← Index | Student Decision Behavior › Student Yield & Enrollment Decisions

Student Yield Behavior: Enrollment Decision Research

Source: student_yield_behavior.md

Student Yield Behavior: Enrollment Decision Research

How admitted students make final enrollment decisions. Data compiled from NBER working papers, NACAC surveys, Common Data Sets, IvyWise/IvyCoach reporting, and institutional press releases (2023-2025 admissions cycles).

Yield Rates by College Tier (Actual Numbers)

HYPSM Tier (75-87% yield)

College Yield (Class of 2029) Yield (Class of 2028) Notes
MIT 86.6% 85.8% Highest yield among all colleges
Harvard 83.6% 83.0% Consistently 82-84%
Stanford ~82% 81.9% Surpassed Harvard briefly in 2020
Princeton 75.4% 72.0% Rising trend, historically 65-72%
Yale ~70% 69.8% Lowest HYPSM, but rising

Cross-admit data (when students get into both):

Ivy+ Tier (55-70% yield)

College Yield (Class of 2029) Yield (Class of 2028) Notes
UChicago ~68% 72% High yield boosted by heavy ED
Brown 73.1% 67.3% Strong rise
UPenn ~67% 67.9% ED fills ~55% of class
Dartmouth 70.9% 63.7% Significant rise
Cornell 63.6% 68.4% Variable year to year
Columbia 61.3% 67.1% Declining trend
Duke 57.3% ~53% ED fills ~50% of class
Northwestern 57.7% ~55% ED fills ~55% of class
Caltech 58.6% 61% Small class amplifies variation

Near-Ivy Tier (40-55% yield)

College Yield (Class of 2029) Yield (Class of 2028) Notes
Notre Dame ~55% 62% High alumni loyalty drives yield
Johns Hopkins 51.4% ~40% Variable, rising
Georgetown ~47% ~47% Stable
Vanderbilt ~47% ~47% Heavy ED reliance
Carnegie Mellon 46.8% 47% Record high in 2024
WashU ~45% 47% ED-dependent
Rice 42.8% 44% Smaller pool

Selective Tier (30-50% yield)

College Yield (est.) Notes
UCLA ~50% High for public; in-state preference
Tufts ~46-50% "Tufts syndrome" / yield protection debate
Boston College 45.1% Strong Catholic/alumni network
UC Berkeley ~44% Similar to UCLA
Emory 37.3% Declining from ~40%
UVA ~38% Public flagship, in-state boost
Michigan ~38% Public flagship
USC ~37% Rising

Top LAC Tier (35-55% yield)

College Yield (Class of 2029) Notes
Bowdoin 53.8% Highest LAC yield
Williams ~47% Strong brand
Middlebury 42.0% Typical for top LACs
Amherst ~39% Lower despite high prestige

Average Yield Rates by Tier (for simulation)

Tier Yield Range Midpoint for Model
HYPSM 70-87% 80%
Ivy+ 55-73% 63%
Near-Ivy 40-62% 48%
Selective 35-50% 42%
Top LACs 35-54% 44%
National average (all 4-year) ~30% 30%

Key Yield Decision Factors

Ranked by Influence (synthesized from NACAC, BestColleges 2023, EAB 2024-25, NCES)

  1. Institutional prestige / academic reputation -- The single strongest predictor at the selective tier. Cross-admit data shows students almost always choose the higher-prestige option.

  2. Financial aid / net cost -- The dominant factor outside the top 20. For families with income <$150K, net price is often the decisive factor. Research consensus: $1,000 additional aid increases enrollment probability by 2-4 percentage points.

  3. Program/major strength -- Students choosing MIT over Harvard often cite STEM program depth. Engineering-focused admits favor Caltech, MIT, CMU, Stanford.

  4. Location / geography -- 47% of students rank location as a top campus factor (BestColleges 2023). Students prefer staying closer to home on average, though prestige overrides distance for elite institutions.

  5. Campus culture / student life -- Student quality of life (38%), campus safety (33%), diversity, and social scene all factor in.

  6. Financial aid type (merit vs need) -- Merit scholarships carry a psychological "scholarship effect" beyond their dollar value. Students feel "chosen" by merit aid in ways need-based aid does not replicate.

  7. Weather / climate -- Surprisingly ranked in top 10 by EAB 2024-25 survey. May explain Stanford/Duke/Vanderbilt appeal vs Northeast competitors.

  8. Family influence -- Parent preferences, legacy connections, and sibling attendance patterns.

  9. Campus visit experience -- Admitted student weekends have measurable yield impact (see Demonstrated Interest section).

