Tax Avoidance and DuPont Measures of Future Performance...tax incomes (a proxy for tax avoidance) provide information to market participants about earnings 6 Pretax return on operating

Electronic copy available at: http://ssrn.com/abstract=2581577

Tax Avoidance and DuPont Measures of Future Performance

Sharon Katz

Associate Professor of Accounting Columbia Business School

[email protected]

Urooj Khan Assistant Professor of Accounting

Columbia Business School [email protected]

Andrew P. Schmidt

Assistant Professor of Accounting Poole College of Management

North Carolina State University [email protected]

March 2015

Abstract: To date, there is mixed evidence on the implications of tax avoidance on firm value as measured by Tobin’s q or stock price reactions. The take-away from prior literature is that increased opportunities for rent extraction associated with tax avoidance (e.g., in low governance firms), might negatively affect the after-tax value of the firm. We revisit this topic by investigating the association between tax avoidance and firm fundamentals (leverage, profitability, and asset utilization), using DuPont analysis. We document that tax avoidance unambiguously lowers future pretax accounting rates of return (i.e., return on equity, return on net operating assets, and return on operating assets), largely due to inefficient utilization of operating assets and operating liabilities. These results also hold in different contexts that mitigate rent extraction, including when firms have foreign operations and good governance. ____________________________

We are thankful to Dan Amiram, Stephen Penman, Trevor Harris, Doron Nissim, Ed Outslay, Sonja Rego, Terry Shevlin, Ryan Wilson, the Iowa Tax Readings Group, the Texas Tax Readings Group, and the Arizona State Tax Readings Group, and workshop participants at the North Carolina State University, and the 2013 AAA annual meeting, for helpful comments and suggestions. We also gratefully acknowledge financial support provided by Columbia Business School. All errors are our own.


1

Tax Avoidance and DuPont Measures of Future Performance

INTRODUCTION

Despite the significant tax savings generated by tax avoidance activities (e.g., Wilson [2009];

Lisowsky, Robinson, and Schmidt [2013]), there is mixed evidence on the implications of tax

avoidance for firm value. The general consensus from prior literature is that increased opportunities

for rent extraction associated with tax avoidance,1 for example in low governance firms, might offset

the increase in the after-tax value of the firm.2 This literature examines the impact of tax avoidance

on firm value by focusing on Tobin’s q or stock price reactions. However, fundamental

characteristics of the firm are key determinants of firm value (e.g., Feltham and Ohlson [1995],

Abarbanell and Bushee [1997]). The accounting literature has largely ignored the effect of tax

avoidance on fundamental characteristics of firms.

We investigate the association among tax avoidance and future firm fundamentals. More

specifically, to better understand the mechanisms underlying the association between tax avoidance

and firm value, as well to shed light on the mixed evidence in the prior literature, we examine how

tax avoidance affects future firm fundamentals – profitability, asset utilization, and the leverage

effects.3 Further, we also investigate the influence of certain cross-sectional firm characteristics that

can affect the ability of management to extract rents (these include foreign operations and

governance structures) on the association between tax avoidance and firm fundamentals.

Tax avoidance activities increase after-tax cash flows, net assets, and financial slack of a

firm; the average tax shelter transaction generates federal tax savings between $206 and $376 million

(Wilson [2009]; Lisowsky, Robinson, and Schmidt [2013]). Desai and Dharmapala [2009] discuss

two alternative perspectives on the motivations and effects of corporate tax avoidance. First, they

1 Rent extraction refers to non-value maximizing activities pursued by the management at the expense of the shareholders/owners, including perk consumption, aggressive financial reporting, and related party transactions (Chen, Chen, Cheng, and Shevlin [2010]). 2 For example, Desai and Dharmapala [2009], Hanlon and Slemrod [2009], Wilson [2009], Koester [2011], Blaylock [2012], DeSimone and Stomberg [2012]. 3 We examine two types of leverage effects in this study, the effects of operating liability leverage (OLLEV_FX) and financial leverage (FLEV_FX). Nissim and Penman [2003] find that operating liabilities (i.e., accounts payable, deferred revenue) have different implications for future accounting rates of return than financial liabilities (i.e., debt); thus, the market prices the two types of leverage differently.


2

treat corporate tax avoidance as a substitute for other tax favored activity, like corporate borrowing

(e.g., Graham and Tucker [2006]). An alternative approach focuses on the agency perspective of tax

avoidance, where tax avoidance enables and masks rent extraction through the use of complex and

opaque structures. Therefore, ex-ante it is not obvious whether tax avoidance will be associated with

an improvement in firm fundamentals.

Desai and Dharmapala [2009], Hanlon and Slemrod [2009], Wilson [2009], and Koester

[2011] find evidence that investors value varying degrees of tax avoidance positively, but only in

well-governed firms.4 Therefore, we complement our analysis by examining whether the association

between tax avoidance and firm fundamentals varies in different contexts that mitigate or exacerbate

rent extraction. Specifically, we use the presence of foreign operations and governance structure to

test for differences in the association between tax avoidance and future firm fundamentals in the

cross-section.

We use DuPont analysis to study the relationships between tax avoidance and future firm

fundamentals. DuPont analysis decomposes return on common equity into return on net operating

assets and the financial leverage effect, and further decomposes the return on net operating assets

into net operating profit margin, net operating asset turnover, and the operating liability leverage

effect (Nissim and Penman [2001], [2003]). Net operating profit margin and net operating asset

turnover measure different aspects of a firm’s operations – pricing power and cost cutting drive the

operating margin, and asset turnover reflects efficient asset utilization (e.g., Nissim and Penman

[2001]; Soliman [2008]).5

We conduct our analysis on a sample of 75,543 firm-year observations from 1970-2012. We

measure tax avoidance as the quintile rank of the difference between a firm’s five-year statutory tax

rate (STR) and five-year effective tax rate (ETR). Overall, the association between tax avoidance and

future accounting rates of return (pretax return on common equity [PTROCE], pretax return on net

4 In contrast, Blaylock [2012] fails to find evidence of an association between tax avoidance and managerial rent extraction for a sample of U.S. firms. 5 In our analysis, we estimate our measures of operating profitability and its components on a pretax basis to control for the direct effect of reduced taxes on profits as a result of tax avoidance activities.

3

operating assets [PTRNOA], and pretax return on operating assets [PTROOA]) is negative.6 When

we decompose PTROCE, we find a positive association between tax avoidance and the future

financial leverage effect (FLEV_FX) but a (much larger) negative association between tax avoidance

and future pretax return on net operating assets (PTRNOA); these effects persist for periods up to

five-years ahead. When we decompose PTRNOA, we find a persistent, negative association between

both tax avoidance and the future operating liability leverage effect (OLLEV_FX) and tax avoidance

and future pretax return on operating assets (PTROOA). Tax avoiding firms utilize financial

(operating liability) leverage better (worse) than firms with less tax avoidance, therefore, the negative

association between tax avoidance and PTROCE is smaller than the negative association between tax

avoidance and PTRNOA, while the negative association between tax avoidance and PTRNOA is

larger than the negative association between tax avoidance and PTROOA.

Upon further decomposition of PTRNOA, we find a positive association between tax

avoidance and future pretax net operating profit margins (PTNOPM) and a (much larger) negative

association between tax avoidance and future net operating assets utilization (NOAT), which leads to

the negative association between tax avoidance and future PTRNOA. In summary, our evidence

suggests that poor asset utilization and operating liability utilization (not lower operating

profitability) drives the negative association between tax avoidance and future accounting rates of

return.7 In the tax literature, Lev and Nissim [2004] find that the ratio of taxable income to book

income predicts earnings growth, while a number of studies (e.g., Hanlon [2005]; Ayers, Jiang, and

LaPlante [2009]; Blaylock, Shevlin, and Wilson [2012]) document that differences between book and

tax incomes (a proxy for tax avoidance) provide information to market participants about earnings

6 Pretax return on operating assets (PTROOA) is an accounting rate of return that reflects operating profitability in the absence of all leverage effects (i.e., the effects of both operating liabilities (OLLEV_FX) and financial liabilities (FLEV_FX). As we show later in the paper, PTROOA + OLLEV_FX = PTRNOA, while PTRNOA + FLEV_FX = PTROCE. 7 Anecdotal evidence is consistent with the negative association between tax avoidance and operating asset utilization (i.e., NOAT). IBM recently reduced its tax rate to a two-decade low as assets increased and revenue declined for seven straight quarters (Barinka and Drucker [2014]). From 2011-2013, IBM’s net operating profit margins (NOPM) remained relatively flat at 20 percent, NOAT decreased from 4.46 to 2.91, and the operating liability leverage ratio dropped from 2.55 to 1.74.

4

quality.8 Our results suggest that the lower persistence of accounting rates of return in tax avoiding

firms is due to relatively poor balance sheet management (i.e., utilization of operating assets and

operating liabilities), rather than earnings quality.

The results of our cross-sectional analyses are similar to those of our primary analysis

reported above, regardless of the scope of foreign operations or the level of firm-specific governance.

We find that tax avoidance unambiguously lowers future pretax accounting rates of return (PTROCE,

PTRNOA, and PTROOA), even in firms with foreign operations or good governance. Among tax

avoiding firms, firms with foreign operations (good governance) have better future PTROOA and

PTRNOA than corresponding domestic firms (poorly governed firms). However, domestic (poorly

governed) firms use financial leverage more effectively than their multinational (good governance)

counterparts, such that the future PTROCE of firms with foreign operations (good governance) is

worse than (similar to) the future PTROCE of domestic (poorly governed) firms. In contrast, prior

studies suggest that tax avoidance can increase firm value when opportunities to extract rents are

relatively limited (e.g., Desai and Dharmapala [2009]).

Our primary contribution to the literature is a comprehensive analysis of the association

between tax avoidance and future performance. Unlike prior research that uses stock prices and

returns to measure performance (e.g., Desai and Dharmapala [2009], Hanlon and Slemrod [2009];

Wilson [2009], Koester [2011]), we provide direct evidence of the association between tax avoidance

and future performance by focusing on how tax avoidance affects future earnings and its

components; these are primary inputs in equity valuation (e.g., Feltham and Ohlson [1995],

Abarbanell and Bushee [1997]). Our analysis is also consistent with the view that an important

objective of accounting research is to discover how firm choices and financial statement information

affect future earnings (Penman [1992]).

8 Hanlon [2005] shows that firms with large temporary book-tax differences, which result in deferred taxes, have less persistent earnings (i.e., return on assets). Blaylock, Shevlin, and Wilson [2012] show that the source of temporary book-tax differences affects the implications of those differences; book-tax differences related to tax avoidance (earnings management) exhibit greater (lower) accruals and earnings persistence.

5

Second, many studies investigate the characteristics of profitability and demonstrate that

understanding the evolution of profitability improves the predictability of earnings (Freeman, Ohlson

and Penman [1982]; Fairfield, Sweeney, and Yohn [1996]; Fairfield and Yohn [2001]; Nissim and

Penman [2003]; Penman and Zhang [2006]). Graham, Raedy, and Shackelford [2012, p. 425] note

that room remains for future research on the association between book-tax differences and earnings

characteristics. We answer Graham et al. [2012]’s call and build on the prior literature that examines

tax avoidance and earnings characteristics by examining the association between tax avoidance and

DuPont measures of financial performance. Specifically, we examine the effects of tax avoidance on

pretax ROCE and its components (RNOA, NOPM, NOAT, OLLEV_FX and FLEV_FX) and

document that the negative association between tax avoidance and future accounting rates of return is

largely due to inefficient utilization of operating assets and operating liabilities.

LITERATURE REVIEW AND RESEARCH QUESTION DEVELOPMENT

Tax Avoidance and Firm Value

Tax avoidance activities increase the after-tax cash flows, net assets, and financial slack of a

firm. Graham and Tucker [2006] find that for their sample of 44 tax shelter cases, the average tax

shelter generates an annual tax deduction sufficient to shield income equal to approximately nine

percent of asset value. Wilson [2009] and Lisowsky, Robinson, and Schmidt [2013] find that the

average tax shelter transaction generates federal income tax savings of $206 to $376 million.

