Demonstrated Interest (DI) in College Admissions

demonstrated_interest.md

Demonstrated Interest (DI) in College Admissions

Overview

Demonstrated interest (DI) refers to the signals a prospective student sends to a college indicating genuine intent to enroll if admitted. Colleges track DI primarily as a yield management tool: admitting students who are likely to matriculate protects yield rates, stabilizes enrollment targets, and (for rankings-conscious institutions) improves U.S. News metrics.

According to NACAC's State of College Admission report, 16% of colleges rate DI as having "considerable importance" and 28% rate it as "moderate importance" in admissions decisions. Combined, 44% of institutions assign meaningful weight to DI. However, this percentage is heavily skewed toward mid-tier privates and enrollment-dependent schools. At the most selective end, the picture is sharply divided.

CDS Section C7: Our 55 Colleges

The Common Data Set Section C7 asks each institution to rate "Level of applicant's interest" on a four-point scale: Very Important, Important, Considered, or Not Considered. Below is the classification for all 55 colleges in our simulation, synthesized from the most recent available CDS filings (2023-2024 and 2024-2025 cycles), official admissions blogs, and multiple cross-referenced secondary sources.

Not Considered (35 colleges)

These schools explicitly state DI plays no role in admission decisions:

Considered (17 colleges)

DI is a factor but typically a minor one, relevant primarily as a tiebreaker for borderline applicants:

Important (3 colleges)

DI is a meaningful factor that can shift outcomes for competitive applicants:

Very Important (0 colleges in our set)

No college in our 55-school simulation rates DI as "very important." Schools that do (American University, Dickinson, Olin College) tend to be in the 30-60% acceptance rate range, well below our simulation's selectivity floor.

How DI Is Tracked

CRM Systems (Slate)

48 of the top 50 U.S. universities use Technolutions Slate CRM. Slate tracks:

• Email engagement: open rates, click-through rates, time spent reading, device/location metadata
• Event registration: virtual info sessions, webinars, campus tours (with check-in verification)
• Portal activity: application status page logins, checklist views, financial aid page visits
• Website behavior: page views on admissions site (limited; not full web tracking)
• Social media follows: some schools track follows/interactions on institutional accounts
• Direct contacts: emails to admissions counselors, phone calls (logged manually)
• Campus visit sign-ins: in-person tour registration, information session attendance

Signal Hierarchy (Cost-Based)

The Lehigh University study (Howell & Patel, Contemporary Economic Policy, 2019) established that costlier signals carry more weight:

1. Campus visit (highest signal strength) -- ~30 percentage point boost in admission probability at DI-tracking schools
2. On-campus interview -- high signal but availability varies
3. Regional info session / college fair contact -- moderate signal (~10-13 pp boost)
4. Virtual event attendance -- lower signal, especially post-COVID
5. Email opens / portal logins -- weakest trackable signal

The Lehigh Study: Quantitative Evidence

The most rigorous quantitative study on DI comes from Lehigh University researchers (Howell & Patel, 2019), published in Contemporary Economic Policy:

• Sample: 12,501 applicants across 2 admission cycles at a "medium-sized, highly selective university"
• Methodology: Logistic regression controlling for SAT, GPA, and other applicant characteristics
• Key findings:
• On-site contacts (campus visits) increase admission probability by ~30 percentage points
• Off-site contacts (info sessions at high schools) increase probability by 10-13 percentage points
• Combined on-site + off-site contacts: 21-24 percentage point increase
• Effect is stronger for higher-SAT applicants: top SAT quartile + campus visit = 34 pp increase; second quartile + campus visit = 22 pp increase
• Costlier signals (requiring travel, time investment) have greater impact than low-cost signals

Important caveat: This study was conducted at a single institution that actively values DI (Lehigh lists DI as "Important" in CDS). The effect sizes would be much smaller or zero at schools that don't track DI.

Estimated Logit Boosts for Simulation

To implement DI in computeAdmissionScore(), we need to translate the research into logit-space additions. The simulation currently uses additive logit terms (pre-sigmoid), where typical academic scores produce logit values in the range of -3 to +3.

Deriving Logit Values from Research

The Lehigh study found campus visits increase admission probability by ~30 pp at a school with ~30-40% baseline admit rate. Converting to logit space:

• Baseline P = 0.35 -> logit = ln(0.35/0.65) = -0.62
• With DI boost P = 0.65 -> logit = ln(0.65/0.35) = +0.62
• Delta logit = ~1.24 for maximum DI at a school that rates it "Important"

For "Considered" schools, the effect is weaker (roughly 1/3 to 1/2 of "Important"):

Recommended di_logit_boost Values

For simulation implementation, we use a single per-college value representing the maximum logit boost a strong-DI applicant receives. The actual boost per student would be modulated by their DI level (0 = none, 0.5 = moderate, 1.0 = strong).

