Unilateral Introduction of Destination-Based Corporate Income Taxation

Transcription

1 Unlateral Introducton of Destnaton-Based Corporate Income Taxaton by Johannes Becker and Markos Jung Abstract We consder source-based and destnaton-based corporate ncome taxaton n a symmetrc two-country model wth heterogeneous frms. Whereas source-based taxaton yelds (fscal) externaltes and s thus neffcent, blateral destnaton-based taxaton s effcent, as proposed by Auerbach & Devereux (2013). Ths paper shows that the unlateral ntroducton of a destnaton-based tax may, frst, ncrease tax revenue n both countres and, second, be Pareto mprovng f certan condtons are met. Thus, even f a multlateral reform s not feasble, a unlateral reform may be desrable both from the ndvdual and the global pont of vew. JEL classfcaton: F23, H21, H25 Keywords: Corporate Taxaton, Destnaton-Based Taxaton, Multnatonal Frms The usual dsclamer apples. Correspondng author, Insttute of Publc Economcs, Unversty of Münster, Wlmergasse 6-8, Münster, Germany, johannes.becker@ww.un-muenster.de. Insttute of Publc Economcs, Unversty of Münster, Germany, markos.jung@ww.unmuenster.de. The author gratefully acknowledges support from the Henrch Böll Foundaton. 1

2 1 Introducton Throughout the world, corporate profts are taxed at source. 1 Decentralzed tax polces yeld tax rate dfferentals whch dstort the allocaton of productve assets. Publc fnance scholars have long thought about ths and come up wth a range of propostons to mtgate the welfare loss due to these dstortons. Most recently, Auerbach & Devereux (2013) proposed the ntroducton of a destnaton-based tax on corporate ncome. If properly mplemented and f all countres adopt such a system, the allocaton of captal and frms could be effcent. However, t may be deemed unrealstc that all countres smultaneously adopt such a system as there wll lkely be wnners and losers. Therefore, ths paper consders the unlateral ntroducton of a destnatonbased corporate ncome tax (DT). We extend the work by Auerbach & Devereux (2013, A&D henceforth) n two ways. Frst, we consder DT n a framework wth mperfect competton whch seems especally sutable n the context of multnatonal frms. Whereas A&D assume that the frm s a prce-taker (snce ths smplfes the analyss of allocatve effcency), we consder Meltz-type heterogeneous frms that supply to ther home market and the foregn market va exports and FDI (as n Helpman et al. 2004). Second, we descrbe the equlbrum after a unlateral ntroducton of DT. A&D consder the margnal ncentve to ntroduce a system wth DT, startng from an equlbrum under the source-based taxaton. Ths allows gnorng the mpact on equlbrum tax rates. It remans unclear whether a full ntroducton leaves the country better-off or not. We show that, under mperfect competton, a multlateral ntroducton of DT yelds constraned effcency n the sense that a dstorton due to mperfect competton perssts. The DT system mples that the export-fdi margn remans unaffected by tax dfferences. Snce, wth cash-flow type taxaton, ths s the only source of tax dstorton n a system wth source-based taxaton, the DT system s constranedly effcent. However, there s a tendency towards overtaxaton whch equally occurs under perfect competton (but has been neglected by A&D). Under plausble assumptons, the unlateral ntroducton of a DT system ncreases tax revenues n both countres and may, under certan crcumstances, mply a Pareto mprovement. That s, even f a coordnated swtch from the current system to the DT s (poltcally) not 1 A small number of countres, ncludng the US, complements source-based taxaton wthn ts terrtoral borders wth taxaton of worldwde ncome. However, as t s often argued, deferral of repatraton and other tax avodance strateges make foregn ncome of corporatons effectvely tax exempt. 2

