The Numbers Game

How the Commentariat Tries to Rig the Scoreboard Before the Votes Are Counted

There is a ritual as old as democracy itself, and it has nothing to do with voting. It takes place in the days before polling, in the offices of think tanks, the studios of broadcasters, and the columns of political magazines. It is the ancient art of expectation management — the careful calibration of what counts as success and failure, conducted not in the interests of accuracy but of narrative. This week, with the May 7th elections bearing down upon us, we have been treated to a masterclass of the genre.

Peter Kellner, former president of YouGov and a man whose estimable intelligence I have no interest in disputing, has published a guide to the upcoming elections in Prospect. It is admirably readable and contains much of interest. But embedded within it is a paragraph about Reform that repays close attention, because it illustrates with almost pedagogical clarity how the expectation game is played.

Kellner deploys the Rallings and Thrasher model to suggest that if Reform win 1,400 seats, they will be “sunk in gloom,” and that anything short of 2,000 should indicate that they are “slipping back.” He frames sub-2,000 as the threshold of adequacy. The implication is clear: a party that currently holds two councillors among the seats being contested should apparently consider 1,400 gains a cause for institutional mourning.

Only?

Let me be direct: I would be happy with 1,000 seats. I would be delighted with anything north of 1,200. And I say this not from false modesty but from an honest reading of the data, weeks of campaigning on the ground, the political landscape, my own politically pessimistic nature, and, perhaps most importantly, from a sceptical eye on the baseline figure Kellner has chosen to make his arithmetic work.

Because here is where the conjuring trick is performed. Kellner anchors his analysis with this claim: “Last year this was more than 30 per cent according to separate estimates for the BBC and Sunday Times.” Those are indeed the figures that certain post-election projections produced. But to treat them as the polling consensus in the run-up to last year’s elections, as the benchmark against which 2026 performance should be measured, is to quietly launder an outlier into an orthodoxy.

The average voting intention polling for Reform in the weeks before the May 2025 elections was running at approximately 24 per cent. The 30 per cent figures were projected vote share estimates produced retrospectively after the votes were counted, using Rallings and Thrasher’s own methodology, not the pre-election polling average. The Rallings and Thrasher post-election estimate put Reform’s projected national vote share at 32 per cent, higher than UKIP’s 23 per cent at the 2013 local elections — making it, by their calculation, the first local election in which neither Labour nor the Conservatives received the highest vote share. That is a significant and praiseworthy achievement. But it is not what the polls predicted going in.

To treat a post-hoc projection as if it were the pre-election polling baseline, and then to set the bar for 2026 “success” at a level that assumes that figure will be replicated or exceeded, is, however inadvertently, to construct a trap. You inflate the starting point, then measure any deviation from it as decline. It is a statistical version of moving the goalposts, except you move them before the game starts and then pretend they were always there.

Other serious analysts take a markedly different view. Stephen Fisher, the psephologist behind the BBC’s general election exit poll, projects Reform to make net gains of around 2,260 seats, but he explicitly caveats that this assumes Reform translates their polling rise into council seats at the same extraordinary rate they managed last year, which was itself a considerable outlier against historical precedent. PollCheck whose ward-level modelling is among the most granular available, places its own projection lower, describing itself as “conservative, particularly in metropolitan boroughs and London where no 2025 calibration data exists.”

Meanwhile British Brief, synthesising a range of projections, suggests Reform gains of over 1,300 seats from a base of just three, more than doubling what would already be the largest single-election gain by any party from near-zero in modern English local government history. Luke Tryl of More in Common, a widely cited figure in this conversation, has been making observations about where Reform’s gains are likely to concentrate, noting that London isn’t Reform’s strongest area but that the party could do well in Havering, Barking and Dagenham, and in parts of south east London, without suggesting that anything short of 2,000 represents failure.

In other words, the range of serious projections runs from roughly 1,300 to 2,260, with a great deal of uncertainty across the piece. That is not a landscape from which one can responsibly extract a “floor” of 1,400 and label it despairing, or 1,700 and label it disappointing.

I do my own projections, and I am under no illusion that they are infallible. Last year I guessed Reform would win two to four councils. We won ten. I was wrong, as was nearly everyone else. That was a genuine surprise to the whole commentariat, not a vindication of some rival model. The lesson I took from it was not that higher expectations are always warranted, but that this terrain is genuinely difficult to predict, which makes confident assertions about what counts as adequate all the more suspect.

The structural bias in pre-election commentary cuts in a predictable direction. For the legacy parties, expectations are lowered so that the result, however catastrophic, can be narrated as within acceptable parameters. For insurgent parties, expectations are inflated so that even a historic performance can be repackaged as a stumble. It flatters editors and programme-makers, because it manufactures a story, “Reform fails to meet expectations”, from a result that would, in any objective historical frame, represent an extraordinary political achievement.

The latest Electoral Calculus MRP puts Reform on 24 per cent nationally — down from 31 per cent in January, which is itself a reminder that the party is navigating a more contested political environment than a year ago, with competition emerging on its right flank and a Conservative party beginning to claw back some of its lost ground. Anyone building an expectation model on peak-Reform polling rather than current-Reform polling is building on sand.

What would genuinely constitute a disappointing night for Reform? That is a more interesting question than the one Kellner poses. A result under a thousand seats would give pause for reflection. Failure to take control of any councils, despite fielding candidates in the overwhelming majority of wards, would be a structural concern. An inability to breach traditional Labour strongholds in the north, after all the ground-level investment made since last year, would be a genuine story.

But 1,400 seats? From two to fourteen hundred? That would be, in the unadorned language of democratic arithmetic, an astonishing result, achieved in a single election cycle, from a standing start, by a party that did not exist in its current form half a decade ago.

We will know the truth of it in a week. The votes will be counted, the results declared, and the commentariat will assemble its retrospective framework. It is worth recording, before that happens, what the expectations actually were, and who was quietly inflating them, and to what end.