Higher-Ed A/G Research Extension

A public IPEDS study of institutional closure risk in a validated higher-ed subdomain.

Research-grade Hardened narrow validation Not a universal higher-ed model Not production

This site documents a research extension of the A/G framework to U.S. higher education data. The validated result is intentionally narrow: FASB-filing private nonprofit four-year institutions, evaluated for closure within five years. The site does not publish institution rankings or operational risk scores.

What the evidence says

Strong narrow closure signal

Joint-weak G+A geometry predicts closure within five years for FASB × Sector 2 private nonprofit four-year institutions.

Robustness checks passed

The result survives institution-cluster bootstrap uncertainty and four sliding temporal folds.

Not universal

Public, for-profit, two-year, and enrollment-collapse claims remain unvalidated or mixed.

Interpretive value

The A/G structure separates financial ground weakness from institutional appearance, but the validated closure lane is narrow.

At a glance

Final status: locked partial domain-specific extension
Validated lane: FASB × Sector 2 private nonprofit four-year
Target: closure within five years
Primary geometry: joint-weak G+A
AUROC: about 0.864
Cluster-bootstrap 95% CI: about [0.806, 0.914]
Temporal folds: about 0.77–0.86 AUROC
Baseline comparison: small positive edge over strongest tested baseline; not definitive universal superiority
Product status: research-grade / not production

Main finding

The higher-ed extension has a hardened, domain-specific validation result: joint-weak G+A geometry predicts closure within five years for FASB × Sector 2 private nonprofit four-year institutions. The result survives institution-cluster bootstrap uncertainty and temporal multi-fold validation. It should not be generalized to all higher education.

Key validation

Finding	Result	Interpretation
Closure AUROC	0.864	Strong held-out result in validated lane
Cluster-bootstrap CI	[0.806, 0.914]	Robust to institution-level resampling
Temporal folds	0.77–0.86	Signal persists across folds
Paired edge vs best baseline	+0.072	Small positive edge; not a universal claim
Enrollment-collapse target	baseline-dominated	Not a validated A/G superiority claim

Claim boundary

This is not a college-failure prediction product, not an accreditation tool, not an admissions tool, not investment advice, not a ranking system, and not a production model.

For the full canonical list of allowed and forbidden claims, see Claim Boundary Note.

Related Public Demos

Related public demos from the same project family. Each demo has its own scope, assumptions, validation status, and limitation notes.