The Emotional Impact of Climate Disasters
Client | Center for Applied Behavioral Health Policy at ASU
New Analysis by Rogue Scholar
In collaboration with Arizona State University’s Center for Applied Behavioral Health Policy, Rogue Scholar completed an analysis that examines the emotional impact of climate disasters among residents along the Gulf of Mexico. This region includes Alabama, Oklahoma, Texas, and Florida.
The study, sponsored by the National Academy of Sciences, is part of a nationwide partnership to create a Zillow-like app that captures flood risk information to potential homebuyers. Researchers believe that proper risk communication (e.g., phrasing to encourage or discourage actions) can help homebuyers in flood-prone areas consider disaster likelihood and effects, as well as encourage adopting disaster mitigation strategies or purchasing disaster insurance. Understanding emotional responses to natural disaster risks may determine which messaging modality (i.e., percent chance of risk, potential cost, or anecdotes from local residents) is the most persuasive in choosing a home.
The survey, developed by economists with the Darla Moore School of Business at the University of South Carolina, exposes Gulf home buyers and renters to homes with a variety of flood risk scenarios. These “choice experiments” seek to determine emotional reactions to risk information and home buying choices.
Dr. Tamara Sheldon (USC) uses a specific methodology to determine the utility function of how people make decisions, or trade-offs between a given set of criteria. In this case, participants were asked to choose between two houses with 5-7 different characteristics for a total of six times. Dr. Sheldon uses those choices to determine if people are willing to accept more disaster risk based on home characteristics such as interior quality, size, or price.
What we Did
Rogue Scholar planned the data analysis, performed data cleaning, analyzed the data, and interpreted and wrote up the results.
The analysis was an ordinal regression, also known as a proportional odds model, which is used when a dependent variable has ordered categories. In this case, we were predicting negative, mixed, and positive emotions. This analysis “discretizes” the odds of each emotion as a function of the independent variable(s). Like a logistic regression, results are interpreted with an exponent of beta, except that rather than a binary outcome, an ordinal regression estimates the odds of being in a higher vs. lower emotional category. The model estimates ‘proportional odds’ because the distance between each category is parallel (that’s one of the model’s assumptions!), hence the proportional moniker.
I enjoyed learning a few new SPSS tricks to complete the analysis, like how to add exponentiated betas to your output . Check out my post on SPSS’s Output Management System (OHS).
Also, I had a personal connection to the project. Growing up along Florida’s Gulf Coast, I experienced a several weather disasters firsthand. The most significant one was Hurricane Andrew, a compact, powerful, and destructive Category 5 hurricane. Being part of a project that can contribute to saving homes and lives in this beautiful part of the United States was rewarding to me.
Have data, need analysis? Contact Rogue Scholar for an estimate!