This is a point made against electionmethods.org on this page, which seems to be part of an electronic voting systems discussion. I’ve just included it for fairness, because it speaks against the site I’ve used above.
This one has a fairly broad discussion of why plurality systems don’t elect everyone’s preferred candidate in some types of elections, and that the problem comes from assuming that a method that works fine for ‘yes or no’ types of questions works for elctions with 3 or more choices.
I’ve discovered there is an initiative here in Washington State to implement Instant Runoff Voting.
I’m having a little trouble deciding wether to back this initiative because IRV is so much better than the current system, or to continue talking about Condorcet as an even better choice.
I’ll probably back it, since would be a vast improvement. Also, the voting process the voter goes through is the same with both IRV and Condorcet, so once IRV is here it’ll be easy to change to the better system.
Sorry to keep posting to my own thread repeatedly, but each of these are sort of separate, and others may want to deal with each one individually.
I have a question for those who know how IRV is used elsewhere (like Australia). One of the problems pointed out with IRV is what electionmethods.org calls the Summability Criterion (SC). I’m wondering how this is dealt with in elections where IRV is used.
Here in the States during a vote on a state-wide issue, each county counts the votes from all its precincts, and the totals are passed up to the state level to find the totals for the state. This pass-up involves only the tabulation, and the actual ballots remain with the county.
With IRV, it seems there’s no way to tabulate the results for such a transfer to the state level. In order to fairly calculate the total for the state, every ballot must be added in at the same time. While this can be done using computers fairly easily, it makes the process more opaque to verification when all totals involve potential millions of ballots, rather than being able to look at each precinct and county separately.
In Australia (or any other use of IRV), how do they do this? Do they count all ballots at once in some central agency? Or is the IRV count done at some lower level and then those winners combined by some other rules at the higher levels?
Interesting ideas being floated around here, but I have a question: Under any form of preferential voting, is there an option for leaving a candidate off my list? For example, could I vote #1 - Libertarian, #2 - Republican, #3 - Independent and stop right there? Not cast a ballot for a Democrat under any circumstances? Or do I have to fill in all spaces?
If there are only 4 candidates, then only rating 3 of them would be equivalent to ranking the remaining one last.
It seems that if you want to effectively vote against all Democrats, you should specify that they are your last choice. In all the systems, I think your vote would never go to your last choice.
If there were 5 or more candidates, there’s an ambiguity if only 3 are ranked. I can’t find any specifics on how those ambiguities must be resolved, so I’d assume that a single implementation of one of these systems would have to specify its own rule for what to do. Does anyone know how these ambiguities are dealt with in Australia?
Me, again. I have one example to provide to silenus.
In the proposed IRV initiative starting here in Washington, they specify the term ‘exhausted ballot’ to mean ‘a ballot on which there are no uncounted choices for remaining candidates, i.e., all choices made on the ballot have been counted as potential votes for the various candidates or contain choices for eliminated candidates or both and no other choices remain to be counted.’
It says that when all choices specified on an exhausted ballot have been used in earlier stages and all have been eliminated, the ballot is discarded. Essentially, it’s like that person had stayed home.
A ballot is also considered exhausted if two candidates have been given the same ranking and that ranking is the most relevent one at a particular stage (neither of the equal-ranked choices counts as a vote).
Thank you very much for that answer, Saltire ! Appreciate it, as it gives me something to chew on this weekend. So I could basically give my top three or four choices, and refuse to vote for anybody if those got eliminated. Hmmmm…much to consider.
But there’s another way to look at it, one that doesn’t require any of those people to be savvy, sneaky, or strategic voters: Just reverse the order of the two hypothetical elections.
We start out the day before election day:
6 people plan to vote Republican, Libertarian, Democrat.
2 people plan to vote Libertarian, Republican, Democrat.
5 people plan to vote Democrat, Republican, Libertarian.
4 people plan to vote Libertarian, Democrat, Republican.
The two “L, R, D” voters read a news article the next morning that convinces them the Republican is a better candidate than the Libertarian, so they rank the Republican first, causing him to lose even though he would’ve won if they’d left him at #2! By voting their true preference, those voters have put their last choice (the Democrat) into office instead of their first choice.
Well, it probably wouldn’t have that sort of activity on purpose, but it could still happen by accident. The problem with IRV is still there: ranking a candidate higher can make him more likely to lose, and ranking a candidate lower can make him more likely to win. Even if no one intentionally votes against their preference, the election results still can’t be taken as an accurate reflection of voters’ preferences.
