A0195
Title: Modeling heterogeneity by structural varying coefficients models in presence of endogeneity
Authors: Stefan Sperlich - University of Geneva (Switzerland) [presenting]
Abstract: High degrees of heterogeneity across the studied units can make the estimates of structural parameters inaccurate and uninformative. First, we reviews how semiparametric varying coefficient models can identify cross-sectional heterogeneity. Next, we discuss how different potential sources of endogeneity problems can be addressed. It is explained thereby why modelling coefficient variation across groups increases the credibility and interpretability of the assumptions needed when using instrumental variable estimation. They are therefore not just a nice compromise between non- and fully parametric models to circumvent the curse of dimensionality, they are also a good compromise between these two extremes regarding causal inference. This will be illustrated along various models taken from applied econometrics. In order to make the econometric specification theory-consistent, all considerations and developments are embedded in the modeling of heterogeneous structural equations rather than in purely theoretical remedies.
