Effect of sample bias on paleodemographic fertility estimates.

Paleodemographers must work to understand how representative any archaeologically recovered skeletal series is and the potential effects of series bias on their demographic reconstructions. We examine two forms of bias: 1) infant underenumeration caused by differential preservation or incomplete archaeological recovery and 2) the underenumeration of individuals over age 45 related to methodological bias. We generated 60 simulated skeletal series of 250 individuals each based on the Brass ([1971] Biological Aspects of Demography (London: Taylor and Francis), pp. 69-110) logit models. In the first test, age bias was introduced deterministically for all individuals with age at death over 40 years using the Lovejoy et al. ([1985] Am. J. Phys. Anthropol. 68:1-14) bias estimates. In the second test, 50% of all individuals under 5 years old were removed from each simulated distribution. The simulated series were analyzed using the model life table fitting procedure developed by the authors (Milner et al. [1989] Am. J. Phys. Anthropol. 80:49-58; Paine [1989] Am. J. Phys. Anthropol. 79:51-62). Forms of adult age estimation bias described by Lovejoy and coworkers inflate estimates by 10-20% of the true crude birth rate (CBR) (the number of births per year per 1,000 population). Overestimation of fertility and birth rates increases both absolutely and as a percentage of the true rate as population growth increases. This bias is very consistent. Because Lovejoy and colleagues have estimated the methodological bias itself, its effects can be estimated. Infant underenumeration is a more serious obstacle. It is not presently possible to estimate infant underenumeration reliably without prior knowledge of fertility rates. This reduces fertility reconstructions based on infant-biased samples to minimum fertility estimates.