Wednesday, May 31, 2017

Problems with Polls

The first thing to say about polls is that the only one that matters takes place on 8th June.

The opinion polls led people to have false expectations in the 2015 general election, in the 2016 referendum, and in the 2016 US Presidential election. The latter was the most understandable since Hillary Clinton did carry the popular vote - although this should warn us not to map polls directly to results in Britain too, since just as the US uses a state-by-state electoral college to elect their president, Britain elects a House of Commons from 650 constituencies each returning one MP.

Part of the problem is that pollsters have to make a large number of assumptions and adjustments to correct for bias in their samples - some types of voter are much harder to reach - adjust for differential turnout, guess what voters who say they have not yet made their minds up are likely to do, etc. These assumptions can radically affect what the poll predicts.

There is a very good article by Andrew Hawkins, chairman of pollster COMRES,

"Who will vote and why?"

which you can read in full here.

His article explains why some polling companies are giving the Conservatives larger leads than others: those which are predicting larger leads tend to be making their adjustments on the basis that voter turnout by age will follow the normal historical pattern (e.g. older people, who are more likely to vote Conservative, are also more likely to vote) and that those who make up their minds late will be heavily influenced by their view of party leaders. The main clue in 2015 that the Conservatives were going to do better than some headline polls was that David Cameron's ratings were much better than Ed Miliband's - and even though her approval ratings are no longer at the extremely high levels with which she entered the campaign, Theresa May has an even larger lead over Jeremy Corbyn than DC had over Red Ed.

I think Andrew Hawkins' arguments are probably more right than not, and what he is saying certainly lines up with what I am finding on the doorstep in West Cumbria.

But we don't know. If the Corbynistas are right and lot of young people who normally do not vote are motivated to do so this time, those pollsters who are predicting a hung parliament or even Jeremy Corbyn as Prime Minister could be right.

Essentially nobody can take the electorate for granted or be certain what is going to happen.

No comments: