Optimal control of a terror queue

The task of covert intelligence agents is to detect and interdict terror plots. Kaplan (2010) treats terror plots as customers and intelligence agents as servers in a queuing model. We extend Kaplan’s insight to a dynamic model that analyzes the inter-temporal trade-off between damage caused by terror attacks and prevention costs to address the question of how many agents to optimally assign to such counter-terror measures. We compare scenarios which differ with respect to the extent of the initial terror threat and study the qualitative robustness of the optimal solution. We show that in general, the optimal number of agents is not simply proportional to the number of undetected plots. We also show that while it is sensible to deploy many agents when terrorists are moderately efficient in their ability to mount attacks, relatively few agents should be deployed if terrorists are inefficient (giving agents many opportunities for detection), or if terrorists are highly efficient (in which case agents become relatively ineffective). Furthermore, we analyze the implications of a policy that constraints the number of successful terror attacks to never increase. We find that the inclusion of a constraint preventing one of the state variables to grow leads to a continuum of steady states, some which are much more costly to society than the more forward-looking optimal policy that temporarily allows the number of terror attacks to increase.