Industrial Groupings and Strategic FDI.

Transcription

1 Octobe Industial Goupings and Stategic FDI. Buce A. Blonigen*, Chistophe J. Ellis**, and Dietich Fausten***. Abstact We show that industial owneship stuctues, such as keietsu goupings in Japan, may significantly impact fims incentives to engage in FDI. While the pevious liteatue has mainly focused on the cost of capital advantages enjoyed by keietsu fims, this pape examines two elatively unexploed channels by which owneship stuctue mattes fo FDI incentives. The fist channel involves the diect incentives geneated via standad poduct and facto maket inteactions wheeby keietsu fims with coss-owneship conside moe diectly the congestion effects of futhe FDI into a maket. The second channel involves the indiect incentives geneated by shaing of infomation acoss keietsu fims which educes enty costs of subsequent FDI. We find that keietsu fims ae moe agessive than non-keietsu fims in thei FDI stategies, that is, fo any given paamete values they undetake FDI with a highe pobability than independent fims. Futhemoe, keietsu fims adopt a moe agessive investment stategy against independent ivals than amongst themselves. Keywods: Foeign diect investment; keietsu. JEL classification: F0, F2, F23. Depatment of Economics, Univesity of Oegon, Eugene OR and NBER buceb@oegon.uoegon.edu Depatment of Economics, Univesity of Oegon, Eugene OR cjellis@oegon.uoegon.edu Depatment of Economics, Monash Univesity, Clayton 368, Victoia, Austalia. Dietich.Fausten@buseco.monash.edu.au Acknowledgements: We thank Ron Davies fo seveal useful comments.

2 Intoduction. It has fequently been suggested that fims in the lage industial goupings of Japan and Koea, known espectively as keietsu and chaebol, may behave diffeently fom thei US o Euopean countepats o independent domestic ivals. Membes of these industial goupings hold owneship shaes in each othe, obtain epeated financing fom associated membe banks, and paticipate on joint committees. Fo each of these easons it has been agued that the stuctue of keietsu and chaebol lead thei membe fims to behave (semi)coopeatively. Consequently, they may be expected to intenalize extenalities and find ways to mitigate the poblems implied by infomation asymmeties. 2 It has been futhe noted that coss-shaeholding stuctues can also weaken a fim s bagaining position and dilute its maket incentives, Flath [6] Flath [7]. The pupose of this pape is to exploe the elationships between industial owneship stuctue and the incentives fo fims to cay out foeign diect investment (FDI). It has been alleged that (semi)coopeative industial owneship stuctues, such as the Japanese keietsu system, yield thei membes advantages in exploiting oppotunities fo FDI. Typically these advantages ae explained as aising fom access to cheap funds fo investment. While these stoies seem quite plausible, the empiical suppot fo them has been somewhat mixed (see Beldebos and Sleuwaegen [2], Hoshi, Kashyap, and Schafstein [0], Fukao, Izawa, Kunimoi and Nakakita [9], and McKenzie [3]). In this pape we take a diffeent appoach. Rathe than concentate on the implications of owneship stuctue fo the financing of FDI, we instead focus, fist, on the implications it has fo the stategic incentives to invest that aise though the inteactions between fims on input and output makets, and, second, on the incentives it povides fo infomation geneation and dissemination. We fist develop an illustative theoetical model simila to that poposed in the liteatue on the adoption of new technology by Fudenbeg and Tiole [8]. FDI decisions ae modelled as Bank epesentatives also sit on the boads of associated fims. 2 See fo example Suzuki [7], Dewente and Wathe[3], Kimua and Pugel [].

3 enty pobabilities in a mixed stategy equilibium to a game in stages. We model the facto and poduct maket inteactions by allowing the fims payoffs to change as successive enty takes place. The infomation aspect of the pocess is captued by assuming that enty costs ae a declining function of the total numbe of pio investments. To futhe captue the salient featues of the FDI pocess we assume some infomation is public, and is geneated as an extenality to be enjoyed by all potential entants, wheeas some of the infomation is pivate, and is only tansmitted between fims engaged in coopeative elationships. Modelling FDI decisions as enty pobabilities in a mixed stategy equilibium has peviously been poposed by Lin and Saggi [2] and Ellis and Fausten [4]. Ou analysis, though still not fully geneal, consideably extends these ealie contibutions. Linn and Saggi examined a single stage game between two competitive fims so as to obtain the compaative statics popeties of the initial enty pobability decision, and the optimal delay between initial and subsequent enty. Ellis and Fausten followed the same path as Linn and Saggi but intoduced ovelapping shae owneship into the model to analyze the implications fo FDI of diffeent owneship stuctues. Ou wok makes two key futhe extensions; we intoduce a thid fim into the analysis and allow fo asymmetic infomation between fims. Intoducing a thid fim might seem mino, yet it is significant in thee ways; () It allows us to conside stategic inteactions between a pai of (semi)coopeative fims and a competitive ival; (2) It allows the FDI enty game to be split into a sequence of stages, each of which is chaacteized by equilibium enty pobabilities, allowing examination of the elationships between enty pobabilities ove time; and, (3) It allows pivate infomation to play asignificant ole, as an ealy entant must conside the subsequent asymmeties in infomation that may be geneated by its enty. 2

4 2 The Model. When fims contemplate locating poduction facilities outside of thei home counties they face adifficult tade-off. If they invest ealy they may gain advantages on both poduct and input makets. Howeve, in moving ealy they also face a host of potential poblems. Fo example, ittakestimetoleanhowtoopeateefficiently in a foeign labo maket and unde a foeign legal system. Thus the initial fixed costs of investment may be high. If, on the othe hand, they delay enty, they will fogo some of the poduct and facto maket advantages enjoyed by ealy entants, but may gain valuable infomation fom obseving thei pedecessos. This infomation will educe the fixed costs of initial enty. The theoetical model we now develop captues this basic tension. 2. Basic Stuctue. We assume that thee ae thee fims that may poduce output eithe in thei domestic economy(ies) o aboad via FDI. The thee fims may be eithe fully independent, as in the case of most US fims o, altenatively, they may be patially coopeative, as fo example when linked via ovelapping shaeholdings, suchasinthecasesofjapanesekeietsu and Koean chaebol membes. We assume that initially all thee fims ae engaged in domestic poduction, and that at each subsequent point in time each must choose eithe to continue in this poduction mode, m = D, o make an ievesible switch to foeign poduction, m = F. At any time t the flow pofit enjoyed by fim n fom choosing a poduction mode, given the modes of poduction chosen by the othe two fims, is witten n (t) = n m n (t) m i (t),m j (t) n =, 2, 3, i=, 2, 3, j=, 2, 3, n6= i 6= j. 3

5 Often we shall adopt shothand notation of the fom D (t) D 2 (t),f 3 (t) DDF.Inthis notation, the fim to whom the pofits accue is always listed fist, we then adopt the convention that othe fims ae listed in ascending sequence, except that will follow 3, that is,2,3,,2 etc. To captue the idea that thee ae advantages to ealy investment we assume that pofits will vay acoss the diffeent combinations of domestic poduction and FDI. Thee ae two main effects involved in these pofit ankings. Theymayeflect eithe Counot o Betand competition in the poduct maket (see Linn and Saggi [2]), whee ealy entants face lowe maginal costs and hence a maket advantage, o, altenatively, they may puely eflect labo costs. In the cost stoy FDI allows the fims to exploit cheap labo in the host county, but epeated enty aises labo demand and hence wages in the appopiate foeign labo pool, but lowes demand and wages domestically 3. The poduct maket stoy geneates the pofit odeing n FDD > n FFD = n FDF > n FFF > n DDD > n DFD = n DDF n DFF n. Howeve, the facto maket stoy geneates the odeing 4 n FDD > n FFD = n FDF > n FFF > n DFF > n DFD = n DDF > n DDD n. Fo both pofit odeings, in the absence of any elocation costs each fim would independently pefe to undetake FDI. This, as we shall see, tuns out to be the cucial featue of the pofit odeings. In what follows the analytical esults obtained ae identical fo both odeings, while the numeical simulation esults display identical qualitative popeties and only elatively small quantitative diffeences. In eality a mix of both stoies is, of couse, possible. The esults epoted 3 Fo example Feensta and Hanson [5] find that fo egions of Mexico in which FDI is concentated, moe than 50% of the incease in the total wages of skilled wokes can be attibuted to the effects of foeign capital inflows. 4 It might be agued that the odeing n DDD > n DFD should be evesed if thee ae disadvantages to being the only poduce in a specific location, as fo example if thee ae positive spilloves between fims. This has no qualitative implications fo ou analysis. 4

