How Far Back Does National Flood Insurance Program Data Go?

Two answers, depending on which side of the ledger you mean. On the losses side, FEMA NFIP Redacted Claims v2 (OpenFEMA) reaches back to loss dates of August 31, 1970 and holds 2,721,780 redacted claim records, each carrying the date of loss, rated flood zone, occupancy type, building and contents damage amounts, deductibles, payments, coverage carried and geography down to census block group.

On the exposure side, FEMA NFIP Redacted Policies v2 (OpenFEMA) holds 73,601,802 policy transactions whose catalog temporal coverage begins January 1, 2009, including effective and termination dates, premiums, fees, surcharges, reserve assessments, deductible codes and rated versus current flood zone. The two files are refreshed monthly and were both frozen at their June 1, 2026 vintage during the v2 deprecation window.

That asymmetry matters when you plan a study window. Fifty-plus years of claim events pair with roughly seventeen years of policy detail, so any premium-versus-loss ratio you compute before 2009 uses exposure proxies (coverage amounts carried on the claims themselves, or counts of claims) rather than true policy-in-force denominators.

A third FEMA product carries no transaction history at all yet still anchors time-series work: the FEMA NFIP Geospatial Hub (ArcGIS) serves the National Flood Hazard Layer as a Map Service with 33 layers covering over 90 percent of the US population. Flood Hazard Zones carry FLD_ZONE codes, SFHA flags, static base flood elevations and depths; FIRM Panels carry panel IDs and effective dates; LOMRs and LOMAs, levees, coastal transects and high water marks round out the set. It describes current effective hazard data with new and revised content added continuously — a snapshot discipline, not an archive.

Ranked by how far back each runs:

Only the first three give you event-level loss history; the rest are price indices, regulatory aggregates and market structure tables.

The long tail is the point: 2,721,780 claims reaching back to 1970 let you cut severity and frequency by decade, by cause-of-damage code, by rated flood zone and by occupancy type, then compare against today's book. Each claim record also carries elevation fields (base flood elevation, lowest floor elevation, lowest adjacent grade, elevation difference), water depth, floodwater duration and named flood events, so vintage effects like pre-FIRM construction can be isolated directly from the file.

Dataset pages worth bookmarking: FEMA NFIP Redacted Claims v2 (OpenFEMA), FEMA NFIP Redacted Policies v2 (OpenFEMA), NOAA NCEI Billion-Dollar Weather and Climate Disasters and FEMA NFIP Geospatial Hub (ArcGIS). Analysts assembling these sources into production workflows will find role-specific guidance on the data scientists use cases page for property & casualty insurance.