Catastrophic health payments in Malawi: analysis of determinants using a zero-inflated beta regression

The existing literature on out-of-pocket (OOP) health payments has tended to use binary models to analyse the determinants of catastrophic health spending. In so doing, the literature ignores the fact that shares of out-of-pocket health care payments which are used to define the binary dependent variables are fractional with a mass point at zero. Further to this, the literature makes no distinction between factors which influence the level and the risk of catastrophic health payments. In order to address these shortcomings, this paper departs from this approach, and uses the zero-inflated beta regression instead. The paper also derives elasticity formalae for the zero-inflated beta regression. These elasticities allow one to talk about both the statistical, and economic significance of the different determinants of health nonpayment, catastrophic health spending, and the risk of catastrophe. Data from Malawi's Third Integrated Household survey are used. The empirical results indicate that the same variable can have a different effect on the levels, and risk of catastrophic health spending as well as OOP health nonpayment.