Conservative College Rankings

Methodology

Data Sources

Twitter / X Data

  • Official university Twitter accounts for 155 US universities
  • Period: 2012–2026 (tweets matching 30 political keywords)
  • Collection: Automated scraping via Nitter proxy instances
  • 15 liberal keywords (diversity, equity, inclusion, social justice, sustainability, etc.) and 15 conservative keywords (veterans, military, faith, constitution, free speech, etc.)

AI Bias Assessment

  • Model: Google Gemini 2.5 Flash
  • Framework: Custom scoring rubric rating each tweet from -100 (strong liberal advocacy) through 0 (neutral) to +100 (strong conservative advocacy)
  • Coverage: ~42,000 tweets assessed with confidence scores
  • Validation: High-confidence subset (>0.7) shows consistent patterns

FEC Donation Data

  • Source: Federal Election Commission individual contributions (2020–2024 cycles)
  • Method: Employer name fuzzy-matching against university name variants
  • Metric: Dem% = Democratic / (Democratic + Republican) contributions

FIRE Free Speech Rankings

  • Source: Foundation for Individual Rights and Expression (2026)
  • Coverage: 138 of 155 schools matched
  • Correlation: r=0.33 (p<0.0001) with our Twitter bias scores

Additional Sources

  • DEI staffing: Web scraping of university DEI office pages
  • Student newspapers: Keyword-based analysis of editorial content
  • Commencement speakers: Wikipedia historical speaker lists with heuristic political classification

Scoring Methodology

Twitter Bias Score (-100 to +100)

-100 to -40
Strong Liberal
-40 to -15
Liberal
-15 to +15
Moderate
+15 to +40
Conservative
+40 to +100
Strong Conservative

Normalization Scale (0–4)

0
Absent
Never appears
1
Exploring
Occasional, informational
2
Accepting
Regular, positive framing
3
Normative
Deeply integrated
4
Mandatory
Institutional requirement

Limitations

  1. Twitter ≠ Reality: University Twitter reflects messaging strategy, not the views of all faculty, students, or administrators.
  2. AI Scoring: While consistent, AI assessments may have systematic biases. Results are relative comparisons, not absolute truths.
  3. FEC Data: Only captures politically active employees who self-report their employer.
  4. Temporal Coverage: Tweet collection density varies by school and time period.
  5. Keyword Approach: Predefined keywords may miss relevant content or capture false positives.

Data Freshness

Twitter Data
Feb 2026
FEC Data
2020–2024
FIRE Rankings
2026 ed.
US News Rankings
2026 ed.