Bailout policy in a globalized economy

Transcription

1 Balout polcy n a globalzed economy Nelly Exbrayat, Therry Madès, Stéphane Rou To cte ths verson: Nelly Exbrayat, Therry Madès, Stéphane Rou. Balout polcy n a globalzed economy. Workng paper GATE <halshs > HAL Id: halshs Submtted on 7 Jan 2014 HAL s a mult-dscplnary open access archve for the depost and dssemnaton of scentfc research documents, whether they are publshed or not. The documents may come from teachng and research nsttutons n France or abroad, or from publc or prvate research centers. L archve ouverte plurdscplnare HAL, est destnée au dépôt et à la dffuson de documents scentfques de nveau recherche, publés ou non, émanant des établssements d ensegnement et de recherche franças ou étrangers, des laboratores publcs ou prvés.

2 GROUPE D ANALYSE ET DE THÉORIE ÉCONOMIQUE LYON - ST ÉTIENNE WP 1340 Balout polcy n a globalzed economy Nelly Exbrayat, Therry Madès, Stéphane Rou December 2013 Documents de traval Workng Papers

3 GATE Groupe d Analyse et de Théore Économque Lyon- St Étenne 93, chemn des Moulles Ecully France Tel. +33 (0) Fax +33 (0) , rue Basse des Rves Sant- Etenne cedex 02 France Tel. +33 (0) Fax. +33 (0) Messagere électronque / Emal : gate@gate.cnrs.fr Téléchargement / Download : Publcatons / Workng Papers

4 Balout polcy n a globalzed economy Nelly Exbrayat, Therry Madès and Stéphane Rou Abstract Ths paper explores how trade ntegraton nfluences the decson by natonal governments to balout manufacturng frms. We develop a 2-country model of generalzed olgopoly wth heterogenous frms and trade costs. Hgh-cost frms are elgble for a balout whle low-cost frms are proftable. Our results show that trade lberalzaton nfluences both poltcal benefts of a balout and ts relatve cost as compared to a lassez-fare polcy. If the fall n trade cost s so large that t allows hgh-cost frms to become exporters, governments mght move away from a balout polcy to a lassez-fare polcy. In contrast, a margnal declne n trade costs that does not affect the export status of hgh-cost frms, always makes governments more prone to adopt a balout decson. Keywords: soft-budget constrant; tax competton; heterogenous frms; trade cost; locaton. JEL classfcaton: F12; F15; D21; H25. We are grateful to partcpants of 3 rd SEBA GATE workhop (Chengde), at 59 th annual RSAI conference (Ottawa), at 4 th UECE Meetngs on Game theory and applcaton (Lsbon), at GATE workshop on Publc Polces and Spatal Economy (Lyon and Sant-Etenne), at CESAER semnar n Djon, and to Jacques-Franços Thsse, Frédérc Robert-Ncoud, Cécle Detang Dessendre, Mare-Laure Breullé and Smon Laponte for helpful comments and suggestons. Unversté de Lyon, Lyon, F-69007, France ; CNRS, GATE Lyon Sant-Etenne, Ecully, F-69130, France ; Unversté Jean Monnet, Sant-Etenne, F-42000, France. Unversty of Frbourg, Faculté des scences économques et socales, boulevard de Perolles, 90, 1700 Frbourg (Swtzerland). Unversté de Lyon, Lyon, F-69007, France ; CNRS, GATE Lyon Sant-Etenne, Ecully, F-69130, France ; Unversté Jean Monnet, Sant-Etenne, F-42000, France. 1

5 1 Introducton A frm or any nsttuton s faced wth a soft-budget constrant f t expects to be baled out n case of fnancal trouble. Ths creates a moral hazard problem as the expectaton of a balout can erode managers effort and thereby encourage falure. 1 Orgnally developed by Korna (1979, 1986), the concept of soft-budget constrant (hereafter SBC) was frst desgned to descrbe the behavor of governments n centrally planned and transton economes. 2 However, the recent crss perod resulted n a wdespread polcy of ndustry balout n most of the developed countres, n addton to tremendous rescue plans n the fnancal and bankng sector. The most emblematc example s certanly the federal balout for Amerca s Bg Three automoble companes n 2009, but many rescue plans have been decded n other countres and ndustral sectors over the past decades. 3 These examples corroborate the clam by Korna, Maskn and Roland (2003, p. 56) that the effects of the SBC syndrome are clearly perceptble n the tradtonally captalst part of the world as well. One reason s that some determnants lke the poltcal desrablty of a balout are not proper to a partcular economc or poltcal system (see. Robnson and Torvk, 2009). Poltcans may be poltcally motvated to grant subsdes to companes n fnancal trouble to save jobs. In addton, polcy-makers may be concerned by the too bg to fal argument when falures cause negatve spllovers on the rest of the economy. The soft budget constrant phenomenon has been studed manly n a closed economy framework (see. Korna, Maskn and Roland, 2003 ). Ths s clearly a lmt of the lterature because corporate balouts n well-establshed market economes have been takng place over the two last decades n a new context of globalzaton characterzed by both ncreasng captal moblty and trade openness. Our paper ams at analyzng how corporate bal-out decsons can be shaped by these two drvng forces of globalzaton. A vast emprcal lterature shows that globalzaton dscplnes governments. It could ncte them to reduce waste and neff cent polces n order to provde a more busness- 1 For example, frms mght under-nvest n order to become unproftable and obtan subsdes (Segal, 1998). 2 Korna, Maskn and Roland (2003) provde a detaled revew of the soft-budget constrant lterature. 3 Even n the European Unon where state ads are forbdden as soon as they threaten to dstort competton by favourng certan undertakngs or the producton of certan goods (artcle 107 of the TFEU), governments can balout frms through the so-called State ads for rescung and restructurng frms n dff culty. Accordng to Chndooroy, Muller and Notaro (2007), 86 rescue and restructurng State ad cases were approved by the Commsson between 1995 and 2003, n varous sectors and mostly n western countres. For addtonnal examples of balouts n developed countres over the perod , see Korna (2009). 2

