One of the central questions for the next general election is how those voters who voted Lib Dem in 2010 will vote next time round, and how this will differ between different categories of voter. This is a big subject, too big to be dealt with in a single post. So I propose dividing it as follows:
2) Lib Dem held seats
3) Conservative/Labour marginals
Let's start with some basics. The Lib Dems have undeniably lost support since 2010. At the last general election, they polled 23% of the national vote share. In the most recent ICM poll (by no means the least favourable to them), they polled 12%. They have lost about half their voters in four years, if the polls are to be believed, and possibly more depending on which pollster you choose to believe.
The scale of this drop raises some important technical questions. Normally when we assess the likely impact on seat numbers, we look at uniform national swing. Past experience has shown that usually is the best rough and ready guide. Is it a reliable tool here? In short, my answer is no, but it's probably still the best starting point in England at least, and if we choose to deviate from it we should need some kind of reason why. I'm very doubtful whether it's any use at all in Scotland.
I want to start looking at the Lib Dem seats in Scotland because they give an extreme illustration of some of the problems we face when trying to work out what is happening in the electorally more important English marginals.
Why is Scotland different? Simply, the Lib Dems have performed very differently electorally in Scotland from south of the border and are now apparently suffering even greater woes in the polls.
In 2010, the Lib Dems took 11 seats in Scotland. Here they are:
They did this on 18.9% of the vote, so they got exactly their proportionate share of the seats. If they had done equally well across the UK in converting votes into seats, the Lib Dems would have had 150 MPs rather than the 57 they actually tallied. So the Scottish Lib Dems far outperformed their southern colleagues on a lower share of the vote.
It should also be noted that, unlike the rest of Britain, the Scottish Lib Dems saw their vote share decline by 3.7% (a sixth of their 2005 votes), but kept all the seats that they had won in 2005. So they already have practice in husbanding their seats against a background of electoral decline.
If the polls are to be believed, however, that was just the hors d'oeuvres for what is coming next year. Few Scottish polls have been taken on Westminster voting intention - for some reason the pollsters are more interested in the independence referendum. So far as I can see, only Survation have conducted surveys on the subject this year. In their two most recent polls this month, the Lib Dems registered 6% both times. Depressingly for the Lib Dems, this is not that particularly out of line with the polling of the last three years:
If we look at Lib Dem prospects on the basis of uniform national swing, then using that last Survation opinion poll there has been a 14.5% swing from the Lib Dems to the SNP, a 3.5% swing from the Lib Dems to Labour and a swing of just over 3% from the Lib Dems to the Conservatives. Uniform national swing would leave the Lib Dems with four seats: Berwickshire Roxburgh & Selkirk, North East Fife, Ross Skye & Lochaber and Orkney & Shetland (some of the seats would be lost to the SNP from third or fourth place).
The problem with this approach is that the maths simply doesn't work for uniform national swing in seats where the Lib Dems polled 13% or less in 2010 if the Lib Dems indeed only secure a national vote share of 6%. In 2010 the Lib Dems polled 13% or less in 33 out of the 59 Scottish constituencies. That is far too many to overlook as a quirk. The decline in vote share is simply too steep for uniform national swing, which is a model designed for smaller swings. Even if the maths works, there's the question of plausibility: is it credible that the Lib Dems will poll as low as 1 or 2% in numerous constituencies? I can't see it. So I would not regard uniform national swing as very helpful unless the Lib Dem vote share in Scotland looks set to exceed 10% by a bit. Right now, that really doesn't look very likely.
What other methods are there? We could work on the basis of proportionate swing - votes are gained or lost in each seat in proportion to the national vote share. But this takes no account of the concept of regional variation, the popularity of a local MP or differential campaigning effort, the last two of which in particular are relevant to the Lib Dems. So I'm not attracted to that.
Alternatively, can the Lib Dems repeat the trick they pulled off in 2010 of defying uniform national swing completely? Not at 6% they can't. In the 11 seats that they won, they tallied 178,534 votes. On the same turnout, the Lib Dems would tally only 148,000 votes on a 6% vote share across the whole of Scotland. Since negative votes have not yet been put into operation in general elections, something has to give. If the Lib Dems poll 6% across Scotland, a lot will give.
We need to start from first principles. Neither uniform national swing nor proportionate swing are helpful in Scotland because the Lib Dems will in effect be fighting a handful of seats and in practice ignoring the rest. But equally, the Lib Dems will get some votes even in the seats that they ignore. They polled nearly 300,000 votes in 2010 in seats which they did not win. Some of them will remain loyal.
The current opinion polling is likely to understate local constituency effects because most voters don't think about them between elections. Some will, of course.
I have no magic answer to this problem. I suggest tackling it in reverse. If the Lib Dems are indeed to tally 6% at a general election, we need to assign a given average vote in seats that the Lib Dems will not seriously contest. I am uncomfortable going below 3% for this, given their past significant voting strength right across Scotland. That implies that the Lib Dems would get around 45,000 votes in the 48 Scottish seats that the Lib Dems do not currently hold.
At 6% across Scotland overall, that leaves around 100,000 votes to be shared among the remaining 11 constituencies. Charlie Kennedy and Alistair Carmichael would presumably account for around 25,000 of these votes. Which leaves nowhere near enough votes to go round for the other MPs. The Lib Dems would be lucky if they salvaged any of the rest.
Now you might say that the Lib Dems would in practice get more than 6% of the vote nationally because of the power of incumbency of some of these MPs. You might well be right - in fact, I expect that you are. But you can't decide to fix the maths to suit your desires. Bear in mind I've set the floor for votes in seats that the Lib Dems will not seriously contest at a very low level. Without any incumbents, the Lib Dems on this basis would be heading for an asterisk in the polling tables. If you are going to assume that the Lib Dems would do better than 6%, you need to assign a percentage, make some assumptions as outlined above and then consider the consequences constituency by constituency.
Where left of centre former Lib Dem voters have a clear alternative to the Lib Dems, it seems reasonable to suppose that many of them will be inclined to take it. Given the strength of feeling on the left, many of them may take such an option even if they don't have a clear alternative. That leaves current MPs relying heavily on personal votes. With Malcolm Bruce and Menzies Campbell stepping down, the Lib Dems will lose even that in two constituencies.
Where might the Lib Dems hang on? They're best placed when they're most closely challenged by Conservatives without a plausible alternative on the left. Enough left of centre voters might hold their noses and choose the lesser of two evils to get them home again. For this reason, the 4/6 on the Lib Dems keeping Berwickshire Roxburgh & Selkirk looks an OK bet, even with everything else going against the Lib Dems. But it's not compelling.
Far better is the 1/2 on Labour taking East Dunbartonshire. These odds seem bafflingly long to me, no matter how good Jo Swinson might be as an MP.