User:Abd/IRV stats

On the San Francisco 2000 election
I think we have some garbage in above, leading to some garbage out. The article currently has, as Ask10Questions noted, "Records of the San Francisco Department of Elections show that before IRV, a runoff round was required far more often than not in that city, where in 2000, 9 out of 10 contested races went to runoffs." First of all, that source does not contain that statement. If it is true, the reader must do the analysis. This claim is not properly sourced by Wikipedia standards. But I'm not, myself, averse to a little original research. I don't attempt to remove claims from the article that I consider true and balanced, placed there by someone else, merely because it is not properly sourced. Others may. So can I confirm it?

The runoff results for the December 12 election show two-candidate runoffs in Supervisor districts 1, 3-6, 7-9, and 11. Out of 11 districts, there were runoffs in 9. One district was not contested, district 2.

These were the election results by district. The first number is the total vote. The rest of the numbers are vote percentages for candidates on the ballot. Write-ins in the contested elections were a small percentage. The ultimate winner is shown in bold.

1. 24211  22%    3%   28%    9%   38%  2. 27070   92% (uncontested) 3. 20714  15%   13%   11%   11%    3%   11%   37%  4. 24617   17%    2%   44%   15%   16%    3%    3%  5. 30125    4%    2%   42%   12%    3%    0%    4%   29%    2%    1%    1%  6. 18738    1%    4%    2%   14%    3%    3%    3%    2%   12%   16%    1%    3%   32%    0%    1%    1%    3%  7. 30229   44%   11%    2%   22%    9%   12%  8. 34178    6%   50%   34%    3%    4%    4%  9. 20972    5%    6%   14%   68%    7% 10. 19764    3%    9%    4%   33%    5%    2%   21%    4%    8%    9%    2%    1% 11. 21409   10%   29%    4%    2%   19%    4%    2%   27%    2%

There is a discrepancy in the report of election data. The candidates shown as for District 10 on the first election are given in the runoffs as District 9 (which did not have a runoff).

I find these numbers extraordinary. Yes, there were seventeen candidates on the ballot in District 6. The reason why there are so many runoffs is that a very unusual number of candidates are running, the vote is being heavily split. Top-two runoff (and IRV, as well as other reforms beyond Plurality) encourage additional candidates to run. Thus a majority failure problem that might be rare with Plurality becomes the norm with Top-two, at least in this environment. What do we see with other elections on that ballot?

U.S. Senator: 307853	0%	72%	0%	1%	0%	15%	11% U.S. Rep, District 8: 274162	9%	2%	66%	1% U.S. Rep, District 12: 58547	2%	78%	3%	16% State Senate, District 3: 146335	14%	6%	80% State Assembly District 12: 128469	18%	82% State Assembly District 13: 146892	6%	78%	15% BART District 7: 31166	27%	53%	10%	10%	1% BART District 9: 100221	76%	23% (I have not reported results for two multiwinner elections.)

Is the statistic true that in 2000, 9 out of 10 contested elections went to runoff? What elections? The report is obviously considering only Supervisor elections. I don't know if a majority is required for the other city elections on the ballot (The two BART districts, though those aren't actually city elections, I think.). I'd guess not. There were 11 supervisor elections, with extraordinary fields of candidates. That's what top-two runoff can do, likewise IRV, and other preferential voting systems and Approval could do the same, particularly with non-partisan elections. However, the Plurality elections (definitely the Federal and State elections would be that) are uniformly generating majority victories in 2000. Looks to me like top-two runoff is creating a need for IRV or other such reform in San Francisco.

Why did they have top-two runoff? Likely because it is considered desirable to have a majority winner; however, almost certainly, they would usually have a majority winner with Plurality. (That is, there would be not so many candidates, or minor candidates would not so effectively split the vote.) But for the exceptions, they want runoffs. Now, comes IRV and promises to do this in one election. Sounds good. However, problem is, IRV not only can fail to deliver a majority winner, in fact, but it actually is doing that in San Francisco. I'll get to the fact of that later, but consider the logic: We have top-two runoff because we want a majority winner, but we can save election costs with IRV, thinking we'll get the same result. But we don't. Instead we get Plurality winners, at least some of the time. One of those runoffs in 2000, it can be noticed, had a 50% vote in the first round for the ultimate winner. That is, it was 50% after roundoff (or exactly 50%, not a majority). They held a runoff for the last half percent, at most, to reach a majority. One can see why some jurisdictions have reduced the necessary win to 40%, only holding a runoff below that point.

Notice that the eleven elections, in eight cases, simply reproduced what Plurality would have done; plus we can assume that with top-two runoff, there is greater sincere voting than with Plurality; quite likely with Plurality, most or all of the other three cases would have had the same winner. The argument for Plurality is stronger than many of us might think.

But there is a better and far cheaper way: let people vote for as many as they like, Count All the Votes, candidate with the most votes wins. Or, if one insists on being able to designate a favorite, Bucklin voting. Precinct summable, counts on standard equipment, no voting machine conversion costs. No method guarantees a majority winner, to be sure, except repeated balloting, which sometimes takes a *lot* of ballots.

As to what is in the article, it's misleading. For starters, to figure out what was happening, I had to analyze the results. This was original research and interpretation. The exclusion of the uncontested race from the statistic is arbitrary. More neutral would be a report that nine out of eleven Supervisor races, due to a majority victory requirement, required a runoff. But saying this without pointing out that top-two runoff could be causing an increase in candidates, thus causing increased incidence of majority failure, introduces bias. It's not going to be simple to get unbiased presentation of this information into the article; I suspect that the situation in San Francisco is quite unusual. No wonder they were the first major city to implement IRV! We will need more information, and I have a strong hunch that we won't have reliable source for it.

I'm taking it out, not because it is not true, but because, without better explanation and context, it's reducing a complex question to a simple summary, making IRV look necessary. (Some claim that SF doesn't have IRV at all, and they sure don't call it that; the limited ranks, with large candidate fields, make exhausted ballots and resulting plurality elections much more common. That is, San Francisco's Ranked Choice Voting encourages more candidates to run, but the ranks are limited. Interesting question: why didn't they call it "Instant Runoff Voting."?

