YouGov/Sun - CON 33, LAB 43, LD 10, UKIP 10

This morning's YouGov poll for the Sun is here and has topline figures of CON 33%, LAB 43%, LDEM 10%, UKIP 10%. Once again, the Lib Dems remain in double figures.

Yesterday Opinium also put up an interesting article clarifying their methodology and pondering why it could be that they consistently show some of the highest levels of support for UKIP. First, it confirms that Opinium use two-stage prompting in their voting intention question, asking people if they'd vote Con, Lab, Lib Dem or Other, and then only giving the options of BNP, Green and UKIP to people who say other (this is the same as YouGov and in contrast to Survation, who prompt for UKIP in their main question).

Opinium's guess seems to be that their higher UKIP scores are down to not politically weighting their sample. As regular readers will know, the majority of polling companies use some sort of political weighting. In most cases they weight their samples so that respondents' recalled 2010 voting behaviour roughly matches what actually happened in 2010, with some allowances made for faulty memory, though YouGov instead weight using party identification. The two exceptions are Ipsos MORI and Opinium.

Ipsos MORI do not weight by past vote because they worry that the levels of false recall can change in response to changing public opinion, in particular they worry about people aligning their recollection of how they voted in May 2010 with how they'd vote now. Other companies like ICM and Populus acknowledge the reality of false recall, but think it is basically pretty constant and changes only slowly over time. In practice it means that MORI's samples are sometimes more Labour inclined than those of other companies, and can be more volatile (although MORI would argue that this is genuine volatility that weighting by recalled vote is disguising).

Anyway, Opinium are the other company that do not politically weight and suggest in their article that this could be why they are showing a higher level of UKIP support than most other companies. This is certainly feasible. As I've written before, the two most obvious explanations for the difference between online and phone polls in terms of UKIP support is either interviewer bias (people are embarrassed to admit to a human interviewer that they are supporting a party outside the main three, less so to a computer screen) or if people who are online are more likely to vote UKIP than those who are not (or, of course, a combination of the two factors). If online panels do get a disproportionately large proportion of the sort of people who'd vote UKIP, then not using some sort of political weighting to control for this could easily produce much higher levels of UKIP support.