DOI: 10.1111/1475-679X.12269Journal of Accounting ResearchVol. 57 No. 4 September 2019

Printed in U.S.A.

Made in the U.S.A.?A Study of Firm Responses to

Domestic Production Incentives

R E B E C C A L E S T E R∗

Received 29 August 2016; accepted 18 March 2019

∗Stanford Graduate School of Business, Stanford University.Accepted by Christian Leuz. This paper is based on my dissertation at the Mas-

sachusetts Institute of Technology, Sloan School of Management, and it received the MITSloan School of Management Thesis Prize, as well as the 2016 American Taxation Associa-tion/PricewaterhouseCoopers Outstanding Tax Dissertation Award. I appreciate commentsfrom my dissertation committee—Michelle Hanlon (chair), S. P. Kothari, and Rodrigo Verdi—as well as from two anonymous referees, Joshua Anderson, Anna Costello, Lisa De Simone,John Gallemore, Joao Granja, Dominika Langenmayr, Patricia Naranjo, Michelle Nessa (dis-cussant), Ralph Rector, Nemit Shroff, Eric So, Joe Weber, Bill Zeile, and seminar participantsat the 2015 AAA Annual Meeting, Duke University, MIT, Harvard Business School, the U.S.Treasury Office of Tax Analysis, Stanford University, the University of Chicago, the Universityof Illinois, the University of Michigan, the University of Notre Dame, the University of Penn-sylvania, and Washington University in St. Louis. I thank Jeff Hoopes for assistance with datacollection and for sharing R&D credit tax data; Michelle Hanlon for sharing AJCA repatri-ation tax data; Scott Dyreng for providing Exhibit 21 data; and Wonhee Lee, Mason Jiang,Christina Becher, and Carl Rodriguez for research assistance. I appreciate institutional in-sight from Harry Moser of the Reshoring Initiative, Joanne Bonfiglio, Jeff Cote, David Hoff-man, Anne-Marie Petersen, Larry Salus, Ryan Stecz, and Todd Videbeck. I gratefully acknowl-edge financial support from the MIT Sloan School of Management, the Stanford GraduateSchool of Business, and the Deloitte Foundation. The statistical analysis of firm-level dataon U.S. multinational companies was conducted at the Bureau of Economic Analysis (BEA),Department of Commerce, under arrangements that maintain legal confidentiality require-ments. The views expressed are those of the author and do not reflect official positions ofthe U.S. Department of Commerce. An online appendix to this paper can be downloaded athttp://research.chicagobooth.edu/arc/journal-of-accounting-research/online-supplements.

1059

C© University of Chicago on behalf of the Accounting Research Center, 2019

1060 R. LESTER

ABSTRACT

How do U.S. companies respond to incentives intended to encourage do-mestic manufacturing? I study the Domestic Production Activities Deduction(DPAD), which was enacted in the American Jobs Creation Act (AJCA) of2004 and was the third largest U.S. corporate tax expenditure as of 2017.Using confidential data from the U.S. Bureau of Economic Analysis, I findgreater average domestic investment spending of $95.5–$143.6 million, butonly within the sample of domestic-only firms and not until 2010, when thegreatest statutory DPAD benefits were available. Additional evidence suggeststhat U.S. multinational claimants invest abroad rather than in the UnitedStates and that the increased investment by DPAD firms is accompanied by areduction in the domestic workforce, consistent with a substitution of capitalfor labor. I also show that the delayed investment response is due to firmsengaging in other responses first, such as changing corporate reporting toshift income across time and borders. Quantifying the extent of these effectscontributes to the literature that studies this tax deduction and informs pol-icy makers as to the effectiveness of both manufacturing incentives and U.S.corporate income tax rate reductions in stimulating real domestic activity.

JEL codes: F23; G38; H25; M40; M48

Keywords: income shifting; investment; employment; tax

1. Introduction

This paper studies how U.S. companies respond to domestic productionincentives. Encouraging domestic manufacturing is a central goal of pol-icy makers, as the industry employs over 12 million U.S. workers and con-tributes $2.1 trillion to the U.S. GDP (Scott [2015], Bureau of Labor Statis-tics [2017]). To evaluate the effectiveness of such incentives in increasingdomestic investment and employment, I study the Domestic Production Ac-tivities Deduction (Section 199 deduction or hereafter the “DPAD”) passedas part of the American Jobs Creation Act of 2004 (H.R. Rep. No. 108-393[2004a]).1 Firms have claimed over $70 billion in DPAD benefits since en-actment, making the deduction the third largest corporate tax expenditureas of the end of 2017 (Joint Committee on Taxation [2017]). However, be-cause the law did not require firms to commit to new domestic investmentor employment, companies could claim the deduction with few, if any, op-erational changes. Firms could also choose other less costly channels, suchas shifting income across time and countries, to increase the amount ofqualifying domestic income. In this paper, I examine (1) the extent of thedomestic investment and employment effects and (2) whether and to what

1 H.R. Rep. 108-393 [2004a] confirms that this incentive was intended to increase domes-tic investment and employment. This report states: “The Committee [on Ways and Means]believes that a reduced tax burden on domestic manufacturers will improve the cash flowof domestic manufacturers and make investments in domestic manufacturing facilities moreattractive. Such investment will create and preserve U.S. manufacturing jobs.”

FIRM RESPONSES TO DOMESTIC PRODUCTION INCENTIVES 1061

extent firms engage in other less costly income shifting activities first so asto maximize the domestic manufacturing benefit claimed.

As a brief overview, the DPAD allows firms to deduct a portion of in-come related to domestic production when determining their U.S. incometax liability. To calculate the amount of the deduction, firms identify do-mestic production gross receipts and the associated direct and indirect do-mestic expenses. The deduction equals net domestic production income(revenues minus expenses) times the statutory deduction percentage of3% (2005–2006), 6% (2007–2009), or 9% (2010–2017). Thus, once fullyphased in, the deduction effectively lowers the corporate tax rate on qual-ifying income by 3.15% (35% × 9% = 3.15%). The DPAD was repealedas part of the 2017 tax legislation H.R. 1 (referred to as the Tax Cuts andJobs Act of 2017, or TCJA), even though the intended goals of stimulatingdomestic manufacturing and creating jobs are central to the new tax law(Oliphant [2017]; Ryan [2017]; White House [2018]). Therefore, under-standing the extent to which the incentive induced changes in corporatereporting, in lieu of the intended real activities, is relevant to anticipatingand evaluating the effects of the 2017 tax law.

The literature (Kemsley [1998]) and other papers that study the DPAD(Blouin, Krull, and Schwab [2014]; Ohrn [2018]) suggest that firms will re-spond to a domestic tax incentive by increasing domestic production. How-ever, Slemrod [1992] states that taxpayers often take less costly actions first.The least costly response is for firms to shift transactions in time (intertempo-ral shifting). Second, firms may use accounting discretion to recharacterizetransactions within the firm (accounting recharacterization). Finally, the costli-est action is to make real operational changes (real changes). I test the extentto which firms claiming the DPAD (“DPAD firms”) engage in each of theseactivities, and I compare these responses to a set of matched control firmsthat never disclose claiming the benefit.

My first hypothesis tests whether DPAD firms engage in intertemporalincome shifting to maximize the amount of the tax benefit. The literatureshows that firms will shift the timing of transactions to recognize income inlower tax rate periods, thereby reducing the associated cash tax payments(Scholes, Wilson, and Wolfson [1992]; Guenther [1994]; Maydew [1997]).The DPAD provides similar incentives for shifting because income relatedto domestic production is effectively subject to a lower tax rate once firmscan claim the tax benefit. I predict that DPAD firms shift more income intothe first year of claiming the DPAD, as well as subsequent years in whichthe statutory DPAD rate increases (i.e., 2007 and 2010), relative to thematched sample of control firms. However, I may not observe intertempo-ral shifting because this action can be costly: the firm may have capital mar-ket or debt contracting incentives that mitigate shifting (Scholes, Wilson,and Wolfson [1992]; Maydew [1997]). Furthermore, firms may not shiftincome forward if they received greater benefits by retaining income inthe pre-DPAD period to claim other tax incentives phased-out in 2005 and2006.

1062 R. LESTER

My second hypothesis tests whether DPAD firms recharacterize internaltransactions to increase the amount of the DPAD tax benefit. One waythis can occur is for firms to recharacterize foreign income as domesticby changing their internal transfer prices, which can be used to facilitateincome shifting across jurisdictions. For example, a firm’s U.S. divisionscould pay a lower price to foreign subsidiaries for inputs used in domesticmanufacturing, thereby shifting less income into the foreign jurisdictionand increasing total domestic income that qualifies for the DPAD (Jenks[2006]; Sherlock [2012]).2 My second hypothesis (H2) predicts that DPADfirms shift less income out of the United States post-DPAD to increase theamount of qualifying domestic revenue, relative to the matched sample ofcontrol firms. However, because statutory foreign tax rates were generallylower than the effective U.S. tax rate even after taking into account theDPAD benefit, I may observe no such effect. Furthermore, altering a firm’sincome shifting strategies results in transfer pricing adjustment costs (De Si-mone, Klassen, and Seidman [2017]) and important cash tax and financialreporting costs otherwise deferred by retaining foreign earnings offshore(Foley et al. [2007]; Graham, Hanlon, and Shevlin [2011]; Blouin, Krull,and Robinson [2012]).

My third hypothesis tests whether DPAD firms increase real domestic ac-tivity, measured with domestic investment and employment, after claim-ing the DPAD, relative to the matched sample of control firms. Empiri-cal work shows that tax rate decreases are associated with increases in ag-gregate corporate investment (Hassett and Hubbard [2002]; Hassett andNewmark [2008]) and firm employment (Giroud and Rauh [2018]). Theliterature studying U.S. export subsidies (Kemsley [1998]; Desai and Hines[2001]) and related DPAD studies (Blouin, Krull, and Schwab [2014]; Fich,Rice, and Tran [2017]; Ohrn [2018]; Dobridge, Landefeld, and Mortenson[2019]) further suggests a positive relation between claiming the DPADand the amount of domestic investment and employment in the post-DPADperiod. However, firms will forego domestic opportunities if instead thetax incentive results in higher prices for the same capital goods (Goolsbee[1998]) or if U.S. labor and input costs exceed the tax benefit (Suarez Ser-rato and Zidar [2016]). It is unclear if the (at most) effective 3.15% DPADcorporate tax benefit is sufficient to induce such real effects.

I identify DPAD firms from companies’ 2004–2013 annual financial state-ments using several search terms (see appendix A). For each firm, I ob-tain public data from Compustat to construct key variables. I also obtain

2 Jenks [2006] notes that the U.S. government “has to be concerned about taxpayers whostructure their affairs to overstate qualified production activities income through transfer pric-ing strategies; if a company were to arrange things in such a way as to understate its true costsin acquiring good or services from abroad, its Section 199 deduction might go up.” Sherlock[2012] states that “given that production activities are tax favored, firms have an incentive toshift profits among divisions, and characterize income as being related to domestic productionactivities, where possible.”

FIRM RESPONSES TO DOMESTIC PRODUCTION INCENTIVES 1063

confidential data from the Bureau of Economic Analysis (BEA) to preciselymeasure domestic investment and employment, which are otherwise pub-licly unobservable. I then match the sample of DPAD firms with a controlsample of firms that are similar in size, performance, and industry affilia-tion in the two years preceding the DPAD firm’s first discussion of the taxbenefit, but that are precluded from claiming the DPAD due only to a statu-tory limitation contained within the tax law (Lester and Rector [2016]).The final sample includes 21,462 firm-years for 1,002 DPAD firms and theirmatched control firms.

Results from testing my first hypothesis confirm that DPAD firms shift in-come into the first year that the firm claims the DPAD as well as subsequentyears in which the statutory benefit increases. Specifically, the DPAD ben-efit is associated with greater operating income and gross margin shiftingequal to 0.077% and 0.098% of firms’ total assets, relative to the matchedsample of control firms. This is equivalent to shifting $5.0–$13.7 million forthe average DPAD firm in the sample. Additional tests show that this shift-ing begins shortly after the benefit was first available in 2005, documentingthat this is a relatively low-cost approach to maximizing the domestic taxbenefit. I interpret these results as evidence that intertemporal shifting isan important means by which firms increase the amount of income thatqualifies for the tax benefit.

Second, I find that multinational DPAD firms engage in accountingrecharacterization by shifting less income out of the United States afterclaiming the DPAD. I first confirm that both DPAD and the matched con-trol firms shifted income out of the United States in the pre-DPAD period.However, once the incentive is in place, DPAD firms shift less income outof the United States, whereas the matched control firms shift even greateramounts to foreign jurisdictions. A 1.0- to 3.15-percentage-point decreasein the domestic tax rate attributable to the DPAD benefit results in DPADfirms reporting approximately $4.9–$20.6 million of additional income inthe United States post-DPAD, relative to the matched sample of controlfirms. Cross-border shifting does not begin until 2007 when the DPAD in-centive increases to 6% of qualified production income. Furthermore, thechange in income shifting behavior occurs in the subset of DPAD firms withrelatively higher foreign effective tax rates. These results are consistent withDPAD firms increasing the amount of domestic income reported to taxingauthorities, thereby maximizing the production tax benefit.

Finally, I observe that domestic investment by DPAD firms increases rel-ative to the matched sample of control firms, but only for the sample ofdomestic-only firms that lack an existing foreign presence in which to in-vest and expand. These firms report greater capital expenditure spendingequivalent to 7.7–8.4% of total plant, property, and equipment (PPE), orapproximately $95.5–$143.6 million of additional spending by the averagedomestic-only DPAD firm, relative to the matched sample of control firms.This effect is delayed until the greatest DPAD benefit is in place in 2010,consistent with real activities being the costliest response and requiring the

1064 R. LESTER

greatest tax incentive to do so. I find that multinational DPAD firms alsoincrease investment spending by 1.9% (equivalent to $58.1 million), butthis occurs in foreign jurisdictions, suggesting that these firms have lowercost or better growth opportunities abroad. Finally, I find that claimingthe DPAD is associated with a lower level of domestic employment, consis-tent with firms substituting capital for existing domestic labor (Arrow et al.[1961]; Chirinko [2002]; Antras [2004]).

My paper makes several contributions to the academic literature studyingproduction incentives and to policy makers evaluating the effectiveness ofsuch incentives. There is little empirical evidence on the DPAD, despite thelarge number of claimants and the large aggregate amount of tax benefits.I quantify the investment and employment responses to the DPAD, using alarger sample of firms and a longer sample period, relative to three papersthat also study this tax benefit (Blouin, Krull, and Schwab [2014]; Fich,Rice, and Tran [2017]; Ohrn [2018]). Specifically, my use of jurisdiction-specific data and firm-level (rather than industry-level) measurement re-sults in a more precise estimate of the investment effects, which I find are6.0 percentage points lower than documented by Ohrn [2018]. Further, Itest and quantify the employment effects. This evidence can be used to as-sess the policy’s two fundamental goals of increasing domestic investmentand creating jobs.

I next explain why the investment responses are delayed—firms initiallyrespond to the incentive through accounting and reporting channels. Al-though I find that domestic-only firms increased investment spending, theDPAD did not provide sufficiently large incentives to encourage this activityuntil 2010. Instead, firms initially engaged in intertemporal shifting begin-ning in 2005 and 2006. For the multinationals in the sample, I observeless income shifting out of the United States from 2007 to 2009, which isthe same period in which I observe greater foreign investment spending.These results suggest that the cross-border shifting generates domestic taxsavings that are associated with investment offshore.

Finally, I provide evidence to inform expectations regarding the antici-pated effects of the TCJA. Among other provisions, the 2017 act loweredthe U.S. corporate income tax rate from 35% to 21%. Because the Section199 deduction was viewed as a proxy for a lower corporate tax rate (Ohrn[2018]), my results suggest how companies may respond to the new law.I show that a subset of DPAD firms with the highest foreign tax burdenswere motivated to shift less income out of the United States in responseto a 3.15 percentage point effective corporate income tax rate reduction.Thus, my evidence implies that the larger U.S. corporate tax rate cut in2017 may be successful in retaining income that multinationals with rel-atively higher tax burdens would otherwise shift offshore. I further showthat the real activity responses to the tax reform will vary based on a firm’sdomestic presence. Although the immediate expensing of fixed assets in-cluded in the TCJA will certainly affect domestic investment decisions,the results show that the reduction in the corporate rate could motivate

FIRM RESPONSES TO DOMESTIC PRODUCTION INCENTIVES 1065

foreign investment spending by some multinational firms. Finally, the empir-ical results support recent press evidence that the United States is unlikelyto experience significant increases in domestic employment, in part to dueto the substitution of capital for labor (Tangel and McGroarty [2018]). Mystudy shows that firms have responded to tax incentives with real changesin corporate investment, but that accounting and workforce effects are sta-tistically and economically significant as well.

2. Institutional Details

Since 1971, the United States has provided special tax breaks to domesticexporters that deferred or exempted some portion of foreign income fromU.S. taxation. These tax incentives were intended to help domestic compa-nies compete with lower taxed foreign firms. (See online appendix A for adiscussion of these regimes.) In response to pressure by the internationalcommunity, which had challenged these tax benefits as uncompetitive ille-gal export subsidies, the United States passed the American Jobs CreationAct in 2004. This law replaced the existing export tax incentive with theDPAD.3 Unlike the export tax subsidies, for which firms had to earn someforeign income to claim the benefit, any U.S. firm that had qualifying do-mestic production income was eligible for the DPAD.

Lester and Rector [2016] discuss the deduction, including the legisla-tive history of the law, the mechanics of the calculation, and details of thefirms that claimed the Sec. 199 deduction in tax year 2012. In brief, thededuction equals a percentage of the firm’s qualified production activitiesincome (QPAI), which is calculated as follows:

Revenues (for tax purposes) from the sale of domestically produced goodsLess: COGS (for tax purposes) attributable to domestic productionLess: Other expenses (for tax purposes) allocable to domestic production

Equals: QPAI

The deduction equals QPAI, times the statutory deduction percentageof 3% (2005–2006), 6% (2007–2009), or 9% (2010 and thereafter). Thebenefit is limited by (1) the firm’s wages and (2) the amount of the firm’staxable income. Specifically, the deduction cannot exceed 50% of the firm’sW-2 wages related to production revenues, and firms must report positive

3 Legislative history confirms that the DPAD was intended to provide relief in response tothe Extraterritorial Income Exclusion (ETI) repeal. H.R. Rep. No. 108-548 [2004b], Part 1,states: “The Committee [on Ways and Means] believes that it is important to use the opportu-nity afforded by the repeal of the ETI regime to reform the U.S. tax system in a manner thatmakes U.S. businesses and workers more productive and competitive than they are today. Tothis end, the Committee believes that it is important to provide tax cuts to U.S. domestic man-ufacturers and to update the U.S. international tax rules, which are over 40 years old and makeU.S. companies uncompetitive in the United States and abroad.” The AJCA also included arepatriation tax holiday (see online appendix B).

1066 R. LESTER

taxable income to claim the deduction. Firms with negative taxable income(from a current year tax loss) or no taxable income (due to use of a prioryear loss carryforward) are ineligible. I exploit this latter limitation, whichprecludes otherwise eligible firms from claiming the DPAD, to constructthe control sample.

Many firms report a Section 199 deduction, and the aggregate amount ofthe benefit is economically significant. For example, approximately 47,000C corporations filed a DPAD tax form (IRS Form 8903) in 2012, and thesecorporations reported $685 million, or 80%, of the total taxable incomereported by all U.S. C corporations in 2012. The aggregate 2012 Section199 deduction claimed by C corporations was $32 billion (estimated taxsavings of $11 billion), of which publicly traded companies accounted for75.0% (Lester and Rector [2016]).

Although Congress intended the tax incentive to help the manufacturingsector, approximately 40.0% of publicly traded C corporations that claimthe deduction are in nonproduction industries and include wholesale, re-tail, information, financial services, utilities, and services firms.4 Given thesize, scope, and complexity of this incentive, the Internal Revenue Service(IRS) identified the DPAD as a key audit area, and the DPAD was one ofthe top four items listed by firms when reporting Uncertain Tax Positionsto tax authorities (Internal Revenue Service [2015]).

3. Research Design

3.1 RESEARCH DESIGN FOR TESTS OF INTERTEMPORAL SHIFTING

My first hypothesis (H1) predicts that DPAD firms will shift more incomeinto the first year of claiming the deduction, as well as any subsequentyear in which the benefit increases, relative to a matched sample of con-trol firms. I test this hypothesis with the following OLS regression:

TimeShifti,t = α + β1DPADFirmi + β2ShiftPeriodi,t

+ β3DPADFirmi ∗ ShiftPeriodi,t + Controlsi,t − 1

+ εi,t . (1)

TimeShifti,t is the amount of shifted gross margin (Gross Margin Shifti,t) orshifted operating income (Op Inc Shifti,t). Following Maydew [1997], I cal-culate a firm’s gross margin and operating income using quarterly data forthe first quarter of year t (e.g., the first quarter of 2005) and for the fourthquarter of year t – 1 (e.g., the fourth quarter of 2004). Gross margin equals

4 Firms in these industries may claim some benefit if they identify certain qualifying incomewithin their business. For example, Starbucks is permitted to claim the deduction for grossreceipts related to the sale of coffee beans that it roasts and packages, whereas receipts fromthe sale of brewed coffee qualify only to the extent of the value of the roasted beans used inthe brewing process (H.R. Rep. No. 108-755 [2004c]).

