Don’t douthat, man! Please give this fallacy a name.

Regular readers of this blog will know that I’m always on the lookout for new items for the lexicon. It’s been a good month on that front. In addition to the Garden of Forking Paths, I’ve encountered two entirely new (to me) fallacies.

The first of the two new fallacies has a name that’s quite a mouthful; I’ll hold off on telling you about it right now, as Eric Loken and I are currently finishing a paper on it. Once the paper’s done, I’ll post it in the usual place (or here, once it is scheduled to be published) and I’ll add it to the lexicon as well.

What I want to talk about today is a fallacy I noticed a couple days ago. I can’t think of a good name for it. And that’s where you, the readers, come in.

Please give this fallacy a name!

Here’s the story. The other day on the sister blog I reported on a pair of studies involving children and political orientation: Andrew Oswald and Nattavudh Powdthavee found that, in Great Britain, parents of girls were more likely to support left-wing parties, compared to parents of boys. And, in the other direction, Dalton Conley and Emily Rauscher found with survey data from the United States that parents of girls were more likely to support the Republican party, compared to parents of boys.

Both these studies came out a few years ago (and I blogged on them way back when), but the Conley and Rauscher paper got a new burst of attention following its recent publication in a sociology journal.

We haven’t reached the fallacy yet, but we’re getting closer.

One thing I noted in my sister blog post was an oddity in the reporting of the Conley and Rauscher paper:

There’s something oddly asymmetrical about how these results are presented, both by the authors and in the media. Consider the following headlines:

“The Effect of Daughters on Partisanship and Social Attitudes Toward Women”

“Does Having Daughters Make You More Republican?”

“Parents With Daughters Are More Likely To Be Republicans, Says New Study”

“Parents Of Daughters Lean Republican, Study Shows”

“The Daughter Theory: Does raising girls make parents conservative?”

To their credit, the study’s authors and many of the journalists make it clear the the claims are speculative (consider, for example, the question mark at the end of the New York Times headline given just above). So that’s all good.
But here’s my question: Why is it all about “the effect of daughters”? Why not “Does having sons make you support the Democrats?” It looks to me like having sons is considered the default. Okay, sure, a bit over 51 percent of babies are boys. Really, though, you can have a boy or a girl, and I think the whole discussion of these claims in the media is a bit distorted by the implicit attitude that the boy is a default. Lots of discussion about how you, as a parent, might change your views of the world if you have a girl. But not so much about how you might change your views if you have a boy. Lots of discussion of how having a girl might affect your attitudes on abortion, not so much discussion about how having a boy might affect your attitudes on issues such as gun control or war, which disproportionately affect young men. This is a real problem, when issues of girls and boys, men and women, are treated asymmetrically.

An illustrative example of this asymmetry came from New York Times columnist Ross Douthat, who recently expressed pleasure about the headline, “Study: Having daughters makes parents more likely to be Republican.” Douthat writes:

Why pleasure? Well, because previous research on this question had suggested the reverse, with parents of daughters leaning left and parents of sons rightward. And those earlier findings dovetailed neatly with liberal talking points about politics and gender: Republicans make war on women, Democrats protect them, so it’s only natural that raising girls would make parents see the wisdom of liberalism …

But the new study undercuts those talking points. Things are more complicated than you thought, liberals! You can love your daughters, want the best for them, and find yourself drawn to … conservative ideas! Especially if you’re highly educated, which is where the effect was strongest!

The fallacy here is that Douthat is thinking unidirectionally. He’s all about what happens if you have a girl. But what happens if you have a boy? Parents of boys are drawn to … liberal ideas! Especially if you’re highly educated, which is where the effect was strongest! This would seem to contradict theories of the feminization of America, the “war on men,” etc.

This fallacy is not special to Ross Douthat or to conservative columnists. Indeed, the political “gender gap” in America is typically framed as an advantage for the Democrats who get these extra votes from women. It could be just as well be framed as a male gender gap in favor of the Republican party, but you don’t usually hear it that way.

Anyway, that’s the fallacy: a comparison could go either way, but people think about it only in one direction, thus not fully understanding the implications (in this case, thinking that a connection between daughters and Republican voting is good news for conservatives, because having sons is implicitly considered as the default case).

I can’t think of a good name for this fallacy. Can you?

