It is relatively rare for an academic paper to both dominate the headlines in the education media and be covered by mainstream outlets, but a new paper by economists Caroline Hoxby and Sarah Turner did exactly that. The paper, benignly titled “Measuring Opportunity in U.S. Higher Education” (technical and accessible versions) raised two major concerns with using the number or percentage of students receiving federal Pell Grants for accountability purposes:
(1) Because states have different income distributions, it is far easier for universities in some states to enroll a higher share of Pell recipients than others. For example, Wisconsin has a much lower share of lower-income adults than does California, which could help explain why California universities have a higher percentage of students receiving Pell Grants than do Wisconsin universities.
(2) At least a small number of selective colleges appear to be gaming the Pell eligibility threshold by enrolling far more students who barely receive Pell Grants than those who have significant financial need but barely do not qualify. Here is the awesome graph that Catherine Rampell made in her Washington Post article summarizing the paper:
As someone who writes about accountability and social mobility while also pulling together Washington Monthly’s college rankings (all opinions here are my own, of course), I have a few thoughts inspired by the paper. Here goes!
(1) Most colleges likely aren’t gaming the number of Pell recipients in the way that some elite colleges appear to be doing. As this Twitter thread chock-full of information from great researchers discusses, there is no evidence nationally that colleges are manipulating enrollment right around the Pell eligibility cutoff. Since most colleges are broad-access and/or are trying to simply meet their enrollment targets, it follows that they are less concerned with maximizing their Pell enrollment share (which is likely high already).
(2) How are elite colleges manipulating Pell enrollment? This could be happening in one or more of three possible ways. First, if these colleges are known for generous aid to Pell recipients, more students just on the edge of Pell eligibility may choose to apply. Second, colleges could be explicitly recruiting students from areas likely to have larger shares of Pell recipients toward the eligibility threshold. Finally, colleges could make admissions and/or financial aid decisions based on Pell eligibility. It would be ideal to see data on each step of the process to better figure out what is going on.
(3) What other metrics can currently be used to measure social mobility in addition to Pell enrollment? Three other metrics currently jump out as possibilities. The first is enrollment by family income bracket (such as below $30,000 or $30,001-$48,000), which is collected for first-time, full-time, in-state students in IPEDS. It suffers from the same manipulation issues around the cutoffs, though. The second is first-generation status, which the College Scorecard collects for FAFSA filers. The third is race/ethnicity, which tends to be correlated with the previous two measures but is likely a political nonstarter in a number of states (while being a requirement in others).
(4) How can percent Pell still be used? The first finding of Hoxby’s and Turner’s work is far more important than the second finding for nationwide analyses (within states, it may be worth looking at regional differences in income, too). The Washington Monthly rankings use both the percentage of Pell recipients and an actual versus predicted Pell enrollment measure (controlling for ACT/SAT scores and the percentage of students admitted). I plan to play around with ways to take a state’s income distribution into account to see how this changes the predicted Pell enrollments and will report back on my findings in a future blog post.
(5) How can social mobility be measured better? States can dive much deeper into social mobility than the federal government can thanks to their detailed student-level datasets. This allows for sliding scales of social mobility to be created or to use something like median household income instead of just percent Pell. It would be great to have a measure of the percentage of students with zero expected family contribution (the neediest students) at the national level, and this would be pretty easy to add onto IPEDS as a new measure.
I would like to close this post by thanking Hoxby and Turner for provoking important conversations on data, social mobility, and accountability. I look forward to seeing their next paper in this area!