“Labour has already won”
I wrote in The Independent on Sunday yesterday about predictions of the next election, again, prompted by the assertion by Nick Harvey, the Liberal Democrat MP (pictured), that “Labour has already won.”
One of the axioms on which he builds this fallacy is this: “It’s very difficult to see why anyone would vote Tory next time who didn’t last time.” Well, no, because there is always churn – people change their vote, move between voting and non-voting, and come onto the register in between every election.
Anyway, I put his best-case Tory scenario into the Electoral Calculus seat predictor thus: Conservative 37% Labour 35% Lib Dems 15% UKIP 7%, and it came up with Labour the largest party, but 19 seats short of majority. In other words, what Alastair “I have called every general election correctly” Campbell, that other Mystic Meg, says: a Lib-Lab government.
There are four known unknowns to apply to this model, however.
1. Lib Dem incumbency. The Lib Dems will hold more of their seats than a uniform swing would suggest. This will reduce Labour gains.
2. UKIP. As you see, I have put them into Electoral Calculus at 7 per cent: twice what they got last time. I think they will be squeezed from where they are now in a general election: anti-Labour voters won’t want to put Ed Miliband in Number 10, so I think UKIP will take fewer Conservative votes compared with Labour ones than many people think, but they will still have a differential effect, reducing Tory seats.
3. The economy. See my column, in which I quote David Cowling, the BBC’s political analyst, who in turn cites a Populus poll that finds that, even if people think the economy is recovering, they don’t feel it, and the MORI economic optimism index, which turned sharply favourable before 2010, which didn’t win the election for Gordon Brown. But I think wages will catch up with prices next year and that this will lift the Tory vote.
4. Swing back to government. Stephen Fisher, of Trinity College, Oxford, has just updated his model of the next election, which takes account of how opinion has tended to swing back towards the government before elections in the past.
Fisher’s central seats forecast is:
Con : 323 (largest party, short of a majority by 3)
Lab : 280
LD : 20
And his probabilities are:
Con majority 48%
Lab majority 22%
Hung parliament 30%
Con largest party 64%
Lab largest party 36%
Note, though, that his model does not include a UKIP effect.
Compare this with the Electoral Calculus probabilities:
Conservative majority 4%
Labour majority 81%
Hung parliament 15%
And the probabilities implied by the betting odds on the party to win the most seats:
Con largest party 40%
Lab largest party 60%
Last word, though, to Lewis Baston on the history of election forecasting:
There have been several predictive models that have performed reasonably well in the past in British elections. The first one (Goodhart and Bhansali) worked well in the 1950s and 1960s and depended on little more than a time-lagged response to changes in unemployment. It came a cropper in 1983, because the rise in unemployment was so extreme that it predicted the Conservatives would win a negative share of the vote. Another econometric model, devised by David Sanders, worked like clockwork for the 1980s and, in advance, 1992. But it was completely off-beam in 1997 when it predicted a Conservative victory.
So there you go. Prediction models are highly reliable … until they are not.Tagged in: 2015 election, opinion polls
Latest from Independent journalists on Twitter