6 frendly envronment (see. Rodrk, 1997, Schulze and Ursprung, 1999). 4 One could therefore expect that globalzaton moderates the use of corporate balouts. Yet, few theoretcal contrbutons explore ths ssue. An excepton s the contrbuton by Alexeev and Jang (2010), that places Segal s (1998) SBC model wthn Meltz (2003) framework characterzed by trade and heterogeneous frms. The man SBC-nduced neff cency arses from the fact that some frms subject to a balout exert a sub-optmal effort. Trade lberalzaton reduces ths neff cency and the number of frms elgble for a balout by rasng the average level of effort. However, ths contrbuton does not specfcally address whether or not trade costs also nfluence balout decsons. In addton, they abstract from the effects of frm moblty. Another mportant contrbuton by Qan and Roland (1998) explores the determnants of the SBC n the context of federal economes. They show that by nducng fscal competton among local governments, factor moblty ncreases the opportunty costs of balout and then serves as a commtment devce (Qan and Roland, 1998, p. 1143). In sum, whle the latter contrbuton provdes a frst nsght regardng the mpact of captal moblty on corporate balouts, the former explores how trade lberalzaton nfluences the number of frms elgble for such balouts. Our contrbuton goes further by nvestgatng the ssue of corporate balouts n economes characterzed by both captal moblty and nternatonal trade. To do so, we develop a trade and locaton model wth two symmetrc countres and heterogeneous frms. Some frms are domestc and mmoble, whle others are moble and owned by foregn nvestors resdng n a thrd country. They all compete on the same olgopolstc market. Importantly, foregn frms are always proftable because they are run by market-orented managers. In contrast, domestc ones are run by poltcally-connected managers and mght be elgble for a balout. Our model thus captures two mportant features. Frstly, the exstence of poltcally-connected frms and ther hgher probablty to be rescued by the government s documented by Facco, Masuls and McConnell (2006) for a large sample of (mostly) developed countres. 5 Secondly, captal nvested n these frms s often partally state-owned and mght therefore be less footloose than captal nvested n competng multnatonal frms. Fnally, our model exhbts varous nteractons embedded n a sequental game à la Qan and Roland (1998). In a frst stage, poltcallyconnected managers choose to exert ether a hgh or low effort. A hgh effort can be vewed as restructurng the frm. In contrast, a low effort results n fnancal dff cultes for the 4 See. Ca and Tresman (2005) for an opposte vew. 5 Frms are defned as poltcally-connected when at least one of ts large shareholders or one of ts top off cers s a member of parlament, a mnster, or s closely related to a top poltcan or party. 3

7 frm so that t cannot survve wthout a balout. In the latter case, the government has to decde n a second stage whether to rescue the frm through a full exoneraton of the corporate tax (balout polcy) or to let t go bankrupt (lassez-fare polcy). Importantly, we retan a too-bg to fal argument by assumng that poltcal benefts from a balout accrung to the government are proportonal to the sze of frms, or equvalently, to ther total labor force. In a thrd stage, governments non-cooperatvely levy a lump-sum tax on all proftable frms. The two last stages descrbe the locaton choce of foregn nvestors and the market outcome. Our model suggests a non-unvocal relatonshp between trade lberalzaton and the decsons to balout, because trade costs can nfluence the relatve cost of a balout and ts poltcal benefts n varous ways. Frstly, the trade cost level nfluences poltcal benefts through ts mpact on the export status of frm, and, n fne, on ther sze. When trade costs are hgh, a frm elgble for a balout s not compettve enough to export and blateral trade s exclusvely drven by foregn frms set up n each country (trade regme 1). Ths s only once trade costs reach a lower threshold value that the former begns to export as well (trade regme 2). Dependng on whether trade regme 1 or 2 prevals, the mpact of trade lberalzaton on the sze of frms elgble for a balout goes n opposte drectons: t ncreases wth a fall n trade costs when they are exporter whereas t shrnks when they only serve the domestc market because of fercer competton. As a result, trade lberalzaton reduces (resp. ncreases) poltcal benefts from a balout under trade regme 1 (resp. trade regme 2). Secondly, a balout also dstorts the market outcome by sustanng domestc compettors. As argued by Slaughter (2008), the evaluaton of the cost of a balout polcy must account for the way foregn and footloose compettors set up n the country may react to such a dstorton. In our model, a balout polcy mantans a hgher number of domestc compettors on the market, so that moble frms are more responsve to a rse n taxaton. Ths often leads governments to set lower taxes and extract less tax revenues f they rescue frms than f they adopt a lassez-fare polcy. Importantly, ths relatve cost of a balout as compared to a lassez-fare polcy s bgger when frms elgble for a balout become exporters, but t dmnshes wth the gradual declne n trade costs whatever the trade regme. By combnng these effects, we obtan our man fndng that can be summarzed as follows. When trade regme 1 prevals, the gradual declne n trade costs makes the balout polcy more and more lkely. Indeed, the decrease n the relatve cost of a balout does 4

8 more than compensate the fall n ts poltcal benefts. When the two countres swtch from trade regme 1 to trade regme 2, the lassez-fare polcy becomes more lkely because of the sudden rse n the relatve cost of the balout. Lastly, as trade ntegraton deepens further, ncentves to bal out domestc frms unambguously grow because the relatve cost of a balout shrnks whle ts poltcal benefts ncrease. The remander of the paper s organzed as follows. The model s descrbed n secton 2. In the followng sectons, we solve the sequental game by backward nducton. Secton 3 s devoted to the presentaton of producton and consumpton outcomes for a gven spatal dstrbuton of moble frms and for gven polcy decsons. Ths allows us to dentfy the range of trade cost values correspondng to each trade regme and to analyze how competton effects are nfluenced by the decson to rescue frms or not. In secton 4, we descrbe the locaton equlbrum and the tax competton outcome n our benchmark case (trade regme 2), for a gven polcy decson regardng frms elgble for a balout. In secton 5, we analyze the determnants of the balout decson by governments across each trade regme and the last secton concludes. 2 The Model The economy conssts of two countres, labelled = A, B, equally populated by L A = L B = L/2 ndvduals. 6 There are two factors of producton labor and captal and two sectors, whch are always actve n the two countres. Each ndvdual n country provdes one unt of labor and s endowed wth an equal share of captal. These factors are employed n the country of resdence of the ndvdual. The stock of domestc captal s completed by foregn captal owned by ndvduals lvng outsde the economy. Contrary to domestc captal, foregn captal s moble and can be nvested ether n country A or B. Whle labor can ether be employed n a manufacturng sector (M sector) or a tradtonal sector (T sector), captal s only needed n the former. The M sector s characterzed by k olgopolstc frms producng a homogenous good x. Exportng ths good nvolves a per-unt cost of τ unts of numerare. These frms enjoy ncreasng returns to scale. A fxed quantty of captal that we normalze to unty and one local manager are always requred to start producton. Moreover, they may dffer n two respects. Frstly, frms relyng on domestc captal are mmoble whereas those usng foregn captal are moble. 7 Secondly, dependng on the orgn of the captal 6 Countres are assumed to be dentcal n all respects n order to control for any comparatve advantage. 7 The wave of prvatzaton that began n developped countres n the late 1970s s not completely 5

