Daan van Soest is grateful to the Netherlands Organization for Scientific Research (NWO) for financial support of the PRET research program. Shelby Gerking acknowledges the hospitality of CentER at Tilburg University where this chapter was written. He would also like to thank the Netherlands Organization for Scientific Research (NWO) for financial support (visiting grant B46-386). This chapter has benefited from careful comments by Erwin Bulte, Arjen Gielen, Henri de Groot, Jan Lambooy, John List, Bart Los and Willem van Groenedaal. Because of the lack of data on sectoral output and the capital stock at the city level an appropriate measure of total factor productivity cannot be constructed. Glaeser et al. (1992) built a small model in which output is produced with only one input, labour, under conditions of decreasing returns to scale. Then, technological progress enhances the marginal value product of labour and hence the demand for labour increases. In that model, assuming constant prices for inputs and outputs, employment growth is an appropriate indicator of output growth. Note that the variable measuring sectoral national growth rates outside the city would be virtually the same for each observation in the Henderson et al. (1995) analysis. Henderson (1997) finds that effects of agglomeration economies on employment growth peak after about 5 years and die out after 6-7 years. Thus, for both data sets, the time interval over which employment growth was measured appears to be long enough to allow measurable data to emerge. See also Combes (2000). Combes (2000) does not agree that the COMPETITION variable as constructed by Glaeser et al. (1992) is a proper measure of the degree of competition an industry faces. However, given that this variable measures the impact of relative firm size on employment growth, he argues that it can be used as a test for the importance of internal economies of scale; he proposes measuring competition by the inverse of a local Herfindahl index of productive concentration. A related analysis was also performed using the 580 municipalities data set (i.e., after including the smallest municipalities) with similar results to those presented in table 3. These and all other results that are described, but not explicitly reported in the text, are available from the authors on request. These results differ from those obtained by Combes (2000) in his analysis of (regional) employment growth in France. For manufacturing industries, he finds that (i) diversity slows down employment growth, (ii) specialisation hardly matters and (iii) smaller firms grow faster (where size is measured in terms of the number of employees per firm, which coincides with Glaeser et al.’s (1992) COMPETITIONN measure). Note that the two wage variables could not be used in the individual industry analysis as they have no variation within a sector. Although no zip code-specific sectoral wage data is available, the Dutch Central Bureau of Statistics distinguishes five regions in this province (so-called COROP regions) for which it calculates average sectoral wages. Pearson correlations of sectoral wages between regions range from 0.76 to 0.86.