1. Wildfire Risk, Salience, and Housing Development in
the Wildland-Urban Interface
Shawn J. McCoy
Department of Economics, UNLV
shawn.mccoy@unlv.edu
Nicholas B. Irwin Katie Jo Black
UNLV Kenyon
2. Heightened costs due to fire a function of (1) Development (2) Fire frequency and severity
1] Rapid, unprecedented expansion of
the WUI “Wildland-Urban Interface”
30.8 million units in 1990
43.4 million units in 2010
(Radeloff et al., 2018)
2] Fire frequency and severity is increasing
a) Changes in global climates
(Abatzogluou and Williams, 2016)
b) Human influence
(Syphard et al., 2018) climate vs. human influence
(Syphyard et al., 2007) for anthropogenic ignition
(Spyratos et al., 2007) for landscape flammability
and development
Brings “development” to the
forefront of this discussion.
We tackle a very specific angle
related to fire risk perceptions
and the rate of development in
the WUI.
3. Overview
Broadly: Drivers of development in wildfire-risk areas of the U.S.
Question: What is the link between fire-risk saliency and development?
Background: Risk perceptions inexorably linked to willingness to undertake
a risky decision.
Natural disasters: evidence that in absence of recent event
agent’s may hold inaccurate beliefs regarding disaster risk.
Challenge: Finding exogenous (random) variation in saliency
Solution: Recent large-scale wildfires as a source of randomness
Estimation: “What is the impact of a recent fire on the rate of construction
in wildfire risk areas?”
“How do the impacts of fire vary over time?”
12. DATA: Latent RiskHow we construct our panel data set: For every county “c” …
𝐷𝑒𝑣 𝑔𝑐𝑡
𝐹𝑖𝑟𝑒𝑠 𝑐𝑡
New houses in group “g” of county “c” in time “t”
Number of fires in county “c” in time “t”
13. DATA: Latent RiskBaseline Econometric Model: Set-up
𝐸 𝐷𝑒𝑣 𝑔𝑐𝑡|𝑥 𝑔𝑐𝑡 = 𝑒𝑥𝑝൫𝛽1 𝐹𝑖𝑟𝑒𝑠 𝑐𝑡−1 × 𝐻𝑖𝑔ℎ 𝑅𝑖𝑠𝑘 𝑔𝑐 + 𝛽2 𝐹𝑖𝑟𝑒𝑠 𝑐𝑡−1+. . .
൯… + 𝛼 𝑔𝑐 + 𝜂 𝑡 + 𝜀 𝑔𝑐𝑡 .
Poisson Fixed-Effects Estimation Strategy
What are we testing for? What are we estimating?
“Testing for / estimating the differential development response to fire in
high-risk vs. low-risk areas over time.”
See. Hallstrom and Smith (2005) among others…
Idea: Agents in the market where are disaster occurred are exposed
to the information conveyed by the shock, but consider the
information relevant to the risky area.
14. DATA: Latent RiskIsolating saliency effects
Main Challenge: Does our High Risk vs. Low Risk comparison isolate pure saliency
effects?
No, for these reasons:
-The fires damage the landscape
-Loss of recreational use near the burn
-Dis-amenity effects associated with the burn scar
-Visual dis-amenities associated with seeing a burned area
(All of these could deter development)
But: McCoy and Walsh (JEEM, 2018) show that these effects aren’t present
once we omit properties within 5km of the burn irrespective of their risk
classification.
16. DATA: Latent RiskSummary of Main Results
Finding 1: Evidence that fire induced changes in saliency deter the rate of new
construction in fire-risk areas (e.g. statistically significant reductions in new
housing units in high risk vs. low risk areas one year after a fire).
Finding 2: Effects are short-lived: After two years, development rates in high risk
areas are the same as they were before a fire
Evidence that in the absence of a recent fire, perceived risks might
diverge from actual risks.
Policy Implication: Policy intervention targeting saliency welfare
improving?
Raises another question:
What is the expected magnitude of the reduction in at risk housing from
policies targeting risk saliency?
17. DATA: Latent RiskSummarizing the policy implications of our findings
Approach:
We have already estimated the saliency effects of a wildfire
Estimates represent a plausible upper bound on the strength of the saliency
effects of a man-made policy
Counterfactual Simulation:
(a) assume “fires” happened in every time period (year) for years
that there was not actually a fire (e.g. proxy for the presence of
“information based policies” with hypothetical fires)
(b) adjust development rates based on our model estimates
(c) compare how development really looked vs. how it would have
looked in the presence of these additional “fires”