Over at politicalbetting.com, Mike Smithson notices something odd about the Greens. They do far better in opinion polls than they do in actual elections. “Looking over polling which can be tested against real results,” he says, “there does appear to be something of a pattern – more Green supporters seem ready to tell pollsters that they will vote for the party than actually do so.” Smithson suspects that people say they’re going to vote Green when really they’re not going to vote at all. I hesitate before contradicting Britain’s most distguished commentator on polling, but surely he is wildly wrong.The poor are least likely to vote and the least likely to say they support the Greens. Environmental politics are a middle class obsession with no appeal in the slums, and the middle classes do vote A far more convincing explanation for the discrepancy, comes social scientists who point out that polls have a bias in favour of the left.
As Roger Jowell of the British Social Attitude Survey explained it to me years ago for a piece for the New Statesman
The gold standard for statisticians is the random sample. Lose randomness, and bias creeps in. Random samples, however, are costly…twice the price of the media’s polls. Addresses are picked at random. If the occupants are out, interviewers go back. If the occupants refuse to answer, they are asked to reconsider. Time is as important as money in these circumstances. You cannot have an instant random poll.
By contrast, the media commissions cheap, fast polls. Their researchers do not bang on people’s doors until they give in, but rely on pedestrians stopping strangers in the street and answering their researchers’ questions. If one shrugs their shoulders and walks on, they move on to the next, until they find someone who is prepared to give them the time of day. Who is more likely to do that, a busy Conservative, probably in work, who thinks that time is money and who does not see why he should give something for nothing, or a left-winger or Green whose ideology instructs them to extend a helping hand to importuning strangers? Tories may not like this explanation that a fair proportion of their political allies are heartless swine, but they should like the consequences which flow from it: polls habitually overestimate left-wing support, not just in Britain but in the States as well. (See a piece I did for the Observer before the 2004 presidential election here )
Smithson can see the phenomena, even if he does not understand the reasons for it. His golden rule of polling is that whenever polls are tested against real election results, the survey that “had Labour in the least favourable position is the one which is most likely to be right”. For Labour read Green as well. The latest poll gives the Tories a 13-point lead. In all likelihood, they are doing better than that.