LOCATION, COMMUNICATION, AND CONTROL WITHIN A VERTICALLY INTEGRATED FIRM Carlos Eduardo Lobo e Silva and Geoffrey J.D. Hewings

Transcription

1 The Regional Economics Applications Laboratory (REAL) of the University of Illinois focuses on the evelopment an use of analytical moels for urban an regional economic evelopment. The purpose of the Discussion Papers is to circulate intermeiate an final results of this research among reaers within an outsie REAL. The opinions an conclusions expresse in the papers are those of the authors an o not necessarily represent those of the University of Illinois. All requests an comments shoul be irecte to Geoffrey J. D. Hewings, Director, Regional Economics Applications Laboratory, 607 South Matthews, Urbana, IL, , phone (217) , FAX (217) Web page: LOCATION, COMMUNICATION, AND CONTROL WITHIN A VERTICALLY INTEGRATED FIRM Carlos Euaro Lobo e Silva an Geoffrey J.D. Hewings REAL 07-T-07 August, 2007

2 Location, Communication, an Control Within a Vertically Integrate Firm Carlos Euaro Lobo e Silva Regional Economics Applications Laboratory (REAL), University of Illinois at Urbana-Champaign Geoffrey J. D. Hewings Regional Economics Applications Laboratory (REAL), University of Illinois at Urbana-Champaign Abstract: The main purpose of the paper is to provie a link between the location ecision an the internal organization of firms. In the moel that focuses on the relationship among agents within the firms, the owner of the firm makes two ecisions: who will have the formal authority (the owner or the controller) an where the manufacturing plant will be locate (near or far from the owner). The results show that those two ecisions are interepenent. More than that, for certain parameters, the corporation goes from the secon best to the first best only if it changes the ecision about both aspects; changing just either the elegation scheme or location may rive it to an inferior outcome. In these cases, a flexible managerial structure turns out to be a necessary conition for firms to benefit from locational avantages. The positive correlation between the ecentralization of ecision-making an geographic ecentralization preicte by the moel is supporte by empirical research. Finally the recent heaquarter location ecision by the Boeing company is use to illustrate the results obtaine. 1. Introuction Vertical integration can be efine as a process in which more than one link of the prouction chain takes place within the bounaries of a corporation. In the international trae literature, many papers on vertically integrate multinationals consier the prouction of final goos as a two-phase process; inputs prouce in the manufacturing plant are combine with the heaquarter services to prouce the final goo. The prouction of inputs may or may not be place near the heaquarters. In the international trae terminology, when inputs an final goos are prouce in ifferent places we have a vertical multinational since each country is seen as a single point. In a seminal paper, Helpman (1984) constructs a moel of monopolistic competition with two countries, two factors, an two sectors one prouces homogeneous goos an the other prouces ifferentiate goos. He assumes that the ifferentiate goo is prouce by two processes: one uses only unskille labor (manufacturing plant) an the other uses only skille labor (heaquarter). Moreover, they can be place in ifferent countries. One of the interesting finings is that, analogous to the results in Heckscher-Ohlin moels, the region of factor price

3 Location, Communication, an Control Within a Vertically Integrate Firm 2 equalization is enlarge. It happens because the worl can now be seen as an economy that prouces three proucts: homogeneous goos, parts of ifferentiate goos with unskille labor, an parts of ifferentiate goos with skille labor. Thus, the specialization can be even more intensive when the heaquarters are concentrate in the country with an abunance of skille labor. 1 In Helpman s moel, the firms ecisions are riven by ifferences in wages. Recent papers have shown evience that the trae of inputs has substantially increase aroun the worl over the last few ecaes, a result of the increase of trae within an between firms. Therefore outsourcing an intra-firm trae seem to have ha their relevance expane in the global economy as a consequence of the fact that the prouction process has become multi-locational. Grossman an Helpman (2004), Antras an Helpman (2004), an Antras (2003) among others have use incomplete contracting to moel the ecision of firms about not only their location but also the ownership structure. Therefore, those papers, with the exception of Helpman (1984), go beyon what we propose here. The present paper takes as given the fact that the whole prouction process takes place within the bounaries of a single firm. At first sight, this restriction coul be seen as a weakness of the moel; however what is claime here is exactly the contrary. Even when the ownership structure is not uner iscussion (as in the case of Boeing s recent ecision), incomplete contracts establishe among iniviuals of a given firm remain important in etermining the optimal location strategy. Incomplete contracting provies an appropriate framework to moel ownership structures. The iea here is to go one step back an, by using the methoology of recent papers, incomplete contracting, aress the issue stuie by Helpman (1984): how o vertically integrate firms ecie their manufacturing plant location? Of course, the aim of this paper is to a one more element (internal organization of firms) to the ebate about vertical multinationals, rather than substituting any aspect previously ientifie. While Helpman s moel focuses on ifferences in factor prices, in this work the variable labele locational avantage will stan for any enhancing feature of a given region (istance from inputs an markets, wages, etc). The moel, rawing on Aghion an Tirole (1997), is base on an information structure among iniviuals of a firm. The information is essential for them an the spatial aspects, which are the focus here, can introuce inefficiency into the channels of communication. Grossman an Helpman (2004), Antras an Helpman (2004), an Antras (2003) o not take the cost of 1 See Feenstra (2004) for references on papers that point out other locational avantages besies wage ifferences.

