Each fresh poll comes with its projections about how the electoral map will be coloured in and the seats divided between the parties. Those projections inevitably involve assumptions about how the swings indicated by the national poll will apply in the 73 constituency seats.
A simple and common assumption is uniform swing: that the swings will be the same everywhere. A more sophisticated approach (e.g. that of Ballot Box Scotland) assumes that the regional distribution of a party’s vote – where it is strong and where it is weak – will stay roughly the same as its overall vote goes up or down.
But neither of these really tracks regional differences in swings for the simple reason that an individual poll contains too few voters from each region to draw reliable conclusions about the way the winds are blowing. To do that, we need to pool several polls. Given that there are eight Scottish electoral regions, combining the eight most recent polls means that the pooled sample in each region (with the exception of the less populous Highlands & Islands) is of roughly the size of a single national poll – with those familiar ±3-point margins of error.
That’s what’s done for the table below. It shows that, not surprisingly for a politically diverse country, the changes in vote shares and the swings to and from the SNP since the last Holyrood election (in May 2016) vary quite a lot. Brexit having occupied much political attention since, the Remain/Leave balance in a region is a likely reason for some of these differences: the SNP gains and Conservative losses are largest in EU-loving Lothian while there is a swing to the Conservatives in the (relatively) Eurosceptic Highlands & Islands.
What does this do to seat projections? Not that much, given that the SNP remains warm favourite for the large majority of seats. But, in a context where every marginal matters for that potential SNP majority, eyes should be on seats like Ayr and Edinburgh Central. Both are on a knife-edge based on uniform swing but, if we base the prediction on regional swing instead, the Conservatives hold off the SNP in Ayr but Angus Robertson takes Edinburgh Central with something to spare.
Of course, winning seats in Edinburgh requires the SNP to hold off some concerted Unionist tactical voting. Even regional swing cannot take account of all local circumstances. But, without enough polls to give us a sample large enough to predict seat-by-seat, it’s all we’ve got – and definitely an improvement on uniform swing.