| Abstract: | When intelligence agencies assess whether a state is pursuing nuclear weapons, how much evidence is enough? I argue that intelligence agencies adopt different standards for rendering definitive judgments in such situations. This, in turn, pushes them toward different kinds of mistakes. Urgent judges reach definitive conclusions about the existence of secret nuclear weapons programs more quickly and with less evidence than their peers. They risk seeing ambitious nuclear schemes where none exist. Skeptical judges wait longer and accumulate more proof before reaching definitive conclusions. They risk erring in the direction of underestimation. Where existing work focuses on intelligence accuracy, I show that variation in judgment is a distinct and important dimension of performance. What, then, drives judgment? I test an explanation based on the dynamic influence of previous intelligence failure. I observe that the judgment of intelligence agencies in two states, the United States and Israel, was differentially affected by failure as they tracked potential nuclear proliferation by Libya and Syria. These controlled comparisons constitute a novel approach to the study of nuclear intelligence performance. I find significant support for my explanation. Fearing repeat failure, intelligence agencies alter their efforts and standards of proof in an area critical to statecraft. |
