Coverage & Provenance¶

How far back each public US bank-data collection reaches, where the data physically comes from, when major schedules were introduced, and where the known gaps are. This is a guide to the data's temporal coverage and vintage, written for an outside researcher assembling a long-run panel from public sources. It does not describe any particular database build — only the public collections themselves and the eras they cover.

All collections referenced here are published by US regulators (Federal Reserve, FFIEC, FDIC, OCC, NCUA, CFPB, SEC) or by the academic digitizations of regulator records cited at the end.

1. The collection universe at a glance¶

Collection	Provider	Native temporal coverage	Granularity
OCC historical national-bank balance sheets	OCC Annual Reports (+ academic digitization)	1863–1941	Bank-level, periodic call dates
Pre-1976 Call Reports ("Luck"/NY Fed dataset)	Federal Reserve Bank of New York (public)	1959Q4–1975Q4 (dataset spans 1959–2025)	Bank-quarter
Call Reports — Fed-direct, pre-2012	Federal Reserve Bank of Chicago archive	1976Q1–2011Q4	Bank-quarter
Call Reports — Fed-direct, 2012+	FFIEC Central Data Repository (CDR)	2012Q1–present	Bank-quarter
FR Y-9C (consolidated BHC financials)	FFIEC / Federal Reserve	1986Q3–present	BHC-quarter
UBPR analytical ratios	FFIEC CDR (UBPR)	2002Q4–present	Bank-quarter
FR Y-15 systemic indicators	FFIEC NIC	2020–present (annual)	G-SIB-year
FDIC SDI financials	FDIC API	1984Q1–present	Bank-quarter
FDIC Summary of Deposits	FDIC API	1994–present (annual)	Branch-year
FDIC failures (BankFind)	FDIC API	1934–present	Event
FDIC institution history	FDIC API	All dates	Event
NIC entity attributes / identifiers	FFIEC NIC	Current snapshot	Entity
DFAST stress-test results	Federal Reserve	2013–present (annual)	Bank-year
Basel III Pillar 3 disclosures	Bank disclosures (PDF)	~2024–present	G-SIB-quarter
NCUA 5300 credit-union call reports	NCUA	2022Q1–present (bulk era used here)	Credit-union-quarter
HMDA mortgage summary	CFPB HMDA Data Browser	2018–present	Institution-year
FRED banking/macro series	FRED (St. Louis Fed)	1919–present	Series-observation
SEC EDGAR identifiers	data.sec.gov	Current frame	Entity

2. Coverage by era¶

1863–1941 — National-bank era (OCC). The earliest machine-usable bank-level balance sheets come from the Comptroller of the Currency's Annual Reports, which published the condition of every national bank at periodic call dates from the founding of the national banking system in 1863 through 1941. These exist publicly today as OCC's own digitized Annual Reports and as academic digitizations of them (see §5). Coverage is national banks only; state banks of this era are not in this corpus.

1942–1958 — Quarterly gap. There is no public, machine-readable, bank-level quarterly balance-sheet series spanning the years between the OCC historical series and the start of the modern Call Report digital record. For these years, regulator holdings exist largely as physical archives. Annual-granularity coverage of this window can be obtained from academic historical bank panels (see §5), but quarterly coverage remains archive-only.

1959Q4–1975Q4 — Pre-digital Call Reports (NY Fed / "Luck"). The Federal Reserve Bank of New York publishes a free dataset of bank balance sheets and income statements ("Balance Sheets and Income Statements", associated with Correia–Luck–Verner) that provides bank-quarter coverage beginning 1959Q4. For the years before the Fed-direct digital record begins (1976), this is the authoritative public source of quarterly bank financials. The NY Fed dataset itself extends forward to the present, but for 1976+ it is redundant with the Fed-direct record (see §3).

1976Q1–2011Q4 — Fed-direct Call Reports (Chicago Fed archive). The Federal Reserve Bank of Chicago hosts the public archive of Commercial Bank Call Report data beginning 1976Q1. This is the start of the continuous modern digital Call Report record (FFIEC 031/041, later also 051).

