Adjusting to the New Normal: River Flood Risk and the Real Estate Market Henri Gruhl; Johannes Brehm; Arijit Ghosh; Philipp Breidenbach; Nils aus dem Moore Environmental and Resource Economics · 2025-07-18 CC BY 4.0 · https://creativecommons.org/licenses/by/4.0/ Source: https://link.springer.com/article/10.1007/s10640-025-00995-x Abstract Do individuals learn from salient river floods and update their risk perceptions accordingly? We exploit the Western European flood in 2021 to estimate changes in property values across German flood risk regions. Flood-prone areas in proximity to the flood or with higher climate change beliefs experience housing price declines, while we observe no statistically significant effects in other flood risk zones. Our results imply that flood risk discounts in the housing market will not adjust uniformly, which may result in suboptimal capital allocation in areas at risk of future flooding. Introduction If the global mean temperature increases by \(1.5^{\circ}\) C compared to pre-industrial temperatures, global human fatalities from river flooding are estimated to rise by 70–83% and flood damages by 160–240% without further adaptation (Dottori et al. 2018 ). According to the most optimistic IPCC pathway, it is highly likely that average global warming will cross the threshold of \(1.5^{\circ}\) C, with temperatures likely exceeding this level by 2100 (IPCC Working Group I 2021 ). Even though the change in river flooding risk is heterogeneous across regions (Blöschl et al. 2019 ), losses from river (fluvial) floods propagate through global trade and supply networks, affecting even less directly exposed regions (Willner et al. 2018b ). In addition to continued mitigation efforts, considerable flood adaptation efforts are needed to limit the damages caused by increasing fluvial flood risk all over the world (Willner et al. 2018a ). However, the literature suggests that the increasing risk of climate change is currently underestimated by markets. On the one hand, studies measuring flood risk valuation in housing prices find economically insufficient discounts (e.g., Bernstein et al. 2019 ; Murfin and Spiegel 2020 ; Hino and Burke 2021 ; Bakkensen and Barrage 2022 ; Gourevitch et al. 2023 ) Footnote 1 , Footnote 2 and show that flood risk discounts depend on people’s belief in climate change (Baldauf et al. 2020 ; Bakkensen and Barrage 2022 ; Gourevitch et al. 2023 ). Further Wagner ( 2022 ) finds that demand for flood insurance in the Eastern United States is low even though prices are below the expected payouts. This insufficient risk capitalization cannot be rationalized by future climate policy scenarios since even present flood risk is underestimated (Bakkensen and Barrage 2022 ). On the other hand empirical studies find increases in risk discounts or flood protection premiums after flood events (Hallstrom and Smith 2005 ; Kousky 2010 ; Bin and Landry 2013 ; Ortega and Taṣpınar 2018 ; Wolf and Takeuchi 2022 ) and the dissemination of new flood information (Gibson and Mullins 2020 ) that tend to disappear over time (e.g. Atreya et al. 2013 ; Bin and Landry 2013 ; Beltrán et al. 2019 ). These findings are unlikely to be based on changes in fundamentals, but consistent with flood risk learning when flood risk is undervalued at present (Bakkensen and Barrage 2022 ). Yet only if flood risk is reflected in prices, excessive development in areas at risk of flooding is discouraged by the markets (Hino and Burke 2021 ). Footnote 3 If the risk is not priced in, future losses from floods are amplified (Steinhausen et al. 2022 ). Therefore, adjustments of local flood risk valuation to rising global temperatures are crucial to limit future flood damages, and it is essential to understand if and how people update their risk valuation. Against this backdrop, this paper investigates whether and how fluvial flood risk valuations are updated in the real estate market after a large river flooding event. In a Bayesian learning model, home buyers are expected to update their expectation of future flood risk after observing a flooding event (DeGroot 1970 ; Viscusi 1991 ; Gallagher 2014 ). The underlying hypothesis of our study thus posits that exposure to a nonlocal flooding event, fostered by extensive media coverage, can influence homebuyers’ perceptions regarding future flood risk. To test this hypothesis, we use the most comprehensive dataset on the German housing market and administrative flood risk maps to identify properties located within and outside of areas vulnerable to future fluvial flooding. We employ a hedonic pricing model with a difference-in-differences design, leveraging the Western European flood in July 2021 as an exogenous shock to estimate any changes in risk perceptions across Germany. The flash flood – caused by heavy precipitation between July 12 and 19, 2021 – serves as a suitable exogenous event to test our hypothesis, resulting in 189 deaths Footnote 4 and estimated damages exceeding EUR 33 billion across public, residential, commercial, and industrial sectors and infrastructure (Koks et al. 2021 , Munich 2022 ). Given these impacts, the flood received significant media attention and was salient across Germany (see Supplementary Material A ). Furthermore, the flood’s intensity and likelihood were attributed to climate change (Tradowsky et al. 2023 ), a connection widely reported in the media (Figs. A.1, A.2). We focus on risk-updating in non-inundated regions, which allows us to disentangle the changes in flood risk valuation from direct flood damages and reduce the influence of damages to local amenities, a considerable challenge for papers investigating property price changes in damaged regions. To that end, we exclude inundated regions identified utilizing high-resolution satellite imagery of the flood. We observe substantial increases in the valuation of flood risk in certain regions. However, averaged across all flood risk regions in Germany, the flood did not lead to a statistically significant change. Our results highlight that regionally heterogeneous secondary impacts of climate change related events must be carefully modeled since average effects might otherwise fail to detect statistically meaningful changes. We find that flood risk areas within close proximity to the inundated regions experienced a significant decrease in house prices, with a decline of approximately 5.9%. However, this effect decreases as the distance from the flooded regions increases. Furthermore, we observe that areas with a relatively higher share of climate change believers display a reduction in house prices of approximately 3% in flood-prone regions. Notably, regions with a below-median share of climate change believers exhibit no change in their valuation of flood risk. We perform several robustness checks to examine the sensitivity of the results to spatial spillovers, choices in the identification strategy, and different trends across regions which confirm the paper’s findings. The observed heterogeneous patterns in flood risk updates have broader implications. This persistent variation in risk perceptions can lead to suboptimal capital allocation, potentially resulting in overinvestment in flood-prone areas (Harrison and Kreps 1978 ; Scheinkman and Xiong 2003 ; Bakkensen and Barrage 2022 ). Ultimately, these suboptimal investments may lead to excessive future losses and damages, exacerbated by climate change (Hino and Burke 2021 ). This paper contributes to the literature on the relationship between flood risk and the real estate market in several ways. First, to the best of our knowledge, we are the first to study risk updating in the real estate market of an entire country following a flood event. Beltrán et al. ( 2019 ) document that prices of inundated properties across England decrease following flooding. We extend this evidence by examining non-inundated properties to isolate the effect on nationwide risk perceptions. The few existing studies on risk updating in non-inundated regions predominantly focus on “near misses,” referring to non-damaged properties located within the same floodplain, and yield unclear evidence (Hallstrom and Smith 2005 ; Ortega and Taṣpınar 2018 ; Hennighausen and Suter 2020 ). Moreover, the existing literature predominantly focuses on coastal flood risk, whereas fluvial flood risk is relatively understudied (Hennighausen and Suter 2020 ). Looking at fluvial flood risk, it is also essential to broaden the scope to an entire country, as fluvial floods not only affect coastal regions, as is the case with sea level rise, storm surges, or hurricanes, but can affect several regions in the country. Analyzing narrowly non-inundated properties, Hennighausen and Suter ( 2020 ) discover that property prices decrease for flood-damaged buildings but increase for ʼnear misses’ in a Colorado case study. In contrast, two studies observing regions in the United States find that property prices decrease after hurricane-caused flooding for “near misses” and non-affected but nearby counties, respectively (Hallstrom and Smith 2005 ; Ortega and Taṣpınar 2018 ). Looking beyond “near-misses,” Muller and Hopkins ( 2019 ) analyze property prices in New Jersey flood risk regions and find that distant but well-publicized hurricanes have a negative property price effect if the town participates in public flood awareness campaigns. Indicative evidence on risk updating in non-damaged regions after a natural disaster also comes from studies showing an increase in insurance take-up (Gallagher 2014 ; Kamiya and Yanase 2019 ) and an increase in capitalization rates of flood protection dams (Wolf and Takeuchi 2022 ). Our second contribution focuses on examining the distance-based heterogeneity in risk updating following environmental disasters. The existing body of literature has employed indicators other than hedonic property pricing models to uncover risk updating (e.g., investment patterns, insurance take-up) and presented mixed evidence. For instance, Alok et al. ( 2020 ) analyze professional money managers and discover an increasing aversion towards stocks located in disaster zones as the distance to a climate disaster decreases. In contrast, Gallagher ( 2014 ) find impacts on flood insurance take-up of non-flooded communities in the same media markets as the flood, but no significant impact of distance on flood insurance uptake within these local television markets. Similarly, Kamiya and Yanase ( 2019 ) did not find evidence that physical distance significantly predicts increased insurance uptake following earthquakes in Japan. We show that following the 2021 European flood, the prices of properties listed for sale in flood-risk areas decreased significantly near the flooded regions, with the price effect diminishing as the distance from the flood increases. We argue that our distance-based findings are likely not only related to regional differences in information about the flood, as we find evidence that information about the flood spread throughout Germany. Therefore, we argue that the distance-based heterogeneity in the treatment effect may also result from availability bias (Tversky and Kahneman 1973 ): the memory of the flood is more salient in close proximity, and individuals may be more likely to have personal contacts in inundated regions. Lastly, this paper contributes to the nascent literature on climate change belief heterogeneity and its impact on risk discounts in property prices. Our study builds on the findings of previous studies conducted by Barrage and Furst ( 2019 ), Bernstein et al. ( 2019 ), Baldauf et al. ( 2020 ), and Bakkensen and Barrage ( 2022 ). These studies demonstrated that house price discounts due to rising natural hazard risks under climate change are predominantly observed in communities with a high level of belief in climate change (Bernstein et al. 2019 ; Baldauf et al. 2020 ; Bakkensen and Barrage 2022 ). Additionally, they highlight that new developments in coastal risk zones are more likely to be reduced in areas where the majority of the population believes in climate change (Barrage and Furst 2019 ). Since these studies focus on baseline differences in coastal communities and long-term sea level rise, our study adds to the literature by providing a potential mechanism and extending the evidence from coastal flooding due to sea level rise to river flooding from flash floods. We observe risk updating in regions where belief in climate change is relatively high. These adaptations of flood risk discounts could explain the systematic differences in discounts between regions. Moreover, it enriches the US-based literature by providing the first evidence from a large European case study, characterized by a different distribution of climate change beliefs (Dechezleprêtre et al. 2025 ). The remainder of this paper is organized as follows. Section 2 presents our data and the setup of the empirical model described in Section 3 . Section 4 provides the main results and additional analyses to uncover the heterogeneity underlying the average effect. Section 5 concludes. Data 2.1 Flood Risk Maps We use a federal flood risk map to identify properties at risk of fluvial flooding in Germany. This map, which forms the basis of our study’s treatment assignment, was obtained from the German Federal Institute of Hydrology (BFG et al. 2020 ). We use the latest map version from 2019. The federal flood risk maps were not updated until the end of our study period, ensuring that our treatment is not confounded by new flood risk information from these maps. Since 2007, the EU Floods Directive requires the uniform application of flood risk management in the European Union ( 2007 /60/EC). Due to the federal system in Germany, the responsibility for flood protection lies with the states and is regulated in the Federal Water Act (Wasserhaushaltsgesetz - WHG). The 16 federal states are in charge of designing strategies and measures to prevent flood damages, including the assessment of flood risks and designation of risk areas (§ 73 WHG), as well as preparing flood hazard and risk maps (§ 74 WHG) (Umweltbundesamt 2022 ). This data is assembled by the German Federal Institute of Hydrology and displayed in the common flood risk zone map for Germany, which we use in this study. These flood risk maps are publicly accessible online. Figure A.5 provides descriptives about the salience of the federal flood risk maps. Property sellers have an obligation to inform potential buyers proactively about known risks, such as the danger of flooding. Footnote 5 Sellers must disclose essential information, such as the property’s location in a flood risk zone, that could influence the buyer’s decision to avoid legal disputes. The map consists of spatial polygons for both fluvial and coastal flood risks. Flood risk levels are categorized as low, medium, and high, corresponding to the probabilities of a flood occurring within the next 200 years, 100 years, and 10–25 years, respectively. The probabilities are calculated based on recommendations by the German government and consider data on the hydrology, topography, and roughness of the terrain (for more details, see Bund/Länder-Arbeitsgemeinschaft Wasser ( 2010 )). Since our treatment is linked to extreme precipitation causing flash floods, it differs qualitatively from coastal floods, which are more susceptible to storm surges or sea-level rise and less affected by heavy precipitation. Consequently, we exclude coastal flood risk areas and focus on fluvial flood risk only. Footnote 6 Panel A of Fig. 1 displays the low-, medium-, and high-flood risk zones in Germany. They can be found nationwide, with a notable concentration along the major rivers. In particular, they extend along different sections of the Rhine (west), Weser (northwest), Danube (south), Elbe (northeast), and Oder (east). Nevertheless, flood-prone areas can also be identified in various other locations throughout the country, caused by, for example, small water bodies nestled in valleys prone to flooding during periods of heavy precipitation. 2.2 High-Resolution Satellite Flooding Data To identify the changes in the valuation of flood risk in the flood risk zones, we need to disentangle the changes in risk valuation from the damages caused by the flood, as both might reduce house prices. This allows us to speak to a considerable body of empirical research that discusses updated flood risk valuation as a potential reason next to damages for observed housing price declines in flooded regions (Bin and Polasky 2004 ; Kousky 2010 ; Atreya et al. 2013 ; Bin and Landry 2013 ; Atreya and Ferreira 2015 ; Beltrán et al. 2018 ; Zhang and Leonard 2019 ; Gibson and Mullins 2020 ; Mutlu et al. 2023 ). To correctly exclude directly affected areas, we obtained high-resolution satellite data on the extent of the flood and building-level impact from the European Commission’s Copernicus Emergency Management Service ( 2021 ) (CEMS). The CEMS compiles geospatial data based on satellite and aerial imagery in the case of natural disasters. For the Western European flood in July 2021, the CEMS created the EMSR517 dataset for Germany, which comprises multiple shapefiles, each containing an area of flood damage. The shapefiles include layers showing areas that experienced heavy rain and were suspected to be damaged, inundated areas as well as damaged buildings. Footnote 7 All levels are identified based on the visual inspection of high-resolution satellite images, which allows the CEMS to distinguish between buildings that have been directly destroyed or damaged and those that are only located in affected regions. In Panels A and B of Fig. 1 , the shaded areas indicate the areas that experienced heavy rain and were suspected to be damaged by the flood. In our main specification, we opt for this broader definition to account for observation errors and negative spillovers from the destroyed infrastructure. By broadly excluding these directly affected regions, we ensure that any potential effects stem solely from changes in risk perception rather than physical flood damage. This also minimizes the influence of depreciation of local public goods and amenities. In our robustness section ( 4.2 ), we address potential measurement error in these satellite images by excluding areas less conservatively (only areas with observed floods/damages) and more conservatively (all counties fully or partially impacted). 