Foregn Drect Investment n a World of Multple Taxes by Mhr A. Desa Harvard Unversty and NBER James R. Hnes Jr. Unversty of Mchgan and NBER July 2001 We thank Yumng Zou for excellent research assstance, and Steve Bond, Mark Robson, and partcpants at the World Tax Competton Conference for helpful comments. Fnancal support from the Los and Bruce Zenkel Research Fund at the Unversty of Mchgan and the Dvson of Research at Harvard Busness School s gratefully acknowledged.
Foregn Drect Investment n a World of Multple Taxes ABSTRACT Whle governments have multple tax nstruments avalable to them, studes of the effect of tax polcy on the locatonal decsons of multnatonals typcally focus exclusvely on host country corporate ncome tax rates and ther nteracton wth home country tax rules. Ths paper examnes the mpact of ndrect (non-ncome) taxes on the locaton and character of foregn drect nvestment by Amercan multnatonal frms. Indrect tax burdens sgnfcantly exceed foregn ncome tax oblgatons for these frms and appear to nfluence strongly ther behavor. The nfluence of ndrect taxes s shown to be partly attrbutable to the nablty of Amercan nvestors to clam foregn tax credts for ndrect tax payments. Estmates mply that 10 percent hgher ndrect tax rates are assocated wth 9.2 percent lower reported ncome of Amercan afflates and 8.6 percent lower captal/labor ratos. These estmates carry mplcatons for effcent tradeoffs between drect and ndrect taxaton n rasng revenue whle attractng moble captal. JEL Classfcaton: H87, H25, F23. Mhr A. Desa James R. Hnes Jr. Harvard Busness School Offce of Tax Polcy Research Morgan 363 Unversty of Mchgan Busness School Solders Feld 701 Tappan Street Boston, MA 02163 Ann Arbor, MI 48109-1234 mdesa@hbs.edu jrhnes@umch.edu
1. Introducton. Governments have at ther dsposal several dfferent tax nstruments that can be used sngly or n concert to fnance ther actvtes. These tax alternatves nclude personal and corporate ncome taxes, sales taxes, value added taxes, property taxes, excse taxes, and numerous others. It s not uncommon for a sngle government to elect to use many or all of these alternatves smultaneously. The lkely mpact on a country s ablty to attract nvestment and stmulate economc actvty typcally ranks hghly among the crtera used n makng choces over these tax nstruments. There s by now extensve evdence that hgh tax rates dscourage foregn drect nvestment (FDI), thereby offerng support for the workng hypothess of many governments that mantan low corporate tax rates n order to encourage nvestment. Emprcal studes of the effect of taxaton on FDI typcally consder the mpact of dfferences n corporate ncome tax rates. Ths lterature has consderably less to say about the effect of taxes other than corporate taxes, even though, from a theoretcal standpont, all types of taxes have the potental to reduce FDI. For example, hgh personal ncome tax rates may be reflected at least n part n hgh pretax wages, whch n turn dscourage FDI f labor and captal are complementary. Smlarly, hgh rates of property taxaton may reduce the demand for FDI by as much as hgh rates of ncome taxes. The role of non-ncome taxes may be partcularly mportant for FDI as governments of many countres (ncludng the Unted States) permt multnatonal frms to clam foregn tax credts for corporate ncome taxes pad to foregn governments, but do not extend ths prvlege to taxes other than ncome taxes. As a result, taxes for whch frms are nelgble to clam credts may have the strongest mpact on frm decson-makng, ncludng decsons of where and how much to nvest.
The purpose of ths paper s to nvestgate the effect of the multple tax nstruments that comprse a host country tax system on the magntude and characterstcs of foregn nvestment actvty by Amercan multnatonal frms. In partcular, the emprcal work focuses on the dfferental mpacts of corporate ncome taxes and other taxes, such as personal ncome taxes, property taxes, and value added taxes. The analyss uses data collected by the Bureau of Economc Analyss of the U.S. Department of Commerce on the actvtes of Amercan multnatonal frms n 1982, 1989, and 1994, the most recent years for whch such comprehensve Benchmark data are avalable. The hgh degree of correlaton between corporate ncome tax rates and other tax rates suggests that prevous work may have conflated the effects of these dstnct tax nstruments. Snce the foregn ndrect tax oblgatons of Amercan multnatonal frms are roughly three tmes ther drect tax oblgatons, there s obvous scope for ndrect taxes to nfluence ther behavor. The results confrm that hgh tax rates dscourage busness actvty by Amercan multnatonal frms, but suggest that there are mportant effects of all types of taxes, ncludng taxes other than corporate ncome taxes. Snce Amercans are nelgble to clam tax credts aganst home-country tax labltes for anythng other than foregn ncome taxes, t stands to reason that other taxes such as labor ncome taxes and property taxes mght have the potental to dscourage busness actvty. The results ndcate that ndrect busness taxes nfluence the locaton of proft-generatng busness actvty, wth an mpact that s partcularly notceable on the captal/labor ratos used n offshore operatons. In addton, the paper provdes evdence that ndrect tax rates affect the locaton of profts even controllng for the use of productve nputs. The magntude and mpact of ndrect taxaton on the actvtes of foregn nvestors carres potentally mportant mplcatons for the dynamcs of tax competton. Countres that are eager to protect ther tax bases, possbly at the expense of others, have ncentves to select drect and 2
ndrect tax rates that reflect ths goal. Gven that large captal exporters such as the Unted States provde explct tax relef for drect taxes through foregn tax credts, the scope for competton on ndrect taxes may be much greater and all the more lkely n the absence of blateral treates or multlateral conventons on ndrect taxaton. Consequently, one of the objectves n studyng the behavoral mpact of ndrect taxaton s to antcpate possble pressure ponts for nternatonal tax competton. Secton 2 of the paper descrbes the tax systems used by the Unted States and other countres, and revews the fndngs of earler research on the effect of taxaton on nvestment and other actvtes of multnatonal frms. Secton 3 presents a model of frm behavor and characterzes ts mplcatons for Amercan frms nvestng abroad n countres wth multple taxes. Secton 3 also descrbes the BEA data used to analyze the actvtes of Amercan multnatonal frms. Secton 4 presents the results of estmatng the determnants of captal/labor ratos and proft locaton. Secton 5 s the concluson. 2. Internatonal ncome taxaton n perspectve. 1 It s useful to revew exstng systems of taxng nternatonal ncome n order to nterpret the ncentves facng Amercan frms nvestng n foregn countres usng multple tax nstruments. Ths summary of nternatonal tax rules provdes not only a bass for the analyss that follows n sectons 3 and 4, but also serves as a framework wth whch to nterpret the studes revewed n secton 2.2. 2.1. Internatonal tax practce. 1 Some parts of ths bref descrpton of nternatonal tax rules and evdence of behavoral responses to nternatonal taxaton are excerpted from Hnes (1991, 1997, 1999) and Hnes and Hubbard (1995). 3
The taxaton of nternatonal transactons dffers from the taxaton of domestc economc actvty prmarly due to the complcatons that stem from the taxaton of the same ncome by multple governments. In the absence of double tax relef, the mplcatons of multple taxaton are potentally qute severe, snce natonal tax rates are hgh enough to elmnate, or at least greatly dscourage, most nternatonal busness actvty f appled two or more tmes to the same ncome. 2.1.1 The foregn tax credt. Almost all countres tax ncome generated by economc actvty that takes place wthn ther borders. In addton, many countres ncludng the Unted States tax the foregn ncomes of ther resdents. In order to prevent double taxaton of the foregn ncome of Amercans, U.S. law permts taxpayers to clam foregn tax credts for ncome taxes pad to foregn governments. 2 These foregn tax credts are used to offset U.S. tax labltes that would otherwse be due on foregn-source ncome. The U.S. corporate tax rate s currently 35 percent, so an Amercan corporaton that earns $100 n a foregn country wth a 10 percent tax rate pays taxes of $10 to the foregn government and $25 to the U.S. government, snce ts U.S. corporate tax lablty of $35 (35 percent of $100) s reduced to $25 by the foregn tax credt of $10. 2.1.2 Tax deferral. Amercans are permtted to defer any U.S. tax labltes on certan unrepatrated foregn profts untl they receve such profts n the form of dvdends. 3 Ths deferral s avalable only on 2 The Unted States s not alone n taxng the worldwde ncome of ts resdents whle permttng them to clam foregn tax credts. Other countres wth such systems nclude Greece, Italy, Japan, Norway, and the Unted Kngdom. Under U.S. law, taxpayers may clam foregn tax credts for taxes pad by foregn frms of whch they own at least 10 percent, and only those taxes that qualfy as ncome taxes are credtable. 3 Deferral of home-country taxaton of the unrepatrated profts of foregn subsdares s a common feature of systems that tax foregn ncomes. Other countres that permt ths knd of deferral nclude Canada, Denmark, France, Germany, Japan, Norway, Pakstan, and the Unted Kngdom. 4
the actve busness profts of Amercan-owned foregn afflates that are separately ncorporated as subsdares n foregn countres. The profts of unncorporated foregn busnesses, such as those of Amercan-owned branch banks n other countres, are taxed mmedately by the Unted States. To llustrate deferral, consder the case of a subsdary of an Amercan company that earns $500 n a foregn country wth a 20 percent tax rate. Ths subsdary pays taxes of $100 to the foregn country (20 percent of $500), and mght remt $100 n dvdends to ts parent U.S. company, usng the remanng $300 ($500 - $100 of taxes - $100 of dvdends) to renvest n ts own foregn operatons. The Amercan parent frm must then pay U.S. taxes on the $100 of dvdends t receves (and s elgble to clam a foregn tax credt for the foregn ncome taxes ts subsdary pad on the $100). 4 But the Amercan frm s not requred to pay U.S. taxes on any part of the $300 that the subsdary earns abroad and does not remt to ts parent company. If, however, the subsdary were to pay a dvdend of $300 the followng year, the frm would then be requred to pay U.S. tax (after proper allowance for foregn tax credts) on that amount. U.S. tax law contans provsons desgned to prevent Amercan frms from delayng the repatraton of lghtly-taxed foregn earnngs. These tax provsons apply to controlled foregn corporatons, whch are foregn corporatons owned at least 50 percent by Amercan ndvduals or corporatons who hold stakes of at least 10 percent each. Under the Subpart F provsons of U.S. law, some foregn ncome of controlled foregn corporatons s deemed dstrbuted, and therefore mmedately taxable by the Unted States, even f not repatrated as dvdend payments to Amercan parent frms. 5 4 In ths example, the parent frm s elgble to clam a foregn tax credt of $25, representng the product of foregn taxes pad by ts subsdary and the subsdary's rato of dvdends to after-tax profts [$100 x ($100/$400) = $25]. 5 Subpart F ncome conssts of ncome from passve nvestments (such as nterest and dvdends receved from nvestments n securtes), foregn base company ncome (that arses from usng a foregn afflate as a condut for certan types of nternatonal transactons), ncome that s nvested n Unted States property, money used offshore to nsure rsks n the Unted States, and money used to pay brbes to foregn government offcals. Amercan frms wth foregn subsdares that earn profts through most types of actve busness operatons, and that subsequently renvest 5
