Appendix — Data Card — FDIC (Institution Directory, SDI financials, and bank failures)

Provider & what it is

The Federal Deposit Insurance Corporation (FDIC) insures deposits at U.S. banks and, as the price of that insurance, collects a quarterly financial report from every insured institution. It publishes three things you will use, all through one free API called BankFind Suite:

Institution Directory — the roster of every FDIC-insured bank and thrift: name, charter type, regulator, headquarters location, active/inactive status, and the all-important CERT number that identifies the institution. Think of it as the phone book of American banking.
SDI (Statistics on Depository Institutions) — the quarterly financials: total assets, deposits, loans, net income, capital ratios, return on assets — the Call Report numbers, reshaped into clean fields. This is how you turn “is this bank healthy?” into a number.
Failures & assistance — the historical list of bank failures and FDIC-assisted transactions, with the failure date and the acquiring institution. The 2008–2010 wave and the 2023 regional-bank episode both live here.

The reveal: banks look opaque, but they are among the most heavily disclosed firms in the economy. Every quarter, every insured bank hands the regulator a standardized balance sheet and income statement, and the FDIC hands it to you for free. The “secret” financial health of a bank is a public download.

Coverage

Institution Directory: all FDIC-insured institutions, active and historical; inactive banks retain their record with a closing reason. Tens of thousands of CERTs across history; roughly 4,000–5,000 active banks today (the count has fallen steadily through consolidation). [CHECK] exact active count for your pinned vintage.
SDI financials: quarterly, with broad coverage from the early 1990s and selected series earlier. Reported each quarter on the Call Report cycle (filings due ~30 days after quarter-end, so the most recent quarter lags). [CHECK] earliest quarter per field.
Failures: complete failure list from 1934 onward; the modern, richly-detailed records cluster in the post-2000 period. The famous spikes are 2008–2012 and the March 2023 events (SVB, Signature, First Republic). [CHECK] First Republic resolution coding (acquired vs. failed).

Key identifiers

CERT — the FDIC Certificate number, the primary key for an institution across the Directory, SDI, and failures. Stable for the life of the charter. This is the join key for everything.
Also present: RSSD (the Federal Reserve’s institution id, useful to merge with Fed/Call Report data), the bank’s name and location (state, city, FIPS). Holding-company structure is partial here — a bank’s parent BHC lives more completely in the Fed’s NIC database, so do not assume CERT alone resolves the corporate parent. [CHECK] whether RSSD ships in all BankFind endpoints.

Access path

BankFind Suite API, base https://banks.data.fdic.gov/api/. JSON, no key required, public. Three endpoints map to the three datasets:
- /institutions — directory records (filter by STALP state, ACTIVE, etc.)
- /financials — SDI quarterly financials (filter by CERT and REPDTE report date)
- /failures — the failure list
Query parameters use a filters expression, a fields list (to limit columns), and limit/offset for paging. The API is generous but paginate for bulk pulls rather than requesting everything at once.

import requests, pandas as pd
BASE = "https://banks.data.fdic.gov/api"
# Active banks headquartered in Virginia (public, no key):
r = requests.get(f"{BASE}/institutions",
                 params={"filters": "STALP:VA AND ACTIVE:1",
                         "fields": "NAME,CERT,STALP,CITY,ASSET",
                         "limit": 50}, timeout=60)
banks = pd.json_normalize(r.json()["data"])   # each row nests under 'data'
# Quarterly financials for one bank by CERT:
r2 = requests.get(f"{BASE}/financials",
                  params={"filters": "CERT:3510", "fields": "CERT,REPDTE,ASSET,DEP,NETINC,ROA",
                          "sort_by": "REPDTE", "sort_order": "DESC", "limit": 8}, timeout=60)

No secrets needed — the API is fully open. If you wrap it in a script, still follow CONVENTIONS §5 and never hard-code anything that looks like a credential.

License & note

Free and public. FDIC data is U.S. government work, effectively public domain; cite the FDIC BankFind Suite and your pinned vintage and you are done. The only real obligation is honesty about which quarter you pulled, because SDI is restated — a bank’s reported assets for a past quarter can change in a later vintage as corrections flow in. Pin REPDTE and the access date.