2.3 Housing Data We analyze the housing market using the RWI-GEO-RED real estate dataset (Schaffner and Thiel 2022 ) from the Research Data Center Ruhr at the RWI (FDZ Ruhr). This dataset provides comprehensive information on the German real estate market sourced from the country’s leading online real estate portal, ImmobilienScout24 . In addition to the asking price in euros, the dataset includes data on the date of the advertisement’s posting and removal, as well as property-specific features, such as the number of rooms, living area, building condition, and year of construction. Since the data originate from a collaboration with ImmobilienScout24 , we can access back-end information, such as the precise geo-coordinates of house offers. Our sample includes single-family and some multi-family houses listed for sale, excluding individual apartments. While transaction prices would be preferable to listing prices, as they better reflect market outcomes after negotiation adjustments (Bishop et al. 2020 ), our data is the most comprehensive available for Germany. Second, we choose to focus solely on the listings for sale of houses and not on listings for rentals for two reasons. First, long-term purchasing decisions are more likely to be influenced by long-term risk perceptions than are renting decisions (Bernstein et al. 2019 ). Second, in Germany, homeowners are more responsible for potential flood damage than are renters (IHK zu Essen 2023 ). To avoid duplication, we ensure that each property appears only once in our sample. This is achieved by including only the latest ad placement for each property. We argue that the most recent listing provides the most up-to-date information about the property and is closest to the transaction price. Additionally, using a single listing for sale is advantageous for event studies, as it focuses on a single marginal point in time, thereby avoiding the need to model shifts in the price function over time (Bishop et al. 2020 ). Although our main analyses exclude multiple listings of the same house, we also present results using an alternative setup that includes all houses listed at least twice, exploiting within-house price variation over time (Table B.4). 2.4 Treatment and Control Group Due to our access to the geocoordinates of houses listed for sale, we can merge the house price data with the German flood risk maps and satellite images. After excluding potentially flooded regions, our treatment group consists of houses located within one of the fluvial flood risk zones. To identify a causal effect in our analysis, we require a control group that closely resembles the treatment group, except for the flood risk status. For our main analysis, we select houses located within a three kilometer (km) distance from the nearest border of a flood risk zone as our control group. Given their proximity, this approach guarantees that regional trends affecting local housing prices are similar in both groups. To ensure the validity of our analysis, we need to address some potential issues that arise from defining the control group in close proximity to flood risk areas. One issue is that the population might not be aware of the exact location of the flood risk zones. Consequently, they may change their risk perception, even when not officially categorized as part of the treatment group. This may cause the control group to be affected in the same direction as the treatment group, leading to a potential underestimation of the effect. To address this concern, we draw a 0.5 km buffer around flood risk areas and exclude houses located inside that buffer. This helps to reduce the impact of spillover effects on the control group, ensuring that any differences in home prices between the control and treatment groups are plausibly due to the flood risk. Figure 2 displays an example of the treatment and control group in the Cologne area. The colored dots represent houses in flood risk zones (treatment group) while the grey dots denote houses between 0.5 and 3 km distance of the zones (control group). Our final dataset includes a pooled cross-section of approximately 1.5 million properties listed for sale between January 2014 and May 2022. Table 1 provides summary statistics. Footnote 8 Most of our covariates are available for almost all houses. These include common predictors of a property’s price, such as size, construction year, number of rooms, and the presence of a cellar. Additionally, for about half of the houses, we have information from the online platform about the quality of the facilities ranging from “simple” to “deluxe.” On average, houses in flood-risk zones have lower listing prices, are larger and older, and tend to have less sophisticated facilities. Empirical Strategy We use a hedonic pricing model (Rosen 1974 ), following a well-established literature that employs this method to reveal individuals’ valuation of non-market characteristics (Greenstone 2017 ). The model attempts to isolate the house price component related to flood risk by controlling for the observable and unobservable characteristics of the house. Since we focus on changes in flood risk valuation following the flood, we employ a difference-in-differences (DiD) strategy to identify the average treatment effect on the treated (ATT). The model can then be expressed in the following form: where \(Y_{i,z,t}\) – the price per square meter \(Y\) of house \(i\) located in zip code \(z\) offered in year-month \(t\) – is regressed on the treatment effect \(flood\;risk_{i}{\times}post_{t}\) , the time-invariant treatment group indicator \(flood\;risk_{i}\) , a vector of house characteristics \({X}_{i}\) , a set of fixed effects on zip code level \(\lambda_{z}\) and year-month fixed effects \(\phi_{t}\) . The main coefficient of interest in our analysis is \(\alpha\) , which corresponds to the interaction between the indicator variable for being located in a flood risk area, denoted by \(flood\;risk_{i}\) , and the indicator variable for the post-flood period, denoted by \(post_{t}\) . Since \(post_{t}\) denotes the period following the Western European flood in July 2021, this interaction captures changes in risk perception in flood-prone areas. The house characteristics \({X}_{i}\) control for the construction year of the house (categorical), Footnote 9 the size of the living space, Footnote 10 the number of floors, the number of rooms, the presence of a cellar, and the conditions of the structure (categorical) Footnote 11 . We include separate indicators for characteristics with missing information. Zip code fixed effects capture all unobserved zip code-specific characteristics that are invariant over time. The average zip code in Germany covers 43.2 km \(^{2}\) . Following Abbott and Klaiber ( 2011 ), in addition to the zip code fixed effects, we include several different spatial fixed effects in our main estimations at the larger county Footnote 12 level or the smaller 1 km \(^{2}\) level. While fine-scale fixed effects effectively control for unobserved spatial heterogeneity, they may discard valuable between-variation, leading to incomplete estimates of broader capitalization effects (Abbott and Klaiber 2011 ). In contrast, coarse-scale fixed effects retain between-variation but risk omitted variable bias by failing to account for fine-scale heterogeneity. The inclusion of varying spatial fixed effects allows us to assess the sensitivity of the results, yielding information about capitalization of public goods and amenities at varying spacial scales. The year-month fixed effects control for all unobserved time-specific effects in a specific month and year, which are the same for all included regions in Germany, such as a general upward or downward shift in the German housing market. Furthermore, we also show specifications in which we control for county-specific time trends or relax the temporal fixed effects by only including year and month fixed effects, capturing temporal general changes between years and general seasonal changes between months. Additionally, we present results when we include county-specific quadratic trends, cubic trends, or county-by-year fixed effects. Standard errors are clustered at the zip code level. The identifying assumption in the canonical difference-in-differences framework is the parallel trends assumption. By comparing changes in