Contents

How Students Fall Into Application Archetypes — and Which One Are You?

archetype_decision_rules.md · 1,617 words · 6 min read

How Students Fall Into Application Archetypes — and Which One Are You?

Most applicants don't pick a college list from scratch. They follow a pattern shaped by who they are and where they come from — and admissions officers can usually spot the pattern within the first page of an application. This guide describes the six recognizable student "types" we see in real admissions data, how each one tends to build a college list, and how family resources and school setting reshape the picture.

Two things together explain most of the variation in how students apply: an identity dimension (the academic spike or interest pattern a student has built around — STEM, humanities, arts, athletics, or a balanced profile) and a circumstance dimension (legacy access, first-gen status, household income, school resources). Read together, they predict everything from how many schools a student applies to, to which colleges feel like reaches versus realistic targets, to how strongly hooks like recruited-athlete or legacy actually move the needle.

The numbers below are drawn from admissions research and verified against real cohort data. As you read, look for the row that sounds most like you or your student. Most people are a blend of two — and that's normal.

1. The Two-Axis System

The simulation decomposes students into 6 behavioral types x 4 structural positions = 24 distinct profiles. Behavioral type drives college preferences and application strategy; structural position determines hooks, resources, and application volume.

Axis 1: Behavioral (Identity & Spike)

  • STEM Spike (stem_spike): Quantitative focus, strong math SAT, targets technical elite
  • Humanities Spike (humanities_spike): Strong writing, essay-forward, prefers LACs
  • Arts Spike (arts_spike): Portfolio-driven, fewest applications, targets creative programs
  • Athletic Spike (athletic_spike): Recruited athlete hook dominates; lowest app count
  • Well-Rounded (well_rounded): Balanced achiever, broadest college compatibility
  • Average Academic (average_academic): Solid but not exceptional; safety-conscious

Axis 2: Structural (Socioeconomic Position)

  • High Advantage: 25% legacy, 15% donor, income bracket 4-5, app mult ×0.9
  • Moderate Advantage: Professional class, baseline resources, app mult ×1.0
  • Neutral: Middle class, standard access, app mult ×1.0
  • Disadvantaged: First-gen, Pell-eligible, URM 40%, app mult ×0.75

2. Generation Parameters

Academic & Application Profiles (sim.js:1366-1387, 1604-1608)

Archetype SAT Adj GPA Adj EC Quality Essay Base Apps Math SAT %
STEM Spike +30 to +60 — +1 pt +0.5 pt 8 52.8%
Humanities Spike — — +1.5 pt +1.0 to +1.5 7 46.3%
Arts Spike -10 to -40 — 8-10 (clamped) 7-9 (clamped) 5 48.5%
Athletic Spike -50 to -120 -0.1 to -0.25 5-9 — 4 49.3%
Well-Rounded — — baseline baseline 7 49.3%
Average Academic — — 3-7 (clamped) baseline 6 49.3%

Math SAT % = fraction of composite allocated to math section (national base: 49.3%). Source: MATH_SKEW_BEHAVIORAL (sim.js:2521-2526).

Behavioral Distribution by School Type (sim.js:908-920)

stem hum arts ath well_r avg_acad
Elite boarding 14% 12% 8% 18% 25% 23%
Elite day 12% 14% 12% 10% 27% 25%
Elite public magnet 28% 10% 5% 4% 22% 31%
Affluent suburban 10% 8% 6% 18% 20% 38%
Average suburban 6% 5% 4% 15% 15% 55%
Urban public 4% 4% 5% 12% 10% 65%

Structural Distribution by School Type (sim.js:922-934)

high_adv mod_adv neutral disadvant
Elite boarding 40% 35% 20% 5%
Elite day 30% 40% 25% 5%
Elite public magnet 5% 25% 45% 25%
Affluent suburban 5% 35% 45% 15%
Average suburban 2% 15% 50% 33%
Urban public 1% 5% 30% 64%

Intended Major Weights (sim.js:1435-1442)

Archetype CS Eng Hum Pre-Med Business Arts
STEM Spike 45% 35% 5% 5% 5% 5%
Humanities 5% 5% 50% 15% 15% 10%
Arts Spike 3% 2% 25% 5% 10% 55%
Athletic 10% 10% 25% 15% 30% 10%
Well-Rounded 20% 20% 20% 20% 15% 5%
Average Academic 20% 20% 20% 20% 15% 5%

3. College Fit Scores (sim.js:941-976)

The FIT_SCORES matrix maps each archetype to a 2-5 affinity score for each college. Score of 2 is the default; only explicit entries shown below.

