The disparate impacts of college admissions policies on Asian American applicants

Amongst candidates to the Ivy-11 schools, we estimate that 16% of East Asian, 8% of Southeast Asian, and 10% of South Asian college students finally attended one in all these establishments, in comparison with 12% of white candidates. Whereas these mixture attendance charges differ by race and ethnicity, they don’t account for variations in {qualifications} throughout teams. For instance, Asian American candidates had, on common, greater standardized take a look at scores than white candidates (Desk S3). As a primary step to account for these variations, in Fig. 1 we present estimated attendance charges by standardized take a look at rating for Asian American candidates and white candidates. Except East Asian candidates on the highest take a look at scores, we discover that Asian American candidates to Ivy-11 colleges finally attended one in all these colleges at constantly decrease charges than white candidates with comparable take a look at scores, with the biggest hole for South Asian candidates. For example, amongst candidates with an ACT (or ACT-equivalent) rating of 34—inserting them within the 99th percentile of take a look at takers—we estimate that 16% of white candidates attended in comparison with 9% of South Asian candidates, a relative hole of 43%.

Standardized take a look at scores are however one issue amongst many who impression admissions selections and subsequent enrollment outcomes. Further standards that we’re in a position to observe embody highschool grade-point common (GPA), participation in extracurricular actions, legacy standing, and the state through which every applicant’s highschool is positioned. To know the extent to which these different issues might clarify the noticed disparities in attendance charges, we match a collection of nested logistic regression fashions of the next kind:

the place (Y_i) is a binary variable indicating whether or not applicant i attended any Ivy-11 faculty; (mathbbm {1}_S), (mathbbm {1}_E), and (mathbbm {1}_{SE}) point out whether or not the applicant recognized as South Asian, East Asian, or Southeast Asian, respectively; and (X_i) is a vector of extra covariates (e.g., take a look at scores and GPA) that we fluctuate throughout fashions, with (beta _X) the corresponding vector of coefficients. Our key coefficients of curiosity are (beta _S), (beta _E), and (beta _{SE}), which yield estimates of the hole in attendance charges between white candidates and Asian American candidates within the three Asian subgroups that we take into account. We discover related outcomes if we match separate fashions evaluating white candidates to candidates in every Asian subgroup individually (Tables S13–S15).

Desk 1 exhibits, for 9 fashions that embody completely different subsets of management variables, the fitted coefficients for every of the three Asian subgroups (see additionally Tables S5–S12). Coefficients are exponentiated for ease of interpretation as odds ratios. The primary mannequin contains solely mounted results for the applying season and the subset of schools (or utility “basket”) to which the scholar utilized—among the many full set of schools we take into account—facilitating comparisons amongst teams of scholars who utilized in the identical 12 months and to the identical subset of schools. The corresponding coefficients are thus akin to uncooked attendance odds ratios throughout teams, with out adjusting for variations in applicant credentials.

The second and third fashions in Desk 1 moreover modify for measures of educational preparation, together with SAT/ACT alone (Mannequin 2) and, moreover, GPA, AP take a look at scores, and SAT II topic take a look at scores (Mannequin 3). These academic-preparation fashions corroborate the visible sample in Determine 1: we estimate that Asian American college students—particularly South Asian college students—had considerably decrease odds of attendance than white college students with related take a look at scores and associated tutorial credentials. These disparities largely persist after we progressively modify for extracurricular actions (Mannequin 4); gender and household traits, like whether or not the scholar acquired an utility payment waiver (Mannequin 5); and whether or not the scholar utilized early (Mannequin 6).

