Bayes plus Brass: estimating total fertility for many small areas from sparse census data

Area of Science:

• Demography
• Biostatistics
• Spatial Analysis

Background:

• Accurate small-area fertility estimates are vital for demographic analysis and local planning.
• Countries with incomplete vital registration systems face challenges in obtaining reliable small-area fertility data from sparse surveys or census data.
• Existing estimation methods often struggle with automation, extreme sampling variability, and data errors.

Purpose of the Study:

• To present a novel two-step procedure for estimating total fertility in small areas, particularly in data-scarce environments.
• To address the challenges of automation, sampling variability, and data errors in small-area fertility estimation.
• To provide a robust method applicable to countries undergoing rapid fertility transitions.

Main Methods:

• A two-step estimation procedure was developed.
• The first step involves smoothing local age-specific fertility rates using Empirical Bayes methods.
• The second step applies a modified Brass's P/F parity correction procedure, designed to be robust to rapid fertility decline.

Main Results:

• The method was successfully applied to estimate total fertility for over 5,000 Brazilian municipalities using 2000 Census data.
• The Empirical Bayes smoothing effectively handled local variations in age-specific rates.
• The adapted Brass's P/F procedure provided reliable estimates even amidst fertility decline.

Conclusions:

• The proposed two-step procedure offers a robust and automated solution for small-area fertility estimation in challenging data environments.
• This method enhances the reliability of demographic analysis and planning at local levels.
• The accompanying supplementary materials allow for replication and application to other datasets.