Geographic Diversity in College Admissions

geographic_diversity.md

Geographic Diversity in College Admissions

Comprehensive research on how geography shapes admissions outcomes at selective colleges. Covers state-of-origin distributions, regional clustering, geographic hooks, public university mandates, private college diversity goals, urban/rural preferences, international competition, and feeder school geography.

1. State-of-Origin Distributions at Selective Colleges

1.1 HYPSM Regional Breakdown (Class of 2028-2029)

Data from official class profiles and institutional research offices:

Harvard (Class of 2029) | Region | % | |:---|---:| | Middle Atlantic | 21 | | New England | 18 | | South | 16 | | Pacific | 14 | | International | 16 | | Midwest | 10 | | Mountain | 3 | | Central | 2 | | Territories | <1 |

Harvard's geographic classification uses "permanent address at time of application" and includes US citizens abroad in the International category. The Middle Atlantic has consistently been the largest domestic region across multiple class years.

Yale (Class of 2028-2029) - Northeast: ~31-32% (largest region, stable across years) - International: ~11% - All other regions not published in granular form - Students from 47 states and 27+ countries

Princeton (Class of 2028) - Midwest: only ~10% despite being 20% of US population - Rural: only 9% despite 19% of US population - Largest clusters: NYC, Bay Area, and Princeton area - 48 of 50 states represented

Stanford (Class of 2028) - All 50 states, 70 countries - International: 14% (down slightly from prior years) - California is the most represented state (proximity effect) - Public high school: 56%, Private: 28%

MIT (Class of 2029) | Region | % | |:---|---:| | Mid-Atlantic | 20 | | West Coast, Alaska & Hawaii | 17 | | South & Puerto Rico | 15 | | Southwest & Mountain | 13 | | New England | 12 | | Midwest & Plains | 10 | | Abroad | 11 | | Territories | <1 |

California is the most represented US state at MIT. 49 states represented.

1.2 Ivy+ Regional Distribution (Class of 2028)

Columbia | Region | % | |:---|---:| | Mid-Atlantic | 35 | | West | 23 | | South | 18 | | International | 17 | | Midwest | 13 | | New England | 10 | | DC & Territories | 1 |

Columbia's Mid-Atlantic concentration (35%) is the highest among Ivies, reflecting its NYC location. This is roughly double that of similarly selective peers.

Cornell - 64.8% from out-of-state (NY is home state) - 11% international - All 50 states + DC, Guam, Northern Mariana Islands, Puerto Rico - 93 countries

Dartmouth - 55% from Southern/Western US or outside US - 14.5% non-US citizens - 15% from rural communities - 49 states, DC, Puerto Rico, 64 countries

Duke - 15% from North Carolina (home state, most of any state) - ~13% international - Top states: NC, NY, FL, TX, CA (tied for 5th with SC) - All 50 states, 87 nations - 60% suburban, 25% urban, 15% rural

UChicago (Class of 2028, College Factual data) | Top State | Students | % | |:---|---:|---:| | Illinois (home state) | 244 | 18.1 | | New York | 195 | 14.4 | | California | 178 | 13.2 | | Texas | 77 | 5.7 | | Florida | 71 | 5.3 | | Top 5 total | | 56.6 |

1.3 Cross-College Pattern: Dominant States

Across all selective private colleges, a consistent "Big 5" states emerge:

The Northeast corridor (NY, MA, NJ, CT, PA) is consistently overrepresented at 35-45% of enrollment vs. ~15% of US population. This reflects both feeder school concentration and proximity effects.

2. Regional Clustering Effects

2.1 The Northeast Gravity Well

The most pronounced clustering effect in US admissions: students from the Northeast disproportionately apply to and attend Northeast colleges.

Evidence: - Princeton: only 10% from Midwest despite 20% of population - Harvard: Mid-Atlantic (21%) + New England (18%) = 39% from Northeast - Columbia: Mid-Atlantic alone is 35% - Cornell: ~35% from NY (home state) due to state-funded colleges

Mechanism: Pipeline recruitment -- elite boarding schools (Andover, Exeter, Choate, Deerfield, Hotchkiss) are concentrated in New England. NYC day schools (Trinity, Collegiate, Horace Mann, Dalton) feed Ivies at 30-40% placement rates.

