|
Abstract:
|
Recent literature has highlighted the usefulness of the functional regression model for estimation of models with endogenous spatial dependence and spatial heterogeneity. The key idea is that, since the functional regression estimator is essentially a method-of-moments estimator, the assumption of independent and identically distributed errors is not required here. We extend this approach to the context where the nature of spatial dependence is estimated from the data, or is otherwise endogenously determined with the dependent variable. Using a distance-based, or some other purely exogenous, spatial weights matrix as the instrument, we develop methods for functional regression with endogenous functional regressors and correlated errors in a large data setting. The proposed methods are illustrated using an application to hedonic house price models.
|
