Getting in is only half the story. Once the acceptance letters arrive, families face a harder question: which one to choose. The patterns are remarkably consistent year after year — and almost entirely predictable from a few numbers.
Yield — the share of admitted students who actually enroll — is the single cleanest signal of how badly students want a school. At MIT, 86.6% of accepted students show up. At the average four-year college, only 30% do.
The pattern is monotonic. HYPSM schools sit between 70 and 87 percent. Ivy+ peers cluster in the high 50s and 60s. Near-Ivy names like Notre Dame and Vanderbilt land in the mid 40s to high 50s, and most Selective privates fall to 35-50%.
These numbers don't move much from year to year. Princeton has been climbing slowly out of the high 60s, Brown jumped sharply in the most recent class, and Columbia's yield has drifted down. But the rank order has been stable for a decade.
When students hold offers from two schools, the higher-yield one usually wins. But by how much?
Cross-admit data — what students who get into both schools actually do — reveals a brutal pecking order. Among the 100,000-plus revealed enrollment decisions Parchment has tracked, the higher-prestige option wins about three-quarters of the time when tiers differ.
Even within HYPSM, the rankings are sharp. Harvard beats Yale 63-37. Harvard beats Princeton 73-27. Stanford has briefly surpassed Harvard in STEM-heavy samples but loses 65-35 in Parchment's broader pool. MIT, when compared head-to-head, beats Harvard 63-37 — the only school that does.
These margins have confidence intervals tight enough to take seriously: Harvard-Yale at 63% has a 95% Wilson CI of [58.8, 68.2]. The gaps are not noise.
So prestige dominates. But it isn't the only lever — and for most families, it isn't the strongest one.
When NACAC, BestColleges, EAB, and NCES surveys ask students why they picked one college over another, the answers cluster into a remarkably consistent ranking. Prestige tops the list. Net cost is right behind it. Major strength, location, and campus culture round out the top five.
Climate of all things shows up in the top 10 — possibly explaining why Stanford, Duke, and Vanderbilt punch above their weight against equally selective Northeast peers. Family influence and admitted-student weekends matter, but less than students themselves often realize.
The order matters more than the exact weights. Fit beats fame at the margins, but only at the margins. Most students choose along the prestige axis when prestige differs.
Then the price tag arrives. And for most families, that's when the decision actually starts.
Three decades of natural-experiment research converge on the same number: an extra $1,000 of grant aid raises the probability a student enrolls by roughly 2 to 4 percentage points. Cal Grant studies push that range as wide as 1.2 to 9.2 pp depending on assumptions. DC's tuition assistance program: 3-4 pp per thousand.
But the effect isn't uniform across the prestige ladder. At HYPSM, an extra $10,000 changes yield by 1-3 percentage points — most admitted families already qualify for need-based aid, and prestige dominates the rest. At ordinary selective colleges, the same $10,000 can swing yield by 8-15 percentage points.
There's also a threshold effect. The jump from $0 to any merit scholarship matters more than the dollar amount. A $5,000 award often produces almost the same yield bump as a $15,000 one — the signal that the school chose you appears to do most of the work.
The dollar number masks something deeper: families with different incomes are weighing entirely different tradeoffs.
Aggregate yield numbers conceal something important. The same college choice looks like three different decisions depending on family income.
For families under $60,000, net cost dominates everything else. Prestige still matters, but a meaningful aid gap can flip a Tier 4 school over a Tier 2 one. Location matters enormously — staying close to home is often nonnegotiable.
For middle-income families ($60K-$150K), prestige and aid trade off roughly evenly, with major strength rising in importance. For families above $150,000, the calculation inverts: prestige becomes dominant, location largely stops mattering, and campus culture ascends to a top-three concern.
The implication for any yield model: a single elasticity number is wrong. Aid sensitivity has to be tier-modulated and income-modulated at the same time.
Schools below the top tier can't rely on prestige alone. They need another lever — and many have found it.
Ask Harvard whether it tracks how many emails you opened. The answer is no. Ask Tulane the same question. The answer is yes — and the data feeds your admissions file.
Roughly 16% of colleges rate demonstrated interest as "moderate" or "considerable" in their admissions decisions. Almost none of them are at the very top. Every HYPSM school explicitly ignores it. Most Ivy+ peers do too. The schools that do track it cluster in the 20-50% acceptance band, where yield management is a daily concern: Tulane, Vanderbilt, WashU, Tufts, Boston College, Carnegie Mellon, GWU.
At those schools, an in-person campus visit raises admission probability by about 30%, according to a Lehigh-based study. Costlier signals beat cheaper ones — a flight to admitted-student weekend matters more than an opened email. Early Decision is the strongest signal of all, because it binds.
Speaking of binding: the rest of yield management runs through one round in particular.
Early Decision is the cleanest yield management tool ever invented: accepted ED, you go. Period. At Northwestern and UPenn, more than half the class is locked in before March. Duke and Vanderbilt sit just behind.
The acceptance-rate gap is enormous. ED rates run 3 to 5 times higher than RD rates at the same school. Northwestern admits about 25% of its ED pool and roughly 5% of its RD pool. That gap is often described as a "boost" for ED applicants, but most of it is composition: ED pools self-select for legacies, recruited athletes, and the unusually well-prepared.
From a yield engineer's perspective, the math is simple. A school filling 50% of its class through ED has guaranteed perfect yield on half its admits before the rest of the cycle even begins. The remaining yield problem shrinks accordingly — and so does the reported yield-rate volatility.
For everyone left over, there's one last queue. Most won't move.
At the most selective schools, the waitlist is mostly a public-relations instrument. Princeton offered ~1,400 waitlist spots for the Class of 2029 and pulled 40. Yale offered 943 and pulled 23. MIT offered 620 and pulled 9. The pull rates run from under 2% to roughly 10% in the top tier — and the year-to-year variance is enormous.
Brown is the recent outlier: a Daily Herald poll suggests ~230 waitlist admits for the Class of 2029, an unusually high pull rate near 19% driven by under-yielding in the regular round. The historical range is 15 to 120. One bad yield year, and the waitlist becomes the budget reconciliation tool.
Mid-tier schools pull in genuinely meaningful numbers — the national average admission rate from a waitlist hovers around 20%, with some Selective schools reaching 40% in light years. But the students admitted from waitlists yield at lower rates and receive worse aid packages, on average, than direct admits.
Every spring, families experience the May 1 deadline as a private, high-stakes choice. From above, the choices look almost mechanical. Prestige differentials predict 70-90% of cross-admit outcomes. Aid elasticities run 2-4 percentage points per thousand dollars. Demonstrated interest barely registers at the schools families care most about, and dominates at the schools that need it most.
None of this should be reassuring or upsetting on its own. It does suggest that the variables that feel like personal taste — climate, vibe, weekend impressions — are mostly small effects layered on top of a much louder signal. The signal is almost always the relative selectivity of the offers and the relative price of the bills.
Read the trade-offs honestly before May 1, and the rest of the answer usually appears.