Title: Space-time autoregressive models
Authors: Charles Saunders - University of Western Ontario (Canada) [presenting]
Abstract: Spatial econometric models are estimated via MLE and GMM methods, since fixed-effects and OLS approaches are biased. Introducing time dynamics into a spatial model leads to an additional layer of ML bias and more complicated moment conditions for GMM. The indirect inference estimator is implemented as an alternative, which employs a relatively simple estimator for complex models. The distance between estimates from the data and simulated spatial model data is minimized. We show that indirect inference methods can provide suitable bias correction when both spatial and time dynamics are present. The two-stage indirect inference is applied to spatial econometric models to construct finite-sample exact confidence sets. The resulting estimates and confidence set are able to side-step complicated likelihood functions and moment conditions.