And wouldn't STV for the Board of Supervisors do a better job representing the city? Single-winner districts leave huge chunks of the population, often over have the voters, not to mention the population, without chosen representation. Here my own opinion would converge with the goals of FairVote. --Abd (talk) 05:48, 31 December 2007 (UTC)

More on San Francisco RCV elections: consistent majority failure.
I've gone over the election results available for San Francisco Ranked Choice Voting elections. They are remarkably consistent. Out of 20 RCV elections that have been held since the referendum establishing it passed, the following statistics obtain:

4 elections were uncontested. 9 elections were won with a majority in the first round. 7 elections went into RCV vote transfer. For every one of the 7 elections using RCV, the candidate with a plurality in the first round went on to win the election after transfers were complete. In every one of these elections, the winner did not obtain a majority of ballots cast. In every one of these elections, the runner-up in the first round remained the runner-up in the last round.

One election had 22 candidates; the winner had 33.65% of the vote 37.63% of ballots containing a vote for an eligible candidate.

Below are more detailed results. What I notice is that the relative position of the winner and the runner-up did not much change with vote transfers.


 * There was an error found in these figures; the total vote reported includes ballots not containing any vote for the office. Commonly, a small percentage of ballots will be like that. Following normal rules, these are not considered "valid votes cast for the office" and would not represent part of the basis for a majority. I'm not sure what the San Francisco rules were, but, pending determining that, I'm assuming that the Robert's Rules standard is being followed. Note that these elections allow write-ins (as did the prior runoff elections), so a protest voter may simply write in the name of any eligible candidate. Voters may even write in their own name, usually, which is why attempts to keep ballots secret even if the voter wants to reveal whose ballot it is are in conflict with write-in rules, which are actually fundamental, a safeguard against serious political manipulation of elections, though this requires an aware electorate. Eternal vigilance is the price of liberty....


 * Well, I started to look at it, and it just goes to show. It appears that San Francisco disallows most write-in votes for these RCV offices. Not good. There is one election with a "qualified write-in," so I was technically correct in what I wrote above, but I really wonder how many voters don't know that rule? And so effectively invalidated their ballot by writing in a sincere vote? The ballot images aren't available for these elections, at least in one archive. This will make the results for IRV appear a little more favorable in terms of approaching a majority. And this means I need to look at that IRV ordinance in detail. I'm pretty sure San Francisco used to report all the write-ins. --Abd (talk) 18:52, 24 January 2008 (UTC) (Later this day added the text of the write-in provision as described in voter instructions in 2002:)

QUALIFIED WRITE-IN CANDIDATES (RIGHTS OF VOTERS) — A Qualified Write-in Candidate is a person who has turned in the required papers and signatures to the Department of Elections. Although the name of this person will not appear on the ballot, voters can vote for this person by writing the name of the person in the space on the ballot provided for write-in votes. The Department of Elections counts write-in votes only for qualified write-in candidates.


 * Holy Bovine refuse! As to what is below, looks like somebody (actually a whole community of people) screwed up big-time! Have any lawsuits been filed? Here is the text from the voter information pamphlet explaining what was being voted on in March 2002 when IRV was implemented:

''THE WAY IT IS NOW: When the offices of the Mayor, City Attorney, District Attorney, Public Defender, Sheriff, Assessor-Recorder, Treasurer, and Board of Supervisors are up for election, voters may select only one candidate for each of these offices. If no candidate receives more than 50% of the votes cast for the office, the two candidates who receive the highest number of votes compete in a run-off election at a later date.''

''THE PROPOSAL: Proposition A is a Charter amendment that would require the City to use an instant run-off voting method that would eliminate separate run-off elections. A winner would still have to receive more than 50% of the vote.''


 * And then the sequential elimination process is explained. Sound familiar? This is deja vu all over again; this is the Robert's Rules issue. The explanation:

[...] This process of transferring votes to the voter’s next-choice candidate and eliminating candidates with the fewest votes would be repeated until one candidate received more than 50% of the votes.


 * The meaning of "votes" is not explicit in the text. From the introduction, however, a reasonable voter would, made aware of the issue, assume that this was the same as "first preference votes" mentioned explicitly in the beginning of the explanation. No statement is made in the ballot summary that the basis for majority is reduced. However, a reasonable voter might also notice that there is no mention of what happens if there is no majority; the statement by the committee, though, that "A winner would still have to receive more than 50% of the vote" would imply that the election could fail. The IRV interpretation would be that it *cannot* fail. In fact, just continue the elimination process one step further, and you can require that the winner get 100% of the vote! Imagine that the eliminations go one more step -- exact same process -- and then it was claimed, "The winner would still have to be elected with 100% of the vote."


 * Even if the ordinance itself is explicit [as it turns out it is, see below], there would then have been an incorrect explanation by the Ballot Simplification Committee, and this *could* be a basis for a challenge to an election. As far as I know, no such challenge has been filed, and the deadlines would have expired. But, come the next election, with the cat out of the bag.....


 * The ballot arguments did not address the majority issue at all. San Francisco was the first successes for the modern preferential voting movement in the United States, and the opposition was totally caught flat-footed, they did not understand the issues, they made very weak arguments. ("IRV will give some voters three votes instead of one." Sound familiar? That's the argument in Brown v. Smallwood." Corrupt argument, in fact, though there is *some* basis for it. IRV, by discarding and totally disregarding votes for candidates eliminated in the process, is, in fact, making some votes count for more than others, particularly in the presence of actual majority failure. But this is a more sophisticated argument than the opponents of the measure were aware of, apparently. Another argument was, "no city uses this." Which was true at the time, and is still *almost* true. Only Burlington and Cary have actually made use of IRV in an election where it made any difference at all. Two elections total as this is written.)


 * What's in the ordinance itself?

''SEC. 13.102. INSTANT RUNOFF ELECTIONS.

''(a) For the purposes of this section:

''(1) a candidate shall be deemed “continuing” if the candidate has not been eliminated.

