Why I give my time to the Yes to AV campaign
Wednesday, 13 April 2011 08:58![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
strange_complexNow that my conference paper is done and I am less ludicrously busy, I'm turning my attention firmly back towards the Yes to Fairer Votes campaign. I've written a fair bit on this journal about my involvement with the campaign, but I haven't yet said very much about why I'm so convinced that a change to AV is worthwhile. I did use the example of the Oldham East and Saddleworth by-election back in January to explain why I think AV enhances the dialogue between voters and prospective candidates, which is certainly one good reason for making the change. But there is much more to say than that alone.
I could, of course, write a long rambling post which attempted to cover all of the reasons why I am supporting a change to AV - not to mention the many, many things which are wrong with FPTP (not all of which AV will fix), or with the No to AV campaign. Believe me, there are plenty of arguments to go into, and I've used most of them during the 1000 or more phone-calls which I've now made for the Yes to Fairer Votes campaign. But many of them have also been rehearsed elsewhere. In the more personal context of my own journal, I've decided instead to whittle things down to the single strongest argument which is convincing me to support a change to AV, and focus on writing about that.
So, are you ready for this?
The most straightforward, truthful and accurate statement of why I am campaigning for a Yes vote in the May 5th referendum?
OK - here we go:
That really is my genuine, number one reason for supporting the change. To me, it's the most persuasive argument. Unfortunately, it also isn't an argument I can use when campaigning. That small percentage of the population who have read up on the subject and know what the Condorcet criterion is might well nod sagely and agree with me - and believe me, I've been hanging out with a lot of those sorts of people in the context of the campaign! But most people would just greet me with a blank look. Should you wish to know more, however, read on...
The basic concept of a Condorcet winner is pretty simple, actually. The Condorcet winner is the candidate in an election contest who would beat any of the other candidates in a straight head-to-head competition between just the two of them. It is a measure of popularity, which identifies the candidate whom the majority of the electorate prefer to any one of the other options.
When there are only two contestants in a race, FPTP of course identifies the Condorcet winner by default. Whoever comes first in that scenario is clearly preferred by the majority of the electorate. It ain't rocket science, and it worked perfectly well in the 19th century. But this is the 21st century, and we now live in a multi-party democracy where only 65.1% of the votes cast at the last General Election were for the two largest parties.
The fundamental problem with FPTP is that where election contests feature more than two candidates, its ability to identify the Condorcet winner becomes distinctly unreliable. It still can, and does, do it in plenty of constituencies. But its ability to do so reliably is badly compromised when two or more candidates run on a similar platform, and thus attract the support of similar voters.
I usually steer clear of examples which involve real political parties when I'm campaigning for AV. I believe very strongly that this is an issue which should be decided not on the grounds of partisanship and party advantage, but on what gives the voters the strongest voice and makes our elected representatives most accountable. But in this case, an example based on real-life political parties can help to show how easily FPTP can fail at identifying the Condorcet winner, because we all have a pretty clear idea about how preferences would be likely to transfer between the parties involved. The example I'm going to use is obviously hugely simplified, but I think it can stand as representative of much more complicated situations which are happening regularly all over the country. The numbers are taken from p. 62 of this book (which is excellent, and which I hope to review here shortly), but the text is mine.
Let's start by imagining a simple election contest with only two candidates: Conservative and Labour. We'll start out with a relatively narrow margin between the two, as follows:
So far, so good. FPTP successfully identifies the Condorcet winner - the candidate whom the majority prefer. The Conservative candidate is elected.
Now let's insert a third candidate into the same scenario. It's 1993, UKIP have just formed, and they field a candidate in the same constituency. The existing preferences of the electorate haven't changed - a majority still prefer the Tories to Labour. But they now have more options. Let's say 10% take up the new option, and vote for the UKIP candidate. Of those 10%, most will be former Tory voters, because both parties are predominantly Eurosceptic. So we'll take 8 out of the 10% from the former Conservative vote, and imagine that the remaining 2 are former Labour voters who have strong enough views about the single issue of the EU to change allegiance. That seems realistic enough, right? Result:
So the Labour candidate now wins. But he or she is not the Condorcet winner. We've already seen that in the absence of an option to vote UKIP, a majority of voters in the constituency had expressed a preference for the Conservatives - and this means that the Labour candidate could not have beaten the Conservative candidate in straight a head-to-head race. In fact, the Labour candidate hasn't become any more popular in his or her own right than in the previous election. He or she has simply benefited passively from the introduction of a new option, which a minority of voters preferred to the existing options. Meanwhile, the Conservative candidate could still beat either the Labour or the UKIP candidates in a head-to-head contest. But he or she is not getting the chance to demonstrate that breadth of popularity under the FPTP system.
