September 21, 2017

Thick Heredetarianism

This article is for people who are already racial heredetarians, but who still hold to an individualistic policy proscriptions. There will be more coming on this topic. If you’re not a racial heredetarian and still a big believer in racial equality in the brain, this article is not for you, since it builds on “thin heredetarianism”.

It is very common, particularly among libertarians, to take the basic data on race and IQ, come to reasonable conclusions, but then promote individualism, or at most IQ-based immigration control.

They don’t take into account differences in cognitive structure, “g factor” and difficulty, regression to the mean over generations, or the fact that other traits

They also don’t consider that, if you bring in IQ 130 blacks, those blacks will have loyalty to their race, and won’t appreciate the IQ-based immigration policy, and probably will refuse to believe in genetic race differences in IQ, and thus will agitate for the end of that policy.

This article will operate purely within IQ. Sean’s article against meritocratic immigration dealt with many topics, bringing in new terms to argue against IQ nationalism. This article will deal with IQ nationalism purely on it’s own term: IQ. But this article will be limited to the black population in the United States, as the US white-black gap is one of the most extreme and has the best data, making it good to do first to establish the basic arguments.

The Nominal Gap Doesn’t Matter Much

On this site we talk a lot about the Flynn Effect, and how it is “hollow for ‘g'” and how it is small in magnitude. When talking about it, we start to bump up against something a bit complex to explain. And that is that “the 15 point IQ gap doesn’t matter much”.

Lets look at a progressive matrices test done in South Africa. The overall race gap on this test was gigantic, but it showed the scores by race and question:

If you look at the questions, you can see that the more difficult questions show larger racial gaps. In the first question, nearly everyone got the answer correct. So there was no racial gap. But, if you look at question 24, 64% of whites answered correctly, and only 21% of blacks answered correctly. On question 34, 53% of whites and 20% of blacks answered correctly.

So, if you wanted to make a test that would show a small black-white gap, what would you do? You’d make the test easy. If you wanted to show a big black-white gap, what would you do? You’d make the test hard.

And so when people say there is a “15 point IQ gap” – that’s an artifact of just how “g-loaded” the IQ tests being used happen to be. Increase the “g-loading”, increase the racial gap (for the most part).

Difficulty and g-loading are closely related, but aren’t necessarily the same thing. “G-loading” just refers to the predictive validity of one task onto another. Highly abstract and difficult to teach abilities, such as picture-pattern completion, block design, mazes, are more predictive of scores on everything else than, say, arithmetic.

And, as Charles Spearman predicted, it is more difficult to raise scores on “g-loaded” questions than on non-g loaded question, which is evidence that this “general intelligence factor” is something of a “genetic intelligence factor”.

Also, if cognitive function is generalized, then basically everything can serve as an IQ test; because to one degree or another, any measure of any single cognitive ability is also, to some extent, a measurement of generalized cognitive ability, or “general intelligence”.

Blacks and Whites with the Same Overall IQ Have Different Cognitive Profiles

Blacks and whites with the same IQ number on, say, a Wechsler test, or the same SAT score, will still have very different cognitive profiles.

When whites and blacks have the same overall IQ score, blacks do better on tests like coding (quickly matching symbols to shapes), arithmetic (simple math), and digit span tests (memorizing a series of numbers). And while blacks with the same IQ have an advantage on both forward and backward digit span tests, their advantage is smaller on the backward one (repeating a series of numbers backwards).

Just for curiosity, I eyeballed Jensen 1989’s data where he compared blacks and whites with an IQ of 89.

Black-white gap by subtest of blacks and whites who all have a full-scale IQ of around 89

Reynolds (FSIQ matched) Gap in SD
Object Assembly 0.51
Mazes 0.5
Block Design 0.27
Similarities -0.11
Vocabulary 0.06
Information -0.16
Comprehension 0.44
Arithmetic -0.36
Picture Completion 0.11
Picture Arrangement 0.34
Digit Span -0.62
Coding -0.49

I then subjectively binned the “white tests” as object assembly, mazes, block design, picture arrangement and comprehension (understanding the meaning of a sentence or paragraph). The black tests were coding, digit span and arithmetic.

That’s a gap of 0.412 standard deviations, or 6.18 points. So a white person, with the same overall IQ, would score 6.18 points higher on these “white tests”. Blacks score 0.49 SDs higher on their tests, or 7.35 points higher on “black tests”.

But even this understates the true differences within IQ. Because remember, whites do better on more difficult questions compared to blacks; and so while whites with the same IQ do 0.5 SDs better on Mazes, that number is merely an average of the easy mazes – where the black-white gap is smaller – and the more difficult mazes – where the black-white gap is larger.

