Invited to revise and resubmit to Conflict Management and Peace Science
Abstract: Latent nuclear capability is a concept of central importance to proliferation scholars. However, because such capacity is inherently unobservable, a number of issues hinder existing measures. To overcome them, we adopt a statistical approach well-suited to the estimation of unobservable quantities and apply it to the study of nuclear capability. Specifically, we utilize a model grounded in item-response theory that recovers an estimate of this unobserved characteristic. By estimating the model in a Bayesian framework, we assess the inherent uncertainty of our measures. Throughout, we demonstrate a number of ways in which our scores improve upon the additive-indexing approach taken in existing measures of nuclear capability. In particular, our estimates provide information about activities related to nuclear production. For example, contrary to existing arguments, the estimates indicate that the possession of enrichment sites is not a strong indicator of nuclear capability.