New research explores how opinion polls influence voting behaviour in the UK

Do pre-election opinion polls simply record voting intentions, or can they actually influence voters’ choices?

Opinion polls began to capture British politicians’ attention in the 1950s, thanks to a rapid rise in the number of UK polling companies and their close ties to newspaper groups which turned their findings into headlines.

Concerns that this new tool shaped voting intentions rather than simply capturing them led to discussions in Parliament and, in the wake of the 1966 General Election, calls for polls to be banned. These were brushed off by both Ted Heath and Harold Wilson. Since the 1970s, polling results have become frequently and widely reported in press and broadcast news.

But were those early critics correct? A new studyLink opens in a new window from Warwick economics researcher Eleonora Alabrese aims to better understand the influence of national opinion polls on voting behaviour by exploring the interaction between polling predictions and the UK phenomenon of the “safe seat”, a seat which changes hands rarely or never thanks to the UK’s first-past-the-post electoral system.

The research draws on electoral results from every general election between 1983, when the Conservative Party led by Margaret Thatcher returned to power for a second term with a landslide majority of 144 seats, and 2017 when Teresa May called a snap election with the aim of strengthening her hand in the Brexit negotiations but ended up losing her party’s House of Commons majority. Alabrese also compiles a dataset condensing polling information from each of the pollsters operating during these nine General Election campaigns, and draws on the longitudinal Understanding SocietyLink opens in a new window study which canvasses UK households on a broad range of topics including individual political engagement. For simplicity, electoral results, and polling information focus on the competition between the two main UK parties: the Conservatives and Labour.

The study finds that when combined with a measure of the ‘safeness’ of the constituency – the presence of a significant and pre-existent level of local electoral support for either Labour or the Conservative party – the margin of victory identified in national poll predictions significantly impacts voters willingness to show up at the ballot. This in turn affects the level of competition across different parties within a constituency, and hence their local chances of victory. The effect of polls on behaviour is also more distinct the nearer the poll is to the date of the election.

Alabrese explains: “In broad terms, if national polls predict a landslide election, turnout tends to be lower. Turnout is also generally lower in safe seats.

“My analysis looks at the joint impact of both factors. I find that if polls predict a non-competitive election, this is associated with lower turnout, and more so in safe seats.

“However, the consequences of that fall in turnout are not uniform. At local level, if a seat is safe, and the polls predict a comfortable national win for the same party, the fall in turnout advantages the local challenger. When the two factors move in different direction, that is if the local opposition party is predicted to win at the national level, incumbents in safe seats gain more as the national poll gap widens.

“Part of the explanation for this is found in the distribution of votes within the constituency. In a safe seat when the polls predict a national win for the same party, votes tend to concentrate around fewer parties – it seems that supporters of smaller opposition parties are less likely to go to the polls, or may vote tactically. When instead the main local opposition party is predicted to win at national level, we see a greater diversity of votes in the constituency. Under first-past-the-post this benefits the incumbent as opposition votes may be fragmented between several parties.

“Even if the local challenger’s party is predicted a national win, their local vote share and chances of victory fall as the constituency’s safeness increases. No matter the national trend, local opponents will struggle to overturn a safe seat.”

The evidence of a causal link between poll predictions and voting intent underlines the importance of responsible reporting of polls. There is very little formal regulation of political polling in the UK - current OFCOM Link opens in a new windowguidelines discourage the publication of opinion polls on polling day, while under Section 66A of the Representation of the People Act 1983 it is a criminal offence to report any information gained from an exit poll until after polls have closed, for fear of improperly influencing people’s voting decisions.

Alabrese also highlights the risk that safe seats which seldom change hands may create a pool of voters who feel disenfranchised and may be tempted to back more extreme policy positions or politicians.

• Read the full paper: National Polls, Local Preferences and Voters’ Behaviour: Evidence from the UK General Elections Eleonora Alabrese Warwick Economics Research Papers No: 1426Link opens in a new window, September 2022