Abstract
Natural disasters have been increasing in intensity while rental affordability has been declining in many cities. This article presents a new dataset at this intersection, merging rental outcomes with disaster incidence and government assistance. We construct two new panel models that estimate the effects of the natural disasters from 2000 to 2020 on zip codes across California and Florida. Results indicate that rents per unit increase by 0.2 to 1.3 percent immediately after a disaster and 6.5 to 12.5 percent in the long run. These findings are driven primarily by hurricanes and wildfires—and to a lesser extent, winter storms. The effects accumulate after multiple disasters. They are persistent and not mean-reverting. We further demonstrate that these impacts are moderated somewhat in communities receiving federal Community Development Block Grant–Disaster Recovery (CDBG-DR) funds, whose rental restrictions and rebuilding incentives temper post-disaster rent escalation. By centering this analysis, we highlight how disaster-recovery design influences long-term housing affordability and equity.
Natural disasters fundamentally disrupt rental housing markets. They destroy properties, displace residents, and alter the balance of power between tenants and landlords, potentially exacerbating income and wealth inequality. At first, disasters reduce both the demand for and the supply of housing, as population and housing units decline. What this means for housing affordability is an open empirical question, especially for long-term affordability as residents return home and property owners rebuild. What is the trajectory of rental housing markets during this multi-year process? And how is it altered by public policies designed to aid vulnerable residents?
In this article, we investigate how rents and vacancy rates evolve in two large states with extensive disaster experience, California and Florida, in the wake of six categories of natural disasters: earthquakes, flooding, hurricanes and tropical storms, tornadoes, wildfires, and winter storms. We situate this question within the growing literature on natural disasters, documenting the forces that push rents up and down and the many persistent, negative real economic outcomes that befall communities exposed to disaster damage despite post-disaster the Federal Emergency Management Agency (FEMA) economic assistance. These findings underscore the importance of understanding how the most vulnerable members of society bear these losses—and the role of public policies that attempt to alleviate this burden.
We begin our analysis by compiling a new database that matches disaster shocks to rental price and vacancy rate outcomes. We obtain zip code-level effective rents per unit and per square foot, asking rents per unit and per square foot, and vacancy rates from CoStar on a quarterly basis for twenty years from 2000 to 2020 in the major metropolitan areas across both California and Florida. We merge these data to zip code-level disaster incidence from the FEMA.
We then construct two types of panel models to analyze these data. First, we estimate a dosage model that counts the number of disasters by quarter and by year in each zip code, allowing us to account for the temporal overlap from multiple treatments in a way that typical difference-in-differences models do not allow. Second, we estimate the new interactive fixed effects model developed by Licheng Liu and colleagues (2024), which matches treated and control units prior to treatment to eliminate the parallel trends concern and allows units to switch in and out of treated status, relevant to our study setting as zip codes recover from one disaster and then are impacted by another.
The results in this article show that rents per unit increase by 0.2 to 0.6 percent immediately after a disaster in Florida and 0.9 to 1.3 percent immediately after a disaster in California. In the long run, the cumulative rental appreciation is 6.5 percent higher in disaster-impacted zip codes in California and 12.5 percent higher in Florida. These findings are driven primarily by wildfires and winter storms. Effects accumulate after multiple disasters. They are persistent and not mean-reverting, and they are often driven by factors other than simply the immediate destruction of housing supply, as the effect on vacancy rates is mixed.
We also investigate trajectories in markets with differential access to federal aid to support renters. In markets that receive funding from the CDBG-DR—funding that comes with certain types of rental requirements designed to protect tenants and improve rebuilding of affordable housing—the long-term increase in rents is slightly lower than the average long-run rent increase in disaster-affected zip codes. While disasters’ housing-market effects are important in their own right, one of our primary analytical foci is on how federal recovery policy shapes those trajectories. The CDBG-DR program offers a quasi-experimental setting to compare similar disaster-affected areas that did and did not receive targeted recovery funds. Accordingly, we interpret all preceding estimates as a foundation for understanding the moderating role of CDBG-DR.
The results of this study offer a meaningful, policy-relevant perspective on the impacts of disasters on rental markets. Prior work on rental impacts focused on one disaster, Hurricane Sandy, in one region (New York City) (Harwood 2025); the remainder of the literature looks at property prices. Our finding that disasters in Florida and California induce sustained rental price appreciation, that is moderated by the presence of CDBG-DR, over a twenty-year period, significantly expands the literature and geographic context. This study provides more generalizable estimates than previous work by covering zip codes in major metropolitan areas of these two large states (by population, rental market, and frequency and strength of disasters). Moreover, the study provides a showcase of two valid, repeatable methods that make future research on disaster impacts more robust. The remainder of the article reviews the economic literature documenting how housing markets adjust to natural disasters and develops hypotheses that can be tested with our new dataset. It then describes the data and methodology, presents the results, and concludes with implications for policymakers tasked with preparing for and responding to natural disasters in ways that prevent socioeconomic inequalities from expanding.
HOW NATURAL DISASTERS CAN IMPACT RENTAL HOUSING MARKETS
The classic economics model of supply and demand does not offer a clear prediction for rental affordability in the wake of natural disasters. In this section, we outline the forces that can drive rents up versus down—and how these forces might differ across geographies, disaster types, and time. We also review the literature documenting the government’s role in mitigating these impacts.
Downward Pressure: Forces That Can Decrease Rents
Natural disasters often drive residents out of their homes, at least temporarily, leading to loss of population, income, wealth, and rental housing demand. If this effect dominates, vacancies should increase, and rents should fall. Katharine WH Harwood (2025) shows that this dynamic took hold in low-income neighborhoods in New York City in the wake of Hurricane Sandy. Since there was heterogeneity across neighborhoods, this study suggests that we might find negative or muted effects of hurricanes on rents, but it is not clear if this finding will extend to other disaster types. There has been little research on rents aside from this study, but researchers have documented this impact on property prices extensively.
Even before a disaster occurs, property prices discount at least some of the risk to which they are exposed. In the Fargo-Moorhead metropolitan area in Minnesota and North Dakota, straddling the oft-flooding Red River of the North, for example, housing prices within the one-hundred-year floodplain are valued 3.5 to 12.2 percent less than comparable properties outside the floodplain (Zhang and Leonard 2019). This risk is often moderated by mitigation features. In Miami-Dade County in Florida, hurricane mitigation features increase housing prices by 4 to 10 percent—but importantly, only if those features are verified, suggesting that the effect is due to the insurance premium value rather than the actual structural protection of the house (Gatzlaff et al. 2017). Similarly, in Japan, Tokyo has successfully limited flooding in low-lying areas, and housing prices have responded positively (Sado-Inamura and Fukushi 2019). When a disaster actually occurs, however, it becomes clear that more decreases in real estate prices—both residential and commercial—are in store (Boustan et al. 2020; Cheung et al. 2018; Eichholtz et al. 2019). Even in Fargo-Moorhead, where the discount was apparent before the floods, Lei Zhang and Tammy Leonard (2019) find that the discount grows significantly after floods.
