Last week, Vox ran a long piece by Max Fisher claiming that “the prospect of a major war, even a nuclear war, in Europe has become thinkable, [experts] warn, even plausible.” Without ever clarifying what “thinkable” or “plausible” mean in this context, Fisher seems to be arguing that, while still unlikely, the probability of a nuclear war between the United States and Russia is no longer small and is rising.
I finished Fisher’s piece and wondered: Is that true? As someone who’s worked on a couple of projects (here and here) that use “wisdom of crowds” methods to make educated guesses about how likely various geopolitical events are, I know that one way to try to answer that question is to ask a bunch of informed people for their best estimates and then average them.
So, on Thursday morning, I went to SurveyMonkey and set up a two-question survey that asks respondents to assess the likelihood of war between the United States and Russia before 2020 and, if war were to happen, the likelihood that one or both sides would use nuclear weapons. To elicit responses, I tweeted the link once and posted it to the Conflict Research Group on Facebook and the International Relations subreddit. The survey is still running, but 100 people have taken it so far, and here are the results—first, on the risk of war:
And then on the risk that one or both sides would nuclear weapons, conditional on the occurrence of war:
These results come from a convenience sample, so we shouldn’t put too much stock in them. Still, my confidence in their reliability got a boost when I learned yesterday that a recent survey of international-relations experts around the world asked an almost-identical question about the risk of a war and obtained similar results. In its 2014 survey, the TRIP project asked: “How likely is war between the United States and Russia over the next decade? Please use the 0–10 scale with 10 indicating that war will definitely occur.” They got 2,040 valid responses to that question, and here’s how they were distributed:
Those results are centered a little further to the right than the ones from my survey, but TRIP asked about a longer time period (“next decade” vs. “before 2020″), and those additional five years could explain the difference. It’s also important to note that the scales aren’t directly comparable; where the TRIP survey’s bins implicitly lie on a linear scale, mine were labeled to give respondents more options toward the extremes (e.g., “Certainly not” and “Almost certainly not”).
In light of that corroborating evidence, let’s assume for the moment that the responses to my survey are not junk. So then, how likely is a US/Russia war in the next several years, and how likely is it that such a war would go nuclear if it happened? To get to estimated probabilities of those events, I did two things:
- Assuming that the likelihoods implicit my survey’s labels follow a logistic curve, I converted them to predicted probabilities as follows: p(war) = exp(response – 5)/(1 + exp(response – 5)). That rule produces the following sequence for the 0–10 bins: 0.007, 0.018, 0.047, 0.119, 0.269, 0.500, 0.731, 0.881, 0.953, 0.982, 0.993.
I calculated the unweighted average of those predicted probabilities.
Here are the estimates that process produced, rounded up to the nearest whole percentage point:
- Probability of war: 11%
- Probability that one or both sides will use nuclear weapons, conditional on war: 18%
To translate those figures into a single number representing the crowd’s estimate of the probability of nuclear war between the US and Russia before 2020, we take their product: 2%.
Is that number different from what Max Fisher had in mind when he wrote that a nuclear war between the US and Russia is now “thinkable,” “plausible,” and “more likely than you think”? I don’t know. To me, “thinkable” and “plausible” seem about as specific as “possible,” a descriptor that applies to almost any geopolitical event you can imagine. I think Max’s chief concern in writing that piece was to draw attention to a risk that he believes to be dangerously under-appreciated, but it would be nice if he had asked his sources to be more specific about just how likely they think this calamity is.
More important, is that estimate “true”? As Ralph Atkins argued in a recent Financial Times piece about estimating the odds of Grexit, it’s impossible to say. For unprecedented and at least partially unique events like these—an exit from the euro zone, or a nuclear war between major powers—we can never know the event-generating process well enough to estimate their probabilities with high confidence. What we get instead are summaries of peoples’ current beliefs about those events’ likelihood. That’s highly imperfect, but it’s still informative in its own way.