A Gini-type concentration index is combined with semiparametric estimation techniques to derive a varying inequality index that works without a priori sample stratification. The new approach is used to investigate the question how income inequalities and income-related gradients in the distribution of health vary across age groups. With health data from the 2005 survey of the German microcensus, it is demonstrated that significant inequalities to the detriment of the deprived evolve in early mid-life and reach their maximum around the age for retirement. Some leveling is found for the elderly. Copyright © 2013 John Wiley & Sons, Ltd.