What does this say about current surveys?
The table above shows all the published opinion polls in the January before the last election in May 2005 and compares them with what actually happened.
First thing to note is the methodological changes that three of the firms have made which makes direct comparisons more difficult.
YouGov introduced a new party ID weighting structure to deal with their pre-May 2005 tendency to over-state the Lib Dems at the expense of Labour.
ComRes introduced past vote weighting in March 2007 and made further changes in June 2009.
MORI was a mix of phone and face-to-face polling. This was changed to all phone polling in June 2008 when the firm also introduced public sector work weighting.
What’s interesting from the table is that all the phone pollsters over-stated Labour and this continued, with one or two exceptions right through to polling day.
The big message I get from these numbers is the way, like today, the Tory numbers show very little variance from pollster to pollster and poll to poll. All the polls bar one under-shot by small amounts.
So if the January 2005 polls are a good pointer then we should be expecting a Tory share in the 40 – 41% region.
It’s the Labour-Lib Dem split which seems to be crucial. When one goes up the other goes down and all the phone surveys overstated Labour and understated the Lib Dems.