Sometimes, interesting discussions sprout in the comments that I fear will seem more buried than planted, if not re-propogated here in the main orchard.
Several of us have been discussing the merits of the Alternative Vote (AV), one of several formulas that might fit under the rubric of what American reformers mean by “Instant Runoff Voting” (IRV).
Just this week, I received an e-mail from a political science contact (based outside the USA) who said that he “loathes AV.” At first, I thought the comment a tad harsh. But the more I think about, the more I wonder if it might not fit my own views, with respect to the suitability of the system for US legislative elections (or for the nominal tier of potential MMP systems).
In the thread on Electoral Reform in Canada Ed made the following observation:
I’m also impressed by the evidence that the effect of AV is to reduce the plurality of voices and parties in the legislature. I used to support use of AV over FTFP until I looked more closely at Australian elections.
Its interesting that the use of runoffs have not had quite the same effect in France.
To which I responded:
Of course, in France there was an existing very fragmented party system into which a two-round system was (re-)introduced, in 1958.
All path dependency aside, there are good logical reasons to expect that a two-round system, especially of the majority-plurality variant used in France, would tend to support a multiparty system, but AV (IRV) would not.
When there is an actual second round of voting if no candidate has won a majority of first votes, parties have much more opportunity to enter as “spoilers” (and all the more so, again, when the runoff is a restricted plurality rule and not 50%, plus 1).
Advocates of AV/IRV often favor it because it avoids spoilers. Yes, and perhaps too well.
This latter comment moved Bob to ask:
By “too well” do you mean that reducing spoiled elections reduces the effect of small party and independent candidates on outcomes? If so, then that’s a good thing for the small parties themselves, because spoiled elections are what prevent people who support these candidates from actually voting for them. Or do you mean that small party and independent candidates win less often? If so, then less often that under what other voting rule(s)?
My response to Bob’s question–and now this is something new to this current planting–is that, regarding his two ways of possibly interpreting what I said about AV dealing too effectively with the “spoiler” problem, I would endorse the first one as closer to my view.
I take a pretty Machiavellian view of interparty dynamics under winner-take-all systems (whether FPTP, AV, two-round, the absurd list-plurality system of the absurd US electoral college, or what have you). I am pretty sure large parties take such a view themselves, when they bother to be worried about the groupuscules that, in most US legislative elections, pass for parties other than the dominant Two. Large parties will take note of smaller when the latter threaten the former. So, go and spoil if you are serious about increasing the influence of small parties. As I have suggested before, that is the most likely route to real electoral reform–a form of PR (which is not, by any means, to suggest that it inevitably leads there).
It is clear from the experience of most FPTP cases that (certain types of) small parties can win seats under FPTP, even if they tend to be under-represented, sometimes seriously (but sometimes not). Of course, the USA is not such a system. It has a 2-party system that is even more solidly so than that for Australia’s AV-elected first chamber.
In the US context, that might imply that AV (or another form of IRV) would be a step forward for pluralism in our legislative bodies. I doubt it, though it is possible. I suspect that it is more likely that AV would enhance the role of single-issue organizations that could make a claim to be able to determine which candidates won through following the preference trail. That is, we might see more candidates, but I wonder about more parties, in a form that is recognizably partisan. If single-issue organizations were more institutionalized in US elections, that would hardly be a step forward.
Maybe my views of smaller US parties and AV are too bleak. I don’t know, honestly. But I am very skeptical of the passion that many reformers have for AV/IRV for US House or state-legislative races. I am actually somewhat agnostic about FPTP vs. AV for these types of contests.* I just do not feel that the difference between them is worth getting too excited over, even if the balance of the comparison is favorable to AV. Which, obviously, I am yet to be persuaded it is.
Beyond that, I’ll just say “what Ed said.” And, so that you do not have to go a-clicking, I will let Ed have the more-or-less last word in this planting (for now). Here is what Ed said:
…the evidence from Australian House of Representatives is pretty clear.
Minor parties such as the Greens, Australian Democrats, and Democratic Labor have existed in Australia and elected candidates to the Senate, which uses STV. None of these parties have ever won a House of Representatives seat in a general election, or come even close. The National/ Country Party has won seats in coalition with the Liberals, though the alliance is so close there is reason not to treat Liberals and National as separate parties.
First-Past-the-Post elected legislatures such as Canada, New Zealand (before the switch) and the UK have all had significant third party representation, from both national third parties and regional third parties. Even in the case of the US House of Representatives the Socialists have won a couple of seats. The PDS, the Greens, and I think also the FDP have won Bundesrat districts at various times. [FDP, I think not; certainly not in recent decades--MSS]
So the record is pretty clear. This could be due to cultural reasons unique to Australia, though its hard to see what the are. Minor parties in Australia seem to be much more accepted than in the US. It could be due to the failure of minor party leaders to cultivate regional bases of support, though the dynamics of AV would encourage that, as these parties can exert influence through second preferences without actually winning a seat in the House. I suspect voters may not want to be in a situation where a minor party is a “finalist” for a seat. Mathematically, its hard for a party polling 10% to get enough a deviation in any one district to get over 50%, but they might reach the 30% mark.
I hope we might have a visitor or two with actual experience voting in AV elections stop by later.
(I still need to address the question of the incompatibility of AV and MMP, which has come up in another thread. I’ll get to it–promise.)
______
* On the other hand, for replacing two-round majority elections at the municipal level, especially in the case of officially non-partisan contests, the superiority of AV is clear to me. By the same token, it is obviously superior to the “top two” proposal (which would replace the partisan primary and restrict general election ballots to just two candidates, even if of the same party) being floated in California, but then so is the existing two-stage FPTP (once in the primary, once in the general).
First, two (somewhat rhetorical) questions . . .
Didn’t AV in Australia precede STV? If so, doesn’t that give some evidence to the argument that it is a stepping-stone to STV? It seems to me that AV accustoms people to ranked ballots, thereby making STV more likely. You may be agnostic as to FPTP vs AV as an end in itself, but it’s worth considering what AV in turns makes politically feasible.
What about FPTP’s impact of limiting the candidates that appear in US televised debates? Perhaps more than anywhere else in the world, third party candidates in the U.S. are completely shut out of televised debates due to collusion between the major parties. If AV were in place, the “need” and justification for excluding third party candidates would be significantly reduced. This is a crucial platform that third parties and their candidates need to get their ideas heard and to grow organically. I suspect this is less of an issue in other countries with more independent media.
Now some commentary . . .
I am a big supporter of IRV/AV, but I do not believe that it, by itself, increases the number of parties, nor have I, for one, ever claimed that. What it allows is for third parties to *participate* without being continually ridiculed and dismissed as spoilers. AV allows them to increase: their share of the vote, their likelihood of getting in the debates, their visibility, and their influence over the national debate. This could allow them to grow, organically, over a number of years and eventually set the stage for PR. Even if we don’t reach PR, with AV we would at least have a more diverse, robust debate of the issues.
Seed planted by Greg — 15 July 2009 @ 22:29
Greg’s states very well the Conventional Wisdom among electoral reform activists in the United States. But suppose that to some highly debatable extent the CW is wrong.
Suppose, for example, that instead of increasing the media attention given to alternative candidates and the influence of their ideas, AV would actually reduce their influence because, as MSS puts it, the two leading candidates only have to pay attention when the spoiler might actually spoil things. I have to believe that reducing the penalty for sincere voting would increase the number of people who vote sincerely, thereby revealing that there there are more Greens, Libertarians, Constitutional fundamentalists, and Socialists among us than appears to be the case now. But suppose not, for the reasons MSS outlines.
A couple of other preliminary comments:
(1) As MSS notes in italics, this discussion is about SMD elections for Congress and the state legislatures. The U.S. has a lot more elected mayors, governors and presidents (and sheriffs, district attorneys, county clerks, and dogcatchers) than just about any place else. That fact gives AV a relevance here that it doesn’t necessarily have elsewhere. This is somewhat mitigated by the fact that many of the minor offices are rarely contested.
(2) MSS distinguishes between AV and IRV: the Alternative Vote (AV), one of several formulas that might fit under the rubric of what American reformers mean by “Instant Runoff Voting” (IRV). Few activists make this distinction, except in one respect. In FairVote’s usage, IRV includes the contingent vote and the supplementary vote as well as AV. (There is one city where the term has been expanded to include STV, but that’s an exception.) I don’t know whether the same terminological problem exists in other countries. I do know that when the Australians talk about AV they are fairly careful to call it “majority preferential voting” to distinguish it from STV.
(3) To Greg’s summary of how small parties might grow from obscurity to influence, I’d like to add one more crucial factor. They have to be able to put up qualified candidates — people who could perform in office if elected. As a result of their marginalization, they mostly do not, a fact which reinforces the marginalization. To work the magic attributed to it by the CW, the alternative vote would have to be a factor contributing to increased candidate turnout as well as increased sincere voting. It’s an article of faith among activists that AV encourages people to run for office who don’t run now because they know they would be spoilers.
Seed planted by Bob Richard — 15 July 2009 @ 23:32
With the greatest possible respect, you need to expand your data-sets both by looking at state results and by rethinking your definition of the National party.
The Nationals are in a pre-electoral coalition with Labor in South Australia and in a post-electoral coalition with the Liberals in Western Australia. 6% of the present NSW legislative assembly, for instance, are independents. In 1998 One Nation won 11 of the 78 seats in the Queensland legislative assembly. Even in the current House of Representatives 2% are independents. The ERS report (which admittedly also seems exclusively focused on the federal parliament) states again and again that you cannot make direct extrapolations from the Australian experience,
I do not think AV is a good path to proportional representation, but the diversity argument advanced in this thread is simply inconsistent with the data. I do think an AV chamber is better than an FPTP chamber.
Seed planted by Alan — 16 July 2009 @ 00:39
AV/IRV is all well and good for offices (executive offices, say) where a majority-seeking method is called for.
For legislatures, and for “plurality” representation in general, not so much. But it seems to me that it’s hard to defend FPTP as being any better.
In the recent MN-Sen Franken-Coleman-Barkley race, the results suggest that, if we conjecture that roughly half of Barkley’s supporters voted strategically for Franken or Coleman to avoid “wasting” their FPTP vote, then Barkley would have been a likely AV/IRV winner.
AV/IRV significantly reduces (though does not eliminate) the incentive to vote insincerely. I see that as a significant advantage over FPTP.
Seed planted by Jonathan Lundell — 16 July 2009 @ 00:47
- Peter Brent, “Sprucing up the horse and buggy: New recommendations for electoral reform are a step in the right direction, but the pace continues to be perplexingly slow,” Inside Story (8 July 2009).
Seed planted by Tom Round — 16 July 2009 @ 00:50
I didn’t address the “Machevellian” view that the best way for third parties to exert their influence in the U.S. is by spoiling. Several points.
1) The people best in a position to know whether that is true — the third parties themselves — don’t believe it. Third parties in the U.S. are actively seeking a voting system that would reduce the spoiling factor.
2) Third parties have been spoiling elections in the U.S. for a long time with nothing positive to show for it. Spoiling brings only ridicule, not respect or influence. Third parties have been unable to extract concessions from he major parties here, while at least some amount of horse trading goes on in the AV elections in Australia.
3) Spoiling is a key factor keeping third parties out of televised debates. The power of this cannot be underestimated. Ross Perot and Jesse Ventura were polling in the single digits before they were allowed to debate. You can campaign every hour of every day and not reach the number of people that watch a televised debate.
Seed planted by Greg — 16 July 2009 @ 09:16
I’m the one who loathes AV, as Matthew pointed out above. I’m a lecturer in politics at the University of Glasgow. We occasionally use AV in Scottish local elections, where it is supposed to be STV, but when there’s a by-election and just one person to be replaced, STV turns into AV. STV was only recently introduced for local elections by the previous Labour-Liberal Democrat coalition government that lost in 2007 to the Scottish National Party (SNP), and Labour really did not want to replace the previous single-member plurality (or first-past-the-post, the term Matthew prefers to SMP!) with STV, but the Liberal Democrats would have ended the coalition if STV had not been introduced (the Lib Dems have long campaigned for STV here in Britain). The Electoral Reform Society here also prefers STV (but we have MMP for the Scottish Parliament).
