Population projections for small areas and ethnic groups - developing strategies for the estimation of demographic rates: A thesis submitted to the University of Manchester for the degree of Ph.D. in the Faculty of Humanities

In the UK, there is a growing need for reliable population projections at the sub-national level for efficient planning and resource allocation. Ideally, such projections are required with sex and single year of age (SYOA) detail and often according to ethnic group.
The thesis identifies and addresses some of the methodological and data barriers in the production of estimates and projections for small populations, including ethnic groups. More specifically, it considers strategies for producing demographic rate estimates for use in creating population statistics with a focus on the fertility component of demographic change. This PhD is funded by ESRC and is CASE sponsored by Bradford Metropolitan District Council (MDC), with Bradford data providing the test area for the research.
‘Natural increase’ population projections are carried out for Bradford wards, first using national Government Actuary Department (GAD) rates, and then using localised fertility and mortality information. This provides a basis for developing more sophisticated strategies for creating localised age-specific fertility rates (ASFRs) for the Bradford wards.
A problem of ‘ragged’ rates resulting from rates based on small numbers is identified when using the local data. The Hadwiger function is subsequently used to smooth the ASFRs for all Bradford wards. However, there are situations where ward level ethnic group populations record insufficient numbers of births to generate reliable ASFRs even after smoothing. Consequently, different strategies are considered for grouping sub-populations together to overcome this problem. These include grouping wards according to demographic and socio-economic criteria.
Ward groupings, based in part on cluster analysis, are then used to generate ethnic-specific ASFRs for use in a projection of births for ward level ethnic group sub-populations. The results are compared against actual recorded births, alongside those estimated by other strategies (including national GAD rates, Bradford district-level rates, rates for ethnic group at the Bradford district-level and ward-level rates). The measurement of error for the assessment is the mean absolute percentage error (MAPE). The overall findings indicate that using more detailed information produces smaller errors.