4 Location, Communication, an Control Within a Vertically Integrate Firm 3 interaction (those of communication, controlling, or transaction) into account as the theory of firm traitionally has one. 2 Therefore, this become a seconary objective of the present paper: preparing a framework that, one ay enlarge an aopte, can be use to answer questions relate not only to location but also to ownership arrangements. Usually, moels for regional economics allow all factors to move across regions 3 that, in turn, ten to reuce the locational avantages within a country. This may explain why the literature on regional economics neglects the existence of multi-locate firms an the importance that their location ecisions might have in boosting the economic evelopment of a given unevelope regions. Fujita an Krugman (1995) provie exceptions; even though they o not inclue multi-locate firms in their analysis about the economic transition from a mono-centric economy to a multi-centric economy, they suggest the inclusion as an interesting extension of their moel. They comment: ( ) suppose that each firm consists of multiple units (e.g. HQs, R&D units, an manufacturing plants) which can be locate separately. Then, since ifferent units will follow ifferent agglomeration forces (e.g. availability of business services an convenience of face-to-face communication for HQs), we will be able to evelop a richer class of spatial moels (p.524). In the present moel, there is only one firm forme by two units: the heaquarters (HQ) an the manufacturing plant. Two iniviuals work in this firm: the owner (also calle as superior or principal throughout the paper) who resies in the HQ, an the controller (suborinate or agent) who manages the prouction an lives at the manufacturing plant. The HQ is place at the central point of the worl surroune by infinite peripheral points. The istance between any peripheral point an the central one is constant. The manufacturing plant can be locate either at the central point near by the HQ or at the peripheral point where the locational avantage for manufacturing plant is the highest one. The owner an the controller have only one task: they have to ecie the best project among n alternatives. This ecision process starts after the owner of the firm ecies two things: who will have the formal authority (the owner or the controller) an where the manufacturing plant will be place (near or far from the corporation s heaquarters). The efinition of formal 2 Even though Coase (1937) argues that there is a cost of using the price mechanism (p.390) an within a firm they are greatly reuce, he says also that the cost of negotiating contracts is still present in the firm. The analysis of transaction cost is the central part of his arguments. Regaring the nees of examining the relationship within the firm, see Simon (1991). 3 See Fujita et al. (1999) an Fujita an Thisse (2002)

5 Location, Communication, an Control Within a Vertically Integrate Firm 4 authority will be better unerstoo in the next section where the ecision-making mechanism is explaine in more etail, but for now it can be unerstoo as the right to have the final ecision about which project will be chosen. Thus, the owner eals with two trae-offs: Placing the manufacturing plant far from the owner entails, on one han, a high cost of controlling the process of prouction an, on the other han, gains in terms of locational avantages (e.g., lower wages or an costs). The elegation of power increases the incentive of the controller; however the owner has less control over the outcome. The results will reveal that the ecisions about both the location of the manufacturing plant an the internal elegation scheme are interepenent. More than that, for certain parameters, the existence of multi-locate firms (here represente by those firms with heaquarters an manufacturing plant geographically separate) can be justifie only by the existence of a flexible managerial structure. In other wors, uner those parameters, the first best is to elegate power to the controller an to place the manufacturing plant far from the owner, whereas the secon best is the opposite strategy: no elegation an locational proximity between heaquarter an manufacturing plant. Therefore, if the owner intens to rive the firm from the secon best to the first best (given a new cost of communication for instance) she has to change both aspects at once; otherwise the firm can experience the worst outcome. In iscussion an conclusion, the Boeing company s recent heaquarter location ecision is analyze an it is claime that it is a goo case stuy where elegation an location ecisions work complementarily. Moreover, the finings of recent empirical works, especially those present in Arita an McCann (2002), are compare to the preictions of the moel expose in the next sections. In section 2, the moel by Aghion an Tirole is presente. Thereafter, some aaptations an spatial aspects are introuce to ajust the moel to the purpose of the paper. Section 3 introuces the moel itself. In Section 4, an illustration is provie that anticipates the general results that are presente in Section 5. Discussion an conclusions are shown in section The Basic Moel an New Developments Formal an real authority- Following Aghion an Tirole (1997), the moel analyzes the interaction between a superior (principal) an a suborinate (agent) uner a particular circumstance: they have n possible projects in han an must ecie which one will be