6 in the est of the pape petain to the fist odeing 5. To chaacteize the diffeent potential foms of industial owneship stuctue we intoduce the paamete β i n which epesents the claim of fim i on the pofits of fim n. 6 So if we denote the total flow pofits of fim n as P n, we may wite the possibilities as P n m n,m i,m j =( X i6=n β i n) n m n,m i,m j + X i6=n β n i i m i,m j,m n n =, 2, 3, i=, 2, 3, j=, 2, 3, n6= i 6= j What distinguishes ou model fom its antecedents is the ability to analyze stategic FDI when thee ae both coopeative and non-coopeative fims in the population. We theefoe concentate on this case and assume that fims and 2 ae membes of a symmetic keietsu, soβ 2 = β 2 > 0, while fim 3 is puely competitive, so β 3 = β 3 = β 3 2 = β 2 3 =0. We may now utilize this stuctue to examine the fims FDI decisions in an economy whee some fims ae linked though industial goupings and othe ae not. 2.2 The Fims Poblem. At some initial date t =0each fim is engaged exclusively in domestic poduction. 7 The poblem each must solve is if and when to switch to FDI given that switching poduction fom one county to anothe is clealy costly 8. Weassumethatthecostafim incus in switching fom domestic to foeign poduction is a deceasing function of the numbe of fims that have aleady switched 9. 5 Numeical simulation pogams fo both odeings ae available fom the authos on equest. 6 In the Japanese keietsu system thee ae othe mechanisms by which coopeation may be induced between membes. The ole of associated commecial banks in poviding epeated funding to membes, and the placement of bank officials in senio positions in the membes hieachies seem paticulaly impotant. β may theefoe be intepeted moe widely as a measue of coopeation athe than simply coss shaeholdings. See also Aoki [], Ou, Hamilton and Suzuki [5], and Ouchi [6]. 7 We might think of this as the time at which FDI became a potentially lowe cost mode of poduction. Eithe because of the elaxation of legal estiction by the host county, an impovement in the host counties labou foce, o an incease (eal o theatened) in taifs fo that counties home maket etc. 8 Hee we ae making the implicit assumption that coss shaeholding between fimsdoesnoteliminatethe diect incentive fo fims to undetake FDI. In the simulations that follow we veify this assumption. 9 We assume that the infomation allows fo the eduction in fixed enty costs. This allows us to model the equilibia in each (sub)game as stationay. 5

7 The idea hee is that thee is cost educing infomation that may be obtained by leaning fom the enty expeiences of peceding fims. Howeve, we assume that enty by a keietsu goup fim lowes the futue enty cost of a fellow goup membe by moe than it lowes the enty cost of a non-membe fim. Seveal intepetations can be given to this assumption. The fist, and ou pefeed, intepetation, is that some of the infomation is publicly available, but some is pivate and will be tansfeed only between fims in the same industial gouping. In the appendix we show that shaing pivate infomation is an individually ational stategy fo keietsu membe fims. A second intepetation of ou asymmetic cost eduction assumption is that all infomation is public, but infomation geneated by a keietsu fim is of geate cost educing value to othe goup fims than to independent fims. Hee we ae suggesting that similaities in financial stuctue, simila labo and management pactices, and simila cultual backgounds make the expeiences of keietsu fims moe valuable to each othe than to outsides 0. Finally we might ague that the stuctue of keietsu povides channels fo the cedible tansmission of infomation between goup membes fims that ae not available between non-membe fims. The goup membe bank may be impotant in this egad. We define the enty date of the fist fim as t = t, the second as t = t,andthethidas 0 Fo this intepetation it makes sense to think of this as a game between keietsu fims and westen competitos. 6

8 t = t, natually t t t. We thus expess the enty costs as C(t) = C whichmustbecommonacossfims C if no pivate infomation is evealed at t C if the entant at t eveals pivate infomation C if no pivate infomation is evealed at t C if the entant at t eveals pivate infomation t t t <t t t <t t with C < C R C ª <C <C The fims maximize expected pofits net of switching costs, which involves each selecting pobabilities of FDI at each point in time given those selected by the othe fims. We wite the pobability of fim n switching to FDI as ρ n. Since this is a game in stages, we also equie notation fo which fim(s) have aleady caied out FDI and which have not, consequently G n,i,j, will indicate the game whee no enty has yet occued, G n,i the (sub) game whee fims n and i have not yet enteed, and G n will be the (sub)game whee only fim n has not enteed. With this notation pobabilities will be witten in the fom ρ n (G n,i ) and so on 2 The value that a fim obtains fom a paticula action (F o D) in the game and each sub-game, given the actions of the othe fims, will be definedinthefom V (DDD G,2,3 ), which is the value to fim of the action D if both othe fims also choose D in the game G,2,3. In a simila vein V 3 (FFD G 2,3 ) would epesent the value to fim 3 of the action F in the subgame Clealy simila fims may lean moe fom each othe than dissimila fims will. Howeve, thee ae common poblems such as leaning to deal with a foeign legal system and foeign labou makets and pactices that ae common to all. We thus abstact fom diffeential leaning in this pape. 2 Since each (sub)game is stationay we do not need any futhe notation to denote time. 7

9 G 2,3,andsoon The Extensive Fom of the Game. We ae now eady to descibe how the pocess of FDI evolves by pesenting the game in extensive fom. Figue illustates the initial situation faced by the thee fims. Fim ρ (G,2,3 ) - ρ (G,2,3 ) Enty No Enty Fim 2 ρ 2 (G,2,3 ) - ρ 2 (G,2,3 ) ρ 2 (G,2,3 ) - ρ 2 (G,2,3 ) Enty No Enty Enty No Enty Fim 3 ρ 3 (G,2,3 ) - ρ 3 (G,2,3 ) ρ 3 (G,2,3 ) - ρ 3 (G,2,3 ) ρ 3 (G,2,3 ) - ρ 3 (G,2,3 ) ρ 3 (G,2,3 ) - ρ 3 (G,2,3 ) Enty No Enty Enty No Enty Enty No Enty Enty No Enty V (FFF G,2,3 ) G 3 G 2 G 2.3 G G,3 G,2 G,2,3 V 2 (FFF G,2,3 ) V 3 (FFF G,2,3 ) Figue : The thee fim enty game G,2,3 in extensive fom. Each fim must choose an enty pobability, ρ n (G,2,3 ), as a best eply to those chosen by the othe two fims. Once a fim (o fims) has enteed we move to the appopiate subgame. Fo 3 Heeafte we shall maintain the assumptions V n (FFF G n ) <V n (DF F G n ) n which ae equied to ensue that the whole stuctue does not unavel backwads with all fims enteing instantaneously at the fist oppotunity. 8

