Event report - RSS getstats in parliament: political polling

Written by Web News Editor on . Posted in Features

The Royal Statistical Society, together with the All Party Group on Statistics, held a lunchtime discussion on political opinion polling on 25 February 2015 at a packed event in Parliament.

The meeting, held less than three months before the 2015 general election, attracted a room full of MPs, peers and parliamentary staff. The expert speakers were John Curtice, Will Jennings and Chris Hanretty. The meeting was chaired by the All Party Group chair Lord Lipsey.

The first speaker, John Curtice, is professor of politics at University of Strathclyde, president of the British Polling Council and hosts What Scotland Thinks – a non-partisan blog on attitudes to how Scotland should be governed. He began by asking if the polling companies were all telling the same story when it comes to the forthcoming general election.

He said that in general, the pollsters were explaining that any similarities or difference in the raw polling data will be changed by three factors: weighting; turnout (assumptions of whether those polled are likely to actually vote); and the so-called ‘spiral of silence’ adjustment, which some pollsters employ to account for individuals who may refrain from voicing their opinion. He questioned whether ICM’s adjustment in this regard has served to boost the Liberal Democrat’s ratings in its polls.

Curtice noted that political pollsters were certainly giving similar figures in terms of their published estimates. Despite the different methodologies employed to collect information (face-to-face interviews, phone and internet polling), there are no major differences in the results when it comes to the main parties. However, he did note that phone polls find fewer UKIP voters than online polls.

He gave a couple of examples of where weighting techniques (which adjust answers to account for over and under-represented groups) is helping to achieve convergence. In recent polls (November 2014 - January 2015), widely used turnout weighting/filtering have served to reduce a Labour lead.

Will Jennings, a professor of political science and public policy at the University of Southampton, talked about ‘Signal, noise and uncertainty in the polls’. He pointed out that we still have an incredibly poor statistical literacy when it comes to understanding the polls. This was evident in the furore that just one poll caused in the Scottish Referendum (ie the YouGov poll which showed the ‘Yes’ vote ahead).

He explained that most short term movements in the polls are just noise, and that there are very few weekly shifts. However, the media often focuses on ‘jumps’ in polls as it makes a good story. He also explained that pollsters often change their methodology in the final weeks leading up to an election.

{mbox:lightbox/jennings-polls-feb-2015.png|width=300|height=278|caption=Click to enlarge|title=Ford, Jennings, Pickup and Wlezien - The Polling Observatory}

Jennings showed how his Polling Observatory polls the polls and controls for random fluctuations and systematic bias in the period since the last election in 2010. This graph (see right) shows how the largest weekly shifts are usually caused by the start of parliament and around the time of the 2012 budget.

Chris Hanretty of University of East Anglia and ElectionForecast.co.uk talked about potential outcomes in the general election. ElectionForecast.co.uk uses national and constituency polling and makes probabilistic forecastsabout the potential outcome.

Firstly he talked about how parties typically bounce back if they have fallen behind in the polls, and how they tend to fall back if they have seen a surge. He also explained that the uniform national swing is somewhat broken, as it won't work with smaller parties such as the Scottish National Party or the UK Independence Party.

He made a snapshot prediction based on a methodology that looks at constituencies and takes into account many variables including the population's religion, education, gender, age, social grade and past voting. His prediction showed no overall majority but put the Conservatives ahead with 284 seats and Labour on 281 (figures are updated regularly). However, these figures came with 90% uncertainty intervals. In checking the model, Hanretty pointed out that ElectionForecasts.co.uk's 'nowcasts' match Lord Ashcroft constituency polls better than uniform national swing modelling. ElectionForecast.co.uk continues to post daily updates as new information comes in.

At the end, questions from the audience were taken, which covered a number of issues but was dominated by the situation in Scotland. Addressing a question about ‘protest votes’, the panel noted that the recent Scottish independence referendum has severely affected Labour’s support in Scotland and served to consolidate the SNP vote, although the situation is still volatile. So where the UNS could still work in England and Wales, it won’t in Scotland.

A question was asked about how a national poll can account for voters in UKIP target seats. The panel answered that they don’t - as long as they don’t over-sample from a target constituency. It was also noted that UKIP support is unlikely to be uniform across the country.

Download John Curtice's slides here (PDF)

Download Will Jennings' slides here (PDF)

Download Chris Hanretty's slides here (PDF)

getstats Opinion polling 2015 General Election

Join the RSS

Join the RSS

Become part of an organisation which works to advance statistics and support statisticians

Copyright 2019 Royal Statistical Society. All Rights Reserved.
12 Errol Street, London, EC1Y 8LX. UK registered charity in England and Wales. No.306096

Twitter Facebook YouTube RSS feed RSS feed RSS newsletter

We use cookies to understand how you use our site and to improve your experience. By continuing to use our site, you accept our use of cookies and Terms of Use.