1 Trade and Foregn Drect Investent n Chna: A Poltcal Econoy Aroach by Lee G. Branstetter Det. o Econocs, Unv. o Calorna, Davs and NBER and Robert C. Feenstra Det. o Econocs, Unv. o Calorna, Davs, Haas School o Busness, Unv. o Calorna, Berkeley, and NBER Revsed, March 999 Acknowledgeents: We thank Chen Chun-La, Wen Ha, Davd L, Barry Naughton, Danel Rosen, Scott Rozelle, Edward Steneld, Wlla Sknner, Erc Thun, Shang-Jn We, Wng Woo, Alwyn Young and senar artcants at the Unversty o Mchgan and the World Bank or helul coents and suggestons. We are ndebted to a large tea o research assstants, ncludng Natasha Hseh, Songhua Ln, Kaoru Nabesha, Shun-l Yao, and Davd Yue. Branstetter wshes to thank the Jaan Foundaton Center or Global Partnersh and the Socal Scence Research Councl or nancal suort o the eldwork coonent o ths roject, and he thanks the any anonyous Chna-based ultnatonal eecutves who rovded ther nsghts concernng the envronent conrontng oregn rs n Chna. He also acknowledges the suort o a Unversty o Calorna Faculty Research Grant. Feenstra thanks the Natonal Scence Foundaton or nancal suort. All errors are the sole resonsblty o the authors.
2 Trade and Foregn Drect Investent n Chna: A Poltcal Econoy Aroach Abstract We vew the oltcal rocess n Chna as tradng o the socal benets o ncreased trade and oregn drect nvestent, aganst the losses ncurred by state-owned enterrses due to such lberalzaton. A odel drawng on Grossan and Helan (994, 996) s used to derve an ercally estable governent objectve uncton. The key structural araeters o ths odel are estated usng rovnce-level data on oregn drect nvestent and trade lows n Chna, over the years We nd that the weght aled to consuer welare s between one-th and one-twelth o the weght aled to the outut o state-owned enterrses. We nd that governental reerences have shted over te, but even n recent erods the weght on consuer welare s only one-hal o the weght on state-owned enterrses. Ths suggests that Chna ay nd t oltcally dcult to ollow through wth lberalzng ts trade and nvestent reges, such as under ts WTO accesson roosal. Lee Branstetter Robert Feenstra Deartent o Econocs Deartent o Econocs Unversty o Calorna Unversty o Calorna Davs, CA 9566 Davs, CA 956 (530) (530)
3 . Introducton In recent years, ew develoents n nternatonal econocs have elcted ore nterest or strred ore controversy than the sudden eergence o Chna as a tradng naton and anuacturng center. Chna s hgh rate o econoc growth snce the adoton o ore lberal econoc olces under Deng Xao-Png n the late 970s, whle erhas eaggerated by the ocal statstcs, s nevertheless a stunnng accolshent. 2 In the sace o less than a generaton, Chna has transored tsel ro a oor naton alost coletely cut o ro the global econoy to one o the world s ost ortant sulers o labor-ntensve anuactures. Its arrval nto the ranks o tradng natons wll be colete when t s has acheved ts ntenton to jon the World Tradng Organzaton (WTO). Nevertheless, a nuber o econosts have recently called nto queston the sustanablty o Chna s current rates o GDP and eort growth. These observers have tended to ont out the shortcongs o Chna s socalst arket econoy, ocusng on the nablty o the governent to rase revenue by tang the new rvate econoy, the robles the governent has encountered n reorng state owned enterrses, the dstortons created by Chna s dualstc trade rege, and the ountng non erorng loans roble n the Chnese bankng sector. 3 In a strkng aer, Alwyn Young (997) has recently argued that Chna s hgh trade/gdp rato s a sgn o ntra-natonal trade barrers rather than nternatonal econoc oenness. Partcularly snce the onset o the East Asan nancal crss, the consensus has shted to a ore essstc outlook or Chna. The realty o ortant dstortons n the Chnese econoy, and artcularly n ts nternatonal trade rege, s ncreasngly evdent. We oer, n ths aer, what s to our For an ecellent overvew o ths eergence, see Naughton (996).
