In Defence of Liberty

Driven by data; ridden with liberty.

ComRes Iterates First

ComRes have updated their polling methodology, to account for disparities in turnouts. (Edited: BBC Poll Tracker)

ComRes have updated their polling methodology, to account for disparities in turnouts. (Edited: BBC Poll Tracker)

Many polls taken before the 2015 General Election showed either an impasse between the Conservatives and Labour, or a slim preference to one party.

The actual result was a substantial win for the Conservatives, with 36.9% of the UK-wide vote, compared to Labour’s 30.4%.

The question arises: why were the polls wrong?

Methodology

The British Polling Council have established their inquiry into these inaccuracies.

Polling companies will be updating their methodologies, to further dispel the difference between a poll of thousands and the true opinion of millions.

ComRes have made their first iteration: the ComRes turnout model.

Elections are won by those who go out to vote.

This is the correlation between a constituency turnout and demographic shares. (Source: ComRes)

This is the correlation between a constituency turnout and demographic shares. (Source: ComRes)

There is a chasm between the percentage of survey respondents who claimed they were absolutely certain to vote and the actual turnout on polling day.

Moreover, this overestimation in turnout likelihood varies by demographic segments.

“Exaggerate their turnout likelihood”

According to ComRes:

In particular, our modelling of the election result, based on constituency-level turnout data as well as ward-level turnout data from local elections held at the same time, strongly suggests that less affluent voters are more likely to exaggerate their turnout likelihood.

This phenomenon has been pejoratively and inaccurately labelled ‘Lazy Labour’, which emphasises its distortion on a single political party.

ComRes have also released their first post-election poll under this new methodology.

The effect of the turnout model is to suppress Labour and UKIP voting percentages, whilst inflating the vote share of the Greens and other small parties.

The new methodology gives a huge lead to the Conservatives. (Source: ComRes)

The new methodology gives a huge lead to the Conservatives. (Source: ComRes)

The Conservatives and Liberal Democrats were unchanged, but that may not remain consistent in future ComRes polling.

However, the applicability of 2011 census data for the 2020 General Election is questionable.

Polling, like all human activities, is open to error. The modelling of election results is an iterative process, requiring further introspection and insight to improve.

Advertisements

Information

This entry was posted on June 6, 2015 by in Statistics and tagged , .
%d bloggers like this: