The article discusses methods for estimating models with binary responses, like employed or unemployed. It compares parametric and semi- and nonparametric methods, explaining how the latter relax assumptions of the former. Semi- and nonparametric methods are more developed for binary responses than for multiple responses. If a parametric model is misspecified, semi- and nonparametric methods may be necessary. The article also covers the problem of identifying behavioral parameters without strict parametric assumptions, rates of convergence, and efficiency in semi-parametric models, and applications of semi- and nonparametric estimators for binary responses.