6 Location, Communication, an Control Within a Vertically Integrate Firm 5 unertaken. Both agent an principal will invest some effort to learn about the project. Each one can learn either everything or nothing about it. The probability of unerstaning everything regaring the project is the amount of effort, which varies between zero an one. The principal can elegate the ecision to the agent, thus transferring, the formal authority to the agent. Otherwise, the principal keeps the formal authority with her. 4 Here, it is important to unerstan in more etail what formal authority an real authority mean in this moel. The ecision process can be seen as a game with (at most) four perios. In the first perio, as mentione above, the principal makes two ecisions: who will have the formal authority an where the manufacturing plant will be place. In the secon perio, each one ecies her own level of effort to be mae in orer to learn about the projects an, then, they learn either everything or nothing. Next, after the learning process, the iniviual who has the formal authority that is efine in perio 1 plays in the thir perio. Assume that player X has the formal authority. If player X has learnt everything in perio 2, she chooses the project an the ecision is mae (in this case, there will not be the fourth perio). Otherwise, player X gives player Y the real authority, i.e., the right to choose the project. Thus, in the last perio, player Y will not choose any project only if he or she has not learnt anything in perio 2. In this case, the outcome will be zero for both players. Otherwise if player Y knows everything about the projects, player X will rubber-stamp player Y s ecision. Note that, in this moel, both players tell the other the truth when they learn nothing. It happens because the authors assume that there is a project with a very negative expecte payoff; this will make the expecte outcome of choosing a project without knowing anything about it negative. Finally, there is some ivergence between the interests of the principal an the agent. If the principal chooses the project, her payoff is B, but when the agent makes the final ecision, the principal s expecte payoff will be a fraction of B. The reverse is true: the agent s expecte payoff is lower for the case in which the principal ecies. The original moel can be represente by the next four equations. Equations (1) an (2) represent, respectively, the utility of the principal an the agent uner no elegation, i.e., when 4 A contract efines who owns the formal authority. The resiual rights whose efinition can be foun in Grossman an Hart (1986) are etermine via allocation of formal authority. The formal authority can be seen as the assets in the relationship. The party that has the formal authority can ecie uner unexpecte circumstances. Regaring the relationship between ownership of assets an incentives, Hart an Moore (1990) provie a very iactic example on page 1122.

7 Location, Communication, an Control Within a Vertically Integrate Firm 6 the formal authority belongs to the principal. Equations (3) an (4) present the utilities uner elegation. Interpreting equation (1) will assist in making all the other equations clear. E an e represent the principal an agent s efforts, respectively. The payoff of the principal is B when the best project is chosen accoring to her preferences an! B when the agent chooses the project. g(e) measures the cost of effort mae to unerstan the project an it is a convex function. The marginal cost is zero when effort is zero an it is infinite when the effort is one. u " EB # (1 $ E) e! B $ g ( E) (1) p p u " E% b # (1 $ E) eb $ g ( e) (2) a a u " e! B # (1 $ e) EB $ g p ( E) p (3) u " eb # (1 $ e) E% b $ g ( e) (4) a a Note that in equation (1), EB is the probability of ientifying the best project times the payoff that this project yiels to the principal. The effect of elegation on the principal s payoff can be seen in equation (3). (1-e) precees EB; which means that the choice of the principal will be unertaken only if the agent oes not learn anything. Otherwise, the agent will choose the project an it is represente by the first term of the equation. Aaptations an spatial aspects As mentione in the introuction, the owner an the suborinate form the firm. We can think of the suborinate as the controller of a manufacturing plant. The controller an the plant are always locate in the same place, whereas the owner is in the HQ, i.e., at the central point. Aitionally, given the ecisions regaring elegation an location, the owner chooses the level of effort for the task of learning about the project. The story takes place in a worl with a central point where the owner of the firm is locate encircle by infinite points. The istance between any peripheral point an the central one is the same. Each point might represent a particular city. In Aghion an Tirole s (1997) moel, the agent ecies the effort base not only on the formal authority arrangement, but also on the effort expene by the principal. The moel that will be presente here oes not take the secon aspect into account. The assumption behin this simplification is that the controller oes not know how much effort the owner is carrying out. Thus, the utility function of the agent will not be consiere in this moel. Nonetheless, although the agent s maximization problem is not presente explicitly, it is assume that she

8 Location, Communication, an Control Within a Vertically Integrate Firm 7 respons to the formal authority scheme efine. Instea of etermining the amount of effort of the agents by solving their maximization problems, it will be exogenously etermine for each formal authority scheme. Appenix A provies some arguments to justify this simplification. A istinction between the present moel an the one constructe by Aghion an Tirole (1997) is that, now, the principal is the owner of the firm an the analyses will be focuse on her utility function. Thus, the utility function of the principal is compose by the monetary return of the projects an the cost of effort carrie out by her. As mentione, the choice of placing the manufacturing plant far away from the owner brings avantages an isavantages to the firm. We assume that the central point is the best location for the owner, but this may not necessarily be the case for the manufacturing plant. The only sure avantage of having the manufacturing plant close to the owner has to o with the efficiency of communication between the owner an the controller. Therefore, in other wors, the isavantage of placing the plant far from the owner is that communication is less efficient or one coul consier this to imply that the cost of implementing an efficient communication system is higher. The creation an use of the Internet is a goo example of increasing the efficiency of long-istance communication, even though it is still less efficient than negotiations between epartments locate sie-by-sie. Jones an Kierzkowski (2003) point out that some profoun prouctivity improvements in service links have been witnesse in the last ecaes. Moreover, they say that the changes in communication costs have probably been the most significant in lowering the service costs require to co-orinate spatially separate prouction fragments (p.16). There is a crucial aspect of the moel that shoul be unerstoo. The owner is seen as a receptor of information coming from the manufacturing plant. The owner learns from that information an, therefore, any problem of efficiency of communication will hurt her unerstaning. Moreover, the only effort involve in this communication is the owner s. The controller oes not nee any communication to learn about an evaluate the projects since the project is relate to the manufacturing plant controlle by him in loco. Therefore, even when the controller an the owner are far from each other, the controller oes not have any aitional cost

