Contents lists available at ScienceDirect

Journal of Accounting and Economics

Journal of Accounting and Economics 56 (2013) 228–250

0165-41http://d

☆ WeFabrizioNissim,UniversConfereTexas TaCooperstitled "T

n CorrE-m1 Te2 Te

journal homepage: www.elsevier.com/locate/jae

The separation of ownership and control andcorporate tax avoidance$

Brad A. Badertscher a,1, Sharon P. Katz b,2, Sonja O. Rego c,n

a Notre Dame University, 371 Mendoza College of Business, Notre Dame, IN 46556-5646, USAb Columbia Business School, Uris Hall, 3022 Broadway, Room 605A, New York, NY 10027, USAc Indiana University, Kelley School of Business, 1309 E. 10th St., Bloomington, IN 47405-1701, USA

a r t i c l e i n f o

Article history:Received 5 July 2011Received in revised form16 August 2013Accepted 22 August 2013Available online 29 August 2013

JEL classification:G30G32H25K34M40M49

Keywords:Ownership structureAgency costsTax avoidancePrivate equity firmsEffective tax rates

01/$ - see front matter & 2013 Elsevier B.V. Ax.doi.org/10.1016/j.jacceco.2013.08.005

are grateful for helpful comments from JeroFerri, Dan Givoly, Cristi Gleason, Michelle HTom Omer, Krishna Palepu, Gil Sadka, Jimity, Columbia University, Tel-Aviv University,nce, 2010 London Business School Accountinx Readings Group. We thank Michelle Shim, the Kelley School of Business, and the Menhe Impact of Private Equity Ownership on Pesponding author. Tel.: þ1 812 855 6356.ail addresses: bbaderts@nd.edu (B.A. Badertsl.: þ1 574 631 5197.l.: þ1 212 851 9442.

a b s t r a c t

We examine whether variation in the separation of ownership and control influences thetax practices of private firms with different ownership structures. Fama and Jensen (1983)assert that when equity ownership and corporate decision-making are concentratedin just a small number of decision-makers, these owner-managers will likely be more riskaverse and thus less willing to invest in risky projects. Because tax avoidance is a riskyactivity that can impose significant costs on a firm, we predict that firms with greaterconcentrations of ownership and control, and thus more risk averse managers, avoidless income tax than firms with less concentrated ownership and control. Our results areconsistent with these expectations. However, we also consider a competing explanationfor these findings. In particular, we examine whether certain private firms enjoy lowermarginal costs of tax planning, which facilitate greater income tax avoidance. Our resultsare consistent with the marginal costs of tax avoidance and the separation of ownershipand control both influencing corporate tax practices.

& 2013 Elsevier B.V. All rights reserved.

1. Introduction

In this study we investigate the impact of ownership structure on corporate tax avoidance. Shackelford and Shevlin(2001) note that little is known about the cross-sectional differences in the willingness of firms to minimize taxes. They

ll rights reserved.

ld Zimmerman (Editor) and an anonymous referee, Ramji Balakrishnan, Jennifer Blouin, Dan Collins,anlon, Shane Heitzman, Paul Hribar, Alon Kalay, Michael Kimbrough, Josh Lerner, Greg Miller, DoronSeida, Joseph Weber, Ryan Wilson, and workshop participants at Baruch College at CUNY, Boston2009 Information, Markets & Organization Conference at Harvard Business School, 2009 JAAF/KPMGg Symposium, University of Colorado at Boulder, University of Iowa, University of Minnesota, and theek for her assistance with the hand-collection of tax footnote data. We also thank Pricewaterhouse-doza College of Business for financial support. All errors are our own. Prior versions of this study wereortfolio Firms' Corporate Tax Avoidance."

cher), sk2280@columbia.edu (S.P. Katz), sorego@indiana.edu (S.O. Rego).

B.A. Badertscher et al. / Journal of Accounting and Economics 56 (2013) 228–250 229

point out that insider control is an important but understudied organizational factor that impacts corporate tax avoidance.We take advantage of a unique sample of firms with privately-owned equity but publicly-traded debt and examine whethervariation in the separation of ownership and control (i.e., the extent of inside ownership) influences income tax avoidance atprivate firms.3

Our sample includes private firms that are majority-owned by the firm's managers (i.e., management-owned firms) andprivate firms that are owned by private equity (PE) firms (i.e., PE-backed firms).4 As a result, our research setting exhibitssubstantial variation in the separation of ownership and control, as reflected in the proportion of stock owned by the firm'smanagers.5 Due to their public debt, sample firms are required to file financial statements with the Securities and ExchangeCommission (SEC). These filings allow us to utilize audited financial information and examine corporate tax practices whileholding financial reporting requirements constant. Moreover, because the private firms in our sample are subject to lesspublic scrutiny than publicly-traded firms (Givoly et al., 2010), they place less weight on financial reporting decisions andmore weight on tax reporting decisions, relative to public firms (e.g., Penno and Simon, 1986; Beatty and Harris, 1998). All ofthese features enhance the power of our empirical tests.

Our first analysis compares the income tax avoidance of management-owned and PE-backed private firms. We predictthat management-owned firms avoid less tax than PE-backed firms because management-owned firms have more highlyconcentrated ownership and control than PE-backed firms. Our prediction is based on Fama and Jensen's (1983) theory thatwhen equity ownership and corporate decision-making are concentrated in just a small number of decision-makers, theseowner-managers will likely be more risk averse and thus less willing to invest in risky projects. Since tax avoidance is a riskyactivity that can impose significant costs on firms and their managers (e.g. Desai and Dharmapala, 2008; Hanlon andHeitzman, 2010; Rego and Wilson, 2012), we conjecture that firms with more highly concentrated ownership and control(and thus more risk averse managers) avoid less income tax than firms with less concentrated ownership and control.Because PE firms do not randomly select firms to acquire (e.g., PE firms generally acquire firms with stable profits and cashflows, low leverage, and little risk of financial distress, as documented in Opler and Titman, 1993), we follow the Heckman(1979) procedure to mitigate potential selection bias in our sample of management-owned and PE-backed private firms.6

Using the Heckman procedure and several measures of corporate tax avoidance, we find that management-owned firmsavoid significantly less income tax than PE-backed firms. These results suggest that firms with more concentratedownership and control tolerate less tax risk.

We evaluate the robustness of our results in a variety of ways. First, we compare the tax avoidance of firms with higherrates of managerial stock ownership to the tax avoidance of firms with lower rates of managerial stock ownership, utilizingsettings that either exclude PE-backed firms (e.g., management-owned firms compared to employee-owned firms) orinclude only PE-backed firms (e.g., minority-owned PE-backed firms compared to majority-owned PE-backed firms). In eachcase we find that firms with higher rates of managerial stock ownership avoid less income tax than firms with lower rates ofmanagerial stock ownership, consistent with tax avoidance increasing in the separation of ownership and control. Second,because PE firms may select firms to acquire based on specific observable attributes, we also perform our tests based onpropensity score matched samples of management-owned and PE-backed firms, both with and without the Heckman(1979) procedure. Inferences based on results for these tests are substantially similar to those for tests that only utilize theHeckman (1979) procedure.

We next consider a competing explanation for why management-owned firms avoid less income tax than PE-backedfirms. Given their reputation for reducing portfolio firms' costs of debt with lenders (e.g., Kaplan and Stomberg 2009;Demiroglu and James, 2010; Ivashina and Kovner, 2011), PE firms may also be able to reduce portfolio firms' marginal costsof tax avoidance, resulting in greater tax avoidance at PE-backed firms than at management-owned firms.7 We examine thiscompeting explanation by identifying subsets of firms that likely have lower marginal costs of tax avoidance. We firstpartition PE-backed firms based on whether the PE-backed private firm is owned by: (1) a PE firm that owns many vs. fewerportfolio firms, and (2) a large vs. a small PE firm (based on total capital under PE firm management). We predict privatefirms owned by PE firms that own many portfolio firms avoid more income tax than private firms owned by PE firms thatown fewer portfolio firms. We also predict that private firms owned by large PE firms avoid more income tax than privatefirms owned by smaller PE firms. Our predictions are based on PE firms' economies of scale and scope, which should reduce

3 For the remainder of this paper we refer to firms with private equity and public debt as “private” firms and firms with public equity and public debtas “public” firms. We note that sample firms are on average larger, have higher credit and earnings quality, and are financially stronger than private firmsthat do not issue public debt (Cantillo and Wright, 2000; Denis and Mihov, 2003; Bharath et al., 2008; Katz, 2009; Givoly et al., 2010).

4 PE firms, such as The Blackstone Group, The Carlyle Group, and Kohlberg Kravis & Roberts, manage investment funds that generally buy maturebusinesses via leveraged buyout transactions.

5 To illustrate, the mean (median) proportion of stock owned by managers at management-owned firms is 66.4 (79.4)% but just 8.9 (4.2)% at PE-backedprivate firms (see Table 2). In contrast, the mean (median) proportion of stock owned by managers at S&P 1500 public firms is 5.7 (2.5)%, with aninterquartile range of 3.8% (based on ownership data obtained from ExecuComp, 1992–2010).

6 If PE firm acquisition choices are correlated with target firm tax planning, then PE firm ownership could be endogenously related to tax avoidance inour sample of private firms. Sections 3.2 and 4.7.1 discuss the endogeneity issues that surround our empirical tests and the various methods we adopt tomitigate potential selection bias.

7 We define the marginal costs of tax avoidance in a manner consistent with Brickley et al. (2009), where marginal costs are the incremental costsassociated with a decision (p. 38). In our study, we predict the incremental costs of implementing a particular tax strategy are lower for some PE-backedfirms, due to the economies of scale and scope of their PE firm owners.

B.A. Badertscher et al. / Journal of Accounting and Economics 56 (2013) 228–250230

the marginal costs of tax avoidance at PE-backed firms.8 We then directly disentangle the dual impact of the separation ofownership and control from the marginal costs of tax avoidance by including empirical proxies for these constructs in thesame regression. We continue to find that while higher concentrations of ownership and control are associated with lesscorporate tax avoidance, our proxies for lower marginal costs of tax planning are associated with greater tax avoidance.These results hold in tests that compare tax avoidance at management-owned and PE-backed firms, and also amongstPE-backed firms only.

We then consider a more common proxy for marginal costs – firm size – and examine the tax avoidance of small- vs.large-sized, management-owned and PE-backed firms. To the extent that PE firms are able to reduce the marginal costsof tax avoidance for their portfolio firms (small and large alike), we expect to find larger differences between small-sizedmanagement-owned and PE-backed firms than between large-sized management-owned and PE-backed firms, since largefirms may enjoy economies of scale to tax planning independent of PE ownership (e.g., Rego, 2003; Dyreng et al., 2008). Theresults are consistent with our predictions and indicate that small-sized firms experience the greatest tax savings from PEownership.

To more directly capture inside stock ownership, we hand-collect data on the proportion of stock owned by all namedexecutive officers for our sample of private firms, where available. The results based on the proportion of managerial stockownership for this subsample of private firms, and also within subsamples of management-owned and PE-backed firms,confirm our main findings that firms with more highly concentrated ownership and control avoid less income tax than firmswith less concentrated ownership and control. Finally, we provide exploratory evidence regarding the methods thatPE-backed firms utilize to avoid more income taxes than management-owned firms. Our results broadly suggest that thelower effective tax rates of PE-backed firms are caused – at least in part – by their use of intangible assets, tax-exempt invest-ments, tax credits, and the use of multi-jurisdictional tax planning, including affiliates in low-tax rate foreign countries.

Because PE firms do not randomly select firms to acquire, we acknowledge that our study addresses a joint hypothesis:(i) our model of PE ownership adequately controls for selection bias in our sample of management-owned and PE-backedprivate firms, and (ii) management-owned firms avoid less tax than PE-backed firms. To the extent our model does notadequately capture the determinants of PE ownership, then our results based on the Heckman (1979) and/or propensityscore matching procedures may be unreliable. Nonetheless, we perform numerous other robustness tests (including thealternative approach taken in Larcker and Rusticus, 2010), such that we are confident our results are highly robust.

Our study extends the accounting and finance literatures in several ways. Prior accounting research considers the impactof different organizational factors, including public vs. private ownership (e.g., Mills and Newberry, 2001), family ownership(Chen et al., 2010), and dual-class stock ownership (McGuire et al., 2012) on corporate tax practices, but these studiesprovide disparate evidence on how ownership structure influences corporate tax avoidance. In contrast, we use Fama andJensen's (1983) theory on the separation of ownership and control to understand how one attribute of ownership structurethat is present in all public and private firms impacts corporate tax practices. Our findings are relevant for future researchon the impact of ownership structure on corporate tax avoidance. They also increase our understanding of how PE firmsgenerate value in their portfolio firms. Prior research documents that PE firms create value in their portfolio firms byimplementing effective financial and operating strategies and by actively monitoring top executives at their portfolio firms(e.g., Cao and Lerner, 2009; Kaplan and Stromberg, 2009; Masulis and Thomas, 2009). However, little is known aboutPE-backed firms' tax practices. Given recent criticisms of PE firm investment practices,9 and the growing significance of PEfirms for the U.S. capital markets,10 our study provides new insights on the extent to which PE firms increase portfolio firmvalue by increasing their tax efficiency relative to other private firms.

2. Background and empirical predictions

2.1. The separation of ownership and control and prior tax research

Corporations exhibit substantial variation in the extent to which equity ownership is separated from control overcorporate decision-making. At the extremes, small closely-held corporations have highly concentrated equity ownershipand control, while large publicly-traded corporations have nearly complete separation of equity ownership and control. Theseparation of ownership and control creates well-known agency problems, including managerial incentives to pursue non-value-maximizing behaviors such as shirking, perquisite consumption, and rent extraction. To reduce these agency costs,

8 We utilize the delegation of authority papers by Aghion and Tirole (1997) and Baker et al. (1999) to provide insights into why some PE firm ownersmight retain decision rights over tax planning at PE-backed firms (by requiring portfolio firms to acquire tax services from a particular tax serviceprovider), but delegate authority over day-to-day operations (including the implementation of tax planning) to portfolio firm managers.

9 The rapid growth of the PE industry has raised concerns regarding anticompetitive behavior, excessive tax benefits, and stock manipulations in thissector (see Katz, 2009 and Section 2 for further discussion).

10 The cumulative capital commitments to non-venture capital PE firms in the U.S. between 1980 and 2006 is estimated to be close to $1.4 trillion(Stromberg, 2008). In addition, approximately $400 billion of PE-backed transactions were announced in both 2006 and 2007, representing over 2% of thetotal capitalization of the U.S. stock market in each of these years (Kaplan, 2009). Despite a decline in PE transactions since 2007, experts maintain that PEfirms have become a permanent component of U.S. investment activity (e.g., Kaplan, 2009; Kaplan and Stromberg, 2009).

