March 26, 2017

Race and IQ: Mixed Populations Part 2

This article should be treated as a continuation of this article.

In the previous article we looked at IQ studies of populations and samples of mixed-race people compared to monoracial subjects within a similar environment. Degrees of similarity of environment vary. For example, comparing identified racial groups in Brazil, or comparing identified racial groups within an adoption study.

In this article, we’re going to be trying to look at finer gradients of admixture and how that impacts IQ and other standardized test scores.

Admixture in Mexican Districts

Using admixture data from Francisco Mauro Salzano and Andrés Moreno-Estrada, we know the racial admixture of 32 districts in Mexico.

We also have PISA scores for those same 32 districts, and Raven’s Progressive Matrices scores for 15 districts, and we can look at the average scores compared to the racial admixture of the districts.

You can look at the data for all 32 districts here, including the 15 that have Progressive Matrices scores and you can read Chuck from HumanVarieties’ full analysis for more in-depth anaylsis.

Progressive Matrices in Mexico

The Progressive Matrice data is available in 15 districts. Here I divided the districts into thirds and looked at the scores that way. You can look at the full data and divide it however you want though. No district was over 64% European. The African percentages in each district didn’t appear variable enough that it would throw off the results too much.

% European # of districts RPM 2002 RPM 2005 Avg.
40-64 5 11.41 11.94 11.68
29-39.9 5 10.85 11.43 11.14
0-28.99 5 10.35 11.43 10.89

PISA Scores in Mexico

PISA scores are available for 32 districts most of the time. Some of the years some districts are missing. Since I go down the list of districts in order of European admixture, and look at average each of the four groups of 8, when a district was missing, I just averaged the remaining 7. There was never more than one missing district from each group each year. You can see my ordering here

Districts by European % Average European Percent “Heredetarian Genetic IQ” 2003 PISA 2006 PISA 2009 PISA 2012 PISA
Top 8 54.25 94.51 394.69 419.19 426.6 426.2
9 to 16 41.05 92.9 410.5 416.19 424.4 423.19
17 to 24 28.24 91.39 401.86 413.86 425 423.5
Bottom 8 17.51 90.1 364.6 385.4 399.25 397.79

The most European districts were below the middle districts the first year of the PISA tests, but quickly took the top spots in the next 3 tests. The least European districts were at the bottom for all 4 tests.

The less European of the middle districts did better than the more European middle districts in 2009 and 2012, but it was very close.

The “Heredetarian Genetic IQ” is simply the European admixture times 100, and everything else times 88. This is a bit rough but I couldn’t be arsed to tabulate the effects of the African proportions of each group, and honestly it’s not going to change the results much.

This is not post-hoc. The “genetic IQ” of Africans being 84, Europeans 100 and Amerindians 90 is a very common estimate.

It’s visually striking how closely the scores line up with the “heredetarian predicted IQ”, even though the expectations for the differences in admixture are rather small.

Admixture in Colombian Districts

Each year Colombia gives 11th graders an exit exam, called the SABER test. These results are available by district.

We also have admixture data by district. And so we can compare racial admixture of districts in Colombia with each districts’ SABER scores.

There are 33 districts in Colombia, one of which is Bogota, the capital. I first listed each district by European admixture, then lumped them into groups of 8 in order of European admixture, except Bogota which is presented separately.

Not only is Bogota a conspicuous result, but it makes the number of districts in Columbia odd. The analysis on which this is based is from Chuck at humanvarieties, which presented the scores of each district in terms of distance in standard deviations from the mean (d).

I simply found the raw means and standard deviations from the 2014 and 2012 SABER scores and converted the (d) values back into real scores. You can see Chuck’s excel file and what my excel file looked like when I finished working with Chuck’s file and the raw SABER scores.

Admixture data was found by combining data from Salzano-Sans (2014)Ruiz-Linares (2014) and Rodriguez-Palau (2007) as described by the original humanvarieties post.

