Thread by @kareem_carr

"The cognitive demands of those jobs mean a lot more white people qualify than black people."

This video is full of misuses of statistics! A thread. 🧵👇

1. First of all. CORRELATION IS NOT CAUSATION. As a statistician, I'm contractually obligated to let you know that these patterns that he's talking about are all correlations. They are not proof of causation.

2. An IQ difference of 12 points is NOT "a lot". The distribution of IQ and the distribution of human height are very similar. They follow what's called a bell curve. So, we can get an intuition for IQs by thinking about heights which we are typically more familiar with.

A difference of 12 IQ points translates into a difference in height of about 2.5 inches for American men and 2 inches for American women. It's not nothing but it's a bit of stretch to call this "a lot".

3. Should IQ always translate into job performance? NO! According to the science of IQ (which I have my problems with), IQ only translates to tasks that are "g loaded" where g stands for "general mental ability".

Another way of saying this is if a job involves a lot of things that IQ tests test for then we expect IQ to predict job performance. Nothing in the science of IQ suggests that IQ should predict EVERYTHING. A randomly chosen job MAY or MAY NOT involve anything IQ related.

I'll leave it to you the reader to decide for yourself if you think the task of completing a paper test is very similar to the kinds of tasks involved in nursing.

4. Even if IQ did translate to job tasks, would we see a difference in job performance between two groups with different IQs? Not necessarily. This logic fails for TWO reasons.

SELF-SELECTION. People often self-select out of careers that they don't think they will be good at. So, if a person felt they couldn't compete cognitively with the majority of people in their field YET STILL continued to pursue that career, it suggests that they probably felt...

...they were bringing other things to the table. This might be persistence, people skills, etc.

SELECTION DUE TO HIRING. Employers are often looking for a basket of talents when they hire people. So if a candidate has a weakness in one area and still got hired, we would expect that this candidate will be strong in other areas.

In statistics, we call this "selection bias". Whenever people can make choices about whether to join a group OR when people are selected to join by some kind of decision-making process, we expect to see a bias.

It's one of the reasons statistics is hard. It's part of the reason why we assign people randomly in clinical trials and keep the assignments secret. Allowing people the freedom to select often creates bias.

This is also why naively making numerical comparisons between human groups is more often that not BAD science and BAD statistics.

5. Even if we agreed black and white nurses (on average) had different talents, does this imply the higher IQ group would get paid more? NO! Let me put it to you this way. Do you think the highest paid person at your job is also the person with the highest IQ?

As I mentioned before, employers value things in addition to technical skills like loyalty, creativity, hard-work, wisdom, people skills, etc. These might be more valuable traits! So it's not obvious to me that a slight IQ advantage will always translate into more promotions.

For instance, would you want a person in leadership to be the absolute best person in terms of technical skills or someone who is 90% as good but was also loyal, creative, wise, hardworking and had good people skills?

6. At the end of the clip, he generalizes from nurses to the whole job market. He claims the science of IQ will mean that you will see whites doing better than blacks in almost every kind of job. This does not logically follow.

For all the reasons I mentioned, each job type requires us to reassess all the points I made before in this new context. Are the job tasks g-loaded? How much selection is going on? What is the relative value to employer of soft skills?

7. When he says that we should expect to see a black white compensation gap due to IQ, he's implying that IQ differences are the dominant driver of the compensation patterns that we see in society. This has not been scientifically proven at all.

As a statistician, I can tell you that not everything that's statistically significant is actually significant. The effect of racial difference in IQ could easily be swamped by other much larger factors (like structural racism for instance).

Just because IQs differ and income differ. It doesn't mean income differences are driven by IQ differences. Again. CORRELATION IS NOT CAUSATION.

Whether you are agree or disagree with the general idea that black-white differences in income are due to racism, I hope you agree that sloppy thinking disguised as "science" just makes this contentious issue even harder to discuss.

Hope you found this thread useful. This kind of long-form content takes extra work so if you like it and want to show support, like and retweet the thread, and give me a follow! 👍