3 feasble, some countres may consder the ntroducton unlaterally. As mentoned above, the lterature provdes a range of reform proposals drected at mprovng the effcency (and, sometmes, the farness) of the nternatonal tax system. Obvously, the proposed cure s a functon of the underlyng dagnoss. The latter has changed over tme, and t may be helpful to quckly recapture the man steps. Peggy Musgrave (née Rchman 1963) and, later on, Roger Gordon (1986) stressed the neffcency of source-based captal ncome taxaton and ponted out that pure resdence based taxaton s superor. These frst studes (as well as the frst treatments of tax competton n Zodrow & Meszkowsk 1986 and Wlson 1986) share the assumpton that captal s homogeneous and perfectly moble across borders, whereas e.g. labor and frms are mmoble. Snce markets equate the after-tax return to captal across borders and allocatve effcency requres equal pre-tax returns n each country, decentralzed tax rate settng almost necessarly leads to dstortons. Pure resdence based taxaton may solve ths problem, but may be mpractcal due to hgh enforcement cost and quckly changng ownershp structures. As a consequence, effcent captal taxaton may be nfeasble and, thus, the optmal tax on the normal return to captal s presumed to be zero. Ths does, however, not prevent government to tax economc rents. One straghtforward mechansm to acheve ths s to grant full tax deductblty of the normal return to captal under the standard source-based tax system. Ths reduces the effectve margnal tax rate on captal to zero. 2 An alternatve way of achevng ths a system of cash-flow taxaton (see e.g. Bond & Devereux 2002). 3 In the 1990s, after observng the surge of foren drect nvestment throughout the world, especally of US FDI n Ireland, and the rse of multnatonal frms, the dscusson shfted towards the moblty of frms or whole producton unts, nstead of margnal unts of captal. As ponted out by Devereux & Grffth (1998), taxes do dstort the allocaton of producton even f the normal return to captal remans untaxed. A crucal nsght of these consderatons s that full deductblty of captal expenses does not abolsh all dstortons. Fnally n the 2000s, the focus of the academc as well as the poltcal debate shfted agan and centered on the tax-nduced shftng of book profts va transfer prcng and fnancal polcy. Wth proft shftng alone, formula apportonment (see, e.g., Gordon & Wlson 1986) or CFC regulaton may be sutable 2 Snce the cost of debt fnance s already deductble, such a polcy reform would requre an allowance for corporate equty (ACE). 3 Snn (1990) supposed that grantng full deductblty of captal expenses would mmunze a country from the pressures of globalzed captal markets. 3

4 solutons, but wth cross-border nvestment n multnatonal frms, such nsttutons may gve rse to addtonal dstortons. In general, the opportunty of proft shftng may exacerbate the tax elastcty of nvestment (because frms nvest n low-tax countres n order to shft profts) or mtgate t (because frms antcpate that they wll not have to tax profts n hgh-tax countres). The latter assumpton seems to be shared by most observers, as nvestment dstortons seems to have somehow lost poltcal (and academc) attenton. Auerbach & Devereux (2013) propose a system that tackles the dstorton drvers from all three past decades: moble captal, moble frms and moble profts. Snce the locaton of frms and profts (or, more generally, value added) has tax consequences as long as taxaton depends on the locaton of producton, they propose to make the tax lablty dependng on the locaton of the consumer. Snce frms cannot (easly) move consumers, such a system could get rd of all three potental dstortons. Such a system would be smlar to a destnaton based system of value added taxaton. In the verson of A&D, under a destnaton based system of corporate taxaton, tax revenue s collected by the country where the consumer resdes. For allocatve effcency, t s rrelevant whch government collects the revenue as long as frms pay the same amount of tax. However, f tax rates are endogenzed (as n our framework), the destnaton country has the ncentve to export some of ts tax burden to foregn producers. Ths may gve rse to overtaxaton as we show below. Another dffculty arses when the DT system s unlaterally ntroduced. Then, wavng the rght to tax to the destnaton country would create untaxed ncome (exporters n the country that has ntroduced DT) and double taxed ncome of mports from foregn producers. We therefore modfy the settng and assume that the DT country stll taxes ts exporters, but at the rate of the destnaton country. In allocatve terms, ths system s equvalent to the one proposed by A&D. The remander of the paper s organzed as follows. Secton 2 presents the model setup. Secton 3 dscusses the benchmark scenaro under source-based taxaton (ST) wth underprovson of publc goods. In Secton 4, we analyze tax polcy under destnaton-based taxaton f multlaterally or unlaterally ntroduced. Secton 5 dscusses several extensons. Secton 6 concludes. 2 Model setup Before we descrbe the detals of the model, t may be useful to justfy some of our modellng choces. Our model s a standard Helpman et al. (2004) 4

