Kaggle Survey: Student Demographics & Diversity Data

Source: kaggle_survey_demographics.md

---

Kaggle Survey: Student Demographics & Diversity Data

Purpose

Identify Kaggle datasets for calibrating demographic parameters in the college admissions simulation: % URM students, % first-gen, gender ratios, and post-SFFA (2023) enrollment shifts.

---

1. College Scorecard (US Dept of Education) -- PRIMARY RECOMMENDATION

URL: https://www.kaggle.com/datasets/kaggle/college-scorecard Alt: https://www.kaggle.com/datasets/thedevastator/u-s-department-of-education-college-scorecard-da

Attribute	Detail
Source	US Department of Education
Years	1997--present (updated annually)
Institutions	~7,000+ Title IV institutions
Size	~589 MB
License	CC0 Public Domain
Format	CSV (Scorecard.csv) or SQLite

Key Demographic Variables

Variable	Description
UGDS	Total undergraduate enrollment
UGDS_WHITE	% White undergraduates
UGDS_BLACK	% Black undergraduates
UGDS_HISP	% Hispanic undergraduates
UGDS_ASIAN	% Asian undergraduates
UGDS_AIAN	% American Indian/Alaska Native
UGDS_NHPI	% Native Hawaiian/Pacific Islander
UGDS_2MOR	% Two or more races
UGDS_NRA	% Non-resident alien
UGDS_UNKN	% Race unknown
UGDS_MEN	% Male undergraduates
UGDS_WOMEN	% Female undergraduates
PCTPELL	% receiving Pell grants (proxy for low-income)
PCTFLOAN	% receiving federal student loans
FIRST_GEN	% first-generation students
PAR_ED_PCT_1STGEN	% parents with no college degree

Simulation Relevance: HIGH

• Directly provides race/ethnicity percentages for all 30 colleges in our simulation

• PCTPELL serves as proxy for socioeconomic diversity

• FIRST_GEN calibrates the first-gen hook (currently 1.4x multiplier)

• Gender ratio data (UGDS_MEN/UGDS_WOMEN) for gender-balance modeling

• Multi-year data allows trend analysis

• Data dictionary: https://collegescorecard.ed.gov/assets/CollegeScorecardDataDictionary.xlsx

---

2. College Tuition, Diversity, and Pay (TidyTuesday)

URL: https://www.kaggle.com/datasets/jessemostipak/college-tuition-diversity-and-pay GitHub: https://github.com/rfordatascience/tidytuesday/tree/master/data/2020/2020-03-10

Attribute	Detail
Source	US Dept of Education, TuitionTracker, PayScale, NCES
Diversity Year	2014
Tuition Year	2018-2019
Historical Range	1985-2016 (tuition trends)
Size	1.9 MB
License	MIT

Files Included

1. tuition_cost.csv -- tuition and fees by school
2. diversity_school.csv -- enrollment by race/ethnicity/gender category
3. tuition_income.csv -- net cost by income bracket
4. salary_potential.csv -- early/mid-career pay
5. historical_tuition.csv -- tuition trends 1985-2016
6. student_diversity.csv -- diversity percentages

diversity_school.csv Columns

Column	Type	Description
name	character	School name
total_enrollment	double	Total enrollment
state	character	State
category	character	Race/ethnicity/gender category
enrollment	double	Enrollment count for category

Data is in long format (one row per school-category pair). Categories include racial/ethnic groups and gender.

Simulation Relevance: MEDIUM

• Easy to use (clean, small, well-documented)

• Good for quick prototyping of diversity parameters

• Diversity data is from 2014 (stale for post-SFFA calibration)

• Joinable with tuition data for economic analysis

---

3. Economic Diversity and Student Outcomes (Opportunity Insights)

URL: https://www.kaggle.com/datasets/umerhaddii/economic-diversity-and-student-outcomes-data

Attribute	Detail
Source	Opportunity Insights (Chetty et al.) via NYT Upshot
Focus	Economic mobility, income quintiles
Institutions	~2,200 colleges
License	CC0

