Our MRP was the most accurate in terms of constituency calls and had the smallest error for party shares at the regional level

The 2026 Scottish Parliament election was set against the backdrop of significant change in the popularity of parties from 2021 to 2026, but with an expected outcome not altogether different from the previous contest – particularly regarding the Scottish National Party and their quest to secure a fifth term in government and only their second ever Holyrood majority.

In the end, the SNP did again take the most seats, albeit on a significantly lower portion of the vote than the last election. Reform UK rose substantially, from zero representation in Holyrood to being the joint-second largest party with Labour on 17 seats. The Greens also made significant gains, cementing a pro-independence majority in Holyrood (if not an SNP majority) for the fourth successive election, while Labour and the Conservatives both lost seats.

Our MRP compared to the actual result

YouGov’s MRP was by various measures the most accurate in the 2026 Scottish Parliament election. All of our calls for the election – seats and shares on both constituency and regional lists – across the parties were comfortably within our upper and lower bounds, and we clearly predicted the scale of the SNP victory, Labour’s struggles, and the surge for both Reform UK and the Green party.

We were most accurate in terms of individual constituency calls, getting the right winner in 63 out of Holyrood’s 73 seats. In all, most final MRP models produced for the 2026 Holyrood election managed to correctly project more than 60 constituency results, which represents strong overall performance by the polling industry.

Our MRP was also the most accurate in terms of minimising error for party shares at the constituency level, with an average absolute mean error of just 2.4 across all parties in all constituencies. YouGov’s next closest competitor on this scored 3.4 on this measure.

In terms of national seat totals, our MRP was the second most accurate. Our biggest errors were a four-seat overestimate on the SNP total, and a five-seat underestimation on the Conservative total.

On national vote shares, our model was (again) second most accurate in terms of estimating the constituency vote Scotland-wide, and the most accurate on estimating the regional list vote across the country. Our root mean squared error on nationwide list vote intention was just 1.25, narrowly edging out our next-closest competitor on 1.31.

Our root mean squared error on constituency vote intention across Scotland was just 1.20, with our largest errors being a two-point underestimate of the Conservative share and a two-point overestimate of the Reform share.

Overall, we are very pleased with the results of our final 2026 Holyrood election MRP projection. As well as telling all the correct stories – including picking up some hyper-specific constituency wins for the Greens and Liberal Democrats – the model proved to be the most accurate across a range of different metrics on different outcomes and the second most for others.

Our Holyrood MRP model

In the context of a rapidly evolving political environment, YouGov created a complex MRP model along the lines that accurately forecasted elections across the world in recent years – including those in Spain (2023), the UK (2024) and Australia (2025).

MRP models work by first establishing the relationship between a broad range of characteristics about prospective voters and their views – in this case, which party they intend to vote in the Scottish Parliament election – within a multilevel model. It then draws on constituency and regional-level data to forecast seat outcomes based on the distribution of different voter types in each area, informed by what the multilevel model indicates about the likelihood of those voters supporting various parties (post-stratification).

We also built strong and detailed models of the likely Scottish electorate for 2026, considering how this might have changed since 2021 and using information from 2024 elections and the Hamilton, Larkhall and Stonehouse by-election in 2025 to properly calibrate expectations on who we thought would turn out for this Holyrood election (and, crucially, who they would be voting for).

Why our results are so accurate

The political landscape has changed considerably since the last election in 2021 following two changes of First Minister, the change of government at the 2024 UK election and the continued rise of both Reform UK and the Green party. All of this, alongside the inherent challenges of polling and forecasting public opinion across a complex electoral system involving both constituency and regional votes particularly demanding. This made the performance of YouGov’s MRP all the more satisfying.

Three factors drive our ongoing record of accuracy.

1. YouGov's world-leading panel provides quality data

Because we own our own panel, we have a distinctive relationship with the people who take our surveys and immediate access to the views and opinions of millions of people around the world. This means YouGov has a deep understanding of who these people are, enabling us to build and execute models at scale that are genuinely market-leading and that help us keep our position as the most accurate research company.

In Scotland, we used our panel to carry out a stratified sampling approach, ensuring we produced high-quality, reliable samples that fed into our MRP model. It also meant we had confidence in our past voting behaviour data, as we capture and store it as close to an election as possible, reducing the risk of false recall bias.

2. YouGov in-house experts

The second is the quality of our expertise across our Political and Social Analytics, Political Research, Data Journalism, and Panel teams, to name but a few. YouGov's experts in political polling, psephology, and advanced analytics worked hand-in-hand with experienced colleagues from our panel and data teams to deliver our accurate results in the 2026 Scottish parliament election.

We are also incredibly grateful to Professor Chris Hanretty, Professor at Holyrood, for producing and publicly releasing notional Scottish 2021 results for new constituencies, which were a significant help in the development of our modelling.

3. YouGov's commitment to innovation

Our pioneering MRP approach relies not only on our gold-standard data and in-house expertise but also on our ongoing use of advanced technology and computing solutions that allow us to make the most of these strengths.

YouGov has been using MRP modelling for nearly a decade, since its landmark application at the 2017 UK general election. In the years since, we have built both a strong track record and – importantly – a methodology grounded in a combination of the latest technology and extensive experience. Around the world, elections and electorates have changed notably over the past decade, and we have refined and improved our MRP models to keep pace – an essential part of our continued ability to get it right.

Subscribe to the YouGov newsletter