# The economic effects of financial derivatives on corporate tax avoidance ☆

- 57 trang
- file .pdf

Author's Accepted Manuscript

The economic effects of financial derivatives

on corporate tax avoidance

Michael P. Donohoe

www.elsevier.com/locate/jae

PII: S0165-4101(14)00072-X

DOI: http://dx.doi.org/10.1016/j.jacceco.2014.11.001

Reference: JAE1042

To appear in: Journal of Accounting and Economics

Received date: 28 August 2013

Revised date: 3 November 2014

Accepted date: 11 November 2014

Cite this article as: Michael P. Donohoe, The economic effects of financial

derivatives on corporate tax avoidance, Journal of Accounting and Economics, http:

//dx.doi.org/10.1016/j.jacceco.2014.11.001

This is a PDF file of an unedited manuscript that has been accepted for

publication. As a service to our customers we are providing this early version of

the manuscript. The manuscript will undergo copyediting, typesetting, and

review of the resulting galley proof before it is published in its final citable form.

Please note that during the production process errors may be discovered which

could affect the content, and all legal disclaimers that apply to the journal

pertain.

The economic effects of financial derivatives on corporate tax avoidance

Michael P. Donohoe

Assistant Professor of Accountancy

University of Illinois at Urbana-Champaign

1206 S. Sixth Street, MC-706

Champaign, IL 61820

[email protected]

October 2014

Abstract

This study estimates the corporate tax savings from financial derivatives. I document a 3.6 and

4.4 percentage point reduction in three-year current and cash effective tax rates (ETRs),

respectively, after a firm initiates a derivatives program. The decline in cash ETR equates to

$10.69 million in tax savings for the average firm and $4.0 billion for the entire sample of 375

new derivatives users. Of these amounts, $8.75 million and $3.3 billion, respectively, are

incremental to tax savings that theory suggests are a byproduct of risk management. Collectively,

these findings provide economic insight into the prevalence of derivatives-based tax avoidance.

Keywords: financial instruments, derivatives, tax avoidance, effective tax rate

Data Availability: Data are available from public sources identified in the article.

JEL Classification: G32; H25; M40

This study is based on my dissertation completed at the University of Florida, which won the 2012 American

Taxation Association/PricewaterhouseCoopers Outstanding Tax Dissertation Award. I am grateful for the

guidance of my committee: Jon Hamilton, Robert Knechel, Sandra Kramer, and especially my chairman, Gary

McGill. I also appreciate helpful comments from Ross Watts (Editor), an anonymous reviewer, Rashad Abdel-

khalik, Bipin Ajinkya, Andrew Bauer, Paul Beck, Jenny Brown, Raluca Chiorean, Will Ciconte, Dan Collins,

Paul Demeré , Joel Demski, Victoria Dickinson, Brian Gale, Carlos Jimenez, Allison Koester, Andy Leone,

Pete Lisowsky, Michael Mayberry, Sean McGuire, Lil Mills, Ed Outslay, Mark Peecher, Sonja Rego, Casey

Schwab, Terry Shevlin, Bill Snyder, Theo Sougiannis, Jake Thornock, Jaron Wilde, Ryan Wilson, and

workshop participants at the Arizona State University, University of Florida, Florida State University,

Southern Methodist University, Temple University, University of Illinois at Urbana-Champaign, University of

Iowa, University of Kansas, University of Miami, and University of Mississippi. Special thanks to Scott

Dyreng for providing tax haven data, Stephen Brown for data collection programming assistance, and Hye Sun

Chang and Shannon Woods for research assistance. I gratefully acknowledge financial support from the

PricewaterhouseCoopers Faculty Fellowship.

The economic effects of financial derivatives on corporate tax avoidance

1. Introduction

Financial derivatives are a leading source of corporate tax noncompliance (Organisation

for Economic Co-Operation [OECD] 2011; U.S. Government Accountability Office [GAO]

2011). Experts claim the legal patchwork of derivatives taxation encourages the development of

tax planning strategies (Raskolnikov 2011), while the Internal Revenue Service (IRS) concedes it

is “falling farther and farther behind” financial innovation (Raghavan 2007). Although it is clear

from government reports (U.S. Treasury 1999; GAO 2011; U.S. Congress Joint Committee on

Taxation [JCT] 2011), anecdotes (Warner 2008), and academic studies (McDonald 2004; Warren

2004) that companies can avoid tax with derivatives, no study directly examines the economic

tax effects of these financial instruments (Shevlin 1999, 2007). I fill this void by answering the

question: How economically significant are the corporate tax savings from derivatives?1

Nearly two-thirds of non-financial firms in the U.S. participate in the $710 trillion

derivatives market (Bank for International Settlements 2013). In general, derivatives help firms

manage risks relating to interest rates, foreign currency exchange rates, and commodity prices.

However, as a notoriously complex and ambiguous area of tax law, derivatives also promote tax

avoidance. Many groups, including academics (Schizer 2000; Warren 2004), lawmakers (Baucus

2011; Hatch 2011), and regulators (GAO 2011, 2012; JCT 2011; OECD 2011), characterize

derivatives as a significant threat to global tax revenue.

Derivatives facilitate two types of tax avoidance. The first type is a benign byproduct of

risk management as theory suggests that hedging (reducing) risk can lower taxes by reducing the

volatility of taxable income when firm-level tax functions are convex (Smith and Stulz 1985).

1

I define corporate tax avoidance as the reduction of explicit taxes (Hanlon and Heitzman 2010). Tax avoidance

does not necessarily imply that firms are engaging in improper behavior as managing tax costs is an appropriate

component of a firm‟s long-term strategy (Atwood et al. 2012). Section 2 describes derivatives-based tax avoidance.

1

The second type results from both the general and aggressive means in which firms navigate the

extensive ambiguity in derivatives taxation (e.g., Schizer 2000, 2004). In particular, the tax code

takes a categorical approach to derivatives taxation by prescribing rules for only a few types of

instruments. However, because derivatives can replicate the economic profile of virtually any

underlying asset and also create innovative economic profiles, these sophisticated instruments do

not fit neatly into a categorical tax system. Instead, similar (and even identical) economic

positions are taxed differently depending on a versatile transactional form. A firm with a specific

economic objective can therefore choose a transactional form that offers an optimal tax outcome

in terms of the timing, character, and source of related gains/losses. Many of these choices are

perfectly legal, while others aggressively push the envelope of tax law (GAO 2011; JCT 2011).

There are no overall tax savings estimates for derivatives; however, prior research

provides an important first step by evaluating the tax byproducts of risk management. In

particular, Graham and Smith (1999) use a simulation to estimate the tax savings that result from

any type of reduction in taxable income volatility, while Graham and Rogers (2002) gauge the

debt-related tax benefits of lower average volatility. I extend this research by directly estimating

the tax savings attributable to derivatives, both overall and incremental to reductions in volatility.

These estimates are important because, while lawmakers are calling for a complete reform of

derivatives taxation (Baucus 2011; Hatch 2011), “our knowledge of [tax] revenue losses from

[derivatives] is largely anecdotal, wholly unsystematic, and woefully incomplete” (Raskolnikov

2011, 1). Moreover, not all firms that use derivatives necessarily engage in derivatives-based tax

avoidance. Surveys indicate that risk management is a primary objective among derivatives users

(Bodnar et al. 2003), and other forms of tax avoidance likely produce comparable tax savings

2

with far less complexity and scrutiny from tax authorities (JCT 2011). Thus, whether the tax

savings of derivatives are economically significant is an open empirical question.

I use both levels and difference-in-differences designs to investigate the economic tax

effects of derivatives. The levels design examines whether firms that use derivatives on an

ongoing basis indeed avoid more tax than firms that do not use derivatives. The difference-in-

differences design examines the tax effects of derivatives initiation and provides a focused

setting in which to estimate the tax savings of derivatives. I measure tax avoidance with three

different forward-looking effective tax rates (ETRs) estimated over a three-year horizon (t to

t+2), where lower ETRs imply higher levels of tax avoidance. Corroborating evidence across

both designs allows for stronger inferences about the link between derivatives and tax avoidance.

The levels analysis reveals that, on average, derivatives users avoid more tax than non-

users. Specifically, after controlling for endogeneity and factors that explain corporate tax

avoidance, I find that derivatives users have three-year current and cash ETRs that are 0.9

percentage points lower than those of non-users. These amounts are economically significant and

indicate that current and cash ETRs for derivatives users are nearly 3.3 percent smaller than

those of non-users. However, I find no difference in book ETRs (i.e., total tax expense reported

under ASC 740, Income Taxes) among derivatives users and non-users, suggesting that the tax

savings of derivatives largely result from tax-timing opportunities.

The difference-in-differences analysis reveals economically significant tax savings for

firms that begin using derivatives in the sample period. Relative to a propensity score matched

control sample of non-users, I find that new derivatives users realize a 3.6 and 4.4 percentage

point reduction in three-year current and cash ETRs, respectively, after derivatives initiation. The

decline in cash ETR equates to $10.69 million in cash tax savings over a three-year period for the

3

average firm and roughly $4.0 billion in savings for the entire sample of 375 new derivatives

users. These effects are not entirely a byproduct of hedging risk as new derivatives users without

a convex tax function or a reduction in taxable income volatility after initiation realize more tax

savings than new derivatives users with these features. Specifically, the incremental decline in

cash ETR equates to an average savings of $8.75 million for each new user and $3.3 billion in

aggregate. Overall, these estimates are comparable to those for other sophisticated tax planning

strategies, including some tax shelters (Wilson 2009; Brown 2011).

Finally, I mitigate alternative explanations by documenting that the economic tax effects

of derivatives are not driven by (1) reductions in earnings or cash flow volatility; (2) changes in

risk management incentives at the time of derivatives initiation; (3) effective management of risk

exposures; or (4) economies of scale. Instead, I find that the tax savings of derivatives increase

with tax aggressiveness and the magnitude of derivatives positions at initiation. Collectively,

these findings provide insight into the economic incentives (i.e., tax savings) that drive the

growing use of derivatives-based tax planning strategies, and help explain why some lawmakers

aim to weaken their “potential for mischief” (Baucus 2011, 1).

