This paper characterizes the bandwidth value (h) that is optimal for estimating parameters of the form η = E ω/f V |U ( V | U) , where f V |U (the conditional density of a scalar continuous random variable V , given a random vector U) is replaced by its kernel estimator. The results in this paper are directly applicable to semiparametric estimators proposed in Lewbel (1998), Lewbel (2000b), Honoré and Lewbel (2002), Khan and Lewbel (2007), and Lewbel (2006). The optimal bandwidth is derived by minimizing the leading terms of a second-order mean squared error expansion of the resulting estimator with respect to h. The expansion also demonstrates that the bandwidth can be chosen on the basis of bias alone, and that a simple 'plug-in' estimator for the optimal bandwidth can be constructed. Finally, the small sample performance of our proposed estimator of the optimal bandwidth is assessed by a Monte Carlo experiment. We then use our methodology in an empirical application of the female labour force participation in Ecuador.