Chasing perfection; getting nowhere.

By Christopher Nicholson of Portland, Oregon. Christopher is a sophomore political science major at Reed College, studying social choice theory.

Our current voting system is terrible. There was a BlueOregon post last month ("Instant Runoff Voting: The Time Has Come") which argued for supporting the Instant Runoff Voting bill before the legislature. There was heated debate, even among Blue Oregon members, about this proposal. The bill never got a work session or hearing, and essentially died in committee.

I supported the IRV bill, even though I don't really like IRV. So why support it? Because it’s a form of preference or ranked voting, where people are allowed to rank their choices in order of preference. Pretty much every voting theorist I've ever read agrees, ranked voting is a much better balloting method than regular voting, because it allows the voter to express more information. The problem arises when someone has to decide what to do with that information.

Here's what makes voting theory so much fun. If you're using ranked voting and you have more than two candidates and more than one voter, there is no "right way" to count the votes. Every ranked voting system is flawed in some way. I'm not going to get into the details of why this is, but if you're interested in learning more about it, you can read about Arrow's Theorem, which is what proves this to be the case.

So that's the mathematical reality. Here's the political one.

Ranked voting gives election officials more information about who voters want. The problem is that no one wants to support a specific ranked voting system when a majority of the population can find a flaw with any given system. Thus, no system gets implemented because whenever a ranked system is proposed, it is shot down by the very valid arguments of those who support a rival ranked system.

The fact is, however, that the system we currently use is probably the second worst possible. (The worst would be randomly selecting a ballot and declaring that ballot's choice the winner). When there are more than two candidates on the ballot, the current system denies voters the ability to fully express their preferences.

We need to begin studying the political ramifications of different systems of ranked voting. None of them are perfect, but almost all of them are better than what we currently use. If we continue to wait for mathematicians and political scientists to present us with "the perfect ranked voting system," we'll be waiting forever.

Comments

  • Brian Wagner (unverified)
    (Show?)

    I've supported IRV for years because of the belief that the system we have is a poor compromise for a supposedly thriving democracy to have to accept even as the nation changes. One of my favorite parts of ranked voting (other than the fact that it substantially alters the ability of third parties to reach out to voters) is, as you say Chris, the amount of information that becomes available on preferences. At the national level, it has always annoyed me when politicians talk about mandates--I'd love to see mainstream voting statistics released in newspapers the day after every election that make it clear to everyone that the people around them have a lot of doubts--knowledge of this sort helps to substantiate public debate and deepen it so conversation moves from my candidate beat your candidate, which often ignores third party and non-voting opinions, to a debate on why the winning candidate might have only won 25% of the first-choice votes.

    Being in the UK now, I can imagine how much of a difference IRV would make in the upcoming May 5 general election. No one seems to want Tony Blair, but people fear the alternatives. How many people would vote for Liberal Democrats if they didn't think it would tip votes toward Conservatives, or would change their opinion about Blair if they saw that he won based entirely on second choice votes. Politicians live and die in part on their popularity, IRV would destroy the facade some try to create of support by taking a step toward more total information for voters.

    Perfect? Of course not, but since little in our nation or state has remained stable over time, what convinces so many people that voting should be exempt from this progressive attitude?

  • LT (unverified)
    (Show?)

    Look at the results for the last legislative election, and how many races just had 2 candidates. Is anyone saying, for instance, that if you lived in District 19 you would have rated Grisham the Democrat first and Doyle second when you wanted Doyle GONE as he was after he resigned? If Rob Brading in District 49 had gotten 1600 more votes than Karen Minnis, she would now be a former state rep.

    Seems to me in the real world more attention should go into the campaigns, such as why money should flow into a group like FuturePac and then out to "selected" candidates rather than flowing directly to candidates, and whether campaigns should be locally controlled or answer to a caucus.

    How did Alan Bates turn a Republican St. Senate district into one represented by a Democrat? Where is the effort to find someone in each district to run who has a chance to win and who will campaign in the whole district rather than just in their town? Should the Democratic Party or the caucus be the organization giving support to the legislative candidates? Not to mention the role the Bus Project played in turning the Senate Democratic.

    Furthermore, explain to the busy person with a full time job, family, other activities why they should learn a whole new system of voting just because someone has a theory.

