Comparing Cardinal and Ordinal Approaches to Measuring Socioeconomic Inequalities in Health

In the last decade, a literature has evolved showing that standard measures of inequality or probit regression are not well-suited for ordinal data such as self-reported health status. Conclusions may be easily reversed via different scaling of the ordinal indicator. This problem also concerns the measurement of socioeconomic inequalities by the concentration index (CI). Using the Survey of Health, Aging and Retirement in Europe (SHARE) we compare the performance of CI against the non-parametric method introduced by Duclos and Echevin (DE) to evaluate joint health-income distributions. We find that Poland is the worst country according to all three methods of computing CI, but it is the best country according to DE dominance curves. This indicates that when applied to ordinal health indicators, the CI approach relies on quite selective information about health-income distribution and should be treated with caution.