I have been supporting a modified form of IRV but am interested in Condorcet. The main problem for me is that I don’t agree with the Condorcet Criterion which states that if one candidate is preferred over each of the other candidates, that candidate should win. I don’t think anyone should win without the approval of a majority of the electorate. I wouldn’t consider someone an “Ideal Democratic Winner” just because they were prefered over the other ten candidates when only 2 or three out of every ten voters may have indicated any preference at all for the candidate. Is there any way to include a majority requirement to a Condorcet voting system? If all else fails perhaps by having the 2 most popular candidates move on to a run off election?
Also, I noticed in the ElectionMethods.org site that ties have to be resolved somehow. What happens if 2 or more candidates have equal support?
2sense, I have 2 comments about your desire for a majority requirement.
1)The current plurality system doesn’t require a majority, and in fact it seems true majorities are rare in, for example, presidential races.
IRV doesn’t require a majority, either. It tries very hard to generate a majority from the preferences of the voters, but it is certainly possible for the winner to merely be the candidate with the most votes after all stages of runoff, even if that doesn’t constitute a majority of all voters.
I’m afraid I can’t answer you on what is done in the case of a tie. Of course, an actual tie would be very rare in any election with more than a few voters. I’m not sure of the actual math, but I’d guess it would be just as unlikely as a tie in IRV. If a tie were to emerge, I’d guess a runoff would be necessary (but I’m really just guessing, as I don’t see any tie-resolution info on the sites).
FYI, in a recent referendum, San Francisco adopted IRV for city elections. The Center for Voting and Democracy has prepared a computer animation of the process, which you can link to here: http://www.fairvote.org/sf/vote/index.html
Oh, and San Francisco’s first election under IRV will be the election to the Board of Supervisors (same thing as a city council) this November. Here’s an article from the San Francisco Chronicle:http://www.fairvote.org/articles/sfrunoff.htm[/ur]
Actually, after IRV was approved in a 2002 referendum, the system was supposed to be used for the November 2003 elections for mayor, district attorney and sheriff, but the elections office dragged its feet on updating equipment and changing regulation, etc., so that election was held under the old system – which meant that a traditional, second runoff election had to be held (and paid for) between the top two finishers for the mayor’s office, the Green Matt Gonzalez and the Democrat Gavin Newsom. Newsom won the runoff.
Yes but I’m not comparing Condorcet with our current presidential electoral method, which you may have noticed I’m not a big fan of, or against regular IRV. As I said I am currently supporting a modified form of IRV. This form does require a candidate to acheive a majority of all ballots to be declared the winner. If not then there must be a seperate runoff election between that candidate and the last one to be eliminated. I’m wondering if something similar ( or better ) can work with Condorcet.
Notice that my system not only guarantees a majority winner but also satisfies another criterion that Condorcet ( or IRV or First Past the Post for that matter ) doesn’t and that’s Ballot Relevancy. In those systems every ballot isn’t necesarily relevant. Under Condorcet if a person doesn’t approve of any candidate and submits a blank ballot it is irrelevant. When comparing preferences the ballot contains none and so is ignored. In my system every ballot counts every single time the votes are counted. Even if you don’t list any candidates your ballot still counts toward the total raising the number a candidate needs to reach a majority.
The difference is that the math behind IRV is simple addition. I understand it and can grasp the consequences of differing tie breaking procedures. Condorcet sounds good as far as it goes but is more complex and I don’t have that assurance.
I’m not convinced that that added runoff really makes the process any more democratic. I’d expect that those that preferred one of the two candidates over the other will vote that way in the runoff, and those that didn’t prefer either will stay home or hand in a blank ballot. Either the runoff comes out for the guy with the plurality in the original election or, if the blank ballots are numerous enough, the election again fails to select anyone. Either way, you’ve spent a bunch of money on a runoff without gaining anything.
What happens if the blank ballots prevent anyone from getting a majority? I don’t think it’s practical to have rules that might make it possible to invalidate the election entirely. While it would be great if all elections ended in a clear mandate (a majority), it isn’t reasonable to expect that to happen every time.
Condorcet is also simple addition, only you need multiple totals running at once.
Have you read the link I provided above about the Summability Criterion? While it doesn’t bear on tie-breaking, it does show how counting votes in IRV is much more opaque than other methods, because all ballots (potentially from an entire state or even the whole country) must be examined during each stage of the counting. There is a lot more data to work through, so it’s harder to be sure there’s no funny business going on.
