Are Modern Insurgencies Predictable? New Evidence from the Afghanistan and Iraq Wars

Substate conflict is the source tremendous human suffering and financial burden worldwide. Research recently published in Science claims to have uncovered predictable patterns in insurgent and terrorist attacks: like various unrelated physical phenomena, the rate of learning by insurgents (and thus rate of conflict escalation) follows a predictable power law process. Governments afflicted by insurgent violence might use such patterns to thwart some attacks. The authors do not, however, consider strategic interaction during conflict and derive their results using data that recent research indicates is likely biased. Using newly declassified, comprehensive, and methodically collected data on insurgent activities covering all years of the Afghanistan and Iraq wars supplied by the Pentagon, we demonstrate that these findings are an artifact of biased microdata. We further show that dynamic learning during conflict varies continually: at times, insurgents learn more quickly than their opponents; at others, they do not. Observable patterns of learning by insurgents in conflict environments are complex and reflect strategic imperatives. The results speak to the need for greater interdisciplinary collaboration. Parsimonious mathematical models used to describe physical phenomena are not clearly suited to modeling strategic political interactions. If physical or social scientists are to inform national security policies, they will likely need to do so collaboratively.