Yesterday’s mayoral primary in New York City was run by ranked choice voting (as I suspect anyone who comes to this site has heard). Analysis of preference flows, when they are available, is going to be interesting. Actually, because it was a highly competitive primary for determining the Democratic nominee, there may not be much discernible pattern. That is, unlike a general election with multiple parties (as in Australian House elations, for example), voters may have had little information or understanding of how to use their ranked choices effectively. This will have been complicated further by what I understand was a relative paucity of explicit preference-exchange deals among candidates (e.g., “vote for me 1st, and then give your second choice to X”). In a primary for a single office, there may not be much incentive for candidates to do such deals. There is little to trade–or at least limited credibility to such trades–unlike in a general election, particularly a partisan one that spans across multiple districts. (There were, of course, also primaries for City Council seats; it is not clear to me how preference trades might work between a mayoral primary and council primaries. Again, the lack of party alignments of the candidates–or, rather, all being of the same party–probably greatly limits effectiveness of any such deals.)

From what we know so far, on first-choice votes, the leading candidate has under a third of the vote, and the next two are in the 20–22% range. That means a healthy lead for the one in the initial first place, Eric Adams, but also a big shortfall from majority.

I have not followed the campaign closely enough to have anything to say about how second preference might break. But I am sure some readers have, so please enlighten us!

101 thoughts on “NYCRCV

  1. Journalists describing various forms of PR as unbearably complex is bad enough, but journalists describing ranked choice voting as unbearably complex should be taken out and shot.
    Sri Lanka elects its president by supplemental vote which is a form of ranked choice voting. The population pf Sri Lanka is a tad larger than New York’s.

    • (Alan) “When [ACT e-voting software] prompts for a second preference you get a dropdown menu that excludes anyone you’ve already preferenced so Shugart does not appear.”
      Now, one could ice the cake by combining this with Robson rotation so that the algorithm keeps track of votes cast in real time and presents you with the candidates (within each party group) ordered from fewest to most votes in current totals. Groups themselves could then be rotated regularly as they appear to voters. This would help to reduce the occasional anomalies that Dr Kevin Bonham refers to as “the Ginninderra Effect” after the northern Canberra electoral district of that name (soft G) and where, coincidentally, I once used to live but never got a chance to vote in.
      It’s basically a vote-splitting problem akin to SNTV that can occur where a party has two candidates still in play for two still-vacant seats, where they are first and third in ordinal vote-rank but would have been first and second had the team votes been divided more evenly.

      • Need to clarify that having Shugart not appearing on one’s ballot menu is a defect in any electoral system ever devised.

  2. “[B]ecause it was a highly competitive primary for determining the Democratic nominee, there may not be much discernible pattern.” My impression is that the candidates were widely viewed as falling into one of two camps — moderate and progressive. So I expect that some patterns will, in fact, emerge from the transfers.

    “… voters may have had little information or understanding of how to use their ranked choices effectively.” This wording suggests that strategic voting can be effective in RCV elections. To the best of my knowledge, there are two situations where this is possible. One is when the voting equipment doesn’t allow voters to rank enough candidates to insure that their ballots are not exhausted involuntary. Here voters need to be sure to rank the least objectionable of the front runners last. This has been a problem at times in cities where only three rankings are possible, but I suspect that five will prove to be enough in practice even though any limitation less than the number of candidates isn’t pretty conceptually.

    A second situation is when your favorite can’t win (and you know that in advance) and ranking her first contributes to the early elimination of an acceptable candidate who otherwise might have won — in other words, risks contributing to the eventual election of your least favorite. This is usually called the “center squeeze”. I believe it’s one species of non-monotonicity. With a field of 13 candidates and ambiguous survey data, I think it’s very unlikely that even very sophisticated voters would be able to manipulate the election in this way and that it would be a mistake to try. It’s hard enough to pull off in a race with three candidates and a lot of reliable survey data.

    In summary, I don’t think strategic voting can play much of a role. But when RCV is new and voters are unfamiliar with it, confusion — aided and abetted by deliberate misinformation spread by opponents of RCV itself — can play a role. There are two issues here. Some voters seem to feel they don’t know enough to have opinions about all of the candidates and/or differentiate between candidates they like a lot and others they like only a little. And some voters have trouble understanding when to stop ranking more candidates. There’s a clear answer to that question (clear to me, anyway): you don’t rank a candidate when you would prefer that your ballot not count at all rather than count for that candidate. But it appears that many voters are not used to thinking this way.

    • Voters were only allowed to cast 5 preferences due to the ballot style chosen. Several newspapers had editorials on how to cast your preferences based on ideal outcome (such as ‘elect a progressive,” “stop Yang” etc).

      I think most voters will have expressed more than one preference. The question is whether those preferences end up being effective or for minor candidates that were already excluded.

      I don’t think in the low 20s on the first count that there are enough left-leaning votes out there for Wiley to win. She might hold on to the final count but if that happens I don’t think she’d get the 75%+ of the preferences from Garcia that she’d need to overtake Adams.

      However, I do think Garcia has a solid shot of winning from third. She and Yang had a formal deal for second preferences (though I doubt they handed out how to vote cards, and they announced it after early voting had already begun), and I expect she’ll get enough second preferences to overtake Wiley. At that point it’s just a question of whether enough Wiley voters put Garcia on their ballot for them to stop Adams. I think most of them would definitely prefer Garcia, but I don’t know if the limitations of 5 preferences (which is a terrible system when there are so many candidates) will prevent preferences from flowing to her.

      If I were a betting man and betting on elections were a thing here, I’d give Garcia about a 60% chance of winning, Adams a 35% chance, and Wiley a 5% chance.

    • And I also think that voters did have enough info to group candidates into informal “parties” of ideological affinity. Wiley, Morales, and Stringer were all clearly left-of-center by Democratic standards; Adams ran on a strongly pro-police platform and McGuire as a pro-business, Wall Street-aligned candidate, and both are on the right of the party (though whether McGuire votes and Adams votes flow to one another is up in the air, they almost certainly wouldn’t flow to the three on the left); Garcia, Yang, and Donovan emphasized technocracy and can probably be identified as being at the center of the party. The remainder were essentially fringe candidates and probably can’t be expected to flow in any reasonable pattern (though I would imagine that there will be a significant percentage of Chang voters that will flow to Yang due to strong traditions of co-ethnic voting in Democratic primaries, that is just a hunch, and it won’t be enough to impact the results); even if they do snowball to each other, the 5 preference limit with optional preferences will keep any of them from winning.

      The key question is whether the left bloc or the technocratic bloc keeps their votes most together within the bloc without leakage. Actually, upon reflection, I’d probably give the slight advantage there to the left bloc, and reduce my odds of Garcia winning somewhat above, but I still think it’s unlikely Wiley has a path to victory unless.

      • Christopher, I guess the question is whether these tendencies were as clear to the typical voter as to you. I would not claim to have an assessment of that. I just have the question.

      • Fair point. They were certainly well-covered in the papers, but at least based on circulation numbers, the typical voter probably isn’t reading the newspaper regularly.

    • Bob, many good points, as per usual. My main concern is with voters not giving their full allotted ranking, maybe because they are not certain which candidates are in which camp. This would be mitigated by either (sub-)party labels or organizations issuing “how to vote” cards. I would not know how common the latter was in NYC.

  3. Thank you for highlighting some very different properties of the NYC elections, as compared to Australian (national) elections.

