There is an updated version of this estimate here.
If I were a certain type of journalist, I might be minded to title this piece something like “SNP to win 35 seats at Westminster” or “Labour to lose more than half its Scottish seats” or “Scottish Lib Dem Wipeout” or maybe even “Edinburgh zoo to triple is panda population”. But I’m not, so I’ll proceed with caution…
The tables for GB-wide polling include regional subsets, primarily to provide transparency with respect to the geographic weighting of the poll’s toplines. Given the particular interest in Scotland, it is tempting to try to extract Scottish voting intention from on the Caledonian crosstabs, particularly given the dearth of Scotland-only Westminster polling. There are two problems with this:
- Sample sizes are tiny (typically N~150, MoE ±8) which make them useless on their own.
- The demographic and political weighting is applied at GB-wide level, not to every individual subset.
The first of these problems is easily overcome by aggregating a number of polls. The second is more troubling because if respondants in Scotland were systematically to be biased in a way that was systematically different from those in England and Wales, the aggregated Scottish figures would also be biased. Electoral Calculus measured this bias against actual Scottish Westminster polls in early 2012 and found that it was “small”, although clearly the risk is that things may have changed since then.
As we can see, the SNP will hope that these figures are representative. They suggest the SNP vote share has surged more than 17 points, the bulk of which came in August and September 2014. Labour will hope the opposite, as their vote share appears to have dropped from 42% to 29%, lower than their share across Great Britain as a whole, and lower even than their mainland vote share in 2010.
Elsewhere, the Tories have dropped back since the summer, more-or-less to where they were at the last election, the Lib Dems and Greens are polling slightly lower than in England and Wales, while UKIP haven’t made the breakthrough that they’ve achieved south of the border.
If these figures are potentially dodgy, then making UNS projections based on them is even dodgier, but since people are going to do it anyway, I might as well save them a job:
These seat totals do look a bit extreme, and I wouldn’t get too carried away. Lib Dem incumbency could be one factor, as could tactical voting. But it’s hard to deny the direction of travel.