Does “party ID” weighting understate UKIP?
Yesterday I posted a couple of Tweets about the impact on the party ID weighting in the new online Populus poll which has just been launched.
Instead of sticking with its traditional past vote weighting approach the firm had adopted party ID weightings based back to what was recorded in British Social Attitudes Survey of 2010, when, of course, UKIP support was much lower than it is today. The effect is quite marked as this Tweet observes.
The raw Populus poll data 205 responses saying UKIP and 113 Lib Dem Headline voting figures are 13% LD and 10% UKIP http://t.co/xBCpO1Kqf3
— Mike Smithson (@MSmithsonPB) July 15, 2013
This is the response of Rick Nye – head of the firm:-
“In all the trials we ran developing this new methodology we found that past vote weighting of an online sample did not produce reliably comparable figures to past vote weighted, spiral of silence adjusted telephone-based voting intention, which Populus had used for the past decade and which we took as our benchmark.
There were a number of potential responses to this: one was simply to live with the difference between the two sets of figures, the second was to eliminate any attempt to introduce political representativeness into our online voting samples, and the third was to try and find a different way to make those samples more politically representative. While each has its merits – and its adherents among British Polling Council members – we opted for the last of these approaches.
As I touch on in the note that accompanied the release of our new voting intention figures, we began constructing our online panel shortly after the 2010 General Election. Had we built it beforehand we would most certainly have screened it for party ID and also compared what people said they did at that Election with what they said they were going to do in the run-up to polling day. In the absence of that information we have opted for what we believe is the next best thing, the party identification people gave in response to the British Social Attitudes Survey in that last General Election year. As you point out trying to make our sample more representative politically in this way has the effect of down-weighting UKIP considerably and at the moment of up-weighting the Liberal Democrats – as it does in YouGov polls which use their own version of Party ID weighting. We have been absolutely transparent about that. It is why we have published the unweighted and weighted percentages for Party ID alongside our voting intention figures.
As we have always done, and in line with those who make no such adjustments or who make different adjustments, we will keep our methodology under review as time goes on. In the meantime I’m sure the debate will continue on your blog and on others about whose most accurately reflects the current political situation. That is the beauty of transparency and disclosure.”
For the latest polling and political betting news