 0.00 0.00Asset Growth 5,692 0.36 1326 0.15 396 0.14 930 0.15 0.00 0.00 0.55 0.42Sales Growth 5,499 0.25 1323 0.14 395 0.14 928 0.14 0.00 0.00 0.75 0.66Capital Exp. 5,877 0.04 1334 0.05 396 0.05 938 0.05 0.00 0.01 0.27 0.12Acquisition Exp. 5,938 0.02 1338 0.03 396 0.03 942 0.03 0.00 0.02 0.46 0.32R&D Exp. 5,938 0.05 1338 0.02 396 0.02 942 0.02 0.00 0.00 0.80 0.74NOL 5,938 0.37 1338 0.42 396 0.39 942 0.43 0.00 0.28 0.33 0.18GAAP ETR 4,239 0.26 1172 0.29 349 0.31 823 0.29 0.00 0.00 0.18 0.093-Yr Cash ETR 4,052 0.27 1165 0.27 354 0.27 811 0.28 0.72 0.67 0.55 0.42MTR 2,097 0.27 756 0.31 227 0.31 529 0.31 0.00 0.00 0.91 0.88
All Compustat (1)
All Public Firms we Contacted with Available Data
(2)
Survey Responders (Public Firms) with
Available Data (3)
Survey Nonresponders (Public Firms) with
Available Data (4)
p-Value
54
Table 3 Survey Responses Describing the Tax Rates used by Firms for Corporate Decision Making
Panel A: What is the primary tax rate your company uses to incorporate taxes into each of the following forecasts or decision making processes?
U.S. Statutory
Tax Rate (STR) GAAP ETR
Jurisdiction Specific STR
Jurisdiction Specific ETR
Marginal Tax Rate
Other N
Merger &Acquisition Decisions 21.1% 24.9% 20.3% 20.1% 10.1% 3.4% 497
Capital Structure 25.9% 29.7% 14.5% 15.3% 12.0% 2.7% 491
Investment Decisions 22.9% 24.5% 21.1% 16.0% 12.5% 3.1% 489
Decision to Purchase vs. Lease 23.9% 23.7% 20.3% 16.4% 12.3% 3.5% 489
Weighted Average Cost of Capital 25.4% 34.3% 13.4% 12.6% 11.8% 2.5% 484
Where to Locate New Facilities 17.0% 16.6% 28.8% 25.9% 8.8% 2.9% 487
Compensation 25.7% 27.2% 19.1% 12.9% 10.6% 4.6% 482
Panel B: Pearson Correlation
1 2 3 4 5 6 7
1 Merger &Acquisition Decisions 1
2 Capital Structure 0.78 1
3 Investment Decisions 0.80 0.80 1
4 Decision to Purchase vs. Lease 0.76 0.80 0.93 1
5 Weighted Average Cost of Capital 0.72 0.78 0.81 0.82 1
6 Where to Locate New Facilities 0.76 0.74 0.82 0.78 0.71 1
7 Compensation 0.66 0.71 0.77 0.77 0.76 0.70 1
Notes: Panel A in this table summarizes the responses to our survey question and Panel B presents the correlations among the survey responses in Panel A. All the correlation coefficients in Panel B are statistically significant at the 1% level.
55
Table 4 Univariate Analyses – Percentage of Firms that Use a Given Tax Rate as an Input into a Given Policy Decision
Panel A: Conditioning based on Survey Data
Panel B: Conditioning based on Compustat Data
Notes: This table presents the responses to our survey questions after partitioning firms based on their characteristics. Panel A presents data from our entire sample of firms (both private and public) and partitions firms using variables constructed from our survey instrument. Panel B presents data from our sample of public firms and partitions firms using variables constructed from our survey instrument and Compustat. Highlighted fields indicate a statistically significant difference in the percentages reported in the columns at the 10% level or better. All partitioning variables are defined in the Appendix. As an example of how to interpret this table, consider the Ownership partition in Panel A. The numbers in that column indicate that 47.4% (37%) of the public (private) firms use the GAAP ETR for capital structure decisions. The remaining numbers are to be interpreted similarly.
