Do more say they were Yellow than actually were?

We all know the big story of the general election and the polls. When the votes were counted the 23.6% GB share that the Lib Dems chalked up was much smaller than all the pre-election polls. At one point YouGov had Clegg’s party with a four point lead on 34%.

Yet eleven weeks on from May 6th an interesting trend has developed.

When pollsters ask how people voted they invariably find that a greater proportion “remember” voting Liberal Democrat than actually did so.

In fact the “memory” of interviewees is more in line with what people were telling pollsters in the final week – not with what actually happened.

A big academic post-election survey has observed the same phenomenon.

After previous elections the false recall element has tended to apply more to declared Labour voters not the other two parties.

The possibility of mis-remembering is the reason that the pollsters that past vote weight their samples factor in an element to deal with false recall. TICM, Populus and ComRes each have a different formula for dealing with this and different target weightings.

In its voting intention polls ComRes, for instance, has targets of C24-L19-LD15 – the balance being other parties, refusers, and those who say they didn’t vote.

Unfortunately, and wrongly in my view, ComRes does not apply these weightings in its surveys for the BBC. So in last Friday’s Newsnight poll a third more people who said they “remembered” voting Lib Dem were included than in their main VI surveys. This can skew the results.

The one past-vote weighting pollster at the general election that did not have a “false recall” corrector was Angus Reid – and we all know what happened.

Mike Smithson