Where the system sits
Of the 4,336 credit unions in the panel as of Q1 2026, 945 — 21.8% of the system — score in the Elevated or High merger-susceptibility band. Susceptibility is not a prediction that a given institution will merge; it is the degree to which its balance sheet shows the combination of scale, membership trend, earnings, capital and charter age that has preceded merger in the historical record.
| Band | Institutions | Share of system | Median assets | Median net worth ratio |
|---|---|---|---|---|
| Low | 1,838 | 42.4% | $336.8M | 11.7% |
| Moderate | 1,523 | 35.1% | $34.9M | 13.4% |
| Elevated | 696 | 16.1% | $16.3M | 13.7% |
| High | 249 | 5.7% | $11.1M | 10.7% |
The gradient down that table is the finding: median assets fall at every step from Low to High, which makes susceptibility overwhelmingly a scale story. The Elevated and High bands are populated by small charters whose fixed compliance and technology cost is spread across too few members — not by institutions in distress.
Concentration by asset tier
| Asset tier | Institutions | Elevated or High | Share of tier |
|---|---|---|---|
| <$10M | 789 | 388 | 49.2% |
| $10M-$50M | 1,142 | 414 | 36.3% |
| $50M-$100M | 577 | 105 | 18.2% |
| $100M-$500M | 1,070 | 34 | 3.2% |
| $500M-$1B | 284 | 2 | 0.7% |
| $1B-$10B | 450 | 2 | 0.4% |
| >$10B | 24 | 0 | 0.0% |
Read the last column, not the second. A tier holding many institutions will hold many elevated ones; the share is what says whether the tier is unusual.
Where the pressure is geographically
States with at least 15 institutions in the panel, ranked by the share scoring Elevated or High. Small-state counts are excluded rather than shown with a caveat, because a share computed on nine institutions moves eleven points when one of them changes band.
| State | Institutions | Elevated or High | Share |
|---|---|---|---|
| NJ | 127 | 48 | 37.8% |
| LA | 136 | 47 | 34.6% |
| DE | 15 | 5 | 33.3% |
| AR | 49 | 16 | 32.7% |
| NE | 50 | 16 | 32.0% |
| CT | 64 | 20 | 31.2% |
| DC | 29 | 9 | 31.0% |
| IL | 190 | 56 | 29.5% |
| KY | 52 | 15 | 28.8% |
| VA | 95 | 27 | 28.4% |
| TN | 126 | 34 | 27.0% |
| OK | 53 | 14 | 26.4% |
What changed this quarter
| Band | Last quarter | This quarter | Change |
|---|---|---|---|
| Low | 1,863 | 1,838 | -25 |
| Moderate | 1,564 | 1,523 | -41 |
| Elevated | 660 | 696 | +36 |
| High | 214 | 249 | +35 |
A band transition is the event worth acting on. An institution that has sat in Moderate for eight quarters is background; one that crossed into Elevated this quarter did so because something on its balance sheet moved, and the score decomposition names which component moved.
How the score is built
Merger susceptibility is a published scorecard, not a black box: seven components — scale, membership trend, loan trend, earnings, capital adequacy against the 7% regulatory threshold, charter age and charge-off rate — each scaled to 0–1 and blended at documented weights. Every score decomposes back into weight × feature, and the product shows that decomposition per institution.
Separately, a fitted model calibrates a one-year merger probability, validated out-of-time against a labeled outcome rather than against itself. The label is narrow on purpose: a merger the regulator attributes to the credit union's own condition, so a strategic combination of two healthy charters counts as a negative. Against that label the published results are 0.86 AUC out-of-time, with the top scored decile capturing roughly 56% of distress mergers — about 5.6× the hit rate of a random list of the same length.
Those figures rest on 89 events, which puts the standard error on the AUC near 0.04, and the held-out window contained no distress merger above $500M in assets. Above that threshold the model is untested rather than weak, and it is also silent: 49 of its top 50 names are under $100M. The method and its limits are published in full.
This post reports the system. It names no institution, by design — the counts and distributions here are the free half, and the ranked, named, exportable list underneath them is what a subscription opens.
Caveats
Every figure above is a model estimate computed from the quarterly panel as of the quarter named at the top of this post. Scores are not investment, credit or merger advice, and nothing here is a recommendation about any institution. The scorecards, their published weights and their known limitations are set out in the disclosures.