Calculation of life tables from survey data: a technical note.

ABSTRACT

Life table calculations from survey data are frequently based on events for which exact dates are not available. When these dates are coded in monthly form (e.g., century months), estimates should take into account the fact that the first duration interval--the interval which captures events occurring in the first month of exposure--is half the length of all remaining intervals. Although failure to do so has a trivial effect on many demographic calculations, estimates which are based on events which occur with high frequency in the first few months of exposure can be substantially biased. Estimates of fecundability for four countries in the World Fertility Survey are used to illustrate this bias.