Source: RESOURCES FOR THE FUTURE, INC. submitted to NRP
WILDFIRE RISKS IN RURAL COMMUNITIES: MEASURING IMPACTS, MODELING CHOICES, EVALUATING POLICIES
Sponsoring Institution
National Institute of Food and Agriculture
Project Status
COMPLETE
Funding Source
Reporting Frequency
Annual
Accession No.
1026083
Grant No.
2021-67023-34483
Cumulative Award Amt.
$425,015.00
Proposal No.
2020-06923
Multistate No.
(N/A)
Project Start Date
Apr 1, 2021
Project End Date
Mar 31, 2024
Grant Year
2021
Program Code
[A1651]- Agriculture Economics and Rural Communities: Environment
Recipient Organization
RESOURCES FOR THE FUTURE, INC.
1616 P ST NW
WASHINGTON,DC 20036
Performing Department
(N/A)
Non Technical Summary
Impacts of wildfires in rural communities in the western US have increased in recent decades due to increasing wildfire frequency and severity, and increased development within high wildfire hazard areas. Increasing risk, as well as policies designed to address that risk, have potential to reshape these areas--many of which are in what is called the wildland-urban interface (WUI), where developed areas abut or intermingle with large areas of wildland vegetation--in coming years. In this project, we study who is affected by wildfire risk, and how WUI communities might change under various policies to address that risk. First, we undertake the first detailed assessment of exactly who is affected by wildfires, using microdata on property values and characteristics merged with household demographic data, including income. This analysis sheds light on the distribution of impacts across income groups within regions of individual states, information that is currently lacking but is important for policymakers. In the second part of the project, we build an equilibrium locational sorting model to analyze how households relocate in response to changing wildfire risks and the implementation of policies such as zoning and defensible space requirements, building codes, and insurance rate hikes. The results of the model will provide guidance on the cost-effectiveness and distributional impacts of alternative approaches to addressing wildfire costs on rural communities.
Animal Health Component
100%
Research Effort Categories
Basic
0%
Applied
100%
Developmental
0%
Classification

