Pre-print, Working paper: This paper proposes a new method to infer gender preferences from birth spacing. We apply it to Africa, where the least is known about gender preferences. We show that son preference is strong and increasing in North Africa. By contrast, most Sub-Saharan African countries display a preference for variety or no preference at all. Further analysis concludes that traditional family systems predict well the nature of gender preferences, while religion does not. Last, the magnitude of preferences is stronger for wealthier and more educated women.