« Back to Results

Weather Forecasts and Adaptation

Paper Session

Monday, Jan. 5, 2026 8:00 AM - 10:00 AM (EST)

Philadelphia Marriott Downtown, Grand Ballroom Salon J
Hosted By: Association of Environmental and Resource Economists
  • Chair: Vaibhav Anand, St. John's University

The Impact of Precipitation Forecasts on Small Business Performance

Abhishek Nagaraj
,
University of California-Berkeley
Tim Maximilian Sels
,
University of California-Berkeley
Weilong Wang
,
Columbia University

Abstract

Weather forecasts are a critical tool for economic decision-making, yet systematic evidence on their real-world value remains limited. This study quantifies the economic impact of short-term precipitation forecasts on small business performance across the United States. Using a natural experiment framework, we exploit variation in forecast accuracy while holding actual weather constant to isolate the causal effect of forecast quality on business outcomes. Specifically, we examine how customer foot traffic and consumer spending at over one million establishments vary when precipitation is predicted accurately versus when forecasts either over- or under-estimate rainfall. Drawing on high-frequency data from SafeGraph and anonymized transaction datasets, we find that forecast errors meaningfully influence small business activity. On days when rainfall is overpredicted, businesses see a measurable decline in visits and spending—even when actual conditions are favorable. Conversely, underpredictions lead to increased business activity, though often at the cost of preparedness. These effects are concentrated on days with moderate to heavy precipitation and vary significantly by industry, with retail, food service, and entertainment sectors most affected. Our results highlight not only the economic cost of inaccurate forecasts but also the value generated by reliable short-term predictions. We estimate that improvements in 1-day precipitation forecasts yield substantial returns in terms of increased consumer activity, especially during inclement weather. These findings suggest that investment in improving forecast accuracy can have significant downstream benefits for local economies. By providing causal evidence on the value of weather information, this study offers a rigorous framework for evaluating forecast innovations and underscores the importance of incorporating user impact into forecasting system design and funding decisions.

Climate Shift Uncertainty and Economic Damages

Gernot Wagner
,
Columbia University
Manuel Linsenmeier
,
Columbia University
Romain Fillon
,
Université Paris-Saclay, France

Abstract

Focusing on global annual averages of climatic variables can bias aggregate and distributional estimates of the economic impacts of climate change. We here empirically identify dose-response functions of GDP growth rates to daily mean temperature levels and combine them with regional intra-annual climate projections of daily mean temperatures. We then disentangle, for various shared socio-economic pathways (SSPs), how much of the missing impacts are due to heterogeneous warming patterns over space. Global damages in 2050 are 25% (21-28% across SSPs) higher when accounting for the shift in the shape of the entire intra-annual distribution of daily mean temperatures at the regional scale.
JEL: D62, Q56

Weather Forecasts and Adaptation in the U.S.: Disparities and Determinants

Vaibhav Anand
,
St. John's University
Tim Philippi
,
University of Hohenheim

Abstract

Weather forecasts play an important role in adaptation for weather and climate-related damages. However, accuracy of weather forecasts shows significant spatial heterogeneity across the United States. This paper documents disparities in forecast skills, identifies potential social and economic determinants, and explores implications for adaptation and damages. We use county-level daily weather observations from the Global Historical Climatology Network and forecasts from the National Digital Forecast Database for 2004-2024. We examine social, economic, and climatological factors potentially influencing these disparities. We assess how these disparities in forecast skill impact adaptation actions, focusing specifically on responses to weather advisories. Advisory accuracy and lead times are measured using advisory archives from Iowa Environmental Mesonet. Adaptation responses are measured through SafeGraph mobile phone location data (2018-2024). Preliminary results focus on 24-hour forecasts of daily temperature. We find substantial regional and seasonal variation in forecast skill. Specifically, skill is lower in the Western and Southwestern US, including Arizona, New Mexico, and California, while it is higher in the Midwest, Northeast, and the Southeast. Results suggest that greater access to weather stations significantly enhances forecast accuracy. The average forecast error of maximum and minimum temperature significantly decreases by 0.0006 °F and 0.0003 °F per weather station within 50 miles of a county center. While preliminary findings from the mobile phone location data indicate that people’s movement responds to weather advisories, our ongoing analysis investigates how spatial variations in numerical forecast skill affect the accuracy and timeliness of weather advisories, and consequently the effectiveness of adaptation. We contribute to the emerging literature on the economics of weather forecasts-based adaptation (Shrader et al. 2022; Shrader 2022; Molina and Rudik 2023; Burlig et al., 2024; Linsenmeier and Shrader 2024; Song, 2024).

Costs of Climate Adaptation: Evidence From French Agriculture

Jeffrey G. Shrader
,
Columbia University
Tristan du Puy
,
Columbia University

Abstract

Adaptation costs are one of the main elements missing from the existing literature on the effects of climate change. Policy to address climate change depends on how costly it is for people to adapt, but a lack of cost-related data means that such estimates are rare. In this paper, we use uniquely rich data on agriculture in France to provide novel, direct estimates of the marginal cost of adapting to temperature shocks. The dataset is a farm-level panel with measures of outputs, inputs, and prices from 1994-2018. We merge the farm data with measures of realized and forecasted weather. Controlling for realized weather, we use forecasts as information shocks to estimate the marginal cost of ex ante adaptation. For the year that heat shocks arrive, we find that, for the average farm in France, the marginal cost of adaptation is low. In contrast, the marginal benefits of adaptation are large and both significantly and economically meaningfully different than the estimated marginal costs. Dynamically, however, the costs of adapting to that original shock rise over subsequent years, eventually leading the marginal benefits of adaptation to equal marginal costs, a result consistent with a dynamic envelope theorem but inconsistent with a static sufficient statistic approach widely used when estimating climate damages. This pattern of adaptation costs is driven by the behavior that farms engage in when responding to forecasts. In the year of the shock, they use the forecasts to change the timing of key growing decisions as well as their crop mix, rather than to change costly inputs. These decisions create dynamic switching costs, which we can relate to soil nutrients and pest control. Finally, we show that forms of adaptation currently implemented by farmers are less likely to remain relevant in a warmer world as farmers face tighter adaptation.

Discussant(s)
Ivan Rudik
,
Cornell University
James Rising
,
University of Delaware
Renato Molina
,
University of Miami
Derek Lemoine
,
University of Arizona
JEL Classifications
  • Q5 - Environmental Economics
  • Q0 - General