outcomes over time between a treatment group (houses in flood risk zones) and a control group (houses outside flood risk zones but in close proximity), DiD designs assume that, in the absence of treatment, both groups would have developed similarly over time (parallel trends). Under this assumption, the effect is identified even if shocks affect the potential outcome as long as they are not correlated with the flood risk zones. Although it is impossible to test this assumption conclusively, it is common practice to perform an event study to show that there is no statistically significant effect on property prices between the groups in the period before the flood occurred. Methodologically, this concept can be expressed as follows: where \(\underline{t}\) is the first pre-treatment period for which we want to rule out anticipation effects, while \(\bar{t}\) represents the last post-treatment period for which we expect adjustment effects. We take \(\alpha_{-1}\) as the reference category. Therefore, the period-specific effects in the first summation (anticipatory effects) and the second summation (reactive effects) are interpreted relative to the period \(t-1\) . This approach enables us to assess whether there are any statistically significant changes in property prices between the treatment and control groups in the periods leading up to and following the flood event, and to quantify the magnitude of these effects over time. Figure 3 displays event studies based on Eq. ( 2 ). We do not detect any statistically significant pre-trends in the quarters leading up to the flood. We also do not find longer-term systematic pre-trends in the years before the flooding. The event studies support the parallel trends assumption, indicating that the potential outcomes for the treated and control groups follow parallel trends when conditioned on the specified model. This further strengthens our analysis, as it suggests that any changes in the outcomes after the flood event can be attributed to the flood and not to other external factors. For identification, we rely on intertemporal variation in prices. However, as noted by Kuminoff and Pope ( 2014 ), capitalization may not precisely reflect households’ willingness to pay if the gradient of the equilibrium price function shifts following the flood. Specifically, if the flood changes residential sorting in flood-risk areas, altering the hedonic gradient, panel variation may yield biased estimates of risk valuation updating. Therefore, our approach assumes a stable hedonic gradient over time. Since this assumption is difficult to test, the estimated effects should be interpreted with caution in terms of welfare. Results 4.1 Main Results Our analysis shows that after the flood, risk perceptions are updated in some regions of Germany while remaining unchanged in others. Specifically, we observe localized effects in regions near the flooding that are statistically significant at the one percent level, which diminish with distance, and significant effects at the five percent level in regions with higher baseline beliefs in climate change. Table 2 presents the average effect of the flood on house prices in flood risk regions that were not directly affected. We do not detect any statistically significant effects at the one, five or ten percent level across model specifications with varying restrictiveness. The result indicates that, on average, the flood did not lead to a statistically significant change in flood risk valuation across German flood risk zones. We run our model using five specifications. Column 1 estimates the effect according to Eq. ( 1 ). In Column 2, we loosen the restrictiveness of the model by removing the year-month fixed effect and only include year fixed effects for unobserved differences between years and month fixed effects to control for seasonality across years. Similarly, Column 3 allows for more spatial variation by expanding the spatial fixed effect to the county level. In Column 4, we restrict the model by including a 1 km \({^{2}}\) fixed effect, capturing more unobserved spatial differences, but in turn reducing variation in the estimation of the effect. Finally, Column 5 includes linear time trends at the county level to account for different trends between regions. While the average effect is a null result, statistically and economically significant effects may be averaged out across heterogeneous groups. For example, economic theory and empirical evidence from the behavioral economics literature suggest that the distance to an event may play a role in how economic agents react (Gallagher 2014 ; Alok et al. 2020 ). Moreover, empirical evidence – primarily from the United States – suggests that climate change belief heterogeneities could play a role in how flood risk information, e.g., induced by a natural disaster, is processed (Barrage and Furst 2019 ; Bernstein et al. 2019 ; Baldauf et al. 2020 ; Bakkensen and Barrage 2022 ). To assess the role of proximity, we interact the treatment effect with a continuous measure of the distance of a property to the closest area that was flooded in July 2021. Our findings, presented in Panel A of Table 3 , show an economically meaningful negative effect of the flood on nearby houses across various specifications that is statistically significant at the one percent level (Column 5). The statistically significant (one percent level) positive interaction term indicates that the discount on house prices decreases as houses are located farther away from the flood. Specifically, the square meter price of at-risk houses immediately next to the flood decreases by about 5.9% (Column 5) compared to houses outside flood risk zones. The effect then decreases by approximately 0.03% points with each kilometer distance to the flood. While the direct comparison of effect sizes from previous studies is challenging due to differences in research settings, designs, and flooding types, our findings align most closely with Ortega and Taṣpınar ( 2018 ). The authors document an 8% price penalty for flood zone properties not directly damaged by Hurricane Sandy. Hallstrom and Smith ( 2005 ) find that Hurricane Andrew conveyed new risk information, causing property prices in high-risk storm zones to decline by at least 19%. Gibson and Mullins ( 2020 ) report an 11% price reduction for properties newly designated as being in a flood zone after narrowly escaping damage from Hurricane Sandy, attributing this reduction to belief updating. These findings are consistent with our proposed learning mechanism increasing perceived flood risk. Since distance can be categorized in various ways with varying assumptions about the functional form, we employ various alternative measures to test the robustness of our main finding. Figure B.1 illustrates the distribution of distances between properties listed for sale and the flooded area. We observe a large concentration of properties near the flooded areas, which gradually declines with distance, a pattern driven by the high population and property density in the state of North-Rhine Westphalia. To avoid relying on a linear functional form assumption, we employ alternative distance bins rather than a continuous measure. This approach effectively captures potential non-linearities and thresholds in the relationship. Table B.1 displays the results relying on the median split, which creates two distance bins, allowing us to estimate effects independently for properties above and below the median distance to the flooded areas, without imposing any functional form. The result holds: we find a statistically significant negative effect for properties geographically close to the flood at the one percent level, but no significant effect for properties located above the median distance (with \(\alpha=0.05\) ). Further, Fig. 4 displays the event studies for the median split-based distance heterogeneity, which show no systematic pre-trends. Figure B.2 displays the event studies for 50 kilometer distance bins. Due to decreasing precision from smaller sample sizes in each distance bin, we refrain from interpreting the precise location where the distance-based effect reaches zero. Nonetheless, Fig. B.2 displays a pattern where the effect size is largest for those properties listed for sale closest to the flood and then tapering off. Smaller bins generally introduce more noise due to fewer observations, this is particularly the case for the distance bins far from the flooded area (see Figs. B.1 and B.2). Consequently, we exclude regions that are more than 400 kilometers away from the flooded areas and re-estimate the main distance result. Table B.2 shows that the results remain statistically significant at the one percent level and are slightly smaller than the main estimates. Lastly, we take inspiration from Callaway et al. ( 2024 ) and treat the distance-based effect as a continuous treatment. Comparing houses in flood risk areas