''(2) a ballot shall be deemed “continuing” if it is not exhausted; and

''(3) a ballot shall be deemed “exhausted,” and not counted in further stages of the tabulation, if all of the choices have been eliminated or there are no more choices indicated on the ballot. If a ranked-choice ballot gives equal rank to two or more candidates, the ballot shall be declared exhausted when such multiple rankings are reached. If a voter casts a ranked-choice ballot but skips a rank, the voter’s vote shall be transferred to that voter’s next ranked choice.

''(b) The Mayor, Sheriff, District Attorney, City Attorney, Treasurer, Assessor-Recorder, Public Defender, and members of the Board of Supervisors shall be elected using a ranked-choice, or “instant runoff,” ballot. The ballot shall allow voters to rank a number of choices in order of preference equal to the total number of candidates for each office; provided, however, if the voting system, vote tabulation system, or similar or related equipment used by the City and County cannot feasibly accommodate choices equal to the total number of candidates running for each office, then the Director of Elections may limit the number of choices a voter may rank to no fewer than three,. The ballot shall in no way interfere with a voter’s ability to cast a vote for a write-in candidate.

''(c) If a candidate receives a majority of the first choices, that candidate shall be declared elected. If no candidate receives a majority, the candidate who received the fewest first choices shall be eliminated and each vote cast for that candidate shall be transferred to the next-ranked candidate on that voter’s ballot. If, after this transfer of votes, any candidate has a majority of the votes from the continuing ballots, that candidate shall be declared elected.

''(d) If no candidate receives a majority of votes from the continuing ballots after a candidate has been eliminated and his or her votes have been transferred to the next-ranked candidate, the continuing candidate with the fewest votes from the continuing ballots shall be eliminated. All votes cast for that candidate shall be transferred to the next-ranked continuing candidate on each voter’s ballot. This process of eliminating candidates and transferring their votes to the next-ranked continuing candidates shall be repeated until a candidate receives a majority of the votes from the continuing ballots.''


 * Classic bait and switch. There is direct conflict between the Ballot summary and the Pro arguments in the ballot, and the actual ordinance. Some of this belongs in the article. Anyone care to put it in? Now, back to our regularly scheduled programming (correcting the figures below). --Abd (talk) 20:04, 24 January 2008 (UTC)


 * Pursuant to a comment from User:Tbouricius, I am including information below on the number of candidates remaining when counting stopped: he pointed out that the 22-candidate election was stopped when the winner had a majority; not all candidates had been eliminated at that point, so the winner might have obtained additional votes. I will also, in the future, add Bucklin results to this analysis, because Bucklin uses the same vote data as is available from San Francisco RCV. (The spreadsheets give first place votes, second place votes, and third place votes. That's all Bucklin needs.


 * (This is a study of how IRV is performing, simply compiling and reporting data from the San Francisco public records. Other users are invited to correct errors here or to add information considered important that I have not reported. --Abd (talk) 05:29, 27 January 2008 (UTC)

The data was taken from the San Francisco Department of Elections Archive

San Francisco election for Board of Supervisor, November 2, 2004.

District 1: 7 candidates. First round plurality: Jake McGoldrick, 11815/28787 votes plus 1934 ballots with no vote for an eligible candidate. runner up: Lillian Sing, 8989 After transfers: Jake McGoldrick, 14011/28787 votes, 48.67%. runner up: Lillian Sing, 11929

District 2: 5 candidates, won with 61.25% first round.

District 3: 4 candidates, won with 62.55% first round.

District 5: 22 candidates. First round plurality: Ross Mirkarimi, 9947/35109 plus 4146 ballots with no vote for an eligible candidate. runner up: Robert Haaland, 5124. After transfers: Ross Mirkarimi, 13211/35109 = 37.63% runner up: Robert Haaland, 7272.


 * Note: this election terminated with three candidates, Mirkarimi having reached 50.596% of the unexhausted vote. Besides Robert Haaland, Lisa Feldstein had 5628 votes. Some of these votes would have probably gone to Mirkarimi, thus raising his vote percentage. It is highly unlikely, unless there were special conditions, that this would have given Mirkarimi a majority; in order to do so, Markarimi would have had to have obtained 77% of the Feldstein voter second and third choices.

District 7: 13 candidates. First round plurality: Sean R. Elsbernd, 10505/31639 plus 3266 ballots with no vote for an eligible candidate. runner up: Christine Linnenbach, 6784 After transfers: Sean R. Elsbernd, 13834/31639 = 43.72%%. runner up: Christine Linnenbach, 10491.

District 9: 6 candidates, won with 50.73% first round. District 11: 8 candidates First round plurality: Gerardo Sandoval, 7477/23176 plus 1726 ballots with no vote for an eligible candidate. runner up: Myrna Viray Lim, 4280 After transfers: Gerardo Sandoval, 10769/23176, 46.47% runner up: Myrna Viray Lim, 7628

November 8, 2005

Assessor-Recorder: 4 candidates First round plurality: Phil Ting, 94062/199224 plus 26146 ballots with no vote for an eligible candidate. runner up: Gerardo Sandoval, 71850 After transfers: Phil Ting, 110053/199224, 55.24% MAJORITY runner up: Gerardo Sandoval, 79261

Treasurer: 4 candidates, won with 61.36% first round.

City Attorney: uncontested.

San Francisco election for Board of Supervisor, November 7, 2006.

District 2: 2 candidates, 80.13% first round.

District 4: 6 candidates. First round plurality: Ed Jew, 5184/19814 votes. plus 2171 ballots with no vote for an eligible candidate. runner-up: Ron Dudum, 5134 votes. After transfers: Ed Jew, 8388/19814 votes, 42.33%. runner-up: Ron Dudum, 5072 votes.

District 6: 8 candidates. First round plurality: Chris Daly, 8746/17941 votes. plus 1974 ballots with no vote for an eligible candidate. runner-up: Rob Black, 7115 votes. After transfers: Chris Daly, 8968/17941 votes, 49.99%. runner-up: Rob Black, 7303 votes.


 * The counting terminated with four candidates, Chris Daly having obtained a majority of unexhausted votes. It is a practical certainty that he would have obtained a majority from the remaining candidates. Davy Jones and Matt Drake, with 1375 votes between them, would have been eliminated. To gain a majority, Daly needed an additional three votes.

