Do Hurricane Forecasts Meaningfully Change Damages and Disaster Spending?

Molina and Rudik (2024) examine how much hurricane forecasts reduce damage and improve decision-making. They study all U.S. hurricanes making landfall from 2005–2022, combining detailed county-level data on forecasted and actual wind speeds, federal protective spending, recovery costs, and damages. They find forecasts strongly influence pre-storm spending and that underestimating wind speed significantly raises damages. For example, a 10 m/s underforecast increases county damages by about $220 million and recovery costs by $20 million. Improvements in forecast accuracy since 2007 reduced total hurricane costs by about 19%, averaging roughly $2 billion in savings per hurricane.

Why This Article Was Selected for The Policy Scientist

The policy relevance of hurricane forecasting extends well beyond meteorology, as it directly shapes how governments allocate resources, protect populations, and limit large-scale economic losses in an era of rising climate risk. As extreme weather events become more frequent and costly, understanding the value of information itself becomes central to public policy design. Molina and Rudik, who have contributed to a growing literature on environmental risk and adaptation, provide timely evidence quantifying these informational benefits. Their novel county-level dataset is unusually rich and credible, enabling precise estimation of behavioral and fiscal responses. While the empirical strategy is rigorous, it relies on observational variation rather than causal identification; future work using quasi-experimental or experimental designs would strengthen inference and external validity across jurisdictions.

Full Citation and Link to Article

Molina, R., & Rudik, I. (forthcoming). The social value of hurricane forecasts. American Economic Journal: Economic Policy. https://www.aeaweb.org/articles?id=10.1257/pol.20240808

Central Research Question

Molina and Rudik investigate a fundamental question at the intersection of environmental economics and public policy: what is the economic value of hurricane forecasts, and how do improvements in forecast accuracy affect total social costs? The analysis is motivated by the premise that forecasts are not merely informational artifacts but active inputs into decision-making by governments and households. The authors focus specifically on whether forecasts influence protective expenditures prior to landfall, whether forecast errors translate into measurable increases in damages and recovery costs, and how to quantify the marginal value of improving forecast precision. The paper therefore integrates behavioral responses, realized damages, and ex ante uncertainty into a unified framework that treats forecasts as economically consequential policy tools.

Previous Literature

The study builds on several strands of literature. Early work examined the role of weather forecasts in sectors such as agriculture and transportation, generally identifying modest but positive economic value. More recent contributions have expanded this perspective by examining how forecasts influence adaptation to environmental risks, including precipitation forecasts affecting construction and traffic accidents, and temperature forecasts influencing mortality outcomes. Closest in spirit are studies that attempt to measure the value of environmental information while accounting for behavioral adaptation, such as research on air pollution monitoring and temperature forecast improvements. Within the hurricane literature, prior work has primarily focused on the economic damages of storms or on stated-preference estimates of forecast value, often relying on survey-based willingness-to-pay measures. Molina and Rudik extend this literature by providing a revealed-preference, outcome-based valuation using observed damages, expenditures, and forecast errors. Their contribution is to link forecast quality directly to realized economic outcomes, thereby advancing beyond aggregate or hypothetical valuation approaches.

Data

The empirical analysis relies on a newly constructed, high-resolution dataset that combines hurricane forecasts, realized storm characteristics, government expenditures, and economic damages at the county level. The sample includes all 31 hurricanes of Category 1 or higher that made landfall in the continental United States between 2005 and 2022. Forecast data are derived from the official models used by the National Hurricane Center, including both deterministic predictions and probabilistic distributions capturing forecast uncertainty. These forecasts are reconstructed at a granular spatial level, allowing the authors to compute county-specific expectations of wind speed and precipitation, as well as forecast errors.

The dataset is supplemented with administrative data on federal emergency expenditures from FEMA’s Public Assistance Grant Program, which distinguishes between pre-landfall protective spending and post-landfall recovery spending. Economic damages are measured using the SHELDUS database, which aggregates property losses and fatalities at the county level. Additional demographic and economic controls are drawn from Census and Bureau of Economic Analysis sources. The resulting dataset is unusually comprehensive, linking ex ante expectations to ex post outcomes across multiple dimensions. While measurement error in damages is acknowledged—particularly in less severely affected counties—the overall data structure represents a significant improvement over prior studies that rely on aggregate or storm-level observations.

