Faculdade de Economia, Administração e Contabilidade de Ribeirão Preto Universidade de São Paulo Texto para Discussão Série Economia TD-E 12 / 2011 A Importância da Organização Interna da Firma para o Processo de Fragmeção Espacial da Produção: uma Simulação Prof. Dr. Carlos Eduardo Lobo e Silva Av. Bandeirantes, 3900 - Monte Alegre - CEP: 14040-900 - Ribeirão Preto-SP Fone (16) 3602-4331/Fax (16) 3602-3884 - e-mail: cebelima@usp.br site:www.fearp.usp.br Av. Bandeirantes, 3900 - Monte Alegre - CEP: 14040-900 - Ribeirão Preto - SP Fone (16) 3602-4331/Fax (16) 3602-3884 3884 - e-mail: cebelima@usp.br site: www.fearp.usp.br
Universidade de São Paulo Faculdade de Economia, Administração e Contabilidade de Ribeirão Preto Reitor da Universidade de São Paulo João Grandino Rodas Diretor da FEA-RP/USP Sigismundo Bialoskorski Neto Cefe do Departamento de Administração Marcos Fava Neves Cefe do Departamento de Contabilidade Adriana Maria Procópio de Araújo Cefe do Departamento de Economia Walter Belluzzo Junior CONSELHO EDITORIAL Comissão de Pesquisa da FEA-RP/USP Faculdade de Economia, Administração e Contabilidade de Ribeirão Preto Avenida dos Bandeirantes,3900 14049-905 Ribeirão Preto SP A série TEXTO PARA DISCUSSÃO tem como objetivo divulgar: i) resultados de trabalos em desenvolvimento na FEA-RP/USP; ii) trabalos de pesquisadores de outras instituições considerados de relevância dadas as linas de pesquisa da instituição. Veja o site da Comissão de Pesquisa em www.cpq.fearp.usp.br. Informações: e-mail: cpq@fearp.usp.br Av. Bandeirantes, 3900 - Monte Alegre - CEP: 14040-900 - Ribeirão Preto-SP Fone (16) 3602-4331/Fax (16) 3602-3884 - e-mail: cebelima@usp.br site:www.fearp.usp.br
Managerial institutions as a determinant of te spatial dispersion of te economic activities Carlos Eduardo Lobo e Silva Abstract Te relationsip between link-service costs and te locational distribution of economic activities as been widely recognized. However, little (if anyting) as been done to formally sow te importance of managerial arrangements for te map of production. In te model, vertically integrated firms make two interdependent decisions: te location of teir establisments and teir managerial arrangements. Once close establisments and delegating decisions work as substitute goods for communication services, tose decisions determine te intensity of use of communication services. One of te interesting features of te model is tat communication services are taken as a substitutable production factor. From te interaction of te firms decisions, urban systems presenting some of te penomena tat ave been identified by te empirical and teoretical literature emerge. Production fragmentation and functional specialization of cities wit eadquarters concentrated in te metropolitan area are more likely to occur under decentralized management, low communication costs, low transportation costs, and standardized production process. Key words: Location Decision, Production Fragmentation, Managerial Structures, Teoretical Approac. JEL: R12, R30, D23.
1. Introduction Te relationsip between link-service costs and te locational distribution of economic activities as been widely recognized. However, little (if anyting) as been done to formally sow te importance of managerial arrangements for te map of production. Tis paper claims tat building te link between management and location decisions not only introduces one more important aspect to te debate, but also allows us to enric te teoretical treatment of intrafirm communication costs, wic will be considered as a substitutable production factor. Wile te consequences of reductions in transportation costs ave been extensively studied and modeled 1, communication costs ave been incorporated into te models as a given production cost tat appears wenever eadquarters and plant are spatially separated. In Ota and Fujita (1993), intrafirm communication costs are fixed and firms decide te intensity of communication wit oter firms. Te autors analyze te configuration of a single city. Duranton and Puga (2005) model intrafirm communication costs by assuming Samuelson s iceberg form to explain wat tey call functional urban specialisation. Bot papers build general equilibrium models to obtain insigtful results. However, none as analyzed ow te relationsip between communication costs and managerial strategies can determine te spatial dispersion of economic activities. In te present model, vertically integrated firms take te communication cost among oter aspects - to make two interdependent decisions: te location of teir establisments and teir managerial arrangements. Once close establisments and delegating decisions work as substitute goods for communication services, tose decisions determine te intensity of use of communication services. From te interaction of te firms decisions, urban systems presenting some of te penomena tat ave been identified by te empirical and teoretical literature emerge. Te production fragmentation and te functional specialization of cities wit eadquarters concentrated in te metropolitan area are correlated to te following conditions: flexible managerial structure, low communication costs, low transportation costs, and standardized production process. 1 Te seminal paper by Krugman (1991) and te more recent work by Tabuci and Tisse (2011) are examples among many oters.