10 example, if fim entes in the game G,2,3 then fims 2 and 3 play the subgame G 2,3 illustated in figue 2 Fim 2 ρ 2 (G 2,3 ) ρ 2 (G 2,3 ) Enty No enty Fim 3 ρ 3 (G 2,3 ) - ρ 3 (G 2,3 ) ρ 3 (G 2,3 ) - ρ 3 (G 2,3 ) Enty No Enty Enty No Enty V 2 (FFF G 2,3 ) G 3 G 2 G 2.3 V 3 (FFF G 2,3 ) Figue 2: The two fim enty game G 2,3 in extensive fom. Hee the emaining two fims must choose the enty pobabilities ρ 2 (G 2,3 ),andρ 3 (G 2,3 ) as best eplies. 2.4 Equilibium. To obtain the equilibium of the model we solve ecusively fo equilibia in each of the potential subgames, stating with {G,G 2,G 3 } then using these values to solve {G,2,G,3,G 2,3 }, and finally using the values fom both {G,G 2,G 3 } and {G,2,G,3,G 2,3 } to solve fo G,2,3. We thus obtain the subgame pefect equilibium as a sequence of mixed stategy equilibia in the 9

11 subgames. Given that fims and 2 ae, by assumption, membes of a symmetic keietsu, and we have assumed symmety between these two fims in all othe espects, it seems natual to conside symmetic equilibia whee ρ (G,2,3 )=ρ 2 (G,2,3 ) ρ 2 (G,2,3 ) The Thid Wave: Enty in the Subgames G, G 2,andG 3. Once two fims have enteed the only poblem faced by the thid is whethe to undetake FDI o to continue poducing domestically. We shall make the assumption V n (DFF G n ) >V n (FFF G n ) n =, 2, 3. That is, the thid fim will always opt fo domestic poduction ove FDI. We choose to adopt this condition fo both technical and conceptual easons. Technically the condition is sufficient to ensue that the model does not unavel backwads. Conceptually, if all fims wish to undetake FDI egadless of the pesence of othe fims, then thee can be no inteesting stategic inteactions in these subgames. One intepetation of this assumption is that pevious FDI by two of the thee fims is sufficient to bid up facto costs in the inbound juisdiction to the point whee futhe investment is no longe pofitable The Second Wave: Enty in the Subgames G,2, G 2,3,andG,3. We tem the subgames G,2 4, G 2,3,andG,3 asthesecondwaveofenty. Heeonefim has enteed and the emaining two fims enjoy infomation geneated by the fist entant. In the games G 2,3,andG,3 pio enty was by a keietsu membe. The emaining keietsu fim enjoys the enty cost eduction C C which eflects both the infomation geneated that is publicly and pivately available, wheeas the cost eduction fo the non-keietsu fim is only C C eflecting only the publicly available infomation. In the game G,2 the non-keietsu fim has enteed and 4 With β 2 = β2 =0this subgame coesponds to the Linn and Saggi op. cit. model. With /2 β 2 = β2 0 it is the Ellis Fausten op. cit. model. 0

12 only eveals the public infomation, giving the cost eduction C C. The equilibia in these subgames ae deived fom a value fo the subgame and an indiffeence condition fo each of the two fims. 5 Fo example, in the subgame G 2,3 (and by symmety G,3 ) these involve fou expessions 6. The two indiffeence conditions ρ 2 (G 2,3 )V 3 (FFF G 2,3 )+( ρ 2 (G 2,3 ))V 3 (FFD G 2,3 ) = ρ 2 (G 2,3 )V 3 (DFF G 2,3 )+( ρ 2 (G 2,3 ))V 3 (DFD G 2,3 ) and ρ 3 (G 2,3 )V 2 (FFF G 2,3 )+( ρ 3 (G 2,3 ))V 2 (FDF G 2,3 ) = ρ 3 (G 2,3 )V 2 (DFF G 2,3 )+( ρ 3 (G 2,3 ))V 2 (DDF G 2,3 ), and the two value functions V 2 (DDF G 2,3 )=ρ 2 (G 2,3 ) ρ 3 (G 2,3 )V 2 (FFF G 2,3 )) + ( ρ 3 (G 2,3 ))V 2 (FDF G 2,3 ) +( ρ 2 (G 2,3 )) ρ 3 (G 2,3 )V 2 (DFF G 2,3 )+( ρ 3 (G 2,3 ))V 2 (DDF G 2,3 ) and V 3 (DFD G 2,3 )=ρ 3 (G 2,3 ) ρ 2 (G 2,3 )V 3 (FFF G 2,3 )) + ( ρ 2 (G 2,3 ))V 3 (FFD G 2,3 ) +( ρ 3 (G 2,3 )) ρ 2 (G 2,3 )V 3 (DFF G 2,3 )+( ρ 2 (G 2,3 ))V 3 (DFD G 2,3 ). 5 Details in appendix 2. 6 In each of these subgames, fo thee to be a mixed stategy equilibia we equie that fo each fim neithe F no D is a dominant stategy. Consistent with ou pio assumptions we assume that each fim pefes to ente if the othe does not, but pefe not to ente if the othe does. Again using the subgame G 2,3 to fix notationthis tanslates into conditions of the fom V 2 (FDF G 2,3 ) >V 2 (DDF G 2,3 ) V 2 (DF F G 2,3 ) >V 2 (FFF G 2,3 )

13 Substituting in and solving, then epeating the pocess fo the othe subgames, povides the equilibium solutions pesented in table. Subgame. Non-keietsu enteed Keietsu enteed Fim G,2 G,3 G 2,3 2 µ ( 2β) DF F FFF ( β) FDF µ ( 2β) DF F FDF +C β DF F βc FDF +C FFF ( β) FDF β DF F 3 N/A DF F FFF FDF βc N/A FDF +C µ ( β) DF F FFD +C FFF FFD DF F µ ( β) FFF N/A FDF FDF +C DF F FFD +C FFF Table : Equilibium enty pobabilities fo the second wave subgames. FFD Notice that in the subgames involving a keietsu and non-keietsu fim, G,3 and G 2,3,theenty pobability of the fome always exceeds that of the latte, that is DF F FFF FDF FDF + C > ( β) ³ DF F FFF FFD FFD + C. This follows both because the keietsu fim faces lowe enty costs, C <C, and because it elinquishes pat of its pofits to the othe keietsu fim. This latte effect follows fom the indiffeence condition that chaacteizes mixed stategy equilibia. The incentives to ente by a second keietsu fim ae diluted both because enty hams the fist keietsu fim and because it loses some of its pofit gain to the fist keietsu fim. Thus, fo the keietsu fim to be indiffeent between enty and delay it must be the case that the enty pobability of the non-keietsu fim is elatively lowe. Keietsu fims also have highe enty pobabilities in the subgames whee they face independent 2

14 ivals, G,3 and G 2,3, than in the subgame whee they face each othe, G,2,i.e. DF F FFF FDF FDF + C > FFF ( 2β) ³ DF F ( β) FDF FDF + C β DF F βc. This can be undestood by consideing the indiffeence conditions that chaacteize mixed stategy equilibia. In the subgames G,3 and G 2,3 the enty pobability of one keietsu fim is equied to make the non-keietsu fim indiffeent between FDI and domestic poduction. In the game G,2, this enty pobability must make the othe keietsu fim indiffeent between domestic poduction and FDI. Eveything else equal a non-keietsu fim has geate incentive to ente than a keietsu fim. It etains all of the gains fom the action, and does not shae in the damage it inflicts on othe fims. Thus to make a non-keietsu fim indiffeent between domestic poduction and FDI equies its opponent, that is the keietsu fim, adopt a highe pobability of enty than it would othewise. Hence keietsu fims ente with a geate pobability in the subgames whee they face non-keietsu fims. It is now staightfowad to deive the compaative statics popeties of these subgames, as shownintable2. 3

