Do Voters Ever Think About The Voting System?

Eric Black has an interesting article about a recent poll on the Minnesota gubernatorial race. The whole thing is worth a read, but here are some relevant excerpts.

The poll, taken by St. Cloud State University (SCSU) and partially sponsored by MinnPost, suggests that if the election was conducted (as it will be) by the traditional plurality-vote-wins system, Dayton leads among likely voters as follows:

Dayton: 40 percent
Emmer: 30 percent
Horner: 19 percent
Other, don’t know, refused: 11 percent

The findings are probably good news for Horner. Recent polls have pegged the Independence Party candidate to around 10-15% of the vote. The results get more interesting.

But this poll also included an explanation of Ranked Choice Voting, then asked respondents to say who their first and second choices would be if Minnesota had that system. The first choices broke down (these numbers are rounded, so they don’t add to 100 percent):

Dayton: 39
Emmer: 30
Horner: 19
Other, don’t know, refused: 13

Another co-sponsor of the poll was FairVote Minnesota, the group that has been pushing for an expansion of Ranked Choice Voting (RVC). Under RCV (also known as Instant Runoff Voting) voters are allowed to indicate their first, second, third preference on the ballot. If no candidate is ranked as the first choice on a majority of ballots, the lower-finishing candidates votes are assigned to the voter’s second (and, if necessary, third, fourth, etc.) choice until someone amasses a majority.

The surprise here, at least to me, is that Horner gets no bump when voters are offered the ranked choice option.

Under RCV, If Dayton and Emmer were the top two candidates, those who gave first preference to Horner or someone else would be assigned to their second choice. This poll corroborates what some other polls have shown, that based on second preferences, Horner appears to be taking more votes from Emmer than from Dayton. But not by a huge margin.

If we reassign the votes to the second choice of those whose first choice was Horner or other, and throw out the votes of those who didn’t give a second choice (which is what would happen under RCV), the horserace looks like this:

Dayton: 46
Emmer: 39

The article goes on to point out this poll, like others, shows that a good deal of Horner’s support is ‘soft’- that is, they are still not completely set upon their choice of candidate and may split for either Emmer or Dayton.

Interestingly, the only change when RCV is used is in Dayton’s numbers; there, Dayton loses ~1% support to minor party candidates. I would guess this means Ecology Democracy candidate Ken Pentel, who has run for the post several times as a Green, or Green Party nominee Farheen Hakim.

But to the question at hand- why does Horner’s support not go up with RCV? After all, with the endorsement of 3/5 living governors and every major newspaper in the state, the candidate certainly has some prominent backers. I will offer two potential reasons:

1) The major party candidates have done a good job of attacking Horner. Black makes the case in another article that one part of Ventura’s success was the lack of major party attention to his campaign (and with it, a lack of attacks). Today, the combination of outside money and Ventura’s example have ensured Minnesota politicos do not neglect the variable of Independence Party candidates.

2) Voters have simply become accustomed to thinking of elections in the context of the lesser of two evils. This reminds me of the Georgia gubernatorial race. In theory, the runoff mechanism in that state should allow voters to vote their conscience for whichever party they may support. However, this does not occur in practice, with John Monds consistently polling between 3-8%. Considering the various missteps of the Republican candidate in that race, one would guess the Libertarian would pull stronger support in the first round of voting. However, voters do not seem to value this nuance.

One may encounter a similar problem in Minnesota. After years of an electoral system that inherently produces vote splitting, voters now sub-consciously factor in third party status to candidates regardless of the individual circumstances of the race.

Oh, and as to support for RCV in Minnesota?

A 60 percent majority of respondents said they had never heard of Ranked Choice Voting before the poll. When they were asked (after the system had been explained to them) whether they would like to use the system, the result was:

No: 48.5.
Yes: 34.
Don’t Know: 17.5

12 thoughts on “Do Voters Ever Think About The Voting System?

  1. Be Rational

    Single member plurality voting is the best possible electoral system. It has the greatest tendency toward liberty and stability. It is the only system simple enough for the overwhelming majority of voters to comprehend.

    We do have far too many elected officials however. Most elected offices are unnecessary and should be eliminated.

  2. Peter Tharaldson

    Actually dropping IRV into the middle of a pulurality race in a poll is silly. While most concentrate on the calculus of IRV, they forget that the biggest net effect is in changing the psychology of the election.

    A bigger and much more profound methodological problem with IRV is that teh first question in a survey instrument is always who you would vote for. THat is done so that your answer on party id or ideology won’t bias your vote (there is a small group who might say I am a conservative or liberal in the ideology question and take that as a reason to be consistent on the preference poll.

    Think about it…if you get the plurality question first….you are more likely to give the same response on the IRV option later. THis was really poor methodology…very poor indeed

  3. Dale Sheldon-Hess

    Horner’s problem is that he’s in the middle. And plurality and IRV both severely disadvantage moderate candidates.

    Horton would beat Dayton in a one-on-one race (by picking up the vast majority of Emmer votes), and beat Emmer in a one-on-one race (by picking up the vast majority of Dayton voters); but when all three are running, then he comes in last.

    In a situation like this, plurality and IRV agree on the winner, but it’s likely that every other commonly suggested voting method (approval, Borda, Bucklin, Condorcet, score) would pick Horner.

    There’s no paradox here, and it’s not the “stupid voter’s” fault. IRV fails to do the “obviously” right thing because IRV, like plurality, is flawed.

  4. Darryl W. Perry

    MOST voters don’t “think” when it comes to voting- voting systems or otherwise… I think non-partisan IRV or approval voting is the best way to hold elections – that way, people are forced to vote for a candidate instead of voting for a party.

