Although birth order is recognized as an important social variable, the incidence of various birth orders is often difficult to assess using vital statistics. A technique is described whereby the incidence of birth orders in a population may be estimated directly from census data on the number of children ever born to post-menopausal women. The technique has the advantages of allowing ready utilization of existing and widely available data, based on very large samples on which considerable supplementary information is available. In addition, it draws attention to a relatively neglected aspect of differential fertility, namely, the child’s perspective as opposed to the mother’s view. Limitations of the technique relating to quality of data, to multiple births, and to differentials in infant and maternal mortality are discussed, and possible applications in demography and sociology are suggested.