The agglomeration forces of national and multinational companies: are they different? The case of Italy

This paper aims at understanding whether national and multinational firms have different agglomeration patterns, which may derive from their different idiosyncratic characteristics that are likely to affect their location choices. By using the Figueiredo, Guimaraes and Woodward (2007) version of the Ellison and Glaeser (1997) index, we analyze the distribution of almost 600.000 plants across 686 geographic units (Local Labor System) in 78 Italian manufacturing industries in the year 2001. This index has been used to computed the agglomeration within multinational enterprises (MNEs), the agglomeration within national companies and the co-agglomeration between national firms and MNEs. By using the WLS econometric technique to control for heteroskedasticity, we correlate the main centripetal and centrifugal forces identified by the literature with the agglomeration index computed for national firms, the agglomeration index computed for MNEs, and the co-agglomeration index computed for national and multinational firms, in order to check whether the signs and the significances of the determinant of agglomeration change across the different categories of firms. We find that competition is the only driver of agglomeration that does not change across the different categories of firms, since it acts as centrifugal force both for national and for multinational enterprises. Conversely, information costs and skilled labor are firm-specific drivers of agglomeration, since they push only MNEs to cluster together while they have no effect on the national companies. Indeed, on the one hand, MNEs tend to replicate the location choices of other MNEs in order to decrease the information costs they support when they enter the host country, while national firms do not face such costs because they already know their home market. On the other hand, MNEs 2 cluster in the sectors with high human capital in order to undertake labor pooling, provided that they can afford to attract the best workers, while national firms are afraid of labor poaching and prefer not to cluster next to MNEs in order to avoid a loss of skilled workers. We also find that, contrary to our expectations, transportation costs are not a significant determinant of agglomeration. Codified spillovers, which arise from knowledge that is reported on official documents such as patents, is not a significant driver of agglomeration either, since patents do not require interaction to acquire their knowledge and, hence, they do not trigger agglomeration. Conversely, spillovers that arise from tacit knowledge, which have been measured through an innovative index that takes into account the absorptive capacities and the interactions that occur within and among sectors, discourage both the clusters of national enterprises and the co-agglomeration between multinational and national companies. Indeed, tacit spillovers imply not only absorption but also a transfer of knowledge, which occurs from the most to the least advanced national firms, in the first case, and from MNEs to national firms, in the second case. At the same time, tacit spillovers do not act as centrifugal force for the agglomeration of MNEs, because the latter constitute a homogenous group of advanced firms that can both gain and lose from the unintentional transfer of knowledge, with a net not significant final impact. These results may interest first of all the industrial organization literature, which hardly ever distinguishes between national and multinational firms when dealing with agglomeration. The identification of the firm-specific drivers of agglomeration may also help policy makers to set up differentiated territorial policies according to the type of firms they want to move. Finally, the managers of national and multinational companies themselves may be interested in such results, which can help them to better understand the dynamics of agglomeration and, hence, to make more consciously their location choices.