2.2 California / West Coast Clustering

• Stanford, Caltech: California students are the single largest cohort
• UC Berkeley, UCLA: 80% California residents (Board of Regents policy)
• Stanford proximity pull: Bay Area and LA drive ~25-30% of California applicants
• West Coast students are less likely to apply to small Northeast LACs

2.3 Southern Expansion

• Duke: 15% from NC, strong draw from FL, TX, SC, GA
• Vanderbilt: ~25-30% from Tennessee and neighboring states (estimated)
• Rice: ~25-35% Texas residents (estimated)
• Wake Forest, Tulane, Emory: Southern regional draws
• Georgetown: draws heavily from Mid-Atlantic (DC-area effect)

2.4 Midwest Under-Representation

The Midwest is consistently the most underrepresented region at elite private colleges: - Harvard: 10% from Midwest (vs. 20.6% of US pop) - MIT: 10% from Midwest & Plains - Princeton: 10% from Midwest - Contributing factors: fewer feeder schools, lower concentration of college-prep culture in rural areas, distance from coastal elites

2.5 Quantifying Regional Clustering

Estimated regional enrollment share at private selective colleges, by school location:

3. Geographic Hooks: The Underrepresented State Advantage

3.1 Evidence of a Geographic Boost

Selective private colleges seek to represent all 50 states in every incoming class. This creates measurable advantages for students from low-population or underrepresented states.

Brown University (2015 data, striking example): - 23 applicants from Montana, 7 accepted = 30% acceptance rate - Compare to overall rate of 8.5% - This 3.5x relative advantage is one of the most documented examples

General pattern: Applicants from states like Montana, Wyoming, North Dakota, or West Virginia face dramatically less competition from their home state. An Ivy League school might receive 5,000+ applications from New York but only 20-30 from Wyoming.

3.2 Important Caveats

• Geographic boost is not a guaranteed ticket -- applicants must still be academically competitive
• Admissions officers note students from underrepresented states "often don't have as strong grades or scores" but may be given more latitude
• The boost is most significant when combined with other diversity factors (rural background, first-gen, unique perspective)
• Being from an overrepresented state (NY, CA, NJ, MA, CT) is a mild penalty -- competition is fiercer within the same geographic pool

3.3 Estimated Logit Boost for Simulation

Based on the Brown/Montana data point (30% vs 8.5% = 3.5x) and admissions consultant estimates:

Note: The current simulation uses hooks.underrepresented_state = true (8% random assignment) with a Math.log(1.3) logit boost. This is a reasonable first approximation. A more sophisticated model would condition on actual student state.

3.4 The "50 States" Constraint

Every HYPSM and Ivy+ school aims to report "students from all 50 states" in their class profile. This creates an implicit floor: at least 1-3 students from every state. For states like Wyoming (population 577K), this means a disproportionate per-capita advantage.

4. Public Flagship Geographic Mandates

4.1 In-State Enrollment Requirements

4.2 Acceptance Rate Differentials (IS vs OOS)

The in-state vs out-of-state acceptance rate ratio reveals the strength of geographic preference:

Virginia Tech is an outlier where the OOS rate exceeds IS rate. This likely reflects that OOS applicants self-select more heavily (only the strongest apply), while IS applicants include a broader range.

4.3 Simulation Impact

The current simulation already implements IS/OOS mechanics via rateInState/rateOOS fields and a logit adjustment in computeAdmissionScore. The in-state logit boost is calculated as:

inStateLogit = Math.log(rateInState / rateOOS)

This ranges from +0.34 (UC Berkeley, mild) to +1.56 (UNC, strong). Students also get an in-state enrollment utility bonus (+8 to +12) reflecting cost savings.

5. Private College Geographic Diversity Goals

5.1 Implicit Targets (Not Publicly Stated)

No elite private college publicly states "no state should be more than X% of our class." However, the data reveals implicit constraints:

• Home state soft cap: ~15-20% for the host state
• UChicago: 18% from Illinois
• Duke: 15% from North Carolina
• Dartmouth: likely 10-15% from New Hampshire (small state)

• Columbia: 20-25% from New York (hard to suppress given NYC scale)

• Any single state: Informally kept below ~20% except for NY at Northeast schools

• "All 50 states" aspiration: Nearly universal at tier 1-3 schools. Princeton achieved 48/50 for Class of 2028.

5.2 How Schools Achieve Geographic Diversity

1. Regional admissions officers: Each elite school assigns admissions officers to geographic territories. These officers build relationships with high schools in their region.

2. Recruitment travel: Officers visit "pipeline" cities (NYC, LA, Chicago, SF, Houston, DC, Boston, Philadelphia, Atlanta, Dallas, Seattle, Miami).

3. STARS network: Coalition of 33 elite colleges specifically recruiting rural and small-town students. Extended 11,000+ admissions offers to Class of 2028 (up 12.9% YoY).

4. Financial aid as a tool: Full-need schools use generous aid packages to attract students from underrepresented regions who might otherwise attend in-state publics.

5.3 LAC Geographic Patterns

Small liberal arts colleges show distinct geographic patterns:

NESCAC schools have the strongest Northeast clustering among all institution types.