relatively close to the flood to flood risk areas far from the flood does not display significant pre-trends (with \(\alpha=0.05\) ) and identifies an imprecise effect of around ten percent (Fig. B.3). These distance-based results may be attributed to greater amount of information about the flood in regions closer to the inundated areas, and cognitive biases. Geographically closer regions likely experienced more media coverage of the event through local news outlets. They were also more likely to be exposed to rainfall during the event. Both likely increased interest in and knowledge of the event. A heightened interest is supported by Fig. A. 3, which display regional Google search intensity in Germany for 2021. The data reveals that the relative search activity for terms related to flooding and flood insurance was highest in states directly affected and bordering states. In addition to pure information effects, our results also indicate that other channels, such as cognitive social biases, might play a significant role in the observed effects. The fact that the treatment effect completely vanishes as the distance increases, even though the flood was extremely salient across all of Germany (see Figs. A.4, A.1 and A.2), suggests that other cognitive biases influenced the risk updating of individuals who live close to the flood. They likely have a higher degree of attachment to the affected regions and know more people who were directly affected by the flood. These individual memories, attachments, and connections likely play a significant role along with more information and higher rain exposure close to the flood. People make probability judgments by recalling experiences from memory, which can lead to several biases affecting many economic decisions (Tversky and Kahneman 1974 ; Bordalo et al. 2023 ). One major concept that may partly explain our results is the availability bias : if an instance is relatively easier to recall, it is assigned a higher probability (Tversky and Kahneman 1973 ). People who are geographically close to a disaster tend to remember it more easily. They have more vivid and emotional memories of it. People with these more salient memories of an event tend to overestimate how likely it is to happen again (Tversky and Kahneman 1973 ). This difference - how present an event is to some individuals compared to others - can have economic consequences. People’s perception of risk is further shaped by how they evaluate and emotionally respond to such salient events. As Slovic ( 1987 ) highlights, individuals often rely on intuitive judgments and emotional reactions when assessing risks, which can amplify their perceived likelihood of rare but memorable events. For example, Bordalo et al. ( 2012 ) model this further, by including the salience of infrequent negative events which lead to greater weight in decision-making, contributing to risk-averse behavior toward assets exposed to such risks. Alok et al. ( 2020 ) empirically studied the impact of climate disasters on professional money managers’ investment behavior. They find that managers’ aversion to stocks in disaster-affected areas decreases as their distance from past climate disasters increases. They provide evidence that this heterogeneity is due to the availability bias of managers located close to the disaster, rather than a difference in information about the disaster. In our context, the more salient memories close to the flooding may explain part of the importance of distance to the flood. The less salient memories away from the flooding could play a role in the lack of impacts that we observe for larger distances. Another possible bias explaining the results relates to people’s tendency to misconceive relatively small sequences or samples as representative (Tversky and Kahneman 1971 ). Therefore, this bias could lead to individuals believing that a wider region, which has already been affected by a natural disaster, such as flooding, is more likely to be hit again in the future than other regions. This perception may explain why the prices of flood-prone houses decreased only in regions close to the flood. Next, to investigate the hypothesis that the effects also depend on belief in climate change, we utilize data on the county-level share of people believing in human-made climate change from Levi et al. ( 2023 ). While our climate change belief measure may be endogenous and subject to measurement error, we attempt to mitigate these issues by using a pre-treatment measure and a sample split. Specifically, we split properties into two groups based on the median climate change belief level in 2019: one group with houses in areas above the median and another with houses in areas below the median. We use a median split rather than a continuous measure to avoid false precision at a smaller geographic scale. Therefore, we focus on the relative differences between regions, that are likely to hold with more fine-grained data. We assume that the distance between below and above median climate change belief remains stationary over time. By comparing the housing market outcomes of these two groups with an interaction term, we aim to identify any heterogeneity in the risk updating effects after the flood that is associated with climate change beliefs. Panel B of Table 3 presents the climate change belief heterogeneity results. By interacting the treatment effect with a dummy indicating above-median climate change beliefs, we find that the average null effect holds for regions with a relatively small share of people believing in human-made climate change. The square meter price of houses in flood risk zones does not change in statistically significant terms (with \(\alpha=0.05\) ) in these regions compared to houses outside of flood risk zones. However, in most specifications, we observe a statistically significant negative effect on house prices in areas where climate change beliefs are above the median (significant at the five percent level in Column 5). The effect on the square meter price of at-risk houses in regions with relatively above-median belief in climate change decreases by about 3.8% (Column 5) compared to those in regions with relatively low belief. Figure 4 displays the event studies for the regional belief in human-made climate change, which show no statistically significant pre-trends. Notably, the statistical significance of the effects does not reach conventional levels when using the most restrictive spatial fixed effect (1 km \(^{2}\) fixed effects) in Column (4). Further, in the repeat listings specification (Robustness section, Table B.4), the coefficients are not statistically significant. This may be driven by insufficient precision when measuring climate change beliefs and warrants further investigation. Repeatedly listed properties require at least one relatively early listing that might not reflect beliefs closer to 2019. These findings suggest that, similar to findings from the United States, beliefs about climate change play an important role in people’s flood risk valuation. These results advance the findings in the literature in two ways. First, it provides evidence that differences in climate change beliefs are also relevant in contexts outside the United States and for flood risk unrelated to sea level rise. Second, we show that belief in climate change not only explains undervaluation at baseline (Baldauf et al. 2020 ; Bakkensen and Barrage 2022 ; Gourevitch et al. 2023 ) but also how flood risk valuation is updated after a flood. This mechanism may explain the different levels of valuation depending on beliefs in climate change if regions with relatively higher beliefs consistently updated their risk perceptions after flooding events. The fall in property prices and the increase in flood insurance take-up in damaged areas are usually observed to be temporary and vanish after some years. The number of years varies from one (Shao et al. 2022 ) to three (Atreya et al. 2015 ), to five years (Bin and Landry 2013 ; Beltrán et al. 2019 ), up to nine years (Atreya et al. 2013 ). Similar to the importance of distance, this temporality is also often (partly) explained by the idea of an “availability bias.” Following Tversky and Kahneman ( 1973 ), people attribute more weight to events that they have recent and vivid memories of. Therefore, we test whether the effects differ over time. We jointly estimate the treatment effect and interaction term for each post-treatment quarter (see Table B.7 and B.8). However, we do not find consistent evidence that the effects in the responding regions – regions close to the flood or with a relatively higher belief in climate change – decrease after ten months. Due to the recency of the flood and data limitations, we cannot test whether the effect decreases after a longer period of several years, as found in the aforementioned empirical studies (Bin and Landry 2013 ; Atreya et al. 2015 ; Beltrán