Will the US return to the Iran deal by the end of 2021?

Time spent: ~7 hours

Or, more specifically: Will the United States return to compliance with the 2015 Iran nuclear deal by December 31, 2021?

This question certainly feels very unique, but reference-class forecasting remains the strongest methodology I’m aware of for generating calibrated forecasts.

In my view, the crucial question is whether the US will rejoin the Iran Deal by the end of 2021. This requires knowing how long these agreements typically take. So, I start by building an outside view then update for inside view considerations.

Whence A Reference Class?

It was hard to come up with a clearly similar set of events to forecast from. Other international institutions that Trump withdrew from and Biden re-entered required substantially less negotiation due to their multilateral nature (e.g. the WHO, the Paris Accords). By contrast, this is very much a (perhaps augmented) bilateral negotiation between the US and Iran. Also, Biden has only been in office for a few months and may rejoin more agreements than he has so far, so this reference class is actually not so similar to what we’re interested in.

But negotiations over arms control agreements are nothing new. I built a dataset of negotiation timelines for major arms control agreements (treaties, protocols, etc) since 1945–this mainly includes negotiations between the US and Soviet Union from a Congressional Research Service report.¹ To the US-Soviet agreements I added: the case of Libyan disarmament of 2003, the North Korean Agreed Framework of 1994, the Iranian Joint Plan of Action of 2013, and the Iranian Joint Comprehensive Plan of Action in 2015. I expect that this dataset includes between two-thirds and three-quarters of all the actually relevant cases–I doubt that its comprehensive! The dataset includes 20 instances. (You can see the dataset at the bottom of this post.)

Taking the average length of negotiations over the entire dataset gives an average of 3.14 years, and just four of the twenty instances took less than one year. But the question at hand is a follow-on negotiation, so if we restrict our scope to only follow-on negotiations then two of eleven instances took less than one year. ²

What were those two instances? One was the Strategic Arms Reduction Treaty (START II) in 1993, negotiations for which took around six months. But it was held up by the Senate in the US, and by the Duma in Russia and never entered into force. The other was the New START Extension in 2021, in which the Biden administration and Putin’s Kremlin basically agreed they wanted to extend New START with little fuss. So neither of these actually seem to suggest what a “Yes” example here would look like.

This gives me a very low probability outside view: the most this data could suggest (because I expect to have missed some relevant cases) is a historical base rate of between 4/15 (if I missed one-third, and all of them are favorable) and 3/14 (if I missed one-quarter, and all are favorable).

I don’t think that’s plausible, so I settle at an outside view of around 10%.

But What About the Chatter?

I give some weight to the positive signs of progress and update accordingly.

I think one should update some on positive comments by Javad Zarif, Hassan Rouhani, and other Iranian diplomats associated with the current Iranian government. That said, if a deal were not imminent, I think we would be likely to hear from them that things are going swimmingly because they have an incentive to project success for upcoming elections. And, a deal does not automatically generate a “yes” because timelines may be too slow. For example, working groups on synchronizing a return have already been established, but lifting hundreds of sanctions and rolling back nuclear enrichment is difficult.³ Small update: 95% to encounter if “yes”, 60% if “no”. 1:9 prior times 95/60 = ~ 3:20 odds, or ~15%.

Notably, signals of commitment or good faith are not reasons to update up from the base rate because in all the cases in the dataset, negotiators were actually trying to strike a deal (and usually did)–it just took longer than one year. And, the dataset is almost certainly biased in favor of successful negotiations, either because secretive negotiations which were unsuccessful remain secret, or because it is easier to find writing on negotiations when they result in a new treaty or agreement.

I update downward on the Guardian Council’s decision to bar moderates from running in the upcoming election. The creates a deadline of early June to reach an agreement, which could be motivating. However, if that deadline is not met (as happened repeatedly in the JCPOA negotiations⁴), then I think the likelihood of success this year falls near zero. Medium update: 10% to encounter if “yes”, 50% if “no”. 3:20 prior times 10/50 = ~3:97 odds, or ~3%.

It’s hard to say what to make of comments by Russian envoy Mikhail Ulyanov that, “As of now, there are no plans for the sixth [round]. The negotiators proceed from the understanding that the current round should be final.”⁵ He seems to be firing from the hip based on previous comments on Twitter being walked back, so I don’t think he’s especially reliable. See also comments above on failed deadlines during JCPOA negotiations. Iran’s spokesman Saeed Khatibzadeh’s comment also undercuts credibility: “Each round of talks in Vienna could have been the final round. We should not rush. We have made significant progress but key issues remain.”⁶ However, this is still useful information and I should still update up on this somewhat. On the other hand, I have not seen any similar comments by US diplomats; interlocutors continue to express caution, and I take that seriously. Small update: 90% to encounter if “yes”, 50% if no. 3:97 prior times 90/50 = ~1:19 odds, or 5%.

I don’t update on oil prices moving; markets move for a ton of reasons, but the Iran deal is hot right now so there are lots of journalistic incentives to attribute movement to negotiations.

Current likelihood of “yes”: 5%.

Possible Alternative Scenarios

This does not mean I think that the Iran deal is unlikely to be revitalized, just that it is unlikely to be revitalized in this year. Here are a few alternative scenarios that seem plausible:

Iranian trust in the US has been crushed by the Trump administration withdraw, so instead of an all-at-once deal, the US will have to lift some sanctions, then Iran will reduce capacity by some amount, and things will proceed iteratively. This may take well over a year. Note that negotiations for the JCPOA itself took 1.5 years despite the JPA roadmap already being agreed to (which took 7.5 years of negotiations).

It could also be that the Iran deal does not succeed by the June Iranian elections (for a variety of reasons), hardliners are unconvinced, and a new Iran deal will have to wait until a more favorable administration in Iran comes to power.

Iranian trust that the US won’t pull out again may be restored, but companies that got burned might not be able to be coaxed back to Iran because of the risk of future sanctions, so Iran might see limited upside anyway.

There are lots of reasons this could happen–I’m just spelling out a few that seem plausible.