9 of the frm they run, the managers profle s not the same. Managers of foregn frms are market-orented : they always exert a hgh effort resultng n a low margnal cost so that ther frm s proftable. In contrast, domestc frms are run by managers that are connected poltcally wth the government. These managers know that ther frm can potentally be baled-out n case of fnancal trouble. Therefore, they can choose among two levels of effort: a low effort results n a hgh margnal cost of producton so that the frm becomes elgble for a balout and survve f and only f ths balout s mplemented, whereas a hgh effort leads to the same margnal cost as foregn frms and guarantees ther proftablty wthout any governmental support. To summarze, domestc and mmoble frms run by poltcally-connected managers compete wth foregn and moble frms run by market-orented managers on the same olgopolstc market. In the T sector, frms produce the numerare commodty z under perfect competton usng workers only. Specfcally, one unt of labour s requred to produce one unt of output, so that wages are equalzed to one n ths sector. Note that our modellng of the labor market shares many common assumptons wth footloose captal models (see, Henderson and Thsse, 2004) : labor supply s nelastc, mmoble across countres but moble across sectors and the numérare sector s always actve n both countres. We make two addtonal assumptons. For convenence and to avod the modellng of two dstnct labor markets, managers and producton workers are perfect substtutes. It allows all wages to be equalzed to unty as n footloose captal models. Thus, our approach voluntarly abstracts from wage consderatons n order to focus on the role of employment for the balout decson. Secondly, we ensure that whatever the balout decson and the resultng market outcome, the total labor supply s hgh enough to make producton possble. These two assumptons are explctly descrbed n the Appendx A. In the followng, we descrbe n more detal the dfferent types of manufacturng frms (secton 2.1) and the sequence of nteractons between frms and governments (secton 2.2). acheved (see Bortolett and Facco, 2009). As a result, n many countres, captal nvested n former state-owned frms s stll mostly owned by the government and less footloose than captal nvested n competng multnatonal frms. Wth our modellng strategy regardng the manufacturng sector, we try to buld a model that fts ths context. 6

10 2.1 Manufacturng frms and the government We set the total number of domestc frms to 4 (2 n each country). 8 Ther managers are strongly lnked to the government. Therefore, ther behavor depends on ther expectatons of fnancal support they mght receve from polcy-makers. Specfcally, poltcallyconnected managers can choose among two alternatves. The frst one conssts n makng a hgh effort whch can be nterpreted as strongly restructurng. In ths case, the frm enjoys a zero margnal cost, s proftable and thus subject to lump-sum taxaton by the government. Alternatvely, the manager can exert a low effort. In ths case, the margnal cost s postve and normalzed to one labor unt. These hgh cost frms are consdered as elgble for a balout. We model ths balout n a smple way by assumng that t takes the form of a full tax exempton, a common mean to rescue frms (see Korna, Maskn and Roland, 2003, and Shlefer and Tresman, 2000). 9 cost frms go bankrupt. Wthout such a tax exempton, hgh Because of ther poston, poltcally-connected managers also enjoy non-monetary benefts. By non-monetary benefts, we consder all advantages resultng from ther relatonshp wth some members of the government, whch gves them some poltcal nfluence. These benefts are assumed to be hgher, the larger the number of workers employed n the frm. Specfcally, each poltcally-connected managers receves (net of effort) benefts equal to E tmes the number of workers employed. Let l denote the total labor force (ncludng the manager) n a hgh-cost frm. Thus, a poltcally-connected manager receves E l f he exerts a low effort, and E f he exerts a hgh effort snce n ths case l = 1. Ths s n lne wth Qan and Roland (1998) who assume that benefts are hgher when managers exert a low effort and antcpate a balout, n order to ensure that a balout decson can arse at the equlbrum. Moreover, ths modelng strategy s a smple way to render the balout decson endogenous to the number of jobs n the manufacturng sector (see, secton 5). Specfcally, our model captures the too bg to fal argument that contrbutes to explan why governments often choose to rescue the largest frms. The remanng k 4 foregn frms are run by market-orented managers. They feel entrely responsble for the survval of the frm and thus always exert a hgh effort nducng 8 We fx the number of poltcally-connected frms n each country to 2 n order to convey our message n a smple way and to smplfy the algebra. However, we could develop our analyss for any gven number of domestc frms wthout changng our qualtatve results. 9 Tax exemptons fall n one of the categores of means of rescue lsted by Korna, Maskn and roland (2003). Ths category conssts of fscal means, that can ether take the form of subsdes or tax concessons. The two other categores are credt and the varous ndrect methods of support. 7

