On the relative importance of intermediate and non-intermediate goods for FDI location : a new approach

Agglomeration economies reflect proximity and are an important explanation for industrial location. They feature prominently in the theories of location, including intermediate inputs and labour of the new economic geography and knowledge spillovers in the new growth theory. However, while there is a good supply of theories, there is much less evidence on the relative importance of these explanations for location. In this paper, the relative strength of agglomeration economies on industrial location is examined by adopting a new approach to measure the non-intermediate goods at a sub-national industry level. It is based on an Input-Output transaction table, which not only incorporates the ‘core Input-Output table', showing the exchange of goods and services between industries, but the flow of goods and services to or from agents outside of the industrial sector. An advantage of the approach is that it permits all possible non-intermediate goods to be included, and measures these on a comparable basis to the intermediate goods. Further, it is shown that basing the basing the measurement of the intermediate goods solely on the core Input-Output table yields biased estimates. The empirical investigation utilises data on 13,000 investments by foreign-owned plants across the regions of Great Britain over the period 1985-2007. These data are available for manufacturing and service industries, and they relate to the location decision. The paper distinguishes between intermediate and non-intermediate goods that are backward or forward in nature to explore the relative importance of production linkages and knowledge spillovers. Calculated as elasticities, the paper finds that backward intermediate goods are relatively more important for manufacturing and forward intermediate goods for services, but that the opposite is the case for non-intermediate goods. Overall, knowledge spillovers are of less importance, but indicate a greater transferability of knowledge across service industries. The empirical results are plausible and consistent with the literature, and show that care should be taken in the construction of the input-output terms. They indicate that when discriminating between competing theories it is important to distinguish between the broad industrial sector, where different processes appear to be at work that reflect the intrinsic nature of these.