I’ve never understood the appeal of AV, whose main selling point in Britain appears to be getting a ‘majority’ of the voters (by means of preference transfers in cases where this does not exist ‘naturally’) behind a single candidate. This point, brought up by some in the Labour Party (like Peter Hain) who keep trying to sell this system, betrays a fundamentally majoritarian view of democracy, one that not everyone shares, and one that is at odds with many of the constitutional changes that the Labour Party itself has brought about in the last decade or so (such as devolution itself!). It is these inconsistencies that open Labour up to criticism (see works by Matthew Flinders, University of Sheffield, for good examples). While Labour has introduced many anti-majoritarian changes (which limit or reduce the power of the executive - an executive that is more powerful in Britain than in almost any other major democracy), the party/government still appears to want to govern as if it has total power.
AV, at least in the very limited number of cases in which we can study it (it is used in very, very few places), does appear quite majoritarian and anti-pluralist. I understand that no ‘minor’ party has ever one a seat in the House of Representatives in Australia, the main example of an AV country. While I take the point about state-level politics in Australia raised by Alan above, I still do not consider the National Party to be distinct enough from its larger national-level counterpart, the Liberal Party, to rate National as a successful ‘minor’ party - indeed, I think the Liberal Democrats in Britain should be quite wary of AV’s potential to turn them into a subsidiary of Labour, should the Lib Dems consider supporting a minority Labour government (if we have a hung parliament next year) in return for electoral ‘reform’.
I also don’t buy into the argument that AV can help get policies of ‘minor’ parties better represented via the deals over preference transfers; small parties will only be heard properly if they can get their candidates elected, and this only happens in Australia via STV in the Senate. The only significant consensual mechanism I have seen in Australian politics is the bargaining that occurs when the governing party does not control the Senate (thanks to STV). Otherwise, the only way minor party candidates seem to have any impact is when they defect to larger parties (like Cheryl Kernot and Peter Garrett in recent years).
In Britain, smaller parties can sometimes manage the 35-40% of the constituency vote they need to win a plurality of the votes and the seat (this is particularly the case in Scotland, where we have a multiparty system that pre-dates the introduction of PR systems here). With AV, that threshold would increase to 50% plus one vote and would almost certainly reduce the party pluralism we have. If you value majoritarian democracy, then fine - you would like AV. If you prefer a more consensual system which embraces pluralism (as I do), then you will probably not like AV and would prefer a PR system (as I do). I think the debate over AV simply boils down to a person’s values, as is the case with the larger issue of electoral system and institutional choice. So, I do not see AV as an improvement, in any way, over SMP (at least in Britain today), and would not support it, even as an ‘interim’ measure (there have been suggestions that Labour might offer AV first with the ‘+’ - the ‘additional members’, as Roy Jenkins recommended - to be added later, if you would trust them to do so…..). In the British case, AV would be a step away from the party pluralism we have seen in recent decades (since the early 1970s).
Seed planted by Thomas Lundberg — 16 July 2009 @ 09:27
Before I forget, everyone following this thread should look at these comments by Australian Green Party activist Ben Raue. He argues that AV has been better for his party than plurality would have been.
One of his reasons is that in Australia AV elections are embedded in a context that includes STV elections for the Senate and for many state and local bodies, a circumstance that would not apply in the U.S. But it seems to me that much of the rest of his post is relevant in the U.S. (I almost wrote “relevant here”, but we’re really talking about at least four countries now).
Seed planted by Bob Richard — 16 July 2009 @ 10:37
I also don’t buy into the argument that AV can help get policies of ‘minor’ parties better represented via the deals over preference transfers; small parties will only be heard properly if they can get their candidates elected, and this only happens in Australia via STV in the Senate.
Actually, no. The pre-election bargaining over preferences focuses, if anything, more on the House than the Senate, indeed Labor has not been above a couple of attempts to knife the Greens in the Senate preferences.
Greens have held seats in the House of Representatives on two occasions, although both were an unusual set of circumstances. The record of non-major party House victories is listed here.
As far as I know Procrustes has never served in the House of Representatives but he seems to have a certain influence on this thread.
Seed planted by Alan — 16 July 2009 @ 10:52
Jonathan, just to be clear: I would put the US Senate in the class of systems for which AV is clearly superior to FPTP. That was why I said “US House or state legislative races.” If you are electing just one (at a time) for a jurisdiction, then I do not think FPTP is worth the risk that nearly happened in Minnesota: a majority disapproved incumbent reelected.
For larger bodies, like the House, one should use (some, any, kind of) PR anyway. But I do not think FPTP is clearly worse than AV for such elections. Now I am repeating myself, and I will continue to read and think about the evidence and arguments presented here by those here who disagree.
Seed planted by MSS — 16 July 2009 @ 12:04
By the way, I am pretty much convinced that supplementary vote and contingent vote are among the very worst electoral systems, down there with SNTV and list-plurality and other widely condemned systems. It makes me quite nervous to use a term like “instant runoff” which clearly encompasses such systems, if advocates of AV do not make the distinction.
And then there is the question of whether the goals of AV/IRV would be better accomplished with Coombs Rule (eliminate first the candidate with the most last-place ranks, rather than the one with the fewest first places).
Seed planted by MSS — 16 July 2009 @ 12:09
Thanks, Alan, for mentioning non-major party Australian House of Representatives victories (above), but the list you refer to lists all the successful candidates as independents, so I’m confused about your mention of Green victories (did the candidate you mentioned actually contest on each occasion as a Green or an independent)? Yes, I was aware of independents winning seats under AV in unusual circumstances, but in terms of candidates actually standing as members of ‘minor’ parties, the record looks bleak. I don’t know if smaller parties in Australia have the territorial concentration that would help them win seats under a plurality system (perhaps in a few places they do), but in Britain, AV would raise the threshold and hurt the smaller parties that are currently able to win under SMP.
Few people here in Britain strongly support AV, apart from some Labour people (like Peter Hain, who wrote a book on this subject) who seem to think British voters are stuck in a 1990s time warp - they expect that anti-Conservative tactical voting would bring Liberal Democrat and other voters over to them in second preferences. The current reality (reflected in many opinion polls in which government or prime ministerial preference is asked) is that there is likely to be a lot of anti-Labour tactical voting: Gordon Brown and his government are deeply unpopular. Some Labour people appear to be in denial about this, or just don’t get it - AV (if implemented ‘unilaterally’ by Labour, something that is entirely possible - a referendum is not, strictly speaking, required, nor do boundaries need to be redrawn) could actually help the Conservatives secure a bigger victory next year, after which the new government would change the system back to SMP (the system that helped the party for most of the 20th century). It would be a delicious irony.
Seed planted by Thomas Lundberg — 16 July 2009 @ 12:14
Regarding Bob’s comment at #8, it makes sense that Greens would generally see AV as far preferable to FPTP, inasmuch as there are very few cases in the world of Greens winning FPTP elections in any national or state/provincial legislature.
The Greens are precisely the sort of party that is most hurt by plurality voting, because they tend to have more dispersed support than other ’secondary’ or ‘minor’ parties (like the NDP and LibDems, which can break 35-40% in some districts, though likely not 50% in many, even with AV, as Thomas noted). In this sense, Greens, in the absence of a PR system, are probably more like the single-issue organizations I referred to above than they are a party. Of course, when PR is available, Greens can act more like a party, and then it is interesting to see their preferences over type of PR. See the BC thread: Greens may have helped defeat STV in 2005, because they really wanted MMP.
(There must be some political science work about how Green Party strategy and organization varies with electoral system. Isn’t there? If not, why am I not writing it? Here I am, a Green-identifier and electoral-systems scholar–the topic seems such a natural!)
Seed planted by MSS — 16 July 2009 @ 12:25
The U.S. differs from the U.K., Australia and Canada in (at least) two important respects relevant to this discussion.
(1) PR is not on the public agenda here. Although three consecutive provincial referenda in Canada (voted on by over half the population) have lost by decisive margins, the rough equivalent in the U.S. would be losing referenda in the nine or ten largest states. That’s unthinkable — at least so far. Instead, we have had a losing ballot measure in the city of Cincinnati, Ohio, a winning ballot measure (for a couple of very minor elected boards and in connection with the adoption of AV for the city council) in the city of Minneapolis, Minnesota, and purely advisory measure in Davis, California.
While this difference may not be a decisive factor in the present discussion, it is clearly a factor.
(2) The U.S. not only has a presidential form of government, but one in which voters (for the most part correctly) perceive executive branch offices and elections as far more important than legislative elections. Thus almost everyone thinks about electoral systems in intrinsically single-winner terms. This is one of many subsidiary factors contributing to (1). It also means that electoral reform has to include some fix for the problems of plurality voting in single-winner elections. If you propose AV for governor but propose nothing for the state legislature (because proposing PR would be premature), you have to explain yourself. At that point, you start to lose people who did not major in political science or social choice theory.
If I were in the U.K. or Canada I would oppose any referendum to adopt AV for a legislative body. In the U.S. I have supported such measures, within limits. One limit is that I would oppose switching a city council from plurality-at-large (MNTV) to district elections in order to use ranked ballots. Another limit is that I do not think reformers should be helping local jurisdictions devise majoritarian alternatives to STV in order to use ranked ballots in multi-seat elections, as in Aspen, Colorado, and Hendersonville, North Carolina.
Seed planted by Bob Richard — 16 July 2009 @ 12:28
It’s clear from the comments here that party positions on electoral reform have much to do with the existing political institutions, party systems, and the democratic values people hold. I suppose we all know this, but sometimes it’s easy to forget and to look at these issues only from our own perspectives.
Regarding the Greens and their electoral strategies, as Matthew pointed out - yes, I think there should be more research on the topic, and an excellent case to look at would be the UK, where we have six different electoral systems. The Greens have succeeded in the PR ones (MMP - Scottish Parliament and London Assembly; STV - Scottish local government; and closed-list PR - European Parliament) as well as the majoritarian ones - SMP (and possibly MNTV/block vote - I’d have to check), with more than 100 councillors in English towns and cities, including Brighton, Lancaster, and Oxford (places where ‘new politics’ do well - lots of younger people who think differently!). They are trying to elect MPs next time in places where they do well locally, so there is some degree of concentrated support for them in the UK.
Yes, Matthew, I would highly recommend a study of British Greens (actually, there are two organisations - one for England and Wales, and one for Scotland; I was part of a team coding party positions on issues for the EU Profiler, and we did two separate codings for them). There is a book on the Scottish Greens by Lynn Bennie (Aberdeen), but I think it focuses on membership and why people join. I doubt whether there is anything on electoral strategy, but could take a closer look. It’s really just the ‘big three’ UK parties that have had most of the academic attention, despite the rise of others - you would be surprised at the state of what passes for political science in Britain. An article I wrote with a friend on UKIP (the UK Independence Party, which wants us to leave the EU) was one of the first on this party. So please do consider something on the Greens - it would be very timely!
Seed planted by Thomas Lundberg — 16 July 2009 @ 14:07
The no minor party has ever won a seat canard really needs to go join the no independents canard. They can both hang their heads in shame while they stand in the corner.
I live in a traditional Labor seat, one of several, where the Green vote has been increasing over the long term.
Labor, to judge by the effort they put into campaigning here, does not appear to share the Shugart/Lundberg certainty that they can never lose the seat because, and I quote, no minor party has ever won a seat in the House of Representatives. Oh wait…
Seed planted by Alan — 16 July 2009 @ 17:17
MSS, we’re probably in fairly close agreement. My preference for AV over FPTP in legislative races has more to do with my conviction that the list of desiderata of election methods is topped by a) ballot expressiveness (the voter should be able to provide enough information for the method to get to some “best” result), b) low strategy incentive, and c) ease of voting.
FPTP comes up rather short by those measures. AV/IRV and Condorcet methods probably do the best, on balance.
Coombs sucks on the strategy criterion, in that it provides a glaringly obvious strategy (burial). AV’s later-no-harm, on the other hand, is a valuable guarantee that’s easily understood by voters. (It has its own strategy problems in center-squeeze contexts, but those appear to be less common, and the strategy is somewhat more subtle.)
Bob@14 writes, correctly, that “PR is not on the public agenda” in the US, which is true enough, but I’d conclude that our priority ought to be to put it on the public agenda, rather than to lead with AV/IRV. Whether AV/IRV is a little better or a little worse than FPTP for legislative elections is a matter of low import. To me, anyway.
