Methodology

How TradeSchoolOutlook scores 12,699+ trade and vocational programs — data sources, formulas, refresh schedule, and known limitations.

Data sources

Earnings and debt

All program-level earnings come from the U.S. Department of Education College Scorecard. Specifically:

• Median 1-year and 5-year post-completion earnings, reported at the program level (CIP code × institution)
• Median debt at graduation
• In-state and out-of-state tuition
• Program enrollment counts (used to suppress thin data)

Scorecard publishes annually. We refresh within two weeks of each release.

Job market and wages

Occupation-level data comes from the U.S. Bureau of Labor Statistics Employment Projections 2024–2034:

• Total employment and 10-year projected growth rate per occupation
• Annual job openings (growth + replacement)
• Median wage and wage percentiles

We map each CIP program code to its corresponding SOC occupation codes using the federal CIP–SOC crosswalk. One program may map to multiple occupations; we weight each by employment share.

AI exposure

AI resilience scores blend two independent research sources:

• OpenAI GPTs-are-GPTs (Eloundou et al.) — task-level exposure to large language model augmentation
• Felten et al. AIOE (AI Occupational Exposure index) — academic exposure index from 2023

Exposure scores are blended using occupation-level weights proportional to employment share in each mapped career path.

The TradeSchoolOutlook Score

Every program receives a composite score from 0–100 combining four factors:

1. Earnings strength — how 1-year and 5-year graduate earnings compare to the national median for the trade
2. AI resilience — how shielded the mapped careers are from automation (blended GPT exposure + AIOE)
3. Job market size — total annual openings and 10-year growth
4. Earnings multiple (ROI) — total projected 10-year earnings divided by tuition

Earnings projections

The 10-year earnings projection on each program page starts from actual Scorecard 1-year earnings (not an estimate) and applies a growth rate derived from Scorecard's own 1-year-to-5-year delta where available, with BLS projected occupational growth as a fallback. We floor growth at 1% and cap at 12% annually to avoid unrealistic compounding.

Refresh schedule

• College Scorecard earnings and debt — annual, refreshed within 2 weeks of each release (typically fall)
• BLS occupational projections — biennial, refreshed with each BLS projections cycle
• AI exposure scores — research-driven, refreshed only when Eloundou or Felten publish updates
• Data version displayed on every page — each program page carries the data vintage so you can check freshness

Known limitations

1. Earnings data has a multi-year lag. Scorecard's 1-year earnings for the current release reflect students who graduated two to three years prior. In fast-moving trades (solar, EV-adjacent), current graduate earnings may differ from the reported figures.
2. Small programs are suppressed. Programs with fewer than ten graduates have their earnings suppressed in Scorecard, so we can't score them. If a program doesn't appear on the site, this is usually why.
3. CIP–SOC mappings aren't perfect. The federal crosswalk is well-maintained but some CIP codes map to broad occupational categories that span a wider earnings range than the program's graduates actually enter.
4. AI exposure research is evolving. We use the best available academic and industry sources, but AI capability frontiers shift quickly. We update when the underlying research updates.
5. State-level wages aren't yet factored into scoring. The composite score uses national medians; state pages show state wages as a separate display. In states with unusually strong or weak trade wages, the national-median-based score may understate or overstate local value.

Sister site

Considering a four-year degree instead? DegreeOutlook applies the same scoring framework to 24,000+ bachelor's and master's programs, with the added dimension of three AI-disruption scenarios. Trade programs don't model scenarios here because the work is inherently hands-on and the scenario range is too narrow to matter.

Corrections

If a number on our site doesn't match the source we cite, or we've mapped a program to the wrong occupation, let us know at our contact page. Corrections process within a week.