District 8: 3 candidates, 66.24% first round.

District 10: 7 candidates, 56.17% first round.

Assessor/Recorder: uncontested.

Public Defender: uncontested.

November 6, 2007. No RCV analysis necessary for any election.

Mayor: 12 candidates. Majority winner first round (73.66%)

City Attorney: uncontested.

Sheriff: 2 candidates, majority winner first round (73.69%)


 * The data above was revised after what is below was written. Note added by --Abd (talk) 05:00, 25 January 2008 (UTC)

It's true that runoff elections are being avoided in San Francisco, but the basic cause of the prior massive need for runoffs was nonpartisan elections with top-two runoff, thus encouraging large numbers of candidates. If the elections had all been plurality, the winners would have remained the same, which is remarkable, I did not expect to see this to be the case so strongly, for IRV, like top-two runoff, will encourage more candidates to run and will cause the first round vote to be split far more than would happen with actual plurality elections. If preference voting is going to be used, it would seem that limiting it to three ranks when there are hordes of candidates creates a problem; on the other hand, it's not clear how many voters would use additional ranks. IRV in Australia is being used for partisan elections. One of the selling points of RCV was to find a majority without holding runoffs; that is not happening, *at all*. It could be claimed that actual runoffs create a false majority just as much as do RCV elections (these elections are reported with majority victories based on exclusion of exhausted ballots); however, with actual runoffs the voters have a chance to make a choice between the top two; whereas with these elections many voters were excluded from making that choice, and we don't know how many of them were excluded because they did not use all the ranks (their fault?) or were excluded because they *did* use the ranks, but all three were used up.

One thing I've learned from this analysis is a bit of an appreciation of why Plurality voting has lasted so long in spite of its serious theoretical shortcomings that sometimes actually bite. It usually works.

Calling RCV "IRV," as well, is problematic. Limiting the ranks causes loss of voting power. It's conceivable that, if full ranking had been allowed, one or more of the results would have changed. In that 22-candidate election, almost half the ballots were exhausted. The article currently uses the 2000 election situation to show how many runoffs were previously needed; that may have been an unusual year. In this series of 20 elections, only seven would have needed runoffs. If the law required a majority (as in the Vermont governor race, from the Vermont constitution), RCV would not have saved even one runoff (or other process as in Vermont, where the legislature votes by secret ballot if there is majority failure in the public election) . --Abd (talk) 05:02, 14 January 2008 (UTC)


 * When the data was corrected, there was one election where RCV did reach a majority. All others that needed RCV would have gone to runoff if the process had been as described in the voter information pamphlet summary. --Abd (talk) 05:00, 25 January 2008 (UTC)


 * What improvements to the article are being contemplated in the discussion of this section? MilesAgain (talk) 15:36, 16 January 2008 (UTC)


 * Wow! Asking the same question *three* times in succession on the same Talk page! If someone wants to see my answer, look at Special:Contributions/Abd, the last two edits before this one. So I guess I'll ask a question myself. MilesAgain, what's your user name?


 * In this case, there has already been editing of the article from the information collected here. The article, of course, summarizes the information concisely, but to make it easy for other editors to check it, the detailed compilation was done here. I've now concluded that I will write an article for publication based on what I'm finding, unless someone comes up with prior publication of similar facts. In that article, of course, I'd come to conclusions, since I think they are stunningly obvious, but I can't put those conclusions in the article at this point -- unless the work has been done before. --Abd (talk) 19:27, 16 January 2008 (UTC)

Cary, NC election result.
I've finally been able to find actual IRV results from Cary, it was hard for a while. This election was nonpartisan, and the real contest was between two challengers to the incumbent, Roseland, who did not make the "runoff." (I've read articles about this election in NC newspapers that don't mention the IRV but which do mention the runoff.) The Cary official results page shows Frantz, 1151, Maxwell, 1075, and Roseland 793, with 3 write-in votes, total votes, 3022. Because Frantz did not receive a majority, the IRV method was used to transfer ranked votes. This is a Ranked Choice Voting scheme, it appears, like San Francisco (which, in this case, should be adequate, even if a voter casts a write-in vote.) The official results page, which only shows first round votes is at. Then there is the IRV page, which shows the results of vote transfers, it's an Excel spreadsheet at.

There is a 1 vote discrepancy, the spreadsheet only shows 1150 first rank votes for Frantz. With Roseland eliminated, and the three write-in votes as well, Frantz received 248 second choice votes and 3 third choice votes, total 1401. Maxwell received 274 second choice votes and 4 third choice votes, total 1353. I've seen this with the San Francisco votes: the IRV distribution tends to not favor one candidate over another, which is why plurality winners are going on to win the election. This was a fairly close election, though. Now, Maxwell won with only 46.36% of the votes; it looks like a good number of the 793 Roseland voters did not rank Frantz or Maxwell; at most 66% did.

The third rank votes are a bit mysterious. There are a total of 7 that were assigned to Maxwell or Frantz, there may have been more assigned to Roseland. There were only 3 write-in votes in first rank (which is what the official results page shows). One possible explanation is that there were 4 write-in candidates in second rank; then the voters voted for Maxwell or Frantz in third rank.

This is another contest which is just the same as all the San Francisco contests: uncontested, won in the first round with a majority, or, as in this case, the plurality winner went on to win, but not with a majority vote (except one obtained by disregarding exhausted ballots, and, note, take any election, discard ballots for all but two candidates, it will either be a tie or one will win with a "majority." That is not what Robert's Rules considers a majority, which would always be a majority of ballots containing a valid vote. Whether or not the 3 or 7 or ? write-ins were valid is not important here, the ballots with Roseland votes were valid.

