After an unexpected hiatus, we’re back to our regularly scheduled blog posting this week! Did you miss us? In this issue, I look at the results of a new report about the updated earnings test and why it may not be what it seems. I then close with three Sparks about monitoring students for AI cheating, rethinking AI detection, and state politics and college enrollment.   

Who is Benefiting from the Earnings Test? 

From New Earnings Test Will Have Limited but Concentrated Impact | Inside Higher Ed 

A new report from the Postsecondary Education & Economics Research (PEER) Center at American University finds that only about 2.5 percent of federally aided students are enrolled in programs that are predicted to fail the new ED earnings test. 

Our Thoughts  

While I’ve linked to the news article from Inside Higher Ed, the report and interactive website are actually much more interesting. Not only is the report deeper, but the interactive website lets you look at how individual institutions and programs are doing. Go search your own or somewhere you know. I’ll wait. #datanerd Okay, now that you’re back, let’s talk about this report, what it means, and where maybe the entire test falls short.  

For most of you reading this, the report is good news. Fewer than 500,000 students, who received federal financial aid for their studies, are in programs projected to fail the new earnings test. Those 500,000 are mainly concentrated in undergraduate certificate programs like cosmetology and medical assisting or enrolled in for-profit colleges. About 70 percent of community colleges, 75 percent of public four-year institutions, and 84 percent of private non-profit four-year institutions are projected not to have a single student in a failing program. To me, that sounds pretty great if it means that programs are serving students and helping them find better lives because of their postsecondary education. Then the authors of the report, one of whom was the first chief economist at the Department of Education, crushed my dreams about this test. 

The authors believe the test is too weak. It targets programs at the very bottom of the earnings range, so the vast majority pass easily and face little pressure to do better. Let’s unpack that a little, because when I read about graduate programs, that’s when I started to ask questions.  

Undergraduate programs are easier to explain. They simply have to beat the median earnings of 25- to 34-year-olds in their state who only have a high school diploma. Maybe not the highest bar because there’s no required margin (say a 10 percent earnings premium), but it’s easy to understand. Graduate programs have to beat 25- to 34-year-olds with only a bachelor’s degree, but they get three metrics to compare against: bachelor’s holders in the same state and field (at the two-digit CIP level), all bachelor’s holders in the same state, or bachelor’s holders nationally in the same field. The program’s median earnings only have to be higher than the lowest of those three numbers. So, the median earnings for your program could be less than the median earnings of bachelor’s holders in that field in your state, but as long as your program’s median is higher than one of the other two options, your graduate program gets a pass.   

I guess I was hoping that these rules would measure value, but they never consider price in their equations. Imagine two MSW programs, one at a public university charging $25,000 in tuition and one at a private university charging $60,000. Graduates of each of those programs go on to work for the county children’s services agency on the same pay schedule. Those programs pass or fail together because earnings are all that is being measured. While people argued about the previous gainful employment framework, it at least paired the earnings test with a debt-to-earnings measure, which the new rules dropped. The report’s discussion section actually suggests bringing back debt-to-earnings limits for graduate programs in low-earning fields, and I wholeheartedly agree with that idea.  

I’m not trying to call out any graduate programs at any institutions; I understand the complexity of higher education finances. But we have to begin drawing more attention to the costs of education compared against expected earnings in some fields. Let’s stick with social work and licensure requirements. In most states, becoming a licensed clinical social worker means roughly 3,000 hours over at least two years of supervised practice after the MSW. Professional counselors are another field who face similar requirements. Comparing MSW graduates to bachelor’s holders means comparing them to people who can’t legally do the job. So, the test ends up measuring whether the profession pays more than generic bachelor’s-level work. As I argued in Issue 101, that’s largely a question of licensure-tiered pay scales and how little we’ve chosen to pay people in majority-female helping professions. Earlier rules accounted for part of this, allowing some programs that require supervised practice early in careers an extension on when they have to begin their comparisons. No such extended window exists in the current law.   

This doesn’t mean that every graduate program that fails is a victim of bad measurement. Among master’s programs in mental and social health services, about 25 percent of federally aided students are in programs projected to fail. Among master’s programs in clinical counseling and applied psychology, many of which prepare students for the same counseling licenses, it’s about 1 percent. Social work doesn’t appear among the 20 graduate fields with the most students in failing programs at all. The failing programs in that first group are concentrated at institutions that operate mostly or entirely online, and online programs drawing students from across the country get measured against national benchmarks. The data can’t tell us whether that pattern reflects program quality, who those programs enroll, or where their graduates end up working. I’d want to know all three before drawing conclusions, because the test may be catching something real in those programs while still measuring the broader field poorly.  

In Issue 101, I wrote that institutions would be wise to have a more nuanced version of the story ready to tell, and this is the moment that advice was for. ED starts calculating results in 2027, and penalties can’t take effect until 2028 at the earliest, so IR and registrar offices have time to get ahead of the federal number if they start now. Look up your programs on PEER’s website. Trace the story of the programs that seem close to the threshold. See if they require licensure that depresses early earnings or if graduates follow a specific path that the data doesn’t capture. Be prepared to have conversations with your legislators and trustees. While none of that will change the federal calculation, you’ll at least have the other half of the story ready when someone asks.   

Sparks 

• The New Nanny State (The Chronicle of Higher Education) - Beth McMurtrie looks at the new ways professors are attempting to monitor students to combat AI cheating. Seems like a lot of work to keep treating the take-home paper as a sign of learning. 
• More U.K. Universities End Turnitin Contracts as Concerns Mount (Inside Higher Ed) - Several U.K. universities are considering ending their relationship with Turnitin as they question whether they still need it. Maybe we should take some notes from our higher ed friends across the pond? 
• More students are ruling out colleges based on politics, poll finds (Higher Ed Dive) - A recent Art & Science Group poll found that 34 percent of students had ruled out a college solely based on the politics of the state where it is located. That’s a new challenge for enrollment managers to try to overcome.