Nuclear Risk Forecasting

Link to Google Doc

On February 28th this year, I posted a couple of forecasts about the Russian invasion of Ukraine. Now, with the recent developments, I think it is time for a formal update to some predictions and I’ll also post my estimates to a couple of new questions. 

First, I’ll admit that it seems like I severely underestimated Putin and Russia’s willingness to extend the war. This was a mistake made on several fronts. I overestimated the effects of the sanctions on the Russian economy and their willingness to keep fighting, I underestimated the Russian willingness to take losses without giving up, and I overestimated the amount of resistance the Russian people would put up against Putin. In the previous post, I wrote:

This leads me to predict a 75% confidence interval for an armistice or peace between March 31st and October 1st.

This has clearly not happened and I have updated this prediction to way farther into the future. Now I think it is quite unlikely that this conflict ends without Putin losing power over Russia. 

Now, onto the new predictions. I have operationalized my views on the conflict and the risk that nuclear weapons are used in a couple of questions. I have stated my estimate after each question and I’ll go into more detail below. 

(i) At least one nuclear weapon used offensively before January 1, 2023? 7.5%

(ii) At least one offensive nuclear weapon used offensively before January 1, 2024? 17%

(iii) Conditional on a Russian nuclear weapon used offensively, will the US/NATO respond with offensive use of a nuclear weapon before January 1, 2024? 15% 

(iv) At least one Russian or US nuclear weapons test before January 1, 2023? 15%

(v) Will Vladimir Putin be the president of the Russian Federation on February 1, 2023? 87%

(vi) Conditional on Putin dying or otherwise being removed from office, will the Ukraine-Russia conflict end within two months? 85%

Even though these are point estimates, I’m have a lot of uncertainty about the situation. The recent weeks have provided much evidence that has made me update towards higher risk. Russian officials have been making explicit threats and North Korea, a Russian ally, has flown a robot over Japan. Explicit tests are of course more likely than offensive use, as they signal determination and threat without risking death. In (iii), I previously thought that it was higher but it currently seems like Nato and the US have much more to lose from using nuclear weapons than Russia has. Part of what drives the risk of offensive nuclear weapons usage is that it may happen due to an act of desperation or miscalculation from Russia—it may be that they can come to believe that the economic ruin and embarrassment from losing is worse than the risk of nuclear war. 

Finally, (v) and (vi) make my views clear on Putin’s role in this war. To me, it seems like Putin is one of the driving forces in continuing the war. If he loses power through a military coup, assassination, or any other way, then it is likely that a more peaceful leader/group takes power and that will likely end the war pretty quickly. With that said, it seems unlikely that Putin will lose his power in the near future. He’s likely on high alert and on the lookout for conspiracies against him. According to Statista, he still enjoys a high degree of approval (though I’m not sure about how confident one should be in the numbers.).

For other forecasts on similar questions, I recommend checking out Samotsvety Forecasting’s recently published forecast and, as always, Metaculus. I’m slightly more optimistic than both Samotsvety (they post an aggregate prediction from their superforecasters’ individual predictions) and Metaculus on the nuclear weapons questions. Samotsvety has an estimate of 6.3% for a nuclear weapon used in the coming month and Metaculus estimates a 9% probability of a nuclear weapon being used in Ukraine before 2023. 

Since the situation is developing all the time, I believe that more updates will come with time.

Leave a Comment

Your email address will not be published.

This site uses Akismet to reduce spam. Learn how your comment data is processed.