Provenance

Where our data comes from

Every number in College Monte Carlo traces to a public institutional source — the Common Data Set, IPEDS, the College Scorecard, NSF, O*NET, BLS, and WICHE. This page maps each metric in the product to its primary source, the data vintage we ship, and how often it refreshes.

Maintained by Petr Kirsanov and Mikhail Kirsanov, founders of College Monte Carlo. How the sources become predictions: Methodology.

Metric-by-metric provenance

One row per family of metrics. Each source link goes to the primary dataset, not a summary of it. Vintages are the versions currently shipped in the product.

College Monte Carlo data provenance, as shipped July 2026.
Metric in the product Primary source Data vintage Refresh cadence
Acceptance rates (overall, early, regular), SAT/ACT middle-50% ranges, yield, class size, admission-factor importance (section C7) Common Data Set — each college's own standardized filing CDS 2024–25 (Class of 2028 entering cohort); Class of 2029 acceptance-rate and yield updates applied where published; headline fields re-verified July 2026 Annual, per CDS release
Early Decision / Early Action admit-rate multipliers Common Data Set early/regular splits + institutional reporting Class of 2029 (2024–2025 cycle) where available; compiled March 2026 Annual
Tuition, net price by household income bracket, financial aid IPEDS via College Scorecard + institutional aid offices 2025–2026 award year; compiled April 2026 Annual
Graduate earnings by major, per college Median earnings of federally aided graduates, 1 and 4 years after completion College Scorecard field-of-study data (U.S. Dept. of Education) Retrieved July 2026; 192/192 colleges with bachelor's-level rows Annual, per Scorecard refresh
Bachelor's degrees conferred by major (program scale) IPEDS Completions survey (NCES) AY 2019–20 (C2020_A) and AY 2023–24 provisional (C2024_A) Annual
University R&D expenditures by field (research investment) NSF Higher Education R&D (HERD) Survey FY 2024 microdata (federal + nonfederal expenditures); retrieved July 2026 Annual survey
High-school profiles (17,884 schools: enrollment, type, academic indicators) NCES Common Core of Data (public) + NCES Private School Universe Survey + 12 state DOE SAT/ACT files CCD 2022–23 (public backbone); PSS 2021–22 (private backbone); enrichment through 2026 sweeps Rebuilt with each pipeline sweep
Career profiles: occupations, wages, education paths, outlook O*NET OnLine (USDOL/ETA, CC BY 4.0) + BLS Occupational Outlook Handbook 112 occupations keyed by O*NET-SOC code; wage bands modeled on BLS wage-distribution percentiles; BLS 2024–34 employment projections Annual (BLS projections cycle)
High-school graduate projections (demographic cliff modeling) WICHE, Knocking at the College Door 11th Edition (December 2024); reported counts 2009–2023, projections 2024–2041 Per edition
Application deadlines (ED, EDII, EA, REA, RD, rolling) Official .edu admissions pages of each of the 192 colleges Fall 2026 entry cycle (latest published) Annual re-verification
Scholarship listings (merit, need-based, state, college-specific) Official provider and institutional pages Last audited May 2026 Periodic audit
AI-Proof Index (experimental): AI-displacement resilience by major family Research synthesis incl. BLS 2024–34 projections, NY Fed graduate outcomes, and academic AI-exposure literature Synthesized June 2026 (60 verified load-bearing claims); full source list on the AI-Proof Index page Periodic

How to read the vintages


How the sources become predictions

Sources are only half the story. The methodology page explains how these datasets feed an agent-based Monte Carlo simulation — logit-space scoring on the CDS C7 admission factors, per-college calibration against published acceptance rates, and separate early and regular rounds — and How It Works shows the calibration charts. For the concepts behind individual metrics, start with college acceptance rates explained, early decision vs regular decision, and what a college's yield rate means.

See the data in action

Two minutes, no signup required.

Run the simulation →