Public PrivateBelow
MedianAbove Median
Yes NoGAAP ETR
Cash Taxes Paid
Below Median
Above Median
GAAP ETR - Capital Structure 45.0% 47.4% 37.0% 49.0% 41.6% 44.2% 44.9% 47.3% 37.2% 45.5% 45.4%
GAAP ETR - Capex 40.5% 42.1% 34.6% 43.7% 37.7% 42.3% 38.6% 41.1% 37.6% 42.4% 38.8%
GAAP ETR - M&A 45.1% 47.9% 35.2% 44.1% 46.1% 44.1% 45.9% 46.7% 38.6% 47.3% 43.2%
MTR - Capital Structure 12.0% 11.8% 13.0% 11.7% 11.8% 11.2% 13.0% 11.1% 15.0% 15.0% 8.4%
MTR - Capex 12.5% 12.6% 12.1% 11.3% 13.1% 11.3% 13.8% 12.1% 13.8% 16.1% 8.0%
MTR - M&A 10.1% 8.8% 14.8% 10.6% 9.0% 10.5% 9.8% 9.1% 13.2% 14.0% 5.7%
STR - Capital Structure 40.3% 38.2% 47.2% 37.7% 43.3% 40.9% 40.2% 39.2% 45.1% 37.2% 43.6%
STR - Capex 44.0% 42.4% 49.5% 43.3% 45.1% 42.7% 44.9% 44.1% 45.0% 38.8% 49.6%
STR - M&A 41.4% 39.7% 47.2% 42.9% 40.8% 41.8% 41.2% 40.8% 45.6% 36.0% 47.6%
Metric Important to Top Management
Proportion of Assets in Foreign Locations
Ownership Firm Size US NOLs% of All
Firms
Below Median
Above Median
GAAP ETR
Cash Taxes
Below Median
Above Median
Below Median
Above Median
Below Median
Above Median
Below Median
Above Median
GAAP ETR - Capital Structure 47.4% 54.0% 40.9% 48.6% 38.2% 48.1% 47.3% 51.8% 42.9% 45.4% 50.3% 47.5% 55.1%
GAAP ETR - Capex 42.1% 47.4% 37.0% 42.5% 37.0% 45.0% 39.6% 44.9% 37.3% 39.1% 44.0% 38.2% 49.0%
GAAP ETR - M&A 47.9% 50.3% 45.6% 48.6% 40.4% 50.3% 46.0% 54.6% 40.9% 47.8% 49.2% 48.5% 56.9%
MTR - Capital Structure 11.8% 11.1% 12.4% 11.5% 14.5% 16.0% 7.4% 10.3% 13.6% 13.2% 10.3% 10.9% 9.2%
MTR - Capex 12.6% 10.5% 14.6% 12.4% 14.8% 17.5% 7.5% 11.2% 13.1% 14.4% 9.7% 13.7% 10.2%
MTR - M&A 8.8% 8.2% 9.3% 8.9% 8.8% 14.1% 3.7% 8.2% 9.4% 11.8% 5.6% 10.9% 5.9%
STR - Capital Structure 38.2% 32.3% 44.0% 37.1% 45.5% 33.7% 42.6% 35.9% 40.9% 38.5% 37.7% 37.6% 35.7%
STR - Capex 42.4% 39.5% 45.3% 41.9% 46.3% 35.4% 49.2% 41.8% 45.8% 43.1% 44.0% 44.1% 39.8%
STR - M&A 39.7% 37.9% 41.5% 38.8% 47.4% 33.0% 46.6% 34.7% 44.7% 36.5% 41.8% 37.6% 35.3%
R&D Intensity Leverage |MTR-GAAP ETR|% of Public Firms
Metric Important to Top Mgmt.
Prop. of Assets in Fgn Locations
Firm Size
56
Table 5 Multivariate Analyses of the Determinants of Firms’ Responses to Survey Questions
Panel A: Determinants of Public and Private Firms Survey Responses using Survey Data as Conditioning Variables
Capital Str. Capex M&A Capital Str. Capex M&A Capital Str. Capex M&A
Public 0.381*** 0.433*** 0.525*** 0.202 0.201 -0.333 -0.474*** -0.449*** -0.356**
(2.74) (3.20) (4.11) (1.03) (0.98) (-1.53) (-3.16) (-2.77) (-2.55)
Log(Assets) -0.163*** -0.186*** -0.060 -0.024 0.079 -0.039 0.167*** 0.105** 0.045
(-2.63) (-2.96) (-0.81) (-0.38) (1.09) (-0.39) (3.21) (2.09) (1.00)
US NOL -0.032 0.148 -0.025 -0.109 -0.113 0.190 0.006 -0.149 -0.089
(-0.12) (0.75) (-0.13) (-0.40) (-0.47) (0.68) (0.02) (-0.75) (-0.43)
GAAP ETR Importance 0.379* 0.163* 0.193 -0.493* -0.309 -0.234 -0.111 0.036 -0.080
(1.90) (1.72) (0.98) (-1.72) (-1.09) (-0.83) (-0.50) (0.14) (-0.35)
Log(Org. Tax Complexity) 0.053 0.094 0.062 0.166 0.115 0.204 -0.155 -0.167 -0.160*
(0.50) (0.75) (0.46) (1.07) (0.78) (1.37) (-1.64) (-1.62) (-1.65)
-0.198 -0.786 -0.349 -1.384 -1.766** -2.943*** 0.792 1.380** 1.174**
(-0.42) (-1.30) (-0.70) (-1.44) (-2.10) (-2.69) (1.64) (2.46) (2.42)
N 441 442 446 441 442 446 441 442 446
Pseudo R-Squared 1.9% 2.1% 1.1% 1.4% 2.1% 3.9% 1.7% 1.7% 1.2%
S.E. Clustered by Industry Yes Yes Yes Yes Yes Yes Yes Yes Yes
Foreign Assets
Effective Tax Rate (ETR) Marginal Tax Rate (MTR) Statutory Tax Rate (STR)Dependent Variable:
57
Table 5 (continued) Multivariate Analyses of the Determinants of Firms’ Responses to Survey Questions