69 thoughts on “Don’t douthat, man! Please give this fallacy a name.

  1. I haven’t read any of the papers, so I’m just making a hypothetical here: The effects need not be symetric. Some people have both sons and daughters. If they look like the parents of daughters, and not like the parents of sons, then I think you would be justified in calling it a daughter effect. If the studies didn’t analyze the data that way — or did, and found symmetric results — then there’s more wrong here than short-sighted labeling. That said, Douthat is just doing his job: He’s on the Times to make David Brooks look smart, and he has one again rushed in to give it another try.

  2. So, I’m not sure I would classify this as a fallacy — basically, the data seem to show a number of correlations, and Douthat provides a possible causal theory behind one of them. I would imagine that he’s not writing about other correlations because of limitations of space, rather than an unconscious bias. But perhaps that is too generous to Douthat. Personally I quite enjoy his columns.

    If I were going to accuse him of a fallacy, it would be ignoring lurking variables/possible spurious correlations. On his blog, when he writes about the relationship between romantic patterns and happiness – example here http://douthat.blogs.nytimes.com/2013/12/17/sex-sociology-and-the-single-girl/ – he repeatedly implies that a more conservative relationship model leaves people happier, but doesn’t mention that people with those models are more likely to go to church, to be part of strong communities, etc.

  3. Seems there are two at work, the male standard, patriarchal reference, and the preferred positivism, selecting the frame that supports your wishes.

  4. It’s the Fisher Exact fallacy! In any 2×2 table with a significant Fisher Exact result, you can use any row or column as the cause and either of the other two rows (if you used a column for the cause) or columns (if you used a row for cause) to describe the effect.

    A more neutral term might be the Pigeonhole fallacy.

    • It like the Fisher Exact Fallacy. For some, unknown reason it also reminds me of the “Prosecutor’s Fallacy”.

      The Pigeonhole fallacy is way too bland.

  5. I think you are being unfair to Douthat. He seems to be ridiculing these same studies. He’s saying look when the studies indicated the other way you liked them because they reinforced your mental model. Now that they contradict your stereotype what will you say?

    I think that’s fair.

    • Rahul:

      I agree that Douthat is appropriately careful in interpreting the published research claims. But I still think he’s committing the One-Way Street Fallacy because he’s not realizing that the results (when taken in the other direction, considering the effect of having a boy) contradict the other side’s stereotypes.

  6. Do you really think the narrative that there is a “war on men” is as strong as the one that says there is a “war on women?” The “war on women” narrative is purported by higher ups in the democratic party, while the “war on men” is much more of a fringe idea.

    The interpretation can go both ways, but it’s more salient (read: sells more newspapers) to note that this finding is contrary to the dominant narrative that there is a war on women.

    • Phil:

      Maybe. But the point here is not to pick on Douthat. Consider those 5 headings:

      “The Effect of Daughters on Partisanship and Social Attitudes Toward Women”

      “Does Having Daughters Make You More Republican?”

      “Parents With Daughters Are More Likely To Be Republicans, Says New Study”

      “Parents Of Daughters Lean Republican, Study Shows”

      “The Daughter Theory: Does raising girls make parents conservative?”

      No headlines of the form: “Does raising boys make parents liberal?” or “The effect of boys” etc. And, if anything, such headlines would be more logical. Given the general presumption that having kids makes you more conservative, the message would be that the effect of having a boy child (compared to the default of girl) goes in the other direction.

      My point is not about the politics, it’s about the interpretation of the scientific claim. Perhaps my invocation of Douthat muddied the waters because he is an outspoken partisan.

  7. “Smurfette fallacy” on the theory that all smurfs have exactly one quality, and simply being female counts as equivalent to being male *and* hefty, brainy etc.

  8. Arbitrary Default Fallacy. Bit pedestrian, maybe, but fits. I also thought about “False Default Fallacy,” which rolls off the tongue very nicely, but might give the wrong impression (ie that the default option is in fact false, rather than it being erroneous to treat one option as the default).

  9. Douthat seemed to be taking sons as the implicit default (or describing the effect as being that of having daughters) because that was how the previous study (with opposite results) framed it. And going back through the links to Oswald’s original paper, the title indeed refers to daughters and left-wing voting.

