What the polls and models might be getting wrong right now

Election forecasts aren’t far out these days. Chances are, when you see some numbers about how Britain would vote in a vote held tomorrow, you can be pretty confident that on the headline level they will bear fruit. Ish.

Britain Predicts has form for calling elections right. Our (absolute) average by-election error has consistently been a few percentage points since launch. That’s good. I’m pleased with that. There have been some admittable howlers. And I’m not sitting on them. But compared to other forecast models, we are all saying pretty much the same thing.

(Although we are better.)

We had some local elections in May however in which a few things went wrong for us.

Labour out-performed our projection by almost 500 council seats. And the Greens under-performed it also by almost 500 seats.

The Labour totals should still be chalked up as a calamity for the party — the third worst losses (as a % of seats up) in recent history in local government terms. May 2026 was not 2001. This was not a typical mid-term parliament polling performance.

And the Green gains, under-performing though they did, were still stonking. A slew of London, Leeds and Manchester wards turning green in a way unseen since the formation of the Labour Party.

Everyone else, meanwhile, did as expected. Labour has a floor, low though it is — that’s the take-away of the 2026 local elections.

But where — how — did I go wrong?

I took a look at the model output for those May elections, took what ward-by-ward estimates we had (because for some new wards/divisions the census data needed wasn’t available), and cross-referenced them with the actual results.

Chances are they would be wildly off, as ward estimates tend to be. (80 per cent of the wards ended up with the main five parties being within 10pts of the forecast, which tells you everything you need to know.) But it did, however, tell me something quite eye-opening, which warrants this write-up.

If you’re just after the “what you need to know” headlines, take these. (And now might be a good time to subscribe, because this is valuable stuff.)

1. Reform is (or was) doing much better than expected in its strongest areas. But they’re doing worse in what we should call “Middle England.”

2. The death of the Tory party may be exaggerated, somewhat. They’re doing better where they need to.