15 Enty Pobabilities Subgames Non-keietsu enteed Keietsu fim enteed G,2 G,3 G 2,3 dρ (G,2 ) dρ 2 (G,2 ) dρ (G,3 ) dρ 3 (G,3 ) dρ 2 (G 2,3 ) dρ 3 (G 2,3 ) dβ dc - a - a dc a : Fo FDF DF F C Table 2: Compaative statics fo the second wave enty pobabilities. We see that in the inteesting subgames, G,3 and G 2,3, those whee a mix of keietsu and nonkeietsu fims emain, an incease in the keietsu coopeation paamete dβ > 0 lowes the pobability of enty by non-keietsu fims, but does not affect the keietsu fims, thus making the elative pobability of enty by keietsu membes highe. This follows immediately fom the natue of the mixed stategy equilibium, which equies that the pobabilities ρ 3 (G,3 ) and ρ 3 (G 2,3 ) satisfy the indiffeence conditions fo the keietsu fims. As the paamete β inceases, the enteing keietsu fim shaes moe of the gain fom enty with the keietsu fim that had peviously enteed, and also shaes moe of the losses its enty imposes on this fim. Thus the etuns to enty fo the new keietsu entant ae educed. To maintain indiffeence it is necessay then that the non-keietsu fim s pobability of enty declines. Fo the inteesting subgames, the effects of changes in the cost paametes dc and dc may be explained in a simila manne. As C inceases, the value of enty to the non-keietsu fim declines, so to maintain the indiffeence condition the enty pobability of the appopiate keietsu fim must fall. As C inceases the value of enty to the appopiate keietsu fim declines, so to 4

16 maintain the indiffeence condition of the mixed stategy equilibium the enty pobability of the non-keietsu fim must fall. Notice also that since dρ 3 (G,3 )/dβ < 0 and dρ 3 (G 2,3 )/dβ < 0 and both ρ (G,3 )=ρ 3 (G,3 ) and ρ 2 (G 2,3 )=ρ 3 (G 2,3 ) if β =0, each keietsu fim always has a highe pobability of FDI than the non-keietsu fim. In the subgame G,2 only keietsu fims emain and only non-keietsu fims have enteed. The effects of β on enty pobabilities ae explained by the dilution of enty incentives as discussed fo the othe subgames. The effects of C ae again detemined be the mixed stategy indiffeence conditions as above. That C does not effect keietsu enty pobabilities follows fom the fact that the only pio entant was not a keietsu membe and does not shae any pivate cost educing infomation The Fist Wave: Enty in the Game G,2,3. Enty in the fist wave, as inspection of figues and 2 might suggest, is vey complex, if no fim has enteed at a time t then thee ae 8 possible stategy choices leading to 8 possible subgames. The mixed stategy equilibium fo G,2,3 is chaacteized by 6 conditions. Fo each fim n we may define a value fo the game and an indiffeence condition that states the fim is indiffeent between FDI and domestic poduction. Fo fims o 2 these conditions take the fom ρ 2 (G,2,3 ) ρ 3 (G,2,3 )V (FFF G,2,3 )+( ρ 3 (G,2,3 ))V (FFD G,2,3 ) +( ρ 2 (G,2,3 )) ρ 3 (G,2,3 )V (FDF G,2,3 )+( ρ 3 (G,2,3 ))V (FDD G,2,3 ) = ρ 2 (G,2,3 ) ρ 3 (G,2,3 )V (DFF G,2,3 )+( ρ 3 (G,2,3 ))V (DFD G,2,3 ) +( ρ 2 (G,2,3 )) ρ 3 (G,2,3 )V (DDF G,2,3 )+( ρ 3 (G,2,3 ))V (DDD G,2,3 ) 5

17 and V (DDD G,2,3 )=ρ 2 (G,2,3 ) 2 ρ 3 (G,2,3 )V (FFF G,2,3 )+( ρ 3 (G,2,3 ))V (FFD G,2,3 ) +ρ 2 (G,2,3 )( ρ 2 (G,2,3 )) ρ 3 (G,2,3 )V (FDF G,2,3 )+( ρ 3 (G,2,3 ))V (FDD G,2,3 ) +( ρ 2 (G,2,3 ))ρ 2 (G,2,3 ) ρ 3 (G,2,3 )V (DFF G,2,3 )+( ρ 3 (G,2,3 ))V (DFD G,2,3 ) +( ρ 2 (G,2,3 )) 2 ρ 3 (G,2,3 )V (DDF G,2,3 )+( ρ 3 (G,2,3 ))V (DDD G,2,3 ) simila conditions hold fo fim 3. Substituting in and solving these equations (togethe with the fim 3 conditions-see appendix 3) yields solutions ρ (G,2,3 )=ρ 2 (G,2,3 )=ψ(c,c,,β, DF F, FDF, FFF) ρ 3 (G,2,3 )=χ(c,c,c,,β, DFF, FDF, FFF). Details of the fom of these expessions ae povided in appendix 3. The solutions ae vey complex and do not easily yield analytical esults, thus we esot to numeical method to exploe thei popeties. Vaiations in the Level of Coopeation and Initial Cost of Enty. Tables 3 and 4 povide illustative examples of simulations un using Mathematica 7. 7 Fo the simulation epoted we assumed that π FDD =00,π FDF = π FFD =96,π FFF =90, π DDD =89,π DDF = π DF D =84,π DFF =79,=5%. Fo vaiations in the fixed cost of initial enty we maintained the diffeential betwen initial and subseqent enty costs by imposing the same changes on C and C. The popeties of the esults epoted wee geneally not sensitive to vaiations in paamete values that satisfied the estictions of the theoy. 6

18 Level of Coopeation β Initial Enty Cost C as a pecentage of FDD Table 3 - Fist wave enty pobabilities fo keietsu fims, ρ 2, (top cell enty) and non-keietsu fims, ρ 3, (bottom cell enty) fo diffeent coopeation β, and initial enty costs C paametes. We see fom table 3 that an incease in the level of coopeation, β, between keietsu fims inceases the fist wave enty pobabilities of keietsu fims and lowes the enty pobabilities of non-keietsu fims 8. The intuition behind these esults is complex. Changes in β affect the enty pobabilities both in the fist wave enty game, and the subsequent second wave enty subgames. Recall that in a mixed stategy equilibium the enty pobabilities ae detemined by the equiement that each fim be indiffeent between undetaking FDI and continuing with domestic poduction. When β inceases, the keietsu fims individually find FDI less attactive in each subgame. This is because an entant shaes a geate popotion of the benefits fom enty with its keietsu patne, and also shaes a geate popotion of the losses its enty imposes on its patne. So to maintain the keietsu fims indiffeence condition the elative pobability of initial enty by the non-keietsu fim must decline. Similaly fo the keietsu fims the elative pobability of initial enty must incease to keep the non-keietsu fim indiffeent between FDI and domestic poduction. Hee the 8 It might appea that the fistcolumnintable3foβ =0eveals an inconsistency in the esults. This is not the case. The simulations wee caied out assuming that keietsu fims shae all pivate infomation (not β pecent of it) thus the enty pobabilites should only be equal when β =0and C C =0. Inspection of table 4 demonstates that this consistency check is satisfied. 7

19 agument is even moe complex. Since an incease in β makes the non-keietsu fim less likely to ente in each subsequent subgame (see the next section fo details), the elative value of enty to the non-keietsu fim in the initial game thus inceases. Hence, to maintain indiffeence fo the non-keietsu fim the elative pobability of enty by the keietsu fims must incease. Inceases in initial enty costs lowe the pobabilities of enty fo both keietsu and nonkeietsu fims 9. Hee, in equilibium, both the keietsu fims and the non-keietsu fim must have lowe enty pobabilities if they ae to emain indiffeent between FDI and continued domestic poduction. Vaiations in the Level of Coopeation and the Value of Pivate Infomation. Table 4 chaacteizes ou simulation esults fo vaiations in the value of pivate infomation, o C C. Level of Coopeation β Value of pivate infomation C C as a pecentage of FDD Table 4 - Fist wave enty pobabilities fo keietsu fims (top cell enty), ρ 2, and non-keietsu fims (bottom cell enty), ρ 3, fo diffeent levels of coopeation, β, and pivate infomation values, C C. 9 These esults obtained fo almost all the paamete space. Fo high costs of initial enty, such that the pobability of enty by keietsu fims became vey small (in the neighbohood of ρ ), the enty pobabilities fo the non- keietsu fims became eatic. We believe this eflects a highly sensitive tade-off, asthekeietsu fim s enty pobabilities become vey small thee is a significant incentive fo the non-keietsu fim to ente, but at the same time the high costs that induced this behavio fom the keietsu fims also povides a stong disincentive to enty fo the non-keietsu fim. 8