This, to me, is why looking for the Condorcet winner is important, and a better measure of popularity than simply who gets the most votes in a single-round contest. FPTP doesn't bother to ask any more questions after that first round - but I suspect that in this scenario, both the Conservative and the UKIP voters would agree that it should. One common complaint against AV is that it can mean the person who 'comes second' can win. But the 'second' place which people are usually thinking of when they make this comment relates to the results as they would be ranked by FPTP - i.e. under first preferences alone. The complaint rests on the assumption that FPTP always ranks candidates correctly in the first place. But my whole point is that I don't think it does.
Meanwhile, AV is much better equipped to cope with multi-candidate scenarios like the one we're using here. In fact, what it basically does is work sequentially through the candidates in a series of head-to-head contests, peeling away the least popular candidates one at a time until a final winner emerges.
If we take the same example above, and now treat it as a set of first preferences, AV would recognise that no candidate has yet demonstrated conclusively that they are the Condorcet winner. That's what the 50% line which AV uses is all about. It represents a majority, and if you can achieve that in the first round, then you clearly are the Condorcet winner. You have beaten not only every other candidate individually, but every other candidate collectively. Well done! But if no-one has done that - as they currently don't in 2/3 of parliamentary constituencies - then AV probes the views of the electorate more thoroughly than FPTP can before deciding on a winner.
In our example, AV would tackle the Conservative and UKIP candidates first, since - on first view at least - those are the two least popular candidates in the field. AV holds them up against each other and effectively asks the electorate - OK, which one do you prefer out of these two? The answer is obviously the Conservative candidate, so the UKIP candidate is eliminated. What the process of elimination is actually doing is removing obvious losers in the popularity stakes stage by stage.
Once that's done, AV then allows the collective opinions of the whole electorate to be consulted again about a second pair of candidates. Here, there's only one pair left, so the question is now simply whom the electorate prefer out of the remaining Labour and Conservative candidates. Those who've expressed a first preference for either of them have already told us what that is, so those preferences can remain in place for the second round. Meanwhile, those who initially voted UKIP also get to comment on this round, using their second preferences. These are now re-allocated to the remaining candidates, and the result should be that the Conservative candidate wins - now the clear victor of two successive head-to-head contests with the other candidates.
In all fairness, it needs to be said that AV is not 100% perfect at identifying the Condorcet winner either, and Wikipedia provides an example of how it can fail when one candidate enjoys broad support, but still comes third in the first round. But it is very obviously a great deal better at doing so that FPTP, because the whole point of it is that it works through the candidates on a head-to-head basis, clearing the field of the unpopular candidates, and allowing the remaining contestants to demonstrate the true breadth of their popularity.
So to me, that basic goal of identifying majority consensus is what AV is all about. FPTP can allow MPs to be elected by the support of the largest minority, even if a majority of their constituents actively loathe them, because the views of that majority are never properly tested. AV insures against that by thoroughly examining people's relative preferences, until it has found a deep, broad seam of support which a majority can agree on.
As far as I'm concerned, that is a much more meaningful measure of popularity in an election contest. From a theoretical perspective alone, it seems more democratic to me - and certainly much fairer than the bizarre results which FPTP can yield when dealing with more than two candidates. Meanwhile, by dint of requiring majority support, AV also:
In other words, at a local level AV (or indeed any system which is better at identifying Condorcet winners than FPTP) really should do exactly what the Yes campaign are saying it will do - give voters more of a say, and ensure more accountable MPs. It won't, of course, save the world, or deliver the fairer representation at national level which a properly proportional system could. In fact, Alan Renwick shows on pp. 68-70 of the same book which I referenced above that over the last seven elections AV would probably have delivered results which were on average almost equally disproportionate to those delivered by FPTP - although the disproportionality would have manifested itself in slightly different ways. But the improvements at a local level are enough to make it obviously preferable to FPTP as far as I'm concerned.
So that's me, and those are my real reasons for voting Yes to AV. As I've said, they aren't necessarily the reasons which are most effective in an actual debate. For most people, saying that I prefer AV because it is better at FPTP at identifying the Condorcet winner in each constituency is meaningless. And even if they're prepared to listen to me explaining it, I still have to acknowledge that that will only achieve a relatively modest improvement in the electoral system, and that AV isn't perfect at identifying the Condorcet winner itself anyway. But nonetheless, that is an honest statement of what is convincing me.
What's really important when I debate the issue with people on the phone is that my own core of conviction is solidly in place. From there, I can leave all talk of Condorcet winners behind, and concentrate on the arguments which are actually accessible and persuasive. It's nothing like as difficult as this post might make it appear. :-)
Click here if you would like view this entry in light text on a dark background.