9 Location, Communication, an Control Within a Vertically Integrate Firm 8 to learn about the project because the source of information is always close to him. 5 As a consequence, no effort of the owner means there is no communication between her an the controller. In that case, the controller will have the real authority. The question, now, coul be why shoul the manufacturing plant be locate far away if this oes not seem to be esirable for the owner? As we have seen in the first section, some reasons can be avance to explain why firms implement the multi-locational system. The ifferences in prices of factors across locations are examples of those reasons. Here, we assume those avantages o exist; however, we o not investigate which ones they are for a particular firm. Instea, in an imaginary process, we take all the possible reasons into account an rank the cities (infinite cities in the continuous case) accoring to their appropriateness for the placement of the manufacturing plant. 6 We can normalize this inex, assuming that the appropriateness of locating the plant in the city of the owner is zero. As note earlier, there is no special reason to believe that the city of the owner presents more competitive avantages than any other in terms of those elements analyze here. Thus, one shoul expect to have some cities with positive an others with negative inices. Since the istance between the central point where the owner is locate an any other point is the same, the ecision turns out to be either keeping the manufacturing plant at the central point where there will be the highest efficiency in communication but at the cost of giving up the chance of choosing the best place for the manufacturing plant, or placing it in the city that presents the highest inex at some cost in terms of the efficiency of communication. This is the basic traeoff for the owner in terms of location choice. Showing how this trae-off interacts with the elegation ecision is the main propose of the next section an the core of the present paper. 3 The Moel Depening on owner s ecision about the formal authority, her utility function can be represente as follows: 7 5 Although the owner nees information that comes from the manufacturing plant, there is no correlation between the learning processes of the owner an the controller. We can justify this fact by assuming that, first, both nee objective information to learn but they interpret it by themselves, separately. Secon, the owner can communicate to any worker an, thus, the controller cannot limit the owner s knowlege. 6 Note that the (es)avantage in terms of internal communication is not taken into account here. This aspect will be treate separately following what was sai in the paragraph above. The criterion of the rank takes in consieration the traitional reasons for vertical multinational. 7 w oes not epen on E since it can be seen as, for example, the ifference in controllers wages.

10 Location, Communication, an Control Within a Vertically Integrate Firm 9 u " E B # (1 $ E ) e! B #& ln(1 $ te ) # w (5) n n n n n u " e! B # (1 $ e ) E B #& ln(1 $ te ) # w (6) If the manufacturing plant is locate far from the owner, part of the effort unertaken by the owner is lost in the process an, then, t is greater than one. Otherwise, t is equal to one an there is no loss of efficiency. w represents the gain of having chosen the best city to place the manufacturing plant without consiering the effects of the loss of effort. Therefore, w will be positive if owner an controller are in ifferent places an zero otherwise. In Equations (5) an (6), & has a positive value an can be seen as a sort of fixe cost of the learning process. It is inclue in the utility function so that the results can be algebraically simpler. From (5) an (6), we erive the level of effort carrie out by the owner uner both formal authority schemes: E E n B(1 $ en! ) $ t& " B(1 $ e!) t B(1 $ e ) $ t& " B(1 $ e ) t n where t = 1 if the manufacturing plant an owner are locate at the same city an t > 1 otherwise. The next step is to check which arrangement will be chosen by the owner. There are four alternatives: 1) no elegation an single location (ns); 2) no elegation an multi-location (nm); 3) elegation an single location (s); an 4) elegation an multi-location (m). First, we will compare the first two. Plugging (7) into (5) yiels: ' & + uns " en! B # B(1 $ en! ) $ & #& ln ( ) B(1 $ en! ), * (9) B(1 $ en! ) ' t& + unm " en! B # $ & #& ln t ( # B(1 $ en! ), w (10) ) * Therefore, the conition that the owner prefers (ns) to (nm), whenever the efforts are positive, 8 is: ( t $ 1) uns $ unm " B(1 $ en! ) #& ln(1/ t) $ w - 0 (11) t From (7), it is possible to see that: 9 8 If t is high enough, the effort of the owner of a multi-locate firm is going to be zero. As a consequence, the ecision turns out to be inepenent on t. (7) (8)

