Parametric, semiparametric and nonparametric models of urban growth

Abstract

This paper discusses parametric, nonparametric, and semiparametric models of urban growth. To illustrate differences across approaches, we test Gibrat's law in the long run, using the three methods and three different datasets: Spanish capital cities and regions (1900–2011, annual data) and US MSAs (1900–2000, decennial data). Our results reveal that the estimation of the relationship between growth and initial size can significantly vary across methods. We suggest and encourage the use of semiparametric methods in future research of urban growth. • This paper discusses parametric, nonparametric, and semiparametric models of urban growth. • We test Gibrat’s law in the long run using the three methods. • Three different datasets are considered: Spanish capital cities and regions, and US MSAs. • The estimated relationship between growth and initial size can significantly vary across methods.