Benjamins - Its all about IQ?

I have no idea how this got by all the control systems of academia and the tech industry.  My hesitation in posting about it is that it is a new substack and I have no awareness of the author, Zach Goldberg, supposedly a PhD candidate.  

How he could be such and publish such a disemboweling article to an article of faith in academia, I don't understand.

From Exposing the group disparities = discrimination fallacy by Zach Goldberg.  

“We have a hard time recognizing that racial discrimination is the sole cause of racial disparities in this country and in the world at large … When you truly believe that racial groups are equal, then you also believe that racial disparities must be the result of racial discrimination.”-Ibram X. Kendi, Stamped From the Beginning

The idea that persistent outcome disparities between racial groups necessarily reflect or result from discrimination rests on a narrow, misleading, and politically motivated portrayal of the empirical reality. In what follows, I introduce and briefly discuss a number of data points that call this narrative into question.

Goldberg appears to be using data from the American Community Survey for the five years 2015-2019.  Respondents to the Census survey self-report their primary ethnic/geographical/cultural origin.  Any ethnic group with fewer than 300 members is omitted.  ACS randomly selects approximately 3.5 million households for participation each year and about 95% of recipients respond.  

Goldberg also uses data from longitudinal surveys conducted by the Department of Labor.  Specifically, he is using data from 

NLSY79

The National Longitudinal Survey of Youth 1979 (NLSY79) is a sample of 12,686 men and women born during the years 1957 through 1964 and living in the United States when the survey began. Survey respondents were ages 14 to 22 when first interviewed in 1979. The U.S. Department of Labor selected the NLSY79 cohort to replicate the NLS of Young Women and the NLS of Young Men, which began in the 1960s. The NLSY79 also was designed to help researchers and policymakers evaluate the expanded employment and training programs for youths legislated by the 1977 amendments to the Comprehensive Employment and Training Act (CETA). Data are available for this cohort through 2014 when the 7,071 men and women in the sample were ages 49 to 58. Data from the 2016-2017 survey will be released in late 2018/early 2019. To supplement the main data collection, survey staff conducted special high school and transcript surveys. NLSY79 respondents also participated in a special administration of the Armed Services Vocational Aptitude Battery.

and from

NLSY97

The National Longitudinal Survey of Youth 1997 (NLSY97), the newest survey in the NLS program, is a sample of 8,984 young men and women born during the years 1980 through 1984 and living in the United States when first interviewed. Survey respondents were ages 12 to 17 when first interviewed in 1997. The U.S. Department of Labor selected the NLSY97 cohort to enable research on youths’ transition from school to the labor market and into adulthood. Data from the first 17 rounds of data collection are available to researchers. Round 17 consisted of 7,103 respondents, age 30- 36, and was completed in 2015-2016 with data made available in fall of 2017. In addition, survey staff conducted special high school and college transcript data collections to supplement the data on schooling provided by respondents. Many NLSY97 respondents also participated in a special administration of the computer-adaptive form of the Armed Services Vocational Aptitude Battery, and scores from that test are available for approximately 80 percent of sample members. 

These are large population sets, as close to random as possible and are extensively used by both government agencies and corporations for consequential and important administrative purposes as well as business and policy decisions.  This is good data.

Goldberg does not, in his post, elaborate on his methodology but this is pretty straightforward data analysis.  His findings are consistent with much of the literature of the past thirty years.  What is different is that he pulls it all together in one place.

From what I see, my greatest methodological concern is the degree to which people are able to accurately estimate their own heritage.  Outside genealogical circles, I would be surprised if most people could name more than their great-grandparents much less identify where they came from.  I would also be comfortable assuming the error rates on origin go up fairly dramatically the more generations back you go.  

At a continental level, of course there is probably reasonably high accuracy, but at a country level?  

And this is especially true for those of heritages with deep roots in America.  Dabbling in genealogy, I have a higher awareness than probably most.  And the great majority of my lines go back 8-12 generations, arriving in the US in the first half of the 17th century.  I could off the top of my head say that roughly 1/3rd of my genetic heritage is likely English, 1/3rd Scottish/Scotch Irish, and 1/3 Palatinate German with relatively small dabs of Dutch, French, Swiss, Danish, Welsh, Native American and Belgian.  I would probably be ballpark accurate.  If I spent a week doing the detailed research, I could probably get within ten feet of home base.

But that is by far the exception.  

Is that usefully accurate for the purposes of Goldberg's analysis, even if everyone could do that?  I am not certain.  

But if we assume that high accuracy of country of origin is not a major impediment to the analysis, what are his results?

I'll use the most surprising one to me because really, the whole piece is worth reading.