4 2 knowledge the rst attet to use the theoretcal odels o oltcal econoy oneered by Grossan and Helan (994) to analyze these dstortons. We vew the oltcal rocess n Chna as tradng o the socal benets o ncreased trade and oregn drect nvestent, aganst the losses ncurred by state-owned enterrses due to such lberalzaton. We wrte down an elct theoretcal odel o the olcy orulaton rocess whch, we argue, catures ortant eatures o Chnese nsttutonal realty. We then use ths odel to derve an ercally estable governental objectve uncton. The key structural araeters o ths odel are estated usng rovnce-level data on oregn drect nvestent and trade lows n Chna n the years These estates allow us to quanty the relatve weghts laced on secal nterests and general econoc welare. In strkng contrast to Goldberg and Magg s (997) results wth U.S. data, we nd quanttatve evdence that Chnese rovncal and central governents have gven uch greater weght to the secal nterests o state-owned enterrses than to general econoc welare. Moreover, our ercal raework allows us to estate how the weghts n the governent s objectve uncton have shted over te. In other words, we can quanty the etent to whch Chnese oltcal econoy has becoe ore arket-orented over the course o the reor erod. Fnally, our raework allows us to estate the act on governental utlty o roosed changes n Chna s trade rege. As an eerent, we estate the act o the tar reductons Chna has oered as art o ts bd or WTO accesson. The lcatons o these ndngs or the oltcal econoy lterature and or the debate over econoc reor n Chna are dscussed n the concluson. The roos o roostons are gathered n a searate Aend. 2 See Sachs and Woo (997) or a dscusson wth soe reerence to the robles o ocal Chnese statstcs. 3 Agan, Naughton (996) rovdes an ecellent overvew o these ssues.
5 3 2. Polces towards Foregn Drect Investent and Trade A corehensve descrton o Chnese econoc reor, even one ocused solely on the evoluton o Chna s trade and FDI rege, s well beyond the scoe o ths aer. The reader s drected to Lardy (992) and Naughton (995) or two well-regarded studes. 4 Here, we quckly suarze soe o the ost ortant olcy shts that drectly relate to FDI and trade. The oral ban on oregn drect nvestent n Chna was lted n 972, n the wake o the vst o U.S. Presdent Rchard Non to Chna. Ths oened the doors or a dloatc rarocheent between Chna and several ajor ndustral natons. However, a nuber o severe restrctons on FDI reaned n lace ncludng a ban on eternal nancng o FDI rojects so that there was very lttle nward nvestent untl olces were draatcally changed n 979. By 979, Deng Xao-Png had consoldated hs ower wthn the Chnese governent and the Counst Party. At hs behest, a new Law on Jont Ventures was assed, rovdng a basc raework under whch oregn rs were allowed to oerate. Restrctons on eternal debt and equty nance were relaed, and controls on oregn trade were reduced. Provncal and local governents were allowed consderable reedo n regulatng the jont ventures that were establshed wthn ther jursdctons. In the sae year, our secal econoc zones (SEZ) were establshed n whch oregn rs were oered reerental ta and adnstratve treatent, and gven an unusually ree hand n ther oeratons. 5 In artcular, these zones charged a reduced ta o 5% on busness ncoe o oregn alated rs (as coared to 33% or doestc rs), but these taes were not leved durng the rst two years o oeraton 4 The ateral n ths secton also draws on Grub and Ln (99) and Poret (99). 5 These SEZs ncluded Shenzhen (across the border ro Hong Kong), Zhuha (across the border ro Macau), Shantou (on the Guangdong coast acng Tawan) and Xaen (drectly across the Tawan Strats ro Tawan).
