The present paper, starting from evidence of low growth-to-poverty elasticity characterizing Africa, purports to identify the distributional changes that limited the pro-poor impact of the last two decades' growth. Distributional changes that went undetected by standard inequality measures were not showing a clear pattern of inequality on the continent. By applying a new decomposition technique based on a non-parametric method-the 'relative distribution'-we found a clear distributional pattern affecting almost all analysed countries. Nineteen out twenty four countries experienced a significant increase in polarization, particularly in the lower tail of the distribution, and this distributional change lowered the pro-poor impact of growth substantially. Without this unfavorable redistribution, poverty could have decreased in these countries by an additional five percentage points.