Knowledge Area (KA)Subject of Investigation (SOI)Field of Science (FOS)Percent
60501993010100%
Goals / Objectives
Goal: To improve understanding of the distribution of wildfire risk across the population of the western US, and how the distribution is likely to change under future risk and policy scenarios.Objectives:Use Zillow ZTRAX data and US Census data to identify how wildfire risk across the population of the western US.Use Zillow ZTRAX data linked to Home Mortgage Disclosure Act data, as well as data describing changes in wildfire risk or perceptions of risk, to evaluate how changes in risk are likely to drive sociodemographic change within high wildfire risk areas.Using data on possible changes in (e.g. data describing insurance premiums), use results from objective 2 to project sociodemographic change in high wildfire risk areas over time as wildfire risk increases.
Project Methods
Objective 1: Evaluate the distributional incidence of wildfire risk in the western USEfforts: Our first objective is to investigate the distribution of wildfire risk across the population of the western US. We will merge American Community Survey data from the US Census Bureau and assessed property values with data describing the spatial distribution of wildfire hazard across the western US. This will allow comparison of property values and other variables of interest across low and high fire hazard areas within each state.Our data on assessed property values come from the Zillow Assessor and Real Estate Database (ZTRAX). ZTRAX includes real estate transaction records and property tax assessment data for approximately 150 million properties across the US. To facilitate examination of the incidence of wildfire risk across the distribution of additional demographic variables, we will supplement our property assessors' data with Census block group-level data from the US Census Bureau American Community Survey (ACS). Specifically, we will collect data on per capita income, percent white and percent elderly within each block group. We will assign ACS demographic variables to properties based on property locations relative to Census block groups.Finally, we will merge property tax assessors' data and American Community Survey with spatial data on wildland fire potential (WFP), a measure of wildfire hazard developed by the US Forest Service. WFP is a spatial index of the relative potential for wildfires that would be difficult to contain and therefore potentially damaging to valued assets such as homes. The result is a spatial dataset that ranks the potential for dangerous uncontrolled wildfires across the US using a continuous ordinal index that varies from 0 to approximately 100,000.Together, these data will allow us to assess where on the distribution of household property value and various demographic measures, wildfire risk tends to fall. For example, it will allow us to assess whether households with high or low value properties are disproportionately likely to bear wildfire risk. In further analysis, we will statistically test whether homes at various points of the property value distribution are over- or under-represented in high fire hazard areas. We will also study the distributional incidence of fire hazard at the sub-state level and compare rural areas to areas more on the urban fringe. Finally, we will use Census data to study the incidence of wildfire hazard over several other demographic variables to provide a more complete picture of the risk of wildfire across geographies and demographics.Evaluation: We will evaluate whether we have contributed to changes in knowledge based on our success in publishing the results of this analysis in a peer-reviewed academic journal, and distributing them to policymakers.Objectives 2 and 3: Investigate housing choices over wildfire riskEfforts: The project's second objective is to evaluate how the demographics of households within high wildfire hazard areas are likely to change under various policy scenarios.Households will respond to changes in wildfire hazard and cost-of-living based on household wealth and preferences over amenities and risk. These factors may vary across demographic groups, implying that shifts in hazard and cost may lead to demographic change within the wildland-urban interface. To study how income and preferences influence choices over whether to live in high wildfire hazard areas of the wildland-urban interface, we will use a horizontal sorting model, an approach that explicitly allows preferences, such as preferences for wildland amenities, to vary across households. Using model estimates, we will simulate consequences of policies that change the costs of living in high wildfire hazard areas for demographic change within the wildland-urban interface.Following previous horizontal residential sorting models (e.g. Bayer, Ferreira and McMillan 2007, Klaiber and Phaneuf 2010, Bakkensen and Ma 2019), our theoretical model is based on random utility theory and facilitates a modern discrete choice estimation strategy. Upon choosing to move, each household selects a residence based on a combination of housing and neighborhood characteristics, including price. Households may choose to live in or outside of the WUI. For homes inside the WUI, location can affect amenities and wildfire risk. We allow homes with similar characteristics in the same neighborhood and approximately the same level of wildfire hazard to share common unobserved characteristics, which we include in household indirect utility in a choice-specific constant. We also allow preferences for home characteristics, including wildfire hazard and wildland amenities, to vary based on observed household characteristics, such as income, and unobserved factors that are idiosyncratic to each household and choice. Households choose a home within a given neighborhood, house-type, and wildfire hazard-level when the utility from living there exceeds that from living in any other choice within their choice set. We estimate heterogeneous household preferences for home characteristics using a two-step estimation procedure.First, we construct the probability each household selects each property as a function of choice-specific constants and household-specific components of preferences. Under the standard assumption that unobserved, idiosyncratic components of household preferences are Type I extreme value-distributed, we can estimate preferences by maximum likelihood using a multinomial logit model. We estimate choice-specific constants using a contraction-mapping algorithm described by Berry (1994) (and used in many other studies) to speed up computation.In the second step of the estimation procedure, we regress our estimates of the property-specific constants on property characteristics, including price and wildfire hazard, in order to estimate components of preferences over these characteristics that are shared across households. Estimating our model of sorting over wildfire hazard will require the following data: (1) property characteristics, prices, and transactions, (2) household characteristics, (3) insurance premiums, and (4) wildfire hazard data. As in objective 1 of this proposal, data describing properties and transactions will come from ZTRAX. To identify demographic characteristics of buyers, we follow a procedure established by Bayer et al. (2016) for merging property transactions data with the Consumer Financial Protection Bureau (CFPB) Home Mortgage Disclosure Act data set. The HMDA requires financial institutions to disclose loan-level data regarding mortgages, including Census tract, loan amount, and applicant income, race, and gender. Finally, we will measure wildfire hazard on each property by spatially joining properties to the wildland fire potential (WFP) data set, which we also use in Objective 1, using property coordinates provided by ZTRAX.Parameters recovered from the sorting model will provide estimates of how people of varying racial and income backgrounds will tend to move, given the choices they face and the associated costs and characteristics of the spatial landscape. With these estimates, we can then predict how various policies that alter these choice conditions will affect where people choose to live and their welfare.Evaluation: We will evaluate whether we have contributed to changes in knowledge based on our success in publishing the results of this analysis in a peer-reviewed academic journal, and distributing them to policymakers.

Progress 04/01/21 to 03/31/24

Outputs
Target Audience:Target audiences reached during this reporting period include academic audiences, policymakers and the general public. We reached academic audiencesthrough academic presentations at University of California, Berkeley, University of Alaska, Anchorage, and the annual conference of the Association of Environmental and Resource Economists, through two published academic papers ("Risk Disclosure and Home Prices: Evidence from California Wildfire Hazard Zones" and "The Distributional Incidence of Wildfire Hazard in the Western United States"), and through two Resources for the Future working papers ("Risk Disclosure and Home Prices: Evidence from California Wildfire Hazard Zones" and "Sorting over Wildfire Hazard").We reached policymakers and the general publicthrough an RFF blog post, and a podcast in which Dr. Ma discussed findings from our paper on risk disclosure and home prices. Changes/Problems: Nothing Reported What opportunities for training and professional development has the project provided?The paper written and published under goal 1 included a junior co-author, RFF Research Analyst Molly Robertson. Papers written under goals 2 and 3 include as a co-authors a recent PhD, Connor Lennon.RFF Research Analyst Emily Joiner is included as a co-author on the second paper written under goals 2 and 3. Robertson, Lennon and Joiner eachplayed significant rolesin assembling and analyzing the data sets used in the project. Robertson contributed to writing the paper written under goal 1, and Joiner and Lennon each also contributed to writing the second paper written under goals 2 and 3. Joiner led writing of an RFF blog post related to the first paper written under goals 2 and 3. How have the results been disseminated to communities of interest?We have disseminated results through two academic articles and a public working paper, presentations in academic seminars and seminars and to congressional and agency staff, outreach to the media, and public-facing communications products (blog posts, podcasts). We will perform additional outreach related to the "sorting" paper (the second paper written under goals 2 and 3) upon its publication in an academic journal. What do you plan to do during the next reporting period to accomplish the goals? Nothing Reported

Impacts
What was accomplished under these goals? Under goal 1, we produced a manuscript, published in Environmental Research Letters, that uses Zillow ZTRAX data and US Census data to analyze how wildfire hazard is distributed across the population of the western US. Ouranalysis found that there is substantial heterogeneity in household demographics and property values within high wildfire hazard areas, but that high property value homesare disproportionately likely to be in high hazard areas, as are high income households.However, because many high-value high wildfire hazard properties are concentrated in high density areas, most of the high wildfire hazard area in the western US is sparsely populated and comprises mainly relatively low-value properties. Under goals 2 and 3, we produced two manuscripts. The first is published in Land Economics, and the second has recently been released as a Resources for the Future working paper. In the first paper, we studied how wildfire hazard disclosure affects home prices in California. We found that between 2015 and 2022, homes facing disclosure requirements in Californiasold for approximately 4.3% less than nearby homes that did not. This finding highlights the use of disclosure requirements for ensuring that risks are reflected in housing markets. In the second paper, we used a similar strategy, which also relies on disclosure requirement boundaries in California, to study household sorting over wildfire hazard in California. In this paper, we again found that households are averse to fire hazard in areas where disclosure is required. Moreover, we found that aversion is increasing in income, which suggests that over time lower-income households may replace high income households in high-hazard areas, raising concerns about distributional justice. This manuscript will be submitted to a peer-reviewedenvironmental economics journal.