2.1.3 Excess foregn tax credts. Snce the foregn tax credt s ntended to allevate nternatonal double taxaton, and not to reduce U.S. tax labltes on profts earned wthn the Unted States, the foregn tax credt s lmted to U.S. tax lablty on foregn-source ncome. For example, an Amercan frm wth $200 of foregn ncome that faces a U.S. tax rate of 35 percent has a foregn tax credt lmt of $70 (35 percent of $200). If the frm pays foregn ncome taxes of less than $70, then the frm would be enttled to clam foregn tax credts for all of ts foregn taxes pad. If, however, the frm pays $90 of foregn taxes, then t would be permtted to clam no more than $70 of foregn tax credts. Taxpayers whose foregn tax payments exceed the foregn tax credt lmt are sad to have excess foregn tax credts; the excess foregn tax credts represent the porton of ther foregn tax payments that exceed the U.S. tax labltes generated by ther foregn ncomes. Taxpayers whose foregn tax payments are smaller than ther foregn tax credt lmts are sad to have defct foregn tax credts. Amercan law permts taxpayers to use excess foregn tax credts n one year to reduce ther U.S. tax oblgatons on foregn source ncome n ether of the two prevous years or n any of the followng fve years. 6 In practce, the calculaton of the foregn tax credt lmt entals certan addtonal complcatons, notable among whch s that total worldwde foregn ncome s used to calculate the foregn tax credt lmt. Ths method of calculatng the foregn tax credt lmt s known as those profts n actve lnes of busness, are not subject to the Subpart F rules, and are therefore able to defer U.S. tax lablty on ther foregn profts untl they choose to remt dvdends at a later date. 6 Foregn tax credts are not adjusted for nflaton, so are generally the most valuable f clamed as soon as possble. Barrng unusual crcumstances, frms apply ther foregn tax credts aganst future years only when unable to apply them aganst ether of the prevous two years. The most common reason why frms do not apply excess foregn tax credts aganst ether of the prevous two years s that they already have excess foregn tax credts n those years. Frms payng the corporate alternatve mnmum tax (AMT) are subject to the same rules, wth the added restrcton that the combnaton of net operatng loss deductons and foregn tax credts cannot reduce AMT labltes 6
worldwde averagng. A taxpayer has excess foregn tax credts f the sum of worldwde foregn ncome tax payments exceeds ths lmt. 7 2.2. Emprcal lessons from nternatonal taxaton. Internatonal tax rules and the tax laws of other countres have the potental to nfluence a wde range of corporate and ndvdual behavor, ncludng, most drectly, the locaton and scope of nternatonal busness actvty. A szable lterature s devoted to measurng behavoral responses to nternatonal tax rules. 8 Ths lterature focuses on the mpact of corporate tax rates on nvestment behavor as well as varous fnancal and organzatonal practces used to avod taxes. 2.2.1 Investment. Tax polces are obvously capable of affectng the volume and locaton of FDI, 9 snce, all other consderatons equal, hgher tax rates reduce after-tax returns, thereby reducng ncentves to commt nvestment funds. Of course, all other consderatons are seldom equal. Countres dffer not only n ther tax polces, but also n ther commercal and regulatory polces, the characterstcs of ther labor markets, the nature of competton n product markets, the cost and local avalablty of ntermedate supples, proxmty to fnal markets, and a host of other attrbutes by more than 90%. It s noteworthy that, snce the AMT rate s only 20%, frms subject to the AMT are consderably more lkely to have excess foregn tax credts than are frms that pay the regular corporate tax. 7 Not all countres that grant foregn tax credts use worldwde averagng. For example, whle Japan uses worldwde averagng, the Unted Kngdom nstead requres ts frms to calculate foregn tax credts on an actvty-by-actvty bass. The Unted States once requred frms to calculate separate foregn tax credt lmts for each country to whch taxes were pad; the current system of worldwde averagng was ntroduced n the md-1970s. 8 See Hnes (1997, 1999) for further elaboraton and crtcal analyss of many of the studes surveyed n ths secton. 9 FDI conssts of changes n the ownershp clams of controllng foregn nvestors. For example, an Amercan parent frm that establshes a wholly-owned foregn afflate wth $100 mllon of equty and $50 mllon of loans from the parent company thereby creates $150 mllon of FDI. In order for foregn nvestment to count as FDI, the Amercan nvestor must own at least 10 percent of the foregn afflate. FDI s the sum of parent fund transfers and Amercan owners shares of ther foregn afflates renvested earnngs, mnus any repatratons to Amercan owners. Pror to 1974, the Unted States reported FDI only for nvestments n whch Amercan owners held at least 25 percent ownershp shares. Reported FDI typcally represents book values. 7
that nfluence the desrablty of an nvestment locaton. Furthermore, the varous tax and regulatory polces that are relevant to foregn nvestors may be correlated wth non-tax features of economes that ndependently affect FDI levels. Consequently, t s necessary to nterpret evdence of the effect of taxaton wth consderable cauton. The emprcal lterature on the effect of taxes on FDI consders almost exclusvely U.S. data, ether the dstrbuton of U.S. drect nvestment abroad, or the FDI patterns of foregners who nvest n the Unted States. 10 The smple explanaton for ths focus s not only that the Unted States s the world s largest economy, but also that the Unted States collects and dstrbutes much more, and hgher-qualty, data on FDI actvtes than does any other country. The avalable evdence of the effect of taxaton on FDI comes n two forms. The frst s tme-seres estmaton of the responsveness of FDI to annual varaton n after-tax rates of return. Implct n ths estmaton s a q-style nvestment model n whch contemporaneous average aftertax rates of return serve as proxes for returns to margnal FDI. Studes of ths type consstently report a postve correlaton between levels of FDI and after-tax rates of return at ndustry and country levels. 11 The mpled elastcty of FDI wth respect to after-tax returns s generally close to unty, whch translates nto a tax elastcty of nvestment of roughly -0.6. The estmated elastcty s smlar whether the nvestment n queston s Amercan drect nvestment abroad or FDI by foregners n the Unted States. The prmary lmtaton of aggregate tme-seres studes s that they are dentfed by yearly varaton n taxes or proftablty that may be correlated wth mportant omtted varables. As a 10 Devereux and Freeman (1995) and Hnes (2001) are recent exceptons. 11 See, for example, Hartman (1984), Boskn and Gale (1987), Newlon (1987), Young (1988), Slemrod (1990), and Swenson (1994). 8
result, t becomes very dffcult to dstngush the effects of taxaton from the effects of other varables that are correlated wth tax rates. Two of the tme-seres studes explot cross-sectonal dfferences that offer the potental for greater explanatory power. Slemrod (1990) dstngushes FDI n the Unted States by the tax regme n the country of orgn. Investors from countres (of whch Slemrod analyzes data for Japan and the Unted Kngdom) wth tax systems smlar to that used by the Unted States receve foregn tax credts for taxes pad to the Unted States. Investors from certan other countres (of whch Slemrod analyzes data for Australa, Canada, France, Germany, and the Netherlands) are more or less exempt from home-country taxaton of any profts earned n the Unted States. Consequently, nvestors from France and Germany have stronger ncentves to nvest n the Unted States durng low-tax years than do nvestors from Japan and the Unted Kngdom, snce Japanese and Brtsh nvestors are elgble to clam tax credts for any U.S. taxes they pay. In hs analyss of data coverng 1962-1987, Slemrod fnds no clear emprcal pattern ndcatng that nvestors from countres that exempt U.S. profts from home-country taxaton are more senstve to tax changes than are nvestors from countres grantng foregn tax credts. Ths evdence suggests ether that home-country tax regmes do not nfluence FDI, or that tme seres varaton n tax rates s nadequate to dentfy tax effects that are nonetheless present. Swenson (1994) consders the tax determnants of ndustry-level FDI n the Unted States over the 1979-1991 perod. U.S. tax changes often affect ndustres to dfferng degrees, based largely on the assets n whch they nvest; ths was partcularly true of tax legslaton enacted n 1981 and 1986. Swenson fnds that ndustres n whch the (U.S.) after-tax cost of captal rose the most after passage of the U.S. Tax Reform Act of 1986 were those n whch foregn nvestors concentrated ther FDI n the post-1986 perod. Ths s consstent wth the tax ncentves of foregn 9
nvestors from countres grantng foregn tax credts, snce such nvestors are the least affected by U.S. tax provsons but t s also possble that foregn nvestors chose to concentrate n such ndustres for any of a number of non-tax reasons. Auerbach and Hassett (1993) lend credence to the latter nterpretaton wth ther fndng that nvestors from countres grantng foregn tax credts were no more lkely than were other foregn nvestors to concentrate ther FDI n taxdsadvantaged ndustres after 1986. Other studes of nvestment locaton are exclusvely cross-sectonal n nature, explotng the very large dfferences n corporate tax rates around the world to dentfy the effects of taxes on FDI. Grubert and Mutt (1991) and Hnes and Rce (1994) estmate the effect of natonal tax rates on the cross-sectonal dstrbuton of aggregate Amercan-owned property, plant and equpment (PPE) n 1982. PPE dffers from FDI n that PPE represents (the book value of) real productve assets held by Amercan-owned afflates, whle FDI equals the book value of ownershp clams of controllng foregn nvestors. 12 Grubert and Mutt analyze the dstrbuton of PPE n manufacturng afflates n 33 countres, reportng a 0.1 elastcty wth respect to local tax rates. That s, controllng for other observable determnants of FDI, ten percent dfferences n local tax rates are assocated wth one percent dfferences n amounts of local PPE ownershp n 1982. Hnes and Rce consder the dstrbuton of PPE n all afflates n 73 countres, reportng a much larger 1.0 elastcty of PPE ownershp wth respect to tax rates. Altshuler et al. (2001) compare 12 The dstncton between FDI and PPE ownershp of foregn afflates s perhaps best llustrated by an example. Consder two Amercan-controlled foregn afflates, each wth $100 mllon of assets entrely nvested n PPE. One afflate s 100 percent owned by ts Amercan parent, whle the other s 60 percent owned by the parent company and 40 percent owned by nvestors n ts host country. Both afflates account for $100 mllon of PPE. Establshng the frst afflate wth $100 mllon of debt and equty from the parent company represents $100 mllon of outbound FDI from the Unted States, whle establshng the second wth parent funds represents $60 mllon of FDI. If half of the afflate fnancng represented funds borrowed from local banks, then establshng the afflates would represent $50 mllon and $30 mllon of FDI respectvely. To the degree that the afflates assets were not entrely nvested n PPE, then the PPE fgures could change wthout any correspondng change n FDI. 10