2012Q1–present — Fed-direct Call Reports (FFIEC CDR). From 2012 onward, the public Call Report bulk data is served from the FFIEC Central Data Repository. The Chicago Fed (pre-2012) and CDR (2012+) records join into one continuous bank-quarter Call Report series from 1976 to the present.

1984+ / 1986+ — Modern supervisory layers. FDIC SDI financials begin 1984Q1; the FR Y-9C consolidated BHC report begins 1986Q3; UBPR analytical ratios are offered from 2002Q4 (the FFIEC CDR does not offer the first three quarters of 2002). FDIC Summary of Deposits (branch-level) begins 1994; FDIC bank-failure records reach back to 1934.

2013+ — Stress-testing and disclosure era. DFAST supervisory stress-test results are published annually from 2013. FR Y-15 G-SIB systemic indicators are available in bulk only from 2020 onward. Basel III Pillar 3 disclosures are bank-published PDFs concentrated in the most recent years (~2024+). NCUA 5300 credit-union call reports are used here from 2022Q1; HMDA institution-year summaries from the CFPB Data Browser begin 2018.

3. The three-corpus architecture (concept)¶

A continuous bank-level balance-sheet spine from the 1860s to the present cannot be built from any single public source. It is assembled by stitching three non-overlapping corpora, each kept only for the era where it is unique:

Corpus A — historical OCC. OCC Annual Report national-bank balance sheets, native 1863–1941. Kept for its unique pre-1942 national-bank coverage; nothing else reaches this era at bank level.
Corpus B — pre-digital Fed (NY Fed / Luck). The NY Fed public bank balance-sheet dataset, native 1959–present. Kept only for its unique 1959Q4–1975Q4 quarters (plus a handful of later gap-fill cells); for 1976+ it duplicates Corpus C.
Corpus C — Fed-direct. Chicago Fed (1976–2011) plus FFIEC CDR (2012+). Native 1976Q1–present, and the spine for everything from 1976 onward.

Why the split is necessary, empirically. The Fed-direct corpus contains zero rows before 1976, so the NY Fed pre-1976 core is irreplaceable. Conversely, for the post-1976 overlap the NY Fed dataset is almost entirely redundant: the overwhelming majority of its 1976+ entity-quarter cells are already present in the Fed-direct record, and the few that are not cluster in the early 2000s and are predominantly non-deposit trust companies — most of which can be recovered directly from FFIEC CDR. On the variables that can be mapped across corpora for 1976+, core aggregates (assets, deposits, equity, liabilities, loans) reconcile to roughly 99.9% between the NY Fed dataset and the Fed-direct record; the handful of variables that reconcile less well (e.g. securities, fed funds purchased, time deposits) differ for definitional/recipe reasons rather than data loss — the underlying line items remain present in the Call Report record. The practical implication for a researcher: take 1976+ from the Fed-direct corpus, take 1959–1975 from the NY Fed dataset, take pre-1942 from the OCC corpus, and treat the small residual NY Fed-only post-1976 cells as targeted gap-fill rather than a parallel source.

4. When major schedules / reports were introduced¶

Even within a continuously-covered collection, individual schedules and line items appear at different dates. Some milestones a long-run researcher should expect:

FR Y-9C consolidated BHC reporting begins 1986Q3.
Call Report form FFIEC 051 (the streamlined report for smaller, non-complex banks) was introduced in 2017; before that, banks filed FFIEC 031 (with foreign offices) or FFIEC 041 (domestic only).
Basel III regulatory-capital schedule (Call Report Schedule RC-R) was substantially rebuilt for the Basel III capital framework phased in from 2014, changing the capital-ratio line items and adding common-equity-tier-1 detail.
Expanded trading / market-risk and related detail in the supervisory reports continued to be added through the mid-2010s; researchers should expect broader trading-book and derivatives detail from roughly 2018 onward relative to earlier filings.
UBPR ratios are offered from 2002Q4; FR Y-15 systemic-indicator bulk data only from 2020.
CECL retitling. Under the current-expected-credit-loss (CECL) accounting standard, the former "Allowance for Loan and Lease Losses" line items were retitled to "Allowance for Credit Losses" across the FR Y-11 / Y-7N and Call Report families.

