Quantification of non-linear effects in agglomeration economies for transport appraisals

Abstract

Agglomeration economies arising from the spatial concentration of economic activity have been known to exist and induce higher productivity for firms. The existing empirical evidence, however, has two key caveats. First, it mostly assumes a pre-specified (mostly log-log) functional form for the relationship between firm productivity and agglomeration. Second, it may lack valid instruments to adjust for potential confounding biases (for instance, from the omission of characteristics of local input and output markets) in the estimation of this relationship. This study adopts a flexible Bayesian Non-Parametric Instrumental Variables based approach to quantify non-linear effects in agglomeration economies. The approach uses innovative external instruments derived from traffic casualty data. We adopt a two-step framework: we first isolate the firm’s total factor productivity from a Cobb-Douglas production function and thereafter estimate the non-linear effects of agglomeration on this productivity. Using data from a sample of firms classed into six key industry sectors in England, we present novel evidence that indicates the presence of significant non-linearities in agglomeration elasticities for most industry sectors. Our results provide critical inputs for the appraisal of transport investments.