However, tax avoidance is costly; firms incur implementation costs, reputational costs, and agency

costs such as rent extraction as tax avoidance increases. Recent research focuses on the agency

perspective of tax avoidance, where tax avoidance enables and masks rent extraction through the use

of complex and opaque structures (Chen et al. [2010]; Balakrishnan et al. [2011]; and Desai and

Dharmapala [2006]; [2009]).

Despite the significant tax savings generated by tax avoidance activities, there is mixed

evidence on the implications of tax avoidance for firm value; these effects vary in the cross-section.

Desai and Dharmapala [2009], Wilson [2009], and Koester [2011] find evidence that investors value

varying degrees of tax avoidance positively, but only in well-governed firms. In contrast, Blaylock

6

[2012] documents that tax avoidance is generally positively associated with future relative

performance, even among poorly governed firms. DeSimone and Stomberg [2012] find that

investors value tax avoidance positively; this positive association more than doubles in income

mobile firms and the positive association in income mobile firms declines following the

implementation of FIN 48 in 2007.9 Desai, Dyck, and Zingales [2007] propose that for a given tax

rate, an increase in tax enforcement can increase the market value of a firm by reducing the rent

extraction of controlling shareholders. Consistent with their hypothesis, they find that the market

values of Russian firms targeted by enforcement actions increased and the voting premium for their

stocks decreased following an increase in tax enforcement. Frischmann, Shevlin, and Wilson [2008]

and Robinson and Schmidt [2013] document positive returns to the first disclosures of uncertain tax

benefits under FIN 48. Robinson and Schmidt [2013] further find that disclosure quality moderates

the positive association between returns and uncertain tax benefits; investors value tax avoidance

positively when managers mask the tax avoidance through poor disclosure.

Hanlon and Slemrod [2009] find that the market reacts negatively to press accounts of tax

shelter participation. However, the reaction varies in the cross-section; the negative abnormal

returns to firms in the retail sector are significant and significantly different from the returns of

non-retail firms, while the returns of firms with cash ETRs below the sample median are

negative and significant and significantly less than the returns of firms with cash ETRs above the

median. Similar to Desai and Dharmapala [2009], Wilson [2009], and Koester [2011], Hanlon and

Slemrod [2009] find that investors value tax avoidance positively in firms with good governance,

firms that have high ETRs, and firms that are not in the retail sector.

DuPont Analysis

The above-mentioned studies that investigate the effect of tax avoidance on firm value focus

on Tobin’s q or stock price reactions. Therefore, these studies provide indirect evidence of the

9 DeSimone and Stomberg [2012] characterize income mobile firms generally as those firms that have long-term, sustainable tax strategies that result from organizing global business operations in a tax efficient manner. Income mobile firms include multinational firms in high-tech industries with high profit margins and large amounts of R&D and advertising (DeSimone and Stromberg [2012]).

7

association between tax avoidance and future performance (Bernard [1993]). Hence, we still do not

fully understand the fundamental drivers of the association between tax avoidance and firm value.

We attempt to fill this void.

Expectations about current and future profitability are a primary input in equity valuation

(Feltham and Ohlson [1995]; Ohlson [1995]; Nissim and Penman [2001]). Beaver [1998] provides a

theoretical link between earnings and firm value by arguing that current earnings and profitability

provide information to predict future earnings. Consistent with this idea, several studies investigate

the association between future stock returns and current earnings and/or other financial statement

information. Lipe [1986] and Kormendi and Lipe [1987] document that transitory earnings

components have a smaller association with stock returns. Ou and Penman [1989], Lev and

Thiagarajan [1993], and Abarbanell and Bushee [1997] use financial ratios to estimate the “future

earnings power” of firms and estimate the association between accounting information and future

stock returns.

Studies in the tax literature document that the ratio of taxable income to book income

predicts earnings growth (Lev and Nissim [2004]; Weber [2009]), and that differences between book

and tax incomes (a proxy for tax avoidance) provide information to market participants and are

informative about earnings quality (e.g., Philips, Pincus, and Rego [2003]; Hanlon [2005]; Hanlon,

LaPlante, and Shevlin [2005]; Ayers, Jiang, and LaPlante [2009]; Blaylock, Shevlin, and Wilson

[2012]). Philips, Pincus, and Rego [2003] show that deferred taxes are a proxy for discretionary

accruals and Hanlon [2005] shows that firms with large temporary book-tax differences, which result

in deferred taxes, have less persistent earnings (i.e., return on assets). Blaylock, Shevlin, and Wilson

[2012] show that the source of temporary book-tax differences affects the implications of those

differences; book-tax differences related to tax avoidance (earnings management) exhibit greater

(lower) accruals and earnings persistence.

To better understand the mechanisms underlying the association between tax avoidance and

firm value, we examine the implications of tax avoidance for future performance using DuPont

analysis. The standard DuPont analysis decomposes return on equity into three components, profit

8

margin, asset turnover, and leverage. We use a modified DuPont decomposition introduced in

Nissim and Penman [2001]; [2003]. The Nissim and Penman [2001]; [2003] DuPont decomposition

provides for a clean distinction between the effects of operating and financing activities, which

translates into accounting rates of return that clearly distinguish the effects of operating profit

margins, operating asset utilization, operating liability leverage, and financial leverage. This

decomposition allows us to provide additional insight on the source of the tax avoidance/earnings

quality association.

Ceteris paribus, to the extent firms re-invest savings from tax avoidance activities in positive

net present value projects that enhance firm value; we expect a positive association between tax

avoidance and future performance components. However, if firms re-invest savings in negative net

present value projects or divert the excess cash towards perquisite consumption that destroys firm

value, we expect a negative association between tax avoidance and future performance components.10

Thus, on average, the direction of the association between tax avoidance and future performance (and

its components) is not obvious. Accordingly, our first hypothesis is as follows (we state all

hypotheses in the null):

H1A: Tax avoidance is not associated with future performance. H1B: Tax avoidance is not associated with future performance components.

Contextual Analysis of Performance and Tax Avoidance

The Effect of Foreign Operations

We first examine how the existence of foreign operations affects the association between tax

avoidance and future performance components. A number of studies find a positive association

between tax avoidance and the extent of foreign operations (Gupta and Newberry [1997]; Rego

[2003]; Wilson [2009]; Lisowsky [2009]; Lisowsky et al. [2013]). One way firms with foreign

operations reduce their tax burden is to shift income and expenses between high- and low-tax

jurisdictions. For example, in 2010, news articles revealed how Google used complex tax planning

10 Our hypothesis implies that managers re-invest or divert tax benefits into activities that affect future pretax earnings. Therefore, we examine pretax profitability components in all of our regression models.

9

strategies like the “Dutch Sandwich” and the “Double Irish” to reduce its overseas tax rate to 2.4%

(see Drucker [2010] for a detailed description of these tax planning strategies). These income-

shifting strategies moved most of Google’s profits through Ireland and the Netherlands to Bermuda.

Further, U.S. GAAP permits U.S. multinational corporations (MNCs) to defer taxes on the

earnings of foreign subsidiaries. The U.S. tax law creates incentives for U.S. MNCs to avoid the

repatriation of foreign earnings and hold greater amounts of cash abroad by designating foreign

earnings as permanently reinvested (PRE) (Krull [2004]; Foley et al. [2007]; Blouin, Krull, and

Robinson [2012]).11 However, research shows that excess cash holdings exacerbate agency costs;

managers use the excess cash to enhance their own power and/or wealth (Jensen [1986]; Harford

[1999]). Edwards, Kravet, and Wilson [2012] examine the potential agency costs for firms with PRE

and find that firms with high levels of PRE held as cash are more likely to make value-destroying

foreign acquisitions using cash.12 To the extent foreign operations aid managerial rent extraction of

the savings from tax avoidance, we should not document a positive association between future

performance and tax avoidance. Accordingly, we investigate whether the association between tax

avoidance and future performance varies for domestic and foreign firms.

H2: The association between tax avoidance and future performance components does not differ between firms with and without foreign operations.

Governance Structure

Several recent empirical studies examine the association between governance and tax

avoidance. Desai and Dharmapala [2006] find that equity compensation aligns managerial interests

and reduces managers’ incentives to invest in tax strategies that facilitate rent extraction. However,

the negative association between equity compensation and tax avoidance only holds when firms have

weak governance. Desai and Dharmapala [2009], Wilson [2009], and Koester [2011] document a

11 The FASB prescribes the accounting treatment for taxes on foreign earnings under Accounting Standard Codification section 740 (ASC 740). Under ASC 740, the designation of foreign earnings as PRE allows U.S. MNCs to defer the recognition of the U.S. tax expense related to foreign earnings. 12 Edwards et al. [2012] hand collect the data on PRE from 10-K filings for their sample of 284 observations, which is relatively small compared to ours. Our initial sample comprises of 75,543 firm-year observations. Instead of hand collecting the information on PRE for our much larger sample, we use foreign operations to proxy for the option of designating foreign earnings as PRE to be tax aggressive.

10

positive association between firm value and tax avoidance, but only for firms with high institutional

ownership. This suggests that shareholders value tax avoidance only when there are constraints on

managerial rent diversion. In contrast, Blaylock [2012] documents a positive association between

future performance and tax avoidance, even among poorly governed firms. Since corporate

governance structure can affect rent extraction and the likelihood that firms channel savings from tax

avoidance into positive NPV projects, we investigate whether the association between tax avoidance

and future profitability varies based on firm-specific governance structures.

H3: The association between tax avoidance and future performance components does not differ between firms with good and poor governance structures.

SAMPLE SELECTION AND RESEARCH DESIGN

Sample Selection

Our sample consists of 75,543 firm-year observations with available COMPUSTAT data for

the years 1970 through 2012.13 We restrict our sample to include U.S.-based domestic and

multinational firms listed on the NYSE, NASDAQ, and AMEX exchanges with common equity in

excess of $10 million (in 2012 dollars), positive net operating assets, and at least five consecutive

years of positive pretax income, cash taxes paid, and positive income tax expense. We further

exclude utilities (SIC 4900-4950) and a broad array of financial institutions (SIC codes 6000-6999).

Measures of Corporate Tax Avoidance

In our reported analysis, we use the industry/year quintile rank of the difference between a

firm’s five-year statutory tax rate (SETR5) and five-year effective tax rate (BETR5) as our measure of

tax avoidance (TAX_AVD).14 We define BETR5 as total tax expense summed over five years, scaled

by pretax income, summed over five years.15 This measure of tax avoidance conveys a firm’s

13 In order to ensure that the results of our tests are not driven by changes in the corporate tax rate and/or differences in the accounting for income taxes, we also perform all of our tests on a sample that begins in 1994 (the FASB adopted SFAS 109 in 1993 and Congress changed the corporate income tax rate to 35 percent in 1993). All of our inferences are qualitatively similar when we use this alternative sample. 14 We also use five-year cash ETRs in our definition of tax avoidance (in place of book ETRs). Our results are qualitatively similar using this alternative measure of tax avoidance. 15 Dyreng et al. [2008] compute long-run cash ETRs over five- and ten-year time intervals; we use a five-year ETR in order to maximize our sample. However, our results are qualitatively similar if we compute our tax avoidance measure over a ten-year interval.

11

average tax cost per dollar of pretax income and captures a broad range of tax planning activities that

have both certain and uncertain outcomes with tax authorities. We use the quintile rank of TAX_AVD

instead of its actual value to mitigate potential measurement error and make the interpretation of the

regression coefficient easier (Patatoukos [2012]).