Te simulations of tis paper extend te model proposed in Silva and Hewings (2012). In te original model, te owner of a vertically integrated firm formed by te owner and te controller decides on (1) its managerial structure (centralized or decentralized) and (2) te location of te manufacturing plant (close to or far from te eadquarters). Bot decisions represent trade-offs for te firm. Once te owner as decided upon tese aspects, te owner and te controller decide teir effort to learn about questions related to te manufacturing process. Since te interests of te owner and te controller are not opposite, toug divergent, delegating te decision may work as a (imperfect) substitute good for linking services. As a consequence, firms wit strict ierarcical system and little individual autonomy tend not to spatially fragment teir production to avoid link-service costs. Te extension of te model incorporates many firms and new parameters. Besides coosing teir delegation sceme, eac firm (owner) decides te locations for bot te eadquarters and te manufacturing plant. Te key parameters of te model are (1) te cost of communication between te owner and te controller, in case tey are located in different cities; (2) te cost of transportation to deliver production to a distant market; and (3) te caracteristics of te activity, wic defines te room te controller as to deviate from te owner s interests. Given te parameters, te owners decide also considering te market size and te amount of business services of eac location, wic are endogenously determined in te simulations. As result, an urban system emerges from te interaction among te locational decisions of firms. Te simulations proposed sow tat lower link-service costs or a more flexible managerial structure tend to generate bot more spatially fragmented economic activities and more functionally specialized cities. Specifically, as production becomes more fragmented, tere is a tendency for eadquarters to agglomerate in te metropolitan area, wic is abundant in business-services, wereas some medium cities end up specialized in manufacturing. Instead of replacing any element pointed out by te literature, te contribution of tis paper is to add one more aspect te internal organization of firms to te debate about te location of vertically integrated production cains witin an urban system. Next section presents te model. Te results are found in te tird section, wic precedes te final considerations.
2. Model and equilibrium Two establisments compose eac firm: te ead-quarters and te plant of manufacture. Te owner lives at te ead-quarters, wereas te controller lives at manufacturing plant. Te owner cooses two tings: wo will ave te formal autority and te location of bot establisments. Woever as te formal autority will ave te final word in any decision. Te production presents constant return to scale and can be represented by a sequence of decisions. Eac decision process follows te steps defined in Agion and Tirole (1997): tere are V possible projects and tey ave to coose one to undertake. Eac project k! {1,..., v} is associated to a monetary gain B k for te owner (principal) and a benefit b k for te controller (agent). After te owners decisions, te owner and te controller decide te amount of effort tey will carry out to learn about te projects. By ypotesis as in Agion and Tirole (1997), tey learn eiter everyting or noting about te projects. If te person wo as te formal autority learns everyting, se cooses te best project for er. Oterwise, se asks te oter part to make te coice, since teir interests are divergent, but not opposite. If te oter part as not learned anyting tey do not coose anyting and te outcome is zero. It is assumed tere is a project wit very negative expected outcome, wic makes tem not to coose anyting in case tey learn noting. Tus, te trade-off of te owner s decision is tat giving te formal autority to te subordinate incentives er to increase er effort to learn, but reduce te owner s control over te final decision. In Silva and Hewings (2012) and ere, tose projects can be seen as possible solutions / answers for problems / doubts tat come out at te manufacturing plant. Te model proposed in Silva and Hewings (2012) interacts te trade-off coming from Agion and Tirole (1997) wit a new one: fragmenting te production into two locations allows te firm to coose te most appropriate place for eac establisment, but also increases te communication costs between te owner and te controller. Hence, te model reveals tat delegating te decision wic tends to reduce te communication between owner and subordinate - is often a condition for te firm to benefit from locational advantages troug fragmenting its production. Te extension incorporates many firms and places. Eac firm (owner) cooses te locations for te eadquarters and te manufacturing plant, and defines its internal
