The Justice, Equity, Diversity, and Inclusion (JEDI) Outreach Group Corner is a regular component of Amstat News in which statisticians write about and educate our community about JEDI-related matters. If you have an idea or article for the column, email the JEDI Corner manager.
Eric J. Daza has been a health data scientist for more than 21 years (BA, neurobiology, Cornell; MPS, applied statistics, Cornell; DrPH, biostatistics, UNC Chapel Hill; and postdoc, Stanford ). He created Stats-of-1, a health innovation newsletter/podcast featured in Forbes and Fortune. He is also a diversity, equity, and inclusion leader; Filipino American immigrant; and trained musician.
Diversity, equity, and inclusion—known as DEI—efforts are statistically compelling. As a statistician, technical challenges like those below helped spark my own work to improve DEI. You might feel likewise.
In June 2023, the US Supreme Court struck down affirmative action in college admissions. Many of the conservative justices opined that colleges should not directly consider an applicant’s race to help decide if they should be admitted. They think the probability of admission (i.e., selection probability) given all applicant-reported factors including merit (as defined by each college) and race should be the same as the probability of admission given all applicant-reported factors other than race.
This admissions policy could be fair in a society that gives everyone the same educational resources. From birth, your racial group’s educational needs are the same as those of any other. You must still struggle to prepare for college, but because educational needs are identically distributed across racial groups, at no point in your life does knowing your racial group help educators understand what resources you need.
This equitable cycle causes the proportions of well-prepared college applicants across racial groups to resemble each other—equality in merit caused by a lifetime of equity. Hence, the racial distribution (i.e., distribution of racial groups) among well-prepared applicants resembles that of the general population. Applicants’ educational needs do not vary by race.
In this hypothetical society, no college considers race in deciding who to admit. For example, suppose an admissions policy involves taking a completely random sample of well-prepared applicants. This sampling scheme causes the racial distribution of students to resemble that of well-prepared applicants. And the latter resembles the racial distribution of the general population. This outcome is fair, so the policy is equitable: Everyone has equal needs, so it distributes resources equally.
But meritocracy is a myth. The hypothetical society above is counterfactual.
In reality, we do not give everyone the same educational resources. From birth, your racial group’s educational needs differ from those of others due to historical inequity. These needs are unequal across racial groups. We compound this inequality by providing fewer resources to racial groups with greater needs, while giving more resources to groups that don’t need them.
This inequitable cycle causes the proportion of well-prepared college applicants in your racial group to differ from those in other groups—inequality in merit caused by a lifetime of inequity. Hence, the racial distribution of well-prepared applicants differs from that of the general population. And applicants’ educational needs vary by race.
What causes this inequity? Racism.
Racism is “a belief that race is a fundamental determinant of human traits and capacities and that racial differences produce an inherent superiority of a particular race,” according to the Merriam-Webster Dictionary. It is a historical, fundamental, integral part of many American institutions, cultures, and social norms.
Under an affirmative action policy, a college is able to consider race in deciding who to admit. Such policies recognize a lifetime of educational inequity causes some racial groups to struggle more than others to achieve the same level of merit and qualifications. These policies seek to remedy the results of such prior discrimination and prevent such discrimination in the future.
One remedy might be to take a disproportionate stratified sample of well-prepared applicants with racial groups as strata. Consider two such applicants with similar merit and qualifications and one applicant is from a more resource-deprived racial group. Under a policy that prioritizes admitting members of resource-deprived groups, this applicant would be given a greater probability of admission.
This sampling scheme causes the racial distribution of students to differ from that of well-prepared applicants—which is inequitably distributed. Over time, it ideally shifts the racial distribution of students to resemble that of the general population. This ideal outcome is fair, so the policy is reasonably equitable: It tries to distribute more educational resources to well-prepared applicants who need them most.
After the Supreme Court decision, an admissions committee can no longer directly consider race in college admissions. This can cause an admissions process to more closely mimic the aforementioned completely random sample.
This sampling scheme causes the racial distribution of students to resemble that of well-prepared applicants. Recall that the latter differs from that of the general population. Over time, this sampling scheme maintains, if not increases, the difference between the racial distribution of students and that of the general population.
This outcome is unfair because the general population isn’t well represented in the student body. Hence, the policy is inequitable; it fails to distribute more educational resources to well-prepared applicants who really need them, thereby prolonging—if not increasing—historical inequity and educational inequality.
By mandating ‘colorblind’ policies, the Supreme Court ruling might violate the Constitution’s equal protection clause, which bars racial discrimination by government entities. To this statistician, the irony is both clear and statistically significant.
Leave a Reply