Issues & Ideas | 17th October 2014

Why we can’t predict elections like we did 50 years ago

We’ve taken a look at how limited election polling is in Britain, but what […]

Photo: Getty


We’ve taken a look at how limited election polling is in Britain, but what can we do with the data we have?

The fraught task is to turn abstract national polling numbers – which we are averaging every day in May2015’s “Poll of Polls”, and which currently put Labour on 33, the Tories on 30, Ukip on 17 and Lib Dems on 8 – into actual seat totals for each party.

It wasn’t always complicated. Psephologists used a method called “Uniform National Swing”, which looks at how national polls have changed since the last election and assumes this national “swing” will happen in every seat.

The seats are ranked in order of how marginal, or close, they were in the last election – as we do for 2010 in our “Seat Calculator” – and the national swing is applied to all of those seats. So the model would hand a seat that the Tories won by 10 per cent at the last election to Labour if the national swing was 5 per cent (the swing is half the change in majority).

Here is David Butler, who spent decades covering election night on the BBC and turned 90 today, explaining how it works in 1964.

“How can we talk so confidently about the swing that’s needed for a Labour government? We hope in the very near future to be able to give you a firm prediction of the final result when only a few constituencies have reported. How can we do it? Well basically it’s because Britain is such a unified nation.”

In past elections there has been an extraordinarily even swing, seats move the same, whether they’ve been safe Conservative or safe Labour or highly marginal seats. So when the first result is in, it’s quite reasonable – not very speculative – to say, ‘If the whole country behaves like Billericay, or Walford, or wherever it is, then the final result will be so and so’.”

Now the first result may of course be a bit off the line. However, when a few results are in, the pattern will be, or should be, clear. We believe we can catch the trend very quickly.”

But this method has been complicated by the rise of minor parties.

For the first three decades after the war, other parties won few votes. Seven in eight voters backed Labour or the Tories in every post-war election until 1974.

That changed when Jeremy Thorpe’s Liberals won 19 per cent that year. For the next quarter of a century Britain’s third party kept the “other parties” vote share at around 25 per cent.

But in the last decade that number has gradually climbed. In each of the past elections, the “other parties” nearly won more votes than the winning party (Labour in 2005, the Tories in 2010).

Now, after another spike for Ukip in the polls over the past week, our Poll of Polls shows “other parties” would win more votes than either of the major parties in an abstract national election tomorrow – which would be a post-war first.

Ukip’s popularity could fade, as the SDP’s did in the early 1980s, but we have shown how, compared to then, twice as many voters now no longer identify with a political party.

Disaffection is unlikely to disappear soon. And that makes Britain’s famed swingometer, more recently manned by Peter Snow and Jeremy Vine on BBC election nights, increasingly limited.