delegation sceme. Ten, owner and controller decide teir effort to learn about te problem. To make teir decisions, firms consider te market size and te amount of business services of eac location, wic are endogenously determined in te simulations; and te following parameters: te cost of communication between te owner and te subordinate in case tey stay at different cities; te cost of transportation to delivery its production to a distant market; and te caracteristics of te activity, wic defines te room te controller as to deviate from te owner s interests. Te utilities of te owner and te controller are defined as follows: j j j j j j U E B + ( 1 E ) e E + ln(1 ce ) + L = α (1) * j * j * j * j * j U e B + ( 1 e ) E B + ln(1 ce ) + L = α (2) u u j * j j j j j = E β b + 1 E ) e b + ln(1 e ) (3) ( * j * j * j * j = e b + (1 e ) E β b + ln(1 e ) (4) Equations (1) and (2) represent te utility of te owner, wic depends on er effort (E) and te controller s effort (e). B is te expected utility of te best solution for te problem; wereasα measures te compatibility between te interest of te owner and te interest of te controller. Te tird term sows te cost of effort and c is te communication costs, wic will be one if te eadquarters and te plant of manufacture are at te same city, and bigger tan one oterwise. Finally L is te locational advantage coming from te owner s location decisions. Asterisk means tat te final decision was delegated to te controller. Note tat, in te case te owner delegates te decision, te controller is allowed to make te decision and, only if e does not learn anyting, te controller asks te owner to participate of te decision process. Equations (3) and (4) sow te utility of te controller. b is te expected gain of te best solution for te problem. Te compatibility between te interests of bot from te controller s point of view is represented by β. terms: Te interaction of firms decisions is captured in L, wic is composed by two L = H + P p (5)
Were: H P p 2 = [ g( pop popin )] (6) N = ( 1 t) d + t d (7) p s= 1 s H and P are te locational advantages tat a firm obtains from placing, p respectively, te eadquarters at city and te plant at city p. Te locational advantage of city for eadquarters is positively related to te number of eadquarters located tere ( ) and negatively related to te congestion costs, wic presents exponential effects. Te difference between te current population of city and te initial population (common to all cities) is ( pop popin). g is a parameter for te congestion costs. Two assumptions justify te positive relationsip wit te number of eadquarters: new eadquarters attract more business services, and firms benefit from bot a great variety of business services and te presence of oter eadquarters (Davis and Henderson, 2008). Equation (7) captures te importance of locating te manufacturing plant close to te market. d n measures te size of te market to be explored at city n, wic is defined as te total population of city n divided by te number of plants in city n plus 1: d n = popn n /( m +1) (8) In equation (7), t represents te efficiency of te transport system - t equals to zero means infinite transport costs. Note in equation (7) tat wen t = 0, firms ave access only to te local market. Te simplification about te potential size of market is based on te idea tat firms profits increase wit te size of te demand of teir products, wereas profitability decreases as more sellers sare te same consumers 2. Finally, te population of city n is a function of te number of eadquarters and te manufacturing plants. 2 Tis simple idea is a common starting point in empirical models of entry, as in Berry (1992) and Bresnaan and Reiss (1991).