the tax senstvty of PPE ownershp n 58 countres n 1984 to that n 1992, reportng estmated tax elastctes that ncrease from 1.5 n 1984 to 2.8 n 1992. One of the dffcultes facng all cross-sectonal studes of FDI locaton s the nevtable omsson of many mportant determnants of FDI that may be correlated wth tax rates and therefore bas the estmaton of tax elastctes. Hnes (1996) ncorporates state fxed effects n comparng the dstrbutons of FDI wthn the Unted States of nvestors whose home governments grant foregn tax credts for federal and state ncome taxes wth those whose home governments do not tax ncome earned n the Unted States. The ncluson of fxed effects mplctly controls for hard-to-measure state attrbutes, as long as the effect of these attrbutes does not vary systematcally between nvestors from countres wth dfferng home-country tax regmes. Tax effects are dentfed by comparng, for example, the extent to whch nvestments from Germany (whch exempts from tax foregn-source ncome earned n the Unted States) tend to be located n lower-tax states than are nvestments from the Unted Kngdom (whch provdes foregn tax credts for state ncome taxes pad). The evdence ndcates that one percent state tax rate dfferences n 1987 are assocated wth ten percent dfferences n amounts of manufacturng PPE owned by nvestors from countres wth dfferng home-country taxaton of foregn-source ncome, and three percent dfferences n numbers of afflates owned. Taken as a structural relatonshp, the estmates mply a tax elastcty of nvestment equal to 0.6. It s worth bearng n mnd, however, that ths estmate reflects the effect of taxaton on the dentty of ownershp of captal as well as on the volume of nvestment. 13 2.2.2. Internatonal tax avodance. 13 Swenson (2001a) estmates separate regressons for dfferng types of transactons (such as the establshment of new plants, plant expansons, mergers and acqustons, and jont ventures) undertaken by foregn nvestors n the Unted States, fndng tax effects to dffer between transacton type. 11
One of the mportant ssues n consderng the mpact of taxaton on nternatonal nvestment patterns s the ablty of multnatonal frms to adjust the reported locaton of ther taxable profts. To the extent that FDI can facltate the advantageous relocaton of profts, then frms wll have ncentves to talor ther nternatonal nvestment strateges wth such relocaton n mnd. Of course ths s not a smple process, nor are ts mplcatons necessarly straghtforward. Frms that are able to use nternatonal transactons to avod the bulk of ther tax oblgatons are n turn unlkely to avod hgh-tax nvestment locatons. Hence any complete analyss of the mpact of taxaton on the operatons of multnatonal frms must necessarly consder the ablty and evdent wllngness of multnatonal frms to undertake actvtes to avod nternatonal tax oblgatons. The fnancng of foregn afflates presents straghtforward opportuntes for nternatonal tax avodance. If an Amercan parent company fnances ts nvestment n a foregn subsdary wth equty funds, then ts foregn profts are taxable n the host country and no taxes are owed the U.S. government untl the profts are repatrated to the Unted States. The alternatve of fnancng the foregn subsdary wth debt from the parent company generates nterest deductons for the subsdary that reduce ts taxable ncome, and generates taxable nterest recepts for the parent company. Smple tax consderatons therefore often make t attractve to use debt to fnance foregn afflates n hgh-tax countres and to use equty to fnance afflates n low-tax countres. 14 The evdence s broadly consstent wth these ncentves. Hnes and Hubbard (1990) fnd that the average foregn tax rate pad by subsdares remttng nonzero nterest to ther Amercan parent frms n 1984 exceeds the average foregn tax rate pad by subsdares wth no nterest payments, whle the reverse pattern holds for dvdend payments. Grubert (1998) estmates separate 12
equatons for dvdend, nterest, and royalty payments by 3,467 foregn subsdares to ther parent Amercan companes (and other members of controlled groups) n 1990, fndng that hgh corporate tax rates n countres n whch Amercan subsdares are located are correlated wth hgher nterest payments and lower dvdend payout rates. Contractual arrangements between related partes located n countres wth dfferent tax rates offer numerous possbltes for sophstcated (and unsophstcated) tax avodance. It s wdely suspected that frms adjust transfer prces used n wthn-frm transactons wth the goal of reducng ther total tax oblgatons. Multnatonal frms typcally can beneft by reducng prces charged by afflates n hgh-tax countres for tems and servces provded to afflates n low-tax countres. OECD governments requre frms to use transfer prces that would be pad by unrelated partes, but enforcement s dffcult, partcularly when prcng ssues concern unque tems such as patent rghts. Gven the looseness of the resultng legal restrctons, t s entrely possble for frms to adjust transfer prces n a tax-senstve fashon wthout even volatng any laws. The evdence of tax-motvated transfer prcng comes n several forms. Grubert and Mutt (1991) and Hnes and Rce (1994) analyze the aggregate reported proftabltes of U.S afflates n dfferent foregn locatons n 1982. Grubert and Mutt examne proft/equty and proft/sales ratos of U.S.-owned manufacturng afflates n 29 countres, whle Hnes and Rce regress the proftablty of all U.S.-owned afflates n 59 countres aganst captal and labor nputs and local productvtes. Grubert and Mutt report that hgh taxes reduce the reported after-tax proftablty of local operatons; Hnes and Rce fnd consderably larger effects (one percent tax rate dfferences are assocated wth 2.3 percent dfferences n before-tax proftablty) n ther data. Whle t s possble that hgh tax rates are correlated wth other locatonal attrbutes that depress the 14 Hnes (1994) dentfes exceptons to ths rule that stem from the benefts of lmtng equty fnance n afflates located n countres wth very low tax rates n antcpaton of renvestng all of ther after-tax profts over long perods. 13
proftablty of foregn nvestment, compettve condtons typcally mply that after-tax rates of return should be equal n the absence of tax-motvated ncome-shftng. The fact that before-tax proftablty s negatvely correlated wth local tax rates s strongly suggestve of actve tax avodance. Smlarly, the reported low proft rates of foregn-owned frms n the Unted States over the last 20 years s a source of concern to observers who suspect foregn nvestors of transferrng profts earned n the Unted States to low-tax jursdctons offshore. 15 Patterns of reported proftablty are consstent wth other ndcators of aggressve taxavodance behavor, such as the use of royaltes to remt profts from abroad and to generate tax deductons n host countres. Hnes (1995) fnds that royalty payments from foregn afflates of Amercan companes n 1989 exhbt a 0.4 elastcty wth respect to the tax cost of payng royaltes, and Grubert (1998) also reports sgnfcant effects of tax rates on royalty payments by Amercan afflates n 1990. Clausng (2001) fnds that reported trade patterns between Amercan parent companes and ther foregn afflates, and those between foregn afflates located n dfferent countres, are consstent wth transfer-prcng ncentves. Controllng for varous afflate characterstcs, ncludng ther trade balances wth unafflated foregners, Clausng fnds that ten percent hgher local tax rates are assocated wth 4.4 percent hgher parent company trade surpluses wth ther local afflates. Ths pattern s suggestve of prcng practces that move taxable profts out of hgh-tax jursdctons. 16 Ths lterature has developed strong evdence that multnatonal frms are hghly responsve to nternatonal ncome tax rate dfferences, undertakng nvestments n low-tax locatons and usng 15 Grubert et al. (1993) use frm-level tax return data to compare the tax labltes of foregn-owned frms n the Unted States wth the tax labltes of otherwse-smlar Amercan-owned frms n 1987. They report that approxmately 50 percent of the dfference n the reported U.S. tax oblgatons of foregn and domestc frms s explanable on the bass of observable characterstcs such as frm szes and ages. The other 50 percent may reflect the use of aggressve transfer prcng by those foregn nvestors wth stronger ncentves than Amercan frms to shft taxable ncome out of the Unted States, though t may also smply capture the effect of mportant omtted varables. 14
varous methods at ther dsposal to avod tax oblgatons on ther exstng nvestments. One noteworthy feature of ths evdence s ts almost exclusve focus on dfferences n corporate ncome tax rates. Whether and to what extent taxes other than corporate proft taxes nfluence the actvtes of multnatonal frms represent, by comparson, almost entrely open questons. The relatve mportance of these alternatve questons can be llumnated by the relatve magntudes of foregn ncome taxes and non-ncome taxes pad by U.S. frms that operate abroad. Fgure 1 provdes the rato of non-ncome taxes to foregn ncome taxes pad by Amercan multnatonal frms from 1982 to 1994. For overall nvestment as well as wthn manufacturng, non-ncome taxes are large relatve to ncome taxes and ncreasng n mportance over the last two decades. The relatve mportance of non-ncome taxes n tax competton dynamcs s also hghlghted n Slemrod (1995) and documented n Desa (1999). Snce non-ncome taxes are typcally non-credtable, the relatve ncentve to use non-credtable versus credtable taxes can be a functon of the tax-credt status of multnatonal frms n large captal-exportng countres such as the Unted States. 3. Behavoral mplcatons of multple taxes. Amercan multnatonal frms are typcally subject to several dfferent types of taxaton n foregn countres; n addton, they must also pay taxes to the U.S. government on any proft repatratons. In order to dentfy the behavoral mplcatons of these multple taxes, t s useful to consder the ncentves facng a frm for whch after-tax profts (Œ ) n country are gven by: (1) π [( t ) p Q ( K, L ) t K w L ]( 1 τ ) 1 1 2. 16 Swenson (2001b) fnds a smlar pattern n the reported prces of goods mported nto the Unted States, n whch 15
In expresson (1), the term t 1 s country s tax rate on gross output (such as a sales tax or value added tax), t 2 s country s property tax rate on ndustral captal, and 2 s the combned host country and home country proft tax. The frm produces output n country wth producton functon Q (K, L ), n whch K s the frm s captal n country, and L ts labor nput. The producton functon s assumed to be strctly concave and twce contnuously dfferentable. As a gesture toward smplfyng the analyss, captal s assumed not to deprecate. The frm sells ts output n country at an average prce of p, and hres labor at a wage of w ; snce (by assumpton) captal does not deprecate, the frm s not enttled to clam deprecaton allowances for captal nvestment. Frms are assumed to fnance ther foregn nvestment wth parent equty rather than local or related party debt. 3.1. Income shftng. As the lterature on proft-shftng and transfer-prcng ndcates, multnatonal frms have access to varous methods of reallocatng tax oblgatons between jursdctons. Some of these methods ental the use of tax-motvated fnancal transactons between related partes, whle others may consst of selectng the locaton of hgh-value-added foregn producton actvtes. Expresson (1) ndcates what a frm s after-tax ncome would be f t were not to aval tself of any of these opportuntes. If, nstead, the frm were to attempt to reallocate profts between jursdctons, then ts after-tax profts mght become: (2) π ( 1 t ) pq ( K, L ) + ( 1 at ) ψ c( ψ, pq ) t K wl ( 1 τ ) 1 1 2, hgh unt tarff rates appear to be assocated wth unusually low prces. 16
