Odd polling trends: UK

Website PoliticalBetting.com reports on an interesting little adjustment made by polling agency Populus, which signals a major shift in public attitudes toward the parties. The latest Populus poll of the British electorate puts the Conservatives ahead of Labour by 4% -- 36 to 32, Labour's lowest percentage in years, with 22% going to the Liberal Democrats, a party that runs to the left of Tony Blair's New Labour.

The raw numbers gave the Conservatives a 6-point advantage over Labour, but Populus adjusted Labour's numbers up by 2 points. Here's why:

Early last year, after a short period of neutrality, the spiral of silence reversed itself. Polls began to find that the proportion of former Labour voters saying that they didn’t know how they would vote next time began to climb, while the proportion of Labour supporters saying that they were sure to vote began to fall. Pollsters have been adjusting Labour’s poll support upwards to take account of this growing number because all empirical data tell us there is about a 60 per cent probability that, however reluctantly, if they vote at all they will end up voting Labour again. Without this adjustment the Conservative lead in today’s poll would have been 2 per cent higher. SO in place of Shy Tories we now have Bashful Blairites, people unwilling to admit to pollsters or their friends that they still support the Prime Minister. Once so fashionable, new Labour has now gone out of fashion. This is very difficult to reverse.

My guess is that this is due to Blair's support of the US position in the war on terror, which has been attacked fervently by the arbiters of popular culture. Backing Tony isn't cool any more. A few years ago, in the latter days of John Major's reign, voters supporting Conservative policies were embarassed into silence by sex and financial scandals involving Conservative MPs.

A question: Is the "spiral of silence" a factor in American politics? How would you adjust a poll result to account for it? I'm guessing that it varies state to state, but generally the effect would reduce stated preferences for Republicans.

All pollsters make adjustments, in part because people aren't always honest with pollsters, for various reasons, in part because a variety of factors (time of day, weather, age of respondent) can skew a voter's willingness to respond. The website for Populus has a frank and fascinating discussion of how they arrive at their adjustment factors.

3. The logical way to try to make sure a poll sample is politically representative is to ask those polled how they voted at the last election and compare what they say with the actual result – a known fact about the political views of the country as a whole that serves as a benchmark, so that if the voters who have been surveyed for the poll prove to be politically unrepresentative, the whole sample can be made representative by weighting it to the election result. For the first few months, Populus polls for The Times were, therefore weighted to the actual result of the last election.

4. But the detailed data of Populus polls bore out research at previous general elections, and surveys re-polling the same people during the course of a Parliament, all of which have shown that when asked after a general election how they voted, a lot of voters – possibly as many as one in five – don’t recall correctly: they may lie, or want to be seen to have backed the winner, or are correcting their past vote to match their future intention, or they may simply forget.

The likeliest date for the next British election is believed to be 05/05/05. Another interesting fact, cited in the PoliticalBetting.com report, is that the Tories will need a 7% popular vote edge over Labour in order to win more seats than Labour. It has mostly to do with the presence of a third major party, which tends to split opposition votes. This is another reversal from the '80s and '90s, when the same phenomenon worked in the Conservatives' favor.