Why do prices respond so negatively? Resident behavior offers some explanation. From 1920 to 2010, natural disasters have driven significant out-migration from US counties despite increasingly active governmental policies of assistance (Boustan et al. 2020). After Hurricane Katrina, over one-third of the displaced residents did not return in the decade that followed, significantly reducing housing demand (Deryugina et al. 2018). These residents who evacuated were more likely to be older, higher-income, and White, leaving the local market with lower wealth and more vulnerability per capita (Bleemer and van der Klaauw 2019). Homeowners who do stay are more likely to get insurance after floods, but they do not appear to distinguish between high-cost and low-cost floods in choosing the optimal policy—and regardless, the effect disappears within a decade (Gallagher 2014). This evidence is consistent with laboratory experiments showing that a large share of the population is “experiential,” meaning they are only willing to sacrifice if they personally experience extreme events. If there is a significant delay between the sacrifice and the payoff—that is, between mitigation and damages—then the group will fail to cooperate in preventing disaster (Ghidoni et al. 2017). Thus, disasters will not be sufficiently anticipated, damages will not be sufficiently mitigated, and costs will not be borne by society equitably.
These failures, in turn, generate real economic outcomes. Comparing observably equivalent blocks in New Orleans, flooded households had higher mortgage delinquency rates, lower credit scores, and lower homeownership rates (Gallagher and Hartley 2017; Bleemer and van der Klaauw 2019). After Hurricane Katrina, labor income fell, and unemployment receipts, nonemployment, retirement account withdrawals, and uptake of disability insurance and unemployment all increased. Most of these outcomes returned to their original level after one year, and wages even became higher for hurricane victims, mostly for the evacuees who did not return (Deryugina et al. 2018). Thus, it is not surprising that county-level natural hazard damages from 1999 to 2013 were associated with increases in wealth inequality between Whites and Blacks, between college-educated households and high school dropouts, and between homeowners and renters (Howell and Elliott 2019).
Upward Pressure: Forces That Can Increase Rents
On the other hand, the physical destruction wrought by natural disasters often eliminates capital stock or renders it uninhabitable, driving down rental supply. If this effect dominates, vacancies should decrease, and rents should rise. In the wake of the 2025 wildfires in the Los Angeles area, for instance, news outlets reported significant rate hikes as displaced households flooded the nearby rental markets searching for immediate housing outside the burned areas (Siegel et al. 2025). Three weeks after the Eaton Fire began, the neighboring city of Pasadena had experienced a 60 percent decline in available units and an 85 percent decline specifically in two-bedroom units that could house entire families (Patap 2025). Economic theory suggests that these declines should have the largest impact in markets with low housing supply elasticities, where it is difficult and expensive to build, funneling these demand pressures more into rents than into new construction (Glaeser and Gyourko 2018).
However, this new equilibrium should be temporary. As tenants return and landlords rebuild, the supply curve should retrace its path back to the original equilibrium—unless something more fundamental has changed about how the market operates. There are at least three ways in which these changes could make rental appreciation more permanent.
First, there could be a permanent increase in rental demand. Tamara L. Sheldon and Crystal Zhan (2019) show that homeownership rates decline after natural disasters, likely because residents update their risk beliefs about the potential losses to property investment. As these residents shift their preferences from owning to renting, the number of renters will increase, putting upward pressure on rents.
Second, insurance companies are likely to raise premiums and deductibles in the wake of natural disasters to cover their losses, protect against future losses, and compensate themselves for the risk associated with new policies. If landlords pass these insurance costs onto tenants, rents will rise. Even in the absence of natural disasters, insurance costs are the fastest growing operating expense that landlords must pay, and nearly half of landlords reported that they were passing some of these insurance hikes on to their tenants (Baar et al. 2024; National Multifamily Housing Council 2023).
Third, it is likely that the disaster destroyed older, more affordable units, allowing developers to build newer, more expensive units in their place. This change in the composition of the housing stock may lead to an increase in rents, and it may attract wealthier residents, changing the composition of the population as well. Ana Varela Varela (2023) finds that, in the aftermath of Hurricane Sandy, flooded coastal properties in areas with higher preflood incomes increased in value, and these areas saw increases in high-income White buyers. This suggests that disaster recovery can lead to increased income segregation via more White high-income enclaves, a pattern documented by Ingrid Gould Ellen and colleagues (2026, this issue). Additionally, disaster recovery aid can exacerbate these inequalities, allocating more capital to wealthier residents, which allows them to remain in the area and rebuild while low-income households are displaced (Rhodes and Besbris 2022). Nicole Lambrou and colleagues (2025) show this pattern after the 2018 Camp Fire in Paradise, California, leading to gentrification.
The Role of Government Intervention
Disaster recovery aid flows in different ways to homeowners and renters, and again in different ways to owners and landlords of rental properties. This possibly creates a disparate effect of disaster recovery effectiveness by housing tenure. Junia Howell and James R. Elliott (2019) find that aid from FEMA is associated with greater wealth inequality in a given county, but Zachary Bleemer and Wilbert van der Klaauw (2019) report that Gulf Opportunity Zone subsidies were larger for young and low-income residents and similar across races, leading to higher homeownership, lower delinquency rates, and greater paydown of outstanding mortgages. What is clear is that these interventions are insufficient to prevent persistently negative outcomes. Even though government transfers exceeded capital damage per capita after the average hurricane from 1979 to 2002, the reduction in employment persisted for five to ten years—with seemingly permanent effects for hurricanes of Category 3 or higher (Deryugina 2017). This negative shock to incomes will tend to reduce rental demand, but it occurs at the same time that landlords may experience greater bargaining power due to the shortage of affordable housing units. On balance, what happens to rents and to the availability of housing, especially when the government intervenes to try to protect renters?
Putting It All Together: Hypotheses Regarding Differential Impacts
This literature review suggests several testable hypotheses that unpack the likely heterogeneity of natural disaster impacts in new ways:
H1: Because the forces of downward pressure and upward pressure compete in different ways for different disasters, there will be significant differences across disaster types, with some disasters leading to increases in rents and others leading to decreases in rents.
H2: Because the forces of downward pressure, upward pressure, and government intervention depend on the local rental market and local and state housing policies, there will be significant differences across states, with California and Florida experiencing different impacts even for the same type of disaster.