Using DuPont Analysis to Investigate the Effect of Tax Avoidance on Future Performance

To investigate whether tax avoidance affects a firm’s future performance, we follow Nissim

and Penman [2001]; [2003] and decompose the return on common equity into its main components:

profit margins, asset turnover, and the financial leverage effect. Specifically, we follow equation (4)

in Nissim and Penman [2001] to decompose ROCE into its components:

ROCE = RNOA + [FLEV ˟ (RNOA – Net Borrowing Cost)] (1) = RNOA + [FLEV ˟ FL_SPRD] = RNOA + FLEV_FX

where ROCE is return on common equity, defined as net income at the end of year t, deflated by

common equity at the end of year t–1; RNOA is return on net operating assets, defined as operating

income after tax at the end of year t (net income before net financial expenses after taxes) divided by

net operating assets (operating assets minus operating liabilities) at the end of year t–1; FLEV is

financial leverage, defined as financial obligations (i.e., short- and long-term debt) minus financial

assets (i.e., cash, cash equivalents, and marketable securities) at the end of year t–1 divided by

common equity at the end of year t–1; and Net Borrowing Cost is net financial expense at the end of

year t, divided by net financial obligations at the end of year t–1. The difference between RNOA and

Net Borrowing Cost (i.e., the financial spread [FL_SPRD]) multiplied by FLEV captures the financial

leverage effect (FLEV_FX) on ROCE.

Nissim and Penman [2003] specify a type of financial statement analysis that distinguishes

leverage that arises in financing activities from leverage that arises in operating activities. This type

of analysis produces two leveraging equations, one for borrowing to finance operations (i.e., long-

term debt) and one for borrowing that arises in the normal course of operations (i.e., accounts

payable and deferred revenue). Nissim and Penman [2003] predict and find that investors price

operating liabilities and financing liabilities differently; operating liabilities typically levers

12

profitability more than financing liabilities and firms with higher operating liability leverage have

higher price-to-book ratios. Therefore, we further decompose RNOA to isolate the effects of

operating liabilities on performance.

RNOA = ROOA + [OLLEV ˟ (ROOA – Market Borrowing Rate)] (2) = ROOA + [OLLEV ˟ OL_SPRD] = ROOA + OLLEV_FX

We define RNOA above. ROOA is return on operating assets, defined as operating income after tax

plus the implicit interest on operating liabilities16 at the end of year t (net income before net financial

expenses after taxes) divided by operating assets at the end of year t–1; OLLEV is operating liability

leverage, defined as operating liabilities at the end of year t–1 divided by net operating assets at the

end of year t–1; and Market Borrowing Rate is the after-tax one year risk-free interest rate at the end

of year t–1. The difference between ROOA and Market Borrowing Rate (i.e., the operating liability

leverage spread [OL_SPRD]) multiplied by OLLEV captures the operating liability leverage effect

(OLLEV_FX) on RNOA.

In addition to decomposing ROCE and RNOA to identify the effects of leverage, we follow

Nissim and Penman [2001]; [2003], and also decompose RNOA into its DuPont components, profit

margin and asset utilization:

RNOA = NOPM ˟ NOAT (3)

where NOPM is net operating profit margin, defined as operating income after tax at the end of year

t, divided by sales at the end of year t; and NOAT is net operating asset turnover, defined as sales at

end of year t divided by net operating assets at the end of the year t–1. NOPM provides insights

about the sensitivity of operating income to product price and cost structure and is often the result of

a firm’s pricing power. NOAT is a measure of asset utilization and captures the firm’s efficiency in

employing operating assets to generate sales.

When we implement our financial statement analysis model empirically, we estimate all of

the ratios on a pretax basis to control for the direct effect of reduced taxes on profits as a result of tax

16 The numerator of ROOA adjusts operating income for the full implicit cost of trade credit.

13

avoidance activities. To investigate the association between tax avoidance and future performance,

we estimate the following equations (suppressing firm subscripts):

PTROCEt+ = 0 + 1TAX_AVDt + 2SIZEt + 3MBt + 4ACC_QUALt + 5EARN_VOLATILITYt + ∑kkINDUSTRYt + ∑kkYEARt + t

(4)

PTRNOAt+ = 0 + 1TAX_AVDt + 2SIZEt + 3MBt + 4ACC_QUALt + 5EARN_VOLATILITYt + ∑kkINDUSTRYt + ∑kkYEARt + t

(5)

PTROOAt+ = 0 + 1TAX_AVDt + 2SIZEt + 3MBt + 4ACC_QUALt + 5EARN_VOLATILITYt + ∑kkINDUSTRYt + ∑kkYEARt + t

(6)

The dependent variables, PTROCE, PTRNOA, and PTROOA represent the one-, three-, and

five-year ahead pretax return on common equity, return on net operating assets, and return on

operating assets, respectively. To investigate the association between tax avoidance and the specific

components of future performance, we replace the dependent variables in equations (4) – (6) with the

following future performance components: the financial leverage effect (FLEV_FXt+), the operating

liability leverage effect (OLLEV_FXt+), pretax net operating profit margin (PTNOPMt+), and net

operating asset turnover (NOATt+). TAX_AVD is the industry/year quintile rank of the difference

between a firm’s five-year statutory tax rate (SETR5) and five-year effective tax rate (BETR5). We

include a number of control variables that are correlated with tax avoidance and various measures of

future firm performance. SIZE is the natural log of total assets (Watts and Zimmerman [1983]), MB

is the market-to-book ratio (Lev and Nissim [2004]), ACC_QUAL is accruals quality, defined

consistent with Francis, LaFond, Olsson, and Schipper [2005], and EARN_VOLATILITY is the

volatility of pretax income, defined as the standard deviation of pretax income from t–1 to t–3

(Dichev and Tang [2008]). We report detailed variable definitions in Appendix A. Finally, we

winsorize all continuous independent variables at 1% and 99% of their respective distributions and

include year (YEAR) and industry (INDUSTRY) fixed-effects to control for fundamental differences

in tax avoidance that may exist across years and industries. We report t-statistics based on

heteroscedasticity-consistent clustered (by firm) standard errors.

14

Cross-Sectional Tests of the Association between Tax Avoidance and Future Performance

In order to gain a better understanding of the association between tax avoidance and firm

fundamentals, we examine two specific characteristics (scope of operations and governance) to help

us analyze whether the association between tax avoidance and firm fundamentals varies in different

contexts that mitigate or exacerbate rent extraction. We partition our sample based on each of these

firm characteristics and conduct our analysis separately for each group of firms (e.g. domestic versus

multinational firms, high governance versus low governance firms) and compare and contrast the

results for both groups of firms.

We examine the effect of the scope of a firm's operations by comparing domestic and

multinational firms. We classify a firm as multinational (FOPS = 1) if the firm reports non-zero

foreign pretax income, non-zero current foreign tax expense, or non-zero deferred foreign tax

expense. Otherwise, we classify a firm as domestic (FOPS = 0). We define governance consistent

with Desai and Dharmapala [2006]; [2009], based on the percentage of outstanding common stock

owned by institutional shareholders. We classify observations equal to or above (below) the median

of the industry/year institutional ownership distribution as firms with good (poor) governance firms

(i.e., [GGOV = 1 (GGOV = 0)]).

EMPIRICAL RESULTS

Descriptive Statistics

We present descriptive statistics in Table 1. For Table 1 only, we classify observations in the

top quintile of TAX_AVD as tax avoiders (TAX_AVD = 1) and firms in the bottom four quintiles of

TAX_AVD as firms that are not tax avoiders (TAX_AVD = 0). By construction, BETR5, the five-year

book ETR, is significantly lower for our sample of tax avoiding firms (we classify a firm as a tax

avoider if TAX_AVD falls in the highest quintile each TAX_AVD industry/year distribution) as

compared to firms that are not tax avoiders (mean of 0.209 vs. 0.442, respectively).

Table 1 further reports the descriptive statistics for a broad range of firm characteristics. Tax

avoiding firms have significantly lower PTROCE, PTRNOA, and PTROOA, relative to firms that are

not tax avoiders. Tax avoiding firms have significantly higher financial leverage effects (mean

15

FLEV_FX of 5.8 percent vs. 4.5 percent), while firms that are not tax avoiders have significantly

higher operating liability leverage effects (mean OLLEV_FX of 2.9 percent vs. 4.7 percent).17 When

we decompose PTRNOA into pretax net operating profit margin (PTNOPM) and net operating asset

turnover (NOAT), we find that tax avoiders have larger PTNOPM (mean of 11.9 percent vs. 11.6

percent), while firms that do not avoid taxes have larger NOAT (mean of 2.080 vs. 2.478). Lastly,

Table 1 shows that tax avoiding firms have significantly lower market-to-book ratios than firms that

are not tax avoiders (mean MB of 2.070 versus 2.307), which indicates that on average, the market

pays a premium for firms that avoid aggressive tax strategies.

[PLACE TABLE 1 HERE]

Tax Avoidance and Future Performance

The Association between Tax Avoidance and Pretax Accounting Rates of Return

We begin our multivariate analyses by examining the association between tax avoidance and

various measures of future performance. We present the results of equations (4) – (6) in Table 2, in

which we use PTROCE, PTRNOA, and PTROOA as our performance measures. The coefficients on

TAX_AVD are significantly negative for the one-, three-, and five-year ahead PTROCE, PTRNOA,

and PTROOA models. After controlling for other firm characteristics and industry membership, there

is an unambiguous negative association between tax avoidance and future performance. The

coefficient magnitudes suggest an economically important spread across firms with different levels

of tax avoidance. For the one-year ahead performance measures, the spread is 3.6 percent in terms of

pretax return on common equity, 4.0 percent in terms of return on net operating assets, and 2.4

percent in terms of return on operating assets. We calculate the rate of return differential as the

product of the TAX_AVD coefficient and the difference between the values of TAX_AVD for the

top and bottom quintile (e.g., -0.009*[5-1] in the PTROCE regression). The coefficients on

TAX_AVD are fairly stable three and five years into the future, which suggest that tax-avoiding firms

17 We focus the leverage discussion on the leverage effect – the amount of ROCE (RNOA) generated by the use of financial (operating liability) leverage, instead of a leverage ratio, which is a measure of the amount of liabilities a firm has in its capital structure. We measure the financial leverage effect (FLEV_FX) as PTROCE – PTRNOA and the operating liability leverage effect (OLLEV_FX) as PTRNOA – PTROOA.

16

earn lower accounting rates of return (relative to firms with less tax avoidance) well into the future.

The one-year ahead results are broadly consistent with the results in Hanlon [2005], who finds that

pretax return on assets declines as temporary book-tax differences increase. However, Hanlon

[2005] suggests that her results reflect earnings quality; the persistence of the PTROCE, PTRNOA,

and PTROOA coefficients into the future suggests that firms engage in tax avoidance activities (in

part) to mask poor pretax operating performance.


Tax Avoidance and DuPont Decompositions of Pretax RNOA: The Effect of Net Operating Profit Margin and Net Operating Asset Turnover

To understand why firms with greater tax avoidance have lower accounting rates of return,

we decompose pretax return on net operating assets to identify the effects of tax avoidance on future

operating profit margin and net operating asset turnover. In Panel A of Table 3, we present the results

of estimating regressions of future (one- three- and five- years ahead) PTRNOA components.

Specifically, we focus on the pretax net operating profit margin (PTNOPM) and the net operating

asset turnover (NOAT). In columns [1]-[3] of Panel A in Table 3, the coefficients on TAX_AVD are

significantly positive in each PTNOPM regression. The one-, three-, and five-year ahead net

operating profit margin of tax avoiding firms is 1.6 percent, 1.2 percent, and 0.8 percent,

respectively, higher than the corresponding net operating profit margin of firms that are not tax

avoiders.

We examine the association among tax avoidance and future net operating asset turnover in

columns [4]-[6] of Panel A in Table 3. The coefficients on TAX_AVD are significantly negative in

each NOAT regression; the one-, three-, and five-year ahead net operating asset turnover of tax

avoiding firms is 50.0 percent, 45.6 percent, and 47.2 percent, respectively, lower than the

corresponding net operating asset turnover of firms that are not tax avoiders. The negative

association between TAX_AVD and NOAT is larger than the positive association between TAX_AVD

and PTNOPM, which explains the overall negative association between tax avoidance and pretax

return on net operating assets that we document in columns [4]-[6] in Table 2. Therefore, the

17

negative association between tax avoidance and return on net operating assets is primarily due to

inefficient asset utilization instead of the profitability operations. This suggests that poor balance

sheet management is responsible for the negative association between tax avoidance and future

accounting rates of return.

