PVI for the UK

[bjornhattan]

Nov 11, 2022 11:51:00 GMT aargauer said:

Odd that Canterbury is "progressive", but Cambridge is "liberal right".

This is more an issue of cluster names. "Liberal right" areas aren't just right wing economically, they're also very left wing socially (far more than "Progressive" areas), and as such somewhere like Cambridge which is economically barely left of centre but socially very left wing will fall under an apparently right leaning cluster.

To be clear, this isn't a dig at clyde1998 - naming clusters is always difficult. This is especially true in "noisy" data where clusters are more a statistical construct with fuzzy edges than natural groups with clean delineations.

[clyde1998]

Nov 11, 2022 12:02:18 GMT bjornhattan said:

Nov 11, 2022 11:51:00 GMT aargauer said:

Odd that Canterbury is "progressive", but Cambridge is "liberal right".

This is more an issue of cluster names. "Liberal right" areas aren't just right wing economically, they're also very left wing socially (far more than "Progressive" areas), and as such somewhere like Cambridge which is economically barely left of centre but socially very left wing will fall under an apparently right leaning cluster.

To be clear, this isn't a dig at clyde1998 - naming clusters is always difficult. This is especially true in "noisy" data where clusters are more a statistical construct with fuzzy edges than natural groups with clean delineations.

Yeah, my names aren't the best - especially when certain terms can mean different things to different people - but give a general impression of the seats in each group.

That explanation is right, the "Liberal right" areas are more socially left wing than "Progressive" areas; the "Liberal right" and "Liberal left" groups are separated by their economic position. I think the two "Liberal" groups contain all of the most socially liberal seats between them - the most 'right' seat in the "Liberal left" group is Brighton, Pavilion at 0.615σ left; this is why Cambridge ends up in the "Liberal right" group despite being somewhat left when it comes to economics.

[bjornhattan]

Nov 11, 2022 12:39:20 GMT clyde1998 said:

Nov 11, 2022 12:02:18 GMT bjornhattan said:

This is more an issue of cluster names. "Liberal right" areas aren't just right wing economically, they're also very left wing socially (far more than "Progressive" areas), and as such somewhere like Cambridge which is economically barely left of centre but socially very left wing will fall under an apparently right leaning cluster.

To be clear, this isn't a dig at clyde1998 - naming clusters is always difficult. This is especially true in "noisy" data where clusters are more a statistical construct with fuzzy edges than natural groups with clean delineations.

Yeah, my names aren't the best - especially when certain terms can mean different things to different people - but give a general impression of the seats in each group.

That explanation is right, the "Liberal right" areas are more socially left wing than "Progressive" areas; the "Liberal right" and "Liberal left" groups are separated by their economic position. I think the two "Liberal" groups contain all of the most socially liberal seats between them - the most 'right' seat in the "Liberal left" group is Brighton, Pavilion at 0.615σ left; this is why Cambridge ends up in the "Liberal right" group despite being somewhat left when it comes to economics.

The other slightly odd label is "Rural Right". Generally this label makes sense, but it takes in some urban seats such as Sutton Coldfield or Uxbridge and South Ruislip, and in one case a "conurbation" is entirely "Rural Right" whereas its more rural surroundings fall into the "Tory Liberal" category (specifically the Blackwater Valley area around Bracknell/Camberley/Aldershot). But it's a good effort and I couldn't have done better myself!

[J.G.Harston]

Nov 11, 2022 12:02:18 GMT bjornhattan said:

Nov 11, 2022 11:51:00 GMT aargauer said:

Odd that Canterbury is "progressive", but Cambridge is "liberal right".

This is more an issue of cluster names. "Liberal right" areas aren't just right wing economically, they're also very left wing socially (far more than "Progressive" areas), and as such somewhere like Cambridge which is economically barely left of centre but socially very left wing will fall under an apparently right leaning cluster.

To be clear, this isn't a dig at clyde1998 - naming clusters is always difficult. This is especially true in "noisy" data where clusters are more a statistical construct with fuzzy edges than natural groups with clean delineations.

A key would probably help, something like this:


Though with a two-dimensional measure, I would have expected the number of colours to be a square: 9, 16, 25... not 10.

