An Application of a Switching Regimes Regression to the Study of Urban Structure

[Introduction:] Analyses of urban structures are traditionally conducted using population and employment density functions, which describe the behavior of these densities as a function of the distance from the Central Business District (hereafter - CBD). The three most critical assumptions underlying this approach posit that the urban area is monocentric, land around the CBD is perfectly uniform in all directions and at all distances and housing capital is perfectly malleable. The assumptions are well rooted in a large body of theoretical modeling, its origin dated back to the seminal works on urban structure by Muth [1969] and Mills [1979]. These authors established that introducing the location choice of economic units in spatial models by using only one variable - the distance from the CBD - rather than two variables - distance and direction - is critical in greatly simplifying the analysis of otherwise extremely cumbersome analytical models of urban structure. Moreover, as pointed out by Mills and Hamilton [1994, p. 127], in spite of this complexity, models which introduce two-dimensional description of location yield only few analytical results. Notwithstanding the theoretical appeal of the strong uniformity and malleability assumptions which have been instrumental in sharpening our analytical understanding of the major forces which operate and interact with each other to shape urban structures, throughout the years an increasing number of scholars began to raise serious questions on the empirical validity of these assumptions. These studies are clear in showing that failure of each assumption to hold true will produce distorted estimates of density functions, thus invalidating the standard approach. (...)