Research

Researchers examine 'race unknown' enrollment data in higher education

UNIVERSITY PARK, Pa. — When a college student self-identifies as “race unknown,” what does that mean in the context of higher education research? According to researchers at Penn State and Michigan State University, the “race unknown” category does not represent random “noise” in data collection but rather can be attributed to some combination of student responses and data collection practices. Additionally, they discovered high concentrations of “race unknown” enrollments in certain institutional types (the most and least selective).

Therefore, the research team suggests that researchers refrain from dropping “race unknown” from their studies and also from interpreting the results for the “race unknown” category because “it is not a conceptually meaningful ethnoracial group.”

“We’re hoping that we change the way (higher education) researchers think about racial groups,” said Karly Ford, an assistant professor of education (higher education) in the Department of Education Policy Studies in Penn State’s College of Education.

Credit: Penn State / Penn StateCreative Commons

According to Ford, higher education researchers often drop the “race unknown” category when examining college enrollments and doing so changes the racial compositions of student bodies. A problem with that approach, she added, is that if “you drop this group it makes the percentages of other groups look larger because you’ve taken a slice out of the pie, so the other pieces of the pie all get bigger.”

“We just can’t ignore the fact that we don’t have great race data on students,” said Kelly Rosinger, also an assistant professor of education in the Department of Education Policy Studies. “If we really want to understand enrollment, persistence, completion, borrowing or other outcomes by race (in higher education), we need to understand the limitations that exist in data collection.”

Kelly Rosinger, assistant professor of education (education and public policy) in the College of Education. Credit: Penn State / Penn StateCreative Commons

Ford and Rosinger, along with Qiong Zhu, a postdoctoral research associate in the Education Policy Innovation Collaborative at Michigan State University, present their findings in their paper, “What Do We Know About ‘Race Unknown,’” which was published recently in Educational Researcher. Their results were based on data obtained from the Integrated Postsecondary Data System (IPEDS), a system of interrelated surveys conducted annually by the National Center for Education Statistics (NCES), a part of the Institute for Education Sciences within the United States Department of Education.

According to the researchers, IPEDS classifies students who elect not to identify themselves under a particular racial or ethnic designation as “race and ethnicity unknown.” In 2009, around 175,000 full-time, first-time undergraduate students, or 7% of those enrolled, were reported as “race unknown.” In contrast, in 2017, “race unknown” represented 3% of enrollment and approximately 69,000 students.

Using IPEDS data, the researchers constructed a 28-year dataset of 4,401 institutions to examine trends and patterns of “race unknown” enrollment from 1990 to 2017. They found that for institutions in the for-profit sector, 5% to 18% of students fell into the “race unknown” category. In addition, almost 10% of students attending the most selective institutions were reported as “race unknown” in 2009 before dropping off precipitously.

“Despite the considerable size and fluctuation of this category, ‘race unknown’ has remained largely unexplored,” the researchers stated in their paper.

Although the researchers’ current study does not specify which racial groups comprise “race unknown,” they attempt to shed light on the phenomenon by examining variation in the proportion of “race unknown” enrollment reported by each institutional type. Because IPEDS does not have more nuanced information about race reporting, they wrote in the paper, the common assumption among researchers is that students categorized as “race unknown” amount to measurement error, or error that is randomly distributed across institutions. This method also assumes that the process by which institutions generate this data is uniform across the board. On the other hand, Ford and colleagues state that both elite universities and the least selective have high rates of “race unknown” students and that there are a number of competing explanations for this phenomenon. 

“I wanted to test that assumption to see is this actually random error, and it turns out that it’s not,” Ford said. “We think there are two processes happening at opposite ends of the selectivity spectrum of institutions.”

Ford was quick to add that they are speculating about what processes might be driving “race unknown” enrollments. The research team describes what is happening, and more work needs to be done to uncover the “how” and “why.”

“We imagine that less selective institutions don’t have bandwidth to follow up with students and don’t have the resources or infrastructure to manage the data,” she said.

According to Ford, there may also be strategic reasons why certain institutions, particularly those with high enrollments of minoritized students, may choose to conceal their racial makeup. Between 1990 and 2017, 140 institutions reported 100% of students in the “race unknown” category, effectively not reporting racial data on any student.

Conversely, the researchers reported, 60% to 70% of institutions in the 1990s and 30% of institutions in the most recent year reported zero “race unknown” enrollment, suggesting that many of these institutions may assign a race and ethnicity to students based on the observations of their personnel. In a previous study, Ford examined “observer identification,” a process by which K-12 school personnel select a racial or ethnic identification for a student. Her conclusion was that “observer identification” poses a potential threat to the validity of self-identified race/ethnicity data largely because evidence from various sources suggests that “about 40% of the time, observer identification does not match self-identification of some of the fastest growing racial/ethnic groups in the K–12 population,” such as Latinx and multiracial populations.

On the other hand, Ford said, in certain highly selective institutions, some students fear that by disclosing their racial identity, they would be less likely to gain admission. In the paper, the authors cite a Boston Globe story in which a director of a college admissions coaching company stated that he advises Asian clients to play down their ethnic identity — e.g., not mentioning that they engage in certain activities such as playing violin — to increase their chances of gaining admission to prestigious universities.

One possible explanation for the drop-off in “race unknown” enrollment that occurred in 2010, Ford said, is the introduction of the “two or more races” category. Prior to the IPEDS reporting changes, students in that category would have been reported as “race unknown.” However, she added, the proportion of students reporting “two or more races” is only 2% to 3% of the population, and therefore does not fully account for the 7% to 15% “race unknown” drop-off in the most selective institutions.

“We think the more important thing is that the wording of the question changed,” Ford said, noting that in 2010 it became much more direct and no longer provided a “prefer not to answer” option — which may prompt more people to disclose their race.

Due to the limitations of their data, Ford and Rosinger emphasized that further study is needed to examine both the institutional challenges of quality data collection and why some students choose to refrain from identifying their own race and ethnicity.

“We need much more inclusive practices when it comes to capturing race data,” Rosinger said.  “It’s so important for students to see themselves reflected in the data.”

Last Updated April 15, 2021

Contact