Scottish election results on different boundaries

[clyde1998]

I've been playing around with some regression models, so I thought I'd start sharing some data.

So far I've got estimates for how each Westminster constituency voted in the 2014 Scottish independence referendum: link.

The previous estimates on this site have unfortunately been lost, so I thought it would be helpful to start doing my own, the intention will be to do Scottish Parliamentary constituencies and ward results - with output area being possibility.

I've included my estimated results for each local authority to show the difference between my estimates and the real results.

---

Westminster constituencies (2005) by 2014 independence referendum vote

• Yes 50%+ (11): Glasgow East; Glasgow North East; Dundee West; Glasgow South West; West Dunbartonshire; Dundee East; Cumbernauld, Kilsyth and Kirkintilloch East; Glasgow Central; Glasgow North West; Coatbridge, Chryston and Bellshill; Glenrothes
• Yes 45%+ (15): Rutherglen and Hamilton West; Glasgow North; Inverclyde; Motherwell and Wishaw; Inverness, Nairn, Badenoch and Strathspey; Glasgow South; Airdrie and Shotts; North Ayrshire and Arran; Edinburgh East; East Kilbride, Strathaven and Lesmahagow; Na h-Eileanan an Iar; Kilmarnock and Loudoun; Livingston; Falkirk; Kirkcaldy and Cowdenbeath
  
• Yes 40%+ (18): Lanark and Hamilton East; Central Ayrshire; Ayr, Carrick and Cumnock; Paisley and Renfrewshire North; Midlothian; Aberdeen North; Caithness, Sutherland and Easter Ross; Angus; Linlithgow and East Falkirk; Argyll and Bute; Ochil and South Perthshire; West Aberdeenshire and Kincardine; Moray; Ross, Skye and Lochaber; Dunfermline and West Fife; Gordon; Perth and North Perthshire; Aberdeen South
• Yes <40% (14): Edinburgh North and Leith; Stirling; Edinburgh South West; East Lothian; North East Fife; East Renfrewshire; Banff and Buchan; Edinburgh South; Orkney and Shetland; Edinburgh West; East Dunbartonshire; Dumfries and Galloway; Dumfriesshire, Clydesdale and Tweeddale; Berwickshire, Roxburgh and Selkirk

---

There are some differences in the estimates between local authorities and UK parliamentary constituencies with the same boundaries; I think the biggest was 1.5pp. The reason for this is to due with some inconsistencies in the census data I've got for local authorities and UK parliamentary constituencies when the boundaries are the same (specifically related to proportion of agricultural workers and proportion of students in the area). This will be corrected once I work out which dataset is correct.

If there are any other issues people have spotted, feel free to point them out.

[iain]

Interesting. Not an 'issue' but I remember there being lots of reports (including I seem to remember an SNP member on here who was at the count) that Ross, Skye & Lochaber voted Yes. Regardless of whether those reports were true, I'd have thought, looking at post-2014 election results, that it was at least close, and certainly closer than Caithness etc.

[clyde1998]

Nov 1, 2022 19:12:55 GMT iain said:

Interesting. Not an 'issue' but I remember there being lots of reports (including I seem to remember an SNP member on here who was at the count) that Ross, Skye & Lochaber voted Yes. Regardless of whether those reports were true, I'd have thought, looking at post-2014 election results, that it was at least close, and certainly closer than Caithness etc.

Thanks. I'll have a look at that to see if there are any factors which I haven't included which could improve the estimates.

FWIW, I've got the Skye, Lochaber and Badenoch seat in the Scottish Parliament as being more Yes than Caithness, Sutherland and Ross on the current regression figures. I've definitely seen suggestions the Isle of Skye voted Yes - I'll be able to see if that's the case when I get down to ward level results.

It could be the case Yes 'won' the constituency in the count, but the local count may not have included postal votes (the postal votes being heavily skewed towards No due to the demographic differences between postal voters and in-person voters); I think this was the method used by North Lanarkshire Council, IIRC. Of course, it could simply be localised issues which I cannot factor in; I remember seeing reports of Banff and Buchan voting 52% Yes, but demographically I'm struggling to see how that's possible without localised factors - most likely the historical support for Alex Salmond and the SNP more generally.

The correlation between my constituency estimates and SNP share in general elections have declined since 2015, possibly due to the interplay with Brexit support. I believe Caithness, etc. was more Brexity than Ross, Skye and Lochaber - so that may explain why those two have gone in different directions in the past couple of elections; they both had fairly similar levels of SNP support in 2015.

