A1526
Title: Regional innovation policy evaluation with endogenous regressors and fixed effects
Authors: Alexandra Soberon - Universidad de Cantabria (Spain) [presenting]
Antonio Musolesi - University of Ferrara (Italy)
Juan Manuel Rodriguez-Poo - Universidad de Cantabria (Spain)
Sylvie Charlot - Universite Lumiere Lyon 2 (France)
Abstract: This paper proposes a nonparametric estimator for the gradient of an unknown structural function in panel data models with endogenous regressors and two-way unobserved effects. Individual fixed effects and latent common factors are jointly eliminated through a cross-sectionally demeaned first-differencing transformation. Endogeneity of the regressors is addressed via a control function approach within a triangular simultaneous equations system, where reduced-form residuals enter the differenced regression additively. The gradient is recovered by locally weighted linear regression using product kernel weights that jointly localise pairs of consecutive observations around each evaluation point. Asymptotic properties of the estimator are established and finite-sample performance is assessed through Monte Carlo experiments under structural functions of varying nonlinearity and endogeneity strength. The proposed methodology is applied to the evaluation of regional innovation policies in Europe, allowing for a flexible assessment of the marginal effects of policy interventions on knowledge production while accounting for simultaneity, endogenous input allocation, and unobserved heterogeneity across regions.