I could, of course, write a long rambling post which attempted to cover all of the reasons why I am supporting a change to AV - not to mention the many, many things which are wrong with FPTP (not all of which AV will fix), or with the No to AV campaign. Believe me, there are plenty of arguments to go into, and I've used most of them during the 1000 or more phone-calls which I've now made for the Yes to Fairer Votes campaign. But many of them have also been rehearsed elsewhere. In the more personal context of my own journal, I've decided instead to whittle things down to the single strongest argument which is convincing me to support a change to AV, and focus on writing about that.
So, are you ready for this?
The most straightforward, truthful and accurate statement of why I am campaigning for a Yes vote in the May 5th referendum?
OK - here we go:
The Alternative Vote is better than First Past the Post at identifying the Condorcet winner in each constituency election contest
That really is my genuine, number one reason for supporting the change. To me, it's the most persuasive argument. Unfortunately, it also isn't an argument I can use when campaigning. That small percentage of the population who have read up on the subject and know what the Condorcet criterion is might well nod sagely and agree with me - and believe me, I've been hanging out with a lot of those sorts of people in the context of the campaign! But most people would just greet me with a blank look. Should you wish to know more, however, read on...
The basic concept of a Condorcet winner is pretty simple, actually. The Condorcet winner is the candidate in an election contest who would beat any of the other candidates in a straight head-to-head competition between just the two of them. It is a measure of popularity, which identifies the candidate whom the majority of the electorate prefer to any one of the other options.
When there are only two contestants in a race, FPTP of course identifies the Condorcet winner by default. Whoever comes first in that scenario is clearly preferred by the majority of the electorate. It ain't rocket science, and it worked perfectly well in the 19th century. But this is the 21st century, and we now live in a multi-party democracy where only 65.1% of the votes cast at the last General Election were for the two largest parties.
The fundamental problem with FPTP is that where election contests feature more than two candidates, its ability to identify the Condorcet winner becomes distinctly unreliable. It still can, and does, do it in plenty of constituencies. But its ability to do so reliably is badly compromised when two or more candidates run on a similar platform, and thus attract the support of similar voters.
I usually steer clear of examples which involve real political parties when I'm campaigning for AV. I believe very strongly that this is an issue which should be decided not on the grounds of partisanship and party advantage, but on what gives the voters the strongest voice and makes our elected representatives most accountable. But in this case, an example based on real-life political parties can help to show how easily FPTP can fail at identifying the Condorcet winner, because we all have a pretty clear idea about how preferences would be likely to transfer between the parties involved. The example I'm going to use is obviously hugely simplified, but I think it can stand as representative of much more complicated situations which are happening regularly all over the country. The numbers are taken from p. 62 of this book (which is excellent, and which I hope to review here shortly), but the text is mine.
Let's start by imagining a simple election contest with only two candidates: Conservative and Labour. We'll start out with a relatively narrow margin between the two, as follows:
Conservative 52%
Labour 48%
So far, so good. FPTP successfully identifies the Condorcet winner - the candidate whom the majority prefer. The Conservative candidate is elected.
Now let's insert a third candidate into the same scenario. It's 1993, UKIP have just formed, and they field a candidate in the same constituency. The existing preferences of the electorate haven't changed - a majority still prefer the Tories to Labour. But they now have more options. Let's say 10% take up the new option, and vote for the UKIP candidate. Of those 10%, most will be former Tory voters, because both parties are predominantly Eurosceptic. So we'll take 8 out of the 10% from the former Conservative vote, and imagine that the remaining 2 are former Labour voters who have strong enough views about the single issue of the EU to change allegiance. That seems realistic enough, right? Result:
Labour 46%
Conservative 44%
UKIP 10%
So the Labour candidate now wins. But he or she is not the Condorcet winner. We've already seen that in the absence of an option to vote UKIP, a majority of voters in the constituency had expressed a preference for the Conservatives - and this means that the Labour candidate could not have beaten the Conservative candidate in straight a head-to-head race. In fact, the Labour candidate hasn't become any more popular in his or her own right than in the previous election. He or she has simply benefited passively from the introduction of a new option, which a minority of voters preferred to the existing options. Meanwhile, the Conservative candidate could still beat either the Labour or the UKIP candidates in a head-to-head contest. But he or she is not getting the chance to demonstrate that breadth of popularity under the FPTP system.
This, to me, is why looking for the Condorcet winner is important, and a better measure of popularity than simply who gets the most votes in a single-round contest. FPTP doesn't bother to ask any more questions after that first round - but I suspect that in this scenario, both the Conservative and the UKIP voters would agree that it should. One common complaint against AV is that it can mean the person who 'comes second' can win. But the 'second' place which people are usually thinking of when they make this comment relates to the results as they would be ranked by FPTP - i.e. under first preferences alone. The complaint rests on the assumption that FPTP always ranks candidates correctly in the first place. But my whole point is that I don't think it does.
Meanwhile, AV is much better equipped to cope with multi-candidate scenarios like the one we're using here. In fact, what it basically does is work sequentially through the candidates in a series of head-to-head contests, peeling away the least popular candidates one at a time until a final winner emerges.
