نقش سرمایه گذاری سرمایه انسانی در تصمیم گیری محل شرکت

We explore the role of human capital investments in the location decisions of firms. We show that whether human capital investments act as a force for or against concentration depends on who is undertaking them and whether they are industry or firm specific. We also discuss the empirical predictions of our theoretical analysis.

In this paper, we explore the role of human capital investments in the location decisions of firms. We argue that the location of firms influences the extent of labor market imperfections, which in turn affect the incentives to invest in workers' human capital. We show that the optimal location of firms depends on who is undertaking the investments and whether they are industry or firm specific. Thus, even a simple model that focuses exclusively on human capital investments can provide rich predictions about the spatial concentration of firms. As an illustrative example, consider the accordion industry which is almost entirely concentrated in Castelfidardo, a small city near Ancona, Italy (see Tappi, 2002).2 If one asked locals why this industry is concentrated there, they would probably reply that it is due to the great skill of the local workforce in producing accordions. But this would just lead to the next question, namely, why are all these skillful accordion workers concentrated in Castelfidardo. One potential answer, and the one we focus on here, is that workers who settle down in this medieval town when they are young have very strong incentives to invest in skills that are specific to the accordion industry. In particular, these workers do not have to worry about being held up by their employers after they have invested in their accordion making skills in as much as they can always threaten to work for another local manufacturer. In as much as they do not worry about being held up when they are old, they have strong incentives to invest in their own skills when they are young. The idea that colocation of firms can mitigate the potential hold-up problem between workers and firms and can thus induce more efficient (industry specific) human capital investments by the workers is not novel and was recently analyzed by Rotemberg and Saloner (2000).3 The main empirical prediction of their paper is that firms that use the same type of labor are located close to each other to protect the workers' human capital investments. Thus, using the terminology introduced by Duranton and Puga (2003), they predict ‘functional’ concentration, i.e., concentration of firms using similar skills, rather than ‘sectoral’ concentration, i.e., concentration of firms producing similar goods. Dumais et al., (1997) test this prediction and find evidence that “plants do seem to locate near other industries when they share the same type of labor. This effect is quite large and suggests that labor market pooling is a dominant force in explaining the agglomeration of industry” (Dumais et al., 1997, 28–31).4 A number of other papers have since confirmed the importance of labor market pooling in explaining the spatial concentration of firms (see, for instance, Rosenthal and Strange (2001), Rigby and Essletzbichler (2002) and, for a survey, see Rosenthal and Strange (2004)). All these papers take it for granted that spatial concentration of firms mitigates the hold-up problem. We agree with the authors of these papers that labor market considerations can play an important role in determining the location of firms. However, we want to caution against the view that spatial concentration by firms unambiguously improves the incentives to invest in the workers' human capital. We show below that human capital investments can act as a force for or against spatial concentration, and that their net effect on the location of firms depends on their characteristics. In contrast to the existing literature, therefore, the forces for and against agglomeration in our model are of the same nature. We hope that taking into account the circumstances under which labor market considerations act as a force against spatial concentration will further clarify the empirical evidence about their importance in explaining the spatial concentration of firms. We focus on two cases in which concerns about human capital investments can act as a force against spatial concentration. First, this will be the case if the firms rather than the workers make the human capital investments. For instance, if the accordion manufacturers had to invest in the workers human capital, it might be optimal for them to locate away from their competitors so as to protect their investments. Second, it may be advantageous for firms to locate away from their competitors if workers can make firm-specific investments. Suppose, for instance, that a particular accordion manufacturer uses some production techniques that are industry specific—that is, they are also used by other firms in the industry—and others that are specific to the manufacturer. Suppose further that a worker can decide how much to invest into learning each production technique. If a worker can work for other local manufacturers in the future, he has an incentive to invest too much into the industry-specific skills relative to the firm-specific skills so as to improve his future bargaining position. To avoid this investment inefficiency, a firm may then want to limit the ability of a worker to join a competitor in the future, and it may be able to do so by locating away from its competitors. A number of recent papers have investigated the agglomeration patterns of industries and show that they vary greatly in the degree to which they are spatially concentrated (Ellison and Glaeser (1997), Maurel and Sédillot (1999) and Devereux et al. (2004)). Using employment data, all these papers use a version of the ‘dartboard’ approach put forth by Ellison and Glaeser (1997) and find that between 75% and 95% of industries are ‘localized;’ that is, they differ significantly from a ‘dartboard’ random location, and only 15% are ‘dispersed.’ Using exhaustive data on UK plants and treating space as continuous, Duranton and Overman (2004) find that 52% of industries are localized, whereas 24% of them are dispersed. There is no shortage of theories that identify economic forces, which may contribute to the differences in agglomeration patterns across industries. In particular, in recent years, the ‘New Economic Geography (NEG)’ literature has focused on the interaction of increasing returns to scale at the firm level, imperfect competition and transportation costs as an explanation for agglomeration patterns ( Krugman, 1991, Fujita et al., 1999 and Baldwin et al., 2003). Assuming that plant size captures the extent of internal scale economies, this literature suggests that localization should be driven mostly by relatively large plants. However, using Italian data, Lafourcade and Mion (2004) show that in some industries, small plants show a pattern of localization. Similarly, Duranton and Overman (2004) stress the importance of small firms in driving the agglomeration patterns of some industries. In as much as, in contrast to NEG models, the mechanism we stress in this paper does not rely on the size of plant—if anything, it is reinforced by plants being small in as much as they are less likely to have monopsony power—it is a candidate to fill the gap left open by these models. It should be very clear though that we view our model as complementing rather than substituting existing theories of agglomeration, including the NEG models. Our paper also contributes to the literature on the microfoundations for localized agglomeration economies.5 The clustering of economic activities involves external agglomeration economies and diseconomies of various sorts. On balance, empirical studies suggest that positive externalities dominate at many levels. For instance, using US state level data, Ciccone and Hall (2002) find that the elasticity of average labor productivity with respect to employment density is equal to 6%. This elasticity is slightly smaller (5%) at the NUTS-3 level for the five largest EU-15 countries (Ciccone (2002)). At the heart of our theory lies the hold-up problem which has received widespread attention in the literature on organizational economics (see, in particular, Williamson (1985) and Hart (1995)). The standard situation that is considered in this literature is one in which a buyer and/or a seller of a good can make relationship investments ex ante and then bargain over the price of the good ex post. If contracts are incomplete in the sense that the parties cannot contract ex ante over the price of the good and investments are noncontractible, the agents are likely to underinvest or overinvest into the relationship. Starting from this observation, the literature has identified a variety of contractual and noncontractual means that allow the agents to mitigate these investment inefficiencies (see Hart (1995)). This paper is also related to the large literature in labor economics on the incentives of firms and workers to make human capital investments. In a seminal contribution, Becker (1964) showed that firms do not invest in workers' general skills if labor markets work perfectly. This theoretical prediction contrasts with the empirical observation that firms sometimes do invest in workers' general skills (see, for instance, Acemoglu and Pischke (1999a)). A number of papers have reconciled the theory with the empirical evidence by showing that, under certain conditions, firms have an incentive to invest in workers' general skills as long as labor markets work imperfectly (see Acemoglu and Pischke (1999b) and the references therein). In this paper, we argue that if it is costly for workers to change locations, then the extent of labor market imperfections and thus the incentive to invest in workers' skills depend crucially on the location of firms. We then analyze the optimal location decisions of firms taking into account the effect that location has on the labor market and thus on investment incentives.