n whch ~ π s reported profts, ψ s the volume of addtonal ncome (postve or negatve) allocated to jursdcton, and c (, p Q ) ψ s the cost of engagng n such ncome-reallocaton. The term ( 1 at 1 ) that multples ψ n (2) reflects that a fracton a of ncome shfted nto jursdcton s subject to gross output taxaton. Hence, to the extent that destnaton-based gross output taxaton operates seamlessly, and proft-shftng takes the form of over- or under-nvocng transactons between related partes located n dfferent countres, then a mght be very small or even zero. To the extent that countres mpose gross output-based taxes, or fal to adhere perfectly to the destnaton bass of destnaton-based taxes (as when, for example, taxpayers receve credts and refunds only after sgnfcant delays), then a wll take a value somewhere between zero and unty. The cost of ncome reallocaton s captured by the term c ( ψ, p Q ). It s commonly assumed (e.g., Hnes and Rce, 1994; Grubert and Slemrod, 1998) that ths functon s ncreasng (and convex) n the absolute magntude of ψ and decreasng n the absolute magntude of p Q. Ths assumpton corresponds to the noton that the cost of engagng n a gven volume of ncome shftng s smaller for frms wth vast amounts of ncome. A convenent specfcaton of the c() functon s: (3) c(, p Q ) ( ψ ) 2 f ψ =. 2 pq The specfcaton (2) of the frm s adjusted proft functon mposes that the costs captured by the c () functon are tax-deductble, whch s sensble to the degree that they nclude top management tme and energy, lawyer fees, forgone output, and other expenses that frms deduct from proft taxes. 17
The ablty to reallocate taxable ncome between jursdctons carres mplcatons not only for the pattern of reported ncome but also for factor demands n dfferent jursdctons. Snce the optmal choces of K, L, and ψ can be evaluated n any order, consder the optmal selecton of ψ, takng K and L as gven. Snce proft-shftng cannot generate aggregate net revenue, t follows that the frm s optmzaton problem s to choose constrant that: ψ to maxmze (2), subject to (3) and the (4) ψ 0. n = 1 The frst-order condton for ths maxmzaton problem s: fψ 1 at1 1 =, pq (5) ( τ ) λ n whch λ > 0 s the lagrange multpler correspondng to the constrant (4). Equaton (5) then mples: (6) ψ = pq ( 1 at1 )( 1 τ ) f ( 1 τ ) λ p Qν, whch n turn carres the mplcaton that frm s pretax ncome, denoted µ, s gven by [ p Q + ψ c( ψ, p Q ) w L ], or: 18
(7a) 2 2 2 1 a t at 1 λ 1 λ µ = pq 1 + + w L pq 1 θ 2 2 f f ( 1 τ ) 2 f ( 1 τ ) [ + ] w L. Takng labor nput to be roughly (for the purpose of ths calculaton only) a constant share ( α ) of gross output, as t s wth a Cobb-Douglas producton functon, (7a) becomes: (7b) µ Q [ + θ α] p 1. = In order to evaluate the mpact of tax rates on ncome allocaton, t s convenent to take logs of both sdes of (7b), whch yelds: (8) ( ) ln( ) θ α ln µ p Q +. The frst term on the rght sde of (8) can be expanded to be a functon of productve nputs (captal and labor compensaton) as well as economy-wde productvty ndcators such as GDP. The second term on the rght sde of (8) reflects the mpact of ncome reallocaton, and s a functon of local drect and ndrect tax rates, as well as the cost of ncome reallocaton. 3.2. Factor demands. Factor demands can be convenently analyzed by combnng (2), (6), and (7a), so that the frm s after-tax profts n jursdcton can be rewrtten as: 19
(9) ~ π [ p Q ( K, L )[ 1+ θ t ( 1+ aν )] t K w L ]( 1 τ ) 1 2. Frms choose nputs K and L to maxmze frm value (V), whch reflects the dfference between profts and the carryng cost of captal; ths dfference s gven by: n n (10) V ~ π ϕ K, = 1 = 1 n whch ϕ > 0 s the opportunty cost of employng a unt of captal abroad. The frst-order condtons characterzng factor nputs that maxmze V are: (11) ( K, L ) Q L = p [ 1+ θ t ( 1+ aν )] w 1 and (12) + t 2 (, ) ( 1 τ ) = K p 1+ θ t ( 1+ aν ) Q K L ϕ 1 It follows from equaton (11) and the usual propertes of producton functons that hgher wages or hgher output tax rates rase requred margnal revenue products of labor and thereby depress labor demand. The proft tax rate does not appear explctly n the frst-order condton for labor demand as wrtten n (11). Dependng on the substtutablty of captal and labor, proft taxaton can 20
nfluence K and thereby affect the demand for labor by changng the margnal product on the left sde of (11). In addton, proft taxaton has the potental to nfluence p. Demand for FDI s captured by (12), whch ndcates that hgher proft tax rates, hgher output or property tax rates, and lower values of p all reduce captal demand. The terms contanng p appearng n the denomnators of the rght sde of (12) are noteworthy n ths respect, snce product prces are very plausbly functons of local proft and other tax rates. Consder, for example, the case n whch a multnatonal frm sells all of ts output n local markets, and local producton s depressed by hgh tax rates. Then p wll be an ncreasng functon of local tax rates, and the net effect of taxaton on labor and captal demand s of uncertan sgn. The ablty of Amercan frms to clam tax credts for foregn ncome tax payments nfluences the mplcatons of (11) and (12). In the case n whch an Amercan frm has defct foregn tax credts and does not beneft from deferral of home country taxes, ts effectve foregn proft tax lablty s gven by the U.S. tax rate, and 2 s the same for all foregn locatons. As a result, local proft tax rates should not nfluence factor demands except nsofar as they affect p, whch case hgher proft tax rates wll be assocated wth greater demand for labor and captal. In the populaton of all Amercan nvestors, some frms have excess foregn tax credts and others have defct foregn tax credts, so the net mplcaton of hgher proft tax rates s n prncple unclear. It s, however, clear that there was an mportant change after 1986, when the steep reducton n the U.S. statutory corporate tax rate suddenly greatly ncreased the number of Amercan companes wth excess foregn tax credts. 17 An mplcaton of ths change s that local tax rates should become more mportant to factor demands after 1986. One of the dffcultes of nterpretng estmated factor demand equatons s that of removng the effect of correlated omtted varables. In partcular, countres that are attractve 21
locatons for FDI may also choose to subject such nvestment to taxaton at hgh rates. One emprcal specfcaton that removes the most obvous form of such endogenety s a regresson of captal/labor ratos on tax rates. Ths specfcaton offers the advantage of dentfyng the mpact of taxaton on one aspect of captal demand (captal-labor substtutablty) n a way that s not a smple functon of the general attractveness of an nvestment locaton. The mplcatons of the model for captal-labor ratos are apparent from consderng the ratos of both sdes of equaton (11) and equaton (12): (13) ( 1 τ) ( 1 τ ) Q L w = Q K ϕ + t 2. Equaton (13) then serves as the bass of the factor demand estmaton descrbed n secton 4 and presented n Tables 4, 5, and 7. 3.3. Total ncome. Frms should demand fewer productve factors n hgh-tax jursdctons (all other consderatons equal), and thereby generate less ncome than they would f the same jursdctons nstead had low tax rates. The same frms also have ncentves to talor ther fnancal and nonfnancal practces to reduce stll further the ncome attrbuted to ther afflates located n hghtax jursdctons. Both of these consderatons therefore mply that the reported ncome of Amercan afflates, not controllng for levels of productve nputs, should be a declnng functon of local tax rates. 17 See the evdence reported by Grubert, Randolph, and Rousslang (1996). 22
From the standpont of host governments, the responsveness of taxable ncome to local tax rates carres wth t the mplcaton that there mght be drect and ndrect tax rates n the (0, 1) nterval at whch tax revenue from foregn nvestors s maxmzed. Whle governments mght reasonably not want to mpose such taxes snce governments have more than tax revenue n ther objectve functons, and, n partcular, mght strve to encourage economc actvty t s revealng to consder the levels of revenue-maxmzng tax rates. 18 In partcular, t s nterestng to evaluate whether revenue-maxmzng tax rates have fallen over tme n response to (possbly) greater tax competton, and whether small economes have the lowest revenue-maxmzng rates. Let B denote the tax base to whch an ndrect tax (denoted t) s appled, and let B(1-t) be the tax base to whch the drect tax (at rate τ ) s appled. Total tax revenue ( ρ ) s then: (14) ρ = Bt + B( 1 t)τ Takng dervatves of (14) wth respect to t and τ yelds: ρ t B t (15a) = B( 1 τ ) + [ t + τ ( 1 t) ] ρ τ B τ (15b) = B( 1 t) + [ t + τ( 1 t) ] Settng these dervatves equal to zero to solve for the revenue maxmzng tax rates (denoted t* and τ *) yelds the condtons: 18 The numerous other consderatons n selectng tax revenue-maxmzng rates nclude the frequent requrement that all foregn and domestc nvestors receve equal tax treatment, and the possblty that domestc resdents bear the ultmate ncdence of taxes mposed on foregn nvestors. 23
(16a) (16b) 1 τ t* = 1 B 1 B t ( τ ) 1 t τ* = 1 B 1 B τ ( t). Expressons (16a) and (16b), together wth emprcal estmaton of sem-elastctes of the tax base wth respect to tax rates, provde smple benchmarks aganst whch actual tax rates can be evaluated. 3.4. Data. 19 The Bureau of Economc Analyss (BEA) provdes data on the fnancal and operatng characterstcs of U.S. frms operatng abroad through perodc benchmark and annual surveys of the actvtes of U.S. frms operatng abroad. These surveys ask reporters to fle detaled fnancal and operatng tems for each afflate and nformaton on the value of transactons between U.S. parents and ther foregn afflates. The Internatonal Investment and Trade n Servces Survey Act governs the collecton of the data and the Act ensures that "use of an ndvdual company's data for tax, nvestgatve, or regulatory purposes s prohbted." Wllful noncomplance wth the Act can result n penaltes of to $10,000 or a prson term of one year. As a result of these assurances and penaltes, BEA beleves that coverage s close to complete and levels of accuracy are hgh. 19 Ths data descrpton draws on Desa and Hnes (1999) and Desa, Foley and Hnes (2001). 24
The three most recent benchmark surveys conducted n 1982, 1989, and 1994 (BEA (1985, 1992, 1997)) report data on country and ndustry bases ncludng detals on ncome statements, balance sheets, employment patterns and parent afflate transactons. BEA reports aggregate fgures of countres n whch there s substantal U.S. nvestment; to protect the confdentalty of survey respondents, BEA suppresses nformaton for countres n whch one or two frms represent large fractons of total U.S. nvestment. U.S. drect nvestment abroad s defned as the drect or ndrect ownershp or control by a sngle U.S. legal entty of at least ten percent of the votng securtes of an ncorporated foregn busness enterprse or the equvalent nterest n an unncorporated foregn busness enterprse. The BEA reports aggregate tax payments by country for both foregn ncome taxes as well as ndrect busness taxes. Snce U.S. frms pay corporate taxes to foregn host countres but on occason receve specal treatment n the form of tax holdays and other local tax concessons, t s necessary to calculate ncome tax rates specfcally avalable to Amercan nvestors. Followng Hnes and Rce (1994) and Desa and Hnes (1999), the ncome tax rates used equal the smaller of the statutory corporate tax rate and the average tax rate pad by Amercan frms n a gven year. The average tax rate s calculated as the rato of foregn ncome taxes pad by local afflates of Amercan frms to ther local pre-tax ncome. Amercan frms are also specfcally asked to report taxes other than ncome and payroll taxes. In the surveys dstrbuted by the BEA, these other taxes are defnes as sales, value-added, consumpton, excse, property, mport and export dutes, lcense fees, fnes and all taxes other than ncome and payroll taxes. In the followng analyss, the corporate ncome tax rate descrbed above s referred to as the drect tax rate and the rato of ndrect busness taxes to the sum of local pre-tax ncome and ndrect busness taxes s referred to as the ndrect tax rate. In order to consder seperately the 25