20 Despite thei appaent complexity the popeties of the equilibia may be quite easily stated. Inceases in the value of pivate infomation C C, geneated by vaying C fo a given C, lowe the pobability of enty by non-keietsu fims at all levels of the coopeation paamete, and do not affect the pobability of enty by the keietsu fims. Hee again the intuition is subtle and follows fom undestanding the natue of a mixed stategy equilibium. An incease in C C makesitmoeattactivefoeachkeietsu fim to delay initial enty in anticipation that the othe will ente and povide them with this eduction in enty cost. Thus, fo the keietsu fims to emain indiffeent between FDI and domestic poduction the elative pobability of enty by the non-keietsu fim must fall. The invaiance of the keietsu fim s enty pobabilities aises because vaiations in C do not affect the payoffs associated with FDI fo the non-keietsu fim. Since the pobabilities of enty by the keietsu fims ae detemined by the condition that the non-keietsu fim is indiffeent between FDI and domestic poduction it then follows that the pobability of entybythekeietsu fims is unaffected by vaiations in C. 20 Notice also that fo all values of β > 0 and C C the pobability of enty by each keietsu fim is lage than fo the non-keietsu fim. Ou esults ae illustated in figue 3 20 The esults epoted wee found to be vey obust and obtained ove all of the paamete space fo which mixed stategy equilibia wee found to exist. Full details and the simulation pogams ae available fom the authos on equest. 9

21 Pobabilities of Initial Enty Fist wave enty pobabilities fo by Keietsu and Non - Keietsu Fims keietsu and in the Game G AP non-keietsu fims. 23 Keietsu fim. Enty Pobability Enty Pobability Non-keietsu fim Coopeation Pecentage pecentage. 0 Enty Cost Pecentage in Incements 0 20 Enty cost with pivate infomation in incements. Fist wave enty pobabilities fo the Pobabilities of Initial Enty non-keietsu fim. by Non - Keietsu Fims in the Game G AP 23 Enty Enty Pobability Fist wave enty pobabilities fo the Pobabilities of Initial Enty keietsu fim. by Keietsu Fims in the Game G AP Coopeation pecentage. Coopeation Pecentage 0 Enty Cost Enty cost with pivate Pecentage infomation Incements in incements Enty Pobability Enty Pobability Coopeation Pecentage Coopeation pecentage. Enty cost Enty with Cost pivate Pecentage infomation Incements in incements Figue 3. 20

22 3 Conclusions. In this pape we have exploed how the existence of coopeative industial goupings such as Japanese keietsu o Koean chaebol may effect the likelihood of FDI. This question has been addessed befoe in the context of a two fim model in which it was found that coopeation dilutes the incentives fo FDI. Hee we allowed fo thee fims, two of which ae membes of a coopeative industial gouping. This allowed two significant innovations. We wee able to exploe FDI enty pobabilities amongst a population of fims some of which ae coopeative and some independent. Futhe, we modeled asymmetic infomation by assuming that keietsu fims shae all petinent infomation with fellow goup membes, wheeas independent fims only eveal that which is publicly obsevable. Ou esults contast stikingly with those of the two fim model (see the subgame G,2 in table 2 o Ellis and Fausten [4]). In the pesence of a heteogeneous pool of fims, and with the infomation asymmeties just descibed, coopeation tends to incease the incentives fo FDI. Despite the complexity of the analysis, the detailed conclusions of ou model ae quite clea. Fo any given infomational advantage, C C, in each subgame involving both keietsu and non-keietsu fims, the keietsu fims undetake FDI with highe pobability. Futhe, in these subgames, the moe coopeative ae the keietsu fims, as chaacteized by the coopeation paamete β, the geate the magin by which thei enty pobabilities exceed those of non-keietsu fims. The effects of an infomational advantage ae also clea. In the initial game, temed the fist wave, the non-keietsu fim is discouaged fom enty by the infomational advantage of the two keietsu fims (as C C inceases, ρ 3 (G,2,3 ) falls). The enty pobabilities of the keietsu fims ae unaffected. In the subgames temed the second wave, the effects ae somewhat diffeent, an incease in the infomational advantage aises the enty pobability of the non-keietsu fim, elative to the enty pobabilities of keietsu fims. This model povides some potentially inteesting policy pesciptions. Juisdictions inteested 2

23 in attacting inwad FDI fequently offe diect financial incentives in the fom of tax beaks, and indiect incentives in the fom of sevices and infastuctue that facilitate the tansition of poduction to thei locations. In the fist wave we may view both types of incentive as loweing the enty cost paamete C 2. As table 3 indicates, both keietsu and non-keietsu fims espond positively to this incentive. Howeve, caeful eading of table 3 eveals that the maginal effect of this incentive on keietsu fim enty pobabilities is deceasing in the level of coopeation between the goup membe fims. In fact at low levels of coopeation they ae moe esponsive to these FDI incentives than the non-keietsu fim, while at high levels of coopeation the convese is tue. In the second wave tax beaks may be viewed as geneating eductions in C and C.Ifthe host juisdiction is limited to giving fims equal tax teatment, then we can see fom table that equal eductions in C and C will incease the enty pobabilities of both keietsu and nonkeietsu fims. Futhe, in the second wave subgames involving both keietsu and non-keietsu fims, a tax cut has a geate effect at the magin on the enty pobability of a keietsu fim. If, altenatively, juisdictions have the ability to offe fim-specific tax beaks then inspection of table eveals an inteesting asymmety, tax cuts ae moe effective at the magin in attacting keietsu fims. But a potential host juisdiction is best advised to spend tax dollas in educing C the enty cost of the non-keietsu fim, because this makes the keietsu fims moe likely to ente. Clealy juisdictions need to be caeful in designing tax incentive schemes in these stategic envionments. 2 By modeling tax incentives in this fashion we ae consistent with both the cost eduction captuing a pemanent tax beak, so the eduction in C must be thought of in PDV tems, o, as epesenting a tax holiday, whee the eduction in C is tempoay. 22

24 4 Appendices. 4. Appendix - Deivation of the value functions fo the sub-games G,G 2,and G 3. We deive the value functions V 3 (FFF G 3 ) and V 3 (DFF G 3 ), V (FFF G ), V (DFF G ), and V 2 (FFF G 2 ), V 2 (DFF G 2 ) may be obtained in an identical fashion. V 3 (FFF G 3 ) Z ( X β i 3) 3 FFF K + β 3 FFF + β FFF e t dt t=0 i=,2 integating the RHS of this expession gives us V 3 (FFF G 3 ) ( X i=,2 3 β i FFF 3) K + β 3 FFF + β 3 2 FFF 2 whee K = C o C as appopiate. V 3 (DFF G 3 ) Z t=0 =( X ( X β i 3) 3 DFF + β 3 FFD + β FDF e t dt i=,2 i=,2 β i 3) 3 DFF + β 3 FFD + β 3 2 FDF 2 23