This study contributes to growing literatures on derivatives (Aretz and Bartram 2010) and

corporate tax avoidance (Hanlon and Heitzman 2010). In particular, prior research evaluates the

tax incentives to hedge risk (Graham and Smith 1999; Graham and Rogers 2002), illustrates the

tax benefits of options and forwards (Titman 1985; McDonald 2004; Warren 2004), describes

why derivatives are useful for corporate tax avoidance (Donohoe 2014a), and develops a profile

of derivatives-based tax avoiders (Donohoe 2014b). By providing tax savings estimates, I move

this literature beyond the notion that firms can use derivatives to avoid tax. Although tax

avoidance may not always be the primary reason firms use derivatives (e.g., relative to risk

4

management), they generate significant tax savings nonetheless. I also extend research on

corporate tax shelters (Shevlin 2002; Graham and Tucker 2006; Wilson 2009; Lisowsky 2010;

Brown 2011; Donohoe et al. 2014; McGuire et al. 2014) by documenting the economic effects of

financial instruments often used to generate tax benefits in aggressive tax planning strategies.

Finally, this study answers calls for research on the role of financial instruments in corporate tax

avoidance (Shevlin 1999, 2007; Hanlon and Heitzman 2010; Raskolnikov 2011), complements

the law literature on the taxation of financial products (Schizer 2000, 2004), and provides insight

into why lawmakers are resolute to reform derivatives taxation (Baucus 2011; Hatch 2011).

Section 2 develops my hypothesis. I discuss data, research design and results in Sections

3, 4 and 5, respectively. Section 6 discusses additional tests, and Section 7 concludes.

2. Background and hypothesis development

2.1 Derivatives-based tax avoidance

A derivative is a contract or security deriving its value based on its relation to something

else, commonly referred to as the “underlying” (Stulz 2004). The underlying is often another

financial instrument or economic good, but can be almost anything.2 For example, the values of

some derivatives are based on familiar stock indices, the heat index in Florida, and that of other

derivatives. Many companies use derivatives to manage interest rate, foreign exchange rate, and

commodity price risks (Bodnar et al. 2003). However, an increasingly common motive for

derivatives usage is corporate tax avoidance (GAO 2011); that is, the reduction of explicit taxes.3

2

Derivatives generally fall into three groups: (1) options; (2) futures and forwards; and (3) swaps. Options involve

the right, but not the obligation, to buy or sell an underlying at a set price within a specified period. A futures or

forward contract involves an obligation to exchange an underlying at a future date for a specific price, and swaps are

agreements to exchange a stream of payments based on an underlying over a predefined period.

3

Explicit taxes are taxes paid to tax authorities whereas implicit taxes reflect the lower before-tax return on tax-

favored assets relative to tax-disfavored assets (Shackelford and Shevlin 2001; Scholes et al. 2015).

5

Tax avoidance represents a continuum of tax planning strategies, where relatively benign

strategies lie at one end and extremely aggressive or illegal strategies (e.g., some tax shelters) lie

at the opposing end (Lisowsky et al. 2013). Derivatives facilitate tax planning strategies along

the entire continuum. In particular, theory suggests some tax avoidance is a benign byproduct of

risk management (Smith and Stulz 1985), while the law literature describes both the general and

aggressive means in which firms navigate the extensive ambiguity in derivatives taxation (e.g.,

Schizer 2000, 2004). Figure 1 incorporates these elements and illustrates my view of the tax

avoidance continuum with respect to derivatives.

The far left-end of the continuum, Byproduct, reflects tax avoidance that is an ancillary

effect of managing risk. Under current tax rules, effective (i.e., successful) hedging reduces the

volatility of taxable income (Keyes 2008), which could result in lower taxes for two reasons.4

First, in a progressive tax system, expected taxes are a convex function of taxable income, where

firms facing an increasing marginal tax rate can reduce taxes by smoothing income (Smith and

Stulz 1985).5 Second, smooth income can increase debt capacity, potentially increasing tax

deductible interest expenditures (Stulz 1996; Leland 1998). Prior empirical tests of these theories

(discussed later) suggest that hedging risk may indeed result in some tax savings (Graham and

Smith 1999; Graham and Rogers 2002). I view this type of tax savings as Benign ([A] in Figure

1) as it is attributable to less volatile taxable income rather than overt attempts to reduce taxes.

4

For tax reporting purposes, a hedge is effective if it actually reduces exposures to economic risk (Yanchisin and

Ricks 2006). In such cases, changes in the value of both the derivative and underlying are recognized in taxable

income together (i.e., offset). Gains and losses on speculative, ineffective, or straddle positions are recognized in

taxable income as they occur (i.e., no offset). A straddle is an offsetting position (not designated or qualifying as a

hedge) against actively traded property substantially reducing risk of loss. See Donohoe (2014a) for details.

5

Graham and Smith (1999, 2243) provide the following example: “Assuming two equally likely outcomes–low

taxable income with no associated tax liability or high taxable income and a $1 million tax liability–the firm‟s

expected tax liability is $500,000. However, if the firm could hedge and completely eliminate the volatility of

taxable income, both its [expected] taxable income and tax bill would be zero.”

6

The remainder of the continuum, Transactional, reflects tax avoidance that is attributable

to three sources of ambiguity in the taxation of derivatives transactions (Weisbach 2005). First,

derivatives taxation is inconsistent in that not all economically comparable transactions receive

the same tax treatment. In essence, existing statutes and case law create a “cubbyhole” system,

where tax treatments are prescribed for only a few types of derivatives. A firm determines the tax

treatment for a derivative by using its broad transactional features to identify the appropriate

cubbyhole (see Figure 2). However, through financial engineering (e.g., put-call parity),

derivatives can replicate the economic profile of virtually any underlying asset, including other

derivatives (Stoll 1969; Merton 1973; Scholes 1998).6 Consequently, similar (and even identical)

economic positions are taxed differently depending on a versatile transactional form. For a given

economic position, a firm can often choose a transactional form that provides an optimal tax

outcome in terms of the timing, character, and source of related gains/losses (JCT 2011).

Second, because the tax law response to financial innovation has been “reactive and

particularized” (Warren 2004, 913), the “correct” tax treatment for many sophisticated, yet

common, derivatives (e.g., swaptions) is indeterminable. In particular, novel transactions and

those with unique economic profiles rarely fit into preexisting cubbyholes, while “compound”

instruments often fit into multiple cubbyholes. Such uncertainty translates into discretion and

variation in the tax reporting of these instruments (JCT 2011).

Third, ambiguity arises from the asymmetric tax treatment of opposing sides to the same

transaction. In addition to cubbyholes, derivatives taxation depends on taxpayer characteristics,

such as motive (hedge or speculate), entity form (corporate or pass-through), jurisdiction (foreign

or domestic), and status (securities dealer). Thus, it is possible for a firm to make transactional

6

Financial engineering refers to the construction of asset portfolios that have precise technical characteristics,

particularly when those characteristics are not available in an existing exchange-traded product (Strong 2005, 346).

7

form and/or tax reporting choices without obligating counterparties to make similar choices. This

transactional autonomy provides added flexibility in tax planning.7 In sum, while smooth taxable

income can lead to Byproduct tax avoidance, Transactional tax avoidance results from the ability

to strategically coordinate the timing, character, and source of taxable income/deduction due to

the extensive ambiguity in derivatives taxation.

Whether Transactional tax avoidance is controversial or not depends on the degree to

which a firm exploits the ambiguity in derivatives taxation. As noted, the tax treatment of

derivatives is based on transactional form rather than economic form. Because transactions with

the same or similar economic characteristics can be treated differently from one another for tax

purposes, a firm with a specific economic objective can choose a transactional form that provides

an optimal tax outcome (JCT 2011).8 When transactional form is chosen from a set of legally

permitted alternatives after weighing tax and non-tax factors, Transactional tax avoidance is not

likely to be controversial. I view the resulting tax savings as General ([B] in Figure 1).

However, controversy can arise when the choice of transactional form and/or its related

tax reporting is aggressive, which is defined as having “weak legal support” (Lisowsky et al.

2013, 590) that “pushes the envelope of tax law” (Hanlon and Heitzman 2010, 137). For

example, a firm can use derivatives to design and operate tax shelters (U.S. Treasury 1999; GAO

2011, 2012). Along these lines, derivatives can serve as either a supportive element within a

larger arrangement or the central element around which a tax shelter is designed (see Donohoe

[2014a] for two examples). By doing so, derivatives not only “turbo charge” tax shelters

7

For example, the fact that one party to a transaction designates a derivative as a “hedge” (i.e., intending to reduce

economic risk) for tax purposes does not necessarily compel a counterparty to follow suit. Symmetry is an important

policy issue in the taxation of financial products because a tax system can better police itself when both sides to a

transaction face equal and opposite incentives (Raskolnikov 2011).

8

That is, by choosing to hold one type of instrument over another, a taxpayer can obtain a desired (1) timing of

income/expense recognition; (2) characterization of income/loss as capital or ordinary; or (3) source of income/

expense as foreign or domestic (JCT 2011). Section 2.2 provides a brief illustration of timing.

8

(Sheppard 1999, 132), but make them especially difficult for tax authorities to detect and contest

(GAO 2012). I view this type of resulting tax savings as Aggressive ([C] in Figure 1).

2.2 A brief illustration of derivatives-based tax avoidance

Although derivatives can be extraordinarily complex, mundane instruments generate all

three types of tax savings. To illustrate, consider the most common derivative: a generic (“plain

vanilla”) interest rate swap in which fixed interest rate payments are exchanged for floating rate

payments. By transforming fixed rate debt into variable rate debt (and vice versa), an interest rate

swap hedges a firm‟s exposure to interest rate risk. Because changes in the value of the swap and

underlying debt are recognized in taxable income together (Keyes 2008), a swap can reduce the

volatility of taxable income and, thus, generate Benign tax savings for firms with convex tax

functions (Smith and Stulz 1985; Graham and Smith 1999).