  10. Peer effects -- Where friends/classmates are going, guidance counselor recommendations.

Factor Weight by Student Income

Factor Low income (<$60K) Middle ($60-150K) High (>$150K)
Net cost Dominant Very high Moderate
Prestige High High Dominant
Location High (stay close) Moderate Low (will travel)
Program fit Moderate High High
Campus feel Low Moderate High

Financial Aid Price Elasticity

Core Research Findings

Dynarski & Scott-Clayton (2013) consensus estimate:

Cal Grant program natural experiment:

DC Tuition Assistance Grant:

Merit Aid Elasticity at Selective Institutions

Research on selective colleges shows important nuances:

Price Elasticity by Tier (estimated from literature)

Tier Yield change per $10K aid Notes
HYPSM +1-3% Prestige dominates; most families already receive aid
Ivy+ +3-6% Moderate sensitivity
Near-Ivy +5-10% Sweet spot for merit aid leverage
Selective +8-15% Merit aid is a primary enrollment tool
National avg +15-25% Aid is often the deciding factor

Simulation Implication

For the college-sim model, financial aid primarily matters at the student decision phase (yield), not the admissions phase. The model should apply a yield modifier based on the gap between a student's expected family contribution and the college's net price, with diminishing sensitivity at higher-prestige tiers.

Demonstrated Interest Effects

Who Tracks It

Do NOT track demonstrated interest (yield already high enough):

DO track demonstrated interest (need yield management):

Quantified Effects

Lehigh University study:

NACAC data:

Signal hierarchy (strongest to weakest):

  1. Early Decision application (binding commitment = ultimate demonstrated interest)
  2. In-person campus visit / admitted student weekend
  3. Interview (alumni or on-campus)
  4. Contact with admissions office (email, phone)
  5. Attending info sessions / college fairs
  6. Opening/clicking email communications

Early Decision as Demonstrated Interest

ED is effectively the strongest form of demonstrated interest because it guarantees 100% yield:

School % of class filled via ED ED acceptance rate RD acceptance rate
Northwestern ~55% ~25% ~5%
Duke ~50% ~18% ~4%
Vanderbilt ~45% ~20% ~6%
Cornell ~40% ~17% ~7%
Brown ~40% ~14% ~5%
Dartmouth ~40% ~18% ~5%
UPenn ~55% ~15% ~5%

Simulation Implication

Demonstrated interest should function as a yield predictor, not an admissions factor, for HYPSM/Ivy tier. For Near-Ivy and Selective tiers, it should boost both admission probability and yield probability.

Waitlist Statistics and Behavior

Acceptance Rates from Waitlist

Tier Avg % admitted from waitlist Range Notes
HYPSM 0-5% 0-16% Princeton: 0.15% (low) to 16.4% (high)
Ivy+ 2-8% 0-15% Dartmouth avg: 4.1% over 21 cycles
Near-Ivy 5-15% 0-25% More variable
Selective 10-25% 0-40% Higher acceptance rates
National avg ~20% varies

Waitlist Timeline

Period Activity
April 1-May 1 Students receive waitlist offers; must accept spot on waitlist
May 1 National Candidates Reply Date -- deposits due
May 1-10 Biggest burst of waitlist admissions (colleges learn actual yield)
May-June Rolling waitlist admissions continue
Late June Most colleges close waitlists
July-August Rare late waitlist movement (melt)

Student Behavior on Waitlists

Simulation Implication

Waitlist mechanics should model: (1) a percentage of under-yield spots filled from waitlist, (2) waitlist admits have lower yield than direct admits, (3) the waitlist pool is drawn from borderline-admit students.

Yield Probability Model for Simulation

Proposed Formula

yield_probability = base_yield[tier] * prestige_factor * aid_factor * round_factor * interest_factor * random_noise

Component Definitions

base_yield[tier] -- Starting yield probability by college tier:

Tier Base Yield
HYPSM 0.80
Ivy+ 0.63
Near-Ivy 0.48
Selective 0.42
Top LACs 0.44

prestige_factor -- Adjustment when student has multiple admits:

aid_factor -- Financial aid modifier:

aid_gap = expected_family_contribution - college_net_price
if aid_gap > 0:  # college is cheaper than expected
    aid_factor = 1.0 + (aid_gap / 50000) * tier_sensitivity
else:  # college costs more than expected
    aid_factor = 1.0 + (aid_gap / 50000) * tier_sensitivity

Where tier_sensitivity:

Tier Sensitivity
HYPSM 0.10
Ivy+ 0.25
Near-Ivy 0.40
Selective 0.60

round_factor -- Admission round impact on yield:

Round Factor Rationale
ED 1.00 (forced) Binding; yield = 100%
EA/REA 1.05 Slight boost from early engagement
EDII 1.00 (forced) Binding; yield = 100%
RD 1.00 Baseline
Waitlist 0.75 Lower yield from delayed admits

interest_factor -- Demonstrated interest (Near-Ivy and below only):

Interest Level Factor
High (visited + ED) 1.15
Medium (visited or emailed) 1.05
Low/None 0.95
N/A (HYPSM/Ivy+) 1.00

random_noise -- Uniform +-15% to capture unpredictable personal factors:

random_noise = 0.85 + Math.random() * 0.30  // range [0.85, 1.15]

Student Decision Algorithm

When a student has multiple admits (non-binding rounds), they should:

  1. Calculate yield_probability for each admitted college
  2. Normalize probabilities across all admits (they must choose exactly one)
  3. Apply a "best option bias": multiply the highest-prestige option by 1.3x before normalizing
  4. Select college via weighted random draw using final probabilities

Edge Cases

Validation Targets

The model should produce aggregate yield rates within 5 percentage points of actual data:

Sources