H3: On average across all disaster types, rent increases rather than rent decreases are more common after natural disasters, consistent with the strong evidence of upward pressure and anecdotal lived evidence from tenants (Martín et al. 2023).
H4: In California, where housing supply elasticities are lower and construction is more expensive and time-consuming, rent increases in the wake of natural disasters will be greater than in Florida, where supply is more responsive to demand.
H5: In the wake of natural disasters, rental increases will be persistent over the long run, even when vacancy rates do not decrease, reflecting the long-term upward pressure forces identified in the literature that are unrelated to short-term vacancy impacts.
H6: Government intervention that encourages construction of affordable housing in the wake of natural disasters, such as CDBG-DR funding, will moderate the rental appreciation over the long run.
These six hypotheses follow directly from the literature reviewed earlier. If the empirical evidence confirms these hypotheses, it improves our understanding not only about how rental markets work in the wake of natural disasters but also how policy can be calibrated more precisely to the needs of different markets after different types of disasters.
DATA
To conduct our analysis, we match zip code-level rental data on multifamily housing with locally geocoded variables indicating disaster incidence. First, we obtain rental data from the commercial real estate data provider CoStar. For each building in their database, CoStar collects data on the distribution of unit types, average monthly rent for unit type, vacancy rates, and building characteristics, transactions, and valuations. From these property-level data, CoStar aggregates up to weighted averages within each zip code, which is useful for matching with other datasets. While CoStar does not collect the entire universe of properties and zip codes, the sample size is sufficiently large for regression modeling in most major markets within California and Florida.
California and Florida are natural choices for study areas due to the frequent incidence of natural disasters and relatively large rental housing sectors. Florida had thirty-six “billion dollar disasters” from 2000 to 2019 (NOAA 2024a), with overall damages estimated at $150–$300 billion and California had twenty-seven “billion dollar disasters” from 2000 to 2019, with overall damages estimated at $60–$120 billion (NOAA 2024b), across multiple disaster types. Together, these account for 13 to 26 percent of damages from these very large disasters from 2000 to 2019, and likely a high share of smaller disaster damage too. Together, California’s six million-unit rental market and Florida’s three-and-a-half-million-unit rental market make up more than one-fifth of all US rentals. The multifamily rental markets in both states are continuously exposed to disasters, providing a relevant and powerful test case for measuring rental market impacts of natural disasters. Studying outcomes in these states across multiple disasters adds to previous disaster-specific studies of Hurricane Katrina (New Orleans and surrounding regions) and Hurricane Sandy (New York City region). We cannot claim that these two states are representative of the experiences of all Americans in all states, but they provide an important window into many typical experiences and a new window into many cities that have not been studied previously in the literature.
We identify all major natural disasters in California and Florida that generated a FEMA Major Disaster Individual Assistance Declarations and had FEMA funds allocated toward rental properties for rebuilding between 2000 and 2020, the time period for which we have rental data available. These disasters fall into the following six categories: earthquakes, flooding, hurricanes and tropical storms, tornadoes, wildfires, and winter storms. We match the disasters to zip codes, based on FEMA emergency declarations at the county level.1
We collect quarterly data on multifamily buildings for every five-digit zip code available in CoStar in Florida and California. CoStar data are available for 1,061 zip codes in California (of 1,768) and 703 in Florida (of 1,476 total). These zip codes include urban, suburban, and even exurban and rural places in both states (see figure 1). We focus here on two dependent variables: effective rent per unit (asking rent per unit and any concessions) and vacancy rate. We find the results to be robust across different versions of these variables, including effective rent per square foot or asking rent that does not account for rental concessions.2
Zip Code CoStar Building Count
Source: CoStar.
For demographic covariates, we merge these data with the following variables longitudinally from the US Census Bureau’s five-year American Community Survey (ACS) (for 2009 to 2020) and Decennial Census (2000): total population size, poverty rate, share of renters, and share of White residents. ACS and Census data are not available for 2001 to 2008 at the zip code level; we linearly interpolate values for the years when data are not available. Table 1 shows these variables annually for California, averaged across all zip codes and all quarters in the dataset. Following national trends, it shows an increase in rents from 2000 to 2008, followed by a decline during the Great Recession and then a long appreciation period to 2020. Vacancy rates peak in 2009 and then begin a long decline that continues until 2018. The total population grows in most years, and the housing stock achieves steady growth almost every year. The poverty rate is steady in the early years but starts rising after the Great Recession, peaking in 2014 before declining to its lowest point by 2020. The percent of renters rises over time from 41.3 percent to 45.2 percent, as affordability challenges have reduced homeownership. The non-Hispanic White share of the population declines from 52.9 percent to 39.8 percent. In Florida, table 2 shows similar cyclical patterns in rents and vacancy rates, though monthly rents are generally $200 to $400 lower and vacancy rates are generally two to three percentage points higher. The total population is growing faster in Florida, as is the housing stock. The poverty rate is slightly lower, and the percent of renters is much lower, rising from 29.3 percent to 35.8 percent over time. The White share is much higher, declining from 69.3 percent to 53.9 percent.
California Descriptive Statistics
Florida Descriptive Statistics
Scope and External Validity
Our study focuses on California and Florida, which together account for roughly one-quarter of US disaster damages and encompass diverse hazard types and housing markets. Nevertheless, these findings should be interpreted as analytically illustrative rather than nationally representative. Similarly, the CDBG-DR analysis conditions on areas that received federal recovery allocations, which may reflect higher verified damage or administrative capacity than other jurisdictions.
METHODOLOGY
Our goal is to identify the effect of natural disasters on the rental market. A standard difference-in-differences approach is insufficient to achieve this goal, even given recent advances in the field. We identify four econometric challenges, only one of which the current difference-in-differences tool kit can resolve.
First, natural disasters are staggered treatments. They affect different zip codes at different times. It is not a simple binary switch that gets flipped for all the treated units simultaneously. Fortunately, the recent difference-in-differences literature proposes useful methods to explore this timing heterogeneity (for example, Callaway and Sant’Anna 2021; Goodman-Bacon 2021). Unfortunately, these methods do not address the other econometric challenges we describe later.
Second, the effects of natural disasters are likely not permanent. In fact, these effects are of indeterminate length. When the econometrician codes a unit as treated, should it stay treated for the entire posttreatment period? Certainly, as rebuilding occurs, the initial effects can dissipate. Perhaps this can be captured by time-varying effects, but this approach leads to a new problem.
This third problem is the temporal overlap of natural disasters. Sometimes, a zip code is hit by a second disaster in the same quarter. Sometimes, the second disaster occurs a quarter later. Sometimes, it occurs a year or more later. This overlap will be greater in the model if the post period is coded as lasting for longer. Thus, it matters greatly whether the effect is assumed to be permanent or to have a finite length. However, the existing difference-in-differences approaches do not address this potential temporal overlap, in which a treatment is not binary but rather varies depending on the extent of overlap.