While an interest rate swap fits into a specific cubbyhole, tax rules governing the timing

and character of income/deduction recognition vary for different types of payments within that

cubbyhole (Keyes 2008). In particular, elaborate regulations require taxpayers to classify swap

payments as periodic (paid at an interval of one year or less), termination (paid to extinguish the

contract), or non-periodic (payments other than periodic and termination payments). Periodic and

termination payments are recognized when realized, while non-periodic payments are recognized

according to “economic substance.” Because the notion of economic substance is ambiguous, in

practice, non-periodic payments are recognized over contract life using one of three permissible

allocation methods, each of which yield different tax outcomes over time. The reason is that

actual cash payments are made to the counterparty in one manner, but allocated for tax purposes

in a completely different manner that is, in essence, chosen by the taxpayer.

9

For example, assume Firm A enters a three-year interest rate swap with Firm B, where

Firm A makes annual payments to Firm B at 11 percent and Firm B makes annual payments to

Firm A at the London Interbank Offered Rate (LIBOR) based on a notional principal of $100

million. Assume the average market rate for similar swaps is 10 percent such that Firm B will

also pay Firm A a yield adjustment fee of $3 million to cover the off-market difference. As a

non-periodic payment, the yield adjustment fee is generally recognized by Firm B over contract

life by allocating it in accordance with the forward rate for a series of cash-settled forward

contracts that reflect a specified index chosen by the firm.9 If the 10 percent average market rate

were based on the following forward rates: (a) one-year at 10.5 percent; (b) two-year at 10

percent; and (c) three-year at 9.5 percent, then this method assumes that Firm B pays Firm A

$500,000 in year 1, $1 million in year 2, and $1.5 million in year 3.

However, because Firm B actually pays $1 million each year, the $500,000 difference in

year 1 generates a tax deduction that is allocated over contract life. As such, the total net taxable

income (deduction) recognized by Firm B is $740,796 in year 1, $214,876 in year 2, and

($363,636) in year 3 (see Donohoe [2014a] for calculations and details). Overall, the resulting

tax benefits are similar to that of prepaid interest expense in that Firm B pays $1 at time t and

recognizes $1 of deduction at t+n. Thus, the choice of payment structure and method used to

account for payment structure allow a firm to coordinate the timing of taxable income/deduction,

regardless of tax function convexity and/or income volatility, resulting in General tax savings.10

9

The yield adjustment fee is derived from the 1 percent difference (11−10 percent) × $100 million × 3 years. As

noted, in addition to this general allocation method, Treasury Regulations also permit non-periodic payment

allocation using one of two alternative methods solely for timing purposes.

10

Moreover, most non-periodic payments contain a contingent component, which reflects changes in the value of an

index, and a non-contingent component reflecting changes in the underlying interest rate. With respect to character,

periodic and contingent non-periodic payments are ordinary, while termination payments receive capital character

treatment if the underlying is a capital asset. However, contingent non-periodic payments have no prescribed tax

treatment, leaving taxpayers to account for them in “a manner consistent with other tax positions” (GAO 2011, 13).

10

Finally, by blurring economic substance and heightening ambiguity in tax reporting, an

interest rate swap can generate Aggressive tax savings. For instance, some firms have used swaps

to intentionally disguise loan proceeds to defer income (Prokup 2011), and to broadly amplify

the tax savings of otherwise non-derivatives-based corporate tax shelters (Finnerty and Pathak

2011).11 Prior research (Titman 1985; Ferguson 1994; Warren 2004) and government reports

(GAO 2011; JCT 2011) describe how other generic derivatives facilitate tax avoidance.

Whether and to what extent a firm‟s derivatives generate tax savings is often hidden from

the public and regulators because transaction-level detail is not disclosed in financial reports.

Instead, financial reporting standards require only a brief discussion of a firm‟s overall hedging

strategy and objectives for using derivatives (FASB 1998, 2008). For example, throughout the

1990s, Shering-Plough Corporation participated in a “swap-and-assign” tax shelter, which relied

on interest rate swaps with foreign banks to assign interest payments to foreign subsidiaries in

exchange for an amortizable lump sum. Although the arrangement generated over $473 million

in tax savings, financial reports for the affected years list only the notional values of swaps and

state that “derivatives were not used to manage overall interest rate risk” but rather as part of an

“international cash management strategy.”12 Therefore, to evaluate the economic tax effects of

derivatives, the generality of these disclosures necessitate a broad focus on derivatives usage.

11

The IRS issued guidance on seven derivatives-based tax shelters during 2001 to 2007 (GAO 2011). Many others

have been highly publicized, including the Bond and Option Sales Strategy (BOSS; Son of Boss), Currency Options

Bring Reward Alternatives (COBRA), and Foreign Leveraged Investment Program (FLIP). Although tax shelter use

may have subsided in recent years (Lisowsky et al. 2013), aggressive tax reporting with derivatives continues to be

an area of potential abuse (GAO 2011).

12

The Third Circuit Court of Appeals later affirmed that the swap-and-assign arrangement was a disguised loan that

gave rise to Subpart F income. The Court also admonished Schering-Plough for the “chutzpah” of its argument that

the IRS was too vigorous on account of the company‟s history of failing to comply with tax laws. See Merck v.

United States, No. 10-2775 (3d. Cir. June 20, 2011), for details about the transaction and the company‟s 1989-1996

Annual Reports for related financial statement disclosures.

11

2.3 Hypothesis development

Despite the vast potential for tax avoidance, there are three reasons the overall tax

savings from derivatives ([A]+[B]+[C] in Figure 1) might not be economically significant. First,

a firm‟s use of derivatives does not imply that tax avoidance is an objective. Firms regularly use

derivatives to hedge financial risks that threaten revenues, costs of goods sold, and various

expenses (Aretz and Bartram 2010). In addition, some firms may be reluctant to use derivatives

for tax avoidance. For instance, as one of the most complex areas of accounting and taxation

(Chang et al. 2014; Donohoe 2014a), derivatives are challenging to understand, implement, and

execute. Derivatives-based tax avoidance is also a key concern of tax authorities as they continue

to combat corporate tax noncompliance (McConnell 2007; Raghavan 2007, 2008). Thus, tax

avoidance strategies that do not involve derivatives likely provide comparable tax savings with

far less complexity and risk of regulatory scrutiny.

Second, prior research evaluates the Benign tax savings ([A]) that result from a reduction

in taxable income volatility and finds only moderate effects. In particular, using a simulation and

data from 1980–1994, Graham and Smith (1999) estimate the firm-level tax savings from a 5

percent reduction in taxable income volatility, whether due to derivatives or other factors, at

$57,184. Graham and Rogers (2002) show that firms use derivatives in response to tax incentives

and infer median debt-related tax benefits (computed as tax rate × debt) equal to 0.72 percent of

firm value for 85 firms during 1994–1995.13 Third, there are currently no empirical estimates of

the tax savings from Transactional tax avoidance ([B]+[C]). Rather, the concerns of tax

authorities and lawmakers about derivatives-based tax avoidance are largely based on anecdotal

13

Note that this estimate does not reflect actual tax deductible interest expenditures, but rather the capitalized tax

benefits provided by greater debt capacity as a result of less volatile income.

12

evidence (Raskolnikov 2011). Nevertheless, because derivatives facilitate lucrative tax planning

strategies along the entire continuum, I test the following hypothesis (in alternative form):

H1: Firms that use derivatives avoid more tax than firms that do not use derivatives.

3. Data and sample selection

I begin with Compustat observations for fiscal-years 2000–2008 meeting the following

criteria: (1) publicly traded; (2) domestically incorporated; (3) non-financial, non-utility industry;

and (4) at least three years of consecutive observations.14 I identify new derivatives users by

searching the entirety of Form 10-K (extracted from the SEC‟s EDGAR database) for keywords

relating to derivatives. Consistent with Guay (1999), a firm is a New User if it did not report a

derivatives position when it first appears in the sample, but did report a derivatives position in a

subsequent year. Firms enter the New User sample only when derivatives usage is first observed

(after first observing no derivatives usage). The resulting sample consists of 526 New Users.

I also identify two samples consisting of derivatives users and non-users. A firm is a User

if it reports a derivatives position at year t and is not a new user, while firms reporting no

derivatives positions are classified as Non-Users. A New User firm can be a Non-User in an

earlier period if it did not use derivatives for at least two consecutive years, and enters the User

sample after using derivatives for at least two years.15 These two samples consist of 12,437 User

14

Fiscal year 2000 follows the enactment of SFAS No. 133, Accounting for Derivatives Instruments and Hedging

Activities (FASB 1998), which made fundamental changes to financial reporting for derivatives. Excluding

observations before 2000 ensures a consistent financial reporting regime during the sample period. Financial firms

have two-digit SIC codes 60-69, and utility firms have two-digit SIC code 49. I remove these firms as they are more

likely to use derivatives primarily for trading purposes or act as a derivatives dealer, both of which involve

substantially different tax reporting rules (Keyes 2008).

15

To illustrate, consider a firm that did not use derivatives until 2006. For 2000-2005, observations for this firm are

classified in the Non-User sample. In 2006 (the first year of derivatives usage), the observation for this firm is

classified in the New User sample. If the firm does not continue to use derivatives in 2007, the observation for 2007

is classified in the Non-User sample. If, however, the firm continues to use derivatives in 2007, the observation for

2007 is classified in the User sample. Thus, New User designation only occurs the first time that derivatives usage is

observed (after first observing no usage). A small number of firms stop and later restart using derivatives; however,

omitting these firms does not influence the results.

13

and 12,505 Non-User observations. I then drop observations without necessary data to estimate

the models below, resulting in 7,458 Non-User, 9,613 User, and 375 New User observations.

Table 1 reports characteristics of Non-Users, Users, and New Users. Panel A illustrates

the temporal distribution of each subsample. While Non-Users and Users are fairly stable across

time, the largest amount of New Users in 2001 (84) coincides with the effective dates of SFAS

Nos. 133 (FASB 1998) and 138 (FASB 2000), which substantially altered derivatives accounting

and incentivized firms to use derivatives (Abdel-khalik and Chen 2014).16 Panels B and C report

the type of instruments held and risks hedged, respectively, by New Users in each year.