  • (Show?)

    LT, you can't assume that the candidates on the ballot would remain exactly the same. An IRV system of some sort would encourage minor parties and independent candidates.

  • Rupert (unverified)
    (Show?)

    The problem with IRV that NEVER gets discussed is that the underlying theory presumes that voters have fully developed preferences. In the case of a presidential election, this assumption may be fair (e.g., 1.Kerry, 2.Nader, 3.Cobb, 4.Bush, etc...)

    However, in most races, voters have very little information about the candidates. If there are five people running for city council or state rep, most voters will like one candidate and know almost nothing about the other four. In this case, you'll have voters arbitrarily picking their later preferences, which can really screw-up the election result. (I heard that in one country with IRV, candidates with last names that were earlier in the alphabet were much more likely to be elected, because voters just ranked their later choices in alphabetical order.) I'd much rather have a true run-off, as we do in Portland City Council races for example, so voters have a chance to learn about the people they are voting for.

    Also, in a hotly contested local partisan race, most voters will have only heard of the two leading contenders -- the Republican and the Democrat. So even if their preference might ACTUALLY be for a minor party candidate, they will just give the No.2 slot to the other candidate they have heard of -- even if that person doesn't share their priorities at all. This could wind-up working against the goal of more accurately reflecting voter priorities.

    If voters were more conscientious, IRV would work. But the bottom line is, most voters are ill-informed. Which makes IRV a bad system.

  • David Wright (unverified)
    (Show?)

    Rupert,

    I would agree that most voters are ill-informed.

    That is a completely separate issue from the voting mechanism, however.

    I don't think you've made the case that ill-informed voters cause worse results under IRV than we get now. There are races now where voters don't know anything about any of the candidates. In such races, I choose to abstain -- but many do not. So the problem of voters expressing a meaningless preference is hardly limited to IRV.

    Working to make the voters better-informed should theoretically give us better results no matter which voting system we use, as it would eliminate some of the "ignorant preferences" expressed in any system.

    However, under IRV a better-informed voter base would theoretically give us better results than the plurality system. Seems like an argument in favor of IRV to me, in spite of any flaws that system might have.

  • panchopdx (unverified)
    (Show?)

    FWIW, If IRV had been in place in 2002, there's a decent chance that a Governor Mannix would have resulted (although I'm not sure whether Mannix would have won the Rep. primary under IRV). Bush might have been re-elected over Clinton in '92 as well.

    Both major parties can point to examples where IRV would have helped or hurt their cause. Minor parties seem divided on the idea as well.

    Right now when a minor party mounts an effective campaign in a closely divided district, they can influence the outcome between the two major parties. This allows minor parties to influence the major parties on certain positions. (The Greens mostly do this with the D's, the Libertarians mostly with the R's).

    IRV transforms a minor party's strongest weapon (spoiling the outcome) into a pea-shooter.

    On the other hand, IRV will be very useful as a public polling device. We will learn a lot from studying the second choices of voters, and the exercise of a second vote may expose some voters to the ideas of candidates from parties they would otherwise ignore.

    I'm divided on this idea, but I'd like to allow the counties to experiment with it for county offices. (As long as the people in each county pass it with a public vote).

  • Sally (unverified)
    (Show?)

    You don't have to rank all the candidates in an IRV election. One can vote for only one, or rank any number up to "all."

    A dollar gets you a krispy kreme krumb that most opposed are devotees of one of the two parties in power.

  • tunesmith (unverified)
    (Show?)

    You know, that thing about Arrow and there never being a perfect ranked voting system is really misleading.

    First, arrow's theorem relies too much on the "independence of irrelevant alternatives criteria". When it turns out, there are cases in real life when violating that criteria is actually rational and sincere behavior for a voting population.

    Second, Condorcet Voting - when there is a Condorcet Winner - actually is perfect if you're talking about real democracy.

    If one candidate in an election would beat every other candidate head-to-head, that's the Condorcet Winner. You can derive that information pretty easily from ranked voting.

    The problem only comes when there are circular paths, like if the candidate collectively prefers A to B, B to C, and C to A. What vote theorists gloss over, though, is that cases like those are very rare in large population voting. So people can get pretty much everything they want by going for Condorcet Voting, otherwise known as "Instant Round-Robin" voting.