[bjornhattan]

Nov 11, 2022 14:14:04 GMT J.G.Harston said:

Nov 11, 2022 12:02:18 GMT bjornhattan said:

This is more an issue of cluster names. "Liberal right" areas aren't just right wing economically, they're also very left wing socially (far more than "Progressive" areas), and as such somewhere like Cambridge which is economically barely left of centre but socially very left wing will fall under an apparently right leaning cluster.

To be clear, this isn't a dig at clyde1998 - naming clusters is always difficult. This is especially true in "noisy" data where clusters are more a statistical construct with fuzzy edges than natural groups with clean delineations.

A key would probably help, something like this:

Though with a two-dimensional measure, I would have expected the number of colours to be a square: 9, 16, 25... not 10.

I believe this might be the sort of thing you were looking for:

(Ignore the names I've used to label the categories - they're generally shortened versions of the actual names but I had to get creative in one or two cases).

I was also able to use the map to analyse how seats in the different categories voted:

• The Conservatives dominated "Rural Right" and "Strong Conservative" constituencies, winning 98% and 93% respectively. They also won 88% of Tory-Liberal seats - a slightly lower figure not helped by this cluster having the most Lib Dem seats (4).
  
• Two clusters were fairly evenly split. 53% of "Conservative Working Class" constituencies supported the Conservatives, while 47% prefered Labour, with nineteen Conservative gains from Labour in this cluster. "Centre" constituencies were a little more diverse with some wins for the SNP, Plaid Cymru, and the Lib Dems, but again the Conservatives were the largest single party among these seats, winning 48% of them.
  
• The other clusters had lower rates of Conservative support. The Conservatives won a quarter of "Liberal Right" seats, all in London, and a fifth of "Conservative Working Class" seats, nearly all of which were gained from Labour. "Progressive" constituencies were overwhelmingly Labour (76%) with the SNP the second largest party, and "Liberal Left" constituencies were even more supportive of Labour (81%).
• The only cluster to predominantly vote for a third party was the "Economic Left" - 58% supported the SNP with the rest all backing Labour.
  

Truro and Falmouth and Colchester both backed the Conservatives despite being "Progressive", while Labour won the "Rural Right" constituency of Ilford North and the "Strong Conservative" constituencies of Batley and Spen, Derby South, Halifax, Hayes and Harlington, Luton North, Oldham East and Saddleworth, and Alyn and Deeside. It was previously noted that some ethnically diverse seats back Labour despite right wing economic views; these surprising Labour seats would all fit that mould except for Alyn and Deeside and possibly Halifax.

[clyde1998]

Nov 11, 2022 14:14:04 GMT J.G.Harston said:

Nov 11, 2022 12:02:18 GMT bjornhattan said:

This is more an issue of cluster names. "Liberal right" areas aren't just right wing economically, they're also very left wing socially (far more than "Progressive" areas), and as such somewhere like Cambridge which is economically barely left of centre but socially very left wing will fall under an apparently right leaning cluster.

To be clear, this isn't a dig at clyde1998 - naming clusters is always difficult. This is especially true in "noisy" data where clusters are more a statistical construct with fuzzy edges than natural groups with clean delineations.

A key would probably help, something like this:


Though with a two-dimensional measure, I would have expected the number of colours to be a square: 9, 16, 25... not 10.

The scatter plot looks like this:



It gives a good guide as to where the divide is between groups, clearing showing there are edge cases. 

---

The 2019 vote shares and change in each group:

