Estimating the Severity of the WikiLeaks U.S. Diplomatic Cables Disclosure

Abstract

In November 2010, the WikiLeaks organization began the release of over 250,000 diplomatic cables sent by U.S. embassies to the U.S. State Department, uploaded to its website by (then) Private Manning, an intelligence analyst with the U.S. Army. This leak was widely condemned, including by then Secretary of State Hillary Clinton. We assess the severity of the leak by considering the size of the disclosure relative to all diplomatic cables that were in existence at the time—a quantity that is not known outside official sources. We rely on the fact that the cables that were leaked are internally indexed in such a way that they may be treated as a sample from a discrete uniform distribution with unknown maximum; this is a version of the well-known “German Tank Problem.” We consider three estimators that rely on discrete uniformity—maximum likelihood, Bayesian, and frequentist unbiased minimum variance—and demonstrate that the results are very similar in all cases. To supplement these estimators, we employ a regression-based procedure that incorporates the timing of cables' release in addition to their observed serial numbers. We estimate that, overall, approximately 5% of all cables from this timeframe were leaked, but that this number varies considerably at the embassy-year level. Our work provides a useful characterization of the sample of documents available to international relations scholars interested in testing theories of “private information,” while helping inform the public debate surrounding Manning's trial and 35-year prison sentence.