Consistent with Bartram et al. (2009), swaps and interest rate risk account for over half of the

instruments held and risks hedged by New Users, respectively.17 Panel D reports industry

distributions for each subsample. Firms in the manufacturing and business equipment industries

comprise more than one-third of each subsample. Thus, where appropriate, I control for industry

fixed-effects and cluster standard errors by firm and/or year (Petersen 2009).

4. Research design

I test H1 using a levels and difference-in-differences design. The levels design examines

whether firms that use derivatives on an ongoing basis indeed avoid more tax, on average, than

firms that do not use derivatives. The difference-in-differences design examines the tax effects of

derivatives initiation and provides a focused setting in which to estimate the economic

16

SFAS No. 133 requires that an entity recognize all derivatives as either assets or liabilities at fair value and

permits hedge accounting whereby effective hedges minimize earnings volatility. SFAS No. 138, Accounting for

Certain Derivative Instruments and Certain Hedging Activities, provides (in part) cash flow hedge accounting rules.

17

Total amounts exceed New Users (375) as some firms hold multiple instruments and/or hedge more than one type

risk. These data are not used in any tests as some classifications are conjectured based on disclosed information.

14

significance of derivatives-based tax avoidance. Corroborating evidence across both designs will

allow for stronger inferences about the link between derivatives and corporate tax avoidance.18

4.1 Cross-sectional levels specification

4.1.1 Correcting for selectivity bias

Selectivity bias is a concern because a firm‟s decision to use derivatives is not random.

Instead, the decision is determined by both observable factors identified in prior research (Aretz

and Bartram 2010) and unobservable factors such as financial policy, firm sophistication, and

business strategy. I use the Heckman (1979) two-step procedure to control for these factors as

they potentially influence the relation between derivatives and tax avoidance.

First, I estimate a treatment effect model that predicts derivatives usage. Prior research

finds that derivatives usage is a function of risk management incentives, such as risk exposures,

likelihood of financial distress and underinvestment, agency conflicts, and available substitutes

(Aretz and Bartram 2010). Accordingly, I model a firm‟s decision to use derivatives with the

following probit regression based on Gay and Nam (1998) and Bartram et al. (2009):

Pr USERit 0 x RMIitx y TAI ity z CTRLitz k INDitk t YRitt it , (1)

x y z k t

where the dependent variable, USER, is coded 1 for Users and 0 for Non-Users.19

RMI is a vector of risk management incentives that influence derivatives usage. I include

exposures to interest rate (IRISK), foreign exchange rate (FRISK), and commodity price (CRISK)

risks as surveys reveal these are the risks most often managed with derivatives (e.g., Bodnar et

al. 2003). By insulating firm value and cash flow from unfavorable changes in risk exposures,

18

For example, derivatives usage is a potential signal of firm sophistication, which likely suggests the presence of

other tax avoidance strategies. By using both levels and difference-in-differences designs, I mitigate concerns that

such unobserved factors drive the relation between derivatives and tax avoidance.

19

I exclude New Users to focus on the tax effects of ongoing derivatives usage and avoid the potential that firm-

level changes upon derivatives initiation (e.g., risk exposures) drive the levels results. I focus on New Users and

derivatives initiation in the difference-in-differences design (Section 4.2).

15

derivatives can thwart financial distress (Mayers and Smith 1982), harmonize financing and

investment objectives (Froot et al. 1993), and reduce agency conflicts (Smith and Stulz 1985).

To capture these three incentives, I include the likelihood of financial distress (ALTZ), likelihood

of underinvestment (USCORE), and sensitivity of executive compensation to firm value

(ECSENS). As substitutes for derivatives, I also include convertible debt (CDEBT) and preferred

stock (PSTOCK).20 Lastly, the volatility in cash flow (CFV) and earnings (EV) capture other

general incentives for derivatives usage (Zhang 2009).

TAI is a vector of tax avoidance incentives.21 I include the volatility in estimated taxable

income (TIV) because marginal tax rates are lower when taxable income is smooth (Graham and

Rogers 2002). I also include equity method income (EQINC), research and development (RDI),

leverage (LEV, ∆LEV), book-to-market ratio (BTM), capital intensity (PPE), and depreciation

(DEP) to capture economies of scale and complexity (McGuire et al. 2012). To account for the

need and ability to avoid tax (e.g., Dyreng and Lindsey 2009), I include net operating losses

(NOL, ∆NOL), cash holdings (CHLD), and operations in tax haven countries (HAVEN).

CTRL is a vector of control variables that likely influence both derivatives usage and tax

avoidance. I include total assets (SIZE), profitability (ROA, ∆ROA), abnormal accruals (ABACC),

foreign income (FI, ∆FI), auditor type (BIGN, SECTIER), and mergers and acquisitions (M&A).

Finally, industry (IND) and year (YR) fixed-effects control for variation in the decision to use

derivatives across industries and time, respectively. All variables are defined in the Appendix.

20

Convertible debt includes an embedded option on firm assets, which reduces the sensitivity of equity value to

changes in firm value. Although some tax controversies involve features of convertible debt (e.g., Warren 2004),

these instruments are not necessarily reported as derivatives under SFAS No. 133. Preferred stock reduces the

probability of financial distress by paying periodic dividends as opposed to interest. Both of these substitutes can

reduce the incentive to hedge with derivatives (Nance et al. 1993).

21

Although not directly related to derivatives usage, I include TAI because the Heckman (1979) procedure generally

requires that the first stage model (Equation [1]) include “all outcome variables from the second stage model

(Equation [2] below) plus exogenous variables” (Wooldridge 2009, 604).

16

Second, I use coefficient estimates from Equation (1) to construct an inverse Mills ratio

(IMR), which I include as a control variable in Equation (2) below. The inverse Mills ratio is a

bias correction term that controls for the effect of observable and unobservable determinants of a

firm‟s decision to use derivatives on the relation between derivatives usage and tax avoidance.

4.1.2 Empirical model

To examine the cross-sectional levels association between derivatives usage and tax

avoidance, I estimate the following OLS regression based on McGuire et al. (2012):

ETRit 0 1USERit 2 IMRit x RMIitx y TAIity z CTRLitz k INDitk it , (2)

x y z k

where ETR is one of three forward-looking ETRs estimated over a three-year horizon (t to t+2):

(1) current ETR (CURR3), worldwide current tax expense per dollar of book income; (2) cash

ETR (CASH3), worldwide cash taxes paid per dollar of book income (Dyreng et al. 2008); and

(3) book ETR (GAAP3), total tax expense per dollar of book income. CURR3 captures tax

deferral strategies and non-conforming tax avoidance. CASH3 captures tax strategies that defer

cash tax payments to later periods, and GAAP3 captures tax avoidance activities other than

deferral strategies that affect book income (Hanlon and Heitzman 2010). Each forward-looking

ETR is necessary to capture the diverse effects of derivatives-based tax avoidance, which could

occur immediately (e.g., t) or with a delay (e.g., t+2).22

For instance, tax avoidance via transactional form choices can manifest in several ways.

As noted, when an interest rate swap contains a non-periodic payment, income (expense)

22

Consistent with prior studies (Robinson et al. 2010; McGuire et al. 2012), I drop observations with negative ETR

denominators as unprofitable firms are unlikely to have a significant tax liability. Moreover, I use three-year ETRs

to control for variation in tax laws over time, long-run strategies employed by the firm, and settlement of disputes

over previously filed tax returns (Dyreng et al. 2008). By doing so, the dependent variable in Equation (2) is a three-

year forward-looking average of ETRs (t to t+2) computed as of year t while the independent variables are measured

in year t. This design aligns with my research question concerning whether derivatives usage (and later, derivatives

initiation) in year t influences future tax avoidance activities. Nevertheless, for completeness, I perform (unreported)

analyses using three-year average values (t to t+2) of independent variables. Inferences remain the same.

17

recognition may be deferred (accelerated) in accordance with “economic substance.” The choice

of swap payment structure can therefore reduce cash taxes paid (CASH3) and possibly current

tax expense (CURR3), but not likely total income tax expense (GAAP3). Alternatively, choosing

to designate loss positions (e.g., regulated futures) as hedges for tax purposes characterizes

otherwise capital losses as ordinary. Because capital losses are not deductible against ordinary

income, offsetting taxable income with ordinary losses (with no change to book income) can

reduce cash taxes (CASH3) and total tax expense (GAAP3). Similarly, derivatives-based tax

shelters, like all tax shelters, can generate both temporary and permanent book-tax differences

(GAO 2012), which may or may not reduce total tax expense (GAAP3).23

The variable of interest, USER, is coded 1 for Users and 0 for Non-Users. Consistent with

derivatives users avoiding more tax than non-users (H1), I expect a negative coefficient for

USER (θ1) as lower ETRs reflect more tax avoidance. To correctly implement the Heckman

(1979) procedure, at least one exogenous independent variable from the first stage (Equation [1])

must be excluded from the second stage (Equation [2]). Thus, in Equation (2), RMI is the vector

of risk management incentives with one or more exclusion restrictions (see Section 5.2). TAI and

CTRL are the vectors of tax avoidance incentives and control variables, respectively. Industry

fixed-effects (IND) control for cross-sectional variation in tax avoidance across industries.24

23

In general, derivatives create book-tax differences due to the wide disparity in reporting rules, particularly after

SFAS No. 133. For example: (1) there is a comprehensive definition of a derivative for book purposes, but no such

definition for tax purposes; (2) gains/losses are recognized at fair value in the period of change for book, but when

realized or deferred for tax; (3) gains/losses on an exposure follow the hedge for book while gains/losses on a hedge

follow an exposure for tax; and (4) net investment hedges are permitted for book, but not for tax purposes.

24

Although firms with large tax liabilities have tax incentives to use derivatives, the inclusion of a contemporaneous

or lagged measure of tax liability in Equations (1) and (2) is problematic because such measures can be influenced

by derivatives usage in those other periods (e.g., Guay 1999). Nevertheless, I find qualitatively similar results when

I separately include a lagged measure of CURR3, CASH3, and GAAP3 in both models. I also address this potential

issue by using a difference-in-differences specification (Section 4.2).

