Resolved: Black people are smarter than white people. (A response to Glenn Loury on the Murray question.)

 It is always important to consider heterodox analysis, and one of the most important sources of heterodox thought on economics, politics, race and culture is Glenn Loury.  Today, I am prompted by some thoughts after listening to his latest discussion with Heather MacDonald.  Loury, despite being African-American, has not shied away from interviewing one of the most controversial writers around, whose mere name is enough to send shivers down the spines of so many.  Charles Murray.  Even Coleman Spock Hughes nearly lost his shit talking to Murray.  Why?  Murray is famous for being the guy who "went there" on race and IQ, claiming that white people are genetically, racially smarter than black people.  If you were not aware of the writings of one, Dr. Charles Murray, then let this moment sink in as your introduction to one, Dr. Charles Murray.  Is there more to what he says than that?  Yes, but in the same way, James Earl Ray did more in his life than assassinate MLK.  I mean, the pulling of that trigger was just one moment, right?  Do we define him by that one instant?

Yes, I am comfortable with that.  I will define him by that one moment in his life.

Here is the thing about Charles Murray.  When I was an undergrad, he was the villain du jour.  Proto-woke professors told us about how evil Murray was, but never gave us the book, The Bell Curve.  I found that strange, and vaguely improper, so I actually read the book because my basic guiding principle is that if you tell me not to read something, I'm-a-gonna read it.  I even teach a class based on that guiding principle.  I teach what some other group tells you not to read because it is evil, wrong and scary.  The first thing on the syllabus is a selection from some guy with a square mustache.  That's right.  Charlie Chaplin.  That guy was fucked up.

I have my own methodological critiques of Murray, and the short version is that I do not believe that he addressed confounds in a compelling way.  That's the polite, scholarly way of putting it.

As an experiment, I once assigned his more recent Facing Reality, to see if a class could have a calm, reasoned, civil discussion of the claims, the social scientific fallacies, and the broader lessons that can be drawn from the statistical errors, using the errors in the book to teach some probability theory.  How did it go?

The next time, I decided to go with my man, Thomas Sowell instead.

Wait a minute.  Why am I always going to Loury, and Thomas Sowell, and John McWhorter, and Shelby Steele, and Coleman Hughes, and what do all of these people have in common?  Hmmm.  Let's return to that.

Later.

Anyway, Loury and MacDonald are made of sterner stuff than I am, and they addressed Murray's questions recently, asking if scholars can even ask the questions.

Here is where I think the data stand on the question, given extant analysis.  If you haven't taken stats, shame on you, and if it has been a while, you may need a refresher, so here is a quick primer/refresher.

Scientific analysis works as follows.  The default position is skepticism, that there is nothing, no pattern, just randomness.  You get an idea, derived from "the literature," or some observation.  This idea is that there is a pattern in the data.  There is a trend.  A difference.  A correlation.  (Actually, one should say, "association," because the word, "correlation" refers to a Pearson's r, whereas an association is a nonparametric pattern).

There is an independent variable (the cause), and the dependent variable (the effect).  You hypothesize a causal relationship between the IV and the DV, and set forth, in advance, the criteria by which you test the hypothesis.

The hypothesis, though, can have two tails, or one, and this is where I am going.  In a one-tailed hypothesis, you specify the direction of the association: you hypothesize that it is positive, or you hypothesize that it is negative.  For example, I hypothesize that an increase in interest rates reduces inflation, not merely that it changes inflation.

In a two-tailed hypothesis, you hypothesize merely that there is a relationship, without specifying the direction.

As you specify your hypothesis, you must specify a "null hypothesis" to which you default if the evidence is insufficiently compelling for your hypothesis.  The null hypothesis, essentially, is randomness, but the null and your hypothesis-- the "alternative hypothesis"-- must be mutually exclusive and exhaustive.  That is, one and only one must logically be true.  We shall return to this.

The analysis proceeds as follows.  Since we default to the null, we ask, how likely is it that we would observe data of the kind that we observe were the null hypothesis true?  If that probability is sufficiently low-- generally, below .05, hence p<.05-- we "reject the null," which logically implies accepting the alternative, but we don't say that for technical, epistemological reasons.  The short version is that we structure proofs around disproofs.  Take two mutually exclusive, exhaustive propositions, disprove one, and by default, you have proven the other.  So, we ask whether or not we can "reject the null hypothesis."

Anyway, that's the basic epistemology of scientific inquiry through statistical analysis.

Enter Murray.  His alternative hypothesis is the claim that white people are genetically predisposed to be smarter than black people.  That, my friends, is a one-tailed alternative hypothesis, making the null a one-tailed hypothesis in which either black and white people are, on average, equally intelligent or black people are smarter.

Yes, his null hypothesis includes the proposition that black people are smarter than white people, on average.  That is the other tail.  There is public discussion of one tail, but not the other.

Fuck that.  Let's ask about the other tail.

We usually ignore the second tail in a one-tailed hypothesis.  Why?  Consider, and consider whether or not I am pulling a sleight of hand.

I am, kind of.  But if I point it out, it isn't dishonest.

Suppose a firm is accused of hiring discrimination.  We might test the claim statistically.  The claim is a one-tailed hypothesis that while the pool of potential employees includes X% from Group A and (1-X)% from Group B, the firm has hired X + k from Group A because of bias.  The statistical challenge will be the baseline, but let's ignore that for the moment.  Suppose I fail to reject the null in my one-tailed hypothesis because while the firm hired X + k > X, random chance would produce X + k more than 5% of the time.  Perhaps notably more.

There is still less than a 50% chance of observing X + k if the firm is biased against Group A.  The firm might be biased against Group A, and that is part of the structure of the null, but only by logical necessity.  Our empirical data are not at all supportive of that claim.  It could be, but that bias would have to be suppressed by something else.

But that's also where I'm going.  So am I pulling a sleight of hand on you?  Yes, but only kinda.

Let's talk about the other tail.  I cannot recall the last time I saw anyone bother with it.  However, if Murray and his coterie wish to push his tail, and Loury says that we should consider the question, then I will pose the question.  What about the other tail?  After all, I have no problem acknowledging that Glenn Loury is far smarter than I am.  Also, Sowell, McWhorter, Steele, Hughes, and indeed, I think that the smartest commentators on race are black, and for the most part, it isn't even their main thing.  If you don't know who Glenn Loury is, or who Thomas Sowell is, go learn.

So let us turn to the data.  Could Murray muster sufficient data to reject the null?  In my assessment, no.  Murray's goal was to assess this thing called "G," or, general intelligence.  Psychologists-- that lowest category of wannabe social scientist-- have devised a variety of tests to assess G, and these tests do have predictive power, but what might predict scores aside from actual intelligence?  And are any such variables statistically associated with race?  If so, those are "confounds," and if not properly addressed in the analysis, one cannot draw conclusions.  One cannot reject the null without controlling for such potential confounds.

So what are these potential confounds?  The biggies are socioeconomic background, but more still, early childhood nutrition, parental activity, such as reading to kids at an early age, and so many other factors that are both hard to measure and necessary to include that I am unconvinced by Murray's analysis.  He did not have sufficient evidence to justify rejecting his null, and remember that his null is the null including the hypothesis that black people are, on average, smarter than white people, with heritability as a partial explanatory factor.

Let's go there.  I'm doing this.  Fuck it.  In tenure veritas.  What if black people are, on average, smarter than white people?

Murray's null cannot be rejected, given available data.  Moreover, we are observing the kinds of suppressor effects that make the other tail interesting in a way that it was not in my hiring firm example from above.  So I want to consider the other tail.  The tail that does not get discussed, except perhaps in the circles that I do not travel.  Maybe black people are, on average, smarter than white people.  With a biological basis.

Dig, dig, dig...  Does cancellation work if you talk about this tail?  We're going to talk about that too.  Can I get cancelled for that?

Before we get there, though, consider the following.  I would love to perform the following impossible, absurdly expensive social science study.  Randomly sample a set of newborns.  True random sample, from across the country.  Just pick their Social Security numbers out of a hat, and require the parents to give me full access.  Track everything.  Early childhood development, nutrition, activities, frequency of parents reading to the kids, I mean everything.

Track them over their lives until they get to an IQ test.  Throw every fucking thing into a statistical model predicting the kids' scores, including all of the impossible-to-measure stuff to which I want access, like exactly what the kids are eating as lil' tots.  Also, include race as a categorical variable (which gets complicated with all of the variations of mixed race, but since I can't actually do this, don't bother trying to pin me down on my measurement scheme, I'm on a batshit roll here, yee-HAW!).

What if, upon performing this ludicrously impossible study, there is a positive coefficient for the...

OK, I need to explain this term before I type it.  "Dummy variable."  A dummy variable is a dichotomous variable, taking on a value of 0 or 1, indicating either the presence or absence of a trait.  Race is a "nominal" variable, meaning that it is categorical with no ordering between the categories.  (If there is an ordering, we call it... wait for it... an "ordinal" variable.)  If we have a nominal variable with k categories, in order to use that variable as an independent variable in a regression model, we create k - 1 dummy variables, with one for each category, leaving one as the excluded category (for technical reasons).  We do this with race all the time in the social sciences.  For the sake of funsies, let's make whitey the "excluded" category, othering me, and if words are violence, I just committed self-harm.  Regardless, every other described racial category gets a "dummy variable."  That means there will be... (cancellation alert, cancellation alert) a "black" dummy variable.  But what if the "black" dummy variable winds up with a positive coefficient, when "white" is the excluded category, after controlling for all of the shit that Murray couldn't measure?  What would that mean?

It would mean that black people are, on average, smarter than white people, and since we should not reject the null on the basis of Murray's evidence, given his omitted variable bias, we must treat this as plausible.

And if we have controlled for everything else, all that remains is heritability.  If we have controlled for everything else, the best interpretation would be a biological predisposition among black people to be smarter than white people.

And we cannot rule out this possibility!  We don't talk about it publicly.  Most people probably don't think about it, and certainly very few white people think about it, but I cannot rule it out.

Here is an exercise.  I would like you go walk through a diverse city, thinking about this tail of Murray's null hypothesis.  I am 100% serious about this.  Walk through a diverse city thinking about the actual null from Chucky Murray, which should not be rejected.  There's an actual tail there.  Don't think about the zero point.  Think about the tail, which is mostly there for technical reasons, but if we discuss one tail, why not discuss the other?

What does this do to your thoughts as you walk through a city?  If you are white?  If you are African-American?

Do this.

Murray's tail is the tail we examine because we focus on psychologists' devised tests, educational attainment, and other such easily quantifiable variables, and statistician that I am, that is where I go first anyway.

Yet, my suggestion is that we examine not just the suppressor effects, but other measures.  It is true that educational attainment, for example, differs across races.  Are there measures of intellectual achievement by which African-Americans disproportionately excel?

Come on.  This is easy.  How about writing?  Shall I turn to neurology and the importance of logic in language processing?  Really, though, you know I'm going for music.  Music is math.  Chord theory, polymeter and polyrhythm, this is all math, and the most intellectually advanced music in the Western tradition is jazz.  Not close.  If you want to argue that Indian classical is the most intellectually advanced, I'm not sure I can argue with that (konnakol fucks with my head in a way that I love), but jazz theory is math.  Would you like to argue that some of its innovators were not music theorists, conservatory trained?  Go for it.  Then how much more impressive is it for the autodidact?

There is even a school of heavy metal built on jazz theory.  Do you know what it is called?  Mathcore.  Go listen to some of it, and try to decode it, understanding the giants on whose shoulders they really stand.  On whose mathematical shoulders.

I could keep going, and some time, perhaps I will, but my only real point is that there is a second tail.

Would I stake money on the genetic superiority of one race over another?  Jew-boy, me, ain't gonna do that.  Bad things come of it.  I am merely noting that one can take a hypothesis, and turn it around, observing the tail that has not been rejected, and discuss it, having failed to reject it.

And when you walk through a diverse city thinking about it, understanding your inability to reject it, what do you think?

Whatever you do think-- whether you are African-American, white, Latino, Asian, or of some other race/ethnicity-- here is how you might choose to respond to those thoughts.  This person whom I have encountered is one person, about whom I know essentially nothing.

Murray tried to defend himself by saying that one should do so, even believing what he believes, but for some, it may be hard.

Is it easier, though, when thinking through the lens of the other tail?  And if so, what does that say?  Something about history, no doubt.

Does any policy follow from the other tail, should you believe it?

What would be the social reaction, should I be allowed to conduct my study, and demonstrate convincingly that black people are genetically smarter?  Plenty of white people would howl with rage, of course.  How would the so-called "anti-racist" crowd respond?

One of the many criticisms of Charles Murray has been that Loury is wrong.  The question is one we do not need to ask, because what if Murray were right.  How gross would that be?  What would be the social implication?  Would you want to know that?  Instead, is it better to treat a null hypothesis of equal intelligence between the races, not merely as a null hypothesis, but as an axiom for the social implications of that axiom?

That's a pretty compelling argument, isn't it?

But if you believe that argument, then you must howl with just as much rage should I conduct my study and find that black people are genetically smarter, which again is quite possible, and indeed, plausible.

Would the Kendi-ites, the DiAngel-ites and the rest do so?  I don't think so.  That asymmetry is something troubling.  Do you want to believe that your people are smarter?

Hmmm.

That speaks to the value of an individualist lens, and in some ways, the critique of Murray, and of Loury.

And perhaps ultimately of asking the question.  Of thinking of either tail.

My framework, as always, is unabashedly individualist.  Properly framed, any analysis of such group differences (if we want to conduct the analysis) should have been framed with two-tailed hypotheses, which means this is all at least partially sleight of hand.

Yet I gave my final lecture last week in a class dealing with controversial books, and my primary lesson was on the value of dissent, and the importance of looking for illusory consensus.  The mere fact that social dialog addresses one tail without addressing the other is telling, and distressing from an individualist and equalitarian standpoint.  A response, then, is to note and introduce the question of the other tail.

Hence, I do so.  Look at the other tail.  Ask about the other tail.  Think about the other tail.

Walk through the world asking if the other tail contains the true parameter, as a cognitive antidote.

Because maybe black people are smarter than white people.  If we are forced to confront the Murray question, consider the other.

And consider the wasted human potential in the world.

How about some Herbie Hancock?  "Survival of the Fittest," from Maiden Voyage.  These guys aren't just virtuoso technicians.  They're smarter than you are, and smarter than I am.

PS:  Did I just cancel myself?  Um... how?