Peter Arcidiacono March 2017
Total Page:16
File Type:pdf, Size:1020Kb
Case 1:14-cv-14176-ADB Document 415-1 Filed 06/15/18 Page 1 of 168 EXHIBIT A Case 1:14-cv-14176-ADB Document 415-1 Filed 06/15/18 Page 2 of 168 EXPERT REPORT OF PETER S. ARCIDIACONO Students for Fair Admissions, Inc. v. Harvard No. 14-cv-14176-ADB (D. Mass) Case 1:14-cv-14176-ADB Document 415-1 Filed 06/15/18 Page 3 of 168 TABLE OF CONTENTS 1 Executive Summary ................................................................................................. 1 2 Background, Data, and Methods ........................................................................... 11 2.1 Background ...................................................................................................... 11 2.2 Data .................................................................................................................. 12 2.2.1 Data Sources .............................................................................................. 12 2.2.2 The Timing and Evaluation of Applications by Harvard ........................ 14 2.3 Methods ............................................................................................................ 17 2.3.1 Measuring the Role of Race in the Selection of Applicants for Admission .................................................................................................. 17 2.3.2 Measuring the Role of Race in the Scoring of Applicants ....................... 19 2.3.3. Selecting the Data for Analysis ................................................................ 21 2.4 Factors Correlated with Admission ................................................................ 23 3. Analysis ................................................................................................................... 24 3.1 Time Trends in the Treatment of Race ........................................................... 24 3.1.1 Admit Rates by Race/Ethnicity and the Quality of the Applicant Pool Over Time .......................................................................................... 24 3.1.2 There is Strong Statistical Evidence that Harvard Employed a Floor for African-American Admits for at Least the Post-2016 Admission Cycles ....................................................................................... 27 3.2 Waitlist, Admission, and Rejection Rates by Race/Ethnicity ........................ 31 3.3 Correlates of Admission: Objective Measures ................................................ 32 3.3.1 Academic Measures .................................................................................. 33 3.3.1 Non-Academic Measures .......................................................................... 34 3.4 Correlates of Admissions: Harvard Ratings ................................................... 35 3.5 Analysis of Harvard’s Ratings by Academic Index Deciles ........................... 40 ii Case 1:14-cv-14176-ADB Document 415-1 Filed 06/15/18 Page 4 of 168 3.5.1 How are Different Races/Ethnicities Distributed Across the Academic Index Deciles? ........................................................................... 41 3.5.2 How Do Admission Rates by Race/Ethnicity Vary Across the Academic Index Deciles? ........................................................................... 42 3.5.3 How Do the Rating Components Vary by Race/Ethnicity Across the Academic Index Deciles? ........................................................................... 46 3.5.4 How Do the Overall Ratings Vary Across the Academic Index Deciles? ...................................................................................................... 50 3.6 The Role of Race in Harvard’s Ratings ........................................................... 53 3.7 Statistical Analysis Shows a Penalty Against Asian-American Applicants in the Selection of Applicants for Admission. .............................. 61 3.8 Removing the Penalties and Preferences Associated with Race Would Significantly Increase the Number of Asian-American Admits .................... 69 4 There Is Additional Supporting Evidence that Racial Penalties and Preferences Work Against Asian-American Applicants and that the Predicted Harm Is an Underestimate ................................................................... 77 Appendix A ........................................................................................................................ Appendix B ........................................................................................................................ Appendix C ........................................................................................................................ Appendix D ........................................................................................................................ Appendix E ........................................................................................................................ iii Case 1:14-cv-14176-ADB Document 415-1 Filed 06/15/18 Page 5 of 168 1 Executive Summary I am a Professor of Economics at Duke University. My area of academic expertise is labor economics; I have published numerous peer-reviewed articles on issues of race/ethnicity and admissions decisions in higher education. I was retained by Students for Fair Admissions, Inc. in this case to review and analyze extensive data and information produced by Harvard in this litigation and to answer several questions about Harvard’s admissions process, using accepted econometric and statistical methods and techniques that I have used repeatedly in my published academic work for the past fourteen years: • Are Harvard’s admissions decisions biased against Asian-American applicants in the scoring and/or selection of applicants for admission? • What role does an applicant’s race/ethnicity play in admissions decisions made by Harvard? • Does Harvard set floors or ceilings on the admission of any racial/ethnic groups in making admissions decisions? To answer these questions, I reviewed a litany of materials provided by Harvard in this case, including: (1) data regarding individual applicants to Harvard from the classes of 2014-2019; (2) aggregate admissions data from the classes of 2000-2019; (3) the deposition transcripts and related exhibits of numerous Harvard officials; (4) training materials from the admissions office; (5) summary sheets and application files for selected applicants; and (6) reports from the admissions office and Harvard’s Office of Institutional Research. Using these materials, I constructed a database that permitted me to analyze how various factors—including race/ethnicity—affect admissions and Harvard’s scoring of the applications. I analyze the data using standard techniques for data where the variable of interest takes on a discrete number of values. For example, in analyzing admissions decisions, I code the dependent variable as one (if the applicant was admitted) or zero (if rejected) and estimate logit models of this decision. For Harvard’s ratings, the ratings are ordered such that lower numbers are associated with higher ratings and I use ordered logit models for the analysis. This approach is 1 Case 1:14-cv-14176-ADB Document 415-1 Filed 06/15/18 Page 6 of 168 consistent with generally accepted principles of econometric and statistical analysis, and has been used by experts in the field for the purposes of studying the influence of race in institutional decision-making generally, and in the field of higher education specifically. To analyze the individual applicant data produced by Harvard, I considered two distinct sets of applicants. The first “baseline” set included all domestic applicants who met each of the following criteria: (i) regular decision applicant; (ii) not a recruited athlete; (iii) not a legacy (i.e., the child of a Harvard alum); (iv) not a person appearing on the Dean’s or Director’s Interest List1; and (v) not the child of a member of the Harvard faculty or staff. Each of these characteristics is associated with a preference by Harvard, and thus an increased chance of admission. Excluding them from the baseline allows me to more easily compare similarly- situated candidates, and thus better perceive the role that race/ethnicity is playing (both positively and negatively) in Harvard’s admissions process. 2 Second, I analyzed an expanded set that included all domestic applicants and thus includes the groups excluded from the baseline dataset. In both datasets, I excluded a small number of individuals who were missing key pieces of information (such as both SAT and ACT scores). Employing statistical and econometric methods of analysis, it is my opinion, to a reasonable degree of certainty, that: • Asian-American applicants as a whole are stronger on many objective measures than any other racial/ethnic group including test scores, academic achievement, and extracurricular activities. • Asian-American applicants suffer a statistically significant penalty relative to white applicants in two of the ratings Harvard’s admissions officers assign to each file (the personal and overall rating). 1 These lists are used to identify candidates of particular interest to Harvard’s admissions office, See Fitzsimmons Depo. 268: 6-14. 2 Harvard previously has defended against claims it discriminates against Asian Americans by arguing that any disparity in admissions arises from its preferences for legacies and athletes, not its consideration of race. See HARV00023651; HARV00023143-44; Fitzsimmons Depo. at 371:19-374:3; Hansen Depo. at 114:7-115:19. 2 Case 1:14-cv-14176-ADB