After the recent discussion on the uniform national swing I dug up a paper from Professor John Curtice of the University of Strathclyde in which he looks at ways of projecting likely commons seat number from opinion poll ratings.
He wrote “..The most commonly used method for doing this is to make a â€˜uniform changeâ€™ projection. If, say, a poll estimates that Labourâ€™s share of the nationwide vote is down three points on what it won at the last general election, while the Conservatives are up two and the Liberal Democrats up one, it is assumed that these changes in vote share occur in each and every constituency. We then establish how many constituencies would change hands as result of such a movement…..
In effect what this procedure does is to assume that a partyâ€™s vote will be distributed across the country at the next election in much the same way as it was at the last one. While it is recognised that in practice the change in each partyâ€™s share of the vote will vary from one constituency to another, it is assumed that this variation will have no net impact on how each partyâ€™s vote is distributed; for every seat that a party fails to win against expectations there will be another that it captures against expectations. The outcome of the 2005 election, however, illustrated the potential fragility of this assumption...”
Curtice goes on to show that the UNS in 2005 had Labour with a majority of 28 seats more than they actually achieved and outlines his approach to dealing with the challenge.
Although Curtice does not express it in these words 1997 and 2005 both point to the party which is losing support doing disproportionately badly in terms of seats won compared with the UNS calculation. Maybe this will happen again next time.
The whole paper is well worth reading.