QuantAdmit Research

What actually moves the needle at elite colleges

At the 25 most selective schools, grades and test scores barely tell admits apart from rejects — everyone applying is already excellent. So we fit a model on tens of thousands of self-reported outcomes to measure what decides the rest. Here’s what the data says, holding the school and everything else fixed.

The levers, ranked by effect

Odds multiplier at equal everything-else. Above 1 helps; below 1 hurts.

What it looks like for a strong applicant

Anchored to each school’s published acceptance rate — a strong applicant, then the same applicant with a geographic hook.

How to read this

Why do GPA and SAT barely predict admission at elite schools?

Because at the 25 most selective colleges, almost everyone who applies already has top grades and scores — the pool is academically bunched. In our fitted model, GPA and SAT alone separate admitted from rejected applicants at an AUC of just 0.67 (0.5 is a coin flip). Adding the fuller set of factors lifts that to 0.73 — a gain roughly six times larger than the same factors add across all colleges. Academics are the price of entry, not the deciding vote.

How were these effect sizes measured?

From a fixed-effects logistic model fit on tens of thousands of self-reported admission outcomes, restricted to the 25 most-selective schools and validated out-of-sample (on a held-out, more-recent slice). Each number is the factor's independent effect with everything else — including the school — held fixed. They're directional heuristics from self-reported data, not causal proof.

Are these odds real probabilities?

The example probabilities are anchored to each school's published acceptance rate, so a strong applicant sits sensibly above the school's overall rate and a geographic hook adds roughly ten points. But elite admission is inherently noisy: within a single top school, no model on observable stats separates admits from rejects much better than a coin — so treat any single number as a directional estimate with a wide honest range, never a promise.

Heuristic estimates derived from self-reported admission outcomes, restricted to the most selective schools and validated out-of-sample; anchored to published acceptance rates. Directional, not causal, and not advice about any individual applicant — elite admissions are inherently uncertain, and no model on observable stats predicts a single decision with confidence.

These outputs are estimates from a baseline model — not guarantees of admission, cost, or outcome.