    Maybe one of the Australian readers knows when party labels first appeared on House ballots. I do not recall and do not have handy a copy of Farrell & McAllister (2006, UNSW Press).

      • Amazing it was so recent! Was the number of candidates typically smaller in those pre-84 days? Was it common knowledge who was with which party?

      • There was a certain amount of skullduggery with dummy candidates being run on the assumption that labor voters would find it more difficult to cast a formal vote than Coalition voters. Thus:

        Whitlam had called a double dissolution election for May 1974 and Labor had hoped it could pick up six of the 10 seats in NSW. To thwart Labor’s chances, a core of conservative opponents set out to stack the Senate ballot paper in NSW and boost the informal vote. There were 73 candidates, and under the formality rules of the day, a complete and accurate sequence of preferences was required for a formal vote.

        The stacking of the ballot paper produced an informal rate of 12.3%. Labor polled 5.5 quotas but fell short of electing six Senators. Informal voting research on the smaller 1977 Senate ballot papers indicated it was Labor that was disadvantaged by informal voting, so by inference it can be argued the 1974 Senate ballot paper stack in NSW deprived Labor of a sixth Senate seat, playing a part in the dooming of the second Whitlam government.

      • I am sure how-to-vote cards were innovated some point before 1984 (though if no one has done any academic research on their origins and development it would be a fascinating area for study).

      • Christopher the question was When party labels first appeared on House ballots not when how-to-vote cards appeared.

        The media do not rely just on how to vote cards. You are treating assumptions as facts. A good scrutineer can give a very accurate estimate of preference flows and exhaustion rates, often within 30 minutes of the count beginning. Those estimates get reported to the parties and then are passed to national media panels.

        It is absolutely standard for a panelist to say they’ve been talking to scrutineers in Upper Middle Bogan and they have more recent figures than the electoral commission is showing. Obviously those estimates are not always accurate but in a surprising number of cases they are.

        I don’t have what’s known as ‘the eye’ to do it. A friend roped me in as a scrutineer in Canberra sometime in the early 90s. I’m good at knowing the act and arguing how ballots should be interpreted but I’m hopeless at seeing preference flows. On the other hand, because I was visiting from Sydney where the all-powerful NSW Right faction of the ALP has its lair, everyone in the counting room (including, I suspect, the returning officer) assumed I must be one of the vat creatures of the NSW Right, specially decanted for the occasion in their secret lab in Sussex St. They paid tremendous attention to everything I said.

      • This how-to-vote reference was for how voters figured out how to vote for the candidates of a certain party without that party’s labels appearing on the ballot in response to “was it common knowledge who was with which party”.

        That was unrelated to how the media use HTVs, but I know for a fact that Kevin Bonham uses HTVs and historic patterns of leakage from HTV order in projecting lower house elections and I’m about 90% sure Antony Green does the same. They may not exclusively rely on them, but they are a part of being able to make election projections in the absence of a formal distribution of preferences. The scrutineers may help pundits give opinions, but they are not, as far as I’m aware, a part of the computerized seat projection process that any Australian media outlet uses.

      • I’ve always wondered how ‘Strayans (better?) feel about the fact that the definitive study on their elections was written by two Irishmen (though I would presume McAllister is nationalized by now).

      • Much the way that, I’m guessing, Americans feel about their system being nutshelled by one Englishperson and one human of Gallicity (Bryce and deTocqueville)

  4. One considerable difference of truncated preferences is that you don’t see “put X last” campaigns, as you sometimes see in Queensland and New South Wales. US RCV advocates like to talk up “elimination of negative campaigning” as an advantage of RCV, and I suppose this assists that in a way-the present system favours “cooperation” between candidates since only high preferences can meaningfully be expressed. On the other hand, I think this shows some of the narrowness of this advantage. Voters often actually strongly dislike certain candidates! Truncated RCV often ends up being a bit of an awkward hybrid between FPTP and full RCV: ‘vote-splitting’ is a less obvious concern, so more candidates can jump in, but more candidates do still ultimately create problems for voters.

    • Ah, yes, eliminating negative campaigning, because Australian elections are well-known for the conciliatory tone the Coalition and Labor take toward one another.

      I think the choice of 5 preferences is really interesting. With two preferences like British SV, a voter picks their first preference and then between the front-runners (though I can’t wait to see how the people and media react when someone that’s not Labor or Tory makes the final count in a prominent election). With 3, like in Sri Lankan or Minneapolis, you can pick a first preference, then rank your top 2 of the three most viable candidates. But increasing it to 5 preferences means it’s harder to judge which candidates will be viable in the late counts, so the voters have to do a sort of complex internal calculus with imperfect information between voting their heart and blocking the candidates they hate the most.

      It almost makes me think that Alaska maybe was onto something with their plurality primary + top-4 AV general election (though I dislike that parties aren’t able to identify a single nominee before the general election), although “avoiding the expense of a runoff election” is usually cited as a key advantage of AV by most American reformers and this system undermines that.

      I do hope the full preference data for NYC will be made available; there has been some polling that FairVote released showing the percent of voters ranking a candidate in their top X preferences (only Adams was a top 3 choice for a majority of voters; Garcia and Wiley were top 4 choices for a majority). I will be interested to see when voters consolidate toward front-runners (and when voters who pick front-runner candidates give preferences to other front runners), and how many voters express all 5 preferences and still exhaust their ballots.

      I haven’t seen what the direct mail or push cards at polling sites looked like, but I would be unsurprised if how-to-vote cards haven’t been innovated in NYC or else will be the next time there is a noteworthy ranked-choice election (probably the Congressional primaries next year).

      • Christopher wrote: “… the next time there is a noteworthy ranked-choice election (probably the Congressional primaries next year).” I don’t think the state of New York has adopted RCV for Congressional elections. RCV was adopted by the city of New York for municipal elections.

      • Ach, that’s right, it is only municipal elections (and only the primaries and not the general elections, at that).

        Applying it in general elections either in the city or the state could end up complicated because of the presence of electoral fusion and how parties must earn a specific portion of the vote on their ballot line in order to maintain registration.

      • “With two preferences like British SV, a voter picks their first preference and then between the front-runners (though I can’t wait to see how the people and media react when someone that’s not Labor or Tory makes the final count in a prominent election)”

        You don’t need to wait. While SV works okay for the London mayoral election, it has been used for all direct mayoral and Police and Crime Commissioner elections since those offices were introduced, and has been decidedly more ropey in these. The 2017 West of England mayoral election produced first count results of CON 27% LAB 22% LD 20% IND 15% GRN 11% UKIP 4%. Only 38% of voters for the defeated candidates indicated a valid preference for the second count: it’s not clear how many of those voters were genuinely indifferent, but it is nonetheless a remarkably small number of voters with a preference between the top two.

        I think it’s probably the case that if you must truncate preferences, more preferences are always better at reducing complexity, since a voter who wants to not elect a certain candidate (a perfectly legitimate and reasonable aim) can rank more candidates they feel unsure about.

      • Henry: Agreed that this form of RCV risks being more FPTP-like than reformers would want to admit.

        Christopher: Yes, I have the same thought that Alaska reformers were on to something. The top-x winnowing is not such a bad idea. However–as you also note–it would mean two rounds for the primary, unless (like Alaska) the winnowing round simply replaces the primary. And if you do that, you essentially have a nonpartisan system. Lots of American reformers would be happy with that, but I do not want to encourage them too much.