11 a zero margnal cost. As these frms are proftable, they are subject to lump-sum taxaton by the government of the hostng country. Recallng that captal nvested n these frms s moble, ther captal owners wll thus compare the net profts across the two countres and decde to nvest where the net-return to captal s the hghest. 2.2 Sequence of events Our model contans varous forms of nteractons that are embedded n the followng sequental game (see Fgure 1): Stage 0 [Effort choce of poltcally connected managers]. Poltcally connected managers choose a hgh or low level of effort gven the value of ndvdual non-monetary benefts, and antcpatng the outcome of all subsequent stages. Stage 1 [Balout or lassez-fare polcy]. If a low effort s chosen at stage 0, governments decde whether to rescue hgh-cost frms through a full tax exempton (balout polcy), or to let them go bankrupt (lassez-fare polcy). When dong so, they perfectly antcpate the tax competton outcome, the locaton of foregn frms and the resultng market outcome. Stage 2 [Taxaton of frms]. Governments smultaneously and non-cooperatvely choose the level of lump-sum tax leved on all low-cost frms, gven decsons taken at stages 0 and 1. At ths stage, governments perfectly antcpate the locaton of foregn frms as well as the market outcome. Stage 3 [Locaton of foregn frms]. Foregn captal owners decde whether to nvest n country A or B gven the observed levels of taxaton n each country and antcpatng the market outcome. Stage 4 [Producton and consumpton]. Survvng frms make ther output choces and consumpton takes place. The remanng frms go bankrupt. As we can see from Fgure 1, the sequental game conssts of 4 or 5 stages dependng on the effort choce made by poltcally-connected managers. Importantly, decsons taken at stages 0 and 1 wll lead to three potental decson paths denoted by θ { H, H, S }, where H and H refer to a lassez-fare polcy (or hard-budget constrant) and S stands for the balout scenaro (or soft-budget constrant). Among these three potental outcomes, only decson paths H and S n sold lnes are perfect subgame equlbrum canddates. Let us frst assume that managers perfectly antcpate a balout. Then, as E l > E, they wll prefer to exert a low level of effort because they wll enjoy hgher non-monetary benefts than f they exert a hgh effort. Let us now consder that they antcpate a lassez- 8

12 Fgure 1: Structure of the game fare polcy. Then, t s always optmal for poltcally-connected managers to exert a hgh level of effort as t yelds a postve non-monetary beneft E. Otherwse, the frm goes bankrupt, managers lose ther job and thus receve no beneft. The decson path H n dotted lnes s therefore off the equlbrum path. Nevertheless, we need to evaluate the outcome along ths decson path n order to determne whch one of the decson paths S or H wll be the subgame perfect equlbrum of the game (see secton 5). As ths s well known, the soluton to such a sequental game s gven by a subgame perfect Nash equlbrum that we obtan by backward nducton begnnng wth the last stage of the game. 10 In the followng secton, we analyze how producton and consumpton choces made n the last stage are nfluenced by the balout vs. lassez-fare decson. 3 Short-run equlbrum (stage 4) In ths secton, we present the outcome at the last stage of the game. We call t the shortrun equlbrum as t descrbes the producton and consumpton outcomes for a gven spatal dstrbuton of foregn frms. 10 Because of the symmetry of our model, ths equlbrum outcome s the same n each country. 9

13 3.1 Consumpton Indvduals share dentcal preferences gven by a quas-lnear utlty functon: u θ = ax θ 1 2 ( x θ ) 2 + z θ {A, B} and θ { H, H, S } (1) where x s the consumpton of the manufacturng good, z the consumpton of the numerare and z s the ndvdual endowment n the numerare. Let T stand for tax revenues from all low-cost frms that are redstrbuted equally and n a lump-sum fashon to the consumers n each country. The budget constrant for a representatve consumer n each country s then: 1 + z + T θ L/2 + 2ρl,θ L/2 = z θ + p θ x θ f θ {H} 1 + z + T θ L/2 = zθ + p θ x θ f θ { } H, S where p s the prce of the good produced n the M-sector and ρ l,θ s the after-tax return to captal nvested n domestc frms that are low-cost. Importantly, ndvduals receve no ncome from captal when managers of domestc frms exert a low effort. Indeed, n such a case, two scenaros can occur. If governments adopt a lassez-fare polcy (θ = H), the decson to produce would nduce a negatve net return to captal. Thus, frms go bankrupt and ndvduals earn no captal ncome. If governments nstead decde to rescue these frms through a tax exempton (θ = S), we assume that the balout just allows frms to survve. 11 In our model, t means that the gross return to captal nvested n hgh-cost frms s equal to zero. Utlty maxmzaton leads to the ndvdual nverse demand functon wth respect to the manufacturng good p θ = a x θ (2) Aggregatng the demand over all consumers yelds market demand curves for each country n the olgopolstc ndustry : X θ = L 2 ( a p θ ) 11 For the sake of smplcty, we do not explctly model ther proftablty condton but the balout can be vewed as necessary for the repayment of loans. 10

14 3.2 Producton We assume manufacturng frms compete n quanttes. 12 Before descrbng ther output choces, two comments are n order. Frstly, we pont out that both the effort choce of managers and the balng-out decson of governments affect the number of compettors on the market, and thus the toughness of competton, as shown n Table 1. Table 1: Number and type of frms The frst path occurs when poltcally-connected managers decde to exert a hgh effort gven ther antcpaton of a lassez-fare polcy (θ = H). As a result, there wll be olgopolstc competton among k low-cost frms, among whch 4 domestc frms and k 4 foregn frms. Along the second path, poltcally-connected managers exert a low effort whle the government adopts a lassez-fare polcy (θ = H). Ths mples olgopolstc competton among k 4 foregn low-cost frms. The last path descrbes the balout polcy (θ = S), and gves rse to olgopolstc competton among k heterogenous frms: 4 domestc hgh-cost frms, and k 4 foregn low-cost frms. Secondly, the level of trade lberalzaton shapes output decsons. Indeed, recall that the cost ncurred by a frm for exportng each unt of the manufacturng good s equal to τ unts of the numerare. 13 Because of these trade costs, frms are able to segment ther markets by choosng the quanttes to sell on the domestc and the foregn market ndependently. We are now equpped to descrbe the product market outcome. Let x c,θ and x c,θ j denote the output choces made by a frm located n country, whch depend on the margnal cost, low or hgh (c {l, h}) as well as the number of survvng frms through θ { H, H, S }. Before-tax profts made by low-cost and hgh-cost survvng frms are 12 The same modelng strategy s used, among others, by Gagné and Wooton (2011), Haufler and Wooton (2010), and Thsse (2010). 13 Ths captures all frctons makng blateral trade costly ncludng transport costs or admnstratve barrers to the free moblty of goods between countres. 11

