2014/01 LEM Working Paper Series

A numerical estimation method for discrete choice models with non-linear externalities

Giulio Bottazzi and Fabio Vanni
Externalities, Heterogeneity, Computational methods, Firm localization

  JEL Classifications
C12, C13, C46, C52, R12

This paper presents a stochastic discrete choice model with non-linear externalities and a related methodology to carry out inferential analysis on its parameters. Such framework allows to disentangle the role of the intrinsic features of alternatives from the effect of externalities in determining individual choices. The stochastic process underlying the model is demonstrated to be ergodic, so that numerical methods are adopted to estimate the parameters by Chi-square minimization and evaluate their statistical signicance. In particular, optimization rests on computational procedures that consider also iterative methods such as successive parabolic interpolation. Comparisons among various computational techniques and simulation implementations are analyzed. For completeness, the Chi-squared is compared to the approach of maximum likelihood optimization, underling the advantages of the first in this model.
download pdf