Alternative Vote in the UK

[Deleted]

What if the UK voted for AV in 2011? How would 2015, 2017 and 2019 have been different?

[J.G.Harston]

Mar 31, 2022 6:52:23 GMT @conservativeestimate said:

What if the UK voted for AV in 2011? How would 2015, 2017 and 2019 have been different?

I did some modelling of the 2010 election. The answer was "very little". link

[willpower3]

First preferences wouldn't be the same as the votes cast under FPTP, which is a key point.

[stb12]

The Lib Dems would have won North East Fife in 2017

[clyde1998]

The Electoral Reform Society have done some estimates based on polling in 2015 and 2017 (they didn't do it in 2019):

	2015	Diff	2017	Diff
Con	336	+6	304	-13
Lab	227	-5	286	+24
SNP	54	-2	27	-8
LDm	9	+1	11	-1
PC	3	nc	2	-2
Grn	1	nc	1	nc
UKIP	1	nc	0	nc
Oth	19	nc	19	nc

In 2015, they note only ten changes:

• Cambridge (LDm over Lab)
• Bath (LDm over Con)
• East Renfrewshire (Lab over SNP)
• Berwickshire, Roxburgh and Selkirk (Con over SNP)
  
• Southport (Lab over LDm)
• Ilford North (Con over Lab)
• Coventry North East (Con over Lab)
• Walsall North (Con over Lab)
• Wolverhampton South West (Con over Lab)
• Dewsbury (Con over Lab)

They didn't put a list of changes in their 2017 report - although there must be a minimum of 24 seats changing hands.

It's unfortunate they didn't do an estimate for 2019, as that may have lead to some very interesting changes - especially given the context of Brexit. Might have to have a look through the BES data myself and do some estimates.

---

That's not to say AV couldn't have a big impact - a BBC article from prior to the 2010 election found the Lib Dems would've been the official opposition in 1997, with the Conservatives losing an additional 95 seats and the Lib Dems gaining an extra 69. (link; however it assumes the SNP/PC and others had 'no net effect' from AV and voters selected the same first preference under AV as their FPTP vote).

One interesting point in that article, Lib Dem voters primarily broke towards the Conservatives in 1983, 1987 and 1992, before switching to break towards Labour from 1997 onwards (based on British Election Study data).

[willpower3]

Mar 31, 2022 8:09:02 GMT willpower3 said:

First preferences wouldn't be the same as the votes cast under FPTP, which is a key point.

[iain]

Mar 31, 2022 9:45:52 GMT clyde1998 said:

The Electoral Reform Society have done some estimates based on polling in 2015 and 2017 (they didn't do it in 2019):

	2015	Diff	2017	Diff
Con	336	+6	304	-13
Lab	227	-5	286	+24
SNP	54	-2	27	-8
LDm	9	+1	11	-1
PC	3	nc	2	-2
Grn	1	nc	1	nc
UKIP	1	nc	0	nc
Oth	19	nc	19	nc

In 2015, they note only ten changes:

• Cambridge (LDm over Lab)
• Bath (LDm over Con)
• East Renfrewshire (Lab over SNP)
• Berwickshire, Roxburgh and Selkirk (Con over SNP)
  
• Southport (Lab over LDm)
• Ilford North (Con over Lab)
• Coventry North East (Con over Lab)
• Walsall North (Con over Lab)
• Wolverhampton South West (Con over Lab)
• Dewsbury (Con over Lab)

They didn't put a list of changes in their 2017 report - although there must be a minimum of 24 seats changing hands.

It's unfortunate they didn't do an estimate for 2019, as that may have lead to some very interesting changes - especially given the context of Brexit. Might have to have a look through the BES data myself and do some estimates.

---

That's not to say AV couldn't have a big impact - a BBC article from prior to the 2010 election found the Lib Dems would've been the official opposition in 1997, with the Conservatives losing an additional 95 seats and the Lib Dems gaining an extra 69. (link; however it assumes the SNP/PC and others had 'no net effect' from AV and voters selected the same first preference under AV as their FPTP vote).

One interesting point in that article, Lib Dem voters primarily broke towards the Conservatives in 1983, 1987 and 1992, before switching to break towards Labour from 1997 onwards (based on British Election Study data).

Southport and Coventry North East (!) are clearly wrong.

