Do the results of the 2024 UK general election signal a decisive break from the constitutionally-aligned voting behaviour of the last decade? While we will only know the true extent of this shift when we analyse individual-level survey data, aggregate-level results strongly suggest this is the case. In this piece, our Research Associate Marta Miori investigates the indy-party connection at constituency level using new estimates of each area’s independence vote in 2014.
In the decade since the 2014 Scottish Independence referendum, election analysts have relied on constitutional preferences as key predictors of voting behaviour north of the border. In the 2019 UK General Election, over 70% of pro-independence voters chose a party that matched their stance on Scotland’s constitutional future. This alignment was even stronger in the devolved context of the 2021 Holyrood vote, with close to 9 out of 10 Yes supporters voting for either the SNP or the Greens.
Fast forward to 2024, and the relationship between support for independence and party choice seems to have significantly weakened. We’ll be able to shed light on the individual-level nuances of this dealignment using our Scottish Election Study campaign and post-election survey data at a later date. But in the meantime, here are three key takeaways about the constituency-level relationship between support for independence and voting in the 2024 UK General Election in Scotland.
These use new estimates of vote choice in the 2014 independence referendum at the UK Parliamentary Constituency level. Official indyref results have only been reported at the Local Authority level, hindering any analysis of constituency-level independence voting in subsequent general elections. The new estimates, made by transposing Local Authority results onto different geographies by breaking them down into smaller data-zones and weighting them by population and turnout, are the approximate equivalent of the referendum results having been counted at the constituency level instead.
For the first time, these allow us to consider the relationship between support for independence among the population living in each constituency in 2014 and vote choice in general elections in the same geographical areas. Between 2015 and 2019, we see a strong constituency-level alignment that is consistent with existing research on independence positions and party preferences. The following aggregate results for 2024 mark a decisive break from this narrative.
Takeaway 1. At constituency level, indyref and party vote are no longer correlated
This first graph measures the association between the estimated percentage of 2014 Yes vote in a constituency and the combined vote share for pro-independence parties in each UK General Election since 2010.
Even before the referendum, constituencies with a higher percentage of eventual Yes supporters were slightly more likely to vote for the SNP and Greens. With a sudden jump to a near-linear correlation coefficient of 0.78, independence preferences and support for pro-independence parties became closely aligned in 2015. Whilst weakening slightly, this relationship held-up over the subsequent two elections.
In 2024, this constituency-level association has ended: the relationship between referendum vote and support for pro-independence parties has become so weak that it is no longer constitutes a meaningful correlation. Going further, the weak relationship that does exist is negative: although only marginally, places with a higher share of Yes vote in 2014 are now less likely to support pro-independence parties than No areas.
Takeaway 2. Yes voters are now making a bigger constitutional compromise
We can learn more about these changes by plotting the estimated percentage of 2014 Yes support and the combined pro-independence parties’ vote share for individual constituencies.
Throughout the period of constitutional alignment, the relationship between support for independence and party choice was stronger on the Yes side. While almost all Yes voters supported a pro-independence party between 2015 and 2019, No voters were more likely to spread out across pro-indy and pro-union parties. This also holds at the constituency level: Yes areas voted for pro-indy parties, whereas seats were more scattered in their support for pro-union parties.
The scatterplot below illustrates this for 2019. Most seats cluster around the line of best-fit – which would approximately represent a 1:1 correspondence between the % of Yes vote and vote share for pro-indy parties in a constituency[1] – reflecting the previously mentioned alignment between 2014 referendum vote and party choice in the general election.
Only two seats significantly deviate into the upper-left quadrant – area a – where indy-supporting parties under-performed relative to the proportion of the constituency having voted Yes in 2014. These are Labour’s then-only seat, Edinburgh South, and the Liberal Democrats’ Caithness, Sutherland and Easter Ross.
In area b, on the other hand, we see pro-independence party systematically over-performing in predominantly No areas. The number of SNP seats in this lower-right quadrant is indicative of the party’s broader appeal of managerial competence drawing-in voters from both sides of the constitutional divide across these aligned elections.
The picture for 2024 is completely different. Relative to the percentage of 2014 Yes support in each seat, pro-independence parties have now under-performed in almost all constituencies. With pro-union parties winning a majority of seats across both Yes and No areas, it is now pro-independence voters that are making a bigger constitutional compromise.
Importantly, the large swings from pro- to anti-independence parties have not resulted in a constitutional alignment forming on the No side. Most pro-union seats may have a low vote share for pro-independence parties, but the constituencies with the highest combined support for the SNP and Greens are also predominantly pro-union. This explains why the correlation coefficient between independence and vote choice is now so low: between Yes areas voting for pro-union parties and No areas voting for both, there are no real predictive patterns at the aggregate level.
An exception to this is the fact that areas with a high estimated percentage of 2014 Yes supporters are not voting as strongly as they once did for pro-independence parties. This absence, coupled with the fact that a handful of No areas have a higher than average SNP and Green combined vote share, has resulted in the slight negative relationship seen in the correlation graph.
Takeaway 3. The SNP has lost more votes in Yes areas than No areas
This vacuum of support for pro-independence parties in areas with a higher 2014 Yes vote share is reflected by the SNP losing more votes between 2019 and 2024 in Yes areas than in No areas.
Although the party had larger majorities in those seats and therefore more voters to lose,
the direction of this relationship suggests that the SNP vote did not better withstand the national swings against it in areas with higher estimated Yes vote. Rather than being the last remaining strongholds of an overall declining SNP support, Yes areas were the most likely to swing away.
A working hypothesis to explain this could be that these were the voters the party has let down the most – paying the price of everyday governance but also of not having delivered independence. But beneath the surface, individual-level dynamics might be moving in the opposite direction: in these very seats, the strength of independence preferences might still predict why some Yes voters have stuck with the party.
Whilst we must wait for individual-level data to take these considerations further, these first constituency-level results suggest a sudden weakening of the overall relationship between support for independence and vote choice in Scotland. Knowing the percentage of a constituency estimated to have voted either in favour or against independence in 2014 is no longer informative of how that seat is going to vote. This is true on the Yes side, with the SNP losing votes at a faster rate in pro-independence areas, but also in predominantly pro-union seats which have swung either in favour or against pro-independence parties in a seemingly indiscriminate way.
An important caveat to these results is that they compare 2024 general election results to the referendum vote of the population that lived in each constituency ten years ago. Although aggregate support for independence has remained relatively stable, migration patterns and generational replacement make this unlikely to have been the case at the constituency level. Seen as the relationship between aggregate 2014 indyref results and general election vote shares appears to be consistent with existing individual-level literature until 2019, it’s improbable for the sudden dealignment between the two to be principally explained by underlying demographic change. However, until we estimate constituency-level present-day support for independence more accurately, this is something to keep in mind.
[1] Approximately, because turnout in the 2014 independence referendum was higher than in the 2019 GE