      • As I understand it the truncated preference systems in the US are justified because hardware and software vendors insist that it’s too difficult to allow a fully preferential vote. That is untrue. The ACT electoral commission gives people the option of voting electronically or on the traditional paper ballot. They developed their system in house and made the operating code publicly available so that it can be verified by the public.

        If you wish to vote electronically in the ACT the system prompts you for a first preference and provides a dropdown menu of all candidates. You vote 1 Shugart. When it prompts for a second preference you get a dropdown menu that excludes anyone you’ve already preferenced so Shugart does not appear. You vote 2 Round and the third preference menu excludes Shugart and Round and so on. The system prevents you voting vote say 5 Mussell followed by 7 Schlechta or voting both 5 Burge and 5 Santucci. Friends tell me that voting electronically is actually faster than voting by paper and you have the assurance that you cannot make common informal mistakes like leaving out a number or using the same number twice.

      • I suspect the real answer is that implementing fully preferential RCV in the way Alan is describing would be more expensive, which is a burden that campaigners for a ballot measure do not really want to bear. In addition, I believe that direct recording electronic machines like those used in the ACT have fallen from favour somewhat in the States due to the somewhat clunkier way in which the paper trail has been implemented (in those states that actually bother with a paper trail).

    • I am hearing from some who worked on campaigns that the “eliminates negative campaigning” message worked exactly as you suggest — possibly to the detriment of left-leaning candidates. I guess we will see, maybe, when research on this election starts to appear.

      • Yes, I suppose that’s how it would have gone. And if RCV were used in, say, the Republican primary in Wyoming’s House of Representatives seat, where anti-Trump Republican Liz Cheney is running against a field of about six or seven pro-Trump Republicans, the system would probably end up disadvantaging the right (depending on how many preferences there were) since pro-Trump voters would still have to engage in guesswork to rank the most popular pro-Trump candidate.

      • Limited-preferential voting can be viewed as something of a bait-and-switch in this way. “Voters can number up to [N] candidates! Having two or more from the same bloc doesn’t necessarily split the vote!” true, But if this then encourages [N + X] candidates from Team Mauve to throw their hats in the ring, you still have a coordination problem. The result may then be the worst of both pure FPTP (the largest bloc loses because its vote was divided over too many candidates) and preferential systems (the winner didn’t have the most first-preferences). Optional-preferential AV can also lead to a “47% after preferences beats 39% on first preferences” result, but in that case it’s a deliberate or at least indifferent choice by the voters, rather than imposed on them by capping maximum preferences. And FPTP famously doesn’t “lure” voters into indulging a sincere first, second or third choice: it makes it clear from the outset that it’s a sudden-death choice. Anything short of “You can number all (or all but one) of the candidates, if you so wish” faces this risk, although I understand that US voting technology with punching holes in columns may limit the number of candidates that can be preferenced

      • I ought to clarify that experience from NSW and Qld with optional-preferential AV suggests that a significant proportion of people who accept “this candidate led on the first count, but ended up defeated because 50%+ of voters preferred someone else” don’t extend this logic to accept “this candidate led on the first count, but ended up defeated because 47% of voters preferred someone else and 7% of ballots exhausted”. As between two sub-majority pluralities, this cohort seem to prefer one composed solely of first choices as somehow “purer”. I seem to recollect this also came up during the UK’s 2011 “Do you think Nick Clegg is a right-wing neo-liberal sellout or a trendy left Bruxellois cosmopolitan?” referendum, where No2AV argued that AV (as offered, with OPV) wouldn’t “ensure every MP ha[d] majority support” either.
        Now, I don’t think this argument is particularly logical – it is easy simply to categorise exhausted votes as dividing equally between the last two standing, or as retrospectively abstaining, and even if 47% falls short of a majority it’s still a bigger plurality than 39%, and if you accept a 50%+ final-count majority that includes second, third and fourth preferences, it seems arbitrary to suddenly flip over and maintain that everything but first preferences has cooties if the number drops below 50%. But the fact it’s not logical doesn’t mean it doesn’t seem to have some political traction in certain minds (seriously, would No2AV have used this talking-point if it wasn’t both illogical and powerfully emotive?)

      • I don’t think anyone in the US still “punches holes” in ballots (most of those systems were phased out after 2000). Most paper ballots involve darkening a circle for the candidate of choice like one is taking a standardized test. I’ve seen a few where the voter draws a line between two arrows as well.

        But the 5-preference limit does not appear to be a technical limitation. The law requires a minimum of 5 preferences, if I recall correctly, but does not specify exactly 5. The sample ballots for the ranked choice special elections had a significant amount of blank space on the ballots where additional preferences could have been expressed. I think for these primaries there were multiple races on the same ballot (which I’m not a fan of to begin with) but I still don’t know if available space for the races is part of the rationale or not.

        In Cambridge, Massachusetts, where they use m=9 STV, the ballot has an (a+9)*a grid, where a is the number of declared candidates; there are also 9 write-in blanks below, but if you use write-ins you have to leave at least one of the declared candidates off the ballot.

        If there are technical limitations, I would suggest that electoral laws include remedying those limitations the next time equipment is purchased. Personally, I see no reason to arbitrarily limit the number of preferences a voter can express, and would even suggest that OCR scanners + hand-written rankings might be better than the grids currently being used in the US.

      • Christopher: “If there are technical limitations, I would suggest that electoral laws include remedying those limitations the next time equipment is purchased. Personally, I see no reason to arbitrarily limit the number of preferences a voter can express, and would even suggest that OCR scanners + hand-written rankings might be better than the grids currently being used in the US.”

        Unfortunately, in the U.S. equipment issues and limitations are as much the result of politics as they are technical. For a long time, opponents of ranked choice voting used the non-availability and cost of compatible equipment as an argument against reform. Opponents have also been able to portray RCV as a threat to election integrity — because it is time-consuming to count by hand, because of the integrity issues associated with voting machines, and because of alleged “voter confusion”

        For the most part, equipment is designed as well as manufactured by for-profit businesses who perceive their customers to be local election officials. Local officials don’t want to take on the extra risk of implementing a new voting method, since they get the all of the blame when things go wrong but none of the credit when they go smoothly. So in order to please their customers, voting machine makers have dragged their feet on implementing RCV, and have made it seem more complicated and expensive than it needs to be. That includes essentially artificial limits on the number of preferences. This situation has changed in the last few years, as jurisdictions have adopted RCV and needed manufacturers to supply something. But it has taken a long time.

        Handwritten rankings scanned by OCR, while they might be technically feasible, would be politically impossible in the U.S. They wouldn’t be acceptable to integrity activists, and few election officials want to risk trying something different without a very powerful reason. For better or worse, we are pretty much stuck with the fill-in-the-ovals design.

        Christopher is right that there is no good reason to limit voters’ preferences. But reform campaigns have accepted those and other limitations in order to make ranked choice voting happen at all. Reformers would have to be much more numerous and more powerful in order make this better in the face of continued opposition to RCV.

      • Chris, thanks for clarifying that. I recall Professor Ralph Summy (US expat who taught at Univ of Qld decades ago) saying that any statement about US electoral rules must always be prefaced with “It varies from State to State, but…”
        In 1991 Australia Post adopted OCR readers to quickly scan postcodes (our 4-digit equivalent of Zip codes) on mailed items, and private suppliers went along by printing the squares on envelopes. Very similar to a ballot-paper, albeit horizontal, and I wondered why the electoral commission couldn’t adopt similar technology for ballot-papers? Once you know how many votes were for ABCD, how many for ABDC, etc, you could run it through a computer and have the result by suppertime.