B.A. Badertscher et al. / Journal of Accounting and Economics 56 (2013) 228–250 231

firms write contracts that align managers' incentives with those of shareholders (e.g., Jensen and Meckling, 1976; Smith andWatts, 1982; Smith and Stulz, 1985).

Fama and Jensen (1983) describe the circumstances in which firms should separate or combine decision managementand decision control with residual risk sharing, where decision management includes the initiation and implementationof decisions by “decision agents” (typically top executives); decision control includes the ratification and monitoring ofdecisions and decision agents (typically by the board of directors); and the residual claimants of a firm (i.e., the commonequity owners) share the residual risk and cash flows of the firm. Fama and Jensen (1983) explain that when residual risksharing is separated from decision management (e.g., in larger organizations), then decision management should also beseparated from decision control to reduce agency costs. In contrast, when decision management and decision control areconcentrated in just a few agents (e.g., in smaller organizations), then residual claims should also be restricted to theseagents. One key factor in their theory is the extent to which equity ownership is concentrated in a few decision agents(i.e., managers). In this case, Fama and Jensen (1983) state that it is rational for the managers to invest in less riskyprojects because their portfolios are likely less diversified than those of managers in organizations with more diffuse equityownership.11

We assume that the diversification of a manager's portfolio and professional reputation are decreasing in the proportionof stock owned in the firm. Thus, greater managerial stock ownership implies greater risk aversion. Consistent withHanlon and Heitzman (2010) and Rego and Wilson (2012), we further argue that tax avoidance is a risky activity in whichundiversified, risk averse managers will minimize their investments. Tax avoidance can impose significant costs on firmsand their managers, including fees paid to tax experts, time devoted to the resolution of tax audits, reputational penalties,and penalties paid to tax authorities. Thus, risk-averse owner-managers likely prefer to undertake less risky tax planning,while relatively risk-neutral shareholders prefer managers to implement all tax strategies that are expected to increase firmvalue, regardless of risk (Rego and Wilson, 2012).

Prior accounting research has examined the impact of different ownership structures on corporate tax practices, but nosingle study has examined how the separation of ownership and control impacts tax avoidance for a broad set of firms.Instead, prior research has investigated tax avoidance at public vs. private firms (e.g., Beatty and Harris, 1998; Mikhail, 1999;Mills and Newberry, 2001), at dual-class stock firms (McGuire et al., 2012), at firms with hedge fund activists (Cheng et al.,forthcoming), and at family-owned firms (Chen, et al. 2010).12 Klassen (1997) documents that when divesting operatingunits, public firms that are subject to higher capital market pressure are more willing to trade-off higher tax costs for thebenefit of higher financial accounting income than public firms subject to less capital market pressure.13 In this study we useFama and Jensen's (1983) theory on the separation of ownership and control to develop empirical predictions for variationin tax avoidance amongst private firms with different ownership structures, all of which are subject to less capital marketpressure than public firms.14

2.2. Private equity firms

Our main empirical tests are based on management-owned and PE-backed private firms. PE firms manage investmentfunds that generally acquire majority control of mature, profitable businesses via leveraged buyout (LBO) transactions. Werefer to these acquired businesses as “portfolio firms” or “PE-backed firms.” Before we develop our empirical predictions, wefirst discuss the organizational structure of PE firms, and then describe how PE firms manage their portfolio firms (i.e., thePE-backed firms). This discussion provides the foundation for several empirical predictions, and ultimately is essential tounderstanding the “ownership and control” of PE-backed firms.

PE firms have received recent attention due to their substantial impact on merger and acquisition activity and theirgenerous tax treatment in the U.S. and other countries. PE firms are typically organized as limited partnerships and most PEfirm executive managers are partners in the PE firm. Thus, we also refer to PE firm managers as “PE firm partners.” PE firmsmanage the PE investment funds that directly acquire mature, profitable businesses via LBO (see Fig. 1). PE funds primarilyfinance portfolio firm acquisitions with the capital contributed by limited partners (i.e., investors in the PE fund) andsubstantial amounts of debt, resulting in highly leveraged portfolio firms. PE firm partners contribute just a small proportion

11 The combination of decision management and decision control with residual risk sharing in a small number of agents also generates “efficiencylosses because decision agents must be chosen on the basis of wealth and willingness to bear risk as well as for decision skills” (Fama and Jensen, 1983,p. 306).

12 Chen et al. (2010) provide evidence consistent with our research question. Specifically, Chen et al. find that family-owned, public firms, whichtypically have more highly concentrated stock ownership than other public companies, avoid less tax than other public firms. However, it is the agencyconflict between the founding family owners and minority shareholders that drives the results in Chen et al. (2010), while it is the concentration ofownership and control that increases the risk aversion of managers and drives the results in our study.

13 Klassen (1997) utilizes inside ownership concentration as his proxy for capital market pressure, where the mean (median) inside ownershipconcentration for his sample of 327 public firms is 15.1 (8.2) percent.

14 For the most part we exclude public firms from our study, since publicly-traded firms are subject to greater financial reporting pressure due togreater scrutiny from investors, analysts, and regulators than private firms, and prior research demonstrates that greater financial reporting pressuredifferentially affects tax avoidance at public and private firms (e.g., Beatty and Harris, 1998; Mikhail, 1999; Mills and Newberry, 2001). Nonetheless, Famaand Jensen's (1983) theory on how the separation of ownership and control should impact a manager's risk aversion can also be applied to a sample thatonly includes public firms. However, public firms generally exhibit substantially less variation in managerial stock ownership compared to our sample ofprivate firms.

Private Equity Firm“Manager” (General Partner)

LP, LLP, LLC

Investors(Limited Partners)

Private EquityInvestment Fund

LP, LLP, LLC

Portfolio Firm #1

Portfolio Firm #2

Portfolio Firm #3

Portfolio Firm #4

80% of gain*

(Upon Sale/IPO of Portfolio Firm)

20% of gain (“carried interest”)(Upon Sale/IPO of

Portfolio Firm)

90 – 95% 5 – 10%

Fig. 1. Diagram of typical organizational structure for a private equity firm with one PE fund and four PE portfolio firms. *Approximately 10% of the totalgain is often distributed to the management team as part of performance-based compensation, reducing the investors' share to approximately seventypercent (Fruhan, 2009).

B.A. Badertscher et al. / Journal of Accounting and Economics 56 (2013) 228–250232

of the PE fund capital (i.e., approximately 1%). The limited partners pay annual management fees (typically 2% of investedcapital) to the PE fund as compensation for PE fund investment operations. The PE fund also receives a 20% share (i.e.,carried interest) of any gains generated by the sale or IPO of portfolio firms (Kaplan and Stromberg, 2009). The taxation of PEfirms and PE firm partners has been criticized as exceedingly unfair.15

The generally negative view of the tax benefits enjoyed by PE firms contrasts other characteristics associated with theirmanagement of portfolio firms. PE firms usually obtain a concentrated ownership stake and control of the board of directorswith the intent of substantially improving portfolio firm performance. Portfolio firm boards are typically comprised of theCEO, PE firm partners, and outside industry experts. Portfolio firms' boards are smaller than comparable public firms' boardsand they meet more frequently via both formal and informal meetings. These board members advise portfolio firmmanagers on strategic considerations, and actively monitor and motivate the management team (Cotter and Peck, 2001;Jensen, 2007; Cornelli and Karakas, 2008; Kaplan and Stromberg, 2009; Masulis and Thomas, 2009). PE firm partnersuse their control over the board of directors to impose performance-based compensation on portfolio firm managers anddo not hesitate to replace them when they underperform (Kaplan and Stromberg, 2009; Acharya et al., 2009). As a result,portfolio firm boards are widely considered more effective than both public and other private company boards (Gilson andWhitehead, 2008; Masulis and Thomas, 2009; Strömberg, 2009). In sum, prior research indicates that PE firms exercisesubstantial control over their portfolio firms' boards of directors and actively monitor the portfolio firm management team.

Large PE firms often hire professionals with operating backgrounds and industry expertise to work with portfolio firmmanagers (Gadiesh and MacArthur, 2008; Acharya et al., 2009). To learn how PE firms influence the tax practices of theirportfolio firms, we spoke with partners at a large public accounting firm that provides tax services to PE-backed firms. Thepartners indicated that PE firms frequently arrange for their portfolio companies to acquire tax services from a specificaccounting firm, with the intention of reducing portfolio firm tax costs through more sophisticated tax strategies thanwould otherwise be used by the portfolio firm (e.g., maximizing the utilization of net operating loss carryforwards and R&Dtax credits). Thus, some PE firms view tax planning as one avenue for increasing portfolio firm value.

While PE firm partners actively monitor portfolio firm operations through their control of portfolio firm boards, theygenerally do not assume management roles in PE-backed firms (e.g., Cao and Lerner, 2009; Kaplan and Stromberg, 2009;Masulis and Thomas, 2009). Instead PE firm partners act as advisors to the portfolio firm management team. In addition, PEfirms typically acquire majority equity stakes in their portfolio companies. This separation of equity ownership (by PE firms)and decision management (by portfolio firm managers) at PE-backed firms leads to an ownership structure that alsoseparates decision management from decision control (by portfolio firm boards).16 In contrast, private firms that are owned

15 While the management fees are generally taxed as ordinary income (i.e., 35% tax rate), the carried interest is taxed as long-term capital gain (i.e., 15%tax rate). This tax treatment of carried interest, as well as the fact that some PE firms have been able to avoid corporate taxation once they file for an initialpublic offering (e.g., The Blackstone Group) has provoked numerous negative press reports, proposed changes to federal income tax laws, and academicstudies on the tax treatment of PE firms (e.g. Fleischer, 2007, 2008; Knoll, 2007; Cunningham and Engler, 2008; Lawton, 2008).

16 This organizational structure is consistent with the prediction of Fama and Jensen (1983) that “when venture equity capital is put into a smallentrepreneurial organization by outsiders, mechanisms for separating the management and control of important decisions are instituted” (footnote 9,p. 306).

B.A. Badertscher et al. / Journal of Accounting and Economics 56 (2013) 228–250 233

by the firm's management often combine decision management, decision control, and equity ownership in a few indivi-duals, which provides the basis for our empirical predictions.

2.3. Empirical predictions

Utilizing a variety of settings where the separation of ownership and control exhibits substantial variation, we empiri-cally test one specific implication of Fama and Jensen's (1983) theory. In particular, we examine whether firms with moreconcentrated ownership and control avoid less income tax than firms with less concentrated ownership and control.We also consider a competing explanation for these findings, that being whether PE-backed firms enjoy lower marginalcosts of tax planning, which in turn facilitate greater income tax avoidance.

2.3.1. Predictions for the separation of ownership and control and tax avoidanceTo test our predictions we utilize a unique sample of private firms with privately-owned equity but publicly-traded debt.

This sample holds financial reporting requirements constant, since all sample firms are required to file financial statementswith the SEC. Nonetheless, sample firms are on average subject to less capital market pressure than similar public firms (e.g.Givoly et al., 2010).17 Our sample also exhibits substantial variation in the separation of ownership and control, making it apowerful setting to test our empirical predictions.

Our primary tests are based on management-owned and PE-backed private firms. Katz (2009) documents that topexecutives at management-owned firms own greater proportions of company stock than top executives at PE-backed firms.As a result, management-owned firms exhibit higher concentrations of ownership and control than PE-backed firms.Consistent with Fama and Jensen (1983), we assume that the diversification of a manager's portfolio (and professionalreputation) is decreasing in the proportion of stock owned in the firm. Thus, the higher concentrations of ownership andcontrol at management-owned firms should cause their owner-managers to be more risk averse and tolerate less tax riskthan managers at PE-backed firms, which leads to our first empirical prediction:

P1. Management-owned firms avoid less income tax than PE-backed private firms.

Prior research provides additional insights into variation in the separation of ownership and control at firms withdifferent ownership structures. Amongst PE-backed firms, Katz (2009) demonstrates that the proportion of stock owned bytop executives at minority-owned, PE-backed firms is significantly greater than managerial stock ownership at majority-owned, PE-backed firms. With respect to other types of firms, Kaplan and Stromberg (2009) and Acharya et al. (2010) assertthat CEOs at PE-backed firms typically own larger proportions of portfolio firm stock than CEOs of public firms, while Bovaet al. (2012a) and Bova et al. (2012b) state that employee-owned private firms generally have stock ownership thatis diffused across many individuals. Taken together, these studies suggest that managerial stock ownership rates varysystematically across firms with different ownership structures and lead to the following empirical predictions that buildon P1:

P1a. Minority-owned, PE-backed firms avoid less income tax than majority-owned, PE-backed firms.

P1b. Management-owned firms avoid less income tax than employee-owned firms.

P1c. Management-owned firms avoid less income tax than public firms.

These predictions are based on Fama and Jensen's (1983) theory that managers at firms with high concentrations ofownership and control likely have less diversified portfolios and thus should be more risk averse than managers at firmswith less concentrated ownership and control, all else equal. We predict greater managerial risk aversion should lead to lessincome tax avoidance.

2.3.2. Predictions for the marginal costs of tax avoidance at PE-backed firmsIt is possible that PE-backed private firms are fundamentally different from management-owned firms (beyond the

differences in ownership and control) and these differences influence the tax practices at management-owned andPE-backed firms. One specific attribute that would allow PE-backed firms to avoid more income taxes than management-owned firms involves the marginal costs of tax avoidance at PE-backed firms. Prior theoretical research examines thecircumstances in which a principal is likely to delegate authority (either formal or informal) to an agent. These studies findthat the principal is likely to retain authority over decision making when the principal is better informed than the agent

17 To address concerns that PE-backed firms are subject to different financial reporting incentives than management-owned firms, since they aretypically sold or taken pubic via IPO within 5–7 years of being purchased, we re-run our main tests separately for the sub-groups of private firms that (1)eventually go public and (2) that once were public but then go private. That is, we first compare the tax avoidance of management-owned and PE-backedfirms during the first five private firm-years (if available) after transitioning from public ownership. We then compare the tax avoidance of management-owned and PE-backed firms during the last five private firm-years (if available) prior to transitioning to public ownership. Our results (untabulated)confirm that management-owned firms avoid less tax than PE-backed firms.

B.A. Badertscher et al. / Journal of Accounting and Economics 56 (2013) 228–250234

(Aghion and Tirole, 1997; Baker et al., 1999).18 In our research setting, PE firms (and PE firm general partners) can beconsidered the “principals” in the authority literature, while portfolio firm management teams are the “agents.” From thisperspective, we can evaluate the extent to which PE firms are likely to “retain authority” over tax planning at their portfoliocompanies. PE firms have substantial experience in owning and monitoring a broad set of portfolio companies and theirtax strategies. Thus, in effect PE firms are better informed than portfolio firm managers about a broad range of tax planningopportunities and extant theory on formal and informal authority in organizations would suggest that PE firms are likely toretain authority over tax planning at portfolio companies. In contrast, PE firms are not likely to retain authority over mostportfolio firm operating decisions, since portfolio firm managers are typically better informed than PE firm partners withrespect to day-to-day operating decisions.

