Abstract
The study of qualitative-response models has received considerable attention in the decision sciences, as well as the econometric literature. Applications of these models have included studies of labour force participation, occupational choice, transportation choice, purchase of consumer durables, analysis of voting behaviour, and many others. As recently as a decade ago, the area of qualitative-response models appeared to be a mature field, with no major breakthroughs appearing on the horizon. In recent years, however, semiparametric and non-parametric methods developed by statisticians in the 1960s and 1970s began to be applied to qualitative-response models. In 1993 a special issue of the Journal of Econometrics included a discussion and comparison of the performance of several alternative parametric and semiparametric specifications applied to modelling binary decisions. This paper extends the comparisons previously investigated to include models with flexible parametric forms having fixed and random coefficient specifications. The random coefficient formulations permit the analysis of a form of heterogeneity which has been found useful in a number of applications. These models are applied to three different data sets.