5 framework wth Meltz (2003)-type heterogeneous frms augmented by an export-fdi decson. We add to ths framework proft taxes whch fnance a publc good. All other ngredents of the model stay the same, whch s why we reduce the model descrpton to the necessary. In the baselne verson of the model, we do not consder moble captal and proft shftng. Gven the above dscussed focus of the exstng lterature on these two aspects, ths may seem surprsng, but t can be made plausble as follows. Frst, snce all consdered tax system here are cash-flow systems, the nvestment decson s always non-dstorted and can therefore be neglected. Thus, ntegratng captal as a productve nput does not provde any new nsghts, but comes at the cost of more complexty (as, for nstance, a world captal market would have to be modelled etc.). Second, proft shftng opportuntes have a smlar effect as choosng between exports and FDI; the man drver s the statutory tax rate. Moreover, snce we consder a symmetrc model wth equal tax rates n equlbrum, proft shftng n the symmetrc equlbrum s zero. Therefore, we abstract from proft shftng n the baselne verson of the model, wll however dscuss ts mplcaton n Secton 5. Consder a world wth two countres, home (h) and foregn (f), ndexed by = h, f. In each country, the representatve consumer derves utlty from a publc good and consumpton of prvate goods. The latter consst of a numerare good y and a CES-aggregate of dfferentated goods ndexed ω. Subutlty functon U captures the utlty from prvate consumpton and s gven by where σ = 1 goods, ρ ρ γ 1 ρ U = y + 1 γ [ q (ω) ρ dω ωɛω s the elastcty of substtuton between the dfferentated s the elastcty of substtuton between both sectors, q (ω) denotes the quantty of varety ω, and Ω s the set of of avalable goods, the mass of whch s gven by M. 4 we assume 0 < γ < ρ < 1. 5 ] γ ρ (1) In order to have postve elastctes of substtuton, Wth p (ω) the prce of varety ω and I the household s ncome, the household s budget constrant s gven by I = y + p (ω) q (ω) dω. (2) ωɛω Income I conssts of wage ncome w and dvdend ncome (.e. effectvely the 4 Explct expressons of q (ω) as well as the prce ndex P are derved n the Appendx. 5 See Chor (2009) for a smlar assumpton. 5

6 sum of after-tax profts of domestc frms). Due to the quas-lnearty of the utlty functon, all ncome effects are absorbed by the numerare sector. 6 There are two sectors of producton. The numerare good sector s assumed to produce under constant returns to scale technology wth a margnal labor productvty of unty. Moreover, competton s perfect and trade s costless, whch pns down the wage to unty n both countres. Profts n the numerare sector are zero. In the CES sector, there s a contnuum of heterogeneous frms, as n Meltz (2003) and Helpman et al. (2004). Pror to market entry, an exogenous number of frms, N, draws a random productvty φ from a dstrbuton G (φ). Frms have the opton to serve the domestc market (D), to serve the other market va exports (E) or FDI (F ), or to not enter the market. For each of these actvtes, there are fxed costs F X nvolved wth X {D, E, F } denotng the actvty. Smlarly, there are ceberg shppng cost, τ X 1,.e. f a quantty q s to be sold, a quantty τ X q has to be shpped. We assume τ E = τ > 1 as well as τ D = τ F = 1 for all. Each frm pays a proportonal tax t X on ts net profts. After-tax profts of a frm wth productvty φ are, dependent on the actvty X, gven by π X (φ) = [( p X (φ) τ ) ] X q X (φ) F X (1 t X ) (3) φ Under source based corporate proft taxaton, we have t X = t for domestc producton and exports and t X = t for FDI. As n Meltz (2003), all frms charge the same markups 1 ρ, whch mples a prce p X (φ) = τ X ρφ. The productvty cutoff for domestc producton s denoted by φ D and defned by π D (φ D ) = 0. Wth productvty ncreasng, the frm may consder exportng ts goods. The assocated cutoff s φ E and defned by π E (φ E ) = 0. Hghly productve frms may consder FDI nstead of exportng. The cutoff s φ F and defned by π E (φ F ) = π F (φ F ). The parameters of the model are chosen such that φ D < φ E < φ F whch, among others, requres F D < τ σ 1 F E < F F. Thus, frms wth a productvty below φ D do not produce, between φ D and φ E only serve the domestc market, between φ E and φ F export, and above φ F nvest abroad nstead of exportng. Governments are assumed to maxmze a welfare functon W = W (T, U ) where T denotes country s corporate ncome tax revenue. Ths ncludes 6 We assume that the ncome s hgh enough such that the household consumes both the numerare and the CES goods. 6

