The increasing frequency and intensity of extreme weather events due to climate change raise concerns about their impact on public perception and behavior. While studies suggest a link between extreme weather and pro-environmental changes, it's unclear whether these events cause the changes or if pre-existing differences exist among affected and unaffected populations. This study addresses this gap by analyzing the causal relationship between the 2021 European floods—a significant flooding event in Central Europe, including Luxembourg—and changes in pro-environmental attitudes and behaviors. The 2021 floods presented a natural experiment, allowing researchers to assess causality more rigorously than observational studies. The study's focus on a younger demographic (16–29 years old) in Luxembourg provides a specific lens on attitudes and behavior change, recognizing that this age group may exhibit greater adaptability in environmental attitudes compared to older populations. The study's importance lies in its potential to inform strategies for addressing climate change by illuminating the connection between climate-related extreme weather events and shifts in environmental consciousness and action. By leveraging a natural experiment and validated survey instruments, the study aims to provide more robust causal inferences compared to previous research.

Existing research explores the relationship between extreme weather events and environmental attitudes, with mixed findings. Some studies demonstrate increased risk perception, climate change beliefs, environmental concern, and pro-environmental intentions and voting behavior following such events. However, other studies find no systematic effects or fail to establish causal relationships. A critical gap in the literature is the lack of rigorous causal inference, often due to observational designs and self-selection bias. Studies frequently utilize single-item measures for environmental attitudes and behavior, limiting the depth of analysis. The limited use of validated, multidimensional survey instruments in longitudinal studies further hinders establishing causal linkages. This study seeks to address these limitations by employing a rigorous natural experiment design and validated measurement tools to examine the causal connections between personal experiences of extreme weather events and attitudes towards climate change, whether extreme weather events help close the attitude-behavior gap, and the mediating role of environmental attitudes on the relationship between natural events and environmental behavior.

This study utilized a natural experiment design, leveraging the 2021 European floods in Luxembourg. Data was sourced from the "Young people and COVID-19" (YAC+) survey, a stratified random sample of Luxembourg residents aged 12–29, conducted between July 13th and October 1st, 2021. The floods began on July 14th–15th, creating a pre- and post-event comparison. The analysis focused on respondents aged 16–29 (n=2058) after excluding those aged 12–15 and those with missing data. Environmental attitudes were measured using a validated nine-item scale by Diekmann & Preisendörfer, encompassing affective, cognitive, and conative dimensions. Pro-environmental behaviors were assessed through seven self-reported actions. Spatial proximity to flood areas was binary coded based on commune-level flood data. Multivariate cluster-robust linear models were employed, adjusting for sociodemographic variables and pre-treatment characteristics to account for potential confounding factors. Cluster-robust standard errors were used, clustered at the commune level. Two-stage instrumental variable regression was utilized to assess the causal effect of attitudes on behavior, using flood exposure and spatial proximity as instrumental variables. The R packages `estimatr`, `ggplot2`, `tmap`, and `modelsummary` were used for data analysis and visualization.

The study revealed that flood exposure significantly increased pro-environmental attitudes, with the strongest effect observed on the affective dimension (*t*(101) = 3.410, *P* < 0.001, β = 0.190). The effect on the overall environmental attitude scale was also significant (*t*(101) = 2.947, *P* < 0.01, β = 0.169). The effect was weaker for the cognitive dimension and not significant for the conative dimension. For individuals living near the flood areas, the differences in attitude dimensions were statistically significant, with the strongest impact on the conative dimension (*t*(101) = 3.206, *P* < 0.01, β = 0.360). While higher environmental attitudes were observed following flood exposure, there was no significant direct effect on self-reported pro-environmental behaviors after adjusting for attitudes and covariates. However, a significant moderating effect of flood experience on the relationship between attitudes and behavior was found for buying organic food (*t*(101) = 2.269, *P* < 0.05, β = 0.230) and buying energy-saving electronics (*t*(101) = 2.745, *P* < 0.01, β = 0.270) among those living near flood areas. Two-stage instrumental variable regression indicated a significant causal effect of environmental attitudes on meat consumption and energy-saving behavior.

The findings support the notion that experiencing significant environmental events, such as severe flooding, can lead to increased pro-environmental attitudes, particularly affective ones. This aligns with experiential learning theory, demonstrating that direct experience can enhance understanding and engagement with climate change. The stronger effect among those geographically closer to the floods highlights the importance of perceived personal risk in shaping attitudes. The lack of a significant direct effect on all reported behaviors suggests that increased awareness doesn't automatically translate into behavioral changes across all domains. However, the moderating effect of flood experience on specific pro-environmental purchasing behaviors indicates that context and proximity to risk matter. The causal effect of attitudes on meat consumption and energy-saving behavior further supports the connection between internal beliefs and some specific behavioral changes. These results suggest that while leveraging climate change impacts to promote public concern can be effective, relying on such events solely for behavior change is insufficient. The study underscores the need for effective policy instruments to guide individuals toward sustainable practices.

This study provides evidence of a causal link between flood exposure and increased pro-environmental attitudes, particularly the affective dimension, with stronger effects closer to the affected areas. The findings highlight the importance of experiential learning and perceived risk in shaping environmental attitudes. While the impact on reported behaviors was less direct, the study reveals a moderating effect for specific purchasing behaviors and a causal relationship between attitudes and some behaviors. Future research should investigate actual behaviors rather than self-reported ones, explore the heterogeneous effects on different environmental behaviors, use larger and more diverse samples, and investigate the potential mechanisms linking attitudes and behaviors more thoroughly.

The study's reliance on self-reported behaviors might overestimate pro-environmental actions. The focus on a younger demographic limits generalizability to broader populations. The relatively weak correlations between attitudes and behaviors in some domains suggest a need for larger sample sizes to more reliably assess causal effects. Finally, further research is needed to explore the specific mechanisms through which environmental attitudes translate into behavior changes.