Estimating with Confidence
This page summarises the achievements of the project, and links
to an article giving further details and references to the main
publications from the project.
The Estimating with Confidence project successfully completed an analysis of the accuracy of small area population estimates in Britain as practised in 1991. The results are of direct relevance to a wide range of planning, marketing and research operations that depend on local population estimates.
Inaccuracies for small areas average around 2%-4% for the total population and 10-30% for 5-year age groups, ten years after the last census. The size of inaccuracy depends on the size of area, amount of migration and social characteristics as well as the method used to estimate the population.
The project found that a local census has clear advantages over other estimation methods, but that this advantage depends on achieving a high response rate.
Less expensive methods are appropriate according to the resources and data sources available, and the types of small area to be estimated. Cohort survival methods, for example, are appropriate for those over 40 years of age. Ratio methods improve the current level of a population indicator such as the electorate by using its ratio to a previous census population. They are more successful than apportionment methods which share out a large population purely on the basis of the current value of the population indicator. However, where official boundaries of small areas have changed since the last census, the data required for the ratio approach may be difficult to obtain.
The practitioner in the field must choose an estimation strategy tailored to the local circumstances and can do so with greater confidence using the outputs of this study. In all cases the quality of data sources is shown to be a major factor in the accuracy of population estimates, a factor that varies in its impact from one part of the country to another and therefore requires careful monitoring by producers of population estimates.
The analyses distinguished two sources of inaccuracy in population estimates: some areas are difficult to estimate whatever the method used, and some methods out-perform others when used on the same types of area. The analytical approaches to distinguish these two influences on inaccuracy involved multilevel statistical modelling of a complex dataset of estimates collected from 48 local authorities in Britain, and the project's own application of different estimation methods to a controlled set of areas.
1991 population estimates were evaluated, to take advantage of the national census of that year as a measure of the true population. The project staff contributed to the public and professional debates concerning the extent, distribution and sources of census undercount in 1991. The project's own undercount estimates for small areas followed extensive consultation with local authority staff and have already been used widely in the government, local authority and academic sectors. These undercount data have been made available on the through MIMAS at Manchester University and deposited with the ESRC archive.
The active collaboration with practitioners was a major and unique strength throughout the project's work. A steering group and series of workshops has helped to put together a unique dataset and library on local estimation methods, and encouraged the project to give practical and relevant advice concerning census undercount and accuracy of various estimation strategies. A subsequent project built on this network to create workshops and a guide to making local population estimates.
The main publications from the project are referenced in a summary article How can a national set of small area population estimates gain acceptance? Lessons from the Estimating with Confidence project.
Staff: Elizabeth Middleton, David Lunn, Rachel Cossey, and considerable support from Southampton Universityís Department of Social Statistics, and University of Manchesterís Centre for Census and Survey Research.