FIRM RESPONSES TO DOMESTIC PRODUCTION INCENTIVES 1067

the firm’s total sales (SALEQ), less total cost of goods sold (COGSQ); op-erating income is the firm’s gross margin, less SG&A expense (XSGAQ),where missing values of SG&A are set equal to 0. I use the amount of third-quarter gross margin or operating income prior to the expected shiftingquarter as the baseline income. I then measure the amount of shifted in-come by separately comparing the following two quarters of earnings tothis baseline. The overall estimate of Gross Margin Shifti,t and Op Inc Shifti,tis one half of the negative of the fourth-quarter shift, plus the first-quarterreversal, scaled by the prior year’s total assets.5 I multiply these values by100 for ease of interpretation and use the scaled variables (GMShift%i,t andOpIncShift%i,t) when estimating equation (1).

DPADFirmi is an indicator equal to 1 for firms that discuss the DPAD in atleast one of their annual financial statements, and 0 otherwise. ShiftPeriodi,t

is a firm-specific indicator equal to 1 for both a DPAD firm and its matchedcontrol firm in the first year in which the former discloses the benefit, aswell as any subsequent year in which the statutory DPAD benefit increases.6

The term DPADFirmi∗ShiftPeriodi,t is the interaction of the prior two terms.

The coefficient β3 captures the different intertemporal shifting by DPADfirms in the periods with an intertemporal shifting incentive and is pre-dicted to be positive.

Control variables follow those of Maydew [1997], including Leveragei,t − 1

and RD Crediti,t − 1, both of which control for other tax benefits the firmmay claim. Leveragei,t − 1 is total firm debt (DLTT + DLC) scaled by the

5 Following Maydew [1997], Gross Margin Shifti,Q4 = G0,4 – G0,3, where the first term isthe amount of shifted gross margin in the fourth quarter of the prior year, and the secondterm is the unshifted income in the third quarter of the prior year (“benchmark” quarter).Gross Margin Shifti,Q1 = G1,1 – G0,3, which is the total shifted income in the first quarterof year t, less the benchmark gross margin. Total shifted income is (Gross Margin Shifti,Q1 –Gross Margin Shifti,Q4)/2. For example, assume a firm would normally report (without any in-tertemporal shifting) gross margin of $50 in the third and fourth quarters of 2004 as well asthe first quarter of 2005. In response to the DPAD, the firm shifts $30 into the first quarterof 2005, such that the gross margin in the third and fourth quarters of 2004 is $50 and $20($50 unshifted minus $30 shifted), respectively, and the gross margin in the first quarter of2005 is $80 ($50 unshifted plus $30 shifted). Total shifted income equals (($80 – $50) – ($20– $50))/2, or $30. Op Inc Shifti,t is calculated in the same manner, and results are robust todropping missing values of XSGA as well as including R&D expense (XRDQ).

6 Although the law permitted firms to begin claiming the DPAD in 2005, they may not ac-tually begin to take the deduction until a later year, due to lack of qualifying revenue, netoperating loss position, or other statutory limitations (such as not meeting the required W-2wage threshold). Therefore, I define ShiftPeriodit based on the initial year in which the firmdiscusses the benefit in its financial statements, as well as any subsequent period in which thefirm has incentives to shift income. For example, if a DPAD firm first discusses the DPAD ben-efit in 2006, then ShiftPeriodit equals 1 in 2006, 2007, and 2010 for both the DPAD firm andits matched control; it equals 0 in years 1997–2005, 2008–2009, and 2011–2013. Defining theindicator as a function of the initial year the benefit is discussed likely results in measurementerror because firms could be claiming the benefit in prior period but not discussing or dis-closing it. Consequently, I may inadvertently define some years as “pre-DPAD,” even thoughthe firm actually claimed the DPAD in that year.

1068 R. LESTER

market value of equity (PRCC F∗CSHO) in year t – 1. RD Crediti,t − 1 is anindicator equal to 1 for firms that report a positive value for the sum ofthe firm’s estimated R&D tax credit over the three preceding years, scaledby domestic tax expense for the same period.7 Sizei,t − 1, calculated as thenatural logarithm of total assets (AT) in year t – 1, and MTBi,t − 1, the ratioof the market value of equity to the book value of equity (SEQ) in year t –1, account for variation in shifting related to size and growth opportunities.ROAi,t − 1 is calculated as operating income (OIADP) scaled by the firm’sassets in year t – 1 and controls for firm performance that may affect oper-ating income and gross margin as well as the firm’s taxable income. I alsoinclude industry-by-year fixed effects and report standard errors clusteredby industry. Online appendix C shows that results are robust to alternativestandard error corrections.

3.2 RESEARCH DESIGN FOR TESTS OF CROSS-BORDER SHIFTING

My second hypothesis (H2) predicts that DPAD firms shift less incomeout of the United States, post-DPAD, relative to a matched sample of con-trol firms. I test this prediction using a model from Collins, Kemsley, andLang[1998] that was adapted by Klassen and Laplante [2012], who measureincome shifting over multiple periods so as to reduce measurement errorrelated to annual fluctuations in reported accounting numbers. I estimatethe following equation:

DomROSi,(t,t+n) = α + β1DPADFirmi + β2PostDPADi,t

+ β3DPADFirmi ∗ PostDPADi,t + β4RateDiff i,(t,t+n)

+ β5PostDPADi,t ∗ RateDiff i,(t,t+n)

+ β6DPADFirmi ∗ RateDiff i,(t,t+n)

+ β7DPADFirmi ∗ PostDPADi,t ∗ RateDiff i,(t,t+n)

+ Controlsi,(t,t+n) + εi,(t,t+n), (2)

where DPADFirmi is as defined above and DomROSi,(t,t + n), PostDPADi,t,RateDiffi,(t,t + n), and Controlsi,(t,t + n) are discussed below. I include industry-by-year fixed effects, and I report standard errors that are clustered byindustry.

7 I calculate the average R&D tax benefit by dividing the sum of the firm’s R&D tax creditover the prior three years by the sum of a firm’s domestic tax expense (TXFED) over thesame period. I use firms’ disclosed R&D tax credit for years 1997–2010 from Hoopes (2018).I supplement these data with estimates of the R&D tax credit for 2011–2013 based on the Al-ternative Simplified Credit methodology under IRC Sec. 41(c)(5). Specifically, for each firm-year from 2011 through 2013, I calculate the average R&D expense (XRD) over the preced-ing three years and multiply this amount by 50%. I then subtract this average (base periodamount) from the current year’s R&D expenses and multiply the difference by 14% to esti-mate a firm’s R&D tax credit.

FIRM RESPONSES TO DOMESTIC PRODUCTION INCENTIVES 1069

The dependent variable is a firm’s domestic return on sales,(DomROSi,(t,t + n)), calculated as total pre-tax domestic income (PIDOM) di-vided by total domestic segment sales measured over both a two-year andfive-year period (n = 1 or 4).8 It implicitly assumes that the amount ofpre-tax domestic income should increase relative to the total amount of do-mestic sales after tax rates decrease because firms have greater incentivesto report income in the reduced-tax jurisdiction. Although the location ofsales does not actually determine the appropriate taxing jurisdiction, I usethis measure because I assume that sales are a valid benchmark for the lo-cation of the firm’s economic activity.

Collins, Kemsley, and Lang [1998] and Klassen and Laplante [2012]show that income reported in a foreign jurisdiction increases with the dif-ference in the U.S. statutory rate and firms’ lower foreign effective taxrates. By extension, DomROSi,(t,t + n) should be negatively associated with thedomestic-foreign tax rate differential prior to the DPAD; that is, as the dif-ference between the U.S. statutory rate and firms’ lower effective foreigntax rate increases, firms should report less income in the United States. Theterm Rate Diffi,(t,t + n) is equal to the difference in the U.S. statutory rate of35% during the sample period and a firm’s average foreign effective taxrate, which is calculated as the sum of foreign income tax expense (TXFO)from year t to t + n, divided by the firm’s foreign income (PIFO) over thesame period. During the pre-DPAD period, I expect that both DPAD firmsand control firms shift income out of the United States; thus, β4 (for con-trol firms) and β4+ β6 (for DPAD firms) should be negative.

After the DPAD is in place, the effective U.S. tax rate on domesticproduction income decreases for qualifying firms. Therefore, I expect toobserve an attenuated negative relation (a positive coefficient) betweenDomROSi,(t,t + n) and Rate Diffi,(t,t + n) for DPAD firms, relative to the matchedset of control firms, post-DPAD. To test this prediction, I include several ad-ditional terms in equation (2). The term PostDPADi,t is a firm-specific indica-tor equal to 1 for both a DPAD firm and its matched control firm beginningin the year in which the DPAD firm first discusses the DPAD benefit.9 I alsoinclude the interactions of DPADFirmi, Rate Diffi,(t,t + n), and PostDPADi,t; thetriple interaction term DPADFirmi

∗PostDPADi,t∗Rate Diffi,(t,t + n) captures the

change in the sensitivity of DPAD firms’ domestic income to the tax ratedifferential post-DPAD, relative to the control sample of firms. I predictβ7 > 0.

8 Klassen and Laplante [2012] measure income shifting over a five-year period. I also testequation (2) using variables measured over a two-year period to allow for estimation on alarger sample of observations.

9 For example, if a DPAD firm first discusses the DPAD benefit in 2006, then PostDPADitequals 1 for 2006–2013 (i.e., all years in the post-period) for both the DPAD firm and itsmatched control. This differs from the research design in equation (1), as I expect intertem-poral shifting to only be present in discrete periods when the effective tax rate on productionincome changes, whereas I expect cross-border shifting effects to persist throughout the post-period.

1070 R. LESTER

Equation (2) includes a measure of a firm’s worldwide return on sales(WW ROSi,(t,t + n)), which controls for the firm’s overall profit margin andis calculated as total pre-tax income (PI) divided by total sales (SALE). Ad-ditional control variables follow Dyreng and Markle [2016] and includeSizei,(t,t + n), Leveragei,(t,t + n), RDi(t,t + n), Cashi,(t,t + n), and the percentage of do-mestic sales (%Dom Salesi,(t,t + n)). RDi,(t,t + n) is calculated as R&D expense(XRD) divided by the firm’s total assets and captures a firm’s ability to shiftincome due to placement of intellectual property (created through R&Dactivity) in foreign jurisdictions. Cashi,(t,t + n) is total cash holdings (CHE) di-vided by the firm’s total assets and reflects that firms’ income shifting maybe affected by offshore cash holdings. %Dom Salesi,(t,t + n) controls for thelevel of domestic sales and is calculated as the amount of domestic segmentsales to total worldwide sales.

3.3 RESEARCH DESIGN FOR TESTS OF REAL ACTIVITY

My third hypothesis predicts that firms will invest more (H3a) and em-ploy more workers (H3b) in the United States, post DPAD, relative to amatched sample of control firms. To test these predictions, I estimate thefollowing OLS regression:

DomActivityi,t = α + β1DPADFirmi + β2PostDPADi,t

+ β3DPADFirmi ∗ PostDPADi,t + Controlsi,t − 1 + εi,t , (3)

where DPADFirmi, PostDPADi,t, and DPADFirmi∗PostDPADi,t are as defined

above, and DomActivityi,t and Controlsi,t − 1 are discussed below. As before,I include industry-by-year fixed effects and report standard errors thatare clustered by industry. The coefficient β3 captures the difference-in-differences estimate and is predicted to be positive.

DomActivityi,t is measured in the following two ways: (1) DomInvi,t, the ra-tio of domestic capital expenditures to total PPE in year t – 1 (Almeidaand Campello [2007]), and (2) DomEmpi,t, the log of the number of do-mestic employees. For firms with only domestic operations, I use Compus-tat data on capital expenditures (CAPX), PP&E (PPENT), and employees(EMP). For multinationals, I use confidential firm-level panel data on theU.S. parent’s domestic capital expenditures, fixed assets, and employeesfrom the BEA benchmark and annual surveys of U.S. Direct InvestmentAbroad. These surveys include financial, trade, and intercompany transac-tion data on U.S. parent companies’ domestic activity as well as the activityof all foreign affiliates in which the U.S. parent owns at least 10% of thevoting securities or an equivalent interest.10

10 The BEA data are collected for the purpose of producing publicly available aggregatestatistics on the operations of U.S. multinational companies. Due to penalties for failure tofile, the BEA believes that the data coverage is substantially complete and accurate. The levelof data collected varies based on certain BEA reporting thresholds.

FIRM RESPONSES TO DOMESTIC PRODUCTION INCENTIVES 1071

I select control variables for investment and employment following Mc-Nichols and Stubben [2008], Biddle, Hilary, and Verdi [2009], and Edger-ton [2010]. I include Sizei,t − 1 and ROAi,t − 1 to capture differences in in-vestment and employment based on size and performance of the firm.Cash Flowi,t − 1, calculated as a firm’s operating cash flows (OANCF) di-vided by total assets in year t – 1, captures a firm’s ability to finance in-vestment and hire employees out of operating cash flows. MTBi,t − 1 (asa proxy for Tobin’s Q) controls for a firm’s investment opportunity set,and Leveragei,t − 1 controls for investment financing. The employment testsalso include Inventoryi,t − 1 and Tangibilityi,t − 1, calculated as total inventory(INVT) or equipment (PPENT) scaled by total assets in year t – 1, as capital-intensive firms may have a different demand for workers.

4. Sample

4.1 DPAD FIRMS

Table 1, panel A, outlines the sample selection steps. I identify firmslikely to claim the DPAD by first obtaining and reading text disclosuresthat contain key search terms from publicly traded firms’ 2004–2013 an-nual financial statements (n = 5,153 firm-years for 1,494 firms). I retainU.S.-incorporated firms and restrict the sample to firm-years with positivepretax income (as a proxy for unobservable taxable income) to reflect therequirement that firms with current year losses cannot claim the incentive.I also require firm-years to have positive total assets, positive total sales, andnonmissing data for construction of control variables and the intertempo-ral shifting variables used to test H1 (n = 3,821 firm-years; 1,125 firms).Finally, I require DPAD firms to have at least $100 million of average assetsover the sample period, as IRS data show that these firms claimed approx-imately 88.0% of the total Sec. 199 deduction reported by all public firmsin 2012 (Lester and Rector [2016]). The final sample is composed of 1,002distinct DPAD firms from 2004 to 2013.

To confirm that this sample of firms represents the true population ofSec. 199 firms, I compare descriptive statistics for this sample to 2012 IRSstatistics.11 Table 1, panel B, shows that the industry distribution is similar;for example, approximately 56.0% (59.5%) of firms in the sample (the IRSsample) are in manufacturing, and 7.4% (7.2%) of firms in the sample (theIRS sample) are in information industries. The samples differ in the pro-portion of wholesale and services firms. In panel C, approximately 56.6%

11 The IRS statistics are calculated by an employee of the U.S. Treasury using 2012 con-fidential tax return data (Lester and Rector [2016]). For purposes of comparing the 2004–2013 sample used in this paper to the 2012 IRS statistics, I select the last year that a DPADfirm is in the sample. Differences across the samples may be attributable to differences inclassification across these two data sources; for example, firms self-report industry codes inCompustat, whereas the IRS assigns codes to some companies.

1072 R. LESTER

TA

BL

E1

Iden

tifica

tion

ofD

PAD

Firm

Sam

ple

and

Firm

Cha

ract

eris

tics

Pan

elA

:Ide

ntifi

cati

onof

DPA

Dfi

rms

DPA

DFi

rm-Y

ears

DPA

DFi

rms

Dat

aR

equi

rem

ents

toId

enti

fySe

c.19

9Fi

rms

Obs

erva

tion

sD

ropp

edO

bser

vati

ons

Rem

ain

ing

Obs

erva

tion

sD

ropp

edO

bser

vati

ons

Rem

ain

ing

Init

ials

ampl

eof

DPA

Dfi

rms

inC

ompu

stat

5,15

31,

494

Elim

inat

en

on-U

.S.i

nco

rpor

ated

firm

s(6

3)5,

090

(19)

1,47

5E

limin

ate

obse

rvat

ion

sw

ith

mis

sin

gda

tato

calc

ulat

eva

riab

les

(1,2

69)

3,82

1(3

50)

1,12

5E

limin

ate

obse

rvat

ion

sfo

rfi

rms<

$100

mill

ion

inas

sets

(367

)3,

454

(123

)1,

002

Tota

lfirm

-yea

rs(fi

rms)

3,45

41,

002

Pan

elB

:Com

pari

son

ofD

PAD

firm

sto

2012

IRS

stat

isti

cs—

indu

stry

dist

ribu

tion

DPA

DFi

rms

Indu

stry

Des

crip

tion

Num

ber

ofO

bser

vati

ons

Obs

erva

tion

s(%

)20

12IR

SSt

atis

tics

Obs

erva

tion

s(%

)

Man

ufac

turi

ng

(SIC

2000

–399

9)56

156

.0%

59.5

%In

form

atio

n(S

IC36

50–3

670,

4800

)74

7.4%

7.2%

Wh

oles

ale

and

Ret

ail(

SIC

5000

–599

9)41

4.1%

10.6

%Fi

nan

cial

(SIC

6000

–620

0,67

00)

212.

1%3.

3%Se

rvic

es(S

IC70

00–8

999)

124

12.4

%8.

1%O

ther

indu

stri

es18

118

.0%

11.3

%

Tota

lfirm

s1,

002

100.

0%10

0.0%

(Con

tinue

d)

FIRM RESPONSES TO DOMESTIC PRODUCTION INCENTIVES 1073

TA

BL

E1—

Con

tinue

d

Pan

elC

:Com

pari

son

ofD

PAD

firm

sto

2012

IRS

stat

isti

cs—

size

cate

gori

es

DPA

DFi

rms

Size

Cat

egor

yN

umbe

rof

Obs

erva

tion

sO

bser

vati

ons

(%)

2012

IRS

Stat

isti

csO

bser

vati

ons

(%)

$100

–$25

0m

illio

n10

910

.9%

11.2

%$2

50–$

500

mill

ion

153

15.3

%13

.5%

$500

mill

ion

–$1

billi

on17

317

.2%

15.6

%G

reat

erth

an$1

billi

on56

756

.6%

59.7

%

Tota

lfirm

s1,

002

100.

0%10

0.0%

Pan

elD

:Com

pari

son

ofD

PAD

firm

sto

2012

IRS

stat

isti

cs—

prop

orti

onof

firm

sw

ith

fore

ign

pres

ence

DPA

DFi

rms

MN

Cs

asPe

rcen

tage

ofSa

mpl

eN

umbe

rof

Obs

erva

tion

sO

bser

vati

ons

(%)

2012

IRS

Stat

isti

csO

bser

vati

ons

(%)

Dom

esti

c-on

lyFi

rms

215

21.5

%18

.5%

Mul

tin

atio

nal

Firm

s78

778

.5%

81.5

%

Tota

lfirm

s1,

002

100.

0%10

0.0%

Th

ista

ble

pres

ents

the

sele

ctio

nst

eps

toid

enti

fyth

eD

PAD

firm

s(p

anel

A)

asw

ell

asde

scri

ptiv

est

atis

tics

tova

lidat

eth

atth

eD

PAD

sam

ple

rese

mbl

esth

etr

uepo

pula

tion

ofpu

blic

lytr

aded

DPA

Dfi

rms

(pan

els

B,C

,an

dD

).In

pan

elA

,afi

rmis

incl

uded

asa

“DPA

Dfi

rm”

ifit

disc

usse

sth

eD

PAD

inan

yof

its

ann

ualfi

nan

cial

stat

emen

tsfr

om20

04(t

he

year

inw

hic

hth

eD

PAD

was

enac

ted)

thro

ugh

2013

.Th

esa

mpl

ein

clud

esfi

rm-y

ear

obse

rvat

ion

sfo

rD

PAD

firm

sth

atm

eet

the

follo

win

gre

quir

emen

ts:(

1)U

.S.-i

nco

rpor

ated

;(2)

data

toca

lcul

ate

inte

rtem

pora

lsh

ifti

ng

mea

sure

san

dco

ntr

olva

riab

les

for

H1,

incl

udin

gpo

siti

veto

tala

sset

s,sa

les,

and

pre-

tax

inco

me;

and

(3)

grea

ter

than

$100

mill

ion

ofas

sets

.Pa

nel

B(C

,D)

com

pare

sin

dust

ry(s

ize,

fore

ign

pres

ence

)st

atis

tics

to20

12IR

Sst

atis

tics

from

Les

ter

and

Rec

tor

[201

6].S

ize

isde

term

ined

base

don

tota

lass

ets.

Afi

rmis

iden

tifi

edas

am

ulti

nat

ion

alin

the

IRS

sam

ple

base

don

repo

rtin

gof

afo

reig

naf

filia

te,s

ubsi

diar

y,or

fore

ign

pare

ntt

oth

eU

.S.t

axau

thor

itie

s;to

iden

tify

asi

mila

rse

tofm

ulti

nat

ion

alfi

rms,

afi

rmin

this

sam

ple

isid

enti

fied

asa

mul

tin

atio

nal

base

don

aba

sed

onei

ther

fin

anci

alst

atem

entd

iscl

osur

eof

atle

asto

ne

mat

eria

lfor

eign

subs

idia

ryus

ing

Exh

ibit

21da

ta(D

yren

gan

dL

inds

ey[2

009]

),n

onm

issi

ng,

non

zero

fore

ign

segm

ents

ales

,or

non

mis

sin

g,n

onze

rofo

reig

nse

gmen

tass

ets.

1074 R. LESTER

(59.7%) of firms in the sample (the IRS sample) report over $1 billion inassets.

Lester and Rector [2016] show that public multinational firms reportapproximately 95% of QPAI in 2012. Given this high percentage and therelative importance of MNCs as DPAD claimants, I next compare the pro-portion of firms reporting an international presence across the two samplesin panel D. IRS statistics report that 81.5% of public DPAD claimants havesome foreign activity, where a firm is identified as an MNC based on re-ported ownership of or by a foreign entity in the tax data. To construct ananalogous measure, I identify a firm as a multinational based on financialstatement disclosure of at least one material foreign subsidiary in Exhibit21 data (Dyreng and Lindsey [2009]). I further identify any firm report-ing nonmissing, nonzero foreign segment sales or foreign segment assetsas a multinational firm. This methodology results in a very similar percent-age of firms with a foreign presence (78.5%), validating that this approachmost closely aligns with the domestic/multinational classification based onreported tax data. However, because this definition uses foreign segmentsales data that may inaccurately identify some domestic exporters as multi-national firms, I alternatively calculate the proportion of firms identified asdomestic-only based on zero or missing values for pre-tax foreign income(PIFO) and foreign tax expense (TXFO). This latter methodology resultsin a much lower 70.7% of the DPAD sample being classified as a multina-tional, implying some misclassification of firms relative to the IRS definitionused in Lester and Rector [2016]. I present results throughout the paperfor both domestic-only samples.

In summary, the statistics presented in table 1, panels B–D, confirm thatthe sample constructed from financial statement disclosures resembles thetrue population of publicly traded firms claiming the Section 199 deduc-tion.

4.2 POOL OF POSSIBLE CONTROL FIRMS

I exploit a statutory limitation in the tax law to construct the controlsample. Recall from section 2 that firms that are otherwise eligible for thededuction—profitable firms with qualifying production income—are pre-cluded from claiming the DPAD if the firm uses a tax net operating loss(NOL) carryforward generated in a prior tax year. IRS statistics that com-pare DPAD firms to nonclaimant firms provide evidence of this controlgroup: in 2012, over 1,000 large, publicly traded C corporations would havebeen eligible for the DPAD, except for the fact that these firms reported noU.S. taxable income due to use of an NOL (table 6 of Lester and Rector[2016]).

The literature shows that many firms have an NOL carryforward at somepoint in a firm’s life cycle (Cooper and Knittel [2010]; Edgerton [2010]),and recent work documents that almost 90% of the largest public com-panies report an NOL in at least one jurisdiction (Heitzman and Lester[2019]). Thus, the existence and use of a loss carryforward should not

FIRM RESPONSES TO DOMESTIC PRODUCTION INCENTIVES 1075

imply that the DPAD firms and control firms inherently differ. This differ-ence instead suggests that the DPAD and control firms have simply incurredand used NOL carryforwards at different times.

To construct the sample of possible control firms, I retain observations inCompustat from 2004 to 2013 that meet the same data restrictions as thoselisted for DPAD firms (U.S.-incorporated, positive pretax income, data forkey variables, and $100 million of assets). I exclude any firm from this sam-ple if it ever discuss the DPAD during the period 2004–2013.12

4.3 MATCHING PROCEDURE AND VALIDATION OF CONTROL FIRMS

I match DPAD firms to control firms in the year preceding a DPAD firm’sfirst discussion of the tax benefit. I require an exact match on industry toensure that control firms have similar types of income that qualify for theDPAD. I use Mahalanobis distance measures to match the DPAD firms tocontrol firms with replacement based on Sizei and firm performance (ROAi)in year t – 1; online appendix D shows that results are robust to alterna-tive matching procedures. The matching produces a final sample of 21,462matched DPAD and control firm-years for the period 1997–2013. Table 2,panel A, provides a summary of this sample by year.13 Table 2, panel B,shows the sample construction for each of the three hypotheses, startingwith the 21,462 firm-years from H1; imposing additional data requirementsand requiring matches for the treatment firms results in samples of 4,605(H2), 8,274 (H3a), and 8,296 (H3b) observations, respectively. This panelalso provides the number of domestic-only and multinational observationsfor each test.

Comparison of descriptive statistics in table 3, panel A, validates that thematching procedure produces control firms similar to the DPAD firms. Inaddition to requiring an exact match on industry, there are no statisticallysignificant differences in the mean values of the matching variables Size orROA between the DPAD and control firm samples in the two pretreatmentyears. Furthermore, the samples have similar means on all other variables

12 Although I attempt to identify all DPAD firms by searching firm financial statements withmany different search terms, this method results in a sample selection bias because I onlyobserve firms that discuss the benefit in their financial statements. As a result, the controlsample may be contaminated if some of these firms actually claim DPAD but do not disclose.Because my tests compare DPAD firms’ responses to the contaminated control sample, thiscontamination is likely to bias against finding results for the DPAD firms, although the actualsign or magnitude of the bias is not estimable.

13 Table 1, panel A, shows that there are 3,454 firm-years in which a firm specifically discussesor discloses the deduction in its financial statements. Because I treat all years following the firstdisclosure of the DPAD as a post-DPAD firm year for tests of H2 and H3, I classify an additional1,292 firm-years (for 4,746 total firm-years in table 2, panel A) as post-DPAD. This assumes that(1) once a firm is eligible, it will continue to claim the benefit in future periods, and (2) to theextent the firm incurs a tax loss that precludes it from claiming the deduction in a subsequentyear, it will return to claiming the DPAD once the tax loss is used. This design also capturescross-border and real activity effects, which could be delayed relative to the initial year in whichthe DPAD is claimed.

1076 R. LESTER

TA

BL

E2

Sam

ple

ofD

PAD

and

Con

trol

Firm

s

Pan

elA

:DPA

Dan

dco

ntro

lfirm

-yea

rob

serv

atio

ns

DPA

DFi

rms

Con

trol

Firm

sTo

talF

irm

-Yea

rs

Year

Num

ber

ofFi

rms

Num

ber

ofFi

rm-Y

ears

inPo

st-D

PAD

Peri

od

Num

ber

ofFi

rm-Y

ears

(Pre

-an

dPo

st-D

PAD

)N

umbe

rof

Firm

s

Num

ber

ofFi

rm-Y

ears

inPo

st-D

PAD

Peri

od

Num

ber

ofFi

rm-Y

ears

(Pre

-an

dPo

st-D

PAD

)

Num

ber

ofFi

rm-Y

ears

Post

-DPA

D

Num

ber

ofFi

rm-Y

ears

(Pre

-an

dPo

st-D

PAD

)

1997

–46

8–

465

–93

319

98–

571

–52

3–

1,09

419

99–

588

–58

4–

1,17

220

00–

616

–59

4–

1,21

020

01–

589

–52

1–

1,11

020

02–

596

–55

6–

1,15

220

03–

658

–64

1–

1,29

920

0452

5273

052

5272

010

41,

450

2005

385

427

763

385

425

789

852

1,55

220

0693

469

761

9344

075

890

91,

519

2007

107

502

733

107

479

727

981

1,46

020

0843

465

642

4340

059

386

51,

235

2009

3945

059

439

401

578

851

1,17

220

1065

563

705

6550

466

11,

067

1,36

620

1166

602

709

6651

264

61,

114

1,35

520

1262

616

684

6252

360

01,

139

1,28

420

1390

600

600

9049

949

91,

099

1,09

9

Tota

l1,

002

4,74

611

,007

1,00

24,

235

10,4

558,

981

21,4

62

(Con

tinue

d)

FIRM RESPONSES TO DOMESTIC PRODUCTION INCENTIVES 1077

TA

BL

E2—

Con

tinue

d

Pan

elB

:Fir

m-y

ear

sam

ple

for

test

sof

H1–

H3

H1

H2

H3a

H3b

Inte

rtem

pora

lSh

ifti

ng

Cro

ss-C

oun

try

Shif

tin

gIn

vest

men

tEff

ects

Em

ploy

men

tEff

ects

Sam

ple

ofD

PAD

and

con

trol

firm

-yea

rs,1

997–

2013

21,4

6221

,462

21,4

6221

,462

Les

s:O

bser

vati

ons

mis

sin

gda

tato

calc

ulat

ede

pen

den

tan

dco

ntr

olva

riab

les

–(1

4,55

9)(9

,014

)(9

,011

)

Les

s:O

bser

vati

ons

wit

hou

tmat

ched

firm

afte

rda

tare

stri

ctio

ns

impo

sed

–(2

,298

)(4

,174

)(4

,155

)

Tota

lfirm

-yea

rob

serv

atio

ns

for

test

sof

H1–

H3

21,4

624,

605

8,27

48,

296

Dom

esti

c-on

lyob

serv

atio

ns

4,45

6–

2,68

92,

731

Mul

tin

atio

nal

obse

rvat

ion

s17

,006

4,60

55,

585

5,56

5

Th

ista

ble

pres

ents

the

sam

ple

com

posi

tion

and

sam

ple

sele

ctio

nst

eps.

Afi

rmis

iden

tifi

edas

a“D

PAD

firm

”if

itdi

scus

ses

the

DPA

Din

any

ofit

san

nua

lfi

nan

cial

stat

emen

tsfr

om20

04(t

he

year

inw

hic

hD

PAD

was

enac

ted)

thro

ugh

2013

.Pan

elA

show

sth

en

umbe

rof

DPA

Dan

dm

atch

edco

ntr

olob

serv

atio

ns

byye

ar.T

he

num

ber

ofD

PAD

firm

-yea

rsin

the

post

-DPA

Dpe

riod

incl

udes

allo

bser

vati

ons

for

the

DPA

Dfi

rmfo

llow

ing

the

firs

tti

me

this

tax

ince

nti

veis

disc

usse

din

the

fin

anci

alst

atem

ents

(3,4

54fi

rm-y

ears

from

tabl

e1,

plus

1,29

2ad

diti

onal

year

sin

the

post

-per

iod)

.Eac

hD

PAD

firm

ism

atch

edto

aco

ntr

olfi

rmon

Size

,per

form

ance

(RO

A),

and

indu

stry

inth

eye

arpr

eced

ing

the

firs

tye

arin

wh

ich

the

firm

disc

usse

sth

eD

PAD

.Pan

elB

outl

ines

the

sele

ctio

nst

eps

for

the

sam

ples

used

tote

stea

chof

the

thre

eh

ypot

hes

esaf

ter

the

corr

espo

ndi

ng

data

rest

rict

ion

sar

eim

pose

d.A

ppen

dix

Bpr

ovid

esde

fin

itio

ns

ofal

lvar

iabl

es.

1078 R. LESTER

T A B L E 3Matching Characteristics and Parallel Trends Assumption

Panel A: Comparison of matching characteristics for DPAD firms and control firms in thepre-DPAD periods

DPAD Firms Control Firms Difference

Variables Mean Median Mean Median Mean Median

Sizei,t − 1 6.935 6.704 6.932 6.746 0.003 −0.042ROAi,t − 1 0.112 0.103 0.110 0.101 0.002 0.002MTBi,t − 1 3.255 2.353 3.317 2.421 −0.062 −0.068Leveragei,t − 1 0.317 0.147 0.424 0.157 −0.107∗ −0.010Rate Diffi,t − 1 0.049 0.089 0.057 0.092 −0.008 −0.003WW ROSi,t − 1 0.139 0.116 0.115 0.100 0.024∗ 0.016∗

RDi,t − 1 0.034 0.019 0.035 0.022 −0.001 −0.003Cashi,t − 1 0.148 0.093 0.170 0.127 −0.022 −0.034∗∗

%DomSalesi,t − 1 0.615 0.614 0.585 0.541 0.030 0.073∗∗∗

Cash Flowi,t − 1 0.136 0.116 0.141 0.125 −0.005 −0.009Inventoryi,t − 1 0.109 0.090 0.108 0.076 0.001 0.014Tangibilityi,t − 1 0.338 0.260 0.316 0.220 0.022 0.040∗

Sizei,t − 2 6.878 6.637 6.874 6.642 0.004 −0.005ROAi,t − 2 0.100 0.092 0.107 0.100 −0.007 −0.008∗

MTBi,t − 2 4.118 2.344 3.358 2.451 0.760 −0.107Leveragei,t − 2 0.330 0.149 0.590 0.157 −0.260 −0.008Rate Diffi,t − 2 0.062 0.084 0.067 0.106 −0.005 −0.022∗

WW ROSi,t − 2 0.138 0.113 0.132 0.105 0.006 0.008RDi,t − 2 0.037 0.021 0.039 0.030 −0.002 −0.009Cashi,t − 2 0.159 0.109 0.173 0.120 −0.014 −0.011%DomSalesi,t − 2 0.627 0.634 0.578 0.544 0.049∗∗ 0.009∗∗∗

Cash Flowi,t − 2 0.131 0.117 0.134 0.120 −0.003 −0.003Inventoryi,t − 2 0.109 0.092 0.107 0.083 0.002 0.009Tangibilityi,t − 2 0.328 0.261 0.308 0.208 0.020 0.053∗

Panel B: Comparison of tax loss carryforwards for DPAD firms and control firms inpre-DPAD periods

DPAD Firms Control Firms Difference

Variables Mean Median Mean Median Mean Median

NOL Indicatori,t 0.413 0.000 0.491 0.000 −0.078∗∗∗ 0.000∗∗∗

NOL Indicatori,t − 1 0.401 0.000 0.473 0.000 −0.072∗∗∗ 0.000∗∗∗

NOL Indicatori,t − 2 0.383 0.000 0.437 0.000 −0.054∗∗ 0.000∗∗

Dom NOL Indicatori,t 0.208 0.000 0.310 0.000 −0.102∗∗∗ 0.000∗∗∗

Dom NOL Indicatori,t − 1 0.211 0.000 0.277 0.000 −0.066∗∗∗ 0.000∗∗∗

Dom NOL Amounti,t 0.019 0.000 0.048 0.000 −0.029∗∗∗ 0.000∗∗∗

Dom NOL Amounti,t − 1 0.023 0.000 0.057 0.000 −0.033∗∗∗ 0.000∗∗∗

(Continued)

FIRM RESPONSES TO DOMESTIC PRODUCTION INCENTIVES 1079

T A B L E 3—Continued

Panel C: Pre-DPAD trends in intertemporal and cross-border shifting, domestic investment,and domestic employment

DPAD Firms Control Firms Difference

Variables Mean Median Mean Median Mean Median

�GMShift%i,(t − 3,t) 0.004 0.028 −0.026 0.029 0.030 −0.001�OpIncShift%i,(t − 3,t) 0.013 0.024 −0.036 −0.008 0.049 0.032�DomROSi,(t − 3,t) 0.039 0.168 0.000 −0.008 0.038 0.176�DomInvi,(t − 3,t) 0.007 −0.002 0.019 0.018 −0.012 −0.020�DomEmpi,(t − 3,t) −0.063 0.008 −0.053 −0.072 −0.010 0.080

This table presents and compares descriptive statistics for DPAD and matched control firms. Panel Apresents mean and median statistics for the matching variables of Size and ROA, as well as for other vari-ables used in the empirical tests, in the two years prior to the DPAD firm first disclosing information relatedto the tax incentive. Panel B compares descriptive statistics for NOL Indicatori,t, which is an indicator equalto 1 for firms that report a tax loss carryforward in Compustat (tlcf) or 0 otherwise. This panel also com-pares descriptive statistics for Dom NOL Indicatori,t, which is an indicator equal to 1 for firms that reporta domestic tax loss carryforward in their financial statements, for treatment and control firms, as well asDom NOL Amounti,t, which is the amount of a firm’s domestic NOL carryforward, scaled by total assets. PanelC presents the average change in the measures of intertemporal shifting (GMShift%i,t and OpIncShift%i,t),cross-border shifting (DomROSi,t), domestic investment (DomInvi,t), and domestic employment (DomEmpi,t)from year t – 3 to year t for purposes of assessing the parallel trends in the DPAD and control firm samplesin the pre-DPAD period. The median amount reflects the median value of the average change; tests of thedifferences in medians reflect whether the samples were drawn from populations with the same medianvalue. Appendix B provides the definitions of all variables. ∗, ∗∗, and ∗∗∗ denote significance at the 10%, 5%,and 1% levels, respectively.

used in the empirical tests, except for Leverage and WW ROS in year t – 1and %DomSales in year t – 2.

Panel B presents descriptive statistics that compare the incidence andlevel of tax loss carryforwards across the two samples. I expect these val-ues to be statistically different if indeed control firms did not claim thededuction due to NOL use. The indicator NOL Indicatori,t equals 1 if a firmreports any NOL carryforward in its financial statements (tlcf) and 0 other-wise. Approximately 41.3% of DPAD firms report some amount of tax losscarryforward (including federal, state, and foreign tax losses) as comparedto 49.1% of control firms, and the difference is statistically significant in thefirst year the firm claims the DPAD, as well as in the two preceding years.Furthermore, while the median value of the indicator is 0.000 for both sam-ples, equality-of-median tests confirm that the samples exhibit a statisticallydifferent distribution.

I construct two additional measures using hand-collected data on firms’federal tax loss carryforwards: Dom NOL Indicatori,t, which is an indicatorequal to 1 for firms that report a domestic NOL carryforward and 0 oth-erwise, and Dom NOL Amounti,t, the dollar amount of the federal NOLcarryforward, scaled by a firm’s total assets. DPAD firms also report signifi-cantly lower values for both of these measures.14 The descriptive statistics in

14 Untabulated statistics show that control firms report over twice the amount of NOLs—$80.3 million as compared to $37.7 million for DPAD firms in year t. There are at least tworeasons why a DPAD firm would report some nonzero level of domestic tax loss carryforwards.

1080 R. LESTER

panel A confirm that the matching produced control firms of similar size,similar financial performance, and similar types of qualifying income butthat differ predictably based on the incidence and level of domestic NOLcarryforwards, as seen in panel B.

4.4 PARALLEL TRENDS ASSUMPTION

A key identifying assumption in using difference-in-differences estima-tion is that treatment and control firms would have exhibited similar out-comes in the treatment period, absent any treatment effect. Although thisassumption cannot be explicitly tested, similar trends in the outcome vari-ables in the period preceding treatment provide insight into the similarityof the groups. Table 3, panel C, shows that pretreatment trends for the out-come variables of interest are not statistically different across the DPAD andcontrol samples in the three years prior to treatment.15

Figures 1–4 present graphical evidence of these trends and the treatmenteffect. Figure 1, panel A plots the amount of gross margin that is shiftedfrom the fourth quarter of a year into the first quarter of the following yearusing annual data, and panel B incorporates quarterly data.16 The plots inthe firm-year graph in panel A at t – 2 and t – 1 correspond to the plots inquarters t – 8 (i.e., two years prior to claiming DPAD) and t – 4 (one yearprior to claiming DPAD) in panel B, respectively. Although treatment andcontrol firms exhibited similar trends in shifting in the pre-DPAD periods,panel A (B) shows that DPAD firms report a higher amount of shifted gross

First, the amount of tax loss carryforward that the DPAD firm can claim in any particular yearcould be limited under IRC Section 382; in this case, the DPAD firm may only be able to claimsome portion of its domestic tax loss carryforward in any year. To the extent that the allowableamount of carryforward does not completely absorb a DPAD firm’s taxable income, then thefirm could also claim the Section 199 deduction in that tax year. Furthermore, some firmsblend state and federal tax loss carryforwards as “domestic” in their financial statements, suchthat even the hand-collection of these data results in some measurement error.

15 Because cross-border income shifting is tested by examining the relation between re-ported income and a firm’s domestic–foreign tax rate differential, an assessment of paralleltrends in the pretreatment period is included in the discussion of the H2 results (see subsec-tion 5.2). For completeness, I also include �DomROSi,(t − 3,t) in table 3, panel C, to mitigateconcerns that the observed results are attributable to differences in the pre-period trend ofDomROSi,(t,t + n).

