Population projections by small areas and ethnic groups – developing strategies for demographic rate estimation

ESRC (CASE) Studentship with the Bradford Council
Supervisors: Mark Brown, Ludi Simpson (Bradford Council), Mark Tranmer
Researcher: Lee Williamson

Research aims and outcomes

Statistical methods will be used to extend robust estimates of demographic rates to small areas and ethnic groups. Different strategies of both methodological approach and data sources will be implemented. The impact of each will be assessed by implementing the estimates in projection software provided by Bradford Council.

The outcomes will be:

• Demographic rates and population projections estimated for wards of Bradford, ethnic groups in wards of Bradford, and ethnic groups in Great Britain.
• Robust and documented estimation strategies for sub-national demographic rates for ethnic groups, applicable to any area in Britain.

Context of project

The need for these estimates within population forecasts for all areas in the Untied Kingdom will give a practical context for the investigations. They are required for policy reviews of race equality and for setting objectives for public services and employment.

Local authorities have traditionally taken the lead in local demography in Britain, to satisfy their responsibilities for land use planning and service planning. Projections of each ethnic group population have been undertaken in Bradford, uniquely in the UK until recently. In 1999, Bradford Council made generally available their demographic software (‘POPGROUP’) which provides a forecast population for single years of age. It has been used for estimation and projection of ward populations and ethnic groups within a local authority District, but not in combination.

Bradford Council’s interest in funding this studentship is the extension of its own work towards robust population estimates and projections for wards and ethnic groups. Of equal importance to the Council is to provide a pool of relevant research for use by other local demographers.

The combination of population statistics and race has provoked a wide range of political responses, some of which are xenophobic and dangerous to the marginalised communities represented by some ethnic group categories. This context will be kept in mind to ensure that the work contributes positively to improved understanding.

Methodology

The most promising method of estimating and projecting the population of small groups – whether ethnic groups or small areas or the two in combination – is the standard approach of ageing forward cohorts with separate estimation of components of population change. These components – rates of births, deaths, and migration – are of varying difficulty to estimate due partly to lack of data and partly to the volatility of rates based on small population numbers.

For projections, estimation of future change in demographic rates requires a study of time series of previous rates. Again, for ethnic groups and small areas there are limited data. For ethnic groups, the experience of those born in Britain is more relevant to the future than the experience of those whose migration has strongly influenced their fertility and mortality and migration. However, the number of second-generation migrants is relatively small. For ethnic groups there are also serious issues of identity changing over time, cohorts and age, and the representation of children of mixed ethnic group parentage. These issues may be dealt with by using age-specific propensities to change ethnic group (including from mother to child) but such propensities are yet to be estimated for any area in Britain, which will give this research another original element.

The problem can then be summed up as one of robust estimation for small sub-populations from a variety of data sources. The combination of data from different sources, and of local and national data from the same source such as the census, involves calibration of datasets against each other, maintaining the most relevant and reliable patterns observable in each. To extend existing methods of:

• structure-preserving estimation,
• raking or Iterative Proportional Fitting,
• multi-level modelling and
• various graduation techniques.

Data and measures of success

The primary sources of data will be output from the 1991 Census, the Longitudinal Study and the Labour Force Survey. These, together with data local to Bradford, will provide the material for a variety of estimation techniques for the demographic counts and rates required by population projections.

These estimates, when used in a projection forward from a population estimated in detail for 1991, will give forecasts for 2001 and 2011 that provide a sensitivity test for the use of estimates from different strategies.

Should the results from the 2001 Census be available on time (they are currently due ‘early in 2003’), they will be used both to assess the accuracy of different estimation strategies, and to provide new population projections based on the 2001 population.