Where Growth Is Outrunning Credit Discipline in Smaller Credit Unions
An early-warning read on the $100M–$250M asset band, three to four quarters out, as of the Q4 2025 call report.
This note reads the deterioration signal building in the peer cohort of federally insured credit unions holding $100M to $250M in assets — 678 institutions, measured against the December 31, 2025 call report. It is a directional analysis: we do not name individual credit unions, but describe where the cohort is leaning three to four quarters ahead.
The read covers the specific, directional risk themes the early-warning system tracks for this cohort — growth, asset quality, portfolio, earnings, liquidity, and efficiency — where the forward signal is most separable and the raw-metric corroboration is strongest. Every theme below carries strong raw-metric evidence: the modeled risk direction is confirmed by the institutions’ own reported NCUA metrics for roughly four-fifths or more of flagged cases. One target under validation review is excluded.
A note on reading the figures: the theme shares are independent, overlapping signals, not a breakdown of a whole. A single credit union can screen elevated on more than one theme, so the percentages are not mutually exclusive and do not sum to the cohort.
01 What we are observing now
This band leans differently from larger community credit unions. The clear lead is growth / expansion stress, elevated for roughly 25% of the cohort as of Q4 2025 — followed closely by asset quality (≈17%) and portfolio deterioration (≈17%), then earnings (≈16%). Liquidity sits lower here (≈13%) than in larger bands — smaller balance sheets carry less wholesale-funding sensitivity, so funding is not the lead risk.
The read is one of growth quality: in smaller institutions the forward signal concentrates where expansion is running ahead of underwriting and reserve build, and where credit-quality and portfolio metrics are already beginning to confirm it. This is a credit-cycle posture, not a funding-structure posture.
02 Why it matters for balance sheet and P&L
Growth quality. The lead signal pairs rapid member-, asset-, and loan-growth with rising charge-off and delinquency context. Fast growth is not itself a problem; growth that outruns underwriting standards and reserve build is. For a $100M–$250M institution, a single fast-growing segment can shift the credit profile of the whole book within a few quarters.
Asset quality and portfolio. The near-tied second-tier themes reflect credit deterioration already surfacing — net charge-off and delinquency rates, allowance coverage, and segment-level delinquency moving off-peer. On the income statement the proximate drag appears through operating-expense-to-assets, provisioning, and ROA.
03 What near-term risk this points to
At the T3/T4 horizon (three to four quarters from the Q4 2025 cycle) the cohort’s lean points to a sequence:
- Growth quality first — the lead theme; expansion outrunning underwriting and reserve build is the earliest-firing pattern in this band.
- Asset-quality and portfolio confirmation second — net charge-off, delinquency, and segment-level credit metrics begin to surface the deterioration, already visible in this cohort.
- Earnings drag third — provisioning and efficiency pressure convert credit deterioration into thinner margins.
This is a credit-cycle vulnerability read, not a distress call: a meaningful slice of the band is expanding in a way that a normal credit cycle would test more than peer-median institutions.
04 Historical validation and model evidence
The signals come from gradient-boosted and glass-box models selected through a governed champion-selection process. Across the themes in scope, held-out discrimination is strong: validation AUC runs from the high-0.70s to low-0.90s (median ≈0.82), KS in the 0.4–0.7 range (median ≈0.50), and top-decile lift between 2.6× and 4.5× (median ≈3.5×), with score-population stability (PSI) green on every model.
| Risk theme | Horizon | Validation AUC | KS | Top-decile lift | Model confidence |
|---|---|---|---|---|---|
| Growth / expansion stress | T4 | 0.81 | 0.45–0.46 | 3.4× | High |
| Asset quality | T4 | 0.86 | 0.55 | 4.4× | High |
| Portfolio deterioration | T4 | 0.77 | 0.40 | 3.0× | Strong |
| Earnings stress | T4 | 0.80–0.89 | 0.47–0.62 | 3.2–3.8× | High |
| Liquidity pressure | T3–T4 | 0.76–0.85 | 0.38–0.54 | 2.6–3.6× | Strong High |
| Management efficiency | T4 | 0.82–0.92 | 0.50–0.69 | 3.5–4.5× | High |
| Composite risk signal | T4 | 0.91 | 0.67 | 4.0× | High |
05 Current-cycle screen
Against the December 31, 2025 call report, across the 678-institution band, growth/expansion stress is the dominant theme, followed by asset quality and portfolio deterioration in a near tie. Unlike the larger asset bands, liquidity is not the lead risk here — it screens elevated for only about an eighth of the cohort. Raw-metric corroboration is high across themes — between roughly 81% and 89% of flagged cases confirmed — which supports the strong-evidence designation for this cycle.
The macro overlay reinforces the read: across the state-quarters in scope, labor-market pressure is the modal macro context (the majority of state-quarters), with a minority showing broader systemic pressure or housing-momentum softening. Macro context supports the risk story far more often than it contradicts it.
| Risk theme | Elevated | Top-decile cutoff | Confirmed evidence | Model confidence |
|---|---|---|---|---|
| Growth / expansion stress | 25.4% | 0.176 | 89.1% | High |
| Asset quality | 17.1% | 0.212 | 84.3% | High |
| Portfolio deterioration | 17.0% | 0.217 | 84.3% | Strong High |
| Earnings stress | 15.5% | 0.205 | 85.3% | High |
| Liquidity pressure | 12.8% | 0.232 | 81.0% | Strong High |
| Management efficiency | 11.7% | 0.100 | 84.6% | High |
| Composite risk signal | 11.2% | 0.289 | 80.7% | High |
06 Raw metrics supporting the signal
The model direction is corroborated by these reported call-report metrics, the most frequently deteriorating across flagged institutions in this cycle:
- Growth quality: member-, asset-, and loan-growth running ahead of peers, paired with allowance-to-delinquent-loans and net charge-off context.
- Asset quality: net charge-off, total and 60–179-day delinquency rates, allowance coverage — already surfacing in this band.
- Portfolio: segment-level delinquency and charge-off (used-auto, other real estate) and intermediate income and charge-off flows.
- Earnings / efficiency: operating-expense-to-assets and cost-to-income elevated; ROA soft.
- Liquidity (secondary): loans-to-shares and cash-to-assets — present but not the lead cluster in this band.
The lead corroboration is growth-quality and credit performance — which is why a forward model adds value: it flags the expansion posture before the credit metrics fully confirm.
07 Model drivers
The features carrying the most weight are overwhelmingly peer-relative and percentile-based, not raw levels — the models flag institutions that are off-peer, not merely large or small:
- Growth: member-, asset-, and loan-growth levels with state-level deterioration context, plus loan-to-share and portfolio-concentration dynamics relative to peers.
- Asset quality: total delinquency and net charge-off relative to peers and as state percentiles, with allowance-coverage context.
- Portfolio: used-auto and other-real-estate delinquency dynamics relative to peers, and intermediate income and charge-off flows.
- Earnings & efficiency: operating-expense-to-assets, ROA peer percentiles, cost-to-income asset-band percentiles, and a negative-ROA flag.
- Liquidity: loans-to-shares and cash-to-assets relative to peers — a secondary driver in this band.
The recurring theme is relative position — growth and credit metrics measured against peer group and state — which makes the signal portable across the band rather than an artifact of any one institution.
08 Practical implications
For the CRO
The lead risk in this band is growth quality, not funding. The usable question is whether loan and asset growth are running ahead of the peer band while allowance coverage and charge-off context move the wrong way — that combination is where the model concentrates its forward signal, confirmed by raw metrics in the large majority of flagged cases.
For lending leadership
This is the front line for this cohort. Underwriting standards and reserve discipline in the fastest-growing segments are the most direct response to a forward growth-stress flag — particularly where asset-quality and portfolio metrics are already beginning to confirm.
For the CFO
Model the provisioning path: if growth continues at the current pace and the credit-quality drift in the book continues, what does the reserve build do to the earnings and net-worth trajectory over the next three to four quarters? At this size, that is the highest-leverage scenario.
For the board
Frame it forward and probabilistic. The models discriminate well (AUC ≈0.82 median, lift ≈3.5×) and the institutions’ own reported metrics corroborate the direction for roughly four-fifths or more of flagged cases — strong raw-metric support — though full-cycle backtesting is still accruing and these remain discrimination signals, not calibrated loss probabilities. As of Q4 2025, growth running ahead of credit discipline is the clearest forward risk in the $100M–$250M peer cohort, visible three to four quarters before credit metrics fully confirm.
Methodology: Signals derive from NCUA call-report data through the December 31, 2025 cycle, modeled at the credit-union/quarter grain with forward targets at the three- and four-quarter horizon. Reported AUC, KS, and top-decile lift are held-out validation metrics; out-of-time backtesting is ongoing. “Strong evidence” denotes model direction corroborated by raw reported metrics — trend, peer position, and macro context — for the large majority of flagged cases. Theme shares are independent and overlapping and do not sum to the cohort. This analysis is directional and cohort-level; it does not identify individual institutions and is not investment, legal, or supervisory advice.
Where Funding Strain Is Building in Mid-Sized Credit Unions
An early-warning read on the $250M–$500M asset band, three to four quarters out, as of the Q4 2025 call report.
This note reads the deterioration signal building in the peer cohort of federally insured credit unions holding $250M to $500M in assets — 391 institutions, measured against the December 31, 2025 call report. It is a directional analysis: we do not name individual credit unions, but describe where the cohort is leaning three to four quarters ahead.
The read covers the specific, directional risk themes the early-warning system tracks for this cohort — funding, growth, earnings, asset quality, and efficiency — where the forward signal is most separable and the raw-metric corroboration is strongest. Every theme below carries strong raw-metric evidence: the modeled risk direction is confirmed by the institutions’ own reported NCUA metrics for roughly four-fifths or more of flagged cases. One target under validation review is excluded.
A note on reading the figures: the theme shares are independent, overlapping signals, not a breakdown of a whole. A single credit union can screen elevated on more than one theme, so the percentages are not mutually exclusive and do not sum to the cohort.
01 What we are observing now
The band’s lean is led by liquidity pressure, which screens elevated for roughly 23% of the cohort as of Q4 2025 — and its elevated tail is the most separated of any theme, carrying the highest top-decile cutoff. Growth / expansion stress follows at ≈18%. The remaining themes — earnings, asset quality, the composite signal, portfolio, and efficiency — each sit in the 3–8% range.
The read is structural: balance-sheet funding tightens ahead of any visible deterioration in credit performance. For mid-sized credit unions, the funding/liquidity dimension is the clearest and most separated forward risk in this cycle.
02 Why it matters for balance sheet and P&L
Funding structure. The liquidity signal is driven by elevated loans-to-shares and loans-to-assets, compressed cash-to-assets, and rising borrowings-to-assets and funding concentration. A larger loan book funded against a thinner liquid cushion makes net worth and earnings more sensitive to any stall in share growth or rise in wholesale-funding reliance.
Growth quality. The growth signal pairs rapid member-, asset-, and loan-growth with rising charge-off and delinquency context. Fast growth is not itself a problem; growth that outruns underwriting and reserve build is. On the income statement the proximate drag appears through operating-expense-to-assets, cost-to-income, and ROA.
03 What near-term risk this points to
At the T3/T4 horizon (three to four quarters from the Q4 2025 cycle) the cohort’s lean points to a sequence:
- Funding and liquidity first — the most prominent and most-separated theme; high loans-to-shares with compressed cash and rising borrowings is the earliest-firing pattern.
- Earnings drag second — elevated operating-expense and cost-to-income ratios with soft ROA convert funding pressure into thinner margins.
- Asset-quality confirmation third — net charge-off and delinquency metrics typically confirm the deterioration later, after the funding and earnings channels.
This is a structural-vulnerability read, not a distress call. The well-separated liquidity tail is the feature most specific to this band: when liquidity flags here, it tends to flag firmly rather than marginally.
04 Historical validation and model evidence
The signals come from gradient-boosted and glass-box models selected through a governed champion-selection process. Across the themes in scope, held-out discrimination is strong: validation AUC runs from the high-0.70s to low-0.90s (median ≈0.82), KS in the 0.4–0.7 range (median ≈0.50), and top-decile lift between 2.6× and 4.5× (median ≈3.5×), with score-population stability (PSI) green on every model.
| Risk theme | Horizon | Validation AUC | KS | Top-decile lift | Model confidence |
|---|---|---|---|---|---|
| Liquidity pressure | T3–T4 | 0.76–0.85 | 0.38–0.54 | 2.6–3.6× | Strong High |
| Growth / expansion stress | T4 | 0.81 | 0.45–0.46 | 3.4× | High |
| Earnings stress | T4 | 0.80–0.89 | 0.47–0.62 | 3.2–3.8× | High |
| Asset quality | T4 | 0.86 | 0.55 | 4.4× | High |
| Composite risk signal | T4 | 0.91 | 0.67 | 4.0× | High |
| Portfolio deterioration | T4 | 0.77 | 0.40 | 3.0× | Strong |
| Management efficiency | T4 | 0.82–0.92 | 0.50–0.69 | 3.5–4.5× | High |
05 Current-cycle screen
Against the December 31, 2025 call report, across the 391-institution band, liquidity pressure is the dominant theme, followed by growth stress. The most striking feature is liquidity’s separation: its top-decile cutoff (≈0.43) is the highest of any theme in the band, meaning its elevated tail is sharply distinct from the cohort median rather than a gradual gradient. Raw-metric corroboration is high across themes — between roughly 80% and 91% of flagged cases confirmed — which supports the strong-evidence designation for this cycle.
The macro overlay reinforces the read: across the state-quarters in scope, labor-market pressure is the modal macro context (the majority of state-quarters), with a minority showing broader systemic pressure or housing-momentum softening. Macro context supports the risk story far more often than it contradicts it.
| Risk theme | Elevated | Top-decile cutoff | Confirmed evidence | Model confidence |
|---|---|---|---|---|
| Liquidity pressure | 23.3% | 0.432 | 85.3% | Strong High |
| Growth / expansion stress | 17.6% | 0.130 | 91.3% | High |
| Earnings stress | 7.9% | 0.138 | 88.0% | High |
| Asset quality | 7.4% | 0.107 | 83.8% | High |
| Composite risk signal | 5.6% | 0.158 | 79.5% | High |
| Portfolio deterioration | 4.3% | 0.148 | 90.4% | Strong High |
| Management efficiency | 3.6% | 0.050 | 87.5% | High |
06 Raw metrics supporting the signal
The model direction is corroborated by these reported call-report metrics, the most frequently deteriorating across flagged institutions in this cycle:
- Funding / liquidity: loans-to-shares and loans-to-assets elevated; cash-to-assets compressed; borrowings-to-assets and funding concentration rising — the most prominent cluster in this band.
- Growth quality: member-, asset-, and loan-growth running ahead of peers, with charge-off and delinquency context building.
- Earnings / efficiency: operating-expense-to-assets and cost-to-income elevated; ROA soft.
- Asset quality (confirming): net charge-off, recovery, and 60–179-day delinquency rates — later-stage confirmations rather than the lead signal.
The lead corroboration is structural — funding and growth quality — with credit-quality metrics confirming later, which is why a forward model adds value over a delinquency-watching approach.
07 Model drivers
The features carrying the most weight are overwhelmingly peer-relative and percentile-based, not raw levels — the models flag institutions that are off-peer, not merely large or small:
- Liquidity: loans-to-shares and loans-to-assets relative to peers and as state percentiles, borrowings-to-assets, and a funding-concentration proxy — the band’s signature theme.
