Bank Reporting Baseline

Methodology

Every number in a baseline report is measured from a public record and links to its source. We don't model, estimate, or convert anything into an hourly cost, because a number you can't trace is a number you can't defend. We compute revision history from FFIEC filing timestamps and read the bank's own Call Report line items against named supervisory screens; we only show a figure when it's genuinely measurable and, for the red-flag KPIs, only when it crosses a line.

Revision detection

Every Call Report filing in the FFIEC Central Data Repository carries a last-updated timestamp. Original submissions are due 30 days after quarter-end (a few filers get up to 35–45). So:

When several quarters share one last-updated date, that is a single multi-quarter correction event (a restatement discovered once and fixed across periods), and the report flags it as one event, not several.

What this can and cannot say. Public data shows that a filing changed and when, not what changed, why, or by how much. Baseline reports phrase these as “revised after original submission,” never as “errors we found.” Validity edits are pre-cleared by the CDR before a filing is accepted, so failed validity edits are not observable in filed data. Only amendments, quality edits, and continuity breaks are.

Longest error-detection gap

When several quarters are restated on the same day, the earliest of them was wrong from its quarter-end until the correction date. We report that interval, in months, as the longest stretch an error is known to have stood in the published record. It only appears when a multi-quarter restatement makes it measurable; otherwise we don't show it.

Revisions vs. same-size banks

We pull active banks on the same Call Report form with assets within roughly 0.6× to 1.6× of the target from the FDIC BankFindAPI, then score each one with the exact revision rule above over the same thirteen quarters, using filing timestamps we already fetched. The report shows the bank's revision count against the peer median and where it lands in the distribution. We only show it when at least eight comparable peers have a near-complete filing history.

CRE / construction concentration

From the bank's own Call Report loan schedule we compute construction & land loans and total commercial real estate (construction + multifamily + income-producing commercial, owner-occupied excluded where reported) as a percent of total risk-based capital. The 2006 Interagency CRE Guidance screens a bank for heightened review at 100% of capital in construction, or 300% in total CRE paired with 50%+ growth over three years. We flag a bank only when it crosses a line, state plainly whether the growth condition is also met, and always call these screening criteria, not limits or a compliance breach.

Reserve coverage of bad loans

Allowance for credit losses divided by noncurrent loans (nonaccrual plus 90+ days past due) — does the reserve still cover the loans already gone bad? We flag only a real collapse: coverage under 100%, on a material noncurrent book, that fell sharply (≥25 points in a quarter or ≥30 over a year) while noncurrent is rising. Because charge-offs lag, this stress often isn't in the P&L yet, which is exactly why it is worth surfacing against allowance-adequacy (CECL) expectations.

Unrealized securities losses vs capital

The mark-to-market gap on the bank's securities (available-for-sale fair value minus amortized cost, plus the same for held-to-maturity) as a percent of Tier 1 capital. For banks that elect the AOCI opt-out, these losses are excluded from reported regulatory capital — so the capital ratios look fine while the economic hole is real. We flag only when the underwater position exceeds 25% of Tier 1; below that we report it as clear. This is the mechanism behind the 2023 regional-bank failures.

Reporting-burden thresholds

When a bank's assets sit within 15% of a threshold that changes its reporting obligations, the profile flags it — the $500M and $1B audit and internal-control lines in 12 CFR Part 363, the $3B exam-cycle change, and the $5B point where a bank loses FFIEC 051 eligibility and moves to the longer 041. These are drawn straight from published rules, not modeled.

Credit unions

Credit union pages use the public data behind NCUA's Research a Credit Union tool for the profile and threshold flags. The measured filing-quality checks banks get are not available for credit unions: NCUA re-publishes corrected 5300 datasets in quarterly bulk files without the per-filing revision timestamps the FFIEC CDR gives for banks. Rather than model a substitute, the credit union report says so and points to the review, which measures those checks directly from the institution's own filings.

Data sources & freshness

Results are cached for 24 hours per institution. If a data source is unavailable we show a retry state; we never substitute an estimate for a measurement.

Back to the Baseline tool