If we grant the possibility of nonpermanent treatment effects, then we introduce a fourth problem: units switching in and out of treatment status. The state-of-the-art difference-in-differences models do not allow for units to be removed from treatment because they are trying to minimize the difficulty in understanding the control groups, which change over time with staggered treatments. The problem of fluid control groups is magnified by units switching between treatment and control status and back again.
To address these four problems, we construct two models. The first is a dosage-type model, as biostatisticians use to understand the effects of different dosages of a medical treatment such as a pharmaceutical drug. In such a case, there is essentially overlap of treatments—for example, where some patients receive one pill, others receive two, and so on. This is analogous to different zip codes experiencing different numbers of disasters in a given time period. In this approach, our model is similar to Boustan et al. (2020):
where Yijt is our outcome variable rents or vacancy rates, Disastersij,t–b is a series of variables counting the number of different kinds of disasters that occur in zip code i within state j at time t, δi is an indicator variable for geographic fixed effects, μt is an indicator variable for time fixed effects, Xijt is a vector of demographic control variables, and (Sj*t) is a series of state-specific linear time trends.
This model allows significant flexibility to vary the outcome variable (rents or vacancy rates), the geographic fixed effects (county or zip code), and the time fixed effects (quarter or year). We cluster standard errors at the level above the geographic fixed effects (county or state, respectively) to account for spatial correlation. In equation (1), we employ four quarterly lags to measure the effects of disasters up to one year afterward in a quarterly model; we will also show a specification with one lag when using annual data. Thus, we can address two econometric challenges with this model: the staggered treatments and the temporal overlap of treatments.
The second model takes aim at the problem of units switching in and out of treatment, as well as the unclear finite persistence of the posttreatment effect. We employ the new interactive fixed effects model developed by Liu and colleagues (2024), which allows for these possibilities. Similar to a synthetic control, it matches the pretreatment trends of the treated and control groups, so there is no need to be concerned about the parallel trends problem that so often plagues difference-in-differences designs.3 Unlike the synthetic control method, however, the interactive fixed effects model allows for multiple treated units.
Briefly, the estimation strategy predicts the counterfactual outcomes in the absence of the treatment—in this case, how rental outcomes would have been different for disaster-impacted zip codes if they had not experienced a disaster—using the observed and unobservable covariates of the untreated observations during our study period (2000–2020), where unobservable attributes are proxied by fixed effects. Specifically, it fits a model of the form:
where the new term λijft captures the unobservable attributes by assigning a time trend to each zip code and allowing the interactive coefficients to differ heterogeneously across the cross section.
Because this model follows a more familiar difference-in-differences type approach, the treatment variable takes the form of a post-disaster indicator, set to 1 in the periods after the disaster when we expect the zip code to be impacted. We show specifications where the treatment lasts for different time periods, but we also estimate a specification that is agnostic to this problem. Rather than trying to impose structure on the posttreatment period, this latter approach makes each treatment permanent but separates the first treatment to impact a unit from the second, third, and so forth. Thus, we can show the effect of the first disaster to strike a zip code during our time period, controlling for the second and third; the effect of the second disaster, controlling for the first and third; and so on. In all these specifications, the interactive fixed effects strategy uses the counterfactuals and the actual treated outcomes to estimate individualistic treatment effects for each unit-time (zip code-quarter) observation and averages across the units for dynamic posttreatment estimates.
It is important to note that CDBG-DR allocations are not random. Jurisdictions become eligible only after presidential disaster declarations and state or local applications verified by the US Department of Housing and Urban Development (HUD), typically prioritizing areas with the greatest documented damage and administrative capacity to implement large grants. These factors may correlate with both pre-disaster rental dynamics and post-disaster reconstruction potential. To partially address this concern, our interactive fixed-effects framework matches CDBG-DR zip codes to comparable untreated zip codes within the same state that share similar pre-disaster rent trajectories. The persistence of differences, if they exist, even after this adjustment suggests—but does not conclusively prove—a mitigating effect of CDBG-DR. Future research with detailed project-level grant data could further test for selection bias using propensity-score or synthetic-control designs.
RESULTS
In this section, we operationalize our two models using two outcome variables: rent per unit and vacancy rate. In each case, we run several different specifications, varying the data frequency as well as the geographic unit of analysis, to test the sensitivity of the model and provide a robust range of treatment effects. Throughout these models, we find sufficiently high F-statistics, R-squared, and degrees of freedom, indicating that the methodology works well in this application for these study areas. First, we present the dosage models. Second, we present the interactive fixed effects models. Finally, we break down the results depending on whether the zip codes received CDBG-DR funding after the disaster.
Dosage Model
Rents per unit change in very different ways depending on the type of disaster, as indicated by the coefficients in table 3. This model focuses on quarterly changes, in which columns (1) and (3) use zip code and quarter fixed effects for California and Florida, respectively, and columns (2) and (4) use zip code and year fixed effects for California and Florida, respectively. Across both specifications, some consistent patterns arise. Generally, natural disasters either have a significant positive effect or a null effect on rents, but there is very little evidence of negative effects. However, confirming hypotheses H1 and H2, there are significant differences across both disaster types and states. Wildfires have the strongest effect, increasing rents significantly in California. Hurricanes also have a significant positive effect in most specifications in Florida. Winter storms have a weakly positive effect in California but not in Florida.4 Tornadoes have a very weak positive effect in Florida. Earthquakes and flooding do not have significant positive or negative effects, consistent with the muted average effects suggested by Harwood (2025), where the decline in rental demand balances out the loss of rental supply.
Dosage Model: Quarterly Rent per Unit
Testing hypotheses H3 and H4, we combine the different disaster types into one general natural disasters variable in table 4. Consistent with H3, we find that disasters significantly increase rents by 0.2 to 0.6 percent per quarter in Florida and by 0.9 to 1.3 percent per quarter in California. This is consistent with a reduced supply, caused by damage and destruction to the properties, as well as homeowners shifting to renting as they update their risk beliefs, as suggested by Sheldon and Zhan (2019). Consistent with H4, these supply effects are more impactful in California than in Florida, consistent with the lower supply elasticities in California cities that generate greater housing shortages and more difficulty in rebuilding (Saiz 2010; Miller et al. 2026).
Testing hypotheses H3 and H4, we combine the different disaster types into one general natural disasters variable in table 4. Consistent with H3, we find that disasters significantly increase rents by 0.2 to 0.6 percent per quarter in Florida and by 0.9 to 1.3 percent per quarter in California. This is consistent with a reduced supply, caused by damage and destruction to the properties, as well as homeowners shifting to renting as they update their risk beliefs, as suggested by Sheldon and Zhan (2019). Consistent with H4, these supply effects are more impactful in California than in Florida, consistent with the lower supply elasticities in California cities that generate greater housing shortages and more difficulty in rebuilding (Saiz 2010; Miller et al. 2026).