Gotchas

Restatements. SDI values for a past quarter can change in a later snapshot. Pin the vintage or your “2019Q4 assets” will not reproduce.
Merged-away banks vanish from the active roster but live in the directory. When you study failures or consolidation, filter on ACTIVE deliberately — a survivorship-biased panel of only-currently-active banks will understate distress.
The bank is not the holding company. CERT identifies a bank charter; a single holding company can own several CERTs. Aggregating to the parent firm needs the RSSD/NIC crosswalk, not CERT alone.
“Failure” has a precise meaning. A bank acquired in a healthy merger is not a failure; only FDIC-resolved closures and assisted transactions appear in /failures. Do not conflate consolidation with distress.
Field-name soup. SDI uses terse Call Report codes (ASSET, DEP, NETINC, RBCT2 …). Keep the FDIC field dictionary open; [CHECK] any field meaning before you regress on it.

First 10 rows (illustrative schema sketch — not real values)

Institution Directory (/institutions), shape only:

CERT	NAME	STALP	CITY	ACTIVE	ASSET
3510	Example National Bank	VA	Fairfax	1	18432109
57890	Sample Community Bank	MD	Bethesda	1	742118
…	…	…	…	…	…

SDI financials (/financials), one row per CERT × quarter, shape only:

CERT	REPDTE	ASSET	DEP	NETINC	ROA
3510	2026-03-31	18432109	14201338	52114	1.12
3510	2025-12-31	18120044	13988201	49870	1.09
…	…	…	…	…	…

Failures (/failures), shape only:

CERT	NAME	FAILDATE	CITY	STALP	ACQINST
57053	Sample Bridge Bank	2023-03-10	Santa Clara	CA	Acquiring Bank, N.A.
…	…	…	…	…	…

Which chapter / lab / capstone uses it

Bank financials and the CERT key support worked examples on firm-level financial-health measures and panel construction — a clean, free alternative to licensed Compustat for banks specifically, useful when a student wants real balance-sheet data without a WRDS seat.
The failures file is the natural dataset for an event-study or survival exercise on bank distress (the 2008–2012 and 2023 waves), and a good demonstration of survivorship bias when contrasting active-only vs. full panels (ties to the inference discipline of Weeks 2–4).
It sits alongside the other public-finance sources (treasury-finra.md, msrb-emma.md) as a “real institutions, free data, messy keys” source students can use for a self-directed capstone extension when they want banking data outside the five canonical capstone datasets.