potental effects of output taxes, value-added taxes obtaned from the Unversty of Mchgan World Tax Database are also employed. Table 1 provdes descrptve statstcs for the relevant varables for 1982, 1989 and 1994. 4. Results and nterpretaton. Ths secton presents estmated regresson coeffcents for equatons explanng the aggregate proftablty of Amercan afflates condtonal on factor nputs, aggregate factor demands, and total taxable ncomes. Snce the results ndcate the mportance of taxes other than corporate ncome taxes, t s useful to start by consderng the extent to whch taxes of dfferent knds are correlated. 4.1. Tax rates. Tax rates facng Amercan multnatonal frms exhbt strong postve cross-sectonal correlatons, meanng that countres wth hgh corporate tax rates are also lkely to have hgh valueadded tax rates, hgh ndvdual tax rates, and hgh rates of other taxes. Ths correlaton may reflect dfferng natonal revenue needs, or perhaps the workng of tax competton n whch some countres are more subject to compettve pressures (on all ther tax rates) than are others. The hgh degree of correlaton also means that t can be dffcult to dentfy the behavoral mpact of one tax aganst another. Table 2 presents tax rate correlatons for 1994, the last year for whch t s possble to obtan detaled nformaton from a benchmark survey. The drect tax rate n 1994 s defned as the rato of ncome taxes pad by all Amercan afflates to the sum of aggregate 1994 after-tax ncome of these afflates plus ncome taxes pad. Consequently, the drect tax rate can be 26
nterpreted as an average corporate tax rate faced by Amercan frms. Snce ths s potentally a nosy varable, ts value s restrcted to le n the (0, 1) nterval, and to equal the statutory corporate tax rate whenever the statutory rate s below the measured average corporate tax rate. The ndrect tax rate n 1994 s the rato of non-ncome and non-payroll taxes pad by all Amercan afflates to the sum of aggregate 1994 pre-tax ncome of these afflates plus ncome taxes pad and non-ncome taxes and non-payroll taxes. The statutory corporate tax rate s the top margnal corporate tax rate, and the statutory ndvdual tax rate s the top margnal ndvdual ncome tax rate. The wthholdng tax rate s the rato of wthholdng taxes on dstrbuted earnngs pad by all Amercan afflates n 1994 to total dstrbuted earnngs of those afflates. The value-added tax rate s the statutory value-added tax rate on typcal consumed goods. Table 2 ndcates that every one of these tax rates exhbts a postve correlaton wth all of the others. Notably, drect tax rates have a greater than 0.5 correlaton wth ndrect tax rates and wth statutory corporate tax rates, whle ndrect tax rates also have a greater than 0.5 correlaton wth value-added tax rates. The statutory ndvdual tax rate has a greater than 0.5 correlaton wth both the statutory corporate tax rate and the value-added tax rate. The hgh degree of correlaton between the statutory rates s not only suggestve of compettve tax-settng pressures, but also offers reassurance that the hgh correlaton of measured drect and ndrect tax rates reflect somethng other than statstcal artfacts. An mportant feature of ndrect tax rates s ther magntude compared to more commonlystuded drect tax rates. Table 1 reports that ndrect tax rates exceed drect tax rates for all three years of the sample, whch n turn mples a sgnfcant revenue dfference between these two tax types, snce the ndrect tax base typcally exceeds the drect tax base. Indeed, Fgure 1 plots the rato of total ndrect tax payments by U.S. majorty-owned foregn afflates to drect tax payments 27
by the same afflates for every year between 1982 and 1994. Ths rato exceeds 1.5 for every year of the sample, and n the 1990s (after a wave of drect tax reductons around the world) began to exceed three. 4.2. Proft locaton. Tables 3a, 3b, and 3c offer estmates of varants of the proft locaton equaton (8), n whch the log of pretax ncome s regressed aganst ts determnants: labor nputs, captal nputs, local GDP, and drect and ndrect tax rates. The data used to estmate these equatons consst of country-year observatons of the actvty of all U.S. majorty-owned afflates. These equatons are consstent wth smple Cobb-Douglas specfcatons of aggregate producton functons, n whch local GDP s ncluded as a productvty scalng factor. Table 3a reports results for 1982, Table 3b reports results for 1989, and Table 3c reports results for 1994. The frst three columns of each table report regressons n whch the dependent varable s the log of the sum of after-tax ncome and proft taxes; columns 4-7 of the same tables report regressons n whch the dependent varable s the sum of after-tax ncome, proft taxes, and non-ncome and non-payroll taxes. Hgher tax rates are assocated wth lower pretax ncome n almost every specfcaton. The results reported n frst three columns of Tables 3a 3c are consstent wth the specfcatons and results reported by Hnes and Rce (1994), who estmate smlar equatons for 1982, and who fnd that hgher drect tax rates are assocated wth reduced pre-tax ncome. Snce drect and ndrect tax rates are postvely correlated, and hgher ndrect tax rates reduce pre-tax ncome as defned n columns 1-3 of Tables 3a 3c, t follows that hgh drect tax rates mght be correlated wth low pretax ncome smply through the effect of ndrect tax rates. It s n part to guard aganst ths possblty that the regressons reported n columns 2 and 3 are re-run wth a dependent 28
varable that adds back ndrect tax payments; these regressons are reported n columns 4 and 5 of Tables 3a 3c. Whle the estmated drect tax rate coeffcents n the regressons reported n columns 4 and 5 ndeed turn out to be smaller than the correspondng tax coeffcents reported n columns 2 and 3, they reman large and sgnfcant. The mpact of ncludng ndrect tax rates n the pretax proft locaton estmaton s apparent from consderng the 1989 results reported n Table 3b. The specfcaton reported n column 8 of Table 3b ndcates that, controllng for levels of productve nputs, hgher ndrect tax rates have negatve and sgnfcant effects on reported pretax ncome: an ndrect tax rate dfference of 10 percent s assocated wth a 9.2 percent dfference n reported ncome. Whle drect tax rates are estmated to have larger effects n ths and other regressons, these estmates are often mprecse when the (hghly correlated) ndrect tax rates are ncluded n the same equatons. The estmated effect of a drect tax rate dfference of 10 percent n the regresson reported n column 7 s an 18.2 percent dfference n reported ncome, but ths estmate has a large standard error. The avalable nformaton s suffcent to use changes between 1982 and 1989, and agan between 1989 and 1994, to estmate the proft reportng regressons n frst dfferences. Dong so offers the beneft of controllng for tme-nvarant locaton-specfc factors that affect reported proftablty. Columns 1-3 of both panels of Table 7 report estmated coeffcents from such frst dfference specfcatons. The results from the 1982-1989 frst dfference are qute smlar to those from the 1989 cross-secton reported n Table 3a. The regresson reported n column 3 of the left panel of Table 7 ndcates that a ten percent ndrect tax rate change between 1982 and 1989 was assocated wth a 7.4 percent change n reported proftablty, controllng for productve nputs. The same regresson ndcates that a 10 percent drect tax rate change was assocated wth a 12.3 percent change n reported proftablty, though a large standard error s assocated wth ths 29
estmate. Columns 1-3 of the rght panel of Table 7 ndcate that ndrect tax changes between 1989 and 1994 had smaller effects on reported proftablty than they dd over the 1982-1989 tme perod, whle drect tax changes had much larger effects. Ten percent ndrect tax changes between 1989 and 1994 are assocated wth 3.7 percent changes n reported proftablty, whle ten percent drect tax changes are assocated wth 29.3 percent proftablty dfferences. 4.3. Factor demands. Taxaton of dfferent knds has the potental to dscourage FDI and to dstort the mx of factors that frms demand. Equaton (13) llustrates that captal/labor ratos are postvely affected by wage rates, and negatvely affected by ncome taxes and property taxes. One of the dffcultes wth mplementng (13) emprcally s that labor does not come n dentfable homogenous unts; as a result, employee compensaton rather than employment s used n the denomnator of the captal/labor rato n an effort to adjust for human captal. The results of estmatng varants of (13) are presented n Table 4. The estmates reported n Table 4 ndcate that captal/labor ratos are decreasng functons of wage rates, whch s sensble when t s recalled that labor s measured as total employee compensaton. 20 The estmated effects of drect tax rates on captal/labor ratos n 1982 are small and not statstcally sgnfcant, whle the effects of ndrect tax rates are consderably larger and are sgnfcant n the quadratc specfcaton reported n column 4. A smlar pattern appears n 1989, wth the dfference that estmated coeffcents on drect tax rates are not unformly negatve. In the regresson reported n column 3 of the 1989 results n Table 4, a ten percent hgher ndrect tax rate s assocated wth an 8.6 percent lower captal/labor rato, presumably reflectng the 20 Data on wage rates are avalable only for producton workers n manufacturng, whch are no doubt correlated, albet mperfectly, wth wage rates pad by all majorty-owned Amercan afflates. 30
property tax and related components of ndrect taxaton. The same regresson reports a postve and nsgnfcant effect of hgher drect tax rates on captal/labor ratos. These results suggest that factor demands are nfluenced by ndrect tax rates, and ndeed, offer stronger evdence that ndrect tax rates affect factor demands than they do evdence that drect tax rates affect factor demands. Whle there s ample other evdence n the lterature of the mpact of drect taxaton on the demand for FDI, the ablty of ndrect taxaton to nfluence factor demands should not be mnmzed, partcularly snce Amercan nvestors are nelgble to clam foregn tax credts for ndrect taxes pad. As a check on the relablty of the factor demand results, and the nterpretaton of ndrect tax rates as reflectng at least n part the mpact of property and property-lke taxes, the equatons reported n Table 4 were re-run usng statutory value-added tax rates n place of ndrect tax rates. The results are reported n Table 5. Snce, as ndcated n equaton 13, value-added taxes should not nfluence captal/labor ratos, t follows that the results of these new specfcatons should look markedly dfferent from those reported n Table 4. And the results reported n Table 5 n fact do dffer greatly from the results of estmatng the same equatons usng ndrect tax rates n place of value-added taxes. Estmated coeffcents on value-added tax rates are postve (though nsgnfcant) n the lnear specfcaton (column 2) n each of the three cross-sectons, whle drect tax rates are estmated to have nsgnfcant negatve effects n 1982 and 1989. Due to data lmtatons there are fewer observatons n the regressons reported n Table 5 than there are n those reported n Table 4; nevertheless, the specfcatons are so otherwse smlar that ther dfferences offer evdence n favor of the prevous nterpretaton of the results reported n Table 4. 4.4. Total ncome. 31