Our understanding is that many PE firms effectively retain decision rights over tax planning at PE-backed firms byarranging tax service providers for their portfolio firms. Thus, because PE firms typically own more than one portfoliofirm, PE firms should be able to reduce the marginal costs of tax avoidance at PE-backed firms by applying similar taxplanning strategies at more than one PE-backed firm and/or by negotiating lower fees with tax service providers onbehalf of their portfolio firms. Lower marginal costs of tax avoidance at PE-backed firms would be consistent with PEfirms' reputation for reducing portfolio firms' costs of debt with lenders (e.g., Kaplan and Stomberg, 2009; Demirogluand James, 2010; Ivashina and Kovner, 2011). Thus, we posit that PE firms have the ability to generate economies ofscale and scope for tax avoidance at PE-backed firms. Conversations with tax partners at a large public accounting firmare consistent with this assertion. These partners explained that some (but not all) PE firm clients effectively retaindecision rights with respect to tax planning at PE-backed firms by arranging a particular tax service provider for most orall of their portfolio firms.19 The centralization of tax accounting services should reduce the marginal costs of taxplanning at PE-backed firms, resulting in greater tax avoidance at PE-backed firms relative to management-ownedfirms.20 Thus, we also examine the extent to which variation in the marginal costs of tax planning impact tax avoidanceat private firms.

To identify which PE-backed firms likely possess lower marginal costs of tax planning while also holding the separationof ownership and control relatively constant, we first restrict our analyses to only PE-backed firms. Within this subsample,we assert that firms owned by PE firms with “many” portfolio companies are likely to have lower marginal costs of taxplanning than firms owned by PE firms with “fewer” portfolio companies. We classify a PE firm as having “many” portfoliofirms if they own more than 200 portfolio firms and their average equity investment is greater than $30 million. We classifyall PE firms not meeting these two requirements as having “fewer” portfolio firms. PE firms that own many portfoliocompanies should enjoy economies of scale and scope with respect to tax planning costs at their portfolio firms, since thesame tax planning strategies can potentially be utilized at a larger number of portfolio firms. Thus, our next empiricalprediction is:

P2a. Firms that are owned by PE firms with more portfolio firms avoid more income tax than firms that are owned by PEfirms with fewer portfolio firms.

Consistent with the discussion above, we also partition PE-backed firms based on whether they are owned by largeor small PE firms, where “large” PE firms include the 15 largest PE firms as measured by total capital under PE firmmanagement during our sample period.21 We classify all other PE firms as “small” PE firms. We expect firms that are ownedby large PE firms to have lower marginal costs of tax planning than firms that are owned by small PE firms, since large PEfirms should enjoy economies of scale and scope with respect to tax planning costs at their portfolio firms. Indeed, priorresearch shows that large PE firms regularly outperform smaller PE firms, consistent with a greater ability to create financialvalue through operational improvements at portfolio firms (e.g., Kaplan and Schoar, 2005; Acharya et al., 2009). Thus, ournext empirical prediction is:

P2b. PE-backed firms that are owned by large PE firms avoid more income tax than PE-backed firms that are owned bysmaller PE firms.

We then consider variation in the marginal costs of tax planning amongst management-owned and PE-backedfirms. Within this larger sample, we continue to predict firms that are owned by large PE firms and/or PE firmswith more portfolio firms avoid more tax than other private firms due to lower marginal costs of tax planning. We also

18 Aghion and Tirole (1997) claim that asymmetric information is the key to understanding the delegation of authority. They also explain that formalauthority is likely to be delegated for decisions that are (1) relatively unimportant for the principal, (2) but important to the agent, (3) for which theprincipal can trust the agent, and (4) are sufficiently innovative that the principal does not have substantial experience or competency.

19 These partners also stated that PE firms similarly reduce other portfolio firm costs by centralizing certain administrative services for their portfoliocompanies. For example, some PE firms require their portfolio firms to purchase legal services from specific law firms and insurance services from specificinsurance firms.

20 However, tax services must be tailored to fit the particular needs of each portfolio company and so it is not clear that PE-backed firms truly enjoylower marginal costs of tax planning.

21 The 15 largest PE firms in our sample are Carlyle Group, Blackstone Group, Warburg Pincus, Kohlberg, Kravis, Roberts and Company, Goldman Sachsand Company, Cerberus Capital Management, Fortress Investment Group, Apollo Global Management, Bain Capital, TPG Capital, 3i Group, Apax PartnersWorldwide, Thomas H. Lee Partners, Morgan Stanley Private Equity, and Welsh, Carson, Anderson, and Stowe.

B.A. Badertscher et al. / Journal of Accounting and Economics 56 (2013) 228–250 235

consider a more common proxy for marginal costs – firm size – and examine the tax avoidance of small- vs. large-sized,management-owned and PE-backed firms. To the extent that PE firms reduce the marginal costs of tax planning bycentralizing tax services for their portfolio firms (small and large alike), we expect to find greater differences between small-sized, management-owned and PE-backed firms than between large-sized, management-owned and PE-backed firms,since large-sized firms may enjoy economies of scale to tax planning independent of PE ownership (e.g., Rego, 2003;Dyreng et al., 2008).22 We define small-sized (large-sized) firms as those in the lowest (highest) quartile of net sales for oursample of private firms and predict:

P2c. The difference in tax avoidance at small-sized, PE-backed and management-owned firms is larger than the differencein tax avoidance at large-sized, PE-backed and management-owned firms.

Lastly, we attempt to disentangle the dual impact of the separation of ownership and control from the marginal costs oftax planning on corporate tax avoidance. Specifically, we hand-collect managerial stock ownership data for our samples ofmanagement-owned and PE-backed firms, which allows us to include empirical proxies for a manager's risk aversion andthe marginal costs of tax planning in the same regression. As previously explained, we assume that the diversification of amanager's portfolio is decreasing in the proportion of stock owned in the firm and lower diversification leads to greater riskaversion. Our final empirical prediction is:

P3. Holding the marginal costs of tax avoidance constant, private firms with managers that own larger proportions of thefirm's stock avoid less income tax than private firms with managers that own smaller proportions of the firm's stock.

3. Research design

3.1. Measures of corporate tax avoidance

We rely on several measures of tax avoidance because different measures capture different aspects of corporate taxplanning. Our first two measures are based on effective tax rates and include GAAP_ETR and CASH_ETR, where GAAP_ETR(CASH_ETR) is total tax expense (cash taxes paid) summed over three years, scaled by adjusted pretax income summed overthree years.23 Both measures convey a firm's average tax cost per dollar of pretax income and capture a broad range of taxplanning activities that can have both certain and uncertain outcomes with tax authorities. Recent research presentsevidence that both effective tax rate measures reflect variation in tax avoidance across firms (Dyreng et al., 2008).

We complement these effective tax rate measures with two additional measures designed to capture more risky taxavoidance: Frank et al.'s (2009) discretionary permanent book-tax difference measure (DTAX) and Wilson's (2009) measureof tax sheltering (SHELTER). While DTAX is the residual from a regression of permanent book-tax differences on non-discretionary sources of those differences,24 SHELTER is the predicted value from a tax shelter prediction model. Frank et al.(2009) demonstrate that DTAX is significantly associated with actual cases of tax sheltering and Wilson (2009) demonstratesthat SHELTER is able to predict tax shelter activity out-of-sample. See the Appendix for details on how we calculate each ofthese measures.

We acknowledge that all four measures reflect income tax avoidance with error. While the effective tax rate measures arecommonly used in accounting research and understood by a broad set of financial statement users, they capture all types oftax avoidance (i.e., risky and non-risky strategies alike). Moreover, GAAP_ETR is confounded by changes in tax reserves andthe valuation allowance, while CASH_ETR is confounded by the timing of tax payments, settlements with tax authorities, andsome types of earnings management. In contrast, DTAX and SHELTERwere designed to capture more risky tax avoidance, andin fact both measures are associated with tax shelter transactions (Frank et al., 2009; Wilson, 2009). But DTAX only capturestax strategies that generate permanent book-tax differences (i.e., not those that create temporary book-tax differences) andboth DTAX and SHELTER are based on cross-sectional empirical models that are subject to criticisms similar to those directedat discretionary accrual models (i.e., the models estimate tax avoidance with error). None of the four measures are clearlysuperior (or inferior) to the other three. Consequently, we rely on all four measures in our empirical tests to evaluate therobustness of our results.

22 Note that the term “large PE” firm refers to economies of scale and scope of the PE firm, while the term “large-sized PE-backed” firm refers toeconomies of scale and scope of the PE-backed firm.

23 Whenever possible we use three years of data to calculate GAAP_ETR and CASH_ETR. However, if data limitations (such as transition years or missingvalues) prohibit us from using three years of data, we next use two years, followed by one year of data. Results are qualitatively similar if we base ourcalculations on one year of data.

24 GAAP_ETR also reflects variation in permanent book-tax differences, where permanent book-tax differences are differences between financial andtaxable income that do not reverse through time (e.g., interest income from municipal bonds is exempt from federal income taxation but included in pre-tax financial income). DTAX is distinct from GAAP_ETR because Frank et al.'s (2009) model is designed to remove non-discretionary sources of permanentbook-tax differences from GAAP_ETR to isolate intentional, more aggressive tax avoidance.

B.A. Badertscher et al. / Journal of Accounting and Economics 56 (2013) 228–250236

3.2. Modeling the impact of separation of ownership and control on corporate tax avoidance

To investigate whether the separation of ownership and control impacts corporate tax avoidance we utilize a sample ofprivate firms that are owned by either PE firms or by the firm's management team. Because PE firms do not randomly selectfirms to acquire, selection bias may be present in our empirical tests. Specifically, if PE firm acquisition choices are correlatedwith target firm tax planning, then PE firm ownership could be endogenously related to tax avoidance in our sample. Thisendogeneity would cause OLS coefficient estimates to be biased. To correct for selection on unobservable differencesbetween management-owned and PE-backed firms, we follow the Heckman (1979) procedure and implement a treatmenteffect model (Lennox et al., 2012). We perform this two-stage estimation procedure for all regression analyses that include bothmanagement-owned and PE-backed private firms. In the first stage, we use Lee's (1979) switching simultaneous equation (seeMaddala, 1983, Chapter 9) to estimate the following probit regression, which predicts whether a private company is owned by aPE firm:

PE_BACKEDi;t ¼ β0þβ1Q_RATIOi;tþβ2OPER_CYCLEi;tþβ3FIRM_AGEi;tþβ4RNOAi;t

þβ5LOSSi;tþβ6NOLi;tþβ7LEVi;tþβ8MNCi;tþβ9INTANGi;t

þβ10EQ_EARNi;tþβ11SALES_GRi;tþβ12AB_ACCRi;tþβ13SOXi;t

þβ14ASSETSi;tþ∑tαtYEARtþ∑kαkINDUSiþεi;t ð1ÞThis equation is based in part on existing models of private investor financing and PE ownership. Importantly, it includes

three exclusion restriction variables (i.e., Q_RATIO, OPER_CYCLE, and FIRM_AGE) that are significant predictors of PEownership but because we do not expect these variables to be directly related to corporate tax avoidance, they are excludedfrom the second stage regression.25 Prior research finds that PE firms tend to acquire firms with lower risk of financialdistress (e.g., Opler and Titman, 1993). Thus, consistent with models of PE and private ownership in Ball and Shivakumar(2005), Katz (2009), and Givoly et al. (2010), we select both Q_RATIO and OPER_CYCLE as exclusion restriction variables foreq. (1). FIRM_AGE is the third exclusion restriction based on recent research that indicates younger firms are more likely tobe taken private because they fail to attract investor recognition as a public firm (Mehran and Peristiani, 2010).26

The remaining equation (1) variables are primarily included because the Heckman (1979) procedure requires that allsecond stage regression variables also be included in the first stage regression. We note, however, that prior research on PEfirm ownership finds that PE firms generally acquire targets that are growing (SALES_GR), have greater profitability (RNOA,LOSS, NOL) but lower leverage (LEV) (e.g., Acharya et al., 2009; Aslan and Kumar, 2011). Hence, prior research suggests thesevariables are also significant determinants of PE ownership.

We compute the inverse Mills' ratio (INV_MILLS) for each firm-year observation based on the estimated coefficients forEq. (1), and then include that variable in Eq. (2), the second stage of the Heckman estimation procedure:

TAXi;t ¼ α0þα1MGMT_OWNEDi;tþα2RNOAi;tþα3LOSSi;tþα4NOLi;tþα5LEVi;t

þα6INTANGivþα7MNCi;tþα8AB_ACCRi;tþα9EQ_EARNi;tþα10SALES_GRi;t

þα11ASSETSi;tþα12SOXi;tþα13INV_MILLSi;tþ∑tatYEARtþ∑kakINDUSiþεi;t ð2ÞP1 predicts that management-owned firms avoid less income tax than PE-backed firms. Thus, the variable of interest inEq. (2) is MGMT_OWNED, which is an indicator variable equal to ‘one’ if a firm is majority-owned by its current and pastnamed executive officers, and ‘zero’ if otherwise. The dependent variable, TAX, represents the four proxies for corporatetax avoidance: GAAP_ETR, CASH_ETR, DTAX, and SHELTER. If management-owned firms avoid less tax than PE-backed firms,then the coefficient on MGMT_OWNED should be significant and positive (negative) in regressions where GAAP_ETR andCASH_ETR (DTAX and SHELTER) are the dependent variables. See the Appendix for detailed definitions of each variableincluded in Eqs. (1) and (2).

Eq. (2) also includes controls for factors that influence a firm's tax avoidance activity, as documented by prior research(e.g., Manzon and Plesko, 2002; Rego, 2003; Dyreng et al., 2008; Frank et al., 2009; Wilson, 2009; Chen et al., 2010). The firstset of control variables, which includes RNOA, LOSS, NOL, and LEV, controls for a firm's need to tax plan. We include anindicator variable, LOSS, and the return on net operating assets (RNOA) as proxies for current profitability, since profitablefirms have greater incentives to tax plan. We include an indicator variable for the presence of net operating losscarryforwards (NOL) at the beginning of the year, since firms with loss carryforwards have less incentive to engage incurrent year tax planning. We include a firm's leverage ratio (LEV) because firms with greater leverage have less need to taxplan due to the tax benefits of debt financing.

We include an indicator variable for foreign operations (MNC) in Eq. (2), since firms with foreign operations have greateropportunities for tax avoidance by shifting income between high and low tax rate locations (e.g., Rego, 2003).MNC equals ‘one’ if afirm reports non-zero foreign income or foreign tax expense, and zero if otherwise. We control for intangible assets (INTANG) and

25 For the selection model to effectively control for endogeneity, the exclusion restriction variables must be important determinants of the dependentvariable in the selection model (i.e., PE_BACKED) but unrelated to the dependent variable in the treatment effect model (i.e., TAX) (Lennox et al., 2012). Wedo not expect our exclusion restriction variables to be related to TAX, and in fact, when we include them in Eq. (2) (results untabulated), none of thecoefficients on the exclusion restriction variables are significant.

