Thanks for your reply. The estimator we construct does not exist in standard software such as Eviews, Stata or SPSS. We have built the code ourselves in Matlab but are looking for a dataset which would allow the estimation of a probit selection model with a binary endogenous variable...
@Luca: thanks, but I am familiar with Lewbel and Dong's simple special regressor estimator: although it allows to account for a binary endogenous variable, it is not suited to deal with sample selection (ie we have a selection and an outcome equation, with a nonrandomly selected samples in the latter).