# Data Card — FDIC (Institution Directory, SDI financials, and bank failures) ## Provider & what it is The **Federal Deposit Insurance Corporation (FDIC)** insures deposits at U.S. banks and, as the price of that insurance, collects a quarterly financial report from every insured institution. It publishes three things you will use, all through one free API called **BankFind Suite**: - **Institution Directory** — the roster of every FDIC-insured bank and thrift: name, charter type, regulator, headquarters location, active/inactive status, and the all-important **CERT** number that identifies the institution. Think of it as the phone book of American banking. - **SDI (Statistics on Depository Institutions)** — the quarterly *financials*: total assets, deposits, loans, net income, capital ratios, return on assets — the **Call Report** numbers, reshaped into clean fields. This is how you turn "is this bank healthy?" into a number. - **Failures & assistance** — the historical list of bank *failures* and FDIC-assisted transactions, with the failure date and the acquiring institution. The 2008–2010 wave and the 2023 regional-bank episode both live here. The reveal: banks look opaque, but they are among the most heavily *disclosed* firms in the economy. Every quarter, every insured bank hands the regulator a standardized balance sheet and income statement, and the FDIC hands it to you for free. The "secret" financial health of a bank is a public download. ## Coverage - **Institution Directory:** all FDIC-insured institutions, active and historical; inactive banks retain their record with a closing reason. Tens of thousands of CERTs across history; roughly 4,000–5,000 active banks today (the count has fallen steadily through consolidation). `[CHECK]` exact active count for your pinned vintage. - **SDI financials:** quarterly, with broad coverage from the early 1990s and selected series earlier. Reported each quarter on the Call Report cycle (filings due ~30 days after quarter-end, so the most recent quarter lags). `[CHECK]` earliest quarter per field. - **Failures:** complete failure list from 1934 onward; the modern, richly-detailed records cluster in the post-2000 period. The famous spikes are 2008–2012 and the March 2023 events (SVB, Signature, First Republic). `[CHECK]` First Republic resolution coding (acquired vs. failed). ## Key identifiers - **CERT** — the FDIC Certificate number, the primary key for an institution across the Directory, SDI, and failures. Stable for the life of the charter. This is the join key for everything. - Also present: **RSSD** (the Federal Reserve's institution id, useful to merge with Fed/Call Report data), the bank's **name** and **location** (state, city, FIPS). Holding-company structure is partial here — a bank's parent BHC lives more completely in the Fed's NIC database, so do not assume CERT alone resolves the corporate parent. `[CHECK]` whether RSSD ships in all BankFind endpoints. ## Access path - **BankFind Suite API**, base `https://banks.data.fdic.gov/api/`. JSON, **no key required**, public. Three endpoints map to the three datasets: - `/institutions` — directory records (filter by `STALP` state, `ACTIVE`, etc.) - `/financials` — SDI quarterly financials (filter by `CERT` and `REPDTE` report date) - `/failures` — the failure list - Query parameters use a `filters` expression, a `fields` list (to limit columns), and `limit`/`offset` for paging. The API is generous but paginate for bulk pulls rather than requesting everything at once. ```python import requests, pandas as pd BASE = "https://banks.data.fdic.gov/api" # Active banks headquartered in Virginia (public, no key): r = requests.get(f"{BASE}/institutions", params={"filters": "STALP:VA AND ACTIVE:1", "fields": "NAME,CERT,STALP,CITY,ASSET", "limit": 50}, timeout=60) banks = pd.json_normalize(r.json()["data"]) # each row nests under 'data' # Quarterly financials for one bank by CERT: r2 = requests.get(f"{BASE}/financials", params={"filters": "CERT:3510", "fields": "CERT,REPDTE,ASSET,DEP,NETINC,ROA", "sort_by": "REPDTE", "sort_order": "DESC", "limit": 8}, timeout=60) ``` No secrets needed — the API is fully open. If you wrap it in a script, still follow CONVENTIONS §5 and never hard-code anything that *looks* like a credential. ## License & note **Free and public.** FDIC data is U.S. government work, effectively public domain; cite the FDIC BankFind Suite and your pinned vintage and you are done. The only real obligation is honesty about *which quarter* you pulled, because SDI is restated — a bank's reported assets for a past quarter can change in a later vintage as corrections flow in. Pin `REPDTE` and the access date. ## Gotchas - **Restatements.** SDI values for a past quarter can change in a later snapshot. Pin the vintage or your "2019Q4 assets" will not reproduce. - **Merged-away banks vanish from the active roster but live in the directory.** When you study failures or consolidation, filter on `ACTIVE` deliberately — a survivorship-biased panel of only-currently-active banks will understate distress. - **The bank is not the holding company.** CERT identifies a *bank charter*; a single holding company can own several CERTs. Aggregating to the parent firm needs the RSSD/NIC crosswalk, not CERT alone. - **"Failure" has a precise meaning.** A bank acquired in a healthy merger is *not* a failure; only FDIC-resolved closures and assisted transactions appear in `/failures`. Do not conflate consolidation with distress. - **Field-name soup.** SDI uses terse Call Report codes (`ASSET`, `DEP`, `NETINC`, `RBCT2` …). Keep the FDIC field dictionary open; `[CHECK]` any field meaning before you regress on it. ## First 10 rows (illustrative schema sketch — not real values) *Institution Directory (`/institutions`), shape only:* | CERT | NAME | STALP | CITY | ACTIVE | ASSET | |---|---|---|---|---|---| | 3510 | Example National Bank | VA | Fairfax | 1 | 18432109 | | 57890 | Sample Community Bank | MD | Bethesda | 1 | 742118 | | … | … | … | … | … | … | *SDI financials (`/financials`), one row per CERT × quarter, shape only:* | CERT | REPDTE | ASSET | DEP | NETINC | ROA | |---|---|---|---|---|---| | 3510 | 2026-03-31 | 18432109 | 14201338 | 52114 | 1.12 | | 3510 | 2025-12-31 | 18120044 | 13988201 | 49870 | 1.09 | | … | … | … | … | … | … | *Failures (`/failures`), shape only:* | CERT | NAME | FAILDATE | CITY | STALP | ACQINST | |---|---|---|---|---|---| | 57053 | Sample Bridge Bank | 2023-03-10 | Santa Clara | CA | Acquiring Bank, N.A. | | … | … | … | … | … | … | ## Which chapter / lab / capstone uses it - **Bank financials and the CERT key** support worked examples on firm-level financial-health measures and panel construction — a clean, free alternative to licensed Compustat for *banks specifically*, useful when a student wants real balance-sheet data without a WRDS seat. - **The failures file** is the natural dataset for an event-study or survival exercise on bank distress (the 2008–2012 and 2023 waves), and a good demonstration of survivorship bias when contrasting active-only vs. full panels (ties to the inference discipline of Weeks 2–4). - It sits alongside the other public-finance sources (`treasury-finra.md`, `msrb-emma.md`) as a "real institutions, free data, messy keys" source students can use for a self-directed capstone extension when they want banking data outside the five canonical capstone datasets.