The assignment
After every major storm, the same letters arrived at the same addresses: polite, final, and wrong. Homeowners across Louisiana’s parishes were being told their flooded houses were worth a fraction of their neighbor’s identical claim. The assignment was to find out whether the pattern was anecdote or architecture — and the only way to answer was to get the data the national flood agency had spent two years refusing to release.
The resulting 8,400-word investigation showed it was architecture: adjuster firms paid by volume, not accuracy, had built an incentive structure in which underpayment was the rational outcome — applied to 23,000 claims across three storm seasons.
Approach & sourcing
The analysis was built claim by claim. Each of the 23,000 records was geocoded, matched to flood-depth mapping, and compared against contractor estimates for equivalent repairs. A consulting statistician reviewed the methodology blind; the newsroom’s data team reproduced every finding independently before publication.
- 23,000 claim records obtained under FOIA after a two-year appeal.
- Methodology reviewed blind by an outside statistician and reproduced in-house.
- Forty-one homeowners interviewed; adjuster firms and the national flood agency given detailed findings twice.
What the file contained
The investigation published with a searchable database: readers could look up their own parish, their own storm, and their own adjuster firm. The full methodology — matching rules, exclusion criteria, confidence intervals — ran alongside the narrative, and the underlying dataset was released for other newsrooms under a research license.
Results & impact
The national flood agency revised its adjuster compensation rules within five months, shifting part of payment from volume to review outcomes. A Senate subcommittee cited the database in hearings on program reform, and two adjuster firms lost their federal contracts. The dataset has since powered follow-up reporting in six states.