6 4 and charged at one-hal o the ull rate n the thrd through th years. 6 In addton, wholly oregn owned enterrses were ertted or the rst te. It s ortant to recognze that none o the orgnal secal econoc zones were develoed ndustral centers n 979. In act, these zones were establshed outsde the state s ndustral centers to revent contanaton o Chnese heavy ndustry by outsde nluences. These eerents n attractng oregn drect nvestent were qute successul. In 984, ourteen addtonal areas known as oen coastal ctes, were granted slar eetons ro taes and adnstratve rocedures n a bd to attract FDI. 7 These ctes leved a ta rate o 24% on oregn alated rs, and were granted local authorty to arove oregn nvestent or rojects under $30 llon (now $50 llon). The end o the 970s and early 980s also arked a erod o roound reor n Chnese trade olcy. U through the 970s, Chnese trade was essentally onoolzed by twelve state-owned tradng coanes, each wth resonsblty or a derent set o coodtes. Trade took lace wthn the contet o a lanned econoy and thereore all trade was subject to very eactng quanttatve barrers. The state onooly on oregn trade was ended at the end o the 970s, and what cae to relace t by the d 980s were, n essence, two searate tradng reges one or the oregn-nvested enterrses (FIEs) and one or natve Chnese enterrses. 8 The trade rege under whch the FIEs oerated was by ar the ore lberal. Unlke doestc enterrses, whch had to ort and eort through state-owned oregn trade coanes, FIEs were allowed to engage n nternatonal trade drectly. In addton, eort-orented FIEs were allowed duty-ree ort o raw aterals, coonents, and 6 Naughton (996,. 30). The dscreancy between the cororate ta or oregn-alated and doestc rs has led any Chnese rs to seek oregn artners, soetes through settng u cororatons n Hong Kong. Ths so-called round trng accounts or soe (unknown) racton o the oregn nvestent nto Chna. 7 The entre sland o Hanan was declared a secal econoc zone n Borrowng ro the Chnese ternology, Naughton (996) reers to these two reges as the eort rocessng and ordnary trade reges, resectvely and he rovdes an ecellent dscusson o the.
7 5 catal equent or eort roducton. They also qualed or concessonary ncoe ta rates and ta holdays. As noted, doestc rs stll had to trade ndrectly, obtanng orts and channelng eorts through oregn trade coanes (FTCs). 9 To be sure, the orgnal twelve onooles were broken u, and by the late 980s there were ore than 5,000 FTCs oeratng throughout Chna. However, ost o these FTCs were lted to trade n certan coodtes, oeraton n a artcular rovnce, and were only allowed to servce certan knds o custoers. Thus, doestc enterrses seekng to nteract wth the global econoy had to work through a legally andated, ar ro erectly coettve layer o ddleen. On to o ths, ost doestc orts are not duty-ree, so that doestc rs have had to contend wth Chna s oral and noral barrers to trade. Whle tars have coe down over the course o the 980s and 990s, they are stll hgh. On to o these tar barrers, soe 20 ercent o Chnese orts are subject to oral non-tar barrers The net ajor regulatory change n FDI cae n 986, wth the leentaton o the so-called Twenty-two Regulatons. These changes reresented a ajor lberalzaton whch aled throughout Chna. Foregn nvested enterrses were ade elgble or reduced busness ncoe ta rates regardless o locaton, and were gven ncreased anageral autonoy. Tght controls on the rettance o rot n oregn currences were lted. In addton, the Twentytwo Regulatons desgnated two categores o oregn nvestents as beng elgble or addtonal secal benets eort orented rojects (dened as rojects eortng 50% or ore o ther roducton value) and technologcally advanced rojects (dened as rojects whch ugrade doestc roducton caacty through the use o advanced technology). 9 Over te, soe larger enterrses were granted ndeendent ort rghts, but ths was generally restrcted to larger enterrses.