• Not Considered: di_logit_boost = 0.0
• Considered (weak): di_logit_boost = 0.1 (Northwestern, UChicago, JHU -- schools that list "considered" but publicly downplay it)
• Considered (standard): di_logit_boost = 0.15 (Rice, Tufts, Middlebury, Wellesley, etc.)
• Important: di_logit_boost = 0.3 (Tulane, Northeastern)
• Important (aggressive): di_logit_boost = 0.4 (Tulane -- yield protection is especially aggressive)

Tufts Syndrome / Yield Protection

"Tufts Syndrome" refers to the practice of rejecting or waitlisting highly qualified applicants who appear unlikely to enroll. It's a form of yield protection -- schools protect their yield rate (and U.S. News ranking) by avoiding admits who will decline.

Schools Most Associated with Yield Protection

From our 55 colleges, those most frequently cited for yield-protection behavior:

1. Tulane -- The most extreme example in our set. ED acceptance rate ~67%, overall ~9%, RD acceptance <3%. Tulane fills the vast majority of its class through ED/EA/ED2, leaving extremely few RD spots.
2. Northeastern -- Heavy ED reliance, aggressive DI tracking, significant yield optimization
3. Tufts -- The namesake, though Tufts has somewhat moderated; DI listed as "considered" not "important"
4. WashU -- Historically associated with yield protection before dropping DI in 2022
5. UChicago -- CDS says "considered" but public messaging says "no"; historically accused of yield protection during its rapid selectivity increase in the 2010s
6. Case Western Reserve -- Not in our 55 but frequently cited alongside these schools

How Yield Protection Manifests

• Waitlisting applicants whose stats far exceed the school's median (suggests they'll attend a more selective peer)
• Favoring ED/ED2 applicants (binding commitment guarantees yield)
• Weighting "Why [School]" essays heavily as a DI signal
• Tracking whether an applicant visited campus or attended events

DI and Early Decision/Action Interaction

Applying Early Decision is widely considered the ultimate demonstrated interest signal. A binding ED commitment:

1. Guarantees 100% yield for that admit
2. Signals the school is the applicant's first choice
3. Reduces uncertainty in enrollment modeling

How DI Interacts with Early Rounds

• ED applicants at DI-tracking schools already demonstrate maximum interest through binding commitment. DI tracking is largely redundant for ED applicants.
• EA/REA applicants signal interest through early application but without binding commitment. DI tracking may matter more here as a tiebreaker.
• RD applicants at DI-tracking schools face the highest bar. Without early commitment, they need other DI signals (campus visit, engagement history, strong "Why Us?" essay).

Implication for Simulation

The existing earlyRoundMultiplier in the simulation already captures the ED/EA advantage. DI should be modeled as an additional factor that primarily affects RD applicants at DI-tracking schools. One approach:

effectiveDIBoost = di_logit_boost * studentDILevel * (round === 'RD' ? 1.0 : 0.3)

This makes DI matter most in RD, with minimal additional effect in early rounds (where applying early is itself the DI signal).

Post-COVID Changes

The pandemic significantly altered DI tracking:

What Changed

1. Virtual events became universal (2020-2022) -- Colleges rapidly built virtual tour, webinar, and info session infrastructure. Virtual attendance became a trackable DI signal.
2. Campus visit accessibility improved -- Virtual options reduced geographic/economic barriers to showing interest.
3. Email and portal tracking intensified -- With in-person signals disrupted, CRM-based digital tracking (email opens, portal logins) gained importance.
4. Some schools dropped DI -- WashU dropped DI in 2022, citing equity concerns that rural and first-gen students were disadvantaged by visit-based DI tracking.

Current State (2024-2026)

1. Virtual event participation declining -- Down from 13% to 8% between 2023 and 2024. Students increasingly prefer in-person engagement.
2. Campus visits resurging -- Post-pandemic, in-person visits have returned as the strongest DI signal.
3. Test-optional amplified DI's role -- With SAT/ACT optional at many schools, DI fills part of the evaluative gap.
4. Video submissions -- Some schools (e.g., Tufts, Villanova) accept optional video introductions, which serve as DI signals.
5. CRM sophistication growing -- Slate's capabilities continue to expand, making passive DI tracking (email opens, web behavior) more granular.