Dosage Model: Quarterly Rent per Unit
Vacancy rate effects also depend on disaster type, as shown in table 5. These columns follow the same specification variations as table 3. The strongest effects here come from flooding and hurricanes, which lower vacancy rates, and wildfires and winter storms, which raise vacancy rates. The other disasters have less significant impacts. Table 6 confirms that natural disasters, taken together, have a mixed effect on vacancy rates, with most coefficients very close to zero. Consistent with hypothesis H5, rental increases are possible even when vacancy rates do not decrease significantly.
Dosage Model: Quarterly Vacancy Rate
Dosage Model: Quarterly Vacancy Rate
We test the robustness of these results in table 7 by aggregating all disasters by zip code annually and then running a one-lag model to estimate the effect of the current year and the previous year’s disasters on both rent per unit and vacancy rate. These results are very similar to the findings from the quarterly model. The effect on rents per unit is a significant increase of 0.5 percent per year in Florida and 1.2 to 1.3 percent per year in California, and vacancy rates do not change significantly.
Dosage Model: Annual Rent per Unit and Vacancy Rate
We also explore heterogeneity by historical disaster exposure. Some counties experience disasters frequently since the federal declaration dataset began in 1964, and others do not. Does it matter how much experience a county has with natural disasters when it is hit by the next one? To answer this question, we divide the dataset roughly into thirds based on the terciles of total disasters declared in each county. Table 8 shows the results for these three subsamples, classified as “low” frequency if they experienced fewer than twelve disasters, “medium” frequency if they experienced twelve to sixteen disasters, and “high” frequency if they experienced more than sixteen disasters. We run a quarterly dosage model with zip code and quarter fixed effects. The coefficients are largest and most significant for the medium and high frequency counties, suggesting that disasters have the greatest impact in counties where they are experienced most often. Rather than adapting to the repeated disasters, these counties seem to experience magnified effects from the compounded damage.
Dosage Model: Quarterly Rent per Unit by Disaster Frequency
Finally, we consider whether rent stabilization policies might have an impact on these post-disaster rent increases. In table 9, we run the quarterly dosage model using zip code and quarter fixed effects and include a dummy variable controlling for any time and place where rent stabilization policies were in place in California. The results are very similar to our original findings, suggesting that the impacts are not driven by these policies. Our model is robust to this control variable.
Dosage Model: Quarterly Rent per Unit with Rent Stabilization Control in California
Interactive Fixed Effects Model
How persistent are these effects over the long run? The interactive fixed effects model provides an answer by matching the treated and control groups in the pretreatment period and then estimating quarterly time-varying posttreatment effects for several years. Because this rich temporal variation can only be visualized with a separate graph for each regression, we aggregate across all disasters for this model, rather than showing the variation across every different type of disaster.
As outlined in the methodology section, there are two ways to estimate the treatment effect in this model. First, we identify each treatment as permanent without any units switching out of treated status; however, we have a separate treatment for the first, second, and third disaster to strike a given zip code. Thus, the temporal overlap is handled here by distinguishing between the number of successive disasters, rather than trying to make an assumption about how long the average treatment lasts.
Figure 2 shows each specification of this model in a separate line connecting quarterly average treatment effects for California (panel A) and Florida (panel B). The first disaster line indicates the effect of the initial event on a zip code, controlling for the second and third disaster. It reveals steadily increasing rents as soon as the disaster occurs, peaking around four years after the disaster at 6.5 percent higher than the original level in California and 12.5 percent higher in Florida. These patterns are consistent with our earlier dosage model, which suggests approximately a 1 percent increase per quarter immediately after the disaster in California and a 0.2 to 0.6 percent increase per quarter in Florida. Importantly, the rents do not return to their original level. In other words, the treated rents remain permanently higher than the control units’ rents, where no disasters occurred. The second disaster and third disaster lines show similar effects for subsequent events affecting a given zip code.5 These results are additive to the first disaster, as the model controls for its effects. Thus, consistent with H5, the effects of natural disasters on rents accumulate over time as multiple disasters strike the same area.
Interactive Fixed Effects Model: Rent per Unit, Sequential Disasters
Source: Authors’ calculations.
CDBG-DR Funding
CDBG-DR funding is one of the largest federal programs in disaster recovery. Since the program’s inception in 1993, until fiscal year 2023, nearly $100 billion has been allocated by Congress. As noted earlier, some zip codes receive CDBG-DR funding after a natural disaster while others do not. This funding often includes rental requirements that tie tenant protections and the building of affordable rental housing to program implementation, thus changing how the rental market might operate during the disaster recovery period. Brian An and colleagues (2024), for instance, documents four major rental requirements that are stipulated in the CDBG-DR programs: Action plan consideration; coordination with public housing authorities (PHAs); minimum amount set aside; and set affordability period.
For action plan consideration, states and localities receiving CDBG-DR funds must develop comprehensive action plans outlining how they will allocate resources to address the housing needs of both homeowners and renters. These plans serve as blueprints for disaster recovery efforts and are subject to HUD approval. Coordination with PHAs requires collaboration between CDBG-DR grantees and local PHAs, which is essential for ensuring efficient and equitable distribution of rental assistance and resources. By coordinating efforts, stakeholders can better identify and address the specific needs of low-income renters and vulnerable populations. The minimum amount set aside prioritizes rental housing development by requiring a minimum percentage of CDBG-DR funds to be allocated specifically for rental housing projects. Lastly, the set affordability period requires CDBG-DR grantees to designate a certain period to ensure that rental units developed with the funds remain affordable. This requirement ensures that rental housing remains accessible to low- and moderate-income households beyond the immediate post-disaster period.
