ABSTRACT
Two principal methodological problems were identified after the polls failed to anticipate the Conservatives' victory in the 1992 general election. First, Conservative voters were less willing to say how they would vote. Second, the samples tended to be too downmarket. Since 1992, quota polls have persistently found that Labour 'won' the 1992 election when they ask their respondents to recall their 1992 vote. This discrepancy cannot be adequately accounted for by the tendency for voters to align their past behaviour with their current preference, a tendency which has in any case diminished. Rather, Conservative voters clearly continued to be less willing to participate in polls while quota controls still failed to ensure that those approached for interview were fully representative. Both problems may be addressed by 'adjustment' techniques, but these may not be adequate and are not always given sufficient prominence.
