Robin Grant of perfect.co.uk draws everyone's attention to a meeting on Wednesday on electoral reform at the House of Commons attended by the great and the good (and, perhaps the less great and the merely adequate, though I won't try to tell you which is which):
Speakers will include Chris (Lord) Rennard (Lib Dem Campaign-Manager-guru-chap), Polly Toynbee (Guardian Columnist), Billy Bragg (singer/songrwriter and political activist) and Martin Linton MP (Chair of Make Votes Count's Labour Wing)
(I doubt I'll go but if people want to encourage me to come and heckle -- by means of bribes of alcoholic beverages, or otherwise -- you know where to ask....)
On the same subject, I should also draw your attention to this piece by John Quiggin on `Crooked Timber', which picks up on a related topic I've mentioned before.
And to hang a fragment of content on this otherwise fairly blatant plug, here's how the late General Election would have come out under PFPTP:
(For those who are just joining us and can't be arsed to read my previous post on this, PFTP, ``proportional first-past-the-post'' is a modification to Britain's current electoral system which achieves proportionality not by modifying the electoral system but by modifying the procedure of divisions in the House of Commons. Instead of getting one vote each, each MP gets a vote proportional to their party's support in the country, divided by the number of MPs in that party. Each constituent gets exactly one MP and each MP exactly one constituency, but power in the Commons is proportional to the parties' support in the country as a whole. Pedants will note that I've computed the above table on the basis of the 645 seats which have declared. In Staffordshire South the Lib Dem candidate died during the campaign; the result will be determined by a byelection later. This, if you are [un]lucky may provoke a post on how byelections should work under PFPTP, since there are several plausible answers.)
Note that despite various bizarre features of this election, none of the parties' voting weights are too out-of-whack. Lady Sylvia Hermon, the remains of the Parliamentary UUP, would get just over three votes, against a bit under six for their nemesis, the DUP. George Galloway picks up a little extra influence from a closely-fought campaign in Bethnal Green and Bow, and support elsewhere (he won about 16,000 votes, while the party as a whole got just over 68,000). Richard Taylor, as befitting a single-issue candidate in one constituency, commands rather little influence: about half a vote, less even than that of a Labour MP.
As expected, this result would mean that, united, any two of the Conservatives, Labour and Lib Dems can between them command a majority in the Commons; no other combination of parties can. Of course, once MPs start rebelling (the rotters!) anything could happen -- but what's new?
That concludes today's test of the Bonkers Electoral Reform Broadcasting System. I'll try to write about something interesting -- or, failing that, bloody ID cards again -- in the near future.
Update: Simon Keal (see comment, below) corrected me on the name of the Kidderminster Hospitals MP -- I had incorrectly called him Richard Thomas. Oops. He also drew my attention to the inexplicable absence of an updated plot of Anne Campbell's Parliamentary Majority. Now, arguably, as an ex-MP Anne is not a public figure and so such taking the piss is out of order, but (a) I've never been one to turn down a cheap laugh, and (b) she did vote for ID cards and the abolition of habeas corpus, so sod that:
Comments
Posted by James Fairbairn, Tuesday, 10 May 2005 22:43 (link):
I know this issue was discussed in the comments on your original PFPTP post, but...
...apart from those that didn't win a single seat. If one of the virtues of any kind of proportional representation is fairer a representation of minorities, then this "rounding error" at the bottom end should probably be fixed somehow. (Disclaimer: I would very much like the Green Party to have 6.somethingish votes, which they would have won this time under a "true" proportional system.)
Posted by Chris Lightfoot, Tuesday, 10 May 2005 23:14 (link):
Well, most `proportional' systems impose cutoffs at the lower end to keep the nutters out, or whatever the politically-correct way to put that is. At least the constituency cutoff in PFPTP is a natural consequence of the system rather than an arbitrary cutoff. (Actually that's not quite true, in that the size of a constituency is an arbitrary parameter, but it still seems preferable to picking, say, 5%, and running with it.)
That said, I didn't start from a list of requirements and come up with the system, but rather by considering small modifications to the existing system. So it's not very surprising that it shares most features with FPTP....
Posted by Simon Keal, Wednesday, 11 May 2005 00:41 (link):
The Kidderminster Hospital chap is Richard Taylor, not Richard Thomas.
I was also rather hoping for a new chart indicating what happened to Anne Campbell's parliamentary majority in 2005. Could you oblige?