Panel B: Determinants of Public Firms’ Survey Responses using both Survey and Compustat Data as Conditioning Variables
Capital Str. Capex M&A Capital Str. Capex M&A Capital Str. Capex M&A
Log(Assets) -0.441*** -0.450*** -0.323** 0.215 0.353* 0.502** 0.313*** 0.222* 0.091(-3.25) (-3.28) (-2.52) (1.12) (1.92) (2.12) (2.67) (1.83) (0.72)
US NOL -0.346 -0.012 -0.155 -0.002 -0.100 0.143 0.263 0.143 0.074(-1.14) (-0.04) (-0.55) (-0.00) (-0.26) (0.32) (0.92) (0.58) (0.28)
GAAP ETR Importance 0.557 0.295* 0.386 -0.432 -0.440 -0.111 -0.347 -0.181 -0.307(1.49) (1.83) (1.12) (-0.77) (-0.77) (-0.14) (-0.88) (-0.53) (-0.85)
Log(Org. Tax Complexity) 0.131 0.221 0.257 0.194 0.074 -0.041 -0.209 -0.233 -0.195(0.78) (1.25) (1.46) (0.67) (0.26) (-0.12) (-1.45) (-1.52) (-1.19)
Foreign Assets -0.050 -1.157 -0.434 -2.689** -2.948** -4.782*** 1.020 2.038*** 1.451** (-0.07) (-1.49) (-0.54) (-2.09) (-2.32) (-2.63) (1.60) (3.04) (2.09)
R&D Intensity -8.920*** -8.591*** -7.504* 9.338** 8.028** 9.411** 3.998 1.858 4.142(-2.82) (-3.08) (-1.72) (2.38) (2.13) (2.09) (1.54) (0.72) (1.29)
Leverage 1.214 0.770 0.299 -0.453 -1.150 -2.002 -0.747 -0.421 0.274(1.35) (0.91) (0.39) (-0.50) (-1.12) (-1.60) (-0.93) (-0.59) (0.38)
ROA -1.636 -1.937 -2.640 2.160 4.178 4.895 1.274 1.720 2.060(-0.70) (-0.90) (-1.29) (0.69) (1.37) (1.16) (0.55) (0.87) (1.05)
Intangible Intensity 0.118 0.709 0.226 -0.024 0.331 0.719 0.102 -0.507 -0.401(0.14) (0.84) (0.28) (-0.02) (0.32) (0.63) (0.13) (-0.69) (-0.55)
Analyst Following 0.061*** 0.056** 0.051** -0.054* -0.049 -0.088** -0.037* -0.042** -0.029(2.76) (2.41) (2.37) (-1.79) (-1.59) (-2.19) (-1.90) (-2.01) (-1.40)
Institutional Ownership -0.357 -0.185 -0.669** 1.102* 0.936 1.898** 0.061 -0.014 0.453(-0.97) (-0.52) (-2.02) (1.77) (1.43) (2.28) (0.18) (-0.04) (1.48)
Sales Growth -0.013 0.658 0.612 -1.549 -0.983 -2.552** 0.325 -0.448 -0.195(-0.02) (1.27) (1.23) (-1.51) (-1.05) (-2.48) (0.59) (-0.86) (-0.40)
Text-Based HHI -0.411 -0.656 -0.312 0.008 0.310 0.405 0.425 0.434 0.169(-0.69) (-1.08) (-0.52) (0.01) (0.34) (0.41) (0.78) (0.84) (0.32)
|MTR - GAAP ETR| 1.634* 1.611 1.965* -2.626 -2.232 -4.252 -0.586 -0.650 -0.880(1.67) (1.63) (1.86) (-1.29) (-1.18) (-1.48) (-1.10) (-1.31) (-1.16) -0.073 -0.060 -0.224 0.529 0.233 0.285 -0.136 -0.068 0.098(-0.22) (-0.18) (-0.74) (1.15) (0.49) (0.53) (-0.46) (-0.24) (0.35)
N 297 299 302 297 299 302 297 299 302Pseudo R-Squared 7.9% 8.4% 6.5% 9.5% 9.0% 16.9% 3.9% 5.0% 4.1%S.E. Clustered by Industry Yes Yes Yes Yes Yes Yes Yes Yes Yes
|MTR - GAAP ETR| - Missing Indicator
Dependent Variable:Effective Tax Rate (ETR) Marginal Tax Rate (MTR) Statutory Tax Rate (STR)
58
Table 5 (continued) Multivariate Analyses of the Determinants of Firms’ Responses to Survey Questions
Notes: This table presents the results from logistic regressions where the dependent variable is an indicator variable constructed based responses to our survey questions and the independent variables are firm characteristics. Panel A presents result from our entire sample of (public and private) firms and thus the independent variables include only those constructed from our survey instrument. Panel B presents data from our sample of public firms and the independent variables include those constructed from our survey instrument as well as those constructed from Compustat, I/B/E/S, and Thomson Reuters. All the independent variables are defined in the Appendix. The standard errors in the above regressions are clustered at the industry-level. ***, **, * signifies statistical significance at the two-tailed 1%, 5%, and 10% levels.