So, sure, as FairVote trumpets for Cary, an "expensive runoff" was avoided, but at what cost? Not only was counting this election problematic (it was counted by hand, and "mistakes were made"), but the reason for the runoff in the first place would be to ensure an actual choice by a majority of voters voting. That happens in a runoff, though the set of voters may be different. (Whether or not a real runoff is superior to an "instant" one is debatable, to be sure; I'll merely note that real runoffs are a kind of ad hoc Range Voting. If the election is important to you, you vote, if not, you don't. But from an expense point of view, Plurality is usually producing the same result, victories due to vote-splitting don't seem to be that common, in spite of the fact that IRV -- as with any reformed voting method on the table -- would encourage more candidates and thus, theoretically, more apparent vote-splitting in the first round. My guess is that Approval voting-- just Count All the Votes, i.e., allow voters to vote for more than one, which is really just another form of alternative vote, or Bucklin voting, with plurality sufficient for election (as is really the case with IRV as all these examples show), would accomplish the same salutary effect as IRV -- runoff avoidance -- at much lower cost; with both methods, just count the votes and add them up. --Abd (talk) 03:17, 15 January 2008 (UTC)

By the way, turnout was quite low in the Wake County election. 56,032 ballots were cast with the number of registered voters being 512,515. For Cary, the election was "top two" runoff. To be elected in this way was the office of Mayor, plus three council seats. There were only two candidates on the ballot for Mayor, it was won with 58.31% of the ballots cast. Council District 1 had four candidates, it was won in the first round with 69.10% of the votes cast, District B had 3 candidates, and is the one described above, and District D was won in the first round with 55.00% of the votes cast.

To find these figures, I looked at the "Canvass" file at

Now, just to show, once again, how effective FairVote propaganda has been, I wrote above that a runoff was avoided. FairVote says "expensive runoff," and this claim has been repeated all over the internet. Wait a minute, I just realized. This election was a *primary*, as with San Francisco, before IRV was implemented, the primary was held, then the top-two faced off, unless someone got a majority. Unlike San Francisco, however, this "primary" was not held at the same time as the regular November general elections, which explains the poor turnout. It was held on October 9, 2007. If Cary were to hold their "primary" two months before the general election in even-numbered years, they would have almost no expense for runoffs.

What was the turnout for this particular election? To find out, I have to identify the precincts in the vote totals and then determine the number of registered voters in those precincts. (The same data is available that I used for vote totals, it is part of the Canvass file.). I just looked at the Mayoral election, so far: Total turnout across all Cary precincts was 20.11%. The overall turnout in Wake County for the November 2006 general election was 40.53%. I haven't looked at the Cary specific turnout figures, at least not yet. It's late....

So Cary runs their municipal elections in odd years. Let me say I'm not surprised to find this. General elections have much higher turnout, typically; the low turnout normally suits the status quo, it keeps the riff-raff out. It can backfire, though, if some constituency that does not normally vote in off-years comes out in droves.... This is why top-two runoff can be interesting, for it will sometimes motivate voters who'd normally skip an odd election. As I wrote above, it's a kind of ad-hoc Range voting, weighting votes according to how much people care about the result. --Abd (talk) 05:22, 15 January 2008 (UTC)


 * Cary also had a contest for "At Large" City Council member. Two at large candidates with drew from the contest because they were afraid they would dillute the vote and throw the contest to the opposition.  One of the candidates who with drew still had her name on the ballot and got a significant number of votes.
 * The raw vote data showing how voters ranked their choices has not been provided. NC has no intention to count these votes nor to account for these votes since there was no need for a "runoff".


 * In the District B, Nels Roseland told his supporters to rank him all three times, and he lost.
 * Don Frantz won in the "runoff". Maybe in part because he spent alot of time correctly explaining IRV to his supporters. --Ask10questions (talk) 08:41, 15 January 2008 (UTC)

Astonishing IRV statistics
Since FairVote began their campaign in the 1990s, and particularly in recent years, quite a few jurisdictions have passed IRV legislation. However, as the article shows, by far, most of them have not implemented it. So far, there have been, as I count it, 25 IRV elections in three cities in the U.S. in recent years (i.e., after Ann Arbor IRV went down in flames). Every single one of these elections was won by the plurality winner from the first round, IRV vote transfers made no difference, except in avoiding mandated runoffs for majority failure, yet, ironically, out of the nine elections that went to runoff, none of them reached a true majority vote as a result of the vote transfers. Where runoffs were avoided, it could be argued in a few cases that the public was deprived of the right to make an informed choice. This is particularly poignant in the Cary, North Carolina, election, which was very close; the incumbent did poorly, and the two "insurgent" candidates split the vote fairly closely. It looks like about one-third of the incumbent's supporters voted second rank for one of the insurgents and one third for the other, and one-third did not add a second rank vote (which could mean that they cared not at all for either of the insurgents; but with an actual runoff, they might have decided there was a difference). Turnout was poor (20% of registered voters), because Cary holds their municipal elections at an odd time, in October of odd numbered years, not in connection with the General election, which seems to have double the turnout (i.e., 40% of registered voters). This was an election crying for an actual runoff, there is no reason to expect that turnout would have been worse; it would only have been held in the precincts for that Council seat. --Abd (talk) 05:50, 15 January 2008 (UTC)


 * Where runoffs were avoided, it could be argued in a few cases that the public was deprived of the right to make an informed choice. This is particularly poignant in the Cary, North Carolina, election, which was very close; the incumbent did poorly, and the two "insurgent" candidates split the vote fairly closely. It looks like about one-third of the incumbent's supporters voted second rank for one of the insurgents and one third for the other, and one-third did not add a second rank vote (which could mean that they cared not at all for either of the insurgents; but with an actual runoff, they might have decided there was a difference). Turnout was poor (20% of registered voters), because Cary holds their municipal elections at an odd time, in October of odd numbered years, not in connection with the General election, which seems to have double the turnout (i.e., 40% of registered voters). This was an election crying for an actual runoff, there is no reason to expect that turnout would have been worse; it would only have been held in the precincts for that Council seat.