6. Urban vs. Rural Preference Patterns

6.1 Rural Underrepresentation

Rural students are significantly underrepresented at selective colleges despite comparable high school graduation rates:

Dartmouth and Duke show the strongest rural recruitment among peer institutions.

6.2 Barriers for Rural Students

• Counseling gap: Rural schools average 1 counselor per 500+ students vs. 1:30 at prep schools
• Course access: Fewer AP courses, no calculus at some rural schools
• Information gap: Less awareness of selective college options and financial aid
• Cultural pull: Strong community ties reduce desire to attend distant colleges

6.3 Growing Rural Recruitment

• STARS College Network: 33 schools (including all Ivies, Stanford, MIT, Duke, etc.) specifically targeting rural applicants
• QuestBridge: Matches high-achieving, low-income students (many rural) with elite colleges
• Yield challenge: Even when admitted, rural students matriculate at lower rates (cost concerns, distance anxiety)

6.4 Simulation Relevance

The simulation could model this via: - A ruralUrban student attribute (rural/suburban/urban) - Rural students: slight admissions boost (+0.15 logit) at private colleges seeking diversity - Rural students: lower base application count (information gap) - Rural yield penalty: lower enrollment probability at distant, expensive privates

7. International Student Geographic Competition

7.1 Top Sending Countries (2024/25, IIE Open Doors)

Total: 1,177,766 international students in US (5% increase YoY). 12 of top 25 countries reached all-time highs.

7.2 International Share at Selective Colleges

7.3 Geographic Competition Within International Pool

International admission is far more competitive than domestic. Key dynamics:

• China/India penalty: The largest applicant pools create the most intense within-country competition. A student from China competes against thousands of other Chinese applicants for a handful of slots.
• Small-country advantage: Analogous to the domestic "Montana effect" -- being the only applicant from Bhutan or Moldova provides distinctiveness
• Regional quotas (informal): Many schools aim for no single country >25-30% of the international cohort
• Visa policy sensitivity: The simulation's intlShareShock slider models this; 2025-2026 visa uncertainty has already reduced international enrollment at some schools

7.4 Simulation Implementation

The existing intlSharePct per college (ranging from 0.05 to 0.25) handles international seat reservation. The phantom applicant system fills these seats. No changes needed for basic modeling, but a future enhancement could model within-country competition for international students.

8. High School Feeder Geography

8.1 Geographic Concentration of Feeder Schools

The Glasener (2022, University of Michigan) dissertation mapped 3,200 secondary schools to 76 selective colleges:

• Less than 20% of ~19,000 high schools qualify as "feeder schools"
• The most connected feeder school had ties to 58 of 76 elite colleges
• Private, predominantly white, affluent schools have more and stronger feeder ties
• Schools with higher student-of-color enrollment had 17% of feeder ties despite being 30% of schools

8.2 Feeder School Geographic Clusters

New England Boarding Schools (Tier 1 Feeders) - Phillips Exeter (NH), Phillips Andover (MA), Deerfield (MA), Choate (CT), Hotchkiss (CT), Lawrenceville (NJ), Groton (MA), Milton (MA), St. Paul's (NH) - Concentrated in MA/CT/NH/NJ corridor - Feed primarily to Northeast Ivies, then Stanford/Duke - Top placement: Andover ~221 Ivy League students, Choate ~212, Lawrenceville ~179, Deerfield ~170

NYC Day Schools (Tier 1 Feeders) - Trinity, Collegiate, Horace Mann, Dalton, Brearley, Spence - Feed heavily to Columbia, NYU, and then other Ivies - Unique: Hunter College HS, Stuyvesant, Bronx Science (public magnets) -- similar feeding pattern to elite privates

West Coast Feeders - Harvard-Westlake (CA), Castilleja (CA), Lakeside (WA) - Feed primarily to Stanford, Caltech, then Ivies - Growing pipeline to UC Berkeley - Less concentrated than NE feeders

Southern Feeders - Relatively fewer elite feeders - TJ (TJHSST, VA) feeds to Georgetown, UVA, then Ivies - American Heritage (FL) feeds to UF, then selective privates

8.3 The Pipeline Effect on Geographic Distribution

Feeder schools explain much of the geographic skew at elite colleges: - ~50% of Harvard's class comes from just ~200 high schools (out of ~27,000+ in the US) - These 200 schools are concentrated in: NYC metro, Boston metro, Bay Area, DC metro, LA, and Connecticut - This creates a self-reinforcing cycle: alumni from these regions send their children to the same feeder schools

8.4 Simulation Alignment

The simulation's 20 core high schools are geographically distributed across feeder clusters: - NE Boarding: Exeter (NH), Andover (MA), Groton (MA), Hotchkiss (CT), Milton (MA), Deerfield (MA), Choate (CT), Webb (CA), Lawrenceville (NJ) - NYC: Trinity (NY), Collegiate (NY), Horace Mann (NY), Hunter (NY), Stuyvesant (NY), Bronx Science (NY) - Day Schools: Harvard-Westlake (CA), Lakeside (WA), American Heritage (FL) - Magnets: TJ (VA), BASIS (AZ)

The existing feederTier system (1-4) already captures the feeder advantage. Geography adds an additional layer: NE boarding schools feed NE colleges most strongly, West Coast feeders feed West Coast colleges most strongly.

