ArticlePDF Available

Abstract and Figures

Structural racism has often been invoked to explain observed disparities in social outcomes, such as in educational attainment and income, among different American racial/ethnic groups. Theorists of structural racism typically argue that racial categories are socially constructed and do not correspond with genetic ancestry; additionally, they argue that social outcome differences are a result of discriminatory social norms, policies, and laws that adversely affect members of non-White race/ethnic groups. Since the examples of social norms and policies commonly provided target individuals based on socially-defined race/ethnicity, and not on genetic ancestry, a logical inference is that social disparities will be related to socially-defined race/ethnicity independent of genetically-identified continental ancestry. In order to evaluate this hypothesis, we employ admixture-regression analysis and examine the independent influences of socially-identified race/ethnicity and genetically-defined ancestry on the educational attainment and income of parents, using data from a large sample of US children. Our study focuses on self-identified Whites, Blacks, Hispanics, and East Asians in the United States. Analyses generally show that the association between socially-identified race/ethnicity and outcomes is confounded by genetic ancestry and that non-White race/ethnicity is unrelated to worse outcomes when controlling for genetic ancestry. For example, conditioned on European genetic ancestry, Americans socially-identified as Black and as Hispanic exhibit equivalent or better social outcomes in both education and income as compared to non-Hispanic Whites. These results are seemingly incongruent with the notion that social outcome differences are due to social policy, norms, and practices which adversely affect individuals primarily based on socially-constructed group status.
Content may be subject to copyright.
Submitted: 16th of June 2023 DOI: 10.26775/OP.2023.09.11
Published: 11th of September 2023 ISSN: 2597-324X
Income and Education Disparities Track Genetic Ancestry
Meng HuEmil O. W. KirkegaardJohn G. R. Fuerst
OpenPsych
Abstract
Structural racism has often been invoked to explain observed disparities in social outcomes, such as in educational
attainment and income, among dierent American racial/ethnic groups. Theorists of structural racism typically argue
that racial categories are socially constructed and do not correspond with genetic ancestry; additionally, they argue that
social outcome dierences are a result of discriminatory social norms, policies, and laws that adversely aect members
of non-White race/ethnic groups. Since the examples of social norms and policies commonly provided target individuals
based on socially-defined race/ethnicity, and not on genetic ancestry, a logical inference is that social disparities will be
related to socially-defined race/ethnicity independent of genetically-identified continental ancestry. In order to evaluate
this hypothesis, we employ admixture-regression analysis and examine the independent influences of socially-identified
race/ethnicity and genetically-defined ancestry on the educational attainment and income of parents, using data from a
large sample of US children. Our study focuses on self-identified Whites, Blacks, Hispanics, and East Asians in the United
States. Analyses generally show that the association between socially-identified race/ethnicity and outcomes is confounded
by genetic ancestry and that non-White race/ethnicity is unrelated to worse outcomes when controlling for genetic ancestry.
For example, conditioned on European genetic ancestry, Americans socially-identified as Black and as Hispanic exhibit
equivalent or better social outcomes in both education and income as compared to non-Hispanic Whites. These results are
seemingly incongruent with the notion that social outcome dierences are due to social policy, norms, and practices which
adversely aect individuals primarily based on socially-constructed group status.
Keywords: Structural racism, education attainment, income, race, genetic ancestry
1 Introduction
In the United States (USA), there are well-documented socioeconomic status (SES) dierences between socially
(self and other)-defined race/ethnic groups. These groups, as defined by the USA Oce of Management and
Budget, include Hispanic/Latin American ethnicity and White, Black, American Indian, Asian, and Pacific
Islander race or origin. Adults who self-identify as either Hispanic or as non-Hispanic Black, American Indian,
and Pacific Islander typically have lower educational attainment and income than those who identify as non-
Hispanic White. Conversely, those who identify as Asian have higher average educational attainment and
income than Whites.
Racial/ethnic SES inequalities are often attributed to structural racism. Jones (2002, p. 10) gives one frequently
cited definition of structural racism: “structures, policies, practices, and norms resulting in dierential access
to the goods, services, and opportunities of society by ‘race’.” Braveman et al. (2022) note, additionally, that
structural racism encompasses laws, policies, institutional practices, and entrenched social norms, as distinct
from individual acts of discrimination.1
Independent Researcher, Email: mh19870410@gmail.com
Ulster Institute for Social Research, London, UK
Cleveland State University, University of Maryland Global Campus, Bioinformatics
1
It should be noted that public policies don’t have to be designed to aect the socio-economic outcomes of ethnic minorities. Policies
which tend to increase, intentionally or not, SES inequalities will aect minorities more negatively since lower SES families are typically
minorities (Noguera & Alicea,2020).
1
Published: 11th of September 2023 OpenPsych
According to Jones (2002) “race is only based on a few phenotypic-related genes, not global genetic ancestry,
since the few genes that determine skin color, hair texture, and facial features are not informative about other
aspects of the genotype at the individual level. Advocates of the structural racism hypothesis frequently
emphasize that race/ethnicity categorization “reflects neither biological nor cultural dierences” and that race
is “often conflated erroneously with biology and ancestry” (Adkins-Jackson et al.,2022, p. 540), that “race is a
social construct and is distinct from ethnicity, genetic ancestry or biology” (O’Reilly,2020, p. 2), that “social
races bear little relationship to the reality of human biological diversity” (Smedley & Smedley,2005, p. 22) and
that race “is a social construct with no biological basis and stems from White supremacy” (Haeny et al.,2021,
p. 889). While the phrase “race is a social construct” can have a range of meanings, a popular one, given by the
American Sociological Association (2003), is that race is a social invention that changes as political, economic,
and historical contexts change”; this social invention is said to be important because “social and economic life is
organized, in part, around race as a social construct.
Part of the reason for the emphasis on the socially constructed aspects of race is rhetorical, as it is believed
that “a long history of work has attempted to link race to genetics in order to justify racial discrimination and
race-based social stratification (Guo et al.,2014, p. 2337). For example, Suyemoto et al. (2022, p. 78) argue that
the myth that race is “genetic or biological” is “important to unravel because it’s the foundation of justifying
racial hierarchies and therefore racism. Additionally, the assertion, “race is a social construct” is used to argue,
fallaciously, that dierences between socially-defined races are necessarily social or environmental in origin
(e.g., Robbins et al. 2022).
Nonetheless, there are two substantive reasons for the adoption of social constructivism in context to discussion
of Oce of Management and Budget-defined race/ethnic groups and social inequalities.
First, laws, policies, institutional practices, and entrenched social norms must aect members of a group
delineated in specific ways to aect social outcome dierences. In the USA, race/ethnic groups are delineated
based on “social and cultural characteristics” (Oce of Management and Budget,1997) in addition to, and not
strictly based on, genetic ancestry.
Second, structural racism is typically conceptualized as including many specific laws, policies, institutional
practices, and social norms which did not discriminate or aect individuals strictly based on ancestry. Examples
frequently given include: voter suppression of Blacks, political gerrymandering, predatory financial services,
mass incarceration, police violence, sending American Indian children to boarding schools (Braveman et al.,
2022), slavery, black code, Jim Crow laws, segregated housing, redlining (Bailey et al.,2021;Erikson et al.,
2022). These policies and practices targeted individuals based on socially-defined race, not strictly based on
genetic ancestry.
For example, Jim Crow impacted all individuals, in the Southern states, with a certain degree of recognizable
African American ancestry. The impact of this policy was not apportioned according to genetic ancestry. Given
that these policies and practices targeted individuals based on socially-defined race/ethnicity, not strictly genetic
ancestry, it makes sense that proponents of the structural racism hypothesis emphasize the social construction
of race. Discussing this point, Gichoya et al. (2022, p. 8) observe, “in the context of racial discrimination and
bias, the vector of harm is not genetic ancestry but the social and cultural construct that of racial identity...
biased decisions are not informed by genetic ancestry information, which is not directly available to medical
decision makers in almost any plausible scenario.”.
Strong social constructionist claims about self-identified racial categories, to the eect that they do not re-
flect biological dierences or that they provide information only about genes related to conspicuous race-
related phenotypes, are probably false. In the USA, at least, there is a moderate to high concordance between
self/parental-identified race/ethnicity and continental-level lineage (Fang et al.,2019;Kirkegaard,2021). Since
continental-level populations - such as Sub-Saharan African, West Eurasian, East Asian, and Amerindian - are
dierentiated with respect to many morphological and physiological traits (Brues,1990), socially-identified
race can also be predicted from, for example, medical imaging data (Gichoya et al.,2022;Kirkegaard & Fuerst,
2023). Nonetheless, self-identified race, let alone ethnicity, in the USA, is also socio-politically constructed in
that classifications are based on complex political and cultural considerations independent of continental-level
genetic ancestry. For example: “Hispanic” is not defined by genetic or biogeographic ancestry, but refers to
a “person of Cuban, Mexican, Puerto Rican, South or Central American, or other Spanish culture or origin,
regardless of race” (Oce of Management and Budget,1997); “Black,” while defined as “a person having origins
in any of the black racial groups of Africa,” includes individuals with a proportion of Black African ancestry
that can range from 2 % to 100 % (Bryc et al.,2015).
2
Published: 11th of September 2023 OpenPsych
Given the focus on the socially-constructed aspects of race/ethnicity in relation to specific laws, policies, and
norms, one obvious scientific prediction is that socioeconomic dierences will relate to socially-constructed
aspects of race/ethnicity substantially above and beyond continental-genetic ancestry. This seems to be what,
for example, Smedley & Smedley (2005, p. 22) predict when they state that although the term race is not useful
as a biological construct... social race remains a significant predictor of which groups have greater access to
societal goods and resources and which groups face barriers - both historically and in the contemporary context
- to full inclusion. Or, what Herd et al. (2021, p. 420) mean when they argue that it would be illogical to focus
on “racial group genetic variation and also argue that “[g]iven what is known about ancestry, it makes little
sense to examine racial group genetic variation, especially in the American context. People share a social and
political category, not a biological category...”.
One method of assessing whether the socially constructed aspects of race/ethnicity have an ancestry-independent
impact on outcomes is to use admixture-regression designs (Connor & Fuerst,2022;Fuerst,2021;Kirkegaard et
al.,2019;Lasker et al.,2019). In these designs, recently-admixed populations are treated as natural experiments,
and genetic admixture is used to disentangle various cultural, environmental, and genetic factors contributing
to variation in a trait. Self-identified racial identity is treated as a “surrogate to an array of social, cultural,
behavioral, and environmental variables” (Fang et al.,2019, p. 764) and included in the regression models
alongside genetic ancestry variables. These designs can disentangle the eects of factors related to social
racial/ethnic identity from the eects of factors related to genetic ancestry. In this paper, we test the hypothesis
that self-identified race/ethnicity (this being an index of social race), is strongly related to two measures of SES
income and education independent of continental-level genetic ancestry. We interpret the structural racism
hypothesis, at least as commonly presented, as predicting that it will be since this is what many proponents of
this hypothesis explicitly state.
2 Method
2.1 Data and Sample
The Adolescent Brain Cognitive Development (ABCD) study is a joint long-term initiative that includes 21
research sites throughout the US, focused on examining brain development and child health to investigate the
psychological and neurobiological foundations of human growth. At baseline, around 11,000 children aged
9-10 years were sampled, using a probability-based sampling approach aimed at establishing a comprehensive
and inclusive sample of US children within that age group. In this current investigation, we utilized the ABCD
3.01 baseline data.
We used parents’ variables, with the exception of the child’s genetic ancestry. As we only had access to the
children’s genetic ancestry, we limited the sample to cases where both parents were the biological parents of the
child. In families with multiple children, we used the genetic ancestry estimates of the first biological child.
We then selected cases with no missing data in our variables, which included income, educational attainment,
parents’ age, and the child’s genetic ancestry. Additionally, we excluded parents who identified as Pacific
Islander, South Asian, or Other Asian so as to focus on individuals primarily of European, African, East Asian,
and Amerindian ancestry. These restrictions resulted in a final sample of 5,073 parent dyads.
2.2 Variables
Several variables were computed for the purpose of the present study. The list of these variables is provided
below.
2.2.1 Genetic ancestry
Genetic ancestry of the biological children was computed using the National Institutes of Health (NIH) estimates.
The process of imputing and genotyping was carried out by the ABCD Research Consortium. To determine
genetic ancestry, the ABCD Research Consortium utilized a k = 4 solution (European, African, Amerindian, and
East Asian). The researchers used the 1000 Genomes populations as reference samples and fastStructure as the
algorithm for this purpose (Hatton,2018). We divided the ancestry estimates by the sum of European, African,
Amerindian, and East Asian ancestries so that the sum of the four ancestries is equal to 1. For this reason, we
have to drop one ancestry variable in the regression equation for the estimation of the other variables. We
selected European ancestry as the reference category.
3
Published: 11th of September 2023 OpenPsych
2.2.2 Educational attainment and income
Two outcome variables, parental education attainment and parental income, were computed for analyses. The
parental education variable was calculated as the average of both parents’ education level, which originally had
22 categories. To create an interval variable, we recoded the variable to have 11 categories, where 0 represented
never attended school, 1-12 represented 1st-12th grade, 12 represented High school graduate, 14 represented
Some college or an Associate degree (Occupational or Academic Program), 16 represented Bachelor’s degree,
and 18 represented Professional School or Doctoral degree.
Parental income was an interval variable that reflected the total combined family income of the parents in the
past 12 months, which was originally coded as 1-10. The categories were recoded as follows: 1 represented
Less than $5,000, 2 represented $5,000 through $11,999, 3 represented $12,000 through $15,999, 4 represented
$16,000 through $24,999, 5 represented $25,000 through $34,999, 6 represented $35,000 through $49,999, 7
represented $50,000 through $74,999, 8 represented $75,000 through $99,999, 9 represented $100,000 through
$199,999, and 10 represented $200,000 or more.
Detailed descriptions of the educational and income variables are provided by Fuerst et al. (2021) and Fuerst et
al. (2023) who used the same coding scheme.
Both the education attainment and income variables were centered around the mean and standardized. However,
some extreme values (i.e., lower than -3 standard deviations (SDs) below the mean) were identified in both
variables. To retain these cases while minimizing the impact of outliers, we winsorized the data using a 3
standard deviation threshold.
2.2.3 Age
Age represents the age of the responding parent. For the regression analyses, the age variable was mean-centered
and standardized.
2.2.4 Immigrant status and English
In the survey, parents were asked whether anyone in the child’s family, including maternal or paternal
grandparents, was born outside of the United States, and this variable was assigned a value of "1" for "Yes"
and "0" otherwise. Additionally, the parent who provided the response was asked if their native language was
English, and this variable was also coded as "1" for "Yes" and "0" otherwise.
2.2.5 Self-identified race/ethnicity
The responding parent was asked 18 questions about his or her race/ethnicity but not the race/ethnicity of his
or her spouse. Based on these responses, we created six dummy-coded variables: Hispanic ethnicity and White,
Black, East Asian, Native American, and Other race. The East Asian category included individuals identifying as
Chinese, Japanese, Korean, Filipino and Vietnamese, while the Native American category included individuals
identifying as American Indian and Alaska Native. Because the race of the nonresponding parent is unknown,
we supplied a supplementary analysis by creating parallel sets of race/ethnicity dummy variables based on
the biological child’s race/ethnicity, as indicated by the parents. If the parents are intermarried, this would be
reflected in the child belonging to two race categories.
2.2.6 State racism and xenophobia
ABCD calculated state-level indicators of both racism and anti-immigrant bias (xenophobia) for the 18 states
in which the recruitment sites were located. These were based on both implicit bias measures and state-level
structural variables. The two indicators correlated at r= .34 (p< 0.05, N= 5,073). Both variables were centered
around the mean and standardized. These variables were only used in supplementary analyses.
4
Published: 11th of September 2023 OpenPsych
2.3 Description of Analyses
2.3.1 Main analyses
To examine the eect of race/ethnic identity on income and education we ran a series of admixture regressions.
We used a multilevel regression model, specifying research sites as a random eect and the following fixed
eects: dummy coded race/ethnicity (Black, Native American, Hispanic, Other Race), native English fluency,
genetic ancestry (African, Amerindian and East Asian), and finally parent’s age. Both White racial identity
and European ancestry are used as benchmark variables and the associated variables are dropped from the
regression models. Because we are only interested in the fixed eects estimates, Full Maximum Likelihood
(FML), instead of Restricted Maximum Likelihood (REML), is used as the estimation method. FML produces
more accurate fixed eects whereas REML produces more accurate random eects. Regression analyses are
carried out using lme4 package for R (Bates et al.,2009).
2.3.2 Supplementary analyses
In order to supplement our main analyses and further explore the robustness of our findings, we conducted
several additional analyses. Firstly, as a robustness check, we reran the analyses excluding all cases with values
of education and income 3 standard deviations (SDs) or more below the mean. This was done in order to ensure
that our results were not primarily driven by the extremely low values of education and income. Secondly, we
used an alternative approach to determine race/ethnicity. Specifically, we used the children’s race/ethnicity
as reported by the responding parent, instead of the responding parent’s race/ethnicity. This approach was
taken because the parent-reported race/ethnicity of the child may provide a better representation of the average
race/ethnicity identity of both biological parents, since the race/ethnicity of the nonresponding parent is
unknown. In a third set of analyses, we restricted the sample to families with at least one biological parent (as
opposed to having strictly two biological parents), which yielded a larger sample size due to the higher rate of
single parenthood among Blacks and Hispanics (N= 7,652). In a fourth set of analyses, we added state-level
indicators of racism and anti-immigrant bias to the regression models to explore the potential impact of these
variables on our findings. A fifth set of analyses were conducted using weighted regressions with the survey
package (“Package ‘survey”’,2020). This approach was used to take into account selection bias, with weights
based on the propensity-based weight of children provided by the ABCD. It is important to note that we did
not use these weight variables in our main analysis because we primarily used parents’ variables. Finally, a
sixth set of analyses was conducted within the combined Hispanics, Black, and Native American subsamples.
This approach was taken to explore potential dierences within disadvantaged non-White groups.
2.4 Data
The complete data set is available to qualified researchers at: https://nda.nih.gov/abcd
3 Result
3.1 Descriptive statistics
Table 1presents the descriptive statistics for the variables used in our study. In this table, age is presented
in the unstandardized form for ease of interpretation. Black, Hispanic, and Native American self-identified
race/ethnicity are associated with lower levels of income and educational attainment, while White and East
Asian self-identified race/ethnicity are associated with higher levels. It is noteworthy that the Hispanic and
East Asian groups are characterized by a higher percentage of immigrant families and also low rates of English
fluency. To note, the total case number in Table 1(N= 5,162) is larger than the dyads (N= 5,073) because some
non-Hispanic individuals self-identified as more than one race. This overlap does not impact our regression
analyses because in these analyses we control for race dummy variables. The high level of European admixture
among East Asian respondents reflects the relatively large number of White-East Asian couples in this sample.
5
Published: 11th of September 2023 OpenPsych
Table 1: Variable Means (and Standard deviations) by the Race/Ethnicity of The Responding Parent.
non-
Hispanic
White
non-
Hispanic
Black
non-
Hispanic
East Asian
non-
Hispanic
Native Amer-
ican
Hispanic
Age 42.51 (5.41) 40.34 (6.49) 43.79 (5.10) 41.37 (6.64) 40.57 (6.14)
Education 0.20 (0.80) -0.53 (0.93) 0.38 (0.68) -0.35 (1.02) -0.82 (1.21)
Income 0.20 (0.75) -0.67 (1.22) 0.37 (0.71) -0.31 (1.11) -0.74 (1.20)
English % 0.97 (0.18) 0.93 (0.26) 0.48 (0.50) 0.99 (0.12) 0.28 (0.45)
Immigrant % 0.22 (0.41) 0.23 (0.42) 0.84 (0.37) 0.16 (0.37) 0.83 (0.38)
*European% 0.95 (0.11) 0.25 (0.17) 0.39 (0.27) 0.80 (0.25) 0.61 (0.21)
*African% 0.02 (0.07) 0.70 (0.18) 0.05 (0.10) 0.09 (0.20) 0.09 (0.11)
*Amerindian% 0.01 (0.04) 0.01 (0.03) 0.01 (0.04) 0.08 (0.14) 0.27 (0.20)
*East Asian% 0.02 (0.07) 0.03 (0.04) 0.55 (0.26) 0.03 (0.09) 0.03 (0.06)
N3861 326 182 67 726
Note: Standard deviations are reported in parentheses; *Genetic ancestry is the average of both biological
parents.
3.2 Main regression results
Table 2displays the results for the analysis involving education. In the first model, excluding ancestry variables,
the unstandardized coecients of Black, Asian, Native American, Hispanic, and Other race/ethnicity are b
= -0.55, 0.22, -0.14, -0.64, and -0.42, respectively. However, in the second model which includes ancestry
variables, the unstandardized race/ethnicity coecients are b= 0.24, -0.01, -0.04, 0.01, and -0.17, respectively.
This finding indicates that ancestry statistically explains the eect of socially-defined race/ethnicity and,
more importantly, that minority race/ethnic categories, apart from Other race, are not associated with lower
educational attainment levels as compared to White identity once genetic ancestry is controlled for. In fact, Black
is associated with higher educational attainment than White race/ethnicity. When examining the coecients
of the genetic ancestry variables, the African and Amerindian ancestry variables are negatively related to
education (b= -1.25 and b= -3.27, respectively), while East Asian ancestry shows a weak positive association
(but non-significant) with education as compared to European ancestry (b= 0.13, p= .363).
Table 3displays the result for the analysis involving income. In the first model, excluding ancestry variables,
the unstandardized coecients of Black, Asian, Native American, Hispanic, and Other race/ethnicity are b=
-0.70, 0.20, -0.21, -0.53, and -0.32, respectively. However in the second model including ancestry variables, the
unstandardized race/ethnicity coecients are b= 0.25, 0.18, -0.16, -0.04, and -0.11, respectively. Once more,
these findings indicate that ancestry statistically explains the eect of race/ethnicity categories, except in the
case of Asian and Native American race/ethnicity. When examining the coecients of the genetic ancestry
variables, African and Amerindian ancestry variables are negatively related to income (b= -1.48 and b=
-2.51 respectively). East Asian ancestry also shows a negative association (but non-significant) with income as
compared to European ancestry (b= -0.23, p= .102).
3.3 Results from the Robustness analyses
We ran a series of robustness tests using variations of the main models, displayed in Tables 2 and 3. Detailed
results from these additional analyses are provided in the supplementary file.
2
First, we exclude all cases with
values of education and income 3 (or more) SD below the mean so as to ensure that our results are not driven
mainly by extreme values. The results from these analyses are very similar to those from the main analyses
except that, in the models without ancestry, Hispanic and Black identity is associated with slightly less worse
socioeconomic outcomes and in the models with genetic ancestry both African and Amerindian ancestries are
somewhat less negatively associated with outcomes.
A second set of analyses repeats the main regression models using child’s race/ethnicity (as reported by the
parents) instead of the responding parent’s race/ethnicity. The results of these analyses are also very similar
2
In the supplementary file, we also provided the bivariate correlation of European ancestry with income and with educational attainment
for two biological Black families (respectively, .19 and .18) and for two biological Hispanic families (respectively, .43 and .54).
6
Published: 11th of September 2023 OpenPsych
Table 2: Admixture Regression Results for Parental Educational Attainment.
Model 1 Model 2
Predictors b P b p
(Intercept) -0.06 (0.06) 0.355 0.20 (0.06) 0.001
Black -0.55 (0.05) <0.001 0.24 (0.10) 0.012
East Asian 0.22 (0.07) 0.001 -0.01 (0.09) 0.957
Native American -0.14 (0.09) 0.145 -0.04 (0.09) 0.656
Hispanic -0.64 (0.05) <0.001 0.01 (0.06) 0.864
Other Race -0.42 (0.06) <0.001 -0.17 (0.06) 0.003
English Fluency 0.21 (0.05) <0.001 0.00 (0.05) 0.993
Age 0.22 (0.01) <0.001 0.18 (0.01) <0.001
Immigrant Family 0.10 (0.03) 0.002 0.15 (0.03) <0.001
African ancestry -1.25 (0.13) <0.001
Amerindian ancestry -3.27 (0.14) <0.001
East Asian ancestry 0.13 (0.14) 0.363
Random Eects
σ20.69 0.61
τ00 0.03sit e_id_l0.02s ite_i d_l
ICC 0.04 0.03
N 22sit e_id_l22s ite_i d_l
Observations 5073 5073
Marginal R2/ Conditional R20.207 / 0.239 0.297 / 0.319
Note: standard errors are reported in parentheses.
Table 3: Admixture Regression Results for Parental Income.
Model 1 Model 2
Predictors b P b p
(Intercept) -0.17 (0.07) 0.013 0.06 (0.07) 0.421
Black -0.70 (0.05) <0.001 0.25 (0.09) 0.010
East Asian 0.20 (0.07) 0.003 0.18 (0.09) 0.051
Native American -0.21 (0.09) 0.019 -0.16 (0.09) 0.076
Hispanic -0.53 (0.05) <0.001 -0.04 (0.06) 0.519
Other Race -0.32 (0.06) <0.001 -0.11 (0.06) 0.067
English Fluency 0.35 (0.05) <0.001 0.17 (0.05) <0.001
Age 0.18 (0.01) <0.001 0.15 (0.01) <0.001
Immigrant Family 0.08 (0.03) 0.005 0.14 (0.03) <0.001
African ancestry -1.48 (0.13) <0.001
Amerindian ancestry -2.51 (0.14) <0.001
East Asian ancestry -0.23 (0.14) 0.102
Random Eects
σ20.66 0.61
τ00 0.05sit e_id_l0.05s ite_i d_l
ICC 0.08 0.08
N 22sit e_id_l22s ite_i d_l
Observations 5073 5073
Marginal R2/ Conditional R20.196 / 0.256 0.259 / 0.319
Note: standard errors are reported in parentheses.
7
Published: 11th of September 2023 OpenPsych
to the main results except that Black identity shows a slightly lower positive coecient in the regression
analysis for educational attainment. A third set of analyses restricts the sample to families with at least one (as
opposed to two) biological parent. Doing so yields a much larger sample (N= 7,652) due to the high rate of
single-parenthood among Black and Hispanic identifying individuals. The results are similar to those from the
main analyses except that Black identity is less positively associated with outcomes and that Native American
identity is more negatively associated with outcomes.
A fourth set of analyses adds state-level racism and anti-immigrant bias. Both racism and anti-immigrant bias
are weak predictors of outcomes and so had little eect on the results. The results for the second through fourth
analyses are summarized in Table 4, alongside those from Tables 2 and 3.
As seen in Table 4, when ancestry is controlled for, Black identity is statistically significantly positively
associated with outcomes in the majority of models. Since a large portion of the structural racism literature
focuses on discrimination against individuals socially identified as Black it would be worthwhile to see if these
results replicate on other samples. Given these counterintuitive results, before such replication is done (our
Black sample is small in our analysis), we are reluctant to speculate on possible causes.
Finally, we replicate the full model of our main analyses within the combined Hispanic, Black, and Native
American sample. In the educational attainment model, the coecients for Black, Asian, Native American,
and Hispanic identity are, respectively, b= 0.56 (p= .034), 0.24 (p= .592), and 0.35 (p= .134), whereas the
coecients for African, Amerindian, and East Asian ancestry are, respectively, b= -1.55 (p< .001), -3.21 (p
< .001), and -1.01 (p= .104). In the income model, the coecients for Black, Asian, Native American, and
Hispanic identity are, respectively, b= 0.05 (p= .872), 0.10 (p= .836), and -0.20 (p= .440), whereas the
coecients for African, Amerindian, and East Asian ancestry are b= -1.59 (p<.001), -2.54 (p<.001), and -0.23
(p= .733), respectively. As seen, the eect of race/ethnicity is similar to that in the main sample. Additionally,
Black identity is not negatively associated with socioeconomic outcomes after controlling for genetic ancestry.
Table 4: Summary of Admixture Regression Results for Parental Education and Income.
Main models
Models with
child care
Models with
=>1 biological
parent
Models with
state-level
racism variables
Educ.
Income
Educ.
Income
Educ.
Income
Educ.
Income
Black dummy 0.24 0.25 0.11 0.25 0.11 0.08 0.24 0.24
Asian dummy -0.01 0.18 0.00 0.19 0.11 0.15 0.00 0.18
Native American dummy -0.04 -0.16 -0.07 -0.17 -0.13 -0.17 -0.04 -0.15
Hispanic dummy 0.01 -0.04 -0.04 0.00 0.01 -0.05 0.01 -0.04
Other Race dummy -0.17 -0.11 -0.09 -0.05 -0.15 -0.17 -0.17 -0.11
African ancestry -1.25 -1.48 -1.13 -1.55 -1.06 -1.59 -1.24 -1.48
Amerindian ancestry -3.27 -2.51 -3.23 -2.57 -2.84 -1.92 -3.28 -2.52
East Asian ancestry 0.13 -0.23 0.13 -0.22 0.02 -0.26 0.13 -0.24
N5073 5073 5073 5073 7652 7652 5073 5073
4 Discussion
The research examined whether racial dierences in educational attainment and income are associated with
socially-identified race/ethnicity independent of genetic ancestry. We found that, after controlling for genetic
ancestry, Hispanic and Native American identities, as compared to White identity, are not associated with lower
educational attainment and that Native American identity is associated with only slightly lower income levels.
Additionally, when genetic ancestry is held constant, Black identity is associated with higher education and
income than White identity. The results hold when the children’s race/ethnicity are used instead of those of
the responding parents. Furthermore, the results hold when subsetting to only Hispanic, Black, and Native
American individuals, a finding which is consistent with the findings of a meta-analysis of American studies
(Kirkegaard et al.,2017).
The structural racism hypothesis, as commonly formulated, clearly predicts that socially-defined race/ethnicity
will have an eect on educational attainment and income independent of genetic ancestry. This is because most
8
Published: 11th of September 2023 OpenPsych
of the specific laws, policies, institutional practices, and entrenched social norms in the USA, which could have
adversely aected non-White groups, did not target individuals based on genetic ancestry, but rather did so
based on socially defined race/ethnicity (Gichoya et al.,2022;Smedley & Smedley,2005). An alternative is
the cognitive meritocracy hypothesis. According to this, educational and income dierences are mostly due
to general cognitive ability and other human capital dierences. Since a number of studies have shown that
general cognitive ability tracks genetic ancestry better than socially-defined race and genetically predicted color
(e.g., Kirkegaard et al. 2019), this model would predict that social outcome dierences, being antecedent to
cognitive ability ones, also follows genetic ancestry.
Kirkegaard et al. (2017) conducted a large meta-analysis of genetic epidemiological studies on the relation
between European, African, and Amerindian ancestry and indices of SES. The authors found that, with a high
degree of consistency, European ancestry was positively associated with better outcomes and that African and
Amerindian ancestry was negatively associated with better outcomes. For the most part, these studies did not
examine the independent eect of socially-defined race on outcomes. However, the authors report results from
one study from Brazil which examines the eects of both interviewer and participant-reported color/race (“cor”)
on household assets, schooling, and income. Independent of European ancestry, racial/color identification was
not statistically significantly associated with outcomes, while European ancestry was strongly associated with
better SES outcomes. These results from Brazil, then, are consistent with the ones reported in this paper.
One major issue related with our data is that, as we only had admixture estimates for children, we have limited
the sample to dyads who were biological parents. As a result the non-White sample sizes were modest. Moreover,
since dual parenthood is positively related to socioeconomic status, the samples used are not representative
of American populations. This should, therefore, be seen as an exploratory study. There are large datasets
containing genetic data and adult educational attainment and income on which admixture regression analyses
could be run in the future to determine if socially-defined race/ethnicity has predictive validity independent of
genetic ancestry. Since many researchers very clearly argue that the constructive aspects of race/ethnicity are
strongly related to socioeconomic outcomes, this research is worth pursuing to better understand the nature of
race/ethnic related socioeconomic disparities in the USA.
A major issue with the study design is that factors other than discrimination, such as assortative mating and
selective ethnic attrition, can potentially induce associations between social outcomes and socially defined
race. For example, mixed-race couples could be socioeconomically selective, and this selectivity could lead to
social race being correlated with SES independent of genetic ancestry. Were these processes influencing our
results, we would generally expect socially identified races to be associated with outcomes independent of
ancestry. However, we generally do not find this to be the case. Nonetheless, it is theoretically possible that
the eects of discrimination could be moderated by countervailing eects of assortative mating and selective
ethnic attrition. This leaves open the question as to whether the pattern of mating could have explained the
positive coecient of Black identity on socio-economic outcomes. It might be argued that Black mothers who
intermarry typically achieve higher SES levels than White mothers who intermarry, therefore weighing up the
impact of Black identity. This requires a rate high enough to create a positive eect of b= 0.24 (as found in this
study). At the same time, not having information about the racial identification of the spouse of the responding
parent does not help to address this issue. Whether this is an empirically plausible scenario depends on the
pattern of mating and identification in American race/ethnic groups and is a subject for future research.
Author contributions
J.F. conceived of the idea. Analyses used the ABCD sample were conducted by J.F. under the supervision of
B. J. Pesta while at Cleveland State University (2020-2021). J. F, M.H, and E.O.W. K edited and revised the
manuscript. Both authors discussed the results and contributed to the final manuscript.
Acknowledgments
Data used in the preparation of this article were obtained from the Adolescent Brain Cognitive Develop-
ment (ABCD) Study (
https://abcdstudy.org
), held in the NIMH Data Archive (NDA). This is a multi-
site, longitudinal study designed to recruit more than 10,000 children aged 9-10 and follow them over 10
years into early adulthood. The ABCD Study® is supported by the National Institutes of Health and addi-
tional federal partners under award numbers U01DA041048, U01DA050989, U01DA051016, U01DA041022,
U01DA051018, U01DA051037, U01DA050987, U01DA041174, U01DA041106, U01DA041117, U01DA041028,
9
Published: 11th of September 2023 OpenPsych
U01DA041134, U01DA050988, U01DA051039, U01DA041156, U01DA041025, U01DA041120, U01DA051038,
U01DA041148, U01DA041093, U01DA041089, U24DA041123, U24DA041147. A full list of supporters is
available at
https://abcdstudy.org/federal-partners.html
. A listing of participating sites and a complete
listing of the study investigators can be found at
https://abcdstudy.org/consortium_members/
. ABCD
consortium investigators designed and implemented the study and/or provided data, but did not necessarily
participate in the analysis or writing of this report. This manuscript reflects the research results and inter-
pretations of the authors alone and may not reflect the opinions or views of the NIH or ABCD consortium
investigators. The ABCD data repository grows and changes over time. The ABCD data used in this report came
from Version 3.01. The raw data are available at
https://nda.nih.gov/edit_collection.html?id=2573
.
Additional support for this work was made possible from supplements to U24DA041123 and U24DA041147,
the National Science Foundation (NSF 2028680), and Children and Screens: Institute of Digital Media and
Child Development Inc. NDA waved the requirement of creating a NDA data project (NDA, Oct 26, 2022).3
References
Adkins-Jackson, P. B., Chantarat, T., Bailey, Z. D., & Ponce, N. A. (2022). Measuring structural racism: a guide
for epidemiologists and other health researchers. American Journal of Epidemiology,191(4), 539-547. doi:
10.1093/aje/kwab239
American Sociological Association. (2003). The importance of collecting data and doing social scientific research
on race [Computer software manual]. Washington, DC.
Bailey, Z. D., Feldman, J. M., & Bassett, M. T. (2021). How structural racism works—racist policies as a root cause
of us racial health inequities. New England Journal of Medicine,384(8), 768-773. doi:
10.1056/nejmms2025396
Bates, D., Maechler, M., Bolker, B., Walker, S., Christensen, R. H. B., Singmann, H., & Grothendieck, G. (2009).
Package ‘lme4’ [Computer software manual]. Retrieved from https://lme4.r-forge.r-project.org/
Braveman, P. A., Arkin, E., Proctor, D., Kauh, T., & Holm, N. (2022). Systemic and structural racism:
Definitions, examples, health damages, and approaches to dismantling. Health Aairs,41(2), 171-178. doi:
10.1377/hlthaff.2021.01394
Brues, A. M. (1990). People and races. Prospect Heights, IL: Waveland.
Bryc, K., Durand, E. Y., Macpherson, J. M., Reich, D., & Mountain, J. L. (2015). The genetic ancestry of african
americans, latinos, and european americans across the united states. The American Journal of Human Genetics,
96(1), 37-53. doi: 10.1016/j.ajhg.2014.11.010
Connor, G., & Fuerst, J. G. (2022). Linear and partially linear models of behavioral trait variation using
admixture regression. In R. Lynn (Ed.), Intelligence, race and sex: Some controversial issues. a tribute to helmuth
nyborg at 85. Budapest, Hungary. doi: 10.1101/2021.05.14.444173
Erikson, C. E., Dent, R. B., Park, Y. H., & Luo, Q. (2022). Historic redlining and contemporary behavioral health
workforce disparities. JAMA Network Open,5(4), e229494-e229494. doi:
10.1001/jamanetworkopen.2022
.9494
Fang, H., Hui, Q., Lynch, J., Honerlaw, J., Assimes, T. L., Huang, J., & Tang, H. (2019). Harmonizing genetic
ancestry and self-identified race/ethnicity in genome-wide association studies. The American Journal of
Human Genetics,105(4), 763-772. doi: 10.1016/j.ajhg.2019.08.012
Fuerst, J. G. (2021). Robustness analysis of african genetic ancestry in admixture regression models of cognitive
test scores. Mankind Quarterly,62(2). doi: 10.46469/mq.2021.62.2.12
Fuerst, J. G., Hu, M., & Connor, G. (2021). Genetic ancestry and general cognitive ability in a sample of
american youths. Mankind Quarterly,62(1), 186-216. doi: 10.46469/mq.2021.62.1.11
Fuerst, J. G., Shibaev, V., & Kirkegaard, E. O. W. (2023). A genetic hypothesis for american race/ethnic
dierences in mean g: A reply to warne (2021) with fifteen new empirical tests using the abcd dataset.
Mankind Quarterly,63(4). doi: 10.46469/mq.2023.63.4.2
3NDA Study Acknowledgments. Retrieved from: https://nda.nih.gov/ampscz/manuscript-preparation.html
10
Published: 11th of September 2023 OpenPsych
Gichoya, J. W., Banerjee, I., Bhimireddy, A. R., Burns, J. L., Celi, L. A., Chen, L. C., & Zhang, H. (2022). Ai
recognition of patient race in medical imaging: a modelling study. The Lancet Digital Health,4(6), e406-e414.
doi: 10.1016/s2589-7500(22)00063-2
Guo, G., Fu, Y., Lee, H., Cai, T., Li, Y., & Harris, K. M. (2014). Recognizing a small amount of superficial
genetic dierences across african, european and asian americans helps understand social construction of race.
Demography,51(6), 2337-2342. doi: 10.1007/s13524-014-0349-y
Haeny, A. M., Holmes, S. C., & Williams, M. T. (2021). The need for shared nomenclature on racism and
related terminology in psychology. Perspectives on Psychological Science,16(5), 886-892. doi:
10.1177/
17456916211000760
Hatton, S. (2018). Preview of the abcd study® data release 3.0 (Tech. Rep.). Department of Neurosciences, UC San
Diego. Retrieved from https://abcdstudy.org/wp-content/uploads/2020/06/ABCD_Release3.0.pdf
Herd, P., Mills, M. C., & Dowd, J. B. (2021). Reconstructing sociogenomics research: Dismantling biological
race and genetic essentialism narratives. Journal of Health and Social Behavior,62(3), 419-435. doi:
10.1177/
00221465211018682
Jones, C. P. (2002). Confronting institutionalized racism. Phylon (1960-), 7-22. doi: 10.2307/4149999
Kirkegaard, E. O. W. (2021). Genetic ancestry and social race are nearly interchangeable. OpenPsych. doi:
10.26775/op.2021.12.22
Kirkegaard, E. O. W., & Fuerst, J. G. (2023). A multimodal mri-based predictor of intelligence and its relation
to race/ethnicity. Mankind Quarterly,63(3). doi: 10.46469/mq.2023.63.3.2
Kirkegaard, E. O. W., Wang, M., & Fuerst, J. (2017). Biogeographic ancestry and socioeconomic outcomes in the
americas: A meta-analysis. Mankind Quarterly,57(3), 398–427. doi: 10.46469/mq.2017.57.3.10
Kirkegaard, E. O. W., Williams, R. L., Fuerst, J., & Meisenberg, G. (2019). Biogeographic ancestry, cognitive
ability and socioeconomic outcomes. Psych,1(1), 1-25. doi: 10.3390/psychology1010001
Lasker, J., Pesta, B. J., Fuerst, J. G., & Kirkegaard, E. O. W. (2019). Global ancestry and cognitive ability. Psych,
1(1), 431-459. doi: 10.3390/psych1010034
Noguera, P. A., & Alicea, J. A. (2020). Structural racism and the urban geography of education. Phi Delta
Kappan,102(3), 51-56. doi: 10.1177/0031721720970703
Oce of Management and Budget. (1997). Revisions to the standards for the classification of federal data on
race and ethnicity. Federal Register,62(210), 58782-58790.
O’Reilly, K. B. (2020). Ama: Racism is a threat to public health. American Medical Association.
Package ‘survey’ [Computer software manual]. (2020). https://cran.r-project.org/.
Robbins, N. M., Charleston, L., Saadi, A., Thayer, Z., Codrington, W. U., Landry, A., & Hamilton, R. (2022).
Black patients matter in neurology: Race, racism, and race-based neurodisparities. Neurology,99(3), 106-114.
doi: 10.1212/wnl.0000000000200830
Smedley, A., & Smedley, B. D. (2005). Race as biology is fiction, racism as a social problem is real: An-
thropological and historical perspectives on the social construction of race. American Psychologist,60(1),
16.
Suyemoto, K. L., Donovan, R. A., & Kim, G. S. (2022). Unraveling assumptions: A primer for understanding
oppression and privilege. Taylor & Francis. doi: 10.4324/9780429059599
11
... These include the Philadelphia Neuroimaging Cohort (PNC), Pediatric Imaging, Neurocognition, and Genetics Study (PING), and the Adolescent Brain Cognitive Development Study (ABCD). These studies predominantly focus on regressing an estimate of European genetic ancestry from the program Admixture against various phenotypes ranging from scores from cognitive test batteries, income, and neuroimaging data (examples include Lasker et al., 2019;Kirkegaard et al., 2019;Fuerst et al., 2021a;Fuerst et al., 2021b;Fuerst et al., 2021c;Kirkegaard and Fuerst, 2023;Fuerst et al., 2023a;Fuerst et al., 2023b;Hu et al., 2023;Shibaev and Fuerst, 2023). Despite known issues with this admixture regression approach to distinguish whether an ancestry-trait correlation is caused by genetic effects or covarying environmental effects (Schraiber and Edge, 2023) these papers make bold claims about genetic differences explaining a substantial portion of racial differences in intelligence, educational attainment, and parental income among Black and white Americans. ...
... Despite this punitive action ostensibly nullifying all pre-existing Data Use Certification Agreements held by Dr. Pesta (and, in turn, cutting off the access of his collaborators to said data), Dr. Pesta's institution, Cleveland State University, concluded that John Fuerst, a graduate student and coauthor of Lasker et al. (2019), had retained an unauthorized copy of the ABCD dataset (Standifer, 2022). Fuerst has since published at least 8 preprints and papers analyzing the ABCD data, with at least 5 other coauthors (Fuerst 2021a;Fuerst et al., 2021b;Fuerst et al., 2021c;Fuerst et al., 2023a;Fuerst et al., 2023b;Hu et al., 2023;Kirkegaard and Fuerst, 2023;Shibaev and Fuerst, 2023). Searching the NIMH Data Archive's database of approved Data Use Certification (DUC) Agreements, we did not find a single DUC requesting access to the ABCD data that had been granted to Fuerst, nor to any of the coauthors of these recent papers. ...
Article
Full-text available
Public genomic datasets like the 1000 Genomes project (1KGP), Human Genome Diversity Project (HGDP), and the Adolescent Brain Cognitive Development (ABCD) study are valuable public resources that facilitate scientific advancements in biology and enhance the scientific and economic impact of federally funded research projects. Regrettably, these datasets have often been developed and studied in ways that propagate outdated racialized and typological thinking, leading to fallacious reasoning among some readers that social and health disparities among the so-called races are due in part to innate biological differences between them. We highlight how this framing has set the stage for the racist exploitation of these datasets in two ways: First, we discuss the use of public biomedical datasets in studies that claim support for innate genetic differences in intelligence and other social outcomes between the groups identified as races. We further highlight recent instances of this which involve unauthorized access, use, and dissemination of public datasets. Second, we discuss the memification, use of simple figures meant for quick dissemination among lay audiences, of population genetic data to argue for a biological basis for purported human racial groups. We close with recommendations for scientists, to preempt the exploitation and misuse of their data, and for funding agencies, to better enforce violations of data use agreements.
... This method is also used to keep track of genetic changes taking place over time in present and historic human populations (e.g., Kong et al., 2017;Piffer et al., 2023): evolution in action. The second method is admixture regression, which tests the hypothesis that in admixed populations, genetic ancestry rather than physical appearance or social classification best predicts the studied outcomes (e.g., Hu et al., 2023). Suffice it to say that the results of both research strategies so far indicate that the genetic measures predict differences in racial averages. ...
Article
Full-text available
Intelligence tests are excellent predictors of school and job performance, and racial/ethnic differences in mean IQ are common. Based on five lines of evidence, Warne (2021) builds a case for partly genetic causes of differences in general intelligence (g) across American self/parental-identified race or ethnicity (SIRE). Based on a careful reading of Warne (2021) and the authors he cites, we generated 15 predictions flowing from a partial genetic hypothesis. These predictions concern (1) mean cognitive differences and Spearman’s hypothesis, (2) measurement invariance across European genetic ancestry, (3) high within-group heritability and low shared environmentality, (4) admixture regression for g, (5) polygenic scores, (6) brain volume, and (7) results based on the method of correlated vectors. We used the Adolescent Brain Cognitive Development Study sample (N = 10,245) to test these hypotheses using classical and state-of-the-art statistical techniques. Decomposition of variance using twins showed that the heritability of intelligence and of brain/intracranial volume estimates were, respectively, moderate and high for both the White and the non-White subsamples, while the variance attributable to shared environment was low. Within SIRE groups, both genetic ancestry and education-related polygenic scores (eduPGS) predicted both brain volume and g. Moreover, brain volume was weakly but statistically significantly related to g (r = .14 to .25). Path and mediation analysis showed that total brain volume explained approximately 15% of the association between European ancestry and g and also explained approximately 8% of that between eduPGS and g. Finally, based on the method of correlated vectors (MCV), a positive manifold was found for genetic, brain volume, and ancestry/SIRE-related variables. We conclude that the results support the hypotheses tested and are in line with a partial genetic hypothesis.
Article
Full-text available
We used data from the Adolescent Brain Cognitive Development Study to create a multimodal MRI-based predictor of intelligence. We applied the elastic net algorithm to over 50,000 neurological variables. We find that race can confound models when a multiracial training sample is used, because models learn to predict race and use race to predict intelligence. When the model is trained on non-Hispanic Whites only, the MRI-based predictor has an out-of-sample model accuracy of r = .51, which is 3 to 4 times greater than the validity of whole brain volume in this dataset. This validity generalized across the major socially-defined racial/ethnic groupings (White, Black, and Hispanic). There are race gaps on the predicted scores, even though the model is trained on White subjects only. This predictor explains about 37% of the relation between both the Black and Hispanic classification and intelligence.
Article
Full-text available
Background Previous studies in medical imaging have shown disparate abilities of artificial intelligence (AI) to detect a person's race, yet there is no known correlation for race on medical imaging that would be obvious to human experts when interpreting the images. We aimed to conduct a comprehensive evaluation of the ability of AI to recognise a patient's racial identity from medical images. Methods Using private (Emory CXR, Emory Chest CT, Emory Cervical Spine, and Emory Mammogram) and public (MIMIC-CXR, CheXpert, National Lung Cancer Screening Trial, RSNA Pulmonary Embolism CT, and Digital Hand Atlas) datasets, we evaluated, first, performance quantification of deep learning models in detecting race from medical images, including the ability of these models to generalise to external environments and across multiple imaging modalities. Second, we assessed possible confounding of anatomic and phenotypic population features by assessing the ability of these hypothesised confounders to detect race in isolation using regression models, and by re-evaluating the deep learning models by testing them on datasets stratified by these hypothesised confounding variables. Last, by exploring the effect of image corruptions on model performance, we investigated the underlying mechanism by which AI models can recognise race. Findings In our study, we show that standard AI deep learning models can be trained to predict race from medical images with high performance across multiple imaging modalities, which was sustained under external validation conditions (x-ray imaging [area under the receiver operating characteristics curve (AUC) range 0·91–0·99], CT chest imaging [0·87–0·96], and mammography [0·81]). We also showed that this detection is not due to proxies or imaging-related surrogate covariates for race (eg, performance of possible confounders: body-mass index [AUC 0·55], disease distribution [0·61], and breast density [0·61]). Finally, we provide evidence to show that the ability of AI deep learning models persisted over all anatomical regions and frequency spectrums of the images, suggesting the efforts to control this behaviour when it is undesirable will be challenging and demand further study. Interpretation The results from our study emphasise that the ability of AI deep learning models to predict self-reported race is itself not the issue of importance. However, our finding that AI can accurately predict self-reported race, even from corrupted, cropped, and noised medical images, often when clinical experts cannot, creates an enormous risk for all model deployments in medical imaging. Funding National Institute of Biomedical Imaging and Bioengineering, MIDRC grant of National Institutes of Health, US National Science Foundation, National Library of Medicine of the National Institutes of Health, and Taiwan Ministry of Science and Technology
Article
Full-text available
It has been claimed that social race and genetic ancestry are at best weakly related. Here we test this claim by applying predictive modeling in both directions, i.e., predicting genetic ancestry from social race(s), and predicting social race(s) from genetic ancestry. We utilize the public Pediatric Imaging, Neurocognition, and Genetics (PING) dataset (n = 1,391), so that others may examine the data as well. In the simple scenario where we are only concerned with self-identified white, black, and mixed (black-white) race individuals (571 whites, 140 blacks, 25 mixed), model accuracy was very high. Predicting social race from genetic ancestry resulted in an area under curve (AUC) of .994, an overall accuracy (concordance) of 98.0%, and a pseudo-R2 of .951. Conversely, predicting genetic ancestry from social race had a model R2 adjusted of .992. Using the full dataset, there are 8 census-type categories of social race. Using cross-validated multinomial regression to predict social race from 6 genetic ancestry variables, we find that the AUC is .89. Using Dirichlet regression to predict ancestries from social race, we find an overall correlation of .94 (R2 = 88.4%). Further analyses using more sophisticated methods (random forest, support vector machine) found similar results. In conclusion, social race and genetic ancestry are nearly interchangeable.
Article
Full-text available
Admixture regression involves the regression of a medical or behavioral trait on sub-population admixture proportions, race and ethnic group identifiers, and other explanatory variables, to measure the environmental and genetic components to trait variation across self-identified racial and ethnic groups. Several recent admixture regression studies find that proportion of African genetic ancestry has strong explanatory power for cognitive ability test scores. This paper applies stratified sub-sampling to explore the possibility that potential model mis-specification and/or variable mis-measurement underlies this empirical finding. The estimated coefficient on African ancestry is not significantly changed when the sample is stratified by low socioeconomic status/high socioeconomic status, Hispanic/non-Hispanic, Black-only/Black-White biracial, or Black/White racial group self-identification.
Article
Full-text available
Black and Hispanic children in the United States have lower mean cognitive test scores than White children. The reasons for this are contested. The test score gap may be caused by socio-cultural factors, but the high heritability of g suggests that genetic variance might play a role. Differences between self-identified race or ethnicity (SIRE) groups could be the product of ancestral genetic differences. This genetic hypothesis predicts that genetic ancestry will predict g within these admixed groups. To investigate this hypothesis, we performed admixture-regression analyses with data from the Adolescent Brain Cognitive Development Cohort. Consistent with predictions from the genetic hypothesis, African and Amerindian ancestry were both found to be negatively associated with g. The association was robust to controls for multiple cultural, socioeconomic, and phenotypic factors. In the models with all controls the effects were as follows: (a) Blacks, African ancestry: b =-0.89, N = 1690; (b) Hispanics, African ancestry: b =-0.58, Amerindian ancestry: b =-0.86, N = 2021), and (c) a largely African-European mixed Other group, African ancestry: b =-1.08, N = 748). These coefficients indicate how many standard deviations g is predicted to change when an individual's African or Amerindian ancestry proportion changes from 0% to 100%. Genetic ancestry statistically explained the self-identified race and ethnicity (SIRE) differences found in the full sample. Lastly, within all samples, the relation between genetic ancestry and g was partially accounted for by cognitive ability and educational polygenic scores (eduPGS). These eduPGS were found to be significantly predictive of g within all SIRE groups, even when controlling for ancestry. The results are supportive of the genetic model.
Article
Black people living in the United States suffer disproportionate morbidity and mortality across a wide range of neurologic conditions. Despite common conceptions to the contrary, “race” is a socially defined construct with little genetic validity. Therefore, racial health inequities in neurology (“neurodisparities”) are not a consequence of biologic differences between races. Instead, racism and associated social determinants of health are the root of neurodisparities. To date, many neurologists have neglected racism as a root cause of neurologic disease, further perpetuating the problem. Structural racism, largely ignored in current neurologic practice and policy, drives neurodisparities through mediators such as excessive poverty, inferior health insurance, and poorer access to neurologic and preventative care. Interpersonal racism (implicit or explicit) and associated discriminatory practices in neurologic research, workforce advancement, and medical education also exacerbate neurodisparities. Neurologists cannot fulfill their professional and ethical responsibility to care for Black patients without understanding how racism, not biologic race, drives neurodisparities. In our review of race, racism, and race-based disparities in neurology, we highlight the current literature on neurodisparities across a wide range of neurologic conditions and focus on racism as the root cause. We discuss why all neurologists are ethically and professionally obligated to actively promote measures to counteract racism. We conclude with a call for actions that should be implemented by individual neurologists and professional neurologic organizations to mitigate racism and work towards health equity in neurology.
Article
This cross-sectional study investigates the association between redlining and behavioral health specialist supply in 2 mental health professional shortage areas.
Article
Racism is not always conscious, explicit, or readily visible-often it is systemic and structural. Systemic and structural racism are forms of racism that are pervasively and deeply embedded in systems, laws, written or unwritten policies, and entrenched practices and beliefs that produce, condone, and perpetuate widespread unfair treatment and oppression of people of color, with adverse health consequences. Examples include residential segregation, unfair lending practices and other barriers to home ownership and accumulating wealth, schools' dependence on local property taxes, environmental injustice, biased policing and sentencing of men and boys of color, and voter suppression policies. This article defines systemic and structural racism, using examples; explains how they damage health through many causal pathways; and suggests approaches to dismantling them. Because systemic and structural racism permeate all sectors and areas, addressing them will require mutually reinforcing actions in multiple sectors and places; acknowledging their existence is a crucial first step.