Testing for causality between productivity and agglomeration economies

The productivity effects of agglomeration economies are often treated as endogenous in empirical work due to the potential for reverse causality. The extent to which these relationships are actually simultaneously determined, however, remains largely unobserved. This paper estimates panel data vector autoregressions for different sectors of the economy to test for bidirectional causality between productivity and both localization and urbanization economies. The aim is to address some key questions that will help to identify the extent of the endogeneity problem. Can we actually observe bidirectionality in the data? Does it feature more for some industries than for others? Is it more prevalent for localization or urbanization economies? The results show that agglomeration economies are not strictly unidirectional and that higher levels of productivity can induce growth in the scale of local urban and industrial environments. The paper discusses the difficulties that these issues pose for the estimation of agglomeration economies.