Friday, 6 February 2015

Timber! The great SNP price crash

It's time to take stock in the wake of Lord Ashcroft's Scottish constituency polls.  It's important to note that not a single vote has been cast in the coming Westminster election yet.  But its contours now seem so much sharper.  Shit just got real. 
How we got to where we are now
In 2010, the story of the Scottish Westminster election was simple: no change.  Not a single seat changed hands.  And for a long time, the expectation was for something similar in May.  This is what I wrote in April last year:
"Two things immediately stand out.  First, the SNP is fighting a battle on three fronts.  And secondly, the size of the swings required to make even limited progress is daunting.  In many of the more promising seats, the SNP would need to come from third or fourth.  Not only would they need to dislodge the incumbent, they would need to muscle past existing established challengers."
I looked at this again at the end of August just before the referendum, with a particular strategy in mind:
By this stage we had a full set of markets, but the SNP were still not expected to make any gains.  Indeed, they were shorter than 5/1 in only 14 constituencies.

That, remember, is less than six months ago.

Things changed rapidly after the referendum.  In the immediate wake of the referendum, the SNP polled well.  The effect was to result in a shortening in the prices of long shots, but betters were reluctant to back the SNP down to odds-on in any constituencies.  So the SNP dropped in just over a month from 33/1 to 5/1 in Glasgow East, for example, but were no shorter than 11/8 in Ochil & South Perthshire, having previously been 7/4:

The SNP then gained and kept a steady large lead in the polls of 20% or more, but the constituency markets did not move anywhere near as much as you would have expected (or as I expected).  This was the position just before Christmas in the Labour/SNP seats:
Despite the SNP benefiting from what pollsters were recording as a 20+% swing against Labour, the SNP were favourites to take only three seats from them:
And the slow shortening in SNP prices continued throughout January.  This was the position on Tuesday night, on the eve of release of Lord Ashcroft's constituency polls:
By this time the SNP were odds-on in 25 seats and favourites or joint favourites in four more.  Not bad, but still far below what you would expect of a party that was 20% ahead in the opinion polls.
Where we are now
What a difference a day makes.  24 little hours later, the SNP saw a price drop worthy of a shopping channel special:

The SNP are now odds-on in 39 constituencies and favourites in one more.  They are no longer than 5/1 in any constituency in Scotland, and longer than 2/1 in only eight constituencies.

This sudden catching-up with the opinion polls is inexplicable on any rational basis.  Lord Ashcroft's polls told us little that we could not have inferred from the Scotland-wide polls.  As Lord Ashcroft is at pains to point out, his opinion polls are snapshots and not predictions.  No votes have been cast as yet.  And the fundamentals remain the same.

But somehow seeing inferences confirmed by hard data has impressed on the betting public the reality of the current state of play that has been evident since at least Christmas.  The markets have finally caught up.

Interestingly, the markets have not slavishly followed the constituency polling.  Lord Ashcroft found the SNP to be ahead in Coatbridge, Chryston & Bellshill, but the SNP is still rated as an odds against proposition in this seat.  It's the overall picture that's had the impact, not the detail.

What does Lord Ashcroft's polling actually tell us?

There are few examples of required reading in political betting, but Lord Ashcroft's summary of his Scottish constituency polls is definitely in that category:

All the things to be concerned about in constituency polling continue to apply.  These polls are more reliable in aggregate than individually - there is always the risk of an outlier, and getting accurate samples on a constituency basis is a greater challenge than balancing samples on a broader basis.  For that reason, I agree with the markets in having a polite mild scepticism about the SNP's chances of taking Coatbridge, Chryston & Bellshill - a small SNP lead in one constituency poll is not something to bet the house on.

This polling is not especially good for the SNP by recent standards.  The SNP averaged 48% in the 16 constituencies polled, which is not so far away from the average tally in recent Scotland-wide polls (for example, the last two YouGovs have recorded them at 47%, the last two Ipsos-MORIs have had the SNP at 52% and 53%, the last three Survations have had the SNP at 46% twice and 48%).  But these constituencies were chosen specifically because they were in areas that were particularly Yes-friendly.  So we might expect the SNP to be recording lower percentages in other areas.  It is noteworthy that the one constituency polled where Yes did not do particularly well, Gordon, shows the SNP benefiting from a markedly smaller (though still hefty 15.5%) swing.  These straws suggest that Lord Ashcroft's polling is perhaps more consistent with the 43% vote share that ICM found in December or even the 41% vote share that Panelbase found in January.

Set against that, there are good technical reasons for thinking that this polling may understate the SNP's position.  Lord Ashcroft weights his polls by reference to respondents' recall of their vote in the 2010 election.  That makes sense in England, but in Scotland the voting pattern in the 2011 Holyrood election was very different from 2010, and there is evidence that some respondents get confused.  Since the SNP did far better in 2011 than 2010, their vote may be being weighted down in these polls.

