xenon
Forum Regular
Posts: 426
|
Post by xenon on Jul 21, 2022 16:23:30 GMT
The Landtag of the German state of Baden-Württemberg recently voted to move away from its unique method of AMS and towards a more standard version with closed party lists, which I think is a shame because it was a good way of enabling proportionality without relying on unaccountable party lists. For those who aren't aware, in Baden-Württemberg voters only have one vote, which is counted towards both the candidate's total in the constituency and their party's total on the regional list. 70 of the 120 seats (excluding any extra overhang seats) are awarded to the winner in each constituency, with the remaining 50 being distributed proportionally among the parties which gained over 5% of the vote across the state, taking into account the number of seats each party already has. These seats are filled by ranking each of the party's candidates by the vote share they gained in their constituencies, with the top n runners up winning a "second mandate" (where n is the number of top-up seats allocated to that party). (There's a more detailed explanation of the system here, along with the results of the last year's election) With this in mind, I thought it might be interesting to see how this system would apply to the UK. It has several benefits over other forms of proportional representation, most notably that voters cast their ballots in exactly the same way as they do at present. It was recommended by the Hansard Society in the 1970s for adoption for Westminster elections (see here), but has rarely been suggested more recently. For this experiment, each parliamentary seat will be merged with a neighbour to create 325 larger constituencies, and will continue to elect one member using first past the post. The remaining seats are distributed to 12 regional lists corresponding roughly to the former European Parliament constituencies, and then allocated to the parties using the D'Hondt method. To start, here's Scotland at the 2019 general election. This was the most noticeably unproportional area of the UK at the last election, with the SNP picking up 48 of the 59 seats (81%) on 45% of the vote, while at the other end of the scale Labour got just one seat for slightly over 18% of the vote. I've opted to merge the two islands constituencies with Caithness etc. to avoid an odd number of seats breaking the border with England, so this results in 29 constituency seats and 30 on the regional list. The results under this system would be as follows: SNP: 28 seats (-20 vs FPTP), comprising 26 first mandates and 2 second mandates Con: 15 seats (+9 vs FPTP), comprising 3 first mandates and 12 second mandates Lab: 11 seats (+10 vs FPTP), all second mandates LD: 5 seats (+1 vs FPTP), all second mandates  As you can see from the map, most constituencies end up with two members each, with three with only one member each, and four with three members each. Labour and the Conservatives pick up second mandates roughly where you would expect (i.e. mainly urban Central Belt seats for the former, and more rural and affluent ones for the latter), and the Lib Dems make up for the loss of first-placed finishes by gaining second mandates in the same areas. The additional members could either explicitly represent the constituency in which they ran (as in Baden-Württemberg), or simply become a list member for the entire region. |
|
|
|
Post by Pete Whitehead on Jul 21, 2022 16:56:22 GMT
Excellent thread. You've done a good job of pairing the constituencies there as well - I suspect some English regions may prove more tricky (we have already looked at Wales in the context of the proposed changes to elections to the Welsh Assembly)
|
|
|
|
Post by johnloony on Jul 21, 2022 19:00:07 GMT
The idea that the additional members would still represent the constituencies in which they stood sort-of sounds good, until you go to the next thought which is that so,e constituencies have 3 MPs and some have only 1. That feature alone makes it unacceptable. If you deem them to be representing the entire region, that is also unfair because they have been voted for by the whole region. In some places, even if there are only 2 MPs in the constituency, the second elected MP might be the one who was in 3rd place rather than the one in 2nd place.
A more sensible method would be to have a single vote (in the constituency) with the additional members coming from a separate list, albeit that the votes for the list are simply the aggregation of the votes from the constituencies.
The flaws in such a version of AMS are so microkerfufflious and contrafibularitative that it makes the whole system not worth the candle it’s written on.