7 the case of revenue maxmzaton, f W U = 0, and the case of a benevolent government that maxmzes ts ctzens utlty whch depends on T because t fnances the publc good. We wll contnue by outlnng the benchmark,.e. taxaton under a system of source-based corporate ncome taxaton. Then, we quckly outlne a system wth destnaton-based corporate ncome taxaton as proposed by Auerbach & Devereux (2013). After that, we analyze the effects of a unlateral ntroducton of a DT system. 3 Benchmark: Tax polcy under source-based taxaton (ST) Under a pure (and stylzed) source-based corporate ncome tax system, the government n taxes all profts that accrue wthn ther jursdcton at source. Country s household has ncome of I ST = (1 t ) (B D + B E ) + (1 t ) B F + w. Tax revenue s gven by T ST = t (B D + B E + B F ) where B D, B E and B F refer to the tax bases from domestc producton, exports and nbound FDI, respectvely. These can be expressed as B X = π X (φ) ω Ω X 1 t X dω where Ω X denotes the set of frms nvolved n actvty X. 7 The government n sets t such that dw = W T dt + W U du = 0 (4) wth dw gven by dw = (W T W U ) (B D + B E ) + W T B F (5) ( +W T t ST dbd + db E + db ) ( F du dp + W U + du ) dp dp dp The frst term on the rght hand sde captures the mechancal reallocaton of funds from the domestc household to the government. If the margnal socal value of tax revenue s larger than the margnal socal value of prvate consumpton, ths effect s postve. The second term represents the socal gan from taxng foregners (B F ). The thrd term captures the behavoral effects on the tax base, wth db D = B D dp P dp > 0, db E = B E dω E Ω E + B E dp P < 0 0. We assume that a tax rate ncrease and db F = B F dω F Ω F + B F P always reduces the tax base under ST as well as n the unlateral DT system. 8 7 A more explct expresson of the tax base as well as the frst dervatves of all cutoffs, prces, and tax bases are gven n the Appendx. 8 Hence, we assume db D + db F < 0, whch s the case f the nter-sectoral elastcty of 7

8 Note that the tax t does not drectly affect the domestc producton cutoff B D φ D and the export cutoff, = 0 and B E φ E = 0. Fnally, prvate utlty s negatvely affected by changes n the prce level, both va consumpton and dvdend ncome. 9 Now, consder a symmetrc Nash equlbrum, startng from whch both countres margnally ncrease ther tax rates by an equal amount. Then, all cutoffs reman unaffected, as do the prce levels. The resultng welfare effect s gven by dw = (W T W U ) (B D + B E ) + W T (B F ) W U (B F ) whch s postve as long as W T > W U. Proposton 1 [Zodrow & Meszkowsk 1986] In the symmetrc equlbrum, competton wth source-based corporate taxes yelds neffcent low tax rates provded that the tax base from nbound nvestment actvty, B F, s suffcently small. Thus, tax competton yelds underprovson of publc goods. Proof. Wth B F suffcently small, W T > W U s a necessary condton for dw = 0 to hold. Then, a coordnated tax ncrease rases welfare whch proves the above Proposton. A foregn owned tax base (due to cross-border ownershp lnks), B F, may gve rse to tax exportng and potentally overtaxaton (see Bucovetsky 1995 and Huznga & Nelsen 1997 for the frst formal treatments of the foregn frm ownershp effect ). Ths effect makes the classcal result of underprovded publc goods (Zodrow & Meszkowsk 1986) vansh. In what follows, we wll suppose that ths effect wll not be too strong,.e. that the classcal dlemma as descrbed by the classcal tax competton models stll exsts. ρ substtuton s suffcently hgh. 9 ρ γ Note that the envelope theorem mples that margnal changes n cutoffs do not affect U. However, snce the sngle frm takes the prce level as gven, a change n prce level has a frst order effect on prvate utlty. Wth U = I + 1 γ as well as db E dp, db F dp γ P γ 1 γ > 0 (see the Appendx), these effects are gven by du = di P dp dp du = dp 1 γ 1 = (1 t ) dbd dp 1 γ 1 P < 0 di dp = (1 t ) dbe dp + (1 t ) dbf dp > 0. and (1 t ) db D dp < P 1 γ 1 8

