Modèles économétriques des configurations des aires urbaines françaises

Abstract

Econometric models of spatial urban structure : the case of French metropolitan areas. This paper presents different econometric models that account for the changes or urban spatial structures in France over the last decades. These changes are characterised by a decline of population densities in city centers and by an urban sprawl for population and employment. The monocentric model offers a relevant framework to study household location choices. This model not only identifies some factors that have an impact on these choices but also predicts that population density is represented by a negative exponential function. A two-pronged empirical approach is adopted in this article. We first estimate negative exponential population densities for each of the 112 French metropolitan areas selected from the last four population censuses (1975, 1982, 1990 and 1999). These metropolitan areas count over 8000 communes (commune is the smallest administrative entity in France). Taking the gradient of this negative exponential density function allows to measure patterns of household location over time. Any econometric work based on spatial data needs to take into account spatial interactions among observations and thus to use appropriate econometric tools. Therefore, we estimate two types of spatial econometric specification for this negative exponential density function : a spatial autoregressive model and a spatial error model with two different spatial weight matrices. Estimated spatial model specifications are assessed using different statistical tests such as Lagrange multiplier and Kelejian-Robinson tests. Compared to the conventional ordinary least squares estimation results (with no spatial autocorrelation), estimated spatial models indicate that the estimates of the density gradients and central den¬ sities are lower, hence resulting in a much smoother urban sprawl trend. In the second stage, we introduce in the previous model of population densities additional explanatory variables such as the size of central city area, income and size of households, housing benefits, the infrastructures and natural amenities of the communes. As it was done previously, different spatial models are estimated. Because of technical constraints associated with the econometric procedure, we select two samples of communes, one with communes belonging to small metropolitan areas and one representing communes of great metropolitan areas. The various statistical test results led us to accept a spatial autoregressive model specification with a spatial autocorrelation of errors. Compared to the foregoing econometric results, we can observe that these new estimated models offer a better goodness-of-fit. Furthermore, these additional explanatory variables work either as agglomerative forces or dispersion forces.