8 6 The Twenty-two Regulatons also set u an aroval rocess or oregn drect nvestent rojects whch reans n lace today, albet wth soe odcaton. The aroval rocess s a cole one nvolvng thrteen agences and branches o the central governent as well as local Foregn Econoc Relatons and Trade Cossons. Whle the oral regulatory raework les substantal centralzaton o ower over the aroval rocess and subsequent regulatory oversght o FIEs, there s consderable debate aong Chna scholars as to how uch the central governent ntervenes n the oversght o FIEs ater they are establshed. In ractce, there sees to be a consderable degree o de acto local autonoy n regulatng FIEs. The net ajor sht n FDI n Chna arked not so uch a regulatory sht as a change n the cooston o oregn nvestors. FDI n Chna slowed brely ater the events o Tananen square n 989, but resued and quckly grew n the 990s. 0 Whereas FDI n Chna n the 980s had been overwhelngly donated by Hong Kong and Tawan-based nvestors seekng to elot relatvely low cost labor n the SEZs or eort rocessng, n the 990s FDI n Chna ncreasngly conssted o nvestents by Western and Jaanese ultnatonals seekng to serve the Chnese arket through local roducton caacty. It s also ortant to ont out that FDI nlows n the ost-990 area were uch larger than those o the 980s, both n absolute quanttes and as a ercentage o GDP. Fnally, n the later 980s and 990s, FDI began to low n sgncant quanttes nto the centers o Chna s (state-owned) heavy ndustry anuacturng and nance. O these, Shangha and the neghborng rovnces are the clear standouts. Fgure llustrates the cuulatve stock o FDI nto Chna s 30 rovnces (not countng the Hong Kong Secal Adnstratve Regon) n 995. The varaton across regons s qute 0 Naughton (996), aong others, suggests that there was a de acto loosenng o the ocal regulatons on oregn drect nvestent whch allowed ultnatonals to skrt the ocal eort requreents. Essentally, eort requreents were ncreasngly gnored or the denton o a technologcally advanced roject was broadened to allow even not artcularly technology-ntensve rs to set u lants to serve the Chnese arket.
9 7 strkng, wth Guangdong rovnce, the ste o three o the ntal our SEZs, and neghborng Fujan, the ste o the ourth, antanng donant ostons as the ost ortant stes o FIE actvty. On the other hand, Shangha and the surroundng rovnces have receved very substantal nlows startng ro a very low base n the late 980s, such that they have collectvely becoe the net-largest recent area by Fr Intervews Gven the enorous varatons n central governent olcy and regonal governent leentaton across rovnces and across te, we elt t was ortant to obtan soe drect qualtatve noraton concernng the regulatory envronent on the oeratons o ultnatonal rs. To ths end, one o the authors traveled to Chna n the wnter o 998, ntervewng eecutves and orer eecutves o ore than 20 ultnatonals, consultng rs, research nsttutons, and advsory agences n Chna. We have also consulted wth Chna eerts n the U.S. and n Chna tsel. One o the ost ortant ressons obtaned n Chna could be suarzed as ollows: For every anecdote n Chna, there s an equal and ooste anecdote. One coes to understand the lts o qualtatve eld research rather quckly, as the eerence o rs vares so wdely across ndustres, rovnces, rs, and te. Ths truth was orcbly drven hoe to the authors when derent rs n the sae ndustry n the sae rovnce would tell strkngly derent stores about ther eerence n the Chnese arket. Nevertheless, soe consstent thees eerged ro our dscussons wth r anagers, whch renorce soe o the key assutons n our odelng raework. The rst s that ost o the oregn nvested enterrses we ntervewed do, n act, coete wth state-owned rs to soe degree. In soe ndustres, the FIEs have already collectvely obtaned a donant arket
10 8 oston, such that the state-owned enterrses have been conned to the less rotable low end o the roduct arket. The second s that the Chnese governent, both natonal and local, s acutely aware o ths coetton, and has taken stes to ede the ablty o oregn rs to coete n the Chnese arket. Scholars o the Chnese econoy wll be alar wth the neus o restrctons on the oeratons o ultnatonals that were descrbed to us by ultnatonal eecutves. They nclude eort requreents, localzaton requreents, requreents or technology transer, restrctons on doestc arket access, and dcultes n recrutng and retanng key ersonnel a eature o the Chnese econoy whch s eacerbated by governent olcy. A nuber o ultnatonal anagers also colaned btterly about attets by rovncal and even local governents to etract unds ro oregn rs through both legal and llegal surcharges and taes. Whle, on aer, oregn rs get avorable ta and ort treatent, n ractce, gurng n all o the restrctons and etralegal surcharges, t s clear that, n any cases, oregn rs are oeratng wth a clear, governent-engneered dsadvantage. Thrd, our resuton o lted econoc ntegraton aong Chnese rovnces drawn ro Young (997) receved surrsngly strong suort ro dscussons wth eatrate anagers and natve Chnese eecutves. Whle the etent o local rotectons vares across ndustres, t s seen as a barrer to growth and econoc ecency by nearly every r we ntervewed. As one Shangha-based advsor to Aercan rs ut t, There s no Chna arket. There are several Chna arkets, and none o the are as bg as you thnk. These varous nerences ro our r ntervews wll be bult nto the odel that we roose to cature the tradeos aced by rovncal and central authortes, as they consder the olces towards oregn drect nvestent and trade.
