How YouGov accurately called the 2026 local elections

Our projections were the most accurate of competing pollsters

In the lead up to the 2026 local elections, YouGov released projected MRP vote intention estimates for 45 out of the 131 lower-tier authorities holding ballots on 7 May, covering 32 London boroughs and 13 council areas in the West Midlands.

Two other research companies published projections this year. JL Partners produced estimates for every council being contested, while More in Common released estimates for all London boroughs, plus Birmingham.

YouGov’s accuracy

Undertaking this modelling is not straightforward. The fieldwork was carried out up to six weeks before polling day and there are significant challenges in capturing and projecting vote intention in second order, low-turnout elections with drastically different patterns of party (and non-party) contest and prior strength from area.

However, YouGov enjoyed strong results. We correctly projected the largest party (in terms of vote share) in 35 out of the 45 councils we modelled.

There is little point in comparing this rate across other pollsters given the large differences in number of councils included in the various projections. It also doesn’t make a make much sense to compare pollster models to other forms of models. This is because the former informs the latter and so the accuracy of latter depends upon the accuracy of the former.

However, we can look at party-level vote share accuracy across the three pollsters as a more even-handed metric and assess YouGov’s position within our own league and we can see how well we fared.

How we compare

Across all five main parties and all council areas projected by each pollster, YouGov’s vote share error was by far the lowest at an average of 1.21 pts. This was followed by More in Common at 1.93pts, and then JLP at 2.01pts.

Our party share accuracy across all 45 councils projected was strong. For example, in only two councils did the eventual Reform UK vote share fall outside our range of projected intervals for the party. In 40 councils, its vote share fell within just five points of our central estimate.

For the Greens these figures were zero and 40 respectively, Labour four and 37, and for the Conservatives nine and 35.

We had greater levels of error in terms of projecting Liberal Democrat shares, particularly in London. We overestimated the party in 12 councils but still got within five points of their vote share in 35. We believe the larger issue for our Lib Dem projections was a slope problem in the models, which did not quite have enough information to properly capture the geographic distribution of the party’s vote share.

As ever, these elections provide us with a great opportunity to test our models and learn against real world events. Overall, we are pleased with how our models performed at the 2026 local elections, especially as the projections were made two weeks (London) and one week (West Midlands) before polling day.

Combined with the outstanding success in our Wales Senedd election and very impressive results in the Scottish Holyrood election showcases the quality of YouGov’s panel, data and expertise and our continued commitment to methodological excellence, public data, and getting it right.