ABSTRACT
The 1992 general election was a disaster for the opinion polls. Four polls published on polling day on average put Labour one point ahead of the Conservatives; in the event the Conservatives were seven and a half points ahead of Labour. Despite some important limitations in the evidence available to it, the Market Research Society’s inquiry into what went wrong identified three main causes – a last minute change of opinion, inaccurate and inadequate quota controls and a greater propensity among Conservative voters to refuse to say how they would vote. Other possible causes, such as lying, the impact of the poll tax on electoral registration, or the votes cast by voters living abroad were of little or no importance. Its conclusions have already led the polling companies to make some changes to their application of the quota sampling methodology, but whether these will be sufficient is far from certain. The reliance on quota sampling in Britain is driven by the media’s assumption that the best poll is the most recent one, but the experience of 1992 suggests that this and other media expectations of opinion polls require re-examination. Geoffrey Evans, Anthony Heath and Clive Payne
