Faulty poll methods blamed for wrong election forecasts

THE FAILURE of opinion polls to predict the outcome of the 1992 general election resulted largely from sampling inadequacies and a greater tendency among Tory voters to refuse to say how they would vote, according to an independent report published yesterday.

Although a late swing to the Tories - occurring after the final interviewing took place - was also a factor, it probably accounted for not more than a third of the total error, the report to the Market Research Society says.

The MRS working party - which included heads of leading polling organisations - in effect lays the blame on the polling operations themselves for a failure which the group's chairman Dr David Butler called the 'most spectacular in the history of British election surveys'.

The report admits that 'the gap between the polls' findings and the final result was greater than their consumers are entitled to expect'.

It says that Conservatives were less likely to reveal their true loyalties than Labour voters - either by refusing to answer voting intention questions or by saying 'don't know'. It also suggests that Conservatives had a greater tendency to refuse to participate in polls at all.

It acknowledges that there was a sampling bias 'partly because quotas and weights did not reflect sufficiently accurately the social profile of the electorate and partly because the variables used as the basis of the quotas and in corrective weighting were not closely enough correlated with voting behaviour . . .'

In some cases polling organisations weighted samples by housing tenure but used the proportion of households in council housing rather than the proportion of adult council tenants in the population as a whole. The result of this was to bias the samples more heavily to Labour than the surveys should have done. The polls also used data underestimating the number of people in two-car households - again resulting in a Labour bias.

The late swing was compounded by a unexpectedly greater tendency on the part of Tories to turn out.

Some pollsters are now experimenting with tentative recommendations in the report, including the use of 'secret balloting' techniques in which those surveyed write down their intentions without showing them to the pollster. Mori's accurate poll of European election voting in London used that technique. The report highlights two minor factors, the choice of sampling points and the interviewing of respondents who were not on the electoral register - also more likely to be Labour supporters.

But it rejects a number of explanations for the failure - including the idea that many voters deliberately lied; the behaviour of postal voters or overseas voters; the use of telephone polling; and sample size.