Prop 13 / Insurance Wedge

Working paper studying how rising wildfire insurance premiums act as a shadow property tax that offsets Prop 13 lock-in incentives for long-tenured California homeowners.

Source Code

California’s Proposition 13 caps property tax assessment growth at 2% per year, creating a large implicit subsidy for long-tenured homeowners. The longer you’ve owned your home, the more you save — and the larger the tax penalty for selling and moving. This is the famous “lock-in” effect.

But home insurance is uncapped. Wildfire risk has pushed California insurance premiums sharply higher in high-hazard zones since the 2017–2018 fire seasons, with many carriers exiting the market or non-renewing policies in CalFire Fire Hazard Severity Zones (FHSZ). For homeowners whose Prop 13 lock-in subsidy was the main reason to stay, this “shadow property tax” may be eroding the incentive.

Research question: Does rising insurance exposure reduce the Prop 13 lock-in effect, particularly for the longest-tenured homeowners in high-risk zones?

Identification strategy: Triple Difference (DDD) exploiting variation across:

  • Tenure length — long-tenured owners have the largest Prop 13 subsidy and the most to lose
  • FHSZ zone — CalFire Very High/High zones as quasi-random treatment assignment
  • Time — pre/post 2017–18 wildfire seasons as the insurance shock

M0 result (Census-only MVP): DiD coefficient (HighRisk × Tenure₂₀₁₀) = +0.093* — treated (high-risk, long-tenure) tracts lost more long-tenure owner-occupants over 2010–2020 than the DDD parallel trends assumption would predict. The positive sign is counter to the hypothesis as stated but consistent with the lock-in eroding: long-tenured owners in fire-risk zones appear to be selling at higher rates. The 2010–2020 window straddles the shock, making the M0 estimate a noisy pre/post comparison; the M1 HMDA panel (in progress) will sharpen the timing.

M1 pipeline (in progress): Annual HMDA purchase originations as numerator; residential structure count denominator from the CA Residential Structure Panel (Overture Maps + BPS hind-cast + CAL FIRE DINS correction + bootstrap). ACS B25001 for 2010–2020; bootstrap P50 for 2021–2024.

Data sources:

  • ACS B25038 (tenure by housing units) — M0 outcome
  • CalFire FHSZ SRA polygons — treatment assignment
  • Keys-Mulder insurance panel (county × year 2014–2024) — treatment intensity
  • HMDA CA bulk CSVs — M1 numerator (in progress)
  • CA Residential Structure Panel — M1 denominator

Tech: Python, pandas, geopandas, statsmodels, pyarrow; spatial clustering with Conley SEs