Economists have acknowledged that the undergraduate population of economics majors underrepresents women and minority groups. Less attention seems to be given to the graduate school admissions process, though disparities persist and perhaps worsen as students move into graduate school. A 2014 survey from the AEA suggests that over the last 10 years about 36% of undergraduate economics majors at top 10 ranked economics departments were female. However, only about 25% of first-year graduate students at the same institutions were female.* A similar trend holds for the top 20 institutions. These facts suggest a need for more analysis into our admissions process, and beg the question of whether our processes contribute to the "leaky pipeline."
The vast majority of people involved in admissions acknowledge the underrepresentation of women and minorities in the field. It is not individual acts of explicit bias or discrimination that seem to produce the lack of growth in representation over the last 20 years - rather it is the failure of our field to put in place systems and institutions that mitigate and counteract the implicit bias for which we are all culpable.
A large body of research by economists suggests that implicit bias generates discrimination in the labor force. For example, Bertrand and Mullainathan (2004) show that when identical resumes are assigned African-American-sounding names, applicants are called-back for an interview at only two-thirds the rate of an applicant with a white-sounding name. More recent work by Milkman, Akinola, and Chugh (2013) reveals the same gender and racial biases in an academic setting with faculty email response rates significantly higher for white and male students.
Given that implicit bias is pervasive, one can infer that the procedures we use in graduate admissions would not be free from this trap. Of course, anyone who has sat in on a conversation about recruiting female graduate students knows that the applicant pool (supply side) is highly unbalanced to begin with. This is an important issue that our field is beginning to address with efforts such as The Undergraduate Women in Economics Challenge. However, it does not preclude the need for faculty to pay attention to the application review and admission process (demand side). In fact, we need to evaluate and understand our process for building the graduate student cohorts so that we can address the supply side imbalances we face at the faculty-hiring stage.
Taking the supply side of graduate admissions as given, departments must ask themselves how they plan to build a cohort. Unfortunately, we do not currently have access to much data or research on how the admissions process works in various departments. Anecdotal evidence suggests that the guidelines given to admissions committees on how to evaluate candidates varies by department and remains vague.
Many departments employ some type of linear model that assigns value to candidates based on defined categories and produces a ranking of candidates from which the department can make admission decisions. However, evidence suggests that constructing a ranking of candidates may not be the best approach for constructing a cohort. Experimental research by Hoogendoorn et al. (2013) concludes that gender balance improves the success of teams, suggesting that departments should consider the composition of a cohort in addition to individual strengths. Scott E. Page, a professor at the University of Michigan, argues that organizations should invest in diversity to achieve their best and most productive workforce. This research implies that admissions models should consider complementarities between students.
Further, many of the measures used to evaluate applicants - GRE scores, GPA, research experience, and recommendation letters - may deliver systemic preference to certain demographic groups. Certainly statistical discrimination poses an issue - candidates from less-conventional backgrounds deliver a noisier signal than candidates that look familiar to those reviewing applications. Additionally, evidence suggests that these measures may include bias themselves. For example, Trix and Psenka (2003) find that recommendation letters written for women in the sciences tend to be shorter and to emphasize research skills less than recommendation letters for men.
More troubling, the measures and categories we use to evaluate applicant strength may not predict success in the field. The presumed objective function of an admissions committee is to admit a cohort of students who will (a) graduate and (b) go on to contribute meaningfully to the field (perhaps as measured by the prestige of the institution where they are hired). The measures used to assess applicants may be optimized to reduce attrition in the program, but may not select for future success. Informal analysis at various departments has shown that, conditional on admission, internal rankings of applicants does not predict future success in the field.
Practical steps that departments can take immediately to improve their admissions processes include collecting data on the admissions rankings of candidates and analyzing the long-term outcomes of students in their cohorts. Admissions committees, which often change year-to-year, can also record their goals and guidelines for building cohorts, so that we can evaluate and learn from past practices. Setting and communicating clear guidelines also helps to reduce bias and increase awareness of gender and racial disparity in the field. Finally, in order to take on and sustain this effort, departments should ensure that the transfer of roles (such as admissions chair) involves creating and transferring institutional memory.
When considering changes to the admissions process, matching the gender (and racial) distribution of incoming cohorts to that of the applicant pool is not the correct measure of success. It may be that female (and minority) applicants apply at lower rates because of discouragement or lack of information on how to apply, causing selection bias that would raise the average quality of female (or minority) applicants. Discussions with women currently in graduate school and in the process of applying corroborate this theory. A better measure of a successfully unbiased process is one where the expected marginal contribution of female (or minority) candidates is equal to the expected marginal contribution of male (or white) candidates. Of course, this would require establishing a measure of an individual’s contribution which, as previously discussed, is difficult in both theory and in practice. Nonetheless, collecting data on current admissions practices and considering what measures of contribution a department values are important steps towards improving the admissions processes.
Departments are intent upon contributing to a solution to the drop-off in diversity among graduate students. While there are checks in place to make sure that excellent candidates do not fall through the cracks, more can be done. With the preceding framework in mind, perhaps departments will engage with some of the fundamental questions important to improving our current systems.
* It should be noted that African Americans, and Hispanics/Latinos, Native Americans, first-generation college students, low-income students, and Pacific Islanders are also underrepresented but we have very limited data on these groups, so our analysis will pertain mainly to women. We intend our main message to extend to these other underrepresented groups.
Author: Emily Eisner