Choix discret dynamique et spatial: Utiliser le GMM à une étape: Application aux décisions opérationnelles dans le secteur minier

In many spatial applications, agents make discrete choices (e.g. operating or product-line decisions), and applied researchers need econometric techniques that enable them to model such situations. Unfortunately, however, most discrete-choice estimators are invalid when variables and/or errors are spatially dependent. More generally, discrete-choice estimators have difficulty dealing with many common problems such as heteroskedasticity, endogeneity, and measurement error, which render them inconsistent, as well as the inclusion of fixed effects in short panels, which renders them computationally burdensome if not infeasible. In this paper, we introduce a new estimator that can be used to overcome many of the above-mentioned problems. In particular, we show that the one-step ('continuous updating') GMM estimator is consistent and asymptotically normal under weak conditions that allow for generic spatial and time series dependence. We use our estimator to study mine operating decisions in a real-options context. To anticipate, we find little support for the real-options model. Instead, the data are found to be more consistent with a conventional mean/variance utility model.