16 To demonstrate the parallel trends in the pre-period, I present the graphs of intertempo-ral shifting (H1) around the first DPAD year but do not include subsequent periods in whichshifting incentives exist. Although an event period graph that aligns all shifting periods at t =0 would provide graphical evidence of the shifting effect consistent with the empirical tests, itwould not provide a clean visual of the parallel trends because the pre-period graphs wouldbe confounded by earlier intertemporal shifting effects. Similarly, the graphs cannot includepost-period years for H1 beyond t = 0 because, while I expect the intertemporal shifting dif-ferences to revert in years without shifting incentives, the graphs will be confounded by anysubsequent shifting. I mitigate these concerns by also graphing the pre-period intertemporalshifting effects at the quarterly level. In contrast, because I expect that the cross-border shift-ing and real activity responses persist throughout the post-DPAD period, figures 2–4 includeall three years following first discussion of the DPAD benefit.

FIRM RESPONSES TO DOMESTIC PRODUCTION INCENTIVES 1081

Intertemporal Shifting

Panel A: Firm-year measures of GMShift%i,t Panel B: Firm-quarter measures of GMShift%i,t

-0.1

00.

000.

100.

20R

esid

ual G

ross

Mar

gin

Shi

fting

-3 -2 -1 0Year Relative to First DPAD Year

Control Firm DPAD Firm

-0.4

0-0

.20

0.00

0.20

0.40

Res

idua

l Gro

ss M

argi

n S

hifti

ng

-8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3Quarter Relative to First DPAD Year

Control Firm DPAD Firm

Panel C: Firm-year measures of OpIncShift%i,t Panel D: Firm-year measures of OpIncShift%i,t

-0.1

00.

000.

10R

esid

ual O

pera

ting

Inco

me

Shi

fting

-3 -2 -1 0Year Relative to First DPAD Year

Control Firm DPAD Firm

-0.4

0-0

.20

0.00

0.20

0.40

Res

idua

l Ope

ratin

g In

com

e S

hifti

ng

-8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3Quarter Relative to First DPAD Year

Control Firm DPAD Firm

FIG. 1.—Intertemporal Shifting. The figures in panels A and B (C and D) plot the treatmenttrends in GMShift%i,t (OpIncShift%i,t). Panel A and C show the pretreatment trends using afirm-year sample and captures the residual intertemporal shifting over the three years preced-ing the first year the DPAD is claimed. Panels B and D incorporate quarterly data; in thesegraphs, t - 8 (t - 4) captures the residual shifting eight (four) quarters prior to the first year theDPAD is claimed and corresponds to years t-2 (t-1) in the firm-year graphs. Statistical tests ofthe pre-treatment trends are presented in table 3, panel C.

margin, relative to the matched sample of control firms in the first year(quarter) in which DPAD is claimed, consistent with the H1 prediction.Figure 1 also presents graphs that correspond to the statistics presented intable 3, panel C for OpIncShift%i,t. Sections 5.2 and 5.3 include the figuresthat correspond with the variables used in testing cross-border shifting, in-vestment, and employment.

4.5 DESCRIPTIVE STATISTICS

Table 4, panel A (B, C), presents descriptive statistics for the variablesused to test H1 (H2, H3). All continuous variables are winsorized at 1%and 99% of the distribution. The average amounts of shifted gross mar-gin (GMShift%) and shifted operating income (OpIncShift%) as a percent-age of the firm’s assets are –0.053% and –0.012%, respectively, evidencethat on average firms shift income from the first quarter into the fourthquarter of the preceding year, perhaps for earnings management purposes

1082 R. LESTER

TA

BL

E4

Des

crip

tive

Stat

istic

s

Pan

elA

:Int

erte

mpo

rals

hift

ing

vari

able

s(H

1)

Vari

able

sO

bser

vati

ons

Mea

nM

edia

nSD

P25

P75

GM

Shift

%i,t

21,4

62−0

.053

−0.0

021.

236

−0.4

590.

396

OpI

ncSh

ift%

i,t21

,462

−0.0

12−0

.003

0.99

7−0

.399

0.34

9D

PAD

Firm

i21

,462

0.51

31.

000

0.50

00.

000

1.00

0Sh

iftPe

riod

i,t21

,462

0.17

30.

000

0.37

80.

000

0.00

0Sh

iftPe

riod

0506

i,t21

,462

0.04

90.

000

0.21

70.

000

0.00

0Sh

iftPe

riod

0709

i,t21

,462

0.05

30.

000

0.22

50.

000

0.00

0Sh

iftPe

riod

1013

i,t21

,462

0.07

00.

000

0.25

50.

000

0.00

0D

PAD

Firm

i∗ Shi

ftPe

riod

i,t21

,462

0.08

80.

000

0.28

40.

000

0.00

0D

PAD

Firm

i∗ Shi

ftPer

iod

0506

i,t21

,462

0.02

50.

000

0.15

50.

000

0.00

0D

PAD

Firm

i∗ Shi

ftPer

iod

0709

i,t21

,462

0.02

70.

000

0.16

30.

000

0.00

0D

PAD

Firm

i∗ Shi

ftPer

iod

1013

i,t21

,462

0.03

60.

000

0.18

70.

000

0.00

0Si

zei,t

−1

21,4

626.

934

6.75

21.

729

5.64

98.

049

RO

Ai,t

−1

21,4

620.

111

0.10

10.

077

0.06

40.

149

MT

Bi,t

−1

21,4

623.

134

2.27

72.

901

1.52

43.

600

Lev

erag

e i,t

−1

21,4

620.

350

0.16

70.

577

0.02

30.

432

RD

Cre

dit i,

t−

121

,462

0.04

20.

000

0.20

10.

000

0.00

0

Pan

elB

:Cro

ss-b

orde

rsh

ifti

ngva

riab

les

(H2)

Vari

able

sO

bser

vati

ons

Mea

nM

edia

nSD

P25

P75

Dom

RO

S i,t

4,60

50.

149

0.09

80.

189

0.05

30.

172

DPA

DFi

rmi

4,60

50.

504

1.00

00.

500

0.00

01.

000

Post

DPA

Di,t

4,60

50.

475

0.00

00.

500

0.00

01.

000

DPA

DFi

rmi∗ P

ostD

PAD

i,t4,

605

0.24

30.

000

0.42

90.

000

0.00

0

(Con

tinue

d)

FIRM RESPONSES TO DOMESTIC PRODUCTION INCENTIVES 1083

TA

BL

E4—

Con

tinue

d

Pan

elB

:Cro

ss-b

orde

rsh

ifti

ngva

riab

les

(H2)

Vari

able

sO

bser

vati

ons

Mea

nM

edia

nSD

P25

P75

Rat

eD

iffi,(

t,t+

n)4,

605

0.05

30.

082

0.19

6−0

.010

0.16

6D

PAD

Firm

i∗ Rat

eD

iffi,(

t,t+

n)4,

605

0.02

20.

000

0.13

50.

000

0.07

3R

ate

Diff

i,(t,t

+n)

∗ Pos

tDPA

Di,t

4,60

50.

033

0.00

00.

141

0.00

00.

095

DPA

DFi

rmi∗ P

ostD

PAD

i,t∗ R

ate

Diff

i,(t,t

+n)

4,60

50.

014

0.00

00.

095

0.00

00.

000

Post

DPA

D05

06i,t

4,60

50.

082

0.00

00.

275

0.00

00.

000

Post

DPA

D07

09i,t

4,60

50.

139

0.00

00.

346

0.00

01.

000

Post

DPA

D10

13i,t

4,60

50.

254

0.00

00.

435

0.00

00.

000

DPA

DFi

rmi∗ P

ostD

PAD

0506

i,t4,

605

0.04

20.

000

0.20

00.

000

0.00

0D

PAD

Firm

i∗ Pos

tDPA

D07

09i,t

4,60

50.

072

0.00

00.

259

0.00

00.

000

DPA

DFi

rmi∗ P

ostD

PAD

1013

i,t4,

605

0.12

90.

000

0.33

50.

000

0.00

0R

ate

Diff

i,(t,t

+n)

∗ Pos

tDPA

D05

06i,t

4,60

50.

005

0.00

00.

056

0.00

00.

000

Rat

eD

iffi,(

t,t+

n)∗ P

ostD

PAD

0709

i,t4,

605

0.01

20.

000

0.07

00.

000

0.00

0R

ate

Diff

i,(t,t

+n)

∗ Pos

tDPA

D10

13i,t

4,60

50.

016

0.00

00.

114

0.00

00.

000

DPA

DFi

rmi∗ P

ostD

PAD

0506

i,t∗ R

ateD

iffi,(

t,t+

n)4,

605

0.00

30.

000

0.04

00.

000

0.00

0D

PAD

Firm

i∗ Pos

tDPA

D07

09i,t

∗ Rat

eDiff

i,(t,t

+n)

4,60

50.

004

0.00

00.

051

0.00

00.

000

DPA

DFi

rmi∗ P

ostD

PAD

1013

i,t∗ R

ateD

iffi,(

t,t+

n)4,

605

0.00

80.

000

0.07

80.

000

0.00

0W

WR

OS i

,t4,

605

0.12

70.

106

0.08

50.

069

0.16

1Si

zei,

(t,t

+n)

4,60

57.

972

7.82

01.

614

6.77

09.

083

RD

i,(t

,t+

n)4,

605

0.03

80.

022

0.04

30.

003

0.05

5C

ash i

,(t,t

+n)

4,60

50.

147

0.09

70.

145

0.04

00.

211

Lev

erag

e i,(t

,t+

n)4,

605

0.22

20.

148

0.26

30.

050

0.29

0%

Dom

Sale

s i,(t

,t+

n)4,

605

0.60

20.

591

0.22

10.

451

0.76

5

(Con

tinue

d)

1084 R. LESTER

TA

BL

E4—

Con

tinue

d

Pan

elC

:Inv

estm

enta

ndem

ploy

men

tvar

iabl

es(H

3)

All

Firm

sD

omes

tic-

On

lyFi

rms

Mul

tin

atio

nal

Firm

s

Vari

able

sO

bser

vati

ons

Mea

nSD

Obs

erva

tion

sM

ean

SDO

bser

vati

ons

Mea

nSD

Dom

Inv i

,t8,

274

0.20

60.

335

2,68

90.

326

0.49

75,

585

0.14

70.

192

Dom

Cap

exi,t

($M

)8,

274

289.

8866

9.99

2,68

924

0.86

570.

005,

585

328.

3773

3.04

PPE i

,t($

M)

8,27

42,

513.

445,

273.

752,

689

1,62

2.75

4,42

6.61

5,58

52,

975.

595,

735.

64#D

omes

ticEm

ploy

ees

8,29

611

,357

.820

,174

.42,

731

3,20

0.77

7,08

9.67

5,56

515

,387

.50

23,5

90.6

4D

omEm

p i,t

8,29

68.

224

1.63

2,73

16.

700

1.54

65,

565

8.82

61.

297

DPA

DFi

rmi

8,27

40.

510

0.50

02,

689

0.50

20.

500

5,58

50.

513

0.50

0Po

stD

PAD

i,t8,

274

0.39

90.

490

2,68

90.

364

0.48

15,

585

0.41

50.

493

Post

DPA

D05

06i,t

8,27

40.

090

0.28

62,

689

0.07

70.

267

5,58

50.

096

0.29

5Po

stD

PAD

0709

i,t8,

274

0.13

40.

341

2,68

90.

102

0.30

25,

585

0.15

00.

357

Post

DPA

D10

13i,t

8,27

40.

174

0.37

92,

689

0.18

60.

389

5,58

50.

168

0.37

4D

PAD

Firm

i∗ Pos

tDPA

Di,t

8,27

40.

205

0.40

42,

689

0.18

50.

388

5,58

50.

214

0.41

0D

PAD

Firm

i∗ Pos

tDPA

D05

06i,t

8,27

40.

046

0.20

92,

689

0.03

50.

184

5,58

50.

050

0.21

9D

PAD

Firm

i∗ Pos

tDPA

D07

09i,t

8,27

40.

068

0.25

22,

689

0.04

90.

216

5,58

50.

078

0.26

7D

PAD

Firm

i∗ Pos

tDPA

D10

13i,t

8,27

40.

091

0.28

82,

689

0.10

10.

301

5,58

50.

086

0.28

1Si

zei,t

−1

8,27

47.

607

1.79

32,

689

6.21

31.

704

5,58

58.

278

1.42

6R

OA

i,t−

18,

274

0.11

10.

071

2,68

90.

095

0.09

25,

585

0.11

80.

065

Cas

hFl

owi,t

−1

8,27

40.

132

0.08

42,

689

0.13

40.

109

5,58

50.

131

0.07

4M

TB

i,t−

18,

274

3.25

42.

934

2,68

92.

396

2.27

35,

585

3.72

53.

594

Lev

erag

e i,t

−1

8,27

40.

369

0.55

62,

689

0.49

40.

703

5,58

50.

313

0.53

7In

vent

ory i,

t−

18,

296

0.11

70.

124

2,73

10.

118

0.18

65,

565

0.11

70.

083

Tang

ibili

tyi,t

−1

8,29

60.

328

0.24

52,

731

0.44

50.

300

5,56

50.

271

0.18

8a T

his

tabl

epr

esen

tsde

scri

ptiv

est

atis

tics

for

the

depe

nde

nt

and

con

trol

vari

able

sus

edto

test

inte

rtem

pora

lsh

ifti

ng

(pan

elA

),cr

oss-

bord

ersh

ifti

ng

(pan

elB

),an

din

vest

men

tan

dem

ploy

men

tac

tivi

ty(p

anel

C).

Inpa

nel

C,d

escr

ipti

vest

atis

tics

are

prov

ided

for

the

full

sam

ple,

asw

ell

asfo

rth

esu

bsam

ples

ofdo

mes

tic-

only

and

mul

tin

atio

nal

firm

s.D

ueto

rest

rict

ion

son

data

disc

losu

refo

rth

eva

riab

les

con

stru

cted

from

BE

Ada

ta,o

nly

mea

nva

lues

and

stan

dard

devi

atio

ns

are

pres

ente

din

this

pan

el.A

llco

nti

nuo

usva

riab

les

are

win

sori

zed

at1%

and

99%

and

are

defi

ned

inap

pen

dix

B.

FIRM RESPONSES TO DOMESTIC PRODUCTION INCENTIVES 1085

(Dechow, Ge, and Schrand [2010], subsection 3.1.5). Panel B shows that av-erage DomROS (WW ROS) is 14.9% (12.7%) for the sample used in testingH2. Panel C presents descriptives for the full H3 sample as well as the sub-samples of domestic-only and multinational firms; I present only the meanand standard deviation in panel C due to BEA restrictions on disclosure.The average firm spends 20.6% of their fixed asset book value on domesticcapital expenditures per year, or approximately $289.9 million. Domestic-only firms spend $240.9 million annually on capital expenditures, whereasmultinational firms spend $328.4 million. The average domestic-only firmemploys 3,200 workers, whereas the average MNC has 15,387 domestic em-ployees.

5. Empirical Results

5.1 TESTS OF INTERTEMPORAL SHIFTING (H1)

Table 5 presents the results obtained from estimating equation (1).Columns 1–4 present results using GMShift%i,t as the measure of in-come shifting, and columns 5–8 present results using OpIncShift%i,t.Column 1 shows a positive and significant coefficient of 0.098 onDPAD Firmi

∗Shift Periodi,t, which means that the DPAD benefit is associatedwith gross margin shifting equal to 0.098% of firms’ total assets. I confirmthis result in column 3, where I re-estimate the results at the firm-quarterlevel; that is, I compare shifting in the first quarter of years with shiftingincentives to all other quarters in the pre- and post-DPAD period for treat-ment and control firms. Estimation at the quarterly level increases the sam-ple size to 85,521 observations. When using this larger sample, I find a sta-tistically significant coefficient of 0.209, which means that the DPAD benefitis associated with gross margin shifting equal to 0.209% of firms’ total as-sets. Columns 5 and 7 repeat the analyses for OpIncShift%; the coefficientsof 0.077 and 0.111 mean that the DPAD benefit is associated with operatingincome shifting ranging from 0.077% to 0.111% of firms’ total assets. Forthe average DPAD firm that reports total assets of $6.5 billion in the yearprior to claiming the incentive, the coefficients across these four columnsare equivalent to shifting of $5.0–$13.7 million per firm.

Columns 2, 4, 6, and 8 present results from tests that examine howfirm responses varied across time. I replace the indicator Shift Periodi,t inequation (1) with three separate indicator variables. For example, Shift-Period 0506i,t is an indicator equal to 1 in the 2005 (2006) firm-year foreach matched pair if the DPAD firm first claimed the incentive in 2005(2006) when the statutory benefit was 3%. The indicator ShiftPeriod 0709i,t

is equal to 1 in 2007 for firms that first disclose the DPAD benefit in 2007, aswell as for firms that previously disclosed the benefit in 2005 or 2006. Theinteraction term captures the additional shifting for DPAD firms in theyear that they first discuss the tax incentive in their financial statementsor in any subsequent year with a shifting incentive. The results show that

1086 R. LESTER

TA

BL

E5

Inte

rtem

pora

lShi

fting

Pan

elA

:Reg

ress

ion

anal

ysis

ofin

tert

empo

rals

hift

ing

Dep

ende

ntV

aria

ble:

GM

Shift

%i,t

Dep

ende

ntV

aria

ble:

OpI

ncSh

ift%

i,t

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

DPA

DFi

rmi

0.08

00.

080

−0.0

34∗∗

−0.0

34∗∗

0.02

30.

023

−0.0

13∗∗

−0.0

12∗∗

(1.0

55)

(1.0

56)

(−2.

500)

(−2.

492)

(0.3

79)

(0.3

81)

(−2.

018)

(−2.

006)

Shift

Peri

odi,t

−0.0

82∗∗

−0.4

01∗∗

∗−0

.042

−0.1

92∗∗

∗

(−2.

303)

(−3.

990)

(−1.

651)

(−2.

679)

DPA

DFi

rmi∗ S

hift

Peri

odi,t

0.09

8∗∗0.

209∗∗

0.07

7∗∗0.

111∗

(2.0

98)

(2.3

60)

(2.2

61)

(1.9

45)

Shift

Peri

od05

06i,t

−0.1

25∗

−0.4

78∗∗

∗−0

.088

−0.2

74∗∗

∗

(−1.

863)

(−5.

018)

(−1.

565)

(−3.

103)

Shift

Peri

od07

09i,t

−0.0

75−0

.387

∗∗∗

−0.0

41−0

.196

∗∗

(−1.

129)

(−3.

225)

(−0.

953)

(−2.

419)

Shift

Peri

od10

13i,t

−0.0

52−0

.352

∗∗∗

−0.0

05−0

.126

(−1.

194)

(−2.

890)

(−0.

157)

(−1.

529)

DPA

DFi

rmi∗ S

hiftP

erio

d05

06i,t

0.16

5∗∗0.

261∗∗

∗0.

143∗∗

0.16

5∗∗

(2.2

02)

(3.1

26)

(2.4

76)

(2.5

66)

DPA

DFi

rmi∗ S

hiftP

erio

d07

09i,t

0.03

40.

127

0.03

10.

053

(0.4

08)

(1.1

00)

(0.5

21)

(0.6

86)

DPA

DFi

rmi∗ S

hiftP

erio

d10

13i,t

0.09

9∗0.

230∗∗

0.06

50.

114∗

(1.7

97)

(2.2

05)

(1.4

85)

(1.7

19)

Size

i,t−

1−0

.043

∗∗−0

.043

∗∗−0

.055

∗∗∗

−0.0

55∗∗

∗−0

.017

−0.0

17−0

.029

∗∗∗

−0.0

29∗∗

∗

(−2.

408)

(−2.

408)

(−10

.266

)(−

10.2

59)

(−1.

231)

(−1.

229)

(−8.

392)

(−8.

413)

RO

Ai,t

−1

−2.8

33∗∗

∗−2

.830

∗∗∗

−0.7

79∗∗

∗−0

.778

∗∗∗

−2.8

67∗∗

∗−2

.865

∗∗∗

−0.8

73∗∗

∗−0

.871

∗∗∗

(−6.

104)

(−6.

123)

(−4.

411)

(−4.

420)

(−7.

627)

(−7.

649)

(−6.

940)

(−6.

967)

MT

Bi,t

−1

0.05

5∗∗∗

0.05

5∗∗∗

0.03

2∗∗∗

0.03

2∗∗∗

0.04

2∗∗∗

0.04

2∗∗∗

0.02

0∗∗∗

0.02

0∗∗∗

(6.4

07)

(6.4

11)

(7.5

93)

(7.5

95)

(5.5

73)

(5.5

76)

(8.4

08)

(8.4

18)

(Con

tinue

d)

FIRM RESPONSES TO DOMESTIC PRODUCTION INCENTIVES 1087

TA

BL

E5—

Con

tinue

d

Pan

elA

:Reg

ress

ion

anal

ysis

ofin

tert

empo

rals

hift

ing

Dep

ende

ntV

aria

ble:

GM

Shift

%i,t

Dep

ende

ntV

aria

ble:

OpI

ncSh

ift%

i,t

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Lev

erag

e i,t

−1

0.06

30.