- Growth: member-, asset-, and share-growth levels with state-level deterioration context, plus auto and real-estate share dynamics relative to peers.
- Earnings & efficiency: operating-expense-to-assets, ROA peer percentiles, cost-to-income asset-band percentiles, and a negative-ROA flag.
- Asset quality: total delinquency relative to peers and as state percentiles, with first-mortgage charge-off context.
- Portfolio & composite: used-auto delinquency dynamics relative to peers, intermediate income and charge-off flows, and delinquency position versus peers.
The recurring theme is relative position — funding and growth measured against peer group and state — which makes the signal portable across the band rather than an artifact of any one institution.
08 Practical implications
For the CRO
The funding/liquidity dimension is the clearest and most-separated risk in this band as of Q4 2025. The usable question is whether loans-to-shares is high relative to the peer band with cash-to-assets compressed and borrowings rising at the same time — that combination is where the model concentrates its forward liquidity signal, confirmed by raw metrics in the large majority of flagged cases.
For lending leadership
Growth quality is the second-order watch item. The signal pairs fast member/asset/loan growth with building charge-off context; underwriting and reserve discipline in the fastest-growing segments is the most direct response to a forward growth-stress flag.
For the CFO
Given the prominence of liquidity, stress-test the funding side first: what happens to the net-worth and earnings trajectory if share growth stalls and borrowing dependence rises into a normal provisioning cycle? That combination is where the band’s headroom is thinnest.
For the board
Frame it forward and probabilistic. The models discriminate well (AUC ≈0.82 median, lift ≈3.5×) and the institutions’ own reported metrics corroborate the direction for roughly four-fifths or more of flagged cases — strong raw-metric support — though full-cycle backtesting is still accruing and these remain discrimination signals, not calibrated loss probabilities. As of Q4 2025, funding/liquidity fragility is the clearest forward risk in the $250M–$500M peer cohort, visible three to four quarters before credit metrics confirm.
Methodology: Signals derive from NCUA call-report data through the December 31, 2025 cycle, modeled at the credit-union/quarter grain with forward targets at the three- and four-quarter horizon. Reported AUC, KS, and top-decile lift are held-out validation metrics; out-of-time backtesting is ongoing. “Strong evidence” denotes model direction corroborated by raw reported metrics — trend, peer position, and macro context — for the large majority of flagged cases. Theme shares are independent and overlapping and do not sum to the cohort. This analysis is directional and cohort-level; it does not identify individual institutions and is not investment, legal, or supervisory advice.
Where Funding Strain Is Building in Larger Community Credit Unions
An early-warning read on the $500M–$1B asset band, three to four quarters out, as of the Q4 2025 call report.
This note reads the deterioration signal building in the peer cohort of federally insured credit unions holding $500M to $1B in assets — 282 institutions, measured against the December 31, 2025 call report. It is a directional analysis: we do not name individual credit unions, but describe where the cohort is leaning three to four quarters ahead.
The read covers the specific, directional risk themes the early-warning system tracks for this cohort — funding, growth, earnings, asset quality, and efficiency — where the forward signal is most separable and the raw-metric corroboration is strongest. Every theme below carries strong raw-metric evidence: the modeled risk direction is confirmed by the institutions’ own reported NCUA metrics for roughly four-fifths or more of flagged cases. One target under validation review is excluded.
A note on reading the figures: the theme shares are independent, overlapping signals, not a breakdown of a whole. A single credit union can screen elevated on more than one theme, so the percentages are not mutually exclusive and do not sum to the cohort.
01 What we are observing now
The band’s lean is led by liquidity pressure, which screens elevated for roughly 32% of the cohort as of Q4 2025 — and its elevated tail is the most separated of any theme, carrying the highest top-decile cutoff in the band. Growth / expansion stress follows at ≈22%. The remaining themes — asset quality, earnings, the composite signal, efficiency, and portfolio — each sit in the 2–12% range.