[clyde1998]

Mar 31, 2022 9:54:02 GMT iain said:

Mar 31, 2022 9:45:52 GMT clyde1998 said:

The Electoral Reform Society have done some estimates based on polling in 2015 and 2017 (they didn't do it in 2019):

	2015	Diff	2017	Diff
Con	336	+6	304	-13
Lab	227	-5	286	+24
SNP	54	-2	27	-8
LDm	9	+1	11	-1
PC	3	nc	2	-2
Grn	1	nc	1	nc
UKIP	1	nc	0	nc
Oth	19	nc	19	nc

In 2015, they note only ten changes:

• Cambridge (LDm over Lab)
• Bath (LDm over Con)
• East Renfrewshire (Lab over SNP)
• Berwickshire, Roxburgh and Selkirk (Con over SNP)
  
• Southport (Lab over LDm)
• Ilford North (Con over Lab)
• Coventry North East (Con over Lab)
• Walsall North (Con over Lab)
• Wolverhampton South West (Con over Lab)
• Dewsbury (Con over Lab)

They didn't put a list of changes in their 2017 report - although there must be a minimum of 24 seats changing hands.

It's unfortunate they didn't do an estimate for 2019, as that may have lead to some very interesting changes - especially given the context of Brexit. Might have to have a look through the BES data myself and do some estimates.

---

That's not to say AV couldn't have a big impact - a BBC article from prior to the 2010 election found the Lib Dems would've been the official opposition in 1997, with the Conservatives losing an additional 95 seats and the Lib Dems gaining an extra 69. (link; however it assumes the SNP/PC and others had 'no net effect' from AV and voters selected the same first preference under AV as their FPTP vote).

One interesting point in that article, Lib Dem voters primarily broke towards the Conservatives in 1983, 1987 and 1992, before switching to break towards Labour from 1997 onwards (based on British Election Study data).

Southport and Coventry North East (!) are clearly wrong.

I've just checked both of those to check they weren't typos on my part, but no that's what their report says!

I would love to see their calculations for it! Perhaps down to the changes in how first preference votes would be allocated between parties, which somehow gave a narrow Conservative win there - which must've been ultra narrow in their calculations given the Labour candidate had over half the votes in 2015.

Southport to Labour does look wrong, although again it could be down to adjustments in how the first preference votes would lie (especially when it comes to tactical voting); Southport to the Conservatives seems more realistic.

[jamie]

Mar 31, 2022 9:45:52 GMT clyde1998 said:

It's unfortunate they didn't do an estimate for 2019, as that may have lead to some very interesting changes - especially given the context of Brexit. Might have to have a look through the BES data myself and do some estimates.

Please do!

[clyde1998]

Mar 31, 2022 13:57:48 GMT jamie said:

Mar 31, 2022 9:45:52 GMT clyde1998 said:

It's unfortunate they didn't do an estimate for 2019, as that may have lead to some very interesting changes - especially given the context of Brexit. Might have to have a look through the BES data myself and do some estimates.

Please do!

I'll hopefully have some numbers over the weekend. 👍

[Deleted]

AV or even SV in the UK opens up a treasure trove of possibilities based on the back of a fag box electoral arithmetic.


Might the Tories have picked different leaders in 1997 and 2001 if they got a mightier bruising than in our timeline?

Might the Lib Dem decapitation strategy have succeeded in more seats (other than Westmorland & Lonsdale) in 2005?

Would the Conservatives have won outright in 2010? Would the Lib Dems have got over 100 seats at that election?

Would Cameron have got a lot of UKIP second preferences and won a large majority, thus negating the need for a 2017 General Election?

Would a 'progressive alliance' have won more seats in 2019? Kensington, Wimbledon and many others saw hugely split opposition.

[clyde1998]

My first estimates, based unchanged first preference votes:

	FPTP	AV	+/-
Con	365	363	-2
Lab	203	195	-8
LDm	11	20	+9
SNP	48	49	+1
PC	4	4	nc
Grn	1	1	nc
Brx	0	0	nc

The seats I have changing hands are:

