the lacour and green retraction

News broke recently of very serious concerns about the data in a high-profile political science study. Not to put too fine of a point on it, it now appears that a UCLA graduate student and rising star in political science, Michael LaCour, fabricated data nearly out of whole cloth. These data led to a surprising, widely-cited finding about the ability of relatively minor sympathetic contact to change attitudes toward LGBT people over the medium term. The original article is here, a very careful forensic investigation that revealed the likely fabrication is here, and Retraction Watch has a timeline and many relevant links here. This post is intended in part as an open thread for conversation about the article, its lessons, and the retraction. A few thoughts to begin:

  • Kieran points out that Andrew Gelman noted in December the size of the effect and expressed surprise at it. Gelman proposed an apt theory for explaining that size (I like the “jumping on a moving train” metaphor).
  • In retrospect, people are suggesting that greater a priori suspicion is in order. LaCour’s coauthor says he is “deeply embarrassed” (and who wouldn’t be?), and says he should have looked at the original data. On Twitter:Screenshot from 2015-05-20 10:52:35I have to say, I don’t think it’s reasonable to have to guard against blatant falsification (assuming that’s what this turns out to be).
  • Psychological social psychology and political psychology, as fields, share a fetish for big effects of isolated causes with cute narratives (think about Larry Sanna’s infamous escalator-and-altruism study). This is a particularly egregious version of publication bias, because in general the social world doesn’t work that way. Most of the time, large aggregate changes (such as the secular increase in support for LGBT rights, which still needs explanation) are not the result of specific, isolated causes, but those are precisely the things that lead to advancement in these fields.
  • I find myself feeling bad for LaCour (whom I don’t know at all, by the way). His heretofore very promising career looks a lot less promising now (as it should). I wonder what led to this kind of misconduct. Kieran, again (jeez, he’s insightful):

    This research was part of his dissertation. Which means that the dissertation, other published work, and future research will all be subject to intense scrutiny. Again, not that it’s unwarranted, but poor guy.

  • All this presumes that the situation is as it currently seems. LaCour is implying he thinks it isn’t:

Screenshot from 2015-05-20 11:10:01 Screenshot from 2015-05-20 11:09:29 Screenshot from 2015-05-20 11:08:56

Author: andrewperrin

University of North Carolina, Chapel Hill

19 thoughts on “the lacour and green retraction”

  1. I have to say, I don’t think it’s reasonable to have to guard against blatant falsification.

    Wait, why not? Would you accept whatever reason you provide for an absence of safeguards against fraud in other domains of life, like lawyers charging billable hours they never worked or doctors billing Medicare for tests that were never ordered or journalists for top magazines making up people they interview?

    Liked by 1 person

  2. Andrew,

    I liked most of your post but was shocked by your sympathy for LaCour. We are not talking about the typical thing of wishful thinking driven p-hacking (eg the Himmicane study), small n work that doesn’t replicate (basically all of social psych), or an unnoticed missing data code (eg Jasso 1985). Rather we are talking about somebody who completely fabricated the data, whole cloth. This is basically the worst thing anyone can do in science.

    Liked by 1 person

    1. I’m with you, Gabriel – it’s not that I actually sympathize with his actions, and of course (assuming all turns out to be as it appears) scientific misconduct doesn’t get much worse than this. And it matters what the motivations were. I just meant to recognize, on a human level, that he’s going to have a crappy time from here on out.

      Liked by 1 person

    2. Do you really view “basically all of social psych” as small-N research that doesn’t replicate? High profile failures to replicate are worth the great deal of concern and attention they are getting now from scholars within and outside social psychology, but I’m not sure how you concluded that all the findings in a field don’t replicate from those cases.
      I can give you a long list of findings you can go replicate yourself very easily: the effect of status on influence, the effect of shared identity on cooperation, basic semantic priming effects, spreading of alternatives, the conjunction fallacy, the Stroop effect, the confirmation bias, etc., etc.

      Liked by 1 person

      1. I mostly had in mind a bunch of priming studies but I acknowledge that availability heuristic is leading to gross exaggeration and I should slander an entire field.

        Like

      2. @gabrielrossman: As a social psychologist, I am eager to claim this as a low-N replication of the availability heuristic.

        But seriously…one positive effect of the conspicuous failed replications in social psychology has been an increase in the N’s that are normatively expected for experiments within the field. Though I think this trend owes just as much to the new availability of cheap online study participants. Fodder for another thread…

        Liked by 1 person

  3. I’m interested in Princeton’s response (which I haven’t seen, if they have issued one). In a just world, if you can lose a job-in-waiting over impolitic tweets – Salaita, jury’s still out on Grundy – you certainly should lose a job-in-waiting over research fraud of this magnitude and apparent intent. But, unless the alleged fraudster was stupid enough to leave the R code that generated the fake data lying around, all the evidence is just highly highly suggestive. I’m sure there are internal due process and contract issues, too, that could drag this out for at least a year.

    Didn’t Marc Hauser keep his Harvard job, after issuing suitably abject apologies?

    Like

      1. Jeremy: You have greater faith in university administration than I do. I hope you’re right.

        Kieran: Thx.

        Like

  4. There’s nudging a p-value and there’s fabricating data. And then there’s faking an entire, large, funded study – with fake grants! Forget it.

    The question about trusting collaborators (students or not) requires an institutional response. I wouldn’t have the expertise to detect this fraud even if I had the raw data. So we will have to rely on expert replication and analysis to check for fraud in import studies.

    Like

  5. Also, score one for data collected by publicly-funded, accountable third parties instead of privately by researchers for their own use. Which is why this case will surely help make the case to restore NSF funding to its proper level.

    Liked by 4 people

  6. Until we get bodycams on social scientists–at least when they’re on the clock conducting research and writing up results–we’re stuck with maintaining strict norms about collection, processing, and reporting data. You can’t depend upon everyone adhering to these norms all the time, but there are all kinds of checks built into our academic communities, including dissertation committees, peer review, and replication. The odd irony here is that had Lacour/Green been less interesting, it’s unlikely someone would have wanted to build on it and find the discrepancies.

    Like

    1. I know a lot of people thought it was interesting but when I heard about it my thought was “we know about the contact hypothesis, and we know it works for LBT acceptance, why is this news?” It probably helped that I was unaware that the effect was absurdly large given the minor intervention, but at least in terms of direction it wasn’t remotely surprising.

      Like

  7. P.S. On the career consequences: there is no shortage of hard-working underemployed social scientists who have not had to retract articles or been charged with fabricating data. Princeton will be fine.

    Like

Leave a Reply

Please log in using one of these methods to post your comment:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s