HOW ACCURATE ARE TRADITIONAL QUOTA OPINION POLLS

Political opinion polls in Britain have traditionally been conducted u sing a face-to-face quota sampling methodology. Each interviewer is gi ven an area in which to work (usually either a parliamentary ward or c onstituency) and then is required to interview within that area a give n number of persons with a particular set of social characteristics. T he characteristics most commonly used in setting these quotas are sex, age, social class, work status and housing tenure. Although the sampl e may then be further weighted after interviewing by additional charac teristics such as car ownership, the key assumption behind quota sampl ing is that the quota controls should be sufficient to ensure that a s urvey contacts a politically representative sample of the population. The performance of the polls in the 1992 general election revived deba te about the adequacy of quota sampling (see particularly Jowell et al 1993; Worcester 1996). Four polls published on polling day, all of wh ich used quota sampling, on average put the Labour Party one point ahe ad; but when the ballot boxes were opened just hours later the Conserv atives proved to be as much as eight points ahead. Although The Market Research Society's own inquiry into the election polls found that par t of the explanation lay in 'late swing', it also found that 'some ina dequacies were found in the operation of the quota system' (Market Res earch Society 1994). This paper revisits this debate, not by looking o nce again at what happened in 1992, but by examining the performance o f the polls since 1992. In particular, by comparing the results of quo ta samples conducted during the course of the 1992-7 parliament with d ata collected from a random panel survey, viz. the British Election Pa nel Survey, we find evidence that quota sampling is still producing sa mples which overestimate Labour's strength. Moreover this conclusion a ppears to be confirmed by the performance of the polls in the 1997 ele ction. True, the adjustment mechanisms introduced by some companies si nce 1992 appear to help to correct the problem, but these may in some cases still be inadequate, and they are certainly often given insuffic ient prominence by their media clients.