Use of sibling data to estimate family mortality effects in Guatemala.

This paper examines the potential bias in estimates of child mortality determinants produced by the questionable assumption that sibling data are independent, and estimates the unmeasured familial effects shared among siblings. The parameter estimates yielded by the multivariate hazard model are very similar to those yielded by the standard hazard model. The standard errors of the parameter estimates, however, tend to be underestimated in conventional analyses. The contribution to child mortality from the familial factors seems modest net of household socioeconomic status, at least in this Guatemalan data set.