QuarterCall Data

Call Report data for research

QuarterCall Data republishes every NCUA Form 5300 Call Report from Q1 2019 as Parquet files at data.quartercall.com. Reading them needs no account or key, and each amendment is resolved to the filing that stands, so an analysis can start from the filings themselves.

What the files hold

A quarter’s filings are in three files: accounts holds every account code each credit union filed, accounts_text the values that are not numbers, and profiles NCUA’s Credit Union Profile, one row per credit union. The submissions index lists every filing collected, amendments and the filings they replaced included. The documentation describes each table column by column, and the gotchas set out where NCUA’s data differs from what a reader would expect.

Read it in Python

The quartercall package, on PyPI, reads a quarter into pandas.

pip install quartercall
Shell
import quartercall as qc

qc.quarters()                                 # every published quarter, newest first
assets = qc.accounts("2026Q2", codes="A010")  # total assets, one row per credit union
Python

The package’s readme lists every reader and its arguments.

Read it in DuckDB

DuckDB reads the same files over HTTP, with nothing to download first.

SELECT cu_number, account_value AS total_assets
FROM read_parquet('https://data.quartercall.com/curated/accounts/quarter=2026Q1/data.parquet')
WHERE account_code = 'A010'
ORDER BY total_assets DESC
LIMIT 10;
examples/quickstart/first-query.sql

The quickstart explains what the query returns.

Citing the data, and keeping a result reproducible

A quarter’s files are overwritten in place whenever they are rebuilt: when a filing arrives, when the collector changes, and once a week. A file’s URL therefore names what the file holds when it is read, not what an analysis read.

The manifest records each file’s SHA-256 and size under its quarter, as accounts_sha256 and accounts_bytes, and likewise for the other files; its guide describes the rest of it. Keep, with a result, the URL of each file it read, the date it was read and its SHA-256: a file whose SHA-256 still matches holds the same bytes. The weekly rebuild gives the accounts files a new SHA-256 even when no filing changed, so keep a copy of any file a published result depends on.

A citation names NCUA as the source, then the file:

NCUA Form 5300 Call Report filings, [quarter], [table] file, republished by QuarterCall Data. [file URL], read [date]. SHA-256 [hash].

QuarterCall has not yet published terms of use for the data. When it does, this page will say how the data may be shared.

Series across quarters

NCUA’s account codes change between cycles, so a series that spans years needs a definition at each change. callseries builds those series, each with a published definition and a validation result. Its catalogue of definitions is free, at Measures and Ratios, and its releases are licensed.