The Media and Opinion Polls

UK political media divides into three sectors, and the differences between them are legal, not just editorial.

Sector	Regulation	Free to take a side?
Newspapers	Self-regulation, mainly through IPSO since 2014	Yes — a title may endorse and campaign for a party
Broadcasters	Ofcom Broadcasting Code; the BBC also by its Royal Charter	No — bound by due impartiality, tightened during elections
Social media and online	Little content regulation; political advertising is not pre-vetted	Yes — and targeting is invisible to non-recipients

That asymmetry is the foundation of every essay on this topic. A newspaper may print a front page urging its readers to vote a particular way on polling day; a broadcaster doing the equivalent would breach the Code.

The balance between the three has shifted decisively. Newspaper circulation is estimated to have fallen by around three-quarters since 1997. Televised debates drew 9.4 million viewers in 2010 and 4.8 million for the first ITV debate in 2024. Meanwhile the 2024 election confirmed the pattern set in 2017 and 2019: a growing share of voters take their political information from social media, where parties spent heavily — Labour £2.9 million and the Conservatives £1.7 million on digital advertising across the six-week campaign.

When you argue about "media influence", say which media. The case for press power in 1992 and the case for social media power in 2019 rest on completely different evidence.

Media bias means the systematic slanting of coverage — through story selection, prominence, language and choice of voices — towards a party or position. In the UK press it is open rather than concealed, expressed in formal endorsements.

For most of the period 1979–2010 the press leaned heavily Conservative. The alignment at the 2024 election was different:

Backed Labour	Backed the Conservatives	No endorsement
The Sun, Sunday Times, Financial Times, Independent, Guardian, Daily Mirror	Daily Mail, Daily Express, Daily Telegraph	The Times

The Sun is the case study worth learning in detail, because its three switches frame the whole debate.

• 1992. The paper campaigned hard against Neil Kinnock, including a polling-day front page inviting the last person leaving Britain to turn out the lights. The Conservatives won 41.9% to Labour's 34.4% after most final polls had suggested a hung parliament, and on 11 April 1992 The Sun claimed the credit for the result in a front-page headline.
• 1997. It switched to Labour on 18 March 1997, the day after the election was called — by which point Labour had led the polls for four years.
• 2024. It backed Labour again on the eve of poll, using a football headline about time for a new manager, having supported the Conservatives since 2010.

The pattern in all three is the same: the paper moved towards a party that was already winning. That is precisely why its 1992 boast is treated sceptically by political scientists.

Three competing models are used to explain media influence. A good answer names them and tests them against evidence.

The direct effects model

Coverage changes minds; readers absorb what they are told. For it: The Sun's 1992 campaign coincided with a late Conservative surge no poll had captured, and readers who switched papers have been shown to shift their voting intentions with them. Against it: readers self-select. Someone who buys the Daily Mail is likely to already lean Conservative, so correlation between readership and vote proves little about causation.

Reinforcement theory

Coverage mostly hardens existing views. For it: newspaper endorsements follow rather than lead the polls, as The Sun did in 1997 and 2024, and Pearson's 2024 election case study reaches the same verdict on the paper: its endorsements track what its readers already believe instead of steering them. Against it: reinforcement still matters electorally, because turnout depends on motivation and a reinforced voter is a voter who actually goes to the polling station.

Agenda-setting

The media may not tell voters what to think, but they strongly influence what voters think about. For it: in 2019 sustained coverage of Brexit deadlock kept a single issue salient, and the Conservatives won 74% of Leave voters. In 2024, coverage of NHS waiting lists, small-boat crossings and the collapse of the Truss mini-budget set the terms on which the government was judged. Against it: parties and events, not editors, generate most of what is covered.

The safest exam judgement is that the media are more powerful at setting the agenda and reinforcing existing loyalties than at converting voters — but that in an era of partisan dealignment, agenda-setting and reinforcement are enough to shift a close result.

Broadcasters operate under the Ofcom Broadcasting Code, whose rules on due impartiality are applied more tightly during an election period. This produces the "stopwatch" coverage familiar from campaigns, and it means a broadcaster cannot endorse a party.