Tax Avoidance and DuPont Decompositions of Pretax ROCE and RNOA: The Effect of Financial Leverage and Operating Liability Leverage

To further understand why firms with greater tax avoidance have lower accounting rates of

return, we decompose pretax return on common equity and pretax return on net operating assets to

identify the effects of tax avoidance on future financial leverage and operating liability leverage. In

Panel B of Table 3, we present the results of estimating regressions of the future (one- three- and

five- years ahead) leverage effect components of PTROCE and PTRNOA. Specifically, we focus on

the pretax financial leverage effect (FLEV_FX) in columns [1]-[3] and the pretax operating liability

leverage effect (OLLEV_FX) in columns [4]-[6] of Panel B. The coefficients on TAX_AVD are

positive in each FLEV_FX regression and generally significant. The one-, three-, and five-year ahead

financial leverage effect of tax avoiding firms is 0.4 percent (not significant), 1.2 percent, and 1.2

percent, respectively, higher than the corresponding financial leverage effects of firms that are not

tax avoiders. This positive financial leverage effect explains why the negative association between

TAX_AVD and PTROCE (presented in columns [1]-[3] of Table 2) is smaller than the negative

association between TAX_AVD and PTRNOA (presented in columns [4]-[6] of Table 2).

We next examine the future operating liability leverage effect of firms with different degrees

of tax avoidance. The coefficients on TAX_AVD are significantly negative in each OLLEV_FX

regression; the one-, three-, and five-year ahead operating liability leverage effect of tax avoiding

firms are all 1.6 percent, lower than the corresponding operating liability leverage effects of firms

that are not tax avoiders. The negative operating liability leverage effect explains why the negative

association between TAX_AVD and PTRNOA (presented in columns [4]-[6] of Table 2) is larger than

the negative association between TAX_AVD and PTROOA (presented in columns [7]-[9] of Table 2).

18

These results provide further evidence that poor balance sheet management is responsible for the

negative association between tax avoidance and future accounting rates of return.


Cross-Sectional Tests of the Effect of Tax Avoidance on Future Profitability

Existence of Foreign Operations

We present the results of tests that examine how tax avoidance affects various summary

measures of future performance (PTROOA, PTRNOA, and PTROCE) of multinational versus

domestic firms in Panels A, C, and E, respectively, of Table 4. We report the one-, three-, and five-

year ahead tax avoidance associations of multinational (FOPS = 1) and domestic (FOPS = 0) firms in

columns [1]-[2], [3]-[4], and [5]-[6], respectively, of each panel in Table 4. In the PTROOA,

PTRNOA, and PTROCE regressions, all of the coefficients on TAX_AVD are negative and

significant, regardless of the performance measure or the scope of firm operations (multinational or

domestic). In addition, the negative associations are largely stable; the negative effects of tax

avoidance manifest early and persist for at least five years. We next examine specific ratio

components to provide insights on the source of these negative associations.

Tax Avoidance and DuPont Decompositions of Pretax ROCE: Pretax ROOA, the

Financial Leverage Effect and the Operating Liability Leverage Effect. We first examine the

association between TAX_AVD and pretax return on operating assets (PTROOA). PTROOA

represents the accounting rate of return on operations, excluding the effects of both operating liability

leverage (OLLEV_FX) and financial leverage (FLEV_FX). The results in columns [1] and [2] of

Panel A indicate that the negative association between TAX_AVD and one-year ahead PTROOA is

significantly smaller for multinational firms relative to domestic firms (2PTROOA = 20.64, p <= 0.01).

However, the multinational advantage dissipates quickly; the association between TAX_AVD and

three- (five-) year ahead PTROOA for multinational firms is no different from the association

between TAX_AVD and three- (five-) year ahead PTROOA for domestic firms (see columns [3] –

[6]).

The results in Panel B of Table 4 show how tax avoidance affects the use of operating

19

liability leverage by multinational and domestic firms. The operating liability leverage effect

(OLLEV_FX) explains how firms lever up pretax return on net operating assets (PTRNOA) relative to

PTROOA. In the OLLEV_FX regressions, all of the coefficients on TAX_AVD are negative and

significant; tax avoiding multinational and domestic firms have lower future OLLEV_FX than their

counterparts that avoid fewer taxes. However, there is no difference in the association between

TAX_AVD and OLLEV_FX for multinational and domestic firms until five years ahead; the

association between TAX_AVD and five-year ahead OLLEV_FX for multinational firms is

significantly different from the association between TAX_AVD and five-year ahead OLLEV_FX for

domestic firms (1FOPS = 1 = -0.005 < 1

FOPS = 0 = -0.003, 2 = 3.82, p <= 0.05). The negative

associations between TAX_AVD and OLLEV_FX suggest that the associations between TAX_AVD

and PTRNOA will be more negative than the associations between TAX_AVD and PTROOA.

We examine the association between TAX_AVD and pretax return on net operating assets

(PTRNOA) in Panel C of Table 4. The coefficients on TAX_AVD are all negative and significant and

are more negative than the coefficients on TAX_AVD from the PTROOA regressions. Similar to the

results we reported on the associations between TAX_AVD and PTROOA in Panel A, we find that the

negative association between TAX_AVD and one-year ahead PTRNOA is significantly smaller for

multinational firms relative to domestic firms (2PTRNOA = 9.56, p <= 0.05), while there is no

difference in the association between TAX_AVD and three-year ahead PTRNOA of multinational and

domestic firms. However, the association between TAX_AVD and five-year ahead PTRNOA for

multinational firms is significantly different from the association between TAX_AVD and five-year

ahead PTRNOA for domestic firms (1FOPS = 1 = -0.012 < 1

FOPS = 0 = -0.009, 2 = 3.81, p <= 0.05). As

we discussed in the previous paragraph, this deterioration in performance, relative to PTROOA, is

due to the inefficient use of operating liabilities.

The results in in Panel D of Table 4 document how tax avoidance affects the use of financial

leverage by multinational and domestic firms. The financial leverage effect (FLEV_FX) explains

how firms lever up pretax return on common equity (PTROCE) relative to PTRNOA. In the

20

FLEV_FX regressions, all of the coefficients on TAX_AVD for domestic firms are positive and

significant, while the coefficients on TAX_AVD for multinational firms are generally not different

from zero. Tax avoiding domestic firms use financial leverage more effectively than tax avoiding

multinational firms; the difference in the coefficients on TAX_AVD between domestic and

multinational firms is highly significant in each future period. The lower FLEV_FX of multinationals

relative to domestic firms is consistent with multinational firms holding more cash abroad by

designating foreign earnings as permanently reinvested, which leads to a less than optimal leverage

structure and a reduction in future performance.18 The positive (lack of) associations between

TAX_AVD and FLEV_FX for domestic (multinational) firms suggest that the associations between

TAX_AVD and PTROCE for domestic (multinational) firms will be less negative than (similar to) the

corresponding associations between TAX_AVD and PTRNOA.

We examine the association between TAX_AVD and pretax return on common equity

(PTROCE) in Panel E of Table 4. The coefficients on TAX_AVD are all negative and significant and

are in general more negative than the coefficients on TAX_AVD from the PTROOA regressions. In

contrast to the results we reported on the associations between TAX_AVD and PTROOA (PTRNOA)

in Panel A (C), we find that the negative association between TAX_AVD and one-year ahead

PTROCE is significantly larger for multinational firms relative to domestic firms (1FOPS = 1 = -0.010

< 1FOPS = 0 = -0.008, 2

PTROCE = 3.94, p <= 0.05). Further, this difference between the magnitude of

multinational and domestic TAX_AVD coefficients grows as we examine the association between

TAX_AVD and three- (1FOPS = 1 = -0.011 < 1

FOPS = 0 = -0.004, 2PTROCE = 31.91, p <= 0.01) and five-

year ahead PTROCE (1FOPS = 1 = -0.012 < 1

FOPS = 0 = -0.004, 2PTROCE = 30.69, p <= 0.01). As we

discussed in the previous paragraph, this relative underperformance of tax avoiding multinationals is

due to the superior use of financial leverage by tax avoiding domestic firms.

18 If we decompose FLEV_FX, and examine the effects of TAX_AVD on future financial leverage ratios (i.e., FLEV), we find that the coefficient on TAX_AVD for multinational firms is significantly lower than the coefficient on TAX_AVD for domestic firms in one-, three-, and five-year ahead regressions (all p <= 0.01). Further, unreported descriptive statistics indicate that multinational firms hold significantly more financial assets (FA) than domestic firms (mean (median) multinational FA = 515.52 (54.76); mean (median) domestic FA = 199.83 (11.30); t = 22.68, p = 0.000).

21

Tax Avoidance and DuPont Decompositions of Pretax RNOA: The Effect of Net

Operating Profit Margin and Net Operating Asset Turnover. The preceding discussion suggests

that tax avoiding multinationals enjoy a short-term operating performance (PTROOA and PTRNOA)

advantage over tax avoiding domestic firms, but this advantage dissipates over time. In order to

better understand the evolution of PTRNOA, we separately examine the association of TAX_AVD and

the two multiplicative components of PTRNOA, pretax net operating profit margin (PTNOPM) and

net operating asset turnover (NOAT).19 We present the results of future PTNOPM (NOAT)

regressions in Panel F (G) of Table 4. In the PTNOPM regressions, all of the coefficients on

TAX_AVD for domestic firms are positive and significant, while the coefficients on TAX_AVD for

multinational firms are positive and significant only for one-year ahead PTNOPM (the coefficients

on TAX_AVD for multinational firms in the three- and five-year ahead PTNOPM regressions are

statistically indistinguishable from zero). In addition, the coefficients on TAX_AVD in the PTNOPM

regressions for domestic firms are all significantly larger than the coefficients on TAX_AVD in the

PTNOPM regressions for multinational firms.

However, the coefficients on TAX_AVD in the NOAT regressions are all negative and

significant; tax-avoiding firms have lower asset turnovers than firms that use less tax avoidance. We

find that the negative association between TAX_AVD and one-year ahead NOAT is significantly

smaller for multinational firms relative to domestic firms (1FOPS = 1 = -0.110 < 1

FOPS = 0 = -0.132,

2NOAT = 2.73, p <= 0.10), while there is no difference in the association between TAX_AVD and the

three- or five-year ahead NOAT of multinational and domestic firms. The magnitudes of the NOAT

coefficients are many times greater than the magnitudes of the PTNOPM coefficients, which suggests

that the negative association between TAX_AVD and PTRNOA (and PTROOA) is primarily due to

inefficient asset utilization, rather than the profitability of sales.


19 Our inferences are similar if we examine the two multiplicative components of PTROOA, pretax net operating profit margin (PTNOPM) and operating asset turnover (OAT).

22

Governance Structure

We present the results of tests that examine how tax avoidance affects various summary

measures of future performance (PTROOA, PTRNOA, and PTROCE) of firms with good and poor

governance in Panels A, C, and E, respectively, of Table 5. We classify observations above (below)

the median of the industry/year institutional ownership distribution as having good (poor)

governance. We report the one-, three-, and five-year ahead tax avoidance associations of firms with

good (GGOV = 1) and poor (GGOV = 0) governance in columns [1]-[2], [3]-[4], and [5]-[6],

respectively, of each panel in Table 5. Similar to all of the previous PTROOA, PTRNOA, and

PTROCE regressions, all of the coefficients on TAX_AVD are negative and significant, regardless of

the performance measure or the firm-specific governance structure (good or poor). In addition, the

negative associations are largely stable; the negative effects of tax avoidance manifest early and

persist for at least five years. Next, we examine specific ratio components to provide insights on the

source of these negative associations for firms with good and poor governance.

