# 17-031/V (2017-03-14)

Laura Rossouw, Stellenbosch University, South Africa; Teresa Bago d'Uva, Erasmus University Rotterdam, The Netherlands; Eddy van Doorslaer, Erasmus University Rotterdam, The Netherlands, and Stellenbosch University, South Africa
self-assessed health, vignettes, health measurement, inequality, South Africa
JEL codes:
D30; D31; I10; I14

In spite of the well-known wide disparities in wealth and in objective measures of health like mortality in countries like South Africa, health inequality by wealth in self-reported health measures appears to be nearly non-existent. We test and correct for reporting heterogeneity in sixteen domains of self-assessed health by wealth and race among elderly South Africans using anchoring vignettes. We find that significant reporting differences between high and low wealth groups lead to severe underestimation of the health-wealth gap: poorer individuals rate the same health relatively higher than richer. Using hierarchical ordered probit (HOPIT) modeling, we show that a significant and substantial health disadvantage of the poor emerges after correction. We also address the question whether and how health inequality and reporting heterogeneity are confounded by race. We find that within race groups - especially among Blacks but also among Whites - reporting heterogeneity leads to the underestimation of the health inequalities between richest and poorest. Finally, we show that the apparent Black (vs White) health disadvantage within the top wealth quintile disappears once we correct for reporting tendencies. All in all, our findings suggest that reporting tendencies are an important source of bias in the measurement of health disparities and that anchoring vignettes and HOPIT models can play a role in correcting for these biases.