59
Table 6 Analyses of Capital Structure Decisions
Panel A: Descriptive Statistics of the Variables Used in the Regression
Variable N Mean SD P25 P50 P75
Debt Kink (JG) 450 3.856 3.257 1.200 3.000 6.000
Debt Kink (BCG) 450 2.071 2.343 0.400 1.200 3.000
MTR - GAAP ETR (JG) 450 -0.021 0.106 -0.051 -0.023 0.019
MTR - GAAP ETR (BCG) 450 -0.011 0.084 -0.056 -0.030 0.010
Log(Assets) 450 7.115 1.364 6.088 6.983 7.930
Tobin's Q 450 1.767 1.317 0.912 1.356 2.290
Sales Growth 450 0.136 0.156 0.037 0.108 0.194
CFO 450 0.029 0.016 0.017 0.028 0.039
PPE 450 0.290 0.190 0.143 0.241 0.392
Capital Expenditures 450 0.015 0.010 0.008 0.013 0.021
R&D Expenditures 450 0.003 0.007 0.000 0.000 0.002
Acquisition Expenditures 450 0.009 0.016 0.000 0.001 0.011
Leverage 450 0.226 0.144 0.120 0.229 0.338
Credit Rating Dummy 450 0.456 0.499 0.000 0.000 1.000
Z-Score 450 2.040 0.933 1.367 1.943 2.752
Dividend Dummy 450 0.556 0.497 0.000 1.000 1.000
Foreign Income 450 0.013 0.026 0.000 0.000 0.014
NOL Dummy 450 0.287 0.453 0.000 0.000 1.000
60
Table 6 (continued) Analyses of Capital Structure Decisions
Panel B: Regression Results
Notes: Panel (A) B in this table presents the results from (descriptive statistics of the variables used in) a regression of the Debt Kink, which measures the conservativeness of a firm’s debt policy, on the difference firm’s MTR and its GAAP ETR, and control variables. The sample is comprised of firms that say they use their GAAP ETR as the tax rate input for capital structure decisions. The regressions include both year and industry fixed effects and the standard errors are clustered at the firm-level. ***, **, * signifies statistical significance at the two-tailed (one-tailed) 1%, 5%, and 10% levels (for the coefficients of interest in the highlighted row). All the variables are defined in the Appendix.