If you want to have 2 MPs per constituency, just do it properly with a closed list system with 2 seats per constituency, allocation by d’Hondt. I think that’s what they do (or used to do) in Chile.
|
|
xenon
Forum Regular
Posts: 426
|
Post by xenon on Jul 21, 2022 20:55:58 GMT
The idea that the additional members would still represent the constituencies in which they stood sort-of sounds good, until you go to the next thought which is that so,e constituencies have 3 MPs and some have only 1. That feature alone makes it unacceptable. I agree about the unequal representation of constituencies being problematic. However it is somewhat remedied by the fact that the winning candidate in a one-member constituency will be more representative of their seat than in a two or three member one – for example in Dundee some 54% of those polled would have voted for their SNP MP, whereas in Mid and East Lothian that figure is under 39%. I don't think this would really matter in practice, as the amount of people who wouldn't vote for a party because their vote might elect someone not on their ballot paper would be vanishingly small – very few are aware of most list candidates for Scottish Parliament elections for example, and the vast majority of people vote for a party on the second ballot without any regards as to the candidates that their vote might elect. While this may occur on occasion (particularly with small parties), none of the constituencies in the example have this particular problem. Even where it does occur, it isn't too dissimilar to the levelling seats used in Scandinavia and Germany, where seats can be back-allocated to a region based on nationwide results that might not match the order of parties in that region but is accepted as necessary as it makes the overall result more proportional. This could work, but if you are going to have separate lists then you may as well have a separate ballot too. There's also the issue that in the "second mandate" system, your vote always has a chance of electing the candidate on the ballot paper as a second member, so you are in effect voting for them twice (it just so happens that on the top up vote you are also voting for all the other candidates from their party across the region too). With a list, there is no guarantee that that candidate will be on the list at all, let alone in a winnable position. This would render the whole system rather pointless, as the district magnitude is in effect 2 (rather than the 30 for the Scotland-wide list), meaning that it's almost as disproportionate as FPTP, and you have the same issue with the inflexibility of closed lists. |
|
|
|
Post by jamesdoyle on Jul 22, 2022 8:10:30 GMT
This is very interesting, I'm looking forward to seeing other regions.
|
|
|
|
Post by swindonlad on Jul 22, 2022 10:43:45 GMT
It may take me a day or 2, but I'm looking at the South West.
I am using the new proposals for constituencies as there's an even number, otherwise I'll need to lose 1.
There will be some odd looking pairings as trying not to cross county border pairings (except where there is a necessity)
|
|
|
|
Post by 🏴☠️ Neath West 🏴☠️ on Jul 22, 2022 12:31:12 GMT
 Here's the 2019 election in Wales, presuming I've understood this right. Quite an interesting map – only ten constituencies end up with two mandates. Five end up with three (Arfon+Ynys Môn, Aberconwy+Dwyfor Meirionnydd, Carmarthen East+Ceredigion, Gower+Llanelli, and Caerphilly+Islwyn) – basically a map of where Plaid Cymru are competitive. Five end up with only one (Powys, Aberavon+Neath, Cynon Valley+Rhondda, Blaenau Gwent+Merthyr, and Cardiff Central+South). The perverse effect here is that the two constituencies with the lowest electorates (Aberconwy+DM 89,061 and Arfon+YM 94,140) end up with three members and the constituency with the highest electorate (Cardiff C+S 142,874) ends up with just one. Powys is also a bit of an oddity, in that the Lib Dems came second, but weren't eligible for any mandates as they only got 4.0% of the vote across Wales. The state of the parties: Labour 19 (-3) (11 first and 8 second mandates) Conservative 16 (+2) (7 first and 9 second mandates) Plaid Cymru 5 (+1) (2 first and 3 second mandates) Very strange electoral system. |
|
|
|
Post by bjornhattan on Jul 22, 2022 15:11:47 GMT
It may take me a day or 2, but I'm looking at the South West. I am using the new proposals for constituencies as there's an even number, otherwise I'll need to lose 1. There will be some odd looking pairings as trying not to cross county border pairings (except where there is a necessity) As you point out, regional seat allocations seem to be a recurring issue here. I had a quick look at the North East - the problem there was that the region had 29 seats (and drops to 27, so your trick doesn't help), but the obvious cross-boundary pairings run into Yorkshire, which has an even number of seats. You could link Penrith and Hexham, but this messes up the obvious pairings in Cumbria and North Lancashire and generally produces a much worse solution.
So either an extra seat will need to be merged somewhere, or there will be a chain of cross-region seats across the North. Scarborough, Whitby, and East Cleveland isn't too bad a pairing, but virtually every seat which merges the North West and Yorkshire is pretty awful (Pendle and Skipton&Ripon may be the least bad, just from a quick glance - at least a lot of the former seat is traditional Yorkshire).
