Misestimation: Explaining US Failures to Predict Nuclear Weapons Programs^†

Abstract

Various policy options have been proposed for slowing or halting the spread of nuclear weapons, yet all rely on sound intelligence about the progress of nuclear aspirants. Historically, the United States's record of estimating foreign weapons programs has been uneven, overestimating the progress made by some proliferators while underestimating others. This paper seeks to catalogue and evaluate the intelligence work surrounding sixteen of the twenty-five states that are thought to have pursued nuclear weapons, and to provide a framework for evaluating the causes of distorted intelligence estimates of nuclear proliferation. In particular, we identify twelve specific hypotheses related to politics, culture, bureaucracy, and organizational culture, then explore how they play out in practice through two case studies (North Korea and Israel). We find that the US has overestimated nuclear programs much more frequently than it has underestimated or correctly estimated them.

Atomic Ambiguity: Event Data Evidence on Nuclear Latency and International Cooperation

“…Conversely, countries that pursue lab-scale facilities while engaging in a campaign to keep them secret will receive less cooperation on the part of the U.S. than countries that lack such facilities-namely because efforts to keep these programs in the dark raise suspicions about states' intentions. Despite best efforts to keep these facilities concealed, and often precisely because of those efforts, the United States often succeeds in gathering some intelligence about latent programs (Montgomery and Mount 2014;Kemp 2014;Smith and Spaniel 2018). Over the past 70 years, the U.S. has primarily focused on preventing the spread of nuclear materials and technologies, especially to rogue actors that may pursue this form of ENR technology in secret, potentially aiming to revise or disrupt the status quo (Carnegie and Carson 2019).…”

“…Major powers and nonproliferators like the United States often must rely on noisy or incomplete assessments of a state's capacity and intentions-especially among proliferators whose programs are decidedly more secretive. Thus, U.S. internal intelligence estimates about the extent of a program can vary dramatically-with some estimates acknowledging a program may exist and others reports indicating the time to break-out or other critical points along the nuclear pathway from latency to weapons (Montgomery and Mount 2014;Smith and Spaniel 2018). While we do not directly engage this debate (as it is outside our scope of analysis), we argue that there is likely to be some relationship between whether latency is pursued by putting in place a concerted campaign to keep the existence of operational facilities secret and the cooperative overtures the U.S. does or does not offer.…”

The Spotlight's Harsh Glare: Rethinking Publicity and International Order

“…96. Montgomery and Mount 2014. directly involved. It could lead to Soviet or Chinese assurances of nuclear weapons support to North Korea in the event of a conflict."…”

The Disclosure Dilemma: Nuclear Intelligence and International Organizations

“…108 The U.S. continued to collect extensive information using satellite, communications, and human intelligence, detecting the components of the program and the personnel Fuhrmann (2012, 336). In 2001, the U.S. found that Libya sought dual-use technologies, and in 2002 the U.S. thought that Libya could produce a nuclear weapon by 2007, though the latter estimate exaggerated the trajectory of Tripoli's program (Montgomery and Mount, 2014). However, the U.S. and partners like the U.K. did possess detailed intelligence about Libya's activities, and Libya "saw how much [the U.S. knew] about what they were doing" in the nuclear arena.…”

“…132 For example, "When the IAEA began its inspections in 1991, South Africa was not obligated to reveal the existence of the critical facility or the other buildings in the valley....The IAEA, however, had learned of Building 5000 from Western intelligence, and asked for and was granted permission to inspect it" (Albright, 1994). 129 See Montgomery and Mount (2014). 130 See Burr and Richelson (2013).…”