The objective of this analysis is to determine whether useful information on birthweight can be obtained from a retrospective survey. The 1986 Peru Demographic Health Survey collected both numerical weights (in grams) and subjective assessments of relative size at birth, for infants born during 1981–1986.

Simple tabulations suggest that reports of both measures are of reasonably high quality. However, a more detailed analysis demonstrates that exclusion of the large proportion (almost one-third) of missing responses on numerical weights can lead to biases in the resulting estimates of the incidence and correlates of low birthweight, of the level of infant mortality, and probably of the relative risk of mortality for low birthweight infants.

The results emphasize the necessity of collecting data on the relative size of infants, in addition to information on numerical weights, in populations in which a substantial fraction of infants are born outside the formal health system. The subjective assessments allow the analyst to examine the extent to which infants with missing numerical weights are select and to assess the degree of bias in estimates which are based on the subsample of births with reported weights.