Flood insurance is a driver of population growth in European floodplains

The study addresses how population exposure to riverine flooding evolves when household settlement decisions respond to environmental (amenities and flood risk) and institutional (insurance) factors. Traditional climate risk assessments often treat exposure and vulnerability as exogenous and static, relying on top-down projections (e.g., SSPs) coupled with changing hazards. Recent work emphasizes the dynamic nature of exposure and vulnerability and the need to incorporate human behavior. The authors aim to integrate a bottom-up behavioral model of household location choice with large-scale flood hazard modeling to more accurately project floodplain population growth and consequent flood risk across Europe. They hypothesize that flood insurance—especially when premiums are not risk-reflective—can offset the disamenity of flood risk, encourage settlement in floodplains, and increase future risk, whereas risk-based premiums may better signal risk and temper exposure growth.

The paper situates itself in literature that distinguishes top-down risk models using static exposure and vulnerability from bottom-up approaches that consider behavioral responses over time and space. Prior global flood risk studies show socio-economic growth as a dominant driver of increasing risk, yet typically downscale population using generic suitability maps insensitive to flood risk and insurance. Empirical and policy literature notes that amenities (aesthetic and recreational qualities) of river proximity can attract settlement, while flood risk and protection measures influence location choices. Insurance design has been debated: the US NFIP has been criticized for encouraging development in high-risk areas via subsidized coverage; in Europe, countries like France, Belgium, and Spain apply flat or bundled systems promoting solidarity and cross-subsidization, while Germany, the UK, and others strive for risk-based pricing and incentives for household-level adaptation. Studies also document how risk-based premiums can stimulate mitigation, and how regulatory measures (land-use planning, managed retreat) and infrastructure interact with exposure dynamics.

The authors extend the Dynamic Integrated Flood Insurance (DIFI) framework to simulate household settlement decisions and their effects on exposure and flood risk at NUTS3 level in Europe. The workflow links: (1) flood hazard/exposure/vulnerability modeling (GLOFRIS), (2) insurance premium estimation, and (3) a household decision model for settlement location capturing environmental amenities, flood risk, and insurance. - Population growth inputs: Exogenous regional growth from SSP scenarios (SSP1, SSP2, SSP5) downscaled via the 2UP model provide total population change per NUTS3. The analysis applies only to regions with positive population growth (declining regions omitted). Household-level iterations are generated by converting population growth to new households (Eurostat household size data). - Choice framework: Each new household chooses between a location inside the 100-year floodplain (1% AEP) and higher ground. The decision maximizes subjective expected utility (SEU) over a 15-year residence horizon, considering three strategies if choosing floodplain: (s1) no action, (s2) insure, (s3) install a household disaster risk reduction (DRR) measure. The floodplain SEU accounts for two states: flood occurrence with probability p_i (derived from regional protection standards from FLOPROS; e.g., 1/100 implies p=0.01) and no flood with probability 1−p_i. Wealth W_ijt (a fraction of income, scaled with regional GDP under SSPs) is adjusted by river amenity value A_ijt and by flood losses L_ijt in the flood state. Utility is logarithmic (constant relative risk aversion). To reflect behavioral biases, flood probability and impact perceptions are modified with parameters β (probability overweighting; N(5,0.5)) and γ (impact underestimation; N(0.8,0.2)). Annual flood probability is multiplied by the 15-year residence horizon. The SEU on high ground includes amenities but no riverine flood loss; whether a premium is paid depends on insurance system (no premium in risk-based systems for zero risk; flat-rate premium applies to all households). - Insurance strategies: For s2, the household pays an aggregated 15-year premium π_ijt; in a flood, insured loss is reduced to a deductible (α=15% of damage). Two stylized national systems are modeled: (a) risk-based (premiums reflect average risk of households within the floodplain at NUTS3; no premium outside the floodplain), used in countries like UK, Ireland, Sweden, Netherlands, etc.; (b) flat-rate (country-averaged EAD divided by number of households), used in France, Spain, Belgium, implying cross-subsidization and premiums insensitive to individual risk. - DRR strategies: For s3, the household pays a one-off cost C and reduces expected perceived flood losses by δ. Two options are considered: dry flood-proofing (C≈€471, δ≈13%, effective up to limited inundation depths) and wet flood-proofing (C≈€2,389, δ≈25%). - River amenities: Amenity values from hedonic pricing literature decline with distance to the river. The model selects a random potential location inside the floodplain and one on higher ground using a 1D transformation of regional area by floodplain share. A piecewise linear amenity function A(x_i) assigns higher values closer to the river with diminishing marginal effects and a cutoff beyond 50 km. Amenity values are scaled by national housing price deviations (θ_i from Eurostat) and adjusted by projected GDP ratios over time. - Flood risk modeling: GLOFRIS provides inundation maps at multiple return periods (2–1000 years) and time points (2010, 2030, 2050, 2080) using HadGEM under RCP4.5 (baseline) and RCP8.5 (sensitivity). Exposure focuses on residential assets (assumed 75% of built environment). Vulnerability uses depth-damage curves. Expected annual damage (EAD) is the area under the probability-impact curve above protection thresholds (FLOPROS). - Simulation and outputs: The simulation runs in 5-year steps from 2010 to 2050 (8 steps), updating flood risk and premiums (interpolated) and household wealth. For each region and step, the share of new households choosing the floodplain determines flood exposure growth. Modified floodplain population projections then proportionally adjust damage per return period, and EAD is recomputed to estimate future flood risk under scenarios: baseline (2UP), amenities+DRR (no insurance), and amenities+DRR+insurance (status-quo and counterfactual risk-based for flat-rate countries). Outputs are aggregated to NUTS3 and compared as factors vs baseline.

