Parity cohort PPRs (cont.)
Calculation from birth histories
The data required to calculate parity cohort PPRs can be obtained from the same sources that provide data for the calculation of age cohort PPRs. In principle, PPRs could be calculated from routine vital statistics data. This is only possible, however, if information is collected on birth certificates about the date of the mother’s previous birth as well as about how many children she has borne previously. Few countries collect this information routinely.
Secondly, one can collect birth histories prospectively in a longitudinal study. Unfortunately, few such studies exist that have followed up women for the long periods of time needed to obtain complete information on their family building histories. Moreover, longitudinal studies such as this can suffer from severe attrition bias if they lose touch with their respondents or the respondents decide that they do not wish to participate in the study any longer. Thus, more commonly, parity cohort PPRs are calculated from birth history data collected retrospectively in a single-round census or survey such as the Demographic and Health Surveys.
The PPRs can be calculated using standard life table methods. While childbearing is a repeatable event, exiting from a specific birth interval is an unrepeatable – and irreversible – single-decrement process. (Of course, women also exit birth intervals by dying or emigrating from the population of interest but birth histories that were collected retrospectively in a survey cannot provide any information on women who have left the population. In most – though not all – populations, however, mortality and emigration of women aged 15 to 49 is sufficiently low that ignoring them has little impact on the representativeness of the estimates).
In life table notation, the PPR at exact duration t, an(t), is the complement of life table survivorship at duration t, lt. Thus, using a superscript to indicate the parity, n, for which the life table is calculated:
In this context, interval durations are usually measured in months rather than years. Thus, for example, l60n would represent the proportion of women of parity n who have not progressed to their n + 1th birth after an interval of exactly five years.