رقابت و انباشت سرمایه انسانی: یک نظریه تخصصی سازی درون منطقه ای و تجارت

We consider a model with several regions whose technological ability and factor endowments are identical and in which transport costs between regions are non-negligible. Nonetheless, certain goods are sometimes produced by multiple firms, all of which are located in the same region. These goods are then exported from the regions in which their production is agglomerated. Regional agglomeration of production and trade stem from two forces. First, competition between firms for the services of trained workers makes it easier for workers to recoup the cost of acquiring industry-specific human capital. Second, the technology of production is more efficient when plants are larger than a minimum efficient scale and local demand is insufficient to support several firms of that scale. We also study the policy implications of our model.

Regional agglomeration in the production of goods is pervasive: the Swiss specialize in the production of watches and chocolates; Silicon Valley in computers; and shoes, haute couture, and movies originate disproportionately in Italy, Paris, and Hollywood respectively. Regional specialization extends even to product subcategories. Within automobiles, luxury sedans tend to come from Germany, sports cars from Italy, and reliable forms of basic transportation from Japan. The existence of regional agglomeration suggests a different rationale for cities than that which is provided in Krugman (1991). He assumes that manufactured goods are costly to transport and subject to economies of scale in production. The result is an equilibrium where the entire array of manufactured goods is produced in small geographical areas called cities. The role of the city is to keep the producers and the consumers in close proximity so as to avoid transportation costs. In fact, as our previous examples suggest, many manufactured goods are produced in one city and consumed in another. It is thus possible that cities are not a consequence of transportation costs but a consequence of having many people do similar work in close proximity. In this view, cities exist only because particular industries demand regional agglomeration. The standard explanation of regional agglomeration (e.g. Marshall, 1920, Melvin, 1969 and Markusen and Melvin, 1981; Ethier, 1982) is that, for given inputs, the output of an individual firm is larger the larger is the aggregate output of other firms producing the same good in the same region. So, for example, the level of inputs required by a new watchmaker to produce a given output is lower if that entrant locates in Switzerland where there are other watch manufacturers. Romer (1986) and Lucas (1988) have shown that external returns of this type can also explain the fact that some nations seem to remain forever more advanced than others. A possible source of external economies of this kind is the spillover of knowledge i.e., the possibility that knowledge acquired by one agent can be used by others. In order for knowledge spillovers to provide a compelling explanation, however, it must be the case that they are somehow localized. If an engineer in Taiwan can reverse engineer a product of Silicon Valley as easily as an inhabitant of Silicon Valley, there is no good reason for regional concentration of computer companies. Marshall (1920) posits instead that the external economies arise from proximity to specialized inputs. As noted by Helpman and Krugman (1985), unless there is a natural comparative advantage for the production of these inputs in the region, this explanation is incomplete. The puzzle is simply rolled back to the previous production stage: Why do the producers of inputs locate in the region? Our theory is that the equilibrium locations of firms and their input suppliers are interdependent. Thus, as in the Heckscher–Ohlin–Samuelson model, exporting firms are located where inputs are abundant. The key novelty is that inputs become abundant wherever there are several exporting firms. The reason is that input suppliers would not choose to invest ex ante in the accumulation of the capital necessary to supply the inputs efficiently unless they have several potential customers. The resulting competition among customers assures suppliers of a fair return on their investment. In the absence of such competition, the relatively immobile suppliers would be subject to the monopsony power of the downstream firms. Foreseeing that this monopsony power would be used to drive down input prices, input suppliers simply refrain from investing. This critical role of competition in securing a return to suppliers is one of the elements in Porter’s (1990) broad treatise on regional agglomeration. For concreteness, the particular input we focus on is industry specific human capital which is costly for individuals to acquire, such as the specific hand–eye coordination needed to cut diamonds or the skills which facilitate the creation of a new chocolate concoction. If trained workers can choose among several potential employers, they will be paid as a function of their marginal product. By contrast, if there is only one potential employer, there is no reason for this monopsonist to pay employees who have obtained training any more than untrained employees earn (in this industry or elsewhere) unless it is able to contractually commit itself to do so ex ante. If it cannot, the hold-up problem described by Williamson (1975) arises. Confronted with the prospect of a single potential employer, workers do not find it worthwhile to accumulate human capital. Moreover, if entry by firms is costly, firms will themselves refrain from entering if they can expect to be the only firm in the industry. The industry can then only exist with several closely located competitors. This conclusion hinges critically on our assumption that workers choose their training before they have had any formal relationship with the firm. We are thus ruling out any initial long term contract which guarantees the workers a high wage if they do become trained. Similarly, we are ruling out arrangements in which the firm trains workers at its own expense. Our assumption that these alternative arrangements are impossible is only a convenient simplification. Similar conclusions follow in the more realistic setting where such contracts are possible but involve a variety of costs which are absent when (as in the case of multiple firms) the workers make their own training decisions. Such costs arise because long-term contracts that specify future payments as a function of worker training are hard to enforce and because it is difficult for firms themselves to provide the appropriate training. Consider first the ‘solution’ where the firms assume responsibility for the training of the workers. The immediate problem with this is that the workers may not all be equally suited for training or may need different types of training. For instance, training may improve the skill of some workers but not that of others. A firm that commits itself to compensate workers for their training