The read is structural: balance-sheet funding tightens ahead of any visible deterioration in credit performance. What distinguishes the $500M–$1B band is that the funding/liquidity dimension is the most pronounced and most sharply separated risk — these institutions carry larger, more wholesale-sensitive balance sheets than smaller community credit unions.
02 Why it matters for balance sheet and P&L
Funding structure. The liquidity signal is driven by elevated loans-to-shares and loans-to-assets, compressed cash-to-assets, and rising borrowings-to-assets and funding concentration. A larger loan book funded against a thinner liquid cushion makes net worth and earnings more sensitive to any stall in share growth or rise in wholesale-funding reliance — and at this balance-sheet size the absolute dollars at stake are material.
Growth quality. The growth signal pairs rapid member-, asset-, and loan-growth with rising charge-off and delinquency context. Fast growth is not itself a problem; growth that outruns underwriting and reserve build is. On the income statement the proximate drag appears through operating-expense-to-assets, cost-to-income, and ROA.
03 What near-term risk this points to
At the T3/T4 horizon (three to four quarters from the Q4 2025 cycle) the cohort’s lean points to a sequence:
- Funding and liquidity first — the most prominent and most-separated theme; high loans-to-shares with compressed cash and rising borrowings is the earliest-firing pattern.
- Earnings drag second — elevated operating-expense and cost-to-income ratios with soft ROA convert funding pressure into thinner margins.
- Asset-quality confirmation third — net charge-off and delinquency metrics typically confirm the deterioration later, after the funding and earnings channels.
This is a structural-vulnerability read. The well-separated liquidity tail is the feature most specific to this band: when liquidity flags here, it tends to flag firmly rather than marginally.
04 Historical validation and model evidence
The signals come from gradient-boosted and glass-box models selected through a governed champion-selection process. Across the themes in scope, held-out discrimination is strong: validation AUC runs from the high-0.70s to low-0.90s (median ≈0.82), KS in the 0.4–0.7 range (median ≈0.50), and top-decile lift between 2.6× and 4.5× (median ≈3.5×), with score-population stability (PSI) green on every model.
| Risk theme | Horizon | Validation AUC | KS | Top-decile lift | Model confidence |
|---|---|---|---|---|---|
| Liquidity pressure | T3–T4 | 0.76–0.85 | 0.38–0.54 | 2.6–3.6× | Strong High |
| Growth / expansion stress | T4 | 0.81 | 0.45–0.46 | 3.4× | High |
| Asset quality | T4 | 0.86 | 0.55 | 4.4× | High |
| Earnings stress | T4 | 0.80–0.89 | 0.47–0.62 | 3.2–3.8× | High |
| Composite risk signal | T4 | 0.91 | 0.67 | 4.0× | High |
| Management efficiency | T4 | 0.82–0.92 | 0.50–0.69 | 3.5–4.5× | High |
| Portfolio deterioration | T4 | 0.77 | 0.40 | 3.0× | Strong |
05 Current-cycle screen
Against the December 31, 2025 call report, across the 282-institution band, liquidity pressure is the dominant theme, followed by growth stress. The most striking feature is liquidity’s separation: its top-decile cutoff (≈0.59) is the highest of any theme in either this band or the smaller mid-sized cohort, meaning its elevated tail is sharply distinct from the cohort median rather than a gradual gradient. Raw-metric corroboration is high across themes — between roughly 80% and 89% of flagged cases confirmed — which supports the strong-evidence designation for this cycle.
The macro overlay reinforces the read: across the state-quarters in scope, labor-market pressure is the modal macro context (the majority of state-quarters), with a minority showing broader systemic pressure or housing-momentum softening. Macro context supports the risk story far more often than it contradicts it.
