Cohorts, overlapping generations and consanguinity estimates

A number of problems in human population genetics involve determining the effects of non-random mating due to cultural or geographic factors, and therefore requires determination of expectations based on random mating. These expectations are usually taken to be the arithmetic mean of the contributions of all possible pairs drawn from a subpopulation of individuals of mating age during a specified period. This cohort approach in effect treats the subpopulation as a discrete generation. In real populations, however, the pool of potential mates with whom any individual may be paired is continually changing, and not all pairs are eligible, on demographic grounds, to mate for equal lengths of time. A simple means of avoiding problems due to this discrepancy between method and reality is suggested and applied to data from an isolated Caribbean population in order to determine whether treating cohorts as discrete generations introduces serious bias into estimates of expected consanguinity. The results suggest that any such biases are likely to be negligible, but may be important in some special circumstances.