You Just Need To Look Beyond Correlation To Causation
You Just Need To Look Beyond Correlation To Causation
If you’ve been here for a while, you’ll know I can’t resist a hot-button issue that people are unwilling or unable to think about carefully. So when faced with a hot-button issue that’s literally about thinking, well, how could I resist?
In my article, What If Black People Are Just Stupid, I explored the question of racial IQ differences.
I asked whether IQ is a neutral way of measuring intelligence or, indeed, whether it’s even possible to measure intelligence in a culturally neutral way. I asked whether apparent “racial” differences were more likely to be cultural than genetic. And I pointed out that our genes influence many, many differences beyond the colour of our skin.
But a) those differences don’t tell you anything useful about an individual selected from that group, and b) we don’t rank IQ by eye colour or by height because we realise it would be meaningless.
Scott wasn’t quite convinced.
Scott:
If you group people by hair colour, you’ll discover that one shade is statistically more intelligent than the others
Not quite - you may observe a difference, but it probably won't be statistically significant. Calculating statistical significance is too complex to delve into here, and involves some level of judgment as the result is a number rather than "yes" or "no," but suffice it to say that if the difference between two groups is smaller than the standard deviation within either group then any apparent difference is not meaningful.
Steve QJ:
Not quite - you may observe a difference, but it probably won't be statistically significant.
I don't see any reason to assume this. And let's be clear, it is an assumption. If we tested IQ by hair colour we'd group all black people and Asian people and dark haired white people together. Would there be a statistically significant difference between that group and redheads, say? Or blondes? Who knows?
But that's the point. We can't say for sure because this research hasn't been done and almost certainly never will be. Most people would be quite comfortable saying that testing IQ by hair colour would be a stupid waste of time. Yet sadly, for skin colour, some people think it's meaningful.
Here, another reader, Ebo, joins the conversation to expand on Scott’s point about statistical significance.
Ebo:
If you associate "hair color" with "race" then what you're really doing is testing IQ differences between races, so you're back where you started and it breaks the analogy.
What Scott said was correct by definition. If there is no correlation between two attributes, then large, random samples where each exhibits one instance of attribute A - would have approximately the same number of occurrences of attribute B.
For example, let's take a random sample of 10000 women, and a random sample of 10000 men, and then checked how many in each group were born on a January 1st.
The numbers of Jan-1st-birthdays in the two groups will most likely not be the same, but they would be close enough to be statistically insignificant (unless there actually is correlation between sex-at-birth and January 1st).
It's technically possible that by pure chance we might pick 10k women who were not born on January 1st, and 10k men of whom 1k were born on January 1st. It's just very, very, very unlikely. It's unlikely enough for us to actually reach conclusions and make decisions based on the result. And anyway, this experiment can be repeated by peers with similar or identical settings, and if it would yield different results - hopefully it will open the door to additional similar experiments, and each repeating result would make it less likely to be a fluke by orders of magnitude.
The situation with IQ is much more complicated, because the way to determine it is not as objective as determining if a person was born on January 1st.
Steve QJ:
If you associate "hair color" with "race" then what you're really doing is testing IQ differences between races
Of course not. Black people tend to have black hair. So do Asian people. The association is a simple observation of reality.
I understand statistical significance. The issue is that even in something like your birthdays experiment, we might find phantom "racial" differences. For example, let's stick with birthdays, but choose a date 9 months after Christmas.
In cultures that celebrate Christmas (mainly white), you might well expect a higher incidence of births as couples enjoyed "the Christmas spirit." But in other cultures, there's be no special correlation. Maybe you'd need to choose a different date. Are white people genetically pre-disposed to have children in late September? Nope. You just need to look beyond correlation to causation.
But the real problem, once again, is that you think that "race" is a real concept, that can be discerned, presumably, by similar looking skin.
As I wrote in the article, back when we didn't understand genetics, this wasn't an entirely unreasonable assumption. But in the year of our Lord 2022, having decoded the human genome decades ago, we know as a matter of fact that intra-"race" differences can be greater that inter-"race" differences.
You cannot look at two people with similarly coloured skin and assume they're genetically similar. Not in athletic ability, not in intelligence, not medically. This is a silly, antiquated idea that we should all know better than by now.
