Dennis Kristensen, University College London

"Nonparametric Identification and Estimation of Multivariate Transformation Models"

Abstract

This paper develops novel results for nonparametric identification of two classes of multivariate transformation models. The identification argument imposes weak restrictions on the models and only requires that (i) a sufficient number of exogenous covariates are available, (ii) the log-density of the errors satisfies a rank condition and, if endogenous covariates enter the model, (iii) either instruments or control functions are available. The proof is constructive in the sense that it leads to natural estimators of the identified components. We apply the general theory to show identification in a class of multivariate demand models a la Berry-Levinsohn-Pakes under weaker conditions compared to existing ones found in the literature.

Contact person: Jesper Riis-Vestergaard Sørensen