Estimation and Inference in Semiparametric Quantile Factor Models

We consider a semiparametric quantile factor panel model that allows observed stock-specific characteristics to affect stock returns in a nonlinear time-varying way, extending Connor, Hagmann, and Linton (2012) to the quantile restriction case. We propose a sieve-based estimation methodology that is easy to implement. We provide tools for inference that are robust to the existence of moments and to the form of weak cross-sectional dependence in the idiosyncratic error term. We apply our method to daily stock return data where we find significant evidence of nonlinearity in many of the characteristic exposure curves.