Methods

The empirical strategy proceeds in three stages. First, the authors estimate how pre-landfall protective expenditures respond to forecasted hurricane intensity. This is implemented using regression models with county fixed effects and state-by-hurricane fixed effects, controlling flexibly for both wind speed and precipitation forecasts. The objective is to isolate the extent to which forecasts causally influence government behavior, holding constant geographic and institutional factors.

Second, the authors estimate the impact of forecast errors on realized damages and post-landfall recovery expenditures. By conditioning on actual hurricane intensity using flexible binning strategies, the analysis isolates the independent effect of forecast errors. This design leverages within-storm variation across counties, allowing the authors to compare areas that experienced similar realized conditions but received different forecast signals. Although not a formal quasi-experimental design, this approach approximates causal inference by conditioning on key confounders and exploiting plausibly exogenous variation in forecast errors.

Third, the paper develops a structural framework to estimate the marginal value of forecast improvements. The model treats agents as minimizing expected total costs, including both protective expenditures and realized damages, under uncertainty about hurricane intensity. The key insight is that the value of improving forecasts can be inferred from the relationship between squared forecast errors and total costs. This yields an estimable expression linking forecast variance to economic outcomes. While the approach is theoretically grounded and internally consistent, it remains observational in nature and would be strengthened by future work incorporating quasi-experimental or experimental variation in forecast quality.

Findings/Size Effects

The results provide consistent evidence that hurricane forecasts have substantial economic value. First, forecasts significantly influence pre-landfall protective expenditures. Counties forecast to experience hurricane-force winds receive approximately $36 million more in protective spending than counties forecast to experience lower wind speeds, equivalent to about 0.8 percent of county GDP or over $300 per capita. This indicates that forecasts materially shape resource allocation decisions prior to landfall.

Second, forecast errors have large and economically meaningful effects on damages and recovery costs. Conditional on actual hurricane intensity, underestimating wind speed by 10 meters per second increases county-level damages by approximately $220 million and post-landfall recovery expenditures by about $20 million. In proportional terms, this corresponds to roughly 15 percent of county GDP or over $5,500 per capita. These findings suggest that inaccurate forecasts lead to under-preparation, which in turn amplifies the economic impact of storms.

Third, improvements in forecast accuracy generate substantial cost savings. A marginal reduction in forecast error reduces total costs by approximately $5.5 million per county per hurricane. When aggregated, improvements in forecasting since 2007—following the implementation of the Hurricane Forecast Improvement Program—are estimated to have reduced total hurricane-related costs by 19 percent, or roughly $2 billion per hurricane. The implied value of forecast improvements exceeds the total federal budget devoted to weather forecasting, indicating large returns to investment in forecast technology.

Conclusion

The analysis demonstrates that hurricane forecasts are a critical component of disaster mitigation, with measurable effects on both pre-storm behavior and post-storm outcomes. By linking forecast accuracy to economic damages and government expenditures, Molina and Rudik provide a comprehensive valuation of information in a high-stakes policy context. The study’s primary contribution lies in its integration of behavioral responses, forecast uncertainty, and realized outcomes into a unified empirical framework.

The findings have broader implications for the economics of information and climate adaptation. As extreme weather events become more frequent and costly, the marginal value of accurate forecasts is likely to increase, particularly in environments where long-term adaptation investments remain limited. The study also highlights the importance of investing in information systems as a complement to physical infrastructure. While the empirical approach is rigorous and supported by high-quality data, the absence of a clearly exogenous source of variation in forecast quality suggests that future research could strengthen causal interpretation through quasi-experimental designs or randomized interventions. Nonetheless, the paper establishes a clear and quantitatively significant link between forecast accuracy and economic outcomes, advancing the literature on environmental risk and public policy.