Group	Seats	Con	Lab	LDm	SNP	PC	Grn	Brx	Oth		Con	Lab	LDm	SNP	PC	Grn	Brx*	Oth
Progressives	54	27%	49%	10%	7%	1%	3%	2%	0%		-1	-7	+3	+2	nc	+2	+2	nc
Economic Left	40	21%	41%	5%	28%	0%	2%	2%	1%		-2	-9	+3	+6	nc	+1	+1	+1
Cons Working Class	51	44%	39%	4%	-	0%	2%	9%	2%		+7	-14	+2	-	nc	+1	+5	+2
Liberal Left	27	15%	58%	11%	6%	-	8%	2%	0%		-2	-5	+2	+1	-	+3	+1	nc
Strong Conservative	101	60%	28%	8%	0%	0%	3%	1%	1%		+6	-8	+3	nc	nc	+1	-2	+1
Tory-Liberals	81	51%	20%	24%	1%	0%	3%	0%	1%		-3	-7	+9	nc	nc	nc	-1	+1
Centre	86	41%	37%	7%	9%	2%	2%	2%	1%		+2	-7	+2	+2	nc	+1	+1	+1
Traditional Labour	58	33%	48%	5%	-	2%	2%	8%	2%		+2	-11	+2	-	nc	+1	+5	+2
Liberal Right	20	29%	40%	23%	4%	-	3%	1%	0%		-4	-7	+9	+1	-	+1	nc	nc
Rural Right	114	59%	20%	16%	-	-	3%	0%	2%		+1	-7	+5	-	-	+1	-1	+1

*2019 Brexit Party vote share compared to 2017 UKIP vote share

If anyone can come up with some better names for the groups, that would be helpful.

[gwynthegriff]

Nov 11, 2022 17:59:38 GMT bjornhattan said:

Nov 11, 2022 14:14:04 GMT J.G.Harston said:

A key would probably help, something like this:

Though with a two-dimensional measure, I would have expected the number of colours to be a square: 9, 16, 25... not 10.

I believe this might be the sort of thing you were looking for:

(Ignore the names I've used to label the categories - they're generally shortened versions of the actual names but I had to get creative in one or two cases).

I was also able to use the map to analyse how seats in the different categories voted:

• The Conservatives dominated "Rural Right" and "Strong Conservative" constituencies, winning 98% and 93% respectively. They also won 88% of Tory-Liberal seats - a slightly lower figure not helped by this cluster having the most Lib Dem seats (4).
  
• Two clusters were fairly evenly split. 53% of "Conservative Working Class" constituencies supported the Conservatives, while 47% prefered Labour, with nineteen Conservative gains from Labour in this cluster. "Centre" constituencies were a little more diverse with some wins for the SNP, Plaid Cymru, and the Lib Dems, but again the Conservatives were the largest single party among these seats, winning 48% of them.
  
• The other clusters had lower rates of Conservative support. The Conservatives won a quarter of "Liberal Right" seats, all in London, and a fifth of "Conservative Working Class" seats, nearly all of which were gained from Labour. "Progressive" constituencies were overwhelmingly Labour (76%) with the SNP the second largest party, and "Liberal Left" constituencies were even more supportive of Labour (81%).
• The only cluster to predominantly vote for a third party was the "Economic Left" - 58% supported the SNP with the rest all backing Labour.
  

Truro and Falmouth and Colchester both backed the Conservatives despite being "Progressive", while Labour won the "Rural Right" constituency of Ilford North and the "Strong Conservative" constituencies of Batley and Spen, Derby South, Halifax, Hayes and Harlington, Luton North, Oldham East and Saddleworth, and Alyn and Deeside. It was previously noted that some ethnically diverse seats back Labour despite right wing economic views; these surprising Labour seats would all fit that mould except for Alyn and Deeside and possibly Halifax.

Ilford North "rural" ?

[bjornhattan]

Nov 11, 2022 18:36:42 GMT gwynthegriff said:

Nov 11, 2022 17:59:38 GMT bjornhattan said:

I believe this might be the sort of thing you were looking for:

(Ignore the names I've used to label the categories - they're generally shortened versions of the actual names but I had to get creative in one or two cases).

I was also able to use the map to analyse how seats in the different categories voted:

• The Conservatives dominated "Rural Right" and "Strong Conservative" constituencies, winning 98% and 93% respectively. They also won 88% of Tory-Liberal seats - a slightly lower figure not helped by this cluster having the most Lib Dem seats (4).
  
• Two clusters were fairly evenly split. 53% of "Conservative Working Class" constituencies supported the Conservatives, while 47% prefered Labour, with nineteen Conservative gains from Labour in this cluster. "Centre" constituencies were a little more diverse with some wins for the SNP, Plaid Cymru, and the Lib Dems, but again the Conservatives were the largest single party among these seats, winning 48% of them.
  