Districts by European Admixture / Bogota Average Score of Districts “Heredetarian Genetic IQ” European %
Top 8 46.51 93.655 47.125
9 to 16 45.58 93.445 45.375
17 to 24 45.03 91.945 32.875
Bottom 8 44.05 90.985 24.875
Bogota 51.95 93.64 47

This is another example of racial admixture in regions within a country predicting variation within a country. In Latin America, the population is so heavily admixed that we’re not necessarily dealing with “47% whites” and “45% amerinds”. We’re dealing with a bunch of light brown people who have varying degrees of European, Amerindian and some African admixture within each person.

Like with Mexico, the genetic differences that should produce very small differences in IQ, produce corresponding differences in scores on standardized tests. The “Genetic IQ” difference of the top 8 and the bottom 8 is 2.67, while the difference in SABER score is 2.46 between those groups.

This is a higher level of precision than I would have expected in terms of racial admixture predicting relative standardized test scores within a country.

Bogota, the capitol city, is an outlier. If it were treated normally it would fit in the most European group of countries and would actually widen the gap between the most European districts and everyone else.

Admixture in Brazil

For Brazil, can look at PISA scores by region, and compare that to the admixture of the regions.

PISA scores are available for each state here, and the admixture for each region is here, and the states in each region is available here, just so you know I didn’t find the data and then group the states in such a way as to confirm a heredetarian theory. The states and regions were pre-defined.

I have not been able to find any admixture data for each state. If I did I would have grouped the states by European % like I did with Mexico and Colombia. You can see what my excel file looked like when I was finished here.

Region % Euro % Amerindian % African “Genetic IQ

Prediction”

Average of state PISA scores
South 77 11 12 96.98 417.3
Southeast 67 10 23 95.32 413.3
Central-West 64 12 24 94.96 404.6
Northeast 58 15 27 94.18 380.9
North 52 32 16 94.24 376.6

“Genetic IQ prediction” is assuming that the European admixture should be 100, Amerindian should be 90, African should be 84. Don’t take this too seriously, as this is just a rough idea of what the scores should be relative to each other.

Perhaps my estimate of 90 for the Amerindian portion is too high for the “Genetic IQ Prediction”, as the North underperforms the Northeast on PISA when my prediction is that it should be slightly higher; or perhaps it is a matter of an environmental disparity in the North, which seems intuitively plausible since the North is literally the Amazon rainforest.

But it’s not a big deal, as the heredetarian prediction is mostly upheld; if anything the exceptionalism of whiteness may be more important than perhaps even I(!) suspected, as it really is the white percentage driving the PISA scores.

Admixture and Average IQ of Latin American Countries

Using admixture data from Humanvarieties and IQ data from table 2.1 of Lynn 2012, we can compare the European admixture of 29 countries in the Americas and their correlation with IQ:

Excel file here

I don’t think this is as compelling as the admixture data within Mexico and Colombia, as there are obvious plausible environmental explanations for between-country differences in IQ. But it still complicates any environmental explanation.

Even when considering the inconsistency of the tests that will introduce noise and thus dampen any correlation, the huge economic disparities between countries, there is still a correlation of 0.4146.

Skin Color

Skin color is a fairly reliable way of determining admixture in a person, especially when we know things already about a specific group, like blacks in the United States.

And so we can look at how dark a Black person is in the United States as a rough indication of their admixture. Obviously some light-skinned blacks have more African admixture than some darker-skinned blacks, but there will also be times when the light-skinned-blacks are even less African than their already light skin would suggest.

Which is why we take groups of people – a lighter group of 20 blacks will probably have more European admixture than a darker group of 20 blacks – and the bigger the groups, the more confident we can be of this. And so we can use skin color as a “rough and ready” proxy for European admixture in blacks in the US.

GSS 2012 PVT Scores

The General Social Survey includes a picture vocabulary test and the examiner would write down the skin color of the person they were examining. These were the results of that:

Color PVT Score Number
White 6.22 947
Light-Black 5.67 69
All Blacks 5.13 191
Dark-Black 4.69 99

The light Black group, which with 69 people is enough to say that they are certainly more European by admixture on average, scored higher than all Blacks and dark Blacks.

NLSY 1997 PVT Scores

The 1997 National Longitudinal Survey of Youth included a picture-vocabulary test and score. Examiners also rated the respondent by skin color sometimes. I’m not sure how often or in what circumstances this was asked, since apparently whites were only asked it 3 times in each wave.

These were the results of that:

Color Wave III PVT Number Wave I PVT Number
Black 89.64 243 91.2 246
Dark Brown 88.92 249 92.09 287
Medium Brown 96.05 294 95.56 287
Light Brown 98.31 109 97.47 111
White 110.33 3 105.33 3

The results are inline with the previous analysis.

GSS 1982

A PVT and skin-color question combination also existed in the 1982 General Social Survey, but in this one the examiner asked the subject what the color of his skin was. This was the result of that:

Color Average PVT score Number
Very Dark 4.43 42
Dark 4.01 106
Medium 4.79 208
Light 5.25 67
Very Light 5 14
No Answer 5.07 14

Odd result is that the very dark scored higher than dark. One possibility is that the very dark are African immigrants and possibly more intelligent than US Blacks of the same skin color.

Skin Color by Country

Templer and Arikawa looked at skin color ratings and average IQ of 55 countries. These are their results aggregated from Table 2 in their paper:

Skin Color Rating Average IQ Number of Countries
0 to 1.99 98.4 18
2 to 3.99 95.4 15
4 to 5.99 84.9 8
6 69.3 14

Correlation between national IQ and skin color was -0.90. By comparison, real GDP per capita correlated with IQ at 0.74.

Which means that the negative relation between skin color and national IQ was STRONGER than the positive relation between real GDP per capita and IQ.

And this is true even when you consider that Japan, China, Korea and Taiwan all have a skin color rating of 2.00 and have IQs above 100. I.e. East Asians are (according to this data) slightly darker than Europeans but have higher IQs.

Admixture and Socioeconomic Status

This review found that across 14 studies that measured African ancestry 13 found that the more African a person’s genome was the lower their socioeconomic status tended to be. The 14th study found no association.

Across 13 studies that measured European ancestry 11 found a positive association between how European a person’s genome was and how high their socioeconomic status was. The other two studies found no association.

There was only 1 study which looked at East Asian ancestry and it was found to be positively associated with how educated someone is.

There were also 15 studies which looked at how Native American ancestry correlated with success among Hispanics. 13 found a negative association. The other 2 found no association.

Image source

Shuey Studies

The purpose of presenting the “Shuey studies” here is not that they will be particularly persuasive. It is simply to provide a view to the past, and show that there really is nothing new under the sun. They also show just how long-going and consistent this admixture data has been.

A.C. Strong 1913

Color % Retarded % Advanced Number
Light 44.2 11.6 43
Medium 31.1 6.6 45
Dark 14.4 8.8 34

Strong’s results suggest that the mean intelligence of darker blacks is higher than that of lighter blacks.

Ferguson 1919

4 studies looking at samples of the Army Alpha and Beta tests used to screen troops prior to WW1. The tests were conducted in 1917.

1st Study – 657 subjects, specific subject breakdown not available.

Color Median Score
Light 51
Dark 40

2nd Study – 667 subjects, specific subject breakdown not available.

Color Median Score
Yellow 59
Brown 45
Black 39

3rd Study – 344 subjects, specific subject breakdown not available, score breakdown not available. Description reads “Dark scored 60% of Light”.

4th Study – 727 subjects, illiterates, specific subject breakdown not available, score breakdown not available. Description reads “Dark scored 80% of Light”.

Koch 1926

Sample from Austin, El Paso and San Antonio Texas. Number of subjects available for each test, no skin color breakdown. Says, “Judgments made on faces with use of color card.”

MMM = Myers Mental Measure

NIT = National Intelligence Test

Color MMM NIT Det P-C
Light 22.7 58.2 39.5 37.2
Intermediate 19.9 51.9 35.8 40.7
Dark 21.2 53.1 35.3 39.3
Number 613 246 132 87

Author notes “No consistency in ranking of dark and intermediate”

Herskovits 1926

Herskovits took correlates of negroid features with scores on the Thorndike College Entrance Exam, 115 subjects.