    My understanding is that the ratio of boys vs girls who are born (NOT the ratio of those currently alive) is nearly, but not exactly, equal. In that case one of them really could be an appropriate “default”. But I’m also under the vague impression this varies from place to place, and not just because of sex-selective abortions in China & India.

    • Wonks:

      1. As noted in one of the comments above, this isn’t about Douthat, he’s just an example. It’s a fallacy that many others do too. Indeed, if it were only Douthat’s error, it wouldn’t be interesting enough to merit a general discussion.

      2. Yes, as noted in linked post, it’s a bit more than 51% boys, doesn’t vary much at all from place to place or time to time except under extreme conditions.

      • Among whites and Asian-Americans, about 105 boys are born for every 100 girls. Among African-Americans, it’s about 103 for every 100.

        Nobody seems to know why, although it strikes me as interesting and potentially significant.

        • Steve:

          Yes, I talk about this in my intro statistics classes. One general pattern is that, when things are tougher in utero, you get more girls, presumably because boys are weaker than girls (at any age, including negative ages, males are more likely than females to die). For example, older mothers are more likely to have boys, and under conditions of famine, girls become more prevalent. I always assumed that African American mothers were relatively more likely to have girls because they are more likely to be poor or to otherwise have poor conditions in utero. But I don’t really know, I was just fitting the data into that general pattern.

        • Rahul:

          Data are available. I happen to have a graph with data since 1970. During that time the proportion girls for black mothers has been pretty constant at 0.493, with the proportion girls for white mothers has very gradually increased from about 0.487 to 0.488. But a quick web search should reveal data for earlier years, I’m sure.

        • If even in utero more males die than females & yet at birth we see more males than females is the actual probability of conception even more grossly skewed to favor sperms carrying Y chromosomes?

          I wonder how this is brought about. Is there a causal mechanism / hypothesis why a Y-sperm may be more virile / motile than a X-sperm?

  10. I have no ideas for a clever name. I’ll cast a vote for Rasmus’ idea, though. (Hej, Rasmus!)

    The OP made me think of symmetry in physics, though. It must be a rich source of metaphors to turn into memorable names. For example, Galileo had that famous argument about being inside of a ship and not being able to tell from any measurement inside whether or not the ship is moving.

    This seems similar to the son/daughter problem, in the sense that we cannot tell from any estimate of the difference between parents of sons or daughters whether sons are moving default conservative attitudes or rather daughters are moving default liberal attitudes. Other data are needed.

    This reminds me of Buridan’s ass, also (http://en.wikipedia.org/wiki/Buridan's_ass). Not sure how to make that into a name for the fallacy, but Xuan’s comment above suggests maybe calling it “Buridan’s half-ass.”

  11. Trying to come up with something not said: “the blinkered horse fallacy” …. sees only what’s in front of him, plus it occurs in a race, so it’s part of a competition to get to some finish line with that process putting on the blinkers.

  12. Readers might like to know that Peter Hegarty has studied for several years the way that psychologists and non-psychologists tend to report male data as the default, both in text and in graphs, and to ‘interpret’ female data in terms of how it deviates from men. Two relevant papers are “Androcentric reporting of gender differences in APA journals: 1965-2004.” and “Graphing the order of the sexes: Constructing, recalling, interpreting, and putting the self in gender difference graphs”.

    It’s very rare that this blog discusses issues I’ve thought hard about before, but my lab actually had a long conversation about the strange phenomenon in which poll results often treat the democratic party as having an advantage in the “gender gap”–as though the real numbers were the men’s polls, and these were thrown off by those pesky females.

    As for a name, I suppose you could use “Androcentrism”, but I would have thought that “Patriarchy” would suffice!

  13. The backhand fallacy. In the abstract, forehand and backhand and equal and opposite, but in practice, it’s a completely different story.

  14. Andrew,

    however you decide to call it, you are prone to it yourself in assuming that having children is the default. Deciding for parentship, giving birth to a child and caring for it over a period of decades, is a much stronger game changer than offspring gender. (This is at least true for those regions of the world were sexual chauvinism has come to a historical minimum.)

    To my mind the fallacy has at least the following aspects:

    One aspect originates in the unduly emphasis on experimentation in social science. We learned to think of exactly one level being the control (placebo in clinical studies), whereas the other is the manipulation, the game changer. In that respect, one could call it the “confusion of natural levels with manipulation” fallacy. In the present example this is just spiced up with some chauvinism.

