Why Do We Need to Rethink STEM PhD Admissions Procedures?
LOW COMPLETION RATES
Six-year completion rates for U.S. STEM doctoral students average around 40% (see graph below) and 10-year rates hover around 60%  suggesting that the selection process may not be optimized for identifying students with the skills and personal characteristics needed to complete STEM PhD degrees.
HIGH SOCIETAL COSTS
Approximately 37,000 students enter STEM PhD programs annually , but 15,000 do not complete. The annual cost of educating a STEM doctoral student generally ranges between $50,000 and $75,000. If the 15,000 students who leave do so after one year, the annual cost approaches $1 billion. Many students stay for 2-3 years and leave with masters’ degrees. Institutions do not generally fund STEM masters training and, instead, charge students tuitions and fees, so institutions lose revenue. If even half of those who do not complete take this route, the cost approaches $2 billion per cohort.
These calculations do not include the personal investments of students who do not complete degrees or the time and resources invested by faculty and staff in training these students.
STEM DOCTORAL TRAINING IS CHANGING
STEM doctoral education is increasingly replacing didactic classroom learning and individual research projects with problem-based learning and interdisciplinary team research to solve complex problems. At the same time, the selection process has not changed dramatically for decades and still relies heavily on Graduate Record Examination (GRE) scores . The GRE measures analytical ability, but not creativity or practical abilities , nor does it measure interpersonal skills and collegiality, persistence, personal and professional integrity or maturity and responsibility, characteristics valued by graduate faculty [5, 6]. This means that the main measure used for STEM doctoral student selection is a test that does not measure skills needed to succeed in modern STEM disciplines.
U.S. DEMOGRAPHICS ARE CHANGING
According to the 2010 census, the resident population of the U.S. was comprised of 64% White, 15% Hispanic, 14% Black and 7% Asian American. The current projection is that by 2050, there will be no group in the majority. But by 2060, 56% of the population will be comprised of groups now considered minority. In fact, in 2015, 40% of the U.S. college students were from minority groups.
ADMISSIONS COMMITTEES MAY BE IGNORING GRE VERBAL SCORES AND SETTING CUTOFFS FOR GRE QUANTITATIVE SCORES
We recently conducted a study (Petersen, Erenrich, Levine, Vigoreaux and Levine; submitted) of over 1800 U.S. STEM doctoral students that related GRE scores and PhD completion institutions in four Northeast Alliance for Graduate Education and the Professoriate (NEAGEP) institutions. One of our findings was that GRE Verbal scores for students are wide-ranging and normally distributed, suggesting that they are not a selection factor. On the contrary, there is clearly a shift to the right for GRE Quantitative scores with most scores above 650.
WOMEN AND STUDENTS FROM LOWER SOCIOECONOMIC OR UNDERREPRESENTED GROUPS DO NOT SCORE AS HIGHLY AS WHITE OR ASIAN COUNTERPARTS ON THE GRE QUANTITATIVE
Despite the perception of the GRE as a fair and equitable test, women and students from underrepresented and lower socioeconomic groups score less well than other groups [7-9]. The graph on the left (from Miller and Stassun, ) shows that those from underrepresented groups historically score below levels major institutions have as official or unofficial cutoffs for admission. Miller and Stassun note, “In simple terms, the GRE is a better indicator of sex and skin color than of ability and ultimate success.”
Sedlacek  suggests that if diverse groups have different experiences and different ways of presenting their capabilities, it is unlikely that a single measure could be developed to produce outcomes fair to all. He also suggests that equality of results, not process, should be the goal of a measure used for admissions.
GRE SCORES ARE NO LONGER A MAJOR FACTOR USED TO EVALUATE THE QUALITY OF GRADUATE STUDENTS OR PROGRAMS
- The NIH has dropped the requirement for T32 training grants to report GRE scores of trainees
- GRE scores are no longer required for individual F30 or F31 fellowship applications
- GRE scores are not part of the NSF Graduate Research Fellowship Program Application
- For the first time in their ranking of graduate education programs, U.S. News and World Report adjusted the weight of GRE scores used to determine the 2019 institutional ranking downward for schools in which less than 50 percent of the entering doctoral students submitted new GRE scores. The GRE score weight was at maximum 0.06 for Verbal and Quantitative scores. For comparison, research activity is weighted by 0.30, Faculty Resources by 0.12 and Peer Assessment by 0.25
COLLATERAL DAMAGE OF RELYING ON GRE SCORES
- Those who have anxiety about high-stakes tests may underperform at the beginning of the GRE so that their scores are lower because of the section-adaptive format
- Those in lower social economic groups cannot afford expensive, but effective, test
- Students with test anxiety may rule out graduate school before even taking the test
- Test scores may determine the programs to which the students apply
- Students with lower than average test scores may define themselves by the numbers
- Faculty expectations may change if they know GRE scores
- Diversity is limited
WHY IS THERE RESISTANCE TO DEEMPHASIZING OR DROPPING GRE SCORES?
Despite evidence that GRE scores are not reliable predictors of STEM doctoral success and ETS recommends deemphasizing the scores, many faculty resist changing the practice of relying heavily on GRE scores. Several reasons have been proposed:
- Few exemplars of successful candidates with low scores if only choose students with high GRE scores
- Current faculty likely had high GRE scores and “we did well”
- When high scorer leaves, faculty accept that “it wasn’t right for that person”, but when low scorer leaves, faculty suggest “it is predictable based on GRE scores”
- The test has been around for nearly 70 years
- They are numbers, so are viewed as objective measures
- Speeds up admission process