My operating assumption is that of course there is variability among population groups and that that is driven primarily by average differences in IQ and by average differences in cultural constructs (values and behaviors).  Among the latter, importance of family creation and sustenance, work ethic, valuation of education attainment, and risk tolerance, some examples of what I believe to be materially consequential cultural differences.  

If you control for IQ, sex, cultural values and age for any population, I believe the marketplace results in terms of income are going to be very similar.

Goldberg doesn't have the data to be that specific, but he is close. 

As shown in Figure 10, when controlling only for age and sex, US-born white panelists earn, on average, over $24,000 and $15,000 more than their black and Hispanic counterparts. When additionally controlling for cognitive ability, though, these disparities are all but eliminated.

Figure 10.

Click to enlarge.

Figure 11 presents data from the 1997 cohort (born 1980-1984) of the NLSY, as arrest histories were not measured for the 1979 cohort. In the left graph, which controls only for age, we see that US-born black and Hispanic male panelists were significantly more likely than their white and Asian counterparts to be arrested at least once. But, these differences all but evaporate when controlling for cognitive ability.

Figure 11.

Click to enlarge.

Figure 12 shows that this pattern extends to disparities in incarceration. Once again, controlling for age and cognitive ability, white, black, and Hispanic male panelists experience incarceration at statistically indistinguishable rates, while rates for Asian men remain significantly lower (apparently, one is better off Asian than white in our ‘white supremacist’ society).

Figure 12. 

Click to enlarge.

Figure 10 says that, contrary to my expectations, only IQ matters for income attainment.

Figure 11 says that, contrary to my expectations, only IQ matters for arrests. With a small exception for Asians.

Figure 12 says that, contrary to my expectations, only IQ matters for incarceration.  With a large exception for Asians.

Hmm.  I am comfortable to accord a heavy weight to IQ as determinative of outcomes, but 100%?  

Is there anyway to rescue culture as an influential variable?  Perhaps.  The marked differences for Asians is a clue.  Asian migration at volume is really only a phenomenon of the past two generations.  Goldberg excluded first generation participants but that might not be sufficient.  The very act of emigration is self-selecting and usually it takes 2-4 generations for full cultural assimilation.

Also, assuming that the great majority of Asian immigration is legal, then they are subject to a very heavy selection bias from admission criteria which heavily bias towards factors highly correlated to IQ (professional attainment, educational attainment, etc.).  Separate from the experience of immigration itself and time gap towards full cultural assimilation, there is also the factor that biologically IQ tends to revert to the mean.  Two high IQ parents are not guaranteed to have equally gifted children owing to reversion.

Asians in Goldbergs data may be anomalous simply because of recency.

Otherwise we are left with those striking graphs.  Once having properly controlled for age, sex, and IQ, everyone achieves the same income, the same probability of arrest and nearly the same probability of incarceration.

That is a huge impediment to the hypothesis of systemic racism and similar constructs.  In fact, if true, it is a huge endorsement that the US has come very close to fulfilling its Age of Enlightenment aspirations of all men being equal under the law in an environment of the rule of law and maximal practical individual freedom.  

If true, this is an enormous cause of celebration!

I am left with the discomfort over the absence of significant impact of cultural behaviors.  It is possible that cultural values are indeed still significant but that owing to multi-causal complexity, they only show up within the race groups.

In other words while controlling for age, sex and IQ, when you look at the dataset for White (and then separately Blacks and then Hispanics) you might still find that cultural values related to work ethic, familial establishment, time discounting, risk tolerance, valuation of education attainment do show up as determinative in distribution of outcomes.  

It could also be that cultural persistence is very high and that race identification is also highly correlated with the key cultural attributes determinative of good life outcomes.

We just can't know this from Goldberg's evidence here.  I am not willing to let go of cultural behaviors as a key driver of good life outcomes but I certainly need to asterisk that assumption.

Because what Goldberg's analysis from pretty high quality data suggests that age, sex, and IQ are the key forecasting variables for life outcomes AND that the US is living up to its ideals of everyone having an equal opportunity for success only constrained by those variables.  Not race.

If Goldberg's findings get any circulation they will be attacked with vigor because they go against a whole set of ideologies (Critical Race Theory, Social Justice Theory), a whole set of political philosophies (communism and socialism), against a political party (Democrats) and against almost the entirety of academia and mainstream media.  

If Goldberg, his data, and his analysis survives that assault, it will be a cause of huge celebration for those who love the ideals of America and finally will allow us to focus on policies which might make lives better rather than always doubling down on the will-o-wisps of Critical Race Theory, Social Justice Theory, Communism and Socialism.