Total ncome earned by Amercan afflates s a functon of factor demands and proft allocaton behavor. Table 6 offers estmates of the determnants of total pre-tax ncome n whch only log(gdp) and tax rates are used as regressors (n varous combnatons), snce the goal s to endogenze factor demands and thereby reveal the choces made by governments n selectng ther tax rates. The results reported n Table 6 ndcate that hgh drect tax rates dscourage ncome producton n all three cross-sectons, and that, n 1982 and 1989, hgh ndrect tax rates lkewse dscourage ncome producton. The coeffcents reported n column 3 of the 1982 panel of Table 6 ndcate that ten percent hgher ndrect tax rates are assocated wth 8.4 percent lower ncome producton, whle ten percent hgher drect tax rates are assocated wth 13.3 percent lower ncome producton (thought the latter effect s estmated wth consderable mprecson). The estmated effect of ndrect taxaton s smlar n 1989, though the estmated mpact of drect taxaton s much larger: ten percent dfferences n drect taxaton are assocated wth 55.8 percent ncome dfferences. Nether drect nor ndrect tax rates sgnfcantly affect reported profts n 1994. Equatons 16a and 16b, the coeffcents n columns 1 and 2 of the 1989 panel, and the mean tax rates for 1989 can be combned to fnd the revenue-maxmzng drect and ndrect tax rates for 1989. These rates, -35 percent for drect tax rates and 33 percent for ndrect tax rates, suggest that ncome tax subsdes n combnaton wth hefty ndrect taxes would rase maxmal revenues from U.S. frms. Such a combnaton of ncome tax subsdes and ndrect taxes reflects the prevalence of excess credt frms amongst U.S. frms n 1989. These results can be mapped to a dynamc of drect and ndrect tax competton from 1982 through 1994. The coeffcents mply that the revenue-maxmzng drect tax rate fell consderably between 1982 and 1989 whle the revenue-maxmzng ndrect tax rate also fell, but by 32
consderably less. The larger relatve declne n the revenue-maxmzng drect tax rate s consstent wth the mportance of the dstncton between credtable and non-credtable taxes n nfluencng how U.S. multnatonals frms respond to taxaton. The rapd ncrease n the share of U.S. frms wth excess foregn tax credts would ncrease ther senstvty to credtable (ncome) taxes relatve to noncredtable (non-ncome) taxes. These changed senstvtes, and the correspondng revenue-maxmzng rates, change the ncentves for tax competton over dfferent tax nstruments. Column 4 n each of the panels of Table 6 report the results of specfcatons that nclude nteractons between tax rates and log(gdp), n order to check whether the responsveness of ncome to tax rates (and therefore also the revenue-maxmzng tax rates) vares wth economy sze. The results confrm that economc sze dampens the responsveness of pretax ncome to tax rates. The reduced responsveness to tax rates n larger countres lkely reflects the mmoblty of specfc factors (such as local markets or local resources) n large countres or the effcency wth whch tax revenue s deployed n those countres. The estmates n column 4 for 1989 can be combned wth equatons 16a and 16b to derve the revenue-maxmzng tax rates for partcular countres. For the country wth the medan GDP n the sample for 1989, Egypt, the coeffcents n column 4 for 1989 along wth Egypt s tax parameters mply a revenue-maxmzng drect tax rate of 24 percent and a revenue-maxmzng ndrect tax rate of 152 percent. 21 Table 7 reports the results of frst-dfference estmates of the determnants of the locaton of total ncome. Columns 4-6 of the left panel of Table 7 report estmated coeffcents from regressons n whch the dependent varable s the growth rate of afflate ncome between 1982 and 1989; columns 4-6 of the rght panel report estmates n whch the dependent varable s the 21 The 24% fgure, for example, was calculated as 1 Indrect Tax Rate 7.88 + 0.88 Log GDP 1-Indrect Tax Rate Egypt Egypt Egypt. 33
growth of afflate ncome between 1989 and 1994. The results are broadly consstent wth those reported n Table 6. In the 1989-1994 regresson reported n column 6, a ten percent change n the ndrect tax rate s assocated wth a 7.5 percent change n reported ncome, whle a ten percent change n the drect tax rate s assocated wth a 25.4 percent change n reported ncome. 5. Concluson. Ths paper offers evdence that taxes other than ncome taxes sgnfcantly nfluence the pattern of ncome producton by multnatonal frms by alterng ther nvestment and transfer prcng ncentves. The hgh degree of correlaton between ncome and non-ncome tax rates suggests that the body of emprcal work exclusvely emphaszng ncome taxes may have nadvertently obscured the role of non-ncome taxes. Snce Amercan taxpayers can clam tax credts for ncome taxes pad to foregn governments, but are unable to clam smlar tax credts for ndrect taxes pad to foregn governments, t follows that foregn ndrect taxes have much greater potental to nfluence ther behavor. Governments that are concerned about the potental for competng jursdctons to lure economc actvty and about ndrect tax recepts may already be competng on non-ncome taxes and are lkely to be ncreasngly aware of relatve ndrect tax burdens n ther own and other jursdctons. Whle a race to the bottom dynamc on ndrect taxaton has not been demonstrated, there s, however, consderable evdence to suggest that multnatonal frms respond to ndrect tax rate dfferences. Snce ndrect tax burdens greatly exceed drect tax burdens, there s ample scope for downward compettve dynamcs as governments respond to greater nternatonal moblty of productve factors. 34
References Altshuler, Rosanne, Harry Grubert, and T. Scott Newlon, Has U.S. nvestment abroad become more senstve to tax rates?, n: James R. Hnes Jr., ed. Internatonal taxaton and multnatonal actvty (Chcago: Unversty of Chcago Press, 2001), 9-32. Auerbach, Alan J. and Kevn Hassett, Taxaton and foregn drect nvestment n the Unted States: A reconsderaton of the evdence, n: Alberto Govannn, R. Glenn Hubbard, and Joel Slemrod, eds. Studes n nternatonal taxaton (Chcago: Unversty of Chcago Press, 1993), 119-144. Boskn, Mchael and Wllam G. Gale, New results on the effects of tax polcy on the nternatonal locaton of nvestment, n: Martn Feldsten, ed. The effects of taxaton on captal accumulaton (Chcago: Unversty of Chcago Press, 1987), 201-219. Clausng, Kmberly A., The mpact of transfer prcng on ntrafrm trade, n: James R. Hnes Jr., ed. Internatonal taxaton and multnatonal actvty (Chcago: Unversty of Chcago Press, 2001), 173-194. Desa, Mhr A., Are we racng to the bottom? Evdence on the dynamcs of nternatonal tax competton, n: Proceedngs of the 91st Annual Conference on Taxaton, Washngton, D.C.: Natonal Tax Assocaton, 1999, 176-187. Desa, Mhr A., C. Frtz Foley, and James R. Hnes Jr., Dvdend polcy wthn the frm, workng paper, Harvard Unversty, 2001. Desa, Mhr A. and James R. Hnes Jr., Basket cases: Tax ncentves and nternatonal jont venture partcpaton by Amercan multnatonal frms, Journal of Publc Economcs, March 1999, 71 (3), 379-402. Devereux, Mchael P. and Harold Freeman, The mpact of tax on foregn drect nvestment: Emprcal evdence and the mplcatons for tax ntegraton schemes, Internatonal Tax and Publc Fnance, May 1995, 2(1), 85-106. Grubert, Harry, Taxes and the dvson of foregn operatng ncome among royaltes, nterest, dvdends and retaned earnngs, Journal of Publc Economcs, May 1998, 68(2), 269-290. Grubert, Harry, Tmothy Goodspeed, and Deborah Swenson, Explanng the low taxable ncome of foregn-controlled companes n the Unted States, n: Alberto Govannn, R. Glenn Hubbard, and Joel Slemrod, eds. Studes n nternatonal taxaton (Chcago: Unversty of Chcago Press, 1993), 237-270. Grubert, Harry and John Mutt, Taxes, tarffs and transfer prcng n multnatonal corporate decson makng, Revew of Economcs and Statstcs, May 1991, 73(2), 285-293.
Grubert, Harry, Wllam C. Randolph, and Donald J. Rousslang, Country and multnatonal company responses to the Tax Reform Act of 1986, Natonal Tax Journal, September 1996, 49 (3), 341-358. Hartman, Davd G., Tax polcy and foregn drect nvestment n the Unted States, Natonal Tax Journal, December 1984, 37(4), 475-487. Hnes, James R., Jr., The flght paths of mgratory corporatons, Journal of Accountng, Audtng, and Fnance, Fall 1991, 6(4), 447-479. Hnes, James R., Jr., Credt and deferral as nternatonal nvestment ncentves, Journal of Publc Economcs, October 1994, 55(2), 323-347. Hnes, James R., Jr., Taxes, technology transfer, and the R&D actvtes of multnatonal frms, n: Martn Feldsten, James R. Hnes Jr., and R. Glenn Hubbard, eds. The effects of taxaton on multnatonal corporatons (Chcago: Unversty of Chcago Press, 1995), 225-248. Hnes, James R., Jr., Altered states: Taxes and the locaton of foregn drect nvestment n Amerca, Amercan Economc Revew, December 1996, 86(5), 1076-1094. Hnes, James R., Jr., Tax polcy and the actvtes of multnatonal corporatons, n: Alan J. Auerbach, ed. Fscal polcy: Lessons from economc research (Cambrdge, MA: MIT Press, 1997), 401-445. Hnes, James R., Jr., Lessons from behavoral responses to nternatonal taxaton, Natonal Tax Journal, June 1999, 52 (2), 305-322. Hnes, James R., Jr., Tax sparng and drect nvestment n developng countres, n: James R. Hnes Jr., ed. Internatonal taxaton and multnatonal actvty (Chcago: Unversty of Chcago Press, 2001), 39-66. Hnes, James R., Jr. and R. Glenn Hubbard, Comng home to Amerca: Dvdend repatratons by U.S. multatonals, n: Assaf Razn and Joel Slemrod, eds. Taxaton n the global economy (Chcago: Unversty of Chcago Press, 1990), 161-200. Hnes, James R., Jr. and R. Glenn Hubbard, Appendx, n: Martn Feldsten, James R. Hnes Jr., and R. Glenn Hubbard, eds. Taxng multnatonal corporatons (Chcago: Unversty of Chcago Press, 1995), 103-106. Hnes, James R., Jr. and Erc M. Rce, Fscal paradse: Foregn tax havens and Amercan busness, Quarterly Journal of Economcs, February 1994, 109(1), 149-182. Newlon, Tmothy Scott, Tax polcy and the multnatonal frm's fnancal polcy and nvestment decsons, Ph.D. dssertaton, Prnceton Unversty.
Slemrod, Joel, Tax effects on foregn drect nvestment n the Unted States: Evdence from a cross-country comparson, n: Assaf Razn and Joel Slemrod, eds. Taxaton n the global economy (Chcago: Unversty of Chcago Press, 1990), 79-117. Slemrod, Joel, Tax polcy toward foregn drect nvestment n developng countres n lght of recent nternatonal tax changes, n: Anwar Shah, ed. Fscal ncentves for nvestment and nnovaton (New York: Oxford Unversty Press, 1995), 289-307. Swenson, Deborah L., The mpact of U.S. tax reform on foregn drect nvestment n the Unted States, Journal of Publc Economcs, June 1994, 54(2), 243-266. Swenson, Deborah L., Transacton type and the effect of taxes on the dstrbuton of foregn drect nvestment n the U.S., n: James R. Hnes Jr., ed. Internatonal taxaton and multnatonal actvty (Chcago: Unversty of Chcago Press, 2001a), 89-108. Swenson, Deborah L., Tax reforms and evdence of transfer prcng, Natonal Tax Journal, March 2001b, 54 (1), 7-25. Young, Kan H., The effects of taxes and rates of return on foregn drect nvestment n the Unted States, Natonal Tax Journal, March 1988, 41(1), 109-121.
Fgure 1: The Rato of Non-Income and Non-Payroll Taxes to Foregn Income Taxes, 1982-1994 4 3.5 3 2.5 2 1.5 1 0.5 0 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 Overall U.S. Investment Abroad Manufacturng U.S. Investment Abroad Note: The fgure provdes the rato of non-ncome and non-payroll taxes to foregn ncome taxes from 1982 to 1994 for all U.S. multnatonals and for the manufacturng sector.
