Abstract

Theoretical contributions on export behaviour highlight important aspects why and to what extend firms export their products to `foreign’ markets. Especially the model of Melitz shows that productivity differentials between firms are a key aspect to enter the export market. In the light of agglomeration economies firms might enjoy productivity gains when they are located close to competitors or upstreaming industries due to higher local demand, which leads to increasing returns at the firm level, knowledge spillovers, or other positive externalities. In such a stimulating environment firms may become more prone to be exporters. This argumentation line is the departure of this empirical research: We ask whether German firms (i.e. establishments) face higher export proportions when they are located within an intra-industrial agglomeration or within a region that offers a large variety of industries. Also, spatial spillovers emerging from the surrounding regions are explicitly taken into account. We thus test for so called Marshall-Arror-Romer and Urbanization Externalities.
For this purpose we make usage of German survey data, the IAB Establishment Panel and link it with a special draw from the entire population of all German establishments that employ people subject social security, the IAB Employment History. From the latter data source data can be constructed which controls for intra-industrial and inter-industrial regional spillover effects. It also allows considering the establishment’s workforce as an internal source for productivity gains; such as the employment of human capital or the occupational and cultural diversity. The former data source provides information on establishment’s characteristics such as revenues, innovation behaviour, the technology employed and legal status but also on export behaviour (measured as the share of total sales).
First evidence shows the existence of MAR and urbanization externalities which is derived by various estimation strategies (OLS, Panel techniques, Tobit and Fractional Response Models).