26 Mehran and Peristiani (2010) also find that the ratio of total income tax expense to net sales is not associated with likelihood of a firm being takenprivate in a LBO transaction. We confirm their results in our research setting (see discussion in Section 4.7.1).

B.A. Badertscher et al. / Journal of Accounting and Economics 56 (2013) 228–250 237

equity in earnings of unconsolidated affiliates (EQ_EARN) because these items often generate differences between book and taxableincome and can thus affect our tax avoidance measures. We include sales growth (SALES_GR) in Eq. (2) because growing firmslikely make larger investments in depreciable assets, which generate larger temporary book-tax differences and can thus affectsome tax avoidance measures. We control for firm size (ASSETS) because large firms likely enjoy economies of scale in tax planning.We include an indicator variable for years following the Sarbanes–Oxley Act of 2002 (SOX), since prior research demonstrates thatthe regulatory environment surrounding corporate financial and tax reporting changed substantially in the post-SOX time period(e.g. Cohen et al., 2008). We further include year (YEAR) and industry (INDUS) fixed-effects to control for fundamental differences intax planning that may exist across years and industries.

Frank et al. (2009) find a strong positive relation between financial and tax reporting aggressiveness. Katz (2009)documents that PE-backed firms report more conservatively and engage in less earnings management compared to non-PE-backed firms. To the extent our test and control firms exhibit different financial reporting quality, we need to controlfor financial reporting quality in Eq. (2). Thus, we control for both timely loss recognition and earnings managementby including AB_ACCR in Eq. (2).27 AB_ACCR is the amount of abnormal accruals after controlling for conservatism inour abnormal accruals calculation (see Ball and Shivakumar, 2006). Our last control variable is the inverse Mills ratio(INV_MILLS) calculated based on the first stage of the Heckman (1979) procedure.

4. Sample selection and empirical results

4.1. Sample selection

Our initial sample consists of private firms that have publicly-traded debt. Because their debt is public, these firms mustfile financial statements with the SEC, even though their equity is privately-held. We follow Katz (2009) and select all firm-year observations on Compustat in any of the 31 years from 1980 through 2010 that satisfy the following criteria: (1) thefirm's stock price at fiscal year-end is unavailable, (2) the firm has total debt as well as total annual revenues exceeding $1million, (3) the firm is a domestic company, (4) the firm is not a subsidiary of another public firm, and (5) the firm is not afinancial institution or in a regulated industry (SIC codes 6000–6999 and 4800–4900). To ensure that the sample includesonly private firms with public debt, we examine each firm and remove public firm observations (details provided in Table 1,Panel A). We further categorize each firm as being in one of the following categories: (1) management-owned, defined asfirms that do not have a PE sponsor and are at least 50% owned by founders, current and past named executive officers,and/or their families, (2) PE majority-owned, defined as firms whose equity is majority-owned (i.e., more than 50%)by PE firms, according to Thomson Financials VentureXpert, and (3) PE minority-owned, defined as firms whose equityis minority-owned (i.e., less than or equal to 50%) by PE firms. The resulting sample consists of 2628 private firm-yearobservations and 549 private firms.

Table 1, Panel B, presents the industry composition of our sample of private firms with public debt (i.e., the 2628 firm-year observations in Panel A). Our sample of private firms with public debt is generally consistent with the broaderCompustat population over the same time period. Only the proportion of private firms classified as retail firms is signifi-cantly different from the Compustat population (25.1 vs. 9.4%).

We hand-collect managerial stock ownership data for our sample of private firms (where available) as an alternativeproxy for the separation of ownership and control. Specifically, we hand-collect from SEC filings the total amount of stockowned by all named executive officers.28 We then calculate the proportion of all outstanding common shares owned bythese executive officers and refer to this variable as MGR_STOCK. Because managerial stock ownership data is only availablefor 374 of our 549 private firms, we use MGR_STOCK as our secondary proxy for the separation of ownership and control,while MGMT_OWNED is our primary proxy.

4.2. Descriptive statistics on stock ownership at private firms

Table 2 presents statistics on the proportions of stock owned by PE firms, all named executive officers (i.e., Managers),and CEOs for 9 different categories of private firms, including different types of PE-backed and employee-owned privatefirms.29 Asterisks indicate significant differences between the mean and median amounts of stock owned by PE firms/Managers (including CEOs)/CEOs Only at management-owned firms (row 1) compared to other types of private firms(rows 2–9). Overall, the results in Panel A indicate that managers not only own the majority of stock in management-ownedfirms (by definition), but their percentage stock ownership (mean¼66.4%, column 2) is also substantially larger atmanagement-owned firms compared to all other types of firms. Amongst PE-backed firms, managers own greater propor-tions of stock at firms that are minority-owned by PE firms (mean¼29.9%, row 4) than at other PE-backed firms. In fact,

27 Results (untabulated) are substantially similar if we replace AB_ACCR with the absolute value of AB_ACCR.28 For each firm, we collect stock ownership data for only one firm-year and assume stock ownership remains relatively constant all years the firm

remains in our sample, unless we determine that the firm experienced a change in ownership structure. In this case we collect stock ownership data for atleast 1 year after the change in ownership structure.

29 Employee-owned firms are private firms that do not have a PE sponsor and whose equity is more than 50% owned by employees. These firms areexcluded from most analyses, except Tables 2 and 5, Panel D.

Table 1Sample selection procedures for private firms with public debt (1980–2010).

Panel A: Private firms with public debt

No. of firm-years No. of firms

“Potential” private firms with public debt (Compustat)a 14,190 3699Eliminate firms that:Do not have historical (non-prospectus) datab (3634) (1475)Are public firms (2357) (380)Are subsidiaries of public firms (585) (108)Are public spin-offs (111) (34)Are involved in bankruptcy proceedings (306) (104)Have insufficient information (1683) (344)Are foreign firms (848) (226)Otherc (957) (400)

Subtotal of private firms with public debt 3709 628Eliminate firms that:Are cooperatives, LPs, government-owned, and firms forwhich ownership structure cannot be ascertained

(1081) (79)

Private firms with public debt: 2628 549Private firms that are majority-owned by PE firms 1559 350Private firms that are minority-owned by PE firms 312 71Private firms that are owned by management 757 128

Panel B: Industry classification for private firms with public debt

Industry classification Firm-years Sample% Compustat%

Agriculture 5 0.20% 0.40%Mining & construction 22 0.80% 3.60%Food 89 3.4% 4.20%Textiles & printing/publishing 355 13.50% 10.10%Chemicals 128 4.90% 4.90%Pharmaceuticals 25 1.00% 3.30%Extractive 61 2.30% 5.70%Durable manufacturers 751 28.60% 32.30%Computers 86 3.30% 8.30%Transportation 67 2.50% 4.10%Utilities 0 0.0% 0.10%Services 380 14.50% 13.60%Retail 659 25.10% 9.40%

Total observations 2628

Industry classification is determined by primary SIC code as follows: Agriculture (0100–0999), Mining & Construction (1000–1999, excluding 1300–1399),Food (2000–2111), Textiles & Printing/Publishing (2200–2780), Chemicals (2800–2824, 2840-2899), Pharmaceuticals (2830–2836), Extractive (2900–2999,1300–1399), Durable Manufactures (3000–3999, excluding 3570–3579 and 3670–3679), Computers (7370–7379, 3570–3579, 3670–3679), Transportation(4000–4899), Utilities (4900–4999), Retail (5000–5999), and Services (7000–8999, excluding 7370–7379).

B.A. Badertscher et al. / Journal of Accounting and Economics 56 (2013) 228–250238

mean (median) managerial stock ownership at majority-owned, PE-backed firms (row 3) is just 7.0 (3.7) percent. We alsonote that CEOs (column 3) account for the majority of stock ownership by Managers (column 2) regardless of firm type. Weconclude that our sample of private firms exhibits substantial variation in managerial stock ownership, making it a powerfulsetting to examine the impact of the separation of ownership and control on corporate tax avoidance.

To more closely evaluate the separation of ownership and control at PE-backed vs. management-owned firms, we hand-collect data on board composition and CEO characteristics from SEC filings and the BoardEx database. To minimize the hand-collection process, we randomly select three minority PE-backed firms for each year in our sample and match them withboth majority PE-backed and management-owned firms in the same year and same four four-digit SIC code. Descriptivestatistics based on this hand-collected data (results untabulated) indicate that while 57% of board members at management-owned firms are insiders, the proportions of insiders on boards at minority- and majority-owned, PE-backed firms aresignificantly smaller at 45% and 30%, respectively. In fact, PE firm representatives account for 62 (39) percent of the boardmembership at majority-owned (minority-owned) PE-backed firms. The chairman of the board is a representative of the PEfirm owner 29 (48) percent of the time, and the CEO is either nominated by or is affiliated with the PE firm owner 58 (44)percent of the time at majority- (minority-)owned PE-backed firms. These statistics clearly demonstrate PE firms' abilities tomonitor and control portfolio firms' management and boards of directors, consistent with the discussion in Section 2.2.

Table 2Stock ownership data for management-owned and PE-backed private firms.

N Percentage of stock owned by:

PE firms Managers (including CEOs) CEOs only

Mean (%) Median (%) Mean Median Mean Median

(1) MGMT_OWNED 92 0.0 0.0 66.4 79.4 52.3 53.8(2) ALL PE_BACKED 282 79.7nnn 83.8nnn 8.9nnn 4.2nnn 5.6nnn 2.0nnn

(3) MAJORITY_PE 258 83.9nnn 85.1nnn 7.0nnn 3.7nnn 4.3nnn 1.8nnn

(4) MINORITY_PE 24 34.7nnn 38.2nnn 29.9nnn 20.7nnn 19.5nnn 13nnn

(5) MANY_PE 49 83.3nnn 87.8nnn 6.5nnn 3.6 nnn 4.6 nnn 2.1 nnn

(6) FEWER_PE 233 79.9nnn 83.2nnn 9.4nnn 4.3 nnn 5.8 nnn 2.0 nnn

(7) LARGE_PE 89 85.4 nnn 87.6 nnn 5.1nnn 2.7 nnn 3.5 nnn 1.5 nnn

(8) SMALL_PE 193 77.0nnn 80.9nnn 10.6nnn 5.1nnn 6.6nnn 2.4nnn

(9) EMPLOYEE_OWNED 13 0.0 0.0 7.2nnn 2.6nnn 3.6nnn 1.2 nnn

n, nn, nnn indicate significance at the 10%, 5%, and 1% level, respectively. Differences in means are tested for significance using a two-tailed t-test; differencesin medians are tested for significance using a two-tailed Wilcoxon signed rank test. An asterisk indicates that the percentage of stock owned by PE firms(column a)/Management (column b)/and CEOs (column c) at MGMT_OWNED firms (row 1) is significantly different than the corresponding percentage ofstock owned by PE firms (column a)/Management (column b)/ and CEOs (column c) at each of the other firm types (rows 2–8). For instance, the meanamount of stock owned by Management at MGMT_OWNED firms (row 1, column b: 66.4%) is significantly different than the mean amount of stock ownedby Management at EMPLOYEE_OWNED firms (row 2, column b: 7.2%).

B.A. Badertscher et al. / Journal of Accounting and Economics 56 (2013) 228–250 239

4.3. Results for the separation of ownership and control and tax avoidance

The evidence in Table 2 indicates that the rates of managerial stock ownership differ significantly between management-owned and PE-backed private firms. Fama and Jensen (1983) assert that when equity ownership and corporate decision-making are concentrated in just a small number of decision-makers, these owner-managers will likely be more risk averseand less willing to invest in risky projects, which we argue includes income tax avoidance. To examine the impact of theseparation of ownership and control on corporate tax avoidance at sample firms, we perform the Heckman (1979) two-stageprocedure to correct for selection on unobservable differences between management-owned and PE-backed firms.

Table 3, Panel A, presents the mean and median values for all variables included in the second stage of the Heckman(1979) procedure. The statistics for our four measures of tax avoidance (GAAP_ETR, CASH_ETR, DTAX, and SHELTER) uniformlysuggest that management-owned firms avoid less income tax than PE-backed private firms. Specifically, the mean andmedian amounts of GAAP_ETR and CASH_ETR are statistically higher, while the mean and median amounts of DTAXand SHELTER are significantly lower for management-owned firm-years. These results are consistent with P1, which predictsthat management-owned firms avoid less income tax than PE-backed private firms. Panel A also indicates thatmanagement-owned firms are significantly different from PE-backed firms in many respects, including higher profitability(RNOA), fewer tax losses (LOSS and NOL), lower leverage (LEV), fewer intangibles (INTANG), less foreign operations (MNC),higher abnormal accruals (AB_ACCR), and fewer total assets (ASSETS). In supplemental analyses we also perform a propensityscore matching procedure to correct for any possible selection bias based on observable differences between management-owned and PE-backed firms (see Section 4.7.1).

Table 3, Panel B, presents Pearson and Spearman correlations between the MGMT_OWNED indicator variable and eachmeasure of tax avoidance. Consistent with Panel A, the correlations in Panel B indicate that management-owned firms avoidless tax than PE-backed firms. In addition, most of the correlations between the measures of tax avoidance are as expected.In particular, the ETR measures are positively correlated with each other, while DTAX and SHELTER are positively correlatedwith each other. However, CASH_ETR is not correlated with DTAX, perhaps because the latter measure is designed to capturemore risky tax avoidance.

Table 4, Panel A, presents results for the first stage of the Heckman (1979) procedure. Recall that Q_RATIO, OPER_CYCLE,and FIRM_AGE are exclusion restriction variables that are excluded from the second stage regression.30 The coefficients oneach of these variables are statistically significant, as are most of the coefficients on the other independent variables in thefirst stage regression (except for the coefficients on RNOA, EQ_EARN, SALES_GR, and SOX). Untabulated results indicate thatinclusion of the three exclusion restriction variables increases the MacKelvey–Zavonia pseudo-R-squared from 57.8% to60.2% and increase McFadden's LRI pseudo-R-squared from 18.4% to 21.1% (see tabulated results at bottom of Panel A). Theselatter statistics are larger than the 18–19% reported in related studies (e.g. Lee and Wahal, 2004; Morsfield and Tan, 2006).We conclude that the exclusion restriction variables and our model have significant explanatory power in predicting PEownership.

30 To confirm that the exclusion restriction variables (Q_RATIO, OPER_CYCLE, and FIRM_AGE) are not associated with tax avoidance, in untabulatedrobustness tests we include these variables in the second stage regressions. None of the estimated coefficients are significant in any of the four taxavoidance regressions.

