We show how differences in aggregate human development outcomes over time and space can be additively decomposed into a pure mean income (growth) component, a component attributed to differences in the distribution of income, and components attributed to 'non-income' factors and differences in the model linking outcomes to income and non-income characteristics. The income effect at the micro level is modelled non-parametrically, so as to flexibly reflect potentially complex distributional changes. Our proposed method is illustrated using data for Morocco and Vietnam, and the results offer some surprising insights into the observed aggregate gains in schooling attainments.