Posted by Chris Lightfoot, Wednesday, 11 May 2005 02:06 (link):
Oops! Now corrected. I've also added the graph you asked for....
Posted by Pete Stevens, Wednesday, 11 May 2005 09:06 (link):
I suspect you'd do better at persuading people if you multiplied everyones weight by the total number of votes cast. Smart people will notice this makes no difference. Not so smart people will appreciate that all fractions and percentages have been removed from the system, and everything reverts to a matter of counting which people in general can understand.
So instead of a Labour MP counting for 0.638% of an MP - which people will claim is unfair since they voted for a whole MP - you'd say,
The labour party got 9,556,183 votes and have 356 seats, so each MP is worth 26843 votes, etc. etc.
Posted by Liz Upton, Wednesday, 11 May 2005 11:24 (link):
>I suspect you'd do better at persuading people if you multiplied everyone's weight by the total number of votes cast.
Surely this means than Nicholas Soames would have become King-Emperor last Thursday, and that nobody else would have had a look-in?
Posted by Chris Lightfoot, Wednesday, 11 May 2005 13:16 (link):
gah! On a tangentially related note, I see that Soames has resigned from the Shadow Cabinet, ``to spend more time with his lunch'', as Nick Barlow put it.
Posted by Chris Lightfoot, Wednesday, 11 May 2005 17:56 (link):
Yes, that is better. Noted.
(NB it's 0.638 votes, not 0.638% of a vote, but as you point out the numbers are pretty arbitrary.)
Posted by Paul Davies, Wednesday, 11 May 2005 12:27 (link):
Not sure UKIP would be too happy - easily the forth largest party but no MPs. Conversely, the DUP have 9 MPs, as theirs is a local vote for local people
Posted by Chris Lightfoot, Wednesday, 11 May 2005 18:14 (link):
Well, that's life, and certainly no worse than we have now. I'm not sure that UKIP would have picked up any seats under STV in single-member constituencies either, which seems to be the usual proposal for those who want to maintain a one-to-one mapping between constituencies and MPs (essential, as Martin points out, if you want constituents to be able to fire their MPs).
Posted by Roy Badami, Wednesday, 11 May 2005 23:48 (link):
Is anyone proposing that? It solves (some of) the problems of tactical voting (but the electorate seem to be pretty good at doing that themselves when they need to) but has nothing to do with PR...
Posted by Peter Clay, Thursday, 12 May 2005 01:19 (link):
Well, if we must change the voting system then that's what I'd prefer. I'm writing a longer post to explain my thoughts on PR, but one hit-and-run point: tactical voting doesn't vanish, it just moves it up to the house; parties must necessarily vote compromise and vote for their second-preference in order to get anything done at all.
Posted by Roy Badami, Thursday, 12 May 2005 22:06 (link):
If we're going to stay with single-member consituencies, and have no PR, then AV (aka IRV, aka STV in single-member constituencies) is not what I'd choose. You want a system that is guaranteed to elect the Condorcet winner when there is one -- that is, the candidate who would win a run-off against any other single candidate. Obviously, no such candidate might exist, but conventional wisdom is that a single-winner election should always elect such a candidate when they exist. AIUI, no conventional formulation of STV guarantees this.
The problem with STV is the elimination round once all the surplusses are allocated. The elimination of the candidate with least votes is essentially an arbitrary hack to make the counting feasible. The problem is that it is eminently possible to eliminate a candidate who should, by rights, win in a subsequent round, because lots of votes would be transfered to them by subsequent eliminations. But because of their standing at this round of the count they are eliminated and never considered again.
See www.electionmethods.org for further discussion.
-roy
Posted by Martin Keegan, Friday, 13 May 2005 12:15 (link):
I'm glad to hear this argument against AV, as it implies that no better argument has yet been found and publicised.
The theoretical problem you're talking about does not occur in practice. There aren't all these Australian politicians going around claiming they would have won but for some irrelevant alternative candidate's presence in the election.
Maybe in the UK this would be more of a problem than in Australia, as there are three-way marginals here between non-cooperating parties, but this starts to sound suspiciously like yet another debating point about UK electoral systems seen through the prism of the interests of the Liberal Democrat Party.
Posted by Roy Badami, Friday, 13 May 2005 17:43 (link):
I think the argument goes like this:
If you accept that you have a two party system, and will always have a two party system, you may as well stay with FPTP, and let people vote for their whichever of the two main parties they prefer, just as they do now.