      • I feel like if large enough jurisdictions make it a requirement that election machines can handle 50-preference STV or whatever, someone is going to want to bid for that contract. There are over 8 million people in New York City. It’s a big contract. If citizens put the pressure on election officials (many of whom, distressingly enough, are directly elected by the public) to adopt ordinal voting and to require the machines to cater to the electoral system and not the other way around, I can’t imagine that all of these multi-billion dollar corporations trying to get their piece of the elections market will go “nope, sorry, can’t make it happen” when the technology to do so almost certainly already exists.

    • Henry: I don’t, myself, have a problem with “negative campaigning” if it takes the form of “Duke/ LePen is so bad that everyone should preference every other candidate (or at least every significant candidate you’ve ever heard of), as well as their first choice, because at least none is Duke/ LePen and you want to make sure Duke/ LePen can’t win”. I think it becomes a problem when it’s “hold your nose and vote for Lyndon Johnson/ Michael Foot, even if he isn’t your first choice, because at least he isn’t Barry Goldwater/ Margaret Thatcher and you can’t afford even the faintest risk that Goldwater/ Thatcher might win”. ( In LBJ’s case the tempting alternative for US left-wingers was not a Henry Wallace-style third party, but staying home/ abstaining, but same principle applies, despite their growing misgivings about Vietnam). Note that this worked (once) for LBJ but not for Foot. Ie, my problem with negative campaigning is that it dragoons or bullies voters into casting insincere ballots. Caveat that with voluntary voting, “stay home and a plague on them both” is always a third option even with two-way races (and referenda), so there’s arguably a disguised .”split vote” even in those cases which mechanisms like turnout thresholds seek to mitigate.
      Edit to myself: “The result may then be the worst of both pure FPTP (the largest bloc loses because its vote was divided over too many candidates and the winner didn’t have 50% support even after preferences) and preferential systems (the winner didn’t have the most first-preferences).

      • This is a fair point, but from what I’ve seen opposition to “negative campaigns” in the American electoral reformer context is more about stopping campaigns spending the GDP of a small Pacific island nation on negative ads, regardless of whether they say “you must vote for X to stop Y” or “you must rank X ahead of Y” . In any case independent organisations aren’t even allowed to make either appeal, at least directly.

      • My read is that purely negative campaigns (“turn out and vote primarily or even solely to keep X out”) are seen as bad if there is one beneficiary who then wins by default even if many voters have to hold their noses. You might get a Martin Bell as the main alternative… or you might get a Jacques Chirac. A vote cast mainly to block the candidate memorably described by Hunter S Thompson as “that pig####er Nixon” is then construed by the winning candidate, “that pig####er Johnson” (HTS was talking about the 1968 race as he predicted it in late 1967), as glowing endorsement of his policies when it is nothing of the sort. Whereas if the “anyone but Chirac” vote is divided fairly evenly between a Gaullist and Giscard, or the “anyone but One Nation” vote between a Liberal, a National and a Labor candidate, it is hard for whoever wins by default to spuriously claim the mandate of heaven.
        We certainly have negative campaigns in Australia at the forming-government level (1993 and 1996 stand out, with Paul Keating as the second-most-hated and the most-hated candidate in Australia on each occasion… I don’t think we’ve had anything quite that… personal since then, at federal level). At the electorate level, though, AV limits its usefulness. Yes, Wilson Tuckey may be well past his use-by date for many conservatives, but that doesn’t mean you have to vote for his Labor opponent just to get rid of him… you can elect his National sort-of-frenemy, sort-of-running-mate instead and still send a Coalition supporter to Canberra.

      • (Which reminds me, I’m still waiting for a reply from the Approvalista/ Range Voting/ Asterisk Voting/ Triple-Underline-and-Gold-Stars voting crowd as to how their preferred systems would have performed better than AV in enabling conservative voters in O’Connor in 2010 to replace Wilson Tuckey with Tony Crook, as they clearly wanted to do,)

      • Sorry, “if the ‘anyone but Mitterrand’ vote is divided evenly between Giscard and a Gaullist…” – was thinking of France 1974

  5. Just quietly, by 48 hours after the close of voting, any electoral commission in Australia that was not producing at least a rough approximation of preference distributions would be, as they say in the classics, in deep dodo.

    The current tally in the Democratic mayoral primary is 798,491. The final tally in the NSW legislative council election in 2019 was 4,752,827. By midnight on Election Day the media and the NSW electoral commission were able to project 17 of the 21 available seats in the council, and that’s under the highest magnitude STV system in the world.

    • Yes, but in Australia you have how-to-vote cards which allow for projections of how most voters are going to vote, which doesn’t exist here. And the projections of the last seats in STV races aren’t happening on the night, because people don’t know how the preferences will flow for the smaller parties. We don’t see actual below-the-line data until after the button push for the upper house.

      This would be the equivalent of if every voter could vote in pre-selection and that the media had to guess how those preferences were going to flow between candidates. I don’t follow Aussie pre-selection that much but the coverage I have seen suggests that when pre-selection is open it’s hardly as clear cut as projecting a general election.

      I think that given the bubble sheet format they SHOULD be able to release voter preferences on the night, but it isn’t in the law and the media haven’t demanded it yet.

      • How-to-vote cards have nothing to do with the way that preferences are counted. No parties use primaries to select candidates. A media whose idea of election commentary is to designate AV as hopelessly complex, and then leap onto the nearest table, lift their skirts and scream a lot is hardly a rational basis for judgment.

        Incidentally I know almost no-one who uses Aussie in regular conversation. It occupies roughly the same linguistic register as Yank. No-one at Fruits and Votes would ever describe the event last November as a Yank election.

      • NY is notorious for having some of the worst-managed elections in the country. It took weeks for them to produce tallies for the Presidency last year, and we have reason to be thankful that the state is solidly Democratic. It’s not terribly surprising that this is going the way it is.

        That being said, the reason the federal Electoral Commission is able to produce such quick tallies is because we do not conduct preference ‘counts’ on the night of an election-the AEC selects the two candidates who they think will finish first in each seat, and distributes votes between these two candidates. I don’t know what the exact procedure is at the state level but I think it is similar in most states. This works fine in most seats, but it is less useful in non-classic divisions where the order of exclusion is important, such as Melbourne Ports in 2016 and Prahran in 2014. State and federal electoral commissions don’t do full preference counts until every vote has arrived, so in those cases we are left hanging usually for a number of weeks.

        Obviously this approach would be problematic in the NY mayoral primary: if the two candidates had been selected too far ahead of time, we could be seeing a Yang-Stringer 2CP count now, and even knowing first preferences we can’t be sure which of Garcia and Wiley will make the top two. The US might do well to adopt this approach for House and Presidency elections, but it may be beyond the capacity of the voting machines they have available.

        More generally, the AEC and state electoral commissions, with the notable exception of the ACT, are perfectly willing to keep people waiting when the count is complex. While we did know 17 of the 21 seats in the NSW Legislative Council a few days after the election, this more reflects the fact that preferences are not terribly important in that election-it took three weeks for every ballot to be entered into the system and the final seats to be declared. This happens now in the Senate more frequently. GVTs allowed Senate results to be roughly divined based on first preference totals: now, although preferences are generally somewhat less important, we have to wait weeks for ballots to be entered into the system also.

      • It would not be a tenable argument that NYC election administration is very poor and therefore NYC election administration is very poor.

