Five Factors That Doom Colleges: What the Data Actually Shows

New Federal Reserve research shows which institutional warning signs are most important—and which boards should be closely monitoring.

The reports of higher education closures have become all too common. Birmingham-Southern. Iowa Wesleyan. The University of the Arts. Fontbonne University. Each one raises the same anxious question among trustees and administrators at other schools: Could that be us?

A December 2024 working paper from the Federal Reserve Bank of Philadelphia provides the most thorough answer yet. Researchers Robert Kelchen, Dubravka Ritter, and Douglas Webber compiled data on over 8,600 institutions from 2002 to 2023—including 1,671 closures—and built predictive statistical models that significantly outperform existing federal monitoring tools. Their machine-learning approach accurately identified 84% of institutions with the highest 3-year closure risk.

What this research reveals is not just a list of warning signs. It presents a weighted framework that highlights which factors are most significant, and, just as importantly, which commonly cited concerns may be less predictive than boards believe.

The Five Critical Risk Factors

Based on the research findings, here are the five factors most strongly associated with institutional closure, ranked by their relative contribution to predictive accuracy:

1. Enrollment Trajectory (Estimated Weight: 30-35%)

The single most powerful predictor is not how many students you have—it is which direction you are heading.

The research found that year-over-year enrollment changes and multi-year enrollment trends were among the strongest predictors across all models. Institutions experiencing a 10% decline from their five-year enrollment high, or three consecutive years of enrollment drops exceeding 5% each year, showed dramatically increased closure risk.

This finding has serious implications for the coming "demographic cliff." The researchers' simulations indicate that a worst-case 15% enrollment decline could lead to up to 80 additional closures each year—a 142% increase over typical closure rates.

Board implication: Enrollment dashboards should clearly display trend lines, not just headcounts. A stable enrollment of 800 students is categorically different from an enrollment of 800 following years at 1,200.

2. Financial Momentum Indicators (Estimated Weight: 25-30%)

Changes in financial position predict closure better than absolute financial metrics.

The study found that trajectory-based measures—year-over-year changes in operating margin, revenue, and days cash on hand—consistently outperformed static measures of financial health. Persistently negative operating margins (three or more of the past five years) appeared in over 35% of institutions that eventually closed, compared with about 16% of surviving institutions.

Notably, the Federal Reserve's Financial Responsibility Composite (FRC) score, while contributing to predictive accuracy, was far less effective as a standalone measure than the machine-learning models that incorporate multiple financial-trajectory indicators.

Board implication: Finance committee reports should highlight directional indicators and multi-year trends. A college with modest reserves but stable finances might be safer than one with larger reserves experiencing rapid P&L deterioration.

3. Revenue Concentration and Tuition Dependency (Estimated Weight: 15-20%)

Institutions that closed derived 77% of revenue from tuition, versus 48% for survivors.

Tuition dependency emerged as a significant risk factor, particularly when combined with enrollment pressure. Colleges that closed had limited revenue diversification, with minimal contributions from auxiliary enterprises, investments, or philanthropic sources.

This finding explains why the for-profit sector has experienced such high closure rates (nearly one-third of two-year for-profits shut down during the study period). These institutions get about 90% of their revenue from tuition and lack other income sources that could help them weather enrollment declines.

Board implication: Revenue diversification is not just nice to have—it is an institutional survival strategy. Boards should track tuition dependency ratios and set explicit targets for reducing concentration risk.

4. Operating Scale and Institutional Capacity (Estimated Weight: 10-15%)

Smaller institutions close at higher rates, but size alone is not destiny.

The median closed institution had around 200 full-time equivalent students and between 20 and 50 employees, compared to much larger medians for surviving institutions. However, size mainly acts as a risk factor through its interaction with other variables. Small institutions have less ability to handle enrollment shocks, less revenue diversification, and fewer economies of scale.

The research showed that the predictive models performed with nearly perfect accuracy for institutions enrolling over 5,000 students—not because large institutions never close, but because their financial distress is more visible and predictable.

Board implication: Small institutions should be especially vigilant about other risk factors on this list. Scale creates vulnerability, but it does not determine outcomes.

5. Federal Accountability Triggers (Estimated Weight: 5-10%)

Heightened Cash Monitoring and failed financial responsibility scores signal elevated risk, but they are lagging indicators.

Institutions on Heightened Cash Monitoring Level 2 (HCM2) or with a failing federal Financial Responsibility Composite score showed increased closure risk. However, these indicators were less predictive than the researchers' machine-learning models—mainly because they are snapshots taken at a single point in time rather than measures of trends, and because they overlook institutions that are declining but have not yet met federal thresholds.

Only about 1% of institutions were on HCM2, while the actual at-risk population is much larger. The research indicates these metrics are helpful but not enough for early detection of troubled institutions.

Board implication: Federal accountability status serves as a serious warning, but the lack of it should not offer false reassurance. Keep in mind that many institutions that ultimately shut down never activated federal monitoring.

What the Research Does Not Capture

The researchers are commendably honest about the limits of even their most advanced models. Birmingham-Southern College, which closed in 2024 after years of public financial difficulties, was not identified as high-risk by the models—mainly because its historic endowment and relatively solid balance sheet hid the governance issues and operational losses that ultimately led to its closure. The study notes that "unobservable (to the researcher) governance and specialized reasons leading to closure are concealed by financial performance that does not necessarily give rise to concerns." For private nonprofit four-year institutions in particular, a significant portion of closure risk remains unexplained by observable financial and enrollment metrics.

This is where board judgment becomes crucial. Quantitative risk factors can spot vulnerable institutions, but they cannot detect dysfunctional board dynamics, leadership failures, or the gradual decline of institutional culture that often leads to financial collapse.

The Geographic Wild Card

The research also uncovers notable geographic differences in predictive accuracy. State-level analysis shows models performing very well in some states (99% accuracy in Connecticut) while nearing coin-flip levels in others (51% in Montana). States with idiosyncratic higher education environments or fewer private institutions proved harder to predict.

For boards, this suggests that regional and local factors—demographic trends, competitive dynamics, state policy environments—deserve attention alongside institutional metrics.

Implications for Board Oversight

This research offers several actionable insights for governance:

Demand trajectory data, not just snapshots. Every major board report should include multi-year trend lines for enrollment, revenue, operating margin, and liquidity. Static dashboards hide the directional signals that matter most.

Stress-test enrollment scenarios. Given the primacy of enrollment trajectory in predicting closure, boards should regularly model the institutional impact of various enrollment decline scenarios—and identify the trigger points at which current strategies become unsustainable.

Evaluate revenue concentration. High tuition dependency is not inherently problematic, but it creates fragility. Boards should understand their institution's revenue concentration relative to peers and consider whether diversification strategies merit investment.

Look beyond federal metrics. While compliance with federal financial responsibility standards is important, it is not enough. Institutions that ultimately shut down often stay in technical compliance until quite near to their closing date.

Recognize the limitations of data. The reasons that set Birmingham-Southern apart from Judson College (which the models successfully identified) were not mainly financial. Governance quality, strategic clarity, and leadership capacity are still crucial—and difficult to quantify.

The Federal Reserve research confirms what many in higher education have suspected: the conditions for institutional distress are measurable, and closures are largely, though not entirely, predictable. For boards willing to confront the data honestly, these findings present both warning and opportunity.

The institutions that succeed in the upcoming decade will be those whose boards and senior leaders understand these risk factors, monitor them closely, and act before the trajectory becomes inevitable.