Could this affect Commons seats projections?
Most people with an interest in politics and election outcomes are familiar with the term – Uniform National Swing. But how many understand how the numbers are made up and that there are two very different approaches to making the calculations.
I raise this because I have to admit that until not so long ago I got it all wrong. My assumption was that if Party A was up by a tenth on the last election; Party B was down by a ninth and Party C was up by another proportion then those changes were applied in each seat to the results from the time before and an overall projection was produced. Not so.
The traditional way that it’s done is to take the actual percentage changes nationally and then add or subtract these figures to the previous result seat by seat. So in simplistic terms if a poll shows a Labour share of 27% you apply the difference on the GB general election result, in this case nine points, and deduct that from the party’s figure in each seat.
Yet in this example Labour’s share is down by a quarter – so why not simply hack off that proportion in every seat? This could make a difference in places where Labour’s share last time was greater than the GB result of 36.2% and includes just about all the marginals that Labour is defending.
The Baxter approach is based broadly on the proportionality of the swing; Wells follows the traditional route of simply applying the same points change in vote shares to every seat.
So which system is best â€“ Wells or Baxter?
The answer is inconclusive. Baxter has tested the data from both approaches from previous elections and found no appreciable differences except at the 1997 general election when, of course, there was a massive swing to Labour. Both approaches to calculating seats underestimated how many Tony Blairâ€™s party would pick up. According to Martin Baxter the proportional model did better then by eight seats.
Like Tony Blair’s great victory twelve years ago it looks as though the next election will produce a step-change. Maybe on this occasion the proportional model will do best?
Labour seats betting.
Lib Dem seats betting.
Conservative seats betting.