11 Location, Communication, an Control Within a Vertically Integrate Firm 10 te E E nm ns ns B(1 $ en! ) $ t& " B(1 $ e! ) B(1 $ en! ) $ & " B(1 $ e! ) n n Equations (12) an (13) yiel to: nm (12) (13) - te (14) Therefore, because of the imperfect communication in (nm), the owner carries out less effort than she oes uner perfect communication. Thus, the probability of learning about the project is smaller in (nm) for two reasons: the communication is imperfect an, as a consequence of that, she carries out less effort. Now, we can analyze the conition escribe in (11). The first term is positive an represents the greater probability of learning uner (ns) than uner (nm). The secon term, the ifference in cost, is negative, since t is greater than one. The reason was alreay mentione; the real effort uner (ns) is greater than uner (nm). Finally, w is positive an, given the sign, the gain of choosing the best place to locate the manufacturing plant favors alternative (nm). It is worth examining closely how the conition (11) respons to variations in t an!. The analysis about! is straightforwar. A high! means that the interest of the controller is similar to the owner s interest. Therefore, as! increases, the incentive of the owner to make an effort ecreases in both cases single an multi-location. Thus, the ifference in terms of efficiency of communication becomes less important, favoring the multi-locate firm. The effect of t on the conition (11) shoul be compute taking into account the restriction that oes not allow the effort of owner to be negative. The partial erivative shown in (15) is true when the conition (15 / ) is satisfie. Otherwise, the effort of the owner is zero an an increase in t oes not affect the firm s outcome.. / 0u. 1 / 0 uns $ unm B $ e nm n! $ t& " $ " t 0t t if. / 1 n (15) B $ e! - t& (15 / ) an 9 Note first that uner no elegation an multi-location (equation 12), te nm is the real effort mae by the owner. Moreover, uner no elegation an single location (equation 13), t =1.

12 Location, Communication, an Control Within a Vertically Integrate Firm 11. u u / 0 $ ns nm 0t otherwise. " 0 As shoul be expecte, a larger t favors the single-locate firm. 10 Again, if the inefficiency of communication is alreay high enough, the effort of the owner of the multilocate firm is going to be zero. As a result, the increase in t will not have any effect on the locational ecision. The intuition is that when the inefficiency of communication is high enough, in a multi-locate firm the real authority always belongs to the controller who analyzes the project in loco. Since there is no communication between the controller an the owner, the value of t oes not matter. The comparison above oes not take into consieration the possibility of elegating formal authority. Now, the iea is the reverse. On the one han, this assumption will be relaxe, allowing firms to ecie their internal structure an, on the other han, the possibility of locating the manufacturing plant in a ifferent city from the owner s is eliminate. In other wors, the iea is to compare the situation (ns) where there is no elegation an a single-locate firm an (s) where there is elegation an a single-locate firm. This new comparison is similar to the one in Aghion an Tirole s (1997) moel since the spatial aspect is not taken into account in either case. Nonetheless, in their paper, the cost function is ifferent an, more importantly, they assume that the effort of the controller epens not only on the formal authority scheme, as is the case here, but also on the effort of the owner. They o not show explicitly the solution of the problem an these two ifferences mentione make the moel algebraically solvable. Using Equation (8), consiering t = 1 an w = 0, effort uner (s) will be: E s B ". 1 e / B. 1$ e / $ $& Plugging (17) in (8): (16) (17) 10 In more etails, initially keeping everything else constant - as the efficiency of the long-istance communication goes own, the expecte pay-off of the project for the owner of a multi-locate firm reuces, say, in k units. Then, the owner iminishes her effort, which, in turn, has two consequences: (1) another reuction in her expecte pay-off of the project an (2) a reuction in the total cost of effort. The secon consequence reuction in total cost overcompensate the first. Finally, the net result for the owner of a multi-locate firm will be negative, but the loss will be less than k.

13 Location, Communication, an Control Within a Vertically Integrate Firm 12 ' & + us " e! B #. 1$ e / B $ & #& ln ( $ ) B. 1 e /, (18) * So, the last step is to check the ifference in terms of utility between the two arrangements, using to that en equations (9) an (18). The comparison yiels the conition for the owner to prefer (ns) to (s): 1 1$ e 2 uns $ us ". 1$! / Be #& ln 3-0 1$ en! / Then, it follows: 0 uns $ us & en " $ e B # 7 0 0! 1$ e!. / 0 uns $ us 0us & " $ ". 1$! / B $ 0e 0 e 1$ e n As coul be expecte, equation (20) inicates that when! increases, the interest of the controller is more similar to the owner s an the elegation of formal authority is more likely to be implemente. 11 Equation (21) has an ambiguous result. The increase of effort of the controller has two effects: (1) the owner has less control over the ecision; an (2) the owner makes less effort an that, in turn, saves costs. We can see that equation (21) may be positive for high e an negative for low e. 12 (19) (20) (21) Basically, the intuition is that the response of the owner to the changes in e takes into account the absolute value of e, since they are uner elegation. Then, the higher the e the less relevant the owner s effort. As a consequence, for large e, any increase in e generates a strong owner s reaction, a large reuction of effort, which saves costs. 4. Introucing the Results: an Illustration Following the sequence presente above, imagine a worl where the internal structure of the firm is given. In other wors, the owner cannot elegate formal authority. The only ecision will be about location. Then, the inverse will be explore: the owner can ecie about elegation, but there is no possibility of placing the manufacturing plant in any other city. Finally, the thir step 11 The proof that the erivative is always negative can be foun in Appenix B. 12 It is clear that if (1 $!)B 7 &, the increase of e always favors the situation with elegation of formal authority.