Leaving aside Gordon, the SNP swing is remarkably uniform in the seats polled (considering the sizes of the swings), ranging from 21% in Airdrie & Shotts to 27% in Dundee West and Motherwell & Wishaw, with tight clustering around the 24% mark in most seats. There is no obvious reason that I can see not to apply this type of percentage across the board in Yes-friendly seats as a starting point.

So where might things go from here?

What are Labour's chances of winning back the defectors?  Quite a bit has been made of Lord Ashcroft's observation that "Just over two thirds (68%) of switchers from Labour to the SNP say they definitely rule out voting Labour again in 2015 – which means nearly one third are at least open to the idea of returning".  This straw has been quite firmly clutched in some quarters.

This looks to me like teasing on the part of his Lordship.  He might equally have pointed out that only 65% of current Labour supporters have ruled out the option of voting for the SNP in 2015.  In other words, things look marginally more likely to get worse for Labour from voter movements than better.  And Labour would in any case have to convert large numbers of people who have changed their allegiance but who haven't finally excluded an idea in three months.  That sounds like a tough challenge to me.

Can Labour pull it back still?  There's no shortage of people making suggestions.  Isabel Hardman of the Spectator suggests that Labour need to pound the pavements:

This is a good idea with one major flaw - Labour doesn't have the members to do this.  Still more worryingly for Labour, the SNP do.  The SNP is much more likely than Labour to be able to make the voter contact in these constituencies (though like Labour, it seems from Lord Ashcroft's polling, they have yet to get in gear on this in most seats), and I expect it to do so, firming up the nationalist vote rather than seeing it fade away.

One weakness for the SNP is that they seem to be selecting some surprising candidates to fight what are now very winnable seats.  Douglas Alexander is projected to lose his seat in this poll.  But his challenger is a 19 year old.  [EDITED TO REMOVE AN ERROR OF FACT] It might make the difference between success and failure for D Alexander (Labour) in a much tighter race.

We still don't have that clear a picture about what is happening in the areas where No was stronger.  Bearing in mind that Gordon showed a sharply lower swing to the SNP (though still very impressive at 15.5%) and that was no doubt increased by the high profile of Alex Salmond, who is standing in that seat, I intend treading carefully in such areas until we have more polling.  Lord Ashcroft has said that he is looking at more Scottish constituency polling this month - I can't wait. 

Betting opportunities

The party is over for those of us who have spent the last few months betting on the SNP in the constituency markets.  With the SNP now favourites or joint favourites in 40 seats, the markets have largely caught up with the pollsters now.

In areas with a strong Yes vote, I am taking that average 24% swing at face value, rating that swing down a bit in areas where Yes did fairly well rather than amazingly.  There may be some return to Labour ranks, but equally as I have noted, this may actually understate the swing to the SNP by a bit.  So using a 24% swing as a starting point seems fair enough.

With that in mind, the following continue to look like good bets on the SNP: Lanark & Hamilton East at 5/6 and Inverclyde at 5/6.  If you're feeling adventurous, Rutherglen & Hamilton West at 11/4 is interesting, being a seat that borders the Yes fortress of Glasgow.  I've put more on all three of these.

Some of the shorter priced bets also look good, especially in the constituencies polled where the SNP have a double digit lead.  The question you have to keep asking yourself is how Labour are going to make up the ground.  And the answer to that is not at all obvious.  How, for example, are Labour going to make up an 18% deficit in Cumbernauld, Kilsyth & Kirkintilloch East?  Yet the SNP remain priced at 1/2.  That is surely too long.  I'm not putting more on this though, simply because I have a fair amount at much longer odds.

What about in areas where Yes did less well?  I'm much more cautious there.  So far we have a constituency sample of one seat, Gordon.  We should not be extrapolating from single examples.  But equally we should not be ignoring them.  This is not the time to start ploughing in at short odds on SNP victories in such seats.  There may be betting opportunities on Labour in some of these seats, but I'm prepared to let them slip by, given the tenuous evidence so far.  If you're feeling brave, the 11/10 on Labour in Midlothian might be such an opportunity.  I'm tempted, but I'm going to pass for now.

What of your own model?

Fair question.  This is how I'd modelled a possible outcome on a constituency basis before we had more detailed results:

Using Lord Ashcroft's polls as the benchmark, it stands up quite well, all things considered.  Lord Ashcroft agreed with it that the SNP would take six out of seven of the Glasgow seats.  Lord Ashcroft found that the SNP would take Coatbridge, Chryston & Bellshill and Paisley & Renfrewshire South, which my model did not predict.  Otherwise, the same results are expected.  Lord Ashcroft finds that the SNP are doing even better than my model suggested, and agrees that the swing to the SNP is apparently less marked in Gordon than elsewhere.

It remains to be seen what will ultimately happen.  But on the evidence that we have so far it seems that my model wasn't too bad a guide to what might happen.  Now we need more polls from less favourable terrain for the SNP to see how they are doing there. 

No comments: