Jordan Peterson and the Dragon That is Scientific Explanation
I would like to be precise: this short essay is is about minutes 12:00 thru 16:25 of The Debate (almost 7 million views on the original video, at the time of writing) and it is written from the perspective of someone training to join Peterson’s profession. The point of debate in this part of the video is differences in job placement and job pay between men and women in the private sector; my main object here is to show that the claims Peterson calmly forwards as psychological or statistical fact in this part of the debate are flawed and disappointments to the science. Those claims are:
- That the fact that Scandinavia — a region that’s really progressive — still has sex differences throughout sectors of its economy clearly illustrates that there are irreducible differences between the sexes
- That these irreducible differences — and specifically irreducible differences in personality between the sexes— account for the lion’s share of the disparities in job placement and pay that Newman talks about in this section
So let’s start with number 1.
The way Peterson frames it, it’s like we have the benefit of a giant social experiment in the form of the cold socialist North; we took away all of the legal differences between the sexes and their treatment, and there are still noticeable differences between them, which allows us to conclude that these remaining differences are natural and ineradicable. QED — What hubris would have to possess us before we try to pervert the homeostasis of the world any further.
The really remarkable thing about Peterson is that, despite his reputation as an indefatigable indicter of postmodernism and its evils, he regularly fails to address even caricature versions of postmodern arguments and analyses. One of the most well-trod and well-known being, of course, that the kinds of sex disparities Peterson points to in Scandinavia — and anywhere else — are the result of acculturation and social conditioning more than biology. This is the obvious, unrebutted rebuttal to Peterson’s argument: that even if the laws of Scandinavian countries promote gender equality, the civil societies of these countries — their actual social fabric, related to but distinct from their laws — still encourages specific differences between the sexes. Sweden, for instance, only enacted women’s suffrage at the end of the First World War, in 1918; there are people alive today who were born before this change was enacted. There are people alive today who were born before this change was enacted. On the glacially slow timescale for fundamental social change, then, it has been next to no time at all since women gained the right to self-govern, let alone be equals in the workplace. The idea that Scandinavia has somehow established a gender utopia, free from all influences aside from those that are natural or essential, is — I’m sorry for the strong language — silly and wrong.
But that doesn’t mean these employment disparities still aren’t caused in part by systematic differences essential to the sex —
Yes, clever boy, you’re right — what I’ve said doesn’t preclude the possibility that these disparities are a partial reflection of systematic differences essential to the sexes themselves, yes. But then that wasn’t the proposition I was denying: I was denying the proposition that, aside from these essential differences, there were no other explanations for Scandinavia’s persisting gender disparities. On the contrary there is a cornucopia of explanations for these social phenomenon — because they’re social phenomenon — because these are events occurring at the levels of countries and societies. Implying that they are the product of any one causal influence, be it irreducible differences or institutional biases or cultural norms — or whatever — is wrong.
Which brings us to number 2.
Peterson touts the importance of multivariate analysis around 5:40 in the video, and derides social scientists who only look for one explanation for the social phenomenon they’re interested in. He does this, granted, in the same breath that he fails to cite the source for his claim that studies have already shown the pay gap doesn’t exist — which isn’t a very academic thing to do — but regardless: Peterson says here, loud and clear, that we shouldn’t narrow our explanations of things to one cause.
And yet Peterson’s point throughout the video is just that men and women are essentially different, and these differences are what produce pretty much all of the disparities Newman is talking about — not the patriarchy or social influences. We can imagine a lot of salient differences between men and women when it comes to employment, but the ones Peterson focuses on are differences in temperament, or personality. This is perhaps best illustrated by the first exchange in this section of the interview, where Peterson and Newman talk about the overwhelming number of men in CEO positions at Britain’s largest companies.
Peterson begins by asking Newman why someone would want to run a Fortune 100 company in the first place. This is a part of what seems to be his usual routine on the issue of executive positions: that they demand their occupants devote all of themselves — all of themselves — to their faithful execution, leaving only auxiliary room for relationships, friends, children, personal pursuits and hobbies. As Peterson states at greater length in the video linked to in this paragraph, women who are in a position to eventually get these CEO positions make the sensible decision to forgo all of this around the time they reach 30, and to instead pursue less hollowing, albeit less compensated jobs. The men who do end up taking these positions then belong to a very specific, very well-delineated subspecies of their sex which is highly intelligent, highly conscientious (from the Big Five Model of personality), and downright obsessive about their professional achievement — as they would have to be to seek out something as foolhardy as an executive position.
So this is one of the reasons there are so few Fortune 100 companies in the UK that are run by women: they’re too canny to fall for the 11 million dollar a year trap. It’s one of the ways their temperament is different, on the whole, from their male counterparts. Their male counterparts — or to be specific: just a special taxon of their male counterparts, these Hank Rearden-type-men Peterson talks about — they can’t help themselves, they’re just too high on FFM facets like achievement-striving and competence to stay put, and so they battle their way to the tops of their organizations, where they can then sit as blameless as triumphant lobsters. It’s just the way of the world.
Now, I’m of the opinion that Cathy Newman overextended herself constantly in the debate, and tried to state Peterson’s side for him in ways that were consistently at odds with what he wanted to say, and that this doubtlessly limited his ability to be as expansive and detailed as I’m sure he wanted to. At the same time, as the video I linked to already shows, and the only other video I could find of him talking about executive positions shows, even when Peterson has the time to talk at length he reduces the issues Newman brings up to differences in temperament between men and women. And before I can work to rebut that claim it’s necessary to understand the logic of the kind of correlational psychology that Peterson studies and I hope to study one day as well.
The object of this psychology’s study does not have the watchmaker’s precision that the object of physics’ does, and accordingly correlational psychologists don’t talk about which variable causes another, but about what proportion of the variance of one thing that another thing explains, or predicts. Take an everyday behavior, like cigarette smoking. We might suspect that one of the influences on smoking is the optimism bias, for instance, and after some research find our suspicions seemingly confirmed when the correlation between the two turns out to be a respectable .30. (This is just an example.) The coefficient of determination for the optimism bias and cigarette smoking in this sample, then, is .09 (.3⁰²). So we say that about 9% of the variance in smoking behavior in our study population is accounted for by the optimism bias. The other ~90% of the variance in smoking behavior must be accounted for (can be predicted by) other things like genetic markers, income level, attitudes, etc. The overall idea of all of this then is that if you know the values of all of the important factors that predict smoking behavior for some population, you can plug them into a multiple regression equation and have a really good chance at predicting (explaining) the smoking behavior of that population.
Okay.
So the claim I’ve been building to — my response to point 2 as outlined at the start of the essay— is that even if there are the chasmic differences between the personalities of women and men that Peterson talks about, it’s doubtful that they would be enough to explain away the significant chunk of the variance in hiring and pay disparities that Peterson is implying they do. Because personality psychology takes place in a “universe of low correlations,” to quote the psychometrist Jane Loevinger. And to quote Jane Loevinger quoting other famous psychometrist J.P. Guilford:
We must face the fact, unpleasant though it may be, that in human behavior, complex as it is, low intercorrelations of utilizable variables is the rule and not the exception. Highly valid predictions must ordinarily be based on multiple indicators. Although each may add a trifle to the total variance of the thing predicted, by summation aggregate prediction can mean a very substantial degree of correlation.
So then just how trifling is personality’s contribution to job placement, or the pay gap? First, it would be good to speak to the issue of systematic differences between men and women’s personalities in the first place. And on that front, yes, the most cited study on the topic found differences between men and women’s FFM scores, but they’re a far way away from chasmic. But okay, okay: it’s established for our discussion here that there are consistent differences between the FFM scores of men and women. So we’re back at the original question: how much of a contribution do these personality differences make to placement and pay gaps?
On the placement side of things, we can see Loevinger’s universe of low correlations evinced in one of the most cited meta-analyses of research on personality traits and executive leadership:
Although results provided some support for the dispositional basis of transformational leadership — especially with respect to the charisma dimension — generally, weak associations suggested the importance of future research to focus on both narrower personality traits and nondispositional determinants of transformational and transactional leadership.
