Jewish IQ: not a myth

Update: the average IQ is probably closer to 108-109, but it’s hard to tell due to the fact that Jewish identity markers are inconsistent and there are large differences between subtests.

############

Several prominent psychologists endorse the claim that Jews are more intelligent than White gentiles. Of them, I count:

• Steven Pinker: Jewish IQ of 108-115

• Richard Lynn: Jewish IQ of 107-115

• Cremieux: Jewish IQ of 110

• Jordan Peterson: Jewish IQ of 115

• Cochran: Jewish IQ of 112-115

There is a tendency for numbers to be cited as gospel after a certain point (e.g. Black IQ of 85, Asian IQ of 105) without doing the necessary statistical/conceptual work to do an estimate that is precise. It would be hilarious if Jews were somehow falsely considered intelligent and fudged the IQ test score averages to make themselves look intelligent. So, why not check?

Fortunately, lots of data these days is public, and have been able to determine that Jews score higher than White gentiles in IQ in various datasets:

• NLSY79: 110 (n = 109) (using sampling weights and ASVAB factor scores - Murray also found a Jewish advantage in this dataset)

• NLSY97: 107.1 (n = 99) (using sampling weights and ASVAB factor scores)

• Project Talent: 106.83 (n = 6944) (my estimate is slightly different from Dunkel’s, who found 106.7)

• MIDUS: 104.5 (n = 96)

• GSS WORDSUM: 108.5 (n =1321)

The NLSY datasets use the ASVAB as a measure of cognitive ability, which loads highly on crystalized ability, but is otherwise a fairly exhaustive test of intelligence.

from here

To calculate IQ scores in the Project Talent, I took the 50+ subtests that test cognitive ability, combined those into 6 major abilities, and used those to calculate IQ. Jews scored above White gentiles, though this varied based on the ability tested.

All scores are normed to the White gentile mean/SD

Jews have mediocre memories and spatial abilities, but perform well on every single other test.

The MIDUS score comes from a set of tests that mostly test executive function and processing speed (backward counting, fluency, number series, word recall, and a few others), which explains why they didn’t score that well (104.9).

I was also able to find other samples in the literature that scored ability exhaustively and scored relative subjects relative to the White mean:

• ABCD (2023): 106.5 (relative to the White mean, according to Fuerst)

• Hirsch 1926: 104.23 (n = 75, Polish Jews)

• Hirsch 1926: 101.23 (n = 627 Russian Jews)

• Seaqo and Koldin (1936): 108.73 (n = 800)

• Rigg (1936): 98.4 (n = 445)

The older samples are all over the place, probably due to data quality/sampling issues, but the ABCD sample generated the same exact score that the Project Talent dataset did. The test used was fairly high quality:

from here

An (unnamed) colleague of mine noted that in two different years (1995 and 2002), Jews scored the equivalent of a 108 on the SAT, which is about as high as you would expect them to score if they have an IQ of 106.5-107 and a small crystalized tilt. I was able to get the source for the 2002 SAT, but not for the 1995 figure. The Unz surname analysis in The Myth of American Meritocracy is also consistent with a difference of roughly 7 points, according to the math the unnamed friend did.

Because of the ability tilt issue, I attempted to do a different meta-analysis where I subtracted 1.5 points from tests that test crystalized ability, and added 1.5 points to tests that tested non-crystalized abilities. I removed all the old samples that had inconsistent results for the sake of data quality. This adjustment increased the precision of the estimate massively.

The heterogeneity is still pretty bad, but the confidence interval does not even include 108, which was the lower bound that most of the other researchers cited. I could add more studies, but many of them (particularly the old ones) look iffy.

On to the particular objections people have with the Jewish IQ:

Objection 1: Israel’s IQ is 92

Cremeiux has already tackled this, and it’s due to the fact Israel has a large Mizrahi, Ethiopian, and Arab population.

OG tweet

Using more realistic numbers (85 for Arabs, 75 for Ethipoians, 97 for Europeans, 107 for Ashkenazi) an estimate of 94.9 is calculated, just 2 points from the estimated average.

> (.318*.79)*107+(.49*.79)*85 + (0.079*0.79)*97 + (.124*.79)*97 + (0.03*0.79)*75 + 0.21*85
[1] 94.96743

Objection 2: researchers use bad measurements of intelligence to test the Jewish IQ.

This criticism doesn’t apply to my work, because most of the datasets I used (particularly the project talent) have high quality measurements of intelligence. Because of the dominance of g, most group differences in low quality measurements of IQ will have differences in the expected direction.

Looking back at the project talent table, all subfactor tests generate differences in the same directions except for four cases:

• Asians score lower on general knowledge, processing speed, and verbal ability despite having slightly higher IQs than Whites. Even so, the ability tilts are still pretty close to the overall average (100.6).

