The SAT is a good thermometer. It tells you how warm a student already is. It cannot tell you how much heat the university adds.

This post is a summary of findings discussed in my article “Selection or Value Added? Earnings Returns to University of California and California State University Attendance“

The University of California dropped the SAT and ACT in 2020. Six years later, the decision is back on the table. Hundreds of UC faculty have called for reinstating tests, at least for STEM applicants. Regent Jay Sures has said his 2020 vote might have been a mistake. The Academic Senate’s admissions committee, BOARS, owes the Senate a testing study by February 12, 2027. The Senate’s recommendation is due by June 2027, and the Regents are expected to vote in July 2027.

The current debate is focused on introductory math grades, first-year GPA, and retention. But families don’t choose a UC for its first-year GPA; both students and their supporters are focused on job prospects. As I have shown in my previous blogs, UC families pay roughly $6,900 more per year for a UC over a CSU because they expect a better career. So what does the SAT tell us about earnings? I used federal tax-based earnings data for UC and CSU campuses, linked to each campus’s average SAT score and family income.

Finding 1: The SAT strongly predicts earnings.

I started with the Opportunity Insights college file, which lists each campus’s average SAT score and its alumni’s median earnings 10 years after entry. I ran a regression of earnings on SAT. SAT alone explained 69% of the earnings differences between campuses, and each 100 SAT points went with about 9% higher earnings. I repeated it across 356 public universities nationwide, comparing schools within the same state, and got 9% to 12% per 100 points.

To check whether SAT accounts for the UC advantage, I added a UC-versus-CSU indicator to the same regression. On its own, the indicator showed UC alumni earning about 17% more. Once SAT was in the model, it dropped to about −6%, which is not statistically significant. I ran the same check on the age-34 earnings from the second file, and the UC premium went from +26% to −6%.

Campus SAT scores explain 69% of the earnings differences between UC and CSU campuses. Every 100 points of average SAT adds about 9% to alumni earnings ten years after entry. Across 356 public universities nationwide, the figure ranges from 9% to 12%.

That single number accounts for nearly all of the UC advantage. At age 34, UC alumni earn 26% more than CSU alumni. But when I use the same SAT score, the advantage disappears. UC campuses then come out about 6% behind CSUs- a gap too small to be statistically meaningful.

The pattern matches research on individual students. Chetty, Deming and Friedman studied applicants to the most selective U.S. colleges. They found that SAT and ACT scores are highly predictive of post-college success.

So the faculty pushing for tests have a point. A test score carries real information about a student’s earnings potential.

Finding 2: The SAT misses what the campus adds

A test score predicts where a student starts. It says nothing about what a campus does with that student. To see that part, I took 356 public four-year universities across the country. For each one, I gathered what its alumni earn ten years after starting college, its students’ average SAT score, its mix of majors, its student demographics, its size, and its state. The state matters because wages differ so much from place to place.

Next, I let a statistical model learn how those factors move earnings. This is a regression, and enrollment leaders use the same logic when they build a yield model. Across those 356 schools, it found that 100 more SAT points goes with roughly 9% to 12% higher alumni earnings, and it learned how much majors and demographics add or subtract. The model never knew which schools were UCs or CSUs.

Then I gave the model each UC and CSU campus’s profile and asked what its alumni should earn. The bars in the chart show the gap between that prediction and what alumni actually earned. Cal State San Bernardino’s students averaged an 890 SAT, and the model predicted about $39,100. Its alumni earned $45,500, which is 16% above the prediction. UC San Diego’s students averaged 1310, and the model predicted about $65,600. Its alumni earned $59,600, which is 9% below.

Finally, I compared the two systems. The average UC campus landed 5% below its prediction, and the average CSU landed 3% above. A gap that size would almost never appear by chance (p = .002). The campuses that add the most are the ones a test-based ranking would put last. Cal State San Bernardino enrolls students with an average SAT of 890. Its alumni earn 16% more than those numbers predict. UC San Diego enrolls students averaging 1310. Its alumni earn 9% less.

The model cannot see everything about students, such as motivation or family wealth beyond what demographics capture, so part of each bar may reflect students rather than the campus. It compares campuses, not individuals. The SAT also misses the shape of a career. UC alumni start slower than their scores predict, then climb faster. At the same SAT, they earn 18% less than CSU alumni at age 28. By 34, the shortfall shrinks to 6%.

Even after controlling for SAT and family income, the UC–CSU gap widens by about one to two percentage points every year.

Two things are true at once. The SAT is a strong signal of who arrives. It is a weak signal of what happens next.

The campuses with the lowest scores add the most to their students’ earnings, so a test that favors students who would succeed anywhere tells UC little about where its degree makes a difference. Scores also travel with family income. In these cohorts, UC students came from families at the 67th income percentile and CSU students from the 61st, so part of what the SAT predicts is what a family could already provide. And the early test-free evidence is encouraging. UCLA reports higher retention and graduation since 2020, along with a larger share of applicants from the top 9% of their high school class.

The testing debate asks whether the SAT predicts success. The earnings data says it does. The SAT measures intake, not impact. If the goal is to admit students who will earn the most, the SAT helps. If the goal is to admit students whose lives a UC degree will change the most, the SAT is a blunt tool. Those are different goals.

I think that the better question is what UC wants its admissions to accomplish. If the goal is to admit students who will earn the most, the SAT helps. If the goal is to admit students whose lives a UC degree will change the most, the SAT is a blunt tool.

The SAT is a good thermometer. It tells you how warm a student already is. It cannot tell you how much heat the university adds.

About the data

Sources: Daily Bruin: UC Regents discuss standardized testing (Sept. 18, 2026) · Compass Education Group: UC testing policy status · Chetty, Deming and Friedman, “Diversifying Society’s Leaders?” NBER w31492 · College Scorecard institution data documentation · Opportunity Insights Mobility Report Cards, Tables 5 and 10 (downloaded data files).