Experimental Testing of Evacuation Messages in Sandy-affected Areas

Helping people understand the threat of coastal storms and persuading them to take appropriate protective actions is an essential job for emergency managers, the National Weather Service, and others who communicate with the public about coastal storms. To determine how to best communicate the risks of coastal storms, a series of experiments were designed to understand key message variables designed to increase the likelihood that coastal residents will take protective measures before a storm, particularly evacuation.

We conducted an online survey in spring of 2015 with randomly selected panel members from GfK's KnowledgePanel®, and a sample of coastal residents who self-selected. The participants included 1,716 residents of NJ, NY, and CT who live in coastal zip codes where at least 1% or greater of the land mass lies in SLOSH model Category 2 storm inundation zones (40% or greater in NY and NJ). GIS analysis of addresses revealed that 20% reside in FEMA 100 year flood zones, and 60% reside in Category 2 storm SLOSH inundation zones.

After a series of questions about experiences with Sandy, message variables were tested using a series of four hypothetical storm scenarios. Factorial ANOVAs were conducted within each scenario. Results indicate that location-based evacuation warnings tailored to include the participant's street name do not significantly increase evacuation intentions when compared to those that only include the name of their municipality (p < .001). A second scenario indicated that “mandatory” evacuation orders result in significantly higher evacuation intentions than “voluntary” orders (p < .001). However, three other types of warning wordings, including “evacuation advisory,” result in evacuation intentions that are in between “mandatory” and “voluntary,” and may be useful when declaring a “mandatory” evacuation is not possible. In another scenario, we found that high levels of guilt-based messages result in higher evacuation intentions (p < .002), and this is most effective when the guilt messages are focused on risks to first responders rather than to the participant's loved ones (p < .001). Finally, participants reported being significantly more likely to evacuate before hurricanes compared to nor'easters, even when the other information about the storm was held constant (p < .001). In addition to evacuation intentions, message variables had effects on intentions to engage in other recommended protective behaviors, risk perceptions, perceived storm severity, emotional responses and message comprehension.

Practical implications of these findings will be discussed, with a special focus on how emergency managers, broadcast meteorologists, the National Weather Service and NOAA can integrate these findings in their communications.