This paper considers two-step estimation of a sample selection model i
n which there is heteroskedasticity of unknown form in the latent erro
rs. We propose an estimator which uses recent developments in nonparam
etric regression estimation involving series approximations. The estim
ator is shown to be consistent and asymptotically normally distributed
under reasonable conditions. A small Monte Carlo experiment demonstra
tes the usefulness of the estimator and highlights the bias inherent i
n the usual Heckman (1979) estimator when there is heteroskedasticity.