We study the roots of violence against women, and propose that it partly originates in cultural norms that derive from (a) characteristics of the traditional subsistence problem in different societies, and (b) differences in the sexual division of labor for solving that problem in each society. We construct this hypothesis on economics and anthropology research showing the potential of traditional livelihoods to shape persistent cultural norms at the local level, and arguing that this concept can be extended to explain outcomes at the domestic level. We test our main hypothesis by examining differences in the incidence of domestic violence across areas with different historical livelihoods in modern-day Tanzania, where we observe a large degree of spatial variation in both attitudes and actions of violence against women. Using rich individual survey and high-resolution georeferenced data, we find systematically less violence against women in traditionally sea-fishing areas vis-à-vis traditionally lake-fishing, agricultural, and pastoralist ones. Our results are consistent with anthropological accounts of the idea that women in sea-fishing societies tend to be comparatively more independent in decision-making, and to acquire skills that are complementary to demands in non-agrarian sectors. We interpret this as evidence for direct mechanisms helping to sustain egalitarian gender norms in general, and less violence against women in particular. By exploiting sub-national variation, this research allows us to move beyond studying the socio-economic and institutional determinants of violence against women, and to analyse the formation of specific cultural traits that explain where and why some women tolerate less violence against them.