Previous research shows that since 2005—when Hurricane Katrina hit New Orleans and its surrounding areas—these four rental requirements were stipulated in CDBG-DR funds, and particularly since 2010, all CDBG-DR grants had at least one of the rental requirements mentioned above (An et al. 2024). Using a case study approach, the analysis also finds that the increase in multifamily rents was less steep in zip codes that received CDBG-DR funds than in other nearby zip codes that were impacted by the natural disaster but did not receive the funds later.6 An and colleagues (2024) further show that one of the mechanisms driving such results is higher rates of construction of multifamily housing (measured by building permits) during disaster recovery among CDBG-DR zip codes compared with non-CDBG-DR zip codes. However, the referenced study is limited to a single case study, and we do not know how these funds alter the impact of natural disasters. To examine CDBG-DR funding’s likely impact on rental markets, we run the interactive fixed effects model in which the treatment group is confined to the zip codes receiving CDBG-DR funds. Following the previous approach, we focus on the first disaster for which zip codes receive these funds, controlling for the second and third such disasters.7
As in the previous results, figure 3 shows that these zip codes experience an increase in rents after the disaster in the “CDBG-DR Only” line. However, consistent with hypothesis H6, the long-term impacts are muted. In California, the CDBG-DR zip codes experience less rental appreciation than we see in our main specification throughout all post-disaster quarters. In Florida, rental appreciation is initially higher in the CDBG-DR zip codes; however, this appreciation peaks after three years and begins receding, while the main specification shows continued appreciation. This is consistent with the effects of increased supply of affordable housing coming online three to four years after the disaster. One of the possible reasons is more and faster construction of multifamily housing as the rental requirements in CDBG-DR often stipulate how much should be spent on affordable rental housing or how many years it should be kept affordable. These affordable rental housing developments, in coordination with local housing authorities, have to be part of comprehensive action plans for renters, not just homeowners. At the end of the four years, the cumulative rental appreciation is lower in the CDBG-DR zip codes than in the main specification.8 Support for the construction hypothesis is provided by Dania V. Francis and Keren M. Horn (2026, this issue), who find small increased upward trends in county-level employment (including short-term gains in construction) starting three quarters after an extreme hazard, with growth slowing by the twelfth quarter.
Interactive Fixed Effects Model with CDBG-DR Funding: Rent per Unit, First Disaster
Source: Authors’ calculations.
Given these rental requirements, we may see more and faster construction of affordable multifamily housing complexes in the areas that receive CDBG-DR funds than those that do not. If new multifamily apartments are operating as affordable housing in the disaster-impacted areas where people would have higher demand for affordable and secure housing, we expect that vacancy rates may go down over time. Whereas, if such affordable rental housing developments are not occurring (that is, areas without CDBG-DR funding), while residents should have urgent demand for housing—hence lower vacancy rates in the short run—over time, the rental market may experience higher vacancy rates due to unaffordable and insecure housing combined with rising rents. These expectations are consistent with our findings for all disaster-impacted areas and those split by the receipt of CDBG-DR funding.
Admittedly, our results cannot be interpreted as strict evidence of the effect of CDBG-DR funding, since it is possible that the zip codes receiving CDBG-DR funding are different in unobservable systematic ways that result in different trajectories regardless of the funding itself. While future research can help disentangle the strict causal impacts, our study presents clear evidence that renters experience a more affordable market in the wake of CDBG-DR funding than they do in its absence. This difference is not necessarily due to weaker rental markets, as these CDBG-DR funds come into markets with persistently lower vacancy rates. The greater affordability may be due to other factors, such as the federal aid, the rental requirements, or other interventions that prevent landlords from raising rents higher in the face of strong demand.
Potential Selection into CDBG-DR Funding
It is important to acknowledge that the allocation of CDBG-DR resources is not random. Federal and state governments jointly determine which jurisdictions receive disaster-recovery funds based on verified damage assessments, unmet needs analyses, and the administrative capacity of local governments to manage large grants. These selection criteria are designed to target the most heavily impacted areas but may also correlate with underlying housing-market characteristics—such as pre-disaster rent levels, local development intensity, preexisting planned public and private investment, or the socioeconomic composition of residents—that could influence post-disaster trajectories independently of the funding itself. Consequently, a concern arises that the observed moderation of rent increases in CDBG-DR locations may reflect government selection rather than the causal effects of the funding.
To mitigate this concern, our interactive fixed-effects framework compares CDBG-DR recipient zip codes to non-recipient zip codes within the same state that share similar pre-disaster rental trends. This approach reduces bias from unobserved, time-invariant factors and from differing baseline trajectories, although it does not eliminate the possibility of residual endogeneity from time-varying unobservables. As an additional diagnostic, online appendix table A.1 reports pretreatment characteristics—poverty rate, population size, vacancy rate, and average rent—for CDBG-DR and non-recipient zip codes.9 The two groups appear broadly comparable along these observable dimensions, providing further reassurance that selection on observables is limited.
We therefore interpret our CDBG-DR findings as strongly suggestive but not definitively causal evidence that federally funded recovery programs with rental requirements can dampen post-disaster rent escalation. Future work linking project-level allocations and timing to micro-market data could further disentangle causal effects from selection dynamics.
CONCLUSION
In this article, we have explored the trajectory of rental housing markets in the wake of a wide variety of natural disasters across two of the largest US states: California and Florida. We have merged zip code-level rental data from CoStar with disaster incidence data from FEMA. We have constructed two panel models—a dosage model and an interactive fixed effects model—that estimate the effect of natural disasters for many years after each disaster event. We have varied the lag structure, the time frequency, the unit of analysis, and the treatment definition for robustness, resulting in several different specifications for each outcome. Our results show that rents per unit increase by 0.2 to 1.3 percent immediately after a disaster and 6.5 to 12.5 percent in the long run, while effects on vacancy rates are mixed. This suggests that the decrease in supply, caused by damage and destruction of the housing stock, as well as increases in rental demand from homeowners who update their risk beliefs and exit homeownership, rising in insurance costs, and changes in the composition of housing units, outweigh the decrease in demand from out-migration. These findings are driven primarily by hurricanes and wildfires—and to a lesser extent, winter storms. The effects accumulate after multiple disasters. The rental appreciation is persistent and not mean-reverting. It is milder, however, in the presence of CDBG-DR funding with rental requirements designed to protect tenants, suggesting more routine access to CDBG-DR funds—such as through a formula-based allocation parametrically triggered by a disaster—could be broadly beneficial to impacted communities and their renters.
There are important differences between the disaster impacts in California versus Florida. Some disasters, like hurricanes and wildfires, only occur in one state or the other, but others, like flooding and winter storms, are experienced by both. In these comparable situations, the results sometimes differ. Winter storms, for instance, increase rents significantly in California but not in Florida. Thus, not all of our findings are generalizable across all types of cities and states. The preexisting rental market conditions, the local and state housing policies, and the demographics all matter, and these factors deserve more careful attention in future work. While California and Florida provide valuable contrasts in hazard type and housing-supply elasticity, extrapolation to other regions should be made with caution, as exposure, governance capacity, and rental market dynamics differ elsewhere. Nonetheless, these heterogeneous effects do not alter the general conclusion that natural disasters have significant effects on rental markets, most often harmful effects for affordability.
