Skip to content

We start by talking reproducible research, then we drift to a discussion of voter turnout

Emil Kirkegaard writes:

Regarding data sharing, you recently commented that “In future perhaps journals will require all data to be posted as a condition of publication and then this sort of thing won’t happen anymore.”

We went a step further. We require public data sharing at submission. This means that from the moment one submits, the data must be public. Two reasons for this setup. First, reviewers may need the data (+code) to review the paper. Some reviewers replicate all analyses in papers they review (i.e. me, hoping to start a trend) which frequently results in mistakes being found in the review. Second, if the data are first shared upon publication, this means that while the submission is in review, they are locked away. This results in a substantial slow-down of science because review times can be so long. A great example of this problem is GWASs which can take >1 year in review, while the manuscripts (usually without data) can be acquired thru networks if one knows the right person.

In your post, you note that you had to dig up the data from the hard drive to the student who requested it (good idea with that study, he should use lower-level administrative divisions too; alas these have lower voter turnout, the opposite of rational voter theory!). Given the fact that hard drives crash, computers get replaced, and humans are bad at backing up, this is a very error-prone method for storing for scientific materials for perpetuity. Would it not be better for you to go thru your old publications and make a project for each of them on OSF, and put all the materials there?

My reply: Yes, I agree with you on the replication thing. But I think you’re wrong regarding rational choice theory and turnout as a function of jurisdiction size; see section 3.3 of this article.

Kirkegaard responds:

I can’t say I’m an expert on RCT or turnout or that I’m interested enough in RCT for turnout to spend a lot of time understanding the math in that paper. A sort of meta-comeback.

However, I did read Section 3 and onwards. EU elections, by the way, have lower turnout than the national ones in EU, and the sub-national ones have lower turnout than the national ones as well (At least, that’s my impression, I did look up the Danish numbers, but did not do a systematic review of turnout by EU country by level). Not sure how the RCTheoist will change up the equations to back-predict this non-linear result, but I’m sure it can be done with appropriate tricks.

Above is a figure of Danish turnout results 1970-2013 [oh no! Excel graphics! — ed.]. Source: The reason the kommunal (communal, second-level divisions, n≈100) and regional (first-level divisions; n=5/14, it changed in 2007, notice no change in the turnout) are so closely tied is that they put them on the same day, so people almost always vote for both. EU, by the way, has grown tremendously in power since the 1970s but voter turnout is steady. As I recall, the reason for the spike in communal/regional turnout in early 2000s was because they put the national election on the same day, so people voted for both while they were there anyway.

Regarding voter intentions. It’s easy to find out why they vote. I have been talking to them about this for years, and they never ever ever cite these fancy decisions models. Of course, normal people don’t really understand this stuff. Instead, they say stuff like “if everybody thought like you, democracy wouldn’t work” (a failure to apply game theory) or “it’s a democratic duty” (not in the legal sense, and dubiously in the moral either). In my unscientific estimate of commoners, non-science regular people, I’d say about 90% of reasons given for why one should vote is one or both of these two.

This discussion reminds me of this one about RCT for voter ignorance. My agreement lies with Friedman.

Just in response to those last two paragraphs: I think these fancy decision models can give us insight into behavior, even if this is not the way people understand their voting decisions. Different explanations for voting are complementary, not competing. See section 5 of our paper for more on this point.


  1. Mitchell says:

    …yeah, the topic here certainly drifts

    drifting further on voter turnout, it seems easiest to watch what voters do, not what they say. Voter turnout is always low in U.S., decreasing sharply with non-national elections. All modern U.S. Presidents took office without the votes of about two-thirds of the general American electorate. Turnout in last year’s Democrat & Republican Primaries was under 20%. American officials assuming elected government positions with a true majority vote of their electorate is extremely rare. What does Game-Theory say about that?

    • Andrew says:


      Rationality is only part of the story of voting. For example, lots of people vote in New York State, even people such as me who live in districts where none of the elections are competitive. The point of our paper is not that all or even most voters are motivated by rational calculation, but rather that it can be rational to vote, and thinking about the rationality of voting helps us understand some aspects of voting decisions.

  2. Daniel Weissman says:

    You know that when Kirkegaard talks about “public data sharing”, he’s referring to things like making a public, non-anonymized database of OkCupid user info, right?

  3. David Marcus says:

    I think a better explanation of why it is rational to vote is to use Douglas Hofstadter’s resolution of the Prisoner’s Dilemma: If you believe that there are other rational voters, then Douglas Hofstadter’s resolution applies to voting. So, you don’t have just one vote, but rather all you rational voters are voting together. This is also very close to the reasons that people give for voting. See

  4. Eoin says:

    Is this the same Emil Kirkegard who runs an “open access journal” that only publishes actually publishes weird 21st-century eugenics stuff, written by the editor himself? (
    He may not be the most reputable figure when it comes to best practices in science.

Leave a Reply