Why Opinion Polls Have Got It Right This Time
N the face of it, Tony Blair could hardly have picked a better time to call the election. Opinion polls published before the announcement put Labour on 50% - well above the vote it received in the 1997 landslide. Taken at face value, we might assume Labour is not only on course for victory, but one of even greater magnitude than last time.
So why do commentators say Blair is nervous? Two reasons: if the polls are right, turnout will probably fall. There is less incentive to vote if the re-sult appears to be a foregone conclusion - turnout dipped in 1983 and 1997 when the result was clear and rose in 1992 when the race was tight.
If turnout falls, this could be bad for Blair in at least two ways. First, polls early this year suggested it would be Labour voters who were more likely to stay at home - though the latest evidence suggests little or no party differential, especially in marginal seats. Second, if turnout does decline to an expected post-war low, it is hardly a ringing endorsement for a government that has pledged to revitalise democracy.
But then we have a second question: are the polls right? And if they are, why are we getting variation in the results from different polling companies? The first point inevitably takes us back to the disaster of 1992. Despite polls collectively suggesting a Labour victory, or at least a hung parliament, the Conservatives won, albeit with a slim majority. Critics took great delight in debunking pollsters and the oft-repeated refrain of "the only important poll is the actual election" appeared to be something more than politico-babble.
Reports of the death of the polls were, however, exaggerated. First, dismissing the polls on the basis of one rogue election was disingenuous. Taking all elections since the war, the polls have actually performed very well - in only two, 1970 and 1992, have there been real hiccoughs. Second, the pollsters engaged in collective reappraisal of their methods. The result was that the performance in 1997 was better, except in one respect - as with 1992, Labour's share was exaggerated.
So why do they vary? Polls are based on samples of the population and some small variation is almost inevitable. Plus, the response of companies to the difficulties of 1992 has varied. Much attention was focused on the "don't knows" or "won't says". These people, some argued, were disproportionately "shy Tories" - people who would vote Conservative, but felt uneasy saying so. Some companies, therefore, decided to "adjust" their polls, for example, by posing supplementary questions - such as previous vote, views on the economy and party leaders - to "don't knows" and "won't says" to try to establish whether their vote could be predicted.
Other companies resisted this trend altogether. The result is that no uniform technique has been established and, consequently, the results are bound to vary slightly. But the important thing is that while the polls do vary to a degree, they all tell the same story - Labour is well ahead. Not so difficult, you might say, but then the polls also told the right story during the fuel crisis last year - they all picked up the rapid decline in Labour's ratings and the equally brisk rise in Tory support. …