Ebo:
you think that "race" is a real concept
I actually don't think that... I was just responding in the context of the statistics experiment.
And I agree with what you say about the possibility to mistake correlation for causation.
The instinct to group people by skin colour feels right. In fact, it feels obvious. Our skin, the largest, most visible organ in our bodies, is an easy way to categorise people. Add in some other superficial similarities, and it’s a no-brainer.
For example, if you came across two red-headed, bearded, bespectacled men, both of whom just so happened to be 6’ 4” and share the same Germanic-rooted name (first name and last name), you’d be sent to an asylum if you claimed they weren’t related.
But thanks to the magic of DNA testing, we learn that our instincts are wrong.
As I said to Ebo, back when we had a limited understanding of genetics, mistakes of this type were inevitable. Just as, before we understood gravity and astronomy, it would have seemed idiotic to say that the Earth was round.
Maybe we’re just in that awkward phase where our cultural knowledge needs to catch up to our scientific knowledge. Maybe the flat Earthers of the future will be race essentialists. Or maybe, by that point in the future, we'll all have become better at thinking.
Measurement of intelligence is a difficult matter. Ignore cultural biases and you will get wildly differing results from people of comparable intelligence. Work too hard to eliminate bias and you will get an IQ of 100 for everyone.
The single metric is just wrong. There need to be at least six distinct metrics. I can't remember them all and Google is hopeless, returning results on intelligence in the espionage sense. A few:
* mathematical ability
* spatial visualization
* language acquisition
* athletic ability
I score high on some, dismally on others.
One big problem is the extent to which intelligence skills have been exercised. I can do arithmetic in my head very well and any time I was stuck at a stoplight I would factor the six-digit number on the license plate in front of me into its prime factors so I got a lot of practice. OTOH I don't have a lot of athletic skill and in high school it took me the entire basketball season to be able to reliably shoot a basket and then the season was over.
3D is an interesting one because it's purely intelligence; we have been out of trees too long for thinking in 3D to be hard-wired. Parrots would score 500 on a 3D test, cats would score 500 on athletic IQ.
It's been a long time since I took an actual intelligence test but I remember being told I am not a genius but a lot closer to genius than to average. So what. Tensors mystify me though they are just one layer of abstraction atop math I know pretty well.
There is no evidence that black people are dumber than white people, but African cultures tend to include a lot more discursive (storytelling) tradition than logical and this inhibits the intellectual skills that lead to high scores on IQ tests. A lot of tests are based on recognition of patterns; a series of numbers, what is the next in the series? Not everyone thinks about numbers.
"The Bell Curve" was a deeply biased book; SJ Gould's "The Mismeasure of Man" is a much better text. Murray set out to show that black people are dumb and his predetermined conclusion drove his analysis.
So let me clear my throat and say I deeply, deeply neither wish nor suspect that Murray is remotely right. My suspicion is that the genetic roots of what we call “intelligence” are multi-layered and ultimately overpowered by cultural factors anyway.
That said, the crude categories of “race” (erroneous as they are) are not just arbitrarily based on skin color. A Cherokee, a Filipino, an Ethiopian, and an Afghani person could all have the exact same shade of skin and hair but we consider them 4 different “races.” To the extent that “race” has any meaning whatsoever, it is on the basis of how far back a group of people share a common genetic ancestor. Members of subpopulations that reproduced in isolation from other groups developed common genetic traits, like skin color but also muscle mass, hair texture, etc, that had evolutionary advantages for their shared circumstances. It seems like wishful thinking to assume that all of these traits could be fine-tuned within a sun population but brain function would remain exactly the same in all mankind. But I’m wishing it anyway. Goodness knows the subpopulations have been mixing and matching quite a bit over the past few hundred years, and any genetic differences are starting to blur to the point that I agree with Scott and Ebo- differences within populations likely outweigh most between them.
Regardless I do very much believe that even the idea of IQ is rooted in a European cultural view and there’s no way to debias it. I certainly, fervently disagree with Murray in his conclusion to not just throw up his hands and say “the differences are genetic, there’s nothing left to be done!” but also say “it’s important to disseminate this idea so that wokeism will stop destroying America.” He is so, so wrong, and I appreciated Coleman Hughes interview that challenged him on this directly.