Dependent Variable:
Method of Simulating MTRs:
Pr. Sign Coefficient t -Statistic Coefficient t -Statistic
MTR - GAAP ETR + 1.805 ** 1.69 2.818 ** 2.20
Log(Assets) 0.855 *** 4.02 0.624 *** 3.36
Tobin's Q 0.289 * 1.82 0.154 1.11
Sales Growth 0.784 1.01 0.426 0.63
CFO 34.345 *** 4.25 12.091 * 1.70
PPE 1.042 0.47 -0.725 -0.37
Capital Expenditures 42.150 ** 2.21 38.279 ** 2.29
R&D Expenditures -85.796 ** -2.01 -4.872 -0.13
Acquisition Expenditures 18.510 *** 2.82 2.713 0.47
Leverage -5.488 *** -4.07 -4.374 *** -3.72
Credit Rating Dummy -0.127 -0.26 -0.358 -0.84
Z-Score 0.029 0.10 0.442 * 1.77
Dividend Dummy -0.960 ** -2.09 0.103 0.26
Foreign Income 0.687 0.09 16.765 ** 2.38
NOL Dummy -0.392 -1.31 -0.143 -0.55
N
Adjusted R-Squared
S.E. Clustered by Firm
Year & Industry Fixed Effects
Debt Kink
Graham/Shevlin Blouin, Core, and Guay (2010)
Yes Yes
450 450
70.0% 55.5%
Yes Yes
61
Table 7 Economic Magnitude of Capital Structure Effects
Panel A: Descriptive Statistics of the Variables Used in the Analyses
Panel B: Univariate Analyses
Panel C: Regression Results with Equilibrium Leverage as Dependent Variable
Variable N Mean SD P25 P50 P75
Equilibrium Factor 440 1.462 1.344 0.736 1.102 1.704
Total Deadweight Loss 440 0.011 0.015 0.001 0.004 0.013
MTR - GAAP ETR 440 -0.018 0.098 -0.050 -0.025 0.015
|MTR - GAAP ETR| 440 0.077 0.093 0.023 0.040 0.078
HP Fin. Constraints Indicator 440 0.434 0.496 0.000 0.000 1.000
BGY Fin. Constraints Indicator 440 0.216 0.412 0.000 0.000 0.000
Non-Distress Indicator 440 0.584 0.493 0.000 1.000 1.000
MTR - GAAP ETR
Mean Mean Median
1 -0.11 1.10 0.90
2 -0.03 1.52 1.16
3 0.02 1.49 1.21
4 0.16 2.10 1.58
MTR - GAAP ETR Quartile Ranks
Equilibrium Factor
Dependent Variable:
Pr. Sign Coefficient t -Statistic Coefficient t -Statistic
MTR - GAAP ETR + 1.427 ** 1.71 1.753 ** 2.06
HP Fin. Constraints Indicator -0.351 -1.00
BGY Fin. Constraints Indicator 0.309 * 1.89
Non-Distress Indicator 0.142 0.63
N
Adjusted R-Squared
S.E. Clustered by Firm
Year & Industry Fixed Effects Yes Yes
49.0% 50.2%
Yes Yes
Equlibrium Factor
440 440
62
Table 7 (continued) Economic Magnitude of Capital Structure Effects
Panel D: Regression Results with Deadweight Cost of Being Under- or Over-Levered as Dependent Variable
Notes: This table examines the economic cost of being sub-optimally levered from using the GAAP ETR in place of the MTR. Panel A presents the descriptive statistics of the variables used in analyses in the later panels. Panel B presents the univariate analyses of the economic magnitudes and Panels C and D present the regression results. Equilibrium Factor is defined as the intersection of the marginal benefit of debt and marginal cost of debt curves. If Equilibrium Factor = 1, the firm's equilibrium leverage is 100% of its actual (i.e., firm is at equilibrium). If Equilibrium Factor = 1.2, the firm should have 120% of its actual leverage (i.e., firm is underlevered). If Equilibrium Factor = 0.8, the firm should have 80% of its actual leverage (firm is overlevered). Total Deadweight Loss is the loss in firm value from being either underlevered or overlevered (see Figure 2). HP Fin. Constraints Indicator is indicator variable for firms in the top tercile of the yearly size-age index developed by Hadlock and Pierce (2010). BGY Fin. Constraints Indicator is an indicator variable for firm-years in the bottom two terciles of all of the following variables: long-term debt issuance, long-term debt reduction, equity issuance, and equity reduction. This variable follows from van Binsbergen, Graham, and Yang (2010). Non-Distress Indicator is an indicator variable for firm-years in the top tercile of the zscore distribution (following van Binsbergen, Graham, and Yang (2010)). The sample comprises of firms that say they use their GAAP ETR as the tax rate input for capital structure decisions. The regressions include both year and industry fixed effects and the standard errors are clustered at the firm-level. ***, **, * signifies statistical significance at the two-tailed (one-tailed) 1%, 5%, and 10% levels (for the coefficients of interest in the highlighted row). All the variables are defined in the Appendix.
