Academic journal article Demographic Research

Sources of Error and Bias in Methods of Fertility Estimation Contingent on the P/F Ratio in a Time of Declining Fertility and Rising Mortality

Academic journal article Demographic Research

Sources of Error and Bias in Methods of Fertility Estimation Contingent on the P/F Ratio in a Time of Declining Fertility and Rising Mortality

Article excerpt


The most commonly used indirect fertility estimation methods rely on the use of the P/F ratio, first proposed by Brass. In essence, the ratio compares cumulated cohort fertility with cumulated period fertility on the basis of three, fairly strong, assumptions. First, that the level of fertility has remained constant over time. Second, that the age distribution of fertility has been constant; and third, that the fertility of women who do not survive to report their numbers of children borne does not differ from those who do survive. This paper interrogates what happens to the results produced by the P/F ratio method as each of these three assumptions is violated, first independently, and then concurrently. These investigations are important given the generally poor quality of census data collected in many developing countries, particularly those in sub-Saharan Africa, and the particular demographic dynamics resulting from the generalised HIV/AIDS epidemic in the region. The investigations suggest that using the P/F ratio for the age group 20-24 to scale the reported fertility schedule is more accurate than the Feeney method and marginally preferable to scaling using the average ratio for the age groups between 20 and 29, although it would overstate fertility while fertility is rising and for some time after period fertility peaks, reaching a maximum of around 10 per cent at the peak of period fertility. In addition differential fertility between HIV-infected and HIV-uninfected women has a trivial impact on the methods, even in an environment with a simulated highly generalised epidemic.

(ProQuest: ... denotes formulae omitted.)

1. Introduction

Improvements in vital registration systems and census methodologies, together with the expansion of coverage of the Demographic and Health Surveys have all but obviated the need for indirect techniques of fertility and mortality estimation in most regions of the world. However, these techniques are still necessary throughout Africa, and since the majority were developed in the 1960s and 1970s on the assumptions that fertility and mortality were constant (which was, roughly, the case in countries to which these techniques were applied at the time the methods were developed), there is a need to question how well they perform in the demographic environment that prevails now in the region.

While some of the indirect techniques were subjected to further refinement in the 1980s (for example, with the development of the Relational Gompertz fertility model; and the variable-r mortality estimation methods), the improvements in data (number of sources, methodologies, and surveys), combined with the shift in research priorities after the 1994 Cairo Conference on Population and Development, have meant that little attention has been paid to the continued validity of the methods under radically altering demographic conditions. In this paper, we address one aspect of this lacuna in our knowledge, at least with regard to the methods of estimating fertility from census-type data.

The relevance of this study may seem obscure to demographers accustomed to working only with developed country data (where vital registration systems and population registers have rendered most forms of demographic survey, not to mention the census, obsolete), or even to demographers whose work with data from developing countries is confined to that available from Demographic and Health Surveys (DHSs). Although the data from DHSs are of high quality (there are exceptions, the 2003 South Africa DHS, for example), their usefulness can be constrained by their comparatively small sample size which makes multivariate analysis of fertility by even a few variables (for example, by age, region and education level simultaneously) unreliable. For such analyses, not to mention the provision of reliable estimates at a sub-national level, and in the absence of a substantially complete vital registration system, the census is an important source of data for demographic estimation and policy formulation. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed


An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.