Monte Carlo evaluation of Moran I for limited dependent variable models (500)

Theme Track: Methods of Spatial Analysis - Spatial Econometrics: Specification and Testing

Authors:
Florax, Raymond ; van der Vlist, Arno

In this paper we study the actual distribution of the Moran I test statistic in finite samples for limited dependent variable models. As data are often observed in binary form, limited dependent variable models has been widely used in applied econometrics. Standard limited dependent models however does not allow for spatial processes, neglecting possible interdependences between actors. The Moran I test statistic is the most popular test for detecting spatial correlation. Though, asymptotic properties have recently been formally derived by Pinkse (1999) or Kelejian and Prucha (2001) the literature on small sample properties of the Moran I for limited dependent variable models are limited. This paper aims to extend the current literature by studying the actual distribution of the Moran I in finite samples.

[Home]    [Sitemap]