Tax Avoidance and DuPont Decompositions of Pretax ROCE: Pretax ROOA, the

Financial Leverage Effect and the Operating Liability Leverage Effect. We first examine the

association between TAX_AVD and pretax return on operating assets (PTROOA) and report the

results in Panel A of Table 5. The results indicate that the negative association between TAX_AVD

and one-, three-, and five-year ahead PTROOA is significantly smaller for firms with good

governance relative to firms with poor governance (2PTROOA = 9.36, p <= 0.01, 2

PTROOA = 7.99, p <=

0.01,2PTROOA = 8.73, p <= 0.01, respectively). Tax avoiders earn lower pretax operating asset

returns relative to firms with less tax avoidance, but good governance moderates this negative

association.

We use the data in Panel B of Table 5 to examine how tax avoidance affects the use of

operating liability leverage by firms with good versus poor governance. In the OLLEV_FX

regressions, all of the coefficients on TAX_AVD are negative and significant; tax-avoiding firms with

good and poor governance have lower future OLLEV_FX than their counterparts that avoid fewer

taxes. However, there is no difference in the association between TAX_AVD and OLLEV_FX for

23

firm-years with different governance in any of the future periods. The negative associations between

TAX_AVD and OLLEV_FX suggest that the associations between TAX_AVD and PTRNOA will be

more negative than the associations between TAX_AVD and PTROOA.

We examine the association between TAX_AVD and pretax return on net operating assets

(PTRNOA) of firms with good and poor governance in Panel C of Table 5. The coefficients on

TAX_AVD are all negative and significant and are more negative than the coefficients on TAX_AVD

from the PTROOA regressions. Similar to the results we reported on the associations between

TAX_AVD and PTROOA in Panel A, we find that the negative association between TAX_AVD and

one-, three-, and five-year ahead PTRNOA is significantly smaller for firms with good governance

relative to firms with poor governance (2PTRNOA = 5.51, p <= 0.05, 2

PTRNOA = 5.11, p <=

0.05,2PTRNOA = 2.71, p <= 0.10, respectively). As we discussed in the previous paragraph, this

deterioration in performance, relative to PTROOA, is due to the inefficient use of operating liabilities.

We next use the data in Panel D of Table 5 to examine how tax avoidance affects the use of

financial leverage by firms with good and poor governance. In the FLEV_FX regressions, all of the

coefficients on TAX_AVD for poorly governed firms are positive, but only the three-year ahead

TAX_AVD coefficient is significant. For firms with good governance, the TAX_AVD coefficient for

one-year ahead FLEV_FX is negative and (marginally) significant, while the three- and five-year

ahead TAX_AVD coefficients are positive, but statistically indistinguishable from zero. However, tax

avoiding firms with poor governance use financial leverage more effectively than tax avoiding firms

with good governance; the difference in the coefficients on TAX_AVD between firms with good and

poor governance is highly significant one- and three-years ahead (2FLEV_FX = 8.15, p <= 0.01,

2FLEV_FX = 7.00, p <= 0.01, respectively). This is an unexpected result; however, the results of

unreported regressions of TAX_AVD on future financial leverage (FLEV) indicate that poorly

governed firms have significantly more net debt in their capital structure than firms with good

governance. Given that tax-avoiding firms with poor governance use financial leverage more

effectively than tax avoiding firms with good governance, we expect that the differences in the

24

association between TAX_AVD and PTROCE for firms with good/poor governance will be smaller

than the respective differences in the association between TAX_AVD and PTRNOA for these firms.

We examine the association between TAX_AVD and pretax return on common equity

(PTROCE) of firms with good/poor governance in Panel E of Table 5. Similar to the PTROOA

(PTRNOA) regression results that we report in Panel A (C) of Table 5, the coefficients on TAX_AVD

are all negative and significant. In contrast to the results we reported on the associations between

TAX_AVD and PTROOA (PTRNOA) in Panel A (C), we find no difference in the association between

TAX_AVD and one-, three-, and five-year ahead PTROCE for firms with good and poor governance.

As we discussed in the previous paragraph, the difference in the association between TAX_AVD and

future PTROCE for firms with good and poor governance is smaller than the corresponding

difference in the association between TAX_AVD and future PTRNOA because tax avoiding poorly

governed firms use financial leverage more effectively than tax avoiding firms with good

governance.

Tax Avoidance and DuPont Decompositions of Pretax RNOA: The Effect of Net

Operating Profit Margin and Net Operating Asset Turnover. The preceding discussion suggests

that tax-avoiding firms with good governance enjoy a short-term operating performance (PTROOA

and PTRNOA) advantage over tax avoiding firms with poor governance. In order to better

understand the evolution of PTRNOA for firms with good and poor governance, we separately

examine the association of TAX_AVD and the two multiplicative components of PTRNOA, pretax net

operating profit margin (PTNOPM) and net operating asset turnover (NOAT).20 We present the

results of future PTNOPM (NOAT) regressions in Panel F (G) of Table 5. In the PTNOPM

regressions, all of the coefficients on TAX_AVD for firms with good governance are positive and

significant, while the coefficients on TAX_AVD for firms with poor governance are statistically

indistinguishable from zero. In addition, the coefficients on TAX_AVD in the PTNOPM regressions

20 Our inferences are similar if we examine the two multiplicative components of PTROOA, pretax net operating profit margin (PTNOPM) and operating asset turnover (OAT).

25

for firms with good governance are all significantly larger than the coefficients on TAX_AVD in the

PTNOPM regressions for firms with poor governance.

However, the coefficients on TAX_AVD in the NOAT regressions are all negative and

significant; tax-avoiding firms have lower asset turnovers than firms that use less tax avoidance. In

addition, there is no difference in the association between TAX_AVD and the one-, three- or five-year

ahead NOAT of firms with different governance structures. The magnitudes of the NOAT

coefficients are many times greater than the magnitudes of the PTNOPM coefficients, which suggests

that the negative association between TAX_AVD and PTRNOA (and PTROOA) is primarily due to

inefficient asset utilization, rather than the profitability of sales.


Discussion of Cross-Sectional Results

The results of the cross-sectional tests largely confirm our main results. Irrespective of the

scope of operations or governance structure, there is an unambiguous negative association between

tax avoidance and future accounting rates of return; this negative association is largely the result of

poor operating asset and operating liability utilization that outweighs the benefits derived from

operating profitability. The negative association between tax avoidance and PTROOA and PTRNOA

is significantly smaller for multinational firms relative to domestic firms. However, these results

reverse as we move from analyzing future PTRNOA to future PTROCE; the negative association

between tax avoidance and PTROCE is significantly larger for multinational firms relative to

domestic firms. This reversal is due to the ineffective use of financial leverage by multinational

firms relative to domestic firms; multinational firms hold more cash abroad by designating foreign

earnings as permanently reinvested, which leads to a less than optimal financial leverage structure

and a reduction in future performance.

The results of the governance tests are largely similar to the results of the

multinational/domestic firm tests. The negative association between tax avoidance and PTROOA and

PTRNOA is significantly smaller for firms with good governance relative to poorly governed firms;

tax-avoiding firms with good governance have better future pretax profit margins but similar asset

26

turnovers than poorly governed tax avoiding firms. However, firms with poor governance make

better use of future financial leverage such that we fail to find a difference in the association between

tax avoidance and PTROCE for firms with good and poor governance.

Robustness Tests

Our tests examine the association between tax avoidance and various measures of firm

performance. However, the decision to engage in transactions considered as tax avoidance is

voluntary; therefore our sample could be systematically biased due to selection-related concerns.

Therefore, we should estimate some form of an endogenous treatment regression model (Heckman

[1976, 1978, 1979]; Maddala [1983]; Wooldridge [2010]) to control for the endogeneity of tax

avoidance. Larker and Rusticus [2008] and Lennox, Francis, and Wang [2012] emphasize that

selection models should include at least one exogenous exclusion restriction in order to successfully

control for endogeneity.21 22

Models of the determinants of tax avoidance (e.g., Gupta and Newberry [1997]; Dyreng,

Hanlon, and Maydew [2008]; Hoopes, Mescall, and Pittman [2012]), from which we would base our

first-stage “selection” regression, almost exclusively include variables (e.g., size, leverage, inventory

intensity, research and development intensity, capital expenditures, advertising intensity, SG&A

intensity, and foreign income) that correlate with the primary outcome of interest in our second-stage

regression, future performance. Therefore, we do not believe that it is feasible to identify an

appropriate exogenous exclusion restriction that would allow us to implement a credible selection

model. Lennox et al. [2012, p. 610] suggest that it is better science when authors admit to having

difficulty identifying proper exclusion restrictions rather than make claims of having controlled for

selection bias using an ad hoc specification of a selection model that corroborates any OLS findings.

21 An exclusion restriction occurs when at least one independent variable that is correlated with the dependent variable in the first-stage model is not correlated with the dependent variable in the second-stage model and is therefore excluded from the second-stage regression. If the variables in each stage of the regression model are the same, the identification in the second stage can be weak; the two-stage least squares approach can be unreliable and researchers should interpret the results with caution (Tucker [2010]). 22 Lennox et al [2012] acknowledge that it is possible to estimate a selection model in the absence of an exogenous exclusion restriction because the inverse Mills ratio is nonlinear in its arguments. However, econometricians advise against using nonlinearities to identify selection bias because of potential model misspecification and multicollinearity.

27

Rather than use an ad hoc specification to try and address any selection related issues, we

follow the suggestion in Lennox et al. [2012, p. 611] and re-estimate all of our regressions using

panel data. As long as any of the unobservable factors that affect the choice of tax strategy remain

constant during our sample period, we can control for them using a fixed effects panel data model.23

When we estimate two-way (firm and year) fixed effects models in place of the OLS regressions, we

find results that are quantitatively and qualitatively similar to the results that we present in Tables 2 –

5. In addition, we repeat the panel data tests using a shorter sample period (beginning in 1994 after

the introduction of SFAS 109 and the change in corporate tax rates). All of our inferences are

qualitatively similar when we use this alternative sample with the panel data estimator.

CONCLUSION

We study the association between tax avoidance and future performance. We use the DuPont

framework to identify the components of future performance. We complement the DuPont analysis

by examining whether there is cross-sectional variation in the association between tax avoidance and

future performance. On average, there is an unambiguous negative association between tax

avoidance and future performance; the negative association between tax avoidance and future

performance is the result of poor operating asset and operating liability utilization and efficiency, not

lower operating profitability. These results persist in various contexts that mitigate or exacerbate rent

extraction, such as the existence of foreign operations and better governance structures.

Prior studies have used stock price related information to assess the impact of tax avoidance

on firm value. Since prices and returns summarize shareholders’ expectation about future

performance of the firm, these studies provide indirect evidence of the effect of tax avoidance on

future performance. Our paper contributes to the literature by providing direct evidence of the

association between tax avoidance and future performance, allowing for a better understanding of the

underlying mechanisms through which tax avoidance affects firm value.

23 However, the time-invariance of unobservables is a strong assumption and researchers cannot empirically validate this assumption (Lennox et al. [2012]).

28

REFERENCES

Abarbanell, J., and B. Bushee. 1997. Fundamental analysis, future earnings, and stock prices. Journal of Accounting Research 35 (1): 1–24.