        Electoral legitimacy has become problematic. The same technique for subverting electoral legitimacy–charges of imaginary electoral fraud–has now been used in the US, Myanmar and Samoa. One of the necessary conditions for restoring legitimacy is for results to be posted as early as possible and in as great detail as possible.

        Electoral commissions in Australia are slowed by the need to physically count actual pieces of paper and then scan their contents into a database. In NYC those steps have already happened. It would be a simple matter for the NYC elections board to design their system to generate an interim distribution of preferences on election night. I did write about a rough approximation not a final count.

        There is at least a nonzero chance that the current leader may be overtaken by the distribution of preferences. If that happens there is a much more than nonzero chance that will be described as a malfunction, that the unbearably complex system of RCV deprived the true winner of their victory.

        Both the general legitimacy of the election, and the particular legitimacy of this first RCV election, would be greatly enhanced by an interim distribution that could be updated as counting proceeded. The Tally Room has similar thoughts.

    • Incidentally I know almost no-one who uses Aussie in regular conversation.
      I’m a Kiwi, and basically everyone I know uses it. Only excluded from more formal settings.

      • Not many answer to “Canuck” except the local NHL hockey team, 22 of whom were born outside Canada.

  6. I thought you were referring to media projections, not electoral commission projections; how-to-vote cards are certainly used to make media projections in three-cornered contests based on the first preference results. I agree that especially given the bubble sheet scanners there should have been a formal interim distribution of preferences, but given how new ordinal elections are here there hasn’t been demand for it. I expect that will change as they become more entrenched here, as the trend seems to indicate.

    I know Australia doesn’t do primaries. I am saying that even in pre-selection meetings it can be difficult to project the winner, and that mass-participation American-style primaries are pre-selection on steroids.

    In American English I would say that “Aussie” is informal but far less informal than “Canuck” or “Yank” (a word that Americans, in general, don’t use). I would put it at the same level of formality as “Kiwi,” and aside from formal academic/journalistic writing I would see no issue with “Kiwi elections” instead of “New Zealand elections.” However, if it legitimately causes offense I’ll avoid using it here in the future.

    • I wouldn’t say it’s offensive, it’s just not as common in Australia itself as in media representations of Australia and jars a bit.

      • The only time I hear it here is when people yell “Aussie, Aussie, Aussie, oi, oi, oi!” at international sporting contests, which is semi-patriotic but also semi-ironic, something like chanting “USA! USA!” while clapping on one’s knees. Otherwise it’s a bit Monty Burns.
        If a non-Australian does use the word, the double -ss- must be pronounced as a voiced S or short Z, as in “rose” – not like “ross”. “Ossi” with an unvoiced S means East German.
        Like the Hungarians and Germans, our name for ourselves in our own language is different from what other nations call us… It’s “Strine” with a long dipthong, longer than one syllable but shorter than two, and the nostrils should be flared as widely as possible. Alan or Henry or Chris C, post a Tiktok or somesuch to illustrate, would youse?

  7. Tom Round: “In 1991 Australia Post adopted OCR readers to quickly scan postcodes (our 4-digit equivalent of Zip codes) on mailed items, and private suppliers went along by printing the squares on envelopes. Very similar to a ballot-paper, albeit horizontal, and I wondered why the electoral commission couldn’t adopt similar technology for ballot-papers.”

    In the U.S. some equipment does exactly this. It’s very controversial because, even though there appears to be a paper trail documenting the voter’s choices, it’s not in human readable form so it is not really a voter-verified paper trail.

    • I think he’s talking about machines that read hand-written postal codes (which I think the US Postal Service uses as well). That would definitely be voter verified if a voter hand-wrote ranks for the races, though if it were like Texas where in some years I literally have 90 races on the ballot (every single judge is elected in a partisan race) that would be quite tedious.

      • Oh, I see. That is different from what I thought. And yes, the U.S. mail system does use that technology. We’d have to look up studies that estimate the error rate in order to talk knowledgably about whether the method could be used to interpret handwritten numerals on election ballots. As a practical matter, this is probably not worth the effort in the U.S. because election integrity is such an emotional issue for everybody.

      • I don’t think there would be an integrity issue because a manual recount is possible.It’s even less of an issue than the voting machines used in many states which have no voter-verifiable paper trail and a “recount” just makes sure that the data entered into the final count matches the receipts from the voting machines.

    • That makes me think that if she loses narrowly on the ballots that are already in, on the ballots that haven’t yet been entered (which are said to be disproportionately white voters and higher educated voters), that Garcia may end up winning the race.

      I do wonder if such a change between the interim and final results will be controversial at all.

  8. A very preliminary tally (with 124,000 ballots not yet counted) is here: Paging through to the 11th round, we find that of 941.832 ballots processed so far, 219,944 — or 23% — didn’t rank either of the two finalists (were “exhausted” in RCV parlance).

    So I might have been wrong earlier in this thread when I said that limiting voters to five choices is not necessarily a major problem. It’s really hard to tell, especially without access to the raw data. At a minimum, you need to know how many of those 219,944 voters used all five of their rankings but were still not included in the final round.

  9. A new, presumably less tainted, tally now shows 14% of the ballots being exhausted in the final round — down from 23% in yesterday’s erroneous report. I think that the relative positions of the four major candidates — Adams, Garcia, Wiley and Yang — have changed very little.. The URL remains the same:

    • Yes, the exhaust problem looks a bit less serious now. I took a look at the exhaust rates on exclusions. Of the votes for the minor candidates (all those besides Yang, Garcia, Adams and Wiley), 17.2% were exhausted. Of votes for Yang (or those that were with Yang when he was excluded), 27.6% exhausted, and 28% of Wiley’s votes exhausted.

      Ultimately I’ll be interested to see if the NYCBOE releases ballot records to further understand the impact of truncation. A ballot only showing 1 Wiley, for example, seems less concerning than, say, 1 Donovan 2 Chang 3 Donovan 4 Yang 5 Wiley, since the latter voter may have had preferences amongst the top four but didn’t guess correctly which two would end up there.

      • In at least one respect, this election is a good example of why truncation can sometimes matter. Supporters of “minor” candidates who are strongly opposed to Adams had to decide whether to rank Wiley in fifth place or Garcia in fifth place. They might reasonably guess that one of those two would be the second finalist, but could not reasonably guess which one. That leaves such voters with only three sincere rankings before using #4 and #5 on the two “stop Adams” candidates. If Yang had performed better and been closer behind the other three, the situation of such voters would have been even more difficult. Fortunately, the number of voters who prefer three or more candidates to any of the front runners might be very small.

      • Well, 1 Donovan 2 Chang 3 Donovan would be particularly concerning because the software is supported to catch voters casting two preferences for the same candidate, but assuming you meant to put someone else that isn’t Donovan, Adams or Garcia in that spot it would still be concerning.

    • I agree truncated preferences are a distortion of the system. That is one reason I and other Australians find support for truncated preference systems where the elector is restricted to expressing a single preference incomprehensible.

      • Touche. Although as noted, “at least” single-preference FPTP is brutally honest that voters cannot afford ever to vote sincerely if that might split the vote. Limited-preference AV by contrast might sucker too many would-be candidates into standing – “let a hundred flowers bloom”, and that always ends up badly.

      • (France and now Latin America suggest that runoffs have a similar problem… yes, the final winner will have 50% plus, unaffected by vote-splitting on the second ballot, but the choice of final winner may be skewed if too many candidates enter, split the first-round vote, and lead to an unrepresentative Hobson’s choice for the runoff).

