In many insect societies, workers can manipulate the reproductive output of their colony by killing kin of lesser value to them. For instance, workers of the mound–building Formica exsecta eliminate male brood in colonies headed by a single–mated queen. By combining an inclusive fitness model and empirical data, we investigated the selective causes underlying these fratricides. Our model examines until which threshold stage in male brood development do the workers benefit from eliminating males to rear extra females instead. We then determined the minimal developmental stage reached by male larvae before elimination in F. exsecta field colonies. Surprisingly, many male larvae were kept until they were close to pupation, and only then eliminated. According to our model, part of the eliminated males were so large that workers would not benefit from replacing them with new females. Moreover, males were eliminated late in the season, so that new females could no longer be initiated, because matings take place synchronously during a short period. Together, these results indicate that workers did not replace male brood with new females, but rather reduced total brood size during late larval development. Male destruction was probably triggered by resource limitation, and the timing of brood elimination suggests that males may have been fed to females when these start to grow exponentially during the final larval stage. Hence, the evolution of fratricides in ants is best explained by a combination of ecological, demographic and genetic parameters.