9 Gven that tax competton yelds neffcent outcomes (be t n terms of publc good provson or captal allocaton), the queston arses why sourcebased taxes stll preval. The answer may be that the alternatves are equally unattractve. Country heterogenety may prevent tax harmonzaton (Bucovetsky 1991) and a swtch to pure resdence taxaton may ether be mpractcal or dstortve as well (see e.g. Becker & Fuest 2011). A novel approach s taken by Auerbach & Devereux (2013) who propose a swtch to a destnaton-based corporate tax system whch follows the prncple used n value-added taxaton. Before we descrbe ths approach, we need to prepare the analyss as follows. For later use, t s helpful to dstngush the tax base generated by sellng product to domestc consumers, B D + B F, and the tax base due to sellng product to foregn consumers, B E. Assume, for a moment, that these two subsets could be taxed at dfferent rates (see Keen 2001 for a broader dscusson on dfferental or preferental treatment of dstnct tax bases). Let t E denote the tax rate on exportng frms and t DF the tax rate on B D + B F. Optmal tax rates, t DF dw dt DF dw dt E and t E, are mpled by = (W T W U ) B D + W T B F + W T t DF = (W T W U ) B E + W T t E db E ( dbd + db F ) + W U du dp dp + W U du dp dp (6) In general, we cannot determne whether t DF or t E other. We can, however, state the followng lemma. s hgher than the Lemma 1 t DF > t E f () W U = 0 or W T s suffcently larger than W U or () prce level effects, dp Proof. See the Appendx. and dp, are suffcently small. The optmal tax rate under ST s a weghted average of t DF 4 Tax polcy under destnaton-based taxaton (DT) and t E. Auerbach & Devereux (2013) propose the ntroducton of destnaton-based corporate proft taxaton. In short, taxes are leved by the country n whch the consumers are. If consumers are suffcently mmoble, such a system leaves the allocaton of captal undstorted and provdes a level playng feld for domestc and foregn compettors on a gven market. 9

10 4.1 Multlateral ntroducton If all countres adopt the DT system, the government n taxes all profts that are assocated wth goods that reman n or are shpped nto ts country. Prvate ncome s now I DT = (1 t ) B D + (1 t ) (B E + B F ) + w (7) and tax revenue T DT = t (B D + B F + B E ). (8) Proposton 2 Destnaton-based corporate ncome taxaton leaves the export- FDI margn unaffected. Proof. By constructon of the model, taxes only dstort the decson between exportng and FDI. Snce, under DT, profts from exports and FDI are taxed at the same rate, ths margn wll not be dstorted. Gven ths, taxaton has no effcency cost and could, n prncple, be effcent gven the nstruments avalable. 10 However, snce each government effectvely taxes foregners ncome, the ncentve remans to export the tax burden. Corollary 1 () If there s no lump-sum transfer to the household, the tax rate s neffcently hgh. () If there s a lump-sum transfer avalable, the optmal tax s effcent but close to 100 per cent. Proof. () Insertng dt = B D +B F +B E and du = B D nto (4) yelds W T < W U. A coordnated decrease of tax rates ncreases welfare. () Wth a lump-sum transfer to domestc households, optmal polcy mples W T = W U. Frms are taxed as hgh as possble,.e. close to full expropraton, snce taxes do not have any behavoral effects. The resultng revenue s dstrbuted among households and publc goods provson such that W T = W U. A coordnated decrease of taxes does not ncrease welfare. Even despte of the overtaxaton result, the DT system seems attractve due to ts allocatve and compettve neutralty. However, t seems equally plausble that a coordnated swtch from ST to DT seems unrealstc. Therefore, the queston arses whether a unlateral ntroducton of DT s desrable both from the ndvdual and the global pont of vew. 10 Note that effcency could be further enhanced by subsdzng FDI, as the nter-sectoral allocaton of labor s not effcent due to dfferent markups n both sectors (see, for example, Dxt & Stgltz 1977). 10

11 4.2 Unlateral ntroducton We wll now assume that one country unlaterally ntroduces destnatonbased taxaton (UDT) whereas the other country keeps the source-based tax system (UST). A unlateral ntroducton requres some adjustments vs-a-vs a multlateral ntroducton. Frst, snce the DT-country cannot tax profts from mports (at least not wthout double taxng the mportng frm), t wll voluntarly forego the tax on these frms. Second, t wll contnue taxng the exportng frms wthn ts own borders, but wll adopt the tax rate of the country of destnaton. Thus, frms headquartered n the DT-country have the same ncentve envronment as under a full-fledged DT system, but the revenue s collected by the source country nstead of the destnaton country. Let country be the DT country. Its household has thus ncome of (7) and tax revenue s gven by T UDT = t (B D + B F ) + t (B E ) (9) A small ncrease n t affects welfare as follows. ( ( dbd (W T W U ) B D + W T B F + t + db )) ( ) F du dp + W U dp (10) wth db D and db F defned above. By effectvely lettng country decde on the tax rate on exportng frms, the domestc tax rate does no longer affect ths tax base or the outbound export-fdi margn. Accordngly, t has no mpact on the prce level P. Effectvely, the DT country makes the same choce as f t had to choose the optmal tax t DF on B D +B F. Evaluated at ( t ST, t ST ), we can conclude that, f t DF > t E, the DT country has a margnal ncentve to ncrease ts tax rate above t ST. 11 Now, let country tax accordng to ST and country havng a DT system. Country s household has ncome of I UST = (1 t ) (B D + B E ) + (1 t ) B F + w,.e. the same as under a multlateral ST system. Tax revenue s stll gven by T UST = t (B D + B F + B E ). 11 t DF > t E mples that, at W U ( du dp dp ) < 0. ( t ST, t ST ), (WT W U ) B E + W T t db E + 11