If your rationale for changing the electoral system is that the current system disadvantages third parties, and you think that is a bad thing, then you should probably choose a system that will continue to work in a 3-party system...
-roy
Posted by Martin Keegan, Sunday, 15 May 2005 23:54 (link):
You've not explained how the number of major parties interacts with this Condorcet winner problem you claim to have identified, but to deal with the new point you raise about motivations for electoral change, I would stress very firmly indeed the point that FPTP does not disadvantage "third" parties or entrench a two-party system per se. The number of major parties thrown up by FPTP is contingent on the geographic distributional efficiency of party supporters, not some arbitrary fixed number like "two". Canada is a good example of this.
Again I infer and impute that this debate is being viewed through strange orangey-yellow tinted spectacles.
Posted by Roy Badami, Monday, 16 May 2005 19:45 (link):
You've not explained how the number of major parties interacts with this Condorcet winner problem you claim to have identified
The suggestion that FPTP entrenches a two party system is a reason sometimes cited for electoral reform, that is all.
I still remain unclear as to the benefits of AV, though (though it is tolerable as part of a PR system such as AV+).
My main point is that AV only works well in a two party system, where you can safely cast a sincere vote, and know that your vote will still be counted towards you're prefered of the two main parties. Once a minor party starts to challenge the two big parties, there's a need to vote tactically -- first place votes are everything, because they determine the order of elimination.
This is a particular manifestation of a well known problem with STV; namely that second (and subsequent) place votes for a candidate that has already been eliminated can never be counted. There are modifications of STV that try to fix this, but they are all incredibly complicated, and arguably more suitable for multi-seat elections.
Again I infer and impute that this debate is being viewed through strange orangey-yellow tinted spectacles.
Infer what you will. My interest is simply in a fair electoral system, though I realise that what consitutes fair is highly subjective. I hope you're not suggesting that any electoral reform that benefits the Lib Dems is intrinisically unacceptable?
-roy
Posted by Roy Badami, Friday, 13 May 2005 18:01 (link):
Incidentally, the problem may well occur in practice, in a three party system. As long as you have a two-party system it does not.
Consider three parties, Left, Centre and Right.
Imagine the votes spit approximately 45% Left, 20% Centre and 35% Right, and that the vast majority of the Left voters rank the candidates:
1. Left 2. Centre 3. Right
and the vast majority of the Right voters rank the candidates in the reverse order. The Centre voters are split as to which of Left and Right they list in second place.
In STV, no party meets the Droop quota in the first stage count, so Centre will be eliminated as the weekest candidate. The Centre voters second choice will effectively decide the election.
However, the Condorcet winner is Centre. If you ask the population, would they prefer Left or Centre, 55% will say Centre (the 20% Centre voters plus the 35% Right voters). If you ask the population, would they prefer Right or Centre, 65% will say Centre (the 20% Centre voters, plus the 45% Left voters).
So STV will elect Left or Right, but whichever it elects a majority of the electorate would have prefered Centre to be elected.
Yes, this will help centre parties. But why is this a bad thing?
-roy
Posted by Martin Keegan, Monday, 16 May 2005 00:58 (link):
Sorry, I am starting to lose the distinction between Condorcet winner and Centre candidate here.
Is your argument that AV should be opposed because it won't favour Centre candidates?
Posted by Roy Badami, Monday, 16 May 2005 19:34 (link):
Is your argument that AV should be opposed because it won't favour Centre candidates?
No. My argument is that if an electoral system elects candidate A, and there exists candidate B such that the majority of the electorate prefer candidate B to candidate A, then the electoral system is flawed.
Any electoral system that can elect a candidate outside the Smith set can suffer from this.
It is certainly true that where support is concentrated on two parties at oposite ends of the political spectrum, and there is also a fairly popular centre party, Condorcet voting is more likely to give power to the centre party... More generally, Condorcet voting is more likely to elect compromise candidates, that are tolerable to the entire electorate, rather than candidates that polarise opinion.
This would certainly have significantly benefited the Liberals in the old days; I'm not really sure what the effect would be now, with many people perceiving the Liberal Democrats as being to the left of New Labour.
-roy
Posted by Roy Badami, Monday, 16 May 2005 20:30 (link):
Actually, I didn't mean to say quite that, because it's unavoidable in the case of a voting paradox.