- Relative to baseline (2UP SSP2), accounting for environmental (dis)amenities and household DRR (no insurance) yields mixed effects: median deviation factor ≈1.1 (slightly higher floodplain population overall) but many regions show lower floodplain growth where flood risk outweighs amenities (e.g., central Sweden, Ireland). Low-risk regions show higher growth (e.g., parts of Spain, Wales, Northern Scotland). - Adding insurance availability substantially increases floodplain population growth. Across regions, the mean deviation factor vs baseline is ≈2.5, with some regions up to 10× higher growth. Insurance can flip outcomes from below-baseline to above-baseline (e.g., Deux-Sévres, Cuenca, West-Surrey). - Insurance design matters: Flat-rate systems (France, Belgium, Spain) show consistently and substantially higher floodplain growth versus baseline; risk-based systems show smaller deviations and outcomes closer to the amenities-only case. Typical 2050 average annual premiums: ≈€13 per household in flat-rate countries vs ≈€400 in risk-based countries (Sweden, Ireland, UK). In the Netherlands and other risk-based countries, many regions remain near or below baseline despite insurance availability. - Flood risk (EAD) amplification can exceed exposure changes: deviations vs baseline reach up to 30× in some regions when jointly considering hazard and exposure changes, compared to up to 10× for exposure alone. Under the amenities+DRR case (no insurance), 80% of regions have lower EAD than baseline, though averages are slightly higher due to regions with large increases. - Insurance availability increases EAD relative to baseline in many regions, especially in flat-rate countries. Average additional EAD growth through 2050 attributable to insurance availability: France ≈€3.9 billion, Belgium ≈€440 million, Spain ≈€1.3 billion. - Converting flat-rate countries to risk-based premiums reduces the insurance-driven rise in EAD by more than 50% (France, Belgium, Spain), aligning regional patterns more closely with local risk. - Eastern Europe shows high increases in EAD despite declining exposed populations under baseline; e.g., Poland’s EAD doubles to ≈€900 million by 2050 with a 13% decline in exposed population, driven by increasing hazard, lower protection standards (e.g., Romania ~1/50 vs France 1/100, Netherlands 1/1000), and faster GDP growth increasing asset values. - Sensitivity: A more severe climate scenario (RCP8.5) with higher population growth (SSP5) raises EAD on average more than baseline, but the average rise in EAD due to more severe climate change through 2050 (~250%) is still less than the rise in EAD due to exposure growth caused by insurance coverage (~360%).

Findings demonstrate that exposure is endogenous to policy and environmental signals: flood risk deters settlement, while insurance—especially with cross-subsidized, flat-rate pricing—offsets the disamenity of risk and attracts households into floodplains, increasing future EAD. Risk-based pricing better aligns private costs with local risk, discouraging new settlement in high-risk zones and moderating exposure and risk growth. Results imply that traditional top-down risk assessments with exogenous exposure and vulnerability can misestimate future risk by ignoring behavioral feedbacks. Policy significance is twofold: (1) insurance market design is a lever for risk communication and adaptation incentives; (2) combining insurance reform with land-use planning, building standards, and protection investments can form robust adaptation portfolios. While higher physical protection standards can yield larger risk reductions, they are costly and can induce safe-development feedbacks; pricing reforms are comparatively cost-effective and can complement hard measures. The study underscores the need for detailed risk mapping and addressing barriers (e.g., bundled policies, premium caps, expectations of ad hoc public compensation) to implement risk-based pricing while preserving affordability via targeted, means-tested support and subsidies for DRR.

The study introduces a behaviorally informed, bottom-up settlement model integrated with flood hazard and insurance to project floodplain exposure and risk across Europe. It shows that insurance availability, particularly under flat-rate, cross-subsidized systems, is a strong driver of increased settlement in floodplains and higher EAD. Shifting to risk-based premiums substantially reduces insurance-driven exposure and risk increases, positioning insurance design as an effective adaptation instrument. The work highlights the limitations of assessments that hold exposure and vulnerability exogenous and provides a framework for integrating behavioral dynamics into large-scale risk models. Future research should extend the framework to include within-region relocations (managed retreat and post-event migration), broader vulnerability dynamics, multiple hazard models/GCMs, finer-scale risk-based pricing, and richer representations of risk perception and amenity heterogeneity.

- Scope limits: Analysis excludes regions with declining populations and does not simulate within-region relocations between floodplain and high ground, potentially omitting important migration responses. - Modeling assumptions: SEU with logarithmic utility, fixed 15-year residence horizon without discounting, and simplified behavioral misperception parameters (β, γ) may not capture full heterogeneity in risk attitudes and perceptions. - Insurance stylization: National assignment to two stylized systems (flat-rate vs risk-based) and risk-based premiums averaged at NUTS3 imply residual cross-subsidization and do not reflect property-level pricing, caps, deductibles, or reinsurance dynamics in detail. - Amenities generalization: Hedonic amenity function is generalized across Europe, scaled by national housing prices and GDP, which may under/overestimate local amenity values. - Hazard/exposure simplifications: Focus on riverine flooding (excludes coastal/pluvial interactions), uses GLOFRIS with a single GCM (HadGEM) for main results and specific protection standards; results may vary with alternative models or datasets. - Data constraints: Risk-based pricing requires detailed flood risk maps often unavailable; some economic and asset valuation assumptions (e.g., residential share, GDP scaling) introduce uncertainty.