Table 1 Descrptve Statstcs for Sample Mean Medan Std. Dev. No. Obs. Log of Pre-Drect Tax Income, 1982 6.0976 6.3570 1.4922 46 Log of Pre-All Tax Income, 1982 6.6426 6.5985 1.6406 44 Log of Net PP&E, 1982 6.7992 6.7397 1.5052 54 Log of Employee Compensaton, 1982 6.0254 5.7120 1.6446 54 Log of Average Wage Rate, 1982 1.5495 1.7047 0.7466 50 Log of GDP, 1982 3.9223 4.2380 1.7304 60 Log of Rato of Net PP&E 0.7738 0.7184 0.7440 54 to Employee Compensaton, 1982 Drect Tax Rate, 1982 0.3507 0.4000 0.1511 59 Indrect Tax Rate, 1982 0.5057 0.5388 0.3541 44 VAT Tax Rate, 1982 0.1015 0.1000 0.0830 29 Log of Pre-Drect Tax Income, 1989 6.4598 6.3279 1.6459 53 Log of Pre-All Tax Income, 1989 7.1259 7.0255 1.5109 49 Log of Net PP&E, 1989 6.8590 6.5861 1.6789 58 Log of Employee Compensaton, 1989 6.0195 5.8051 1.8842 54 Log of Average Wage Rate, 1989 1.6653 1.5271 0.9094 58 Log of GDP, 1989 4.4077 4.6579 1.7359 59 Log of Rato of Net PP&E 0.8395 0.6756 0.9073 58 to Employee Compensaton, 1989 Drect Tax Rate, 1989 0.2794 0.3000 0.1486 60 Indrect Tax Rate, 1989 0.3444 0.3370 0.2943 50 VAT Tax Rate, 1989 0.1022 0.1000 0.0775 41 Log of Pre-Drect Tax Income, 1994 6.5866 6.4427 1.5309 58 Log of Pre-All Tax Income, 1994 7.2158 7.0510 1.4992 55 Log of Net PP&E, 1994 7.4023 7.1910 1.6327 58 Log of Employee Compensaton, 1994 6.5019 6.3918 1.8137 58 Log of Average Wage Rate, 1994 1.9712 2.0082 0.9189 57 Log of GDP, 1994 4.7812 4.9954 1.7526 59 Log of Rato of Net PP&E 0.9004 0.7775 0.7966 58 to Employee Compensaton, 1994 Drect Tax Rate, 1994 0.2431 0.2672 0.1480 59 Indrect Tax Rate, 1994 0.3709 0.3515 0.2593 55 VAT Tax Rate, 1994 0.1217 0.1400 0.0727 48 Note: For each year, "Log of Pre-Drect Tax Income" s the logarthm of the sum of net ncome and foregn ncome taxes; "Log Pre-All Tax Income" s the logarthm of the sum of net ncome, foregn ncome taxes, and non-ncome and non-payroll taxes; "Log of Net PP&E" s the logarthm of net property, plant and equpment; "Log of Employee Compensaton" s the logarthm of employee compensaton; "Log of Average Wage Rate" s the logarthm of the average wage for a producton worker n the manufacturng sector; "Log of GDP" s the logarthm of gross domestc product as reported n the Penn World Tables (1982 and 1989) and by the EIU (1994); "Log of Rato of Net PP&E to Employee Compensaton" s the logarthm of the rato of net property, plant and equpment to employee compensaton; "Drect Tax Rate" s the rato of foregn ncome taxes to the sum of net ncome and foregn ncome taxes; "Indrect Tax Rate" s the rato of non-ncome and non-payroll taxes to the sum of net ncome, foregn ncome taxes, and non-ncome and non-payroll taxes; "VAT Tax Rate" s the value-added tax rate as reported n the Unversty of Mchgan World Tax Database.
Table 2 Correlaton Matrx for Country Tax Rates, 1994 Drect Tax Rate Indrect Tax Rate Statutory Corporate Tax Rate Statutory Indvdual Tax Rate Wthholdng Tax Rate Indrect Tax Rate Statutory Corporate Tax Rate Statutory Indvdual Tax Rate Wthholdng Tax Rate Value-Added Tax Rate 0.5365 1.0000 0.0002 na 55 55 0.5387 0.1578 1.0000 0.0000 0.2544 na 57 54 57 0.1668 0.3483 0.5684 1.0000 0.2106 0.0099 0.0000 na 58 54 57 57 0.3066 0.2323 0.3167 0.2117 1.0000 0.0341 0.1161 0.0301 0.1468 na 48 47 47 48 48 0.2536 0.5152 0.1513 0.5009 0.1343 0.0820 0.0003 0.3047 0.0003 0.4027 48 45 48 48 41 Note: For each par, the frst cell s the correlaton coeffcent, the second cell s the level of sgnfcance for that correlaton coeffcent, and the thrd cell s the number of pars. "Drect Tax Rate" s the rato of foregn ncome taxes to the sum of net ncome and foregn ncome taxes. "Indrect Tax Rate" s the rato of non-ncome and non-payroll taxes to the sum of net ncome, foregn ncome taxes, and non-ncome and non-payroll taxes. "Statutory Corporate Tax Rate" s the top margnal corporate rate as reported n the Unversty of Mchgan World Tax Database. "Statutory Indvdual Tax Rate" s the top margnal ndvdual rate as reported n the Unversty of Mchgan World Tax Database. "Wthholdng Tax Rate" s the rato of wthholdng tax payments to dstrbuted drect earnngs. "VAT Tax Rate" s the value-added tax rate as reported n the Unversty of Mchgan World Tax Database.
Table 3a The Locaton of Pretax Profts and Drect and Indrect Tax Rates, 1982 Dependent Varable: Log of Pre-Drect Tax or Pre-All Tax Income Log of Pre-Drect Tax Income Log of Pre-All Tax Income (1) (2) (3) (4) (5) (6) (7) (8) (9) Constant 1.3170 1.6871 2.2637 1.5298 1.9579 1.6305 1.5778 1.6971 2.0994 (0.8815) (0.8454) (0.7124) (0.9080) (0.9685) (0.9756) (1.0532) (0.9582) (1.0159) Log of Employee 0.0822 0.0185 0.0272 0.1385 0.1196 0.2893 0.2534 0.2325 0.1351 Compensaton (0.2878) (0.2882) (0.3085) (0.1716) (0.1738) (0.1651) (0.1832) (0.1630) (0.1765) Log of Net PP&E 0.7768 0.7949 0.8189 0.6664 0.6783 0.5182 0.5485 0.5687 0.6451 (0.2621) (0.2556) (0.2797) (0.1805) (0.1773) (0.2039) (0.2324) (0.1940) (0.1958) Log GDP -0.2571-0.0543-0.0200 0.0925 0.1230 0.0264 0.0209 0.0898 0.1145 (0.1433) (0.1635) (0.1488) (0.1290) (0.1376) (0.1143) (0.1202) (0.1310) (0.1468) Drect Tax Rate -2.8568-12.0265-1.9252-6.8273-1.2778-7.0749 (1.1482) (4.7173) (0.9691) (4.7623) (0.9121) (5.0286) Drect Tax Rate 15.8832 8.3664 9.9265 Squared (7.1792) (7.3871) (7.6983) Indrect Tax Rate -0.7592-0.3921-0.5515 0.2373 (0.3859) (1.1449) (0.4049) (1.0563) Indrect Tax Rate -0.2598-0.5814 Squared (0.6761) (0.6356) R-Squared 0.5794 0.6270 0.6861 0.7546 0.7692 0.7576 0.7585 0.7664 0.7867 No. of Obs. 46 46 46 44 44 44 44 44 44 Note: In columns 1, 2, and 3, the dependent varable s the logarthm of the sum of net ncome and foregn ncome taxes. In columns 4 through 9, the dependent varable s the logarthm of the sum of net ncome, foregn ncome taxes and non-ncome and non-payroll taxes. "Log of Net PP&E" s the logarthm of net property, plant and equpment. "Log of Employee Compensaton" s the logarthm of employee compensaton. "Log of GDP" s the logarthm of gross domestc product as reported n the Penn World Tables. "Drect Tax Rate" s the rato of foregn ncome taxes to the sum of net ncome and foregn ncome taxes. "Indrect Tax Rate" s the rato of non-ncome and non-payroll taxes to the sum of net ncome, foregn ncome taxes, and non-ncome and non-payroll taxes. Heteroskedastcty-consstent standard errors are presented n parentheses.
Table 3b The Locaton of Pretax Profts and Drect and Indrect Tax Rates, 1989 Dependent Varable: Log of Pre-Drect Tax or Pre-All Tax Income Log of Pre-Drect Tax Income Log of Pre-All Tax Income (1) (2) (3) (4) (5) (6) (7) (8) (9) Constant 1.2191 2.0081 2.0752 2.4900 2.6825 2.4701 2.6467 2.6776 2.8364 (0.7902) (0.7770) (0.7661) (0.7547) (0.7579) (0.7520) (0.7617) 0.7761 (0.7701) Log of Employee 0.0517-0.0068-0.0122 0.2624 0.2272 0.4411 0.5524 0.3715 0.4380 Compensaton (0.1771) (0.1819) (0.1807) (0.1758) (0.1776) (0.1601) (0.1756) (0.1684) (0.2165) Log of Net PP&E 0.8822 0.7552 0.8361 0.5142 0.5415 0.4328 0.3460 0.4150 0.3760 (0.1784) (0.1785) (0.1877) (0.1717) (0.1709) (0.1859) (0.1939) (0.1806) (0.2205) Log GDP -0.2755 0.1033 0.0872 0.0009 0.0676-0.1690-0.1502-0.0027 0.0149 (0.1163) (0.1399) (0.1547) (0.2105) (0.2190) (0.1533) (0.1473) (0.2055) (0.1974) Drect Tax Rate -4.5990-11.5453-2.5526-7.9546-1.8234-3.9003 (1.3619) (4.3539) (1.3420) (4.2433) (1.3354) (4.7202) Drect Tax Rate 14.5299 10.8599 4.5269 Squared (7.7192) (8.0467) (9.2903) Indrect Tax Rate -1.1289-3.0160-0.9171-2.2765 (0.4332) (1.2589) (0.3943) (1.5104) Indrect Tax Rate 1.6963 1.3080 Squared (0.9597) (1.0839) R-Squared 0.6262 0.7024 0.7276 0.7016 0.7165 0.7095 0.7252 0.7203 0.7343 No. of Obs. 53 53 53 49 49 49 49 49 49 Note: In columns 1, 2, and 3, the dependent varable s the logarthm of the sum of net ncome and foregn ncome taxes. In columns 4 through 9, the dependent varable s the logarthm of the sum of net ncome, foregn ncome taxes and non-ncome and non-payroll taxes. "Log of Net PP&E" s the logarthm of net property, plant and equpment. "Log of Employee Compensaton" s the logarthm of employee compensaton. "Log of GDP" s the logarthm of gross domestc product as reported n the Penn World Tables. "Drect Tax Rate" s the rato of foregn ncome taxes to the sum of net ncome and foregn ncome taxes. "Indrect Tax Rate" s the rato of non-ncome and non-payroll taxes to the sum of net ncome, foregn ncome taxes, and non-ncome and non-payroll taxes. Heteroskedastcty-consstent standard errors are presented n parentheses.
Table 3c The Locaton of Pretax Profts and Drect and Indrect Tax Rates, 1994 Dependent Varable: Log of Pre-Drect Tax or Pre-All Tax Income Log of Pre-Drect Tax Income Log of Pre-All Tax Income (1) (2) (3) (4) (5) (6) (7) (8) (9) Constant 1.3506 1.5205 1.3992 1.6435 1.7820 1.6288 1.5292 1.6172 1.5949 (0.6770) (0.6069) (0.5328) (0.5187) (0.5331) (0.5276) (0.4960) (0.5163) (0.4782) Log of Employee 0.4562 0.4042 0.2982 0.5491 0.5538 0.5952 0.5698 0.5312 0.4609 Compensaton (0.1453) (0.1512) (0.1547) (0.1388) (0.1289) (0.1328) (0.1242) (0.1375) (0.1281) Log of Net PP&E 0.5479 0.5652 0.7166 0.4329 0.4779 0.4034 0.4778 0.4492 0.6018 (0.1616) (0.1573) (0.1393) (0.1417) (0.1368) (0.1364) (0.1255) (0.1366) (0.1225) Log GDP -0.3745-0.2726-0.1243-0.2084-0.2032-0.2704-0.2333-0.2065-0.1665 (0.0956) (0.1242) (0.1296) (0.0882) (0.0828) (0.0819) (0.0757) (0.0907) (0.0950) Drect Tax Rate -1.8731-11.0059-1.2881-9.0644-1.4563-9.6463 (1.6435) (2.5435) (0.8121) (2.8593) (0.9360) (3.5058) Drect Tax Rate 15.0167 18.3473 18.0955 Squared (3.1810) (6.0589) (6.8525) Indrect Tax Rate -0.1768-3.7263 0.1389-2.1850 (0.3274) (1.1308) (0.3489) (1.3592) Indrect Tax Rate 4.4150 3.2190 Squared (1.4161) (1.4983) R-Squared 0.6809 0.7032 0.8024 0.8172 0.8446 0.8110 0.8373 0.8176 0.8660 No. of Obs. 58 58 58 55 55 55 55 55 55 Note: In columns 1, 2, and 3, the dependent varable s the logarthm of the sum of net ncome and foregn ncome taxes. In columns 4 through 9, the dependent varable s the logarthm of the sum of net ncome, foregn ncome taxes and non-ncome and non-payroll taxes. "Log of Net PP&E" s the logarthm of net property, plant and equpment. "Log of Employee Compensaton" s the logarthm of employee compensaton. "Log of GDP" s the logarthm of gross domestc product as reported n the Penn World Tables. "Drect Tax Rate" s the rato of foregn ncome taxes to the sum of net ncome and foregn ncome taxes. "Indrect Tax Rate" s the rato of non-ncome and non-payroll taxes to the sum of net ncome, foregn ncome taxes, and non-ncome and non-payroll taxes. Heteroskedastcty-consstent standard errors are presented n parentheses.
