YouGov’s latest poll for the Sunday Times has caused a stir. It shows the Conservatives just 2 points ahead of Labour; given the bias in Britain’s political geography this would leave Labour with more seats than the Tories – even if the Tories do secure a bigger swing in Labour marginals than the rest of Britain. (My own guesstimate is that there would be around 300 Labour MPs, 270 Conservatives and 50 Liberal Democrats; if we assume a uniform swing, the figures are: 317-263-41.)
Since the Sunday Times published our results, a number of people have asked, quite reasonably, are our figures right? Could ours be a rogue poll?
Well, we haven’t repealed the laws of probability. All surveys, however they are conducted, are subject to a margin of error. Normally, figures are presented at the 95% confidence level. So, if the sampling error is given as “plus or minus 3%”, this means that the figures for each party should be within 3 points of the published figure 19 times out of 20. But one time in 20, the truth will lie outside that range.
However, conventional sampling calculations don’t apply to YouGov. Our panel-based methods allow us to sample our respondents, and weight the raw data, with greater sophistication than other companies. In particular we go to great lengths to ensure that our weighted data accurately reflect the political loyalties of British electors. That doesn’t mean we are subject to no sampling error, simply that it can’t be calculated the normal way. Taking account of YouGov’s record, I believe that two times out of three (the “one sigma” level) our results will be accurate to within one point, and nineteen times out of twenty (the “two sigma” level), accurate to within two points.
That is certainly consistent with our record of accurate eve-of-election prediction polls, such as the final poll in the 2005 general election, London’s mayoral election and last year’s Euro-election.
What we can be sure of, then, is that the Conservative lead has narrowed since late January. What we can’t be sure is whether it is precisely two points – or, even if it is, whether it will remain that low. One of the reasons for conducting daily polls is to monitor the fluctuations in party support. Over the next few days, our polls for the Sun will tell us whether our Sunday times poll reflects a blip or a trend – just as it takes us a few days before we can judge whether a one-day 100-point movement in the FTSE 100 share index is a short-term wonder or part of a longer-term change in market sentiment.
What we can say is that the recent dip in Tory fortunes is connected with the economy. YouGov has not had a double-digit Tory lead – the margin they need to be sure of a working overall majority – since the second week of January. It looks as if the economy is the main reason. As Britain emerges slowly from recession, doubts are mounting that David Cameron and George Osborne are the best people to steer Britain back to prosperity. For the first time since Gordon Brown’s honeymoon period as Prime Minister, Labour has edged ahead of the Tories as having the better team to run the economy. The Tories’ most urgent task is to reverse this slide in perceptions of their economic competence.