Subsequent, with Mannequin 7, we account for whether or not a scholar is the kid of an alum. After adjusting for legacy standing—along with the entire above talked about components—we see massive reductions within the estimated disparities in attendance charges for all three Asian subgroups we take into account. Determine 2 helps clarify this end result. The highest panel of the determine exhibits estimated attendance charges for Asian American candidates and white candidates conditional on take a look at scores and legacy standing, which we outline on this determine to imply an applicant had at the least one mum or dad who attended an Ivy-11 as an undergraduate, and the scholar utilized to the Ivy-11 establishment(s) that their mum or dad(s) attended. For a given take a look at rating, we estimate that candidates—each white and Asian American—with legacy standing at an Ivy-11 have been greater than twice as prone to attend an Ivy-11 than candidates with out legacy standing. Within the backside panel of Fig. 2, we current prevalence of legacy standing amongst candidates with an ACT-equivalent take a look at rating of 32 or above, mirroring the main focus of the higher panel. Right here, we observe that white candidates have been roughly thrice extra prone to have legacy standing than East Asian and Southeast Asian candidates, and virtually six occasions extra possible than South Asian college students. Thus, although estimated attendance charges conditional on take a look at rating and legacy standing are related throughout race and ethnicity, white college students seem to profit from being considerably extra prone to have legacy standing.

The upper estimated attendance charges that we observe for legacy candidates might stem from both greater admission charges or greater yield charges. Although we are able to solely speculate, it appears possible that each components play a job. Additional, to the extent that our outcomes replicate disparities in admissions selections, these findings might in principle be pushed partially from the doubtless better social capital of legacy college students, quite than specific preferences for legacy candidates. We word, nonetheless, that Mannequin 5 adjusts for whether or not an applicant had a mum or dad who attended a top-50 establishment (based mostly on the 2019 U.S. Information rankings) not included within the subset of schools on which we focus, or attended an Ivy-11 faculty to which the scholar didn’t apply—proxies for having excessive social capital distinct from legacy standing particularly. The change in disparities that we observe shifting from Mannequin 5 to Mannequin 7 thus seems attributable to legacy standing particularly, quite than the extra generalized impacts of excessive social capital.

Estimated fee of attendance at any of the Ivy-11 schools for white candidates and Asian American candidates with excessive ACT or SAT scores. Throughout take a look at scores, we estimate that candidates with a mum or dad who attended an Ivy-11 as an undergraduate are greater than twice as prone to attend than non-legacy candidates with the identical take a look at scores. The underside panel exhibits the proportion of candidates with excessive take a look at scores who’ve legacy standing, disaggregated by race. Excessive-scoring white candidates are three to 6 occasions extra prone to have legacy standing than high-scoring Asian American candidates, suggesting white candidates disproportionately profit from legacy standing.

For every U.S. state, estimated fee of attendance at any Ivy-11 faculty for non-legacy white candidates with an ACT-equivalent rating at or above 32, with the proportion of high-scoring white and Asian candidates who establish as Asian on the horizontal axis. We report attendance charges of non-legacy white candidates to raised isolate the impression of geography on attendance from the potential impacts of legacy standing and race itself. Bigger level sizes point out a better variety of high-scoring white and Asian candidates from the state. The purple least-squares regression line is weighted by the identical rely of candidates. States with a better share of Asian American candidates have, on common, decrease estimated attendance charges for non-legacy white candidates with excessive scores.

Lastly, we look at the connection between estimated attendance charges and geography. For every state, Fig. 3 shows the estimated attendance fee of high-achieving candidates—with ACT-equivalent scores of 32 or above—to the fraction of candidates from that state who have been Asian American. When computing attendance charges, we restrict to non-legacy white candidates to regulate for the the potential results of legacy and race on enrollment and subsequently higher isolate the impression of geography. Level sizes are proportional to the whole variety of high-scoring white and Asian American candidates in every state. The negatively sloped regression line exhibits that states with a bigger fraction of Asian American candidates tended to have decrease estimated attendance charges. Additional, states with a better proportion of Asian American candidates tended to have greater common take a look at scores, suggesting the geographic pattern shouldn’t be pushed by a spot in tutorial achievement (Fig. S2). This geographic sample additionally persists after we exclude candidates from California, and after we disaggregate the information to the extent of highschool as an alternative of state (Figs. S1 and S3). Desk S27 shows the information used to assemble Fig. 3.