11 9 4. The Model The odel we develo draws heavly ro Grossan and Helan (996), who aly the oltcal-econoy raework develoed n ther 994 aer to the ssue o ultnatonals. Seccally, they nvestgate whether the entry o ultnatonals wll aect the level o rotecton, and conversely, n a odel where the doestc ndustry s akng caagn contrbutons. In our odel, we wll suose nstead that the doestc ndustry s owned by the governent, whch gves etra weght to ndustry rots n ts objectve uncton. The resence o ultnatonals resents a otental threat to the state-owned enterrses through roduct arket coetton. L and Chen (998) use a raework lke ths to eane the governent s ablty to etract rents ro the ultnatonals, and we wll use olcy nstruents slar to thers. Our key ndngs concernng the shae o the governent s objectve uncton, deendng on the entry o ultnatonals, are not develoed n ether o these related aers, however. Because soe eatures o our odel are alar, we wll try to resent t as concsely as ossble. The ollowng onts suarze the an coonents: Regons: We wll treat the regons wthn the country as dstnct, ollowng Young s (997) thess that the rovnces wthn Chna have only lted trade wth each other. Thus, the odel below descrbes each regon, whch wll der n ters o ther underlyng araeters. Products: Each regonal econoy roduces a nuerare good denoted by 0, and a derentated roduct whch s a CES aggregate o the varous varetes. There are three sources As Naughton (996) and several other authors have noted, the Chnese governent stll reles on rettances ro state-owned enterrses or about two-thrds o ts revenue. Ths s also true at the rovncal level. The rovncal governent o Yunnan s ruored to obtan nearly all o ts revenue ro the rovncal tobacco onooly.
12 0 or the roducts: n h varetes are roduced by hoe rs; n varetes are roduced by oregn rs, where o these are roduced locally by ultnatonals, and the reanng (n -) are orted nto the country. Consuers: Consuer reerences are gven by, U θ ( θ) / θ = 0, θ > 0, θ, () θ where the CES aggregate s, ( )/ ( )/ ) / /( ) [ n (n ) ] h h ( =, >. (2) The values j denote consuton o the derentated varetes ro source j=h (hoe rs), (orted), and (ultnatonals). Lettng j denote the rces o each goods j=h,,, we can aze utlty subject to the budget constrant 0 n h h h (n -) =I to obtan: θ j = j q, j = h,,, θ and = q, (3) where q s a rce nde, q = /( ) [ n (n ) ]. h h (4) Note the > s the own-rce elastcty o deand or each varety, whle θ > 0 s the elastcty o deand or the CES aggregate. We add the restrcton that > θ, whch ensures that the cross-rce elastcty or each varety s ostve.