9. Simulation Implementation Recommendations

9.1 Current State

The simulation already has: - student.state attribute (from high school state or national pool for boarders) - STATE_WEIGHTS array for population-weighted random state assignment - hooks.underrepresented_state (8% random, +log(1.3) boost) - college.isPublic, college.state, rateInState, rateOOS for public colleges - In-state logit boost in computeAdmissionScore() - In-state enrollment utility bonus in buildCollegeLists()

9.2 Potential Enhancements (Priority Order)

P1: State-aware underrepresented_state hook Instead of random 8% assignment, derive from actual student state:

const UNDERREP_STATES = new Set(['MT','WY','ND','SD','AK','WV','ID','NM','AR','MS']);
hooks.underrepresented_state = UNDERREP_STATES.has(studentState);

P2: Regional affinity for private colleges Add regional enrollment targets to private colleges and a mild boost/penalty:

// Student's region
const REGION_MAP = {
  'CT':'NE','ME':'NE','MA':'NE','NH':'NE','RI':'NE','VT':'NE',
  'NJ':'MA','NY':'MA','PA':'MA',
  'IL':'MW','IN':'MW','MI':'MW','OH':'MW','WI':'MW','MN':'MW','IA':'MW','MO':'MW','ND':'MW','SD':'MW','NE':'MW','KS':'MW',
  'DE':'S','FL':'S','GA':'S','MD':'S','NC':'S','SC':'S','VA':'S','DC':'S','WV':'S','AL':'S','KY':'S','MS':'S','TN':'S','AR':'S','LA':'S','OK':'S','TX':'S',
  'AZ':'W','CO':'W','ID':'W','MT':'W','NV':'W','NM':'W','UT':'W','WY':'W','AK':'W','CA':'W','HI':'W','OR':'W','WA':'W'
};

P3: Rural/suburban/urban attribute Assign based on school type: - elite_boarding, elite_day_school: suburban - elite_nyc_private, elite_public_magnet: urban - average_suburban_public, affluent_suburban_public: suburban - urban_public: urban - Some boarding schools serve rural students (20% of their pool)

P4: Regional clustering in college list building Students slightly prefer colleges in their region (lower distance = higher utility):

const REGION_PROXIMITY = { same: +3, adjacent: +1, distant: 0 };

9.3 Calibration Targets

If implementing regional distribution modeling, these are the targets:

Note: Tier 5-6 publics are dominated by their home state (not captured in regional averages). Tier 6 publics in the Midwest (UIUC, UW-Madison, Purdue) naturally have high MW%.

10. Sources

• Harvard Admissions Statistics (college.harvard.edu/admissions/admissions-statistics)
• MIT First-Year Class Profile (mitadmissions.org/apply/process/profile/)
• Yale Office of Institutional Research (oir.yale.edu)
• Princeton Daily Princetonian Frosh Survey 2028
• Stanford Facts Class of 2028 Profile (facts.stanford.edu)
• Columbia Class of 2028 Profile
• Duke Today: "Duke's Newest Undergraduates by the Numbers" (2024)
• College Factual Geographic Diversity data
• UNC Media Hub: "The Admissions Ratio: the UNC System's 82-18 split"
• Texas Tribune: "UT-Austin tightens automatic admission threshold to 5%"
• UC Board of Regents nonresident enrollment policy
• UVA Admissions: Quotas page (admission.virginia.edu)
• College Transitions: "Cast a Wide Net: Geographic Diversity and College Admissions"
• College Transitions: "Origins of College Students: Understanding Geographic Diversity"
• Glasener (2022): "Shaping Elite College Pathways: Mapping the Field of Feeder High Schools in the United States" (U of Michigan dissertation)
• IIE Open Doors Report 2024/25
• STARS College Network data
• Quartz: "If you want to get into an elite college, you might consider moving to one of these states"
• Great College Advice: "Does Where You Live Affect Your Chances of Admission?"
• Ivy Coach: "Geographic Diversity in College Admissions"
• Daily Cardinal: "Out-of-state students are an increasingly prominent part of the Wisconsin experience"
• UW News: "Washington residents make up nearly three-fourths of incoming class" (2025)