List plurality (if I understand the term) is pretty rare in the US, but at-large plurality is fairly common in local elections (school boards, city councils, etc), and it’s nearly as bad.
Seed planted by Jonathan Lundell — 16 July 2009 @ 18:29
Jonathan,
For legislative bodies, my list of desiderata of election methods is topped by proportionality. I gather that yours is too, even though the list you give here, like all statements meant to apply equally to legislative and executive branch elections, doesn’t reflect our shared preference for PR.
Suppose it turns out in practice that the results of AV elections are, on average, less proportional than the results of plurality elections, holding everything else constant. Apparently, some experts believe this to be true — although I’m having trouble finding a comprehensive and authoritative source on the question. This is not just a question of whether small parties win slightly more or slightly fewer seats, which has been our focus so far in this discussion. It is also affected by whether AV has a tendency to increase the seats/votes ratio of a majority party even more than plurality does, and how it performs when there are already more than two “major” parties, as in the U.K. and Canada.
Wouldn’t making proportionality even worse offset — to some highly debatable extent — the advantages of expressiveness, resistance to strategic voting and ease of use?
On the other side of the ledger, saying that our priority ought to be to put it [PR] on the public agenda, rather than to lead with AV/IRV doesn’t, by itself, explain why leading with AV/IRV is not a strategy for putting PR on the public agenda. It also does not say what other strategy would be better.
I am obviously thinking out loud here, and no doubt planting some weeds along with the occasional seedling. I am extremely grateful to our host for this opportunity.
Seed planted by Bob Richard — 16 July 2009 @ 19:04
Bob@18: Yes, I’d put PR at the top, but I’m thinking of PR vs plurality-take-all as a kind of different dimension than the good-to-bad scale of methods. I’d like to establish the general desirability of proportional representation as a principle without necessarily talking about methods at all. While I have my own preferences, I wince a little at the abundance of detail in the NAF proposal for California, for example, in that it begs the underlying question of why proportionality is desirable in the first place.
My sense is that any difference between AV/IRV and plurality with respect to proportionality is in the noise, especially in the US, and that the question isn’t very interesting. If we’re going to have single-seat elections, I’d prefer them to meet my method criteria, though, and given a choice between AV and FPTP, I’d choose the former.
A slight digression: taking California as an example, it’s interesting that electing the Assembly by single-seat districts leads to a surprising (to me) degree of proportionality as measured by R/D registration. That proportionality is largely accidental, it seems to me, and as proportionality rather weak tea, since the range of political views from R to D is pretty limited. Nonetheless, imagine what the Assembly would look like if we had at-large elections….
Seed planted by Jonathan Lundell — 16 July 2009 @ 22:12
Jonathan @19: … any difference between AV/IRV and plurality with respect to proportionality is in the noise, especially in the US.
If true, this would simplify my life greatly.
… electing the Assembly by single-seat districts leads to a surprising (to me) degree of proportionality.
Apparently so. The numbers for the Assembly (lower house) seem to following something closer to a square rule (or less) than the usual cube rule or 2.5 rule. (If you don’t know what this refers to, see Taagepera and Shugart, Seats & Votes, pp. 158-167.) This might be the effect of gerrymandering; safe seats are more representative than competitive seats.
That proportionality is largely accidental …
Probably not. See the immediate preceding comment about gerrymandering.
Seed planted by Bob Richard — 16 July 2009 @ 23:18
It’s difficult to evaluate AV, if only because of small sample size. I am very hesitant to consider Australia the necessary result of AV. Take a look at the results of recent .au Senate elections: Seat allocation is proportional; votes are preferential and thus not wasted; so we would expect people to vote their true preferences. Still, the two-party* share is always around 80%, and no third-party has managed to exceed ten percent of vote share. It seems that Australia may not be a two-party system due to Duverger’s Law, but rather because voters actually prefer the two largest parties. Using Australia to evaluate AV is like using Malta to evaluate STV.
With that out of the way, what’s my opinion on AV? I think it’s risky! Were there a referendum on AV in Canada**, I’d be quite unsure which way to vote. Obviously it would be nice if voters could express their true preferences, since vote-splitting and strategic voting is so awful in Canada. I’ve seen estimates of 25-30% strategic votes, if all of them got together the Strategic Vote Party would become the Opposition!
The risk here is not only that the minor parties might have trouble getting to 50% + 1, but also that AV is very path-dependent. It’s easy to imagine a left-wing riding voting NDP-Grit-Tory, and the Tory second-preferences putting the Grits over the top; and then a centrist riding voting Tory-NDP-Grit, with the Grit preferences getting the NDP elected. Much more data is needed to determine what AV would do to a multi-party assembly.
* If you consider the Nationals and Liberals to be actually separate parties, read “major alliance” and “non-aligned” for “two-party” and “third-party”.
** Oh god, not another referendum!
Seed planted by Vasi — 16 July 2009 @ 23:52
WRT gerrymandering, I’m doubtful that it makes all that much difference (wrt proportionality). Fortunately (for that particular question, anyway) we’ll soon see in California how much difference it makes. Whatever the drawbacks of the system, it won’t be gerrymandered for much longer.
I mean “accidental” in a very loose sense–more accurately, I suppose, the proportionality that a single-seat-district system yields is dependent on factors outside the voting method: how populations are distributed, how district lines are drawn, and such.
But if I’m a Republican in a safe Democratic district, I’m voiceless as a voter. That is, even if we could arrange for some factor–line drawing, AV vs FPTP, whatever–to give us de facto proportional results, we wouldn’t be addressing what seems to me to be the other core benefit of PR systems: that every voter has an effective voice in choosing their representation.
It’s not just that the resulting legislature happens to be proportional: it’s now that proportionality is achieved with respect to the electorate.
Seed planted by Jonathan Lundell — 17 July 2009 @ 11:35
Vasi,
If you go back a little further, you will find examples of Senate elections in which minor parties have exceeded 10 per cent of the vote.
In the 1970 Senate election, 11.2 per cent of the nation and 19.1 per cent of Victorians voted for the DLP. The total non-major party vote in Victoria was 25.2 per cent.
In the 1977 Senate election, 11.1 per cent of the nation and 16.1 of Victorians voted for the Democrats. The total non-major party vote in Victoria was 24.8 per cent.
In the 1990 Senate election, 12.6 per cent of the nation voted for the Democrats.
In the 2007 Senate election, the total non-major party vote in South Australia was 28.8 per cent.
In response to the quotation in the original re minor parties not coming close:
Minor parties have come close in general elections on occasions. In 1955, the first post-Split election, the DLP candidate, Bill Bourke, came close to winning Burke, with a final result of DLP 47.7 per cent, Liberal 53.3 per cent. The DLP candidate, John Cremean, came close to winning Scullin, with a final vote of DLP 48.3 per cent and ALP 51.7 per cent. The DLP candidate, Stan Keon, came close to winning Yarra, with a final vote of DLP 49.0 per cent, ALP 51.0 per cent.
At the state level, the DLP won Richmond, Victoria, in 1955. It came close in a number of other seast. It won Aubigny, Queensland, until the MP retired in 1972. It came close in Warrnambool, Victoria, in 1967: it only missed out on because the ALP preferenced the Liberal Party.
In the SA Assembly, the Liberal Movement won at least one seat. When the holder of that seat switched to the Democrats, he retained it.
(All figures are from http://psephos.adam-carr.net/countries/a/australia/)
Seed planted by Chris Curtis — 18 July 2009 @ 10:04
Vasi (#21): The risk here is not only that the minor parties might have trouble getting to 50% 1, but also that AV is very path-dependent. … [examples omitted] …
Vasi’s general approach here is similar to Lewis Baston’s in this pamphlet. Baston concludes that, if the smallest of three major parties is ideologically in the middle (Lib Dems in the U.K.), it would do much better under AV than FPTP. This would increase proportionality. He echoes Vasi’s implication that the NDP in Canada would be hurt by AV because it is to the left of the two larger parties. This would decrease proportionality.
Baston also argues convincingly (to me) that the 1997 Jenkins Commission’s view that AV would have hurt the Tories badly was specific to one set of circumstances. Thus, he thinks that the often-mentioned tendency of AV to make landslides even more lopsided than FPTP does is specific to certain elections rather than general.
His treatment of small parties is equally nuanced. They might be helped somewhat by more sincere voting, but the benefit would be indirect.
Baston’s pamphlet is the most thorough and convincing discussion I have found so far. I don’t (yet) know how I would apply his approach to the circumstances of the United States.
Seed planted by Bob Richard — 18 July 2009 @ 17:37
We talk a lot about tactical voting in plurality elections, but (1) we have little (maybe even no) data on how many voters actually do vote for their lesser evils, and (2) we have little (maybe even no) understanding of how many voters, under AV or some other single-winner improvement on plurality, would come to realize that their currently perceived favorite is really a lesser evil. In other words, whether voters would expand their horizons.
I’m starting to think that these questions are the key to this discussion. If there’s not much tactical voting to begin with, then AV is pretty pointless and a diversion from the path toward PR. If there’s a lot of tactical voting and/or voters are eager to expand their horizons, then AV would change things. In the U.S., where there is no third major party and little prospect of one, the changes would be positive and would be very much on the path to PR.
Seed planted by Bob Richard — 18 July 2009 @ 18:03
@Bob
You’d also need to consider that AV changes campaigning.
Unfortunately Australia does not work as AV City because we have very high levels of turnout for other reasons. Labor, for example, cannot knife the Greens when they are hunting for Green second preferences. FPTP/voluntary voting encourages a candidate to try and and find the largest voting group and prevent others voting at all by methods ranging from demonising their candidates to actually excluding them from the polls.
There is anecdotal evidence from San Francisco tat the tone of their campaigns is already shifting in this way. A shift from demonising to policy advocacy would be a very good thing for the US.
[Edited in very minor ways that I am pretty sure Alan would approve of.--MSS]
Seed planted by Alan — 19 July 2009 @ 02:04
Chris,
Thanks for the correction. There have indeed been a few Senate elections in Australia where a third-party has received slightly higher than 10% vote share, I withdraw my assertion to the contrary. As to the South Australian third-party vote in 2007 it is indeed large, but an outlier. It indicates that Xenophon is a popular independent, not that Australians will vote for more parties.
Perhaps I have high standards for third-parties, in Germany most of them would be happy to get to ten percent. But how common is a persistent ~80% two-party vote in a proportional system? I still have to consider Australian elections non-typical, and don’t think their results under AV can be generalized to other countries.
Bob,
Wow, thanks for the link to the Baston pamphlet. It’s very clear and well thought-out. I’m impressed that he included examples of strategic voting in AV, with voters trying to influence the order of elimination.
My only quibble is the theory that AV will exaggerate landslides. The reasoning is that “a strong tide towards a party…[will] also affect second preferences”. I suspect that most landslides occur because voters are disgusted with the losing party, not because they’re thrilled with the winner. In this case AV could be a moderating factor.
For example, many UK voters want to punish Labour right now. Under FPTP, they must vote for whoever is likely to beat Labour, so we expect a Conservative landslide. Under AV, they could vote against Labour without voting for just the Tories, so we might see many votes like Green-LibDem-Tory-Labour or UKIP-EDem-Tory-Labour. The LibDems, Greens and UKIP might pick up a fair number of the ex-Labour seats. This would accomplish the voters’ goal of punishing Labour without unduly rewarding the Tories, and without eliminating any semblance of an opposition.
Regarding data on strategic voting, there have been a number of studies in Canada, though their results differ enormously.
Seed planted by Vasi — 20 July 2009 @ 02:03
The US political system is so non-pluralistic that just about any change to the electoral system will increase voter options.
Seed planted by Ed — 20 July 2009 @ 13:49
AV isn’t even a very good single-winner method. Any Condorcet method is better.
1) Because it’s Condorcet. That is, whenever the AV winner and the CW are different, by definition a majority would prefer the CW. IRV supporters claim that in that case the CW is a “turkey”, yet the addition or removal of completely irrelevant candidates, or even changes in the order in which they are eliminated, can make the CW a “turkey” or not a turkey without any changes to any of his preferences.
