Individuals living in cooperative animal societies face numerous conflicts of interest with other group members. In many vertebrate societies within-group conflict culminates in the forced eviction of group members, leading to the permanent dispersal of individuals. Eviction, therefore, is an important determinant of population structure and demography, but little is known about what factors trigger such events. Three main explanations for patterns of eviction are (1) the reproductive competition hypothesis; (2) the coercion of cooperation hypothesis; and (3) the adaptive forced dispersal hypothesis. The last hypothesis proposes that dominant individuals use eviction as an adaptive strategy to propagate copies of their alleles through a highly structured population. Here we test these three hypotheses as explanations for observed eviction events in cooperatively breeding banded mongooses (Mungos mungo), using a 16 year database on life history, behaviour and genetic relatedness.  In this species groups of females, or mixed sex groups of males and females, are periodically evicted en masse. Our results suggest that reproductive competition is the main proximate trigger for eviction for both sexes. We find no evidence to suggest that eviction is used to coerce helping, or that eviction is an adaptive mechanism to force dispersal of related individuals into the wider population. Eviction of females changes the landscape of reproductive competition for remaining males, which can explain why males are often evicted (or inclined to leave) en masse alongside females, but then disperse in separate groups. Our results show that the means of resolving within-group conflict resonate through groups and populations to affect population structure, with important consequences for social evolution.