Armstrong, C., A. Jagolinzer, and D. Larcker. 2010. Chief executive officer equity incentives and

accounting irregularities. Journal of Accounting Research 48: 225-271. Angrist, J. and J. Pischke. 2009. Mostly harmless econometrics. Princeton University Press, Princeton,

NJ. Ayers, B., J. Jiang, and S. LaPlante. 2009. Taxable income as a performance measure: The effects of tax

planning and earnings quality. Contemporary Accounting Research 26 (1): 15-54. Balakrishnan, K., J. Blouin, and W. Guay. 2011. Does tax aggressiveness reduce transparency? Wharton

University, Working Paper. Barinka, A, and J. Drucker. 2014. IBM uses Dutch tax haven to boost profits as sales slide. Bloomberg. Barth, M., W. Beaver, J. Hand, and W. Landsman. 1999. Accruals, cash flows, and equity values. Review

of Accounting Studies 4 (3-4): 205-229. Beaver, W. 1998. Financial reporting: An accounting revolution. Third edition. Upper Saddle, NJ:

Prentice Hall. Bernard, V, 1993. Discussion of an investigation of revaluations of tangible long-lived assets. Journal of

Accounting Research 31: 39-45. Blaylock, B. 2012. Do managers extract economically significant rents through tax aggressive

transaction? Oklahoma State University, Working Paper. Blaylock, B., T. Shevlin, and R. Wilson. 2012. Tax avoidance, large positive temporary book-tax

differences, and earnings persistence. The Accounting Review 87 (1): 91-120. Blouin, J., L. Krull, and L. Robinson. 2012. Where in the world are “permanently reinvested” foreign

earnings? University of Pennsylvania, University of Oregon and Dartmouth College Working Paper.

Chen, S., X. Chen, Q. Cheng, and T. Shevlin. 2010. Are family firms more tax aggressive than non-

family firms? Journal of Financial Economics 95: 41-61. Desai, M., and D. Dharmapala. 2006. Corporate tax avoidance and high-powered incentives. Journal of

Financial Economics 79 (1): 145-179. Desai, M., and D. Dharmapala. 2009. Corporate tax avoidance and firm value. Review of Economics and

Statistics 91 (3): 537-546. Desai, M., A. Dyck, and L. Zingales. 2007. Theft and taxes. Journal of Financial Economics 84 (3): 591-

623. De Simone, L., and B. Stomberg. 2012. Do investors differentially value tax avoidance of income mobile

firms? Stanford University and University of Georgia, Working Paper. Available at http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2102903

29

Dichev, I. D., and V. W. Tang. 2008. Earnings volatility and earnings predictability. Journal of

Accounting and Economics 47 (1-2): 160-181. Drucker, J. 2010. Google 2.4% rate shows how $60 billion lost to tax loopholes. Bloomberg. Dyreng, S., M. Hanlon, and E. Maydew. 2008. Long-Run Corporate Tax Avoidance. The Accounting

Review 83 (1): 61-82 Edwards, A., T. Kravet, and R. Wilson. 2012. Permanently reinvested earnings and the profitability of

foreign cash acquisitions. University of Toronto, University of Texas at Dallas and University of Iowa working paper.

Fairfield, P., R. Sweeney, and T. Yohn. 1996. Accounting classification and the predictive content of

earnings, The Accounting Review 71 (3): 337-355. Fairfield, P., and T. Yohn. 2001. Using asset turnover and profit margin to forecast changes in

profitability. Review of Accounting Studies 6 (4): 371-385. Feltham, G. and J. Ohlson. 1995. Valuation and clean surplus accounting for operating and financial

activities. Contemporary Accounting Research 11: 689–732. Foley, F., J. Hartzell, S. Titman, and G. Twite. 2007. Why do firms hold so much cash? A tax-based

explanation. Journal of Financial Economics 86: 579-607. Francis, J., R. LaFond, P. Olsson, and K. Schipper. 2005. The market pricing of accruals quality. Journal

of Accounting and Economics 39: 295-327. Freeman, R., J. Ohlson, and S. Penman. 1982. Book rate of return and prediction of earnings changes: An

empirical investigation. Journal of Accounting Research 20: 3-42. Frischmann, P., T. Shevlin, and R. Wilson, 2008. Economic consequences of increasing the conformity in

accounting for uncertain tax benefits. Journal of Accounting and Economics 46 (2-3): 261-278. Graham, J. R., J. Raedy, and D. Shackelford. 2012, Research in accounting for income taxes. Journal of

Accounting and Economics 53: 412-434. Graham, J.R., and A. L. Tucker. 2006. Tax shelters and corporate debt policy. Journal of Financial

Economics 81(3): 563-594. Gupta, S., and K. Newberry. 1997. Determinants of the variability in corporate effective tax rates:

Evidence from longitudinal data. Journal of Accounting and Public Policy 16 (1): 1-34. Hanlon, M. 2005. The persistence and pricing of earnings, accruals, and cash flows when firms have large

book-tax differences. The Accounting Review 80: 137-166. Hanlon, M., K. LaPlante, and T. Shevlin. 2005. Evidence on the information loss of conforming book

income and taxable income. The Journal of Law and Economics 48: 407-442. Hanlon, M., and J. Slemrod. 2009. What does tax aggressiveness signal? Evidence from stock price

reactions to news about tax shelter involvement. Journal of Public Economics, 93 (1-2): 126-141. Harford, J. 1999. Corporate cash reserves and acquisitions. The Journal of Finance 54 (6): 1969-1997.

30

Heckman, J. 1979. Sample selection bias as a specification error. Econometrica 47 (1): 153–162. Hoopes, J. L., D. Mescall, and J. A. Pittman. 2012. Do IRS audits deter corporate tax avoidance? The

Accounting Review 87 (5): 1603-1639. Jensen, M. 1986. The agency costs of free cash flow: Corporate finance and takeovers. American

Economic Review 76: 323-330. Koester, A. 2011. Investor valuation of tax avoidance through uncertain tax positions. Working paper,

Georgetown University. Kormendi, R., and R. Lipe. 1987. Earnings innovations, earnings persistence, and stock returns. Journal

of Business 60: 323–345. Krull, L. 2004. Permanently reinvested foreign earnings, taxes, and earnings management. The

Accounting Review 79 (3): 745-767. Larcker, D. and T. Rusticus. 2010. On the use of instrumental variables in accounting research. Journal of

Accounting and Economics 49 (3): 186-205. Lennox, C., J. Francis, and Z. Wang. 2012. Selection models in accounting research. The Accounting

Review 87 (2): 589-616. Lev, B., and D. Nissim. 2004. The persistence of the accruals anomaly. Contemporary Accounting

Research 23 (1): 193-226. Lev, B., and S. Thiagarajan. 1993. Fundamental information Analysis. Journal of Accounting Research

31 (2): 190-215. Lipe, R. 1986. The information contained in the components of earnings. Journal of Accounting Research

24 (1): 37-63. Lisowsky, P., L. Robinson, and A. Schmidt. 2013. Do publicly disclosed tax reserves tell us about

privately disclosed tax shelter activity? Journal of Accounting Research, Forthcoming. Maddala, G. 1983. Limited-dependent and Qualitative Variables in Econometrics. Cambridge University

Press, Cambridge, UK. Nissim, D., and S. Penman. 2001. Ratio analysis and equity valuation: From research to practice. Review

of Accounting Studies 6: 109 - 154. Nissim, D., and S. Penman. 2003. Financial statement analysis of leverage and how it informs about

profitability and price-to-book ratios. Review of Accounting Studies 8: 531 - 560. Ohlson, J. 1995. Earnings, book values, and dividends in security valuation. Contemporary Accounting

Research 11: 661–687. Ou, J., and S. Penman. 1989. Accounting measurement, P/E ratios and the information content of security

prices. Journal of Accounting Research 27: 111-144. Patatoukos, P. 2012. Customer base concentration: Implications for firm performance and capital

markets. The Accounting Review 87 (2): 363-392.

31

Penman, S. 1992. Financial statement information and the pricing of earnings changes. The Accounting

Review 67 (3): 563-577. Penman, S., and X.J. Zhang. 2006. Modeling sustainable earnings and P/E ratios with financial statement

information. Columbia University, Working Paper. Phillips, J. D., M. Pincus, and S. O. Rego. 2003. Earnings management: New evidence based on deferred

tax expense. The Accounting Review 78 (2): 491-521. Rego, S.O. 2003. Tax-avoidance activities of U.S. multinational corporations. Contemporary Accounting

Research 20 (4): 805-833. Robinson, L., and A. Schmidt. 2013. Firm and investor responses to uncertain tax benefit disclosure

requirements. The Journal of the American Taxation Association 35 (2): 85-120. Soliman, M. 2008. The use of DuPont analysis by market participants. The Accounting Review 83 (3):

823-853. Tucker, J. 2010. Selection bias and econometric remedies in accounting and finance research. Journal of

Accounting Literature 29: 31-57. Watts, R. L., and J. L. Zimmerman. 1983. Agency problems, auditing, and the theory of the firm: Some

evidence. Journal of Law and Economics 26 (3): 613-633. Weber, D. 2009. Do analysts and investors fully appreciate the implications of book-tax differences for

future earnings? Contemporary Accounting Research 26 (4): 1175 – 1206. Wilson, R. 2009. An examination of corporate tax shelter participants. The Accounting Review 84 (3):

969-1000. Wooldridge J.M. 2010. Econometric analysis of cross section and panel data. Second edition. Cambridge,

MA: MIT Press.

32

Appendix A Variable Definitions

Variable

Definition

SETR5 Statutory Tax Rate [five-year ratio of (Pretax Income (PI) * Statutory Federal Tax Rate) / Pretax Income (PI), measured contemporaneously (t–4 to t)]. The Statutory Federal Tax Rate equals 49.2 percent in 1970, 48 percent from 1971-1978, 46 percent from 1979-1986, 39.5 percent in 1987, 34 percent from 1988-1992, and 35 percent from 1993-current.

BETR5 = Book ETR [five-year ratio of Total Tax Expense (TXT) / Pretax Income (PI), measured contemporaneously (t–4 to t)].

TAX_AVD = the industry/year quintile rank of the difference between a firm’s five-year statutory tax rate (SETR5) and five-year book effective tax rate (BETR5)

Main Regression Models

FA = Financial Assets [Cash and Short-term Investments (CHE) + Investments and Advances-other (IVAO)]

OA = Operating Assets [Total Assets (AT) – FA] FO = Financial Obligations [Debt in Current Liabilities (DLC) + Long-term Debt (DLTT) +

Preferred stock (PSTK) – Preferred Treasury Stock (TSTKP) + Preferred Dividends in Arrears (DVPA) + Minority Interest (MIB)]

CSE = Common Equity [Common Equity (CEQ) + Preferred Treasury Stock (TSTKP) – Preferred Dividends in Arrears (DVPA)]

NOA = Net Operating Assets: [FO – FA + CSE + Income Taxes Payable (TXP) + Deferred Taxes and Investment Tax Credit (TXDITC)]

NFO = Net Financial Obligations: [FO – FA] OL = Operating Liabilities: [OA – NOA] IO = Implicit Interest on Operating Liabilities: [The One-year Risk-Free Rate * (Average OL)] COA = Current Operating Assets: [Total Current Assets (ACT) – Cash and Short-term Investments

(CHE)] COL = Current Operating Liabilities: [Total Current Liabilities (LCT) – Debt in Current Liabilities

(DLC)] NCOA = Non-Current Operating Assets: [Total Assets (AT) – Current Assets (ACT) – Investments

and Advances-other (IVAO)] NCOL = Non-Current Operating Liabilities: [Total Liabilities (LT) – Current Liabilities (LCT) –

Long-term Debt (DLTT)] PTROCE = Pretax Return on Common Equity [Pretax Income (PI) + Interest Expense (XINT)) /

Average CSE] PTRNOA = Pretax Return on Net Operating Assets [Pretax Income (PI) + Interest Expense (XINT)) /

Average NOA] PTROOA = Pretax Return on Operating Assets [Pretax Income (PI) + Interest Expense (XINT) +

Implicit Interest on Operating Liabilities) / Average OA] FLEV_FX = Pretax Financial Leverage Effect: [PTROCE – PTRNOA] OLLEV_FX = Pretax Operating Liability Leverage Effect: [PTRNOA – PTROOA] FLEV = Financial Leverage: [Average NFO / Average CSE] OLLEV = Operating Liability Leverage: [Average OL / Average NOA] FL_SPRD = Financial Leverage Spread: [FLEV_FX / FLEV] OL_SPRD = Operating Liability Leverage Spread: [OLLEV_FX / OLLEV] PTNOPM = Pretax Net Operating Profit Margin [Pretax Income (PI) + Interest Expense (XINT)) / Sales

(SALE)] NOAT = Net Operating Asset Turnover: [Sales (SALE) / Average NOA] WCTO = Net Working Capital Turnover: [(COA – COL) / Average NOA] NCOAT = Net Non-current Operating Asset Turnover: [(NCOA – NCOL) / Average NOA]

33

Appendix A (Continued) Other Variables

SIZE = Size [Natural log of Total Assets (AT)] MB = Market-to Book Ratio [Market Value of Equity (CSHO * PRCC_F) / Common Equity

(CEQ)]. ACC_QUAL = Accrual Quality [the standard deviation of the residual from years t–4 to t of a regression

of pretax current accruals on lagged, current, and future pretax cash from operations (OANCF), the change in revenues (SALE), and the gross value of PP&E (PPEGT) (Francis, LaFond, Olsson, and Schipper [2005])].