2) AV reproduces one of the objectionable features of the current American plurality primary and plurality general election system, which is that most of the work is done by self-selected segments of the electorate, rather than by the people as a whole. For example, whether Clinton or Obama should be President properly concerns the whole people. Under the current rules it was unjustifiably entangled in a separate question, which is who should lead the Democratic Party. Under IRV/AV, even that justification is gone (although admittedly the self-selected segment is given a slightly more rational means of making its choice); once you get down to one of the top two, any remaining preferences are ignored, which means that there are really two votes running parallel, choosing one each from the right and left, and then a vote between them. If you pick the loser in the top two, you get nothing, even though a different elimination order might have reversed the right vs. left vote.
AV implicitly assumes that unless it involves your first preference, you don’t really care. But then there’s no reason to even ask: just revert to plurality. If the voter has the option of truncating his ballot, we know that every listed preference matters.
Seed planted by Aaron Armitage — 22 July 2009 @ 12:40
Aaron: Whenever the AV winner and the CW are different, by definition a majority would prefer the CW.
I don’t think this is quite right. Consider a three-candidate election with the following preference votes:
3 ABC
4 BCA
2 CAB
If this is an AV election, C is eliminated, and A then wins 5-4.
If it’s instead a Condorcet election, then A defeats B 5-4, B defeats C 7-2, and C defeats A 6-3. The weakest defeat is A>B, so most (all?) Condorcet methods would elect B.
The result is an AV winner, A, who a majority prefers to the Condorcet winner, B. I’m not arguing that the IRV result is better–just pointing out that your claim above doesn’t seem to be true in this case.
By the way, what exactly do you mean by a “turkey”? Is that the idea that a candidate who’s not well-liked but whom nobody hates could squeak through?
[O]nce you get down to one of the top two, any remaining preferences are ignored, which means that there are really two votes running parallel, choosing one each from the right and left, and then a vote between them.
I’m sure that’s what the French thought of the two-round system until 2002. Anyhow, I’m not sure we can even assume that parties won’t still choose leaders via primary even when AV or Condorcet is available–that seems to be the pattern in Australian elections, and Irish Presidential elections.
But I’m going on. I think Condorcet elections are great, in situations where they can be accepted. The open-source projects where it’s used now are perfect: The voters have all become involved by choice, they’re often mathematically inclined, and a strong consensus is needed. I’m not sure I can see Condorcet being adopted in any geographic political entity though. I would hate to be the one to try explaining Cloneproof Schwartz Sequential Dropping to people who think MMP is complicated! If you’re enamored of the theoretical properties of Condorcet, but want to advocate a system that is at least marginally likely to ever be adopted, perhaps you should argue for Approval Voting instead?
Seed planted by Vasi — 23 July 2009 @ 05:37
Vasi;
By a Condorcet winner, I mean a candidate who beats every other candidate one-on-one, which means that there might not be one.
My favorite completion method is ranked pairs, where you start from the strongest defeat and keep adding them, skipping any defeats that would contradict previously added ones. So for your example, B>C is added, then C>A, but A>B is ignored, leaving a final ordering of B>C>A.
Also, applying AV to the Smith set is a perfectly valid Condorcet method. I like it because some kinds of strategic voting that work in Condorcet become irrelevant; but it isn’t summable, which seems undesirable, but not necessarily fatal.
By the way, what exactly do you mean by a “turkey”? Is that the idea that a candidate who’s not well-liked but whom nobody hates could squeak through?
It’s not my term, it’s the term I always seem to see FairVote people using when they argue for AV over Condercet. They seem to mean any candidate who doesn’t have enough first place votes, although they never really define “enough”. I think they just want a single-winner method that can be directly generalized into STV, and that they’ve gotten distracted by other single-winner reforms (e.g., tinkering with Presidential primaries). But the biggest problem with our executives is that most of them have too much power and more responsibility than one person can handle. The legislatures can no longer credibly attempt to reclaim the authority they’ve abdicated, so the first order of business must be to fix them so that they can resume their responsibilties. Once people begin thinking of deliberative bodies (especially the lower chambers) as the primary venue for translating their policy preferences into actual policy most of the problems with the executive will resolve themselves.
I’m sure that’s what the French thought of the two-round system until 2002.
Sure, but it turned out that way because the electoral system, like pure plurality, couldn’t handle multiple candidates. AV would have put Jospin over Le Pen, so I give it that much. On the other hand, it’s possible that under AV the 1991 gubernatorial election in Louisiana would still have seen Edwards and Duke as the last two, but Roehmer was probably the CW and was also the only sane choice running that year.
Anyhow, I’m not sure we can even assume that parties won’t still choose leaders via primary even when AV or Condorcet is available–that seems to be the pattern in Australian elections, and Irish Presidential elections.
You’re right, of course. What I’m advocating is really two reforms: replace plurality with Condorcet and also abolish primaries. As far as I know, the (few) jurisdictions in the United States which have adopted AV at the urging of FairVote have also abolished primaries, so I suppose my position is really the same as theirs, but with Condorcet in place of AV.
There’s reason to think the parties would prefer to avoid having only one candidate in a Condercet election. If we assume that for any given district, the Democrats will be to the left of the median voter and Republicans to the right, both candidates will lose to a more centrist independent, if voters rank according to issue positions. But without primaries, both parties will have a range of candidates, so the most centrist candidate (centrist relative to that district, which may not be centrist at all from someone else’s perspective) is more likely to be a member of a major party. You see that my reform is not necessarily favorable to third parties, but it is as fair as any single-winner method will ever be.
I’m not sure I can see Condorcet being adopted in any geographic political entity though. I would hate to be the one to try explaining Cloneproof Schwartz Sequential Dropping to people who think MMP is complicated! If you’re enamored of the theoretical properties of Condorcet, but want to advocate a system that is at least marginally likely to ever be adopted, perhaps you should argue for Approval Voting instead?
Actually I’ve had reasonably good success at persuading ordinary people that a Condorcet method should replace plurality, but I don’t even attempt to explain completion methods. I explain the basic idea and mention that cycles and completion methods exist. A typical voter will never need more information than that even if a Condorcet method is adopted.
I would vote for AV over plurality, but I won’t jump on the IRV bandwagon.
Seed planted by Aaron Armitage — 23 July 2009 @ 13:00
My great problem with Condorcet is the same as Vasi’s, opacity. I have both counted and scrutineered AV and STV elections.
The nature of an AV/STV scrutiny is a series of counts where the ballots undergo a finer and finer sort. That is a transparent process and easy to count, scrutineer and recount. If necessary you can retrace the count if you get an error in the math.
To count a Condorcet ballot you must re-arrange the papers onto 2 parcels for every pairwise contest. That would be extraordinarily difficult to count or scrutineer and there is no way to detect an error. Transaction costs do matter when you consider electoral systems.
I also had the nasty experience of being the Condorcet winner, but AV loser in an election for a university senate in my early 20s. (The math school got hold of the ballots and conducted a Condorcet count) The problem with that unhappy experience is that I got 18% of first preferences.
I would not like to explain to the supporters of the other candidates why 18% had beaten candidates who each received more than twice that on first preferences. Frankly I can hear the hanging chads and fuzzy math go thud right now.
I would vote for AV over plurality in most circumstances, specially when there the other house is STV. Proportional representation in multimember districts is obviously a superior system.
Seed planted by Alan — 23 July 2009 @ 13:08
Alan;
To count a Condorcet ballot you must re-arrange the papers onto 2 parcels for every pairwise contest.
Even if we were counting by hand the votes would be reduced to matrices as soon as possible and summed, probably even below the precinct level. For large-scale elections, computers are necessary. Just imagine trying to do an AV count by putting every ballot cast in a Presidential election in a big room and physically carting them around. And you don’t have until January, you have to get it done in time for the 270 Electors belonging to the NPV compact to recieve their instructions.
If you don’t trust the current infrastructure, the software can be made open-source and the hardware devoloped and built under closer supervision or even by the government. If that still doesn’t help, I don’t know why the physical ballots are any better, since those can be tampered with too.
I also had the nasty experience of being the Condorcet winner, but AV loser in an election for a university senate in my early 20s. (The math school got hold of the ballots and conducted a Condorcet count) The problem with that unhappy experience is that I got 18% of first preferences.
I would not like to explain to the supporters of the other candidates why 18% had beaten candidates who each received more than twice that on first preferences.
The problem is that it was an AV election, not a Condorcet one. You don’t get to change the rules after you see what the outcome is, even though Condorcet was better all along.
If you’re willing to replace plurality with a ranked choice method, then you’re already committed to the idea that someone can beat a candidate with twice the first preferences. Examples of AV elections where this happens would be trivial to construct.
I agree completely on PR being better for legislative races.
Seed planted by Aaron Armitage — 23 July 2009 @ 15:24
I am very much enjoying, and learning from, this AV- Condorcet discussion. Thanks to those participating.
Seed planted by MSS — 23 July 2009 @ 16:02
Aaron
For entirely different reasons, I prefer paper ballots. You can throw an electronic election by manipulating the operating code. It takes a vanishingly small number of people to do that. Of course you can throw a paper election as well, but you need a considerable number of people to do it and the chances of detection rise astronomically.
Even leaving those issues aside, and I assure you the count of a paper election is a somewhat more sophisticated activity than you seem to envisage, saying ti can all be captured on matrices only proves my point.
Elections require transparency to prove their legitimacy. It is one thing to say that Mr 18% can overtake other candidates on latter preferences. It is an order of magnitude different to say that Mr 18% can overtake candidates supported by 80% of the electorate and ascend to viceroy. The proposition is frankly ludicrous.
You entirely misunderstand the sad tale of my youth. I did not contest that result or even object to it. It was funny, nothing more, to discover some weeks later that there existed a voting system by which I was the allegedly legitimate winner.
Seed planted by Alan — 23 July 2009 @ 19:57
Discussions of the Condorcet principle (such as #29 to #34) rarely mention the objection to it that I find most compelling. The majorities (note the plural) who favor the Condorcet winner over each of the other candidates are sometimes overlapping but nevertheless different groups of voters. To me, these are mathematical majorities rather than a cohesive political one.
Contrast this with the principle inherent in AV: if there exists a group of voters all of whom prefer every candidate in a set over every candidate not in the set, the winner should be a member of that set. In social choice theory this principle is called the mutual majority criterion, or sometimes the majority coalition criterion. Majority coalitions are political, not mathematical, groups.
There can be elections that have a Condorcet winner but not a majority coalition, and vice versa. There can be elections that have both, in which case the Condorcet winner is necessarily one of the candidates in the majority coalition. In such elections AV does not necessarily chose the Condorcet winner. In elections that have a majority coalition but no Condorcet winner, some Condorcet completion methods always pick a winner who is part of the majority coalition, while others sometimes do not. If I were to propose any Condorcet method at all, it would have to be one that does.
Both principles produce a few results whose legitimacy can (and therefore will) be attacked by the loser. I don’t think this consideration provides much basis for choosing between them.
When AV (or any other non-Condorcet method) picks someone other than the Condorcet winner, that candidate’s supporters cry foul, saying “but my candidate beats the winner in a head-to-head match up”. Well, yes, but the supporters who helped him beat A are not the same folks as the supporters who helped him beat B. Who, exactly, does he have a mandate from? The rejoinder, of course, is that when opinion is polarized a centrist, compromise candidate should always win.
When a Condorcet method elects someone because she is hated by no one but is almost no one’s first choice, supporters of the candidates with more first preferences cry foul, describing the winner as colorless and without any real positions on the issues. (As an aside, I’m more familiar with the term “turkey” in the related by very distinct context of voting insincerely for a weaker opponent of your favorite in order to try to eliminate a stronger opponent, hence the strategy of “turkey raising”. Some Condorcet methods are vulnerable to this strategy.)
Methods should be evaluated in terms of the strategic pressures they present for candidates and potential candidates as well the strategic pressures they present to voters. We have no data on whether this is significant in practice, but it stands to reason that Condorcet encourages candidates to crowd together in the middle of the political spectrum and, perhaps, to be as inoffensive as possible. The incentives that AV presents to candidates (as already mentioned in this thread) are, if they exist in practice, mostly positive.
Adding Condorcet’s lack of transparency (relative to AV) to these conceptual considerations, I think AV is clearly better.
Seed planted by Bob Richard — 26 July 2009 @ 15:46
> “supporters of the candidates with more first preferences cry foul, describing the winner as colorless and without any real positions on the issues”
Add to this that Condorcet has the same problem as Approval Voting: supporters of a larger party (with more first preferences) have motive, means and opportunity to vote tactically.