Table 3Descriptive statistics that compare the tax and non-tax characteristics of management-owned private firms and PE-backed private firms.

Panel A: Comparison of tax avoidance and non-tax characteristics

Management-owned private firm-years PE-backed private firm-years Difference between

# Obs Mean Median # Obs Mean Median Means Medians

GAAP_ETR 563 0.360 0.342 1138 0.305 0.308 0.055nn 0.034nnn

CASH_ETR 419 0.329 0.299 1022 0.275 0.279 0.054nnn 0.020nn

DTAX 757 0.020 �0.002 1871 0.071 0.018 �0.050nnn �0.020nn

SHELTER 757 �1.198 �1.212 1871 �0.986 �1.012 �0.212nnn �0.200nnn

RNOA 757 0.142 0.125 1871 0.106 0.098 0.037nnn 0.026nn

LOSS 757 0.336 0.000 1871 0.559 1.000 �0.223nnn �1.000nnn

NOL 757 0.231 0.000 1871 0.365 0.000 �0.133nnn 0.000LEV 757 0.581 0.592 1871 0.706 0.665 �0.125nnn �0.073nnn

INTANG 757 0.129 0.008 1871 0.279 0.204 �0.150nnn �0.196nnn

MNC 757 0.292 0.000 1871 0.464 0.000 �0.173nnn 0.000AB_ACCR 757 0.002 0.000 1871 �0.030 �0.019 0.031nnn 0.019n

EQ_EARN 757 0.007 0.000 1871 �0.001 0.000 0.008 0.000SALES_GR 757 0.239 0.034 1871 0.293 0.026 �0.054 0.008ASSETS 757 5.774 5.723 1871 6.021 5.998 �0.247nnn �0.275nnn

SOX 757 0.221 0.000 1871 0.262 0.000 �0.042 0.000

Panel B: Pearson (Spearman) Correlation Coefficients for MGMT_OWNED and Tax Avoidance Measures

MGMT_OWNED GAAP_ETR CASH_ETR DTAX SHELTER

MGMT_OWNED 0.065 0.053 �0.069 �0.158GAAP_ETR 0.064 0.484 �0.067 �0.113CASH_ETR 0.056 0.509 0.020 �0.115DTAX �0.070 �0.078 0.003 0.023SHELTER �0.135 �0.157 �0.150 0.099

Panel A: *, **, *** indicate significance at the 10%, 5%, and 1% level, respectively. Differences between means are tested for significance using a two-tailedt-test; differences in medians are tested for significance using a two-tailed Wilcoxon signed rank test. All variables are as defined in the Appendix.All continuous variables are winsorized at the 1st and 99th percentile.Panel B: Bold indicates significance at the greater than 10% level based on a two-tailed t-test. All variables are as defined in the Appendix.

B.A. Badertscher et al. / Journal of Accounting and Economics 56 (2013) 228–250240

Table 4, Panel B, presents results for the second stage of the Heckman (1979) procedure, which tests whethermanagement-owned firms avoid less income tax than PE-backed private firms as predicted by P1.31 The coefficients onall four measures of tax avoidance are in the predicted directions and are statistically significant based on two-tailedp-values, providing support for P1.32,33 The coefficient on MGMT_OWNED in the CASH_ETR (GAAP_ETR) regression indicatesthat management-owned firms pay on average 7.5 (4.2) cents more income tax per dollar of adjusted pre-tax income thanPE-backed private firms. This result suggests a large economic difference in tax avoidance between management-owned andPE-backed firms. Given mean pretax income of $27.8 million (untabulated) for our sample, the coefficient on MGMT_OWNEDin the CASH_ETR (GAAP_ETR) regression translates into greater tax costs of approximately $2.1 ($1.2) million for management-owned firms compared to PE-backed firms. These results are economically smaller than those in Chen et al. (2010), whichfind that on average family-owned, public firms avoid $6.7 ($2.8) million less income tax than non-family-owned, publicfirms. Nonetheless, our results should be interpreted with caution as endogeneity can influence coefficient estimates and infact the MGMT_OWNED coefficients vary in magnitude across different estimation methods (e.g., Heckman vs. propensityscore matching procedures). We instead focus on the consistency of the signs and significance levels of the MGMT_OWNEDcoefficients across different estimation methods.

The coefficients on INV_MILLS are not significant in Panel B, consistent with selection bias having little impact on ourestimates. Stolzenberg and Relles (1997) argue that if selection bias is moderate then the two-step estimation approach cangenerate coefficients that are inferior to those from ordinary least squares (OLS) estimation. Thus, we also use OLS to re-estimate Eq. (2), excluding INV_MILLS. We present only the coefficients onMGMT_OWNED in Panel C of Table 4. Although themagnitudes of the coefficients are somewhat attenuated (e.g., the coefficient on MGMT_OWNED in the CASH_ETR regression

31 The number of observations differs across most regressions due to different data requirements. The GAAP_ETR and CASH_ETR regressions are basedon fewer observations (1701 and 1441, respectively) because these measures require firms to have positive pretax income.

32 Regressions where DTAX (SHELTER) is the dependent variable do not include INTANG and EQ_EARN (RNOA, LEV, MNC, AB_ACCR, and ASSETS) becausethose variables are included in the estimation of DTAX (SHELTER), and thus are orthogonal to DTAX (SHELTER), by design.

33 We include LOSS in the GAAP_ETR and CASH_ETR regressions because GAAP_ETR and CASH_ETR are scaled by the sum of pretax net income over yearst, t�1, and t�2, while LOSS captures whether year t's net income is less than zero.

Table 4Results for regressions that compare the tax avoidance of management-owned private firms and PE-backed firms.

Panel A: Results for first stage of Heckman procedure

Coefficient Pr4χ2

Intercept �8.160 0.999Q_RATIO �0.289 0.000OPER_CYCLE 0.002 0.000FIRM_AGE �0.054 0.052RNOA �0.316 0.302LOSS 0.191 0.006NOL 0.272 0.000LEV 0.590 0.000INTANG 1.347 0.000MNC 0.259 0.000AB_ACCR �1.900 0.000EQ_EARN �3.781 0.218SALES_GR �0.033 0.351ASSETS 0.079 0.003SOX 0.036 0.877

Fixed effects included Industry, YearMacKelvey_Zavonia Pseudo-R2 0.602McFadden's LRI Pseudo-R2 0.211N of PE-backed firm-years 1871N of non-PE-backed firm-Years 757

Panel B: Results for second stage of Heckman procedure

GAAP_ETR CASH_ETR DTAX SHELTER

Coeff t-stat Coeff t-stat Coeff t-stat Coeff t-stat

Intercept 0.419nnn 13.84 0.300nnn 7.08 0.068nn 2.45 �0.671nnn �16.61MGMT_OWNED 0.042nn 2.29 0.075nnn 2.84 �0.061nnn �4.84 �0.232nn �2.44RNOA �0.028 �0.50 0.095 1.38 0.093 1.50LOSS 0.032nn 2.42 0.078nnn 4.95 �0.009 �0.68 �0.954nnn �22.82NOL 0.002 0.13 0.000 �0.01 0.030nn 2.46 �0.091nn �2.12LEV �0.048nnn �2.83 �0.034 �1.60 0.003 0.17INTANG �0.003 �0.13 0.023 1.04 0.142 1.63MNC �0.140 �0.72 �0.145 �1.16 0.017 1.56AB_ACCR �0.072 �0.77 �0.041 �0.38 0.397nnn 4.28EQ_EARN �0.869nn �2.51 �1.560nn �1.97 0.859nnn 10.01SALES_GR �0.004 �0.86 �0.016nnn �2.65 0.004 0.61 0.120nnn 3.41ASSETS �0.014nnn �3.06 �0.012 �1.90 �0.007 �1.62SOX �0.090nnn �4.42 0.026 1.08 �0.013 �0.60 0.805nnn 8.53INV_MILLS 0.100 1.53 0.101 1.57 0.001 0.83 �0.069 �1.48Adjusted R2 0.0861 0.0818 0.0813 0.2548

N 1701 1441 2628 2628

Panel C: Results for OLS regression of tax avoidance measures on MGMT_OWNED without Heckman (1979)'s Correction

GAAP_ETR CASH_ETR DTAX SHELTER

Coeff t-stat Coeff t-stat Coeff t-stat Coeff t-stat

MGMT_OWNED 0.023nn 2.29 0.051nnn 4.18 �0.044nnn �3.19 �0.224nnn �7.01Adjusted R2 0.0740 0.0737 0.0775 0.2514N 1701 1441 2628 2628

Panel A: All variables are as defined in the Appendix.Panel B: n, nn, nnn indicate significance at the 10%, 5%, and 1% level using a two-tailed t-test, respectively. Regressions include industry and year indicatorvariables, which have not been tabulated. The t-stats have been adjusted to control for the clustering by multiple firm observations. The VIFs for theINV_MILLS is 9.24, 11.52, 8.97, and 8.96 respectively. All variables are as defined in the Appendix.Panel C: n, nn, nnn indicate significance at the 10%, 5%, and 1% level using a two-tailed t-test, respectively. Regressions include the following control variables:RNOA, LOSS, NOL. LEV, INTANG,MNC, AB_ACCR, EQ_EARN, SALES_GR, ASSETS, SOX, INDUS, and YEAR variables, which have not been tabulated. The t-stats havebeen adjusted to control for the clustering by multiple firm observations. All variables are as defined in the Appendix.

B.A. Badertscher et al. / Journal of Accounting and Economics 56 (2013) 228–250 241

is now just 0.051), inferences remain the same. Management-owned firms avoid significantly less tax than PE-backed firms,consistent with the separation of ownership and control having a significant impact on the tax avoidance practices ofprivate firms.

Table 5Summary of results for supplemental regressions that examine the impact of the separation of ownership and control on corporate tax avoidance.

GAAP_ETR CASH_ETR DTAX SHELTER

Coeff t-stat Coeff t-stat Coeff t-stat Coeff t-stat

Panel A: Management-owned vs. majority PE-backed firms

MGMT_OWNED 0.038n 1.91 0.068nnn 4.86 �0.056nnn �5.06 �0.206nnn �3.46Adjusted R2 0.0788 0.0958 0.0786 0.2306N 1488 1257 2316 2316

Panel B: Management-owned vs. minority PE-backed firms

MGMT_OWNED 0.022n 1.77 0.027n 1.68 �0.008 �0.96 �0.152nnn �2.02Adjusted R2 0.1348 0.1086 0.0172 0.2067N 739 566 1069 1069

Panel C: Minority PE-backed vs. majority PE-backed firms

MINORITY_PE 0.034nn 2.27nn 0.040nn 2.28nn �0.053nnn �3.67 �0.148nn �2.05Adjusted R2 0.0784 0.0518 0.0541 0.2123N 1138 1022 1871 1871

Panel D: Management-owned vs. employee-owned private firms

MGMT_OWNED 0.015 1.63 0.041n 1.74 �0.035 �1.48 �0.580nnn �5.56Adjusted R2 0.1285 0.1040 0.0387 0.2800N 768 559 997 997

Panel E: Management-owned private firms vs. public firms

MGMT_OWNED 0.043n 1.77 0.059nnn 3.98 �0.016nn �2.21 �0.813nnn �14.74Adjusted R2 0.0868 0.0715 0.0193 0.3250N 33,178 23,802 40,303 40,303

n, nn, nnn indicate significance at the 10%, 5%, and 1% level using a two-tailed t-test, respectively. Regressions include the following control variables: RNOA,LOSS, NOL. LEV, INTANG, MNC, AB_ACCR, EQ_EARN, SALES_GR, ASSETS, SOX, INV_MILLS (Panels A and B), INDUS, and YEAR variables, which have not beentabulated. The t-statistics have been adjusted to control for clustering by multiple firm observations. All variables are as defined in the Appendix.

B.A. Badertscher et al. / Journal of Accounting and Economics 56 (2013) 228–250242

In Table 5, we present results for supplemental tests that further evaluate the impact of the separation of ownership andcontrol on corporate tax avoidance (control variables included but not tabulated). In each of these tests we estimate Eq. (2)but vary the sample composition and/or utilize a different proxy for the separation of ownership and control. Panels A and Bextend the analyses in Table 4 and separately compare the tax avoidance of management-owned firms to that of majority-owned (Panel A) and minority-owned (Panel B) PE-backed firms. As expected, the coefficients on MGMT_OWNED indicatethat management-owned firms avoid less income tax than both majority-owned and minority-owned, PE-backed firms.Given the greater managerial stock ownership percentages in Table 2 for minority-owned, PE-backed firms relative tomajority-owned, PE-backed firms, we compare the tax avoidance of these two types of private firms in Panel C. We findthat minority-owned, PE-backed firms avoid significantly less income tax than majority-owned, PE-backed firms, consistentwith P1a. The coefficients on MGMT_OWNED in Panel D indicate that management-owned firms avoid less income tax thanemployee-owned firms, which have more diffuse stock ownership. These results support P1b. Panel E compares the taxavoidance of management-owned firms and public firms. Consistent with empirical prediction P1c, the coefficients onMGMT_OWNED in Panel E suggest that management-owned firms avoid less income tax than public firms.34 In sum, Table 5provides additional evidence that our findings for P1 are robust to different proxies for the separation of ownership andcontrol and for firms with different ownership structures.

4.4. Results for the marginal costs of tax avoidance at PE-backed firms

Empirical predictions P2a–P2c consider alternative explanations for our findings in Table 4 that management-ownedfirms avoid less income tax than PE-backed private firms. Specifically, P2a (P2b) predicts that private firms owned by PEfirms with many portfolio firms (large PE firms) have lower marginal costs of tax avoidance and, as a result, avoid moreincome tax than other firms. We first test these predictions by re-estimating Eq. (2) based on our sample of PE-backed firmsonly, while including proxies for the marginal costs of tax avoidance. MANY_PE (LARGE_PE) is an indicator variable for

34 We acknowledge that public and private firms are subject to substantially different financial reporting incentives, which may influence our results.In untabulated analyses, we also re-estimate the Panel E regressions based on propensity score matched samples of management-owned private firms andpublicly-traded companies. Inferences from those regressions are qualitatively similar to those based on the results in Panel E.

Table 6Results for regressions that examine the dual impact of the separation of ownership and control and the marginal costs of tax planning on corporate taxavoidance.