    At any rate, ranked voting in general has a much bigger problem than that. For most ranked voting system, it is very hard, if not impossible, to determine how much someone won by. There are times you want to get an easy feel for how much proportional support a candidate or issue got in an election, so you can track support levels over time. That gets very difficult when you aren't tracking first-place votes anymore.

  • afs (unverified)
    (Show?)

    I'd support IRV if it included a "do over" option. Voters need to have the ability to say, "You know... I'd really prefer to have a new election than have some/all of the candidates on the ballot get elected to public office.

    The Gray Davis/Ah-nohd disaster. There's an election screaming for both IRV and do-over. I would really liked to have told the powers that be that I preferred both a porn star and Gary Coleman to either major candidate, would have tolerated Davis over Ah-nold, but I'd really preferred that the whole election was such a mess they needed to start over at square one.

  • LG (unverified)
    (Show?)

    I was going to opine that this column suggests someone has way too much time on their hands. Doubly so when I saw LT weigh in. But then the irony struck when I realized that I was, in fact, reading the column. And replying for that matter.

    PS to the author - don't waste too much time with that there book lernin'. Not too much practical application. Maybe that's why most of the country can't connect with the progressive movement.

  • (Show?)

    I don't have a comment about the piece...but I'm wondering what "social choice theory" is?

    I've never heard of such a thing.

  • Christopher Nicholson (unverified)
    (Show?)

    Hey all, Thanks for commenting. I'll go in reverse order. Carla: Social Choice Theory is the study of decision-making processes for groups of people. As Arrow's theorem proves, there is no perfect selection method, so Social Choice Theory is all about figuring how groups which have disparate preferences can make a decision about something, be it who to elect, or where to go get food.

    The two other main issues that were brought up in posts were IIA (Independence from Irrelevant Alternatives), and the inability of individuals to form opinions about candidates (and thus fully complete a ranked ballot).

    In terms of individuals not being able to fully fill out their ballot, from all the examples I've ever seen, most voters simply fill out their ballot with the candidates they do care about, and leave off all the ones they don't know of or don’t care about. So obviously some people's ballots will be more complete than others. Some will rank more people. This leads me to the conclusion that we should use a system which does not unfairly weight one ballots more heavily than another, if one is more filled out than another. For example, a system which gave points to each candidate on a ballot if they were ranked above a certain number, so 5 points for first place, 4 for second, 1 for 5th, and zero after that, would not be fair to those people who don't rank their first five choices, because they'd get less total say (there is however, a modification which can be made to this system which allows every vote to be equally weighted.

    My point is, as long as we choose a system which gives equal weight to every ballot, regardless of how filled out it is, it wont matter that some individuals don't know anything about the minor party candidates.

    In terms of IIA, I'd like to see some material on what you're saying. I've never seen a good example where one should violate IIA, but then again I'm of the viewpoint that if you have a set of circular preferences, that's not because of a bad voting system, its because you have a group of people whose preferences, when combined, are circular. I think a good voting system should reflect that fact in its outcome and not try to arbitrarily make a choice where one doesn’t exist. I'm a fan of Condorcet, however there are serious problems with it not choosing a winner and with a set of circular preferences being able to change the outcome of an election. (So if 3 people get together and decide on A, and then three more people with a combined circular preference of A>B, B>C, C>A join the group, the winner can change from A to C.

    But again, my main point is that no voting system is perfect. I mean, if someone thinks Condorcet or IRV or Borda is perfect, that's fine with me. They're all good systems in that they're better than the system we have now.

    Also, to the person who said it was impossible to determine margins of victory, in IRV you count the votes for each candidate at each round. In Borda, you just count the total number of votes for each candidate. In Condorcet, you look at each candidate's slimmest victory (the place where he won against another candidate, but it was with the smallest margin). If there's a voting system, I can assure you, there's a way to calculate the margin of victory.

    So here's my idea. There needs to be a statewide initiative to allow preference voting. Not to pick a particular form, but just to allow counties and cities to use ranked ballots like the constitution makes provision for. If Portland wants to use Condorcet and Eugene wants to use IRV, that’s fine with me. The important thing is that all the people who see that ranked ballots are better, come together and say its time for a change.