Outside the world of Brams-Fishburn simulations, compromise candidates are not going to get elected under Approval. Greens or Naderites have an incentive to give a second tick (or, if Condorcet, a second preference) to Democrat or Labo[u]r candidates. Dems and Labs have no long-term incentive to return the favour. As a result, Approval in practice (at least in partisan elections) will function like a crude form of AV anyway, where Green voters tick GRN and DEM/LAB while Dem/Lab supporters tick DEM/LAB only. Condorcet would have a similar problem (not quite as crude).
That’s for elections. Condorcet and Approval would work better for referenda, which have a different set of incentives for “candidates” (ie, propositions and their sponsors).
Seed planted by Tom Round — 27 July 2009 @ 21:57
Alan;
I wrote a long reply on Sunday but I lost it. Since I didn’t feel like rewriting it I put it off. My reply to you is a little shorter than the original, although this post in general is quite a bit longer because of the other reply included.
For entirely different reasons, I prefer paper ballots. You can throw an electronic election by manipulating the operating code. It takes a vanishingly small number of people to do that. Of course you can throw a paper election as well, but you need a considerable number of people to do it and the chances of detection rise astronomically.
You seem to be assuming the current system of proprietary software. If the programs are open source and there are independent checks to guarantee that the correct software is loaded, stealing an election by corrupting the software will be no easier than stealing it by replacing the authentic ballots. The person who doesn’t know how to read programming languages will be in the same position as someone who is not an observer or participant in a hand count, except that because more people can check the software, the confidence would be proportionately greater.
Even leaving those issues aside, and I assure you the count of a paper election is a somewhat more sophisticated activity than you seem to envisage, saying ti can all be captured on matrices only proves my point.
Perhaps it does prove your point, but you’ll have provide some explanation as to why before I can follow the point.
Unless there’s been a breakthrough you could share, AV is not summable. I said above that summability is nice but not required, but I was assuming there would be computers involved. If we’re doing everything by hand, it becomes a necessity.
Elections require transparency to prove their legitimacy. It is one thing to say that Mr 18% can overtake other candidates on latter preferences. It is an order of magnitude different to say that Mr 18% can overtake candidates supported by 80% of the electorate and ascend to viceroy. The proposition is frankly ludicrous.
But saying “supported by 80% of the electorate” treats as one thing an aggregate of completely separate and probably contradictory options, none of which are more appealing to the electorate than “Mr. 18%”, otherwise one of them would have beaten him.
Any ranked choice system allows a candidate to win with 18% first preferences (as does plurality, given enough candidates), even if others have more, and this is always because of the effect of the lower preferences. You seem to be saying that this is ludicrous unless it happens using your favored mechanism. But this is arbitrary.
You entirely misunderstand the sad tale of my youth. I did not contest that result or even object to it. It was funny, nothing more, to discover some weeks later that there existed a voting system by which I was the allegedly legitimate winner.
No, I didn’t think you had contested it. I answered the way because any relevance to this discussion comes from the fact that contesting it would have been absurd had you done it. You try to connect this absurdity to the number of first preferences you had. I’m saying that the absurdity would have been in changing the rules several weeks afterward. If you and the AV winner had been the only candidates but everything else been the same, would your victory have been absurd? Now what if there were an electoral college, and it flipped the election? In both cases, the rules are the rules. But having said that, public elections in the United States have meta-rules.
Bob Richard;
Discussions of the Condorcet principle (such as #29 to #34) rarely mention the objection to it that I find most compelling. The majorities (note the plural) who favor the Condorcet winner over each of the other candidates are sometimes overlapping but nevertheless different groups of voters. To me, these are mathematical majorities rather than a cohesive political one.
Ranking one candidate over another is a political choice, so a majority consisting of those who made that choice is a political one. It may not be a cohesive one, but some of us share Madison’s suspicion of cohesive majorities.
In any case, the criterion you made the standard for a due respect for cohesive majority preferences is one which, you fully admit, some Condorcet completion methods comply with, and which all Condorcet methods comply with as long as a CW exists. Ranked pairs, which I described above, is one of the mutual-majority compliant completion methods. There’s no reason to choose AV instead of Condorcet here.
If there’s no mutual-majority set, the final majority in the final round need not be any more cohesive than any of the majorities that are ignored. There is nothing in this pair of candidates or in the underlying preferences to set these two candidates apart as the final two. It’s entirely possible that the choice between those two might be felt less and excite less passion than some other pairwise choice. It’s only your procedure that sets them apart, and your procedure is non-monotonic and not independent of irrelevant alternatives, which means that at least sometimes it behaves erratically.
Both principles produce a few results whose legitimacy can (and therefore will) be attacked by the loser. I don’t think this consideration provides much basis for choosing between them.
I’m going to comment a little more on the complaints you expect from CW supporters and AV winner supporters, but I have a general observation I’ll make here: the complaints are radically different. CW supporters would complain that the will of the voters was ignored for purely procedural reasons. AV winner supporters, on the other hand, would be complaining that the voters got it wrong. If that complaint is to be a challenge the legitimacy of the outcome, they must first refuse to recognize electoral democracy itself and substitute another standard, and then denounce the people as incompetent expositors of this other standard. It seems a rather drastic move. To say nothing else, the first complaint may be justiciable under American law, while second definitely is not.
When AV (or any other non-Condorcet method) picks someone other than the Condorcet winner, that candidate’s supporters cry foul, saying “but my candidate beats the winner in a head-to-head match up”. Well, yes, but the supporters who helped him beat A are not the same folks as the supporters who helped him beat B. Who, exactly, does he have a mandate from?
He has a mandate to beat the AV winner from the people who ranked him higher.
I don’t think “mandate” is a very useful concept, at least in voting system design. I can’t think of a coherent concept of it that would allow us to move mandates around based on a haphazard elimination order; if we could, the implication would be that CW supporters whose votes counted for the winner in the last AV round but are infuriated by the reversal vs. the CW have given their mandate to someone whose election they consider illegitimate.
The fact that a number of overlapping but not identical majorities contribute to selecting the winner is one of my favorite facts about Condorcet methods. It means that almost everyone will have contributed to the outcome; even if you don’t especially like the CW you probably at least helped bury someone you like even less. I understand that you would rather the winner represent fewer voters but represent them better, a preference I share. I just don’t think single-winner elections are the place for that. Multiple winners, each representing fewer but all of them together representing more, are the place for depth.
Every choice is between given alternatives. A first preference only has meaning by its relation to the alternatives. I alluded earlier to Madison’s distrust of standing coalitions of factions, especially if these are majorities. In most countries there are two, in America the two parties, and in others coalitions of parties, and each is largely sustained by the fact that the only real alternative is the other. To some extent these are unavoidable. But in a Condorcet election, instead of two standing choices with ossified factional alignments, every candidate who puts himself forward is directly juxtaposed with every other, calling out new alignments for each. The fact that the CW has a different coalition supporting him against each of the other candidates is healthy. This opportunity will only ever be partially realized. Most of the time the CW’s coalitions will really mostly be two, one being the median voter and everyone to his right, the other the median voter and the left, with some variation related to personality-driven votes.
The rejoinder, of course, is that when opinion is polarized a centrist, compromise candidate should always win.
I suppose that depends what you mean by centrist and compromise. To the extent that voters choose based on issue positions, the median position will tend to be reflected by the CW. But that will only look like a compromise if none of the factions are a majority. Many Condorcet methods comply with mutual majority, after all. Very often the CW would be the mutual-majority set member who is most acceptable to the minority. I consider that a positive.
When a Condorcet method elects someone because she is hated by no one but is almost no one’s first choice, supporters of the candidates with more first preferences cry foul, describing the winner as colorless and without any real positions on the issues.
If the CW really is colorless and has no real positions, it’s because neither side has a majority and voters on both sides chose colorless middle candidates over the other side, and chose candidates with no positions over candidates with articulated centrist positions. The last seems unlikely to me, but either way it’s not our choice, its the voters’.
Methods should be evaluated in terms of the strategic pressures they present for candidates and potential candidates as well the strategic pressures they present to voters. We have no data on whether this is significant in practice, but it stands to reason that Condorcet encourages candidates to crowd together in the middle of the political spectrum and, perhaps, to be as inoffensive as possible. The incentives that AV presents to candidates (as already mentioned in this thread) are, if they exist in practice, mostly positive.
I don’t understand evaluating methods based on something which might not exist, and if it does exist, are not understood. To the extent that we can make any valid conjectures, I would posit that since Condorcet and AV usually elect the same person, the strategic incentives are almost identical.
Two election methods should be compared on those elections where they choose different candidates. Where they elect the same person, there’s nothing to compare. Now, when AV and Condorcet differ, it’s more probably that the CW has fewer first preferences than the AV winner, because of how the eliminations are done. But there are scenarios where the CW actually has more first preferences. He might be less definite on issues, but maybe not. The only difference there always is, is that the CW was favored over the AV winner by most voters.
Adding Condorcet’s lack of transparency (relative to AV) to these conceptual considerations, I think AV is clearly better.
You and Alan both said that, but I don’t think either of you have explained it very well. Ranked pairs actually strikes me as less confusing than AV.
Tom Round;
I took so long writing the above I missed you. I’ll comment later.
Seed planted by Aaron Armitage — 27 July 2009 @ 23:29
ERS’ Ken Ritchie has an op-ed in today’s Guardian. He is critical for several reasons of the AV and AV+ proposals circulating in Britain. (No, none of the reasons derive from social choice theory.)
Seed planted by Jack — 28 July 2009 @ 12:45
Jack (#39): as I’ve said here several times, I agree with Ken Ritchie that in the U.K. and Canada the adoption of AV would be a defeat for electoral reform. I think circumstances in the U.S. are different, quite possibly different enough that AV for the federal House and/or state legislatures would be a step forward rather than a step sideways or backward. MSS and others disagree, and have some compelling arguments on their side. And, yes, social choice theory has almost nothing to say on this matter.
Seed planted by Bob Richard — 28 July 2009 @ 16:20
I’d agree with the Ritchie editorial unless Brown offers a referendum, as is just possible, for an AV house and an STV senate. I also agree AV would be a step forward in the US and Canada.
Seed planted by Alan — 28 July 2009 @ 20:02
Aaron (#38) presents the most persuasive case for the importance of the Condorcet principle that I have encountered. I don’t think I accept all of his points. In particular, it strikes me as a play on words to say that AV winner supporters, on the other hand, would be complaining that the voters got it wrong (see #38 for context). Also, I think incentives for candidates do exist even if they are poorly understood, and that we should try to understand them better and account for them rather than writing them off. But, at least for the time being, I’m going to have to accept that he is winning the conceptual argument.
Then Aaron says, Ranked pairs actually strikes me as less confusing than AV.
Huh? I invite anyone who has pursued the discussion to this point to study the Wikipedia explanation of ranked pairs along with Tideman’s description of it (Collective Decisions and Voting, pp. 219-223). Then set up an appointment with your own county elections department. At that meeting, teach the head election administrator to count both AV elections and ranked pairs elections. I have done this for AV and know for a fact that it is fairly easy. I suspect that no more than one election administrator in ten could master the procedure for ranked pairs. The proportion of voters who could follow the explanation well enough for voter confidence/transparency purposes would be smaller. Even Tideman, who invented ranked pairs, calls it “quite complex”.
The one Condorcet-compliant voting rule that I think might be explainable to voters and election administrators is maximin (Tideman), a.k.a. minimax. In 12 words, “elect the candidate whose worst loss in paired comparisons is least bad”. But this rule does not guarantee selection of a member of a majority coalition when there is one, and has other choice-theoretic weaknesses.
We can have very erudite discussions of the Condorcet principle, the utility maximization principle underlying range/score voting, and a vast literature full of theorems and conjectures. But, as Iain MacLean wrote in 1988, social choice theory and electoral reform are like two ships passing in the night. While there are a lot of reasons for this, transparency is one of the important ones. The fact that social choice theory has nothing to say about proportionality is another.