GAAP_ETR CASH_ETR DTAX SHELTER

Coeff t-stat Coeff t-stat Coeff t-stat Coeff t-stat

Panel A: Includes PE-backed firms only

MINORITY_PE 0.033nn 2.44 0.028 1.59 �0.061nnn �3.93 �0.102 �1.54MANY_PE �0.021 �1.40 �0.017 �0.96 0.015 1.18 0.087 1.23LARGE_PE �0.033nn �2.23 �0.043nn �2.73 0.045nn 2.47 0.376nnn 7.06

Adjusted R2 0.0841 0.0617 0.0574 0.2436N 1138 1022 1871 1871

Panel B: Includes management-owned and PE-backed private firms

MGMT_OWNED 0.032 1.55 0.059nnn 4.32 �0.064nnn �7.17 �0.119nn �2.07MANY_PE �0.027n �1.84 �0.018n �1.67 0.020 1.07 0.054 0.86LARGE_PE �0.028nn �2.02 �0.042nnn �2.76 0.042nnn 2.79 0.321nnn 6.40

Adjusted R2 0.0992 0.1026 0.0779 0.2334N 1701 1441 2628 2628

Panel C: Difference-in-difference regression analysis that compares the tax avoidance of management-owned and PE-backed, small- andlarge-sized private firms, where small- (large-) sized private firms are in the bottom (top) quartile of net sales for all private firms (excludesfirms not classified at small- or large-sized)

PE_BACKED� LARGE 0.317nnn 5.48 0.149nn 2.00 0.031 1.56 �0.005 �0.08PE_BACKED� SMALL 0.323nnn 7.77 0.168nn 2.75 0.045 1.31 �1.10nnn �17.24MGMT� LARGE 0.336nnn 5.35 0.164nn 2.10 0.006 0.62 0.585nnn 6.49MGMT� SMALL 0.384nnn 9.84 0.271nnn 4.57 �0.037 �0.99 �1.899nnn �25.69

Adjusted R2 0.7545 0.6594 0.1010 0.7149N 846 693 1327 1327

Tests for differences between PE- and non-PE-backed firms within firm size categoriesPE_BACKED� SMALL �0.061n �0.103nnn 0.082nnn 0.802nnn

-MGMT� SMALLPE_BACKED� LARGE �0.019 �0.016 0.025 �0.590nnn

-MGMT� LARGE

Difference �0.0424 �0.0874 4 0.056 1.3924 4 4

F-test (p-value) 0.097 0.021 0.328 0.001

n, nn, nnn indicate significance at the 10%, 5%, and 1% level using a two-tailed t-test, respectively. 4 ,4 4 ,4 4 4 indicate significant at the 10%, 5%, and 1% levelbased on an F-test. Regressions include the following control variables: RNOA, LOSS, NOL. LEV, INTANG, MNC, AB_ACCR, EQ_EARN, SALES_GR, ASSETS, SOX,INV_MILLS, INDUS, and YEAR variables, which have not been tabulated. The t-statistics have been adjusted to control for clustering by multiple firmobservations. All variables are as defined in the Appendix.

B.A. Badertscher et al. / Journal of Accounting and Economics 56 (2013) 228–250 243

whether a PE-backed firm is owned by a PE firm with many portfolio firms (large PE firm). If firms owned by PE firms withmany portfolio firms (large PE firms) have lower marginal costs of tax avoidance, the estimated coefficients on MANY_PE(LARGE_PE) should be negative (positive) in the GAAP_ETR and CASH_ETR (DTAX and SHELTER) regressions.

Table 6, Panel A, summarizes the OLS results where MINORITY_PE is our proxy for the separation of ownershipand control (since managers own significantly more stock at minority-owned PE-backed firms than at majority-ownedPE-backed firms). Although all four coefficients on MINORITY_PE have the predicted signs, just two are statisticallysignificant based on a two-tailed t-test, providing mixed evidence in support of P1a (i.e., minority-owned, PE-backedfirms avoid less tax than majority-owned, PE-backed firms). With respect to the marginal costs of tax avoidance, all fourcoefficients on LARGE_PE are significant in the predicted direction but none of the coefficients on MANY_PE are statisticallysignificant. Thus, the results in Panel A suggest firms owned by large PE firms enjoy lower marginal costs of tax avoidancebut firms owned by PE firms with many portfolio companies do not.

We now use our sample of management-owned and PE-backed firms to test whether private firms that are owned bylarge PE firms or by PE firms with many portfolio firms have lower marginal costs of tax avoidance (P2a and P2b), followingthe Heckman (1979) estimation procedure. Panel B summarizes the results where MGMT_OWNED is now our proxy for theseparation of ownership and control and we also include our proxies for the marginal cost of tax avoidance (i.e., MANY_PEand LARGE_PE). The results in Panel B reveal that three of the four coefficients on MGMT_OWNED are statistically significantin the predicted directions based on two-tailed t-tests even after controlling for the marginal costs of tax avoidance(the fourth coefficient is marginally significant based on one-tailed t-test). In addition, all of the coefficients on MANY_PEand LARGE_PE are significant in the predicted direction, except for the coefficients on MANY_PE in the DTAX and SHELTERregressions. Overall, the results in Panels A and B indicate that the separation of ownership and control (as proxied

B.A. Badertscher et al. / Journal of Accounting and Economics 56 (2013) 228–250244

by MGMT_OWNED and MINORITY_PE) has a significant, predictable impact on tax avoidance, even after controlling for themarginal costs of avoidance across different types of firms.

Empirical prediction P2c predicts that the difference in tax avoidance between small-sized, PE-backed and management-owned firms is larger than the difference in tax avoidance between large-sized, PE-backed and management-owned firms,since large firms may enjoy economies of scale to tax planning that are independent of PE ownership. We perform adifference-in-difference analysis based on our sample of management-owned and PE-backed private firms to empiricallytest this prediction. We first classify firms as small- vs. large-sized, where small-sized (large-sized) firms are those in thelowest (highest) quartile of net sales for all private firms. We then remove firms that are not small- or large-sized andestimate the following equation:

TAXi ¼ α1PE_BACKED� SMALLiþα2MGMT � SMALLiþα3PE_BACKED� LARGEiþα4MGMT � LARGEiþα5RNOAiþα6LOSSiþα7NOLiþα8LEViþα9INTANGi

þα10MNCiþα11AB_ACCRiþα12EQ_EARNiþα13SALES_GRiþα14ASSETSi

þα15SOXiþα16INV_MILLSiþαj∑iYEARiþαk∑lINDUSiþεI; ð3Þ

In this model specification, the coefficients on the four indicator variables (α1–a4) capture the average value for each taxavoidance measure for each type of firm (e.g., small-sized vs. large-sized, PE-backed firms), after controlling for numerousfirm characteristics. We predict that the difference in coefficients between small-sized, PE-backed and management-ownedfirms is larger than the difference in coefficients between large-sized, PE-backed and management-owned firms.

Table 6, Panel C, summarizes the results of our difference-in-difference analyses (control variables included but not tabulated).The differences in tax avoidance (see bottom of Panel C) are in the predicted directions, although the results based on DTAX are not

Table 7Summary of results for regressions that use the proportion of stock owned by managers (MGR_STOCK) as the proxy for separation of ownership and control(rather than MGMT_OWNED).

GAAP_ETR CASH_ETR DTAX SHELTER

Coeff t-stat Coeff t-stat Coeff t-stat Coeff t-stat

Panel A: Sample includes management-owned and PE-backed private firms (test of P1)

MGR_STOCK 0.034nn 2.17 0.058nn 2.44 �0.034nnn �4.76 �0.625nnn �8.76Adjusted R2 0.0476 0.0642 0.0436 0.2529N 1262 1164 1939 1939

Panel B: Sample includes management-owned firms only

MGR_STOCK 0.023n 1.68 0.065n 1.94 �0.033 �1.59 �1.08nnn �7.40Adjusted R2 0.1516 0.1474 0.0296 0.3594N 410 360 561 561

Panel C: Sample includes PE-backed firms only

MGR_STOCK 0.030n 1.74 0.054n 1.81 �0.129nnn �3.09 �0.287nn �2.17Adjusted R2 0.0436 0.0372 0.0454 0.2335N 852 804 1378 1378

Panel D: Sample includes PE-backed firms only

MGR_STOCK 0.016 1.31 0.025 1.55 �0.141nnn �3.42 �0.125n �1.65MANY_PE �0.035nn �2.01 �0.039nn �2.18 0.017 1.10 0.039 1.01LARGE_PE �0.042nn �2.46 �0.036nn �2.06 0.047nn 2.11 0.357nnn 6.16

Adjusted R2 0.0665 0.0569 0.0486 0.2642N 852 804 1378 1378

Panel E: Sample includes management-owned and PE-backed private firms (test of P3)

MGR_STOCK 0.018n 1.75 0.030n 1.75 �0.050nnn �6.00 �0.560nnn �7.54MANY_PE �0.038nn �2.16 �0.053nnn �2.95 0.027 1.56 0.088 1.58LARGE_PE �0.036nnn �3.01 �0.043nnn �2.58 0.076n 1.86 0.192nnn 3.36

Adjusted R2 0.0752 0.1027 0.0681 0.2882N 1262 1164 1939 1939

n, nn, nnn indicates significance at the 10%, 5%, and 1% level using a two-tailed t-test, respectively. Regressions include the following control variables: RNOA,LOSS, NOL. LEV, INTANG, MNC, AB_ACCR, EQ_EARN, SALES_GR, ASSETS, SOX, INV_MILLS (Panels A and B), INDUS, and YEAR variables, which have not beentabulated. The t-statistics have been adjusted to control for clustering by multiple firm observations. All variables are as defined in the Appendix.

Table 8Analysis of items that cause GAAP_ETR to differ from the statutory tax rate for management-owned and PE-backed private firms.

Panel A: Statutory reconciliation items

PE-backed firms (N¼76) Management-owned firms (N¼31) T-statistic for difference

Foreign tax rate differential �0.042 �0.002 �1.145State tax rate differential 0.012 0.013 �0.135Intangible assets �0.020 0.036 �1.988nn

Tax-exempt income items �0.013 0.014 �1.860n

Nondeductible expenses 0.013 �0.001 1.774n

Change in tax reserve 0.010 0.002 0.518Tax credits �0.021 0.000 �2.140nn

Other items 0.014 0.001 0.655

Panel B: Tax haven analyses

PE-backed firms (N¼465) Management-owned firms (N¼93) T-statistic for difference

Number of subsidiaries in tax havens 3.01 1.36 3.28nnn

n, nn, nnn indicate significance at the 10%, 5%, and 1% level. Differences between means are tested for significance using a two-tailed t-test; differences inmedians are tested for significance using a two-tailed Wilcoxon signed rank test.

B.A. Badertscher et al. / Journal of Accounting and Economics 56 (2013) 228–250 245

significant. For example, the difference in CASH_ETR between small-sized firms (�0.103) is significantly larger than the difference inCASH_ETR between large-sized firms (�0.016); P-value for F-test is 0.021. The results in Panel C are consistent with small-sizedfirms experiencing the greatest tax savings from PE ownership. Overall, the results in Table 6 indicate that large firms (Panel C) andcertain PE-backed firms (Panels A and B) have lower marginal costs of tax avoidance than other private firms. Nonetheless, theseparation of ownership and control (as proxied by MGMT_OWNED and MINORITY_PE) has a significant impact on tax avoidanceeven after controlling for variation in the marginal costs of avoidance across firms.

4.5. Results for tests based on managerial stock ownership rates

A more direct proxy for the separation of ownership and control is the proportion of stock owned by top executives(MGR_STOCK). Because this hand-collected data is not available for our entire sample of private firms, we view this measureas an alternative proxy for the separation of ownership and control. Table 7 summarizes results for supplemental teststhat are based on MGR_STOCK rather than MGMT_OWNED. First, we re-estimate Table 4 Heckman regressions, replacingMGMT_OWNED with MGR_STOCK. The coefficients on MGR_STOCK are presented in Panel A and they consistently indicatethat corporate tax avoidance is decreasing in managerial stock ownership. Next, we partition our sample into management-owned (Panel B) and PE-backed firms (Panel C) and estimate Eq. (2) using ordinary least squares regression. The results inboth panels also generally indicate that tax avoidance is decreasing in managerial stock ownership within each subsampleof private firms.

To test empirical prediction P3 (i.e., holding the marginal costs of tax avoidance constant, income tax avoidance isdecreasing in managerial stock ownership), we re-estimate Table 6, Panels A and B regressions. However, we replaceMGMT_OWNED with MGR_STOCK and we include our proxies for the marginal costs of tax avoidance. Table 7, Panel D,summarizes the results for OLS regressions based on our sample of PE-backed firms only, while Panel E summarizes resultsbased on our sample of management-owned and PE-backed firms. Across Panels D and E, all of the coefficients onMGR_STOCK have the predicted signs and 6 of the 8 coefficients are statistically significant. In addition, all of the coefficientson LARGE_PE are significant in the predicted directions while only 4 of the 8 coefficients on MANY_PE are significant in thepredicted directions. Overall, the results for MGR_STOCK in Table 7 confirm our findings in Tables 4 and 6 and indicate thatseparation of ownership and control has a significant influence on corporate tax avoidance, even after controlling forvariation in the marginal costs of avoidance across different types of firms.

4.6. Tax avoidance strategies and the utilization of subsidiaries in foreign tax havens

To gain a better understanding of the tax strategies used by our sample of private firms, we hand-collected detailedincome tax data from SEC filings for subsamples of management-owned and PE-backed firms matched on industry and year.Specifically, we collected data from Form 10-K statutory reconciliation schedules, which reveal material sources ofdifferences between effective and statutory tax rates, and thus sources of variation in our tax avoidance measures. Theresults in Table 8, Panel A, indicate that compared to management-owned firms, PE-backed firms report more negativestatutory reconciliation items related to foreign taxes, intangible assets, tax-exempt income (e.g. corporate-owned lifeinsurance policies), and tax credits, consistent with PE-backed firms relying on a variety of tax reduction strategies.

B.A. Badertscher et al. / Journal of Accounting and Economics 56 (2013) 228–250246

We further investigate the use of tax avoidance strategies that involve foreign subsidiaries. Multinational corporationscommonly reduce their worldwide tax burdens by strategically locating operations in low tax countries, including “taxhavens.”35 Following the methodology in Dyreng et al. (2011), we calculate the number of subsidiaries located in tax havensfor management-owned vs. PE-backed sample firms. The results in Table 8, Panel B, indicate that management-owned firmshave significantly fewer subsidiaries in tax haven countries than PE-backed private firms. Overall, the results in Table 8suggest that PE-backed firms use a variety of tax strategies to reduce their tax liabilities, including tax planning throughforeign operations.36

4.7. Supplemental analyses

4.7.1. Alternative approaches to addressing and evaluating endogeneityBecause PE firms may select firms to acquire based on specific observable attributes (as opposed to unobservable

attributes), we also perform our tests based on propensity score matched samples of management-owned and PE-backedfirms. In our study, the propensity score matching procedure would be effective if the selection bias (i.e., PE firm ownership)can be entirely explained by observable factors (Tucker, 2010). In case the selection bias in our sample is based on bothobservable and unobservable factors we also perform our tests based on both the propensity score matching and Heckman(1979) procedures (see discussion below).