Dependent Variable:
Pr. Sign Coefficient t -Statistic Coefficient t -Statistic
|MTR - GAAP ETR| + 0.037 *** 3.42 0.038 *** 3.39
HP Fin. Constraints Indicator -0.001 -0.27
BGY Fin. Constraints Indicator 0.001 0.71
Non-Distress Indicator 0.002 0.99
N
Adjusted R-Squared
S.E. Clustered by Firm
Year & Industry Fixed Effects
Yes Yes
Yes Yes
Total Deadweight Loss
440 440
51.7% 51.7%
63
Table 8 Analyses of Capital Investment Decisions
Panel A: Descriptive Statistics of the Variables Used in the Regression
Panel B: Regression Results
Notes: Panel (A) B in this table presents the results from (descriptive statistics of the variables used in) a regression of investment on investment opportunities, the difference between a firm’s MTR and its GAAP ETR, an interaction between the two variables, and cash flows. The sample comprises of firms that say they use their GAAP ETR as the tax rate input for capital expenditure decisions. The regressions include both year and industry fixed effects and the standard errors are clustered at the firm-level. ***, **, * signifies statistical significance at the two-tailed (one-tailed) 1%, 5%, and 10% levels (for the coefficients of interest in the highlighted row). All the variables are defined in the Appendix.
Variable N Mean SD P25 P50 P75
Capital Expenditures 648 0.088 0.092 0.027 0.058 0.111
|MTR A.I. - GAAP ETR| (JG) 461 0.151 0.211 0.028 0.052 0.255
|MTR A.I. - GAAP ETR| (BCG) 648 0.106 0.196 0.032 0.057 0.100
Tobin's Q 648 1.753 1.233 0.916 1.354 2.295
Sales Growth 601 0.190 0.259 0.040 0.147 0.267
CFO 648 0.139 0.112 0.072 0.127 0.200
Dependent Variable:
Investment Opportunities (INV_OPP) Proxy:
Method of Simulating MTRs:
Pr. Sign Coefficient t -Statistic Coefficient t -Statistic Coefficient t -Statistic Coefficient t -StatisticINV_OPP 0.019 *** 3.63 0.016 *** 3.48 0.062 ** 2.55 0.033 * 1.81
|MTR A.I. - GAAP ETR| 0.025 ** 1.97 0.011 1.33 0.018 1.58 0.005 0.54|MTR A.I. - GAAP ETR| × INV_OPP + -0.019 *** -3.99 -0.006 *** -2.54 -0.118 ** -1.99 -0.020 * -1.57CFO 0.186 *** 2.82 0.156 *** 3.48 0.244 *** 4.19 0.683 *** 3.63
N
Adjusted R-Squared
S.E. Clustered by Firm
Year & Industry Fixed Effects
Capital Expenditure
Yes
Yes
Yes
Yes
435
65.1%
630
67.2%
461
65.4%
Yes
Yes
Yes
Yes
648
71.2%
Tobin's Q
Graham/Shevlin
Sales Growth
Blouin, Core & Guay (2010)
Tobin's Q
Blouin, Core & Guay (2010)
Sales Growth
Graham/Shevlin
64
Table 9 Analyses of M&A Decisions
Panel A: Descriptive Statistics of the Variables Used in the Regression
Variable N Mean SD P25 P50 P75
Acquisition Announcement Returns 381 0.008 0.050 -0.022 0.006 0.037
|MTR A.I. - GAAP ETR| (JG) 307 0.107 0.137 0.019 0.035 0.128
|MTR A.I. - GAAP ETR| (BCG) 381 0.066 0.075 0.019 0.040 0.080
Pre-Acquisition Return Run Up 381 -0.064 0.411 -0.326 -0.082 0.185
Log(Assets) 381 7.442 1.573 6.239 6.912 8.831
Tobin's Q 381 2.777 2.517 1.227 1.740 2.955
Leverage 381 0.146 0.197 0.002 0.070 0.202
Cash Acquisition 381 0.193 0.206 0.033 0.101 0.254
Diversifying Acquisition 381 0.475 0.500 0.000 0.000 1.000
Public Target 381 0.433 0.496 0.000 0.000 1.000
Foreign Acquisition 381 0.223 0.417 0.000 0.000 0.000
Relative Value of Target 381 0.058 0.072 0.008 0.032 0.074
No. of Bidders 381 1.000 0.000 1.000 1.000 1.000
65
Table 9 (continued) Analyses of M&A Decisions
Panel B: Regression Results
Notes: Panel (A) B in this table presents the results from (descriptive statistics of the variables used in) a regression of acquisition announcement returns on the difference between a firm’s MTR and its GAAP ETR and control variables. The sample comprises of firms that say they use their GAAP ETR as the tax rate input for acquisition decisions. The regressions include both year and industry fixed effects and the standard errors are clustered at the firm-level. ***, **, * signifies statistical significance at the two-tailed (one-tailed) 1%, 5%, and 10% levels (for the coefficients of interest in the highlighted row). All the variables are defined in the Appendix.