14 Location, Communication, an Control Within a Vertically Integrate Firm 13 is to allow her to make a ecision about both aspects. The parameters are chosen in orer to provie the main result of the paper. It might mean a loss of generality, but it will not be claime here that the location ecision is always etermine by internal structure; rather, as seen in the introuction, the argument inicates that the internal structure may be an important source of ispersion of a given firm. Assume the following values for the parameters: & " 0.5; B " 4; w " 0.8; e e n " 0.4; " 0.6;! " 0.75; t " 1.8 B is equal to 4 in orer to guarantee a positive effort. Therefore, w shows that placing the manufacturing plant in another city yiels for the firm a gain of 20% of the maximum obtaine by it. The effort of the controller ranges from 0.4 to 0.6 when the formal authority is transferre to him. The choice of the controller provies 75% of the total to the owner. The efficiency of communication within a multi-locate firm has a loss of aroun 45% if it is compare to the face-to-face communication, that is, about 45% of the owner s effort are waste through longistance communication. In this worl, the formal authority is given to the owner. The ecision is about whether or not to place the manufacturing plant in another city. As will be reveale below, the parameters favor the single-locate alternative. u u " 2.49 (22) ns " nm Since we have several parameters, there are multiple ways to interpret this result. The result shown in (22) might happen because either t is high or w an! are low. Now compare the two single-locate options: with an without elegation of formal authority. u u " 2.32 (23) ns " s Again, the ns alternative presents better results.

15 Location, Communication, an Control Within a Vertically Integrate Firm 14 Therefore, at first sight, we coul conclue that, given those parameters, single-locate an no elegation are more esirable than, respectively, multi-location an elegation. However, if we allow the owner to ecie about both aspects, the results are ifferent as shown in (24): u u " 2.7 (24) ns " m Hence, given these parameters, the best alternative is to place the manufacturing plant far from the owner as long as the owner is allowe to elegate the formal authority to the controller. Otherwise, she will locate the plant close to her. Table 1: Utilities of the four arrangements for ifferent values of t T Non elegation single Non elegation multi Delegation single Delegation multi Assuming that t is the only parameter that significantly changes over time, table 1 shows the best option for ifferent levels of communication efficiency; it also helps with the unerstaning of the intuition behin the results of the illustration. Note that single-locate firms have constant utility because the value of t oes not matter. In aition, for t = 1, the ifference between utilities uner the same elegation scheme is w since there is no inefficiency of communication.

16 Location, Communication, an Control Within a Vertically Integrate Firm 15 The key point of table 1 is that u m > u nm for large t. This happens because when the formal authority belongs to the owner, she carries out more effort than uner elegation as reveale in equations (7) an (8) an, thus, the loss of efficiency has more impact on u nm. Note that uner no elegation an for t between 2.2 an 2.5, the gain w of moving the plant to the other city oes not offset the loss of efficiency in communication, but the possibility of elegating formal authority combine with the gain w oes offset. 5. Results In this section, the main result will be presente an it will be seen that the interaction between formal authority an location of manufacturing plants always moves in a unique irection as communication costs change, i.e., the possibility of elegating power is never a reason to have a single-locate firm. This property oes not epen on the parameters. Finally, there is a iscussion about the range of the parameters that provies us the special transaction: from nonelegation an single-locate to elegation an multi-locate arrangement. Graphically, we can see how each type respons to variations of t. Note first that: 2. $! / & 0u nm B 1 en " $ # 2 0t t 2. $ / & 1 " $ # 2 t t t 0u m B e 0 t (25) (26) It easy to check that (25) an (26) will never be positive because the conition for having positive effort is:. / B $ e! - t& (25 ) 1 n. / B $ e - t& (26 ) 1 Otherwise, (25) an (26) will be zero. Since & is less than one, the erivatives will be negative for any positive value of effort an zero otherwise. Note that (25) is greater than (26) in absolute value (25 is more negative than 26). This result happens because uner no elegation, the effort is bigger than uner elegation an, consequently, a ecreasing t has a stronger impact on the no elegation arrangement. Thus, the example presente before can be illustrate in figure 1, where any

17 Location, Communication, an Control Within a Vertically Integrate Firm 16 evelopment in communication ecreases t, therefore, the tenency over time is to go through the graph from the right to the left. Utility a c U nm U m b U s U ns h g f e t=1 t t j t Figure 1: From ns to m The horizontal axis shows the value of t. When t is equal to one there is no inefficiency of communication. For t > t, E m is zero an for t > t, E nm is zero. Note also that: a " bc " w (27) jg " B! e # w (28) je " B! e # w (29) n eg " B! e $ e ) (30) ( n First, (27) reveals that when there is no inefficiency in communication, multi-locate firms are always better off than single-locate ones for obvious reasons. It is possible to see in (30) that when t is high enough, elegation is always better than no elegation since the owner s effort is zero in both cases. In that case, the best the owner can o is to provie incentives to the controller to work as much as possible.

