The median proportion of investment in females among 11 populations of seven bumble bee (Bombus) species was 0.32 (range 0.07 to 0.64). By contrast, two species of workerless social parasites in the related genus Psithyrus had female–biased sex allocation, the reasons for which remain unclear. Male–biased sex allocation in Bombus contradicts the predictions of Trivers and Hare's sex ratio model for the social Hymenoptera, which are that the population sex investment ratio should be 0.5 (1:1) under queen control and 0.75 (3:1 females:males) under worker control (assuming single, once–mated, outbred queens and non–reproductive workers). Male bias in Bombus does not appear to be either an artefact, or purely the result of symbiotic sex ratio distorters. According to modifications of the Trivers—Hare model, the level of worker male–production in Bombus is insufficient to account for observed levels of male bias. There is also no evidence that male bias arises from either local resource competition (related females compete for resources) or local mate enhancement (related males cooperate in securing mates). Bulmer presented models predicting sexual selection for protandry (males are produced before females) in annual social Hymenoptera and, as a consequence (given some parameter values), male–biased sex allocation. Bumble bees fit the assumptions of Bulmer's models and are protandrous. These models therefore represent the best current explanation for the bees' male–biased sex investment ratios. This conclusion suggests that the relative timing of the production of the sexes strongly influences sex allocation in the social Hymenoptera.