Three structural problems in 2025 bank credit-quality data show up clearly across three quarters of FDIC Call Report filings published through Q3 2025: net charge-off ratios reading at or rounding to 0.00% across multiple quarters at three of the largest U.S. banks holding combined assets of $846 billion, a $222 billion bank's nonperforming loan field appearing for the first time in Q3 2025 after two prior quarters of zero readings, and a 95.8% systematic missingness rate across the broader bank universe. The combined effect of the three is that peer-comparison analytics built on this data are systematically incomplete, and the incompleteness has been stable across every quarter EF News examined this year.

Anomaly 1: Three Top-25 Banks Reported Net Charge-Off Ratios At Or Rounding To Zero For All Of 2025

The vendor-aggregated FDIC Call Report data for the three quarters Q1 2025, Q2 2025, and Q3 2025 shows the following readings for three of the largest U.S. banks by asset size:

  • Fifth Third Bank ($212B): NPL 0.00% / 0.00% / 0.00%; NCO 0.00% / 0.00% / 0.00%
  • Santander Bank ($98B): NPL 0.00% / 0.00% / 0.00%; NCO 0.00% / 0.00% / 0.00%
  • Truist Bank ($536B): NPL 1.03% / 0.96% / 1.04%; NCO 0.00% / 0.00% / 0.00%

Combined assets across the three institutions: $846 billion. Fifth Third and Santander show zero readings across both nonperforming loans and net charge-offs for every quarter of 2025 to date. A $212 billion bank with literally zero nonperforming loans is not a credit reading; it is a reporting reading. Truist shows the same pattern selectively: NPL is populated and reads near 1%, but the corresponding net charge-off ratio reads at 0.00% across all three quarters. A $536 billion bank carrying 1% nonperforming loans either cures, modifies, or charges off those loans, and sustained 0.00% charge-off ratios across three consecutive quarters at that scale are not plausible under any of those resolution paths.

The most likely explanation is vendor-pipeline coverage. The FDIC Call Report system captures bank charge-off data across multiple schedules depending on charter type and reporting election, and vendor data products that scrape the FDIC public APIs may pull from a subset of those schedules. Banks whose primary charge-off reporting flows through unscraped schedules show as zero in the aggregated dataset.

For peer benchmarking, the implication is direct: any comparison built on the vendor-aggregated dataset is comparing populated banks against three of the largest U.S. peers reading at or near zero. Medians, quartiles, and percentile rankings produced by that comparison are wrong, and the direction of the wrongness is favorable to the populated banks.

Anomaly 2: Citizens Bank's Nonperforming Loan Field Appeared For The First Time In Q3 2025

Citizens Bank ($222B) reported the following Nonperforming Loan ratios across the first three quarters of 2025 per the same FDIC Call Report dataset:

  • Q1 2025: 0.00%
  • Q2 2025: 0.00%
  • Q3 2025: 1.79%

A real credit event at a $222 billion bank does not produce a 0.00% to 1.79% jump in a single quarter. Real credit deterioration at this scale shows as gradual increases in 30-89 day delinquency, then 90+ day delinquency, then nonaccrual classification, with the lag between front-end stress and NPL classification typically running 2 to 4 quarters. Citizens Bank's two consecutive 0.00% readings followed by a 1.79% reading in Q3 2025 is more consistent with the bank starting to populate the NPL field in Q3 2025 than with real deterioration occurring in that quarter.

A practitioner running peer-comparison analytics in November 2025, when Q3 2025 data first published, would have flagged Citizens Bank as a deteriorating institution. That flag may be wrong. The benchmarking literature contains no explicit guidance on how to handle reporting onsets in long-running vendor datasets.

EF News did not contact Citizens Bank for comment on the underlying classification timeline. The bank's parent, Citizens Financial Group (NYSE: CFG), publishes 10-Q disclosures that would resolve the question definitively. Anyone using vendor-aggregated FDIC Call Report data for peer credit benchmarking should verify the Q3 2025 onset against the institution's own filings before treating the change as a credit signal.

Anomaly 3: 95.8% Of Banks Show All-Zero Credit Quality Fields, And It Has Been Stable For Three Quarters

Of the 4,379 U.S. banks in the Q3 2025 FDIC Call Report dataset, 4,194 (95.8%) report values of zero across the full set of credit-quality columns: 30-89 day past due, 90+ day past due, nonaccrual, nonperforming, charge-offs, recoveries, and net charge-offs. The pattern has been stable across all three 2025 quarters published to date, with all-zero rates running 95.8% to 95.9% in every period.