I shall be very interested to see how xenon deals with this issue. |
|
xenon
Forum Regular
Posts: 426
|
Post by xenon on Jul 22, 2022 16:15:07 GMT
It may take me a day or 2, but I'm looking at the South West. I am using the new proposals for constituencies as there's an even number, otherwise I'll need to lose 1. There will be some odd looking pairings as trying not to cross county border pairings (except where there is a necessity) As you point out, regional seat allocations seem to be a recurring issue here. I had a quick look at the North East - the problem there was that the region had 29 seats (and drops to 27, so your trick doesn't help), but the obvious cross-boundary pairings run into Yorkshire, which has an even number of seats. You could link Penrith and Hexham, but this messes up the obvious pairings in Cumbria and North Lancashire and generally produces a much worse solution.
So either an extra seat will need to be merged somewhere, or there will be a chain of cross-region seats across the North. Scarborough, Whitby, and East Cleveland isn't too bad a pairing, but virtually every seat which merges the North West and Yorkshire is pretty awful (Pendle and Skipton&Ripon may be the least bad, just from a quick glance - at least a lot of the former seat is traditional Yorkshire).
I shall be very interested to see how xenon deals with this issue. I'm currently doing the North West (I'll post the result later when I have a chance), and I've managed to get an even number of seats by adding High Peak to Stalybridge and Hyde. However that means that the East Midlands now has an odd number of seats, so I may have to shuffle about with Northern Lincolnshire and Teesside to get the North East and Yorkshire within the rules. Edited to add: Here's the 2019 election in Wales, presuming I've understood this right. Quite an interesting map – only ten constituencies end up with two mandates. Five end up with three (Arfon+Ynys Môn, Aberconwy+Dwyfor Meirionnydd, Carmarthen East+Ceredigion, Gower+Llanelli, and Caerphilly+Islwyn) – basically a map of where Plaid Cymru are competitive. Five end up with only one (Powys, Aberavon+Neath, Cynon Valley+Rhondda, Blaenau Gwent+Merthyr, and Cardiff Central+South). The perverse effect here is that the two constituencies with the lowest electorates (Aberconwy+DM 89,061 and Arfon+YM 94,140) end up with three members and the constituency with the highest electorate (Cardiff C+S 142,874) ends up with just one. Powys is also a bit of an oddity, in that the Lib Dems came second, but weren't eligible for any mandates as they only got 4.0% of the vote across Wales. The state of the parties: Labour 19 (-3) (11 first and 8 second mandates) Conservative 16 (+2) (7 first and 9 second mandates) Plaid Cymru 5 (+1) (2 first and 3 second mandates) Very strange electoral system. According to your calculations would the Lib Dems be in line for a seat if there was no threshold, or does the number of seats create a natural threshold anyway? If the former is the case then it could be worth adjusting the threshold to 3% (as in Spain), or 4% (as in Sweden and Norway). I've come across a similar issue in the North West, as the Brexit Party had a 3.8% share, and as such should be entitled to 2 or 3 seats if the threshold didn't apply. I suppose instead of lowering the threshold you could have an exemption for parties that win a certain share (say over 10%) in a given number of constituencies, which I believe is similar what occurs in the Scandinavian countries. |
|
|
|
Post by 🏴☠️ Neath West 🏴☠️ on Jul 22, 2022 16:53:26 GMT
