A new way of thinking about Holyrood projections
Behold, my latest creation of inspired madness...
What have I done here? Essentially, three things:
A) Take Holyrood regions/seats, and clusters of their coterminous Westminster constituencies - not just arbitrarily by Holyrood region, but where voting patterns have shown some degree of similarity (e.g. taking Dundee out of the NE and adding Moray to it).
B) Compare the average SNP vote and margin across them, and how they relate to the SNP’s national performance in that given year, using what is in essence something known as the PVI (partisan voting index), pioneered by Cook Political Report in the USA as a means of analysing how republican or democrat-leaning a particular district or state is relative to the nation as a whole. Of course this doesn’t translate quite as neatly to the multi-party context of Scottish politics, but stripped down to a measure of just SNP strength alone it still tells us quite a lot.
C) Identify trends by seeing which regions shifted away from, and towards the SNP - in both absolute terms and relative to the national environment.
So, how does this relate to the current models that are used to project seat results at the upcoming election? Well, those work off Universal Swing, which assumes that there are absolutely no regional particularities. Of course we know that in reality the SNP is going to outperform its modelled fundamentals in certain areas and underperform them in others. While I don’t have enough information to accurately gauge how off the projected SNP voteshare may be in Banffshire or Edinburgh Central, these shifts ought to give one a good idea of where the party can expect to underperform and overperform expectations.