Executive summary

View the most recent version.

The growth accounting framework that is used to parse out the various determinants of economic growth focuses on changes in labour, capital, intermediate materials and an unexplained residual that is usually referred to as multifactor productivity (MFP). MFP is often attributed either to unmeasured inputs, such as managerial experience and innovative capabilities, or to externalities arising from the environment. Externalities include freely available knowledge, the social infrastructure of the economy and supporting economic structures.

Marshall's agglomeration (localization) economies fall within the latter category. Agglomeration economies include the advantages of having the correct labour mix readily available for a firm, of having other firms that can readily supply specialized inputs and having an information flow from other firms in the same industry that reduces costs or improves the quality of the product. In all cases, distance is seen to provide cost advantages in having the right mix of labour, suppliers or information available close at hand.

Measurement of the impact of these externalities is difficult. Many early studies were hampered by their use of aggregate data and poor proxies for agglomeration. To overcome these problems, this paper makes use of detailed microdata on Canadian manufacturing plants and firms that permit both productivity and associated characteristics of the production entities to be measured. The database essentially covers the entire population, thereby reducing the selection bias associated with less comprehensive databases and it is tracked over a 10-year period.

Plants can be located precisely using constant geographic codes and tracked over time. This allows us not only to examine differences across plants at a particular point in time but, more importantly, to study how changes over time in urban characteristics have influenced productivity. Examining these changes allows us to ask whether recent growth in urban economies and changes in industrial structure have in turn fed back into changes in productivity. It also allows us to ask whether simple cross-sectional results might have been the result of selection bias — that higher productivity in the firms or plants in certain areas might arise from special characteristics of those firms (fixed effects) as opposed to the characteristics of some urban economies that give rise to agglomeration economies.

The study also links two other sources of information to the microdatabase. Census data provide information about the occupational distribution of the labour force in urban areas to test the extent to which occupational matches between firms and their urban areas are related to MFP. Input–output data are used to describe the nature of linkages that are important to different industries and then the microdata are used to ask whether suppliers in industries that the input–output tables identify as suppliers of importance are located in close proximity to each plant and whether the impact on MFP of that plant suggest that supplier links at the urban level contribute to MFP. Finally, the number of plants in the same industry located in close proximity is used to test whether there are intra-industry spillovers that arise from knowledge transfer of various sorts. This transfer could come through employees moving from one plant to another and from knowledge transfer that occurs from informal or formal contacts.

The study finds that all three sources of agglomeration economies are important. At the aggregate level, our results show that plant productivity is significantly influenced by the occupational distribution of workers, the density of the buyer-supplier network and the count of own-industry establishments within the region in which the plant is located. The labour-matching effect is empirically the largest. These results substantiate and extend earlier findings from cross-sectional investigation in the United States (Rigby and Essletzbichler 2002) and Canada (Baldwin et al. 2007).

Following Rosenthal and Strange (2001, 2003) and Henderson (2003), we explore the geographical extent of the benefits that derive from the co-location of plants. As with Rosenthal and Strange (2003), these results indicate that the benefits of own-industry co-location attenuate rapidly with distance.

The results also suggest that the impact of urban agglomeration economies with regard to labour supply, specialized suppliers and knowledge spillovers is broadly felt across industrial sectors — though the impact does differ by sector, both in terms of the size of its impact and its statistical significance.

View the publication Agglomeration Economies: Microdata Panel Estimates from Canadian Manufacturing in PDF format.