As you point out, regional seat allocations seem to be a recurring issue here. I had a quick look at the North East - the problem there was that the region had 29 seats (and drops to 27, so your trick doesn't help), but the obvious cross-boundary pairings run into Yorkshire, which has an even number of seats. You could link Penrith and Hexham, but this messes up the obvious pairings in Cumbria and North Lancashire and generally produces a much worse solution. So either an extra seat will need to be merged somewhere, or there will be a chain of cross-region seats across the North. Scarborough, Whitby, and East Cleveland isn't too bad a pairing, but virtually every seat which merges the North West and Yorkshire is pretty awful (Pendle and Skipton&Ripon may be the least bad, just from a quick glance - at least a lot of the former seat is traditional Yorkshire). I shall be very interested to see how xenon deals with this issue. I'm currently doing the North West (I'll post the result later when I have a chance), and I've managed to get an even number of seats by adding High Peak to Stalybridge and Hyde. However that means that the East Midlands now has an odd number of seats, so I may have to shuffle about with Northern Lincolnshire and Teesside to get the North East and Yorkshire within the rules. From the persepective of the West Midlands, it would be helpful to pair Cheshire with Shropshire, even though the resultant North Shropshire and Eddisbury would be one for the Pitchfork Bait thread – all of the other West Midland counties have even numbers of constituencies. That should then let Merseyside and SELNEC be a self-contained pair. |
|
|
|
Post by 🏴☠️ Neath West 🏴☠️ on Jul 22, 2022 17:08:34 GMT
Here's the 2019 election in Wales, presuming I've understood this right. Quite an interesting map – only ten constituencies end up with two mandates. Five end up with three (Arfon+Ynys Môn, Aberconwy+Dwyfor Meirionnydd, Carmarthen East+Ceredigion, Gower+Llanelli, and Caerphilly+Islwyn) – basically a map of where Plaid Cymru are competitive. Five end up with only one (Powys, Aberavon+Neath, Cynon Valley+Rhondda, Blaenau Gwent+Merthyr, and Cardiff Central+South). The perverse effect here is that the two constituencies with the lowest electorates (Aberconwy+DM 89,061 and Arfon+YM 94,140) end up with three members and the constituency with the highest electorate (Cardiff C+S 142,874) ends up with just one. Powys is also a bit of an oddity, in that the Lib Dems came second, but weren't eligible for any mandates as they only got 4.0% of the vote across Wales. The state of the parties: Labour 19 (-3) (11 first and 8 second mandates) Conservative 16 (+2) (7 first and 9 second mandates) Plaid Cymru 5 (+1) (2 first and 3 second mandates) Very strange electoral system. According to your calculations would the Lib Dems be in line for a seat if there was no threshold, or does the number of seats create a natural threshold anyway? If the former is the case then it could be worth adjusting the threshold to 3% (as in Spain), or 4% (as in Sweden and Norway). I've come across a similar issue in the North West, as the Brexit Party had a 3.8% share, and as such should be entitled to 2 or 3 seats if the threshold didn't apply. I suppose instead of lowering the threshold you could have an exemption for parties that win a certain share (say over 10%) in a given number of constituencies, which I believe is similar what occurs in the Scandinavian countries. Dropping it to 3% gives: Labour 17 (the Aberconwy+DM and Ceredigion+Carmarthen E dots disappear) Conservative 15 (the Swansea dot disappears) Plaid Cymru 4 (the Caerphilly dot disappears) Lib Dems 2 (Powys, Cardiff C+S) Brexit 2 (BG+Merthyr, Caerphilly) Dropping it altogether does not give the Green Party a seat by quite a margin: seats 41–50 would be 41) LD, Ceredigion+Carmarthen E 42) Lab, Aberconwy+DM 43) C, Swansea 44) Lab, Ceredigion+Carmarthen E 45) PC, Caerphilly 46) C, Cardiff C+S 47) Brexit, Cynon+Rhondda 48) Lab, Powys 49) C, Aberavon+Neath 50) Green, Pontypridd+VOG |
|
xenon
Forum Regular
Posts: 426
|
Post by xenon on Jul 22, 2022 17:56:16 GMT
