Basics and Caveats
Some of the work I do here on Ballot Box Scotland relates to what are known as “projections”. Projection involves taking a national poll of voting intention and seeking to translate it to a sub-national level. For obvious reasons, people are usually keen to have an indication of what a given poll suggests for their area and, in the case of parliamentary polling, how many seats each party would win.
The two most common forms of simple projection are “Uniform” and “Proportional” swing models. In Uniform Swing, the change in each party’s vote since the previous election is simply replicated nationwide. So, a party that is up 5% the polls has 5% added to its projected share in every constituency.
In Proportional Swing, the relative size of the change in each party’s vote is used instead. So, if the party up 5% had won 20% at the last election, its projected share is increased by a quarter in every constituency. For example, if they had won 12% somewhere else, that would give a projected share of 15%.
More complex forms of projection may use sophisticated demographic modelling or run a range of simulations to give a % likelihood of a party winning a seat. A widely discussed form of projection in recent elections has been “Multilevel Regression and Post-stratification (MRP)” polling, though the record of MRPs have not been any better than traditional swing based models, and often worse.
Regardless of your model the reality is that isn’t how elections actually work. Local issues, personalities, and campaigns have a substantial impact, as do changing demographics. Parties can also stay steady in national vote share, as the Lib Dems did in Scotland in 2011, 2016 and 2021, but have significant changes in individual constituencies. In short, it simply isn’t possible to take a national poll and use it to predict with perfect accuracy the result in each constituency. Projections are useful indications and a lot of fun, but not reliably more than that.
This should be quite obvious, and the Ballot Box Scotland ethos is to try and improve understanding of elections whilst also assuming people know enough to start with that they don’t need to be beaten over the head with caveats and health warnings every time I post something. Good plain fun is also part of the ethos.
Not doing something lots of people enjoy and find informative because some people might occasionally misinterpret or misunderstand it isn’t in keeping with that ethos. And that’s where this page comes in – a quick and easy explanation of some of the issues with projection, just in case people do need it.
Scottish Parliament
Holyrood projections are one of my bread and butter bits of content, and amongst the most popular. With the limitations of both Uniform and Proportional swing models in mind, the BBS approach is to effectively average out between the two approaches. For the Proportional swing element, a projection of vote share in each Region is estimated first, and then Constituency level results are projected from that.
Again, this is absolutely not a perfect approach. A party doing very well in a Region last time might suggest they’ve hit saturation point there and they won’t grow much further even with a big positive national swing, or that even with a small negative national swing they’ve got a lot further to fall there than in regions they might already be rock bottom.
To further emphasise the fuzzy nature of projections, for the Constituencies I also provide a map showing the runner-up in each, and one showing what band the winning margin falls into. Effectively, you can assume that a constituency in the 20%+ projected margin band is highly likely to go the way the main map indicates, whereas one in the 0-5% band could still very easily actually be won by the projected runner-up.
Green Constituency Shares
Due to the relatively limited presence of Green candidates in previous Holyrood elections, my model effectively ignores their polled constituency vote. Green constituency figures are instead driven by a combination of their regional vote and a very rough-and-ready estimate of how their constituency vote in a given seat would compare if they had stood. This approach was highly accurate in 2026, projecting within 3% of their final result for all bar one constituency, with the largest miss being coming up 5% below what they actually received in Glasgow Southside.
Further, I’ve set a toggle for the Greens in each constituency. By default, I’ve chosen to toggle on the 6 constituencies the Greens contested in 2026, as well as Glasgow Central which they did not contest but did win the most Regional votes within. Whether the Greens are present (negative for other party shares) or not (positive) impacts the SNP the most, then Labour and the Lib Dems.
UK Parliament
As I explain in detail in this post, I don’t do projections for Westminster. The short version is as follows;
- Westminster is widely covered by others as it is
- Pure First Past the Post is difficult to reliably project
- First Past the Post projections dramatise small shifts in vote share yet are intensely boring due to non-impact of new and/or smaller parties
- I don’t have an in-house calculator for Westminster, and don’t want to develop one for the reasons above
Independence
Independence projections are now only done for annual reviews, though they were previously a normal part of polling coverage. These are simple UNS, as described above, and are even more “just for fun” than other projections. In reality, even with an identical national result to 2014, the distribution of votes between areas would have some difference.
In particular, following the Brexit Referendum, there has been some degree of alignment between Yes/Remain and No/Leave. Areas with higher than average 2016 Remain votes are therefore likely to be more supportive of Independence than in 2014 even on an equivalent national vote share. On the other hand, areas with higher than average 2016 Leave votes are likely to be more supportive of the Union than in 2014 in those circumstances.