海角社区

Professor Thiemo Fetzer has researched the relationship between political polling and election outcomes. He comments: 

鈥淭he Reform UK sting highlights the vulnerability of the British political system from outside money and influence. This is well known, well understood, but yet, the British system is struggling to meaningfully improve its defenses. Channel 4 News reported three polls that were effectively commissioned by Reform, costing 拢32,500, were ultimately paid for by a US company that Reform figures believed belonged to a wealthy American backer. This construct effectively circumvents the already quite weak protections that British democracy has from outside influence.鈥�

Eleonora Alabrese adds: 鈥淭here are three separate failures here. A party seeking money from abroad to pay for polling. A pollster accepting an undisclosed commission, which sits uneasily with the industry鈥檚 own transparency rules. And newspapers running the results as independent findings without checking, or telling readers, who was behind them. Our work on UK general elections shows that polls published in the final weeks of a campaign can affect turnout: a poll is not a neutral piece of news, it is something people act on. There is a double standard too: had the same construct been used to move money for other purposes, a criminal investigation would follow. Here the cost falls on trust in the media, already low, and unlikely to recover if front-page numbers turn out to have been quietly paid for by a party. The practical answer is not necessarily to restrict polling, but to require disclosure along the chain.鈥� 

These polls effectively were commissioned in a way that was informed by Reform鈥檚 political strategy. The polls subsequently generated major newspaper coverage without Reform鈥檚 involvement being disclosed. As usual, Reform denies wrongdoing; two senior officials have stepped down, and the Metropolitan Police is assessing the allegations.

The episode reveals something more fundamental about the information architecture of elections: opinion polls do not simply measure politics. They can change politics.

This is precisely what Eleonora Alabrese and I have been studying.

In our paper, , we compare polls published at similar points before UK general elections. Polls appearing in Conservative-leaning newspapers systematically report lower Conservative vote shares than polls appearing in other news sources—and the difference becomes larger as election day approaches. In the final week, the estimated Conservative vote share reported by Conservative-leaning newspapers is 1.47 percentage points lower. In the paper we write that this pattern could suggest that polling is being used for 鈥渢urnout engineering whereby aligned voters may be mobilized by a perceived worse expected performance in the run up to an election.鈥�

That is not proof of deliberate manipulation by newspapers. Selection of which polls newspapers commission and publish may explain the pattern. 

Polls affect turnout.

When national polls predict an uncompetitive election, turnout falls. The effect is stronger in constituencies already perceived as safe and becomes more pronounced closer to election day. In Britain鈥檚 first-past-the-post system, national information therefore interacts with local electoral geography to change participation and party performance.

The intuition is straightforward. If your team appears not to be doing well, that can become 鈥�a voter mobilization mechanism.鈥� Conversely, if victory looks assured, some supporters may conclude that the personal cost of voting is no longer worth paying.

Now consider what the Reform sting appears to show.

According to Channel 4 News, the first of the polls appeared on the Telegraph front page under the headline 鈥�Labour would face wipeout if elections went ahead.鈥� Reform鈥檚 role in commissioning it was not disclosed to readers. It is not unlikely that such media coverage may have had impact on the internal politics within the Labour party, as scrutiny of Reform UK鈥檚 affairs was working its way through government. The poll cost 拢18,000, with payment subsequently arranged through the supposed US-backed company used in the sting.

That matters because a poll is not merely a number. It is also a news-producing technology. It is potentially the conversion of political expenditure into apparently independent public information: cash-strapped media organisations have a strong appetite for clean, authoritative content: a poll offers a number, a headline and an immediate political narrative. 

This is much more valuable than buying advertisement. This is precisely how dark influence operations can be orchestrated. While an advertisement says: a political party would like you to believe this, an apparently independent opinion poll may suggest to the public: this is what reality looks like.

Such information can affect political participation and may produce electoral surprises.

There is a second reason to take this seriously. 

In , we study electoral surprises by comparing opinion-poll expectations immediately before elections with eventual results across 233 elections in 51 countries.

Those surprises have economic consequences. A one-percentage-point larger electoral surprise is associated with about 0.37 percentage points lower year-on-year growth one year later, with the effects concentrated in strong democracies and operating primarily through increased policy uncertainty and weaker investment.

That paper is emphatically not evidence that manipulating polls allows somebody to manipulate the business cycle. But put the two papers together and they make an important point.

Polls help construct political expectations and may affect intra-party politics. Political expectations affect participation. And surprises relative to those expectations affect economic behaviour. The provenance of those expectations therefore matters.

This is where the Reform episode exposes a genuine regulatory blind spot. We tend to imagine political influence as:

donor 鈫� political party 鈫� campaign expenditure 鈫� voter.

The legal system and political convention are built around this dynamic. The Reform UK episode highlights that information can travel by another route:

funder 鈫� pollster 鈫� media 鈫� voter.

The party does not necessarily need to receive cash for somebody else鈥檚 expenditure to generate political value. This is very much akin to how sanction evasion can be orchestrated with indirect benefits occurring to third parties that appear unconnected. 

What democracy therefore needs is not simply more disclosure of party donations. It needs information provenance.

Who initiated a poll? Who selected the questions? Who paid the invoice? Was a political party involved? Who had the results before publication? And what did the newspaper know about the commissioning chain?

The point of such an audit layer is not to decide which narratives citizens are permitted to hear. Quite the opposite - the key thing for democracies to be able to protect themselves is through the ability of contesting a narrative. And you can only contest information properly if you can establish where it came from.

This should be the lasting lesson from the Reform sting.

The democratic question is no longer simply whether an opinion poll is statistically well executed. Once polls themselves can affect political outcomes, and participation metrics such as turnout, political viability and expectations, we also need to know why that particular poll came into existence in the first place.

Citizens need to know not only what a poll says.

They need to know who caused the poll to exist.

4 September 2026