We first calculate propensity scores derived from a probit model, where the dependent variable is a PE-backed indicatorvariable (PE_BACKED), and the model includes variables that are significantly different between management-ownedand PE-backed firms, including RNOA, LOSS, NOL, LEV, MNC, INTANG, AB_ACCR, SALES, and ASSETS. We then match eachmanagement-owned firm-year, one-to-one, to the PE-backed firm-year with the closest propensity score without replace-ment. To ensure that each management-owned firm-year and its match are similar to each other, we restrict the twofirms to have propensity scores within 0.10 of each other. We then use OLS regression to re-estimate Eq. (2) based on thepropensity score matched samples of management-owned and PE-backed firms. The results (untabulated) are similar tothose shown in Table 4, Panel B. Specifically, the coefficients on MGMT_OWN are all significant in the predicted directions.We also re-estimate Eq. (2) based on both the propensity score matched samples and following the Heckman (1979)two-stage procedure. The results are substantially similar to those for OLS regressions based on propensity score matchedsamples, as described above.

Larcker and Rusticus (2010) explain that it is sometimes difficult to obtain reliable instrumental variables whenaddressing endogeneity issues. An alternative approach is to assess how large the endogeneity problem must be to overturnthe results generated based on OLS (i.e., without controlling for endogeneity). We adopt their approach and calculatethe “impact threshold for a confounding variable” (ITCV) for eachMGMT_OWNED coefficient in the OLS regressions shown inTable 4, Panel C. In general, higher ITCVs indicate that OLS results are robust to omitted variable concerns. The ITCVs forMGMT_OWNED in Table 4, Panel C, are much higher than the benchmarks described in Larcker and Rusticus (2010) Table 7,i.e., Impact and ImpactRaw.37 We conclude it is unlikely that endogeneity concerns can overturn our main finding, i.e., that taxavoidance is increasing in the separation of ownership and control.

Lastly, we perform two final tests to evaluate whether tax is a significant predictor of PE ownership (e.g., if PE firmsfrequently acquire firms with relatively high tax burdens). First, we create a sample of public firms in the 5 years before theyare taken private by either management or a PE firm. We then estimate Eq. (1) – the PE ownership regression – and includeeach of our tax measures (i.e., GAAP_ETR, CASH_ETR, DTAX, and SHELTER). None of the coefficients on the tax measures isa significant predictor of PE ownership (results untabulated). Second, we create a sample of public firms that were eithertaken private by a PE firm or were never taken private. We again estimate Eq. (1) and include each of our tax measures.As before, none of the coefficients on the tax measures is a significant predictor of PE ownership (results untabulated).We conclude that tax is not a significant predictor of PE ownership for our sample of private firms, consistent with Mehranand Peristiani (2010). Overall, we conclude that our findings are strongly robust to a variety of econometric techniques thatevaluate and/or correct for endogeneity between PE firm ownership and corporate tax avoidance.

35 In this paper, the term “tax haven” refers to a country that has been designated a “tax haven” by the Organization for Economic Cooperation andDevelopment (OECD), due to its exceptionally low income tax rates and other favorable tax attributes relative to other countries. We thank Scott Dyreng forallowing us to use his database.

36 Untabulated analysis further indicates that PE-backed firms have a significantly higher mean tax fees paid to auditors than management-ownedfirms, consistent with PE-backed firms investing more resources in tax planning.

37 We calculate the ITCVs in a manner consistent with Frank (2000), which derives the minimum correlations necessary to turn a statisticallysignificant result into a borderline insignificant result. Specifically, the ITCV for MGMT_OWNED in the CASH_ETR regression is 0.069, which implies that thecorrelations between MGMT_OWNED and CASH_ETR and the unobserved confounding variable each need to be about 0.263 (¼√0.069) to overturn the OLSresults. This correlation is significantly higher than the “Impact” of the variable with the greatest influence on the MGMT_OWNED coefficient in theCASH_ETR OLS regression: LOSS (Impact¼0.016). Thus, the confounding variable would need to have a larger impact on CASH_ETR than LOSS to overturn thecoefficient on MGMT_OWNED. Nonetheless, similar to Larcker and Rusticus (2010), we acknowledge that impact thresholds are difficult to establish and soour results should be viewed with caution.

B.A. Badertscher et al. / Journal of Accounting and Economics 56 (2013) 228–250 247

4.7.2. Deletion of firms with negative pre-tax incomeAlthough our calculation of GAAP_ETR and CASH_ETR require the deletion of firm-years if the sum of pre-tax income over

years t�2 to year t is negative, we do not impose a similar data requirement on the other measures of tax avoidance (i.e., DTAXand SHELTER). To further evaluate whether our results are sensitive to the exclusion of firms with negative pre-tax income, weimpose a 3-year, positive pre-tax income data requirement on regressions where DTAX and SHELTER are the dependentvariables. Our results (untabulated) are qualitatively similar for this smaller, more profitable sample of firms relative to thoseshown in all tabulated analyses. We also note that the correlations between our four tax avoidance measures strengthenwhenwe require all sample observations to have positive, cumulative pre-tax income over a three year time period.

4.7.3. Tax benefits from employee stock optionsGraham et al. (2004) find that employee stock options (ESOs) generate significant tax savings and reduce marginal tax

rates for large firms, and thus are important non-debt tax shields. While tax deductions related to ESOs reduce cash effectivetax rates, they are not directly reflected in GAAP_ETR, DTAX, or SHELTER. Consistent with PE firms tying portfolio firmmanagement compensation to performance, the CEOs of PE-backed portfolio firms more frequently receive stock optionsthan the CEOs of non-PE-backed firms (Katz, 2009). However, as pointed out by Kaplan and Stromberg (2009), the equitystake of a portfolio firm manager is illiquid because the manager cannot sell portfolio firm equity or exercise stock optionsuntil the firm is publicly-traded. As a result, we do not expect stock options to generate tax benefits for PE-backed firms,which should bias against finding our predicted results.

To empirically evaluate the impact of stock options on corporate tax burdens, we utilize ESO tax benefit data (TXBCOand TXBCOF) that is available on Compustat for fiscal years 2005 and thereafter. Although less than 15% of our sampleobservations report non-zero ESO tax benefits, the amounts that are reported are not statistically different betweenmanagement-owned and PE-backed firms. Overall, we conclude that ESO tax benefits do not significantly influence ourresults.

5. Conclusions

In this study we investigate the impact of ownership structure on corporate tax avoidance. We take advantage ofa unique sample of firms with privately-owned equity but publicly-traded debt and examine whether variation in theseparation of ownership and control influences the tax avoidance of private firms. Fama and Jensen (1983) assert that whenequity ownership and corporate decision-making are concentrated in just a small number of decision-makers, these owner-managers will likely be more risk averse and thus less willing to invest in risky projects. Because income tax avoidance isa risky activity that can impose significant costs on a firm, we predict that firms with greater concentrations of ownershipand control (and thus more risk averse managers) avoid less income tax than firms with less concentrated ownership andcontrol. Our results are consistent with expectations. However, we also consider a competing explanation for these findings.In particular, we examine whether certain PE firms are able to reduce portfolio firms' marginal costs of tax avoidance,resulting in greater tax avoidance at PE-backed firms than at management-owned firms. Our results are consistent with themarginal costs of tax avoidance and the separation of ownership and control both influencing corporate tax practices.Overall, these findings increase our understanding of whether and how ownership structure influences corporate taxpractices.

Our findings are subject to several limitations. First, PE firms do not randomly select firms to acquire. To the extent PEfirm acquisition choices are correlated with target firm tax planning, then PE firm ownership could be endogenously relatedto tax avoidance in our sample, which would cause OLS coefficient estimates to be biased. To mitigate potential selectionbias in our empirical tests, we employ a variety of econometric techniques, including the Heckman (1979) two-stageestimation procedure, a propensity score matching procedure, and the alternative approach described in Larckerand Rusticus (2010). Regardless of whether or how we correct for endogeneity in our empirical tests, inferences from theresults are always the same: tax avoidance is increasing in the separation of ownership and control. Nonetheless, we cautionreaders to not place substantial weight on the magnitudes of our coefficient estimates but to instead focus on theconsistency of the signs and significance levels of our results.

Second, our main results are based on a sample of management-owned and PE-backed private firms that are required tofile financial statements with the SEC. These firms provide a powerful research setting for our research question because oursample exhibits substantial variation in the separation of ownership and control but holds financial reporting requirementsrelatively constant across all firms. Although our sample of private firms is subject to less financial reporting pressure thanpublic firms, we acknowledge that PE-backed firms are likely subject to somewhat greater financial reporting pressure thanmanagement-owned firms (since PE-backed firms are typically sold or taken public 5–7 years after they are taken private).Thus, our results could be influenced by differences in financial reporting pressure at management-owned and PE-backedprivate firms. We note, however, that we continue to find a negative association between managerial stock ownershipand income tax avoidance when we repeat our tests using the proportion of stock owned by managers as our proxy for theseparation of ownership and control within subsamples that contain only management-owned or only PE-backed firms. Weconclude that financial reporting incentives do not drive our main results.

Our study seeks to understand the fundamental firm characteristics that influence corporate tax avoidance by relying onprinciple-agent theory to build a framework for understanding how one specific feature of ownership structure, namely the

B.A. Badertscher et al. / Journal of Accounting and Economics 56 (2013) 228–250248

separation of ownership and control, impacts corporate tax practices. Our findings contribute toward a better understandingof the impact of insider control on corporate tax avoidance (e.g., Shackelford and Shevlin, 2001) and complementsrecent research that examines how agency costs and managerial incentives influence corporate tax practices (e.g., Desai andDharmapala, 2006, 2008; Hanlon and Heitzman, 2010; Rego and Wilson, 2012).

Appendix. Variable measurement

Measures of tax avoidance

GAAP_ETR

Firm i's GAAP effective tax rate, which equals total income tax expense (Compustat TXT), over years t�2 to t, divided by the sumof pre-tax income (PI) minus special items (SPI) in year t�2 to t. If data limitations prohibit us from using years t�2 to t, we nextuse years t�1 to t, followed by year t. GAAP_ETR is set to missing when the denominator is zero or negative and we winsorizeGAAP_ETR to the range [0,1]

CASH_ETR

Firm i's cash effective tax rate, which equals cash taxes paid (TXPD), over years t�2 to t, divided by the sum of pretax net income(PI) minus special items (SPI) in years t�2 to t. If data limitations prohibit us from using years t�2 to t, we next use years t�1 tot, followed by year t. CASH_ETR is set to missing when the denominator is zero or negative and we winsorize CASH_ETR to therange [0,1]

DTAX

Firm i's residual from the following regression, estimated by industry and year: PERMDIFFit¼β0þβ1 INTANGitþβ2 UNCONitþβ3MIitþβ4 CSTEitþβ5 ΔNOLitþβ6 LAGPERMitþeit; where PERMDIFF¼Total book-tax differences�temporary book-tax differences¼[{BI�[(CFTEþCFOR)/STR]}�(DTE/STR)], scaled by beginning of year assets (AT); BI¼pretax book income (PI); CFTE¼currentfederal tax expense (TXFED); CFOR¼current foreign tax expense (TXFO); STR¼statutory tax rate; DTE¼deferred tax expense(TXDI); INTANG¼goodwill and other intangible assets (INTAN), scaled by beginning of year assets (AT); UNCON¼ income (loss)reported under the equity method (ESUB), scaled by beginning of year assets (AT); MI¼ income (loss) attributable to minorityinterest (MII), scaled by beginning of year assets (AT); CSTE¼current state tax expense (TXS), scaled by beginning of year assets;DNOL¼change in net operating loss carryforwards (TLCF), scaled by beginning of year assets (AT); LAGPERM¼PERMDIFF in yeart�1. From 1980 to 1986 the STR is 46%, for 1987 the STR is 40%, from 1988 to 1992 the STR is 34%, from 1993 to 2005 the STR is35%.We winsorize DTAX to the range [�1,1]

SHELTER

Probability that firm i engages in a tax shelter as defined by Wilson (2009), where Compustat Tax Shelter¼�4.86þ5.20n BookTax Differencesþ4.08nDiscretionary Accruals�1.41nLeverageþ0.76nSizeþ3.51nROAþ1.72nForeign Incomeþ2.42nR&D

Private firm indicator variables

MGMT_OWNED 1 if the firm does not have a PE sponsor and at least 50% of the firm is owned by founders, current and past named executive

officers, and/or their families, and 0 otherwise

MGR_STOCK The ratio of stock owned by founders, current and past named executive officers, and/or their families, to common shares

outstanding

PE_BACKED 1 if a PE firm has a majority or minority ownership stake in a private company, and 0 otherwise MAJORITY_PE 1 if 50% or more of the firm is backed by PE firms, and 0 otherwise MINORITY_PE 1 if less than 50% of the firm is backed by PE firms, and 0 otherwise LARGE_PE 1 if the private equity firm that owns the portfolio firm is one of the following: Carlyle Group, Blackstone, Warburg Pincus, KKR,

Goldman Sachs Private Equity, Cerberus Capital, Fortress Investment, Apollo Global, Bain Capital, TPG Capital, 3i Group, ApaxPartners, Thomas H. Lee, Morgan Stanley Private Equity, and Welsh Carson Anderson & Stone and 0 for all other PE firms. PEfirms are ranked according to total U.S. dollar investment during the years 1980–2009. (Source: Thomson Financials,VentureXpert)

MANY_PE

1 if the number of firms owned by the PE firm is greater than 200 and the ratio of equity invested divided by number of firmsowned is greater than $30 million and 0 otherwise

EMPLOYEE_OWNED

1 if the firm does not have a PE sponsor and more than 50% of the equity is owned by the firms' employees, and 0 otherwise

LARGE

1 if the firm's sales are in the top quartile of net sales (SALE) for all private firms and zero otherwise

SMALL

1 if the firm's sales are in the bottom quartile of net sales (SALE) for all private firms and zero otherwise

Control variables and other variables of interest

AB_ACCR Firm i's abnormal total accruals in year t computed derived from the modified cross-sectional Jones (1991) model. To estimate

the model yearly by two-digit SIC code, we require that at least 10 observations be available. The regression is: TACCj,t/TAj, t�1¼a1n[1/TAj, t–1]þa2n[(ΔREVj, t�ΔTRj, t)/TAj, t–1]þa3n[PPEj, t/TAj, t–1] where TACC is total accruals for firm j in year t, which is definedas income before extraordinary items (IBC) minus net cash flow from operating activities, adjusted to extraordinary items anddiscontinued operations OANCF�XIDOC). For the years prior to 1988, TACC is defined as Δ(current assets ACT)�Δ(currentliabilities LCT)�Δ(cash CHE)þΔ(short-term debt DLC)�(depreciation and amortization DPC). To correct for measurement errorsin the balance-sheet approach, we eliminate firm-year observations with “non-articulating” events (Hribar and Collins, 2002).TA is the beginning-of-the-year total assets (lagged AT). ΔREV is the change in sales in year t (SALE), PPE is gross property, plant,and equipment in year t (PPEGT), and ΔTR is the change in trade receivables in year t (RECTR). To control for the asymmetricrecognition of gains and losses, the modified Jones model is augmented with the following independent variables: cash flowfrom operations in year t (CFt), a dummy variable set to 1 if CFt o1 and 0 otherwise (DCFt), and an interactive variable, CFt�DCFt(as suggested by Ball and Shivakumar, 2006). CFt is defined, for years after 1988, as cash from operations in year t adjusted forextraordinary items and discontinued operations (OANCF–XIDOC), and prior to 1988 as funds from operations (FOPT)–Δ(currentassets ACT)þΔ(cash and cash equivalent CHE)þΔ(current liabilities LCT)–Δ(short-term debt DLC). All variables are standardizedby total assets at year-end t�1

