Skip to content

Ballot order effects in the news; I’m skeptical of the claimed 5% effect.

Palko points us to this announcement by Marc Elias:

BREAKING: In major court victory ahead of 2020, Florida federal court throws out state’s ballot order law that lists candidates of the governor’s party first on every ballot for every office. Finds that it gave GOP candidates a 5% advantage. @AndrewGillum lost in 2018 by .4%

Is anyone really saying that being first on the ballot is worth 5% in a high-profile general election for governor?

OK, following the links, I see this post by Joanne Miller linking to some pages from the court decision. Here’s a key passage:

I’m skeptical that ballot order would swing the vote margin in the Florida governor’s race by 5 percentage points.

I discussed the general topic a couple years ago, in the context of the presidential election:

Could ballot order have been enough to cause a 1.2% swing? Maybe so, maybe not. The research is mixed. Analyzing data from California elections where a rotation of candidate orders was used across assembly districts, Jon Krosnick, Joanne Miller, and Michael Tichy (2004) found large effects including in the 2000 presidential race. But in a different analysis of California elections, Daniel Ho and Kosuke Imai (2008) write that “in general elections, ballot order significantly impacts only minor party candidates, with no detectable effects on major party candidates.” Ho and Imai also point out that the analysis of Krosnick, Miller, and Tichy is purely observational. That said, we can learn a lot from observational data. Krosnick et al. analyzed data from the 80 assembly districts but it doesn’t look like they controlled for previous election results in those districts, which would be the obvious thing to do in such an analysis. Amy King and Andrew Leigh (2009) analyze Australian elections and find that “being placed first on the ballot increases a candidate’s vote share by about 1 percentage point.” Marc Meredith and Yuval Salant (2013) find effects of 4-5 percentage points, but this is for city council and school board elections so not so relevant for the presidential race. A Google Scholar search found lots and lots of papers on ballot-order effects but mostly on local elections or primary elections, where we’d expect such effects to be larger. This 1990 paper by R. Darcy and Ian McAllister cites research back to the early 1900s! . . .

Based on the literature I’ve seen, a 1% swing seems to be on the border of what might be a plausible ballot-order effect for the general election for president, maybe a bit on the high end given our current level of political polarization.

Given that I thought that a 1% effect was on the border of plausibility, you won’t be surprised that I think that 5% is way overstating it, at least for a major election. Sure, governor is less major than president. But, again, in this era of polarization I doubt there are so many more swing voters available for that race either.

0.4%, though? Sure, that I believe.

So, yes, I believe that ballot order was enough to swing the 2018 Florida governor’s election. And, more generally, I favor rotated or randomized ballot orders to eliminate the ballot order effects that are there. This is a bias that can be easily and inexpensively removed; it seems like a no-brainer to fix it. Whether this is a matter for the courts, I don’t know.


  1. Michael Nelson says:

    The Supreme Court has ruled that political gerrymandering is okay, even when it shifts representation by a whole lot more than 5% (see North Carolina). Given that, and Roberts’s infamous skepticism of social science statistics, I find it hard to believe the Court would uphold this verdict against Florida. Andrew, maybe your next book should make the case for a Constitutional amendment guaranteeing the right to vote! Most people don’t even realize it’s not already in there…

    • dl says:

      The Court has long been clear that there is a fundamental constitutional right to vote. So that’s not the problem. The problem is what the Court thinks that right entails.

      • Michael Nelson says:

        Alas, while the Court has clearly endorsed the principle of “one person one vote,” it has not backed “each person one vote.” If it did, the justices would be making it up–the constitution only says what criteria can’t be used to exclude people from voting, and then only via amendments. If the Court recognized a fundamental right to vote, states couldn’t prohibit voting by felons, or require people to register a month in advance or to present approved IDs when voting, or refuse to count absentee ballots of anyone (including active military) whose absentee ballot arrives after an arbitrary time on election day. Shockingly, I don’t think the Court has ever ruled against a single voting restriction other than those motivated by race, gender or religion. Even poll taxes were only unconstitutional because they were motivated by race–it’s as constitutional now as it was in 1776 to disenfranchise the poor.