STEM Spike (5 = perfect, 4 = strong, 3 = good)

  • 5: MIT, Caltech, Harvey Mudd
  • 4: Stanford, CMU, Georgia Tech, UC Berkeley
  • 3: Princeton, Cornell, UChicago, Rice, JHU, Purdue, UIUC, UW Seattle, UT Austin, Northeastern, Swarthmore, CMC

Humanities Spike

  • 5: Yale
  • 4: Harvard, Columbia, Princeton, UChicago, Wesleyan, Amherst, Swarthmore
  • 3: Brown, Stanford, Williams, Georgetown, Bowdoin, Hamilton, Colby, Middlebury, Pomona, Emory, NYU, Davidson

Arts Spike

  • 4: Yale, Northwestern, CMU, NYU, USC
  • 3: Columbia, Brown, Tufts, Amherst, Middlebury, Wesleyan, Wellesley, Pomona, Bowdoin

Athletic Spike

  • 4: Dartmouth, Cornell, Duke, Williams, Colgate
  • 3: Brown, Stanford, Princeton, UPenn, Middlebury, Notre Dame, Davidson, Colby, UNC, Virginia Tech, BC, Hamilton

Well-Rounded

  • 3: Dartmouth, Brown, UPenn, Duke, Northwestern, Rice, Vanderbilt, Notre Dame, Georgetown, Wake Forest, UNC, UVA, Michigan, Pomona, Bowdoin, Emory, BC, CMC, Colgate, Tufts, Colby, Middlebury, Hamilton, Davidson

No school has fit=5 for well-rounded — broad compatibility, no strong signal.

Average Academic

  • 3: Northeastern, NYU, BC, USC
  • 2: Everything else (default)

4. College List Building (sim.js:1602-1750)

Application Count (Lognormal)

baseMean = APP_MEANS_BEHAVIORAL[archetype] × APP_MULT_STRUCTURAL[position]
K = round(exp(log(baseMean) + z × 0.4))     z ~ N(0,1)
K = clamp(K, 3, 20)

HSLS:09 adjustments (sim.js:1631-1639): - Income gradient: INCOME_APP_SCALE = [0.85, 0.92, 1.00, 1.07, 1.15] - Asian-American: +1 app (HSLS Table 4: 4.92 vs 3.66 White) - URM: +1 app (HSLS: Black 4.46, Hispanic 4.35)

Utility Function (sim.js:1642-1676)

Each student evaluates every college:

U(s,c) = prestige + fitBonus + legacyBonus + lacBonus + inStateBonus + 4 × log(P_admit)
Component Formula Range
Prestige (6 - tier) × 5 + noise 0-25
Fit bonus FIT_SCORES[archetype][college] 2-5
Legacy bonus 15 if legacy at this college 0/15
LAC bonus 5 if top_lac + humanities/arts/well_rounded 0/5
In-state bonus 12 if public + same state; +5 if income ≤3 0-17
Admission odds 4 × log(P_admit) (academic-only estimate) ≤0

The 4 × log(P_admit) term means students moderately prefer schools they can get into, but prestige dominates. A 10× lower admission chance reduces utility by ~9 points, roughly equivalent to dropping 2 tiers.

Category Classification (sim.js:1613-1618)

Colleges sorted by utility are classified by log-odds of admission:

Category Threshold Target Allocation
Dream log(P) < -2.5 25%
Reach -2.5 ≤ log(P) < -1.0 35%
Target -1.0 ≤ log(P) < -0.3 25%
Safety log(P) ≥ -0.3 15%

This mirrors standard counselor advice (Hossler & Gallagher 1987).

5. GPA-SAT Correlation (sim.js:1276-1293)

Uses Cholesky decomposition with ρ = 0.65:

[z₀, z₁] = Box-Muller()
gpa = school.gpa.mean + school.gpa.sd × z₀
sat = school.sat.mean + school.sat.sd × (ρ × z₀ + √(1 - ρ²) × z₁)

Source: Empirical r ≈ 0.6 for GPA-SAT in selective populations (supply-side.md).