      • Also, could there please be a moratorium on pundits saying things like “This RCV system would have been great when it was Hillary Clinton vs Barack Obama or Bernie Sanders, but it’s no good in this particular contest because there are [too many/ too few] candidates from the political faction I support.” Write out 100: “WE HAVE TO PICK A VOTING SYSTEM THAT APPLIES TO ALL ELECTIONS FOR THAT OFFICE. WE DON’T CHOP AND CHANGE THE VOTING SYSTEM AT EACH ELECTION OT MAXIMISE THE CHANCES OF THE FACTION THAT YOU, BLOGGER, SUPPORT.”
        [“too many candidates” = more than the number of preferences, so you can’t back them all. “too few candidates” = see the above complaint about how “I only liked one candidate and hated all the others, so I didn’t get to use all my preferences, so I had less influence on the election, totally not fair!” (Hold your nose and express more preferences, crybaby, or nominate more candidates, but otherwise this is like a fighter pilot complaining that she never had the fun of deploying her ejector seat, or a Borda supporter complaining that Princess Margaret never got a chance to succeed her sister).

      • Tom, I didn’t know it was still possible for one to have a blogroll. I also agree with your comments on limited-preference RCV being in some ways more concerning than FPTP. This is particularly important I think in the US context, where RCV is generally introduced by referendum and where supporters of RCV in such a referendum are likely to say things like “RCV will mean you can’t waste your vote!”. I certainly can’t point to a causal effect, which probably wouldn’t have shown up in NYC anyway since RCV is replacing a runoff system, but it seems like a valid concern.

        The tactical appeals are also more straightforward, even if they are obviously less necessary. For a stop-Adams voter who likes the minor candidates, saying “after your first four preferences don’t rank Yang fifth since Yang won’t make it to second place, instead rank Garcia” is arguably less clear than the FPTP appeal, and probably cannot be expressed in the form of a bar chart.

  10. Rest of the world: “Ho-hum, another US electoral stuff-up.”
    American pundits: “Whoa, this crazy new system has caused a logjam that is totally unprecedented in our Nation’s shining history of orderly and uncomplicated elections.”

  11. NYC 2021 has, however, had the silver lining that I can now save precious reading time by purging my blogroll of any pundit who complained “I only liked one of the candidates, so I had less influence on the outcome than other votes who liked three or four of them, which is discriminatory”, because this is a quick and reliable litmus test for whether someone grasps the essentials of RCV enough to deserve to publish any opinion about it. It’s up there with “the Hare quota is higher than the Droop quota so it’s less favourable to small parties” and “if only the abolitionists had had more influence at Philadelphia in 1787, they could have insisted that the US Constitution count slaves as 100% of a full person instead of only 60%”.

  12. (Bob Richard) “That leaves such voters with only three sincere rankings before using #4 and #5 on the two “stop Adams” candidates”
    That is definitely one area in which Approval has an advantage over limited-preferential RCV. At least you can actually put someone “last”, not merely hope to pick which strong horse is best-placed to outpoll that person.
    Team Approval are out exulting that this one disastrous, no-good, undemocratic result has finally discredited sequential preferences as against simultaneous equal approvals for all time (Aaron Burr not being reachable to comment).
    And I see the “RCV might be justifiable if it elected a runner-up with over 50% on the final count, but if the last candidate standing is under 50% as well, the screw it, may as well elect the plurality winner” school of thought is very widespread.
    I also see much hankering for runoffs as better because they focus the mind on a clear two-horse race. No consideration apparent of those two horses being chosen by the lottery of first-past-the-post.

    • One of the biggest issues I have with approval voting (aside from the fact that it does not allow a majority of voters to elect the candidate of their choice unless they’re acting in a coordinated strategic manner) is that , while it does allow one to put a specific candidate “first” or “last,” one can only do so by voting tactically and by potentially voting insincerely, either by only voting for their first preference or else by voting equally for every single candidate other than their least favorite. It is no different in that regard than FPTP except that voters who don’t realize the strategic imperative are more likely to cast a “sincere” ballot which does not benefit their preferred result.

      Heck, for all we know, if this primary had been conducted using approval voting then Paperboy Love Prince could be the mayor-presumptive right now.

    • Finding a solution to this problem–that in a crowded field, two-round majority is basically a FPTP “lottery” in its first round–has been an obsession of mine for some time. Once upon a time I thought allowing two votes in the first round would help. I have been completely disabused of that notion (St. Louis used such a system for the first time recently: approval vote in the first round). The downside with this rule is the risk that a large minority bloc has two candidates and both get into the runoff, shutting out other forces pretty much entirely.

      The other idea I have had is to determine the two runoff contenders via an open-list, with alliances being possible among candidates in that round). I made this “modest proposal” in 2013 for Egypt and other transitional countries, but there is no reason it could not be applied elsewhere in cases where parties are weak or non-existent, as they pretty much by definition are when the election in question is a primary.

      Yet another possibility is to allow three candidates to compete in the runoff and hope that elites and voters can figure out how to coordinate in the second round. Yes, I immediately recognize the risks, but the point is to resolve the fragmented first-round field problem, which might imply tolerating more than two runoff contenders.

      A useful addendum to that last idea is Miguel Madeira’s suggestion: have the runoff include the top N candidates whose combined votes top 50%+1.

      • The open list system (I think Honduras has used it without a runoff as well as Uruguay) is quite sensible, though to some extent that just shifts the “lottery” part to the Alliance level (for instance, there’s no real evidence that Jospin was supported by a majority of the left in 2002, just that he was the plurality winner of the left).

        Argentina does this to some extent with their PASO system where all parties must compete in an open primary, and parties earning less than 1.5% of the total votes are excluded from the general election. There’s no reason you couldn’t increase that threshold significantly.

        I do think the French may be onto something with their non-presidential races; parliamentary candidates need to earn the support of 12.5% of eligible voters in the first round to advance to the plurality second-round, and regional elections require a list to earn 10% of votes cast to advance to the second round where the plurality bonus is doled out.

        However, I’m not sure I approve of having the final round be plurality when there are more than two candidates; you avoid the issue of fringe candidates splitting the vote but when you have 3-4 candidates with a shot to win (like here in New York) the tactical voting pressure is still there.

  13. This was what I had meant by opinion pieces serving as virtual how-to-vote cards, though I also see that it was written after most early voting and absentee votes would have been cast. I think it’s interesting the head of FairVote would openly associate himself with a piece on how to vote tactically when one of FairVote’s big selling points has always been that ranked choice eliminates the need for tactical voting.

    • I think you’ll find Richie was talking about ranked choice voting where the elector has the right to express a full set of preferences.

      • Alan: “I think you’ll find Richie was talking about ranked choice voting where the elector has the right to express a full set of preferences.”

        Actually, Richie’s New York Times piece deals with involuntary truncation pretty thoroughly even though it (deliberately) avoids the jargon we use to talk about the subject.

      • Richie and FairVote in general talk about AV with full preferences, but in this specific piece he’s essentially offering a variety of how-to-votes for various voter profiles. It is a very well-written reaction to the reality that many voters could not rank their top 5 preferences without exhausting and disenfranchising themselves on the meaningful count.

        The potential risk there is that this piece is used to say “RCV doesn’t end tactical voting and therefore is bad” when that only applies to this particular form of RCV with limited preferences and a large number of candidates.