  2. Anonymous says:

    Point: Australia has compulsary voting and STV, which would probably increase the number of donkey votes quite a bit. (I suspect the typical case of this is someone who is frustrated with the major parties and lists minor parties/independents first in order, then the majors). I’d expect ballot order in a high-profile US election to have much less of an effect than 1%.

  3. jim says:

    From a political / democracy point of view, a 0.5% swing due to ballot order is not a serious concern. It’s hard for me to believe that democracy has been suppressed by ballot order. Anyone who really wants to vote for someone can easily find the right name on the ballot.

    The big question is not so much that ballot order matters. The big question is why does it matter? I’d be willing to bet that it matters more in local elections because most people don’t know anything about the candidates so they vote the first one. More than likely they’re picking up the ballot to vote for presidents, senators and governors, but are happy to fill out the rest of the ticket while they’re at it.

    What might be really interesting from a democracy standpoint is, rather than voting for one candidate or other, give each voter the opportunity rate the desirability of each candidate on 1-5 scale.

    That would be a really good exit poll strategy: get voters opinions on *both* candidates, rather than just which one they voted for. Would all the Rep / Dem candidates be at opposite extremes, or just a shade off the middle?

    I’m sure that’s been done….

  4. Steve says:

    I think that this discussion like many discussions on gerrymandering or other attempts by parties to suppress the vote miss the real point. The party in power do a lot of things to gain an advantage not just one. We should think of all of these things cumulatively. We should allow any kind of unfair tampering. The fact that the courts treat each issue separately is the problem. I suggest that you think about efforts to manipulate the vote the same way you treat efforts to manipulate research which so much of this blog is devoted to. There are so many degrees of freedom for those in power to manipulate the vote, that we should take away where we can any path the manipulate it.

  5. jim says:

    “It seems like a no-brainer to fix it. “

    It’s a serious mistake to open the door to using statistics to compensate for purported voter failings.

    Mr. Some at NPR would have a field day with that:

    “Some say putting women at the top of the ballot would compensate for bias against women”
    “Some say ballots are biased against (enter progressive constituency group) candidates because they’re not at the top of the ballot”

    Mr. Some will demand larger text for Group X, bold text for Group Y, different ballots for men and women, different ballots for election districts based on income, shoe size, frequency of each type of sexual orientation – it will never stop until Mr. Some gets what he wants.

    • jim says:

      I mean if you think people abuse statistics now just in the hope of getting a stupid paper published, wait until there’s a precedent for using it to manipulate voters. that *will* be the end of democracy.

  6. Thanatos Savehn says:

    I finally got around to reading the Court’s Order. Want to know how the judge concluded that ballot order has a big effect? Here you go: “For example, clustering by governor cycle reduced the p-value to .14 for Republican candidates and .16 for Democratic candidates, reflecting an 84% and 86% chance, respectively, that the increased vote share was a result of candidate name order effects.” See? There’s an 86% chance you’re wrong about the effect size Dr. Gelman. The law says so. Yes, the p-values calculated were pretty big but when the standard is “more likely than not” p less than .5 is all you really need. Legal statistics are lit!

    Also, you may enjoy this application of “statistically significant”: “Florida may not continue to order candidates’ names on its ballots in a way that systematically awards the statistically significant advantage conferred by candidate name order effects to candidates of one political party on the basis of their partisan affiliation.” Statistical significance wins elections! Just in case you thought the court was kidding there’s this: “By systematically awarding a statistically significant advantage to the candidates of the party in power, Florida’s ballot order scheme takes a side in partisan elections.” Possible headline: “2020: Which party will wield statistical significance?”

Leave a Reply