Here's how the North West would look like under this system:  As mentioned earlier, High Peak is merged with Stalybridge and Hyde to give a total of 38 constituencies, and a corresponding 38 list seats. Five seats have one member each, five have three members each, and the remaining 28 have two members each. You'll notice that every seat has a Labour member, and Liverpool Walton and Bootle has two - this has occurred with the CDU in Baden-Württemberg on several occasions, with the solution being that a substitute is named on the ballot paper along with the candidate, who then takes the second mandate. This total of 39 seats gives Labour a majority in the region, even though they only received 46.5% of the vote - had the threshold been 3%, then they would have lost two seats to the Brexit Party and therefore their majority (the Conservatives also would have lost one, with the Brexit Party picking up three second mandates in Wigan and Makerfield, St Helens, and Rochdale, Heywood and Middleton). If there were no threshold at all then the Green Party would pick up a seat in Chorley and Blackburn at the expense of the Conservatives, but this is down to the Speaker standing in Chorley (their next strongest performance would have been in Liverpool Riverside and Wavertree). On the topic of the Speaker, theoretically he loses out on his seat, as Labour's votes in Blackburn total some 2,000 more than Hoyle's in Chorley. However it is likely that if this system was implemented the Speaker would continue to get a free run from the main parties, or alternatively he might be awarded an automatic second mandate in his constituency, with the number of available list seats in the region decreased by one accordingly. The results (with the 5% threshold) were: Lab: 39 seats (-2 seats vs FPTP), comprising 20 first mandates and 19 second mandates Con: 31 seats (-2 seats vs FPTP), comprising 18 first mandates and 13 second mandates LD: 6 seats (+5 seats vs FPTP), all second mandates |
|
J.G.Harston
Lib Dem
Leave-voting Brexit-supporting Liberal Democrat
Posts: 14,755 
|
Post by J.G.Harston on Jul 22, 2022 18:30:51 GMT
It may take me a day or 2, but I'm looking at the South West. I am using the new proposals for constituencies as there's an even number, otherwise I'll need to lose 1. There will be some odd looking pairings as trying not to cross county border pairings (except where there is a necessity) As you point out, regional seat allocations seem to be a recurring issue here. I had a quick look at the North East - the problem there was that the region had 29 seats (and drops to 27, so your trick doesn't help), but the obvious cross-boundary pairings run into Yorkshire, which has an even number of seats. You could link Penrith and Hexham, but this messes up the obvious pairings in Cumbria and North Lancashire and generally produces a much worse solution.
So either an extra seat will need to be merged somewhere, or there will be a chain of cross-region seats across the North. Scarborough, Whitby, and East Cleveland isn't too bad a pairing, but virtually every seat which merges the North West and Yorkshire is pretty awful (Pendle and Skipton&Ripon may be the least bad, just from a quick glance - at least a lot of the former seat is traditional Yorkshire).
I shall be very interested to see how xenon deals with this issue. I think there's three seats in lost Cleveland and three in north Lincolnshire. Add lost Cleveland to Yorkshire and take lost Lincolnshire out, would that sort it? |
|
|
|
Post by bjornhattan on Jul 22, 2022 19:55:10 GMT
As you point out, regional seat allocations seem to be a recurring issue here. I had a quick look at the North East - the problem there was that the region had 29 seats (and drops to 27, so your trick doesn't help), but the obvious cross-boundary pairings run into Yorkshire, which has an even number of seats. You could link Penrith and Hexham, but this messes up the obvious pairings in Cumbria and North Lancashire and generally produces a much worse solution.
So either an extra seat will need to be merged somewhere, or there will be a chain of cross-region seats across the North. Scarborough, Whitby, and East Cleveland isn't too bad a pairing, but virtually every seat which merges the North West and Yorkshire is pretty awful (Pendle and Skipton&Ripon may be the least bad, just from a quick glance - at least a lot of the former seat is traditional Yorkshire).
