Field names — MDRM + UBPR catalog

A machine-readable catalog of every non-confidential field the FFIEC collects across federal bank regulatory reports, together with the derivation trees for UBPR ratios built on top of those fields. Downloadable today as newline-delimited JSON; library integration is planned.

At a glance

Item	Value
Current version	`2026-04-21` (published 2026-04-21)
Records	29,908 non-confidential fields
Refresh cadence	Periodic — refreshed against each new quarterly MDRM release, plus ad-hoc reprocessing when the normalization pipeline improves
License	Mozilla Public License 2.0
Publisher	Call.Report catalog built from FFIEC primary sources

The -latest download URL always points at the newest version; historical snapshots may be published at versioned paths in the future.

Downloads

File	Size	Notes
`field-names-latest.jsonl`	30.2 MB	Newline-delimited JSON; one record per field.
`field-names-latest.jsonl.gz`	4.0 MB	Same content, pre-gzipped (`Content-Encoding: gzip` — your HTTP client will decompress transparently).

One record per line. Read with jq, pandas.read_json(lines=True), polars.read_ndjson, jsonlines, or any streaming JSON parser.

Example record

{
  "mdrm_series": "RCON",
  "mdrm_series_name": "Report of Condition — domestic offices",
  "mdrm_item": "2170",
  "concat": "RCON2170",
  "is_confidential": false,
  "original_title": "TOTAL ASSETS",
  "long_normalized_title": "Total Assets",
  "original_description": "The sum of all asset items. ..."
}

Fields

Field	Type	Notes
`mdrm_series`	`str`	4-character MDRM series (e.g. `RCON`, `RCFD`, `UBPR`).
`mdrm_series_name`	`str`	Human-readable name of the series / report population.
`mdrm_item`	`str`	4-character item code within the series (e.g. `2170`).
`concat`	`str`	`mdrm_series + mdrm_item` — the identifier used throughout the library (e.g. `RCON2170`).
`is_confidential`	`bool`	Always `false` in this public catalog.
`original_title`	`str`	FFIEC’s terse, all-caps title (e.g. `TOTAL ASSETS`).
`long_normalized_title`	`str`	Readable English rewrite (e.g. `Total Assets`).
`original_description`	`str`	Full MDRM definition, including historical revisions and cross-series notes.

Methodology

Sources. The catalog is built from two FFIEC primary sources: the Micro Data Reference Manual (MDRM) — the authoritative index of every data field collected across federal bank regulatory reports — and the Uniform Bank Performance Report (UBPR) definitions, which specify how analytical ratios are computed from those fields. Together they cover 87,687 field definitions spanning commercial banks, bank holding companies, credit unions, savings associations, foreign banking organizations, and several auxiliary report populations.

Structure. Every field is connected at three levels:

Source provenance — the regulatory context a field lives in: which form it appears on, which institution type reports it, its scope (domestic offices, consolidated, parent-only), and its reporting cadence. This lets analysts cross-reference the same economic measure as reported by different institution types.
Calculation graph — UBPR ratios and derived measures are parsed into full expression trees with typed argument roles (numerator, denominator, addend). The composition of any derived value is traceable to its underlying fields, at arbitrary depth.
Business concept taxonomy — a hierarchy of approximately 49,000 concepts derived from the structural patterns in FFIEC Item Names, linking each field to what it measures independent of who reports it.

All three structures anchor on a stable (mnemonic, item code) identifier — the same identifier the regulator uses. Historical revisions are preserved, with an explicit pointer identifying the current-truth record at any point in time.

Title normalization. FFIEC field titles are terse, all-caps, and abbreviation-heavy (e.g., NET INT INC-TE TO AVG EARN ASSET). We convert them into readable English — Net Interest Income (Tax-Equivalent) to Average Earning Assets — through a three-stage pipeline:

A deterministic ruleset applies case conventions, punctuation repair, and a curated glossary of regulatory abbreviations.
A large-language-model review pass inspects every distinct title and corrects where the ruleset falls short, constrained to preserve content faithfully — expansion only, never summarization.
A consistency post-pass unifies small variations between reviewers so identical phrases normalize uniformly across the dataset.

Verification. Every record is audited against automated checks: no source content is dropped, identical phrases normalize identically, acronyms retain case fidelity, and regulatory citations (schedule references like 1.C.(2)(a)) remain literal. Confidential fields are filtered prior to publication; the public catalog contains 29,908 non-confidential fields. The pipeline is idempotent — refreshing against a new quarterly MDRM release updates only changed records.

License

Released under the Mozilla Public License 2.0 — the same license as ffiec-data-connect. You can use it commercially, modify it, redistribute it, and build products on top of it. Derivative works that modify the catalog records themselves must carry the MPL-2.0 license forward for those modifications; applications that merely use the catalog do not.

The underlying FFIEC MDRM and UBPR source material is a US government work and not subject to copyright in the United States. The catalog in this form — normalized titles, computed derivation trees, taxonomy linkages — is a derivative authored by Call.Report and is what the MPL-2.0 covers.

Library integration

ffiec-data-connect does not yet consume this catalog at runtime. Planned integration: a lookup helper that resolves an mdrm code returned by collect_data() to its normalized title and description without a separate fetch. Until then, this page is the download surface.