et al. 2019 ; Shao et al. 2022 ). The distance to the flooding and regional belief in climate change may be correlated with each other. Figure A.6 displays the houses in our main sample and the distance to the flooded areas in kilometers as well as whether they lie within regions with above median belief in human-made climate change. The correlation of the two heterogeneity variables is 0.25. To assess how much this correlation conflates our results and whether regions with below median climate change belief update risk perception at all, we estimate the effect of the flood based on distance for a subsample of only those regions with low belief in climate change (Table B.6). We find similar results with statistically significant effects at the one percent level close to the flood that decrease with distance. Furthermore, given that the variables might be correlated with other dimensions, we aim to investigate whether the observed heterogeneous effects are influenced by socioeconomic factors that vary across Germany. Consequently, we examine whether effect heterogeneity exists concerning the economic affluence of regions, as well as the density of buildings and the share of unemployment. All variables are measured at the 1 km \(^{2}\) level. Table B.9 displays the results of these interactions. Our analysis reveals no statistically significant heterogeneous effects ( \(\alpha=0.05\) ) related to purchasing power per capita, suggesting that economic affluence does not strongly influence the observed effects. Similarly, we find no statistically significant effects ( \(\alpha=0.05\) ) with respect to the density of buildings or the share of unemployment. The flood could have also changed the number of house listings in different regions in Germany. Therefore, we analyze in Table B.5 how the number of listings per zip code changed in response to the flood and whether this effect differs with the distance to the flooding or climate change belief. We find no statistically significant effect ( \(\alpha=0.05\) ) on the number of listed apartments. Additionally, we test whether there are heterogeneous effects based on the risk level of flood risk areas, ranging from low to high. While we observe patterns consistent with theoretical expectations – i.e. effect estimates increase with the risk level, they do not reach statistical significance at conventional levels (Table B.9). The floods’ effect in low-risk areas is not statistically significant ( \(\alpha=0.05\) ), and while the estimate for the medium-risk level is larger (more negative), it is not statistically significantly different ( \(\alpha=0.05\) ). Similarly, the coefficient for the high-risk areas is even larger but also not statistically different. The lack of statistical significance is likely due to the relatively smaller number of houses in the more severe flood-risk areas, especially those with the highest flood risk. Only 12.74% of the houses in flood risk zones are situated in high risk areas, which are expected to experience flooding every 10–25 years. 4.2 Robustness Checks We conduct several robustness checks to examine the sensitivity of the results to spillovers, choices in the identification strategy, choices in the econometric model, and different trends across regions (Table 4 ). First, we test whether spillovers affect our results. For instance, following a flood event, flood risk may also be perceived to be higher just outside the flood risk areas since the spatial demarcation of the flood risk zones may not be sharp in individuals’ purchasing decisions. These negative spillovers from the treatment to the control group may bias our treatment effect towards zero. In our main specification, we exclude a 0.5 km buffer zone between the treatment and control group to mitigate the influence of spillovers. Columns (1) and (2) of Table 4 present the results when the excluded buffer zone is widened to 1 km and 1.5 km, respectively. We observe slightly stronger negative effects with larger buffer sizes, indicating that our main results are likely lower-bound estimates of the true effect. Second, we examine the robustness of our results to variations in the choice of the control group. In our main analysis, we include houses located between 0.5–3 km away from the flood risk area in the control group to ensure that the trends are comparable to those of the treatment group. However, it may be argued that the control group should not include areas close to flood risk zones in which no houses are located. To address this concern, we define an alternative control group in Column (3) which consists of houses located more than 0.5 km away from a flood risk area, but within 3 km of the nearest house in the treatment group. Another potential issue could be that our main control group, which extends up to 3 km outside the flood risk areas, may include houses that are too distant and subject to dissimilar time trends. We restrict the control group in Column (4) to houses located between 0.5–2 km from the treatment area to test the robustness of our results to this concern. For both alternative control groups, the results remain qualitatively the same. Third, excluding potentially flooded regions, which allows us to disentangle the effect of risk updating and physical damage, may be too generous or conservative. In our main analysis, we exclude flooded regions based on CEMSs relatively generous definition of “potentially impacted” regions (as shown in Figure 1 ), excluding substantial areas around flooded rivers. In Column (5), we restrict the definition of inundated areas and only exclude properties located within areas of confirmed flooding. This change slightly increases the observed treatment effect in the distance heterogeneity analysis, which aligns with our previous findings as this specification includes more properties close to the flood. On the other hand, even though we generously exclude areas around the flood in our main specification to assure the exclusion of all damaged properties, it may still be possible that we include areas that were indirectly affected by the physical consequences of the flood due to damages to local amenities. For instance, it may have been the case that public infrastructure or recreational spaces in close proximity to the property were destroyed or damaged. These physical flood impacts could have led to a reduction in nearby house prices. If these damages are correlated with flood risk areas, our analysis may capture not only changes in flood risk valuation, but also changes in amenities. To address this concern, Column (6) excludes entire counties if any part was flooded. Overall, the results from both alternatives are very similar in significance and magnitude to the main results. Fourth, we assess the robustness of our findings to alternative choices in the econometric model, specifically different types of regional trends and a model exploiting variation in repeated listings. Table B.3 presents results using not only county-specific linear trends, but also county-specific quadratic trends, cubic trends, or county-by-year fixed effects. The results remain largely consistent, except for the heterogeneous effects based on climate change beliefs, which become statistically insignificant at conventional levels when cubic trends or county-by-year fixed effects are included. Table B.4 presents analyses using houses listed for sale multiple times in our sample period, allowing for house fixed effects to control for time-invariant characteristics and exploit within-house price variation over time. The average effect of the flood remains consistent with our main findings, and the heterogeneous effects by distance are also robust. Specifically, we estimate a smaller but statistically significant (at the five percent level) negative effect of approximately 2.6% for houses closer to the flood, which decreases, statistically significant at the ten percent level, for houses farther away. However, the climate change belief-based heterogeneous effects are not robust in the repeat-listings approach, as we find no statistically significant effects or interactions ( \(\alpha=0.05\) ). These results indicate that differences related to climate change beliefs warrant further investigation across various contexts and models to validate the indicative evidence. Finally, we aim to ensure that our results are not sensitive to different time trends across regions in Germany. Although we control for county-specific time trends in the regressions, we confirm that we do not pick up any remaining regional differences by conducting placebo tests in time. In Columns (7) and (8), we estimate the effects of hypothetical floods that occurred one and two years before the actual flood, respectively. If our estimation strategy is biased by differences in time trends across space, we would observe (heterogeneous) treatment effects in these tests. However, we find no effects