Key Variables

Variable	Description
super_opeid	Institution cluster ID
name	College name
par_income_bin	Parent household income group (percentile)
par_income_lab	Income group label
attend	Test-score-reweighted attendance rate
rel_apply	Relative application rate
rel_attend	Relative attendance rate

Simulation Relevance: MEDIUM-HIGH

• Income quintile data is excellent for calibrating socioeconomic hooks

• Mobility rate = access (% from bottom quintile) x success rate

• Complements Pell Grant data from College Scorecard

• Does NOT include race/ethnicity breakdowns directly

• Key finding: highest-mobility colleges are mid-tier publics

---

4. College Enrollment Demographics 2021 (IPEDS)

URL: https://www.kaggle.com/datasets/nivedithavudayagiri/college-enrollment-demographics-2021

Attribute	Detail
Source	IPEDS (NCES)
Period	July 2020 -- June 2021
Institutions	All IPEDS-reporting institutions
License	CC0 Public Domain

Demographics Tracked

• Gender: Male/female headcounts

• Race/Ethnicity: Nine IPEDS categories (White, Black, Hispanic, Asian, AIAN, NHPI, Two or more, Non-resident alien, Unknown)

• Level: Undergraduate vs. graduate

• Attendance: Full-time vs. part-time

• Student type: First-time, transfer, continuing

Simulation Relevance: MEDIUM

• Provides institutional-level race/ethnicity enrollment counts

• More granular than College Scorecard (full-time/part-time, grad/undergrad splits)

• Single year (2020-21), so no trend analysis

• IPEDS UNITID enables joining with other NCES datasets

---

5. "American University Data" IPEDS Dataset

URL: https://www.kaggle.com/datasets/sumithbhongale/american-university-data-ipeds-dataset

Attribute	Detail
Source	IPEDS via Tableau Public
Size	1.2 MB
Demographics	SAT/ACT, ethnicity, immigration, gender

Simulation Relevance: LOW-MEDIUM

• Combines test scores with demographics (useful for cross-referencing)

• Documentation is sparse; exact columns unclear without download

• May overlap with College Scorecard data

---

6. Historically Black Colleges and Universities (HBCU)

URL: https://www.kaggle.com/datasets/paultimothymooney/historically-black-colleges-and-universities

Attribute	Detail
Source	NCES / TidyTuesday
Years	1910-2016 (degree attainment), 1976-2015 (HBCU enrollment)

Files

• hbcu_all.csv -- HBCU enrollment data

• hs_students.csv / bach_students.csv -- completion rates by race

• Gender-disaggregated files (male/female)

Race/Ethnicity Categories

White, Black, Hispanic, Asian/Pacific Islander, American Indian/Alaska Native, Two or more races

Simulation Relevance: LOW

• Only covers HBCUs (none in our 30-college list)

• Useful for understanding long-term racial trends in higher education

• Degree attainment data (1910-2016) provides historical context

---

7. College Admissions (Samson Qian)

URL: https://www.kaggle.com/datasets/samsonqian/college-admissions

Attribute	Detail
Alt name	"Admission/Class Demographics by University"
Size	223 KB
License	CC0
Year	~2018

Simulation Relevance: LOW-MEDIUM

• Name suggests class demographics by university

• Small dataset, likely limited institution coverage

• Column details unavailable without download

---

8. Post Secondary Education Data (IPEDS)

URL: https://www.kaggle.com/datasets/hark99/post-secondary-education-data-ipeds

Attribute	Detail
Source	IPEDS/NCES
Size	19.3 MB
Focus	Student loan repayment, low-income students
Updated	January 2020

Simulation Relevance: LOW

• Focused on financial aid/loans rather than demographics

• May contain some enrollment demographics as secondary data

• Useful for financial aid yield modeling

---

Post-SFFA (Students for Fair Admissions) Enrollment Data

Background

The Supreme Court's SFFA v. Harvard/UNC decision (June 2023) ended race-conscious admissions. Two admissions cycles have occurred since: Class of 2028 (Fall 2024) and Class of 2029 (Fall 2025).

