Capital humain, productivité agricole, et travail féminin : variables latentes et séparabilité dans les modèles de ménage

It is frequent, especially in the African context, to find a negative and statistically significant coefficient associated with human capital when it is introduced into the estimation of a plot-level production function. We provide an explanation for this paradox in the context of a non-separable household model. A new approach to testing for separability, based on the estimation of a switching regression, is implemented. Our results suggest that the impact of human capital on the demand for female labor is positive in the context of the separable regime whereas, in the case of the non-separable regime, human capital increases output only when the household is subject to a binding credit constraint.