Scottish Parliament 2026 — regional list vote correlations

This is an analysis of how party support correlates across the regional list ballot in the 2026 Scottish Parliament election, computed from per-constituency results in all eight regional list contests. The intention was to find out whether or not the IGV vote was just "stolen" from Scottish Greens.

I'm not a data scientist or a statistician, so there may be mistakes, but none of this is very complex, so you would hope not!

Methodology

For each of the 73 constituencies (actually 61 - we are missing data from East Lothian and all of Central Region) that contribute to a regional list, the votes cast for every party on the regional list were combined into a constituency × party matrix. Pearson correlations were then computed across constituencies — one correlation value for each pair of parties — both on the raw vote counts and on each constituency's vote shares (each constituency normalised to fractions of valid ballots cast). The share-based correlations remove the effect of differing constituency size and turnout, so they reflect the shape of party support rather than its absolute scale.

The resulting Pearson correlations are then discussed below to understand the implications.

Heatmaps

These heatmaps are the core outputs from which most of the other conclusions are drawn

Constituency-level vote-share correlation across the six main parties plus IGV:

The same view extended with Scottish Family Party and Alliance to Liberate Scotland:

Results

The assumption that started the analysis was that IGV votes were obtained by a small fraction of SGP voters putting their mark in the wrong box.

Hypotheses

The summary below outlines the hypotheses in more detail.

Every vote for IGV was created by a small fraction of all voters mistakenly selecting the wrong "green" party.

The signature of this would be that the IGV vote would be a multiple of the SGP vote, and so there would be a very strong correlation (r close to 1) between vote shares.

However, this does not match the data well at all. The value of r is only 0.34, which is actually less than the correlation with Scottish Labour.

Thus this is shown to be false. The correlations suggest that IGV gets votes in similar places to the SGP, but no more than other urban based parties.

IGV vote share is wildly higher than is consistent with a party that did not campaign

The signature of this would be that IGV did better than other tiny parties with no significant basis.

The parties that ran in every region (other than the big six) got the following total votes.

Of these, Scottish Family Party and Scottish Socialist Party campaigned, which presumably helped them, but also have baggage that may have hurt them. I had never even noticed Alliance to Liberate Scotland before seeing the final election results, and had only heard of IGV because of the concerns about mistaken voters. It would appear that the results are consistent with the hypothesis that "if you just create a small party with no baggage, give it a catchy name, and get it alphabetically in the top part of the list, you get somewhere in the 0.5% to 1% of the votes range without doing any campaigning".

I think this is therefore also false.

IGV did not get a significant number of votes from SGP mistaken votes

We absolutely know that some green voters ticked the wrong box (from anecdotal evidence), and the data seems to show that we'd expect IGV to get a number of votes roughly in line with what they got.However, there is no evidence in this data either way on whether a measurable fraction of IGV votes came from mistaken votes.

This is definitely unknown.

Commments

There are a couple of important compounding factors here.

1. It is likely that in areas where SGP campaigned hard the proportion of voters who picked a party by mistake is lower. That would wash out some of the signal from correlations (if there is one).

2. We are trying to find the "mistaken vote" count by subtracting an unknown baseline from the total, where the baseline is comparable in size to the total. That requires some assumptions and means we are only going to get rough numbers at best.

The correlations between IGV and Alliance to Liberate Scotland are really interesting, and do suggest that the IGV vote total might be made up of a random vote background (about equal to ALS) plus some number of mistaken votes. Running a linear regression should show up how well that fits, but it seems likely to find a small number for the mistaken votes (as a percentage of total IGV votes), and will have large uncertainties.

I'll come back to that when I have time.

Conclusion

The data is not consistent with the argument that the IGV vote is made up entirely of mistaken SGP voters. It seems to show a model where some voters pick a party they like the sound of, and IGV and other small parties can get some baseline level of votes just by being on the ballot with a plausible name, ideally near the top.

The data is consistent with the some level of Scottish Greens to IGV vote leakage, but cannot quantify it beyond that it is a small proportion of the total IGV vote share. It's also not clear how much of that leakage is of the form "I wanted to vote Scottish Greens and ticked the wrong box" versus "I like the sound of parties with the word green in them, and ticked one at random" (which would have the same effect on votes but have different implications for the electoral commission).

Files available