Dependent Variable:
Method of Simulating MTRs:
Pr. Sign Coefficient t -Statistic Coefficient t -Statistic
|MTR A.I. - GAAP ETR| - -0.103 ** -2.01 -0.079 *** -2.51
Pre-Acquisition Return Run Up -0.003 -0.35 -0.007 -0.83
Log(Assets) -0.013 -0.92 -0.012 ** -2.07
Tobin's Q 0.003 ** 1.96 0.003 * 1.84
Leverage 0.028 0.88 0.033 * 1.72
Cash Acquisition -0.106 -0.81 -0.044 -0.78
Diversifying Acquisition -0.009 -1.12 -0.007 -0.86
Public Target -0.004 -0.40 -0.006 -0.72
Foreign Acquisition 0.007 1.04 0.011 1.52
Relative Value of Target 0.028 0.47 0.010 0.16
No. of Bidders 0.008 0.23 0.002 0.07
N
Adjusted R-Squared
S.E. Clustered by Firm
Year & Industry Fixed Effects
Acquisition Announcement Returns
Graham/Shevlin Blouin, Core, and Guay (2010)
Yes
Yes
Yes
Yes
381
1.8%
307
0.5%
66
Table 10 Falsification Tests
Panel A: Descriptive Statistics of the Variables Used in the Regressions
Variable N Mean SD P25 P50 P75
Debt Kink (JG) 211 4.016 3.373 1.600 3.000 6.000
Debt Kink (BCG) 211 2.105 2.447 0.400 1.200 3.000
MTR - GAAP ETR (JG) 211 -0.015 0.113 -0.052 -0.005 0.059
MTR - GAAP ETR (BCG) 211 -0.001 0.089 -0.053 -0.014 0.055
Log(Assets) 211 7.683 1.526 6.446 7.327 8.622
Tobin's Q 211 1.567 1.024 0.867 1.153 1.991
Sales Growth 211 0.106 0.145 0.016 0.072 0.149
CFO 211 0.029 0.016 0.017 0.027 0.039
PPE 211 0.300 0.191 0.148 0.241 0.397
Capital Expenditures 211 0.013 0.008 0.007 0.011 0.017
R&D Expenditures 211 0.007 0.010 0.000 0.002 0.011
Acquisition Expenditures 211 0.008 0.016 0.000 0.000 0.005
Leverage 211 0.250 0.137 0.143 0.247 0.348
Credit Rating Dummy 211 0.668 0.472 0.000 1.000 1.000
Z-Score 211 1.548 0.609 1.052 1.639 2.007
Dividend Dummy 211 0.573 0.496 0.000 1.000 1.000
Foreign Income 211 0.017 0.031 0.000 0.000 0.021
NOL Dummy 211 0.365 0.483 0.000 0.000 1.000
Capital Expenditures 346 0.061 0.060 0.025 0.044 0.077
|MTR A.I. - GAAP ETR| (JG) 243 0.153 0.199 0.032 0.066 0.226
|MTR A.I. - GAAP ETR| (BCG) 346 0.096 0.176 0.027 0.054 0.098
Tobin's Q 346 1.681 1.332 0.866 1.206 1.907
Sales Growth 310 0.135 0.236 0.019 0.078 0.175
CFO 346 0.127 0.092 0.070 0.116 0.174
Acquisition Announcement Ret. 157 0.007 0.043 -0.013 0.003 0.031
|MTR A.I. - GAAP ETR| (JG) 130 0.123 0.122 0.042 0.068 0.114
|MTR A.I. - GAAP ETR| (BCG) 157 0.080 0.082 0.030 0.054 0.076
Pre-Acquisition Return Run Up 157 -0.070 0.376 -0.357 -0.055 0.138
Log(Assets) 157 7.727 1.737 6.256 7.136 9.693
Tobin's Q 157 2.336 1.392 1.316 1.738 3.234
Leverage 157 0.140 0.179 0.021 0.058 0.229
Cash Acquisition 157 0.137 0.141 0.026 0.098 0.198
Diversifying Acquisition 157 0.510 0.502 0.000 1.000 1.000
Public Target 157 0.497 0.502 0.000 0.000 1.000
Foreign Acquisition 157 0.159 0.367 0.000 0.000 0.000
Relative Value of Target 157 0.074 0.092 0.007 0.031 0.102
No. of Bidders 157 1.000 0.000 1.000 1.000 1.000
Variable Used in the Investment Sensitivity Analyses
Variable Used in the Kink Analyses
Variable Used in the M&A Analyses
67
Table 10 (continued) Falsification Tests
Panel A: Analyses of Capital Structure Decisions for Firms using MTRs for Decision Making
Dependent Variable:
Method of Simulating MTRs:
Coefficient t -Statistic Coefficient t -Statistic
MTR - GAAP ETR -0.541 -0.39 2.107 1.12
Log(Assets) -0.130 -0.23 -1.115 ** -1.96
Tobin's Q 0.409 * 1.64 1.110 *** 4.45
Sales Growth 0.069 0.06 -0.043 -0.04
CFO 34.796 ** 2.47 13.106 0.93
PPE -6.355 ** -2.10 -5.470 * -1.81
Capital Expenditures 10.192 0.39 -5.737 -0.22