in either placebo exercise, indicating that our main results are not picking up on general regional developments. Conclusion This study investigates the impact of a significant fluvial flooding event on flood risk valuation in the German real estate market. Specifically, we focus on the Western European flood in July 2021. Using a hedonic pricing model and a difference-in-differences approach, we explore the changes in flood risk valuation across different flood risk areas in Germany. By excluding physical flood damages using satellite imagery, we disentangle changes in flood risk valuation from direct flood damages and find that, on average, flood risk valuation across flood-prone areas in Germany did not change. However, the average effect across the entire country masks substantial regional heterogeneity. First, house prices declined by approximately 5.9% in flood risk areas near inundated areas, with the effect diminishing as distance from the flooding increases. Second, house prices decrease by approximately 3% in flood-prone regions, with a relatively higher share of people believing in climate change. Regions with below-median belief in climate change show no statistically significant change in flood risk valuation. Our findings highlight the importance of considering regional differences in the flood risk valuation updating following fluvial floods. These findings suggest that flood risk undervaluation may persist despite increasing fluvial flood frequency, particularly in regions distant from past floods or with lower belief in human-induced climate change. This heterogeneity in flood risk perceptions may lead to suboptimal capital allocation in regions at risk of future flooding (Harrison and Kreps 1978 ; Scheinkman and Xiong 2003 ; Bakkensen and Barrage 2022 ). The lack of change in flood risk valuation in distant regions or areas with below-median climate change beliefs may necessitate government intervention. The policy implications hinge on diverging perspectives of who is responsible to adapt to future climate change: the government or the market. The government-centric perspective argues that governments should act by implementing protective measures against fluvial flash floods, enforcing land-use and zoning restrictions to discourage development in flood-prone areas, or mandating elementary insurance for at-risk properties. In contrast, the market-centric perspective assigns responsibility to individuals to internalize flood risk costs. However, mandatory flood risk disclosures or information campaigns may serve as soft-touch policy measures to address information gaps and facilitate informed decision-making (Gallagher 2014 ; Gibson and Mullins 2020 ). Future research could further explore heterogeneity in flood risk perception updates following fluvial floods. While our study provides comprehensive nationwide data, further research on other countries and flood events would be valuable. Our study focuses on house prices as an indicator of flood risk perception, rather than flood insurance or stated preferences, which merit further exploration in future research. Notes See Beltrán et al. ( 2018 ) for a meta-analysis of the earlier flood risk valuation literature up to May 2014. Bosker et al. ( 2019 ) provide evidence for an interesting exception. They show that flood risk valuation in the Netherlands is arguably larger than official expected flood damages if the government’s commitment to uphold the stated flood protection and damage bailout is credible. The Netherlands is a special case with one of the world’s best flood defenses, reducing the likelihood of flooding to up to once every 10,000 years. Excessive development refers to development where market signals underprice flood risk costs, resulting in an inefficient allocation of resources. In Germany, the death toll from water-related hazards is the highest since a storm surge in February 1962 along the North Sea Coast (Thieken et al. 2022 ). As clarified by the Federal Court of Justice (BGH) in ruling Az. V ZR 193/90. To provide indicative evidence that these official flood risk maps are salient to home buyers, Table A.1 shows that at baseline, before the 2021 flood, flood risk in these locations is associated with moderate price discounts. In the CEMS shapefiles, the three levels are called “area of interest,” “observed event,” and “built up,” respectively. We excluded extreme outlier observations in preparing the dataset. Specifically, we removed the top and bottom 1% of the selling price and living space distribution for each year, followed by the overall top and bottom 1% of the price per square meter distribution. The categories for the construction year are years “before 1900,” “1900–1945,” “1946–1959,” “1960–1969,” “1970–1979,” “1980–1989,” “1990–1999,” “2000–2009,” and “after 2009.” We control for property size on the right-hand side of the equation as we suspect a non-linear relationship between price per square meter and property size. In Germany, smaller properties tend to have higher price-per-square-meter values than larger ones, especially in urban settings. The categories for the condition of the house range from “simple” over “normal” and “sophisticated” to “deluxe”. Counties in Germany correspond to the NUTS-3 level and are called “Landkreise” or “Kreisfreie Städte”. There are 400 counties in Germany. References Abbott JK, Klaiber HA (2011) An embarrassment of riches: confronting omitted variable bias and multi-scale capitalization in hedonic price models. Rev Econ Stat 93(4):1331–1342 Article Google Scholar Alok S, Kumar N, Wermers R (2020) Do fund managers misestimate climatic disaster risk. Rev Financ Stud 33(3):1146–1183 Article Google Scholar Atreya A, Ferreira S (2015) Seeing is believing? Evidence from property prices in inundated areas. Risk Anal 35(5):828–848 Article Google Scholar Atreya A, Ferreira S, Kriesel W (2013) Forgetting the flood? An analysis of the flood risk discount over time. Land Econ 89(4):577–596 Article Google Scholar Atreya A, Ferreira S, Michel-Kerjan E (2015) What drives households to buy flood insurance? New evidence from Georgia. Ecol Econ 117:153–161 Article Google Scholar Bakkensen LA, Barrage L (2022) Going underwater? Flood risk belief heterogeneity and coastal home price dynamics. Rev Financ Stud 35(8):3666–3709 Article Google Scholar Baldauf M, Garlappi L, Yannelis C (2020) Does climate change affect real estate prices? Only if you believe in it. Rev Financ Stud 33(3):1256–1295 Article Google Scholar Barrage L, Furst J (2019) Housing investment, sea level rise, and climate change beliefs. Econ Lett 177:105–108 Article Google Scholar Beltrán A, Maddison D, Elliott R (2019) The impact of flooding on property prices: a repeat-sales approach. J Environ Econ Manage 95:62–86 Article Google Scholar Beltrán A, Maddison D, Elliott RJR (2018) Is flood risk capitalised into property values? Ecol Econ 146:668–685 Article Google Scholar Bernstein A, Gustafson MT, Lewis R (2019) Disaster on the horizon: the price effect of sea level rise. J Financ Econ 134(2):253–272 Article Google Scholar BFG, WasserBlick, and Zuständige Behörden der Länder (2020) German flood risk zones. https://geoportal.bafg.de/inspire/download/NZ/riskZone_flood/datasetfeed.xml . The German Federal Institute of Hydrology Bin O, Landry CE (2013) Changes in implicit flood risk premiums: empirical evidence from the housing market. J Environ Econ Manage 65(3):361–376 Article Google Scholar Bin O, Polasky S (2004) Effects of flood hazards on property values: evidence before and after hurricane Floyd. Land Econ 80(4):490–500 Article Google Scholar Bishop K, Kuminoff N, Banzhaf H, Boyle K, Gravenitz K, Pope J, Smith V, Timmins C (2020) Best practices in using hedonic property valu e models for welfare measurement. Rev Environ Econ Policy 43 Blöschl G, Hall J, Viglione A, Perdigão RA, Parajka J, Merz B, Lun D, Arheimer B, Aronica GT, Bilibashi A et al. (2019) Changing climate both increases and decreases European river floods. Nature 573(7772):108–111 Article Google Scholar Bordalo P, Conlon JJ, Gennaioli N, Kwon SY, Shleifer A (2023) Memory and probability. Q J Econ 138(1):265–311 Article Google Scholar Bordalo P, Gennaioli N, Shleifer A (2012) Salience theory of choice under risk. Q J Econ 127(3):1243–1285 Article Google Scholar Bosker M, Garretsen H, Marlet G, van Woerkens C (2019) Nether lands: evidence on the price and perception of rare natural disasters. J Eur Econ Assoc 17(2):413–453 Article Google Scholar Breidenbach P, Eilers L (2018) Rwi-geo-grid: socio-economic data on grid level. Jahrbücher für Nationalökonomie und Statistik 238(6):609–616 Article Google Scholar Bund/Länder-Arbeitsgemeinschaft Wasser (2010) Empfehlungen zur aufstellung von hochwassergefahrenkarten und hochwasserrisikokarten. https: //www.lawa.de/documents/hwgk15062010_1552299609.pdf. Accessed 29 Nov 2024 Callaway B, Goodman-Bacon A, Sant’Anna PH (2024) Difference-in-differences