If we take the same example above, and now treat it as a set of first preferences, AV would recognise that no candidate has yet demonstrated conclusively that they are the Condorcet winner. That's what the 50% line which AV uses is all about. It represents a majority, and if you can achieve that in the first round, then you clearly are the Condorcet winner. You have beaten not only every other candidate individually, but every other candidate collectively. Well done! But if no-one has done that - as they currently don't in 2/3 of parliamentary constituencies - then AV probes the views of the electorate more thoroughly than FPTP can before deciding on a winner.
In our example, AV would tackle the Conservative and UKIP candidates first, since - on first view at least - those are the two least popular candidates in the field. AV holds them up against each other and effectively asks the electorate - OK, which one do you prefer out of these two? The answer is obviously the Conservative candidate, so the UKIP candidate is eliminated. What the process of elimination is actually doing is removing obvious losers in the popularity stakes stage by stage.
Once that's done, AV then allows the collective opinions of the whole electorate to be consulted again about a second pair of candidates. Here, there's only one pair left, so the question is now simply whom the electorate prefer out of the remaining Labour and Conservative candidates. Those who've expressed a first preference for either of them have already told us what that is, so those preferences can remain in place for the second round. Meanwhile, those who initially voted UKIP also get to comment on this round, using their second preferences. These are now re-allocated to the remaining candidates, and the result should be that the Conservative candidate wins - now the clear victor of two successive head-to-head contests with the other candidates.
In all fairness, it needs to be said that AV is not 100% perfect at identifying the Condorcet winner either, and Wikipedia provides an example of how it can fail when one candidate enjoys broad support, but still comes third in the first round. But it is very obviously a great deal better at doing so that FPTP, because the whole point of it is that it works through the candidates on a head-to-head basis, clearing the field of the unpopular candidates, and allowing the remaining contestants to demonstrate the true breadth of their popularity.
So to me, that basic goal of identifying majority consensus is what AV is all about. FPTP can allow MPs to be elected by the support of the largest minority, even if a majority of their constituents actively loathe them, because the views of that majority are never properly tested. AV insures against that by thoroughly examining people's relative preferences, until it has found a deep, broad seam of support which a majority can agree on.
As far as I'm concerned, that is a much more meaningful measure of popularity in an election contest. From a theoretical perspective alone, it seems more democratic to me - and certainly much fairer than the bizarre results which FPTP can yield when dealing with more than two candidates. Meanwhile, by dint of requiring majority support, AV also:
- Rewards candidates who are prepared to reach out and appeal to a wide range of voters, including those who aren't planning to put them as a first preference. This should mean that it favours moderate candidates campaigning on a positive platform, and punishes extremists who simply slag off their rivals.
- Makes the real opinions of the electorate much clearer by removing the obfuscating effects of tactical voting. That was the point I was making in my Oldham East and Saddleworth post, but I think this postcard makes it rather nicely, too.
- Makes some seats contestable which currently aren't, helping to mitigate against the current culture of political parties targeting all their efforts on swing voters in marginal seats, while largely ignoring the views of those who live in safe seats. Of course, the same strategy will still be used, but if a higher proportion of seats are marginal, then a higher proportion of the electorate should find that their opinions are being listened to.
- Makes it harder for MPs who don't do a good job of representing their constituents to get re-elected, because they won't manage to do so if they have alienated a majority of their constituents.
In other words, at a local level AV (or indeed any system which is better at identifying Condorcet winners than FPTP) really should do exactly what the Yes campaign are saying it will do - give voters more of a say, and ensure more accountable MPs. It won't, of course, save the world, or deliver the fairer representation at national level which a properly proportional system could. In fact, Alan Renwick shows on pp. 68-70 of the same book which I referenced above that over the last seven elections AV would probably have delivered results which were on average almost equally disproportionate to those delivered by FPTP - although the disproportionality would have manifested itself in slightly different ways. But the improvements at a local level are enough to make it obviously preferable to FPTP as far as I'm concerned.
So that's me, and those are my real reasons for voting Yes to AV. As I've said, they aren't necessarily the reasons which are most effective in an actual debate. For most people, saying that I prefer AV because it is better at FPTP at identifying the Condorcet winner in each constituency is meaningless. And even if they're prepared to listen to me explaining it, I still have to acknowledge that that will only achieve a relatively modest improvement in the electoral system, and that AV isn't perfect at identifying the Condorcet winner itself anyway. But nonetheless, that is an honest statement of what is convincing me.
What's really important when I debate the issue with people on the phone is that my own core of conviction is solidly in place. From there, I can leave all talk of Condorcet winners behind, and concentrate on the arguments which are actually accessible and persuasive. It's nothing like as difficult as this post might make it appear. :-)
Click here if you would like view this entry in light text on a dark background.
