On age-specific variations in income-related inequalities in diabetes, hypertension and obesity

Abstract

Objective

This article examines whether the social health gradients in diabetes, hypertension and obesity for men and women vary significantly across different age groups.

Methods

We use a pooled sample of German survey data from the years 2002 and 2006 with a total of 87,601 observations. We employ a varying Wagstaff index derived from the class of Gini-type concentration indices to estimate age-specific income-related health inequalities.

Results

We find significant health disadvantages among poor women in mid-age, but no significant age-specific income-related health inequalities among men. Some leveling of inequalities in diabetes is observed.

Conclusions

The results suggest that variations in age-specific inequalities are unlikely to be a purely artificial result of health-related selection into retirement or mortality.

Appendix: Statistical inference

The local standard error σ _C (z) of the varying concentration index can be approximated by estimating β ^*₀ and β ^*₁ from Eq. (1) with the untransformed health variable y in place of $\left(2\sigma_{r}^{2}(z)/\mu_{y}(z)\right)y.$ One may then take advantage of the fact that β ^*₀ (z) + μ _r (z)β ^*₁ (z) = μ _y (z) and apply the δ method to $C(z)=2\sigma_{r}^{2}(z)\beta^{*}_{1}(z) \left[\beta^{*}_{0}(z)+\mu_{r}(z)\beta^{*}_{1}(z)\right]^{-1}$ (Kakwani et al. 1997; Siegel and Mosler 2010; Wildman 2003). The standard error σ _W (z) for W(z) can be estimated analogously (Siegel and Mosler 2010). Note that μ _r (z) and σ ² _r (z) need not be considered as stochastic as they are sample independent. The error term may likely be heteroscedastic and autocorrelated (Kakwani et al. 1997; Wildman 2003; McKinnon et al. 2011); we therefore follow Siegel and Mosler (2010) and estimate local Newey-West type standard errors.