25 4.2 Appendix 2 - Deivation of the Equilibium Mixed Stategy Enty Pobabilities fo the Subgames G,2, G 2,3,andG, Subgame G,2. Fo this subgame we utilize the expessions fo the value of the (sub)game and the indiffeence conditions to solve fo the enty pobabilities as ρ 2 (G,2 )V (FFF G,2 )+( ρ 2 (G,2 ))V (FDF G,2 ) = ρ 2 (G,2 )V (DFF G,2 )+( ρ 2 (G,2 ))V (DDF G,2 ) and V (DDF G,2 )=ρ 2 (G,2 ) 2 V (FFF G,2 )) + ρ 2 (G,2 )( ρ 2 (G,2 ))V (FDF G,2 ) +( ρ 2 (G,2 ))ρ 2 (G,2 )V (DFF G,2 )+( ρ 2 (G,2 )) 2 V (DDF G,2 ) Multiplying though the indiffeence condition by ρ 2 (G,2 ) then manipulating the two expessions eveals V (DFF G,2 )=V (DDF G,2 ) substituting this back into the indiffeence condition and solving povides ρ 2 (G,2 )= V (DFF G,2 ) V (FDF G,2 ) V (FFF G,2 ) V (FDF G,2 ) substituting in fo the tems V (DFF G,2 )=( β) DFF + β µ 2 FFD C =( β) DFF + β µ FDF C 24

26 V (FDF G,2 )=( β) µ FDF C + β 2 DFF =( β) µ FDF C + β DFF and µ V (FFF G,2 )=( β) FFF µ C 2 + β FFF C = FFF C povides ρ 2 (G,2 )= 2 FFF ( 2β) ³ DF F ( β) FDF FDF β DF F + C βc as epoted in the text Subgame G 2,3. >Fom the text we have the equations fo the values of the game V 2 (DDF G 2,3 )=ρ 2 (G 2,3 )ρ 3 (G 2,3 )V 2 (FFF G 2,3 )) + ρ 2 (G 2,3 )( ρ 3 (G 2,3 ))V 2 (FDF G 2,3 ) +( ρ 2 (G 2,3 ))ρ 3 (G 2,3 )V 2 (DFF G 2,3 )+( ρ 2 (G 2,3 ))( ρ 3 (G 2,3 ))V 2 (DDF G 2,3 ) V 3 (DFD G 2,3 )=ρ 2 (G 2,3 )ρ 3 (G 2,3 )V 3 (FFF G 2,3 )) + ρ 3 (G 2,3 )( ρ 2 (G 2,3 ))V 3 (FFD G 2,3 ) +( ρ 3 (G 2,3 ))ρ 3 (G 2,3 )V 3 (DFF G 2,3 )+( ρ 3 (G 2,3 ))( ρ 2 (G 2,3 ))V 3 (DFD G 2,3 ) and the indiffeence conditions ρ 2 (G 2,3 )V 3 (FFF G 2,3 )+( ρ 2 (G 2,3 ))V 3 (FFD G 2,3 ) = ρ 2 (G 2,3 )V 3 (DFF G 2,3 )+( ρ 2 (G 2,3 ))V 3 (DFD G 2,3 ) 25

27 ρ 3 (G 2,3 )V 2 (FFF G 2,3 )+( ρ 3 (G 2,3 ))V 2 (FDF G 2,3 ) = ρ 3 (G 2,3 )V 2 (DFF G 2,3 )+( ρ 3 (G 2,3 ))V 2 (DDF G 2,3 ) multiplying the indiffeence conditions by ρ 3 (G 2,3 ) and ρ 2 (G 2,3 )ρ 2 (G 2,3 ) espectively yields ρ 3 (G 2,3 )ρ 2 (G 2,3 )V 3 (FFF G 2,3 )+ρ 3 (G 2,3 )( ρ 2 (G 2,3 ))V 3 (FFD G 2,3 ) = ρ 3 (G 2,3 )ρ 2 (G 2,3 )V 3 (DFF G 2,3 )+ρ 3 (G 2,3 )( ρ 2 (G 2,3 ))V 3 (DFD G 2,3 ) ρ 2 (G 2,3 )ρ 3 (G 2,3 )V 2 (FFF G 2,3 )+ρ 2 (G 2,3 )( ρ 3 (G 2,3 ))V 2 (FDF G 2,3 ) = ρ 2 (G 2,3 )ρ 3 (G 2,3 )V 2 (DFF G 2,3 )+ρ 2 (G 2,3 )( ρ 3 (G 2,3 ))V 2 (DDF G 2,3 ) substitution the RHS of these expessions into the values of the game gives V 2 (DDF G 2,3 )=ρ 2 (G 2,3 )ρ 3 (G 2,3 )V 2 (DFF G 2,3 )+ρ 2 (G 2,3 )( ρ 3 (G 2,3 ))V 2 (DDF G 2,3 ) +( ρ 2 (G 2,3 ))ρ 3 (G 2,3 )V 2 (DFF G 2,3 )+( ρ 2 (G 2,3 ))( ρ 3 (G 2,3 ))V 2 (DDF G 2,3 ) V 3 (DFD G 2,3 )=ρ 3 (G 2,3 )ρ 2 (G 2,3 )V 3 (DFF G 2,3 )+ρ 3 (G 2,3 )( ρ 2 (G 2,3 ))V 3 (DFD G 2,3 ) +( ρ 3 (G 2,3 ))ρ 3 (G 2,3 )V 3 (DFF G 2,3 )+( ρ 3 (G 2,3 ))( ρ 2 (G 2,3 ))V 3 (DFD G 2,3 ) simplifying these educe to V 2 (DDF G 2,3 )=V 2 (DFF G 2,3 ) V 3 (DFD G 2,3 )=V 3 (DFF G 2,3 ) 26

28 using this infomation the indiffeence conditions may be ewitten ρ 3 (G 2,3 )V 2 (FFF G 2,3 )+( ρ 3 (G 2,3 ))V 2 (FDF G 2,3 ) = ρ 3 (G 2,3 )V 2 (DFF G 2,3 )+( ρ 3 (G 2,3 ))V 2 (DFF G 2,3 ) ρ 2 (G 2,3 )V 3 (FFF G 2,3 )+( ρ 2 (G 2,3 ))V 3 (FFD G 2,3 ) = ρ 2 (G 2,3 )V 3 (DFF G 2,3 )+( ρ 2 (G 2,3 ))V 3 (DFF G 2,3 ) ewiting these in tems of ρ 2 (G 2,3 ) and ρ 3 (G 2,3 ) gives ρ 2 (G 2,3 )= V 3 (DFF G 2,3 ) V 3 (FFD G 2,3 ) V 3 (FFF G 2,3 ) V 3 (FFD G 2,3 )) ρ 3 (G 2,3 )= V 2 (DFF G 2,3 ) V 2 (FDF G 2,3 ) V 2 (FFF G 2,3 ) V 2 (FDF G 2,3 )) now V 3 (DFF G 2,3 )= 3 DF F, V 3 (FFD G 2,3 )= 3 FFD C,and V 3 (FFF G 2,3 )= 3 FFF C,so ρ 2 (G 2,3 )= 3 FFF 3 DF F 3 FFD + C = C 3 FFD + C 3 DFF 3 FFD + C 3 FFF 3 FFD futhe V 2 (DFF G 2,3 )=( β) 2 DF F + β FDF, V 2 (FDF G 2,3 )=( β) h 2 FDF C i + β FFD,and 27

