MONTE CARLO SIMULATION OF INTERSTATE WAR FROM 1816 TO 2007

摘要

Probability density functions were synthesized from Correlates of War Project data to create a compact Monte Carlo algorithm able to approximate the total number of interstate combat deaths recorded from 1816 to 2007. Results suggest that three random variables are required (the annual occurrence of war, number of alliances, and combat deaths per state) and that there is a marked transition from more common wars involving less than ten allies (regional) to uncommon wars having ten or more allies and a composite severity distribution (bimodal). Combat deaths associated with regional conflicts are distributed according to a log-gamma probability density function whereas deaths associated with bimodal conflicts are distributed according to two overlapping log-normal probability density functions. Incorporated into a Monte Carlo simulator, the algorithm can be used by a national security enterprise to generate random numbers to help estimate the risks of war in terms of combat deaths, to weigh potential outcomes against other types of risks, and to explore strategies to minimize risk.