Haplodiploidy (encompassing both arrhenotoky and paternal genome elimination) could have originated from coevolution between male-killing endosymbiotic bacteria and their hosts. In insects, haplodiploidy tends to arise in lineages that rely on maternally transmitted bacteria for nutrition and that have gregarious broods in which competition between siblings may occur. When siblings compete, there is strong selection on maternally transmitted elements to kill males. I consider a hypothetical bacterial phenotype that renders male zygotes effectively haploid by preventing chromosome decondensation in male-determining sperm nuclei. By causing high male mortality, such a phenotype can be advantageous to the bacterial lineage. By eliminating paternal genes, it can also be advantageous to the host female. A simple model shows that the host female will benefit under a wide range of values for the efficiency of resource re-allocation, the efficiency of transmission, and the viability of haploid males. This hypothesis helps to explain the ecological correlates of the origins of haplodiploidy, as well as such otherwise puzzling phenomena as obligate cannibalism by male Micromalthus beetles, reversion to diploidy by aposymbiotic male stictococcid scale insects, and the bizarre genomic constitution of scale insect bacteriomes.