Insurance company statutory financial data: the filings behind the benchmarks, delivered

The numbers solvency oversight runs on. Every insurer admitted in a US state files an annual financial statement with its home-state regulator on the statutory accounting basis - with quarterly filings alongside - and those statements are the substrate every other American insurance number sits on. More than 4,800 US insurance companies file them, and the National Association of Insurance Commissioners, the association of the fifty state regulators, is the body whose members receive each one.

Datadory catalogs that stack as two regulator-native records among the 10 pooled multi-line-insurance datasets, out of the 1,744 cataloged site-wide. NAIC Resource Center - Regulatory Data & Publications scores 8 out of 10 and carries 164 publication files - the published surface of the filing system, from Market Share Reports to Profitability by Line by State. NAIC Insurance Data and Regulatory Tools scores 9 out of 10 and reaches beneath that surface: statutory statement pages per carrier, up to ten years of history across annual and quarterly periods. Both arrive through Datadory already typed and keyed, delivered daily, weekly, or hourly.

Three shapes cover the whole subject, and they all hang off the same identifiers:

# DIRECTORY ROW - PROP.csv, edition LOC-JUN-2026     grain: one insurer per row
COMPANY NAME       : 1842 INS CO            STATE OF DOMICILE : MD
GROUP CODE         : 447                    COMPANY CODE      : 16862
FEIN NUMBER        : 84-4224428             COMPANY STATUS    : 1
CONTACT BLOCK      : BEL AIR, MD 21014-3544 + phone

# STATUTORY STATEMENT PAGE                           grain: one carrier x statement period
key                : NAIC Company Code | Company Group | Statement Period
depth              : up to 10 years of filings, annual and quarterly

# DERIVED SCREENS                                    grain: one carrier x measure x period
IRIS               : twelve standard financial-ratio screens, with means and medians
PROFITABILITY      : premiums, losses, expenses, estimated profit ratios - per line per state

One layer observes individual carriers at line-item depth, one grades them, one ranks them. Teams that pick the wrong layer usually do not notice until the deliverable is due - a valuation built on aggregate ratios, or a leaderboard rebuilt by hand from statement pages.

The test is simple. If the deliverable explains why one carrier's combined ratio moved, only statement-page detail answers - aggregate ratios summarize away exactly the schedule and exhibit lines a failure analysis needs. If the deliverable asks whether the carrier is normal, the IRIS screens answer, because they are the same early-warning arithmetic regulators apply to the filings. If it asks who leads, the Market Share Reports settle it outright on the 2024 data year. The table below routes the common cases.

Geography - the complete US regulatory map: fifty states, the District of Columbia and five territories, 56 supervisory jurisdictions in all. Concentration lands at the jurisdiction grain because that is where insurance competition is actually decided. The offshore extension is the quiet advantage: roughly 17,000 alien insurers and reinsurers ride in the same sixteen-table drop as the 5,000-odd domestic carriers, an enumeration no domestic register offers.

Temporal - each family keeps its own clock. Rankings stand at the 2024 data year; the carrier directory issues named December and June editions (LOC-DEC-2025, LOC-JUN-2026); IRIS documentation covers 2024/2025; and the statement layer reaches up to ten years back on both annual and quarterly cycles. Date-stamping the edition is part of working with the data, not bookkeeping after it.

Granularity - three axes at once: one row per company in the directory, one cell per group per state per line in the rankings, one observation per line per state in the profitability tables, and statement-page depth beneath everything. Pick the axis your question lives on; the same delivery carries all of them. When the question turns macro, World Bank World Development Indicators - Insurance & Financial Sector supplies the country-year lens described below.

Done by hand, a defensible single-carrier screen is a weekend of report archaeology; the assembly happens upstream here, and the file arrives already keyed:

1. Resolve the entity. Pull the target's group code, company code, FEIN and domicile from the directory - one row settles who you are actually looking at, including which holding group stands behind the brand.
2. Place it in the market. Match its group to the Market Share Report for each line it writes - top 10 by state, top 125 countrywide - to see where premium volume really concentrates.
3. Run the early-warning screens. Read the twelve IRIS ratios beside Profitability by Line by State to see whether weakness clusters in particular lines or jurisdictions.
4. Check the asset side. NAIC's Capital Markets Bureau has tracked insurer invested assets monthly since 2016 - $9.6 trillion at year-end 2025 - so asset concentrations read directly against liability-side results.
5. Escalate deliberately. If ratios and aggregates clear the carrier, stop. If they do not, that is the moment to drop to statement-page detail for the specific schedules that moved.

