A little more than two weeks out from the General Election, the Labour Party has increased its list of Scottish target seats to 36, in a sign of growing confidence about its prospects north of the border. While Scotland polling data has indicated the wind blowing in Labour’s favour for some time now, a new MRP model from Ipsos highlights the degree of uncertainty that remains in Scotland – and how far relatively small shifts in votes could yet change the outcome here.

Estimating the number of seats from opinion polls has become something of a sport for political geeks, even though most conventional polls are not designed to estimate this. Rather, they are a snapshot of how the share of the vote might fall, were the election held today.

Of course, where those votes fall is crucial to how they translate into seats, particularly in a first past the post system, as used in General Elections in the UK. The Liberal Democrats are acutely aware of this – in 2019, they won 11.6% of the vote UK wide, but just 11 seats (under 2% of the total number).

Recent elections have seen more voting intention polls than ever, but also increasing use of a statistical technique – Multiple Regression and Post-Stratification, or MRP – to estimate the share of the vote in each constituency.

Simply put, MRP involves taking a very large poll of voting intentions (typically 10,000+) and using it to analyse how different types of people in different types of seat say they will vote.

For example, it estimates the probability that a woman, aged 25-34, educated to degree level, living in an SNP/Labour marginal seat, who voted SNP in 2019 will vote for each party running in that constituency. The model creates thousands of these estimates, and then applies them to the demographic profiles of each constituency (based on data from, for example, the Census) to estimate how the votes might fall out for each party.


Poll: SNP faces wipeout in Glasgow and losses to Labour

Struggling public services are influencing voters in Scotland

Ipsos has this week released its new MRP model for the UK, uniquely based on data from its online random probability KnowledgePanel.

At the overall UK level, the picture is bleak for the Conservatives, who are estimated to be on course for just 115 seats. Labour, meanwhile, would take as many as 453 seats – even higher than its tally of 418 in 1997.

However, MRP models, as with any poll, come with a degree of uncertainty attached – 117 seats have a ‘winning margin’ in the estimates of less than 5 percentage points, and are therefore considered too close to call. The Conservatives are second in 50 of these and Labour in 43, so the overall balance between the Conservatives and Labour could be narrower – or even wider – than the headline figures indicate.

This uncertainty is even more apparent in the Scotland data. While the SNP are only a little behind Labour on overall vote share, the distribution of votes across constituencies in our MRP model indicates that this may translate into 34 seats for Labour, 15 seats for the SNP, 3 for the Conservatives and 5 for the Liberal Democrats. The implied vote share from our KnowledgePanel data puts the Labour on 36%, the SNP on 33%, the Conservatives on 13%, the Liberal Democrats on 8%, Reform UK on 5%, and Greens on 3%.

On these figures, the SNP could be set to lose 29 of their notional 2019 seats to Labour and 3 to the Liberal Democrats. And in Aberdeenshire North and Moray East – the subject of much controversy after Douglas Ross announced he would stand, the Scottish Conservatives having taken the decision that David Duguid, the sitting MP, was too ill to stand (against Mr Duguid’s own assessment) – the Conservatives looked set to lose out to the SNP.

Just hours after our MRP model was published, the electoral landscape in that seat has changed yet again, with the withdrawal of Labour party support for its candidate. This may or may not benefit Mr Ross, depending on whether voters in the seat who had been planning on supporting Labour switch to the Conservatives or to the SNP.

However, 12 of the 57 seats in Scotland are too close to call in our MRP model. This includes three SNP/Conservative battlegrounds (Gorden and Buchan, Berwickshire, Roxburgh and Selkirk, and Clydesdale and Tweedale) – seats to watch as the results come in overnight on the 4th.

A further 14 seats ‘lean’ towards one party or another but the gap is still relatively modest - between 5 and 10 percentage points. Add to this the fact that our Scotland telephone poll last week showed that as many as 42% of voters may change their minds ahead of election day, and the picture could yet look quite different come July 5th.

So while overall the story from our MRP – as in the polls overall – is positive for Labour, worrying for the SNP, and pretty bleak for the Conservatives, it also shows that the battle for Scottish votes is far from over.

Rachel Ormston and Chris Martin are Research Directors at Ipsos Scotland