But when it is avoidable, it's surely desirable?
Posted by Peter Clay, Friday, 13 May 2005 18:13 (link):
Do people want a system that elects the Condorcet winner? The complaints I hear from non-experts seem to be of three forms:
Unfortunately, the first two can't be implemented without weakening the consituency link and transferring power from voters to party list selectors. The third would be fixed in practice by AV with sensible voters (who put their minority party choice first). The minorities would be eliminated as everyone expects - unless they've turned into a serious contender.
I'm trying to understand Condorcet; it seems that given 5 ballots (ranks left-to-right):
2x CON-LD-LAB 1x LD-CON-LAB 2X LAB-LD-CON
I think LD is the Condorcet winner but CON are the AV winners. It would be instructive for someone to do a survey of example ballot sets like this and see who people intuitively think should win, then reverse-engineer a voting system from that. It might make for nice graphs ;)
Posted by Roy Badami, Friday, 13 May 2005 20:39 (link):
I think LD is the Condorcet winner but CON are the AV winners.
Correct.
Do people want a system that elects the Condorcet winner?
Consider your example. Imagine that you, as an AV advocate, conduct an AV count and declare CON duely elected.
I, as a Condorcet advocate can then say to the electorate Wouldn't you have prefered it if LD had been elected instead? and a majority (60%) would say yes. So in some sense LD is what the people want.
Whether it's a useful sense is open to debate, of course.
-roy
Posted by Roy Badami, Saturday, 14 May 2005 01:28 (link):
The minorities would be eliminated as everyone expects - unless they've turned into a serious contender.
Indeed. But the people at electionmethods.org argue that this only works whilst the minor parties are unelectable, and that things go horribly wrong when a minor party becomes a serious contendor.
Unfortunately they don't give an example with numbers, and I was beginning to be somewhat sceptical. But consider the following scenario:
14 votes: 1. Con 2. Lab 3. LD --(a)
10 votes: 1. Lab 2. Con 3. LD --(b)
8 votes: 1. LD 2. Lab 3. Con --(c)
3 Votes: 1. LD 2. Con 3. Lab --(d)
We have 35 voters, so the Droop quota is 18.
Stage 1: No candidate reaches quota; eliminate Lab.
Stage 2: Con has 24 votes; elected by quota
Now, consider that voters (c) were aware of this possiblity, and chose to vote tactically for Lab to keep Con out:
Now we have:
14 votes: 1. Con 2. Lab 3. LD --(a)
10 votes: 1. Lab 2. Con 3. LD --(b)
8 votes: 1. Lab 2. LD 3. Con --(c)
3 Votes: 1. LD 2. Con 3. Lab --(d)
Stage 1: Labour have 18 votes, and are elected by quota.
(In this example, even if voters (d) voted tactically for Con, Lab would still win.)
So, if there is any possibility that your symbolic first place vote might not be for one of the weakest parties -- in particular if there's any possibility it might beat your second place choice -- it's in your interests to vote tactically, and essentially pretend you just have FPTP. Once everyone starts doing this, you may as well just have FPTP.
I don't know how plausible these scenarios are in practice, but I'm inclined to give at least some credence to the suggestion that AV only works in a two party system, and as soon as the third party becomes significant, you have an incentive to revert to FPTP tactical voting...
But the argument is that AV only works when you're casting a symbolic vote for a party you're confident is going to be eliminated early on. As soon as it might make a difference, you probably want to vote tactically.
-roy
Posted by Peter Clay, Monday, 16 May 2005 22:25 (link):
I still don't understand what's so bad about tactical voting by the public. Could you explain your opposition to it?
Posted by Roy Badami, Tuesday, 17 May 2005 22:34 (link):
I don't particularly like a system that requires electors to cast a tactical vote to gain their prefered result, but that isn't a point I was arguing here...
Alegedly the benefit of AV is that it removes the need for tactical voting, and I don't believe that AV in practice achieves this.
My point here is whether AV is an appropriate way to count ranked ballots, given it fails to achieve the usually stated objective for moving to ranked ballots.
-roy
Posted by Martin Keegan, Tuesday, 24 May 2005 14:31 (link):
Why don't you believe that AV in practice "achieves" the "need" for removing tactical voting, or present any actual evidence from countries with an established AV system (or even fleeting experimentation with it) that this isn't the case?