costs may have to compensate all workers who are trained including those who turn out to be less capable ex post. 1 By contrast, competitive firms who are not contractually committed to any workers can effectively screen out less capable ones. Knowing they will be subject to effective screening, the workers in turn self select according to their abilities. That is, if workers know whether training will improve their skill ex ante while firms only discover this ex post, competition induces the right workers to obtain training. Thus a monopsonistic industry may be less efficient at production than a competitive one. Similar problems arise if the firm signs a contract committing it to pay a high wage to workers who obtain training. The difficulty here is in defining ‘training’ in a way that is contractually implementable. If the contract only specifies that a specific training course must be taken, then the difficulty is the same as when the firm provides the training course itself. Instead, the firm might try to write a contract that specifies a required level of acquired skill. The problem here is that it is much more difficult for a third party to verify skill itself than the completion of some training course. So, the workers cannot be sure that the firm will not attempt to exploit them ex post by claiming that they are insufficiently skilled and therefore do not qualify for a skilled wage. For most of the paper we focus on the case where firms are unable to commit themselves contractually so that monopsonists are unable to produce. Later, we expand our model to allow for contracts so that monopsonists can produce but are inefficient and show that our main results are unaltered. We follow this sequence in the introduction as well, describing first the no-contracting setting. If there is a minimum efficient scale below which each firm cannot operate profitably, the necessity of having several competing firms for an industry to be viable implies that the region’s output may have to be substantial. In particular, demand in the producing region itself may be insufficient to accommodate the requisite number of firms. Then, the only way of ensuring competition among firms is to have several of them locate in one region and produce for the world market. In equilibrium, trade emerges between spatially separated regions with the same endowments and access to the same technology even though there are transport costs and it is technologically feasible to produce all of a region’s consumption locally.2 Competition plays a role in our model of regional agglomeration that is related to the role that it plays in Dudey’s (1990) model of the clustering of retail stores. Whereas in our model competition stimulates the supply of inputs, it stimulates demand in Dudey’s model; consumers searching for a product are attracted to shopping areas where there are many vendors of the good because they rationally anticipate that competition will cause prices to be lower there. Despite the greater competition, stores may prefer to be located near each other because the costs of competition are outweighed by the benefits of a larger number of customers. Since Dudey’s model operates on the demand side of the market, however, it cannot provide the basis for a model of trade. Because each firm needs other firms to be present for workers to become trained, it might be thought that our model has multiple equilibria in some of which no firm produces. This would be true if we insisted that the entry decisions of firms all be made simultaneously. Instead, following Farrell and Saloner (1985), we model firms as making this choice sequentially so that, through their actions, they can communicate their intentions to each other. This makes it impossible for equilibria with no production to coexist with others in which production is positive. One of the most important implications of the traditional theory of trade based on external returns is that nations can be made worse off as a result of trade (see Graham, 1923 and Ethier, 1982). In our model such losses from trade are possible as well. These losses are intimately linked to the existence of multiple equilibria. In some equilibria (the agglomerated equilibria) only one region produces the good even though, in autarky, the good is produced in both regions. At these agglomerated equilibria the importing region can be worse off. In our model this happens because there are transportation costs so that an imported good costs more than a locally produced good. In Ethier (1982) it occurs when one region produces nothing other than the good subject to external returns. This can lead its price in terms of the other good to rise relative to autarky (even though it is produced more efficiently) because factor demands rise in the producing region. We show that, in our context at least, the equilibria with losses from trade are not robust. They, again, depend crucially on the absence of any mechanism that allows the agents to tell each other that they would like to produce the good subject to the externality.3 Formally, losses from trade can occur in our model if workers must make their decision whether to become trained simultaneously. To capture the possibility that workers can communicate their intentions to each other we consider a variant of our model in which workers become trained in sequence. In this case, the equilibrium is unique and trade can only be beneficial. One difference between our theory and the traditional external returns approach is in the role ascribed to antitrust policy. In the traditional theory, relaxation of antitrust policy can be socially desirable. Cooperation among firms can lead them to internalize whatever externality leads the production by one firm to lower the input requirements of the others. This logic has led Jorde and Teece (1989), for example, to conclude that antitrust exemptions are essential for certain US high-technology industries to succeed in a world scale. By contrast, in our theory as well as in Porter (1990), society benefits from competition. The more competition among firms the potential suppliers of labor expect, the more willing they are to make industry specific investments. Thus, a vigorous antitrust policy can play an important role in promoting the creation of viable export industries. Since the presence of other firms is necessary for each firm to have access to suitable workers in our model, our theory involves an externality. However, unlike the alternative theory of trade based on external returns, we do not require that the presence of other firms lower the input requirement for producing output. Instead, the technology for producing output is the same in all regions. An unrealistic feature of the version of the model that does not allow for contracting is that it is impossible for a firm to exist as a monopsonist. This seems to fly in the face of such examples as Boeing (the sole manufacturer of commercial aircraft in Washington State) and Hershey (the sole chocolate manufacturer in Hershey, Pennsylvania). The version of the model with contracting can