 * Why is this different than the normal case for runoffs? *The runoff was, by design, held as part of the general election.* Now, it is not a *major* general election, but actual statistics for runoffs from Cary showed no significant decline in turnout for the runoffs they had. (It was slight decline in two and slight increase in one). Turnout, *period*, was poor, probably because these are off-year elections. I'll stand with what I wrote, it's supported by the evidence I gave, which, it appears, MilesAgain is so concerned with trying to stop it from being presented in talk that he hasn't actually read and considered it. The comment about "only ... in the precincts" was related to expense, not to turnout, per se. Let me repeat this: the common wisdom that runoffs have lower turnout did not apply to Cary, and it's obvious why.
 * --Abd (talk) 04:38, 17 January 2008 (UTC)

about election detail in the United States section
I have been collecting information about actual use of IRV in the U.S., placing what I find, sourced, here in Talk. Then, in the article, I've continued what I did previously, and which was allowed to remain: I described briefly each example of adoption. There are a number of categories in the political adoptions:

(1) An adoption formally took place, but implementation was postponed, typically indefinitely. It's questionable whether or not these should even be mentioned, but FairVote definitely likes to promote the idea of a wave of successes, I'm sure it helps their bottom line and maintains the impression of "momentum" (which is the argument that they make when discussing possible cooperation with supporters of other election reforms: those other reforms are "impractical" -- even though one of them is zero-cost and widely respected among experts -- and IRV "has momentum.")

(2) It's actually been used for one or more elections, but the elections were with only one or two candidates on the ballot, so the IRV ranking was almost useless. (A write-in voter, however, could find it useful).

(3) There were more than two candidates, but one obtained a majority in the first round and so the runoff provisions were not exercised. Note that the existence of a top-two runoff or IRV makes it more likely, almost certainly, for additional candidates to run, because they can do so without spoiling an election (as long as they don't get upwards of one-third of the vote, where center-squeeze becomes a problem); so it becomes harder for a candidate to gain a majority in the first round, or at all.

(4) There were more than two candidates, no candidate obtained a majority, so the runoff provisions were used, but did not change the result (plurality winner) from the first round, however, the candidate did gain a majority of votes through the runoff process. *This is an empty category.*

(5) There were more than two candidates, runoff provisions were exercised, but the leader did not change from the plurality winner in the first round, and no majority was obtained. All IRV elections I found fall in this category, so far, and I was quite surprised at this, it is not at all what I expected. As far as I know, indeed, this has largely escaped notice, but it isn't "original research," because it can be directly verified by any reader, and no interpretation is necessary.

(6) The election went to runoff, and the IRV winner was not the winner of the first round, and a majority was obtained for the IRV winner. This is an empty category.

(7) The election went to runoff, the IRV winner was not the first round plurality winner, and no majority was obtained. This is an empty category.

Then, in the data on the Talk page, above, there is another interesting fact which I did not put into the article. For the San Francisco elections -- I did not check out the Burlington and Cary runoffs -- the *second place* candidate did not change from the Plurality results in the first round.

Take the 2005 citywide race for city-assessor. Phil Ting won 47% of the first choice vote. He then won substantial number of second choices from another Asian candidate and won with an absolute majority of the first-round vote. But Abd has him as not winning a majority of the vote, because he counts in people who didn't vote in the race at all. See:

http://www.sfgov.org/site/elections_index.asp?id=61607

Similiarly, in 2006, Chris Daly won in the 5th round of counting with a majority of continuing votes. They stopped the count at that point with him something like 4 votes away from a majority of the valid first-count vote, so one could argue he technically didn't win a majority of all votes cast in the race, but they stopped the count with his three strongest opponents still in the race, because he had won a majority of continuing ballots (discounting exhausted) -- meaning he obviously would have won a majority of the first-round vote if they had reduced the field to two.

It is true that the hotly contested race in District 4 was won by a majority in the last round, but that majority was less than 50% of the valid first choice vote. That also has happened in a couple other Supervisor races, but people have accepted the results just as much as they would have if someone ran a runoff with the voter turnout a little lower in the runoff. It is never proper to describe the percentage in the runoff by using the denominator from the first round of voting.

Abd really messes up the race he profiles in the write-up - -the one with 22 candidates. He suggests the winner had only 37% of votes, but he skips over the MAJOR fact that the count ended with three candidates still in the race -- the winner had a majority of the final round vote with two other candidates still in the field so they didn't need to eliminate the third-place finisher, because no other candidate had any mathematical chance to win. So Abd is really comparing apples and oranges. True, Mirkarimi would not have won 50% of the first-choice vote if they had gone to the final two, but he almost certainly would have been a good bit over 40% -- and his point just is highly misleading and does not belong in a NPOV article. Tbouricius (talk) 21:20, 26 January 2008 (UTC)


 * First of all, I did, in fact, overlook the "abstentions", drop-outs from particular races However, before Mr. Bouricius wrote the above, I'd corrected it. Perhaps the edits crossed. There is one error that was in the text of the article: there was, in fact, one election which found a majority, the election of the Assessor. I'd intended to correct that in the article. What Bouricius wrote above *was* correct, initially. Note that nobody caught this for quite a while. As soon as I saw the objection, I investigated. I had misunderstood the way that the city was presenting the data. There are, in fact, an astonishing number of people who, with some of the races, "skipped" it. I have not yet compared this with the prior situation, but more than 10% abstentions.... that's a lot, though maybe Bouricius is right. After all, he's been an elected politician. After correcting for this, all the elections *still* failed to find a majority, and, now, what is being considered is the standard of "majority of all ballots containing a valid vote for any eligible candidate." Which Bouricius surely knows is the Robert's Rules standard. If there remain errors, by all means, identify them. The information will be going back in the article, see the next section in Talk.


 * (One of the oddities of San Francisco is that true write-ins are not allowed, unlike most elections I've seen. My guess is that some of those "dropouts" are write-in votes in some place or other. In Cary, the write-ins are all tabulated.


 * The candidate in the 22-candidate race won with 37% of the vote, as far as we know. We might speculate that if counting had continued, he'd have gotten more. Yes, Bouricius has a technical point. But he's totally overlooking the core issue: San Francisco voters were explicitly promised an election that would still require candidates to gain a majority. I quoted the proposition analysis that voters saw in 2002, above. They got something else. Bait and switch. So the winner might have gotten over 40%? In that case, one of the other elections could be the one with the worst results! However, I will re-examine the results in the light of this comment by Bouricius. Note that what would be involved would be vote transfers among frontrunners. There is, however, another analysis which can be done with the vote data, which might be even more interesting: Bucklin voting analysis can be done; the data is provided to do it. It's just a matter of adding up the votes.