pop = σp + n n n (9) Were! is a parameter greater tan one. In terms of location (wit N places) and delegation sceme, te owner as 2N 2 coices. For eac of tese coices, it is possible to find levels of (no-negative) effort of te owner and te controller tat determine Nas equilibrium. Terefore, two-step process determine te outcome: firstly, te owner cooses bot te locations of te establisments and te delegation sceme and, secondly, te owner and te controller decide teir level of effort. After te first decisions, i.e., given (, i) - were p and are te locations of te plant and te eadquarters respectively, and i is te delegation sceme and as long as some conditions are satisfied, tere will be an unique equilibrium defined by te coices of effort (E i, e i ). Tus, a backward-induction strategy solves te owner s maximization problem: knowing te outcome of te equilibrium of eac locationdelegation coice, te owner cooses te set of (, i) tat maximizes er utility. Tis process describes te decision of a single firm, given te location of all oters. However, firms jointly reac te equilibrium if and only if tey ave made teir best coices given te oter firms coices. Simulations procedure - In te simulation, te number of cities and firms was arbitrarily defined as 50 and 100, respectively. Te only purpose of tis coice was to make te simulation able to reveal wat te model claims. Te distance between any two cities is te same. Besides, in te first period, tey ave te same population size and no firms. As consequence, none of te cities as business services in period zero. Tus, for te first firm, te locational advantage is zero for all locations. At te beginning of te first period, te first firm appears. Since tere is no business service in tis economy yet and te populations across cities are te same, te eadquarters of firm 1 will be located in te city were te plant is located (say, city 1) in order to avoid unnecessary costs of communications. Given tis coice, before te second period starts, te simulation updates te population and te amount of business services in city 1. Te size of population increases and te business service sector becomes positive. Now, te economy is ready to receive te second firm. Te options for te second firm are: (1) locating bot establisment at city 1 to avoid internal communication costs and to benefit from bot bigger population and
positive business services. Te disadvantage of tis coice is te presence of a competitor (firm 1) at te same city. An alternative (2) is to locate bot establisments at city 2. In tis case, it saves internal communication costs and explores te local market as te unique local firm. Te negative aspect of tis coice is to give up of a bigger population and positive business services of city 1. Finally, option (3) is to split production, locating plant of manufacture at city 2 to explore its market witout local competitor and eadquarters at city 1 to benefit from positive business services. Te disadvantage is te internal communication cost tat turns out to be positive since te production becomes spatially fragmented. After te decision of te second firm, new updating occurs and te tird firm appears. Note tat te tird firm (and all next firms) faces te trade-off described above, once no coice satisfies all te desirable objectives: placing te manufacturing plant at te city wit te biggest market to be explored, considering te number of local competitors; placing te eadquarters were te amount of business services is te maximum; and avoiding internal communication costs. Terefore, considering 50 cities and te possibility of delegating te final decision, te simulation calculates te outcomes of 5,000 options (2x50 2 ). As already mentioned, eac outcome is a Nas equilibrium, in wic te level of effort of te owner is te best response to te level of effort of te controller and vice-versa. Te location of te economic activity will be in equilibrium wen all firms ave cosen teir best strategy in terms of location, delegation sceme, and effort, given te decisions of te oter 99 firms. Tus, after all firms making teir first decisions, a new round starts. Differently from te first round, in te second round te first firm will decide, knowing te locations of oters. Te equilibrium is reaced wen te decisions of all firms made in round z were te same as tose made in round z+1. Tere is no cost of migration. It is wort empasizing tat te objective is to sow ow and wy some aspects - namely te managerial arrangements, communication and transportation costs, and te degree of standardization of te production process - can affect te distribution of production over a given region. Te parameters used do not come from estimations; teir role is to introduce into te simulation te assumptions made in te model construction. Tus, te results coming from te simulations do not aim to measure any effect; rater tey only suggest tendencies.