29 V 2 (FFF G 2,3 )= 3 FFF C so ρ 3 (G 2,3 ) = V 2 (DFF G 2,3 ) V 2 (FDF G 2,3 ) V 2 (FFF G 2,3 ) V 2 (FDF G 2,3 ) = = ( β) 2 DF F + β FDF ( β) 2 FFF ( β)c ( β) + C ( β) ³ DF F FFF FDF FDF h 2 FDF h 2 FDF C i β FFD C i β FFD which ae the solutions epoted in the text Subgame G,3. The solutions fo this subgame ae deived exactly as in the pevious case except fims 2 and change oles, we immediately have ρ (G,3 )= 3 FFF 3 DF F 3 FFD + C = C 3 FFD + C 3 DFF 3 FFD + C 3 FFF 3 FFD ρ 3 (G,3 )= ( β) ³ DF F FFF FDF FDF + C 4.3 Appendix 3 - Deivation of the Equilibium Mixed Stategy Enty Pobabilities fo the Game G,2,3. Exploiting symmety so ρ 2 (G,2,3 ) ρ (G,2,3 )=ρ 2 (G,2,3 ) 6= ρ 3 (G,2,3 ), we adopt the same method as used in appendix 2 to obtain solution equations fo ρ 2 (G,2,3 ) and ρ 3 (G,2,3 ) of the 28

30 fom (2 ρ 2 (G,2,3 )) ρ 2 (G,2,3 ) 2 V 3 (FFF G,2,3 ) +2 (2 ρ 2 (G,2,3 )) ρ 2 (G,2,3 )( ρ 2 (G,2,3 ))V 3 (FFD G,2,3 ) +(2 ρ 2 (G,2,3 )) ( ρ 2 (G,2,3 )) 2 V 3 (FDD G,2,3 ) ρ 2 (G,2,3 )V 3 (DFF G,2,3 ) 2( ρ 2 (G,2,3 ))V 3 (DFD G,2,3 )=0 V (FFF G,2,3 )ρ 2 (G,2,3 )ρ 3 (G,2,3 )[ρ 2 (G,2,3 )ρ 3 (G,2,3 ) ρ 2 (G,2,3 ) ρ 3 (G,2,3 )] +V (FFD G,2,3 )ρ 2 (G,2,3 )[ ρ 3 (G,2,3 )] [ρ 2 (G,2,3 )ρ 3 (G,2,3 ) ρ 2 (G,2,3 ) ρ 3 (G,2,3 )] +V (FDF G,2,3 )[ ρ 2 (G,2,3 ])ρ 3 (G,2,3 )[ρ 2 (G,2,3 )ρ 3 (G,2,3 ) ρ 2 (G,2,3 ) ρ 3 (G,2,3 )] +V (FDD G,2,3 )[ ρ 2 (G,2,3 )] [ ρ 3 (G,2,3 )] [ρ 2 (G,2,3 )ρ 3 (G,2,3 ) ρ 2 (G,2,3 ) ρ 3 (G,2,3 )] +V (DFF G,2,3 )ρ 2 (G,2,3 )ρ 3 (G,2,3 )+V (DFD G,2,3 )ρ 2 (G,2,3 )[ ρ 3 (G,2,3 )] +V (DDF G,2,3 )[ ρ 2 (G,2,3 )] ρ 3 (G,2,3 )=0 We Deive fist the expession fo ρ 2 (G,2,3 ) and thus need to obtain expessions fo V 3 (FFF G,2,3 )= 3 FFF C V 3 (FFD G,2,3 )= 3 FFD C V 3 (FDD G,2,3 )=V 3 (FDD G,2 ) C V 3 (DFF G,2,3 )= 3 DF F V 3 (DFD G,2,3 )=V 3 (DFD G 2,3 ) It now follows that we need to obtain V 3 (FDD G,2 ) and V 3 (DFD G 2,3 ) fom the appopiate subgames 29

31 The subgame V 3 (FDD G,2 ) the value to playe 3 of this subgame may be witten V 3 (FDD G,2 )=ρ (G,2 )ρ 2 (G,2 )V 3 (FFF G,2 )) +ρ (G,2 )( ρ 2 (G,2 ))V 3 (FFD G,2 ) +( ρ (G,2 ))ρ 2 (G,2 )V 3 (FDF G,2 ) +( ρ (G,2 ))( ρ 2 (G,2 ))V 3 (FDD G,2 ) exploiting symmety and V 3 (FFD G,2 )=V 3 (FDF G,2 ) this educes to V 3 (FDD G,2 )= ρ 2(G,2 )V 3 (FFF G,2 )) + 2( ρ 2 (G,2 ))V 3 (FFD G,2 ) 2 ρ 2 (G,2 ) now V 3 (FFF G,2 )= 3 FFF V 3 (FFD G,2 )= 3 FFD,so V 3 (FDD G,2 )= ρ 2(G,2 ) 3 FFF +2( ρ 2 (G,2 )) 3 FFD 2 ρ 2 (G,2 ) fom appendix 2 we have ρ 2 (G,2 )= 2 FFF ( 2β) ³ DF F ( β) FDF FDF β DF F + C βc 30

32 We may now conclude that V 3 (FDD G,2 )= = 2 2 FFF 2 µ ( 2β) DF F FDF +C ( β) FDF µ ( 2β) DF F 2 FFF β DF F βc FDF +C ( β) FDF β DF F µ ( 2β) DF F FDF +C 2 FFF 2 FFF ( β) FDF β DF F βc µ ( 2β) DF F FDF +C ( β) FDF β DF F µ ( 2β) DF F FDF +C 2 FFF 2 2 ( β) FDF β DF F 2 FFF βc βc µ ( 2β) DF F FDF +C ( β) FDF β DF F µ ( 2β) DF F FDF +C 2 FFF 2 2 FFF ( β) FDF β DF F 3 FFF βc 3 FFD FFF βc βc µ ( 2β) DF F FDF +C ( β) FDF β DF F βc FDF The same techniques yield V 3 (DFD G 2,3 )=V 3 (DFF G 2,3 ) and again fom appendix 2 we have ρ 2 (G 2,3 )= 3 FFF 3 DF F 3 FFD C 3 FFD + C + C and ρ 3 (G 2,3 )= ( β) ³ DF F FFF FDF FDF + C thus we have all the components epoted in the text and used in the numeical simulations. 3

33 Now we may deive the expession fo ρ 3 (G,2,3 ) >Fom ou pio calculations we have V (FFF G,2,3 )ρ 2 (G,2,3 )ρ 3 (G,2,3 )[ρ 2 (G,2,3 )ρ 3 (G,2,3 ) ρ 2 (G,2,3 ) ρ 3 (G,2,3 )] +V (FFD G,2,3 )ρ 2 (G,2,3 )[ ρ 3 (G,2,3 )] [ρ 2 (G,2,3 )ρ 3 (G,2,3 ) ρ 2 (G,2,3 ) ρ 3 (G,2,3 )] +V (FDF G,2,3 )[ ρ 2 (G,2,3 ])ρ 3 (G,2,3 )[ρ 2 (G,2,3 )ρ 3 (G,2,3 ) ρ 2 (G,2,3 ) ρ 3 (G,2,3 )] +V (FDD G,2,3 )[ ρ 2 (G,2,3 )] [ ρ 3 (G,2,3 )] [ρ 2 (G,2,3 )ρ 3 (G,2,3 ) ρ 2 (G,2,3 ) ρ 3 (G,2,3 )] +V (DFF G,2,3 )ρ 2 (G,2,3 )ρ 3 (G,2,3 )+V (DFD G,2,3 )ρ 2 (G,2,3 )[ ρ 3 (G,2,3 )] +V (DDF G,2,3 )[ ρ 2 (G,2,3 )] ρ 3 (G,2,3 )=0 so we need expessions fo V (FFF G,2,3 )=( β) ³ FFF C + β ³ 2 FFF C = FFF C ³ ³ V (FFD G,2,3 )=( β) FFD 2 C + β FDF C = FFD C ³ V (FDF G,2,3 )=( β) FDF C + β 2 DF F V (FDD G,2,3 )=V (FDD G 2,3 ) ( β)c V (DFF G,2,3 )=( β) DF F ³ 2 + β FFD C =( β) DF F ³ + β FDF C V (DFD G,2,3 )=V (DFD G,3 ) ( β)c V (DDF G,2,3 )=V (DDF G,2 ) So we need to solve fo the values of subgames V (FDD G 2,3 ), V (DFD G,3 ), and V (DDF G,2 ) The subgame G 2,3 >Fom pio calculations we have ρ 2 (G 2,3 )= 3 DFF 3 FFD + C 3 FFF 3 FFD 32