15 descrbed by equatons 3 and 4 respectvely: π l,θ = p θ x l,θ + ( p θ j τ ) x l,θ j 1 rl,θ wth θ { H, H, S } (3) π h,s = ( p S 1 ) x h,s + ( p S j 1 τ ) x h,s j 1 r h,s (4) where r l,θ s the gross return to captal. In the long run, ths return to captal absorbs all operatng profts. Maxmzng (3) and (4) wth respect to quanttes, we get the followng output levels for a low-cost frm: x l,θ = L 2 pθ and x l,θ j = L ( p θ 2 j τ ) wth θ { H, H, S } (5) and for a survvng hgh-cost frm: x h,s = L ( p S 2 1 ) and x h,s j = L ( p S 2 j 1 τ ) (6) Equlbrum prces are obtaned by nsertng the equlbrum output choces (5) and (6) n the market clearng condtons. Let δ θ denotes the share of foregn frms located n country under θ, then equlbrum prces are as follows: [ ] a + τ δ H p H j (k 4) k 3 p H a + τ [ δ H j (k 4) + 2 ] k + 1 p S a + τ [ δ S j (k 4) + 2 ] + 4 (9) k + 1 Three comments are n order. Frstly, observe that the toughness of competton s captured by the denomnator of each expresson (equal to the number of survvng frms plus one). Secondly, p S > p H at the symmetrc locaton equlbrum where δ j = 1/2. Indeed, the average margnal costs s hgher under balout polcy because hgh-cost frms survve. Moreover p H > p H at δ j = 1/2, because the number of compettors s lower when governments let hgh-cost frms go brankrupt. Fnally, as the number of survvng frms resultng from H dffers from the one arsng from S and H, the prce responsveness to the spatal dstrbuton of foregn frms vares accordngly. Specfcally, we verfy that dp H /dδ j > dp H /dδ j = dp S /dδ j. We now determne the trade feasblty condtons ensurng that exportng s always proftable for frms along the two equlbrum paths θ {H, S}. We obtan the followng condton for low-cost frms located n : x l,h j > 0 τ < τ H = 2 a k (7) (8)

16 and the same condton apples for frms located n country j. 14 condton ensurng that hgh-cost frms export s gven by: The trade feasblty wth τ S > 0 for all a > a mn k x h,s j > 0 τ < τ S = 2 a k + 3 k + 2 Clearly, these two threshold values can be ranked n the followng manner: 16 τ S < τ H Thus, the level of trade costs affects the dstrbuton of survvng frms on the export market as stated by Lemma 1. Lemma 1 Let us defne trade regme 1 by τ ( τ S, τ H) and trade regme 2 by τ ( 0, τ S). Then: Under trade regme 1, low-cost frms are exporters whle hgh-cost frms only serve ther domestc market; Under trade regme 2, all frms serve both markets. Startng from the hghest possble level of prohbtve trade cost (that s, τ H ), trade lberalzaton wll frst allow low-cost frms to export and t s only once trade costs reach a lower threshold value (that s, τ S ) that hgh-cost frms rescued by the government wll also begn to export. 17 Therefore, we explore the effects of lberalzaton over a wder range of trade cost values than n Okubo, Pcard and Thsse (2010), who focus on the spatal selecton of frms nduced by trade lberalzaton under trade regme The model beng symmetrc, all trade feasblty condtons are evaluated at δ θ = δ θ j = 1/2 and apply also to frms located n country j. 15 Hereafter, ths condton s consdered as fulflled. The demand parameter a s hgh enough to make the export by baled-out frms dependent on the trade cost level. 16 The condton ensurng that low-cost frms are proftable on the export market wrtes τ < 2a/ (k 2) τ H. Ths s the less restrctve condton as we verfy that τ H s always hgher than τ S and τ H. 17 Note also that p S < p H when τ < τ S. 18 In ther paper, all frms are moble and export whatever ther cost level and the market sze dffers across countres. 13

17 4 Tax Polcy and Locaton under Trade Regme 2: Lassez-fare vs. Balout In the followng subsectons, we consder trade regme 2 as a benchmark case to solve both the locaton and the tax competton equlbra. Ths wll allow us to compare the mpact of a lassez-fare versus balout polcy on the locaton of foregn frms (secton 4.1) and on the tax competton outcome (secton 4.2). Trade regme 1 wll be analyzed n secton 5, where we solve the stages 1 and 0, and study the effect of trade lberalzaton on the balout decson for both trade regmes. 4.1 Locaton equlbrum (stage 3) In the long run, the equlbrum rental rate to foregn captal n each country s determned by a bddng process, whch ends when no foregn frm can earn a strctly postve proft at the equlbrum market prce. Therefore, before taxaton, the equlbrum rental rate to captal wrtes: r l,θ r h,s = p θ x l,θ + ( p θ j τ ) x l,θ j 1 wth θ { H, H, S } = ( p S 1 ) x h,s + ( p S j 1 τ ) x h,s j 1 As governments non-cooperatvely mpose a lump-sum tax t θ on low-cost frms set up wthn ther respectve jursdctons, the after-tax return to captal nvested n low-cost frms n country s gven by: ρ l,θ = r l,θ t θ (10) Owners of captal nvested n foregn frms decde to nvest n the most proftable country. Therefore, the locaton ( of foregn ) ( frms s governed ) by the spatal dfference n net returns to captal l,θ = r l,θ A tθ A r l,θ B tθ B. After nsertng equlbrum prces and quanttes n (10), we obtan: ( 2δ θ t B t A Lτ 2 A 1 ) (k 4) for θ {H, S} l,θ = ( k + 1 2δ θ t B t A Lτ 2 A 1 ) (k 4) for θ = k 3 H The locaton equlbrum for each θ can be defned as the share of foregn frms located n country A (δ θ A) such that θ = 0, that s: 14