• Batley and Spen (Lab -> Con)
• Bridgend (Con -> Lab)
• Cambridgeshire South (Con -> LDm)
• Cheltenham (Con -> LDm)
• Cities of London and Westminster (Con -> LDm)
• Coventry North West (Lab -> Con)
• Delyn (Con -> Lab)
• Doncaster North (Lab -> Con)
• Dumfries and Galloway (Con -> SNP)
• East Dunbartonshire (SNP -> LDm)
• Finchley and Golders Green (Con -> LDm)
• Hartlepool (Lab -> Con)
• Hemsworth (Lab -> Con)
• Kensington (Con -> Lab)
• Lewes (Con -> LDm)
• Moray (Con -> SNP)
• Normanton, Pontefract and Castleford (Lab -> Con)
• Rotherham (Lab -> Con)
• Sheffield, Hallam (Lab -> LDm)
• Sheffield, South East (Lab -> Con)
• Stockton North (Lab -> Con)
• Wentworth and Dearne (Lab -> Con)
• Wimbledon (Con -> LDm)
• Winchester (Con -> LDm)

Some very close contests in the final round were:

• Coventry South (Lab over Con by 0.1%)
• Sheffield, South East (Con over Lab by 0.3%)
• Lewes (LDm over Con by 0.3%)
• Heywood and Middleton (Con over Lab by 0.3%)
• Guildford (LDm over Con by 0.5%)
• Weaver Vale (Lab over Con by 0.5%)
• Blyth Valley (Con over Lab by 0.5%)
• Chipping Barnet (Con over Lab by 0.7%)
• Bedford (Lab over Con by 0.9%)
• Warrington North (Lab over Con by 0.9%)
• Coventry North West (Con over Lab by 0.9%)
• Finchley and Golders Green (LDm over Con by 1.0%)
  
• Winchester (LDm over Con by 1.1%)
• Stoke-on-Trent Central (Con over Lab by 1.2%)
• Kensington (Lab over Con by 1.3%)
• Cheltenham (LDm over Con by 1.3%)
• West Aberdeenshire and Kincardine (Con over SNP by 1.4%)
• Dagenham and Rainham (Lab over Con by 1.4%)
• High Peak (Con over Lab by 1.5%)
• Bury North (Con over Lab by 1.5%)

There's a lot I can improve in the calculations:

• Producing a 'preference profile report' to better distribute third/fourth/etc preferences.
• Adjusting the first preferences to better reflect how people may have placed their first preference under STV. 418 seats had a candidate win an outright majority under FPTP.
• Dealing with major independents. The BES data doesn't ask about how likely the voter is to vote for a non-major party candidate (Con, Lab, LDm, SNP, PC, Brx, UKIP, Grn).
• Using 'seat-types' - particularly categorising how strongly Remain or Leave the constituency was (ie. Conservative voters in remain seats may have different second preferences that Conservative voters in leave seats).

However, for an estimate without adjusting the first preferences, these seat estimates look about right.

[greenchristian]

Apr 3, 2022 15:33:58 GMT clyde1998 said:

My first estimates, based unchanged first preference votes:

The seats I have changing hands are:

• Coventry North West (Lab -> Con)

I don't see how 2717 Lib Dem votes, 1956 Brexit Party votes, and 1443 Green votes break for the Conservatives, even by the small margin that's needed. None of the candidates had any measurable personal vote (I was the only candidate who was at all known in the constituency, and I definitely didn't get one). So the Green and Lib Dem votes would have been heavily remain leaning. And in 2019 Remainers were not going to preference the Conservatives over Labour.

[J.G.Harston]

Apr 3, 2022 15:33:58 GMT clyde1998 said:

My first estimates, based unchanged first preference votes:
...

Some very close contests in the final round were:

• Sheffield, South East (Con over Lab by 0.3%)

When I did my 2010 calculations, Sheffield Central ended up flipping going to 49.122% LAB / 50.877% LD. The real election re-counted into the early hours resulting in a 0.4% majority, an AV recount that close would have been draining.

[Deleted]

Who wins Bristol West in 2015 with AV?

[clyde1998]

Apr 3, 2022 16:26:26 GMT greenchristian said:

Apr 3, 2022 15:33:58 GMT clyde1998 said:

My first estimates, based unchanged first preference votes:

The seats I have changing hands are:

• Coventry North West (Lab -> Con)

I don't see how 2717 Lib Dem votes, 1956 Brexit Party votes, and 1443 Green votes break for the Conservatives, even by the small margin that's needed. None of the candidates had any measurable personal vote (I was the only candidate who was at all known in the constituency, and I definitely didn't get one). So the Green and Lib Dem votes would have been heavily remain leaning. And in 2019 Remainers were not going to preference the Conservatives over Labour.

Specific seats may have odd looking results due to the way the transfers were calculated.

