Endogenous inclusion in the Demographic and Health Survey anthropometric sample: Implications for studying height within households.

Development economists study both anthropometry and intra-household allocation. In these literatures, the Demographic and Household Surveys (DHS) are essential. The DHS censors its anthropometric sample by age: only children under five are measured. We document several econometric consequences, especially for estimating birth-order effects. Child birth order and mothers' fertility are highly correlated in the age-censored anthropometric subsample. Moreover, family structures and age patterns that permit within-family comparisons of siblings' anthropometry are unrepresentative. So strategies that could separate birth order and fertility in other data cannot here. We show that stratification by mother's fertility is important. We illustrate this by comparing India and sub-Saharan Africa (SSA). Children in India born to higher-fertility mothers are shorter, on average, than children of lower-fertility mothers. Yet, later-born children in India are taller, adjusted for age, than earlier-born children of the same sibsize. In SSA, neither of these associations is large.