 * My *point* isn't relevant. The facts are. If there is an error in the facts, it is totally legitimate to correct them. Not to exclude them because they supposedly support some POV point. That's up to the reader, and excluding this is POV.


 * I've fixed the data above. There is, as Bouricius noted, another majority winner. Chris Daly. He would have had to get three more votes and he had a pool of about 1300 voters to get them from, as second or third-choice votes from the last two eliminated candidates. I consider this election to have been won by a majority, though it remains quite significant that the RCV method doesn't actually determine the fact, we don't know how great the majority is. As Bouricius notes above, Mirkarimi almost certainly did not gain a majority. (He would have had to get 77% of the Feldstein votes. Of the total Feldstein votes, 42% were already second or third rank choices from other candidates.) Bouricius was incorrect about my report on the Assessor-Recorder election, that is, I had already corrected this. (The article hadn't reflected it. The article was technically correct in reporting not a majority of ballots, but this was misleading and would have been corrected. And, indeed, could have been corrected by anyone.) So, we do have two elections where the winner obtained a majority through IRV. Apparently, it is actually possible.... --Abd (talk) 05:45, 27 January 2008 (UTC)

significance of election results presented, cherry-picking of data?
Abd. You might want to redefine the academic consensus that runoffs and instant runoff voting are majority voting systems, but you know you are doing so outside that academic consensus. Runoff elections also can have significant changes in turnout -- when one of two rounds can be decisive, significant differences in turnout obviously can raise questions about whether it was a "majority result", but the academic consensus is that it is. Choosing to make your case within the Wikipedia article that it's not is a highly political decisions, reflecting your own point of view.

In San Francisco, ithin the city, there was absolutely no controversy about the result of the Mirkarimi win in 2004. People were very used to candidates winning majorities in runoffs with a drop in turnout, and that's how this was seen --- with the added fact that the turnout in a real December runoff would have been much, much larger. Mirkarimi had a maority of continuing votes, people knew those were the rules, and he was generally seen as winning a majority in the instant runoff because he was the majority choice of the top candidates. You obviously can dispute that based on your own numbers, but it's one definition of majority versus another.

Furthermore, your selective highlighting of the 37% number in a short summary of data is highly misleading when there were three candidates left. Did you not know that when you did this or chose to overlook it?

As a final point, every winner in the thousands of election in Australia's house of representatives has won with an absolute majority of the valid first-count ballot. That's a product of their rules, which invalidate ballots that don't rank everyone. So if one wants absolute majorities of first-round ballots, it can be done through such a requirement. But if voters don't want to rank everyone, allowing them to truncate their ballot is reasonable policy decision. It also has been seen as a reasonable decision to give voters three times the chance to cast an effective vote with IRV as they would have with a plurality ballot, although most IRV backers would prefer a ballot with unlimited rankings (as indeed will be what San Franciso uses when it has voting equipment that allows it). —Preceding unsigned comment added by RRichie (talk • contribs) 14:35, 27 January 2008 (UTC)


 * A number of issues were raised by Mr. Richie. By the way, welcome!


 * I'm not sure what User:RRichie means by a consensus that IRV is a "majority voting system." IRV satisfies every variation on the Majority criterion, which has nothing to do with what I've found and reported with the elections. These were elections where, apparently, there was no majority choice among the candidates. Had there been a real runoff, the majority would have made a choice, at least among the top two. What I've found is that in similar elections prior to the series reported here, with top-two runoff, the election shifted from the plurality winner in the first election, to the original runner-up in the runoff, about one-third of the time. It's a complex issue and I can't put this into the article at this point, because it is a conclusion, but it appears that IRV is frustrating the purpose of having top-two runoff. I have not done enough analysis to know why, but it is pretty clear. It will be clearer when I look at the runoff series in certain North Carolina towns that, like Cary, had primary elections in October and then the runoff in November. I noticed that this same schedule was proposed for San Francisco, by the way, as allegedly superior to switching to IRV. Might have been.


 * Richie appears to be making a common assumption, that differences between results in primaries and actual runoffs are due to differential turnout. Cary and the other towns in North Carolina are particularly interesting because of their schedule: they have similar turnout for both elections. (In Cary, sometimes the runoff had higher turnout, sometimes the primary, but they were close to each other.) When we look at San Francisco in 2000, we see runoff results switching from the plurality winner to the runner-up. With IRV, it isn't happening: in every election, the plurality winner went on to become the IRV winner after vote transfers. Just as interesting interesting, the runner-up in each election remained the runner-up. In this substantial series of elections, it seems we are seeing sincere plurality producing the same result as IRV. Top-two runoff, like IRV, tends to encourage sincere voting, that is a benefit of both of these methods, to be sure. (That strategic voting can exist with IRV and top-two does not negate that sincere voting becomes more likely than with plurality.) (And, of course, if these elections were not IRV, or top-two runoff, voters would probably do what voters usually do: choose the favorite of the frontrunners. *Usually* the result will be the same. This is, indeed, why Plurality, as defective as it is, has lasted so long. It *usually* works.


 * The implication was, when IRV was passed in San Francisco, that (1) the results would be similar to top-two runoff, but at lower cost, but (2) results might change if the effect of differential turnout were eliminated. So is what is happening in San Francisco due to voters not having to come to the polls again? I.e., are the IRV results, which are confirming the plurality winner, more fair than the reversals that were reasonably common before? This is why I want to study the other North Carolina elections, because from a larger sample, I may be able to get a better understanding of how often top-two runoff reversed the result in the absence of differential turnout.


 * Now, the legitimacy of the Mirkarimi win isn't the point. I did not dispute it, none of the study above is judgemental in that way. (I do make a few comments about the implications as I see them, but these are dicta. However, we don't know if some other candidate might have beaten Mirkarimi in a direct face-off. There were *many* exhausted ballots. The point is that the electorate in San Francisco was actually promised, in 2002, that candidates would "still be required to attain a majority." Now, since IRV essentially *guarantees* a majority result (in the final round), there is no attainment, there is actually no requirement at all. If I were of a cynical mind, I'd ask how much was paid for that ballot summary. Sure, any voter could have read the proposition itself, but, as you know, most don't. In any case, the San Francisco results will be of great interest, I'm sure, to other cities considering Instant runoff voting or alternate reforms.


