An Evaluation of the Postcode Address File as a Sampling Frame and its Use within OPCS

Abstract

The main objective of the Social Survey Division of the Office of Population Censuses and Surveys (OPCS) is to conduct or advise on the conduct of social surveys for Government Departments. OPCS is responsible for most of the major government household surveys and several specialised ad hoc surveys. Between 100,000 and 200,000 addresses could be approached in a year, the exact number depending on the work in the ad hoc program. This paper describes various evaluation studies which led OPCS to adopt a new sampling frame, the Postcode Address File (PAF), for most of its household surveys. In many OPCS Surveys, including the Family Expenditure Survey, the General Household Survey and the Labour Force Surveys, the population sampled is that of private households in Great Britain. So long as data are collected about all persons in the selected households, this enables the survey results to be analysed at both the household and individual level. Although a small proportion of the population of individuals live permanently at non-private addresses such as hospitals, hotels and prisons, this group is disproportionately costly and complicated to cover and is usually excluded from the sampling frame for general population surveys. In fact, there exists no comprehensive listing of private households as such, which can be used in survey sampling. Fortunately, however, various lists of addresses do exist and these can be used for sampling purposes by taking advantage of the fact that, in the great majority of cases, each address is inhabited by a single household. Traditionally the address sampling frame most often used by OPCS and other survey organisations has been the Register of Electors. This is actually a listing of persons qualified to vote and is often used as a frame for directly sampling individual adults. However, where samples of households are required OPCS have found it is preferable to identify these via addresses rather than via named individuals, a proportion of whom move or die during the life of a Register. Special procedures must be followed in order to obtain equal-probability samples of addresses from the Electoral Register and these are relatively costly and time consuming as the Registers are not yet available centrally in computerised form. Another disadvantage of the Registers is that they tend to under-represent some sub-groups such as ethnic minorities and recent movers and this is a serious drawback for many government policy studies. Another address sampling frame used from time to time, particularly on housing surveys, is the rating valuation list (valuation roll in Scotland), (see Breeze, 1981). Although good in