  5. Daniel Surman

    @Dale, there is no reason Horner wouldn’t have as much potential as Emmer or Dayton to contest the lead in an “all things equal” IRV scenario. Horner is suffering from a wave of attack ads that he has limited ability to answer with a limited warchest.

    But notice he still hit almost 20% in this poll. If Horner had fundraising parity, this race would likely be anybody’s game.

    Personally, I think range is the best voting system on the grounds that it forces candidates to account for the entire population rather than only mobilizing an ideological base. However, IRV still seems a stronger system than the status quo.

  6. pete healey

    Alright, so 60% weren’t familiar with the IRV system and were the only ones questioned about how they felt about it? Is it safe to asssume that the other 40% were comfortable with it and have a positive view of ranked choice voting? And if we can add 40% and one-third (34%) of the remaining 60% who expressed a positive view of it (which comes to 20% of the total sample) then we have 60% of people who like it, 30% who don’t and about 10% undecided? Does that make sense?

  7. Dale Sheldon-Hess


    IRV (like plurality) over-emphasizes the support of the top-two candidates… which helps them get more money… which helps them stay one of the top-two candidates. Sure, if Horner had as much money, he’d be doing better. But he doesn’t, and the system (plurality or IRV) pushes the situation further out of balance.

    I agree that range (AKA score) is the best alternative (although I’m currently pushing for approval, since it gets most of the advantages with very low complexity.) As a range supporter though, surely you must have seen this graphic, which shows how poorly IRV actually performs?

    In the worst case, it is NO BETTER than plurality.

  8. Jack

    @Dale. You don’t seem to make realistic assumptions about approval voting – or don’t have any real sense of how campaigns really work. Have you seen how hard candidates and their backers work to win for themselves? Having a vote count equally for two people at a time seems a recipe for mess.

    @Peter – Good point about the poll.

  9. Daniel Surman

    Good debate, in the format of a job interview. I know a few students at Macalester who were really impressed with Horner’s performance. The moderators targeted Emmer, who had to necessarily follow a delicate balancing act between conservatives and being perceived as another Tim Pawlenty and thus was sometimes vague.

    BTW Dale, I remember that chart from Poundstone’s book. It clearly shows that IRV can be perceived as just as bad as the status quo (of course it has its tradeoffs), but that it also has the potential to dramatically outperform plurality voting. So that is probably not the best proof for your case.

  10. Dale Sheldon-Hess


    And you seem not to have looked at any of the research or studies on the website I’ve been pointing out. Having a vote count equally for multiple candidates is the best thing we can do for society.

    I don’t particularly care how hard the campaigns or the candidates work; I care about society.

    Perhaps you’d be happier with range (AKA score) voting, since that gives voters the option to differentiate the strength of their separate votes. That would be fine by me; I’m promoting approval over range only because it’s a little simpler (approval is just range with a range from 0 to 1.)


    That bar isn’t a some kind of variance or margin of error; it’s the range of outcomes from varying the %honesty input.

    I’m perfectly satisfied to stand behind that image, and proud to point it out. Sure, IRV could out-perform plurality; and it probably would, since it only is as-bad as plurality when 100% of voters are tactical. And it does best when 100% of voters are honest, but that’s true for all voting systems.

    But, hold %honesty constant, and at every point, approval and score are better than IRV. And, as pointed out in that particular graph (although it was not the case in all 720 simulations) even with 100% honest voters, IRV was worse than 100% tactical approval and range voters.

    I never said (or meant to imply) that IRV was WORSE than plurality; only that, AFTER plurality, IRV is the WORST possible choice.

    Polls of support for 3rd parties versus how many votes they actually get suggest that about 75-95% of voters are tactical. This suggest that the improvements from IRV would be relatively small (and the real-world backs that result up), while the improvements from approval or score would be significant.

    (This ignores the cost component (IRV is more expensive than approval), the complexity component (IRV is more complex than approval), and the ballot-spoilage component (IRV has about ten times as many spoiled ballots as approval.) I’m focusing purely on outcomes here.)

  11. Clay Shentrup


    I don’t see how Dale makes any unrealistic assumptions about Approval Voting. Nor do I see how it’s a “recipe for a mess” to have votes of equal weight for multiple candidates.

    Approval Voting actually has several properties that make it preferable to the alternatives.

    Approval Voting Passes the Favorite Betrayal Criterion. This means that a voter is NEVER harmed for giving his full support to his favorite candidate. By contrast, Instant Runoff Voting, Borda, and all Condorcet methods, fail that. This has tremendous beneficial effects. For one, it means that money matters much less in elections, because candidates no longer have to prove they’re “electable” in order to win — thus indicators of electability (e.g. being at the front of the pack in terms of fund-raising, and gaining the nomination of one of the two major parties) have a dramatically smaller effect on election outcomes.

    Another beneficial side-effect of passing the FBC is what we call the “Pleasant Surprise Theorem”.

    Approval Voting is also far simpler for voters and election administrators. IRV elections in my home of San Francisco, for instance, typically result in about 7 times as many spoiled ballots. Voters literally mess up more often. Approval Voting experimentally LOWERS ballot spoilage.

    IRV also cannot be counted by summing precinct totals, which means it must be centrally tabulated. This raises election integrity issues. And doing Condorcet sub-totals is considerably more complicated than with our present system. Approval Voting is quite simple in this regard.

    Approval Voting can also be counted on simple “dumb totaling” Plurality voting machines, whereas systems like IRV effectively cannot be.

    And as Dale pointed out, Approval Voting performs extremely well in Bayesian regret measurements, which are the “gold standard” of voting method performance or “representative-ness”. That holds under any amount of strategic voting.

Leave a Reply

Your email address will not be published. Required fields are marked *