Estimating an Earnings Function from Coarsened Data by an Interval Censored Regression Procedure

This paper estimates an earnings function where the dependent variable is a mix of point and interval data using an interval regression model based on a pseudo-maximum likelihood estimation procedure. The analysis uses the 1999 OHS, and takes into account point and interval income observations, as well as design features of the survey including stratification, clustering and weights. In developing and applying the methodology, it is shown that researchers interested in analysing the determinants of income in a meaningful way need not be hampered by the presence of both point and interval observations, and can in fact account for these simultaneously using a generalised Tobit model. By incorporating survey design features into the analysis of the variance, some changes were needed to the estimation procedure and this is where the pseudo-likelihood becomes useful. However, this then affects how the coefficients of the model are interpreted, and researchers are encouraged to focus attention on the confluence of these factors.