I shall be very interested to see how xenon deals with this issue. I think there's three seats in lost Cleveland and three in north Lincolnshire. Add lost Cleveland to Yorkshire and take lost Lincolnshire out, would that sort it? The East Midlands would then have 49 seats - but if High Peak is moved to the North West, then that issue is resolved too. The only minor snag is Brigg and Goole, which is mostly in Lincolnshire but partly in Yorkshire too. On your numbers that would need to be paired with a Yorkshire seat, and none of them work that well - Haltemprice is least bad but means you have to combine Thirsk and Bridlington. Fortunately other than that monster there don't seem to be any other issues - parts of Stockton South are in Yorkshire, but creating a paired seat coterminous with Stockton borough seems reasonable regardless of county boundaries (cue pitchforks in Yarm). |
|
|
|
Post by Pete Whitehead on Jul 23, 2022 6:07:34 GMT
It may take me a day or 2, but I'm looking at the South West. I am using the new proposals for constituencies as there's an even number, otherwise I'll need to lose 1. There will be some odd looking pairings as trying not to cross county border pairings (except where there is a necessity) Isn't the point of pairing the existing constituencies that it makes it possible to calculate the result of the last election? I know there are various notional results available on the proposed new boundaries but that is not quite the same. I would guess (or hope at least) that if such a system were adopted here that entirely new constituency boundaries would be drawn up for the direct mandates which would not be constrained by the existing boundaries used under a different system. For that matter it should not necessaril have to be exactly 325 seats as the additional member element could be say 40% rather than 50. It wouldn;t then be necessary to cross any regional boundaries and if we didn't impose too tight a quota it should also be possible to avoid too many county crossings as well. |
|
|
|
Post by Pete Whitehead on Jul 23, 2022 6:16:46 GMT
I mean if you look at London and start in the East (I always start from Havering for some reason), you have a bunch of very natural and obvious pairings: Romford/Hornchurch & Upminster; Dagenham & Rainham/Barking; Ilford North/South; Chingford/Walthamstow; Leyton/East Ham. You are then left with West Ham which would have have to pair with a seat on the West bank of the Lea and therefore break up all the natural pairings of Tower Hamlets, HAckney, Islington etc. It would make no sense to cross the river Thames at that point and the other alternative would be to link with a seat in Essex (eg Chingford and Epping Forest), but both Essex itself and the Eastern region as a whole have an even number of seats. In fact the five boroughs East of the Lea and North of the Thames have an entitlement of 5.8 (ie 6) seats rather than 5.5 so that problem would be solved by devising entirely new boundaries.
|
|
|
|
Post by Pete Whitehead on Jul 23, 2022 6:53:17 GMT
So, for example, in the Eastern region you would be able to devise a plan something like this which better links natural communities than an arbitrary pairing of existing seats could do  |
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Jul 23, 2022 9:12:34 GMT
Here's how the North West would look like under this system: As mentioned earlier, High Peak is merged with Stalybridge and Hyde to give a total of 38 constituencies, and a corresponding 38 list seats. Five seats have one member each, five have three members each, and the remaining 28 have two members each. You'll notice that every seat has a Labour member, and Liverpool Walton and Bootle has two - this has occurred with the CDU in Baden-Württemberg on several occasions, with the solution being that a substitute is named on the ballot paper along with the candidate, who then takes the second mandate. This total of 39 seats gives Labour a majority in the region, even though they only received 46.5% of the vote - had the threshold been 3%, then they would have lost two seats to the Brexit Party and therefore their majority (the Conservatives also would have lost one, with the Brexit Party picking up three second mandates in Wigan and Makerfield, St Helens, and Rochdale, Heywood and Middleton). If there were no threshold at all then the Green Party would pick up a seat in Chorley and Blackburn at the expense of the Conservatives, but this is down to the Speaker standing in Chorley (their next strongest performance would have been in Liverpool Riverside and Wavertree). On the topic of the Speaker, theoretically he loses out on his seat, as Labour's votes in Blackburn total some 2,000 more than Hoyle's in Chorley. However it is likely that if this system was implemented the Speaker would continue to get a free run from the main parties, or alternatively he might be awarded an automatic second mandate in his constituency, with the number of available list seats in the region decreased by one accordingly. The results (with the 5% threshold) were: Lab: 39 seats (-2 seats vs FPTP), comprising 20 first mandates and 19 second mandates Con: 31 seats (-2 seats vs FPTP), comprising 18 first mandates and 13 second mandates LD: 6 seats (+5 seats vs FPTP), all second mandates One of your seats has two Labour MPs |
|
xenon
Forum Regular
Posts: 426
|
Post by xenon on Jul 23, 2022 9:16:56 GMT
One of your seats has two Labour MPs Yes, this can occur if a party is entitled to more seats than there are constituencies. In this case, you simply go down the list of constituencies ordered by vote share again, and as Labour require one more seat they pick up a second mandate in their strongest constituency of Liverpool Walton and Bootle. |
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Jul 23, 2022 9:22:20 GMT
One of your seats has two Labour MPs Yes, this can occur if a party is entitled to more seats than there are constituencies. In this case, you simply go down the list of constituencies ordered by vote share again, and as Labour require one more seat they pick up a second mandate in their strongest constituency of Liverpool Walton and Bootle. No it can't because there would only be one candidate per party per seat (as in BaWü). Basically it's a flaw in the system |
|
