How sensitive is city size distribution to the definition of city? The case of Spain

Abstract

In this paper we want to test whether the choice of different types of urban data for the same country exerts an influence or not on the selection of the best parametric density function (among the Pareto, truncated lognormal, the double Pareto lognormal and mixtures of lognormals) to describe the city size distribution. We have employed four different definitions of city for Spain. We have concluded that the outperforming density is different for each type of data.