As Hurricane Irma bore down on Florida, AP data journalist Meghan Hoyer saw an opportunity to tell an important data-driven story in the moment by looking at flood insurance rates for homeowners who could be affected by the storm. Using Federal Emergency Management Agency data that no other news organization had accessed in relation to the 2017 storms, Hoyer plotted the coverage rates for homes in flood hazard areas. The results were deeply troubling: Florida’s overall flood insurance rate for hazard-zone homes was just 41 percent. In just five years, the state’s total number of federal flood insurance policies had fallen by 15 percent.
Hoyer took the story further, though, looking not only at homes in Irma’s path but at the national rate of flood insurance across the country. Maureen Linke jumped in to visualize the change since 2012, providing a county-level map and lookup with dynamic text describing the state of flood insurance for each county. Florida-based reporter Terry Spencer pursued on-the-ground reporting and expert sources for a strong team effort that exposed dramatic potential risk.
What makes this project innovative?
This story is an example of using data to make the potential impact of local disaster more relevant to readers across the country. Typically, this is the sort of enterprise data journalism arrives days or weeks after the event, but with the lessons of Hurricane Harvey fresh in mind, Hoyer was able to provide that insight while the storm still lurked offshore. While other news organizations had reported on the flood insurance program and its deficiencies, Hoyer was the first to obtain data that put those deficiencies in the context by showing the lack of coverage for those most vulnerable to flooding.
What was the impact of your project? How did you measure it?
The story was prominently featured by members across Florida and beyond. Hoyer packaged the data for distribution, leading to localized member stories in San Francisco and Jacksonville, as well as a follow-up by USA TODAY.
Source and methodology
Hoyer accessed three separate FEMA datasets -- one that provided current coverage details for each community in the US, another that provided similar data from 2012, and a third that looked at county-level coverage rates for high-risk and flood-prone residences. The three datasets weren’t standardized to be joined or used together, so Hoyer did that work in R, and ultimately added census data and NOAA data to flag coastal counties as well.
All data cleanup and analysis was done in R. The interactive map that provided coverage levels for each county in the U.S. was built in D3.