The implications of these findings are important for policymakers interested in preventing inequality from widening in the wake of these disasters. While governments have been increasingly proactive with disaster assistance over the past century, it is not clear whether these interventions have been sufficient to overcome the lingering negative impacts of the disasters (Boustan et al. 2012; Boustan et al. 2018). Nor does it appear that governments at the local, state, or federal level have been able to prevent the displacement of residents or encourage optimal take-up of insurance and other mitigation precautions (Berke et al. 2014; Gallagher 2014; Highfield et al. 2014; Kousky et al. 2020). Most residents do not even seem to be aware of the resources that are available for them to improve their situation (Hamel et al. 2018). To the extent that they are aware of government interventions, residents often feel that their actual needs are given less priority than other political concerns (Ford et al. 2019). By exploring the dynamics in rental markets, we hope to bring their needs to the forefront of public debate and help policymakers shape more targeted solutions to the socioeconomic inequities that they face.
FOOTNOTES
↵1. All these disasters have been officially declared at both the state and federal levels, and impacted areas are shown at the county level.
↵2. The coefficients do not change significantly when we change from effective rent per unit to effective rent per square foot, indicating that the results do not depend on the disasters’ systematic impact on units of smaller or larger sizes.
↵3. Carlos Martín and colleagues (2023) use this more simplistic difference-in-differences model and find very similar results to our findings here. However, the interactive fixed effects model is far more robust and rigorous for the reasons mentioned in this section, and it reveals more persistent long-term impacts.
↵4. It may surprise some readers to see winter storms in Florida, but they do occur. In February 2001, for instance, FEMA declared a major disaster due to a “severe freeze” event that impacted most of the state.
↵5. Over this twenty-year period, 98 percent of the zip codes in our California and Florida dataset are impacted by at least one disaster, 98 percent are impacted by at least two disasters, and 94 percent are impacted by three or more disasters.
↵6. FEMA disaster declaration DR-4145-CO (2013 Colorado floods).
↵7. We do not focus on the effect of the second and third disasters because the sample size is too small for zip codes that receive these funds multiple times.
↵8. However, we do not track rental requirements of CDBG-DR funds in current research as it is beyond the scope of our study. We encourage future research to examine this issue further.
↵9. The online appendix for table A.1 can be found at https://www.rsfjournal.org/content/12/4/54/tab-supplemental.
- © 2026 Russell Sage Foundation. An, Brian Y., Andrew Jakabovics, Anthony W. Orlando, Seva Rodnyansky, and Raphael W. Bostic. 2026. “The Trajectory of Rental Housing in the Wake of Natural Disasters: Evidence from California and Florida.” RSF: The Russell Sage Foundation Journal of the Social Sciences 12(4): 54–76. https://doi.org/10.7758/RSF.2026.12.4.03. We thank seminar participants at the Russell Sage Foundation, Fannie Mae, and the Urban Economics Association for helpful comments. We are especially grateful to Jessica Dill and the Center for Housing and Policy at the Federal Reserve Bank of Atlanta for research assistance and feedback. The funders had no involvement in the conduct of this research or the preparation of this article. All errors are the authors’ alone. Direct correspondence to: Anthony W. Orlando, aworlando@cpp.edu, 1380 Cresthaven Drive, Pasadena, CA 91105, United States.
Open Access Policy: RSF: The Russell Sage Foundation Journal of the Social Sciences is an open access journal. This article is published under a Creative Commons Attribution-NonCommercial-NoDerivs 3.0 Unported License.
REFERENCES
- ↵An, Brian, Jennifer Moody, Rachel Drew, Andrew Jakabovics, Anthony Orlando, and Seva Rodnyansky. 2024. “Community Development Block Grants Disaster Recovery, Rental Requirements, and Rental Market Impacts.” Cityscape 26(3): 137–156. https://www.huduser.gov/portal/periodicals/cityscape/vol26num3/ch8.pdf.
- ↵Baar, Kenneth, Patrick Burns, Daniel Flaming, and Anthony W. Orlando. 2024. Equitable Rent: Rent Stabilization Standards in the City of Los Angeles. Economic Roundtable. September. https://economicrt.org/wp-content/uploads/2024/09/Equitable-Rent-Economic-Roundtable-report-on-the-RSO-9-12-2024-compact.pdf.
- ↵Berke, Philip R., Ward Lyles, and Gavin Smith. 2014. “Impacts of Federal and State Hazard Mitigation Policies on Local Land Use Policy.” Journal of Planning Education and Research 34(1): 60–76.
- ↵Bleemer, Zachary, and Wilbert van der Klaauw. 2019. “Long-Run Net Distributionary Effects of Federal Disaster Insurance: The Case of Hurricane Katrina.” Journal of Urban Economics 110: 70–88.
- ↵Boustan, Leah Platt, Matthew E. Kahn, and Paul W. Rhode. 2012. “Moving to Higher Ground: Migration Response to Natural Disasters in the Early Twentieth Century.” American Economic Review 102(3): 238–44.
- ↵Boustan, Leah Platt, Matthew E. Kahn, Paul W. Rhode, and Maria Luca Yanguas. 2020. “The Effect of Natural Disasters on Economic Activity in US Counties: A Century of Data.” Journal of Urban Economics 118: 103257.
- ↵Callaway, Brantly, and Pedro H. C. Sant’Anna. 2021. “Difference-in-Differences with Multiple Time Periods.” Journal of Econometrics 225(2): 200–30.
- ↵Cheung, Ron, Daniel Wetherell, and Stephan Whitaker. 2018. “Induced Earthquakes and Housing Markets: Evidence from Oklahoma.” Regional Science and Urban Economics 69: 153–66.
- ↵Deryugina, Tatyana. 2017. “The Fiscal Cost of Hurricanes: Disaster Aid Versus Social Insurance.” American Economic Journal: Economic Policy 9(3): 168–98.
- ↵Deryugina, Tatyana, Laura Kawano, and Steven Levitt. 2018. “The Economic Impact of Hurricane Katrina on Its Victims: Evidence from Individual Tax Returns.” American Economic Journal: Applied Economics 10(2): 202–33.
- ↵Eichholtz, Piet, Eva Steiner, and Erkan Yönder. 2019. “Where, When and How Do Sophisticated Investors Respond to Flood Risk?” Unpublished paper. https://ecommons.cornell.edu/server/api/core/bitstreams/d6dc0bb4-6fc2-4d25-b3ac-88e2a2b8a6a3/content.
- ↵Ellen, Ingrid Gould, Katharine Harwood, and Rachel Meltzer. 2026. “Tale of Two Storms: Neighborhood Racial Change After Hurricanes Sandy and Harvey.” RSF: The Russell Sage Foundation Journal of the Social Sciences 12(4): 77–101. https://doi.org/10.7758/RSF.2026.12.4.04.
- ↵Ford, Rebecca M., Andrea Rawluk, and Kathryn J. H. Williams. 2019. “Managing Values in Disaster Planning: Current Strategies, Challenges and Opportunities for Incorporating Values of the Public.” Land Use Policy 81: 131–42.
