Heterogeneity Within and Across Households in Hurricane Evacuation Response

A survey of Houston-area households reveals responses to Hurricane Ike in 2008 were as diverse as the households themselves. Review of evacuation literature shows this remains a fundamental problem. In our analysis no clear correlations between household attributes and evacuation motivators emerge unless the respondents are organized into subpopulations based on household attributes and the stated concerns of survey respondents. These subpopulations overlap so that most households fall within multiple classifications, evidence that heterogeneity across households is also present within them. To address heterogeneity within households, an information content metric (information entropy) is considered a proxy for issue saliency. Focusing on the most salient responses to survey questions makes it possible to isolate some of the factors important in the decision to evacuate and the characteristics of the households for which those factors are most important. Regression analysis of the most salient issues of the most concerned respondents informs the creation of behavioral rules for an agent-based model populated with the survey data. The relative strengths of the risk-averting behavior rules are tuned through Monte Carlo simulations using the actual evacuation time of each household as the fitness metric.