• The other clusters had lower rates of Conservative support. The Conservatives won a quarter of "Liberal Right" seats, all in London, and a fifth of "Conservative Working Class" seats, nearly all of which were gained from Labour. "Progressive" constituencies were overwhelmingly Labour (76%) with the SNP the second largest party, and "Liberal Left" constituencies were even more supportive of Labour (81%).
• The only cluster to predominantly vote for a third party was the "Economic Left" - 58% supported the SNP with the rest all backing Labour.
  

Truro and Falmouth and Colchester both backed the Conservatives despite being "Progressive", while Labour won the "Rural Right" constituency of Ilford North and the "Strong Conservative" constituencies of Batley and Spen, Derby South, Halifax, Hayes and Harlington, Luton North, Oldham East and Saddleworth, and Alyn and Deeside. It was previously noted that some ethnically diverse seats back Labour despite right wing economic views; these surprising Labour seats would all fit that mould except for Alyn and Deeside and possibly Halifax.

Ilford North "rural" ?

I suppose it's more rural than Sutton Coldfield...

Personally, I might be inclined to rename the "Rural Right" seats to "Traditional Right" (mirroring the "Traditional Left" seats which have almost diametrically opposite views). The "Strong Conservative" seats could also be renamed to "Populist Conservative" or something like that - quite a few of them are economically left of centre and many backed Labour fairly recently.

[timrollpickering]

Nov 7, 2022 19:14:16 GMT Andrew_S said:

Nov 7, 2022 12:28:16 GMT clyde1998 said:

For Labour: Ilford North (0.820σ right). A lot of the most right-leaning Labour seats are in London - Harrow West; Brent North; Ealing Central and Action - or have a large amount of students - Sheffield, Hallam; Warwick and Leamington; Reading East. I imagine most of these will be socially liberal.

I'm surprised it isn't Battersea or Putney. (The most right-leaning Labour seat).

Where was Green Badge Valley?

[timrollpickering]

Nov 9, 2022 19:31:40 GMT Andrew_S said:

And of course Banff and Buchan is believed to have been the only constituency to have voted for Brexit.

We went into this a bit on the Scottish election results on different boundaries thread. The estimates vary a bit and Chris Hanratty's final set also put Caithness, Sutherland and Easter Ross in the Leave column.

But certainly these maps confirm Banff and Buchan as a clear outlier in Scotland. ISTR some previous discussion here that it would be the natural heartland for a viable Yes/Leave party.

[timrollpickering]

Nov 11, 2022 18:36:42 GMT gwynthegriff said:

Ilford North "rural" ?

Welcome to Hainault Farm.

[J.G.Harston]

Nov 11, 2022 17:59:38 GMT bjornhattan said:

Nov 11, 2022 14:14:04 GMT J.G.Harston said:

A key would probably help, something like this:

Though with a two-dimensional measure, I would have expected the number of colours to be a square: 9, 16, 25... not 10.

I believe this might be the sort of thing you were looking for:

(Ignore the names I've used to label the categories - they're generally shortened versions of the actual names but I had to get creative in one or two cases).

I was also able to use the map to analyse how seats in the different categories voted:

• The Conservatives dominated "Rural Right" and "Strong Conservative" constituencies, winning 98% and 93% respectively. They also won 88% of Tory-Liberal seats - a slightly lower figure not helped by this cluster having the most Lib Dem seats (4).
  
• Two clusters were fairly evenly split. 53% of "Conservative Working Class" constituencies supported the Conservatives, while 47% prefered Labour, with nineteen Conservative gains from Labour in this cluster. "Centre" constituencies were a little more diverse with some wins for the SNP, Plaid Cymru, and the Lib Dems, but again the Conservatives were the largest single party among these seats, winning 48% of them.
  