18 Location, Communication, an Control Within a Vertically Integrate Firm 17 Aitionally, it can be seen in figure 2 why the reverse never happens. In other wors, figure 2 shows that the possibility of elegation never makes the owner of a multi-locate firm change to a single-locate one. In other wors, the transition never goes from (nm) to (s). Utility U nm U m U ns U s t=1 t t t Figure 2: The impossibility of going from s to nm As can be seen from figure 2, if U nm > U m for any t, implies that u ns > u s for all t since (27) an u ns an u s are not functions of t. Therefore, if no elegation an multi-locate firm is the best option for any t, then, there are no values of t that make elegation an single-locate firm the best alternative. 6. Discussion an Conclusion Puga an Trefler (2003) analyze carefully, from the perspective of their moel, the change in the organizational structure of the Boeing Company announce in March Briefly, the company ( ) promote the three existing unit heas to chief executive officers an geographically separate the corporate heaquarters from all three business units (p.23). Then, they conclue that the control of innovation was given to the unit heas as they were promote to chief executive officers. It is worth noting that, in contrast to the moel presente here, their moel aims to link the elegation of control over the prouction process with knowlege

19 Location, Communication, an Control Within a Vertically Integrate Firm 18 creation. They are not concerne about locational avantages. 13 Nonetheless, the moel presente here iffers from the one propose by Puga an Trefler (2003), even though both moels are base on Aghion an Tirole (1997). 14 In their moel, Boeing s ecision to relocate part of the firm is seen as a way to guarantee the elegation of power to the business units. In contrast, it is claime that in the present moel, rawing on the comments of the presient an vice-presient of the company to support the argument, that the elegation an the location ecisions are interconnecte aspects an their net effects are taken jointly by the company as the present moel preicts. Puga an Trefler (2003) note that the simultaneous announcement of these two ecisions [relocation an promotion] an the explanation given by Boeing s Chairman at the news conference mae it clear that the relocation of Boeing s corporate heaquarters ( ) was not just about locating more centrally within the Unite States; it was mainly a commitment to elegating control over incremental knowlege creation (p.23). In orer to show that their interpretation is correct, Puga an Trefler recommen reaing the recoring of Boeing s news conference of 21 March 2001 in The recoring reveals the Chairman s concerns about having a creative an innovative company. It is worth saying that the investors of the company forme the auience at that time. He comments: Yesteray, we took another step own this path of transformation. First, we announce that we will operate from a new Worl Corporate Center. Secon, we name the heas of our three largest business units as CEOs of those core units. These moves, like the ones before, are aime at only one thing; i.e., to create value. On 18 September 2002, one year after Boeing s corporate managers ha move to Chicago, the Executive Vice Presient of the company Laurette Koellner analyzes the future of 13 As in Munroe et al. (1999), Locational avantages are realize by locating prouction accoring to access to particular markets or by taking avantage of regional wage ifferentials (p.17). 14 Other recent working paper base on Aghion an Tirole (1997) has been one by Acemoglu et al. (2005). They construct a very sophisticate continuous-time moel to analyze the relationship between technology an ecentralization. Geographic aspects with ifferent costs of learning are not aresse. Besies, Marian an Verier (2002) also use Aghion an Tirole (1997) to iscuss market power in a general equilibrium approach.

20 Location, Communication, an Control Within a Vertically Integrate Firm 19 Boeing in a talk entitle Boeing: Yesteray, Toay an Tomorrow. About the new location, she says: Here at Worl Heaquarters we also have the office of ethics an business conuct, the law epartment, communications, finance, international relations, an the office of technology. ( ) In May of 2001, when we announce the selection of Chicago as the site of our new Boeing Worl Heaquarters we marke an important milestone in the transformation of the company. Boeing is working towar long-term growth an value creation. ( ) We chose Chicago in part because after the mergers of 1996 an 1997, Chicago became a location central to Boeing operating units as well as close to the financial community. It provie easier access to our customers all over the worl, an it provie a iverse professional talent base in a business frienly environment ( ) it was separate from existing operations, so that leaership woul not be ientifie with any one business unit. This was part of a strategy in which the business unit CEOs base in St. Louis an Washington State - are responsible in an autonomous way to manage their business, while allowing Boeing Worl Heaquarters to concentrate on strategy an the evelopment of people. Puga an Trefler point out that our aim is not to explain the specific choice of Chicago (p.23). However, as is evient from the speech above, there was a clear reason for moving to Chicago, to where the office of technology was move as well. Therefore, it is ifficult to figure out which reason was ominant: the necessity of being closer to the financial community with easier access to their customers or the necessity of giving autonomy to the unit heas. It seems to be clear that there were two avantages of moving to Chicago an Puga an Trefler s moel incorporates just one of them. Inee, it may be an oversimplification to see the ecision of moving the HQ to Chicago only as a commitment to elegating control over incremental knowlege creation (p.23) in the business units locate in Seattle. Therefore, what we can erive from this story is that: (1) the geographic separation between unit heas an managers generates more autonomy for the former an that, in turn,

