(Nuclear) Change of Plans: What Explains Nuclear Reversals?

ABSTRACT

What explains a state’s decision to give up its nuclear weapons program? While instances of nuclear reversal figure prominently in international politics, evidence in the literature has been largely piecemeal. We offer a novel conceptualization of the proliferation process as nonlinear, potentially including instances of reversal, as well as pursuit of a nuclear program and acquisition of nuclear weapons. Employing this theoretical framework, we consider states’ cost-benefit calculations in each phase of the proliferation process, and we test our theory using a multistate model. Two counterintuitive findings emerge from this framework. First, nuclear latency increases the likelihood of pursuit and acquisition but also increases the likelihood of reversal by reducing the costs of restarting a program in the future. Second, the nonproliferation regime discourages states without a nuclear program from pursuing and acquiring nuclear weapons while at the same time making states with nuclear programs less likely to reverse course.

KEYWORDS:

• Event history analysis
• Non-Proliferation Treaty
• nuclear hedging
• nuclear latency
• nuclear reversal
• nuclear weapons

Acknowledgments

Author names appear in reverse alphabetical order, and both authors contributed equally to the process. For helpful comments, we thank Andrew Little, Jeffrey Kaplow, Rupal Mehta, and three anonymous reviewers.

Supplemental Material

A supplemental appendix is available at the authors’ websites, www.benjamintjones.com and www.eleonoramattiacci.com

Notes

¹ “Romney Stumbles in Explaining Iran Policy,” New York Times, September 14, 2012, cited in Fuhrmann and Tkach (2015).

² We thank an anonymous reviewer for this suggestion.

³ “The Nuclear Talks With Iran, Explained,” New York Times, March 2, 2015.

⁴ Some disagreement exists over how long it would take Japan to assemble a nuclear weapon. See for instance Holmes (2012) and Lewis (2014). For its part, Japan seems to release information on this issue in a very strategic way, sometimes to reassure allies and other times to scare rivals (see Kase 2001).

⁵ Kase (2001:55) documents how US concerns over a possible new Japanese nuclear program can be recorded all the way back to 1955.

⁶ For a recent example, see “China Demands Japan Explain Plutonium Underreporting,” ChinaDaily.com, June 9. Available at http://www.chinadaily.com.cn/china/2014-06/09/content_17573854.htm.

⁷ Notation adopted from De Wreede, Fiocco and Putter (2010).

⁸ Cumulative transition hazards estimated as: .

⁹ The cumulative hazard estimates are aggregated into an SxS matrix, A(t), where S in the number of phases in the model. The transition probabilities can then be estimated as , where (s, t] represents the desired time interval, and u represents all event times in the interval (s, t) (De Wreede, Fiocco, and Putter, 2010).

¹⁰ Because we rely on a gap time approach, we simulate these transition probabilities. For more information on this simulation algorithm, see De Wreede, Fiocco, and Putter (2010).

¹¹ Results presented in the article are also robust to a 10-year specification of this variable.

¹² We test for violations of the proportional hazards assumption, including time interactions where appropriate (Box-Steffensmeier and Jones 2004).

¹³ Although the proportion of system members that are NPT signatories increases over time, these findings do not simply reflect the effect of time. This is apparent in the fact that higher rates of NPT ratification have the opposite effect of the passage of time alone (Figure 2): The underlying rate at which states dismantle nuclear programs and acquire nuclear weapons increases over time, while higher numbers of NPT signatories decrease the likelihood of both of these transitions. Moreover, we control for the potentially confounding effect of the diffusion of nuclear technology over time (see the appendix), and our results are robust, indicating that our measure of the nonproliferation regime is not simply capturing the effect of time or the diffusion of nuclear technology.

¹⁴ All other variables are held at their median values for this and all subsequent plots.

¹⁵ The positive effect of nuclear latency on two mutually exclusive outcomes (reversal and acquisition) may appear counterintuitive; however, it is neither statistically nor logically necessary for a variable to display opposite effects on mutually exclusive outcomes. Rather, these effects indicate that nuclear latency is associated with relatively short nuclear programs, as it enables states to either rapidly complete their programs or reverse course. Indeed, other studies employing competing risks models often identify similar results. Studying the determinants of a congressional representative’s exit from office, Box-Steffensmeier and Jones (2004:170–171) find that a scandal has the same, positive effect on three mutually exclusive outcomes: losing in a primary, in a general election, and retiring.

¹⁶ For a recent review, see Kroenig 2015.