• The other clusters had lower rates of Conservative support. The Conservatives won a quarter of "Liberal Right" seats, all in London, and a fifth of "Conservative Working Class" seats, nearly all of which were gained from Labour. "Progressive" constituencies were overwhelmingly Labour (76%) with the SNP the second largest party, and "Liberal Left" constituencies were even more supportive of Labour (81%).
• The only cluster to predominantly vote for a third party was the "Economic Left" - 58% supported the SNP with the rest all backing Labour.
  

Truro and Falmouth and Colchester both backed the Conservatives despite being "Progressive", while Labour won the "Rural Right" constituency of Ilford North and the "Strong Conservative" constituencies of Batley and Spen, Derby South, Halifax, Hayes and Harlington, Luton North, Oldham East and Saddleworth, and Alyn and Deeside. It was previously noted that some ethnically diverse seats back Labour despite right wing economic views; these surprising Labour seats would all fit that mould except for Alyn and Deeside and possibly Halifax.

With this sort of data representation I would have expected the colouring to be symetrical, something like this:


You seem to have got some data points in the centre coloured as eg Con or Whig for no reason. Why are some data points less than 0.7 from (0,0) coloured Centre, but some data pointes less than 0.7 from (0,0) not coloured Centre? Why are there points coloured LibR in three different sectors?

I could understand it if it was a plot of data points measured by some third measure plotted by these two measures, if it was, say, a flat plot of the two PVI variables and a colour plot of who each constituency voted for. But it's already been said the colours represent values of the the PVI variables, and that the polot is just the two PVI variables.

[bjornhattan]

Nov 11, 2022 22:49:12 GMT J.G.Harston said:

Nov 11, 2022 17:59:38 GMT bjornhattan said:

I believe this might be the sort of thing you were looking for:

(Ignore the names I've used to label the categories - they're generally shortened versions of the actual names but I had to get creative in one or two cases).

I was also able to use the map to analyse how seats in the different categories voted:

• The Conservatives dominated "Rural Right" and "Strong Conservative" constituencies, winning 98% and 93% respectively. They also won 88% of Tory-Liberal seats - a slightly lower figure not helped by this cluster having the most Lib Dem seats (4).
  
• Two clusters were fairly evenly split. 53% of "Conservative Working Class" constituencies supported the Conservatives, while 47% prefered Labour, with nineteen Conservative gains from Labour in this cluster. "Centre" constituencies were a little more diverse with some wins for the SNP, Plaid Cymru, and the Lib Dems, but again the Conservatives were the largest single party among these seats, winning 48% of them.
  
• The other clusters had lower rates of Conservative support. The Conservatives won a quarter of "Liberal Right" seats, all in London, and a fifth of "Conservative Working Class" seats, nearly all of which were gained from Labour. "Progressive" constituencies were overwhelmingly Labour (76%) with the SNP the second largest party, and "Liberal Left" constituencies were even more supportive of Labour (81%).
• The only cluster to predominantly vote for a third party was the "Economic Left" - 58% supported the SNP with the rest all backing Labour.
  

Truro and Falmouth and Colchester both backed the Conservatives despite being "Progressive", while Labour won the "Rural Right" constituency of Ilford North and the "Strong Conservative" constituencies of Batley and Spen, Derby South, Halifax, Hayes and Harlington, Luton North, Oldham East and Saddleworth, and Alyn and Deeside. It was previously noted that some ethnically diverse seats back Labour despite right wing economic views; these surprising Labour seats would all fit that mould except for Alyn and Deeside and possibly Halifax.

With this sort of data representation I would have expected the colouring to be symetrical, something like this:


You seem to have got some data points in the centre coloured as eg Con or Whig for no reason. Why are some data points less than 0.7 from (0,0) coloured Centre, but some data pointes less than 0.7 from (0,0) not coloured Centre? Why are there points coloured LibR in three different sectors?

I could understand it if it was a plot of data points measured by some third measure plotted by these two measures, if it was, say, a flat plot of the two PVI variables and a colour plot of who each constituency voted for. But it's already been said the colours represent values of the the PVI variables, and that the polot is just the two PVI variables.

That would be one way of dividing the points - as would using a grid like you suggested earlier to divide the constituencies into 9 (or 16, or 25). However the approach used here is a clustering algorithm which tries to identify groups of constituencies with relatively similar attributes to each other and distinct attributes from other clusters. Because the data here is fairly evenly distributed this is quite messy and there are plenty of points on the edge which could belong to two (or even three catgeories), but the areas of highest density have been separated in a way that a simple grid might fail to do.

