Score contribution per author:
α: calibrated so average coauthorship-adjusted count equals average raw count
Predicting catastrophes involves heavy‐tailed distributions with no mean, eluding proactive policy as expected cost‐benefit analysis fails. We study US government counterterrorism policy, given heightened risk of terrorism. But terrorism also involves human behavior. We synthesize the behavioral and statistical aspects in an adversary‐defender game. Calibration to extensive data shows that where a Weibull distribution is the best predictor, US counterterrorism policy is rational (and optimal). Here, we estimate the adversary's unobserved variables, e.g., difficulty of an attack. We also find cases where the best predictor is a Generalized‐Pareto with no finite mean and rational policy fails. Here, we offer “work‐arounds”. (JEL H56, D81, C46)