063

−0.0

37∗∗

∗−0

.037

∗∗∗

0.03

40.

034

−0.0

20∗∗

∗−0

.020

∗∗∗

(1.2

63)

(1.2

63)

(−4.

452)

(−4.

450)

(0.9

32)

(0.9

33)

(−4.

006)

(−3.

999)

RD

Cre

dit i,

t−

1−0

.023

−0.0

230.

018

0.01

8−0

.002

−0.0

020.

014

0.01

4(−

0.29

4)(−

0.29

9)(1

.290

)(1

.296

)(−

0.02

2)(−

0.02

8)(1

.053

)(1

.053

)Ye

ar-b

y-in

dust

ryFE

?Y

YY

YY

YY

YO

bser

vati

ons

21,4

6221

,462

85,5

2185

,521

21,4

6221

,462

85,5

2185

,521

R2

0.05

90.

059

0.02

30.

023

0.07

10.

072

0.01

60.

016

Pan

elB

:Ana

lysi

sof

inte

rtem

pora

lshi

ftin

gby

firm

s’ge

ogra

phic

pres

ence

and

othe

rsh

ifti

ngin

cent

ives

Dep

ende

ntV

aria

ble:

Inte

rtem

pora

lIn

com

eSh

ifti

ng

Mea

sure

s

GM

Shift

%i,t

GM

Shift

%i,t

GM

Shift

%i,t

OpI

ncSh

ift%

i,t

Dom

esti

cFi

rms

MN

Cs

Dom

esti

cFi

rms

MN

Cs

ET

IFi

rms

Oth

erFi

rms

ET

IFi

rms

Oth

erFi

rms

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

DPA

DFi

rmi

0.12

60.

072

0.09

50.

072

0.21

2∗0.

052

0.10

2−0

.007

(0.8

83)

(0.9

32)

(1.0

67)

(0.8

43)

(1.7

46)

(0.6

13)

(1.1

78)

(−0.

096)

Shift

Peri

odi,t

−0.1

31−0

.067

∗−0

.159

∗∗−0

.046

−0.1

36−0

.061

−0.1

22∗∗

−0.0

23(−

1.58

1)(−

1.70

6)(−

2.42

5)(−

1.32

7)(−

1.47

8)(−

0.81

5)(−

2.12

1)(−

0.41

9)D

PAD

Firm

i∗ Shi

ftPe

riod

i,t0.

127

0.08

7∗0.

170∗∗

0.06

00.

136

0.14

0∗∗0.

111

0.10

3∗∗

(1.2

64)

(1.7

13)

(2.5

92)

(1.1

63)

(1.3

06)

(2.2

32)

(1.4

52)

(2.1

22)

(Con

tinue

d)

1088 R. LESTER

TA

BL

E5—

Con

tinue

d

Pan

elB

:Ana

lysi

sof

inte

rtem

pora

lshi

ftin

gby

firm

s’ge

ogra

phic

pres

ence

and

othe

rsh

ifti

ngin

cent

ives

Dep

ende

ntV

aria

ble:

Inte

rtem

pora

lIn

com

eSh

ifti

ng

Mea

sure

s

GM

Shift

%i,t

GM

Shift

%i,t

GM

Shift

%i,t

OpI

ncSh

ift%

i,t

Dom

esti

cFi

rms

MN

Cs

Dom

esti

cFi

rms

MN

Cs

ET

IFi

rms

Oth

erFi

rms

ET

IFi

rms

Oth

erFi

rms

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Size

i,t−

1−0

.045

−0.0

41∗∗

−0.0

59∗∗

∗−0

.033

−0.0

71∗∗

−0.0

10−0

.041

∗∗0.

007

(−1.

668)

(−2.

315)

(−2.

691)

(−1.

565)

(−2.

335)

(−0.

608)

(−2.

055)

(0.4

65)

RO

Ai,t

−1

−2.7

12∗∗

∗−2

.891

∗∗∗

−3.1

70∗∗

∗−2

.668

∗∗∗

−1.2

94−3

.951

∗∗∗

−1.8

98∗∗

−3.6

85∗∗

∗

(−3.

244)

(−6.

404)

(–4.

389)

(−5.

979)

(−1.

520)

(−4.

994)

(–2.

601)

(−5.

928)

MT

Bi,t

−1

0.06

3∗∗∗

0.05

4∗∗∗

0.05

5∗∗∗

0.05

4∗∗∗

0.04

8∗∗0.

061∗∗

∗0.

036∗∗

0.04

8∗∗∗

(2.9

09)

(6.9

44)

(2.6

70)

(7.0

65)

(2.6

21)

(6.9

82)

(2.3

00)

(5.9

65)

Lev

erag

e i,t

−1

0.02

90.

076

0.04

20.

078

0.11

40.

033

0.07

90.

016

(0.5

36)

(1.4

17)

(0.6

04)

(1.5

62)

(1.0

00)

(0.5

33)

(1.1

23)

(0.3

29)

RD

Cre

dit i,

t−

10.

124

−0.0

420.

052

−0.0

40−0

.111

0.03

8−0

.100

0.06

8(0

.614

)(−

0.58

4)(0

.361

)(−

0.50

8)(−

0.93

4)(0

.399

)(−

1.24

2)(0

.670

)|D

iffe

ren

ce|

0.04

00.

110

0.00

40.

008

p-Va

lue

0.54

60.

162

0.98

00.

923

Year

-by-

indu

stry

FE?

YY

YY

YY

YY

Obs

erva

tion

s4,

456

17,0

067,

025

14,4

373,

788

8,29

33,

788

8,29

3R

20.

088

0.05

70.

090

0.05

00.

072

0.08

90.

078

0.10

5

Th

ista

ble

pres

ents

resu

lts

ofO

LS

regr

essi

ons

that

test

inte

rtem

pora

lin

com

esh

ifti

ng

byD

PAD

firm

s,re

lati

veto

am

atch

edsa

mpl

eof

con

trol

firm

s,in

the

firs

tye

arin

wh

ich

the

DPA

Dfi

rmdi

scus

ses

the

tax

ince

nti

ve,a

sw

ella

san

ysu

bseq

uen

tye

arw

ith

anin

crea

sed

stat

utor

yD

PAD

ben

efit.

Pan

elA

pres

ents

resu

lts

usin

ga

firm

-yea

rsa

mpl

e(c

olum

ns

1,2,

5,an

d6)

and

afi

rm-q

uart

ersa

mpl

e(c

olum

ns

3,4,

7,an

d8)

.Pan

elB

test

sin

tert

empo

rals

hif

tin

gus

ing

afi

rm-y

ear

sam

ple

afte

rpa

rtit

ion

ing

firm

sba

sed

onth

eir

geog

raph

icpr

esen

ce(c

olum

ns

1–4)

and

onw

het

her

the

DPA

Dfi

rmcl

aim

edth

eE

TI

(col

umn

s5–

8),a

nex

piri

ng

tax

subs

idy

that

prov

ided

ince

nti

ves

tosh

ifti

nco

me

into

the

pre-

DPA

Dpe

riod

.In

pan

elB

,firm

sar

eid

enti

fied

asdo

mes

tic-

only

inco

lum

ns

1an

d2

base

don

no

repo

rted

fore

ign

subs

idia

ries

per

Exh

ibit

21da

taan

dze

roor

mis

sin

gva

lues

for

both

fore

ign

segm

ent

sale

san

dfo

reig

nse

gmen

tas

sets

.Col

umn

s3

and

4id

enti

fydo

mes

tic-

only

firm

sba

sed

onze

roor

mis

sin

gva

lues

for

pre-

tax

fore

ign

inco

me

(PIF

O)

and

fore

ign

tax

expe

nse

(TX

FO).

All

vari

able

sar

ede

fin

edin

appe

ndi

xB

.t-s

tati

stic

sar

epr

esen

ted

inpa

ren

thes

es.

Eac

hsp

ecifi

cati

onin

clud

esye

ar-b

y-in

dust

ryfi

xed

effe

cts,

wit

hst

anda

rder

rors

clus

tere

dby

indu

stry

.∗ ,

∗∗,a

nd

∗∗∗

indi

cate

stat

isti

cals

ign

ifica

nce

atth

e10

%,5

%,a

nd

1%le

vels

,res

pect

ivel

y.T

he

spec

ifica

tion

isas

follo

ws:

Tim

eShi

fti,

t=

α+

β1D

PAD

Firm

i+

β2Sh

iftPe

riod

i,t+

β3D

PAD

Firm

i∗S

hiftP

erio

d i,t+

Con

trol

s i,t

−1

+ε i

,t.

FIRM RESPONSES TO DOMESTIC PRODUCTION INCENTIVES 1089

there is a statistically significant effect in the first years of the tax benefit,suggesting that intertemporal shifting was a relatively low-cost way to max-imize the amount of the DPAD. I also observe statistically significant shift-ing in the more recent period since 2010 once the greatest DPAD benefit isavailable.

I present results from two additional tests of intertemporal shifting inpanel B. I first partition the sample based on whether the DPAD firm is adomestic-only or multinational company in columns 1–4 to mitigate con-cerns that multinationals (whose intertemporal shifting measures may cap-ture activities other than domestic shifting) are driving the results. As intable 1, panel D, I identify domestic-only firms in column 1 as those thatreport no foreign subsidiary in Exhibit 21 (Dyreng and Lindsey [2009]),no foreign segment sales, nor any foreign segment assets. In column 3,I alternatively identify domestic-only firms as those that report zero ormissing values for pre-tax foreign income and foreign tax expense, whichresults in more firms being classified as domestic-only. I observe a posi-tive coefficient of similar magnitude for the domestic-only firms in bothcolumns 1 and 3, although the effect is only significant in column 3. Notethat the coefficients on the interaction terms are not significantly differ-ent between the domestic-only and multinational subsamples, suggestingthat both groups contribute to the overall significance observed in panelA. The results provide some evidence that the intertemporal shifting effectis present for firms in which the intertemporal shifting measures reflectpurely domestic activity.17

I also test how shifting differs based on an expiring tax incentive, theETI, which was repealed as part of the American Jobs Creation Act of 2004.Firms claiming this incentive may not have shifted income into the post-DPAD period if they received greater tax savings from the ETI incentive. Ilimit the sample to firms first claiming the DPAD incentive in years 2004–2006, as these were years in which firms could still receive an ETI benefitprior to its phase-out, and I partition the sample based on an indicator ETIFirmi, which equals 1 for firms that discuss this incentive in their financialstatements or 0 otherwise.18 In columns 5 and 7, I find statistically insignif-icant coefficients for intertemporal shifting by ETI-DPAD firms relative tothe sample of matched control firms. In contrast, the coefficients on theinteraction terms in columns 6 and 8 for the ETI firms are statistically sig-nificant, although the coefficients are not statistically different from thosein columns 5 and 7.

17 Multinationals have an additional channel to engage in intertemporal shifting: deferringsales to (or accelerating purchases from) their foreign subsidiaries. However, because neitherCompustat nor the BEA collects quarterly data on domestic segment revenues and expenses,I cannot empirically test this possible channel.

18 I search financial statements for the phrases “extraterritorial,” “ETI,” “foreign sales corpo-ration,” “FSC,” “export tax,” and “export subsidy” and retain those observations that correctlyrelate to the ETI.

1090 R. LESTER

Cross-border Shifting

-0.1

00.

000.

100.

20R

esid

ual D

omes

tic R

etur

n-on

-Sal

es

-3 -2 -1 0 1 2 3 4Year Relative to First Year DPAD Disclosed

Control Firm DPAD Firm

FIG. 2.—Cross-border shifting. This figure plots the pretreatment trends in DomROSi,t, theproxy used in the tests of cross-border shifting (H2). The graph includes firm-year observa-tions with sufficient data to measure DomROSi,t over two subsequent periods for the multiyearspecification following Klassen and Laplante [2012]. DomROSi,t is first regressed on controlvariables, and the residual amount is then plotted for DPAD firms and the matched controlfirms. The graph shows the pretreatment trends in DomROSi,t over the three years precedingthe first year the DPAD is claimed. Statistical tests of the pretreatment trends are presented intable 3, panel C. Further, section 5.2 discusses differences in the pre-treatment sensitivity ofreported income to the tax rate for DPAD and the matched control firms as a direct analysisof trends in income-shifting.

The collective evidence from table 5 shows that DPAD firms shift moreincome into the post-DPAD period relative to the matched sample ofcontrol firms to maximize the tax benefit, and that this activity varies acrossthe sample period and based on DPAD firms’ geographic presence and ex-piring tax incentives.

5.2 TESTS OF CROSS-BORDER SHIFTING (H2)

Figure 2 and table 6, panel A, present results from testing whether DPADfirms changed their reported income post-DPAD. Columns 1–3 report re-sults using variables measured over a two-year period; columns 4–6 reportresults using variables measured over a five-year period. In columns 1 and 4,I first test whether DPAD firms reported a greater difference in the amountof income reported in the United States per dollar of domestic sales post-DPAD, relative to control firms. I find statistically significant coefficients of0.022 and 0.027 on the interaction term DPAD Firmi

∗PostDPADit in columns1 and 4, respectively, which means that claiming the DPAD is associated withan increase in firms’ domestic return-on-sales of 2.2–2.7 percentage pointsrelative to the control sample. Based on the average amount of domesticsales of $4.06 billion, this is equivalent to the average DPAD firm reporting

FIRM RESPONSES TO DOMESTIC PRODUCTION INCENTIVES 1091

TA

BL

E6

Cro

ss-B

orde

rSh

iftin

g

Pan

elA

:Reg

ress

ion

anal

ysis

ofcr

oss-

bord

ersh

ifti

ng

Dep

ende

ntV

aria

ble:

Dom

RO

S i,(

t,t+

1)D

epen

den

tVar

iabl

e:D

omR

OS i

,(t,t

+4)

(1)

(2)

(3)

(4)

(5)

(6)

DPA

DFi

rmi

β1

0.02

0∗∗0.

019∗∗

0.03

2∗∗0.

029∗∗

(2.1

49)

(2.2

70)

(2.1

54)

(2.3

80)

Post

DPA

Di,t

β2

0.00

10.

006

0.00

10.

007

(0.1

24)

(0.6

29)

(0.1

09)

(0.6

78)

DPA

DFi

rmi∗ P

ostD

PAD

i,tβ

30.

022∗∗

∗0.

012∗

0.02

7∗∗0.

014

(3.3

01)

(1.8

46)

(2.6

12)

(1.6

44)

Rat

eD

iffi,(

t,t+

n)β

4−0

.107

∗∗∗

−0.0

85∗∗

∗−0

.130

∗∗∗

−0.1

21∗∗

(−4.

100)

(−3.

547)

(−4.

604)

(−2.

553)

Post

DPA

Di,t∗R

ate

Diff

i,(t,t

+n)

β5

−0.0

79−0

.085

∗∗

(−1.

647)

(−2.

072)

DPA

DFi

rmi∗ R

ate

Diff

i,(t,t

+n)

β6

−0.0

150.

011

(−0.

345)

(0.1

34)

DPA

DFi

rmi∗ P

ostD

PAD

i,t∗ R

ate

Dif

f i,(t

,t+

n)β

70.

121∗∗

0.16

1∗

(2.0

75)

(1.7

90)

WW

RO

S i,(

t,t+

n)β

81.

416∗∗

∗1.

491∗∗

∗1.

468∗∗

∗1.

427∗∗

∗1.

517∗∗

∗1.

488∗∗

∗

(11.

893)

(11.

095)

(11.

576)

(6.1

92)

(6.0

13)

(6.1

88)

Size

i,(t,t

+n)

β9

−0.0

08∗∗

−0.0

07∗

−0.0

07∗

−0.0

09∗

−0.0

07−0

.007

(−2.

164)

(−1.

903)

(−1.

892)

(−1.

813)

(−1.

423)

(−1.

582)

RD

i,(t,t

+n)

β10

0.31

20.

336

0.33

90.

228

0.33

50.

311

(1.3

36)

(1.4

86)

(1.4

92)

(0.7

06)

(0.9

90)

(0.9

34)

Cas

h i,(

t,t+

n)β

110.

033

0.02

90.

041

0.01

50.

002

0.02

5(0

.518

)(0

.488

)(0

.717

)(0

.152

)(0

.017

)(0

.285

)L

ever

age i,

(t,t

+n)

β12

0.04

6∗∗∗

0.04

4∗∗∗

0.04

5∗∗∗

0.04

80.

045

0.04

6

(Con

tinue

d)

1092 R. LESTER

TA

BL

E6—

Con

tinue

d

Pan

elA

:Reg

ress

ion

anal

ysis

ofcr

oss-

bord

ersh

ifti

ng

Dep

ende

ntV

aria

ble:

Dom

RO

S i,(

t,t+

1)D

epen

den

tVar

iabl

e:D

omR

OS i

,(t,t

+4)

(1)

(2)

(3)

(4)

(5)

(6)

(3.1

30)

(3.2

42)

(3.2

65)

(1.4

80)

(1.4

83)

(1.5

04)

%D

omSa

les i,

(t,t

+n)

β13

−0.2

27∗∗

∗−0

.214

∗∗∗

−0.2

21∗∗

∗−0

.274

∗∗∗

−0.2

43∗∗

∗−0

.261

∗∗∗

(−4.

591)

(−4.

310)

(−4.

434)

(−3.

053)

(−3.

069)

(−3.

131)

Sepa

rate

grou

pco

eff.

from

colu

mns

3an

d6

Gro

upco

eff.

Gro

upco

eff.

Con

trol

firm

-yea

rs,p

re-D

PAD

(β4)

−0.0

85−0

.121

DPA

Dfi

rm-y

ears

,pre

-DPA

D(β

4+

β6)

−0.1

00−0

.110

Con

trol

firm

-yea

rs,p

ost-D

PAD

(β4+

β5)

−0.1

64−0

.206

DPA

Dfi

rm-y

ears

,pos

t-DPA

D(β

4+

β5+

β6+

β7)

−0.0

58−0

.034

Year

-by-

indu

stry

FE?

YY

YY

YY

Obs

erva

tion

s4,

605

4,60

54,

605

3,01

83,

018

3,01

8R

20.

515

0.51

80.

526

0.44

00.

436

0.45

1

Pan

elB

:Ana

lysi

sof

cros

s-bo

rder

shif

ting

base

don

high

tax

fore

ign

subs

idia

ries

Dep

ende

ntV

aria

ble:

Dom

RO

S i,(

t,t+

1)

Abo

veM

edia

nFo

reig

nE

TR

Bel

owM

edia

nFo

reig

nE

TR

Abo

veM

edia

nFo

reig

nE

TR

Bel

owM

edia

nFo

reig

nE

TR

Vari

atio

nby

DPA

DR

ate

(1)

(2)

(3)

(4)

(5)

DPA

DFi

rmi

β1

0.02

5∗∗0.

026

0.01

9∗∗0.

035

0.01

9∗∗

(2.5

27)

(1.0

65)

(2.1

74)

(1.5

40)

(2.2

63)

Post

DPA

Di,t

β2

−0.0

180.

024

−0.0

150.

031

(−0.

669)

(1.4

53)

(−0.

608)

(1.6

51)

DPA

DFi

rmi∗ P

ostD

PAD

i,tβ

30.

038∗∗

−0.0

050.

032∗

−0.0

34(2

.074

)(−

0.35

9)(1

.971

)(−

1.15

7)

(Con

tinue

d)

FIRM RESPONSES TO DOMESTIC PRODUCTION INCENTIVES 1093

TA

BL

E6—

Con

tinue

d

Pan

elB

:Ana

lysi

sof

cros

s-bo

rder

shif

ting

base

don

high

tax

fore

ign

subs

idia

ries

and

the

dom

esti

c-fo

reig

nta

xra

tedi

ffer

enti

al

Dep

ende

ntV

aria

ble:

Dom

RO

S i,(

t,t+

1)

Abo

veM

edia

nFo

reig

nE

TR

Bel

owM

edia

nFo

reig

nE

TR

Abo

veM

edia

nFo

reig

nE

TR

Bel

owM

edia

nFo

reig

nE

TR

Vari

atio

nby

DPA

DR

ate

(1)

(2)

(3)

(4)

(5)

Rat

eD

iffi,(

t,t+

1)β

4−0

.085

∗∗∗

−0.1

42∗

(−2.

791)

(−1.

915)

Post

DPA

Di,t

∗ Rat

eD

iffi,(

t,t+

1)β

5−0

.075

−0.0

69(−

1.56

0)(−

0.62

8)D

PAD

Firm

i∗ Rat

eD

iffi,(

t,t+

1)β

6−0

.011

−0.0

25(−

0.32

0)(−

0.27

7)D

PAD

Firm

i∗ Pos

tDPA

Di,t

∗ Rat

eD

iffi,(

t,t+

1)β

70.