R&D Expenditures 19.024 0.30 -45.081 -0.71
Acquisition Expenditures 1.948 0.18 5.945 0.55
Leverage -5.900 *** -2.67 -2.383 -1.08
Credit Rating Dummy -0.709 -0.92 -0.169 -0.22
Z-Score -0.398 -0.56 1.334 * 1.86
Dividend Dummy 0.225 0.31 -0.286 -0.39
Foreign Income 3.467 0.45 1.777 0.23
NOL Dummy -0.644 * -1.35 -0.377 -0.80
N
Adjusted R-Squared
S.E. Clustered by Firm
Year & Industry Fixed Effects
Debt Kink
Graham/Shevlin Blouin, Core & Guay (2010)
211 211
79.9% 62.0%
Yes Yes
Yes Yes
68
Table 10 (continued) Falsification Tests
Panel B: Analyses of Capital Investment Decisions for Firms using MTRs for Decision Making
Dependent Variable:
Investment Opportunities (INV_OPP) Proxy:
Method of Simulating MTRs:
Coefficient t -Statistic Coefficient t -Statistic Coefficient t -Statistic Coefficient t -StatisticINV_OPP 0.013 *** 2.62 0.009 ** 2.34 0.013 0.38 0.045 1.37|MTR A.I. - GAAP ETR| -0.016 -1.34 -0.007 -0.61 -0.011 -0.85 -0.003 -0.19|MTR A.I. - GAAP ETR| × INV_OPP 0.003 0.33 0.008 0.83 0.089 1.23 -0.030 -0.20CFO 0.060 1.23 0.145 ** 2.00 -0.022 -0.21 0.155 1.62
N
Adjusted R-Squared
S.E. Clustered by Firm
Year & Industry Fixed Effects
Tobin's Q
Yes Yes Yes Yes
Yes Yes Yes Yes
223 311
69.9% 66.5% 36.2% 49.6%
243 346
Graham/Shevlin Blouin, Core & Guay (2010) Graham/Shevlin Blouin, Core & Guay (2010)
Capital Expenditure
Sales Growth Sales GrowthTobin's Q
69
Table 10 (continued) Falsification Tests
Panel C: Analyses of M&A Decisions for Firms using MTRs for Decision Making
Notes: The sample in this table is comprised of firms that use the MTR for decision-making. Panel A in this table presents the descriptive statistics of the variables used in the regressions presented in Panels B, C, and D. Panel B presents the results from a regression of the Debt Kink, which measures the conservativeness of a firm’s debt policy, on the difference firm’s MTR and its GAAP ETR, and control variables. Panel C presents the results from a regression of investment on investment opportunities, the difference between a firm’s MTR and its GAAP ETR, an interaction between the two variables, and cash flows. Panel D presents the results from a regression of acquisition announcement returns on the difference between a firm’s MTR and its GAAP ETR and control variables. The sample in Panel B (C, D) comprises of firms that say they use their MTR as the tax rate input for capital structure (capital expenditure, acquisition) decisions. All regressions include both year and industry fixed effects and the standard errors are clustered at the firm-level. ***, **, * signifies statistical significance at the two-tailed 1%, 5%, and 10% levels. All the variables are defined in the Appendix.
Dependent Variable:
Method of Simulating MTRs:
Coefficient t -Statistic Coefficient t -Statistic
|MTR A.I. - GAAP ETR| -0.016 -0.31 0.016 0.54
Pre-Acquisition Return Run Up -0.016 ** -2.52 -0.002 -0.21
Log Assets 0.001 0.09 -0.006 -1.26
Tobin's Q 0.018 * 1.74 0.012 ** 1.98
Leverage 0.090 1.13 0.018 0.37
Cash Acquisition -0.055 -0.64 -0.023 -0.52
Diversifying Acquisition -0.010 -1.04 -0.006 -0.54
Public Target -0.019 -0.97 -0.001 -0.07
Foreign Acquisition -0.018 * -1.76 -0.006 -0.62
Relative Value of Target -0.082 ** -2.36 -0.039 -0.73
No. of Bidders -0.111 *** -8.63 -0.118 *** -8.66
N
Adjusted R-Squared
S.E. Clustered by Firm
Year & Industry Fixed Effects
Acquisition Announcement Returns
Graham/Shevlin Blouin, Core & Guay (2010)
Yes Yes
130 157
19.9% 13.7%
Yes Yes