(This isn’t a study directly about the correlation between personality traits and being hired to an executive position, because I couldn’t find any studies with that specific research question, surprisingly. The study is instead about the correlations between personality trait scores and ratings on transformational and transnational leadership, important concepts in management. At any rate one can imagine that people who appear to display transformational and transnational leadership would be the kind of people to be hired to executive positions.)
The study found some theoretically expected but again very modest correlations between personality traits and executive leadership. This, compounded with the fact that personality differences between men and women are modest themselves, throws a great deal of doubt on the idea that the deciding factor in this issue is disposition.
So what about pay? Another widely cited study, following up on a similar program of research, found that
Overall, only 3 to 4 percent of the gender gap is explained by differences in personality including differences in traits and trait returns.
(Page 21).
So to return to our lesson from above, statistically this implies that the rest of the difference here is because of other stuff. Not something else — but other stuff, a collection of other factors with teeny effect sizes like this one, all coming together in confluence to create this phenomenon. That’s not exactly an exciting, book-selling kind of explanation, but it’s a well-earned one. At the very least it’s not the kind of explanation that Peterson gives in this interview and H.L. Mencken once immortally quipped about: the kind of explanation which has always existed for every human problem: “neat, plausible, and wrong.”
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
This is a coda. There is a coda because A), the substance of it was conceived months after this essay was published, and I did not feel like doing the work of going back and organically integrating it into the essay, and B), I think it makes it at least a little more clear that this is an addition that came later, and generally it’s better to be more transparent than not.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
A little while ago I finished a freelance project; the project was to help research and write a thickish little booklet (~30 pages long) about women in the workplace, for a conference. Over the course of working on that project I learned a lot more about the gender pay gap. Because this essay continues to be read — and, indeed, seems to be attracting more and more views by the week — I figured I should put that new knowledge to good use, and include it as a kind of bonus feature to the essay. So what follows are some mostly unstructured points about the pay gap that further repudiate Peterson’s lazy, grave pronouncements about this issue.
Mostly Unstructured Point That Further Repudiates etc. etc. #1: Even more substantial evidence that differences in personality traits do not explain the pay gap
So the main thrust of this essay is that, in spite of the lip service he pays to multivariate explanations for things, Peterson offers a thoroughly univariate explanation for the pay gap: men and women are essentially different, and their essential differences are what’s behind the gap. Men are just less agreeable than women, and more competitive, and less neurotic, and this makes them more likely to do things like work longer, and seek promotions more actively, and instantiate the archetype of the King or whatever the fuck — leading to the differences in pay that the feminists have gotten themselves so bloody worked up about.
In the essay I work backwards from there, I suppose, and point out that, in the first place, the distribution of personality traits is not that different for men and women. I say that, on face, these differences, modest as they are, do not seem like they could be enough to explain away the amount of the pay gap that Peterson is saying they can. I provide studies to substantiate this point: differences in personality traits themselves are shown to have only a very modest effect (>8% variance explained in the studies cited) on differences in pay and ratings of leadership qualities. So, I concluded, essential differences in personality traits probably aren’t driving the differences in men and women’s pay or job placement.
As is so often the case I’ve been vindicated in my convictions. While helping to write that booklet, I found this absolute unit of a review article about the wage gap, by Blau and Khan. It is almost 80 pages long with references and tables included and it is difficult to overstate how thorough it is in its coverage of the matter. One of its sections, section 4 (pg. 35), focuses in specific on the contribution that “psychological attributes” and “non-cognitive skills” make to the pay gap. Being a review article, they summarize the results of other studies of this subject:
Table 7 summarizes the results of several studies that examine the importance of psychological factors or noncognitive skills on the gender pay gap where, if needed, we estimated this impact based on data presented in the paper. The notable finding from this table is that, in each case, gender differences in psychological factors account for a small to moderate portion of the gender pay gap. The proportion of the total gender pay gap accounted for by gender differences in psychological factors ranges from 2.5% to 28%, with all of the studies except for Manning and Swafford (2008) finding that these traits account for 16% or less of the gender pay gap.