| Risk theme | Elevated | Top-decile cutoff | Confirmed evidence | Model confidence |
|---|---|---|---|---|
| Liquidity pressure | 31.9% | 0.586 | 85.3% | Strong High |
| Growth / expansion stress | 21.6% | 0.135 | 89.3% | High |
| Asset quality | 11.7% | 0.138 | 82.3% | High |
| Earnings stress | 11.3% | 0.175 | 83.2% | High |
| Composite risk signal | 7.8% | 0.208 | 79.7% | High |
| Management efficiency | 6.0% | 0.053 | 82.3% | High |
| Portfolio deterioration | 2.1% | 0.112 | 88.3% | Strong High |
06 Raw metrics supporting the signal
The model direction is corroborated by these reported call-report metrics, the most frequently deteriorating across flagged institutions in this cycle:
- Funding / liquidity: loans-to-shares and loans-to-assets elevated; cash-to-assets compressed; borrowings-to-assets and funding concentration rising — the most prominent cluster in this band.
- Growth quality: member-, asset-, and loan-growth running ahead of peers, with charge-off and delinquency context building.
- Earnings / efficiency: operating-expense-to-assets and cost-to-income elevated; ROA soft.
- Asset quality (confirming): net charge-off, recovery, and 60–179-day delinquency rates — later-stage confirmations rather than the lead signal.
The lead corroboration is structural — funding and growth quality — with credit-quality metrics confirming later, which is why a forward model adds value over a delinquency-watching approach.
07 Model drivers
The features carrying the most weight are overwhelmingly peer-relative and percentile-based, not raw levels — the models flag institutions that are off-peer, not merely large:
- Liquidity: loans-to-shares and loans-to-assets relative to peers and as state percentiles, borrowings-to-assets, and a funding-concentration proxy — the band’s signature theme.
- Growth: member-, asset-, and share-growth levels with state-level deterioration context, plus auto and real-estate share dynamics relative to peers.
- Earnings & efficiency: operating-expense-to-assets, ROA peer percentiles, cost-to-income asset-band percentiles, and a negative-ROA flag.
- Asset quality: total delinquency relative to peers and as state percentiles, with first-mortgage charge-off context.
- Portfolio & composite: used-auto delinquency dynamics relative to peers, intermediate income and charge-off flows, and delinquency position versus peers.
The recurring theme is relative position — funding and growth measured against peer group and state — which makes the signal portable across the band rather than an artifact of any one institution.
08 Practical implications
For the CRO
The funding/liquidity dimension is the most pronounced risk in this band as of Q4 2025, and its elevated tail is sharply separated. The usable question is whether loans-to-shares is high relative to the peer band with cash-to-assets compressed and borrowings rising at the same time — that combination is where the model concentrates its forward liquidity signal, confirmed by raw metrics in the large majority of flagged cases.
For lending leadership
Growth quality is the second-order watch item. The signal pairs fast member/asset/loan growth with building charge-off context; underwriting and reserve discipline in the fastest-growing segments is the most direct response to a forward growth-stress flag.
For the CFO
Given the prominence of liquidity, stress-test the funding side first and hard: what happens to the net-worth and earnings trajectory if share growth stalls and borrowing dependence rises into a normal provisioning cycle? At this balance-sheet size the absolute dollars make that the highest-leverage scenario to model.
For the board
Frame it forward and probabilistic. The models discriminate well (AUC ≈0.82 median, lift ≈3.5×) and the institutions’ own reported metrics corroborate the direction for roughly four-fifths or more of flagged cases — strong raw-metric support — though full-cycle backtesting is still accruing and these remain discrimination signals, not calibrated loss probabilities. As of Q4 2025, funding/liquidity fragility is the clearest, most-separated forward risk in the $500M–$1B peer cohort, visible three to four quarters before credit metrics confirm.
Methodology: Signals derive from NCUA call-report data through the December 31, 2025 cycle, modeled at the credit-union/quarter grain with forward targets at the three- and four-quarter horizon. Reported AUC, KS, and top-decile lift are held-out validation metrics; out-of-time backtesting is ongoing. “Strong evidence” denotes model direction corroborated by raw reported metrics — trend, peer position, and macro context — for the large majority of flagged cases. Theme shares are independent and overlapping and do not sum to the cohort. This analysis is directional and cohort-level; it does not identify individual institutions and is not investment, legal, or supervisory advice.