6. Admission Score Formula (sim.js:2561-2710)

All factors additive in logit space, combined via sigmoid:

P(admit) = σ(academic + feeder + hooks + yieldPenalty + inState + round + DI)

Academic Component (sim.js:2562-2574)

raw = 20 + aiDelta×0.75 + (ec/10)×20 + ecBonus + (essay/10)×10 + fitScore
logit = (raw - 46) / 20

Median student → raw ≈ 46 → logit = 0 → 50% at threshold school.

Hook Multipliers (multiplicative in odds = additive in logit)

Hook T1 T2 T3 T4 T5 Source
Donor 12.0× 10.0× 8.0× 6.0× 4.0× SFFA v. Harvard
Legacy 5.7× 4.0× 3.0× 2.0× 2.0× SFFA v. Harvard
Athlete 4.5× 4.0× 3.5× 2.5× 2.0× Arcidiacono et al.
First-gen 1.4× 1.4× 1.4× 1.4× 1.4× Institutional priority
Pell 1.25× 1.25× 1.25× 1.25× 1.25× Need-aware boost

Post-SFFA Demographic Multipliers (sim.js:2614-2618)

Group T1 T2 T3 T4 T5 T6
URM 0.85× 0.80× 0.80× 0.85× 1.08× 1.05×
Asian-Am 1.35× 1.25× 1.18× 1.12× 0.98× 0.99×

Gender Multiplier (sim.js:2621-2627)

  • STEM-heavy (Caltech, MIT): Female ×1.9
  • Balanced: Female ×1.05
  • LACs: Male ×1.25 (fewer male applicants)

Major Difficulty Multipliers (sim.js:2640-2658)

  • CS at MIT/CMU/Stanford: ×0.50 (2× harder)
  • CS at Cornell/Michigan/UCLA: ×0.65 (1.5× harder)
  • Engineering at MIT/CMU: ×0.75
  • Humanities at MIT/Caltech: ×1.25 (easier)
  • Pre-Med at JHU/Duke/Emory: ×0.80

Round Multipliers (sim.js:2683-2698)

  • ED: log(clamp(rateE/rate, 1.2, 8.0)) — per-college from ED_MULT_DATA
  • EDII: log(clamp(edMult × 0.65, 1.1, 5.0))
  • EA: log(clamp(rateE/rate × 0.8, 0.5, 4.0))
  • RD: 0 (baseline)

7. Enrollment Preference (ELO_ARCHETYPE_BONUS, sim.js:2067-2091)

When choosing which acceptance to enroll at, archetypes apply Elo bonuses:

College STEM Humanities Arts Athletic Well-Rounded
MIT +100 — — — —
Caltech +80 — — — —
Stanford +60 — — +40 +20
Yale — +40 +20 — —
Harvard — +30 — — +20
NYU — — +50 — —
Brown — +30 +40 — —
Duke — — — +30 +15
CMU +40 — — — —
USC — — +30 — —

8. Validation Gaps

  1. No external validation: Archetype distributions calibrated intuitively, not against HSLS:09 or ELS:2002 microdata which have student-level activity profiles.

  2. Linear utility: The model uses prestige + fit + 4×log(P) — no quadratic penalty for overreach. Reardon (2016) uses U(s,c) = Q_c - λ×(Q_c - C*_s)² which penalizes both undermatching and overmatching.

  3. No archetype switching: Students can't pivot mid-simulation (e.g., STEM rejected → humanities pivot).

  4. EC quality conflation: A painting portfolio and a varsity letter both map to the same EC quality score. The formula doesn't distinguish EC type.

  5. Missing archetypes: Pre-med spike, business spike, and social impact spike are common real-world patterns not represented. Current 6 types chosen for parsimony.

  6. Hook multiplier sources: Legacy (5.7×) and donor (12.0×) from SFFA v. Harvard trial only — no cross-college validation. First-gen (1.4×) lacks a specific published source.

  7. MODEL_SCALE = 0.013: The phantom applicant scaling factor has no published justification in any research document.