Seed planted by Bob Richard — 29 July 2009 @ 14:21
I’m a fan of Approval Voting and Instant Runoff Voting. Here’s what I prefer:
a) Approval Voting Plus: Voters are presented with a ballot full of candidates. On the ballot, you have 2 columns, Yes and No. All you do is mark accordingly. So, if you like Nader and Gore but hate Buchanan and Bush, you can vote for Nader and Gore and against Bush and Buchanan. The winner would be the candidate with the highest balance of Yes/No votes.
b) Instant Runoff Voting Plus: Voters are presented with a ballot full of candidates. On the ballot, you have 3 columns, 1st choice, 2nd choice and 3rd choice. All 3rd choices are counted first. If a candidate has a majority, it’s over. But if nobody does, the candidate with the least 3rd choices is eliminated and all the voter’s 3rd choices are transferred up to their 2nd choices. This repeats until someone has a majority of the votes.
I’m more declined to go with the first proposal. It wouldn’t be interesting to only see who voted for a candidate but also who voted against a candidate too.
Seed planted by Derek — 01 August 2009 @ 01:21
Vasi,
The other difference re Australia and PR is that it is not used to elect lower houses, except in Tasmania, but upper houses in the Commonwealth and in four states, so government is not at stake when people vote, and some do vote differently for the two houses. (The ACT is more like municipal council.)
Seed planted by Chris Curtis — 02 August 2009 @ 15:56
I’ve sometimes wondered about a combination of AV and Condorcet for single-seat elections, and referenda… ie, first you work out who beats who pairwise; then you exclude any candidate who’s beaten pairwise by every other remaining candidate; if there’s a cycle, exclude the candidate with fewest preferred votes; continue til only one remains.
Unlike AV, this would be certain to elect the Condorcet winner if one exists. Unlike motion-and-amendment Condorcet, this avoids cycles; unlike “highest benchmark” (maximin) Condorcet, it means that of the final two candidates, whoever’s Condorcet-preferred will win. (Whereas under benchmark, A may defeat B 52-48% but B is still elected because s/he’s the only candidate who never gets below 48%… this tends to nullify Condorcet’s claimed advantage that it ensures the anti-B majority can always keep B out).
No idea how to apply this to multiple seats and PR (probably not necessary anyway).
Seed planted by Tom Round — 02 August 2009 @ 22:03
I’ve been thinking about AV and MMP. I may have a solution.
I propose extracting any votes (including any fractional votes) that have not contributed to electing a district MP, consolidating the wasted votes across a wider superdistrict, preferably the whole country, and then applying the regular threshold and using those votes to elect list MPs.
Seed planted by Alan — 03 August 2009 @ 23:00
Alan (#46),
(1) What does “contributed to electing a district MP” mean in this context? At first blush it would seem to mean “not exhausted”, in which case there wouldn’t be a lot of “wasted votes” to consolidate. I suspect you have something else in mind.
(2) The district votes classified as wasted were cast for individuals, not parties. If those individuals happen to have run as representatives of a party, that’s not a major problem — just assign them to the corresponding party list. But what are you going to do with the votes that were wasted on losing independent candidates?
(2) For this purpose, what is the “regular threshold”? At first blush, it seems that the this would have to depend on how many votes turn out to be wasted in the districts. And in any case, I don’t think you can answer this question without first specifying whether your list seats are parallel or compensatory.
Seed planted by Bob Richard — 04 August 2009 @ 02:00
Let us have an election in the District of Whatthe with Candidates Shugart, Richard, Round and Alan and with 100 electors.
First round
Shugart 30
Richard 40
Round 20
Alan 10
Second round on Alan’s elimination
Shugart 40
Richard 40
Round 20
Third round on Round’s elimination
Shugart 50
Richard 40
Exhausted 10
The votes that do not contribute to Shugart’s election are:
40 Richard first preferences
10 Round first preferences which exhausted on Round’s elimination.
Note that if all Round first preferences had flowed to Shugart so that his final tally was 60 it would be necessary to calculate a fractional value for the Shugart votes to ensure that his surplus votes went into the election for list MPs.
I am assuming a uniform threshold which would be calculated by dividing the average district enrolment by 2 and increasing that result to the next highest whole number. There would, as you note, need to be some refinement to allow wasted votes for independents to aggregate away from the party lists if that was what those electors desired.
We are not ‘assigning’ any votes. The wasted votes are being counted according to their next available preference in the list election.
There would be one less list MP than the number of district MPs because of the operation of the uniform quota.
NB Somewhat unusually for an election methods junkie I have quite serious dyscalculia so I do not guarantee any figures)
Seed planted by Alan — 04 August 2009 @ 06:20
Thank you, Alan, for the very helpful example. I was, of course, thoroughly wrong in the first paragraph of #47 but didn’t realize it until long after my 10 minute grace period had expired.
Thinking out loud here: perhaps independents (but not party candidates) could be allowed to run in more than one district, so that their (otherwise wasted) votes could be aggregated across districts. But, then, what would you do when an independent wins outright in more than one district? This part does need work.
Seed planted by Bob Richard — 04 August 2009 @ 15:13
An alternative, or possibly an addition, to Bob’s proposal would be a list option with instructions in large letters at the top saying ‘Your list vote will only be counted to the extent that your district vote does not elect a representative’.
Seed planted by Alan — 04 August 2009 @ 15:54
As far as any candidate (not just an independent) winning in more than one district, I’d think you’d return them in the district where they achieved the highest vote in terms of the two party preferred vote and treat their votes in other districts as wasted and count them in the list election.
Seed planted by Alan — 04 August 2009 @ 16:40
#48: if all Round first preferences had flowed to Shugart so that his final tally was 60 it would be necessary to calculate a fractional value for the Shugart votes to ensure that his surplus votes went into the election for list MPs.
I’m not sure this is true. Suppose instead that Shugart gets 10 (any 10 at random) and the remaining 10 become votes for Round’s party in the list tally. Since there are going to be no further transfers of either parcel, I can’t see why it matters which 10 go where (which is, of course, quite different from STV).
In the specific example, the election result is the same: Shugart gets the district seat and Round’s party list gets 10 votes. That may not be true in general, however, in which case my suggestion is not a direct substitute for Alan’s. But I’m not sure it wouldn’t be roughly as good. Given the difficulty of convincing people that fractional transfers in STV are OK, I’d want to put some effort into avoiding them here if possible.
#50: Well, okay, but one of the attractions of your proposal is that it does away with the need for two ballots (as does the “best losers” form of MMP). This still deserves some more effort.
By the way, I don’t harp on making PR fair for independent candidates and voters because I am one myself (I’m not, and in fact believe that those who blame political parties as such are badly mistaken). I do so because in the U.S., independents are an extremely important constituency for electoral reform. They have to be catered to.
Seed planted by Bob Richard — 04 August 2009 @ 16:55
Throughout this thread there has been a tendency to define a good electoral system as that which returns more independent or minor party candidates. Indeed, much of the Lundberg critique of AV as against TRS turned on the greater likelihood f such candidates being returned under TRS. My answer would be that TRS essentially removes that decision from the electors, to whom it rightly belongs, and gives it to the parties who negotiate withdrawals before the second round.
My view is that AV is superior to TRS precisely because it leaves the decision to the electors. I would argue strongly, while accepting the strictures of the Arrow theorem, that an electoral system should be as neutral as possible between different kinds of candidates and different kinds of parties. For the same reason under this rapidly evolving proposal I would allow both independents and party candidates to run in more than one district.
The Chirac/Le Pen election also gives gives a case where AV would clearly have been superior to TRS. It is extremely unlikely that Le Pen could have drawn sufficient second preferences to remain the second candidate under AV. The electors would then have had an actual choice instead of merely selecting between Chirac and le déluge.
It seems to me an exact analogy is the frequent, and misguided, call to limit dual candidacy in MMP systems.
Seed planted by Alan — 04 August 2009 @ 16:57
I thought the debate here was AV vs. FPTP (with some other proposals like Condorcet occasionally in the mix), not AV vs. majority runoff (or majority-plurality).
On the latter choice, I would agree with Alan.
Besides France 2002, there are probably several presidential elections under majority runoff that would have been different under AV. (I think we talked about Peru’s last election in this space as potentially such a case.)
By the way, I would eschew “TRS” (two-round system) because majority-runoff and majority-plurality are quite different. Both are used in France, but only the latter (for assembly elections) has much of an issue with elite negotiations over second-round participation. And then there are, of course, other two-round systems like Costa Rica’s 40% rule or the double complement rule, etc.
It seems to me that TRS is a family of systems, not a system, per se. (Its equivalent would be “ranked-choice single-winner”–RCSW!–which is a family containing AV, Supplementary Vote, Condorcet, etc.)
Seed planted by MSS — 04 August 2009 @ 17:36
Sorry for the threadjacking. Is there a chance of a thread for Mixed Member Preferential?
Seed planted by Alan — 04 August 2009 @ 18:19
As someone who writes open-source software, I’d still greatly prefer hand-counted elections.
Computers are complicated beasts with many layers: BIOS, firmware on every piece of hardware, an operating system, drivers, software libraries, and of course the actual counting application. In all this complexity it’s very easy for problems to go undetected. I do trust open-source software much more than proprietary software. But I’ve fixed plenty of bugs, and even caused a few, and this was in software that nobody had a particular interest in undermining.
I’ve worked in a local election, and it’s hard for me to imagine the amount of effort that would have to go into stealing thousands of votes in a hand-counted vote. There are multiple people involved at every stage, recounts can be done on the actual physical votes to check results, party representatives can observe each step. You would need hundreds of people in on the scam. In a computerized election, a malicious person with five minutes of access to a piece of firmware or a signing certificate could change millions of votes, and it would be very hard to detect.
Perhaps I’m being over-cautious. After all, computers do run important things like the stock markets, communication networks and nuclear missiles. But in an area where we can get along without them, and transparency is so important, I’m very hesitant to just go ahead with computer counting without much more thought than has been given to the issue by most governments.
Seed planted by Vasi — 05 August 2009 @ 01:31
In most Australian jurisdictions the votes are recorded and counted by hand as the first stage of the scrutiny. At a later time they are captured into a database to speed up the complexities of distributing preferences, but even then the physical count is generally already known.
The process is fundamentally one of recording and sorting. Every paper ballot that is issued is recorded (obviously the unique serial numbers are in now ay linked to the elector’s identity) and every ballot must be acquitted against the register of issued ballots.
On election night a series of finer and finer sorts are conducted. The ballots are tied up in bundles of 100 (that number may have changed since I last worked for the AEC). Obviously there are going to be 1 or more bundles that contain mixed papers, but all bundles are clearly marked. Random checks are conducted to ensure that the bundles are 1. pure and 2 contain the right numbers. Party scrutineers sign off on the accuracy of the bundle certificates along with electoral officials.
In the mythical District of Whatthe they might use bundles of 10, so there would 3 bundles marked Shugart 1 and so on. Lower preferences are also recorded on the bundle certificates. Reducing the number of papers moving around by a factor of 100 makes a huge difference to accuracy and transparency. As the bundles are signed off they are also assigned unique numbers and recorded in a register. You can track a bundle of votes that was cast in polling booth X in subdivision Y in electoral district Z and there would be hell to pay if the contents of that bundle did not match what was recorded on election night.
Once the election night scrutiny is completed, and obviously in an AV/STV system that is often not definitive because preferences must sill be worked out, the bundled ballots are transferred to regional offices and then to state offices, again each movement is recorded and acquitted.
In 2000 one read frequently that no count can be perfect. I simply do not accept that. A professional electoral organization can account for every ballot just as a bank can account for every cheque. A well-run paper election certainly does not involve some guy in a back room quietly saying 2673477, 2673478 and so on.
Any stage of the scrutiny can be challenged as a matter of right before the court of disputed returns, which is the High Court for federal elections and varies in the states and territories. Court challenges are not denounced as interference with the election and the courts have clear rules laid down by law for how to resolve disputes.
Seed planted by Alan — 05 August 2009 @ 03:00
Let’s see if I get Alan’s proposal straight.
Say you have 2 districts. And these are the results (with each district having 100,000 voters):
District 1
Republican 51%
Democrat 35%
Green 10%
Libertarian 4%
District 2
Democrat 45%
Republican 44%
Green 6%
Libertarian 5%
In district 2, there’s no winner. The Libertarian is eliminated and the 2nd choices go: 2% for the Republican and Democrat and 1% for the Green.
After this, there’s still no winner. The Green is eliminated and the 2nd choices go: 5% for the Democrat and 2% for the Republican. Now, the Democrat has won the seat.