    Another aspect is that people people seem to readily jump to conclusions regarding the causality. Unless the study shows a strong and even effect, it is rather likely that multiple, interwoven and indirect causal chains form the observed association. This is highlighted by the apparent interaction with culture.

    How comes? Most people perceive gender as one of the most salient partitionings of the human population. Perhaps, this causes the cognitive illusion that any effect associated with gender, must be as clear-cut.

    –Martin

  15. I think you need two names, one for the general case of having a biased default or privileged perspective or whatever, and one for the specific case of males being the default.

    I never saw the Smurfs but I gather from the comments above that the male Smurfs each exemplified a characteristic (old smurf, grumpy smurf, funny smurf, etc.) whereas the female smurfs were just generic Smurfettes. Thus the “generic male smurf” was the implicit standard to which all other Smurfs of both genders are compared. If that’s the case then I understand the appeal of Smurfette Fallacy, it seems perfect for the specific case of taking males as the default for comparison.

  16. This might be called the marked pole fallacy.

    This takes some explaining.

    One of the concepts of linguistics is called marking. In pairs of adjectives, often one side is unmarked. We can ask “How good is X?” and X can be rated anywhere from +3 (good) to -3 (bad). “Good” is unmarked.

    But if we ask how bad something is, we are presuming it lies on the bad side, from not -1 (not very bad) to -3 (very bad). “Bad” is marked.

    See the Battistella quote on this page for more explanation. http://grammar.about.com/od/mo/g/Markedness.htm

    Let’s return to the context here. For many years, we have used man / mankind at the unmarked part of the pair, with woman as the marked part of the pair. Our tendency is to report the marked side. That may be shifting, but it hasn’t shifted all the way yet. (Consider “she” is definitely marked, “he/she” and “they” are unmarked, and “he” is growing more marked over time.

    Similarly, headlines report “Asian Americans overachieve”. “Americans” is unmarked. All hyphenated Americans are marked.

    Citing the marked pole is more informational, so I would guess the marked pole will tend to be used in headlines.

    • I like this quote from your link:

      “…the notion of markedness posits that the terms of polar oppositions at any level of language are not mere opposites, but rather that they show an evaluative nonequivalence that is imposed on all oppositions.”

      I think that pretty much nails the origin of the fallacy. I think that was what I was intuitively trying to get at with my more-than-half-full glass suggestion earlier. I guess in that instance “full” is unmarked and “empty” is marked, barring any additional information (e.g., if it was full of something nasty).

  17. It looks a bit like a special case of the fallacy of composition:

    People use themselves when constructing the default (the only female author involved was the junior person in joint work), then (possibly others) commit the fallacy by not considering the contingency of the baseline when going to the aggregate level.

    • Oh, sure, once we get a comment from Rasmus Bååth, it’s only natural to hear from someone named Øystein! Looks like the empty set to me, but I recognize that this just represents my ignorance of other languages.

  18. What do we know?

    Married voters are much more likely to vote Republican than unmarried voters. This effect dwarfs the celebrated “gender gap” in most Presidential elections.

    The marriage effect appears to mostly — but not totally — be marriage rather than children.

    Other explanations like gender don’t work well, either: it’s not age or education. Homeownership is the one alternative explanation that can’t be dismissed:

    http://www.vdare.com/articles/the-gop-s-other-problem-marriage-gap-huge-in-2012-but-marriage-declining

    There is some evidence that having a son rather than only daughters reduces the chances of divorce, so this would suggests that male children tend to keep people married and voting Republican, although this is a bankshot effect.

  19. Pingback: The female audience (and other sexism-related links) | Fraser Sherman's Blog

  20. The gravitational pull fallacy. Both objects are actually attracted to each other but we see just one directional movement due to the asymmetrical distribution of the masses, in the form of preconceived notions. (Pun of masses not intended)

  21. Pingback: Ticket to Baaaath « Statistical Modeling, Causal Inference, and Social Science Statistical Modeling, Causal Inference, and Social Science

  22. Pingback: Research benefits of feminism - Statistical Modeling, Causal Inference, and Social Science Statistical Modeling, Causal Inference, and Social Science

  23. Might be too cute, but “the default default fallacy”. That is to say, when choosing which option of a binary variable to treat as the default, pundits make the default choice without thinking about it.

Comments are closed.