18

The economic effects of financial derivatives

on corporate tax avoidance

Michael P. Donohoe

www.elsevier.com/locate/jae

PII: S0165-4101(14)00072-X

DOI: http://dx.doi.org/10.1016/j.jacceco.2014.11.001

Reference: JAE1042

To appear in: Journal of Accounting and Economics

Received date: 28 August 2013

Revised date: 3 November 2014

Accepted date: 11 November 2014

Cite this article as: Michael P. Donohoe, The economic effects of financial

derivatives on corporate tax avoidance, Journal of Accounting and Economics, http:

//dx.doi.org/10.1016/j.jacceco.2014.11.001

This is a PDF file of an unedited manuscript that has been accepted for

publication. As a service to our customers we are providing this early version of

the manuscript. The manuscript will undergo copyediting, typesetting, and

review of the resulting galley proof before it is published in its final citable form.

Please note that during the production process errors may be discovered which

could affect the content, and all legal disclaimers that apply to the journal

pertain.

The economic effects of financial derivatives on corporate tax avoidance

Michael P. Donohoe

Assistant Professor of Accountancy

University of Illinois at Urbana-Champaign

1206 S. Sixth Street, MC-706

Champaign, IL 61820

[email protected]

October 2014

Abstract

This study estimates the corporate tax savings from financial derivatives. I document a 3.6 and

4.4 percentage point reduction in three-year current and cash effective tax rates (ETRs),

respectively, after a firm initiates a derivatives program. The decline in cash ETR equates to

$10.69 million in tax savings for the average firm and $4.0 billion for the entire sample of 375

new derivatives users. Of these amounts, $8.75 million and $3.3 billion, respectively, are

incremental to tax savings that theory suggests are a byproduct of risk management. Collectively,

these findings provide economic insight into the prevalence of derivatives-based tax avoidance.

Keywords: financial instruments, derivatives, tax avoidance, effective tax rate

Data Availability: Data are available from public sources identified in the article.

JEL Classification: G32; H25; M40

This study is based on my dissertation completed at the University of Florida, which won the 2012 American

Taxation Association/PricewaterhouseCoopers Outstanding Tax Dissertation Award. I am grateful for the

guidance of my committee: Jon Hamilton, Robert Knechel, Sandra Kramer, and especially my chairman, Gary

McGill. I also appreciate helpful comments from Ross Watts (Editor), an anonymous reviewer, Rashad Abdel-

khalik, Bipin Ajinkya, Andrew Bauer, Paul Beck, Jenny Brown, Raluca Chiorean, Will Ciconte, Dan Collins,

Paul Demeré , Joel Demski, Victoria Dickinson, Brian Gale, Carlos Jimenez, Allison Koester, Andy Leone,

Pete Lisowsky, Michael Mayberry, Sean McGuire, Lil Mills, Ed Outslay, Mark Peecher, Sonja Rego, Casey

Schwab, Terry Shevlin, Bill Snyder, Theo Sougiannis, Jake Thornock, Jaron Wilde, Ryan Wilson, and

workshop participants at the Arizona State University, University of Florida, Florida State University,

Southern Methodist University, Temple University, University of Illinois at Urbana-Champaign, University of

Iowa, University of Kansas, University of Miami, and University of Mississippi. Special thanks to Scott

Dyreng for providing tax haven data, Stephen Brown for data collection programming assistance, and Hye Sun

Chang and Shannon Woods for research assistance. I gratefully acknowledge financial support from the

PricewaterhouseCoopers Faculty Fellowship.

The economic effects of financial derivatives on corporate tax avoidance

1. Introduction

Financial derivatives are a leading source of corporate tax noncompliance (Organisation

for Economic Co-Operation [OECD] 2011; U.S. Government Accountability Office [GAO]

2011). Experts claim the legal patchwork of derivatives taxation encourages the development of

tax planning strategies (Raskolnikov 2011), while the Internal Revenue Service (IRS) concedes it

is “falling farther and farther behind” financial innovation (Raghavan 2007). Although it is clear

from government reports (U.S. Treasury 1999; GAO 2011; U.S. Congress Joint Committee on

Taxation [JCT] 2011), anecdotes (Warner 2008), and academic studies (McDonald 2004; Warren

2004) that companies can avoid tax with derivatives, no study directly examines the economic

tax effects of these financial instruments (Shevlin 1999, 2007). I fill this void by answering the

question: How economically significant are the corporate tax savings from derivatives?1

Nearly two-thirds of non-financial firms in the U.S. participate in the $710 trillion

derivatives market (Bank for International Settlements 2013). In general, derivatives help firms

manage risks relating to interest rates, foreign currency exchange rates, and commodity prices.

However, as a notoriously complex and ambiguous area of tax law, derivatives also promote tax

avoidance. Many groups, including academics (Schizer 2000; Warren 2004), lawmakers (Baucus

2011; Hatch 2011), and regulators (GAO 2011, 2012; JCT 2011; OECD 2011), characterize

derivatives as a significant threat to global tax revenue.

Derivatives facilitate two types of tax avoidance. The first type is a benign byproduct of

risk management as theory suggests that hedging (reducing) risk can lower taxes by reducing the

volatility of taxable income when firm-level tax functions are convex (Smith and Stulz 1985).

1

I define corporate tax avoidance as the reduction of explicit taxes (Hanlon and Heitzman 2010). Tax avoidance

does not necessarily imply that firms are engaging in improper behavior as managing tax costs is an appropriate

component of a firm‟s long-term strategy (Atwood et al. 2012). Section 2 describes derivatives-based tax avoidance.

1

The second type results from both the general and aggressive means in which firms navigate the

extensive ambiguity in derivatives taxation (e.g., Schizer 2000, 2004). In particular, the tax code

takes a categorical approach to derivatives taxation by prescribing rules for only a few types of

instruments. However, because derivatives can replicate the economic profile of virtually any

underlying asset and also create innovative economic profiles, these sophisticated instruments do

not fit neatly into a categorical tax system. Instead, similar (and even identical) economic

positions are taxed differently depending on a versatile transactional form. A firm with a specific

economic objective can therefore choose a transactional form that offers an optimal tax outcome

in terms of the timing, character, and source of related gains/losses. Many of these choices are

perfectly legal, while others aggressively push the envelope of tax law (GAO 2011; JCT 2011).

There are no overall tax savings estimates for derivatives; however, prior research

provides an important first step by evaluating the tax byproducts of risk management. In

particular, Graham and Smith (1999) use a simulation to estimate the tax savings that result from

any type of reduction in taxable income volatility, while Graham and Rogers (2002) gauge the

debt-related tax benefits of lower average volatility. I extend this research by directly estimating

the tax savings attributable to derivatives, both overall and incremental to reductions in volatility.

These estimates are important because, while lawmakers are calling for a complete reform of

derivatives taxation (Baucus 2011; Hatch 2011), “our knowledge of [tax] revenue losses from

[derivatives] is largely anecdotal, wholly unsystematic, and woefully incomplete” (Raskolnikov

2011, 1). Moreover, not all firms that use derivatives necessarily engage in derivatives-based tax

avoidance. Surveys indicate that risk management is a primary objective among derivatives users

(Bodnar et al. 2003), and other forms of tax avoidance likely produce comparable tax savings

2

with far less complexity and scrutiny from tax authorities (JCT 2011). Thus, whether the tax

savings of derivatives are economically significant is an open empirical question.

I use both levels and difference-in-differences designs to investigate the economic tax

effects of derivatives. The levels design examines whether firms that use derivatives on an

ongoing basis indeed avoid more tax than firms that do not use derivatives. The difference-in-

differences design examines the tax effects of derivatives initiation and provides a focused

setting in which to estimate the tax savings of derivatives. I measure tax avoidance with three

different forward-looking effective tax rates (ETRs) estimated over a three-year horizon (t to

t+2), where lower ETRs imply higher levels of tax avoidance. Corroborating evidence across

both designs allows for stronger inferences about the link between derivatives and tax avoidance.

The levels analysis reveals that, on average, derivatives users avoid more tax than non-

users. Specifically, after controlling for endogeneity and factors that explain corporate tax

avoidance, I find that derivatives users have three-year current and cash ETRs that are 0.9

percentage points lower than those of non-users. These amounts are economically significant and

indicate that current and cash ETRs for derivatives users are nearly 3.3 percent smaller than

those of non-users. However, I find no difference in book ETRs (i.e., total tax expense reported

under ASC 740, Income Taxes) among derivatives users and non-users, suggesting that the tax

savings of derivatives largely result from tax-timing opportunities.

The difference-in-differences analysis reveals economically significant tax savings for

firms that begin using derivatives in the sample period. Relative to a propensity score matched

control sample of non-users, I find that new derivatives users realize a 3.6 and 4.4 percentage

point reduction in three-year current and cash ETRs, respectively, after derivatives initiation. The

decline in cash ETR equates to $10.69 million in cash tax savings over a three-year period for the

3

average firm and roughly $4.0 billion in savings for the entire sample of 375 new derivatives

users. These effects are not entirely a byproduct of hedging risk as new derivatives users without

a convex tax function or a reduction in taxable income volatility after initiation realize more tax

savings than new derivatives users with these features. Specifically, the incremental decline in

cash ETR equates to an average savings of $8.75 million for each new user and $3.3 billion in

aggregate. Overall, these estimates are comparable to those for other sophisticated tax planning

strategies, including some tax shelters (Wilson 2009; Brown 2011).

Finally, I mitigate alternative explanations by documenting that the economic tax effects

of derivatives are not driven by (1) reductions in earnings or cash flow volatility; (2) changes in

risk management incentives at the time of derivatives initiation; (3) effective management of risk

exposures; or (4) economies of scale. Instead, I find that the tax savings of derivatives increase

with tax aggressiveness and the magnitude of derivatives positions at initiation. Collectively,

these findings provide insight into the economic incentives (i.e., tax savings) that drive the

growing use of derivatives-based tax planning strategies, and help explain why some lawmakers

aim to weaken their “potential for mischief” (Baucus 2011, 1).

