FEMA NFIP Redacted Policies v2 (OpenFEMA)

It is the policy-side ledger of the National Flood Insurance Program - every transaction that wrote, renewed, endorsed or cancelled flood cover, 73,601,802 of them, one row each. Where the claims file records what happened after water arrived, this one documents the book that stood ready before it: effective and termination dates, rated and current flood zone, occupancy type, floors, original construction date and post-FIRM status.

The money is unusually well decomposed for a government file. Each row splits building versus contents limits, then separates base premium from policy cost, federal policy fee, HFIAA surcharge, reserve fund assessment and ICC premium - so a premium is an anatomy rather than a single number. Rating context completes it: deductibles, rate method, subsidized rate type, grandfathering code, Community Rating System class and the mandatory purchase flag.

Within Datadory's shelf this record owns the in-force exposure layer: who held what, where, at what price.

One row per policy transaction, typed identically whether the property is a barrier-island cottage or a downtown condominium. The block above shows a real one: a Sunset Beach, North Carolina policy rated in zone AE, effective 10 July 2016 for a one-year term, carrying $250,000 of building cover and $100,000 of contents.

The money columns reconcile on the row - $407 premium plus the $50 federal policy fee, $25 HFIAA surcharge and $59 reserve fund assessment land exactly on the $541 policy cost, which makes audit work straightforward. Context columns finish the picture: Community Rating System class 7 for the community, mandatory purchase flagged true, post-FIRM construction true, FEMA region 4.

Samples populate the same columns across whichever states, communities or flood zones you name - nothing is synthesized to fill gaps.

Eighty-two verified fields, organized into six families:

• Lifecycle - effective, termination, endorsement, cancellation and property purchase dates
• Risk and rating - rated and current flood zone, occupancy type, building description, post-FIRM and elevated-building indicators, rate method and grandfathering codes
• Premium economics - building and contents limits, premium, policy cost, federal policy fee, HFIAA surcharge, reserve fund assessment, ICC premium, deductible codes, replacement cost
• Elevation - base flood elevation, lowest floor elevation, lowest adjacent grade, elevation difference
• Program and community - Community Rating System class, mandatory purchase flag, community numbers and name, Regular or Emergency program indicator
• Geography - state, ZIP, county FIPS, census tract, census block group, rounded latitude and longitude

Thirty-one anchors are tabulated below; the remaining fifty-one ship in the same dictionary format with your sample. Redaction is structural and disclosed: recent vintages withhold reported city, and coordinates round so a row narrows to a neighborhood, never a doorstep.

Geography - the United States and its territories wherever NFIP policies are written. The join ladder descends from state through ZIP and county FIPS to census tract and block group, with rounded latitude and longitude on the bottom rung - fine enough to aggregate to a coastline, coarse enough to honor the program's privacy rules.

Temporal - the catalog window opens at 2009-01-01 for policy effective dates, and the v2 vintage holds at its June 2026 freeze while the program's successor version succeeds it. Original construction dates on the structures themselves reach back decades further, so building-age analytics survive the shorter transaction window.

Granularity - one row per policy transaction across 73,601,802 records. Counting in-force policies rather than transactions wants a deduplication pass on community and property keys, which the delivery applies.

Set against the wider catalog - average quality score 7.81 across all 1,744 datasets - this record scores 9/10, carried by field depth, verified definitions and national scope.

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

Your cadence runs independently of the program's own rhythm. Take a one-time snapshot of the national flood book, or keep a warehouse table current so renewals and endorsements diff cleanly into yesterday's rows. Deliveries arrive normalized to the field dictionary above with community identifiers intact and keyed to census geographies your warehouse already speaks, so a join against your own property or mortgage book needs no fuzzy matching.

Two postures work well in practice. Take the policy slice alone as the in-force exposure layer - limits, premiums and zones by geography. Or take it joined to the claims file, aligned on shared state, ZIP, county and community keys, to produce take-up rates and loss ratios in one table. Most production users end up wanting both, in that order.

• Take-up-rate mapping - policy counts per ZIP, tract or county set against the claims file's loss counts put a penetration number on every floodplain.
• Premium adequacy analysis - premium decomposed against coverage limits by flood zone shows where the rated price and the carried risk diverge.
• Portfolio exposure screening - mandatory purchase flags, rated zones and elevation differences screen a mortgage or property book at block-group grain.
• CRS community benchmarking - Community Rating System classes rank which communities buy mitigation into their rates.
• Actuarial and reinsurance exposure bases - an in-force view of the national flood book feeds accumulation and catastrophe-loading work.
• Climate resilience research - post-FIRM indicators, construction dates and elevation readings trace how the built coast changed.

Investors and quants lead fit: an observable, transaction-level national flood book is rare ground for insurance-market analysis; see investors & quants use cases. Data scientists and ML engineers get eighty-two typed columns with verified definitions - feature-store material for flood-risk models; see data scientists use cases. Market researchers and consultants size the flood insurance market and benchmark carrier books; see market researchers use cases. Competitive intelligence teams track which communities and zones grow or shrink their insured base; see competitive intel use cases. Developers and builders stand exposure monitoring on the fixed schema; see developers builders use cases. Journalists, academics and students draw citable federal-program figures; see the citation-grade research workflow.

This record owns the in-force book. FEMA NFIP Redacted Claims v2 owns the loss experience - millions of claim transactions with damage and payment detail; the two join on shared geography and community keys, and the head-to-head lives at claims vs policies. First Street Foundation Climate Risk Data models forward-looking hazard per property, where this file records priced, purchased cover as it actually stood. NAIC Research and Insurance Data aggregates insurer filings market-wide but not property by property. FEMA NFIP Geospatial Hub supplies the mapped hazard layers that contextualize both. Observed book versus modeled peril is the axis the whole industry runs on.

Four things worth settling upfront.

First, transactions are not policies. The 73.6 million figure counts transactions; a single property renewing ten times contributes ten rows. In-force counts want the deduplication pass described above, and samples ship with it applied.

Second, redaction shapes every spatial join. City names withdrew from recent vintages and coordinates round deliberately, so the census block group is the finest geography that behaves consistently across vintages. Plan aggregations there.

Third, the version clock is real. This v2 vintage froze at June 2026 and leaves service after 15 October 2026 in favor of the program's successor version; deliveries transition onto the new version with schema differences documented rather than discovered.

Fourth, codes arrive decoded. Occupancy types, deductible tiers, rate methods and grandfathering categories travel as integers and letters in the raw artifact; every delivery carries the decode references beside them, and the sample confirms the mappings against your scope.

Because the raw artifact is 73.6 million rows of coded columns with a version sunset attached - formidable to wrangle, hostile to join casually. Datadory normalizes the slice to the field dictionary above, decodes the categorical codes, keys every row to census geographies your warehouse already carries, and schedules the feed so consecutive pulls diff cleanly through the version transition.

Browse the rest of the industry on the property casualty insurance data hub, the best property-casualty-insurance datasets ranking, or the full catalog; the source itself is profiled at the FEMA OpenFEMA source page.