de Groot, H. L. F., Poot, J. & Smit, M. J. (2008). Agglomeration externalities, innovation and regional growth: Theoretical perspectives and meta-analysis. (Department of Economics Working Paper Series, Number 1/08). Hamilton, New Zealand: University of Waikato.
Abstract
Technological change and innovation and are central to the quest for regional development. In the globally-connected knowledge-driven economy, the relevance of agglomeration forces that rely on proximity continues to increase, paradoxically despite declining real costs of information, communication and transportation. Globally, the proportion of the population living in cities continues to grow and sprawling cities remain the engines of regional economic transformation. The growth of cities results from a complex chain that starts with scale, density and geography, which then combine with industrial structure characterised by its extent of specialisation, competition and diversity, to yield innovation and productivity growth that encourages employment expansion, and further urban growth through inward migration. This paper revisits the central part of this virtuous circle, namely the Marshall-Arrow-Romer externalities (specialisation), Jacobs externalities (diversity) and Porter externalities (competition) that have provided alternative explanations for innovation and urban growth. The paper evaluates the statistical robustness of evidence for such externalities presented in 31 scientific articles, all building on the seminal work of Glaeser et al. (1992). We aim to explain variation in estimation results using study characteristics by means of ordered probit analysis. Among the results, we find that the impact of diversity depends on how it is measured and that diversity is important for the high-tech sector. High population density increases the chance of finding positive effects of specialisation on growth. More recent data find more positive results for both specialization and diversity, suggesting that agglomeration externalities become more important over time. Finally, primary study results depend on whether or not the externalities are considered jointly and on other features of the regression model specification.
