This paper utilizes for the fi rst time age-structured human capital data for economic growth forecasting. We concentrate on pooled cross-country data of 65 countries over six 5-year periods (1970-2000) and consider specifi cations chosen by model selection criteria, Bayesian model averaging methodologies based on in-sample and out-of-sample goodness of fi t and on adaptive regression by mixing. The results indicate that forecast averaging and exploiting the demographic dimension of education data improve economic growth forecasts systematically. In particular, the results are very promising for improving economic growth predictions in developing countries. Copyright parameter uncertainty, and suggest exploiting the correlations between age structure and GDP growth in the framework of demography-based models for long-run predictions. Owing to their relative stability, demography-based forecasts of GDP have caught the attention of forecasters recently. In line with the research of Lindh and Malmberg (2007), Bloom et al. (2007), for instance, examine whether age structure improves forecasts of economic growth. The authors fi nd that including a simple variable summarizing the age structure improves income growth forecasts. While the size and differential dynamics of each age group for a country are commonly interpreted in this literature as a gross indicator of aggregate productivity effects, no study hitherto, to the knowledge of the authors, explicitly considers differential effects of human capital (in the form of education) across age groups.Indeed, the importance of human capital on economic growth has been highlighted systematically in the theoretical literature on the determinants of long-run income growth. However, the empirical evidence of the impact of human capital on economic growth has yielded ambiguous results (see, for instance, Benhabib and Spiegel, 1994;Pritchett, 2001;Krueger and Lindahl, 2001). Data quality has been deemed at least partly responsible for the lack of a signifi cant positive correlation between GDP per capita growth and human capital variables (see De la Fuente and Domenech, 2006;Cohen and Soto, 2007). Recently, a new database has been developed which for the fi rst time summarizes educational attainment fi gures in different age groups (IIASA-VID dataset; see Lutz et al., 2007Lutz et al., , 2008. 1 While the relative size of age groups can contain information which is useful for economic growth forecasts, age-structured human capital information, by disentangling 'quantity' and 'quality' effects, can lead to further improvements. 2 An important reason illustrating the possible performance differential is that while age structure demographic information can be argued to render level effects on GDP per capita, age-structured human capital may induce both level and growth effects (through its effect on technology adoption; see Spiegel, 1994, 2005) on the latter. Age-structured human capital introduces the role of fi rst, demographic change (age-structured population change and i...