Publications

  • Type: Journal Articles Status: Published Year Published: 2024 Citation: Ma, L., Walls, M., Wibbenmeyer, M., & Lennon, C. (2024). Risk Disclosure and Home Prices: Evidence from California Wildfire Hazard Zones. Land Economics, 100(1), 6-21.
  • Type: Journal Articles Status: Published Year Published: 2022 Citation: Wibbenmeyer, M., & Robertson, M. (2022). The Distributional Incidence of Wildfire Hazard in the Western United States. Environmental Research Letters, 17(6), 064031.
  • Type: Other Status: Published Year Published: 2024 Citation: Ma, L., Wibbenmeyer, M., Joiner, E., Lennon, C., & Walls, M. (2024) Sorting over Wildfire Hazard. Resources for the Future Working Paper 24-05.
  • Type: Conference Papers and Presentations Status: Other Year Published: 2023 Citation: Wibbenmeyer, M. (2023) Sorting over Wildfire Hazard. Presentation at the Annual Conference of the Association of Environmental and Resource Economists, June 1.
  • Type: Conference Papers and Presentations Status: Other Year Published: 2023 Citation: Wibbenmeyer, M. (2023) Sorting over Wildfire Hazard. Presentation at the 2023 University of Alaska, Anchorage Summer Workshop on Alaska's 21st Century Economy, June 16.
  • Type: Conference Papers and Presentations Status: Other Year Published: 2023 Citation: Wibbenmeyer, M. (2023) Sorting over Wildfire Hazard. Presentation to the Department of Agricultural and Resource Economics, University of California, Berkeley.


Progress 04/01/22 to 03/31/23

Outputs
Target Audience: Target audiences reached during this reporting period include academic audiences (reached through academic presentations) and policymakersand thegeneral public (reached through press coverage related to our article, "The Distributional Incidence of Wildfire Hazard in the Western US", published in Environmental Research Letters). Changes/Problems:In 2021, Zillow announced the termination of the ZTRAX program, which we are relying on for data. Zillow allowed ZTRAX users to renew Data Use Agreements a final time in July 2021, and use of Zillow data will no longer be permitted after September 2023. As a result, it may no longer be feasible--due to timing of the program's termination--to publish separate papers estimating a household sorting model and a follow-up paper including policy analysis using model estimates. We will focus on completing the paper describing the household sorting model, and will include policy analysis in that paper where possible. In place of the third paper, which was to be focused on policy analysis, we have instead pursued a second paper--the "hedonic" paper--which also provides insights into how wildfire hazard affects households' choices about where to live. What opportunities for training and professional development has the project provided? The paper written and published under Goal 1 included a junior co-author, RFF Research Analyst Molly Robertson. Papers in progress under goals 2 and 3 include as co-authors a recent PhD, Connor Lennon, and RFF Research Analyst Emily Joiner. How have the results been disseminated to communities of interest?We have disseminated resultsthrough an academic article, presentations in academic seminars and to congressional and agency staff, and through outreach to the media. We will perform additional outreach related to the "hedonic" and "sorting" papers upon release of those papers in the Resources for the Future working paper seriesand as published journal articles. We will present the "sorting" paper in two academic conference presentations in June 2023. What do you plan to do during the next reporting period to accomplish the goals?During the next reporting period, the team will work toward two goals. First, we will work toward publication and outreach related to our "hedonic" paper estimating the effects of wildfire hazard disclosure on property sales prices. Second, we will use the econometric strategy identified in this paper for separating the effects of wildfire hazard on home prices from the effects of correlated amentities to estimate a household sorting model describing effects of wildfire hazard on household residential location choices. Initial results from the "sorting" paper will be presented in two conference presentation during June 2023, and the paper will be submitted to a journal by September 2023.

Impacts
What was accomplished under these goals? We have provided a broad-scale assessment of the distribution of wildfire hazard across households in the western US, which will provide policymakers and the public with critical context when assessing the fairness of various policy proposals aimed at reducing wildfire risk. Currently, we are working on two papers that contribute to understanding how wildfire risk affects household decisions of where to live. This research will inform housing policy in high wildfire hazard areas, including building codes, policy related to insurance costs, and policy affecting rates of development within and outside high risk areas. Goal 1 was completed in a previous performanceperiod. Overthe past performance period, we have discerned a viable econometric strategy for use in identifying effects of wildfire hazard, as distinct from correlated environmental amenities (e.g. forest cover),on home values. We used this strategy, which relies on discrete changes in disclosure requirements that occur across fire hazard zone boundaries in California, to estimate a "hedonic" model ofeffects of hazard disclosure on home price. Results indicatethat disclosure leads to an approximately 5% decline in home value at the boundary. A revision to our paper summarizing these results has been requested at Land Economics. In the past performance period, we have also begun work to estimate a household "sorting" model. This model will build on the identification strategy used in our hedonic paper to evaluate how household demographic characteristics influence preferences over neighborhood characteristics, including fire hazard. This work will be useful for informing expectations of how demographics within high fire hazard areas may change as the costs of living in these areas increase.