3. Results Before presenting te main results, it is wort to divide te outcomes into four categories for te purpose of explanation: (1) divergent urban system: all te economic activity is concentrated at a single location; (2) convergent urban system: te opposite, economic activities are evenly spread across cities; (3) poorly fragmented urban system: some firms are fragmented, but tere is ig concentration of spatially integrated firm in te metropolitan area; (4) igly fragmented urban system: many firms are fragmented, establisments concentrated in te metropolitan area are essentially eadquarters, and tere are some medium cities tat received more tan one manufacturing plant. To capture te effect of some parameters on te fragmentation process, tables 1 to 5 sow tat canges of one of te parameters transform a poorly fragmented urban system into a igly fragmented one or vice-versa. Eac table below presents te results of a simulation, and te parameters used appear in te title of te respective table. Besides, eac column sows ow many cities ave a certain caracteristics. For instance, te simulation wose results are in table 1 considers te following parameters: α = 0.75; c = 1.1; B = 4; b = 2; popin = 10; t = 0; g = 0; σ = 3 3 (t = 0 means infinite transport costs). Te tird column tells us tat 11 cities ave te same caracteristics: no eadquarters, two manufacturing plants, and 16 people living in eac one. Note tat te sum of te first line is 50, wic is te total number of cities. Moreover, te number of manufacturing plant can be obtained troug te following computation: 40x1 + 2x11 + 1x38 = 100. Table 1: Situation 1 (S1): α = 0.75; c = 1.1; B = 4; b = 2; popin = 10; t = 0; g = 0; σ = 3 # of cities 1 11 38 HQ 100 0 0 MP 40 2 1 Pop 230 16 13 Assuming tat table 1 represents te igly fragmented urban system, it is possible to find values for parameters c, t, andα tat ave te same effect on te 3 β is assumed to be zero. Tis simplifying assumption means tat te owner s coice is not better tan a random coice for te controller.
production fragmentation. Tables 2-5 incorporate tese canges. Table 2 assumes all te parameters of situation 1, except te cost of communication tat becomes equal to 1.4. Table 3 brings te results of simulation tat considers te values of situation 1, except te transport costs tat become equal to 0.7. Te next simulation (table 4) canges te value of α. Table 2: S 1 + c = 1. 4 # of cities 1 49 HQ 100 0 MP 51 1 Pop 263 13 Table 3: S 1 + t = 0. 7 # of cities 1 49 HQ 100 0 MP 51 1 Pop 263 13 Table 4: S 1 + α = 0. 4 # of cities 1 49 HQ 100 0 MP 51 1 Pop 263 13 Te values of c, t, andα were cosen in suc way tat te new values generate exactly te same effect on te distribution of te economic activities. Tables 2-4 sow tat lower communication costs, lower transportation costs, and iger degree of standardization of te production process (iger α ) incentive te fragmentation process. Wat seems to be remarkable is tat te results of tables 2-4 can also be obtained by not allowing te owners of firms to delegate te final decision to te controller (table 5). Hence, aving a flexible managerial structure - in wic te decisions can be decentralized may be a condition for te production fragmentation occurring or becoming more intense. Table 5: S 1+ no delegation # of cities 1 49 HQ 100 0 MP 51 1 Pop 263 13
Combining two canges may lead to a divergent urban system wit all production concentrated in te metropolitan area, as sown in table 6. Table 6: S 1+ no delegation + c = 1.4 # of cities 1 HQ 100 MP 100 Pop 410 Even in tis case, introducing a new value for any parameter tat favors te fragmentation process, te total divergence disappears. Table 6 and 7 reproduce te results of te same simulations, except te value of α. Table 7: S 1 + α = 0. 9 + no delegation + c = 1.4 # of cities 1 13 HQ 100 0 MP 87 1 Pop 371 13 Te second set of results comes from simulations tat include positive congestion costs (g). Because of it, te results do not converge any more, i.e., te firms decisions of round z are different from decisions of round z + 1. Tus, table 8 presents te results after 10 rounds. Te main aspect to be igligted is tat metropolitan areas cannot be as big as before, once te congestion costs offset te gains of agglomeration. As consequence, instead of only one metropolitan area, a few of tem emerge. Te number of big cities reduces (increases) as congestion costs become smaller (bigger), as is seen in table 9 (table 10). Te results presented in table 10 represent te total convergence of te urban system: all 50 cities ave te same caracteristics. Table 8: Situation 2 (S2): α = 0.75; c = 1.1; B = 4; b = 2; popin = 10; t = 0; g = 0. 1 (ten rounds) # of cities 1 1 1 1 6 40 HQ 32 30 29 9 0 0 MP 13 13 13 9 2 1 Pop 81 79 78 46 16 13
Table 9: S2 + g = 0.05 (ten rounds) # of cities 1 1 11 37 HQ 51 49 0 0 MP 21 20 2 1 Pop 124 119 16 13 Table 10: S2 + g = 0.15 (ten rounds) # of cities 50 HQ 2 MP 2 Pop 18 Even toug simulations do not converge wen congestion costs are ig enoug, table 11 suggests tat te number of rounds does not modify te results significantly. Comparing tables 9 and 10, it is possible to see tat ten and twelve rounds get almost te same results. Table 11: S2 + g = 0.05 (twenty rounds) # of cities 1 1 10 38 HQ 50 50 0 0 MP 21 21 2 1 Pop 121 121 16 13 Finally, te total convergence also occurs wen a low value of congestion costs is combined wit parameters tat make fragmentation undesirable (table 12). In tis case, firms tend to keep teir production spatially integrated. Because of congestion costs, no big city emerges, and production activities turn out to be evenly distributed across cities. Table 12: S2 + g = 0.05 + no delegation or c = 1.4 or α = 0. 4 (ten rounds) # of cities 50 HQ 2 MP 2 Pop 18
