An Integrated Approach to Modeling Evacuation Behavior: Hyperbolic Discounting and Peer Effects

A spate of recent hurricanes and other natural disasters have drawn a lot of attention to the evacuation decision of individuals. Here we focus on evacuation models that incorporate two economic phenomena that seem to be increasingly important in explaining human behavior: hyperbolic discounting and peer effects. The first part of this research explores the behavior of the naïve or myopic agent in deciding whether to perform a mandatory task whose cost is immediate but reward received only in the future. Following the literature for hyperbolic discounting, we say that a player is naïve if her inter-temporal preference for whether to complete an assigned task is represented by the Phelps and Pollack's hyperbolic discounting utility model. We show that a naïve agent, whose present bias is below a certain game-dependant bound, is meant to complete the task in the last period. This bound offers two new insights about the naïve player. First, "not all naïve players are equal" in that the long run discount factor decides the degree of naivety sufficient to procrastinate. Secondly, this shows that the range of the payoff structure plays a role in favoring procrastinating behavior. Finally, an application of naive hyperbolic discounting for an evacuation model is constructed. The peer effects research is concerned with testing the hypothesis that an agent's decision of whether to evacuate during a hurricane is influenced by the fear propensity of his or her peers. To explore this human aspect of an evacuation, a simple random utility model is set up where the utility of the "agent who is not scared" is allowed to depend on the fear propensity of the group she identifies with. The resulting binary choice model derived contains a real key parameter measuring the peer effect. Using data from Hurricane Floyd, we estimate that a positive peer effect exists in the sense that the larger the fear propensity of the peer group, the more attractive an evacuation. This finding suggests that policies aimed at creating strong awareness of hurricane dangers before the hurricane season can have a substantial effect on the population's evacuation rate via their multiplying effects.