12 A small ncrease n t affects welfare accordng to ( ) ( ) db E du dp (W T W U ) (B D + B E ) + W T B F + t + W U dp (11) wth db E defned above. Now, t does not drectly affect the nbound export- FDI cutoff whch s why the domestc prce level remans constant. Evaluated at ( t ST, t ST W T t ( dbd ), an ncrease n t ncreases welfare f + db F ) ( ) du dp + W U < 0 (12) dp whch s the case due to du dp dp < 0 and the assumpton of db D We can now state the followng proposton. + db F < 0. Proposton 3 If the desred tax on exportng frms s lower than t ST, a unlateral ntroducton of DT provdes both countres wth a margnal ncentve to ncrease ther tax rate. ), the ST country has an ncentve to ncrease ts tax Proof. At ( t ST, t ST rate, as shown above. The DT country does so too f (10) s postve. At ( t ST, t ST ), ths s the case f the desred tax on exportng frms s lower than t ST. In other words, by foregong the rght to determne the tax rate on exportng frms, the DT country sets ts tax rate only on domestc frms and profts from nbound nvestment. That s, t actually chooses ts t DF (see Secton 3). Snce the optmal tax rate under a pure ST system s a weghted average of t E tax on exporters, t E and t DF, t follows that ths tax rate s hgher f the desred, s lower. Proposton 4 Suppose that tax rates are strategc complements (.e. reacton functons are upward-slopng), and the desred tax on exportng frms s lower than t ST. Then, a unlateral ntroducton of DT unambguously ncreases tax revenue. Proof. Wth upward-slopng reacton functons, a margnal ncentve to ncrease taxes for both countres mples that equlbrum tax rates ncrease. How does ths affect welfare? 12

13 Corollary 2 If the socal value of tax revenue s large enough (e.g. when governments are revenue maxmzers or frms are foregn owned), the unlateral ntroducton mples a Pareto mprovement. Proof. Ths follows from the above proposton and lemma 1. 5 Extensons 5.1 Integratng proft shftng opportuntes In the above model, there s no ntra-frm trade n any way. When a frm swtches from exportng to FDI, the whole producton unt s shfted to the foregn country. Ths makes proft shftng hard. However, ntra-frm trade (and, thus, proft shftng) s an mportant aspect of multnatonal frm actvty, and a reform proposal must deal wth t. Auerbach & Devereux (2013) who use a rcher model than the one presented above allow for ntra-frm mports of the basc product whch serves as an nput good for fnal good producton at the destnaton locaton. These mports need to be prced, and the prce can be manpulated for proft shftng purposes. The perhaps easest way of ntegratng proft shftng nto the above model s to assume that the foregn plant uses a patent located at the headquarters. A prce z for usng the patent s pad from the foregn subsdary to the headquarters. We assume that a far prce s gven by z, but the frm may devate from ths prce n order to save taxes. Transfer prce manpulaton, however, comes at a convex cost c (z z) wth c =sgn(z) and c > 0 and c (0) = c (0) = 0. Proft shftng s only possble f the frm has a foregn plant,.e. f t s n the FDI mode. Its profts are then π F = π F (φ) + z (t t ) c (z z) By manpulatng z, the frm may shft profts from one locaton to the other. The proft-maxmzng choce of z requres c = t t,.e. transfer prce manpulaton only takes place wth non-zero tax dfferentals. That s, n the symmetrc equlbrum, the ncentve to shft profts s zero. The payment of z reduces the tax base of the foregn subsdary and ncreases t at the domestc headquarters. In the symmetrc equlbrum, these two cancel out. However, when settng the tax rate, the government accounts for the elastcty of z wth regard to the tax rate. In other words, proft shftng makes ncreasng tax rates more costly n terms of tax revenue, whch 13