The table in question:
It’s just — I mean, it’s a dogpile at this point. I try to maintain the proper terminology and tone of scientific discourse, talking about probabilities, averages — “most people most of the time”. I’m not like Peterson, anyway, who makes all-inclusive pronouncements about the world with the nonchalance of a man folding his Sunday laundry, nary a citation in sight. But at this point I’m about as close to that kind of conclusiveness as I can get: the evidence has only continued to heap: if you think essential differences in personality or temperament are what’s behind the wage gap, you’re wrong.
Which leads into the other bit:
Mostly Unstructured Point That Further Repudiates etc. etc. #2: Even if you were right you’d still be wrong
David Hume is my favorite philosopher, I’d say. I’ve written about him before. One of the reasons he’s my favorite is how he argues: he is incredibly lucid and evenhanded in his analyses; he preempts criticisms of his positions pretty much invariably, and usually steelmans them such that they fall all the harder; he is a masterful piler-on. What I mean by this last bit is that, when arguing against some opposing point of view, Hume thoroughly savages the key positions of that view — before revealing that it doesn’t even matter if those key positions were overturned or not in the first place, because there’s an even bigger, more fundamental problem with the opposing view. Now that I’ve finished explaining why you’re mistaken on these essential issues, I’ll cede them to you — and you’ll still be wrong. What a mad lad.
Here is my own amateurish attempt at the same.
Lets say that there wasn’t a small mound of papers showing that personality traits only have a teeny effect on pay differentials. Things immediately start looking up for Peterson: personality traits are after all highly heritable, at both the level of the Big Five themselves and at the level of their more specific facets, so if personality traits didn’t account for such a pittance of the variance in pay between men and women —if they accounted for say, 40% or 60% — the idea that immutable, biological differences between the sexes are what’s responsible for all of this would be in good shape. Right?
Because the thing is personality traits are a lot like the genes that are supposed to underlie them: they need and, in a very metaphorical way, they seek out situations in which they can express themselves. Being assertive is not enough to climb up one of the good Dr.’s beloved social hierarchies — indeed, being assertive is not even enough to act assertively. Rather you need situations in which you can act assertively in an adaptive way; you need the opportunity to express this personality trait and be reinforced for expressing it. Now, the causality here is definitely circular in that, while situations beget the expression of personality traits, personality traits also beget the occurrence of situations in which they can be expressed. To continue with our example an assertive person will often create situations in which they can assert themselves.
But therein lies the rub: women are not put in these situations as much as men are, and when they are in them an expression of assertiveness on their part will often be perceived in a negative, rather than positive light. Blau and Khan talk about this as well:
The issue we raised earlier, of gender differences in returns to psychological attributes is highlighted by Mueller and Plug’s (2006) study of the reward to the “big five” personality traits–openness to experience, conscientiousness, extroversion, agreeableness, and neuroticism. One of the most consistent gender differences in personality traits has been found for agreeableness, with women being found to be more agreeable than men (Bertrand 2011). Agreeableness refers to being more trusting, straightforward, altruistic (warm), compliant, modest, and sympathetic. Perhaps not surprisingly given labor market realities, Mueller and Plug (2006) find, in a regression context, that men earned a premium for being disagreeable. However, this attribute was not found to be related to women’s wages. Thus, the gender difference in agreeableness contributed to the gender earnings gap both because men were considerably more disagreeable than women, but also because only men were rewarded for this trait (Mueller and Plug 2006). These findings hint at a double bind for women. As in the case of negotiation, women face potential penalties for not engaging in this behavior but, if they do, may elicit negative or less positive responses than men
Do you see? Even if personality trait differences explained a lot of the pay gap, that wouldn’t necessarily mean that the pay gap is due to immutable biological differences, because the effect these biological differences might have on pay and hiring is itself mediated by social differences, which are themselves eminently mutable. So add another thing to the heap, the mound. Sad!
Note: If you respond to this essay and don’t address any of the empirical evidence I’ve gone through all the trouble of finding, reading, and communicating, I’m just going to make fun of you and then ignore you — and it will be completely right for me to do so.