Criteria Correlation w/ test score
Nose width 0.01
Lip thickness -0.2
Black element (pigmentation) -0.14
White element (pigmentation) 0.17

Author notes, “Lighter Negroes given preferment within race, making dark show up to ‘ill advantage on tests”. It seems like Herskovits presumed an environmental / discrimination hypothesis.

Davenport 1928

Davenport gives qualitative descriptions of scores on 5 tests. Says of method “Attempted to select whites, browns and blacks from same social stratum. All 3 groups from several towns in Jamaica. Some whites in addition from Grand Cayman Island.”

Subjects were all adults.

Cube – Whites best, browns intermediate

Know – whites best, browns intermediate

Manikin – Whites best, browns intermediate but more failure than blacks

Aplha – Whites best, browns intermediate

Draw-a-man – Whites best, others same

Author comments “Browns perhaps were more highly selected than others. Browns more variable and generally made somewhat better scores than blacks.”

Klineberg 1928 – West Virginia

Correlation of negroid features with scores on the Pintner-Paterson test. 139 Subjects gathered by “House-to-house canvass of villages. Skin color measured by Milton-Bradley color top.” Tests done in northeastern West Virginia.

Feature Correlation with test score
Nose width -0.06
Lip thickness -0.1
Black pigment -0.12

Author notes, “On whole, no definite evidence for inverse relation between test scores and degree of Negro blood indicated”.

In my opinion Klineberg is misinterpreting his own data.

Klineberg 1928 – New York City

Subjects were 200 black “boys from PS 139, divided into 4 groups using Ferguson’s technique”.

Group Score
Less Negroid 99
Most Negroid 97

Klineberg misinterprets his results again saying, “No correspondence between Negro blood, as judged by inspection and general Negro appearance and test results.”

Young 1929

Subjects were from Baton Rouge and Lake Charles Louisiana, “all 9 and 10 year olds in 3rd grade and above. With aid of teachers and principals, subjects were divided into darker and lighter.”

Obvious methodological problem already is that it doesn’t count students held back, which may be darker or lighter (probably darker) than those on track.

9 year olds Median Mean
Light 35 42
Dark 35 30
10 year olds
Light 45 46
Dark 35 41

Peterson 1929 – New York City

Peterson looked at 75 boys from an elementary and a junior high school in New York City. It was an another correlation between IQ (Yerkes Pre-Adolescent Point Scale) and Negroid features:

Trait Correlation with test score
Nose width -0.11
Lip thickness 0.07
Ear height -0.15
Interpupillary Span 0.01

The author notes, “Only slight evidence that these Negro traits correlate negatively with intelligence. Possibly skin color may be best criterion of degree of Negro blood.”

Peterson 1929 – Nashville Tennessee, Chicago Illinois

Peterson describes the subject selection methods:

Nashville – “All 12 year olds in 2 elementary schools and 9 from a junior high school.” N = 83.

Chicago – “From parks and playgrounds in the summer” N = 75.

Results are combined. Two tests were used: Stanford-Binet 1916, Myers Mental Measure.

Skin Color Stanford-Binet MMM
Light 17 35.8
Dark 14.2 25.8

Correlations with lightness:

Stanford-Binet 0.18
Myers Mental Measure 0.3

Jenkins 1936 comparison to Herskovits’

Jenkins was an analysis of 63 gifted students in Chicago. Results are reported as percentage with IQs above 125. This was compared to 1,551 cases reported by Herskovits.

Jenkins % above IQ 125 Herskovits % above IQ 125
Full Negro 22.2 28.3
3/4 Negro 46.1 31.7
Half Negro 15.9 25.2
1/4 Negro 15.9 14.8

Jenkins notes “Chicago percentages strikingly similar to those of Herskovits for general population. These superior children are not atypical in racial composition.”

Tanser 1939 – Kent County, Ontario

Method described as “Identification on basis of information given by teachers and local residents. Some of them colored.” Test used was Pintner-Cunningham, 54 subjects tested.

