Financial Aid Yield Elasticity: Research & Simulation Calibration

financial_aid_yield.md


Financial Aid Yield Elasticity: Research & Simulation Calibration

2. Literature Review: Financial Aid Yield Elasticity

2.1 Foundational Research

Avery & Hoxby (2003/2004), NBER WP #9482 The most-cited study on aid and college choice among high-aptitude students.

Aid Type Effect per $1,000 Notes
Grants +11pp enrollment probability Strongest effect
Loans +7pp enrollment probability Weaker than grants
Work-study +7pp enrollment probability Similar to loans
Tuition increase -2pp enrollment probability Asymmetric: losing $1K hurts less than gaining $1K helps
Room & board increase -10pp enrollment probability Students treat R&B as more "real" than tuition

Additional behavioral findings: - "Named scholarships" generate stronger enrollment response than equivalent dollar grants (price illusion / prestige signaling) - Front-loaded grants (more money freshman year) have significantly stronger effects than flat grants - ~30% of high-aptitude students respond irrationally -- reducing lifetime NPV of their choice - Students respond to gross tuition, not just net price (tuition + equal aid increase still lowers demand)

Limitation: Data from 1999–2000 cohort of high-aptitude students. The 11pp figure is for students choosing among multiple selective options, not for the general population.

2.2 Price Elasticity Estimates Across Studies

Study Context Price Elasticity Income Elasticity
Campbell & Siegel (1967) All 4-year -0.44 1.20
Hight (1970) Public universities -1.058 0.977
Hight (1970) Private institutions -0.641 1.701
Hoenack (1967) UC campuses -0.85 0.70
Moore et al. (1991) Occidental College -0.72 --
Buss, Parker & Rivenburg (2004) Selective LACs, full-pay -0.76 1.21
Buss, Parker & Rivenburg (2004) Selective LACs, aid recipients -1.18 --

Key insight: Aid-receiving students are 55% more price-elastic than full-pay students at selective institutions (-1.18 vs -0.76).

Grant elasticity at selective LACs: +0.31 (grants boost enrollment). Loan elasticity: +0.12.

2.3 Dynarski's Work on State Aid Programs

Dynarski (2003), AER Georgia's HOPE Scholarship: each $1,000 in aid increased college attendance by 3.7–4.2 percentage points. This is the most-replicated finding in the literature and represents the "consensus estimate" for broad-based aid programs.

Dynarski, Page & Scott-Clayton (2022), NBER WP #30275 Review of 50 years of policy experimentation. Concludes: "Typical financial aid programs have impacts in a range of about 3 to 4 percentage points per $1,000 increase in aid eligibility." More targeted programs (like Pell) show smaller effects due to complexity and take-up issues.

2.4 LaSota et al. (2024/2025) Meta-Analysis

"Does Aid Matter?" -- Review of Educational Research Most comprehensive meta-analysis to date: 709 effect sizes from 86 studies representing 7.66 million students.

Key findings: - Grants produce small but meaningful positive effects on enrollment, credit accumulation, persistence, and completion - Translated estimate: grants increased enrollment by 2.8 percentage points on average - Grants increased persistence by 2.0pp and completion by 0.4pp - Effects did not vary significantly by: eligibility criteria, grant program type, early commitment requirements, award duration, average award amount, or cost coverage type - Grants had larger effects at 2-year institutions than 4-year institutions for credit accumulation - No significant effect on academic achievement or post-college labor market outcomes

2.5 Endowment and Institutional Wealth Effects

Bulman (2022), NBER WP #30404 Counter-intuitive finding: wealthier colleges do NOT increase the number of students served or the fraction receiving aid. Instead, they: - Only modestly increase aid generosity - Offset higher yield rates by becoming MORE selective - Enroll FEWER low-income students and students of color - Use resources to increase spending and institutional rankings

Implication for simulation: At wealthy schools (HYPSM), increased endowment leads to higher selectivity rather than broader access. The current COLLEGE_AID_QUALITY=5 for HYPSM is correct in that these schools meet full need, but the simulation should not assume this translates to higher low-income enrollment rates.

2.6 Merit Aid vs Need-Based Aid

Research on how merit and need-based aid affect yield differently:

Strategic equilibrium (Social Choice & Welfare, 2023): When one college ranks above another, the dominant strategy is for the higher-ranked college to offer need-based aid and the lower-ranked college to offer merit aid. This maps directly to observed behavior: - HYPSM/Ivy+: pure need-based aid, no merit scholarships - Tier 3–4 (Vanderbilt, WashU, Emory, etc.): heavy merit aid to compete for cross-admits - Tier 5–6 (state flagships): mix of merit and need-based

Price illusion / scholarship naming effect: Students are more likely to enroll when they receive a "named scholarship" vs. an equivalent dollar grant, even at identical net prices. A simplified letter affirming belonging while making cost calculations salient increased enrollment in the lowest-cost option by 10.4pp (Behavioural Public Policy, Cambridge).

NACUBO Tuition Discounting Study (2024–25): - Average discount rate at private colleges: 56.3% for first-time undergrads - 83.4% of all undergrads receive some institutional grant aid - Grants cover 63% of tuition and fees for first-time students - Net tuition revenue increased only 1.4% (inflation-adjusted) despite higher discount rates

2.7 Income-Differentiated Yield at Elite Colleges

Chetty et al. (2023), "Diversifying Society's Leaders?"

Children from the top 1% are 2x more likely to attend Ivy-Plus colleges than middle-class students with comparable test scores. Two-thirds of this gap comes from higher admission rates; the remaining third comes from differences in application and matriculation rates.

