Math professor: top two primary is flawed

By Dr. Joseph Kolacinski, an assistant professor in Elmira College’s Department of Mathematics at pressconnects.com:

Recently, California passed Proposition 14, replacing the state’s old primary election system with the “top-two” or “jungle” primary. It works like this: There is a single, open primary where every voter chooses among all the candidates. The top two vote-getters then compete with each other in the general election.

This seems reasonable, and a “jungle primary” certainly sounds exciting. But in judging an election system, you must ask how well the system will produce a true societal choice. By that measure, the jungle primary has serious drawbacks.

The article concludes,

France’s 2002 presidential election used a system similar to the jungle primary. Because there was a large number of left-wing candidates in the initial round of voting, the unpopular incumbent, Jacques Chirac, and the extremely right-wing Jean-Marie Le Pen led in the polls. Each was supported by fewer than one-fifth of the voters, but that was enough to make them the only two candidates in the final round. There was widespread dissatisfaction with these choices, and slogans such as “vote with a clothespin on your nose” appeared.

Every election system has drawbacks. Because the jungle primary system can prevent large portions of voters from having an acceptable choice in a general election and can determine a winner by something other than voters’ decisions, it seems inferior to the system it replaces and to a traditional open primary. Hopefully, California will discard the top-two primary system and consider different and better options.

Read the whole thing

16 thoughts on “Math professor: top two primary is flawed

  1. Bryan

    This is an example of what I said in an earlier comment.

    Could there be Intentional multiple filings, which could spread the vote thin, allowing a minority party to be the only party on the General ballot.

  2. Michael H. Wilson

    It will be interesting to see the outcome of the race for U.S. Senate in Washington State. There are 15 candidates running and only two will make it to the top in November.

    In one state legislature race there are seven candidates. In a U.S. Congress race there are six candidates for the third district.

    What the spread is like will be closely watched.

  3. Melty

    The artile mentions a single voting criterion called “independence of irrelivant alternatives”, or in layman’s talk “spoiler effect”. Range voting is immune to that problem, Danny, but instant runoff is not. For this and lots of other reasons, instant runoff is definitely not a step up.

  4. Michael H. Wilson

    here’s a interesting combination:

    Congressional District 4
    U.S. Representative

    Leland Yialelis
    (Prefers Independent No Party)
    Shane Fast
    (Prefers Republican Party)
    Rex A. Brocki
    (Prefers Tea Party)
    Mary Ruth Edwards
    (Prefers Constitution Party)
    Doc Hastings
    (Prefers Republican Party)
    Jay Clough
    (Prefers Democratic Party)

    With 2 Republicans, an Independent, Constitution Party and Tea Party candidates. That should make for fun analysis. It is presently held by a Republican, Doc Hastings, who is not really a doctor.

  5. paulie Post author

    Could there be Intentional multiple filings, which could spread the vote thin, allowing a minority party to be the only party on the General ballot.

    It doesn’t really work that way in practice. In California’s 2003 recall/special election, there were 135 candidates, and the biggest parties still managed to find their way to the top.

  6. Bryan

    True, however, the majority of those candidates couldn’t be considered “legitimate”, even by fringe groups.

    What happens when EVERY body is listed/files under the “duopoly”? If a democrat in a republican district wants to win…

    Insure there is a “constitution” party candidate, and a “libertarian” candidate, along with at least one other “legitimate” republican…running against the republican that is the “parties” choice…all running as republican.

    If two more or less equal democrats split the dem vote, along with a splintered “right” vote, the odds are pretty good that either dems finish 1/2 or one of the right fringe is able to come out into the general.

    I guess my overall opinion is that if you live in a state with “top two” voting, the “third parties” will still be able to play a role in primaries…and when the “duopoly” gets shafted a time or two, there will be “mass opposition” to this type of elections.

  7. Jeff Vanke

    Range voting runs the risk of those local council elections in the “pick 3″ model. A voter might have three clear first choices out of a field of seven, say. But that voter has a clear #1 choice, who may or may not be polling third place. So the voter risks pushing #1 into fourth by voting for their 2nd & 3rd choices, too, so they vote for just #1.

    The same dynamic would play out in range voting — you possibly hurt your #1 choice’s chances if you give a nearly as high ranking to one or more other candidates. That is a serious practical problem with range voting.

  8. Clay Shentrup

    Jeff,

    What you call a “serious practical problem” with score voting (aka range voting) is actually a benefit. If a voting method ignores an increase in support for Y, as long as Y is less preferred than X, then it necessarily discards an enormous amount of information. This is explained for IRV here:
    http://scorevoting.net/IgnoreExec.html

    The tactic with score voting is (put a little simplistically) to give a maximum score to your favorite of the two frontrunners, as well as to all the candidates you like better. So if you preferred Nader back in the 2000 election, but voted for Gore, you would want to give Gore and Nader a “10” with score voting.

    The fact that people often do not vote sincerely with plurality voting refutes the (quite absurd) suggestion by many opponents of score voting, that it would degrade into (sincere) plurality voting, as voters would just give a max score to their favorite candidate.

    No, they absolutely would not.

  9. Erik G.

    Clay @9,

    Wouldn’t score voting be ruled unconstitutional, however, under ‘one man, one vote’?