      • Christoper: “The potential risk there is that this piece is used to say “RCV doesn’t end tactical voting and therefore is bad” when that only applies to this particular form of RCV with limited preferences and a large number of candidates.”

        For single-winner elections, I would much rather have RCV with only five choices than any other method that I’m aware of, except of course RCV with more than five choices. But that’s not going to stop opponents of RCV from using this argument.

        I once heard a prominent advocate for RCV argue in favor of increasing barriers to ballot access — filing fees, numbers of petition signatures, and so on — in order to reduce the number of what s/he called “zombie candidates”. (Such barriers are already much higher in the U.S. than most other developed countries. For more on this, see Ballot Access News.) The problem for me is that most of the candidates I can support are “zombie candidates” to a lot of other people.

        Involuntary truncation is going to be a fact of life in the U.S. for a while longer. That’s because, even where electoral reform activists are strong enough to win adoption of RCV, they aren’t strong enough to win it on the best possible terms.

  14. On the topic of truncated ranking, in the process of looking into this I found that Minneapolis conducts a STV election with DM=3 (for the Parks Board) while only giving voters three rankings.

    • Oh, wow. I knew there were some local bodies there with M=3 STV, but I had no idea the voters were limited to M preferences! That is very bad. By extension, this rule would require plurality for a single winner.

      • msshugart wrote: “… By extension, this rule would require plurality for a single winner.”

        Minneapolis has the same three choices for single winner offices (mayor, city council). The ballot paper looks the same for both types of contests.

        I don’t think I was ever aware that the STV contests in Minneapolis were limited to three rankings. It is indeed very bad. But is it as bad as MNTV for those offices? And should FairVote Minnesota have waited for more modern equipment to arrive before campaigning for STV? If you answer “yes” to the second question, you should propose a solution to the perennial chicken-and-egg question: if election officials are not required to adopt new voting methods, why would they ever buy compatible equipment?

        Minneapolis illustrates another practical issue facing reformers. The reason the city council is elected by IRV and some minor boards and commissions are elected by STV is that the council has always been elected from single-member districts and the minor commissions have always been elected at large. FairVote Minnesota made a conscious decision not to challenge that arrangement, hoping that experience with STV would lead voters to be more open to it for the council later on. That was 12 years ago and I’m still not sure how I feel about this strategy.

      • They also have an m=2 race (for the Board of Estimate and Taxation) still with 3 preferences. Most races (mayor and city council) are single-winner AV; it doesn’t “require” plurality for the single winner, since votes can transfer, but it increases the chances that the winner will be elected with only a plurality of first-count votes rather than a majority due to exhaustion.

        It’s pretty fascinating because non-partisan Parks Board and Taxation Board elections in a large city aren’t usually races where one would expect most voters to be well-informed about the candidates, so the dynamics of ranking candidates is quite unique. They also have CSV files with every ballot’s preferences available, which could provide interesting research on how voters use their limited preferences (though whether voters have any idea who the front-runners are in a race like that is another question).

        I also found a potentially unique clause when reading their election ordinance: any candidate who is “mathematically eliminated by the sum of all ranked-choice votes comparison” is excluded. That is defined as: “For multiple-seat elections: The candidate could never win because the sum of all ranked-choice votes for that candidate would not be enough to equal or surpass the current vote total(s) of any of the top “x” continuing candidate(s) with the highest current vote total(s), where “x” equals the number of seats yet to be filled for the office.”

        So for the 1st count in the 2017 Parks Board election, in addition to write-in candidates, two candidates who wouldn’t be excluded under traditional STV counting rules (because the total of the votes of the candidates beneath them could allow them to pass the candidate above them) were excluded under this system, because the total number of ballots they’re mentioned on means that they could never pass the top 3 candidates.

        I can’t think of another STV system that has a similar rule.

        I find this rule fascinating because it avoids unelectable candidates acting as spoilers, and could result in significantly different results for the last few seats in some larger-m STV races. It wouldn’t solve the GVT issue in Australia because those snowballs preference every candidate within them, but it could change results in places that use the Australian Senate above-the-line rule (probably favoring lower-ranked candidates in the bigger parties over the lead candidates of minor parties) or traditional STV.

      • and the other part of that “eliminate those candidates who aren’t on enough papers to win” rule is that by limiting to 3 preferences the Minneapolis system greatly increases the chances that a candidate won’t be listed on enough ballot papers, while a larger limit or (better) no limit reduces that factor to voters voluntarily deciding not to vote for those particular candidates.

      • Bob and Christopher, good to know they do not in fact extend a rule of “M rankings” to M=1 (or M=2).

        Also that elimination rule is indeed very interesting.

  15. The media have called the race in favor of Adams. There are some absentee ballots which still may be “cured” but the number is not enough to overturn the election.

    Based on the interim results posted yesterday:

    On the penultimate count, Adams had 40.5%, Garcia 30.4%, and Wiley 29%. 7% of ballots were exhausted at that count. 60% of the exhausted ballots had exhausted after Yang was eliminated.

    Of Wiley’s ballots, 51.1% went to Garcia, 18.4% went to Adams, and 29.2% exhausted. Even if all of those exhausted votes from that count went to Garcia, she still would have narrowly lost (though if she had received more of Yang’s exhausted votes it would be a different story).

    The final count was 50.5% Adams, 49.5% Garcia. 14.9% of ballots exhausted, and I will be very interested to see what share of those exhausted after choosing 5 preferences and what share simply didn’t fully express preferences.

    • They also had the following statement posted on the results page:

      “The June 29th ranked choice voting reporting error was unacceptable and we apologize to the voters and to the campaigns for the confusion.

      Let us be clear: RCV was not the problem, rather a human error that could have been avoided. We have implemented another layer of review and quality control before publishing information going forward.

      We can say with certainty that the election night vote counts were and are accurate and the RCV data put out today is correct as well.

      As we continue to count absentee ballots and run further RCV tabulations, we will do so with a heightened sense that we must regain the trust of New Yorkers. We will continue to hold ourselves accountable and apologize to New York City voters for any confusion.”

    • I think that result, despite the efforts to blamer RCV for the incompetence of the election management body, is actually very good for RCV. Adams ultimately won so there is no case that the true winner was unfairly deprived of their win. The final order was different from the first preference order, so there is no case that RCV makes no difference. Assuming the system survives until next election, I think you’d see candidates making much more explicit preference swaps well before the last weeks of the campaign.

      I’d be interested to know the level of donkey voting. The donkey vote, voting 1, 2, 3, etc in straight ballot paper order is fairly well-known. What’s less well-known is that before 1983 you would occasionally see ‘women’s donkey voting’, where people would vote a donkey for all women before all men and then a donkey among the men. Robson Rotation cures donkey voting. The women’s donkey was common enough that expert scrutineers would look for it and factor it into their calculations.

      In 2013 the federal senate election for Western Australia had to be rerun because (literally) a box of ballots fell off the back of a truck. The electoral commissioner and his deputy in Western Australia both resigned. There was no particular campaign for their resignation, it was just accepted as something that had to happen.

      • The Australian senate uses a system called bulk exclusion that is more or less the consequence of a crazy decision to abandon the traditional way to transfer preferences in STV in favour of transferring all ballots tot he next available preference. It’s an unnecessary feature that should be abolished. The way it works is (in theory):

        In most systems, candidates are excluded one at a time. Only the Senate allows for more than one candidate to be excluded at one time. This is called “bulk exclusion”. Under bulk exclusion more than one candidate can be excluded at the same time if it can be proven that none of the votes of these candidates could possibly affect the order of election of the next candidate to be elected.