153∗∗

∗0.

220

(2.7

48)

(1.2

70)

DPA

DFi

rmi∗ P

ostD

PAD

0506

i,t∗ R

ate

Diff

i,(t,t

+1)

0.04

0(0

.926

)D

PAD

Firm

i∗ Pos

tDPA

D07

09i,t

∗ Rat

eD

iffi,(

t,t+

1)0.

185∗∗

(2.2

08)

DPA

DFi

rmi∗ P

ostD

PAD

1013

i,t∗ R

ate

Diff

i,(t,t

+1)

0.11

8∗

(1.9

41)

|Dif

fere

nce

|0.

043

0.06

7p-

Valu

e0.

009

0.63

7O

ther

inte

ract

ion

term

s?N

NN

NY

Con

trol

s?Y

YY

YY

Ind-

by-y

ear

FE?

YY

YY

YO

bser

vati

ons

2,17

02,

170

2,17

02,

170

4,60

5R

20.

528

0.53

90.

529

0.53

90.

527

Th

ista

ble

pres

ents

resu

lts

ofes

tim

atin

geq

uati

on(2

),w

hic

hte

sts

cros

s-bo

rder

inco

me

shif

tin

gby

DPA

Dfi

rms,

rela

tive

toa

mat

ched

sam

ple

ofco

ntr

olfi

rms,

inth

epo

st-D

PAD

peri

od.P

anel

Apr

esen

tsre

sult

sfr

omes

tim

atin

geq

uati

on(2

)fo

rth

esa

mpl

eof

firm

sw

ith

requ

isit

eda

ta.P

anel

Bpr

esen

tsre

sult

saf

ter

part

itio

nin

gD

PAD

firm

sba

sed

onth

efi

rms’

fore

ign

effe

ctiv

eta

xra

tes

(col

umn

s1–

4).C

olum

n5

pres

ents

resu

lts

from

test

ing

how

the

effe

cts

vary

acro

ssti

me.

All

vari

able

sar

ede

fin

edin

appe

ndi

xB

.t-s

tati

stic

sar

epr

esen

ted

inpa

ren

thes

es.E

ach

spec

ifica

tion

incl

udes

year

-by-

indu

stry

fixe

def

fect

s,w

ith

stan

dard

erro

rscl

uste

red

byin

dust

ry.∗ ,

∗∗,a

nd

∗∗∗

indi

cate

stat

isti

cals

ign

ifica

nce

atth

e10

%,5

%,a

nd

1%le

vels

,res

pect

ivel

y.

1094 R. LESTER

approximately $89.3–$109.7 million more in income in the United States,post DPAD. Figure 2 depicts this effect graphically.

To explicitly test how much of this additional income is attributable tochanges in DPAD firms’ tax-motivated income shifting, I estimate equa-tion (2). For validation, I first include only Rate Diffi,(t,t + n) in columns 2and 5, which captures the difference in firms’ domestic and foreign taxrates. The negative and statistically significant coefficients of –0.107 and –0.130 in columns 2 and 5, respectively, show that reported domestic incomeis negatively associated with the domestic–foreign tax rate differential. Thisassociation is consistent with prior literature and confirms that DPAD andcontrol firms shifted income out of the United States over the sampleperiod.

Columns 3 and 6 present results from separately measuring DPAD firms’income shifting behavior relative to the matched control firms in the pre-and post-DPAD periods; for ease of interpretation, the separate group coef-ficients are presented at the bottom of the table. In column 3, the negativecoefficient of –0.085 on Rate Diffi,(t,t + 1) confirms that, pre-DPAD, controlfirms shifted income out of the United States. DPAD firms also shiftedincome out of the United States in this pre-DPAD period (β4 + β6 =–0.100). Untabulated tests confirm that the difference of 0.015 was not sta-tistically different (p = 0.73), further evidence that the DPAD firms andthe matched control firms exhibited similar pretreatment trends in incomeshifting.

After the DPAD was in place, control firms shift even more income outof the United States; the negative association between reported domesticincome and Rate Diffi,(t,t + 1) changes from –0.085 to –0.164, and this differ-ence of –0.079 is statistically significant (p = 0.10). In contrast, DPAD firmsshift less income out of the United States, as the negative association be-tween reported income and the tax rate changes from –0.100 to –0.058. Al-though this change for DPAD firms (0.042) is not statistically significant, thedifference in the DPAD and control firms in the post-period is (p = 0.03).The positive and statistically significant coefficient on β7 of 0.121 confirmsthat the change in income shifting for DPAD firms is statistically differ-ent than the change in income shifting for control firms. Results frommeasuring income shifting over the five-year period in column 6 yield sim-ilar inferences.

In terms of economic magnitude, a one-percentage-point change in thedomestic-foreign rate differential attributable to the DPAD incentive wouldresult in DPAD firms shifting approximately $4.9–$6.5 million less incomeout of the U.S. post-DPAD, relative to the matched sample of controlfirms.19 A 3.15-percentage-point change in the domestic-foreign rate dif-ferential (equivalent to the full DPAD rate benefit if 100% of DPAD firms’

19 The amounts are calculated as follows: in the pre-DPAD period, DPAD firms reportedDomROSi,t of 19.2%. Assuming a 1.0-percentage-point change in the domestic-foreign tax ratedifferential, the coefficient of –0.100 for DPAD firms implies a 0.1-percentage-point reduction

FIRM RESPONSES TO DOMESTIC PRODUCTION INCENTIVES 1095

income qualified for the incentive) would have resulted in approximately$15.5–$20.6 million less income shifted out of the U.S. post-DPAD, relativeto the matched sample of control firms.

A natural question relates to which types of firms would be willingto forego presumably lower foreign taxes and shift less income out ofthe United States in response to the DPAD. Columns 1–4 in table 6,panel B, present results from cross-sectional tests that study the type offirms that respond to the DPAD by changing their cross-border shift-ing. I partition the sample based on DPAD firms’ average foreign effec-tive tax rates over the sample period and present results in columns 1–4. I find that the increase in the amount of reported income in theUnited States occurs within the sample of firms that have a relativelyhigher foreign effective tax rates (coefficient of 0.038). Further, in col-umn 3, I observe a statistically significant coefficient on the triple inter-action term for this subsample, confirming that firms with relatively highforeign effective tax rates exhibit an attenuated negative association withthe domestic–foreign rate differential post-DPAD. However, the differencein the triple interaction terms across columns 3 and 4 is not statisticallydifferent.20

Finally, column 5 presents results from testing how firm responses var-ied based on the eligible DPAD benefit. I create indicators for each pe-riod with a differing tax rate benefit (PostDPAD 0506i,t, PostDPAD 0709i,t,PostDPAD 1013i,t) and interact these with the other terms in equation (2).I present only the triple interaction terms in these columns for brevity. Theresults show that there is a significant income shifting effect beginning inthe period 2007–2009 when the DPAD percentage increased to 6%, as wellas in the period 2010–2013 when it increased to 9%. These tests show thata tax benefit of at least 2.1% (6% × 35%) is sufficient to induce at leastsome multinationals to report more income in the United States, but thatthis response does not occur immediately following the enactment of theincentive because it is costlier, relative to intertemporal shifting. Results arerobust to estimating over a five-year period as well.

5.3 TESTS OF DOMESTIC INVESTMENT AND EMPLOYMENT (H3)

Figures 3 (4) and Table 7, panel A (B), presents results for studyingthe investment (employment) effects of the Section 199 deduction. The

in reported DomROSi,t pre-DPAD. Post-DPAD, the reported DomROSi,t decreases by only 0.058percentage points. Because the average DPAD firm reported domestic sales of $4.06 billionduring the pre-DPAD period, the results imply an increase in reported income of $1.7 million.By comparison, applying the coefficients for the control firm results in an additional amountof income shifted out of the United States in the post-DPAD period of $3.2 million. Thus, thedifference-in-differences amount is approximately $4.9 million. The coefficients from column6 imply an estimate of $6.5 million for a one-percentage-point change in the tax rate.

20 I also observe similar results (higher reported income; statistically significant coefficienton the triple interaction term) when partitioning based on the proportion of DPAD firms’subsidiaries that are in a high tax jurisdiction.

1096 R. LESTER

Domestic Investment Panel A: Domestic-only Firms

Panel B: Multinational Firms

-0.0

50.

000.

050.

10R

esid

ual D

omes

tic C

apE

x

-3 -2 -1 0 1 2 3 4Year Relative to First DPAD Year Disclosed

Control Firm DPAD Firm

-0.0

4-0

.02

0.00

0.02

0.04

Res

idua

l Dom

estic

Cap

Ex

-3 -2 -1 0 1 2 3 4Year Relative to First DPAD Year

Control Firm DPAD Firm

FIG. 3.—Domestic investment. This figure plots the pretreatment trends in DomInvi,t, the proxyused to test domestic investment (H3). DomInvi,t is first regressed on control variables, and theresidual amount is then plotted separately for DPAD firms and the matched control firms. Thetop graph shows the pretreatment trends for the domestic subsample of firms over the threeyears preceding the first year the DPAD is claimed; the bottom graph presents the pretreat-ment trends for multinational firms. Statistical tests of the pretreatment trends are presentedin table 3, panel C.

results in columns 1 and 2 of panel A indicate that across the full samplethere is no significantly different average investment spending by DPADfirms after the DPAD, relative to the matched sample of controls. How-ever, I observe different effects based on whether the firm is domestic-onlyor multinational. Figure 3 and columns 3–8 present graphs and statistical

FIRM RESPONSES TO DOMESTIC PRODUCTION INCENTIVES 1097

Domestic Employment Panel A: Domestic-Only Firms

Panel B: Multinational Firms

-0.4

0-0

.20

0.00

0.20

0.40

Res

idua

l ln(

Dom

estic

Em

ploy

men

t)

-3 -2 -1 0 1 2 3 4Year Relative to First DPAD Year

Control Firm DPAD Firm

-0.4

0-0

.20

0.00

0.20

0.40

Res

idua

l ln(

Dom

estic

Em

ploy

men

t)

-4 -3 -2 -1 0 1 2 3 4Year Relative to First DPAD Year

Control Firm DPAD Firm

FIG. 4.—Domestic employment. This figure plots the pretreatment trends in DomEmpi,t, theproxy used to test domestic employment (H3). DomEmpi,t is first regressed on control variables,and the residual amount is then plotted separately for DPAD firms and the matched controlfirms. The top graph shows the pretreatment trends for the domestic subsample of firms overthe three years preceding the first year the DPAD is claimed; the bottom graph presents thepretreatment trends for multinational firms. Statistical tests of the pretreatment trends arepresented in table 3, panel C.

results after partitioning the sample based on firms’ geographic presence. Ifirst identify the multinational subsample based on those observations thatcan be matched to the BEA multinational data; then, within the sample ofCompustat-BEA observations, I retain the matched DPAD-control pairs thathave the domestic investment data necessary for these tests (n = 5,585).

1098 R. LESTER

Of the remaining observations not matched to the BEA data, I constructthe domestic-only subsample. As discussed in subsection 4.1 when compar-ing the sample to IRS reported tax data, I drop any observation from thisgroup that reports (1) a foreign subsidiary in Exhibit 21 data (Dyreng andLindsey [2009]); (2) any nonzero, nonmissing foreign segment sales; or (3)any nonzero, nonmissing foreign segment assets.21 The final domestic-onlysample for the investment tests is composed of 227 DPAD-matched controlpairs (n = 2,689).

I plot the average domestic investment spending in panel A of figure 3 forthe domestic-only firms. The graph demonstrates that the DPAD and con-trol firms exhibited similar trends in capital expenditure spending priorto the DPAD. Once the incentive is in place, the amount of spendingby DPAD firms increases relative to control firms. These results are con-firmed with the results presented in table 7, panel A, columns 3 and 4.Although the coefficient for domestic-only firms is statistically insignifi-cant in column 3, I observe a positive and statistically significant effectin column 4 after refining the interaction term based on the differentDPAD rates across time. The investment effect occurs from 2010 through2013, consistent with this activity being the costliest response and thus re-quiring the greatest tax benefit.22 Specifically, the coefficient of 0.084 onDPAD Firmi

∗PostDPAD 1013i,t in column 4 means that claiming the DPADfirm is associated with greater capital expenditures equivalent to 8.4%of total PPE. Given the average DPAD firm in the domestic-only subsam-ple reports fixed assets of $1.71 billion in the year prior to claiming theDPAD (untabulated), this is equivalent to greater spending of $143.6 mil-lion by the average domestic-only DPAD firm over the four-year period2010–2013.23

21 These steps result in the exclusion of some multinational DPAD firms from the analysisin table 7 because there are multinational observations without the requisite BEA data due toBEA reporting thresholds and/or missing BEA data.

22 Results are robust to using a measure of incremental capex, which captures investmentspending on projects beyond that required to maintain the firm’s fixed asset base. I measurethis as total capex, less annual depreciation, following Richardson [2006]. I also find similarresults when testing the effect of domestic R&D: I observe greater R&D spending of 0.9%by domestic-only firms in the post-DPAD period, but no effect for multinationals. Finally, Iexamine the investment behavior of DPAD firms that repatriated under the American JobsCreation Act of 2004; see online appendix B for a discussion of these tests.

23 Ohrn [2018] calculates an increase in investment of 10.06% for a one-percentage-pointincrease in the effective DPAD benefit. He uses industry-level measures to identify the DPADbenefits and measures investment with worldwide (not domestic) capital expenditures. Giventhat he calculates an effective DPAD tax rate benefit (industry-level measurement of QPAI,multiplied by the effective DPAD rate, multiplied by the corporate income tax rate) of1.432 percentage points in the period from 2010, his results imply approximately 14.4%greater investment for firms. Refining the measurement of the effects to the firm-level andusing jurisdiction-specific capital expenditures, I find an effect that is 6.0 percentage pointslower (8.4%).

FIRM RESPONSES TO DOMESTIC PRODUCTION INCENTIVES 1099

TA

BL

E7

Rea

lEffe

cts:

Inve

stm

enta

ndEm

ploy

men

t

Pan

elA

:Reg

ress

ion

anal

ysis

ofdo

mes

tic

inve

stm

ent

Dom

Inv i

,tFo

rInv

i,t

All

Firm

sD

omes

tic-

On

lyFi

rms

Mul

tin

atio

nal

Firm

sM

ulti

nat

ion

alFi

rms

Dep

ende

ntV

aria

ble:

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10)

DPA

DFi

rmi

0.00

10.

001

0.00

10.

001

−0.0

31−0

.031

−0.0

00−0

.000

−0.0

30∗∗

∗−0

.030

∗∗∗

(0.0

92)

(0.0

92)

(0.0

24)

(0.0

22)

(−1.

085)

(−1.

063)

(−0.

028)

(−0.

023)

(−3.

956)

(−3.

992)

Post

DPA

Di,t

−0.0

13−0

.026

−0.0

04−0

.006

−0.0

12(−

0.73

3)(−

0.79

4)(−

0.15

1)(−

0.31

5)(−

1.34

2)D

PAD

Firm

i∗ Pos

tDPA

Di,t

−0.0

040.

033

0.04

0−0

.014

0.01

1(−

0.26

4)(1

.149

)(1

.294

)(−

0.73

9)(1

.180

)Po

stD

PAD

0506

i,t−0

.016

−0.0

070.

042

−0.0

260.

004

(−0.

540)

(−0.

101)

(0.6

64)

(−1.

020)

(0.2

98)

Post

DPA

D07

09i,t

0.00

60.

001

0.01

30.

018

−0.0

26∗∗

(0.2

72)

(0.0

23)

(0.4

47)

(0.8

52)

(−2.

115)

Post

DPA

D10

13i,t

−0.0

24−0

.050

−0.0

92∗∗

−0.0

18−0

.018

∗

(−1.

110)

(−1.

595)

(−2.

015)

(−0.

744)

(−1.

695)

DPA

DFi

rmi∗ P

ostD

PAD

0506

i,t−0

.008

–0.0

080.

060

0.00

90.

008

(−0.

253)

(−0.

103)

(0.7

34)

(0.4

95)

(0.8

53)

DPA

DFi

rmi∗ P

ostD

PAD

0709

i,t−0

.031

−0.0

26−0

.030

−0.0

310.

019∗

(−1.

541)

(−0.

594)

(−0.

856)

(−1.

380)

(1.6

50)

DPA

DFi

rmi∗ P

ostD

PAD

1013

i,t0.

019

0.08

4∗∗∗

0.07

7∗∗−0

.011

0.00

7(0

.967

)(2

.836

)(2

.266

)(−

0.40

0)(0

.591

)Si

zei,t

−1

−0.0

43∗∗

∗−0

.043

∗∗∗

−0.0

43∗

−0.0

44∗

−0.0

46∗∗

∗−0

.046

∗∗∗

−0.0

15∗∗

∗−0

.015

∗∗∗

0.00

10.

001

(−5.

387)

(−5.

377)

(−1.

762)

(−1.

780)

(−2.

744)

(−2.

731)

(−3.

392)

(−3.

372)

(0.4

83)

(0.5

36)

RO

Ai,t

−1

0.01

10.

014

0.03

50.

036

0.14

60.

140

−0.0

000.

003

−0.0

37−0

.039

(0.0

42)

(0.0

54)

(0.1

04)

(0.1

08)

(0.4

52)

(0.4

37)

(−0.

007)

(0.0

42)

(−0.

904)

(−0.

947)

(Con

tinue

d)

1100 R. LESTER

TA

BL

E7—

Con

tinue

d

Pan

elA

:Reg

ress

ion

anal

ysis

ofdo

mes

tic

inve

stm

ent

Dom

Inv i

,tFo

rInv

i,t

All

Firm

sD

omes

tic-

On

lyFi

rms

Mul

tin

atio

nal

Firm

sM

ulti

nat

ion

alFi

rms

Dep

ende

ntV

aria

ble:

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10)

Cas

hFl

owi,t

−1

0.31

30.

311

0.39

30.

388

0.21

90.

223

0.30

2∗∗∗

0.30

1∗∗∗

0.18

1∗∗0.

182∗∗

(1.3

09)

(1.3

03)

(1.0

37)

(1.0

27)

(0.6

87)

(0.7

13)

(3.7

92)

(3.8

38)

(2.2

50)

(2.2

62)

MT

Bi,t

−1

0.00

8∗∗∗

0.00

8∗∗∗

0.02

2∗∗∗

0.02

2∗∗∗

0.03

4∗∗∗

0.03

5∗∗∗

0.00

4∗∗0.

004∗∗

0.00

10.

001

(3.9

25)

(3.9

39)

(2.8

45)

(2.8

77)

(3.3

54)

(3.3

53)

(2.0

87)

(2.1

21)

(1.5

06)

(1.5

26)

Lev

erag

e t−

1−0

.029

∗∗∗

−0.0

29∗∗

∗−0

.040

∗∗∗

−0.0

40∗∗

∗−0

.036

∗∗∗

−0.0

36∗∗

∗−0

.020

∗∗∗

−0.0

20∗∗

∗−0

.009

∗∗−0

.009

∗∗

(−3.

008)

(−3.

035)

(−3.

126)

(−3.

149)

(−3.

517)

(−3.

597)

(−2.

985)

(−2.

800)

(−2.

354)

(−2.

508)

Year

-by-

indu

stry

FE?

YY

YY

YY

YY

YY

Obs

erva

tion

s8,

274

8,27

42,

689

2,68

94,

360

4,36

05,

585

5,58

54,

607

4,60

7R

20.

167

0.16

70.

174

0.17

50.

174

0.17

50.

155

0.15

60.

126

0.12

8

Pan

elB

:Reg

ress

ion

anal

ysis

ofdo

mes

tic

empl

oym

ent

Dom

Emp i

,tFo

rEm

p i,t

All

Firm

sD

omes

tic-

On

lyFi

rms

Mul

tin

atio

nal

Firm

sM

ulti

nat

ion

alFi

rms

Dep

ende

ntV

aria

ble:

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10)

DPA

DFi

rmi

0.07

60.

077

−0.1

22−0

.121

−0.0

34−0

.035

0.12

10.

121

−0.5

41∗∗

∗−0

.542

∗∗∗

(1.0

10)

(1.0

16)

(−1.

037)

(−1.

034)

(−0.

305)

(−0.

311)

(1.3

08)

(1.3

10)

(−4.

250)

(−4.

263)

Post

DPA

Di,t

−0.0

33−0

.021

0.02

8−0

.047

−0.3

07∗∗

(−0.

585)

(−0.

145)

(0.2

20)

(−0.

801)

(−2.

221)

DPA

DFi

rmi∗ P

ostD

PAD

i,t−0

.144

∗−0

.116

−0.0

73−0

.150

∗0.

183

(−1.

977)

(−1.

072)

(−0.

800)

(−1.

868)

(1.1

60)

Post

DPA

D05

06i,t

−0.1

46−0

.231

∗∗−0

.199

−0.0

89−0

.243

∗

(−1.

