We adopt an empirical approach to analyse, measure and decompose inequality of opportunity (IOp) in health, based on a latent class model. This addresses some of the limitations that affect earlier work in this literature concerning the definition of types, such as partial observability, the ad hoc selection of circumstances, the curse of dimensionality and unobserved type-specific heterogeneity that may lead to biased estimates of IOp. We apply our latent class approach to measure IOp in allostatic load, a composite measure of biomarker data. Using data from Understanding Society: The UK Household Longitudinal Study (UKHLS), we find that a latent class model with three latent types best fits the data, with the corresponding types characterised in terms of differences in their observed circumstances. Decomposition analysis shows that about two thirds of the total inequalities in allostatic load can be attributed to the direct and indirect contribution of circumstances and that the direct contribution of effort is small. Further analysis conditional on age-sex groups reveals that the relative (percentage) contribution of circumstances to the total inequalities remains mostly unaffected and the direct contribution of effort remains small.
Keywords: biomarkers; decomposition analysis; equality of opportunity; finite mixture models; health equity; latent class models.
© 2020 John Wiley & Sons, Ltd.