While economic growth in Africa has resurged substantially since the mid-to-late 1990s, the amount of poverty reduction seems much less spectacular. Building on other studies, the paper explores the translation of the recent growth to poverty reduction using 1985-2013 PovcalNet (World Bank) data. It assesses the relative abilities of various panel-data methodologies to predict poverty changes based on income-inequality decompositions. Surprisingly, SYSGMM performs substantially worse than Fixed Effects and Random Effects. The analysis is conducted for both the $1.25 and $2.00 poverty lines, and for the 'spread' and 'depth' of poverty, as well as for the usual popular measure, the headcount ratio. Although income growth appears to be the main force behind poverty reduction in Africa, the decomposition reveals striking differences, across countries and poverty measures, with respect to the relative roles of inequality and income.