This study contributes to growing literatures on derivatives (Aretz and Bartram 2010) and

corporate tax avoidance (Hanlon and Heitzman 2010). In particular, prior research evaluates the

tax incentives to hedge risk (Graham and Smith 1999; Graham and Rogers 2002), illustrates the

tax benefits of options and forwards (Titman 1985; McDonald 2004; Warren 2004), describes

why derivatives are useful for corporate tax avoidance (Donohoe 2014a), and develops a profile

of derivatives-based tax avoiders (Donohoe 2014b). By providing tax savings estimates, I move

this literature beyond the notion that firms can use derivatives to avoid tax. Although tax

avoidance may not always be the primary reason firms use derivatives (e.g., relative to risk

4

management), they generate significant tax savings nonetheless. I also extend research on

corporate tax shelters (Shevlin 2002; Graham and Tucker 2006; Wilson 2009; Lisowsky 2010;

Brown 2011; Donohoe et al. 2014; McGuire et al. 2014) by documenting the economic effects of

financial instruments often used to generate tax benefits in aggressive tax planning strategies.

Finally, this study answers calls for research on the role of financial instruments in corporate tax

avoidance (Shevlin 1999, 2007; Hanlon and Heitzman 2010; Raskolnikov 2011), complements

the law literature on the taxation of financial products (Schizer 2000, 2004), and provides insight

into why lawmakers are resolute to reform derivatives taxation (Baucus 2011; Hatch 2011).

Section 2 develops my hypothesis. I discuss data, research design and results in Sections

3, 4 and 5, respectively. Section 6 discusses additional tests, and Section 7 concludes.

2. Background and hypothesis development

2.1 Derivatives-based tax avoidance

A derivative is a contract or security deriving its value based on its relation to something

else, commonly referred to as the “underlying” (Stulz 2004). The underlying is often another

financial instrument or economic good, but can be almost anything.2 For example, the values of

some derivatives are based on familiar stock indices, the heat index in Florida, and that of other

derivatives. Many companies use derivatives to manage interest rate, foreign exchange rate, and

commodity price risks (Bodnar et al. 2003). However, an increasingly common motive for

derivatives usage is corporate tax avoidance (GAO 2011); that is, the reduction of explicit taxes.3

2

Derivatives generally fall into three groups: (1) options; (2) futures and forwards; and (3) swaps. Options involve

the right, but not the obligation, to buy or sell an underlying at a set price within a specified period. A futures or

forward contract involves an obligation to exchange an underlying at a future date for a specific price, and swaps are

agreements to exchange a stream of payments based on an underlying over a predefined period.

3

Explicit taxes are taxes paid to tax authorities whereas implicit taxes reflect the lower before-tax return on tax-

favored assets relative to tax-disfavored assets (Shackelford and Shevlin 2001; Scholes et al. 2015).

5

Tax avoidance represents a continuum of tax planning strategies, where relatively benign

strategies lie at one end and extremely aggressive or illegal strategies (e.g., some tax shelters) lie

at the opposing end (Lisowsky et al. 2013). Derivatives facilitate tax planning strategies along

the entire continuum. In particular, theory suggests some tax avoidance is a benign byproduct of

risk management (Smith and Stulz 1985), while the law literature describes both the general and

aggressive means in which firms navigate the extensive ambiguity in derivatives taxation (e.g.,

Schizer 2000, 2004). Figure 1 incorporates these elements and illustrates my view of the tax

avoidance continuum with respect to derivatives.

The far left-end of the continuum, Byproduct, reflects tax avoidance that is an ancillary

effect of managing risk. Under current tax rules, effective (i.e., successful) hedging reduces the

volatility of taxable income (Keyes 2008), which could result in lower taxes for two reasons.4

First, in a progressive tax system, expected taxes are a convex function of taxable income, where

firms facing an increasing marginal tax rate can reduce taxes by smoothing income (Smith and

Stulz 1985).5 Second, smooth income can increase debt capacity, potentially increasing tax

deductible interest expenditures (Stulz 1996; Leland 1998). Prior empirical tests of these theories

(discussed later) suggest that hedging risk may indeed result in some tax savings (Graham and

Smith 1999; Graham and Rogers 2002). I view this type of tax savings as Benign ([A] in Figure

1) as it is attributable to less volatile taxable income rather than overt attempts to reduce taxes.

4

For tax reporting purposes, a hedge is effective if it actually reduces exposures to economic risk (Yanchisin and

Ricks 2006). In such cases, changes in the value of both the derivative and underlying are recognized in taxable

income together (i.e., offset). Gains and losses on speculative, ineffective, or straddle positions are recognized in

taxable income as they occur (i.e., no offset). A straddle is an offsetting position (not designated or qualifying as a

hedge) against actively traded property substantially reducing risk of loss. See Donohoe (2014a) for details.

5

Graham and Smith (1999, 2243) provide the following example: “Assuming two equally likely outcomes–low

taxable income with no associated tax liability or high taxable income and a $1 million tax liability–the firm‟s

expected tax liability is $500,000. However, if the firm could hedge and completely eliminate the volatility of

taxable income, both its [expected] taxable income and tax bill would be zero.”

6

The remainder of the continuum, Transactional, reflects tax avoidance that is attributable

to three sources of ambiguity in the taxation of derivatives transactions (Weisbach 2005). First,

derivatives taxation is inconsistent in that not all economically comparable transactions receive

the same tax treatment. In essence, existing statutes and case law create a “cubbyhole” system,

where tax treatments are prescribed for only a few types of derivatives. A firm determines the tax

treatment for a derivative by using its broad transactional features to identify the appropriate

cubbyhole (see Figure 2). However, through financial engineering (e.g., put-call parity),

derivatives can replicate the economic profile of virtually any underlying asset, including other

derivatives (Stoll 1969; Merton 1973; Scholes 1998).6 Consequently, similar (and even identical)

economic positions are taxed differently depending on a versatile transactional form. For a given

economic position, a firm can often choose a transactional form that provides an optimal tax

outcome in terms of the timing, character, and source of related gains/losses (JCT 2011).

Second, because the tax law response to financial innovation has been “reactive and

particularized” (Warren 2004, 913), the “correct” tax treatment for many sophisticated, yet

common, derivatives (e.g., swaptions) is indeterminable. In particular, novel transactions and

those with unique economic profiles rarely fit into preexisting cubbyholes, while “compound”

instruments often fit into multiple cubbyholes. Such uncertainty translates into discretion and

variation in the tax reporting of these instruments (JCT 2011).

Third, ambiguity arises from the asymmetric tax treatment of opposing sides to the same

transaction. In addition to cubbyholes, derivatives taxation depends on taxpayer characteristics,

such as motive (hedge or speculate), entity form (corporate or pass-through), jurisdiction (foreign

or domestic), and status (securities dealer). Thus, it is possible for a firm to make transactional

6

Financial engineering refers to the construction of asset portfolios that have precise technical characteristics,

particularly when those characteristics are not available in an existing exchange-traded product (Strong 2005, 346).

7

form and/or tax reporting choices without obligating counterparties to make similar choices. This

transactional autonomy provides added flexibility in tax planning.7 In sum, while smooth taxable

income can lead to Byproduct tax avoidance, Transactional tax avoidance results from the ability

to strategically coordinate the timing, character, and source of taxable income/deduction due to

the extensive ambiguity in derivatives taxation.

Whether Transactional tax avoidance is controversial or not depends on the degree to

which a firm exploits the ambiguity in derivatives taxation. As noted, the tax treatment of

derivatives is based on transactional form rather than economic form. Because transactions with

the same or similar economic characteristics can be treated differently from one another for tax

purposes, a firm with a specific economic objective can choose a transactional form that provides

an optimal tax outcome (JCT 2011).8 When transactional form is chosen from a set of legally

permitted alternatives after weighing tax and non-tax factors, Transactional tax avoidance is not

likely to be controversial. I view the resulting tax savings as General ([B] in Figure 1).

However, controversy can arise when the choice of transactional form and/or its related

tax reporting is aggressive, which is defined as having “weak legal support” (Lisowsky et al.

2013, 590) that “pushes the envelope of tax law” (Hanlon and Heitzman 2010, 137). For

example, a firm can use derivatives to design and operate tax shelters (U.S. Treasury 1999; GAO

2011, 2012). Along these lines, derivatives can serve as either a supportive element within a

larger arrangement or the central element around which a tax shelter is designed (see Donohoe

[2014a] for two examples). By doing so, derivatives not only “turbo charge” tax shelters

7

For example, the fact that one party to a transaction designates a derivative as a “hedge” (i.e., intending to reduce

economic risk) for tax purposes does not necessarily compel a counterparty to follow suit. Symmetry is an important

policy issue in the taxation of financial products because a tax system can better police itself when both sides to a

transaction face equal and opposite incentives (Raskolnikov 2011).

8

That is, by choosing to hold one type of instrument over another, a taxpayer can obtain a desired (1) timing of

income/expense recognition; (2) characterization of income/loss as capital or ordinary; or (3) source of income/

expense as foreign or domestic (JCT 2011). Section 2.2 provides a brief illustration of timing.

8

(Sheppard 1999, 132), but make them especially difficult for tax authorities to detect and contest

(GAO 2012). I view this type of resulting tax savings as Aggressive ([C] in Figure 1).

2.2 A brief illustration of derivatives-based tax avoidance

Although derivatives can be extraordinarily complex, mundane instruments generate all

three types of tax savings. To illustrate, consider the most common derivative: a generic (“plain

vanilla”) interest rate swap in which fixed interest rate payments are exchanged for floating rate

payments. By transforming fixed rate debt into variable rate debt (and vice versa), an interest rate

swap hedges a firm‟s exposure to interest rate risk. Because changes in the value of the swap and

underlying debt are recognized in taxable income together (Keyes 2008), a swap can reduce the

volatility of taxable income and, thus, generate Benign tax savings for firms with convex tax

functions (Smith and Stulz 1985; Graham and Smith 1999).