13 Frs: Costs or the locally roduced roducts are c j, j=h,. Costs or roducts roduced abroad are c, but also ace a secc tar o τ, so that argnal costs becoe c τ. Then rs at hoe and abroad aze rots gven by, π j = ( j c j ) j, j=h,, π = [ (c τ)] h. (5a) (5b) Usng the deand equatons n (3), and otzng over the choce o j, we obtan the rces: = c j, j h,, (6a) j = = ( c τ). (6b) Substtutng these rces nto (5), we obtan the rots o each tye o r: π j = j j /, j=h,,, (7) Factors: We wll suose that labor s the only actor o roducton, and one unt o the nuerare s roduced wth one unt o labor, so wages are unty. Whle workers n the hoe ndustry roducng derentated roducts earn unty, we shall assue that the ultnatonal rs ay a wage reu o (w) > 0. 2 Deste ths reu, we wll stll suose that s t cheaer or the ultnatonal to suly local deand ro local roducton, than ro orts whch are subject to the tar. That s, we assue, 2 The estence o ths wage reu s strongly conred n our data and n our ntervews wth ultnatonal anagers. The anagers suggested that, snce ultnatonals tycally eloy ore catal and technology ntensve roducton rocesses, they requre the best workers n the local labor ool. Hence, they ay what the labor lterature has reerred to as ecency wages. Other case studes have docuented the need to ay Chnese workers eloyed n ultnatonals a wage reu to oset the envy created by the even hgher wage earned by ther Aercan counterarts n these rs.
14 2 c < c τ. (8) Ths assuton s needed to ensure that the ultnatonal rs have any nterest n enterng the local arket. Polcy towards ultnatonals: Each oregn r aces the decson o whether to suly locally through orts, or through settng u a local lant whch requres a ed cost o F>0. We suose that the governent also charges the ultnatonal an rot ta o λ > 0. Ths nstruent s suosed to relect the vast range o actual olces used n Chna to etract rents ro ultnatonals, and not just the nonal ta on ultnatonals. For eale, the act that ost ultnatonals have had to use local artners relects an lct ta on ther rots, whch are shared wth the artner; slarly, the land-use ees that are coonly charged reduce the ultnatonals rots. By odelng these olces as a ta on rots, we are abstractng ro the necences caused by actual olces (such as local content restrctons, or eale). 3 Ths s slar to the entry ee used by L and Chen (998), and s n the srt o Grossan and Helan (994), who draw on Bernhe and Whnston (984) to argue that the outcoe o oltcal contests can be ecent. The rots earned locally are thus (-λ)π -F, wth π dened by (5a) and (6a). Alternately, the ultnatonal could just eort to the hoe country, and earn π dened by (5b) and (6b). Thus, entry wll occur and only (-λ)π -F > π. Usng (7), ths condton s wrtten: ( λ) F. (9) 3 O course, brbes ad to allow ultnatonals to enter are another eale o the rot ta. We (998) argues that corruton n Chna, whch ncludes the need or questonable ayents, acts as a sgncant deterrent to oregn drect nvestent.
15 3 We wll assue that when =λ=0 then (9) holds as a strct nequalty. Ths eans that or soe ostve λ, entry o ultnatonals wll occur. Governent Objectve Functon: We can now state the total returns to each nterest grou, begnnng wth consuers/workers, who receve a weght o α n the objectve uncton. It s readly vered that azed utlty ro () s U = I q -θ /(θ), where I denotes labor ncoe. Wth a workers used er unt outut n the ultnatonals, the total wage reu s (w)a. The ultnatonal s rce s =wa /(-), so the total wage reu can be wrtten [(-)(w)/w]. Includng ths wthn labor ncoe, we obtan utlty o, U = = w L e w w L w q ( θ ) [n ( θ ) θ (n ) h h h ] (0) where L s the labor endowent o the regon. To obtan the second lne o (0), we cobne (3) and (4) to solve or [ nh hh (n ) ] = θ q. We wll suose that the doestc rs are state-owned, so these rots accrue to the regonal and natonal governent. Prots o the doestc rs are n h π h = n h h h /, ro (6). We gve revenue ro state-owned rs a weght β n the objectve uncton. Fnally, the governent etracts rents λπ ro the ultnatonals, and also collects tar revenue o τ(n ). These two sources o revenue are each gven weghts o unty. The objectve uncton or each regon s dened as,
16 4 G(,T,τ) αu βn h π h λπ τ(n ) h h h ) (n ) ( n ) ( L τ θ α β θ α α = () w ) )(w ( ) ( λ α θ α, where the equalty ollows usng (0). Several roertes o ths uncton are suarzed by: Lea θ β α = ) ( s ) ( ) ( G h θ τ ) ( s (2) θ λ α ) ( s w w where s j, j=h,, denotes the share o doestc sendng on hoe, orted or ultnatonal roducts. Then: (a) τ β λ α, a w w, then G s quas-concave. I θ> these condtons ensure that G s ncreasng, whle θ< then slghtly stronger condtons are needed; 4 (b) τ β λ α, n w w, then G s quas-conve. 4 When θ< we need to ultly the rght-hand sde o the nequalty n (a) by (-θ)/(-)>. Then or s sucently sall, G wll be ncreasng.