Observed yield patterns (Class of 2029): - Harvard yield: 84% overall - Princeton yield: 78% - Yale yield: 68% (70% in some sources) - Stanford yield: 82%

Low-income enrollment trends (2024–2025): - Yale, Duke, JHU, MIT all set records for Pell-eligible students - MIT: Pell-eligible students rose 43% over two years, now >25% of freshman class - MIT policy: free tuition for families earning <$200K/year - Princeton: ~25% of Class of 2029 is Pell-eligible

These trends suggest that at full-need schools, yield among low-income admitted students is extremely high (likely 85–95%), while wealthy students with multiple elite options yield at lower rates (65–80%).


3. Synthesis: Yield Elasticity by Student Type

Combining all research sources, here are calibrated elasticity estimates for the simulation:

3.1 Yield Elasticity Table

Student Type Price Elasticity Per-$1K Grant Effect Per-$1K Loan Effect Yield Range at HYPSM Yield Range at Tier 3–4
Full-pay ($200K+) -0.76 +1–2pp +0.5–1pp 65–75% 35–50%
Upper-middle ($110K–200K) -0.85 +2–3pp +1–2pp 70–80% 40–55%
Middle income ($48K–110K) -1.00 +3–4pp +2–3pp 75–85% 45–60%
Low-income, partial aid ($20K–48K) -1.18 +4–6pp +3–4pp 80–90% 40–55%
Low-income, full aid (<$20K) -1.30 +5–8pp +3–5pp 85–95% 35–50%
First-gen (any income) -1.25 +5–7pp +3–5pp 80–90% 35–50%

Notes: - "pp" = percentage points of yield change - Price elasticity is own-price elasticity of enrollment demand - First-gen students have higher price sensitivity due to greater cost uncertainty (SSRN #4816609) and lack of parental guidance on aid negotiation - Low-income yield at Tier 3–4 schools can be LOWER than at HYPSM because Tier 3–4 often don't meet full need, creating unmet cost gaps

3.2 Key Asymmetries

  1. Grants > Loans > Work-study: Students weight grant aid roughly 1.6x more than loans in enrollment decisions (Avery & Hoxby)
  2. Gains > Losses: An additional $1K in aid boosts yield more (+3–4pp) than a $1K tuition increase reduces it (-2pp). This asymmetry is robust across studies.
  3. Need-based > Merit at top schools: At HYPSM, need-based aid signals belonging and eliminates cost barriers. At Tier 3–5, merit aid signals "we want you specifically" which boosts yield through prestige signaling.
  4. Named > Unnamed: A $20K "Presidential Scholar" award yields better than a $20K "institutional grant" at the same net cost. Estimate: +3–5pp boost from naming alone.
  5. Front-loaded > Flat: Grants with higher freshman-year value yield better than equivalent total grants spread evenly. Estimate: +2–3pp boost from front-loading.

3.3 The Chetty Data Paradox

The simulation's CHETTY_YIELD_BY_INCOME shows bracket 5 ($80K+) with values >1.0 at nearly every school (e.g., Stanford 3.06, Harvard 1.95). This seems to suggest wealthy students are MORE likely to attend. But this is not a yield rate -- it is rel_att_cond_app, relative attendance conditional on being in the applicant pool compared to the national average.

The high values for bracket 5 at elite schools reflect that: - Wealthy students apply to elite schools at much higher rates - Once admitted, their yield is actually LOWER than low-income admits (who have fewer alternatives) - But the sheer volume of wealthy applicants and their higher admission rates (via legacy/donor hooks) means they are overrepresented in the enrolled class

Simulation implication: Using rel_att_cond_app directly as a yield modifier (current approach) is reasonable as a first approximation, but it confounds application behavior with enrollment decisions. A better model would separate the two effects.


5. Summary of Key Numbers for Implementation

Consensus Estimates from Literature

Parameter Value Source
Grant effect on enrollment +3–4pp per $1K Dynarski (2003, 2022); consensus across studies
Grant effect (high-aptitude cross-admits) +11pp per $1K Avery & Hoxby (2003); ceiling estimate
Grant effect (meta-analysis average) +2.8pp per $1K LaSota et al. (2024)
Loan effect +7pp per $1K Avery & Hoxby (2003)
Named scholarship bonus +3–5pp vs unnamed Avery & Hoxby; behavioral econ literature
Front-loading bonus +2–3pp Avery & Hoxby (2003)
Cost framing / salience effect +10.4pp Behavioural Public Policy (Cambridge)
Aid-receiving vs full-pay elasticity ratio 1.55x Buss, Parker & Rivenburg (2004)
NACUBO avg discount rate (2024-25) 56.3% NACUBO TDS 2025
% students receiving institutional aid 83.4% NACUBO TDS 2025

HYPSM Yield by Income (Estimated from Multiple Sources)

Income Bracket Estimated Yield Evidence
<$20K (Pell-eligible, full aid) 88–95% Harvard Pell yield ~90%+; MIT 43% increase in Pell enrollment
$20K–$48K (near-full aid) 82–90% Full-need schools eliminate cost barrier
$48K–$110K (partial aid) 75–85% Some sticker shock remains
$110K–$200K (minimal aid) 70–80% Multiple competitive options
$200K+ (full pay) 65–75% Lowest yield; most alternatives

Tier 3–4 Yield by Income (Estimated)

Income Bracket With Merit Aid Without Merit Aid
<$20K 55–65% 35–45%
$20K–$48K 50–60% 35–45%
$48K–$110K 45–55% 35–45%
$110K–$200K 40–50% 30–40%
$200K+ 35–45% 25–35%

Sources

Academic Papers

Data Sources

Industry / Journalism


Some sections containing simulation-specific implementation details have been omitted from this public version. The research data and analysis above is based on publicly available sources.