Ashcroft Scottish polling - a primer

On average the four Scottish polls so far this year show Labour on 27%, the Lib Dems on 5% and the SNP on 47%. This is the equivalent of a swing of 21 percent from Labour to the SNP and 20.5%

the Liberal Democrats to the SNP. On a uniform swing these polls would results in the SNP taking nearly all the seats in Scotland, leaving Labour and the Liberal Democrats with a rather pathetic rump of around about five seats (depending on the poll). People seem rather reticent about believing that though... it somehow seems too sweeping to be true. That said, this is first past the post, it's the sort of unbalanced result that First Past the Post can produce if you pump the right votes into it. However uniform swing really does becomes nonsensical when there is a sea-change of this magnitude in party support, a strict uniform swing calculation would be predicting many seats with Labour and the Lib Dems getting a minus share of the vote, for example, which clearly can't happen. Hence there been lots of anticipation for Lord Ashcroft's constituency polling in Scotland, waiting to see if it tells us anything about how the huge shift in Scottish opinion will translate into Parliamentary constituencies.

Below is a chart of all Labour's seats in Scotland, by their majorities over the SNP. The first thing that is readily apparent is how few traditional Lab -v- SNP marginals there are. Ochil and South Perthshire, with a Labour majority of 10% over the SNP, is the only seat that would be regarded as a realistic target under ordinary circumstances. These are not ordinary circumstances. The dotted line across the chart shows the Lab>SNP swing that has been suggested by the four Scottish opinion polls so far this year, as you can see, it's enough for the SNP to take the large majority of Labour's Scottish seats.

Below is a similar chart for the Liberal Democrat seats in Scotland. In this case the swing implied by the national Scottish polls would be enough for the SNP to win all but one of the Liberal Democrats' seats in Scotland. The surviving seat would be Orkney and Shetland, the party's safest seat and, incidentally, the only area that managed to retain Lib Dem constituency MSPs in their 2011 Holyrood rout.

We don't know what seats will be included in Lord Ashcroft's poll yet and we can but hope he goes up into some of the "safer" Labour and Lib Dem seats, and with a broad enough spread to give us some area where the SNP surge is happening. We should expect to see the SNP well ahead in polls in any of the constituencies polled that are near the start of the target list (if not, then something is either wrong with the Scottish national polls or with Lord Ashcroft's Scottish constituency polls - those SNP votes need to be coming from somewhere!), but it will be interesting to see how the SNP are doing in those seats with the largest Labour and Lib Dem majorities - will they actually be falling short of their implied national swing in those safest Lab and Lib seats, or will they be outperforming the implied national swing? Also look at the two-stage question, and whether Liberal Democrat incumbents are being insulated against the swing to the SNP at all. A third thing to watch is whether there is any obvious regional differences - in the referendum YES did better in Glasgow, West Dunbartonshire and North Lanarkshire. Will that be reflected in the shift to the SNP at all (certainly it includes some of Labour's most solid seats, so it would be bad for them if it did).

UPDATE: Lord Ashcroft's results accidentally escaped overnight, so I've had time to look at them properly. In the end Ashcroft didn't go for seats all over Scotland, but concentrated on Labour held seats in those areas that voting YES in the referendum, plus two high profile Lib Dem seats (both of which show the SNP ahead, returning Alex Salmond to Westminster in Gordon and ousting Danny Alexander). All the details are here.

Labour seats in Glasgow and North Lanarkshire tend to be some of their very safest, so Ashcroft's polling included some of the SNP's most difficult targets. Four of the seats polled were ones that Labour would have held on a uniform swing... but the Ashcroft polling shows them losing three of them. The average swing across the fourteen Lab-SNP seats polled was 25%, compared to the 21% swing implied by national Scottish polls.

There are three obvious potential reasons for that gap. One is just methodology - perhaps Ashcroft's method is just picking up a bigger swing per se. Unless we have constituency polls from every seat in Scotland to tot up and compare to the national polls we can't tell. The other two would be more interesting. It could be a sign of a "floor effect" - uniform national swing projections work on the assumption that parties gain or lose the same amount of vote in every seat in the country. This is very crude of course, but normally produces pretty good estimates of party seats. In a case like this when one or more parties have seen their support really collapse it runs into obvious practical problems - Labour and the Lib Dems have lost 15% or so support, yet there are seats where they didn't have 15% support to begin with, so they can't have lost that much support there. They must therefore have lost more support in seats where they began with more in order to make up the deficit. If things work this way then Labour should expect to perform even worse than uniform swing would suggest... just like these polls show.

The other possibility is that Labour are doing worse in those areas where YES did very well in the Scottish referendum. This makes a lot of sense in theory, but as Lord Ashcroft only polled in YES areas we don't have the data necessary to make the comparison. We'll have to wait for some Scottish constituency polling in places like Edinburgh to see if Labour are performing any less badly there.

The bottom line of the polling is that Labour are doing awfully in Scotland and face losing the large majority of their seats. We already knew from the national Scottish polls that the SNP were performing extremely strongly, so these figures shouldn't come as a surprise... but there's a certain reluctance to believe that astounding things will actually happen in real life. "Sure, the polls all show the SNP with enough support to win 50 seats, but they won't actually win that many", "Sure, the SNP are twenty points ahead, but they can't win in ultra safe Labour places like that" is the sort of train of thought. Well, if things keep going as they are, they might. Strange things do happen, especially with the vagaries of First Past the Post (for an example that looks mind boggling in hindsight, look at London in 1968). Landslide election results can result in extremely skewed seat distributions under FPTP, we may be about to see one in Scotland. Labour have three months to pull it back.