Table 4 Captal/Labor Ratos and Drect and Indrect Tax Rates, 1982, 1989 and 1994 Dependent Varable: Log of Rato of Net PP&E to Employee Compensaton 1982 1989 1994 (1) (2) (3) (4) (1) (2) (3) (4) (1) (2) (3) (4) Constant 1.37793 1.5009 1.4695 1.6329 1.6468 1.8850 1.8280 1.5867 1.6417 1.4763 1.4725 1.2105 (0.2290) (0.2379) (0.2094) (0.3374) (0.2151) (0.2601) (0.2300) (0.2973) (0.2511) (0.3122) (0.3795) (0.5161) Log Average -0.3993-0.3901-0.3412-0.2895-0.4871-0.4880-0.4290-0.3702-0.3839-0.3956-0.3221-0.3361 Wage Rate (0.1369) (0.1451) (0.1364) (0.1552) (0.1110) (0.1144) (0.1113) (0.1112) (0.1007) (0.0973) (0.0964) (0.0908) Drect Tax Rate -0.4033 0.3064 0.6392-0.8589 0.3448 4.5889 0.7783 1.5824 4.9684 (0.5329) (0.6414) (2.9581) (0.7668) (0.7220) (2.5437) (0.6699) (1.0400) (4.3989) Drect Tax Rate -0.1609-9.1674-8.0138 Squared (5.0639) (4.8338) (8.1526) Indrect Tax Rate -0.5426-2.0519-0.8556-1.9260-0.9339-0.3045 (0.3815) (0.7508) (0.3807) (0.6154) (0.3736) (1.7237) Indrect Tax Rate 1.1206 0.9975-0.8821 Squared (0.3920) (0.4478) (1.6378) R-Squared 0.1564 0.1619 0.2893 0.4095 0.2360 0.2545 0.4262 0.4957 0.1970 0.2161 0.3240 0.3554 No. of Obs. 50 49 42 42 58 58 50 50 57 57 54 54 Note: In the three panels labelled 1982, 1989, and 1994, the dependent varable s the log of the rato of net property, plant and equpment to employee compensaton n the respectve year. "Log of Average Wage Rate" s the logarthm of the average wage for a producton worker n the manufacturng sector. "Drect Tax Rate" s the rato of foregn ncome taxes to the sum of net ncome and foregn ncome taxes. "Indrect Tax Rate" s the rato of non-ncome and non-payroll taxes to the sum of net ncome, foregn ncome taxes, and non-ncome and non-payroll taxes. Heteroskedastcty-consstent standard errors are presented n parentheses.
Table 5 Captal/Labor Ratos and VAT Tax Rates, 1982, 1989 and 1994 Dependent Varable: Log of Rato of Net PP&E to Employee Compensaton 1982 1989 1994 (1) (2) (3) (1) (2) (3) (1) (2) (3) Constant 1.6318 1.752 2.1546 2.2442 2.2640 2.0724 1.8644 1.8250 1.4344 (0.5327) (0.5976) (0.6776) (0.2975) (0.3139) (0.3151) (0.2739) (0.2883) (0.4521) Log Average -0.5212-0.6969-0.7389-0.6258-0.7262-0.7200-0.5213-0.5692-0.4960 Wage Rate (0.3068) (0.3045) (0.3106) (0.1343) (0.1990) (0.1882) (0.0852) (0.0805) (0.0957) Drect Tax Rate -0.2733-1.0897-3.6842-1.3343-1.6151 2.4158 0.4092 0.1172 3.5077 (0.6963) (0.9048) (4.2842) (0.9679) (1.2435) (2.9911) (0.8039) (0.7603) (3.2210) Drect Tax Rate 4.9687-8.4544-8.9861 Squared (7.5030) (7.5321) (7.5628) VAT Tax Rate 4.1315-2.6171 2.2455-2.0194 1.6526 5.2485 (2.6309) (8.0751) (2.9584) (7.2038) (1.3032) (4.4600) VAT Tax Rate 33.7131 18.2180-19.1819 Squared (36.0733) (27.8948) (19.9154) R-Squared 0.1479 0.2544 0.3002 0.3751 0.3961 0.4147 0.3485 0.3650 0.3938 No. of Obs. 26 26 26 41 41 41 46 46 46 Note: In the three panels labelled 1982, 1989, and 1994, the dependent varable s the log of the rato of net property, plant and equpment to employee compensaton n the respectve year. "Log of Average Wage Rate" s the logarthm of the average wage for a producton worker n the manufacturng sector. "Drect Tax Rate" s the rato of foregn ncome taxes to the sum of net ncome and foregn ncome taxes. "VAT Tax Rate" s the value-added tax rate as reported n the Unversty of Mchgan World Tax Database.
Table 6 Pre-All Tax Income and Drect and Indrect Tax Rates, 1982, 1989 and 1994 Dependent Varable: Log of Pre-All Tax Income 1982 1989 1994 (1) (2) (3) (4) (1) (2) (3) (4) (1) (2) (3) (4) Constant 4.9871 4.9212 5.0416 5.9887 4.9061 4.9985 4.9697 6.6973 5.2282 5.1738 5.1443 6.8569 (0.5302) (0.5157) (0.5222) (0.5944) (0.4901) (0.5337) (0.4902) (0.4747) (0.6197) (0.6426) (0.6500) (0.7152) Log GDP 0.6149 0.5714 0.6281 0.2347 0.8510 0.5962 0.8678 0.3076 0.5675 0.4048 0.5605 0.0971 (0.1369) (0.1356) (0.1368) (0.2670) (0.1208) (0.1100) (0.1173) (0.1733) (0.2106) (0.1385) (0.2123) (0.2618) Drect Tax Rate -2.2608-1.3302 1.7345-6.0901-5.5827-7.8758-3.4246-4.3974-5.3985 (1.0706) (1.0586) (4.3455) (1.4148) (1.4696) (3.5404) (2.7007) (2.7273) (6.4225) Indrect Tax Rate -1.0279-0.8380-5.5803-1.3872-0.7350-6.1886 0.1520 0.6910-4.5669 (0.4430) (0.4024) (2.2267) (0.4973) (0.3423) (2.1696) (0.7115) (0.9279) (2.7408) Interacton of Drect -0.3906 0.8792 0.4260 Tax Rate and Log GDP (1.0986) (0.7647) (1.3655) Interacton of Indrect 1.1514 1.2290 1.1610 Tax Rate and Log GDP (0.5111) (0.4494) (0.5640) R-Squared 0.3615 0.3824 0.3924 0.4622 0.5090 0.3881 0.5228 0.6569 0.2695 0.2186 0.2878 0.4298 No. of Obs. 44 44 44 44 49 49 49 49 55 55 55 55 Note: In the three panels labelled 1982, 1989, and 1994, the dependent varable s the log of pre-all tax ncome defned as the sum of net ncome, foregn ncome taxes, and non-ncome and non-payroll taxes. "Log GDP" s the logarthm of the gross domestc product n the respectve year. "Drect Tax Rate" s the rato of foregn ncome taxes to the sum of net ncome and foregn ncome taxes. "Indrect Tax Rate" s the rato of non-ncome and non-payroll taxes to the sum of net ncome, foregn ncome taxes, and non-ncome and non-payroll taxes. Heteroskedastcty-consstent standard errors are presented n parentheses.
Table 7 Changes n Pre-All Tax Income and Drect and Indrect Tax Rates, 1982-1989 and 1989-1994 Dependent Varable: Pre-All Tax Income Growth, 1982-1989 Dependent Varable: Pre-All Tax Income Growth, 1989-1994 (1) (2) (3) (4) (5) (6) (1) (2) (3) (4) (5) (6) Constant -0.1990-0.0519-0.1097-0.6802-0.5627-0.6231-0.5731-0.3454-0.5223-0.2607-0.0520-0.1862 (0.3083) (0.2712) (0.2829) (0.2972) (0.2829) (0.3031) (0.0878) (0.0920) (0.0880) (0.1419) (0.1039) (0.1317) GDP growth rate 0.8317 0.5875 0.5881 2.1907 2.0455 2.0466 1.2296 0.8932 1.1400 1.2866 0.8742 1.0959 (0.6718) (0.6848) (0.6782) (0.5113) (0.5062) (0.5011) (0.2342) (0.2238) (0.2298) (0.3967) (0.3248) (0.3840) Net PPE growth rate 0.1083 0.2631 0.1963 0.7704 0.7384 0.7466 (0.3599) (0.3179) (0.3352) (0.1737) (0.2360) (0.1736) Employee Comp. 0.4639 0.3570 0.4160-0.1348-0.1825-0.1408 growth rate (0.3180) (0.2342) (0.2662) (0.1385) (0.1660) (0.1281) Change n Drect -2.2464-1.2279-2.0084-1.1771-3.1585-2.9308-2.9770-2.5414 Tax Rate (0.5886) (0.8556) (0.7720) (1.2419) (0.6873) (0.6407) (1.0785) (1.1093) Change n Indrect -1.0143-0.7369-0.8718-0.6082-0.6730-0.3709-0.9897-0.7504 Tax Rate (0.2097) (0.3501) (0.2868) (0.4706) (0.3397) (0.3329) (0.3674) (0.3676) R-Squared 0.5909 0.6186 0.6416 0.4096 0.4233 0.4450 0.6581 0.5460 0.6699 0.3428 0.2990 0.3937 No. of Obs. 40 40 40 40 40 40 47 47 47 47 47 47 Note: In the left and rght panel, the dependent varable s the growth rate n pre-all tax ncome from 1982 to 1989 and from 1989 to 1994, respectvely. "GDP growth rate" s the percentage change n gross domestc product from 1982 to 1989 and from 1989 to 1994. "Net PPE growth rate" s the percentage change n Net PP&E from 1982 to 1989 and from 1989 to 1994. "Employee Compensaton growth rate" s the percentage change n employee compensaton from 1982 to 1989 and from 1989 to 1994. "Change n Drect Tax Rate" s the change n the rato of foregn ncome taxes to the sum of net ncome and foregn ncome taxes from 1982 to 1989 and from 1989 to 1994. "Change n Indrect Tax Rate" s the change n the rato of non-ncome and non-payroll taxes to the sum of net ncome, foregn ncome taxes, and non-ncome and non-payroll taxes from 1982 to 1989 and from 1989 to 1994.