While an interest rate swap fits into a specific cubbyhole, tax rules governing the timing

and character of income/deduction recognition vary for different types of payments within that

cubbyhole (Keyes 2008). In particular, elaborate regulations require taxpayers to classify swap

payments as periodic (paid at an interval of one year or less), termination (paid to extinguish the

contract), or non-periodic (payments other than periodic and termination payments). Periodic and

termination payments are recognized when realized, while non-periodic payments are recognized

according to “economic substance.” Because the notion of economic substance is ambiguous, in

practice, non-periodic payments are recognized over contract life using one of three permissible

allocation methods, each of which yield different tax outcomes over time. The reason is that

actual cash payments are made to the counterparty in one manner, but allocated for tax purposes

in a completely different manner that is, in essence, chosen by the taxpayer.

9

For example, assume Firm A enters a three-year interest rate swap with Firm B, where

Firm A makes annual payments to Firm B at 11 percent and Firm B makes annual payments to

Firm A at the London Interbank Offered Rate (LIBOR) based on a notional principal of $100

million. Assume the average market rate for similar swaps is 10 percent such that Firm B will

also pay Firm A a yield adjustment fee of $3 million to cover the off-market difference. As a

non-periodic payment, the yield adjustment fee is generally recognized by Firm B over contract

life by allocating it in accordance with the forward rate for a series of cash-settled forward

contracts that reflect a specified index chosen by the firm.9 If the 10 percent average market rate

were based on the following forward rates: (a) one-year at 10.5 percent; (b) two-year at 10

percent; and (c) three-year at 9.5 percent, then this method assumes that Firm B pays Firm A

$500,000 in year 1, $1 million in year 2, and $1.5 million in year 3.

However, because Firm B actually pays $1 million each year, the $500,000 difference in

year 1 generates a tax deduction that is allocated over contract life. As such, the total net taxable

income (deduction) recognized by Firm B is $740,796 in year 1, $214,876 in year 2, and

($363,636) in year 3 (see Donohoe [2014a] for calculations and details). Overall, the resulting

tax benefits are similar to that of prepaid interest expense in that Firm B pays $1 at time t and

recognizes $1 of deduction at t+n. Thus, the choice of payment structure and method used to

account for payment structure allow a firm to coordinate the timing of taxable income/deduction,

regardless of tax function convexity and/or income volatility, resulting in General tax savings.10

9

The yield adjustment fee is derived from the 1 percent difference (11−10 percent) × $100 million × 3 years. As

noted, in addition to this general allocation method, Treasury Regulations also permit non-periodic payment

allocation using one of two alternative methods solely for timing purposes.

10

Moreover, most non-periodic payments contain a contingent component, which reflects changes in the value of an

index, and a non-contingent component reflecting changes in the underlying interest rate. With respect to character,

periodic and contingent non-periodic payments are ordinary, while termination payments receive capital character

treatment if the underlying is a capital asset. However, contingent non-periodic payments have no prescribed tax

treatment, leaving taxpayers to account for them in “a manner consistent with other tax positions” (GAO 2011, 13).

10

Finally, by blurring economic substance and heightening ambiguity in tax reporting, an

interest rate swap can generate Aggressive tax savings. For instance, some firms have used swaps

to intentionally disguise loan proceeds to defer income (Prokup 2011), and to broadly amplify

the tax savings of otherwise non-derivatives-based corporate tax shelters (Finnerty and Pathak

2011).11 Prior research (Titman 1985; Ferguson 1994; Warren 2004) and government reports

(GAO 2011; JCT 2011) describe how other generic derivatives facilitate tax avoidance.

Whether and to what extent a firm‟s derivatives generate tax savings is often hidden from

the public and regulators because transaction-level detail is not disclosed in financial reports.

Instead, financial reporting standards require only a brief discussion of a firm‟s overall hedging

strategy and objectives for using derivatives (FASB 1998, 2008). For example, throughout the

1990s, Shering-Plough Corporation participated in a “swap-and-assign” tax shelter, which relied

on interest rate swaps with foreign banks to assign interest payments to foreign subsidiaries in

exchange for an amortizable lump sum. Although the arrangement generated over $473 million

in tax savings, financial reports for the affected years list only the notional values of swaps and

state that “derivatives were not used to manage overall interest rate risk” but rather as part of an

“international cash management strategy.”12 Therefore, to evaluate the economic tax effects of

derivatives, the generality of these disclosures necessitate a broad focus on derivatives usage.

11

The IRS issued guidance on seven derivatives-based tax shelters during 2001 to 2007 (GAO 2011). Many others

have been highly publicized, including the Bond and Option Sales Strategy (BOSS; Son of Boss), Currency Options

Bring Reward Alternatives (COBRA), and Foreign Leveraged Investment Program (FLIP). Although tax shelter use

may have subsided in recent years (Lisowsky et al. 2013), aggressive tax reporting with derivatives continues to be

an area of potential abuse (GAO 2011).

12

The Third Circuit Court of Appeals later affirmed that the swap-and-assign arrangement was a disguised loan that

gave rise to Subpart F income. The Court also admonished Schering-Plough for the “chutzpah” of its argument that

the IRS was too vigorous on account of the company‟s history of failing to comply with tax laws. See Merck v.

United States, No. 10-2775 (3d. Cir. June 20, 2011), for details about the transaction and the company‟s 1989-1996

Annual Reports for related financial statement disclosures.

11

2.3 Hypothesis development

Despite the vast potential for tax avoidance, there are three reasons the overall tax

savings from derivatives ([A]+[B]+[C] in Figure 1) might not be economically significant. First,

a firm‟s use of derivatives does not imply that tax avoidance is an objective. Firms regularly use

derivatives to hedge financial risks that threaten revenues, costs of goods sold, and various

expenses (Aretz and Bartram 2010). In addition, some firms may be reluctant to use derivatives

for tax avoidance. For instance, as one of the most complex areas of accounting and taxation

(Chang et al. 2014; Donohoe 2014a), derivatives are challenging to understand, implement, and

execute. Derivatives-based tax avoidance is also a key concern of tax authorities as they continue

to combat corporate tax noncompliance (McConnell 2007; Raghavan 2007, 2008). Thus, tax

avoidance strategies that do not involve derivatives likely provide comparable tax savings with

far less complexity and risk of regulatory scrutiny.

Second, prior research evaluates the Benign tax savings ([A]) that result from a reduction

in taxable income volatility and finds only moderate effects. In particular, using a simulation and

data from 1980–1994, Graham and Smith (1999) estimate the firm-level tax savings from a 5

percent reduction in taxable income volatility, whether due to derivatives or other factors, at

$57,184. Graham and Rogers (2002) show that firms use derivatives in response to tax incentives

and infer median debt-related tax benefits (computed as tax rate × debt) equal to 0.72 percent of

firm value for 85 firms during 1994–1995.13 Third, there are currently no empirical estimates of

the tax savings from Transactional tax avoidance ([B]+[C]). Rather, the concerns of tax

authorities and lawmakers about derivatives-based tax avoidance are largely based on anecdotal

13

Note that this estimate does not reflect actual tax deductible interest expenditures, but rather the capitalized tax

benefits provided by greater debt capacity as a result of less volatile income.

12

evidence (Raskolnikov 2011). Nevertheless, because derivatives facilitate lucrative tax planning

strategies along the entire continuum, I test the following hypothesis (in alternative form):

H1: Firms that use derivatives avoid more tax than firms that do not use derivatives.

3. Data and sample selection

I begin with Compustat observations for fiscal-years 2000–2008 meeting the following

criteria: (1) publicly traded; (2) domestically incorporated; (3) non-financial, non-utility industry;

and (4) at least three years of consecutive observations.14 I identify new derivatives users by

searching the entirety of Form 10-K (extracted from the SEC‟s EDGAR database) for keywords

relating to derivatives. Consistent with Guay (1999), a firm is a New User if it did not report a

derivatives position when it first appears in the sample, but did report a derivatives position in a

subsequent year. Firms enter the New User sample only when derivatives usage is first observed

(after first observing no derivatives usage). The resulting sample consists of 526 New Users.

I also identify two samples consisting of derivatives users and non-users. A firm is a User

if it reports a derivatives position at year t and is not a new user, while firms reporting no

derivatives positions are classified as Non-Users. A New User firm can be a Non-User in an

earlier period if it did not use derivatives for at least two consecutive years, and enters the User

sample after using derivatives for at least two years.15 These two samples consist of 12,437 User

14

Fiscal year 2000 follows the enactment of SFAS No. 133, Accounting for Derivatives Instruments and Hedging

Activities (FASB 1998), which made fundamental changes to financial reporting for derivatives. Excluding

observations before 2000 ensures a consistent financial reporting regime during the sample period. Financial firms

have two-digit SIC codes 60-69, and utility firms have two-digit SIC code 49. I remove these firms as they are more

likely to use derivatives primarily for trading purposes or act as a derivatives dealer, both of which involve

substantially different tax reporting rules (Keyes 2008).

15

To illustrate, consider a firm that did not use derivatives until 2006. For 2000-2005, observations for this firm are

classified in the Non-User sample. In 2006 (the first year of derivatives usage), the observation for this firm is

classified in the New User sample. If the firm does not continue to use derivatives in 2007, the observation for 2007

is classified in the Non-User sample. If, however, the firm continues to use derivatives in 2007, the observation for

2007 is classified in the User sample. Thus, New User designation only occurs the first time that derivatives usage is

observed (after first observing no usage). A small number of firms stop and later restart using derivatives; however,

omitting these firms does not influence the results.

13

and 12,505 Non-User observations. I then drop observations without necessary data to estimate

the models below, resulting in 7,458 Non-User, 9,613 User, and 375 New User observations.

Table 1 reports characteristics of Non-Users, Users, and New Users. Panel A illustrates

the temporal distribution of each subsample. While Non-Users and Users are fairly stable across

time, the largest amount of New Users in 2001 (84) coincides with the effective dates of SFAS

Nos. 133 (FASB 1998) and 138 (FASB 2000), which substantially altered derivatives accounting

and incentivized firms to use derivatives (Abdel-khalik and Chen 2014).16 Panels B and C report

the type of instruments held and risks hedged, respectively, by New Users in each year.