17 5 In (2), we show that G/ conssts o three ters. The rst nvolves s h, the share o hoe rs n doestc sendng, and relects the roduct arket coetton between the ultnatonals and the hoe rs: βs h s hgh as coared to α, then ths ter s negatve. The second ter relects the loss n ort revenue as ultnatonals enter, as rst analyzed by Brecher and Daz-Alejandro (977). The thrd ter relects the wage reu generated by eloyent n the ultnatonals and the rot ta leved on these rs, both o whch generate a gan n welare. Part (a) dentes condtons under whch G s an ncreasng and quas-concave uncton o, as llustrated n Fgure 2 on the curve labeled Hgh wage reu or taes. It s well known ro Brecher and Daz-Alejandro (977) that when oregn catal s taed at a sucently hgh rate, ths can reverse the declne n welare that otherwse occurs ro ther entry. Our odel adds the gan due to the wage reu ad by ultnatonals. Conversely, the wage reu and taes are low as n art (b), then G s nstead a quas-conve uncton o, as ndcated by the curve labeled Low wage reu and taes. In ths case, t s relatvely easy to choose the value o β or o the tar τ such that G / = 0 n (2). Such a crtcal ont wll be a global nu o G, as llustrated. 5 Whle Prooston gves us the roertes o G wth resect to, we do not suose that the governent drectly controls entry o ultnatonals. Instead, entry s nluenced through the ta rate λ and the tar rate τ, so the nuber o ultnatonal rs s wrtten as a uncton (λ,τ). Multnatonals react n the eected anner to changes n these olces: 5 A conve curve between hoe welare and the nlow o oregn catal was llustrated by Brecher and Daz- Alejandro (977, Fgure 2), when oregn catal was not beng taes. In ther odel, there was no wage reu.
18 6 Lea 2 When (9) holds as an equalty, then d < 0 and dλ d > 0. dτ We are now n a oston to setu and solve the governents roble. We suose that the central and the regonal governents jontly deterne the rents arorated ro the ultnatonals n each regon. The central governent also chooses the tar rate. Denotng regons by the subscrt, we let G [ (λ,τ),λ,τ] denote the objectve uncton or each regon. Then the rot ta and tar are chosen to solve: a λ, τ G [ ( λ, τ), λ, τ], (3) subject to (λ,τ) < n, whch s the au nuber o oregn rs wllng to enter any regon. Our strong assuton n wrtng the objectve uncton as n (3) s that the ta rate λ or one regon s searable ro that n another regon. Ths s an etree verson o Young s (997) thess that the rovnces n Chna have only lted trade wth each other. I there s no trade between rovnces, then they wll not be n coetton wth one another or oregn nvestent, because ultnatonals can only serve the arket where they locate. Whle ths case does not lterally aly n Chna, t s certanly true that oregn r ace restrctons on ther ablty to arket outsde ther edate area. 6 By assung an etree or o these restrctons, we greatly sly the governental decson roble. 6 Soe natve Chnese rs ace these restrctons as well. In ntervews, one Chnese eecutve rueully noted that whle the governent has been brngng n ultnatonal enterrses or twenty years, ew truly ultrovncal enterrses are allowed to oerate wthn Chna tsel!
