Estimating a non-neutral production function

In this article we estimate a production function that allows us to depart from the standard hypothesis of Hicks neutrality (HN) while also coping with the endogeneity of a dummy innovation variable. We consider specifications that relax HN, and we derive the testable conditions for common parametric approximations under which HN holds. The model is estimated using instrumental variable methods, allowing innovation to have a heterogeneous effect on the production process. The econometric analysis rejects HN and highlights three main features: a biased technical change (TC), with a higher ratio between the marginal product of labour and the marginal product of capital for innovative firms; a locally progressive TC; fully heterogeneous technologies when comparing innovative to non-innovative firms. We also address other issues, taking a closer look at the effect of process innovation and the role of sector heterogeneity and considering the potential endogeneity of labour input.