        The supplemental vote systems to elect the president of Sri Lanka and the mayor of London use the ultimate version of bulk exclusion in that all but the two leading candidates are eliminated as the first stage of the count.

        Neither Hare-Clark system uses it. No preferential system should.

      • I’d wager donkey voting is not going to be particularly common in the States because neither voting nor expressing full preferences when one does turn up to vote is compulsory.

        Bulk exclusion can’t change the outcome of a single-winner race, so I’d see no reason not to use it in order to save time. It can change the outcome in exceptionally rare circumstances in multi-winner races by changing when candidates get a quota and the size of their surplus.

        The Minneapolis variant is on another level because it can change who gets excluded and ultimately who gets elected by removing potential spoilers, allowing candidates who would get excluded early in the count due to low first preferences to survive into later rounds.

        Whether or not that is a good thing is an open question.

      • I would distinguish two forms of “bulk exclusion”:
        (a) “expedited exclusion” where the two or more get eliminated simultaneously only when and because they could not possibly change the result. Eg, in traditional STV, assuming no one is over quota, if the lowest three candidates have 30 votes each and the fourth-lowest has 100, you can safely exclude all three at once. This version is specifically allowed in the Proportional Representation Society of Australia’s “PR Manual” so I don’t think its merits are tied to the Australian Senate using Uniform Gregory transfers since 1984.The point is, those candidates were doomed anyway, so you’re just hastening the inevitable.
        As opposed to what I’d call (b) “summary exclusion” where all candidates who aren’t one of the N highest, or who are below N% of the votes or of a quota, or who have fewer than N votes, get the chop even if they might in hindsight have been able to build up a quota had they been left in the count.
        RCV with truncated preferences usually goes together with summary exclusion (eg, London, Sri Lanka) but that’s not inherent. You could, eg, allow or even require preferencing all candidates but then stipulate that only the two with the most first preferences survive the first count. This was the “contingent vote” system used in Qld 1892-1942 Variants can be imagined, eg, stop the count after five or ten stages of counting and declare the highest candidate elected by default.
        Alternatively, you could limit every voter to, say, six preferences maximum but still conduct a simulacrum of a full AV-STV count. Depending how evenly later preferences are distributed among candidates, this might go past six stages of election/ exclusion but eventually you would reach a point where all votes are non-transferable and you would (again) have to elect the last candidate(s) standing by default, just like a normal AV-STV count if all voters voluntarily truncated.
        So, the takeaway here is that “limited preferences”, “expedited exclusion” and “summary exclusion” are all distinct features.
        PS: One could combine “expedited” with “summary” exclusion for a runoff system that avoids the final two being chosen by a FPTP lottery, while retaining the careful deliberation for which (we are assured after the Great NYCRCV Debacle Of 2020) second ballots are renowned, by providing that if no candidate has 50%, but the two highest together have over 66.67%, those two go into the runoff; if not, but the three highest together have over 75%, those three are shortlisted; otherwise, the top four proceed. You could use plurality on the second ballot if three or four qualify (that’s a lesser evil of runoffs than using SNTV on the first ballot), or you could allow for up to three runoffs to narrow down to one, if you’re really hardcore. Not sure if anywhere uses this version of “top-down” expedited exclusions but the “bottom-up” type of expedited exclusions is, as noted, quite common in STV and IRO-AV systems.

      • “I’d wager donkey voting is not going to be particularly common in the States ” (Chrs. Burge)
        In which case I’m curious why US States expend so much time and attention on ensuring that ballot positions are drawn by lot, or rotated equally, or (as in California, I believe… MSS?) varied among different counties, and why the rules I’ve viewed usually stipulate that the largest parties (Dems and GOP) get the topmost and leftmost positions on the ballot-paper…
        Voluntary FPTP voting for a single position might conceivably reduce donkey voting – there is no conceivable incentive to turn out and tick that candidate unless you want to support her, whether sincerely or tactically – but the equation changes with multiple positions, whether one race with MNTV or multiple single-seat races. Certainly the US, UK and NZ have noticed that MTNV for local council wards tends to favour the early names in the alphabet, albeit not to the extent of the Dail Eireann.
        I don’t think compulsory and/or preferential voting create donkey voting so much as reveal it. (This extreme transparency is a recurring PR problem with PR… you can go back through the spreadsheets in hindsight and work out that Smith would have win if just 4,367 to 4,985 of her supporters had voted for her main rival, but if Approval produces a winner favoured by 57% of ballots, this is indistinguishable from 57% enthusiastic support, whereas for all we know it could well be “Oh. So it turns out both Bernie and Hillary putpoll Trump. So if only me and mine had just voted for one of those two, instead of both, we could have elected our first choice instead of our second…”).

  16. Re “Voluntary FPTP voting for a single position might conceivably reduce donkey voting …” (me) – This is actually pretty rare. In the US, of course, elections may have short rather than long ballots but there is almost always at the very minimum a mayoral plus a council race and at maximum everything from President, Senator, Representative, Governor, State Assembly, State Senate, State Comptroller of Insurance, etc, down to Municipal Bonds Assessor plus likely one or more referendum propositions. But even in the UK, national House of Commons elections are not uncommonly held on the same day as local elections (1979, 1997, 2005, 2010, 2015) while NZ used to hold its largely symbolic alcohol/ temperance plebiscites at the same time as its national polls. India also often runs some States’ elections on the same day(s) as Lok Sabha polls (eg Andhra Pradesh, Arunachal Pradesh, Odisha and Sikkim in 2019). Canada would be an outlier here for federal and provincial elections, but seems to have directly-elected mayors in a lot of cities. And even in local elections for councillors only, with no directly elected mayor, wards are often multi-member.
    So – in a lot of “FPTP” systems, even with voluntary voting, there is an incentive to “come for” one favoured candidate but “stay for” others whom you vaguely think you kind of sort of know about and like. Which can create an incentive for donkey-voting avant la lettre.

    • Very true on the “donkey vote” for down-ballot races. At least in Texas this most obviously manifests itself in the form of a straight-ticket vote for all of the candidates of one party. Previously, there was a straight ticket option where voters could simply mark the party of their choice and have it apply to all races, but in an attempt to increase their chances in judicial races in large urban areas where there are dozens of races on the ballot, that system has now been abolished and voters must mark every candidate individually (presumably the Republicans hoped to win some of those races by literal “exhaustion” as Democratic voters grew weary of filling in circles on their paper).

      In non-partisan and primary races it often manifests itself in the form of identity-based voting; there have been primary elections in Dallas County, Texas where literally every woman, regardless of ideology, ethnicity, incumbency, experience, or any other factor, won races contested against men. There is also a strong tendency to co-ethnic voting (which has a larger impact on Democratic primaries and non-partisan elections in diverse jurisdictions).

      As far as the ballot order, random drawing is used in Texas for primaries and non-partisan races. I believe there was/is a down-ballot donkey effect favoring those at the top of the order, and if I’m not mistaken alphabetical order was banned. In races being held across multiple counties, each county has a separate random drawing.

      For partisan races, the parties are in the order of the votes received in the last statewide election. There are only 3-4 parties listed on the ballot so it would be hard to think there’s much of a donkey effect, but it does make it easier to find that candidate in every race.

  17. Maybe there is a previous F&V thread that reached 100 comments. I can’t recall. I am posting this one only because I noticed it would be the 100th on this thread!

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.