The empirical content of spatial spillovers: Identification issues
Résumé
The notion of spatial spillovers has been widely used in applied spatial
econometrics. In this paper, we consider their empirical content under the
lens of identification in both structural and causal reduced-form models.
Discussing the various threats to identification in structural models, we
point out that the typical estimation framework proposed in the applied
spatial econometric literature boils down to considering spillovers as a
side-effect of a data-driven chosen specification, which strongly hampers
their empirical content. We also discuss the limits of using interaction
matrices purely based on geography to identify the source and content of
spillovers. Then, we present reduced forms impact evaluation models for
spatial data and show that the current spatial versions of usual impact
evaluation models are not fully satisfactory when considering the identification
issue. Finally, a set of recommendations for applied papers aimed
at identifying spatial spillovers is proposed.
Origine | Fichiers produits par l'(les) auteur(s) |
---|