[The Bishop]

The difference between Scotland and England in the above maps is indeed intriguing - the supposed "myth" of their being similar in left-right terms came from numerous social attitudes surveys that showed less difference between them in public opinion regarding several issues than the contrast in voting behaviour might suggest. Given that, might the PVI findings be partly explained by their having different conceptions of what "left" and "right" actually is?

[clyde1998]

Nov 12, 2022 11:53:17 GMT The Bishop said:

The difference between Scotland and England in the above maps is indeed intriguing - the supposed "myth" of their being similar in left-right terms came from numerous social attitudes surveys that showed less difference between them in public opinion regarding several issues than the contrast in voting behaviour might suggest. Given that, might the PVI findings be partly explained by their having different conceptions of what "left" and "right" actually is?

Potentially, however I deliberately used a calculation based on five specific questions for economic and social position, as opposed to what people say they are, for this reason: different people have different perceptions of what 'left' and 'right' means. The individual perception of left and right has no meaningful correlation with their responses to specific questions.

The average response in Scotland for each individual question isn't hugely different from the average response in England: the biggest difference is on the 'Government should redistribute income from the better off to those who are less well off' question, with 'strongly agree' being 6pp higher in Scotland than England (using BES weights). Scotland is to the left of England on all questions; Wales is to the left of England too on all questions, but to a significantly smaller degree.

I guess the reality is somewhere in the middle - Scotland is, on average, to the left of England, but it's not to a scale that means Scotland has completely different values to England: the difference is noticeable when it comes to whether someone would consider voting Labour or Conservative (for example), but not noticeable on the entirety of the political spectrum.

The difference may be in the demographic makeup of Scotland compared to England, rather than individual groups having noticeably different responses, too - but that's me speculating.

[bjornhattan]

Nov 12, 2022 16:09:56 GMT clyde1998 said:

Nov 12, 2022 11:53:17 GMT The Bishop said:

The difference between Scotland and England in the above maps is indeed intriguing - the supposed "myth" of their being similar in left-right terms came from numerous social attitudes surveys that showed less difference between them in public opinion regarding several issues than the contrast in voting behaviour might suggest. Given that, might the PVI findings be partly explained by their having different conceptions of what "left" and "right" actually is?

Potentially, however I deliberately used a calculation based on five specific questions for economic and social position, as opposed to what people say they are, for this reason: different people have different perceptions of what 'left' and 'right' means. The individual perception of left and right has no meaningful correlation with their responses to specific questions.

The average response in Scotland for each individual question isn't hugely different from the average response in England: the biggest difference is on the 'Government should redistribute income from the better off to those who are less well off' question, with 'strongly agree' being 6pp higher in Scotland than England (using BES weights). Scotland is to the left of England on all questions; Wales is to the left of England too on all questions, but to a significantly smaller degree.

I guess the reality is somewhere in the middle - Scotland is, on average, to the left of England, but it's not to a scale that means Scotland has completely different values to England: the difference is noticeable when it comes to whether someone would consider voting Labour or Conservative (for example), but not noticeable on the entirety of the political spectrum.

The difference may be in the demographic makeup of Scotland compared to England, rather than individual groups having noticeably different responses, too - but that's me speculating.

I think there may be a split between urban Scotland and rural Scotland. The Central Belt seems to have fairly similar attitudes to comparable areas in England (such as the North East or the urban parts of the M62 corridor) - perhaps it's slightly more socially left wing but it's not obviously out of place. What does seem to be more left wing than you'd expect is rural Scotland - particularly the constituencies in the Highlands which seem far more economically left wing than an English seat with similar demographics would be. The only rural Scottish seat which doesn't stand out in this way is Dumfries and Galloway, which is relatively similar to places like Cumbria - but culturally there is far more English influence and there may be a split between Dumfries town (which I suspect is very economically left wing and pretty socially right wing) and the rest of the constituency which looks more like other parts of rural Scotland.