19 7 We also denote all other varables n regon wth that subscrt, though or slcty we suose ths does not aly to the oregn rce, or the nuber o oregn rs n. The ollowng result descrbes the soluton to (3) over the choce o the ta rate λ, whle a dscusson o the soluton or tars τ s ostoned untl secton 7. Prooston Suose that G s ncreasng n. Then regon wll choose the ta rate λ to allow entry o ultnatonals to the ont where ether =n or, G = π λ = θ ( ) ( λ )s n s. (4) A ont satsyng (4) s a global au o G over the choce o λ rovded that w α w λ β τ a,, and also that s < (-)/(-θ) when (4) holds. The otalty condton (4) s llustrated at ont A n Fgure 2, where the sloe o the curve G equals π /( / λ ). Ths agntude relects the all n the revenue λ π when the ta rate s lowered to attract ore ultnatonals. That s, the regons are actng as onosonsts as they attract oregn catal. Ths can be contrasted wth the coetton or oregn catal that s oten odeled between ndustral countres or regons, and that s ruled out n our raework by the assuton o lted trade between the regons.
20 8 Condton (4) leads to a global otu under the condtons o Lea(a), and also s < (-)/(- θ), whch s autoatcally satsed when θ> so that (-)/(-θ)>. I nstead the condtons o Lea(b) hold, so that G s quas-conve n, then t s stll ossble or the roble (3) to have an nteror au: the act that the regons are actng as onosonsts, and recognze that ultnatonal entry s aected by the rot ta, leads to urther concavty n G [ (λ,τ),λ,τ] as a uncton o λ. In our estaton, we wll conr that the second-order condtons are n act satsed. We can now use the results ro Prooston to develo an estatng equaton. We add the subscrt to the varables n (2), and relace s wth -s h -s on the rght-hand sde. Ths s set equal to the rght-hand sde o (4), as requred or otalty. The resultng equaton can be sled by usng (9) as an equalty. Fnally, we also add a te subscrt to all relevant varables. We obtan the ollowng equaton or the share o sendng on roducts o ultnatonals n regon and year t: s t = βs ht w t w t η ( ) (s w α t w t ht s t τ ) t t s t t n t s t α ( ) λtπt θ) ( θ)(ft λtπ α( ) ) ( θ) w τ t t w t t (F ( t t t t 2 π λ t π ) (5) The rst ter on the rght-sde o (5) s the share o sendng on doestc state-owned rs, whch enters wth the coecent -β. Thus, the weght on state-owned rs n the regonal objectve uncton s sly obtaned as the coecent on ther share n the regresson (5). A
21 9 hgh weght on the state-owned rs ndcates that n regons where these rs are ore revalent, the share o ultnatonal rs wll be corresondngly reduced. The net ter on the rght o (5) s the wage reu ad by ultnatonals, whch has the coecent η α(-)(θ-)/(-θ). Followng ths s the wage reu tes the share o sendng on state-owned rs lus orts. When the wage reu s hgher, we eect that regons would be ore wllng to accet ultnatonals, and ths s conred by havng a ostve coecent α(-) on that varable. The estate o tsel coes ro the net ter, whch s the ad valore tar rate tes the share o sendng on orts. Ths ter relects the loss n tar revenue as ultnatonals enter, and the coecent s -. Thus, cobned wth the orer coecent we can recover an estate o α. The nal ter on the rst lne o (5) relects the nuber o ultnatonals tes the share o orts. For slcty we treat n, whch s the nuber o oregn rs wantng to eort or nvest n Chna, as constant over regons and te, and estate t as a coecent. The ters on the second lne o (5) are rather cole eressons o the rot ta collected n each regon, and the ed costs o entry, enterng theselves and nteracted wth the wage reu and tars. We have no way o easurng these ters, but t sees very lkely that they wll vary systeatcally across regons, and ossbly also over te. Thus, we wll odel the ters on the second lne o (5) as: ( ) λtπt ( θ)(ft λtπ t α( ) ) ( θ) 2 w t w t τ t t (F t π λ t t π t = γ ) δ t u t, (6) where γ and δ t and rovnce and year ed-eects, and u t s a rando error assued to be uncorrelated wth the varables on the rst lne o (5). Gatherng all these ters nto ed-