Mannequin 8 in Desk 1—which adjusts for location in addition to tutorial and extracurricular efficiency however not legacy standing—exhibits that these obvious geographic preferences account for a lot of the attendance hole between white and Asian American candidates. Mannequin 9, the final one we take into account, adjusts for all utility data obtainable to us, together with each legacy standing and geography. After adjusting for this wealthy set of covariates, we see that the estimated attendance hole between Southeast Asian and white candidates largely disappears, although we nonetheless discover that white candidates have greater estimated odds of attendance than in any other case related East Asian and South Asian candidates. It’s unclear what might account for these remaining disparities, although it bears repeating that admissions officers have entry to extra full utility supplies than can we—together with letters of advice, essays, and interview assessments—and scholar enrollment selections can also be impacted by components unobserved in our knowledge.

We conclude our evaluation by exploring how the relative share of Asian American college students on the establishments we take into account may change beneath numerous hypothetical admissions insurance policies. To simplify this train, we assume that college students admitted to an Ivy-11 faculty finally attend an Ivy-11 faculty. In keeping with our evaluation above, we prohibit our consideration to white college students and Asian American college students. Particularly, we maintain mounted the mixed variety of college students in these teams (roughly mirroring historic enrollment, as proven in Fig. S4), and so any will increase in Asian American enrollment essentially suggest decreases in enrollment of white college students. Any train of this kind is inherently speculative—partially as a result of modifications in admissions insurance policies might alter utility habits—however we nonetheless imagine it’s informative to gauge the approximate magnitude of results.

As a baseline, the highest row of Fig. 4 exhibits the estimated share of attendees in our knowledge from the three Asian subgroups of curiosity. The remainder of the determine exhibits the estimated share of attendees from these subgroups beneath eight hypothetical admissions insurance policies which might be divided into 4 classes. Within the first class—which we name “top-okay” insurance policies—we think about admitting college students with the very best ACT-equivalent scores, with ties damaged randomly. Within the second class, “random above threshold,” we take into account insurance policies that randomly admit college students above an ACT-equivalent rating t such that admitted college students have a imply rating equal to that of precise enrollees³⁰. For each of those classes we take into account two variants: the “ACT” variant selects from your complete applicant pool of the colleges we take into account, whereas the “ACT+ECs” variant selects solely from candidates with at the least as many hours of reported extracurricular (EC) actions over 4 years of highschool because the median of the hours reported by all enrollees. Underneath all 4 insurance policies, we estimate the identical or bigger shares of Asian American college students in comparison with what we observe within the knowledge. Asian American college students report, on common, fewer extracurricular hours than white candidates, so the ACT+ECs coverage variant leads to fewer Asian American attendees than the ACT variant.

The ultimate two classes we take into account examine outcomes beneath hypothetical insurance policies that preserve each the present variety of attendees from every state and the whole variety of attendees with legacy. Particularly, we first divide our historic knowledge into 102 (2 x 51) cells consisting of legacy and non-legacy candidates from every U.S. state and Washington, D.C.; we then in flip apply every of the 4 insurance policies described above to every of the 102 cells, guaranteeing for every cell that the variety of college students admitted beneath the hypothetical insurance policies matches the historic enrollment numbers. With these added legacy and geographic constraints, the share of Asian American attendees is smaller than beneath the unconstrained analogs, as anticipated given our outcomes above. However, even with these constraints, the variety of Asian American attendees throughout insurance policies remains to be just like or bigger than the established order.

Estimated enrollment of Asian American college students on the Ivy-11 beneath eight hypothetical admissions insurance policies, with the highest panel exhibiting the truly noticed demographic composition in our historic knowledge. In all circumstances, we take into account solely the subset of Asian American college students and white college students, and so will increase in Asian American enrollment correspond to decreases within the enrollment of white college students. In most situations, the hypothetical insurance policies we take into account result in a rise in enrollment of Asian American college students, together with those who protect the variety of legacy college students and the variety of enrollees from every state within the historic knowledge.