Examining patterns of intended response to tornado warnings among residents of Tennessee, United States, through a latent class analysis approach

In the past five years, the southern region of the United States has had a large number of fatal tornadoes. Previous research indicates that residents of this area may not be taking appropriate shelter. The present study uses a random sample of Tennessee residents (N = 1126) and the latent class analysis (LCA) technique to explore discrete types of responders according to their pattern of intended behaviors when presented with a tornado warning scenario in the daytime or nighttime. LCA revealed three distinct groups in the day subsample - Tech Users, Typical Actors, and Passive Reactors - and three in the night subsample - Tech Users, Typical Actors, and Non-Reactors. Being a Tech User or Typical Actor was positively associated with intending to seek safe shelter, although being a Passive Reactor or Non-Reactor was not. Further, Tech Users/Typical Actors were seeking and obtaining more warning information from other sources compared to Passive Reactors/Non-Reactors. While few demographic variables were associated with class assignment, bivariate and multivariate analyses illustrated that cognitive factors, such as previous experience with tornadoes and perceived accuracy of warnings, are significantly associated with class membership when controlling for non-cognitive factors. The distinctions made within and between the subsamples can support the National Weather Service's efforts to better target the public with future messages about tornado safety as well as guide researchers on future studies.