18 δ H A = k 3 2 Lτ 2 (k 4) (t A t B ) δ H A = δ S A = (11) k Lτ 2 (k 4) (t A t B ) The above locaton equlbra are the result of two forces. The frst one s standard and depcts a pro-compettve effect. When a country hosts new frms, ncumbent frms face more compettors n ther domestc market and fewer n the other one. Thus, the domestc prce falls whle t rses n the other market (see equatons 7, 8 and 9). Because domestc sales generate more revenues n the presence of trade costs, ths effect acts as a dsperson force. Importantly, ths effect s proportonal to the number of survvng frms and s captured by the terms k 3 and k + 1. Ths number beng the same under θ (H, S), the locaton equlbrum s dentcal. The second force results from the mpact of the tax wedge on the locaton choce. 19 A unlateral rse n corporate taxaton n country leads to an outflow of captal (dδ θ /dt θ < 0). Moreover, as competton s fercer when θ {H, S}, frms are more responsve to tax varatons and the tax base eroson effect s stronger. We also verfy that the tax base elastcty (defned as ε θ = δ θ / t θ t θ /δ θ ) ncreases when trade costs fall because prces become less and less responsve to the spatal dstrbuton of frms. In other terms, gradual trade ntegraton weakens the pro-compettve effect whch ncreases the weght of taxes n the captal locaton choce. 4.2 Tax competton (stage 2) Governments decde non-cooperatvely and ndependently upon ther tax polcy. assume that ther objectve s to maxmze: We W θ = R θ + ϕe θ where R θ denotes the overall net ncome of resdents. E θ stands for total non-monetary benefts receved by poltcally-connected managers and ϕ > 0 s a parameter amed at capturng how much the government cares about them. Interestngly, as E θ ncorporates the labor force employed by domestc frms n the manufacturng sector, ϕ can also be vewed as the extent to whch governments care about those jobs. 19 Despte ther lump-sum form, taxes dstort the nvestment choce through the spatal dstrbuton of foregn frms. 15

19 Each one of the component of ths objectve functon vares across the decson paths θ. The overall ncome of resdents s gven by: R H = L 2 (1 + z) + th [ (k 4) δ H + 2 ] + 2ρ l,h R θ = L 2 (1 + z) + tθ [ (k 4) δ θ ] for θ { H, S}. The frst term stands for the labor ncome and the ntal endowment n numerare. The second terms denote tax revenues redstrbuted n a lump-sum fashon to resdents, and the thrd term n R H frms (see eq. 3). stands for the net return to captal nvested n low-cost domestc In order to descrbe total external benefts accrung to poltcally-connected managers, recall that x h,s + x h,s j gves the number of producton workers employed by domestc hgh-cost frms. Total employment n each frm ncludng the manager s equal to l = 1 + x h,s + x h,s j when θ = S whereas t s equal to 1 n domestc low-cost frms when θ = H. 20 Therefore, we obtan: E S = 2E E H = 2E ( 1 + x h,s ) + x h,s j In the followng, we present the tax outcome under lassez-fare (secton 4.2.1) and balout (secton 4.2.2) polces Tax equlbrum under lassez-fare polcy We present the tax outcome when poltcally-connected managers exert a hgh effort and a low effort. Poltcally-connected managers exert a hgh effort Let us frst assume that poltcally-connected managers exert a hgh effort n stage 0. All frms beng eff cent and proftable, no publc nterventon s needed n stage 1 whch s equvalent to a lassez-fare polcy (θ = H). Hence, the frst-order condton for each government at stage 2 wrtes: dw H dt H 1 ) = δ H (k 4) ( ε H }{{} +/ ρl,h t H }{{} = 0 (12) 20 Recall that there are no external benefts when θ H snce domestc hgh-cost frms go bankrupt. 16

20 The frst term captures the taxaton effect passng through tax revenues collected from foregn frms. For a gven tax base, a margnal ncrease n t H rases tax revenues redstrbuted to resdents. Nevertheless, ths effect can be counterbalanced by a tax base effect capturng the captal outflow that results from a margnal ncrease n t H. The total effect wll be negatve (resp. postve) f the tax base elastcty s hgher (resp. lower) than 1 n absolute value. 21 The second term gves the tax revenues collected from domestc frms. The last term captures the effect on the net return to captal nvested n domestc frms. Evaluated at the locaton equlbrum, ths net return to captal amounts to: wth ρ l,h = r l,h 1 2 (t + t j ) + 1 (t t j ) 2 (13) 4 Lτ 2 r l,h = 1 4 L4a (a τ) + τ 2 (k (k + 2) + 2) (k + 1) 2. Observe that both a hgher average tax rate and a lower tax wedge between countres reduce the after-tax return on captal, as n Ottavano and Van Ypersele (2005). Dervatng (13) wth respect to t H, we verfy that a unlateral rse n busness taxaton n country always reduces the net return to captal ( ρ l,h / t H < 0). Solvng the frst order condton for each government and crossng the reacton functons, we obtan: t H t H = Lτ 2 k 2 k + 1 Poltcally-connected managers exert a low effort {A, B} (14) Let us now consder that poltcally-connected managers exert a low effort n stage 0 and governments choose the lassez-fare polcy n stage 1. Gven that all domestc frms ext the market, there s no ncome from domestc captal and government s objectve functon sums up to maxmzaton of tax revenues from foregn frms. Therefore, at stage 2, the government s frst-order condton s: dw H dt H ( ) = δ H ε H (k 4) 1 }{{} +/ Interestngly, the mpact of a margnal rse n busness taxaton on tax revenues has a dfferent magntude than when θ = H. Indeed, the captal outflow nduced by a margnal rse n taxaton (and thus, the tax base elastcty) s now lower because competton s relaxed by the lqudaton of hgh-cost frms. 21 We easly check that ε H > 1 at the equlbrum. 17 = 0