The British Election Study questions don't directly include a 'what would you're second preference be?' question, rather questioning how likely (a score between 0 and 10) you are to vote for each party. This leads to the potential of multiple parties each receiving the same score: in that scenario, I've assumed 1/n votes were allocated to each tied party (where 'n' is the number of tied parties).

The Lib Dem, especially, votes broke fairly evenly between Labour and the Conservatives in this case - with Labour having a ~10pp advantage. Whereas the Brexit Party second preferences broke ~65pp more in favour of the Tories. These figures were done as a ratio of the second preference percentage of the remaining parties - which may boost less popular parties, although that appears to largely even itself out over all seats.

This is one area I do want to improve my figures though, which is why I mentioned a 'preference profile report' in my previous post. That may be a technical term, but it would basically ensure the third preferences (for example) of an excluded party are accurately mapped - as I've used the second preferences for parties the previous votes were transferred to (ie. if a Green voters' second preference was the Lib Dems, I've treated them as a Lib Dem voters when it comes to transfer Lib Dem votes). [If that makes any sense at all].

For Coventry North West, the transfer pattern was:

#	Con	Lab	LDm	Grn	Brx	NT
1	20,710	20,918	2,717	1,443	1,956	0
2	20,908 (+198)	21,393 (+475)	3,336 (+619)	0	2,039 (+83)	68 (+68)
3	22,449 (+1541)	21,607 (+215)	3,440 (+104)	0	0	247 (+179)
4	23,878 (+1429)	23,436 (+1829)	0	0	0	430 (+182)

(This should've been on the list of close results, so I'll edit that in after posting this; I missed a page of Northern/Midlands results.)

Once a more accurate picture of third/fourth preferences are mapped, I'd suggest you're right in saying this would probably be a Labour hold once that's sorted.

---

I just produced a quick matrix of hard Labour vs Conservative preference among all Lib Dem voters: 49.7% of Lib Dem voters said they were more likely to vote Labour, 26.4% Conservative, 17.6% equally likely to vote for both, 6.4% gave neither a preference.

If I assume respondents giving both a score under 5 as wouldn't give a preference to either party in an AV election, you get Labour 37.5%, Conservative 22.0%, Equal 6.0%, Not transferred 34.5%. It probably would've been more pro-Labour had Jeremy Corbyn not been Labour leader though, but the result wasn't as one-sided as people may expect.

[clyde1998]

Apr 3, 2022 17:51:40 GMT J.G.Harston said:

Apr 3, 2022 15:33:58 GMT clyde1998 said:

My first estimates, based unchanged first preference votes:
...

Some very close contests in the final round were:

• Sheffield, South East (Con over Lab by 0.3%)

When I did my 2010 calculations, Sheffield Central ended up flipping going to 49.122% LAB / 50.877% LD. The real election re-counted into the early hours resulting in a 0.4% majority, an AV recount that close would have been draining.

I imagine if AV was used, they'd look towards electronic counting machines - similar to what get used in Scotland for local elections; saves huge amount of time when it comes to transfers and recounts. Ireland's STV elections can take days to get a final result, due to them hand counting a STV election .

[Deleted]

Apr 3, 2022 17:56:17 GMT @conservativeestimate said:

Who wins Bristol West in 2015 with AV?

And Brighton Pavilion in 2010?

[jamie]

Apr 3, 2022 18:30:29 GMT clyde1998 said:

If I assume respondents giving both a score under 5 as wouldn't give a preference to either party in an AV election, you get Labour 37.5%, Conservative 22.0%, Equal 6.0%, Not transferred 34.5%. It probably would've been more pro-Labour had Jeremy Corbyn not been Labour leader though, but the result wasn't as one-sided as people may expect.

In a circumstance like 2019 where one party losses by 12% and has a leader with horrific personal ratings, those who vote for smaller parties (and therefore don’t have significant partisan attachment, or in many cases differential partisan aversion, to the big 2 parties) are not going to break for the losing party in massive numbers. If anything, it says something about the socially liberal nature of the modern Lib Dem vote and Labour identity that Lib Dem voters, those supporting a centrist/protest party, are breaking for the unpopular opposition that lost in a landslide. 

[jamie]

clyde1998 That’s a fantastic effort. A lot of those seats make sense eg; the Brexit Party preferences, but I’d forgot about Plaid Cymru breaking for Labour which of course makes sense. Could you post the Labour/Conservative/Non-transfers for each party in GB according to your method?