The purpose of this paper was to analyze the role of human capital investments in the location decisions of firms and to caution against the view that human capital investments always act as a force for spatial concentration. The model we developed is stylized and simple. This allows us to highlight important, basic economic forces and analyze their effect on the location decisions of firms. Also, in spite of its simplicity, the model generates a number of hypotheses that are empirically testable. First, if human capital investments are industry specific, we would expect firms to be less concentrated if they, rather than the workers, are making the investments. Second, we would expect industries to be less concentrated if workers can make firm-specific investments than if they can only make industry-specific investments. Straightforward extensions of our model can provide additional testable implications. Suppose, for instance, that workers can make both industry-specific and general-purpose investments. We would then expect industries to be concentrated but located away from other industries. This induces workers to invest in the useful industry-specific skill and not waste effort acquiring general-purpose skills that they do not use. In the case of the accordion industry, this suggests a reason why this industry is not only very concentrated but also located far away from other industries (a typical pattern, for instance, for the Italian ‘industrial districts’). The difficulty with actually testing the hypotheses of this model is that location and human capital investments are both endogenous. One possible solution to this problem may be to exploit the exogenous differences across countries with regards to legislation that obliges firms to provide training for their workers. For instance, the model would predict that in countries such as Germany, in which firms are obliged to provide some general training, firms are more dispersed than in countries in which this obligation does not exist.11 The model could be extended in a variety of ways. For instance, one could relax our stark assumptions about the labor market by modeling labor market competition more explicitly. Also, it would be interesting to endogenize the specificity of the investments. Finally, one could analyze how the location of firms affects their willingness to pay for worker trainings.12 We leave these extensions for future research. To conclude, we believe that labor market considerations play a potentially important role in the location decisions of firms. We hope that the above analysis sheds more light on these issues and will lead to future empirical research that further evaluates their importance.