Only 185 banks (4.2%) at Q3 2025 report any non-zero credit-quality data. The 185 do not include 14 of the 15 banks holding $100 billion or more in assets that read as all-zero:

  • JPMorgan Chase Bank ($3.81 trillion)
  • Citibank ($1.84 trillion)
  • Capital One ($652 billion)
  • Goldman Sachs Bank USA ($644 billion)
  • The Bank of New York Mellon ($367 billion)
  • State Street Bank and Trust Company ($366 billion)
  • Morgan Stanley Bank ($250 billion)
  • Charles Schwab Bank ($243 billion)
  • Morgan Stanley Private Bank ($240 billion)
  • American Express National Bank ($211 billion)
  • The Northern Trust Company ($170 billion)
  • HSBC Bank USA ($162 billion)
  • UBS Bank USA ($115 billion)
  • USAA Federal Savings Bank ($105 billion)

Combined assets across the 14 large banks reading as all-zero: approximately $9.2 trillion. The exclusion pattern correlates with charter type and business model. Card-issuing banks (Capital One, American Express NB), custody banks (BNY Mellon, State Street, Northern Trust, Schwab), money-center banks (JPMorgan Chase, Citibank), foreign-affiliated U.S. banks (HSBC USA, UBS Bank USA), and investment-bank affiliates (Goldman Sachs Bank USA, Morgan Stanley Bank) are systematically more likely to read all-zero than commercial-bank peers.

The exclusion is almost certainly not because these banks have zero credit-quality data to report. Capital One holds nearly $200 billion in credit card receivables, a portfolio with industry-typical charge-off rates running 4% to 6%. American Express National Bank specializes in card lending. The credit-quality data exists at these institutions. The vendor pipeline is not capturing it.

What Practitioners Should Do With This

For credit analysts, portfolio managers, and competitive intelligence functions building benchmarks from vendor-aggregated FDIC Call Report data:

  • Filter to populated banks before computing peer medians. A 185-bank populated peer set produces materially different benchmarks than a 4,379-bank universe with 95.8% zeros mixed in.
  • Cross-check large reported QoQ changes against the institution's 10-Q. Reporting onsets, classification changes, and vendor-pipeline gaps produce step-function changes that look like credit events. Most are not.
  • Treat zero charge-off readings at large banks as missing data, not as performance signals. A $536 billion bank with 0.00% NCO over multiple quarters is a data flag.
  • For the largest banks (JPMorgan, Citi, Capital One, Goldman, BNY Mellon, etc.), do not rely on vendor-aggregated FDIC data for credit benchmarking. Use SEC 10-Q filings directly.

The Gap

Industry coverage of bank credit performance leans heavily on vendor-aggregated FDIC Call Report data because it is fast, comprehensive in name, and historically reliable for the commercial-bank universe. The three anomalies above are not new; the patterns have been visible across every 2025 quarter EF News examined. They have not been named in the industry press, and no major data provider has published guidance on how to handle the systematic missingness in peer-comparison work.

The structural problem is that the dataset is described as if it covers the bank universe when in practice it covers commercial banks above a certain asset threshold whose charter type and business model align with the schedules the vendor pipeline scrapes. Foreign-affiliated banks, custody banks, card-issuing banks, money-center banks, and investment-bank affiliates are systematically under-represented. The 95.8% missingness number is the headline. The structural skew in which 95.8% are missing is the unstated implication.

Inference: The combination of zero NCO readings at three of the largest U.S. banks, a single-quarter NPL onset at a $222 billion bank, and 95.8% systematic missingness across the bank universe is consistent with vendor-pipeline coverage gaps in how FDIC Call Report credit-quality fields are captured rather than gaps in the underlying regulatory filings; the FDIC almost certainly has the data, but the peer-comparison datasets practitioners use do not fully reach it.

What To Watch In Q4 2025 And Beyond

Q4 2025 FDIC Call Report data will publish in February 2026. Three signals are worth tracking:

  • Whether Truist, Fifth Third, and Santander's NCO ratios remain at 0.00% in Q4 2025, or whether the readings begin to populate. Continued zeros confirm the vendor-pipeline gap; non-zero readings would suggest a reporting change at the banks rather than a vendor issue.
  • Whether Citizens Bank's NPL trajectory continues at the 1.79% level or shows another step-function change. Stable readings would suggest the Q3 2025 onset reflected a true classification change at the bank; volatility would suggest the field remains unreliable at the vendor level.
  • Whether the 95.8% all-zero rate moves at all. The pattern has been stable across three quarters of 2025. Any meaningful change would indicate a vendor-pipeline update.

The patterns above do not invalidate every conclusion drawn from FDIC Call Report data. They do invalidate any peer-comparison analytic that treats the dataset as if it covers the bank universe.