Publications

  • Type: Journal Articles Status: Under Review Year Published: 2023 Citation: Ma, L., M. Walls, M. Wibbenmeyer, Joiner, E., and C. Lennon. Are Wildfire Risks Capitalized into Home Prices? Evidence from California.


Progress 04/01/21 to 03/31/22

Outputs
Target Audience:We have reached academic and policy audiences through presentations of our research in seminars, congressional briefings, and meetings with agency staff. We will reach a broad audience of social scientistics studying wildfire and other natural hazards, policymakers, and the general public with our article "The Distributional Incidence of Wildfire Hazard in the Western US", which has been accepted for publication by Environmental Research Letters and will be released there in the coming months. In addition to being accepted for publication by ERL, ERL selected the article for press release;we anticipate that this will help the article reach a broader audience, which includes the general public. Changes/Problems:In 2021, Zillow announced the termination of the ZTRAX program, which we are relying on for data. Zillow allowed ZTRAX users to renew Data Use Agreements a final time in July 2021, and use of Zillow data will no longer be permitted after September 2023. As a result, it may no longer be feasible--due to timing of the program's termination--to publish separate papers estimating a household sorting model and a follow-up paper including policy analysis using model estimates. We will focus on completing the paper describing the household sorting model, and will include policy analysis in that paper where possible. What opportunities for training and professional development has the project provided?The paper written and published under Goal 1 included a junior co-author, RFF Research Analyst Molly Robertson. As well, we have worked closely on progress toward Goal 2 with University of Oregon graduate student Connor Lennon. How have the results been disseminated to communities of interest?We have disseminated results through academic articles and throughpresentations in academic seminars and to congressional and agency staff. Our soon-to-be released Environmental Research Letters article was selected by ERL for press release, and the RFF communications team will do coordinated outreach to press outlets at the time of the article's release. What do you plan to do during the next reporting period to accomplish the goals?During the next reporting period, the team will work toward two goals. First, we will estimate a "hedonic" model of the effect of wildfire hazard on property values, and we will write a paper describing the results. We will submit this paper to a joint special issue of the journals Land Economics and Journal of Housing Economics. Second, we will use theeconometric strategy identified in this paper for separating the effects of wildfire hazard on home prices from the effects of correlated amentities to estimate a household sorting model describing how wildfire hazard affects household location choices. A draft manuscript of this paper will be completed by January 2023. This paper was included in a conference session submitted to the 2023 American Economic Association annual conference, and if accepted we will present the paper there in January 2023.

Impacts
What was accomplished under these goals? Under goal 1, we have completed a paper measuring how wildfire hazard is distributed across households in the western US. The paper uses property-level tax assessmente data from Zillow's ZTRAX database, as well as wildfire hazard data from the USFS and demographic data from the US Census, to measure how wildfire hazard is distributed across demographic groups in the western US. This paper has been accepted and will soon be published in Environmental Research Letters. Under goals 2 and 3, we have begun work toward estimating a household "sorting" model that will describe how wildfire hazard and costs of living in high wildfire hazard areas affect housing location choices. A significant econometric challenge in estimating this model is identifying the effect of wildfire hazard, as distinct from correlated environmental amenities (e.g. forest cover), on housing location choices. Our work thus far has focused on testing various econometric strategies.

Publications

  • Type: Journal Articles Status: Awaiting Publication Year Published: 2022 Citation: Wibbenmeyer, M. and M. Robertson. The distributional incidence of wildfire hazard in the Western United States. Environmental Research Letters, forthcoming.