Ipsos MORI/Reuters - CON 32, LAB 39, LDEM 11
Ipsos-MORI's monthly political monitor is out, and has topline figures of CON 32%(-5), LAB 39%(nc), LDEM 11%(nc), Others 18%. Changes are from last month's poll, and clearly show a significant drop for the Conservatives and a boost for minor parties.
On leader ratings David Cameron's net approval stands at minus 15, a signifcant drop from last month's minus 2. Ed Miliband's rating is minus 7, up from minus 15 last month and, I think, the first time a poll has shown him with an approval rating above David Cameron's. Scrap that bit - Miliband certainly had more positive ratings when he first became leader and it's actually been quite common for Miliband to have better net ratings than Cameron in MORI's version of the question
One caveat about this though, it has a slightly odd sample. MORI's sample contained significantly fewer people who said they voted Tory in 2010 than it usually does, in fact the weighted sample still had more people who claimed they voted Labour in 2010 than claimed they voted Tory. Regular readers will know that ICM, Populus and YouGov all use political weighting to make sure how people in their sample claim they voted at the last election roughly reflects what actually happened (with some variation due to assumptions about false recall). MORI do not (for reasons which I'll come to below), therefore the political make-up of their sample can be significantly different from one month to the next.
Looking at the polls over the last few days and based on the information available from each company's tables (only ICM gave full details in their tabs, so the others make some assumptions about the proportion of 2010 "others"):
In the last ICM poll, people's recalled 2010 vote was CON 37%, LAB 30%, LDEM 24% In the last Populus poll, people's recalled 2010 vote was roughly CON 37%, LAB 30%, LDEM 24% In the last YouGov poll, people's recalled 2010 vote was roughly CON 36%, LAB 31%, LDEM 25% In the last Ipsos-MORI poll, people's recalled 2010 vote was roughly CON 34%, LAB 35%, LDEM 23%
The reason MORI do not weight by party recall or past vote is because they are concerned that people's past recall of how they voted could change rapidly, and therefore weighting by it risks dampening out genuine volatility in public opinion. ICM and Populus also think that levels of false recall can move, but think it changes only slowly over time, something which their model takes into account. YouGov are panel based, so store respondents' answer to how they voted in 2010 and don't need to worry about false recall changing.
These are legitimate differences of opinion, and obviously the different companies each believe that they are doing what is correct... but we shouldn't be surprised if they result in different answers, and if a poll that has far more 2010 Labour voters in it is better for Labour and Ed Miliband than a poll with more 2010 Conservative voters in it.