Going to my point, we have the following. First, in district 1, the Republican has 1,000 votes in surplus; let’s say that 600 chose the Libertarian, 300 the Green and 100 the Democrat. Second, the Democrat has 2,000 votes in surplus. Here’s my doubt with the second part of this. Would the Greens get 2,00 votes for their list or not?
Seed planted by Derek — 08 August 2009 @ 02:46
Nobody gets any votes for their list. Wasted votes are counted in the list election according to their next available preference.
In each district the number of wasted votes will be the total number of votes less the quota. The value of those votes will vary in some cases according to their transfer value. I cannot do precise calculations for the transfer values on percentages and there is quite a comp;ex argument on which votes to transfer. Let us assume the last parcel rule in the Senatorial Rules which says that where a candidate is elected by transferred votes, only the votes in the last parcel are transferable.
In District 1 the Republican votes will count in the list tier at (roughly) 1/51st of their original value. The other votes will count in the list tier at full value.
In District 2, in your figures the Green appears to have 6% of the vote but on transfer they jump to 7%. Either way, those votes will be counted in the list election, along with all votes credited to the Republican (both original and transferred votes). The Republican’s votes will be counted at full value. The Green votes will have a reduced value because some of them have been used to elect the Democrat.
I would vary my original proposal to say that there are to be as many list seats as district seats and the quota is to be calculated in the usual way.
It would be much, much easier to tell what’s happening if you could give examples in the format I used for the district of Whatthe.
Seed planted by Alan — 08 August 2009 @ 06:02
The government of the Northern Territory is likely to change hands on the floor of the house next Friday. Despite wishful thinking to the contrary after the US midterm elections in 2006, this is actually a fairly rare event in parliamentary systems. The ABC will carry the debate live.
It would be unkind for me to say that the government faces defeat at the hands of independent MPs, because, as we all know from the original post, independent MPs are never elected under AV.
Seed planted by Alan — 12 August 2009 @ 13:15
Regarding that “wishful thinking” in the USA 2006, well, IF the USA were parliamentary and IF you had an election for the entire first chamber in which the governing party was defeated, I would submit that a change of government would have been fairly likely.
Of course, one does not often have such an election two years into the term of a government, at least not in “majoritarian” systems.
As for the NT, of course, when you have one of those crazy electoral systems that makes it easy for independents to win, these sort of floor defeats (or threats thereof) might happen from time to time. Independents are somewhat less predictable than party members, more or less by definition.
Other than one quotation above from someone else, it does not appear that the post/planting made much of the chances of independents in AV. I would not conflate independents and third-party candidates. I would also guess that smaller jurisdictions have a somewhat greater tendency to elect independents even in electoral systems generally less favorable to their chances otherwise. But that is a guess (or, in fancier terms, a hypothesis).
I do concede that I should learn a lot more about Australian subnational elections, and for that, I owe Alan and Tom thanks.
Thanks also for the link on the NT floor vote.
Seed planted by MSS — 12 August 2009 @ 16:54
If 2006 has been a parliamentary election I suspect Prime Minister Bush may have lost them by a larger majority. Whether Governor-general Cheney would have accepted his resignation and sent for the Leader of the Opposition is another question entirely.
I suspect any electoral system is more likely to return independent and minor party MPs as the the population of each district gets smaller.
Seed planted by Alan — 13 August 2009 @ 00:30
The NT vote of no confidence is interesting because generally the results of these things are known in advance, which is not the case in the territory. Just for the record, and in case of any last minute changes to the book, Australian territories have administrators rather than governors and chief ministers rather than premiers.
‘Chief executives and chief ministers’ has a certain ring)
Seed planted by Alan — 13 August 2009 @ 11:03
just for people’s edification, i have written a 6 page essay regarding how IRV *failed* to live up to its promise in the 2009 mayoral race in Burlington Vermont. like Aaron Armitage, i am a Condorcet proponent, and for the same reasons. i spell those reasons out in the essay and use the 2009 mayoral election in Burlington as a case study. (FYI, Burlington uses IRV since 2006. 3 years ago the IRV and Condorcet winner were coincidental. but this year, they were not.) if anyone wants a pdf copy of this essay, please email me at rbj@audioimagination.com and i would be happy to reply with a copy.
Seed planted by robert bristow-johnson — 15 August 2009 @ 02:40
If by “failed to live up to its promise”, Robert (#64) means failed to elect the Condorcet winner, well, AV doesn’t promise that. But I agree that the recent Burlington election is an excellent real-world example of the difference between AV and the Condorcet principle.
Condorcet proponents need to ask themselves the following question. Would I rather contribute to better elections, or wait literally forever for the best possible elections? I say literally forever because, in my opinion, no Condorcet-compliant voting rule is feasible outside of professional organizations, where they may be quite successful.
Seed planted by Bob Richard — 16 August 2009 @ 14:15
In #65 above, I should have said “wait literally forever for the best possible elections from my own point of view.” Otherwise, it appears that I am endorsing the Condorcet principle myself. In spite of Aaron’s fairly persuasive arguments earlier in this thread, I am not.
Seed planted by Bob Richard — 17 August 2009 @ 11:33
I’m actually somewhat attracted to Condorcet but there are 3 great problems to overcome before it can be taken seriously.
1. We would need to a script where someone like Matt Lauer successfully explains the way Condorcet cycles happen and get resolved.
2. There is the question of transparency raised up thread.
3. There is a larger underlying problem. In the election where I was the Condorcet winner but a spectacular loser by any other standard I would not have been the Condorcet winner if a small number of those who gave first preferences to the other candidates had not expressed second preferences. There is an ongoing debate in AV?STV about whether it is always true to say that expressing lower preference cannot harm ca candidate for whom you have expressed a higher preference.
My run for university senator shows that expressing lower preferences can spectacularly and decisively harm candidates for whom you have expressed a higher preference. It follows that Condorcet would seem to carry all of the same weaknesses as AV, but to a spectacularly greater degree. Advising the electorate how to vote becomes no more than a gamble.
Seed planted by Alan — 19 August 2009 @ 01:57
I’m not an expert on the differences among the multitude of Condorcet-compliant voting rules and can’t evaluate them in terms of the later-no-harm criterion. But I’m under the impression that, just as they differ with respect to the majority coalition criterion, they also differ with respect to later-no-harm, in degree if not in kind.
Alan is right. Later-no-harm is critical to voters, and must be critical to any reformer who wants to get her proposals adopted. But many of those who base their proposals on social choice theory think it is not at all essential and a few even believe that satisfying it is harmful.
Seed planted by Bob Richard — 19 August 2009 @ 15:11
Indeed. I recall one range voting advocate arguing on this blog that the absence of later-no-harm is desirable because it requires electors to deliberate about the possible voting choices of their fellow citizens.
I strongly suspect that, like range voting and approval voting, Condorcet would rapidly collapse to FPTP because it seems to me the only rational strategy for a candidate is to ask electors not to give any lower preferences at all.
And I’m still dying to see: (1) Matt Lauer’s teleprompter cards for Cloneproof Schwartz Sequential Dropping. (2) the fuzzy math talking points for candidates who are not part of the Smith set but are part of the Karl Rove set.
Seed planted by Alan — 19 August 2009 @ 18:00
FairVote Canada weighs in on the topic of this thread here. They suggest that in several Canadian elections AV would have increased the seat bonus given to the largest party beyond that party’s margin under FPTP. Asking (rhetorically) whether AV might lead to the ultimate adoption of PR, they write:
No. Societies rarely change their voting systems for parliamentary, legislature or council elections. When those scarce opportunities arise by popular demand, proposals for cosmetic change are diversionary and may make the legislatures even less representative.
Seed planted by Bob Richard — 24 August 2009 @ 19:12
The question of whether half a loaf is better than no bread - whether a winner-take-all system with less vote-wastage is a stepping-stone towards, or dissipates the pressure for, a proportional system - is vexing.
(1) On the one hand, in Australia it is certainly common for politicians and commentators to make a Goldilocks argument for AV - ie, “UK and USA use FPTP which gives minor parties no say on election day. On the other hand, Europe and Israel use PR which give minor parties too much say throughout the whole term of office. AV gets the balance just right.” Throw in “… and we even have PR for the upper house!” and you have a strong “best of both worlds” argument for the Australian mainland’s “norm” (4 States + federal model).
(Now of course this is contestable - I would say AV is not 50-50 PR/ WTA, but more 20/80 - and of course it may even be less proportional than FPTP; instead of the largest party winning 55% of seats on 45% of the votes, it might be cut back to 33%, which is no improvement. Likewise, some commentators have argued that AV is worst of both worlds regarding minor parties; they have great power over who forms the govt (Exhibit A: The DLP), but can avoid responsibility for the decisions and actions of that govt. But still, it is a very common meme in Australian politics.)
(2) On the other hand, as Andrew McLaren Carstairs noted in 1984, those European nations that adopted PR almost all had prior experience with “voting more than once” under winer-take-all - either MNTV, or a second ballot in single-seaters. Almost none had used the UK/ US system where you tick one candidate only in a contest for one seat only. He suggested that this may have made European voters more congenial to the idea of PR.
Moreover, Australia does seem to give empirical support for the proposition that party-list PR can be a stepping stone to STV, with list systems being replaced by STV very quickly after they are used (S Aust, 1970s: ACT, 1990s) or mooted (NSW, 1970s). Having said that, even if one attributes the cause to some Anglo-Saxon ethos, it doesn’t seem to have brought NZ, Scotland or South Africa closer to STV (at least at the national level).
So I’m still undecided, although leaning slightly towards (1).
Seed planted by Tom Round — 24 August 2009 @ 21:33
The fair vote Canada thing piqued my interest. I went back and took a look at the 2006 election results (I haven’t downloaded the 2008 results yet). At the time, I remember doing some rough calculations to see how the elections would have turned out using a single member majority system.
I think assumptions often made about voters’ second choices tend to be too facile, especially in Canada where the party system in one region may have different dynamics than in other regions. Would Liberals and Conservatives vote for each other in Quebec to elect federalist MPs over Bloc MPs? Probably, but the CPC in Quebec attracts very “soft” federalist support, so I’m not so sure.
Anyway, of the 306 ridings, 122 were won by outright majorities, so the results wouldn’t have changed using single member majority. However, there were an additional 62 ridings where the plurality winner got over 45% of the vote. I think its likely that in these situations the plurality winner would have drawn just enough votes leaked from other parties to get over the 50% mark, particularly under AV (with runoffs there is an extension of the campaign, so there is the chance of some new development changing the result).
The CPC won 54 ridings with a majority, and got over 45% in another 26, so they would have won at least 80 ridings using AV. The Liberals won majorities in 37 ridings plus came close in another 20, for a total of 57. The Bloc won majorities in 25 and came close in 7, for a total of 32. The NDP won majorities in 6 ridings and came close in another 9, for a total of 15.
This leaves another 122 ridings where AV could likely have changed the result. Of these in 56 the top two finishers were the Liberals and the CPC. Another 27 were Liberal-NDP contests, 17 were Liberal-BQ contests, 12 were NDP-CPC contests, and 10 were CPC-Bloc contests.
If you max out the Liberal victories in these 122 ridings, under the assumption that as the center party the Liberals would win the most second choice votes, then they would have won 157 ridings and a parliamentary majority, 37 by getting a majority of first choice votes, another 20 where they came close, then another 100 by prevailing in every case the counting went several rounds. This would have given them a majority of six. To do this they would have had to win just about every place they finished in the top two. Presumably under AV they could have converted a few third place first choice finishes into victories.
Actually I think the likeliest scenario would have been the hung parliament that actually happened. Maybe the Liberals could have formed a minority government with NDP support, but remember the 27 ridings where the two two parties were the Liberals and the NDP. The NDP maxes out at 27 ridings under this scenarios without starting to beat the Liberals in ridings where you might have gotten a changed result. Plus there is the issue that moving away from single member plurality will change the party dynamics in all sorts of subtle ways.
I think the likeliest result of a single member majority system in Canada would have been fewer Liberal -NDP coalitions, since in some of the parliaments where they historically needed NDP support they SMM would have given them enough extra seats to put them over the top.
Seed planted by Ed — 25 August 2009 @ 12:47
Way up at the top of this thread, MSS wrote:
On the other hand, for replacing two-round majority elections at the municipal level, especially in the case of officially non-partisan contests, the superiority of AV is clear to me.
