December 12, 2019
The Use of Opinion Polls in General Elections
Written byGeof Andrews

In the weeks before the electorate heads to the polls, news cycles were filled with the results and analysis of polling. Media across the political spectrum have used them to make arguments based on the promise of what may happen, could happen, or will happen.

What is a poll? From a quick show of hands, to an online survey optimised for your smartphone, polls can take many forms. They all share the same ‘aim’: “to ask a representative sample of people a question and then show the results to that question, broken down by demographics.” They serve to illustrate the wider opinion towards almost anything. Commonly they are used in market research to see how the public feels about different brands. At election time it is attitudes to politicians, political parties and voting intentions that grabs the media attention the most.

There is a host of companies that run polls, such as Yougov, ICM, ComRes, Kantar, Survation, Opinium and many others, not to mention smaller scale regional and constituency polls being carried out, as well as the polling by the political parties themselves.

Sampling, simply the way of choosing who answers the polled question, can be random, stratified, opportunistic or a myriad of other forms. The sample is often the easy target for criticising a poll. Truly representative samples are difficult to achieve, and participants may have biases from self-selecting themselves. In election opinion polls the samples are always small when compared to the electorate, typically 1000-2000. The Yougov MRP model (itself not a poll but an analysis of polls) is based on 100,000 replies, which sounds impressive until compared to the c.46 million eligible to vote in 2015 where it becomes miniscule. It breaks down to 150 responses in each constituency.  Despite the potential for major errors polls tend to be reasonably accurate. Margins for error are very good and results generally fall within them.

Before getting carried away about predicting the future there is a large difference between polling percentages and the results of a General Election. For a first-past-the-post system, polling results are problematic in trying to predict the actual outcome. In 2017 the Labour Party had their greatest share of the vote since 2001, just two per cent behind the Conservative Party, and yet totalled eight per cent fewer seats in Westminster. Liberal Democrats have perennially complained about suffering under the system. How the votes fall geographically has just as much impact as the number of votes. Greater support in a safe seat will absorb increases in polling numbers to no effect on the overall result that matters, seats in Westminster.

Much attention is given to polls for the ability to predict the outcome of an election, yet this is not their purpose. Instead they provide a window into meaning during a campaign. Polls help to demonstrate, to journalists, politicians and members of the public, what issues are important and are leading to an impact on general opinion.

Next we will look at the difference between what the polls are predicting and the actual results. At the time of writing on the day of the 2019 General Election an average of polls is giving:

Conservative Party 43%
Labour Party 33%
Liberal Democrats 13%

The Guardian

 

With two analysis models (more on those next time) giving seat predictions of:

Conservative Party 339
Labour Party 231
Liberal Democrats 15
SNP 41

YouGov MRP

 

Conservative Party 351
Labour Party 224
Liberal Democrats 13
SNP 41

Electoral Calculus

Categories

Related posts

‘The Museum of the Carbon Age’, Green Man Festival 2026

‘The Museum of the Carbon Age’, Green Man Festival 2026

In this year’s Green Man Festival (August 2026), our planned session, ‘the Museum of the Carbon Age’ didn’t go at all to plan. And it was much, much better for it. Our premise was simple enough. In a large tent, we curated an exhibition where each exhibit is an item,...