Televised leaders' debates arrived in the UK in 2010, decades after the United States. Their record is a useful corrective to claims that campaigns decide elections.

Worked example: the 2010 debates and Cleggmania. About 9.4 million people watched the first debate on 15 April 2010, in which David Cameron and Gordon Brown competed to agree with the Liberal Democrat leader Nick Clegg. Liberal Democrat poll ratings rose sharply, in some polls into the low thirties, and commentators speculated about a realignment. The actual result: the Liberal Democrats won 23.0% of the vote, up one point on 2005, and 57 seats — five fewer than the 62 they had won before the debates existed. A campaign event dominated three weeks of coverage and moved the seat total backwards.

Two further points about debates:

• They are negotiable, and leaders duck them. Theresa May declined to take part in the 2017 leaders' debates. In 2019 Boris Johnson refused an interview with Andrew Neil that other leaders had granted, and was replaced by a melting ice sculpture in a Channel 4 climate debate, prompting a bias complaint from his party. None of this visibly damaged the Conservative campaign, which won an 80-seat majority.
• Reach is falling. 9.4 million in 2010, 6.7 million for the Johnson–Corbyn debate in 2019, 4.8 million for the first Sunak–Starmer debate in 2024. Snap polls after the 2019 debates split 52% to 48% in Johnson's favour on who had won — close enough to settle nothing, which is the usual outcome.

Social media differs from press and broadcast in three ways that matter constitutionally: it is targeted, so different voters see different messages; it is cheap, so smaller parties can compete; and it is largely unregulated for political content, so claims are not subject to any impartiality duty.

Its use has developed election by election:

Election	What changed
2017	Labour's use of shareable video and public meetings outpaced the Conservatives; estimated 18–24 turnout rose to 58% from 43% in 2015
2019	Dominic Cummings and Isaac Levido applied Vote Leave methods, using focus groups and polling to target Facebook advertising at marginal seats in the final week; the Conservative press account was temporarily renamed "factcheckUK" during a debate
2024	Digital advertising spend of £2.9m by Labour against £1.7m by the Conservatives, with targeting by age, gender and locality

The democratic argument cuts both ways. Targeting lets parties reach voters who ignore traditional news, and lets a party without press support build an audience. Against that, messages seen by a few thousand voters in a marginal seat escape the scrutiny that a front page receives, factual claims face no impartiality requirement, and the 2019 renaming episode shows how easily a partisan account can be dressed as a neutral one.

Section 4.2 asks about the media "during and between" general elections. The between-elections role is where the media's constitutional function is clearest: with a large Commons majority weakening parliamentary scrutiny, investigative journalism is often the most effective check on government.

Three worked examples, in ascending order of consequence:

1. Cash for questions, 1994. The Guardian alleged that MPs had accepted payment for tabling parliamentary questions. John Major responded by establishing the Committee on Standards in Public Life under Lord Nolan, which produced the Seven Principles of Public Life — a permanent change to the ethics machinery of British government, triggered by a newspaper investigation.
2. MPs' expenses, 2009. The Daily Telegraph published leaked details of members' expense claims over several weeks. The Speaker, Michael Martin, resigned — the first to be effectively forced out in over three centuries — a large number of MPs stood down or were deselected, several were prosecuted, and the Independent Parliamentary Standards Authority was created to take expenses out of Parliament's own hands.
3. Partygate, 2021–22. Reporting of gatherings in Downing Street during Covid-19 restrictions produced sustained public anger and, combined with the handling of the Chris Pincher affair, contributed to Boris Johnson's announcement of his resignation in July 2022. A Prime Minister with an 80-seat majority was removed between elections, with the media supplying the evidence and Conservative MPs acting on it.

The same dynamic was visible in 2026. After Labour's losses at the May 2026 local elections, coverage of the resulting leadership crisis ran alongside around 103 Labour MPs calling publicly for Sir Keir Starmer to go; he announced his resignation on 22 June 2026, and Andy Burnham became Prime Minister on 20 July 2026. As of September 2026 the UK had changed Prime Minister without a general election, under sustained media scrutiny throughout.

Scrutiny has limits. Journalists chase what sells, coverage concentrates on personality and scandal rather than legislation, and public bill committee stages of major Bills pass with almost no reporting at all.