 * Mr. Richie asked a question. "Furthermore, your selective highlighting of the 37% number in a short summary of data is highly misleading when there were three candidates left. Did you not know that when you did this or chose to overlook it?"


 * No, I did not know that. I overlooked that there were three candidates left. I just looked at the victory count for Mirkarimi and divided it by the total votes. As was noted above, I initially used the wrong total for "total votes," quite simply forgetting that there would be dropouts, i.e., complete abstentions from the race. I'm astonished by the number of these: people who live in the districts and who either did not vote on the race at all, or who *voted for a write-in candidate who was not eligible," which requires prior process, you cannot just write someone in who would be eligible in general, they must have filed a candidacy. I wonder how many SF voters don't know that? Or whose ballots were otherwise voided for this election, perhaps they overvoted. Over 10%. At some point, I'll compare that with other elections.


 * As to highlighting this number, the question I had in mind was "how far was IRV coming from finding a majority winner?" So I reported the race with the lowest percentage, and I think the comment reflected that. It's obvious that, then, the results ranged from that minimum, up to races with a majority in the first round (and one got 80%). Going over the results, I found, in fact, two elections where IRV *did* find a majority winner. Thus the original report, that there were none, wasn't POV, it was simply wrong.


 * What the article had, didn't have for a while, and now has again, until it's changed, was a list of "adoptions," which serve the purpose of promoting IRV by making it appear that there is this huge momentum. However in fact, there have been very few jurisdictions that have actually held IRV elections, so far. And, only now, are the results starting to show how IRV actually performs. I'm sure we are all eagerly awaiting this knowledge.


 * I can't report any analysis of the results and put it into the article, that would be Original Research. But I can report numbers from reliable sources and put them into the article, and as to the charge that this introduces bias, the typical remedy is not to eliminate such reports, but to balance them. (If they aren't notable, as with fringe science, sometimes sourced fact can simply be excluded, particularly if the "fact" is that so-and-so, a fringe theorist, holds such and such an opinion.) An example from the debates that have raged here: does Robert's Rules "recommend" IRV? Well, one *could* very briefly summarize it in that way, though it's inaccurate, it doesn't actually make a recommendation, it describes some options. In any case, IRV advocates here have attempted to insist on having notice *in the introduction*, which is a prominent position in the article, that Robert's Rules mentions IRV, and says something that could be read as a compliment of IRV or recommendation *if* read incautiously and divorced from the context. But the same section in Robert's Rules goes on to note a major defect of IRV: its capacity to pass over a compromise candidate, and instead elect the most popular of two candidates with fanatic support. (my language, of course). Robert's Rules *actually* suggests "preferential voting," notes that there are many forms, and only describes one, which it then specifically criticizes. There are other forms of Preferential Voting which don't have that particular problem. Bucklin, for example. The point is that mentioning the "recommendation" in the introduction, *or* mentioning the criticism, either of them alone, would be POV imbalance, taking a source and distorting it by selective quotation.


 * Now, if my reporting from the elections is imbalanced, I can assure you that it was not from cherry-picking. I did choose what to study (performance with regard to attaining a majority) based on what seemed interesting to me, and, of course, that choice may be related to my POV. But I did not conceal any results, I reported everything I found, here, and, of course, gave the sources, inviting review. The "cherry-picking" charge has been made ever since I started working on this part of the article, yet, mysteriously, no balancing facts have appeared, which is what would be the proper response if, indeed, I had cherry-picked. Editors with a POV always complain that an editor perceived as being on the other side has "cherry-picked" facts, and that putting in detail would make the article "confusing." That might be correct. Don't confuse me with facts. If the presented facts can be pared down to just those necessary to arrive at a clear conclusion, why, miracle of miracles, we have a very clear article.


 * And a very misleading one, designed through that process to manipulate opinion. Some people are quite good at that.


 * As to points of agreement, certainly having increased ranking would be an improvement; three ranks with 22 candidates is awfully limited. In elections with only a few candidates, five or six or so, it should be plenty. However, I'd suggest Bucklin, and, given that the SF data lends itself to Bucklin analysis, it's going to make some interesting spreadsheets. With Bucklin, as in Duluth, three ranks is *plenty*. Bucklin was working there, you know, and would have continued to do so if not for court intervention on spurious grounds, not supported elsewhere. None of this nonsense about "people not using the lower ranks"! They used them. Again, this is not for this article, this is just background. At least I don't see how to put it in, or how that would be appropriate. It might be appropriate in the Instant-runoff voting controversies article. --Abd (talk) 23:15, 27 January 2008 (UTC)

United States results details
This level of detail is senseless to me. Why is this needed here?! I just moved content to a subarticle! Tom Ruen (talk) 04:48, 28 January 2008 (UTC)


 * Well, do read what I wrote above. Why is this needed here? Has Tom been paying any attention? It's been explained quite a few times over the last few weeks! Moving content needed for balance to subarticles is POV forking, not allowed. However, the detail doesn't have to stay if the *summary* can stay, which is just something like three sentences. What Ruen did was to move *part* of the detail to the subarticle, the part that might be used to criticize IRV (though it is not itself critical, it's just a report), leaving what makes it look like IRV is fantastically successful. So far, it is probably producing exactly the same results as Plurality, at higher expense. And there is a lurking suspicion that it is producing worse results than top-two runoff, though it *might* be cheaper. And it might not.


 * I'd expect more reversals in the runoffs than are occurring, based on a preliminary study of runoff elections, part of which was reported here (the SF 2000 results). Again, that's a complex question, and moot at the moment. I can put the raw data in the article, but not a conclusion like "IRV is frustrating the point of requiring majorities." Not yet.


 * One problem that Tom should be aware of. He did not simply move content to the new article. He *deleted* content that was sourced (that is, he took it out of this article and did not put it in the new one). Perhaps that was inadvertent. —Preceding unsigned comment added by Abd (talk • contribs) 05:03, 28 January 2008 (UTC)