What is it about local councils and boards, but not state legislatures and the federal House, that makes AV worthwhile for one but not the other? Is it the (nominally) non-partisan format? Is it something about two-round majority?
Or is it that PR is less important in local government, so that AV is less of a (potential) diversion from the real goal? If so, what is it about local government that makes PR less important there?
Or is it that AV is more likely to be a step toward PR rather than a diversion at the local level? If so, why?
Seed planted by Bob Richard — 25 August 2009 @ 21:37
Bob: simply the status quo against which AV is being compared, including the problem of (often very) low turnout in one or both rounds, and the often very long inter-round period, of existing municipal majority-runoff systems.
Seed planted by MSS — 26 August 2009 @ 12:33
MSS (#74), does this imply that you wouldn’t put any effort into AV for local legislative offices currently elected by plurality in single-member districts (city council, board of supervisors, school board)?
I’m assuming that AV for executive and administrative offices is worth the effort at any level of government, to replace both plurality and two-round majority.
Seed planted by Bob Richard — 26 August 2009 @ 17:09
In my limited experience (various cities and counties in southern California), FPTP is not used at the municipal level for councils and boards. All are MNTV or else majority runoff. I would prefer to replace either with STV (maintaining the existing district magnitudes, which would mean AV in the case of current majority systems).
That is a way of dodging your main question, Bob!
However, while I am still rather on the fence about whether AV is an improvement over FPTP–this thread has almost persuaded me that it is, even though it likely is not a step towards PR–it seems to me that the case in favor of AV over FPTP might well be greater under nonpartisan elections. My (not well thought out) logic here is that there is probably a higher probability of 3-way races in nonpartisan elections, and in the absence of partisan cues to assist voters in gathering information about who the leading candidates are, the ability to rank-order is a pretty big advantage.
Seed planted by MSS — 28 August 2009 @ 13:43
MSS is right, only a few California cities have single-member plurality councils — about 20 out of about 480, the largest being Santa Ana and Bakersfield. Somewhere between 75% and 90% use MNTV. Of the 58 counties, 57 use two-round majority (San Francisco uses AV).
I also think MSS is probably right about the frequency of three-way contests in non-partisan elections. I would wager that such contests are most common in partisan primaries, least common in partisan general elections (in a two-party system, that’s true by definition), with non-partisan local elections somewhere in between.
But I don’t think I agree that the absence of partisan cues is the main reason for wanting AV. I think it’s the damage done by vote-splitting when there there are more than two serious contenders. Add to that the facts, all previously mentioned, that the two round majority suffers from low turnout, high cost and second round campaigns that are either too short or too long.
But the elephant in the room is still whether AV increases or decreases the likelihood of adopting PR at a later time. We haven’t yet considered the possibility that it does neither, at least not to any material extent. Or that it sometimes does one and sometimes does the other in ways that we can’t predict and therefore shouldn’t worry about.
Seed planted by Bob Richard — 28 August 2009 @ 17:07
> “probably a higher probability of 3-way races in nonpartisan elections”
Doubtless because party preselection/ endorsement processes (whether a panel of half-a-dozen, a mass primary with a million voters, or something in between) both offer a de facto “first round” and also give some guidance as to how big each party’s support is. In non-partisan races, both are lacking (c/f Papua New Guinea).
Seed planted by Tom Round — 29 August 2009 @ 03:11
Those interested in this thread might also be interested in one by John Quiggin at Crooked Timber, prompted by a panel he attended at a recent conference on Logic, Game Theory and Social Choice.
Seed planted by MSS — 31 August 2009 @ 19:38
> “Think about an IRV election, and suppose that there is no strategic voting (I’ll argue that it won’t be needed, so voters will always vote sincerely. Now suppose that , after the votes have been cast, any candidate has the option to withdraw (there are some potential complications about the order in which this option becomes available to candidates, but I don’t think they matter in the end). Suppose that a candidate will only withdraw if by doing so, they will ensure the election of a candidate preferred by the majority of their voters to the candidate actually elected. I claim that this procedure is a Condorcet method. That is, it always selects the Condorcet winner, the candidate who would beat each of the other candidates in a run-off election, if such a candidate exists.”
I should be flattered that Qui-Ginn Jonn reinvented the wheel that I (thought I) discovered in 1992 (scroll down to Chapter 4, “Risk of Pyrrhic Victory”). As far as I could tell in that innocent, pre-Internet age, I was the first to think of adding a standing-down option to AV.
By the way, I have no idea who runs that German website, and as far as I know they never asked my permission to re-publish my Hons thesis (e-copies of which I had emailed to a dozen or so electoral reform anoraks on a mailing list), but again, I s’ppose, should be flattered.
Seed planted by Tom Round — 31 August 2009 @ 21:50
I would question the stand-down option. The electorate could reasonably ask why their votes got cancelled and replaced by an outcome negotiated by the parties.
Seed planted by Alan — 01 September 2009 @ 19:19
Would AV be better for societies with diverse political views and parliamentary systems than FPTP? (I would not presume to suggest what American reformers should do.)
Fair Vote Canada’s research has already been noted above. I will just add a simple comment: bad as Canada’s system is, we do have a House in which 28% of MPs are from third parties, while Australia has zero — clearly worse than FPTP. They have a thinly disguised two-party system: Labor or the Coalition.
An Australian Law Professor makes the case against AV: “Australian parliaments are controlled by a Coalition/Labor duopoly, whose participants happily alternate in power, and whose ministers need fear no independent thinking or scrutiny from their cowed backbenchers. We have one of the weakest legislatures in the democratic world. The Parliament here is under a degree of democratic control that would not be tolerated elsewhere. Australian parliamentary discipline is notorious: the last government to fall to a revolt by its own MPs was the Scullin government of 1931.”
Seed planted by Wilf Day — 05 September 2009 @ 11:13
Bede Harris is of course right that proportional representation for the House is an essential reform. The more senate-like the house becomes, and the sooner, the better. However, the house of representatives is not devoid of independent and third-party members although 3 independents is not nearly enough to disprove the anti-AV thesis. However, the Coalition is increasingly shaky both federally and in the States and territories.
The Labor Government in South Australia depends on National support in the assembly for its working majority. The Liberal government of Western Australia depends on National support for its working majority. The Labor government of the Northern Territory depends on independent MPs for its working majority.
Of the 5 AV lower houses in Australia, 3 have governments that do not enjoy a majority in their own right.
There is a long tradition in Australian political commentary of decrying the supine posture adopted by the lower houses and it’s broadly accurate, although no more so, I strongly suspect, than Canada or the UK. I cannot, for instance, see an Australian assembly where the progressives held a large majority deciding they hated each other so much that a conservative minority government was preferable to coalition.
I do not know that the fall of the Scullin government, where a section of the Labor caucus defied their electors, joined the conservatives, and then went down to defeat in their own districts at the subsequent election, is really one I’d cite as a glorious day for democracy. The reality is that in parliamentary systems governments just do not fall on the floor of the House all that often.
Seed planted by Alan — 07 September 2009 @ 01:58
What Alan said. I doubt that Bede (whom I once edited) is calling for FPTP as any improvement over AV.
Also, Australia is more politically and socially homogeneous than Canada. Language isn’t an issue, and apart from the Nationals being dominant (always in seats, sometimes even in votes too) over the Liberals in Queensland, the basic choice is between a Liberal and a Labor premier/ minister in all nine State and/or mainland [*] lower houses. Contrast Canada, with different govt/ opposition line-ups in nearly every Province and Federally. No Australian State has credibly threatened to secede apart from WA in 1931; but it remains a live issue in many regions of Canada. [**]
Despite Australia’s large area, more than five-sixths of the voters here live in the south-eastern triangle (from the Sunshine Coast to eastern Melbourne). Regionalism isn’t as big a factor here, and where it is, it favours Independents with local roots, rather than minor parties with national organisation but regional strongholds (eg UK Lib-Dems).
While there is a good case for Prof Domenico Fisichella’s argument that AV, like Runoff, can be used to “lock out” “anti-system” parties that are viewed as beyond the pale (the German Social Democrats before 1918, the French Communists, and One Nation in Australia 1996-2001), it is not often used in Australia this way. It might have been had the Democrats ever looked like seriously coming second in any electorates. It may yet if the Greens do. Yet I wouldn’t underestimate Coalition opportunism; the Libs/ Nats may yet preference the Greens just to make life harder for Labor. (Labor has at times done the same for the DLP).
===================
[*] ie, Commonwealth, all States, and the two mainland Territories. Norfolk Island is a world unto itself… non-party mini-council elected by semi-cumulative vorte (like the Swiss system without the party lists), controls its own immigration, doesn’t really want to be part of Australia: a sort of transplanted-Cornish Puerto Rico.
[**] If you ask me, the “Red” States/ Provinces of North America should form one federation, while the “Blue” Provinces/ States should form another, with a customs union of course. But no one asks me…
Seed planted by Tom Round — 07 September 2009 @ 21:35
I think Bede is, to a certain extent, a victim of the Whig Theory of Parliament. Where the Whig Theory of Hisotry claimed that everything in England was wonderful until it became even more wonderful, the Whig theory of parliament is that there was once a golden age when parliaments regularly overthrew cabinets on the highest of principles and MPs were Platonic guardians of the good, the true, and the beautiful.
The reality could not be more different. The right of the House of Commons to overthrow a cabinet was not clearly established until the Bedchamber Crisis of 1839. Moreover the British parliament, even in 1839, even after the Great Reform Act, contained large numbers of rotten boroughs whose MPs voted precisely as their patrons directed and was elected on an extraordinarily narrow franchise.
Australian parliaments were among the first in the world to introduce payment of MPs. The Labor Party is actually one of the oldest continuous parties in the world. Pledged MPs were seen as a democratic reform, not a grubby departure from the golden age of independence.
None of which justifies the tight party discipline that now prevails on both sides of the aisle in the House of Representatives, but it needs a better alternative than hurrying back to the glory days of parliament in 1750 or something.
Seed planted by Alan — 07 September 2009 @ 22:35
It helps to analyse the incentives, from the average voter’s perspective, to elect an independent vs a party MP.
I work on the assumption that:
(1) In an assembly with a disciplined majority party or coalition,
(1.1) an MP in that majority has (say) 7 “power units” as far as their constituency goes (via lobbying ministers, voting in caucus, possibility of Ministerial appointment, etc).
(1.2) An opposition or cross-bench MP has only, say, 2 “power units” (mainly via asking awkward Questions, and pleading with Ministers over non-party “roads and rats” details).
However,
(2) In a “hung” assembly (where no party/ coalition has a majority),
(2.1) an MP in the largest party caucus (which presumably forms a minority government) has 4 “power units”
(2.2) an MP in the opposition caucus has 3 “power units”; but
(2.3) a cross-bench MP now has, say, 6 “power units”, because s/he can credibly threaten to bring down the government.
In other words, if you elect an Independent but majority govt is the norm, your MP will be perpetually sidelined with less influence than a major-party MP. On the other hand, In other words, if you elect an Independent but hung parliaments are common, your MP may end up a kingmaker with arguably more influence than a major-party MP. (I’m sure many ALP MPs in ACT or SA would have liked a Ministerial appointment too, but the Labor chief minister had to appoint Michael Moore and Karlene Maywald to keep a majority).
It is perhaps significant that, in Australia, Independents who have gained a high profile did so when the government had either a very slim majority (John Hatton in NSW, mid-1970s) or none at all (Martyn Evans in Sth Aust, 1990s; Liz Cunningham in Qld, 1995-98).
Seed planted by Tom Round — 07 September 2009 @ 23:07
The other motive for electing an independent MP in rural and regional Australia is the continued weakening of the National Party outside Queensland. Areas like New England, in northern NSW, once National strongholds (once you left the Brisbane outskirts you drove through National electorates all the way to the Hunter Valley just north of Sydney) now elect independents at both state and federal levels. Their feeling seems to be the Nationals get rolled by the Liberals in government and even more by Labor when the Coalition is in opposition. It’s notable that the WA and SA Nationals are both pursing a much more independent line and less socially conservative polices, although the two branches are in coalition with different parties at the State level.
Seed planted by Alan — 08 September 2009 @ 08:13
Seed planted by Tom Round — 14 January 2010 @ 19:55