An opinion poll surveys a sample designed to represent the electorate. UK polls are published freely throughout a campaign — there is no publication blackout — but section 66A of the Representation of the People Act 1983, inserted by the Representation of the People Act 2000, makes it an offence to publish exit poll findings before the close of poll.

The accuracy record is uneven.

Election	What the polls said	What happened
1992	Final polls pointed to a hung parliament or narrow Labour lead	Conservatives won 41.9% to 34.4% and a majority of 21
2015	A hung parliament	Conservative majority of 12
2017	Ipsos MORI an 8-point Conservative lead, ICM 12	Actual lead about 2.5 points; May lost her majority
2019	Average of Con 43, Lab 33, LD 12, Brexit 3	Con 43.6, Lab 32.1, LD 11.6, Brexit 2.0
2024	Labour lead of 15–20 points	Labour lead about 10 points

2019 shows that polls can be accurate on vote share. The failures have specific causes: the 1992 inquiry pointed to late swing and to Conservative supporters declining to disclose their intention, and the inquiry into 2015 commissioned by the British Polling Council and the Market Research Society, chaired by Professor Patrick Sturgis, identified unrepresentative samples as the main problem. Even an accurate poll converts badly into seats, because first past the post rewards concentration rather than volume — which is why Reform UK's 14.3% of the national vote in 2024 returned only five seats.

Exit polls are a different instrument and have a much better record. Respondents report how they have voted, not how they might. The 2019 exit poll predicted a Conservative majority of 86 within minutes of polls closing, against an actual 80. In 2024 the Ipsos exit poll gave Labour 410 seats and the Conservatives 131, against an actual 411 and 121.

Why polls matter politically, whether or not they are accurate:

• They shape party strategy. Labour's 2024 campaign concentrated resources on marginal seats identified by private polling, while the Conservatives were tied to defending seats won heavily in 2019.
• They influence election timing. Callaghan's decision not to call an autumn 1978 election and Sunak's decision to call one in May 2024 were both judgements about which way the polls would move next, and both went badly for the incumbent.
• They enable tactical voting. Voters can only vote tactically if they know which candidate is best placed locally, which is what constituency polling and the guidance issued by campaign groups in 2024 supplied.
• The bandwagon and underdog arguments cut in opposite directions. A published lead may attract support to the likely winner or, equally plausibly, encourage complacency and depress its turnout. Neither effect is firmly established, so hedge any claim about them.

As of early September 2026 the polls describe genuine four- and five-party competition: YouGov on 6–7 September 2026 had Labour 23%, Reform UK 23%, the Conservatives 20%, the Greens 12% and the Liberal Democrats 11–12%. Attribute and date every poll you cite.

1. Asserting that the media decide elections

The strong claim is hard to defend and easy to attack. The Sun backed the winner in 1992, 1997 and 2024, but in each case it moved towards a party the polls already favoured. Argue for agenda-setting and reinforcement, which the evidence supports, rather than direct persuasion, which it does not.

2. Writing only about newspapers

Section 4.2 covers the media as a whole. An answer with no broadcast regulation and no social media is answering a 1997 question in a 2026 exam. Newspaper circulation has fallen by roughly three-quarters since 1997.

3. Ignoring the "between elections" half of the requirement

The specification says "during and between general elections". Cash for questions in 1994, MPs' expenses in 2009 and partygate in 2021–22 all produced institutional consequences without an election taking place.

4. Treating all polling failures as identical

The 1992 miss is attributed to late swing and non-disclosure; the 2015 miss to unrepresentative samples; the 2017 miss to underestimating Labour's turnout. Naming the specific diagnosis is precise AO1 and lifts the analysis.

5. Confusing voting intention polls with exit polls

They ask different questions at different times and have different records. Any argument that "polls are unreliable" needs to acknowledge that the exit polls of 2019 and 2024 were close to the result.

6. Leaving the judgement implicit

The 30-mark command word is Evaluate, with 10 marks for AO3. State early which factor you think is decisive — for example that media influence is real but secondary to governing competency — then use each paragraph to test that claim, and restate it in the conclusion.