Because line items are added, retired, and renumbered over time, any item-level series should be checked against the variable dictionary's first-observed / last-observed dates before being treated as continuous.

5. Known public-data gaps and limits¶

1942–1958 quarterly bank financials are not available publicly at bank-quarter granularity; only annual-granularity academic panels and physical archives cover the window.
FR Y-9LP / Y-9SP parent-only BHC reports have no public bulk product; they are available only as per-institution filings, and the FFIEC NPW interface that serves them has at times blocked automated access.
Pre-2018 HMDA is not available through the modern CFPB Data Browser (2018+ only); legacy HMDA is keyed by respondent ID rather than LEI, complicating linkage to other bank identifiers.
FR Y-15 before 2020 has no bulk product (only individual NIC snapshots from 2020 on).
Basel III Pillar 3 coverage is thin and bank-published: only the largest institutions (G-SIBs) disclose, and only for recent periods.
Identifier coverage floor. Linking historical and modern records depends on the Federal Reserve's RSSD identifier system and the FFIEC NIC entity directory, which do not reach institutions that ceased to exist before the NIC era; a structural residual of historical entities cannot be matched to a modern RSSD.
Confidential reports out of public scope. Several supervisory collections (e.g. FR Y-14, FR 2052a) are confidential and are not part of any public data product.

6. Provenance tiers¶

A simple way to weight sources by directness:

T1 — primary regulator self-serve bulk: the regulator publishes the data directly for download (e.g. Federal Reserve MDRM, FFIEC CDR bulk, FRED, SEC EDGAR data API).
T2 — regulator API / portal: programmatic access to a regulator's own records (e.g. FDIC BankFind/SDI, FDIC Summary of Deposits, NCUA 5300, CFPB HMDA).
T3 — Fed archive: a Federal Reserve archive hosting older records (e.g. Chicago Fed Call Report archive).
T3.5 — public academic-Fed dataset: a Federal Reserve Bank's public research dataset (e.g. the NY Fed balance-sheet / income-statement dataset).
T4 — academic digitization: scholarly digitizations of regulator records (e.g. the OCC Annual Report digitizations and historical failing-bank panels).
derived: values computed from one or more of the above rather than reported directly.

7. Sources¶

For an annotated directory of these suppliers — what each provides, what it covers, what it is best used for, and any citation requirements — see Data Sources.

Federal Reserve MDRM (variable dictionary): https://www.federalreserve.gov/apps/mdrm/
FFIEC Central Data Repository (Call Report / UBPR bulk): https://cdr.ffiec.gov/public/pws/downloadbulkdata.aspx
FFIEC NIC / NPW (entity attributes, FR Y-15): https://www.ffiec.gov/NPW/
Federal Reserve Bank of Chicago Call Report archive: https://www.chicagofed.org/banking/financial-institution-reports/commercial-bank-data
Federal Reserve Bank of New York — Balance Sheets & Income Statements dataset: https://www.newyorkfed.org/research/banking_research/balance-sheets-income-statements
OCC Annual Reports: https://www.occ.treas.gov/publications-and-resources/publications/annual-report/index-annual-report.html
FDIC BankFind / SDI: https://banks.data.fdic.gov/ · https://banks.data.fdic.gov/api/
FDIC Summary of Deposits: https://www.fdic.gov/bank-financial-reports/summary-deposits
Federal Reserve DFAST / stress tests: https://www.federalreserve.gov/supervisionreg/dfa-stress-tests.htm
FRED (St. Louis Fed): https://fred.stlouisfed.org/
SEC EDGAR: https://data.sec.gov/
CFPB HMDA: https://ffiec.cfpb.gov/
NCUA 5300 call-report data: https://ncua.gov/analysis/credit-union-corporate-call-report-data

Academic digitizations of regulator records. The OCC historical national-bank series and the pre-digital Call Report dataset are associated with Correia, Luck & Verner, "Failing Banks" (QJE, DOI 10.1093/qje/qjaf044), and the OCC digitization is archived at Harvard Dataverse (DOI 10.7910/DVN/Q22XR1).