When flood warnings fail: Psychological predictors of information clarity and evacuation intention

Flood warnings are often assumed—particularly within traditional average-effect models—to prompt uniform behavioral responses, yet real-world evidence suggests that message effectiveness varies widely across the population. This study empirically identifies and explains such heterogeneity using a factorial survey experiment ( n = 1,671) conducted among residents in flood-prone regions of Japan. We estimate finite mixture beta regression models to disentangle the distinct behavioral patterns underlying information clarity and evacuation intention. Results reveal two psychological segments: a Responsive majority, whose decisions are shaped by concrete and credible warning attributes (e.g., specific breach point, forecast basis, large-font display), and a Non-responsive minority, who remain disengaged regardless of message design. Membership in the latter group is strongly predicted by distrust of weather information and habitual neglect of local hazard data rather than demographic factors. These findings underscore the limitations of one-size-fits-all communication strategies and highlight the urgent need for audience segmentation in early warning systems. By linking cognitive orientations to message responsiveness, this study advances the behavioral foundations of risk communication and offers practical guidance for designing more inclusive and trusted flood warning systems.