- ↵Francis, Dania V., and Keren M. Horn. 2026. “Impacts of Climate Hazards on Employment and Earnings by Hazard Type, Industry, Race, and Ethnicity.” RSF: The Russell Sage Foundation Journal of the Social Sciences 12(4): 134–55. https://doi.org/10.7758/RSF.2026.12.4.06.
- ↵Gallagher, Justin. 2014. “Learning About an Infrequent Event: Evidence from Flood Insurance Take-Up in the United States.” American Economic Journal: Applied Economics 6(3): 206–33.
- ↵Gallagher, Justin, and Daniel Hartley. 2017. “Household Finance After a Natural Disaster: The Case of Hurricane Katrina.” American Economic Journal: Economic Policy 9(3): 199–228.
- Gatzlaff, Dean, Kathleen McCullough, Lorilee Medders, and Charles M. Nyce. 2018. “The Impact of Hurricane Mitigation Features and Inspection Information on House Prices.” Journal of Real Estate Finance and Economics 57(4): 566–91.
- ↵Ghidoni, Riccardo, Giacomo Calzolari, and Marco Casari. 2017. “Climate Change: Behavioral Responses from Extreme Events and Delayed Damages.” Energy Economics 68(S1): 103–15.
- ↵Glaeser, Edward, and Joseph Gyourko. 2018. “The Economic Implications of Housing Supply.” Journal of Economic Perspectives 32(1): 3–30.
- ↵Goodman-Bacon, Andrew. 2021. “Difference-in-Differences with Variation in Treatment Timing.” Journal of Econometrics 225(2): 254–77.
- ↵Hamel, Liz, Bryan Wu, Mollyann Brodie, Shao-Chee Sim, and Elena Marks. 2018. One Year After the Storm: Texas Gulf Coast Residents’ Views and Experiences with Hurricane Harvey Recovery. Kaiser Family Foundation; Episcopal Health Foundation. August. https://files.kff.org/attachment/Report-One-Year-After-the-Storm-Texas-Gulf-Coast-Residents-Views-and-Experiences-with-Hurricane-Harvey-Recovery.
- ↵Harwood, Katherine WH. 2025. “The Impact of Natural Disasters on Rental Markets: Heterogeneous Effects, Rental Subsidies, and Equity in Disaster Recovery.” Unpublished paper. http://papers.kateharwood.net/The%20Impact%20of%20Natural%20Disasters%20on%20Rents_2_7_25.pdf.
- ↵Highfield, Wesley E., Walter Gillis Peacock, and Shannon Van Zandt. 2014. “Mitigation Planning: Why Hazard Exposure, Structural Vulnerability, and Social Vulnerability Matter.” Journal of Planning Education and Research 34(3): 287–300.
- ↵Howell, Junia, and James R. Elliott. 2019. “Damages Done: The Longitudinal Impacts of Natural Hazards on Wealth Inequality in the United States.” Social Problems 66(3): 448–67.
- ↵Kousky, Carolyn, Howard Kunreuther, Michael LaCour-Little, and Susan Wachter. 2020. “Flood Risk and the U.S. Housing Market.” Journal of Housing Research 29(sup1): S3–S24.
- ↵Lambrou, Nicole, Crystal Kolden, and Anastasia Loukaitou-Sideris. 2025. “Disaster Recovery Gentrification in Post-Wildfire Landscapes: The Case of Paradise, CA.” International Journal of Disaster Risk Reduction 118: 105235.
- ↵Liu, Licheng, Ye Wang, and Yiqing Xu. 2024. “A Practical Guide to Counterfactual Estimators for Causal Inference with Time-Series Cross-Sectional Data.” American Journal of Political Science 68(1): 160–76.
- ↵Martín, Carlos, Brian An, Rachel Drew, et al. 2023. Disasters and the Rental Housing Community: Setting a Research and Policy Agenda. Brookings Institution. October. https://www.brookings.edu/wp-content/uploads/2023/09/Disasters-and-the-Rental-Housing_final.pdf.
- ↵Miller, Wesley A., Anthony W. Orlando, and Christian L. Redfearn. 2026. “Housing Supply Elasticities: A Structural Vector Autoregression Approach.” Unpublished paper.
- ↵National Multifamily Housing Council. 2023. Skyrocketing Insurance Costs Are Affecting Rents and Housing Affordability. National Multifamily Housing Council. https://www.nmhc.org/globalassets/research--insight/research-reports/insurance/nmhc_mf_risk_flyer.pdf.
- ↵National Oceanic and Atmospheric Administration (NOAA) 2024a. “Billion-Dollar Weather and Climate Disasters: Florida Summary.” National Centers for Environmental Information. https://www.ncei.noaa.gov/access/billions/state-summary/FL.
- ↵National Oceanic and Atmospheric Administration (NOAA) 2024b. “Billion-Dollar Weather and Climate Disasters: California Summary.” National Centers for Environmental Information. https://www.ncei.noaa.gov/access/billions/state-summary/CA.
- ↵Patap, Ryan. 2025. “Pasadena Apartment Availability Nosedives Because of Wildfires.” CoStar, January 31. https://www.costar.com/article/160018192/pasadena-apartment-availability-nosedives-because-of-wildfires.
- ↵Rhodes, Anna, and Max Besbris. 2022. Soaking the Middle Class: Suburban Inequality and Recovery from Disaster. Russell Sage Foundation.
- ↵Sado-Inamura, Yukako, and Kensuke Fukushi. 2019. “Empirical Analysis of Flood Risk Perception Using Historical Data in Tokyo.” Land Use Policy 82: 13–29.
- ↵Saiz, Albert. 2010. “The Geographic Determinants of Housing Supply.” Quarterly Journal of Economics 125(3): 1253–96.
- ↵Sheldon, Tamara L., and Crystal Zhan. 2019. “The Impact of Natural Disasters on US Home Ownership.” Journal of the Association of Environmental and Resource Economists 6(6): 1169–203.
- ↵Siegel, Rachel, Andrew Ba Tran, Federica Cocco, and Caroline O’Donovan. 2025. “Rent Rose by 20 Percent Across L.A. County After Fires. That’s Illegal.” Washington Post, January 25. https://www.washingtonpost.com/business/2025/01/25/la-county-rent-gouging-wildfires/.
- ↵Varela, Ana Varela. 2023. “Surge of Inequality: How Different Neighborhoods React to Flooding.” Working Paper. https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4396481.
- ↵Zhang, Lei, and Tammy Leonard. 2019. “Flood Hazards Impact on Neighborhood House Prices.” Journal of Real Estate Finance and Economics 58(4): 656–74.
