Semiparametric estimation of dynamic discrete choice models

We consider the estimation of dynamic binary choice models in a semiparametric setting, which per-period utility functions are known up to finite number parameters, but distribution shocks is left unspecified. This setup differs from most existing identification and literature for discrete models. To show we derive exploit new recursive representation unknown quantile function shocks. Our estimators straightforward compute, resemble classic closed-form on regression average derivative estimation. Monte Carlo simulations demonstrate that our estimator performs well small samples.