EARN_VOLATILITY = Pretax Earnings Volatility [standard deviation of Pretax Income (PI) from t–1 to t–3 Cross-Sectional Partitioning Variables FOPS = Indicator variable that equals one if a firm-year includes either non-zero foreign pretax

income (PIFO), non-zero current foreign tax expense (TXFO), or non-zero deferred foreign tax expense (TXDFO), zero otherwise

GGOV = Indicator variable that equals one if the percentage of institutional shareholders (as reported in Thompson-Reuters 13f database) is equal to or above the median of the industry/year institutional shareholders distribution, zero otherwise. The first year of coverage in the Thompson-Reuters 13f database is 1979.

34

TABLE 1 Descriptive Statistics: 1970-20121

Variables Top Quintile of TAX_AVD

(n = 14,938)

Bottom Quintiles of TAX_AVD (n = 60,605) Mean t

Mean SD Q1 Median Q3 Mean SD Q1 Median Q3 Diff Statistic

Panel A: Summary Statistics

PTROCE 0.219 0.220 0.111 0.214 0.320 0.262 0.211 0.158 0.260 0.363 0.043 21.90 PTRNOA 0.161 0.215 0.083 0.137 0.219 0.216 0.226 0.108 0.181 0.286 0.056 27.24 PTROOA 0.132 0.152 0.075 0.120 0.178 0.170 0.157 0.095 0.149 0.224 0.038 26.32

FLEV_FX 0.058 0.189 -0.004 0.050 0.131 0.045 0.186 -0.010 0.048 0.125 -0.013 -7.66 FLEV 0.502 0.930 -0.059 0.353 0.841 0.363 0.811 -0.112 0.228 0.659 -0.138 -18.08

FL_SPRD 0.155 1.628 0.081 0.150 0.248 0.239 9.009 0.104 0.192 0.324 0.084 1.14 OLLEV_FX 0.029 0.103 0.002 0.012 0.036 0.047 0.108 0.008 0.024 0.057 0.018 18.53

OLLEV 0.398 0.839 0.151 0.253 0.421 0.440 0.776 0.185 0.287 0.468 0.042 5.84 OL_SPRD 0.064 0.209 0.014 0.058 0.113 0.101 0.458 0.038 0.088 0.157 0.037 9.58 PTNOPM 0.119 0.173 0.047 0.096 0.173 0.116 0.133 0.057 0.101 0.164 -0.003 -2.40

GM 0.356 0.198 0.222 0.321 0.463 0.358 0.184 0.232 0.327 0.458 0.002 1.10 OTM 0.251 0.170 0.129 0.214 0.336 0.247 0.157 0.134 0.220 0.328 -0.004 -3.01

NOAT 2.080 2.655 0.908 1.566 2.414 2.478 2.550 1.248 1.937 2.857 0.398 16.93 WCTO 5.512 39.669 2.784 5.253 10.036 7.156 37.252 3.339 5.585 10.653 1.644 4.77

NCOAT 3.775 5.614 1.033 2.278 4.353 4.791 6.080 1.559 3.151 5.635 1.016 18.56 TAX_AVD 0.194 0.117 0.095 0.172 0.297 -0.041 0.164 -0.042 -0.007 0.027 -0.235 -165.12

SIZE 5.827 1.983 4.272 5.686 7.315 5.834 2.000 4.318 5.668 7.170 0.007 0.38 MB 2.070 2.114 0.918 1.478 2.464 2.307 2.310 1.075 1.697 2.736 0.237 11.42

ACC_QUAL 0.036 0.030 0.015 0.027 0.045 0.033 0.026 0.015 0.025 0.042 -0.003 -11.37 EARN_VOLATILITY 48.657 200.099 1.869 6.695 28.413 70.570 424.913 1.979 6.880 27.264 21.913 6.14

SETR5 0.402 0.066 0.350 0.350 0.469 0.401 0.067 0.350 0.350 0.468 -0.001 -1.79 BETR5 0.209 0.136 0.086 0.230 0.317 0.442 0.162 0.361 0.409 0.474 0.234 162.70

3

1 This table reports descriptive statistics for the sample of tax avoiding firms (top quintile of TAX_AVD) and firms that are not tax avoiders (bottom quintiles of TAX_AVD) from 1970-2012 for U.S.-based firm-years (excluding utilities and financial institutions) listed on the NYSE, NASDAQ, or AMEX exchanges with common equity in excess of $10M (in 2012 dollars), positive net operating assets, and at least five consecutive years of positive pretax income, cash taxes paid, and positive income tax expense. Panel A reports descriptive statistics (in $ millions) on firm characteristics. We define a firm as a tax avoider if the difference between its five-year statutory tax rate (SETR5) and its five year book ETR (BETR5) is in the top quintile of each industry/year distribution (we form industries using an 18 industry classification system adapted from Barth, Beaver, Hand, and Landsman [1999]). We define the variables in Appendix A.

35

TABLE 2

Tax Avoidance and Firm Performance: Return on Equity, Return on Net Operating Assets, and Return on Operating Assets

PTROCEt+1 PTROCEt+3 PTROCEt+5 PTRNOAt+1 PTRNOAt+3 PTRNOAt+5 PTROOAt+1 PTROOAt+3 PTROOAt+5 [1] [2] [3] [4] [5] [6] [7] [8] [9] TAX_AVD -0.009*** -0.007*** -0.007*** -0.010*** -0.011*** -0.011*** -0.006*** -0.006*** -0.007*** (-11.58) (-6.99) (-6.48) (-10.69) (-9.52) (-8.59) (-8.67) (-7.35) (-7.33) SIZE 0.012*** 0.016*** 0.018*** -0.007*** -0.003** -0.001 -0.008*** -0.005*** -0.004*** (15.19) (15.99) (15.18) (-6.68) (-2.18) (-0.94) (-10.05) (-4.96) (-3.69) MB 0.039*** 0.026*** 0.021*** 0.037*** 0.026*** 0.022*** 0.026*** 0.018*** 0.016*** (35.61) (21.45) (17.59) (26.55) (18.87) (15.53) (23.06) (16.32) (15.05) ACC_QUAL -0.255*** -0.451*** -0.463*** -0.243*** -0.472*** -0.478*** -0.284*** -0.439*** -0.417*** (-4.71) (-6.67) (-6.41) (-3.61) (-5.93) (-5.49) (-5.89) (-7.66) (-6.95) EARN_VOLATILITY 0.000*** 0.000** 0.000 0.000*** 0.000*** 0.000*** 0.000*** 0.000*** 0.000*** (4.62) (2.44) (1.53) (5.78) (5.81) (5.88) (5.15) (4.87) (5.48) Industry/Year FE YES YES YES YES YES YES YES YES YES N 75543 65701 56948 75543 65179 56263 75543 65680 56928 Adj R2 0.2469 0.1568 0.1339 0.1693 0.0924 0.0784 0.1857 0.1212 0.1142 1 This table presents the results of estimating OLS regressions of future pretax return on common equity (PTROCE), return on net operating assets (PTRNOA), and return on operating assets (PTROOA), on the

quintile rank of a variable for firm-specific tax avoidance (TAX_AVD) and control variables from 1970-2012 for U.S.-based firm-years (excluding utilities and financial institutions) listed on the NYSE, NASDAQ, or AMEX exchanges with common equity in excess of $10M (in 2012 dollars), positive net operating assets, and at least five consecutive years of positive pretax income, cash taxes paid, and positive income tax expense. We include industry (measured using an 18 industry classification system adapted from Barth, Beaver, Hand, and Landsman [1999]) and year indicator variables in all regressions. In each panel, we report robust, clustered (by firm id) t-statistics in parentheses below fully standardized coefficients. The symbols ***, **, and * denote significance at the 0.01, 0.05, and 0.10 (two-tailed) levels, respectively. We define the variables in Appendix A

36

TABLE 3

Tax Avoidance and Firm Performance: Operating Profit Margin, Operating Asset Turnover, the Financial Leverage Effect and the Operating Liability Leverage Effect1

Panel A: The Effect of Operating Profit Margins and Operating Asset Turnover

PTNOPMt+1 PTNOPMt+3 PTNOPMt+5 NOATt+1 NOATt+3 NOATt+5 [1] [2] [3] [4] [5] [6]

TAX_AVD 0.004*** 0.003*** 0.002*** -0.125*** -0.114*** -0.118*** (6.50) (4.01) (2.77) (-9.37) (-8.02) (-7.49) SIZE 0.003*** 0.004*** 0.005*** -0.114*** -0.086*** -0.077*** (3.78) (5.20) (5.12) (-7.78) (-5.73) (-4.70) MB 0.014*** 0.010*** 0.010*** 0.118*** 0.051*** 0.023 (18.86) (12.59) (12.54) (10.27) (4.15) (1.54) ACC_QUAL -0.439*** -0.566*** -0.502*** 9.597*** 8.071*** 6.637*** (-10.80) (-11.16) (-9.00) (10.40) (8.40) (6.22) EARN_VOLATILITY 0.000 0.000 0.000 0.000*** 0.000*** 0.000*** (1.48) (1.27) (1.01) (3.12) (4.57) (3.93)

Industry/Year FE YES YES YES YES YES YES N 75524 65632 56882 75543 65141 56229 Adj R2 0.1566 0.1245 0.1163 0.1766 0.1512 0.1366 Panel B: The Effect of Financial Leverage and Operating Liability Leverage

FLEV_FXt+1 FLEV_FXt+3 FLEV_FXt+5 OLLEV_FXt+1 OLLEV_FXt+3 OLLEV_FXt+5

[1] [2] [3] [4] [5] [6]

TAX_AVD 0.001 0.003*** 0.003*** -0.004*** -0.004*** -0.004*** (1.07) (3.61) (3.05) (-8.76) (-7.57) (-5.89) SIZE 0.019*** 0.019*** 0.019*** 0.001*** 0.002*** 0.003*** (20.29) (19.35) (18.07) (2.68) (3.91) (3.99) MB 0.001 -0.000 -0.001 0.011*** 0.007*** 0.006*** (1.18) (-0.03) (-0.96) (19.73) (10.88) (7.70) ACC_QUAL -0.012 0.023 0.019 0.041 -0.032 -0.059 (-0.21) (0.37) (0.28) (1.23) (-0.71) (-1.16) EARN_VOLATILITY -0.000*** -0.000*** -0.000*** 0.000*** 0.000*** 0.000*** (-3.37) (-3.32) (-3.23) (5.05) (3.14) (2.97)

Industry/Year FE YES YES YES YES YES YES N 75543 65179 56263 75543 65179 56263 Adj R2 0.1474 0.1104 0.0945 0.0922 0.0350 0.0278

1 This table presents the results of estimating OLS regressions of the future pretax financial statement analysis components on the quintile rank of a variable for firm-specific tax avoidance (TAX_AVD) and control variables from 1970-2012 for U.S.-based firm-years (excluding utilities and financial institutions) listed on the NYSE, NASDAQ, or AMEX exchanges with common equity in excess of $10M (in 2012 dollars), positive net operating assets, and at least five consecutive years of positive pretax income, cash taxes paid, and positive income tax expense. We include industry (measured using an 18 industry classification system adapted from Barth, Beaver, Hand, and Landsman [1999]) and year indicator variables in all regressions. In each panel, we report robust, clustered (by firm id) standard errors in parentheses. The symbols ***, **, and * denote significance at the 0.01, 0.05, and 0.10 (two-tailed) levels, respectively. We define the variables in Appendix A