4. Discussion and final considerations Te first set of results brings te main findings of te work. Under low transport costs, more firms agglomerate teir establisments at metropolitan areas in order bot to benefit from business services and to avoid internal communication costs, once te cost of reacing distant market is low. Krugman (1991) gets similar correlation. Even toug te present model considerably simplifies te market mecanisms, wic are incorporated into te insigtful general equilibrium model proposed by Krugman (1991), ere tere is a new reason for firms to agglomerate. Considering (1) tat te production process is composed by more tan one activity - wic can be spatially separated - and (2) tat tose activities require managerial coordination, it is desirable to keep manufacturing plant and eadquarters at te same location so tat te owner can control te production process witout spending too muc on internal communication costs. Parameterα is also important for te fragmentation process. It can be understood as te room for te controller to deviate from te owner s interest. Tus, parameterα sould be a function of te difficulty of specifying in a contract te expected outcome of te controller s tasks. It typically appens to tose activities at te ends of te smile of value creation proposed by Mudambi (2007, 2008): planning, design, R&D, marketing, etc. Following te same idea, Constantin et al. (2010) justify te migration of some activities of te woodworking-furniture and footwear industries in Europe saying tat, te ig degree of standardization of operations, te specific kind of tecnologies used, and te way te process is organized, allow for te slicing of te production cain into relatively autonomous operation wic do not require spatial continuity (p.830). Additionally, Duranton and Puga (2000) utilize te empirical literature to support te idea tat innovations results of R&D activities - are made in metropolitan areas and tat city size as a positive and igly significant effect on innovative output. Te model s results fit very well in bot te case studies of Constantin et al. (2010) and te stylized fact presented by Duranton and Puga (2000). Jones and Kierzkowski (2005) explain te fragmentation process pointing out te importance of increasing returns in service-link activities instead of on te factory floor. Te autors igligt te profound productivity improvement in services links occurred over te last decades, especially in communication services: te canges in communication costs ave probably been te most significant in lowering te service
costs required to co-ordinate spatially separated production fragments (p. 16). Te present model captures te effect of lower communication costs on te managerial strategy of firms, and te results sow tat reducing tem favor te production fragmentation. Differently from te framework of Jones and Kierzkowski (2005), ere firms are formed by only two parts. However, Jones and Kierzkowski (2005) model neiter te internal organization of firms nor te interaction of firms and its effects on agglomeration forces. In te present paper, a deep fragmentation appens wen many firms spatially separate teir production, instead of - as in Jones and Kierzkowski (2005) - a single firm tat divides its production into many parts. It is claimed ere tat te effects of communication costs cannot be understood witout incorporating into te framework te processes of learning, coordinating, and negotiating since te flow of information only affects te economy trougout interactions among agents. For tis reason, te boundaries of te maximizing blackboxes ave to be broken up and Agion and Tirole (1997) offer an insigtful model to do tat. Wen managers can coordinate te production from anywere, and firms are not seen as maximizing black-boxes, firms can take advantages by placing eac establisment and department were tey benefit most from immobile resources and markets. However, te strategy of fragmenting te production is often wort it as long as te decision is delegated to te subordinate. Terefore, for some sets of parameters, a flexible managerial structure is a condition for te fragmentation process. Tat is te new and most important result of tis paper. Te intuition explaining te main result goes in te following way: in order to benefit from bot (1) agglomeration of business services in te metropolitan area and (2) local market wit low competition, firms may ave to fragment teir production. However, fragmenting te product means positive communication costs. Ten, once controlling te production process turns to be costly, owners will ave incentive to decrease teir effort and, consequently, delegate te final decision to teir subordinates. Tis model also confirms te findings coming from empirical work suc as Arita and McCann (2002) about te positive relationsip between decentralization of location and decentralization of decision. Arita and McCann (2002) analyze te electronics and semiconductor industry and conclude tat firms wose institutions allow ierarcical systems wit a greater degree of decision-making latitude tend to be muc more spatially differentiated tan a more strict ierarcical system.