ASSETS

Natural logarithm of the total assets (AT) for firm i, at the end of year t EQ_EARN Firm i's equity income in earnings (ESUB) in year t, scaled by lagged total assets FIRM_AGE Firm i's age (years since first appearance on Compustat) INTANG Firm i's intangible assets (INTAN) in year t, scaled by lagged total assets INV_MILLS The inverse mills ratio from Heckman (1979) two-stage sample selection correction procedure. In the first stage, we estimate the

following probit model:

B.A. Badertscher et al. / Journal of Accounting and Economics 56 (2013) 228–250 249

PE_BACKED¼ β0þβ1Q_RATIOþβ2OPER_CYCLEþβ3FIRM_AGE

þβ4RNOAþβ5LOSSþβ6NOLþβ7LEVþβ8INTANGþβ9MNCþβ10AB_ACCRþβ11EQ_EARNþβ12SALES_GRþβ13ASSETSþβ14SOXþ∑tatYEARtþ∑kakINDUSiþεi

All other variables as defined in the Appendix. We use the estimates from the first-stage probit model to compute the inverseMills' ratio for each sample firm-year. The inverse Mills' ratio serves as a control variable in Eq. (1), which is the second step ofthe Heckman estimation procedure. (Inverse Mills ratio is defined as λ(Z)¼φ(Ζ)/Ф(Z) if private or PE-backed¼1, and λ(Z)¼�φ(Ζ)/(1�Ф(Z)) if private or PE-backed¼0, where φ(Ζ) is the standard normal pdf, Ф(Z) is the standard normal cdf, and Z are theestimates of the first stage probit model)

LEV

Firm i's leverage in year t, measured as total long-term debt (DLTT) divided by total assets LOSS 1 if firm i reports a loss, where loss is net income before extraordinary items (IBC) and 0 otherwise MNC 1 if firm's foreign pre-tax income (PIFO) or foreign income taxes (TXFO) is positive or negative and 0 otherwise NOL 1 if firm i has net operating loss carryforwards (TLCF) available at the beginning of year t, and 0 otherwise OPER_CYCLE Firm i's length of operating cycle, calculated as (yearly average accounts receivable (RECTt))/(total revenues (SALEt)/360)þ(yearly

average inventory (INVTt))/(cost of goods sold (COGSt)/360)

Q_RATIO Firm i's quick ratio, calculated as cash and short-term investments (#CHEt)þtotal receivables (RECTt), scaled by current liabilities

(LCTt)

RNOA Firm i's operating income divided by net operating assets, where operating income is net income (NI)þΔ(cumulative translation

adjustment RECTA)þafter-tax interest expense (XINT)�after-tax interest income (IDIT)þminority interest in income (MII). Netoperating assets (NOA) are common equity (CEQ)þdebt in current liabilities (DLC)þtotal long-term debt (DLTT)þpreferredstock (PSTK)�cash and short-term investments (CHE)� investments and advances (IVAO)þminority interest (MIB) (see Nissimand Penman, 2003)

SALES_GR

Firm i's sales growth, where sales growth is sales (SALE) at the end of year t less sales at the beginning of year t divided by salesat the beginning of year t

SOX

1 if the fiscal year is 2004 and thereafter Σk INDUS 1 (0) if firm i is (is not) in industry k in year t, based on three-digit SIC codes Σt YEAR 1 (0) if firm i is (is not) in year t

References

Acharya, V., Hahn, M., Kehoe, C., 2009. Corporate governance and value creation evidence from private equity. Working Paper. New York University.Aghion, P., Tirole, J., 1997. Formal and real authority in organizations. Journal of Political Economy 105 (1), 1–29.Aslan, H., Kumar, P., 2011. Lemons or cherries? Growth opportunities and market temptations in going public and private. Journal of Financial and

Quantitative Analysis 46 (2), 489–526.Baker, G., Gibbons, R., Murphy, K., 1999. Informal authority in organizations. Journal of Law, Economics, & Organization 15 (1), 56–73.Ball, R., Shivakumar, L., 2005. Earnings quality in UK private firms: comparative loss recognition timeliness. Journal of Accounting and Economics 39 (1),

83–128.Ball, R., Shivakumar, L., 2006. The role of accruals in asymmetrically timely gain and loss recognition. Journal of Accounting Research 44 (2), 207–242.Beatty, A., Harris, D., 1998. The effects of taxes, agency costs and information asymmetry on earnings management: a comparison of public and private

firms. Review of Accounting Studies 4 (3/4), 299–326.Bharath, S.T., Sunder, J., Sunder, S.V., 2008. Accounting quality and debt contracting. The Accounting Review 83 (1), 1–28.Bova, F., Y. Dou, O.-K. Hope, 2012a. Employee ownership and firm disclosure. Working Paper, University of Toronto.Bova, F., K. Kolev, J. Thomas, F. Zhang, 2012b. Non-executive employee ownership and corporate risk-taking. Working Paper, University of Toronto.Brickley, J., Smith, C., Zimmerman, J., 2009. Managerial Economics & Organizational Architecture, 5th ed. McGraw-Hill/Irwin.Cantillo, M., Wright, J., 2000. How do firms choose their lenders? An empirical investigation. Review of Financial Studies 13 (1), 155–189.Cao, J., Lerner, J., 2009. The performance of reverse leveraged buyouts. Journal of Financial Economics 91 (2), 139–157.Chen, S., Chen, X., Cheng, Q., Shevlin, T., 2010. Are family firms more tax aggressive than non-family firms? Journal of Financial Economics 95 (1), 41–61.Cheng, C., Huang, H., Li, Y., Stanfield, J., 2012. The effect of hedge fund activism on corporate tax avoidance. The Accounting Review 87 (5), 1493–1526.Cohen, D., Dey, A., Lys, T., 2008. Real and accrual-based earnings management in the pre- and post-Sarbanes Oxley periods. The Accounting Review 83 (3),

757–787.Cornelli, F., O. Karakas, 2008. Private equity and corporate governance: do LBOs have more effective boards? Working Paper, London Business School.Cotter, J., Peck, S., 2001. The structure of debt and active equity investors: the case of the buyout specialist. Journal of Financial Economics 59 (1), 101–147.Cunningham, N., M. Engler. 2008. The carried interest controversy: let's not get carried away. Tax Law Review, Cardozo Legal Studies Research Paper

No. 211.Demiroglu, C., James, C., 2010. The information content of bank loan covenants. Review of Financial Studies 23 (10), 3700–3737.Denis, D., Mihov, V., 2003. The choice among bank debt, non-bank private debt and public debt: evidence from new corporate borrowings. Journal of

Financial Economics 70 (1), 3–28.Desai, M., Dharmapala, D., 2006. Corporate tax avoidance and high-powered incentives. Journal of Financial Economics 79 (1), 145–179.Desai, M., Dharmapala, D., 2008. Tax and corporate governance: an economic approach. In: Schon, W. (Ed.), Tax and Corporate Governance, Springer-Verlag,

Berlin, pp. 13–30.Dyreng, S., Hanlon, M., Maydew, E., 2008. Long-run corporate tax avoidance. The Accounting Review 83 (1), 61–82.Dyreng, S., Lindsey, B., Thornock, J., 2011. Exploring the role Delaware plays as a domestic tax haven. Working Paper, Duke University.Fama, E., Jensen, M., 1983. Separation of ownership and control. Journal of Law and Economics 26 (2), 301–325. (Corporations and Private Property:

A Conference Sponsored by the Hoover Institution).Fleischer, V. 2007. Taxing Blackstone. University of Illinois Law & Economics Research Paper No. 07-036.Fleischer, V., 2008. Two and twenty: taxing partnership profits in private equity funds. New York University Law Review 83 (1), 1–59.Frank, K.A., 2000. Impact of a confounding variable on a regression coefficient. Sociological Methods & Research 29, 147–194.Frank, M., Lynch, L., Rego, S., 2009. Tax reporting aggressiveness and its relation to aggressive financial reporting. The Accounting Review 84 (2), 467–496.Fruhan, W., 2009. Role of private equity firms in merger and acquisition transactions. Harvard Business School Technical Note # 9-206-101 (July), pp. 1–15.Gadiesh, O., MacArthur, H., 2008. Lessons from Private Equity any Company Can Use. Bain & Company Inc, Harvard Business PressBoston, MA.

B.A. Badertscher et al. / Journal of Accounting and Economics 56 (2013) 228–250250

Gilson, R., Whitehead, C., 2008. Deconstructing equity: public ownership, agency costs, and complete capital markets. Columbia Law Review 108 (1),231–264.

Givoly, D., Hayn, C., Katz, S., 2010. Does public ownership of equity improve earnings quality? The Accounting Review 85 (1), 195–225.Graham, J., Lang, M., Shackelford, D., 2004. Employee stock options, corporate taxes, and debt policy. Journal of Finance 59 (4), 1585–1618.Hanlon, M., Heitzman, S., 2010. A review of tax research. Journal of Accounting and Economics 50 (2–3), 127–178.Heckman, J., 1979. Sample selection bias as a specification error. Econometrica 47 (1), 153–162.Hribar, P., Collins, D., 2002. Errors in estimating accruals: implications for empirical research. Journal of Accounting Research 40 (1), 105–134.Ivashina, V., Kovner, A., 2011. The private equity advantage: leveraged buyout firms and relationship banking. Review of Financial Studies 24 (7),

2462–2498.Jensen, M., 2007. The Economic Case for Private Equity (and Some Concerns), Keynote SlidesHarvard University.Jensen, M., Meckling, W., 1976. Theory of the firm: managerial behavior, agency costs and ownership structure. Journal of Financial Economics 3 (4),

305–360.Jones, J., 1991. Earnings management during import relief investigations. Journal of Accounting Research 29 (2), 192–228.Kaplan, S., 2009. The future of private equity. Journal of Applied Corporate Finance 21 (3), 8–20.Kaplan, S., Schoar, A., 2005. Private equity performance: returns, persistence, and capital flows. The Journal of Finance 60 (4), 1791–1823.Kaplan, S., Stromberg, P., 2009. Leveraged buyouts and private equity. Journal of Economic Perspectives 23 (1), 121–146.Katz, S., 2009. Earnings quality and ownership structure: the role of private equity sponsors. The Accounting Review 84 (3), 623–658.Klassen, K., 1997. The impact of inside ownership concentration on the tradeoff between financial and tax reporting. The Accounting Review 72 (3),

455–474.Knoll, M., 2007. The taxation of private equity carried interests: estimating the revenue effects of taxing profit interests as ordinary income. Working Paper,

University of Pennsylvania Law School.Larcker, D., Rusticus, T., 2010. On the use of instrumental variables in accounting research. Journal of Accounting and Economics 49 (3), 186–205.Lawton, A., 2008. Taxing private equity carried interest using an incentive stock option analogy. Harvard Law Review 121 (3), 846.Lee, L., 1979. Identification and estimation in binary choice models with limited (censored) dependent variables. Econometrica 47 (4), 977–996.Lee, P., Wahal, S., 2004. Grandstanding, certification, and the underpricing of venture capital backed IPOs. Journal of Financial Economics 73 (2), 375–407.Lennox, C., Francis, J., Wang, Z., 2012. Selection models in accounting research. The Accounting Review 87 (2), 589–616.Maddala, G., 1983. Limited-Dependent and Qualitative Variables in Econometrics. Cambridge University Press, Cambridge, UK.Manzon, G., Plesko, G., 2002. The relation between financial and tax reporting measures of income. Tax Law Review 55 (2), 175–214.Masulis, R., Thomas, R., 2009. Does private equity create wealth? The effects of private equity and derivatives on corporate governance. University of

Chicago Law Review 76 (1), 219–259.McGuire, S., Wang, D., Wilson, R., 2012. Dual class ownership and tax a avoidance. Working Paper, Texas A&M University.Mehran, H., Peristiani, S., 2010. Financial visibility and the decision to go private. Review of Financial Studies 23 (2), 519–547.Mikhail, M., 1999. Coordination of earnings, regulatory capital and taxes in private and public companies. Working Paper, Massachusetts Institute of

Technology.Mills, L., Newberry, K., 2001. The influence of tax and nontax costs on book-tax reporting differences: public and private firms. Journal of the American

Taxation Association 23 (1), 1–19.Morsfield, S., Tan, C., 2006. Do venture capitalists influence the decision to manage earnings in initial public offering? The Accounting Review 81 (5),

1119–1150.Nissim, D., Penman, S., 2003. Financial statement analysis of leverage and how it informs about profitability and price-to-book ratios. Review of Accounting

Studies 8 (4), 531–559.Opler, T., Titman, S., 1993. The determinants of leveraged buyout activity: free cash flow vs. financial distress costs. The Journal of Finance 48 (5), 1985–1999.Penno, M., Simon, D., 1986. Accounting choices: public versus private firms. Journal of Business Finance & Accounting 13 (4), 561–569.Rego, S., 2003. Tax-avoidance activities of U.S. multinational corporations. Contemporary Accounting Research 20 (4), 805–833.Rego, S., Wilson, R., 2012. Executive compensation, equity risk incentives, and corporate tax aggressiveness. Journal of Accounting Research 50 (3), 775–809.Shackelford, D., Shevlin, T., 2001. Empirical tax research in accounting. Journal of Accounting and Economics 31 (1–3), 321–387.Smith, C., Stulz, R., 1985. The determinants of firms' hedging policies. Journal of Financial and Quantitative Analysis 20, 391–405.Smith, C., Watts, R., 1982. Incentive and tax effects of executive compensation plans. Australian Journal of Management 7 (2), 139–157.Stolzenberg, R., Relles, D., 1997. Tools for intuition about sample selection bias and its correction. American Sociological Review 62, 494–507.Stromberg, P., 2009. The economic and social impact of private equity in Europe: Summary of research findings. Working Paper, Stockholm School of

Economics.Tucker, J., 2010. Selection bias and econometric remedies in accounting and finance research. Journal of Accounting Literature 29, 31–57.Wilson, R., 2009. An examination of corporate tax shelter participants. The Accounting Review 84 (3), 969–999.