Group IQ
Mixed-Bloods 86.4
Full-Bloods 79.1

Bruce 1940

Study from rural Virginia. Used Kuhlmann-Anderson and Stanford-Binet 1916 tests. Selection method states “Representative sample from original 432 by taking every 5th in IQ order. Examiner rated each subject according to skin color.”

Color Median Mean
Light 76 78.7
Dark 73 74.35

Author notes “41% of dark equals or surpasses median of light Negroes.”

Tanser 1941

Same methods as in 1939, also in Kent County, Ontario.

Group NIT Pintner Non-Language Pintner-Paterson
Mixed 91.7 97 96.7
Full Negro 87 94.5 87.8
Number 103 102 162

Codwell 1947

Subject selection described as “480 of 680 boys in P. Wheatley High School. Age range of the 480 was 11-18. Test used was Otis. Done in Houston, Texas.

Group IQ Average Number
Strong Evidence of White 91.9 94
Intermediate 90.9 210
Dominantly Negroid 87 176

Author notes, “Except for capacity to learn new motor skills (where SEW’s were superior to DN’s) DN’s excelled in most factors of motor function and were most inferior in intelligence.”

Grinder – 1964

Test used was the draw-a-man test in Jamaica.

“Subjects drawn from approximate 50 schools throughout island after schools had been rated on basis of social class milieu. In the main, every child “who was in middle child-hood” was used as a subject, except in large schools where children tested was about 35. Color rating assigned to each subject based on observation of skin color, hair texture and nose breadth.”

All Subjects Low Score High Score Total
Light 29 (27.35%) 77 (72.64%) 106
Mixed 89 (45.18%) 108 (54.82%) 197
Dark 387 (60.66%) 251 (39.34%) 638
Middle Class
Light 23 (26.44%) 64 (73.56%) 87
Mixed 53 (41.09%) 76 (58.91%) 129
Dark 208 (56.83%) 158 (43.17%) 366

Conclusion

The totality of evidence is most easily explained by a heredetarian standpoint on race and IQ. Certainly one could come up with a string of environmental hypothesis and theories to explain admixture data, but what I find with the average person is that when they see this data, and a seasoned environmentalist starts putting forth all sorts of exotic environmental explanations for things like this, the layman thinks “I didn’t sign up for this”.

And it’s true. The layman has no clue just how much information there is on this topic, and just how complicated environmental explanations have to become to account for it.

Of course you can always create a web of environmental effects that accounts for all of the observed data. You can “control for” all sorts of things and see that IQ differences all vanish. Sure, we could also control for how confident they are while taking the IQ test and see a few points from the gap vanish as well.

Normally we would say to obviously allow both genetics and environment to explain varying parts of racial disparities, depending on where and when we’re talking about. There is no “genetic determinism” to speak of, and as far as I know never was. It’s a battle between one view of the world that allows genetics, and another that doesn’t allow genetics – either in certain instances or sometimes in total.

One view says that environment and genetics explains racial disparities in IQ and IQ-like things, and another says only environment can explain race gaps. Normally these very same people allow for genetics to distinguish individuals or even groups of people SO LONG AS those groups don’t correspond to races.

And once you recognize this, the denial of heredetarianism is revealed as baldly political.

Facebook Comments
  • Emil Kirkegaard

    “Even when considering the inconsistency of the tests that will
    introduce noise and thus dampen any correlation, the huge economic
    disparities between countries, there is still a correlation of 0.4146.”

    No, your plot has r^2. The correlation for those data is .64.

    This topic was covered at fairly extreme detail in our 120 page target article. The reply to the critics covers some more data (about 40 pages). The overall results are around .70. This is true no matter if one pools all the subnational data or not.

    https://www.researchgate.net/publication/298214364_Admixture_in_the_Americas_Regional_and_National_Differences

    We are expanding to more countries. The pattern also holds in Argentina at the first-level division level, n=24. It does not hold in Peru, which is not surprising because there is not actually very much variation in ancestry between the units.