with a continuous treatment. Technical report. National Bureau of Economic Research Copernicus Emergency Management Service (2021) EMSR517: flood in Western Germany. http://emergency.copernicus.eu/mapping/list-of-components/EMSR517 . Accessed 4 May 2023 Dechezleprêtre A, Fabre A, Kruse T, Planterose B, Chico AS, Stantcheva S (2025) Fighting climate change: international attitudes toward climate policies. Technical report. National Bureau of Economic Research DeGroot MH (1970) Optimal statistical decisions. McGraw-Hill, New York Dottori F, Szewczyk W, Ciscar J-C, Zhao F, Alfieri L, Hirabayashi Y, Bianchi A, Mongelli I, Frieler K, Betts RA et al. (2018) Increased human and economic losses from river flooding with anthropogenic warming. Nat Clim Change 8(9):781–786 Article Google Scholar European Parliament and Council of the European Union (2007) Directive 2007/60/ec of the European parliament and of the council of 23 october 2007 on the assessment and management of flood risks. https://eur-lex.europa.eu/legal-content/EN/ALL/?uri=celex:32007L0060 Gallagher J (2014) Learning about an infrequent event: evidence from flood insurance take-up in the United States. Am Econ J Appl Econ 6(3):206–233 Article Google Scholar Gibson M, Mullins JT (2020) Climate risk and beliefs in New York flood plains. J Assoc Environ Resour Econ 7(6):1069–1111 Gourevitch JD, Kousky C, Liao Y, Nolte C, Pollack AB, Porter JR, Weill JA (2023) Unpriced climate risk and the potential consequences of overvaluation in us housing markets. Nat Clim Change 13(3):250–257 Article Google Scholar Greenstone M (2017) The continuing impact of sherwin rosen’s “hedonic prices and implicit markets: product differentiation in pure competition”. J Polit Econ 125(6):1891–1902 Article Google Scholar Hallstrom DG, Smith VK (2005) Market responses to hurricanes. J Environ Econ Manage 50(3):541–561 Article Google Scholar Harrison JM, Kreps DM (1978) Speculative investor behavior in a stock market with heterogeneous expectations. Q J Econ 92(2):323–336 Article Google Scholar Hennighausen H, Suter JF (2020) Flood risk perception in the housing market and the impact of a major flood event. Land Econ 96(3):366–383 Article Google Scholar Hino M, Burke M (2021) The effect of information about climate risk on property values. Proc Natl Acad Sci USA 118(17) IHK zu Essen (2023) Hochwasser– wer zahlt den schaden? es muss doch weitergehen. https://www.ihk.de/meo/recht-und-steuern/steuerrecht/downloads/beitraege/hochwasser-wer-zahlt-den-schaden-esmuss-doch-weitergehen-5199572 . Accessed 23 May 2023 IPCC Working Group I (2021) Climate change 2021: the physical science basis. Contribution of working group I to the sixth assessment report of the intergovernmental panel on climate change. In: Masson-Delmotte V, Zhai P, Pirani A, Connors SL, Péan C, Berger S, Caud N, Chen Y, Goldfarb L, Gomis MI, Huang M, Leitzell K, Lonnoy E, Matthews JBR, Maycock TK, Waterfield T, Yelekçi O, Yu R, Zhou B Technical report. Cambridge University Press, Cambridge, UK and New York, NY, USA. https://doi.org/10.1017/9781009157896 Kamiya S, Yanase N (2019) Learning from extreme catastrophes. J Risk Uncertain 59:85–124 Article Google Scholar Koks E, VanGinkel K, VanMarle M, Lemnitzer A (2021) Brief communication: critical infrastructure impacts of the 2021 mid-July western European flood event. Nat Hazards Earth Syst Sci Discuss 1–11 Kousky C (2010) Learning from extreme events: risk perceptions after the flood [Publisher: University of Wisconsin Press]. Land Econ 86(3):395–422 Article Google Scholar Kuminoff NV, Pope JC (2014) Do “capitalization effects” for public goods reveal the public’s willingness to pay? Int Econ Rev 55(4):1227–1250 Article Google Scholar Levi S, Wolf I, Sommer S, Howe PD (2023) Local support of climate change policies in Germany over time. Environ Res Lett 18(6):064046 Article Google Scholar Muller NZ, Hopkins CA (2019) Hurricane Katrina floods New Jersey: the role of information in the market response to flood risk. Working paper No. 25984, National Bureau of Economic Research. https://doi.org/10.3386/w25984 . Munich RE (2022) Hurricanes, cold waves, tornadoes: weather disasters in USA dominate natural disaster losses in. https://www.munichre.com/en/company/media-relations/media-information-and-corporate-news/mediainformation/2022/natural-disaster-losses-2021.html . Accessed 14 June 2022 Murfin J, Spiegel M (2020) Is the risk of sea level rise capitalized in residential real estate? Rev Financ Stud 33(3):1217–1255 Article Google Scholar Mutlu A, Roy D, Filatova T (2023) Capitalized value of evolving flood risks discount and nature-based solution premiums on property prices. Ecol Econ 205:107682 Article Google Scholar Ortega F, Taṣpınar S (2018) Rising sea levels and sinking property values: Hurricane Sandy and New York’s housing market. J Urban Econ 106:81–100 Article Google Scholar Rosen S (1974) Hedonic prices and implicit markets: product differentiation in pure competition. J Polit Econ 82(1):34–55 Article Google Scholar Schaffner S, Thiel P (2022) Fdz data description: real-estate data for Germany (rwi-geo-red v6. 1)-advertisements on the internet platform lmmobilienscout24 2007-12/2021. https://www.rwi-essen.de/forschung-und-beratung/fdz-ruhr/datenangebot/regionaldaten/rwi-geo-red . Research Data Center Ruhr (FDZ) Scheinkman JA, Xiong W (2003) Overconfidence and speculative bubbles. J Polit Econ 111(6):1183–1220 Article Google Scholar Shao J, Hoshino A, Nakaide S (2022) How do floods affect insurance demand? evidence from flood insurance take-up in Japan. Int J Disaster Risk Reduct 83:103424 Article Google Scholar Slovic P (1987) Perception of risk. Science 236(4799):280–285 Article Google Scholar Steinhausen M, Paprotny D, Dottori F, Sairam N, Mentaschi L, Alfieri L, Lüdtke S, Kreibich H, Schröter K (2022) Drivers of future fluvial flood risk change for residential buildings in europe. Glob Environ Change 76:102559 Article Google Scholar Thieken AH, Bubeck P, Heidenreich A, von Keyserlingk J, Dillenardt L, Otto A (2022) Performance of the flood warning system in Germany in July 2021: insights from affected residents. EGUsphere 1–26. https://doi.org/10.5194/egusphere-2022-244 . Tradowsky JS, Philip SY, Kreienkamp F, Kew SF, Lorenz P, Arrighi J, Bettmann T, Caluwaerts S, Chan SC, DeCruz L et al. (2023) Attribution of the heavy rainfall events leading to severe flooding in western europe during July 2021. Clim Change 176(7):90 Article Google Scholar Tversky A, Kahneman D (1971) Belief in the law of small numbers. Psychological Bull 76(2):105 Article Google Scholar Tversky A, Kahneman D (1973) Availability: a heuristic for judging frequency and probability. Cogn Psychol 5(2):207–232 Article Google Scholar Tversky A, Kahneman D (1974) Judgment under uncertainty: heuristics and biases: biases in judgments reveal some heuristics of thinking under uncertainty. Science 185(4157):1124–1131 Article Google Scholar Umweltbundesamt (2022) Identifying and managing flood risks. https://www.umweltbundesamt.de/en/topics/water/extreme-eventsclimate-change/floods/identifying-managing-flood-risks . Accessed 8 May 2023 Viscusi WK (1991) Economic theories of decision making under uncertainty: implications for policy analysis. In: Policy analysis and economics: developments, tensions, prospects. Springer, pp 85–109 Chapter Google Scholar Wagner KR (2022) Adaptation and adverse selection in markets for natural disaster insurance. Am Econ J Econ Policy 14(3):380–421 Article Google Scholar Willner SN, Levermann A, Zhao F, Frieler K (2018a) Adaptation required to preserve future high-end river flood risk at present levels. Sci Adv 4(1):eaao1914 Article Google Scholar Willner SN, Otto C, Levermann A (2018b) Global economic response to river floods. Nat Clim Change 8(7):594–598 Article Google Scholar Wolf D, Takeuchi K (2022) Holding back the storm: dam capitalization in residential and commercial property values. J Environ Econ Manage 116:102732 Article Google Scholar Zhang L, Leonard T (2019) Flood hazards impact on neighborhood house prices. J Real Estate Finance Econ 58:656–674 Article Google Scholar Zubayr C, Denise H, Hartmann L (2020) Nutzungsgewohnheiten und reichweiten im jahr 2019. Tendenzen im zuschauerverhalten. Media Perspektiven 50:110–125 Electronic supplementary material Below is the link to the electronic supplementary material. Supplementary Material 1 (download PDF ) Rights: © 2025 The Author(s) Rights and permissions Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ . Reprints and permissions Edition changes: Licensed narrative text edition. Original headings, paragraphs and references are preserved with normalized whitespace and layout. Figures, tables, display equations and separately licensed third-party material are not reproduced; use the linked original full text for those elements and definitive pagination. This is an attributed republication, not newly conducted research.