A CRITICAL NOTE ON RECENT EMPIRICAL STUDIES BASED ON SPATIAL AGGLOMERATION MODELS

Abstract

 We analytically study a discrete-space version of the spatial model of agglomeration proposed by Allen and Arkolakis.¹⁾ Employing the bifurcation analysis method developed by Akamatsu et al.,¹⁴⁾ we show that the model exhibits, if any, only mono-centric or unimodal spatial concentration patterns of mobile agents. The properties we reveale in this paper pose one simple but significant issue: Can we justify fitting the model to the actual spatial patterns, which are typically not mono-centric but poly-centric? We argue that, the unobserved amenity approach, which Allen and Arkolakis and other recent studies employ to circumvent the issue, is an inappropriate one as it can replicate any observed data with any model.