All in all, I don’t think tie-breaking is a big problem. While it would be important in a situation where there were only 20 voters or so, I don’t think ties would ever occur in large elections. I don’t have the math to prove it, but I expect the odds of a tie in a statewide election is probably similar to the chance of being struck by lightning while holding a winning lottery ticket.
My fault. I didn’t note that blank ballots are verboten in the runoff. There are only 2 choices: the 2 most popular candidates from the general election. There aren’t any blanks so one of them must receive a majority.
I am uninterested in speculating about voter behavior. My concern here is for guaranteed outcomes. ( Well, guaranteed in theory anyways. ) We can’t know how people will vote in any given election. Some people will change their minds between elections, others will rethink their decision to go to the polls. I don’t see how any of that matters. We are dealing with a seperate electorate in a runoff election. How the first electorate voted is only important to determine the 2 candidates. Sure that’s manufacturing a majority somewhat but it’s the most democratic system I have come up with. If you know a better way I’m all ears.
Yes. I agree, that’s troubling.
Noted. Anyone else care to weigh in on possible tie breaking procedures?
Isn’t there a method of voting, I think called the Borda voting method, that is a preference-rank system that’s based on Game Theory, and tends to optimize outcome by essentially making everyone as happy as posssible with it? By that, I mean, you get the maximal amount of votor satisfaction, even if the majority of people don’t get their first choice?
I’d much prefer that to the idiot system we have now, where the majority can actually get their least-preferred, e.g. the 2000 election.
Borda voting is a method where the rank of each candidate on each ballot is translated into a number of points. If you vote Republican, Democrat, Libertarian, then the Republican gets two points from you, the Democrat gets one, and the Libertarian gets none. You must rank each candidate on your ballot, and whoever has the most points at the end is the winner.
Australian here. We solve that particular problem by requiring that all boxes be numbered in order - so 1, 2, 3, (blank), (blank) or 1, 2, 2, 3, 3 are both unacceptable. Such “donkey votes” aren’t counted.
Saltire’s question about state-wide issues:
We have what you call IRV and we call preferential voting only for single electorates. 1 electorate, 1 count, 1 candidate chosen. All votes are hand-written and counted by hand. As election night continues, the latest tallies (500 votes for X, 100 for Y, 700 for Z) are sent electronically to a central organisation. Based on general voting patterns, it’s generally possible to guess most seats at this point if the margins are wide enough. So you can predict many elections only an hour or two after they start counting, but you know that there’s still a few seats remaining to be decided. In a close election, these may be critical, so we may only know the final state-wide or nation-wide result a few days later.
For the national Senate, we have a form of preferential proportional representation. It gets complex, but each state selects 12 senators and the territories select 2. That means that 10 or so senators from each state probably represent the major parties (Labor and Coalition), but the last few senators from each state can be chosen by minor parties or independents.
To change the constitution, only yes / no questions are put to the voters (ie “Do you approve the proposed constitional change of _____”, please write yes or no"), so there’s no preferences required. Does this answer your question?
Here in Norn Iron we use a system which seems very similar to that described by Cunctator. Proportional Representation. We usually have about 20 candidates in each area, and there are no rules about how many candidates you have to put in preferential order - you can just choose your no. 1 preference if you want. I rarely choose more than 4 or 5.
The recounts sometimes go on into the night, but at least one has the satisfaction of having some say in who is elected - even if your first preference candidate gets eliminated. The adoption of this system has brought more of the minority candidates in the Assembly.
Another Australian here. Firstly, those aren’t what are usually called “donkey votes”. A donkey vote is where a ballot paper is number in the order of the candidates, apparently to make the numbering by the voter easier regardless of the actual preferences of the voter. In Australia, where voting is compulsory, such behaviour occurs often enough to make a difference, though there have been some steps taken to reduce the effect (such as arranging names randomly on ballot papers rather than in alphabetical order).
Secondly, in many jurisdictions in Australia, such “exhausted ballots” are counted up to the point where the preferences are not given or are ambiguous. So 1,2,3, (blank), (blank) would have three valid preferences, but if it came to a count where the decision was between the last two, it would be put aside as “exhausted” (and not “informal”). And 1,2,2,3,3 would be a valid vote for the first candidate, but if she or he dropped out of the count, it would be “exhausted” because of having two 2’s.
There is a well-known voting paradox, which says that given some collections of preferences by electors, you can’t get an unambiguous answer as to which one candidate is preferred. But if you have to choose one candidate out of three or more, tyhen some kind of run-off system will give a fairer result than the “first-past-the-post” system used in most elections in the US.