21 Location, Communication, an Control Within a Vertically Integrate Firm 20 provies greater incentives to them to achieve greater prouctivity an value for the company; an (2) it was important to locate the HQ close to the financial community (locational avantage). Both conclusions fit well in the present moel s assumptions. 15 The fact that both ecision were mae jointly, increasing at the same time the autonomy of unit heas an the geographic ecentralization, fits well in the moel s preictions. There are some stuies that, instea of examining a specific case, have empirically investigate the heaquarter-subsiiary relations in some inustries (see, for example, Muambi 2002). Of particular importance for the propose here work by Arita an McCann (2002) that analyzes the electronics an semiconuctor inustry an tries to link the internal structure of the American an Japanese firms with the location of their assembly plants. Accoring to the authors (p. 360), the Japanese organizational arrangements are constructe within a strict hierarchical system with very little iniviual autonomy, whereas US firms have a greater egree of ecision-making latitue. Therefore, it may be seen that there is more elegation in the American structure than in the Japanese corporations. 16 With this assumption, the moel propose here woul expect there woul be a tenency for the Japanese plants to be locate closer to heaquarters, whereas American plants woul be more aggressive in placing their plants where the locational avantage appears to be higher. Arita an McCann (2002) comment that ( ) the US firms are much more spatially ifferentiate an internationally integrate than the Japanese firms, in the sense that the activities are istribute more wiely accoring to both location an activity types (p. 359). Stuying the US an Japanese automobile inustry, Shear (1983) foun the same results in terms of spatial organization. The present paper claims that the internal organization of firms is an important aspect to efine location even when the ownership structure is not taken in consieration. It has been shown that the introuction of communication inefficiency between owners an controllers offsets a small gain of placing the manufacturing plant far from the heaquarters. Therefore, the owner will ecie to move the plant only if the gain is greater than a minimum value. On the other han, the possibility of elegating the ecision to the controller reuces, for a range of the 15 Moreover, even though the financial community oes not belong to the firm, it is interesting to note that the nees of placing the HQ in Chicago captures the iea of the importance of face-to-face communication incorporate here as well. 16 If we assume that the Japanese managers elegate less power than the Americans because of the management culture in Japan, we coul introuce, in the moel, an aitional cost for the Japanese managers to elegate power. By oing so, the moel obtains the same result as we see in semiconuctor inustry: the US firms will be more spatially ifferentiate an internationally integrate.

22 Location, Communication, an Control Within a Vertically Integrate Firm 21 parameter set, the minimum gain require. As has been emonstrate, the elegation of formal authority may make multi-locate firms much more likely. From the results, one shoul expect that (1) plants an epartments place far from the owner have more autonomy than those close to the heaquarters not only because of the inefficiency of communication per se but also because the owner carries out less effort as a response to this inefficiency. Moreover, (2) there is a tenency that plants place in ifferent cities than the owner s have more not only real but e facto formal authority. Therefore, the framework incorporates some aspects that shoul be taken in consieration when multinationals (regional) ecie their strategies in terms of location. The results of the present moel call attention to the fact that there is a tenency for units locate far from the HQs to have greater autonomy. It coul provie an aitional reason to believe that for a less evelope region that wants to expan its inustrial sectors via attraction of manufacturing plants, investment in services (e.g., universities to eucate the labor force an services in general to attract highly-eucate workers) coul prove to be more efficient than investment either in infrastructure or in incentives programs given irectly to inustrial corporations (e.g. tax abatements). Finally, as mentione, there are some stuies that have empirically ientifie the positive relation between geographic ispersion an ecentralization of ecision-making. This relation seems to be quite intuitive. In contrast, the present paper tries to offer a formalization of this phenomenon. Appenix A In Aghion an Tirole (1997), Equations (A1) (A4) represent the utility functions of the principal an the agent. u " EB #. 1$ E/ e! B $ g. E/ (A1) p p u " E% b # (1 $ E) eb $ g ( e) (A2) a a

23 Location, Communication, an Control Within a Vertically Integrate Firm 22 u " e! B # (1 $ e) EB $ g p ( E) p (A3) u " eb # (1 $ e) E% b $ g ( e) (A4) a a The authors assume that g p (E) an g a (e) are increasing an strictly convex. Then, when the principal has the formal authority, the controller maximization problem is: ' (1 E) b ga ( e) $ " (A5) If the controller has the formal authority: ' (1 % E) b ga ( e) $ " (A6) Thus: e - e (A7) n Aghion an Tirole justify the result as follows: Delegation thus increases the agent s initiative; because the principal cannot overrule the agent, the agent has more incentives to become informe (p.12). This interpretation is incorporate in the present moel. The ifference is that, in equations (A5) an (A6), E woul be a constant. It can be unerstoo from the perspective that the agent ha an expectation about the effort of the owner uner those two elegation schemes. Thus, the expectation of E woul be incorporate in e, instea of its actual value. Note that, in their moel, E is lower when the agent has the formal authority; therefore, in this case, not only can the principal not overrule the agent, but also E is low. Both consequences of the elegation increase the agent s initiative an, consequently, e. In other wors, incluing the effect of the actual effort of the owner on the agent s effort woul only intensify the ifference between e an e n. Hence, assuming that e > e n seems to change neither the iea of the original moel nor the main results of the present work. Equation (21) inicates that:. / 0 uns $ us & en " $ e B # 7 0 0! 1$ e! We have seen that: n Appenix B (B1)