Skip to content

An (almost) testable assumption on dogmatism, and my guess of the answer, based on psychometric principles

Tyler Cowen hypothesizes a “dogmatism portfolio” or a “quota of dogmatism”: in his words,

If you’re very dogmatic in one area, you may be less dogmatic in others.

OK, well “may be” is pretty vague. There’s not really anything to disagree with, yet. But then Cowen continues:

There’s a lesson here. If you wish to be a more open-minded thinker, adhere to some extreme and perhaps unreasonable fandoms, the more firmly believed the better and the more obscure the area the better. This will help fulfill your dogmatism quota, yet without much skewing your more important beliefs.

He seems to be making a testable prediction here, that levels of dogmatism on two randomly chosen issues should be negatively correlated. I guess I should call this “almost testable,” as it still requires an issue-by-issue measure of dogmatism. (Is it dogmatic to believe that there was this guy called Jesus who walked on water . . . or is it dogmatic to say that Jesus didn’t walk on water and that you’re right and 2 billion Christians are wrong? And so forth.)

In case you’re wondering, my guess is that, if you do manage to reasonably define “dogmatism” on a variety of issues, and if you do manage to measure these on a bunch of people, I suspect that dogmatism on different issues will be positively, not negatively correlated. That is, I think Cowen is wrong.

Yes, I realize that Cowen’s hypothesis fits various stories of people who were disillusioned by religion and transferred their allegiance to Communism, or ex-communists who found religion, etc etc etc. Still, I’m gonna go with the psychometric principle taught to me many years ago by Don Rubin, which is that (almost) all test scores are positively correlated with each other.

Remember: Psychometrics is the most underrated science.

P.S. It’s possible Cowen is kidding. I say this because he concludes the blog with the statement, “I believe in portfolio models of dogmatism very very strongly,” which, in combination of the above, suggests that he’s just holding this view in order to feel less dogmatic elsewhere. I don’t know if Cowen is kidding, though, so I’ll take his post seriously.

P.P.S. Cowen has more here. I continue to be completely and utterly skeptical of his hypothesis.


  1. You might be interested in my colleague Gerard Saucier's work on the structure of -isms. In that paper he reviews some of Rokeach's work in the '50s on dogmatism. Rokeach tried to purge any overt ideological content from his dogmatism scale, and nevertheless found that it correlated with authoritarianism and conservatism. It's not clear whether that means there's a fundamental link, or if the purge was unsuccessful. But Jost's more recent meta-analysis showing that conservatives are less open to new experiences and less tolerant of ambiguity could be taken to support the former position.

  2. michael webster says:

    One of my philosophy professors use to hold the following position.

    1. Metaphysical realism – there is only one right answer.

    2. Epistemic relativism – but we are so limited in our abilities, we will never know the answer in 1.

  3. LemmusLemmus says:

    Let me submit the view that your two positions do not actually contradict each other.* I'll say that, yes, if you compare individuals, you'll find dogmatism in various areas positively correlated with each other. That's because individuals differ in what we might call "need for dogmatism". This view predicts that you'll find the positive correlations mentioned above if you compare a cross section of individuals; it also predicts that within individuals over time, increases in dogmatism in one area go together with decreases in other areas.

    *by which I mean yours and Cowen's as represented in this post by you. His own view in his post is actually more complicated, thresholds and all.

  4. Anonymous says:

    Why is this tagged "sociology"? From a disciplinary perspective, sociologists are uninterested in individual differences or psychological dispositions of any sort. This is social psychology, as you seem to recognize with your reference to psychometrics. And social psychology is the discipline of Saucier, Rokeach, Jost, and the others mentioned in comments as having done relevant work.

    Time to initiate a new tag?

  5. William Ockham says:

    I'd like to ask how you're using the word dogmatic here. In Cowen's post, he's using the popular definition (having strongly held opinions that are seemingly impervious to new information), but your examples seem to relate more to the technical definition of dogmatic (related to the core belief of a religion or ideology). It's easy to hold either of the beliefs you described in relation Jesus and walking on water in a completely non-dogmatic way (in both senses of the word).

  6. Steve Sailer says:

    The most obvious example of bigoted dogmatism in the intellectual world today is the hounding of heretics who express Doubts about the absolute equality in intelligence of all identity politics groups. The hammer comes down hardest on the most prestigious and therefore credible unbelievers, such as the co-discoverer of the structure of DNA or the president of Harvard.

    Now, what does dogmatism on this subject correlate with?

  7. DaveG says:

    Blue Genes?

    I agree about the stupidity of the intelligence police.
    I heard Watson talk in Basel not long before that fuss erupted and he said some things approaching those holy cows, but it was clear when you listened to the whole talk and took it in the context that what he was saying was more subtle.
    Genetics is important, it is how we got here, and seeing that knowledge we will have about our make-ups should make us more understanding of each other.

    And the real point is missed by the IP because they do not understand statistics and the difference between population means and the variability among individuals. In the case of (biological) characters like this the individual ranges are so big that even knowing the race tells you almost nothing about a random individual. So as methods of policy they are worthless.


  8. Andrew says, as an aside, "Remember: Psychometrics is the most underrated science." I was just in a very frustrating conversation about some psychometrics. And I have generally been frustrated by psychometrics not fully explaining their underlying assumptions which eventually seem to be that there is some single latent trait. So, this sentence made me wonder if it's me. Andrew: care to suggest a good psychometrics reference?

  9. Andrew Gelman says:

    Dahlia: I don't really know what are the good psychometrics textbooks. I just think that psychometricians often do good work, and they work on inherently difficult problems. Concepts such as personality, ability, motivation, and so forth are clearly real but they are also clearly defined only in the context of indirect measurement.

  10. Peter Flom says:

    Hi Dahlia

    The all time classic (but impossible to find for sale, and stolen from some libraries) is Lord and Novick: Statistical Theories of Mental Test Scores.

    Psychological Testing by Annastasi is good

    Psychometric Theory by Nunnally and Bernstein, but get the 2nd edition – from the 1970s or 80s, not the 3rd. I used the 2nd, back in grad school. Many reviewers say the third edition introduces errors and reduces clarity

    Hope that helps. There are also lots of books on specific topics, like Item Response Theory, Scaling, etc.

  11. Wild Guesser says:

    It correlates with the incorrigible urge to quasi-cryptically flog a dead horse and shoehorn it into any tenuously related subject.