Hamilton's rule provides the foundation for understanding the genetic evolution of social behavior, showing that altruism is favored by increased relatedness and increased productivity of altruists. But how likely is it that a new altruistic mutation will satisfy Hamilton's rule by increasing the reproductive efficiency of the group? Altruism per se does not improve efficiency, and hence we would not expect a typical altruistic mutation to increase the mean productivity of the population. We examined the conditions under which a mutation causing reproductive altruism can spread when it does not increase productivity. We considered a population divided into temporary groups of genetically similar individuals (typically family groups). We show that the spread of altruism requires a pleiotropic link between altruism and enhanced productivity in diploid organisms, but not in haplodiploid organisms such as Hymenoptera. This result provides a novel biological understanding of the barrier to the spread of reproductive altruism in diploids. In haplodiploid organisms, altruism within families that lowers productivity may spread, provided daughters sacrifice their own reproduction to raise full-sisters. We verified our results using three single-locus genetic models that explore a range of the possible reproductive costs of helping. The advantage of female-to-female altruism in haplodiploids is a well-known prediction of Hamilton's rule, but its importance in relaxing the linkage between altruism and efficiency has not been explored. We discuss the possible role of such unproductive altruism in the origins of sociality. We also note that each model predicts a large region of parameter space were polymorphism between altruism and selfishness is maintained, a pattern independent of dominance.