An invitation to participants to do a PB guest slot

Tomorrow an academic conference starts in Manchester on forecasting the general election. A series of papers are going to be presented and all of them, as far as I can gather, suggest that there will be a hung parliament.

I’ve already invited Matt Lebo of Stony Brook University in New York to do a guest slot on his model that seeks to predict outcomes from from prime ministerial approval ratings. To give you a flavour this was from an email he sent to me last night:-

“We now have some more recent numbers – MORI’s Feb. poll and YG’s recent polls. So we have redone the forecasts and now have a near certainty of a hung parliament with Labour ahead in seats based on a 3.4% conservative vote lead.”

This was his forecasting model for the 2005 election. Part of my reply was over the narrow choice of pollsters for his study – MORI and YouGov

I noted “..Only ICM and Populus are operating in the same broad way as they did in 2005. The ICM approach has been consistent since the mid 1990s and for you to be putting forward something that deviates so much from ICM’s current polling is brave. All the party HQs will tell you that they regard the firm as “the gold standard”.

Hopefully we can have a good session with Matt in the next few days.

On the previous thread Andy Cooke raises concerns about the approach in a separate paper to the polling over-statement of Labour He wrote:-

“..Having come to the conclusion that the industry overstates Labour and understates the Tories, they adjust to the “ICM house effect” – and boost Labour, reduce the Tories, and boost the Lib Dems. There’s no immediately obvious combination of arithmetical operations that end up with the final figures in the table. If they can’t anchor the house corrections to the 2005 results (correct), then they can’t anchor the estimated house effects of those pollsters to ICM’s….

So the correction for pro-Labour/anti-Tory bias is to increase the Labour score and decrease the Tory score? Are we sure that they haven’t got the sign inverted in this rather important stage?”

UPDATE: It has been pointed out to me that the cover sheet for the paper referred to by Andy contains the following: ““DRAFT – PLEASE DO NOT QUOTE OR CITE. THE FIGURES CONTAINED HERE ARE PRELIMINARY ILLUSTRATIONS OF THE TECHNIQUES DISCUSSED AND SHOULD NOT BE TREATED AS OUR COLLECTIVE FORECASTS OR PREDICTIONS”.

My invitation to participants: If you can produce a summary of you argmuents in under 1,000 words I would be delighted to publish it here so we can open it up to discussion. Please email me.

Mike Smithson