Do test scores still matter? The colleges' own numbers say yes

The finding, in plain English

The arithmetic is simple. If 36% of a college's applicants submitted a score but 51% of its admitted students did — USC's own published figures for the Class of 2028 — then submitters must have been admitted at a meaningfully higher rate than non-submitters. Work the arithmetic through and the advantage comes to about 1.9×. That's derivation from the school's own numbers, not modeling. The same calculation runs 3.2× at Cornell Arts & Sciences, 2.0× at Notre Dame, and 1.2× at Vanderbilt (table below).

The colleges have started saying it out loud. 23 well-known colleges — including Harvard, Stanford, MIT, Yale, Caltech, UT-Austin, and Cornell (effective Fall 2026) — have reinstated a testing requirement after studying their own test-optional results. Across the 1,396 four-year colleges in our database that report a test policy (federal IPEDS, 2023), 1,264 remain test-optional, 101 require scores, and 31 — including every UC and CSU campus — do not consider scores at all.

The research explains why. The Opportunity Insights team found that among students with the same high-school grades and background, moving from a 1200 to a 1600 SAT predicts a first-year college GPA about 0.43 points higher — while moving from a 3.2 to a 4.0 high-school GPA predicts a difference of less than 0.1. As grade inflation has flattened the GPA signal, the score has become one of the clearest pieces of information a college has.

Four takeaways for a family deciding whether to test

1. At selective test-optional colleges, a good score still helps. The schools' own numbers show submitters admitted at 1.2 to 3.2 times the rate of non-submitters. "Optional" describes the checkbox, not the outcome.
2. Testing keeps the whole list open. With 101 colleges requiring scores — 23 of them well-known schools that brought the requirement back — a student who never tests has quietly crossed schools off the list before applying. A student who tests can still choose, college by college, whether to submit.
3. When in doubt, submit — even slightly below the range. Dartmouth's faculty study found that less-resourced students were withholding scores around 1400 that would actually have helped their applications — because a missing score doesn't read as neutral. Admissions research suggests that when no score is sent, readers tend to assume the worst — a score near the bottom of what submitters bring — so a real number slightly under the 25th percentile usually beats the blank. The practical rule: compare the score to each college's middle-50% band and submit when it's in the range, above it, or a little below. Our searchable score-range table covers 985 colleges' bands.
4. Scores now carry more signal than grades. Opportunity Insights: a 1200-to-1600 SAT difference predicts about 0.43 points of first-year college GPA; a 3.2-to-4.0 high-school GPA difference predicts less than 0.1. UT-Austin found score submitters earned a first-semester GPA 0.86 points higher — a key reason it reinstated its requirement. To see what a given score does to admit odds at a specific college, our free score-odds calculator runs the numbers.

Method & sources: Submission shares are each school's own published figures: Cornell Office of Institutional Research & Planning (Test Score Task Force, Fall 2023); USC Undergraduate Admission and Common Data Set 2024-25 (Class of 2028); University of Notre Dame Admissions (Class of 2026 Restrictive Early Action); Vanderbilt Office of Undergraduate Admissions (Class of 2025). The submitter-advantage multiple is the odds ratio implied by the applicant and admit submission shares — it requires no assumptions beyond the school's own two numbers. Test-policy counts: federal IPEDS Admissions (2023), with publicly confirmed corrections for test-blind systems (the UC and CSU campuses) and reinstated requirements. Predictive-validity findings: Opportunity Insights (Friedman, Sacerdote & Tine, 2024), Dartmouth faculty study (2024), UT-Austin announcement (2024) — all linked with full context on our test-scores page. See the full methodology.