I've now got the demographic data for Glasgow's Scottish Parliament constituencies ready, which was how Glasgow City Council counted them, so I'll add them to the regression to see what impact that has (I'd already included the Edinburgh UK Parliament data).

[timrollpickering]

Nov 1, 2022 22:47:07 GMT clyde1998 said:

I remember seeing reports of Banff and Buchan voting 52% Yes, but demographically I'm struggling to see how that's possible without localised factors - most likely the historical support for Alex Salmond and the SNP more generally.

Isn't Banff and Buchan Scotland's "world unto its own" constituency?

EDIT:

Nov 1, 2022 22:47:07 GMT clyde1998 said:

The correlation between my constituency estimates and SNP share in general elections have declined since 2015, possibly due to the interplay with Brexit support. I believe Caithness, etc. was more Brexity than Ross, Skye and Lochaber - so that may explain why those two have gone in different directions in the past couple of elections; they both had fairly similar levels of SNP support in 2015.

Chris Hanretty's estimates went through two sets of revisions and had Caithness, Sutherland and Easter Ross on a knife edge. His final estimates put at 51.3% (and so Banff & Buchan was not the sole estimated Leave constituency in Scotland). Ross, Skye & Lochaber was estimated at 43.4% and Inverness, Nairn, Badenoch & Strathspey at 40.1%

[bjornhattan]

Nov 2, 2022 0:15:45 GMT timrollpickering said:

Nov 1, 2022 22:47:07 GMT clyde1998 said:

I remember seeing reports of Banff and Buchan voting 52% Yes, but demographically I'm struggling to see how that's possible without localised factors - most likely the historical support for Alex Salmond and the SNP more generally.

Isn't Banff and Buchan Scotland's "world unto its own" constituency?

It does seem to defy most demographic trends and models. At the EU referendum, for instance, this BBC article states that a cluster of six wards in Banff and Buchan voted 61% Leave. Given this, it seems likely that the constituency of Banff and Buchan voted very strongly Leave - assuming the other ward in the constituency (at a guess, Turriff) voted narrowly to Remain then a figure of around 59% would be plausible.

But demographic models predict that it was quite close between Leave and Remain. Chris Hanretty's estimates it was 54% Leave, Electoral Calculus also says 54%, and my own random forest model has it at 51% Leave. It seems to me that these models aren't just wrong by chance - there are localised factors meaning the area is more pro-Leave (and at a guess, more pro-Yes) than it should be on paper.

[clyde1998]

Nov 2, 2022 0:40:48 GMT bjornhattan said:

Nov 2, 2022 0:15:45 GMT timrollpickering said:

Isn't Banff and Buchan Scotland's "world unto its own" constituency?

It does seem to defy most demographic trends and models. At the EU referendum, for instance, this BBC article states that a cluster of six wards in Banff and Buchan voted 61% Leave. Given this, it seems likely that the constituency of Banff and Buchan voted very strongly Leave - assuming the other ward in the constituency (at a guess, Turriff) voted narrowly to Remain then a figure of around 59% would be plausible.

But demographic models predict that it was quite close between Leave and Remain. Chris Hanretty's estimates it was 54% Leave, Electoral Calculus also says 54%, and my own random forest model has it at 51% Leave. It seems to me that these models aren't just wrong by chance - there are localised factors meaning the area is more pro-Leave (and at a guess, more pro-Yes) than it should be on paper.

Banff and Buchan is showing as 61.4% leave on the Hanretty figures that was in the results list from the British Election Study, which is probably adjusted for the ward results from that BBC article. I imagine there would've been a strong correlation between fisheries workers and leave vote.

[clyde1998]

I've updated the estimated to take into account the Glasgow mini-counts, correct for data inconsistencies (coterminous areas all show the same estimate now) and I've added Scottish Parliamentary constituency estimates.

[No Offence Alan]

Nice piece of work.
Paisley and Renfrewshire South appears to be missing from your list, though.
I assume it is slightly more "Yes" than P&R North.

On the point about seats moving in different directions post-2014, I think polling has shown that the net Yes/No position is similar to where it was in 2014, but there has been movement within that, with Remainers moving to Yes and Leavers moving to No.

[afleitch]

Really good work!

I might call on a favour at some point; at least with helping with a regression calculation for much older results :/