• Jews have lower spatial abilities/memories in comparison to what would be expected from their IQs.

Objection 3: Jewish eminence is due to nepotism/corruption

Perhaps, but this does not explain why they are overrepresented in basically every single position of power, achievement, and ability.

You’re free to check all of the sources (I provided them in the article), and the figures for the intellectual movements, which were assessed using the famous early life check.

Objection 4: The Jewish IQ is a myth that Jews use to justify Jewish power.

That doesn’t mean the theory that Ashkenazis have higher IQs than Whites is false.

Objection 5: Jewish IQ cannot explain Jewish eminence.

For what it’s worth, almost everybody who has examined this question quantitatively has come to the same conclusion - the higher Jewish IQ alone cannot explain Jewish eminence. For my own work, it appears somewhere between 17% and 40% of the extremely eminent in America are Jews, though it’s difficult to tell what the exact percentage is. Based on current population estimates, this table estimates the distribution of IQ within Whites and gentiles in modern America:

Even at extremely high levels of ability (160 > IQ > 145), there are still about 6 gentiles for every Jew, which is below the observed percentage of Jewish high achievers.

Instead, noncognitive explanations such as general factor of personality or mental instability must be considered instead. It would be unlikely (for historical/evolutionary) that White/Jewish differences in personality are larger than the ones in cognitive ability. If eminence is assumed to be due to a sum of cognitive and noncognitive factors that are uncorrelated, and Jews have a 0.45 SD advantage in cognitive ability, and a 0.35 SD advantage in noncognitive ability, and that these factors contribute equally to eminence, the distribution of “latent eminence” within Whites and Jews will look something like this:

Jews are now 28% of the people who score within 4 to 5 SD above the mean in eminence, which is very consistent with the observed number of Jewish geniuses, despite the small advantages of Jews in both cognitive and noncognitive abilities. This, of course, is not an accurate model of reality, as this model does not accurately consider extrinsic factors like luck, location, beliefs, motivations, and the like.

Personally, I doubt Jewish psychological differences/advantages can wholly explain their achievement, though they almost certainly play a large role in it. Others like Sean Last have documented that social networks may play a role in Jewish achievement, which seems plausible enough to me.

Appendix:

Effect sizes used in the meta-analysis of Jewish IQ

adjustment already made in the mean column

I tried latent analysis in the Project Talent upon Cremieux’s request. Personally, I think latent analysis is useless when dealing with precise measurements, as they tned to derail work because of arguments over fit indices/cross-loadings/model accuracy/MGCFA and so on. Regardless, I did some simple lavaan models in project talent to attempt to test whether the g difference is of a different magnitude than the observed one, and it was marginally larger (107-108 points vs 106.8).

(Jewish g of 107.4)

Jewish g of 108

(CULTURAL KNOWLEDGE/TECH KNOWLEDGE NOW MODELED SEPARATELY)

Jewish g of 108

Modeling every subtest as all falling under one g, aka using this model:

lat0 <- "

G =~ mech + twod + threed + abst + arith + bmath + amath + spell + capital + punct + english + express + words + reading + create + disgwords + arithfast + table + cler + obj + sentmemory + wordmemory + phsk + aerok + eleck + mechk + sportk + huntk + fishk + litk + musk + sstudiesk + mathk + biok + sciattk + farmk + homek + artk + lawk + healthk + engk + archk + jourk + fork + militk + salesk + prack + clerick + biblek + colorsk + etiqk + outdk + photok + gamesk + theatk + foodk + misc

G ~ race

"

Jewish g of 108

From what I’ve heard from Woodley, g loads more heavily on nonverbal factors than verbal ones when hierarchical models are used, which is also in line with what I have seen in the literature. Because of this, I tried a model with no verbal or information factors first, and then a model with no verbal factors (but with info).

lat0 <- "

G =~ VIS + MATH + SPEED + MEM

VIS =~ mech + twod + threed + abst

MATH =~ arith + bmath + amath

SPEED =~ arithfast + table + cler + obj

MEM =~ sentmemory + wordmemory

G ~ race

"

lat0 <- "

G =~ VIS + MATH + SPEED + MEM + TK + CK

VIS =~ mech + twod + threed + abst

MATH =~ arith + bmath + amath

SPEED =~ arithfast + table + cler + obj

MEM =~ sentmemory + wordmemory

TK =~ phsk + aerok + eleck + mechk + sportk + huntk + fishk

CK =~ litk + musk + sstudiesk + mathk + biok + sciattk + farmk + homek + artk + lawk + healthk + engk + archk + jourk + fork + militk + salesk + prack + clerick + biblek + colorsk + etiqk + outdk + photok + gamesk + theatk + foodk + misc

G ~ race

"

Still roughly the same number. I wonder what it would look like with a proper model.