Since most bird species are socially monogamous, variation among species in social mating systems is determined largely by variation in the frequency of mate desertion. Mate desertion is expected to occur when the benefits, in terms of additional reproductive opportunities, outweigh the costs, in terms of reduced reproductive success from the present brood. However, despite much research, the relative importance of costs and benefits in explaining mating system variation is not well understood. Here, we investigate this problem using a comparative method. We analyse changes in the frequency of mate desertion at different phylogenetic levels. Differences between orders and families in the frequency of desertion are negatively associated with changes in the potential costs of desertion, but are not associated with changes in the potential benefits of desertion. Conversely, differences among genera and species in the frequency of desertion are positively associated with increases in the potential benefits of desertion, but not with changes in the potential costs of desertion. Hence, we suggest that mate desertion in birds originates through a combination of evolutionary predisposition and ecological facilitation. In particular, ancient changes in life‐history strategy determine the costs of desertion and predispose certain lineages to polygamy, while contemporary changes in the distribution of resources determine the benefits of desertion and thereby the likelihood that polygamy will be viable within these lineages. Thus, monogamy can arise via two very different evolutionary pathways. Groups such as albatrosses (Procellariidae) are constrained to social monogamy by the high cost to desertion, irrespective of the potential benefits. However, in groups such as the accentors (Prunellidae), which are predisposed to desertion, monogamy occurs only when the benefits of desertion are very limited. These conclusions emphasise the additional power which a hierarchical approach contributes to the modern comparative method.