Using the data from South African Engineering school applicants (an elite sample), we can look at the white percentage correct / black percentage correct on items by difficulty:

Difficulty White % / Black % correct Conversion to general population IQ gap White advantage above baseline
Whole Test 1.51 15 0
Top Half 1.84 18.28 3.28
Top Quarter 2.16 21.46 6.46
Top 8.3% 2.68 26.62 11.62

To take a sports analogy, there are those teams who are more “dangerous” – i.e. they have the ability to beat teams that are supposedly much better than them. For example, Michigan may have a terrible year, having a losing record, doing about as well as Purdue or Rutgers by record. However, we all know that Michigan has a much better chance of beating Ohio State, even in their down year, than Purdue or Rutgers does. Why? “Because it’s Michigan”. There is an underlying strength in any Michigan program, even when the stats to date say they suck.

For world of tanks players, think of it like players’ whose average WN8 across all tiers are the same, but one player’s WN8 is propped up by doing really well at tiers 1-3, while the other is more even and actually rises a bit at tier 10. Most consider the player who is better at the higher tiers “better”, and don’t think tier 2 sealclubbing really well means much. Similarly, most don’t think doing well at remembering a string of numbers (digit span) or simple math (arithmetic) is as important as comprehending the meaning of a sentence or being able assemble an object in a certain way.

And that’s what you see with whites. While the overall score on the test may say a white person has an IQ of 85, that white person has a much better chance of solving a truly difficult problem than a black person with the same IQ. Michigan in it’s down year vs. an average year for Rutgers. “Because he’s white”.

If you took blacks and whites with the same “overall IQ”, and gave them a test which had the same types of questions as the one they already took, but only had the top quarter of difficulty, whites would have about a 6.46 point advantage. And if you further limited the test to the kinds of abstract problems that whites do better on, you would end up with whites, who have “the same IQ” on a normal IQ test, scoring about 12.76 points higher.

And given all of this, how meaningful is “overall IQ” anyway? The dirty little secret is “not much”. Questions that are more A) abstract and B) difficult, whites do better on. I will label these questions the “High-End IQ”.

And that’s fine. But the point I’m making is that there is far less overlap in cognition between blacks and whites once you dehomogenize intelligence.

Regression to the Mean

Before explaining the mechanism, first let me point to the brute fact that this happens: the offspring of parents regress to the mean of their breeding population.

On average, if two black parents with an IQ of 100 have a child, the best prediction for that child’s IQ is 92.5. Now it could be higher than this number, or even lower, or they could have a child is who is even smarter than the parents. These numbers all look “unrealistically clean” because and IQ score is actually a rank-order score, and so regression toward the middle of a population probably wouldn’t regress to an intermediate score; but it will necessarily regress to an intermediate RANK, which is what IQ is.

Regression to the Mean by Mid-Parent IQ for Full-Scale IQ

Mid-Parent 70 85 100 115 130
W. Regression 85 92.5 100 107.5 115
B. Regression 77.5 85 92.5 100 107.5

– The numbers look clean here because the white raw score is set to 100 by design, and the black IQ, just as it happens, happens to be precisely one standard deviation below the white score. These two facts result in regression numbers that appear “too clean”.

But what is the regression effect on the “white IQ” I talked about earlier – the top 25% difficulty on more abstract questions? Well, those show much larger gaps, obviously:

Regression to the Mean by Mid-Parent “High-End IQ”

Mid-Parent 70 85 100 115 130
W. Regression 85 92.5 100 107.5 115
B. Regression 71.14 78.62 86.14 93.62 101.12

What this would suggest is that for blacks whose mid-parents have a “High-End IQ” of 100, their offspring will probably have a “High-End IQ” of 86.14.

If you assume the same standard deviations for black and white “white IQ” scores as they have on full-scale IQ scores, which is 13 for blacks and 15 for whites, we can estimate how many blacks and whites there will be at each IQ and at each “white IQ”:

Percent at or Above Each Full-Scale IQ

FSIQ % White at or Above % Black at or Above
100 50 12.43
115 15.87 1.05
130 2.28 0.03

Percent at or Above Each “HEIQ”

HEIQ % White at or Above % Black at or Above
100 50 1.64
115 15.87 0.05
130 2.28 ~0

Now, with all of this, we have a very interesting result. The percent of blacks scoring 100 or above on the “white IQ” is 1.64%.

For black parents with a HEIQ of 100, only about 14.33% of their offspring will have a WIQ of 100 or above.

For black parents with a HEIQ or 130, about 53.43% of their offspring will have a WIQ of 100 or above.

And so the number of black families who can maintain a “High End IQ” of 100 or above, we would expect to be basically zero.

For “High-End” IQ, the black-white gap is something like 27 points, not 15; and the black standard deviation is probably smaller than 13 points. However in the above calculations I used a 13 points standard deviation for blacks, as that is the black SD for full-scale IQ and I am only speculating that they have a lower standard deviation for the “High-End IQ” that I’ve categorized.

Hopefully this has explained why looking at nominal full-scale IQ gaps for a single generation is going to diminish the importance of racial differences, and make it seem like there would be more overlap in cognitive traits in a multiracial society than there really is. There is an artificially high level of overlap that is a result of aggregating tests that blacks do relatively better on with tests that whites do relatively better on; separate these categories, and the real difference is revealed as much more extreme.

Facebook Comments