Consistent with Bartram et al. (2009), swaps and interest rate risk account for over half of the

instruments held and risks hedged by New Users, respectively.17 Panel D reports industry

distributions for each subsample. Firms in the manufacturing and business equipment industries

comprise more than one-third of each subsample. Thus, where appropriate, I control for industry

fixed-effects and cluster standard errors by firm and/or year (Petersen 2009).

4. Research design

I test H1 using a levels and difference-in-differences design. The levels design examines

whether firms that use derivatives on an ongoing basis indeed avoid more tax, on average, than

firms that do not use derivatives. The difference-in-differences design examines the tax effects of

derivatives initiation and provides a focused setting in which to estimate the economic

16

SFAS No. 133 requires that an entity recognize all derivatives as either assets or liabilities at fair value and

permits hedge accounting whereby effective hedges minimize earnings volatility. SFAS No. 138, Accounting for

Certain Derivative Instruments and Certain Hedging Activities, provides (in part) cash flow hedge accounting rules.

17

Total amounts exceed New Users (375) as some firms hold multiple instruments and/or hedge more than one type

risk. These data are not used in any tests as some classifications are conjectured based on disclosed information.

14

significance of derivatives-based tax avoidance. Corroborating evidence across both designs will

allow for stronger inferences about the link between derivatives and corporate tax avoidance.18

4.1 Cross-sectional levels specification

4.1.1 Correcting for selectivity bias

Selectivity bias is a concern because a firm‟s decision to use derivatives is not random.

Instead, the decision is determined by both observable factors identified in prior research (Aretz

and Bartram 2010) and unobservable factors such as financial policy, firm sophistication, and

business strategy. I use the Heckman (1979) two-step procedure to control for these factors as

they potentially influence the relation between derivatives and tax avoidance.

First, I estimate a treatment effect model that predicts derivatives usage. Prior research

finds that derivatives usage is a function of risk management incentives, such as risk exposures,

likelihood of financial distress and underinvestment, agency conflicts, and available substitutes

(Aretz and Bartram 2010). Accordingly, I model a firm‟s decision to use derivatives with the

following probit regression based on Gay and Nam (1998) and Bartram et al. (2009):

Pr USERit 0 x RMIitx y TAI ity z CTRLitz k INDitk t YRitt it , (1)

x y z k t

where the dependent variable, USER, is coded 1 for Users and 0 for Non-Users.19

RMI is a vector of risk management incentives that influence derivatives usage. I include

exposures to interest rate (IRISK), foreign exchange rate (FRISK), and commodity price (CRISK)

risks as surveys reveal these are the risks most often managed with derivatives (e.g., Bodnar et

al. 2003). By insulating firm value and cash flow from unfavorable changes in risk exposures,

18

For example, derivatives usage is a potential signal of firm sophistication, which likely suggests the presence of

other tax avoidance strategies. By using both levels and difference-in-differences designs, I mitigate concerns that

such unobserved factors drive the relation between derivatives and tax avoidance.

19

I exclude New Users to focus on the tax effects of ongoing derivatives usage and avoid the potential that firm-

level changes upon derivatives initiation (e.g., risk exposures) drive the levels results. I focus on New Users and

derivatives initiation in the difference-in-differences design (Section 4.2).

15

derivatives can thwart financial distress (Mayers and Smith 1982), harmonize financing and

investment objectives (Froot et al. 1993), and reduce agency conflicts (Smith and Stulz 1985).

To capture these three incentives, I include the likelihood of financial distress (ALTZ), likelihood

of underinvestment (USCORE), and sensitivity of executive compensation to firm value

(ECSENS). As substitutes for derivatives, I also include convertible debt (CDEBT) and preferred

stock (PSTOCK).20 Lastly, the volatility in cash flow (CFV) and earnings (EV) capture other

general incentives for derivatives usage (Zhang 2009).

TAI is a vector of tax avoidance incentives.21 I include the volatility in estimated taxable

income (TIV) because marginal tax rates are lower when taxable income is smooth (Graham and

Rogers 2002). I also include equity method income (EQINC), research and development (RDI),

leverage (LEV, ∆LEV), book-to-market ratio (BTM), capital intensity (PPE), and depreciation

(DEP) to capture economies of scale and complexity (McGuire et al. 2012). To account for the

need and ability to avoid tax (e.g., Dyreng and Lindsey 2009), I include net operating losses

(NOL, ∆NOL), cash holdings (CHLD), and operations in tax haven countries (HAVEN).

CTRL is a vector of control variables that likely influence both derivatives usage and tax

avoidance. I include total assets (SIZE), profitability (ROA, ∆ROA), abnormal accruals (ABACC),

foreign income (FI, ∆FI), auditor type (BIGN, SECTIER), and mergers and acquisitions (M&A).

Finally, industry (IND) and year (YR) fixed-effects control for variation in the decision to use

derivatives across industries and time, respectively. All variables are defined in the Appendix.

20

Convertible debt includes an embedded option on firm assets, which reduces the sensitivity of equity value to

changes in firm value. Although some tax controversies involve features of convertible debt (e.g., Warren 2004),

these instruments are not necessarily reported as derivatives under SFAS No. 133. Preferred stock reduces the

probability of financial distress by paying periodic dividends as opposed to interest. Both of these substitutes can

reduce the incentive to hedge with derivatives (Nance et al. 1993).

21

Although not directly related to derivatives usage, I include TAI because the Heckman (1979) procedure generally

requires that the first stage model (Equation [1]) include “all outcome variables from the second stage model

(Equation [2] below) plus exogenous variables” (Wooldridge 2009, 604).

16

Second, I use coefficient estimates from Equation (1) to construct an inverse Mills ratio

(IMR), which I include as a control variable in Equation (2) below. The inverse Mills ratio is a

bias correction term that controls for the effect of observable and unobservable determinants of a

firm‟s decision to use derivatives on the relation between derivatives usage and tax avoidance.

4.1.2 Empirical model

To examine the cross-sectional levels association between derivatives usage and tax

avoidance, I estimate the following OLS regression based on McGuire et al. (2012):

ETRit 0 1USERit 2 IMRit x RMIitx y TAIity z CTRLitz k INDitk it , (2)

x y z k

where ETR is one of three forward-looking ETRs estimated over a three-year horizon (t to t+2):

(1) current ETR (CURR3), worldwide current tax expense per dollar of book income; (2) cash

ETR (CASH3), worldwide cash taxes paid per dollar of book income (Dyreng et al. 2008); and

(3) book ETR (GAAP3), total tax expense per dollar of book income. CURR3 captures tax

deferral strategies and non-conforming tax avoidance. CASH3 captures tax strategies that defer

cash tax payments to later periods, and GAAP3 captures tax avoidance activities other than

deferral strategies that affect book income (Hanlon and Heitzman 2010). Each forward-looking

ETR is necessary to capture the diverse effects of derivatives-based tax avoidance, which could

occur immediately (e.g., t) or with a delay (e.g., t+2).22

For instance, tax avoidance via transactional form choices can manifest in several ways.

As noted, when an interest rate swap contains a non-periodic payment, income (expense)

22

Consistent with prior studies (Robinson et al. 2010; McGuire et al. 2012), I drop observations with negative ETR

denominators as unprofitable firms are unlikely to have a significant tax liability. Moreover, I use three-year ETRs

to control for variation in tax laws over time, long-run strategies employed by the firm, and settlement of disputes

over previously filed tax returns (Dyreng et al. 2008). By doing so, the dependent variable in Equation (2) is a three-

year forward-looking average of ETRs (t to t+2) computed as of year t while the independent variables are measured

in year t. This design aligns with my research question concerning whether derivatives usage (and later, derivatives

initiation) in year t influences future tax avoidance activities. Nevertheless, for completeness, I perform (unreported)

analyses using three-year average values (t to t+2) of independent variables. Inferences remain the same.

17

recognition may be deferred (accelerated) in accordance with “economic substance.” The choice

of swap payment structure can therefore reduce cash taxes paid (CASH3) and possibly current

tax expense (CURR3), but not likely total income tax expense (GAAP3). Alternatively, choosing

to designate loss positions (e.g., regulated futures) as hedges for tax purposes characterizes

otherwise capital losses as ordinary. Because capital losses are not deductible against ordinary

income, offsetting taxable income with ordinary losses (with no change to book income) can

reduce cash taxes (CASH3) and total tax expense (GAAP3). Similarly, derivatives-based tax

shelters, like all tax shelters, can generate both temporary and permanent book-tax differences

(GAO 2012), which may or may not reduce total tax expense (GAAP3).23

The variable of interest, USER, is coded 1 for Users and 0 for Non-Users. Consistent with

derivatives users avoiding more tax than non-users (H1), I expect a negative coefficient for

USER (θ1) as lower ETRs reflect more tax avoidance. To correctly implement the Heckman

(1979) procedure, at least one exogenous independent variable from the first stage (Equation [1])

must be excluded from the second stage (Equation [2]). Thus, in Equation (2), RMI is the vector

of risk management incentives with one or more exclusion restrictions (see Section 5.2). TAI and

CTRL are the vectors of tax avoidance incentives and control variables, respectively. Industry

fixed-effects (IND) control for cross-sectional variation in tax avoidance across industries.24

23

In general, derivatives create book-tax differences due to the wide disparity in reporting rules, particularly after

SFAS No. 133. For example: (1) there is a comprehensive definition of a derivative for book purposes, but no such

definition for tax purposes; (2) gains/losses are recognized at fair value in the period of change for book, but when

realized or deferred for tax; (3) gains/losses on an exposure follow the hedge for book while gains/losses on a hedge

follow an exposure for tax; and (4) net investment hedges are permitted for book, but not for tax purposes.

24

Although firms with large tax liabilities have tax incentives to use derivatives, the inclusion of a contemporaneous

or lagged measure of tax liability in Equations (1) and (2) is problematic because such measures can be influenced

by derivatives usage in those other periods (e.g., Guay 1999). Nevertheless, I find qualitatively similar results when

I separately include a lagged measure of CURR3, CASH3, and GAAP3 in both models. I also address this potential

issue by using a difference-in-differences specification (Section 4.2).

18