    Oh, and finally, to the guy who said I shouldn't waste my time with book learning: Practical applications are what I live for. I change the system at my school to a ranked system, because its better, and I was tired of having to vote in a system where I didn't get to fully express my views and be counted. That's why this is important; it's about representing everyone, equally, and making sure their full preferences, not just their first preferences, are counted. Its about making it so our democracy is truly democratic.

  • gus (unverified)
    (Show?)

    Christopher:

    A suggestion if I might. Your final sentence just above, ("Its about making it so our democracy is truly democratic.") is lacking in accuracy. Our form of government is a Republic while our elections are supposed to be democratic. It would have been more accurate to push for making our democratic elections more truly democratic. Not a big deal. But attention to specificity may save you from having some points deducted on an exam or paper.

    I am with you on proportional voting. But I see a lot of resistance from Democrats and Republicans who have built themselves a cozy system for government financing of primary elections as well as power sharing among the majority and minority caucuses in the legislature.

    I would also like to see a line for "none of the above" (nota) for each position on the ballot. In some contested elections and a lot of uncontested elections I would opt for nota. When nota gets the highest vote total, provisions for a new election would come into play and anyone gathering a prescribed number of valid signatures in 3 or 4 weeks would gain a place on the ballot.

  • Monty Bates (unverified)
    (Show?)

    Actually our present system does allow voters to fully express themselves. We vote for the one WE WANT TO WIN! Anything else is just plain bullsh*t!

  • (Show?)

    "PanchoPDX":

    Re: Clinton potentially losing in 92 under IRV- For years after his first election, the Republicans clinged to the belief that 1) because he got under 50% his election was somehow not legit 2) Without Perot in the race, Bush would have won.

    You seem to be implying that a little too, but the numbers show otherwise. The results were Bush 38%, Clinton 43%, Perot 19%.

    For Bush to win, he would have to carry Perot voters by 68%- which according to every poll at the time was highly unlikely.

    I realize this is a minor point to the discussion, but it's still inaccurate nonetheless.

  • LT (unverified)
    (Show?)

    I agree with Charlie: For Bush to win, he would have to carry Perot voters by 68%- which according to every poll at the time was highly unlikely.

    An old family friend who was a lifelong Republican said to me one day in 1990 that she knew Bush was in trouble when so many of her Republican friends started supporting Perot.

    Do you really think her friends would have adopted IRV/ put Bush as a second choice?

    BTW, even lifelong party members don't always vote straight party. I remember how surprised I was when this same friend spoke of voting for Lonsdale in 1990 because Hatfield "had been there long enough".

  • Anne (unverified)
    (Show?)

    Pretty much every voting "theorist" has never run an election. Thus, they're theorists. Why don't you ask people who actually run elections what they think of IRV? (most don't like it, know it doesn't work in reality like it does in theory, and take issue with the unconstitutional effect of giving minor party voters more than one vote).

  • Sally (unverified)
    (Show?)

    IRV does not give anyone more than one vote. Party hack fear factor at work.

  • (Show?)

    Hooray for preference ranking!

    I like this idea of simply authorizing preference-ranked ballots across the state, and allowing individual counties or districts to implement it. That would be especially good because it would allow Oregonians to experiment with different systems, and discover what works best. Then we'd have elections officials who had actually tried using these systems, and could speak about them based on their experiences.

    Rupert wrote about liking the Portland City Council model. In the case of nonpartisan races, I can suggest a slight variation that I think would work better - STV primaries, the two winners of which face each other in the general, unless one candidate gets more than two-thirds plus one vote.

    (STV is a variation on IRV, in which people whose favorite candidate is lowest-ranked get their votes transferred to their next preference when that candidate is eliminated, and people whose favorite candidate is given a seat have those portion of their votes that were unneccesary to the candidates election transferred to their next preference. Candidates for a two-seat election, like the one I suggest above, need one third of the votes to gain a seat.)

    <h2>Of course, I'm not sure this (STV primary for a two-person general) is desirable, since you already have the preference data from the first ballot. The main effect would be to allow an underdog another chance at winning, and to extend the campaign season by another several months (thereby maintaining higher costs, etc). The preference data already being in hand, one might as well hold the final count on the basis of the ballots cast in the primary.</h2>
guest column

connect with blueoregon