The Provision Lag — Q4 2025
The Widening Gap
Delinquency and charge-offs came apart through 2025. System delinquency climbed from 0.80% in Q1 2025 to 1.03% in Q4 2025, while the annualized net charge-off rate stayed roughly flat: 0.82%, 0.79%, 0.76% and 0.77% in the four quarters. Charge-offs ran above delinquency in Q1 2025; by year-end delinquency sat 26 basis points higher. We read the difference as past-due balances accumulating in late buckets rather than resolving into losses.
Source: NCUA Call Report, Q4 2025 · release v2026.2.27 · QuarterCall
Provisioning tracked realized losses rather than the delinquency build. Full-year 2025 provision expense was $14.4 billion, essentially flat against 2024 (up 1%), and the allowance grew to $23.2 billion. Provisions sized to flat charge-offs fit the losses taken so far and lag a delinquency pipeline that is rising.
Provisions and Earnings
We tested whether flat provisioning against rising delinquency flattered 2025 net income, with reserves lower than they should be. The data do not show it. Provision expense fell as a share of net interest income over the year, to 18.0% from 20.1%, because net interest income grew faster than provisioning, and ROAA rose year over year. Measured against the charge-offs taken in 2025, reserves kept pace and earnings were not overstated.
We read the exposure as one of timing: 2025's earnings are not overstated against the losses already taken, and they remain exposed to the losses today's delinquency has yet to produce. The provision lag is the gap between a delinquency rate that has already moved and the charge-offs and provisions that have not yet caught up.
By Asset-Size Cohort
Size splits the system aggregate. The over-$10B cohort carries both the highest delinquency (1.49%) and the highest charge-offs (1.31%). The aggregate net charge-off rate stays low because the mid-size and small tiers, charging off 0.47% to 0.60%, have not followed their delinquency higher. We read the largest credit unions as a quarter or two ahead in a conversion the rest of the system has yet to make.
Delinquency vs. Charge-offs by Asset-Size Cohort, Q4 2025 (annualized)
| Cohort | Delinq. (60+) | NCO Ratio | ROAA |
|---|---|---|---|
| Under $100M | 1.09% | 0.47% | 0.77% |
| $100M–$500M | 0.89% | 0.50% | 0.75% |
| $500M–$1B | 0.88% | 0.57% | 0.70% |
| $1B–$10B | 0.85% | 0.60% | 0.78% |
| Over $10B | 1.49% | 1.31% | 0.81% |
The Deferred Cost
We expect the aged delinquencies in the mid-size and small tiers to convert in 2026, as the largest cohort's appear to have done: net charge-offs rise, provisioning rises to match, and ROAA gives back part of 2025's gain. The lag moves that cost into 2026 without avoiding it, which is why we do not extrapolate 2025's earnings gain. The credit bill is in the mail, not yet delivered.
Connecting the Lines
This breakdown sits behind two of the System Signals — Q4 2025 (March 15, 2026): the breach of the 1% delinquency mark (Signal 1) and the earnings discussion (Signal 3). The provision lag connects them: a quarter can post a record margin and rising full-year earnings while the delinquency we expect to weigh on both is still building. The gap between delinquency and charge-offs is the measure to follow; we expect earnings to turn in the quarter it closes.
- That delinquent loans are aging into late buckets is inferred from the gap between delinquency and charge-offs; the report shows no delinquency by bucket.
- Reserves are judged against charge-offs already taken; no coverage of delinquent loans by the allowance is computed.
- The tier comparison is one quarter's cross-section; that the largest tier leads the others in time is a reading of it, not a measured sequence.
- The conversion of delinquency into charge-offs in 2026 is the report's expectation; its data ends at Q4 2025.