Finally, Duranton and Puga (2005) offer an insigtful explanation for te fact tat some urban areas ave become functionally specialized instead of aving teir activities concentrated on a small number of sectors. Tis transition, tey say, "is inextricably interrelated wit canges in firm's organization" (p. 1). Following tis idea, te present work tries to formalize a mecanism in wic managerial institutions combined wit te development of linking services - is relevant for te process of functional specialization. Te main limitation of te model is te significant simplification proposed for te market mecanisms.
References Agion, P.; Tirole, J. (1997) Formal and real autority in organizations. Te Journal of Political Economy, v. 105, n. 1: 1-29. Arita, T.; Mccann, P. (2002) Te relationsip between te spatial and ierarcical organization of multiplant firms: observations from te global semiconductor industry. In: MCCANN, Pilip. (Ed.). Industrial location economics. Massacusetts: Edward Elgar Publising limited, p. 319-363 Berry, S. (1992) Estimation of a Model of Entry in te Airline Industry," Econometrica, 60: 889-917. Bresnaan, T.; Reiss, P. (1991): Entry and Competition in Concentrated Markets," Journal of Political Economy, 99: 977-1009. Constantin, F., Giusti, G. e Tattara, G. (2010) Strategies of Italian Firms in Romenia: Evidence from Selected Case Studies. Transit Studies Review. V. 16: 829-847. David, James C.; J. Vernon Henderson. (2008). Te agglomeration of Headquarters. Regional Science and Urban Economics. 38(5): 445 460. Duranton, G; Puga, D. (2000) Diversity and Specialisation in Cities: Wy,Were and Wen Does it Matter? Urban Studies, Vol. 37, No. 3: 533 555.. (2005) From sectoral to functional urban specialization. Journal of Urban Economics, v. 57, n. 2: 343-370. Jones, R.; Kierzkowski, H. (2005) International trade and agglomeration: An alternative framework Journal of Economics Volume 10, Supplement 1: 1-16, Krugman, Paul. (1991) Increasing return and economic geograpy. Te Journal of Political Economy, v. 99, n. 3: 483-499. Mudambi, R. (2007). Offsoring: economic geograpy and te multinational firm. Journal of International Business Studies, 38(1): 206. Mudambi, R. (2008). Location, control and innovation in knowledge-intensive industries. Journal of Economic Geograpy, 8(5): 699-725.
Ota, M.; Fujita, M. (1993). Communication Tecnologies and Spatial Organisation of Multi-unit Firms in Metropolitan Areas. Regional Science and Urban Economics, 23: 695-729. Silva, C. E. L.; Hewings, G. (2012) Locational and managerial decisions as interdependent coices in te eadquarter-manufacturing plant relationsip: a teoretical approac. Te Annals of regional Science, 48: 703-717. Tabuci, T.; Tisse, J. (2011) A New Economic Geograpy Model of Central Palces. Journal of Urban Economics. 69(2): 240 252.