Name the carriers, states and filing years when you request a sample and the extract arrives cut to that scope, shaped identically to the standing feed - so anything prototyped on the sample survives into production unchanged.

Not at statement level. Per-carrier statutory filings are an artifact of US state regulation, so cross-country work starts one level up, from aggregates. The closest systematic equivalent is the World Bank's WDI cut: among 1,498 indicators, 12 carry insurance in their names - balance-of-payments trade shares such as BX.GSR.INSF.ZS, the per_si_allsi.* social-insurance family, and the lending risk premium - measured annually for 217 economies back to 1960, with the insurance trade series verified current through 2025. One honest scope note travels with it: classic premium-volume and penetration ratios were retired from WDI into other World Bank collections, so confirm which series hosts a figure before building a panel on it. The head-to-head between the two lenses is written up at World Bank WDI vs NAIC Resource Center.

For European supervisory holdings specifically, EU Open Data Portal Advanced Search indexes the continent's harvested public-sector metadata - 2,979 of its 1,927,787 records matched 'insurance' at the August 2026 verification pass, each carrying publisher, coverage and distribution detail in up to 24 languages. It is a discovery layer rather than a ledger: precise pointing, then national sources underneath.

Ranked by how directly the filings answer the day job:

1. Solvency and credit screeners run the twelve IRIS screens and state-line profitability against watchlists of carriers, escalating to statement pages only where a screen trips - see credit risk screening.
2. Investors and quant researchers read insurer strength off regulatory ratios and the asset-allocation record instead of press releases - the sell side of insurance measured by its referee. Sketched further at investors and quants in property & casualty insurance.
3. Market researchers and consultants take the 'us insurers by state count' question straight off the directory census - roughly 5,000 admitted carriers across all 56 jurisdictions - and hang sizing on it. The workflow map lives at market researchers in multi-line insurance.
4. Data scientists and modelers train entity resolution and classification on a labeled universe of authorized carriers keyed by stable identifiers; patterns collected at data scientists in multi-line insurance.
5. Journalists, academics and students cite the regulator's own compilation when a carrier's condition becomes a story, rather than someone's chart of it. For loss-trend narrative beside the numbers, Triple-I Research & Data streams weekly statistics across auto, homeowners, life, business, cyber and workers-compensation lines.

Five notes that save rework.

1. Decide company versus group first. A group's footprint spans many filing subsidiaries, and the choice changes every downstream number - including leaderboard order itself.
2. Statutory is not GAAP. These figures are filed on insurance accounting conventions; reconciling them to the GAAP revenue in a public filer's earnings release produces nonsense. Treat them as regulatory truth, not a second copy of the 10-K.
3. Snapshot dates differ by family. Rankings stand at 2024 while the directory turns on December and June editions. Decide which date governs any join between them before you build on it.
4. Part of the shelf is definitions, not observations. Statement blanks and the accounting practices manual tell filers how numbers must be built; only the rankings, screens and profitability tables observe the industry itself. Name which half your question lives in.
5. Domestic and offshore records are not symmetric. The roughly 17,000 offshore entities are an enumeration with identifiers; their contact detail runs thinner than the domestic carriers' - treat that layer as a register, not a CRM extract.

API, files, or your warehouse. Daily, weekly, or hourly.

Because the quirks live in the typing, and typing happens upstream of you here. Snapshot dates differing by family, identifier grammars that survive renames and redomiciliations, definitions sitting beside observations on the same shelf - delivered, each is a settled fact: typed tables keyed on company code and jurisdiction, field dictionaries traveling with the rows, every family stamped with its own edition date.

Start with a sample cut to your named carriers, states and filing years - it arrives shaped identically to the standing feed, so nothing needs re-plumbing between prototype and production. Field-level detail sits on the NAIC resource center dataset page and the NAIC insurance data record; get a sample to see your cut before anything else gets decided.

The multi-line insurance data guide maps all 10 pooled datasets and shows how the regulatory, benchmark and macro layers fit together; the multi-line insurance hub browses the same pool as records, and best multi-line insurance datasets scores them side by side.