Key Data Sources (Non-Kaggle)

1. Post-SFFA Enrollment Dashboard (Class Action)

• URL: https://www.joinclassaction.us/post/the-post-sffa-enrollment-dashboard

• Interactive dashboard tracking racial demographic shifts

2. James Murphy's Post-SFFA Enrollment Tracker

• URL: https://jamessmurphy.com/2026/01/07/the-2025-post-sffa-enrollment-tracker/

• Covers 29+ institutions, 3 data points (2022-23 average, 2024, 2025)

• Categories: Hispanic, Black, Asian, White, Multiracial, Unreported

• No downloadable dataset; manually compiled from university press releases

3. College Transitions Enrollment Demographics

• URL: https://www.collegetransitions.com/dataverse/enrollment-demographics/

• 300+ institutions, Common Data Set source

• Enrollment by gender, race, international status (2024-25)

Class of 2028 (First Post-SFFA Cycle) Key Numbers

Institution	Black (pre)	Black (post)	Hispanic (pre)	Hispanic (post)	Asian (pre)	Asian (post)
Harvard	15.3%	14.0%	16.0%	11.3%	--	--
MIT	--	5%	--	11%	--	47%
Yale	14%	14%	18%	19%	30%	24%
Amherst	--	-8pp	--	-4pp	--	--
UNC	10.5%	7.8%	10.8%	10.1%	--	--
WashU	--	-5pp (all POC)	--	--	--	--
Tufts	--	-3pp	--	--	--	--
UVA	--	-1.4pp (Black+Asian)	--	stable	--	--
Duke	--	+1pp (Black+Hisp combined)	--	--	--	-6pp

Class of 2029 (Second Post-SFFA Cycle)

• Cornell: Black 4.3% -> 4.8%, Hispanic 9.5% -> 11.1% (vs. 25-27% BHI pre-SFFA)

• Most institutions show partial recovery but remain below pre-SFFA levels

• Increasing % of students declining to report race, complicating analysis

Aggregate Trends

• At 76 formerly race-conscious schools: average Black share dropped from 6.4% (Fall 2023) to 5.3% (Fall 2024) -- lowest since 2015

• Highly selective institutions overall: URM freshmen declined 7% from 2023 to 2024

• Black students: -16.3%

• Hispanic students: -1.8%

• Asian enrollment generally increased at most selective schools

• Public flagships showed minority enrollment increases (per Inside Higher Ed, Feb 2026)

---

Recommendations for Simulation Calibration

Tier 1: Must-Use

1. College Scorecard -- Race/ethnicity %, first-gen %, Pell %, gender for all 30 colleges
2. Economic Diversity (Opportunity Insights) -- Income quintile access rates

Tier 2: Supplementary

1. TidyTuesday Diversity -- Quick baseline (2014 data)
2. IPEDS Enrollment 2021 -- Cross-validate Scorecard numbers

Tier 3: Post-SFFA Calibration

1. James Murphy's Tracker -- Manual data for Class of 2028/2029 shifts
2. College Transitions -- CDS-sourced 2024-25 enrollment

Specific Calibration Parameters

Parameter	Data Source	Current Sim Value	Calibration Approach
% URM by college	College Scorecard (UGDS_BLACK + UGDS_HISP + UGDS_AIAN + UGDS_NHPI)	1.2x hook	Set per-college URM % from data; adjust post-SFFA
% First-gen	College Scorecard (FIRST_GEN, PAR_ED_PCT_1STGEN)	1.4x hook	Set per-college first-gen % from data
Gender ratio	College Scorecard (UGDS_MEN/UGDS_WOMEN)	Not modeled	Most selective schools ~50/50; LACs skew female
Low-income access	Opportunity Insights (par_income_bin)	Not modeled	Bottom-quintile access rate varies 3-20% by college
Post-SFFA shift	Murphy Tracker + institutional CDS	Not modeled	Black enrollment -16%, Hispanic -2%, Asian +5-7% at selective schools

Data Gaps

• No single Kaggle dataset has post-SFFA (2024+) demographic data

• First-gen data is inconsistently reported across institutions

• URM definition varies (some include Asian subgroups, some don't)

• "Decline to state" race category growing post-SFFA (complicates %)

• No Kaggle dataset directly maps to the 30 specific colleges in our simulation -- manual filtering required from College Scorecard