Stimson has just published a report on the history of nuclear threat-making that was a project of mine for a few months earlier this year. It’s about 30 pages long, including the data set, so I thought I’d provide a summary of the findings, with a focus on a few things of particular interest to this community.
I focused on threats made during crises or wars between 1970 and 2010. Threats were defined so as to include both observable actions taken (e.g. increasing alert levels, forward-deploying nuclear-capable forces) and statements by high-ranking executive branch or military officials. The choice to study threats made in crises or wars was driven in part by a fear that including threats made in less tense times would include a lot of cheap talk – if making a threat doesn’t bear a risk of inadvertent escalation, the party making it doesn’t face any costs. So the focus on crises and wars helps avoid some “threats” that have more to do with domestic politics than anything else (though obviously any public statement by a politician reflects at least some consideration of the statement’s domestic impact).
There are lots of caveats which are discussed at length in the full report, but there’s one that I want to note here as well. In any quantitative study of nuclear threat-making, you’re going to face a choice: to include all signals that might plausibly have been interpreted as being nuclear threats by the target, or to rely on case studies to restrict the list of possible nuclear threats to ones that were demonstrably intended or interpreted as such. In the first case you’ll be able to stick to a set definition. In the second you’ll have a smaller data set that is more reliable, and a lot of borderline cases, because the burden of proof will be high and the evidence will likely be thin, especially for political systems in which information of this sort is not readily available. I opted to stick to a set definition, which led to the inclusion a few strange cases (the 1977 threat against Guatemala comes to mind) but also provided some benefits. These included being able to use a definition which provided a clear reason for including some threats and excluding others, rather than making a judgment call on each and every data point. The inclusive approach also yielded a perspective which focuses more on the vast majority of countries that don’t possess nuclear arms (since it includes all events that they might plausibly have interpreted as being nuclear threats).
All that being said, I’ll turn to the findings. The most significant of them, to me at least, was how infrequently non-nuclear weapons states that are in compliance with their nonproliferation obligations are threatened with nuclear weapons. There were 9 threats against NNWS in compliance with their nonproliferation obligations in the 1970-1990 time period, 36% of all nuclear threats during that period. There were 7 between 1991 and 2010, and these comprised just 13% of all nuclear threats in the last two decades. The significance of this finding is even more important given the political positions taken by these states. The NNWS regularly demand legally-binding assurances against the threat or use of nuclear weapons against them. Yet they have never been freer from nuclear threats, at least during crises or wars, than they are now.
So what do we make of the NNWS’s demand for legally-binding negative security assurances? Two-thirds of all nuclear threats against NNWS in the last two decades were against states that possessed chemical or biological weapons or against states that failed to live up to their IAEA safeguards commitments. It’s unclear to me how the vast majority of NNWS which are in compliance with their nonproliferation obligations benefit from demanding benefits which would mainly protect the least compliant of their colleagues. One possibility is that the NNWS have so little confidence in the ability of the nonproliferation regime to prevent their neighbors from acquiring WMD that they are trying to remove obstacles to being able to do the same, if necessary. Another possibility is that the NNWS don’t think that nuclear threats have much of an effect on the actual ability of their colleagues to acquire WMD, and would simply prefer to keep regional tensions as low as possible. Thirdly, it could be that it is the existential threat from nuclear weapons rather than nuclear threats during crises or wars that most concerns the NNWS.
The bigger picture isn’t something discussed much in the report. A smaller group of states is responsible for a much larger share of all nuclear threats during the 1991-2010 time period than during the preceding two decades. And furthermore, in the last twenty years, nuclear threat-making has been concentrated across a smaller number of crises and wars involving this same small group of states. What appears to be happening is that nuclear threat-making is becoming confined to a small number of political conflicts that are more removed from the normal course of international relations than before.
Two specific types of conflicts – the narrow one between India and Pakistan and the broader one involving America’s efforts to prevent WMD proliferation – were responsible for a much, much larger proportion of all nuclear threats between 1991 and 2010 than any other pair of conflicts in the previous two decades. If nuclear threats are less dispersed across countries and conflicts now than they were before, but the overall number is up, the crises or wars that do involve nuclear threats will involve more of them per crisis or war. I would argue that when a crisis involves multiple threats to use nuclear weapons, this is a symptom and/or cause of increased escalatory potential. Thus, those with a stake in preventing the future use of nuclear weapons would do well to pay attention to these particularly dangerous political conflicts.