506)

(−2.

125)

(−1.

610)

(−0.

867)

(−1.

682)

Post

DPA

D07

09i,t

0.01

1−0

.044

0.00

1−0

.003

−0.4

27∗∗

(0.1

19)

(−0.

177)

(0.0

07)

(−0.

029)

(−2.

124)

Post

DPA

D10

13i,t

0.03

70.

185

0.27

5−0

.036

−0.2

14(0

.442

)(1

.170

)(1

.457

)(−

0.44

1)(−

1.05

3)

(Con

tinue

d)

FIRM RESPONSES TO DOMESTIC PRODUCTION INCENTIVES 1101

TA

BL

E7—

Con

tinue

d

Pan

elB

:Reg

ress

ion

anal

ysis

ofdo

mes

tic

empl

oym

ent

Dom

Emp i

,tFo

rEm

p i,t

All

Firm

sD

omes

tic-

On

lyFi

rms

Mul

tin

atio

nal

Firm

sM

ulti

nat

ion

alFi

rms

Dep

ende

ntV

aria

ble:

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10)

DPA

DFi

rmi∗ P

ostD

PAD

0506

i,t−0

.026

0.23

00.

182

−0.1

320.

176

(−0.

267)

(1.1

41)

(1.3

75)

(−1.

378)

(0.9

65)

DPA

DFi

rmi∗ P

ostD

PAD

0709

i,t−0

.185

∗∗−0

.131

−0.0

25−0

.202

∗∗0.

137

(−2.

176)

(−0.

750)

(−0.

194)

(−2.

500)

(0.8

52)

DPA

DFi

rmi∗ P

ostD

PAD

1013

i,t−0

.177

−0.2

72∗

−0.2

30∗

−0.1

150.

228

(−1.

604)

(−1.

909)

(−1.

794)

(−1.

098)

(1.0

59)

Size

i,t−

10.

754∗∗

∗0.

754∗∗

∗0.

740∗∗

∗0.

743∗∗

∗0.

741∗∗

∗0.

743∗∗

∗0.

806∗∗

∗0.

805∗∗

∗0.

921∗∗

∗0.

921∗∗

∗

(35.

553)

(35.

543)

(13.

645)

(13.

831)

(14.

165)

(14.

220)

(20.

931)

(21.

016)

(19.

498)

(19.

727)

RO

Ai,t

−1

0.56

60.

572

1.16

6∗∗∗

1.17

3∗∗∗

1.53

1∗∗∗

1.53

1∗∗∗

0.46

70.

478

1.86

6∗∗∗

1.87

7∗∗∗

(1.2

63)

(1.2

73)

(2.7

55)

(2.7

36)

(4.5

00)

(4.4

69)

(1.1

22)

(1.1

45)

(2.8

66)

(2.8

97)

Cas

hFl

owi,t

−1

0.16

40.

164

0.78

1∗0.

798∗

0.43

90.

443

−0.4

19−0

.426

−0.8

63−0

.859

(0.3

57)

(0.3

58)

(1.9

06)

(1.9

72)

(1.0

05)

(1.0

19)

(−0.

861)

(−0.

877)

(−1.

365)

(−1.

380)

Inve

ntor

y i,t

−1

1.44

8∗∗∗

1.44

8∗∗∗

1.25

4∗∗∗

1.24

5∗∗∗

0.65

40.

641

2.15

1∗∗∗

2.14

7∗∗∗

2.73

8∗∗2.

745∗∗

(3.2

41)

(3.2

41)

(2.7

44)

(2.7

44)

(1.1

88)

(1.1

67)

(2.8

87)

(2.8

87)

(2.2

31)

(2.2

45)

Tang

ibili

tyi,t

−1

−0.2

40−0

.241

−0.2

15−0

.225

−0.7

79−0

.789

−0.1

98−0

.197

0.13

90.

143

(−0.

572)

(−0.

576)

(−0.

388)

(−0.

409)

(−1.

248)

(−1.

273)

(−0.

468)

(−0.

465)

(0.1

69)

(0.1

74)

Year

-by-

indu

stry

FE?

YY

YY

YY

YY

YY

Obs

erva

tion

s8,

296

8,29

62,

731

2,73

14,

366

4,36

65,

565

5,56

55,

337

5,33

7R

20.

740

0.74

00.

647

0.64

90.

616

0.61

70.

685

0.68

50.

469

0.47

0

Th

ista

ble

pres

ents

resu

lts

ofO

LS

regr

essi

ons

that

test

dom

esti

cac

tivi

ty,i

ncl

udin

gdo

mes

tic

capi

tale

xpen

ditu

res

(pan

elA

)an

ddo

mes

tic

empl

oym

ent(

pan

elB

),by

DPA

Dfi

rms,

rela

tive

toa

mat

ched

sam

ple

ofco

ntr

olfi

rms,

inth

epo

st-D

PAD

peri

od.I

nbo

thpa

nel

s,co

lum

ns

1an

d2

incl

ude

allfi

rms;

colu

mn

s3–

6in

clud

edo

mes

tic-

only

firm

s;an

dco

lum

ns

7–10

incl

ude

mul

tin

atio

nal

firm

sin

the

BE

Ada

ta.

An

yob

serv

atio

nth

at(1

)is

mat

ched

tom

ulti

nat

ion

alB

EA

data

,(2

)re

port

sat

leas

ton

em

ater

ial

fore

ign

subs

idia

ryin

Exh

ibit

21da

ta,

or(3

)re

port

sn

onze

ro,

non

mis

sin

gfo

reig

nsa

les

orse

gmen

tfo

reig

nas

sets

base

don

Com

pust

atse

gmen

tda

taar

edr

oppe

dfr

omth

esa

mpl

efo

rpu

rpos

esof

iden

tify

ing

dom

esti

c-on

lyfi

rms

inco

lum

ns

3an

d4.

Col

umn

s5

and

6ex

clud

ean

yob

serv

atio

nth

at(1

)is

mat

ched

tom

ulti

nat

ion

alB

EA

data

,(2)

repo

rts

posi

tive

pre-

tax

fore

ign

inco

me

(PIF

O),

or(3

)re

port

spo

siti

vefo

reig

nta

xex

pen

se(T

XFO

).C

olum

ns

9an

d10

inpa

nel

A(B

)pr

esen

tre

sult

sfr

omte

stin

gfo

reig

nin

vest

men

t(e

mpl

oym

ent)

.All

vari

able

sar

ede

fin

edin

appe

ndi

xB

.t-s

tati

stic

sar

epr

esen

ted

inpa

ren

thes

es.E

ach

spec

ifica

tion

incl

udes

year

-by-

indu

stry

fixe

def

fect

s,an

dst

anda

rder

rors

are

clus

tere

dby

indu

stry

.∗ ,∗∗

,an

d∗∗

∗in

dica

test

atis

tica

lsig

nifi

can

ceat

the

10%

,5%

,an

d1%

leve

ls,r

espe

ctiv

ely.

1102 R. LESTER

Columns 5–6 present results using an alternative definition of domestic-only firms; the sample used omits any DPAD-control pairs with positive,nonmissing values for pre-tax foreign income (PIFO) and foreign tax ex-pense (TXFO), resulting in a larger sample (n = 4,360). I observe similarresults; based on reported fixed assets of $1.24 billion for this sample inthe year prior to claiming the DPAD, the coefficient of 0.077 in column 6implies greater spending of $95.5 million by the average firm.24

In contrast, I do not observe any additional domestic capital expendi-ture spending by multinational DPAD firms, relative to their matched con-trol firms, in figure 3, panel B. The pre-period trends are generally parallel(with the exception of year t – 3); two years after the incentive is in place,control firms begin to invest more relative to the DPAD firms. However,the graphs do not show amounts that are statistically different. Results formultinationals are presented in columns 7 and 8 of table 7 and are consis-tent with the graphical evidence.

This lack of domestic investment activity could be attributable to thesemultinationals instead investing offshore.25 The results in column 10confirm this: multinationals report a positive and statistically significanteffect for foreign investment, relative to the matched control firms, in the2007–2009 period. The coefficient of 0.019 suggests that being a multina-tional DPAD firm is associated with greater foreign capital expendituresequivalent to 1.9% of total PPE. Given the average DPAD firm in the multi-national subsample reports fixed assets of $3.06 billion in the year priorto claiming the DPAD (untabulated), this is equivalent to greater foreigninvestment spending of $58.1 million by the average multinational DPADfirm. Notably, the period in which this effect occurs is the same period inwhich multinationals report different cross-border shifting (table 6, panelB), suggesting that the domestic tax savings from the DPAD are associatedwith this foreign investment spending.

Table 7, panel B, presents the employment results. For the pooled sampleof firms, I observe a statistically significant reduction in the level of employ-ment by DPAD firms post-DPAD, relative to the sample of matched control

24 Online appendix E shows results of similar significance and magnitude using two addi-tional definitions of domestic-only firms: one based on zero or missing values for PIFO, TXFO,and foreign segment sales (n = 3,418), and one in which all five fields indicating a foreignpresence are equal to 0 or are missing (PIFO, TXFO, foreign segment sales, foreign segmentassets, and Exhibit 21; n = 2,181).

25 Desai, Foley, and Hines [2005] use data on domestic and foreign capital expenditures inthe 1980s and 1990s to show that foreign and domestic investment by U.S. multinational firmsare complements rather than substitutes. Observing a substitution effect for the multinationalDPAD firms suggests that (1) the relation between foreign and domestic investment spendingis different in the more recent period (as seen in the literature on the repatriation tax, whichshows that U.S. multinationals were more likely to invest offshore rather than in the UnitedStates; see Edwards, Kravet, and Wilson [2016] and Hanlon, Lester, and Verdi [2015]), and/or(2) that DPAD firms’ investment patterns differ from other multinationals (which could be animpetus for providing this targeted type of tax incentive).

FIRM RESPONSES TO DOMESTIC PRODUCTION INCENTIVES 1103

firms. The coefficient of –0.144 in column 1 means that claiming the DPADis associated with a 14.4% lower level of employment after the DPAD, rela-tive to the matched sample of control firms. Given that the average DPADfirm employs approximately 11,400 workers, this coefficient translates intoapproximately 1,640 fewer domestic workers at the average DPAD firm inthe post-DPAD period, relative to the matched sample of control firms.26

I also test how these results vary based on the geographic presence ofthe firm. Figure 4 first shows that the DPAD and control firms within thedomestic-only sample exhibited very similar trends pre-DPAD; however, af-ter the DPAD was in place, the lines diverge, with DPAD firms reportingsignificantly fewer workers relative to the control firms. Consistent with thisgraphical evidence, I observe a negative and statistically significant coeffi-cient of –0.272 in column 4 of table 7, panel B, implying a 27.2% decreasein the level of employment; based on the average employment of 3,200workers for this subsample, this coefficient means that domestic-only DPADfirms employed 870 fewer workers since 2010, relative to the matched sam-ple of control firms.

Columns 5 and 6 present results using the alternative definition ofdomestic-only firms based on zero or missing values for pretax foreignincome and foreign tax expense. In this larger sample, I continue to find acoefficient of similar significance and magnitude.27

Notably, the negative coefficient in column 4 occurs in the same periodin which I observed greater domestic investment by this same sample offirms (panel A, column 4). One interpretation is that these domestic-onlyfirms substitute capital for labor; the greater domestic investment spendingobserved in panel A is associated with a lower demand for domestic em-ployees. This finding is consistent with studies in economics examining therelation between the demand for capital and labor (Arrow et al. [1961];Chirinko [2002]; Antras [2004]) as well as a recent study of the DPAD us-ing matched employee-firm IRS data (Dobridge, Landefeld, and Morten-son [2019]). Assuming that the relative prices of labor and capital remainconstant, these results imply a change in the log of the capital/labor ra-tio of approximately 0.403. This amount resembles recent estimations ofthe elasticity of factor substitution between capital and labor; for example,Leon-Ledesma, McAdam, and Willman [2010] estimate a range of 0.491–1.702, and Chirinko and Mallick [2016] provide a range from 0.12 to 0.55with a preferred point estimate of 0.40.28

26 Using the unlogged number of employees, untabulated tests of differences-in-meansshow that the control firms increased employment by approximately 1,100 workers, whereasthe DPAD firms reduced their labor force by 921 workers; thus these tests (the regressionspecification) implies a difference-in-differences estimate of 2,201 (1,640) workers.

27 Online appendix E confirms the robustness of the employment results to two additionalways of identifying domestic-only firms.

28 The CES production function, introduced by Dickinson [1954] and Solow [1956], in-cludes a term σ that measures the elasticity of substitution between capital K and labor N. This

1104 R. LESTER

Additional untabulated results reveal two important findings related tothis result. First, I confirm that the statistically significant investment andemployment effects occur within the same subset of firms—those in thetop quartile based on the number of domestic employees. Second, I ob-serve that the industry distribution for the domestic-only firms differs fromthat for the full sample. Although manufacturing firms continue to com-prise the largest proportion of this subsample, it is lower relative to thefull sample (36.6% as compared to 56.0% from table 1, panel B). Further,two additional industries—Oil & Gas and Utilities—comprise a much largerproportion of this subsample (27.8% and 14.1%, respectively).29

Multinational DPAD firms also report a decline in domestic employmentrelative to control firms, as seen in figure 4, panel B. However, the pre-period trends for these firms are not as clearly parallel, with the differencenarrowing in year t – 1. This graph suggests that there could possibly beother factors contributing to the differences in multinational DPAD firmemployment, and therefore, the results should be interpreted with this inmind. The statistically significant coefficient of –0.150 in column 7 meansthat claiming the DPAD is associated with a 15.0% lower level of employeespost-DPAD, relative to the matched sample of firms; given that the averagemultinational DPAD firm employs approximately 15,800 domestic workers,this is equivalent to 2,370 fewer domestic workers for this subsample ofDPAD claimants. The results imply a change in the log of the capital/laborratio of approximately 0.18, which is also within the range from Chirinkoand Mallick [2016].

I observe no statistically significant effect for foreign labor in columns 9and 10. Given the foreign investment results in panel A, these results meanthat there is not a substitution of foreign capital for foreign labor. One pos-sible explanation is that the foreign investment relates to firm expansionin foreign jurisdictions, and that the DPAD firms shift or transfer existing

ratio is defined as the following: dlog(K/N)/dlog(FN/FK), where F captures the factor prices.That is, the elasticity of substitution is the percentage change in capital and labor factor pro-portions due to a change in the marginal products (or factor price ratio). I make a simplifyingassumption that the change in the factor price ratio equals 1.

29 Anecdotal evidence confirms increased automation in these three sectors. For example,a 2012 article on manufacturing stated that “firms benefit from new materials, cheaper robots,smarter software, and abundance of online services and 3D printers” (Economist [2012]). Asanother example, a 2009 article stated that “today’s power plant control room is evolving intoan almost office-like setting, typically quiet and with few staff . . . decisions are made by a teamof experts, most of whom will not be on site” (Leimbach [2009]). More recent articles discusspersistent job losses in the energy sector since 2010, attributing the layoffs to “automation andthe rise of computer technology” (DiChristopher, Wells, and Schoen [2016]; Eaton [2016]).As a specific example, DTE Energy, a Michigan power company, began claiming the DPAD in2007. In the post-DPAD period, the company announced a $2 billion annual capital invest-ment program to “upgrade its aging power plants, improve its pollution controls, and add1,000 megawatts of renewable energy.” However, financial statement employment data showthat the firm’s employment dropped by 1,500 workers from the pre-DPAD period to the endof 2013.

FIRM RESPONSES TO DOMESTIC PRODUCTION INCENTIVES 1105

foreign employees to new foreign establishments. Unfortunately, becauseforeign establishment-level details are not available in the BEA data, I amunable to empirically test this possible explanation.

The domestic employment results are subject to one important caveat.The BEA data do not include information on contract labor; therefore, tothe extent that the DPAD firms substitute domestic employees for domesticcontractors or foreign outsourced workers, I would be unable to detect andmeasure such effects.

6. Robustness Tests

6.1 POSITIVE SHOCKS AT DPAD FIRMS AND FALSIFICATION TESTS

One potential alternative explanation for the results I document is thatthe DPAD firms experience a positive performance shock that drives differ-ences in intertemporal shifting, cross-border shifting, domestic investment,and domestic employment. For this to occur, these positive shocks must oc-cur at the same time as the firm is eligible for and claims the DPAD, andthe shocks must disproportionately affect the DPAD firms, relative to thematched control firms. While matching on performance and controllingfor ROAi,t throughout the regression specifications mitigates the potentialeffects of these shocks, I perform two tests to rule out this alternative expla-nation. First, I construct three measures of good news and positive shocksand test whether the proportion of DPAD firms reporting these positiveshocks is greater relative to the matched control sample. I find that theoverall proportion of both DPAD and control firms reporting these shocksis low (approximately 1.5–3.2% of the samples) and is not statisticallydifferent across the two groups, such that it is unlikely these shocks aredriving the results I document. Furthermore, all results hold after droppingfirms reporting such good news shocks. This analysis is presented in onlineappendix F.

Second, I conduct a falsification test in which I re-estimate all results overthe pre-DPAD period from 1997 through 2003 and use placebo treatmentdates of 1999 and 2002. When using these alternative dates, I do not observesimilar effects in these earlier years as those documented in tables 5–7 (seeonline appendix G). Thus, the effects I find are specific to the true post-DPAD period and do not appear to be persistent or systematic performancedifferences between the DPAD and control firms.

6.2 ADDITIONAL ANALYSIS OF REAL ACTIVITIES: BONUS DEPRECIATION

Another alternative explanation is that the bonus depreciation tax in-centive, which permits firms to accelerate the depreciation of their fixedassets even more quickly than regular tax rules stipulate, could drive theinvestment results. In 2011, the bonus depreciation rules permitted firmsto expense the entire cost of new assets (Edgerton [2010]; Zwick and

1106 R. LESTER

Mahon [2017]).30 To address concerns that DPAD firms were more likelyor able to claim bonus depreciation relative to control firms and that, byextension, this alternative tax incentive could drive the investment results,I test the difference in DPAD firms and control firms’ deferred tax ex-pense in 2011 when the bonus depreciation incentive was the greatest. Iexamine deferred tax expense because this financial accounting accrualcaptures the amount of book/tax difference attributable to deferred taxassets and liabilities and therefore should increase when firms claim bonusdepreciation. If DPAD firms claimed significantly more bonus depreciationin 2011, then I would expect to observe much higher levels of deferredtax expense in that year. However, I find no statistically significant differ-ence in deferred tax expense reported by DPAD firms relative to the con-trol firms in 2011 (t = 1.14), nor when testing the difference in deferredtax expense over the entire sample period (t = 0.97). Although this doesnot completely refute the possible alternative explanation for the results(given that actual tax depreciation is unobservable), it supports the con-clusion that the DPAD is associated with the greater domestic investmentspending.

7. Conclusion

This paper studies how firms have responded to the DPAD and whetherthis tax incentive resulted in the intended increases in U.S. investment andemployment. I find that the DPAD is associated with an increase in theamount of domestic investment spending, but that this result only occurswithin the sample of domestic-only firms. Multinationals claiming the in-centive exhibit an increase in foreign investment spending post-DPAD. Oneexplanation for the delayed investment spending is that firms initially en-gage in other accounting responses, such as shifting income across timeand across borders. These channels enable firms to quickly respond to theincentive and maximize the tax savings in the first years that the DPAD ben-efit was available.

These results contribute to the literature that studies the DPAD by ex-tending the analysis beyond investment decisions. The results inform ex-pectations of how firms will respond to the lower domestic corporateincome tax rate: I show that some firms respond with greater domestic in-vestment, but this result is concentrated within the sample of firms thatdo not have foreign operations. Furthermore, I show that these results areaccompanied by economically significant amounts of intertemporal and

30 Bonus depreciation was first allowed in 2001 at a rate of 30%. This means that 30%of the cost of an asset could be depreciated immediately, with the remaining amount to bedepreciated using the Modified Accelerated Cost Recovery System (MACRS) depreciationschedule, based on the life of the asset. During the sample period, the bonus depreciationpercentage was 0% (1997–2000, 2005–2007), 30% (2001–2002), 50% (2003–2004, 2008–2010,2012–2014), and 100% (2011).

FIRM RESPONSES TO DOMESTIC PRODUCTION INCENTIVES 1107

cross-border shifting as well as changes in domestic employment. The re-duction in domestic employment suggests a capital–labor substitution, in-consistent with the objective of the tax incentive to create and preserve U.S.jobs.

This incentive was repealed as part of the TCJA. Ignoring the otherchanges implemented by this law, a repeal of the DPAD alone may haveresulted in lower investment spending by some firms, as well as a reducedincentive to retain income in the United States. However, the TCJA imple-mented several features that magnify the incentives previously introducedby the Section 199 deduction. For example, the corporate tax rate droppedto 21%, thereby offering a 14% statutory tax rate reduction (as opposedto 3.15% for DPAD firms). Further, the TCJA introduced immediate ex-pensing of capital expenditures, thereby providing very large tax incentivesfor investment. In addition to studying the effects of these changes, I lookforward to future research that further studies how accounting rules andtax policies interact to affect U.S. domestic production and, more broadly,worldwide corporate location decisions.