Schools That Explicitly Do NOT Track DI

The following schools from our simulation have made explicit public statements that DI plays no role:

Tier 1 (All five)

• Harvard: "We do not track demonstrated interest"
• Yale: Does not track DI
• Princeton: Does not track DI
• Stanford: Does not track DI
• MIT: "We don't consider demonstrated interest in our admissions decisions"

Tier 2 (Most)

• Columbia, UPenn, Brown, Cornell: None of these Ivies track DI
• Duke: "Demonstrated interest as most people use the term is not a plus factor"; values demonstrated knowledge instead
• Caltech: Does not track DI

Notable Exceptions

• Dartmouth: The only Ivy to list DI as "considered" in CDS
• UChicago: CDS says "considered" but Dean of Admissions says it's not meaningfully factored

Why These Schools Don't Track DI

1. Yield is already high: Harvard (84%), Stanford (80%), MIT (76%) -- these schools don't need DI to predict who will attend
2. Applicant pool is self-selected: Applicants to HYPSM are already demonstrating maximum interest by applying
3. Equity concerns: DI tracking disadvantages low-income, rural, and first-generation students who cannot afford campus visits
4. Volume: Schools receiving 50,000+ applications cannot meaningfully track individual engagement

Correlation Between DI Weight and Yield Rate

There is a strong inverse correlation between a school's yield rate and how much it weights DI:

Schools with low yield rates have the most to gain from DI tracking because: 1. They need to predict which admitted students will actually enroll 2. Admitting DI-strong students improves yield numbers 3. Yield improvement reduces the need for waitlist activity 4. Higher yield rates improve U.S. News ranking inputs

Interview Policies Across Our 55 Colleges

Interviews are a related but distinct form of engagement. They can serve as both an evaluative tool and a DI signal.

Required Interview

• Georgetown: Alumni interview required for all applicants (the only school in our set that requires it)

Evaluative Alumni Interviews (Optional/Offered)

• Harvard, Yale, Princeton, Columbia, Dartmouth, UPenn: Offer alumni interviews; reports are read alongside application materials. These are evaluative but availability is not guaranteed.
• Brown: Alumni interview recommended, not required
• Duke: Optional interviews
• MIT: Alumni interviews offered but stated as informational

Campus/Staff Interviews

• Rice: Offers on-campus evaluative interviews
• Swarthmore, Williams, Amherst, Middlebury, Bowdoin, Wellesley, Colby, Hamilton, Davidson: Many LACs offer campus interviews that carry evaluative weight

Informational Only

• Cornell: Interview is informational, not evaluative
• Stanford: No interviews offered
• Caltech: No interviews offered
• Most public universities: Do not offer interviews

Implementation Recommendations for Simulation

1. Add di_importance field to college data

Values: "not_considered", "considered", "important"

2. Add di_logit_boost field to college data

The maximum logit boost for a strong-DI applicant at that school.

3. Generate student DI level

Each student gets a diLevel (0.0 to 1.0) when building college lists: - Students applying ED/ED2 to a school: diLevel = 1.0 (binding commitment is maximum DI) - Students applying EA to a DI-tracking school: diLevel = 0.6-0.8 - RD applicants: diLevel drawn from a distribution correlated with advantage (high-advantage students more likely to visit campus, attend events) - Disadvantaged students: lower diLevel on average (reflects real-world access barriers)

4. Apply in computeAdmissionScore()

// After other logit terms
if (college.di_logit_boost > 0) {
    let roundMultiplier = (round === 'ED' || round === 'EDII') ? 0.2 :
                          (round === 'EA' || round === 'REA') ? 0.5 : 1.0;
    logit += college.di_logit_boost * student.diLevel * roundMultiplier;
}

This ensures: - DI has no effect at "not_considered" schools (boost = 0) - DI matters most for RD applicants at DI-tracking schools - ED applicants get minimal additional DI boost (their early application is already captured by round multiplier) - Disadvantaged students face a realistic DI penalty at tracking schools

Sources

• Howell, J.S. & Patel, R. (2019). "Signaling Behavior in College Admissions." Contemporary Economic Policy, 37(4). DOI: 10.1111/coep.12216
• NACAC State of College Admission Report (2023-2024)
• Common Data Set filings, 2023-2024 and 2024-2025 cycles
• Individual college admissions websites and blogs (Tufts Inside Admissions, CMU Admissions, WashU Admissions, Emory Admissions FAQ)
• CollegeVine, Scholarships360, College Transitions DI databases
• Technolutions Slate CRM documentation
• College Kickstart DI rankings