21 After solvng the frst-order condtons, we obtan the followng tax equlbrum: t H t H = Lτ 2 k 4 k Tax equlbrum under balng-out decson {A, B} (15) We now turn to the tax outcome that occurs f governments decde to balout frms n fnancal trouble at the prevous stage of the game (θ = S). The frst-order condton of each government at stage 2 s gven by: dw S dt S = δ S (k 4) ( ε S 1 ) }{{} +/ + 2ϕE dl = 0 dt S }{{} 0 The frst term of the frst-order condton s qualtatvely and quanttatvely smlar to the one arsng when θ = H, because locaton equlbra are dentcal. The second term descrbes a new relatonshp between tax polcy and the total non-monetary benefts arsng from the government s balout. Interestngly, ths ncentve exsts f and only f trade costs are hgher than the threshold τ S (trade regme 1). In ths case, the total requrement n producton labor of a rescued frm s equal to domestc sales only (x h,s ). By rasng the busness tax, the government nduces a captal outflow that relaxes competton on the domestc market and ncreases the output per baled-out frm. However, ths relatonshp dsappears when τ < τ S, whch corresponds to trade regme 2 under scrutny here, because the sum of domestc sales and foregn ones (x h,s on the spatal dstrbuton of foregn frms. + x h,s j ) does not depend Solvng the frst order condton for each government and crossng the reacton functons, we obtan: Tax comparson t S = Lτ 2 k 4 k + 1 {A, B} and τ < τ S (16) Let us now comment on the propertes of the tax outcome. Frst, observe that equlbrum taxes are always ncreasng and convex n the level of trade costs : dt θ /dτ > 0 for all τ > 0 and d 2 t θ /dτ 2 > 0 for all θ. Low trade costs weaken the pro-compettve effect and make footloose frms more responsve to busness tax dfferences. Ths encourages governments to engage n a race to the bottom n taxaton. Ths s a standard result n the tax competton lterature assumng mperfectly ntegrated and mperfectly compettve economes (see, among others, Ottavano and Van Ypersele 2005, Haufler and Wooton, 2010, Gagné and Wooton, 2011). 18

22 More mportantly, the level of taxaton at the equlbrum s also shaped by the effort made by poltcally-connected managers n stage 0 and the balng out decson n stage 1. From (14), (15) and (16), we can rank the tax equlbra n the followng way under trade regme 2: t H > t H > t S > 0 τ < τ S (17) Two mechansms contrbute to explan ths rankng. Recall that a gven margnal ncrease n busness taxaton n a country leads to a lower captal outflow when θ = H than when θ = H because competton s weakened by the ext of hgh-cost frms n the former case. Consequently, governments are more able to rase taxaton. Ths effect contrbutes to explan why t H > t H. The sgn of the dfference between t H and t S s a pror not obvous. On the one hand, governments have an addtonal ncentve to cut taxes when θ = H n order to rase the net return to captal nvested n domestc frms (see eq. 12). On the other hand, domestc frms beng an mmoble tax base when θ = H, governments are more ncted to rase taxes. The fact that t H > t S suggests that the latter effect domnates. Gven the symmetry of the model, ths result mples that for all τ < τ S, governments collect more tax revenues under a lassez-fare polcy than when they decde to balout neff cent frms. We are now equpped to analyze the bal-out decson of governments (stage 1). Ths decson beng perfectly antcpated by poltcally-connected managers, we can determne ther optmal effort (stage 0) and select the subgame Nash perfect equlbrum accordngly. 5 Balout decson and trade costs We frst present the government s trade-off between the balout and the lassez-fare polces (subsecton 5.1). Ths secton beng dedcated to the effect of trade ntegraton, we go beyond the benchmark case we consdered so far (that s, trade regme 2) and descrbe how a shft from trade regme 1 to trade regme 2 mpacts the equlbrum polcy decson. Sad dfferently, we analyze whether or not governments are more prone to adopt a balout polcy when trade ntegraton s so deep that hgh-cost frms become exporters. Then, the last subsecton (subsecton 5.2) completes the analyss by dscussng the mpact of a gradual declne n trade costs wthn each trade regme. 19

23 5.1 The government s problem The choce between a lassez-fare polcy and a balout makes sense f and only f poltcallyconnected managers exert a low effort. In such crcumstances, the government selects the polcy opton maxmzng the overall gan of resdents ncludng non-monetary benefts accrung to managers, by antcpatng all the effects of ts choce on the rest of the game (stages 2, 3 and 4). If ths overall gan of resdents W θ s hgher when θ = S than when θ = H, a balout s preferred to the lassez-fare polcy and poltcally-connected managers exert a low effort n order to enjoy a hgher level of benefts (as E.l > E). Therefore, the decson path S s the subgame perfect equlbrum of the game. Otherwse, the government adopts a lassez-fare polcy and poltcally-connected managers exert a hgh effort. In that case, the decson path H s the subgame perfect equlbrum. Solvng the government s problem therefore requres to evaluate the dfference n the overall gan of resdents between a balout polcy and a lassez-fare polcy. 22 If W S denotes the overall gan of resdents that results from a balout polcy when τ (τ S, τ H ), ths dfference amounts to: W S W S W H = 1 (k 4) ( t S 2 t H ) + 2ϕE l under trade regme 1 W H = 1 (k 4) ( t S 2 t H ) (18) + 2ϕEl under trade regme 2 wth t S the equlbrum corporate tax and l the equlbrum labor demand of a hgh-cost frm that occur under a balout polcy for all τ (τ S, τ H ) (see next subsecton). Note that the outcome at stages 4, 3 and 2 when θ = H (and therefore W H ) s the same across the two trade regmes as hgh-cost frms ext the market. Therefore, the trade regme nfluences the dfference n the overall gan of resdents only through the outcome n the presence of a balout decson ( W S W S ). Moreover, whatever the trade regme, the dfference n the overall gan of resdents s decomposed n two terms. The frst term s the dfference n tax revenues resultng from the two polcy optons, that depends on the tax dfferental between the two polcy optons. Therefore, t stands for the relatve cost of a balout polcy as compared to a lassez-fare polcy. In subsecton 4.2.3, we showed that t S < t H under trade regme 2. We extend the analyss to regme 1 n next subsecton and show that the sgn of t S t H can be ether postve or negatve. The second term captures the total poltcal benefts arsng from the amount of jobs saved thanks to the balout. Through ths term, our model ratonalzes the too-bg-to-fal motve for a balout. Ths term s postve whatever the trade regme and proportonal to 22 Ths dfference s evaluated at the equlbrum values of taxes and output, as governments perfectly antcpate the outcomes of stages 2, 3 and 4. 20

