,

Principal Investigator: Joshua Epstein (Brookings)

Investigators:

• Ronald Brookmeyer (JHSPH)
• Edbert Hsu (JHO-CEPAR)

Project Objectives:

Economic policy modeling typically assumes homogeneous agents; people in these models are assumed behave optimally, using perfect (global) information. Effects of space and of social networks are typically ignored. Finally, static equilibria are computed; shocks, “tipping phenomena,” and dynamics are seldom considered. These classic assumptions are likely to be profoundly misleading in projecting the economic effects of terrorist events. Particularly, in terror crises, information will be poor, people will not behave rationally, epidemics of panic will spread through networks, and spatial patterns will abruptly change (e.g., flight patterns after 9/11). These effects will be cumulative and ramify in highly nonlinear ways not captured by standard economic modeling techniques. The agent-based computation approach is expressly designed to handle situations of this sort. Space is explicit in these models (where it is largely absent from standard approaches) and our focus is expressly on dynamics, not the usual static equilibria.

We plan to integrate the most advanced and powerful techniques of micro-simulation (agent-based modeling), to produce a unified computation epidemiology that is (a) behaviorally sound (b) highly visual and user-friendly, and (c) flexibly responsive to public health and bio-security issues of long- range and immediate policy concerns.

Project Summary & Planned Activities:

We will develop a suite of agent-based computational models exploring behavioral adaptations to epidemic (BT) crises; model the interactions of crisis behavior and infectious disease dynamics, and their impact on economics, extracting guidelines for policy and tools for training and education

In year one, we will develop a simple idealized model. We will populate social networks with agents who are psychologically heterogeneous: some prone to huddle, some to flee, some to follow government directives, some to imitate neighbors, etc. We will simulate the release a BT agent into the network and model the diffusion of both the epidemic and behavioral response (e.g. flight) as news spreads through the network via local channels. We will explore how predominance of different behaviors affects disease spread and economic dynamics. We will attempt to validate and build models based on known behavior dynamics.

By March 1, 2008, we would scale up the models to replicate social networks on the level of cities, regions and entire countries, which can be further linked.

By December 31, 2008, we will develop a beta version of a web-based software tools to allow participants to participate in the same WEB-X crisis simulation, all experiencing the same highly visual model, each implementing and evaluating alternative policies in real time. Such simulations and game playing will be considered by the Education Domain for dissemination as it becomes available. We will consult with the UCE for Behavioral and Social Aspects of Terrorism and Counter-Terrorism to explore collaborations in this area.

As of December 12, 2006, we have begun work on a Global Agent-based Model. In addition, we have drafted first version of Behavioral Epidemic Model and implemented it in Mathmatica, both in the form of differential equations and in Java as an agent-based computational model.

In April 2007, we compared the two and evaluated the preliminary results against historical cases. We plan to submit a working paper by June 30, 2007, detailing our results.

Review other PACER Research projects