How changing poll methodology is adding to Gord’s problems

How changing poll methodology is adding to Gord’s problems


    Was “public sector weighting” behind MORI’s Tory 52%?

One of the biggest polling changes that we’ve seen in years was introduced by Ipsos-MORI in the summer following a review after the London Mayoral elections. For in trying to work out why its surveys seemed to be over-stating Labour the firm discovered that people working in the public sector were being over-sampled.

So a new calculation has been created and this has been one of the reasons why the firm is now showing the worst position for Labour of any of the firms.

Just look at the detailed table above to see the impact. Highlighted are the responses from those who worked in the public sector and from the raw sample almost a quarter (244 from 1017) were in this category. This is far higher than the actual proportion in the adult population and the pollster scales that back by more than half to 119.

But look further at the numbers and you can see what a big impact this had on the overall outcome. For the public sector workers split C32%-L32%-LD20%. Compare that with the non-public sector workers and you get a split of C49%-L29%-LD14%. (These shares, it should be noted, are before MORI has applied its certainty to vote filter which normally provides a further uplift in the Tory position).

So by reducing the importance of the public sector segment the poll gives higher figures for the Tories and a lower one for Labour. But the biggest sufferer is the Lib Dem party. What I find amazing is how differently public sector workers responded compared with the rest of the sample.

My respect for MORI increased enormously following their reaction to their London Mayoral election performance. They stopped publishing political polls while they tried to assess what went wrong. The firm remains the only one of the main pollsters not to weight by party ID or by what people say they did at the last election. It could be that public sector weighting is a better solution.

I have asked the firm what the outcome of this week’s poll would have been without this element. I’ll publish it when they’ve worked it out.

Mike Smithson

Comments are closed.