In raw data in the UK, the income loss on separation for women who were cohabiting is less than the loss for those who were married. Cohabitees lose less even after matching on observable characteristics including age and children. This difference is not explained by differences in access to benefits or labour supply responses after separation. We show that the difference arises because of differences in access to family support networks: cohabitees' household income falls by less because they are more likely to live with other adults, particularly their family, following separation, even after matching on age and children. Divorced women do not return to living with their extended families. The greater legal protection offered by marriage does not appear to translate into economic protection.