14 mples that equlbrum tax rates are lower. How does proft shftng affect the effcency propertes of the DT system? The answer to ths queston depends on how exactly proft shftng s modelled. If the patent s used only from producton drected at consumers n the foregn market, a DT system abolshes all proft shftng ncentves snce the proceeds of supplyng to ths market are taxed at the destnaton country s rate ndependent of the transfer prce. In ths case, the welfare-enhancng effect of a DT system s renforced both for multlateral and unlateral ntroducton. Thngs get more complcated f the patent s used for both, producton for domestc and foregn markets. Ths s the case of frm-specfc assets, also covered by Auerbach & Devereux (2013), whch we deal wth n the next subsecton. 5.2 Frm-specfc assets The most productve frms n the above model supply both to the domestc and the foregn market. Now, consder the case n whch there s a frm-specfc asset, e.g. a patent, that s used for producton drected at both markets. In order to mplement the DT system, ncome accrung to the use of the patent has to be assgned to ether the domestc market or the foregn market, n order to attrbute the cost for developng the patent. To see ths, consder the followng example. Assume that a frm s actve n a low-tax country and a hgh-tax country. Both actvtes requre the use of a patent whch has been developed by the frm tself at consderable cost. Under DT system, the frm has to ndcate whch part of the proft s due to actvty n each country. It has now a strong ncentve to underreport the use of the patent for ts low-tax country actvty and to overreport t for the hgh-tax country. Then, t may assgn a larger part of the patent development cost to hgh tax country actvty, thus decreasng the proft from ths actvty. Smlarly, ts proft from low-tax country actvty are ncreased. Overall, ts tax lablty s reduced. In other words, the necessty to assgn the use of the frm-specfc asset gves rse to the same problems dscussed n the context of transfer prcng under standard source-based tax systems. Whether manpulaton of ths knd ncreases or decreases the attractveness of the DT system s hard to tell; t wll depend on whether shftng between locatons s easer than shftng between assgned actvtes or vce versa. 14

15 5.3 Dversfed ownershp Broad ownershp (Auerbach & Devereux 2013),.e. dverse shareholders from (many) dfferent countres, s consdered a defnng feature of multnatonal frms. In the above model, cross-border ownershp lnks occur endogenously when a frm swtches from exportng to FDI. As an alternatve, domestc households could buy shares from foregn households and vce versa. As Bucovetsky (1995) and Huznga & Nelsen (1997) have ponted out, an ncrease n foregn frm ownershp reduces the cost of taxaton as perceved by a natonal government, due to the assumpton that foregners ncome s not part of the government s objectve functon. Thus, an ncrease n foregn frm ownershp s lkely to ncrease equlbrum tax rates, potentally up to a pont where neffcent overtaxaton occurs. To the degree that broad ownershp reduces the sze of the fscal externalty and thus the underprovson of publc goods, t reduces the value of swtchng to a DT system (although t should be kept n mnd that a system wth multlateral DT s prone to overtaxaton, too). Ths s also true for a unlateral swtch to a DT system. 5.4 One (unlateral) step further: Treatment of foregn nvestors In the above symmetrc two-country model wth unlateral ntroducton of DT, the remanng dstorton concerns the foregn nvestors decson between exportng and FDI. There s, however, an easy way of fxng ths dstorton. The reform country (.e. the DT country) could tax the foregn nvestors at the foregn rate,.e. B F at a rate of t, thus unlaterally transformng the two-country world nto an effectve DT system. One potental caveat here s that the tax system would have to condton on ownershp structures (and not, as n the current verson, on the destnaton of product). Snce ownershp structures vary frequently and can easly be manpulated, such a tax system would be prone to neffcent avodance actvtes. 5.5 One (multlateral) step further: Mtgatng overtaxaton The ncentve to overtax n a multlateral DT system s due to the fact that foregners ncome s part of the destnaton country s tax base. For the allocatve effcency, though, t does not matter whch government actually receves the revenue. Therefore, t mght be a soluton f the revenue s redstrbuted accordng to the natonal ownershp. Ths would n any case elemnate the ncentve to overtax snce both the gan (n terms of publc goods) and the cost 15

