Bivariate alternatives to the Tobit model

Abstract

This paper discusses some generalisations of the Tobit model that allow for distinct processes determining the censoring rule and the continuous observations. The effect of different behavioural assumptions on the econometric model are examined and the alternatives are constructed. The paper concentrates on diagnostic tests for misspecification that are appropriate in each of our examples. These include tests of independence and normality. The empirical illustrations relate to household demand for clothing and to married women's labour supply. The data are drawn from the Family Expenditure Survey for 1981.