I asked you for this before, but you didn't come up with anything, though you did say you were concerned for a fair electoral system. I think I must concede that to be democratic enough for my liking, an electoral system has to be a little bit unfair.
Posted by Roy Badami, Wednesday, 25 May 2005 00:16 (link):
I don't claim to be an expert on this, but I find the arguments on against AV reasonably convincing... I accept that they are largely theoretical -- how many countries are there (if any) that use AV and don't have a two-party system?
I could live with AV for constituency elections -- indeed I used to be a proponent of AV, but I've now become convinced of it's flaws in a three party race. I should add that Lord Alexander attached a note of reserveration to the report of the Jenkins Commission for (as far as I can tell) precisely this reason.
As I said, though, I can live with AV for constituency elections since it's better than what we've got, and the odd perverse result won't skew parliament too much. It's just that there are other ways of counting results in single-winner elections that seem far more satisfactory to me....
Of course, the people at electionmethods.org are mainly talking about presidential elections, where an occaisional perverse result is arguably rather more problematic... (Though the possibility of the US being anything other than a two party system in the foreseeable future seems rather remote, but that's by the by...)
-roy
Posted by Peter Clay, Wednesday, 25 May 2005 12:51 (link):
Australia uses AV and has a three-main-party system with influential minor parties, where the Liberal and National parties operate in coalition at Federal level. See http://www.aceproject.org/main/english/es/esy_au.htm, also http://www.aceproject.org/main/english/es/esy_pg.htm is interesting.
I've seen claims that the French two-round presidential election system is very similar to an AV election where the voters get to write down their second preferences after the first are counted. That certainly increased the electoral support for Le Pen over what would have happened in an American-style presidential election. I'd really like to see what would happen in a US presidential election that used a non-FPTP system.
Posted by Roy Badami, Monday, 13 June 2005 21:55 (link):
I'd be interested to know how this works out in practice.
My expectation is that if the Liberals and Nationals have a history of coalition, then voters will think in terms of voting for the coalition rather than specifically for the Liberals or Nationals, so I would expect most voters to rank Liberals and Nationals consecutively - or at least not to rank Labour between them.
If that's the case, then the flaws in AV will never affect the balance of power between Labour and the coalition, though it could well affect the relative strengths of Liberals and Nationals. If there's any significant policy difference between Liberals and Nationals then I might also expect some Labour voters to cast a tactical first preference vote for their prefered of these parties in safe coalition seats, since if they vote sincerly, their second place votes will not be transferred [1].
If voting patterns were ever such that splitting the first preference vote damaged the coalition's electoral success, then I would expect to see an arrangement where only one coalition party contested each marginal seat, but such voting patters are probably unlikely.
-roy
[1] Consider a safe coalition seat where the coalition vote is fairly evenly split between Liberals and Nationals, but Labour is the strongest single party. For simplicity assume that no minor parties are contesting the seat, and that voters split into Labour supporters (who rank Labour first) and coalition supporters (who rank their prefered coalition party first, and the other one second).
Labour get the largest number of first preference votes, but fail to get 50%, so the weaker of the coalition parties is eliminated, and their votes transfered to the other coalition party which is then duely elected.
Note that the Labour supporters have no say in which of Liberals and Nationals are elected, theire votes are not transfered. To influence the outcome in this safe coalition seat a Labour supporter would have to cast a tactical vote.
Posted by Martin Keegan, Thursday, 2 June 2005 22:48 (link):
Does Australia have a two-party system? And if it does, is that the result of the electoral system or the self-destruction of the Democrat Party and the absence of a real support base for anything other than Catholic corporatism and fiscal conservatism?
In Australia minor parties formally command influence over policy through preference deals, so examining the number of parties represented in the federal Lower House is misleading.
Posted by Roy Badami, Wednesday, 25 May 2005 00:17 (link):
I think I must concede that to be democratic enough for my liking, an electoral system has to be a little bit unfair.
Would you care to elaborate?
Posted by Martin Keegan, Thursday, 2 June 2005 22:51 (link):
Oh, all I mean to say is that I regard the presence in the Lower House of anything other than representatives of single member constituencies to be undemocratic (or insufficiently democratic). Obviously, excluding PR systems, STV, and top-up systems and so on, which tend to be fairer, is the price one pays for this.
Post a new comment.
Comments copyright (c) contributors and available under a Creative Commons License. See also the comments policy.