16 (n terms of net ncome) are born by domestc agents (whch, by assumpton, are part of the governement s objectve functon. As n the prevous subsecton, there s the caveat that such a system would requre the tax system to condton upon ownershp. However, ths tme ownershp only affects who receves the revenue and not how much the frm pays. Ths mples that there s no ncentve to manpulate ownershp for tax savng purposes. Wth dverse ownershp, e.g. n publcly lsted frms, and quckly changng ownershp structures, condtonng on ownershp may be dffcult, though. One should keep n mnd, however, that t s not the exact ownershp structure at a gven pont n tme whch affects the effcency of tax rate settng, but the government s percepton that t only affects domestc agents. Therefore, measurement problems may not be as severe as one could thnk. 6 Concluson In ths paper, we reconsder destnaton based corporate ncome taxaton as proposed by Auerbach & Devereux (2013) n a framework wth heterogeneous frms and mperfect competton. Our analyss provdes support for such a reform proposal n two ways. Frst, we show that a DT system performs well under mperfect competton although t cannot correct for the dstortons nduced by mperfect competton. Second, we show that a multlateral ntroducton s not a necessary prerequste for such a system to enhance welfare. Instead, we show that a unlateral ntroducton may not only enhance welfare n the reform country, but also globally (at least f certan condtons are met). There are three caveats wth regard to the robustness of our results. Frst, real world tax competton s characterzed by large asymmetres. The large, economcally powerful countres levy relatvely hgh tax rates whereas small perphercal countres usually have low tax rates. Under these crcumstances, a unlateral swtch to DT wll not necessarly be desrable. The results derved n ths paper should be understood as showng that, f asymmetres are not excessvely large, destnaton based taxaton makes sense even f not all countres partcpate. It should be noted, though, that the same caveat s there for a multlateral ntroducton. Even f the allocaton of captal, frms and profts s restored, such a polcy may stll be unappealng for countres wth strong exporters and small markets, as they wll lose a large part of ther revenue. Second, frms make a large number of decsons, many of whch are tax 16

17 revelant. As every model, we have abstracted from ths complexty and, of course, we cannot exclude that there are decsons whch are not only unaffected by a swtch to DT but also aggravated. A case n pont s the attrbuton of ncome to cost dscussed n Secton 5.4. Agan, ths caveat concerns more DT ntroducton n general than a unlateral swtch. Thrd, an ntroducton of DT be t multlateral or unlateral may strongly affect a country s ncentve to enforce taxaton. Auerbach & Devereux (2013) propose that the destnaton country receves all tax revenue from actvty related to supplyng to ts nhabtants. Ths would requre that the source based countres enforce taxes that they do not get. The abundant experence wth e.g. nformaton exchange n captal ncome taxaton does not warrant an overly optmstc vew on ths. References [1] Auerbach, A. J., & Devereux, M. P. (2013). Consumpton and cash-flow taxes n an nternatonal settng. NBER Workng Papers 19579, Natonal Bureau of Economc Research. [2] Becker, J., & Fuest, C. (2011). Source versus resdence based taxaton wth nternatonal mergers and acqustons. Journal of Publc Economcs, 95(1), [3] Bond, S.R. & Devereux, M.P. (2002). Cash Flow Taxes n an Open Economy, CEPR Dscusson Paper [4] Bucovetsky, S. (1995). Rent seekng and tax competton. Journal of Publc Economcs, 58(3), [5] Chor, D. (2009). Subsdes for FDI: Implcatons from a model wth heterogeneous frms. Journal of Internatonal Economcs, 78(1), [6] Devereux, M. P. & Grffth, R. (1998). Taxes and the locaton of producton: Evdence from a panel of US multnatonals, Journal of Publc Economcs 68(3): [7] Dxt, A. K., & Stgltz, J. E. (1977). Monopolstc competton and optmum product dversty. The Amercan Economc Revew, [8] Gordon, R.H. (1986). Taxaton of Investment and Savngs n a World Economy. Amercan Economc Revew 76(5): [9] Gordon, R.H. & Wlson, J.D. (1986). An Examnaton of Multjursdctonal Corporate Income Taxaton under Formula Apportonment, Econometrca 54: [10] Helpman, E., Meltz, M. J., & Yeaple, S. R. (2004). Exports vs. FDI wth 17

18 heterogeneous frms. The Amercan Economc Revew, [11] Huznga, H., & Nelsen, S. B. (1997). Captal ncome and proft taxaton wth foregn ownershp of frms. Journal of Internatonal Economcs, 42(1), [12] Keen, M. (2001). Preferental regmes can make tax competton less harmful. Natonal Tax Journal, [13] Meltz, M. J. (2003). The mpact of trade on ntra-ndustry reallocatons and aggregate ndustry productvty. Econometrca, 71(6), [14] Rchman, Peggy B. (1963). Taxaton of Foregn Investment Income - An Economc Analyss, The Johns Hopkns Press, Baltmore, MD. [15] Snn, H.-W. (1990). Tax Harmonzaton and Tax Competton n Europe, European Economc Revew 34(2/3): [16] Zodrow, G. R., & Meszkowsk, P. (1986). Pgou, Tebout, property taxaton, and the underprovson of local publc goods. Journal of Urban Economcs, 19(3),