Gone with the Wind: A Statistical Analysis of Hurricane Perceptions from Coastal Connecticut Residents in the Wake of Hurricane Sandy

As global temperatures rise, extreme weather events such as hurricanes and other large storms are predicted to become more frequent and severe. Although there is a widely accepted scientific understanding of this disaster trend, the response from the general public has not been proportional to the risk. Even with greater scientific innovations to assessing the risks and accurately predicting hurricanes, evacuation notices and calls for alarm have often been ignored. As storms become more severe, it has become increasingly important to understand the gap between the dangers coastal residents face and their perceptions and behaviors around those dangers. A survey was conducted by the Yale Program on Climate Change Communication of 1,130 Connecticut coastal residents a year after Hurricane Sandy, one of the costliest extreme weather events in the United States’ history. We used this dataset to identify crucial knowledge gaps in the public’s understanding of storm risks and answer the following questions: 1) are the evacuation predictors found similar to prior research 2) what hurricane perception groups exist within the data 3) can a simple model be created to predict evacuation behavior and 4) how can these findings inform on hurricane communication strategies. In order to identify these gaps and the causes of them, we conducted several statistical analyses including decision trees, principal components analysis, cluster analysis, and multiple linear regression. Some of the most important predictors related to receiving an evacuation notice, perceived likelihood of a hurricane making landfall, and whether or not they would evacuate based on their hurricane knowledge, their desire to stay at home, and their confidence in their decision. Our model – containing variables relating to risk perception, evacuation barriers, demographics, prior experience, decision difficulty and whether or not an individual received an evacuation notice – was found to have high prediction accuracy with an AUC of 0.88 (Area Under the Curve, a statistical measure of accuracy). We found a three-cluster analysis to be the best way of grouping the respondents, with three identities forming: informed and ready, hurricane avoiders, and unprepared. The results indicate significant populations in need of accurate and effective hurricane risk and evacuation communications. Our analysis shows the importance of timely evacuation notices from trusted messengers, such as government authorities, over commonly used media platforms. For the unprepared and hurricane avoiders, there is a critical need for greater communication of risks and preparation strategies for hurricanes. There must be greater research and planning of these communication strategies, but this analysis shows clear groups that exist in regard to how they respond to hurricanes and a path toward reducing loss of life from hurricanes.