1108 R. LESTER

A P P E N D I X A

Search Terms to Identify DPAD Firms

The following search terms are used to identify DPAD firms based on dis-cussion of the tax benefit in firms’ publicly filed financial statements. I thenread each item and retain those that relate to the Section 199 deduction.

Search Terms:

Section 199/Sec. 199

Section 199 manufacturing deduction/credit

DPAD

DMD

Domestic Production/Domestic Production Activities/Domestic Produc-tion Deduction

Domestic Manufacturing Deduction

Domestic Production Activity Deduction

Manufacturing Deduction/Manufacturer’s Deduction/Manufacturers De-duction/Manufacturers’ Production Deduction

Manufacturing Benefit/Manufacturers Benefit/Manufacturer’s Benefit/

Activities Deduction/Activity Deduction

Production Benefit/Deduction

Production Activities

Qualified Production

Qualified Domestic Production

Qualified Activities

QPAI

Qualified Manufacturing Activities

Production Exclusion

Manufacturing Exclusion

FSP 109-1

A P P E N D I X B

Variable Definitions

Cashi,(t,t + n) Total cash holdings (CHE) from t to t + n, divided by total assets fromyear t to t + n

Cash Flowi,t − 1 Operating cash flow (OANCF), divided by total assets in year t – 1DomEmpi,t The log of the number of domestic employees in year t. For

domestic-only firms, this variable equals the number of employees(EMP) in Compustat. For multinational firms, the variable equalsthe number of employees reported by U.S. parent companies on theBEA Survey of U.S. Direct Investment Abroad.

(Continued)

FIRM RESPONSES TO DOMESTIC PRODUCTION INCENTIVES 1109

A P P E N D I X B—Continued

�DomEmpi,(t − 3,t) The difference in the level of DomEmpi,t from year t – 3 to year t.DomInvi,t Domestic capital expenditures in year t, divided by net plant,

property, and equipment (PPENT) in year t – 1. Fordomestic-only firms, domestic capital expenditures equal totalcapital expenditures (CAPX) in Compustat. For multinationalfirms, domestic capital expenditures equal the amount ofdomestic capital expenditures reported by U.S. parentcompanies on the BEA Survey of U.S. Direct InvestmentAbroad.

�DomInvi,(t − 3,t) The difference in the level of DomInvi,t from year t – 3 to year t.Dom NOL Amounti,t The amount of a firm’s domestic tax loss carryforward reported

in its financial statements in year t, scaled by total assets in yeart. Domestic tax loss carryforwards are hand-collected fromfirms’ financial statements.

Dom NOL Indicatori,t An indicator equal to 1 if the firm reports a domestic tax losscarryforward in its financial statements in year t, and 0otherwise. Domestic tax loss carryforwards are hand-collectedbased on a firms’ financial statements.

DomROSi,(t,t + n) Domestic return on sales, measured as pre-tax domestic incomefrom t to t + n (PIDOM), divided by total domestic segmentsales from year t to t + n.

�DomROSi,(t − 3,t) The difference in the level of DomROSi,t from year t – 3 to year t.%DomSalesi,(t,t + n) The percentage of sales that are domestic, calculated as domestic

segment sales from t to t + n, divided by total sales from yeart to t + n (SALE).

DPAD Firmi An indicator equal to 1 if the firm ever discusses the DPADbenefit in its annual financial statements and 0 otherwise.

ForEmpi,t The log of the sum of the number of foreign employees in year t.The number of foreign employees is reported by each foreignaffiliate on the BEA Survey of U.S. Direct Investment Abroad.

ForInvi,t The sum of foreign capital expenditures in year t, divided by netplant, property, and equipment (PPENT) in year t – 1. Foreignexpenditures are equal to the amount of capex reported byeach foreign affiliate reported on the BEA Survey of U.S.Direct Investment Abroad.

GMShift%i,t The amount of shifted gross margin, where gross margin isdefined as quarterly sales (SALEQ), less quarterly cost of goodssold (COGSQ) following Maydew [1997]. The shifted amountis one half of the negative of the fourth-quarter shift, plus thefirst-quarter reversal, scaled by total assets in year t – 1.

�GMShift%i,(t − 3,t) The difference in the level of GM Shift%i,t from year t – 3 to yeart.

Inventoryi,t − 1 Total inventory (INVT) in year t – 1, divided by total assets in yeart – 1.

Leveragei,t − 1 Total debt in year t – 1 (DLTT + DLC), divided by the firm’smarket value of equity in year t – 1 (PRCC F∗CSHO). For H2,this is measured from year t to t + n.

MTBi,t − 1 The ratio of the market value of equity in year t – 1(PRCC F∗CSHO) to the book value of equity in year t – 1(SEQ).

NOL Indicatori,t An indicator equal to 1 if the firm reports a tax loss carryforward(TLCF) in its financial statements in year t and 0 otherwise.

(Continued)

1110 R. LESTER

A P P E N D I X B—Continued

OpIncShift% The amount of shifted operating income, where operatingincome is defined as quarterly sales (SALEQ), less quarterlycost of goods sold (COGSQ) and quarterly SG&A (XSGAQ).The shifted amount is one half of the negative of thefourth-quarter shift, plus the first-quarter reversal, scaled bytotal assets in year t – 1.

�OpIncShift%i,(t − 3,t) The difference in the level of OpInc Shift%i,t from year t – 3 toyear t.

PostDPADi,t An indicator equal to 1 for each DPAD firm and its matchedcontrol beginning in the year that the DPAD firm firstdiscloses/discusses the benefit in its annual financialstatements, and 0 otherwise.

PostDPAD 0506i,t An indicator equal to 1 for each DPAD firm and its matchedcontrol if the DPAD firm first discloses/discusses the DPADbenefit in its 2005 (2006) annual financial statements, and 0otherwise.

PostDPAD 0709i,t An indicator equal to 1 for each DPAD firm and its matchedcontrol if the DPAD firm first discloses/discusses the DPADbenefit in its 2007 (2008, 2009) annual financial statements,and 0 otherwise. The indicator is also equal to 1 for the DPADfirm and its matched control firm if the DPAD firm firstdiscusses the DPAD in its 2005 or 2006 financial statements.

PostDPAD 1013i,t An indicator equal to 1 for each DPAD firm and its matchedcontrol if the DPAD firm first discloses/discusses the DPADbenefit in its 2010 (2011, 2012, 2013) annual financialstatements, and 0 otherwise. The indicator is also equal to 1 forthe DPAD firm and its matched control firm if the DPAD firmfirst discusses the DPAD in its 2005, 2006, 2007, 2008, or 2009financial statements.

Rate Diffi,(t,t + n) The difference in the U.S. statutory rate in effect during thesample period (35%) and each firm’s average foreign effectivetax rate, calculated as the sum of foreign income tax expense(TXFO) from year t to either t + 1 or t + 4, divided by thefirm’s foreign income (PIFO) over the same five-year period.

RDi,(t,t + n) Total R&D expense (XRD) from year t to t + n, divided by thefirm’s total assets from year t to t + n.

RD Crediti,t − 1 An indicator equal to 1 if the firm had estimated R&D tax creditsin year t – 1. For years 1997–2010, firms are identified as havingR&D tax credits based on financial statement disclosurescollected by Hoopes [2018]. For years 2011–2013, firms areidentified as having R&D tax credits based on a nonzeroestimated amount of credit. This amount is equal to theaverage R&D tax credit over the prior three years, divided bythe sum of a firm’s domestic tax expense (TXFED) over thesame period based on the Alternative Simplified CreditMethodology under IRC Sec. 41(c)(5). The amount ismultiplied by 50% and then subtracted from the current year’sR&D expenses. The difference is multiplied by 15%. If thisamount is positive, then the indicator is equal to 1 and 0otherwise.

(Continued)

FIRM RESPONSES TO DOMESTIC PRODUCTION INCENTIVES 1111

A P P E N D I X B—Continued

ROAi,t − 1 The ratio of pre-tax income in year t – 1 (OIADP), divided by thefirm’s total assets in year t – 1.

ShiftPeriodi,t An indicator equal to 1 for each DPAD firm and its matchedcontrol for the year in which the DPAD firm first discusses theDPAD benefit in its annual financial statements, as well as anyother year in which the statutory DPAD benefit increases, and 0otherwise.

ShiftPeriod 0506i,t An indicator equal to 1 for each DPAD firm and its matchedcontrol in 2005 (2006) if the DPAD firm first discusses theDPAD benefit in its 2005 (2006) annual financial statements,and 0 otherwise.

ShiftPeriod 0709i,t An indicator equal to 1 for each DPAD firm and its matchedcontrol in 2007 (2008, 2009) if the DPAD firm first discussesthe DPAD benefit in its 2007 (2008, 2009) annual financialstatements, and 0 otherwise. The indicator is also equal to 1 in2007 for each DPAD firm and its matched control if the DPADfirm first discusses the DPAD benefit in its 2005 or 2006financial statement, and 0 otherwise.

ShiftPeriod 1013i,t An indicator equal to 1 for each DPAD firm and its matchedcontrol in 2010 (2011, 2012, 2013) if the DPAD firm firstdiscusses the DPAD benefit in its 2010 (2011, 2012, 2013)annual financial statements, and 0 otherwise. The indicator isalso equal to 1 in 2010 for each DPAD firm and its matchedcontrol if the DPAD firm first discusses the DPAD benefit in its2005, 2006, 2007, 2008, or 2009 financial statement, and 0otherwise.

Sizei,t − 1 Log of total assets in year t – 1. For H2, this is measured from yeart to t + n.

Tangibilityi,t − 1 Total plant, property, and equipment (PPENT) in year t – 1,divided by total assets in year t – 1.

WW ROSi,(t,t + n) Worldwide return on sales, measured as total pre-tax incomefrom year t to t + n (PI), divided by total sales from year t to t+n (SALE).

REFERENCES

ALMEIDA, H., AND M. CAMPELLO. “Financial Constraints, Asset Tangibility, and Corporate In-vestment.” Review of Financial Studies 20 (2007): 1429–60.

ANTRAS, P. “Is the U.S. Aggregate Production Function Cobb-Douglas? New Estimates of theElasticity of Substitution.” Contributions to Macroeconomics 4 (2004): 1–34.

ARROW, K.; H. CHENERY; B. B. MINHAS; AND R. SOLOW. “Capital-Labor Substitution and Eco-nomic Efficiency.” The Review of Economics and Statistics 43 (1961): 225–50.

BIDDLE, G.; G. HILARY; AND R. VERDI. “How Does Financial Reporting Quality Relate to Invest-ment Efficiency?” Journal of Accounting and Economics 48 (2009): 112–31.

BLOUIN, J.; L. KRULL; AND L. ROBINSON “Is U.S. Multinational Intra-Firm Dividend Policy In-fluenced by Reporting Incentives?” The Accounting Review 87 (2012): 1463–91.

BLOUIN, J.; L. KRULL; AND C. SCHWAB. “The Effect of the Domestic Production ActivitiesDeduction on Corporate Payout Behavior.” Working paper, University of Pennsylvania,2014.

BUREAU OF LABOR STATISTICS. “Employment by Major Industry Sector.” 2017. Available athttps://www.bls.gov/emp/ep table 201.htm.

1112 R. LESTER

CHIRINKO, R. “Corporate Taxation, Capital Formation, and the Substitution Elasticity betweenLabor and Capital.” National Tax Journal 55 (2002): 339–55.

CHIRINKO, R., AND MALLICK, D. “The Substitution Elasticity, Factor Shares, Long-Run Growth,and the Low-Frequency Panel Model.” CESifo Working paper series 4895, 2016.

COLLINS, J.; D. KEMSLEY; AND M. LANG. “Cross-jurisdictional Income Shifting and EarningsValuation.” Journal of Accounting Research 36 (1998): 209–29.

COOPER, M., AND M. KNITTEL. “The Implications of Tax Asymmetry for U.S. Corporations.”National Tax Journal 63 (2010): 33–61.

DE SIMONE, L.; K. KLASSEN; AND J. SEIDMAN. “Unprofitable Affiliates and Income Shifting Be-havior.” The Accounting Review 93 (2017): 113–36.

DECHOW, P.; W. GE; AND C. SCHRAND. “Understanding Earnings Quality: A Review of the Prox-ies, Their Determinants and Their Consequences.” Journal of Accounting and Economics 50(2010): 344–401.

DESAI, M., AND J. HINES. “The Uneasy Marriage of Export Incentives and the Income Tax,”in Tax Policy and the Economy, edited by J. M. Poterba. Cambridge, MA: MIT Press, 2001:41–94.

DESAI, M.; F. FOLEY; AND J. HINES. “Foreign Direct Investment and the Domestic Capital Stock.”American Economic Review 95 (2005): 33–38.

DICHRISTOPHER, T.; N. WELLS; AND J. SCHOEN. “Oil-Field Workers Get No Relief As the EnergyIndustry Cuts Deeper in 2016.” CNBC, November 17, 2016.

DICKINSON, H. D. “A Note on Dynamic Economics.” Review of Economic Studies 22 (1954): 169–79.

DOBRIDGE, C.; P. LANDEFELD; AND J. MORTENSON. “Corporate Taxes and the Wage Distribution:Effects of the Domestic Production Activities Deduction.” Working paper, Joint Committeeon Taxation, 2019.

DYRENG, S., AND B. LINDSEY. “Using Financial Accounting Data to Examine the Effect of For-eign Operations Located in Tax Havens and Other Countries of U.S. Multinational Firms’Tax Rates.” Journal of Accounting Research 47 (2009): 1283–316.

DYRENG, S., AND K. MARKLE. “The Effect of Financial Constraints on Income Shifting by U.S.Multinationals.” The Accounting Review 91 (2016): 1601–27.

EATON, C. “U.S. Oil Job Cuts Reach About 118,000.” Houston Chronicle, April 8, 2016.ECONOMIST. “Automation: Making the Future.” Economist, April 21, 2012.EDGERTON, J. “Investment Incentives and Corporate Tax Asymmetries.” Journal of Public Eco-

nomics 94 (2010): 936–52.EDWARDS, A.; T. KRAVET; AND R. WILSON. “Trapped Cash and the Profitability of Foreign Ac-

quisitions.” Contemporary Accounting Research 33 (2016): 44–77.FICH, E.; E. RICE; AND A. TRAN. “Corporate Tax Cuts, Merger Activity, and Shareholder

Wealth.” Working paper, University of Washington, 2017.FOLEY, F.; J. HARTZELL; S. TITMAN; AND G. TWITE. “Why Do Firms Hold So Much Cash? A

Tax-Based Explanation.” Journal of Financial Economics 86 (2007): 579–607.GIROUD, X., AND J. RAUH. “State Taxation and the Reallocation of Business Activity: Evidence

from Establishment-Level Data.” Journal of Political Economy (2018): Forthcoming.GOOLSBEE, A. “Investment Tax Incentives, Prices, and the Supply of Capital Goods.” Quarterly

Journal of Economics 113 (1998): 121–49.GRAHAM, J.; M. HANLON; AND T. SHEVLIN. “Real Effects of Accounting Rules: Evidence from

Multinational Firms’ Investment Location and Profit Repatriation Decisions.” Journal of Ac-counting Research 49 (2011): 137–85.

GUENTHER, D. “Earnings Management in Response to Corporate Tax Rate Changes: Evidencefrom the 1986 Tax Reform Act.” The Accounting Review 69 (1994): 230–43.

HANLON, M.; R. LESTER; AND R. VERDI. “The Effect of Repatriation Tax Costs on U.S. Multina-tional Investment.” Journal of Financial Economics 116 (2015): 179–96.

HASSETT, K., AND G. HUBBARD. “Tax Policy and Business Investment,” in Handbook of PublicEconomics, edited by A. Auerbach and M. Feldstein, volume 3. Amsterdam: Elsevier Science,2002: 1294–343.

FIRM RESPONSES TO DOMESTIC PRODUCTION INCENTIVES 1113

HASSETT, K., AND K. NEWMARK. “Taxation and Business Behavior: A Review of the Recent Liter-ature,” in Fundamental Tax Reform: Issues, Choices and Implications, edited by J. Diamond andG. Zodrow. Cambridge, MA: MIT Press, 2008: 191–214.

HEITZMAN, S., AND R. LESTER. “Net Operating Loss Carryforwards and Corporate Savings Poli-cies.” Working paper, Stanford University, 2019.

HOOPES, J. “The Effect of Temporary Tax Laws on Understanding and Predicting CorporateEarnings.” Working paper, University of North Carolina, 2018.

H.R. REP. NO. 108–393. (2004a).H.R. REP. NO. 108–548. (2004b).H.R. REP. NO. 108–755. (2004c).INTERNAL REVENUE SERVICE. “UTP Filing Statistics.” 2015. Available at https://www.irs.gov/

businesses/corporations/utp-filing-statistics.JENKS, C. “Domestic Production Deduction: FAQs and a Few Answers.” Tax Practice 51 (2006):

166–72.JOINT COMMITTEE ON TAXATION. “Estimates of Federal Tax Expenditures for Fiscal Years 2016–

2020.” Joint Committee on Taxation Report JCX-3-17, 2017.KEMSLEY, D. 1998. “The Effect of Taxes on Production Location.” Journal of Accounting Research

36 (1998): 321–41.KLASSEN, K., AND S. LAPLANTE. “Are U.S. Multinational Corporations Becoming More Aggres-

sive Income Shifters?” Journal of Accounting Research 50 (2012): 1245–86.LEIMBACH, R. “Power Plant Automation: Where We Are and Where We’re Headed.” Power

Magazine, February 1, 2009.LEON-LEDESMA, M.; P. MCADAM; AND A. WILLMAN. “Identifying the Elasticity of Substitution

with Biased Technical Change.” American Economic Review 100 (2010): 1330–57.LESTER, R., AND R. RECTOR. “Which Companies Use the Domestic Production Activities De-

duction?” Tax Notes 152 (2016): 1269–92.MAYDEW, E. “Tax-Induced Earnings Management by Firms with Net Operating Losses.” Journal

of Accounting Research 35 (1997): 83–96.MCNICHOLS, M., AND S. STUBBEN. “Does Earnings Management Affect Firms’ Investment De-

cisions?” The Accounting Review 83 (2008): 1571–1603.OHRN, E. “The Effect of Corporate Taxation on Investment and Financial Policy: Evidence

from the DPAD.” American Economic Journal: Economic Policy 10 (2018): 272–301.OLIPHANT, J. “Trump Touts Tax Reform, Overlooks White House Aide Cohn.” Reuters. Au-

gust 30, 2017. Available at https://www.reuters.com/article/us-usa-trump-tax/trump-touts-tax-reform-overlooks-white-house-aide-cohn-idUSKCN1BA14X.

RICHARDSON, S. “Over-Investment of Free Cash Flow.” Review of Accounting Studies 11 (2006):159–89.

RYAN, P. “House Speaker Paul Ryan Tours Delco Factory, Talks Tax Reform.” 2017. Availableat http://philadelphia.cbslocal.com/2017/09/28/house-speaker-paul-ryan-tours-delco-factory-talks-tax-reform/.

SCHOLES, M.; P. WILSON; AND M. WOLFSON “Firms’ Responses to Anticipated Reductionsin Tax Rates: The Tax Reform Act of 1986.” Journal of Accounting Research 30 (1992):161–85.

SCOTT, R. “The Manufacturing Footprint and the Importance of U.S. Manufacturing Jobs.”Economic Policy Institute Briefing Paper 388, 2015.

SHERLOCK, M. “The Section 199 Production Activities Deduction: Background and Analysis.”Congressional Research Service Report R41988, 2012.

SLEMROD, J. “Do Taxes Matter? Evidence from the 1980s.” American Economic Review 82 (1992):250–56.

SOLOW, R. M. “A Contribution to the Theory of Economic Growth.” Quarterly Journal of Eco-nomics 70 (1956): 65–94.

SUAREZ SERRATO, J. C., AND O. ZIDAR. “Who Benefits from State Corporate Tax Cuts? A LocalLabor Markets Approach with Heterogenous Firms.” American Economic Review 106 (2016):2582–624.

1114 R. LESTER

TANGEL, A., AND P. MCGROARTY. “Tax Incentive Puts More Robots on Factory Floors.” WallStreet Journal, January 26, 2018.

WHITE HOUSE. “Fact Sheet: President Donald J. Trump’s Tax Cuts are Delivering ForHardworking Americans and Our Manufacturers.” 2018. Available at https://www.whitehouse.gov/briefings-statements/president-donald-j-trumps-tax-cuts-delivering-hardworking-americans-manufacturers/.

ZWICK, E., AND J. MAHON. “Tax Policy and Heterogeneous Investment Behavior.” AmericanEconomic Review 107 (2017): 217–48.