  10. Clay Shentrup

    Erik,

    The phrase “one man, one vote” has to do with universal suffrage. It would be better put “one person, one ballot”.
    http://en.wikipedia.org/wiki/One_man,_one_vote

    Actually, score voting is more compliant with the principle of fairness, since it gives every single voter equal consideration in an A-vs-B contest.

    By contrast in plurality, if you vote for Nader, your preference for e.g. Gore over Bush is totally ignored, because you only get one vote.

    And in IRV, an X>Y>Z voter can have his X>Z preference counted, while IRV can IGNORE a Y>Z>X voter’s Z>X preference. This is the point about “ignoring ballot data” that I made above.

    For instance, in the last mayoral election in Burlington, Vermont, a majority of voters preferred Democrat to Progressive, but the Progressive still won. There was a bloc of Republican voter who preferred the Democrat over the Progressive, and their D>P vote was totally ignored. IRV never counted it, because it passes the “later-no-harm” criterion, which is a really big flaw with IRV.

  11. Clay Shentrup

    I should amend that last sentence. I do not know for a certainty that passing the LNH criterion forces a voting method not to consider relative preferences that come later. That is, LNH implies that an X vote or an X>Y vote or an X>Z>Y vote should not be able to make a difference as to whether X wins or loses — but perhaps they can still change whether Z or Y wins or loses. My point was that the particular property of IRV that causes it to satisfy LNH (that it only considers later preferences after a candidate has been eliminated) also forces it to ignore lots of ballot data.

  12. Melty

    Erik, I gather that in some bad old days in some states a man got another vote for every three slaves (or thereabouts) he owned.

    I’ll one-up the problem you describe, Jeff. What if you vote higher for somebody you don’t like and vote lower for somebody you like? That should diminsh the chances for your favorite and enhance the chances for the one you deride, right? Yes, it will, if your Range voting, but not necessarily under IRV. There are realistic scenarios in which the best strategy in an Instant Runoff vote is to rank your favorite second in order to eliminate a crowd-pleasing dark horse first. One might arrive at a proper strategy of this sort from pre-election poll inf, and conspire with the like-minded for desired results. Of course, I don’t know what kind of voting methods you support, Jeff, but that sort of strategizing when rating under Range or Approval won’t work. See Clay for more details. Clay has a deeper understanding of voteology than I.

  13. Erik G.

    Clay,

    I disagree on a number of levels. First of all, study any court case involving IRV, and you’ll see that “one man, one vote” was often the centerpiece of the case. IRV was ruled constitutional, however, since it doesn’t give one multiple votes, it merely transfers their vote. This isn’t to say I back IRV, it’s merely to say it’s constitutional. They constitutionality of score voting, however, is doubtful.

    Now, in a score voting system, let’s use the commonly accepted idea of ‘voting for a capital of Tennessee’ (look at almost any wikipedia page on voting methods and this is the example given). Cities = Memphis (M), Nashville (N), Chattanooga (C), and Knoxville (K).

    If you live near Chattanooga, you may choose to score C 10, K 8, N 5, M 0. Say I live in Memphis, however, and score C 0, K 0, N 5, M 10. Whether you accept it or not, we both have multiple votes here for a singular decision, and your vote has more weight than mine does.

    Yes, you could attempt to argue that this occurs in IRV when voters don’t list their additional preferences, however, that’s not necessarily the same thing. In an IRV system where a voter doesn’t list their second preference, it simply means their vote didn’t transfer to a runoff round, i.e., they wouldn’t have cared to show up for a runoff. Score voting, however, is not involving any transfer of votes, and actually could give some voters more pull over others. This is particularly apt with strategic voting, as ‘strategic voters’ wind up with more pull than honest ones.

    Now, as for your argument in Burlington, are you making a score voting argument or a Condorcet one? Because it sure sounds like a Condorcet one. The fact is, with score voting, for all you know, the D’s & P’s in the race could have voted strategically enough to yield the same result as what actually happened. You’re going off of list-based criterion and trying to apply it to score voting, which simply isn’t intellectually honest. Now, were you advocating the Condorcet method instead of score voting, it’s true – Burlington would have wound up with a Democratic mayor.

    Now, Condorcet obviously isn’t perfect either, as it has numerous tactical voting problems itself, not to mention the possibility of circular ambiguities, but it’s likely more constitutional than score voting, as all votes are cast equally.

    [For the record, I don't have a huge preference on which system is implemented with reform, I merely want to see the end of FPP/plurality voting.]

  14. Melty

    To my notion, one vote for each candidate, rather than each office, makes perfect sense. Although, some judge might see it differently. Alternative voting methods do have a history of getting deemed unconstitutional.
    I don’t see this one person’s vote has more weight than the other’s stuff in the way you do. It could be done in reverse too, such that the candidate who gets the fewest votes wins. It’s entirely voter’s choice to bullet vote or anybody-but-him vote or something inbetween. If a court deems it a problem, one could rate it like maybe -n to n, with n as your number of candidates minus one, and your vote has to add up to zero, but that would not be desirable what with the voter confusion.

  15. Pingback: Mathematics Professor Refutes Top Two Primary Claims | Rise of the Center

Leave a Reply