Obviously we’re all a little wary of using the polls since the General Election, but it should be remembered that while the polls then underestimated the Tories and over-estimated Labour, they got our tiny percentage pretty much spot on. The only problem was that, because the Tory vote was higher in actuality than predicted, that we ended up losing a few more Tory/LD marginals than we’d ever expected to.
But Pollsters have been amending their weighting…and the ‘Shy Tory Voter’ doesn’t really seem to exist anymore. They’re out and proud! So I’ve been having a look at what the polls have been predicting for a General Election held today.
My reading of the polls uses averages to fill in the gaps on the days when there aren’t polls, and then to run seven day rolling figures based on both those averages and the polls themselves.
The model I use is modelled heavily on that of Martin Baxter’s Electoral Calculus model, as featured often in the press. The idea is that one takes the ‘base’ vote of each constituency (i.e. Southport 2010 36% Con, 9% Lab, 50% LD etc), place that against the GE figure for that year (i.e. 37 Con, 30 Lab, 24 LD) and then see how the constituency figure moves with regards to latest polling (i.e. the final poll pre 2015 GE of 33 Con, 33 Lab and 9 LD) and then see how the initial constituency figure moves (in this case to 33 Con, 12 Lab and 35 LD) thus showing how John Pugh kept his seat.
I’ve been using my model since 2013 and modelled the SNP increases in as it became apparent that they were going to sweep some serious seats up. As an idea of how well it worked, using the 2010 baseline for seats with the actual 2015 GE percentage provided a result of Con 327 seats, Lab 254, LD 11 and SNP 35. The only real issue here was of an under-performing SNP, but an event of such political magnitude was always going to be difficult to capture. The key thing though is that even operating so far from the 2010 election, my model predicted that the LDs would hold six of the eight seats they *did* hold (barring Carshalton and Southport) and gave them just an extra five they didn’t win (among them the very close losses of Twickenham, Yeovil etc.)
So I think my model works well. As I say, it uses a similar method to Martin Baxter’s, but has the benefit in that the user gets percentage
splits in individual seats, whereas Baxter’s model merely gives gains/losses.
With that in mind, then, what have the polls been doing since the election?
Well, the Tory vote of 38% rose swiftly to 40% within a few days of the election, and then 41% by 18th May. It’s fallen since then,arriving back at 39% by 9th June and then 38% by 11th July and down to 37% as of 20th July.
Labour saw their 31% election figure reduced to 30% by 22nd May and 29% by 5th June, before rallying, before dropping back to 28% by 27th
June. It’s been rising swiftly since then though, and reached 31% by 8th July, 32% by 10th July and 33% by 12th July, before retreating by
a percentage point by the 20th.
The Lib Dems have also seen quite some flux since our 8% at the election, falling to 7% within two days, before climbing back to 8% by 3rd June, 9% by 17th June and 10% by 21st June. However, the resurrection of Labour has seen out vote fall back to 9% by 26th June,
8% by 4th July and finally 7% by 9th July. However, the recent Ipsos-MORI double figure polled has pulled our average back up to 8%
as of the 18th and finally 9% by the 20th July.
What does this mean in terms of seats? Well, the continued SNP rise post-election would have seen Alistair Carmichael lose Orkney by 10th
May, with the continued Tory increases seeing both John Pugh and TomBrake lose their seats too, a few days later. Five seats seems to be
the nadir so far, and Orkney was theoretically taken back by the 25th May. The demise of Labour in early June saw us theoretically take back
Julian Huppert’s old Cambridge seat by 4th June, with our own increase in the polls seeing the return of both Carshalton and Southport a day
or so later. Net gain = 1!
Sadly, the resurgent Labour party saw Cambridge fall again by 10th June, before it returned again on the 18th – which is just before theSNP ‘retook’ Orkney, although it had returned again by the 27th. Cambridge, however, fell victim again to a highly resurgent Labour
party by 7th July. Since then, a momentary drop in SNP support saw Jo Swinson’s East Dunbartonshire theoretically come back to us on 10th July, but lost again eight days later.
The current state of play as of the last poll on 10th July is that there are eight seats for us, with Eastbourne within 0.5%, Lewes
within 1.3%, Julian Huppert’s Cambridge 1.5%, Steve Webb’s Thornbury 2.2%, Twickenham 2.4% (although one could argue about name recognition/incumbency with regards to Vince – did this pull his GE score up…or down?), Jo Swinson’s Dunbarton 2.4%, Ed Davey’s Kingston
3.9%, and St. Ives and Mike Crockart’s Edinburgh 4.3%
From a National point of view, the Conservatives lost their absolute majority as of 13th July, at which point a Labour/LD/Green/SNP/PC/SDLP
coalition would be within nine seats of government. Excluding SNP/PCleaves the opposition 65 seats short.
There’s a long, long way to go, but there are seats within (notional) grasping even with just a couple more percentage increases in the
polls. For example, if we gained 3% and the Tories, Labour and SNP lost 1% each, we’d be back up to 20 seats again.
Let the #LibDemFightback continue!
* Stephen is a lifelong Lib Dem voter, but one of many 8/5/15 members. He analyses and recommends for a major city Policy and Strategy organisation. He has two young children and used to have time for making and writing about music.