Harvard Arcidiacono Expert Report Asian Admissions
https://campusquest.blogspot.com/2019/04/harvard-arcidiacono-expert-report-asian.html
EXPERT REPORT OF PETER S. ARCIDIACONO
Students for Fair Admissions, Inc. v. Harvard
No. 14-cv-14176-ADB (D. Mass)
https://samv91khoyt2i553a2t1s05i-wpengine.netdna-ssl.com/wp-content/uploads/2018/06/Doc-415-1-Arcidiacono-Expert-Report.pdf
Asians have highest academic ratings, Asian rejects higher than African admits
To illustrate just how strong the Asian-American pool is, in the baseline
dataset Asian-American applicants have academic indexes that are over 0.2
standard deviations higher than whites, almost one standard deviation higher than
Hispanics, and over 1.5 standard deviations higher than African Americans. Indeed,
Asian-American rejects have academic indexes that are higher than AfricanAmerican admits.
Whites 44% higher than Asian, Blacks 3X White admit rates
• Race plays a significant role in admissions decisions. Consider the example of
an Asian-American applicant who is male, is not disadvantaged,3 and has
other characteristics that result in a 25% chance of admission.
Simply
changing the race of this applicant to white—and leaving all his other
characteristics the same—would increase his chance of admission to 36%.
Changing his race to Hispanic (and leaving all other characteristics the same)
would increase his chance of admission to 77%. Changing his race to AfricanAmerican (again, leaving all other characteristics the same) would increase
his chance of admission to 95%
Black admissions rates consistent with a [illegal]floor quota
For the three most recent admissions cycles, a period during which Harvard’s
Admissions Office has tracked admission rates by race using the federal
IPEDS (Integrated Postsecondary Education Data System) methodology,
Harvard has maintained African-American admission rates at nearly exactly
the same level as the admission rates for all other domestic applicants
(within 0.00064). The
probability that the difference in admission rates would
be smaller than 0.00064 in each of the three years is less than 0.2% absent
direct manipulation, and is consistent with Harvard having a floor on the
African-American admit rate.
Asians are qualified enough to be 50% not 20% of students
Despite being more academically qualified than the other three major racial/ethnic
groups (whites, African Americans, and Hispanics), Asian-American applicants
have the lowest admissions rates. In fact, data produced by Harvard show that this
has been true for every admissions cycle for the classes of 2000 to 2019.
A closer examination of the six years for which Harvard produced applicant-level
admissions data shows that even removing those who receive some other form of
preferences (such as legacy, athletic, or early action) still r
esults in Asian
Americans having the lowest admit rates over this period. For the Class of 2014
through the Class of 2019, Asian Americans made up roughly 22% of domestic
students admitted to the Harvard freshman class. If Harvard relied exclusively on
the academic index it assigns to each applicant in making domestic admissions
decisions, the
Asian-American share of its domestic admitted freshman class over
those same six years would be over 50%.
In particular, the most competitive applicants receive a 1 or 2 (the best
scores) on the Academic Rating. In the baseline dataset, 58.6% of Asian-American
applicants receive a 1 or 2, compared to 44.7% of whites, 14.7% of Hispanics, and
7.3% of African Americans. Asian-American applicants likewise have very strong
Extracurricular Ratings, again ranking higher on average than any of the other
three groups.
Asian-American applicants, however, do not score as well on the Personal Rating
and the Overall Rating
Race Blind Admissions Would Increase Asians by 50%
More stark are the
effects of removing all racial preferences for under-represented minorities, penalties against Asian Americans, and legacy and athlete preferences. The number of Asian-American admits would increase by 1,241 over the six-year period, a 50% increase.6
consistent with Harvard recruiting students from these ethnicities with
lower test scores.35
Second, Asian-American applicants have higher test scores than each of the other
racial groups. In every year, Asian applicants and admits have higher test scores
than white applicants and admits. And over the course of this period, AsianAmerican applicants had test scores between 88 and 125 points higher than African
Americans per section 36 and between 70 and 87 points higher than Hispanic
applicants per section. Indeed,
in every year Asian-American applicants had higher
test scores than either African American or Hispanic admits.
s prompted concern at Harvard that the new reporting would
understate the number of African-American admits to Harvard.37 The portion of the
admitted class that was single-race African American was below 7% for each of the
last three cohorts and the lowest fraction of the admitted class that coded as African
American under the old methodology in the last 19 admissions cycles was above 8%.
Table 1.1 reports admit rates for African-American applicants and all other
domestic applicants.
I can say with 99.8% confidence that Harvard has manipulated its
admissions process to ensure that the African-American admissions rate tracks the
overall admissions rate—it operates as a floor for African-American admit rates
over at least those three admission cycles.
simulation is set up so that the average probability of admission is
exactly the same for each group, regardless of where Harvard sets the cutoff for
admission: racial preferences for single-race African Americans exactly counteract
differences in the quality of the applicants across single-race African Americans and
other domestic applicants. In so doing, I maximize the probability that the two
admit rates will be close together.
38 Notably, the admit rate for single-race African-American applicants did not exhibit this
behavior before the admissions cycle for the Class of 2017 when Harvard’s Admissions
Office began using the IPEDS methodology. Because Harvard’s Admissions Office did not
code for race/ethnicity using the IPEDS methodology before the admissions cycle for the
Class of 2017, this type of data is unavailable for the Classes of 2014, 2015, and 2016. But
using the measures that are available, I am able to mimic the single-race African-American
admit rates in 2017, 2018, and 2019 and use this data to create similar single-race AfricanAmerican (and all other domestic applicant) admit rates for the Classes of 2014, 2015, and
2016. These results are reported in the second set of columns of Table B.1. The minimum
difference in admit rates for the years 2014, 2015, and 2016 are significantly higher. The
average difference between the pre-2017 cycles is 12.7 times higher than the average
difference in the post-2017 cycles.
Case 1:14-cv-14176-ADB Document 415-1 Filed 06/15/18 Page 33 of 168
30
However, for the three-year period
since Harvard began employing the IPEDS methodology to code race/ethnicity (i.e.,
for the Classes of 2017 through 2019), the admit rates for single-race African
Americans begin below that of other domestic applicants, then rise until they
approximate or exceed the admit rates for all other domestic applicants in midMarch through the end of the admissions cycle. In the 2017 and 2019 cycles, there
are points in June where the admissions rate for single-race African Americans are
as close to the domestic non-African American admit rate as they can possibly be
given the size of the admitted class and the number of applicants in each group.
This analysis further supports the conclusion that Harvard has imposed a floor for
African-American admit rates for at least the admissions cycles for the Classes of
2017 through 2019.
I show that Asian-American applicants are stronger on almost all
academic measures than those of other races/ethnicities, so much so that AsianAmerican rejects are stronger on some academic measures than African-American
admits. A
To illustrate just how strong the Asian-American pool is, in the baseline
dataset Asian-American applicants have academic indexes that are over 0.2
standard deviations higher than whites, almost one standard deviation higher than
Hispanics, and over 1.5 standard deviations higher than African Americans. Indeed,
Asian-American rejects have academic indexes that are higher than AfricanAmerican admits.
Asian-American applicants are less likely to be disadvantaged than
African-American or Hispanic applicants, but are more likely to be disadvantaged
than white applicants.44
