Limited dependent variables in willingness to pay studies: applications in health care

Abstract

The appropriate technique for econometric analysis of WTP (willingness to pay) data is an issue which has not been addressed in many studies of WTP for health and health care. This paper argues that, whether an open-ended question or a payment scale approach is adopted, the way in which WTP is recorded means that limited dependent variable models are more appropriate than standard regression analysis. Data from an open ended question on WTP for maternity care contain a large proportion of zeros and the evidence suggests that a two-part specification performs better than OLS or a standard Tobit model. If the payment scale method is adopted, our argument suggests that grouped data regression is an appropriate econometric technique. In practice, with data from a study in Northern Norway, the results from OLS and grouped data regression are broadly similar.