accommodate monopsony. Indeed monopsony in each region is the unique equilibrium in that model if demand in each region is not too large. Once demand exceeds a certain threshold, however, because monopsony is inefficient, the unique equilibrium is agglomeration and trade as in the model without contracting. To see this intuitively, start from a position of monopsony in each region. The entry of a second firm in one region increases the efficiency of this region so that it can export the good at lower cost (including transportation) than the cost of the monopsonist in the other region. Efficient agglomeration drives out monopsony. Section 2 presents the simple partial equilibrium setting in which workers decide simultaneously whether to acquire industry specific human capital. Section 3 embeds this model in general equilibrium and considers trade among ex ante identical regions. That section has several subsections in which we discuss the patterns of trade that emerge as the number of goods and the number of regions that trade varies. In one of these subsections we present our argument that if workers decide whether to become trained in sequence, every region benefits from trade. In Section 5 we consider the policy implications of our theory. In particular we study industrial policy, tariffs, and antitrust. Industrial policy encompasses those policies that governments pursue to affect the location of industries. In our model, such policies can raise welfare in the region imposing them. The reason is that the goods subject to the externality are sometimes produced disproportionately in one region. But, the presence of transport costs implies that regions benefit from having such goods produced locally. Therefore, policies that ensure local production of these goods can be desirable from the region’s point of view. While tariffs can be a tool of industrial policy, they can also be imposed in situations where they do not affect the regional pattern of production. The usual ‘optimum tariff’ argument implies that, after workers in the other region have become trained, importers benefit from such tariffs because they improve the importer’s terms of trade. However, workers in the exporting region who foresee that tariffs will be levied, have a smaller incentive to become trained. So, the perception that tariffs will be imposed raises the equilibrium price of the good in the exporting region. We show that, as a result, tariffs which are foreseen when workers seek training unambiguously lower welfare in both regions. This strengthens Lapan’s (1988) argument against tariffs. Section 6 presents the version of the model with contracting and Section 7 concludes.

We have presented a model of regional agglomeration in the production of specific goods where the principal motor behind a region’s exports is the healthy competition among many suppliers located there. Competition ensures that workers earn high wages if they acquire industry-specific human capital which, in turn, makes human capital accumulation attractive and the industry viable. The main message from the model is that even where it is technologically possible to obtain the same allocation with trade as without trade, trade serves a useful role. It allows industries to operate on a sufficiently large scale that it is possible to have several firms producing the same goods in one location and thereby reap the benefits that flow from regional agglomeration. While we have focused on the salutary effect of regional agglomeration on the abuse of monopsony power, there may be other reasons why regional agglomeration enhances human capital accumulation. That is, it is an open question whether the mechanism by which having several local firms creates an incentive for human capital accumulation is through the increased competition they generate. For example, a different advantage of regional agglomeration may be that it provides some assurance to workers that they will remain employable in the industry if conditions change in the future. That is, workers may prefer it if there is a diverse range of activities in the area that use their industry specific skills in case demand conditions or production techniques change in a way that eliminates the activity the are initially employed in. This preference for diversity might exist even if long-term contracts could be written in such a way that the monopsony inefficiency we have analyzed does not arise. Consider just three specific examples. Workers employed in the production of mid-sized automobiles may want to be located near plants that produce small automobiles in case demand shifts in the direction of the latter. Or workers employed producing computers based on a proprietary operating system, but whose skills are not specific to that operating system, may prefer it if plants producing computers based on alternative operating systems are located nearby in case theirs becomes obsolete. Finally, workers could simply be concerned with the possibility that they might not get along with their supervisor or co-workers. They would then like to know that if, if this were to occur, they could easily shift to another job. While the existence of such diversity may be important to workers making their training decisions, it is not clear why it cannot be achieved within a single enterprise. That is, one firm could encompass a range of technologies, products, plants, and divisions. Multiple firms might have an advantage if workers are concerned about the possibility of bankruptcy and if two firms somehow have a combined probability of bankruptcy that is lower than those firms would have when rolled into one. Conceivably, workers might like to have access to a diversity of ‘corporate cultures’ and this might be difficult to achieve in separate divisions of a single company. Another open area for research is the extent to which competition is actually associated with high wages, extensive industry specific training and exports. One problem is that, in practice, it is hard to gauge when an industry is relatively competitive. Nonetheless it is worth providing some anecdotal examples which appear to support the model. First, centrally planned economies have little actual competition for workers between the various firms and are notorious for their inability to export high wage goods. Second, consider the automobile industry. Japan, the most successful exporter of high quality mass-produced cars, has a relatively large number of firms in this industry. Similarly, Italy has several producers of high performance cars which are very successful exporters. In contrast, Italy has only one large mass-production auto manufacturer, FIAT, whose exports to the US are minimal. Interpreted within the context of our model, the high performance Italian cars and the mass-produced Japanese cars can be thought of as high quality goods which use relatively skilled workers while FIAT can be thought of as a lower quality producer who employs less skilled workers. These anecdotes suggest that a more careful exploration of the empirical validity of the model is warranted.