Leader ratings have proved a far better guide to election outcomes
A month today, on January 19th, the investigation into what went wrong with the general election polling will be announcing its findings at a special event in London.
No doubt all sorts of tweaks will come out of it but the main factor. I’d argue, is that we (and I include myself in that) paid far far too much attention to the voting intention numbers. In almost every case these are asked right at the start of the interview/questionnaire and are almost akin to finding out what tribe those polled think they are in. Whether they’ll actually vote that way is a different matter.
2015 was the second general election in the past six when the voting intention findings have been wrong – but when the leader ratings questions were asked these proved to proved spot on in identifying the winner
Look at the chart above showing the MORI then Ipsos MORI LAB and CON leader ratings in every election since 1979. The satisfaction “leader” in each case went on to head the party that won the election.
In 1992, the previous massive polling fail, the leader rating differential between Major and Kinnock was far larger than what we saw in May.
Regular PBers might have noticed that since the election I’ve not been putting much emphasis on voting numbers but have been giving the bulk of the site’s polling coverage to leader ratings. That will continue.