What Ipsos MORI almost showed
At the weekend MORI had a poll in the Observer that showed voting intentions of CON 34%, LAB 41%, LDEM 16%. They also published the findings from their regular monthly monitor showing thing like the most important issues facing the country (currently topped by immigration and crime, which has fallen from the previous month when Rhys Jones's murder was in the news).
However, as I've mentioned before, the turnaround rate for face-to-face polling tends to be a bit slower than phone or internet polling, so the figures were almost a week old by the time the Observer came to public and hence the voting intention figures that came out were actually more up to date ones from a separate phone poll. So as not to confuse matters with two sets of voting intention data, the Observer held back the older voting intention data from MORI's monthly monitor until now, and it now appears on their website here.
The monthly monitor was conducted between September 20th and September 26th, so most of the fieldwork occured at the same time as the Labour conference, with Gordon Brown's speech bang in the middle of it. Boy did it have an effect - topline voting intentions were CON 31%, LAB 44%, LDEM 15%, Labour's best performance since 2003.
Since we now have more recent figures from Ipsos MORI that show a much smaller lead we can be relatively confident that they was just a blip, or outlying figures at the top of the margin of error, but all the same it was an astonishing lead. Just think about how the first day of the Conservative conference might have been different if it was those figures that had been on the newspaper front pages as delegates travelled to Blackpool rather than the more managable 7 point lead the Observer did report. A lot of the time the polls are more important for the effect they have on the political weather rather than their predictive quality (after all, there isn't an election tomorrow), David Cameron is probably lucky he avoided the squall from these figures popping up on the frontpage.