37

TABLE 4 Tax Avoidance and Firm Performance: Multinational vs. Domestic Firms1

Panel A: Pretax Return on Operating Assets (PTROOA)

PTROOAt+1 PTROOAt+1 PTROOAt+3 PTROOAt+3 PTROOAt+5 PTROOAt+5

FOPS = 1 FOPS = 0 FOPS = 1 FOPS = 0 FOPS = 1 FOPS = 0 [1] [2] [3] [4] [5] [6]

TAX_AVD -0.004*** -0.007*** -0.006*** -0.006*** -0.007*** -0.006*** (-4.55) (-8.07) (-6.11) (-5.42) (-6.57) (-5.07)

2 [(FOPS = 1) – (FOPS = 0)] 20.64*** 0.03 1.50 N 33041 42502 28290 37390 24241 32687 Adj R2 0.1924 0.1900 0.1340 0.1204 0.1286 0.1128

Panel B: The Operating Liability Leverage Effect (OLLEV_FX)

OLLEV_FXt+1 OLLEV_FXt+1 OLLEV_FXt+3 OLLEV_FXt+3 OLLEV_FXt+5 OLLEV_FXt+5


TAX_AVD -0.004*** -0.004*** -0.005*** -0.004*** -0.005*** -0.003*** (-5.02) (-7.89) (-5.15) (-5.72) (-4.38) (-4.01)

2 [(FOPS = 1) – (FOPS = 0)] 0.00 2.03 3.82** N 33041 42502 28094 37085 24005 32258 Adj R2 0.1135 0.0681 0.0550 0.0205 0.0426 0.0168

Panel C: Pretax Return on Net Operating Assets (PTRNOA)

PTRNOAt+1 PTRNOAt+1 PTRNOAt+3 PTRNOAt+3 PTRNOAt+5 PTRNOAt+5


TAX_AVD -0.008*** -0.012*** -0.011*** -0.010*** -0.012*** -0.009*** (-5.92) (-9.39) (-7.01) (-7.03) (-6.82) (-5.98)

2 [(FOPS = 1) – (FOPS = 0)] 9.56** 0.73 3.81** N 33041 42502 28094 37085 24005 32258 Adj R2 0.1773 0.1656 0.1038 0.0864 0.0887 0.0749

Panel D: The Financial Leverage Effect (FLEV_FX)

FLEV_FXt+1 FLEV_FXt+1 FLEV_FXt+3 FLEV_FXt+3 FLEV_FXt+5 FLEV_FXt+5


TAX_AVD -0.002* 0.003*** -0.000 0.006*** -0.000 0.005*** (-1.95) (2.73) (-0.15) (4.75) (-0.03) (3.90)

2 [(FOPS = 1) – (FOPS = 0)] 35.87*** 28.55*** 16.83*** N 33041 42502 28094 37085 24005 32258 Adj R2 0.1449 0.1551 0.1047 0.1220 0.0921 0.1062

Panel E: Pretax Return on Common Equity (PTROCE)

PTROCEt+1 PTROCEt+1 PTROCEt+3 PTROCEt+3 PTROCEt+5 PTROCEt+5


TAX_AVD -0.010*** -0.008*** -0.011*** -0.004*** -0.012*** -0.004*** (-9.33) (-8.08) (-8.01) (-3.09) (-7.31) (-2.86)

2 [(FOPS = 1) – (FOPS = 0)] 3.94** 31.91*** 30.69*** N 33041 42502 28295 37406 24246 32702 Adj R2 0.2984 0.2106 0.1892 0.1399 0.1670 0.1199

38

TABLE 4 (Continued)

Panel F: Working Capital Turnover (WCTO)

PTNOPMt+1 PTNOPMt+1 PTNOPMt+3 PTNOPMt+3 PTNOPMt+5 PTNOPMt+5


TAX_AVD 0.002*** 0.005*** 0.000 0.005*** -0.001 0.004*** (3.25) (5.43) (0.07) (4.21) (-1.50) (3.65)

2 [(FOPS = 1) – (FOPS = 0)] 14.48*** 27.05*** 37.85*** N 33036 42488 28272 37360 24225 32657 Adj R2 0.1415 0.1731 0.1092 0.1427 0.1074 0.1324

Panel G: Net Operating Asset Turnover (NOAT)

NOATt+1 NOATt+1 NOATt+3 NOATt+3 NOATt+5 NOATt+5


TAX_AVD -0.110*** -0.132*** -0.116*** -0.107*** -0.109*** -0.120*** (-5.92) (-7.41) (-5.95) (-5.58) (-4.93) (-5.73)

2 [(FOPS = 1) – (FOPS = 0)] 2.73* 0.36 0.46 N 33041 42502 28081 37060 23994 32235 Adj R2 0.1321 0.2113 0.1039 0.1894 0.0910 0.1755

1 This table presents the results of estimating OLS regressions of the future pretax financial statement analysis components on the quintile rank of

a variable for firm-specific tax avoidance (TAX_AVD) and control variables from 1970-2012 for U.S.-based firm-years (excluding utilities and financial institutions) listed on the NYSE, NASDAQ, or AMEX exchanges with common equity in excess of $10M (in 2012 dollars), positive net operating assets, and at least five consecutive years of positive pretax income, cash taxes paid, and positive income tax expense. We include industry (measured using an 18 industry classification system adapted from Barth, Beaver, Hand, and Landsman [1999]) and year indicator variables in all regressions. In each panel, we report robust, clustered (by firm id) standard errors in parentheses. We classify a firm-year as multinational (FOPS = 1) if foreign tax expense (TXFO), foreign deferred tax expense (TXDFO), or foreign pretax income (PIFO) is non-zero, domestic (FOPS = 0) otherwise. The symbols ***, **, and * denote significance at the 0.01, 0.05, and 0.10 (two-tailed) levels, respectively. We define the variables in Appendix A.

39

TABLE 5 Tax Avoidance and Firm Performance: Good vs. Poor Governance 1

Panel A: Pretax Return on Operating Assets (PTROOA)

PTROOAt+1 PTROOAt+1 PTROOAt+3 PTROOAt+3 PTROOAt+5 PTROOAt+5

GGOV = 1 GGOV = 0 GGOV = 1 GGOV = 0 GGOV = 1 GGOV = 0 [1] [2] [3] [4] [5] [6]

TAX_AVD -0.002** -0.005*** -0.003** -0.006*** -0.003** -0.007*** (-2.34) (-5.35) (-2.49) (-4.96) (-2.30) (-4.64)

2 [(GGOV = 1) – (GGOV = 0)] 9.36*** 7.99*** 8.73*** N 25618 25333 22125 21372 19132 18028 Adj R2 0.2003 0.1602 0.1237 0.0898 0.1157 0.0732

Panel B: The Operating Liability Leverage Effect (OLLEV_FX)

OLLEV_FXt+1 OLLEV_FXt+1 OLLEV_FXt+3 OLLEV_FXt+3 OLLEV_FXt+5 OLLEV_FXt+5


TAX_AVD -0.003*** -0.003*** -0.003*** -0.004*** -0.003** -0.002** (-3.08) (-4.13) (-3.03) (-3.63) (-2.53) (-2.10)

2 [(GGOV = 1) – (GGOV = 0)] 0.28 0.14 0.36

N 25618 25333 21907 21192 18855 17772 Adj R2 0.1316 0.0671 0.0473 0.0278 0.0334 0.0214

Panel C: Pretax Return on Net Operating Assets (PTRNOA)

PTRNOAt+1 PTRNOAt+1 PTRNOAt+3 PTRNOAt+3 PTRNOAt+5 PTRNOAt+5


TAX_AVD -0.005*** -0.008*** -0.006*** -0.010*** -0.006*** -0.009*** (-3.24) (-5.66) (-3.29) (-5.54) (-2.94) (-4.47)

2 [(GGOV = 1) – (GGOV = 0)] 5.51** 5.11** 2.71* N 25618 25333 21907 21192 18855 17772 Adj R2 0.1978 0.1410 0.1023 0.0774 0.0850 0.0611

Panel D: The Financial Leverage Effect (FLEV_FX)

FLEV_FXt+1 FLEV_FXt+1 FLEV_FXt+3 FLEV_FXt+3 FLEV_FXt+5 FLEV_FXt+5


TAX_AVD -0.002* 0.001 0.000 0.004*** 0.001 0.002 (-1.71) (0.70) (0.07) (2.62) (0.35) (1.04)

2 [(GGOV = 1) – (GGOV = 0)] 8.15*** 7.00*** 0.43 N 25618 25333 21907 21192 18855 17772 Adj R2 0.2002 0.1416 0.1400 0.1035 0.1187 0.0846

Panel E: Pretax Return on Common Equity (PTROCE)

PTROCEt+1 PTROCEt+1 PTROCEt+3 PTROCEt+3 PTROCEt+5 PTROCEt+5


TAX_AVD -0.007*** -0.008*** -0.006*** -0.006*** -0.006*** -0.007*** (-5.98) (-6.16) (-3.64) (-3.74) (-2.80) (-4.03)

2 [(GGOV = 1) – (GGOV = 0)] 0.01 0.01 0.71 N 25618 25333 22138 21374 19140 18030 Adj R2 0.2893 0.1987 0.1695 0.1269 0.1455 0.1069

40

TABLE 5 (Continued)

Panel F: Pretax Net Operating Profit Margin (PTNOPM)

PTNOPMt+1 PTNOPMt+1 PTNOPMt+3 PTNOPMt+3 PTNOPMt+5 PTNOPMt+5


TAX_AVD 0.004*** 0.001 0.003*** -0.001 0.002** -0.000 (4.31) (1.01) (2.91) (-0.96) (2.03) (-0.29)

2 [(GGOV = 1) – (GGOV = 0)] 10.35*** 13.68*** 4.73** N 25618 25321 22133 21347 19130 18004 Adj R2 0.1555 0.1252 0.1228 0.0992 0.1224 0.0906

Panel G: Net Operating Asset Turnover (NOAT)

NOATt+1 NOATt+1 NOATt+3 NOATt+3 NOATt+5 NOATt+5


TAX_AVD -0.085*** -0.099*** -0.081*** -0.093*** -0.088*** -0.092*** (-3.80) (-4.29) (-3.39) (-3.61) (-3.27) (-3.20)

2 [(GGOV = 1) – (GGOV = 0)] 0.66 0.33 0.03 N 25618 25333 21903 21183 18848 17764 Adj R2 0.1858 0.1558 0.1418 0.1287 0.1221 0.1117 1 This table presents the results of estimating OLS regressions of the future pretax financial statement analysis components on the quintile rank of

a variable for firm-specific tax avoidance (TAX_AVD) and control variables from 1970-2012 for U.S.-based firm-years (excluding utilities and financial institutions) listed on the NYSE, NASDAQ, or AMEX exchanges with common equity in excess of $10M (in 2012 dollars), positive net operating assets, and at least five consecutive years of positive pretax income, cash taxes paid, and positive income tax expense. We include industry (measured using an 18 industry classification system adapted from Barth, Beaver, Hand, and Landsman [1999]) and year indicator variables in all regressions. In each panel, we report robust, clustered (by firm id) standard errors in parentheses. We define governance consistent with Desai and Dharmapala [2006], based on the percentage of outstanding common stock owned by institutional shareholders. We define observations equal to or above the median of the industry/year institutional ownership distribution as having high (low) governance [GGOV = 1 (GGOV = 0)]. The symbols ***, **, and * denote significance at the 0.01, 0.05, and 0.10 (two-tailed) levels, respectively. We define the variables in Appendix A.

Documents

Tax Avoidance and DuPont Measures of Future Performance...tax incomes (a proxy for tax avoidance) provide information to market participants about earnings 6 Pretax return on operating