Skip to main content

Which Way the Wind Blows: Climate Risk and U.S. Housing Markets

An interactive story connecting county housing trends, FEMA climate-risk groups, extreme events, and a county Climate Playbook using Python, DuckDB, and D3.

Housing is local. So is exposure to extreme weather. This interactive story follows the evidence from a decade of U.S. county housing trends to a Climate Playbook that brings together a county’s risk exposure, housing performance, and characteristics.

Explore the interactive story ↗ View code on GitHub ↗

The question

Do counties with different climate-risk ratings show different housing-market patterns? And what can we learn by looking at the months before and after extreme weather events?

The project focuses on Redfin’s median sale price per square foot, measured as year-over-year change. Calendar-history charts use the latest ten complete calendar years available in the housing data, then compare counties within FEMA National Risk Index groups.

From scattered sources to a shared geography

The pipeline combines Redfin housing data, FEMA risk ratings and disaster declarations, NOAA storm events, NCEI weather, Census ACS characteristics, and StatsAmerica economic data.

Python handles data acquisition and preparation. DuckDB separates raw records, normalized tables, domain features, and event-window analysis. Source metadata and file hashes make it possible to trace inputs through the pipeline. Related FEMA declarations and qualifying NOAA records are reconciled into canonical county incidents while preserving their source identifiers.

What I built

  • A layered analytical database: raw records remain separate from the tables used for comparisons and publication.
  • Event-window analysis: housing observations are aligned from twelve months before event start through three or five years after event end. Each view uses its own complete county-event trajectories, with median trends and interquartile ranges.
  • An interactive web-based story: county geography, housing histories, risk-group comparisons, and within-group characteristics are published as a static D3 experience.
  • A county Climate Playbook: county search connects local housing history and past events with risk ratings, performance against peers, county makeup, and an overall scorecard. Coverage checks identify counties with insufficient housing history.
  • Supporting notebooks and methodology: quality checks, feature exploration, and event diagnostics sit alongside the final story.

The story: Housing from a climate risk perspective

The story starts with county housing trends, groups counties by climate risk, and compares their growth trajectories around extreme events. It then examines economic and demographic characteristics associated with stronger or weaker housing performance within each risk group. The Climate Playbook brings those views together for a selected county.

NoteBoundaries of the evidence

The comparisons describe observational associations. Complete-history requirements change the counties and events represented in each view, and overlapping event windows can reuse housing observations. The Playbook’s equally weighted scorecard summarizes risk exposure and observed associations. Factors labeled “significant” on the story are selected by correlation magnitude and a minimum-count rule, so that label reflects the display rule rather than a formal significance test.

Explore further

Next: Do County Characteristics Explain Migration Flows? →