34 ³ ( β) DF F ρ 3 (G 2,3 )= FFF FDF FDF + C so + V (FDD G,2,3 )=V (FDD G 2,3 ) ( β)c ρ = 2 (G 2,3 )ρ 3 (G 2,3 ) FFF βc [ ρ 2 (G 2,3 )] [ ρ 3 (G 2,3 )] ρ + 2 (G 2,3 )[ ρ 3 (G 2,3 )] FFD βc [ ρ 2 (G 2,3 )] [ ρ 3 (G 2,3 )] ( β) FDF + β DFF [ ρ 2 (G 2,3 )] ρ 3 (G 2,3 ) [ ρ 2 (G 2,3 )] [ ρ 3 (G 2,3 )] ( β)c The subgame G,3. Again we peviously obtained ρ (G,3 )= 3 DF F ³ 3 FDF 3 FFF 3 FDF C ( β) ρ 3 (G,3 )= ³ DF F FFF FFD FFD + C so V (DFD G,2,3 )=V (DFD G,3 ) ( β)c ρ = (G,3 )ρ 3 (G,3 ) FFF ( β)c [ ρ (G,3 )] [ ρ 3 (G,3 )] ρ + (G,3 )[ ρ 3 (G,3 )] FFD ( β)c [ ρ (G,3 )] [ ρ 3 (G,3 )] [ ρ + (G,3 )] ρ 3 (G,3 ) ( β) DFF + β FDF [ ρ (G,3 )] [ ρ 3 (G,3 )] The subgame G 2 As befoe we have ρ (G,2 )=ρ 2 (G,2 ) ρ 2 (G,2 )= FFF ( 2β) ³ DF F ( β) FDF FDF β DF F + C βc 33

35 so = µ V (DDF G,2,3 )=V ρ2 (G,2 ) (DDF G,2 )= 2 ρ 2 (G,2 ) µ µ ρ2 (G,2 ) + V ρ2 (G,2 ) (FDF G,2 )+ 2 ρ 2 (G,2 ) 2 ρ 2 (G,2 ) µ µ FFF C ρ2 (G,2 ) + ( β) 2 ρ 2 (G,2 ) µ µ ρ2 (G,2 ) + ( β) DFF + β FDF 2 ρ 2 (G,2 ) µ ρ2 (G,2 ) 2 ρ 2 (G,2 ) V (FFF G,2 ) V (DFF G,2 ) µ FDF C C + β DFF this supplies all the tems necessay to compute and simulate ρ 3 (G,2,3 ). 4.4 Appendix 4 - Poof that Infomation Shaing is an Individually Rational Stategy fo keietsu Fims. We shall demonstate this esult fo the subgame G 2,3 the poof fo the subgame G,3 diffes only in notation and is omitted. We need to show that. V (FDD G 2,3 ) C < 0. Following the methods used above the value function V (FDD G 2,3 ) may be constucted as V (FDD G 2,3 )=ρ 2 (G 2,3 )ρ 3 (G 2,3 ) ( β) FFF + β 2 FFF βc +ρ 2 (G 2,3 )( ρ 3 (G 2,3 )) ( β) FFD + β 2 FDF βc +( ρ 2 (G 2,3 ))ρ 3 (G 2,3 ) ( β) FDF + β 2 DFF +( ρ 2 (G 2,3 ))( ρ 3 (G 2,3 ))V (FDD G 2,3 ) 34

36 Reaanging and exploiting the symmety between the two goup fims allows us to simplify this to + V ρ (FDD G 2,3 )= 2 (G 2,3 )ρ 3 (G 2,3 ) ( ρ 2 (G 2,3 ))( ρ 3 (G 2,3 )) FFD βc + ρ 2 (G 2,3 ) ( ρ 2 (G 2,3 )) ρ 3 (G 2,3 ) ( ρ 3 (G 2,3 )) FFF βc ( β) FDF + β DFF. Now the enty pobabilities ae as in appendix 3 ρ 2 (G 2,3 ) = ρ 3 (G 2,3 ) = DF F FFF FDF FDF ( β) ³ DF F FFF + C, FFD FFD + C. Diffeentiating ρ 3 (G 2,3 ) with espect to C gives, ρ 3 (G 2,3 ) C = FFF ( β) FFD < 0. Diffeentiating V (FDD G 2,3 ) with espect to C simplifying a little and using ρ 3 (G2,3) C < 0 gives, + V (FDD G 2,3 ) C = ( ρ 2 (G 2,3 )) ρ 2 (G 2,3 ) ( ρ 2 (G 2,3 )) 2 ( β) FDF + β DF F FFF βc ρ 3 (G 2,3 ) ρ 3 (G 2,3 ) C β C ρ 2 (G 2,3 ) ( ρ 3 (G 2,3 )) < 0. Hence shaing infomation is individually ational fo a keietsu fim. 35

37 Refeences Aoki, M., 984, Aspects of the Japanese Fim, in: M. Aoki., (Ed) The Economic Analysis of the Japanese Fim, Noth-Holland, Amstedam:. Beldebos, R. and L. Sleuwaegen, 996, Japanese Fims and the Decision to Invest aboad: Business Goups and Regional Coe Netwoks, Jounal of. Industial Oganization , 77, Dewente, K. L, and V. A.Wathe, 998, Dividends, Asymmetic Infomation, and Agency Conflicts: Evidence fom a Compaison of the Dividend Policies of Japanese and U.S. Fims, Jounal of Finance; 53 n3, Ellis, C. J, and D. Fausten, 2002, Stategic FDI and Industial Owneship Stuctue Canadian Jounal of Economics, 35 3, Feensta, R. C and G. H. Hanson, 997, Foeign Diect Investment and Relative Wages: Evidence fom Mexico s Maquiladoas, Jounal of Intenational Economics, , Flath, D., 993, Shaeholding in the keietsu, Japan s Financial Goups, Review of Economics and Statistics, 74, Flath, D., 996, The keietsu Puzzle, Jounal of the Japanese and Intenational Economies; 0 2, 0-2. Fudenbeg D., and J. Tiole, 985, Peemption and Rent Equalization in the Adoption of New Technology, Review of Economic Studies, LII, Fukao, K., T. Izawa, M. Kunimoi and T. Nakakita, 994, R7D Investment and Oveseas Poduction: An Empiical Analysis of Japan s Electic Machiney Industy based on Copoate Data, Bank of Japan Monetay and Economic Studies, 2, -60. Hoshi, T., A. Kashyap, and D. Schafstein,99, Copoate Stuctue, Liquidity, and Investment: Evidence fom Japanese Industial Goups, Quately Jounal of Economics, Kimua, Y and T. A. Pugel,995, keietsu and Japanese Diect Investment in US Manufactuing, Japan and the Wold Economy; 7 4, Linn, P. and K. Saggi, 999, Incentives fo Foeign Diect Investment unde Imitation Canadian Jounal of Economics, 32 5, McKenzie, C. R., Japanese Diect Foeign Investment: keietsu, Woking Pape Seies, Macquaie Univesity. The Role of Technology and Miwa, Y. and J. M. Ramseye, 2002, The Fable of the keietsu, Jounal of Economics and Management Stategy, 2, Ou, M., G. G. Hamilton and M. Suzuki, 989, Pattens of Inte-Fim Contol in Japanese Business, Oganization Studies, 0, Ouchi, W., 980, Makets, bueaucacies, and clans, Administative Science Quately, 25, Suzuki, K., 993, R&D Spilloves and Technology Tansfe among and within Vetical keietsu Goups: Evidence fom the Japanese Electical Machiney Industy, Intenational JounalofIndustialOganization; 4,