24 the total labor force employed n domestc frms. Therefore, governments are more lkely to rescue domestc manufacturng frms when the number of jobs nvolved s large. Importantly, the level of trade ntegraton nfluences the government s trade-off (18) n two respects. Frst, a dscrete shft from trade regme 1 to trade regme 2 nfluences the outcome of a balout decson and then, the relatve costs and benefts of each polcy opton. We explore these effects n the next two subsectons. Secondly, a gradual declne n trade costs wthn each trade regme affects the magntude of the relatve cost of each polcy decson as well as the amount of total external beneft resultng from a balout. Ths mpact of gradual trade lberalzaton s analyzed n secton Balout decson under trade regme 1 Let us assume that trade costs le wthn the range (τ S, τ H ), so that low-cost frms partcpate to blateral trade whereas hgh-cost frms serve ther domestc market only. In order to determne whether θ = S or θ = H wll be the equlbrum decson path, we frst need to present the outcome at stages 4, 3 and 2 of the game (hereafter p S, δ S, t S ) n the presence of a balout. In the short-run, equlbrum prces are gven by: p S = a τ δ S j (k 4) k 1 for all τ (τ S, τ H ) (19) In the long-run, the share of moble frms set up n country A for gven tax polces wrtes: level: δs A = k 1 2 Lτ 2 (k 4) (t A t B ) for all τ (τ S, τ H ) (20) Solvng the tax competton game at stage 1, we obtan the followng tax equlbrum t S = Lτ 2 k 4 k 1 + ϕ τl k 1 E {A, B} for all τ (τ S, τ H ) It s worth stressng that t S > t S. There are two reasons for that. Frstly, baled-out frms beng not compettve enough to export when τ (τ S, τ H ), competton s relaxed for foregn frms whch are less responsve to a gven rse of taxaton ( dδ S A/dt A ). > d δs A/dt A Secondly, the busness tax n each country now depends postvely on E. To understand ths result, recall that the total non-monetary benefts accrung to poltcally-connected managers are proportonal to the total output and jobs of the frms they manage. When trade regme 1 prevals, ther total output bols down to output on ther domestc market ( x h,s ). Ths output x h,s s obvously ncreasng n the domestc prce level, but the latter goes down wth the share of foregn frms located n the country (see eq. 19). Ths creates a new ncentve to rse taxes as the resultng captal outflow relaxes competton on the 21

25 domestc market, expands producton from domestc frms and therefore ncreases total non-monetary benefts. As a consequence, busness taxes under trade regme 1 may become hgher n the presence of a balout than under a lassez-fare polcy. As the followng expresson shows, ths s more lkely to happen when E and/or ϕ reach hgh values: t S t H (k 3) ϕe 2τ (k 4) = Lτ 0 (k 1) (k 3) 2τ (k 4) when E ϕ (k 3) = Ẽ for all τ (τ S, τ H ) (21) Recallng that t H > t S when τ (0, τ S ), the followng proposton summarzes our results. Proposton 2 Consder the tax competton subgame. For ntermedary trade costs (τ (τ S, τ H )), corporate taxes are hgher under balout than under lassez-fare f and only f the amount of non-monetary benefts per job saved s hgh enough (.e., E > Ẽ). For low trade costs (τ < τ S ), corporate taxes are hgher under a lassez-fare polcy whatever the amount of non-monetary benefts per job saved. Now let us analyze government s behavor at stage 1. Gven the nequalty (21), the S frst component of W W H (see eq. 18) can be vewed as the opportunty cost of the balout when E < Ẽ, and the opportunty cost of the lassez-fare polcy when E > Ẽ. As the second term s always postve, there s no trade trade-off for governments when E > Ẽ: they always choose the balout polcy. Ths decson beng perfectly antcpated by poltcally-connected managers, they choose a low effort. However, when E < Ẽ, governments are faced wth a trade-off between the loss n tax revenues and the poltcal benefts resultng from a balout. Specfcally: W S W H f and only f E Ê wth Ê = Lτ 2 (k 4) 2 ( ) ϕ (L ((a + 2) + (k 4) τ) + (2 (k 1))) (k 3) 0; Ẽ The above nequalty stpulates that poltcal benefts from a balout more than compensate for the loss n tax revenues for all E (Ê, Ẽ). In ths case, poltcally-connected managers antcpate a balout n stage 1 and adopt a low effort. In contrast, governments wll adopt a lassez-fare polcy and poltcally-connected managers wll choose a hgh effort when E < Ê. To understand ths result, one has to analyse the mpact of E on the relatve cost of a balout and ts poltcal benefts. On the one hand, a declne n E reduces 22

26 the relatve cost of a balout as the tax gap (21) s proportonal to E. However, on the other hand, the reducton of per-job non-monetary beneft also reduces the total amount of poltcal benefts accrung to managers. 23 The former effect s predomnant snce we show that a balout equlbrum s more lkely to occur when the beneft per job saved s hgh Balout decson under trade regme 2 Let us now assume that trade costs fall below the threshold τ S so that hgh-cost frms start exportng as well. By replacng t H, t S, x h,s (18), we get: W S W H when E and x h,s j by ther equlbrum values nto 4Lτ 2 (k 4) 2 ϕ (k 3) ((2a 2k τ + 6) L + 2 (k + 1)) = Ē (22) Above a threshold Ē, governments decde to bal-out domestc frms and poltcallyconnected managers choose a low effort. Otherwse, the decson path H descrbes the subgame perfect equlbrum. The ntuton for ths result s the followng. A rse n the level of E ncreases total poltcal benefts, whereas t does not have any mpact on the relatve cost of a balout as nether t H nor t S depend on E. Thus, governments are more prone to balout frms when E rses because t ncreases total poltcal benefts from the balout wthout any mpact on ts relatve cost. Usng the expressons of Ē and Ê, t s now possble to analyze to what extent the equlbrum outcome depends on the level of trade costs. We easly check that Ê < Ē. (see Appendx). Ths leads us to formulate the followng lemma. Lemma 3 The polcy choce of governments depends on non-monetary benefts of managers as follows: f E < Ê, governments choose the lassez-fare polcy and poltcally-connected managers exert a hgh effort whatever the trade regme; f E (Ê, Ē), governments bal out hgh-cost frms and poltcally-connected managers exert a low effort under trade regme 1, whle the opposte result apples under trade regme 2; 23 Gven the symmetry of the model, equlbrum taxes are equalzed across countres and the locaton equlbrum bols down to δ S = 1/2. As a consequence, the ndvdual output of a hgh-cost frm x h,s s not nfluenced by the amount of external beneft per job saved. The latter only nfluences poltcal benefts at the country level, as they are proportonnal to the number of frms run by poltcally-connected managers. 23