Ukraine Post #11: Longer Term Predictions

post by Zvi · 2022-04-25T14:10:01.119Z · LW · GW · 6 comments

Contents

  Metaculus
    How much foreign aid will USA provide Ukraine in 2022? 16.4 billion (unchanged).
    Will 5mm+ Ukrainian refugees seek assistance overseas? Yes.
    Will Ukraine have a security guarantee from another country before 2024? 40%.
  Polymarket
  Synthesis
    5%: Total Russian victory
    65%: Russia keeps some additional territory beyond Donbass/Crimea
    19%: Ukraine does not give up additional territory but fails to reclaim Donbass
    11%: Total Ukrainian victory, they reclaim 90% of Donbass
None
6 comments

The pace of these Ukraine posts is slowing down, and will probably slow further if things go as I expect them to go. Given how much slower things are going, this will likely be my last update on Ukraine for a few weeks.

I did want to record how various things are tracking for future reference and there are some interesting things to think about, but this post proved inessential. If you’re not interested in the details you can safely skip it.

Early in the war, the situation was fluid and unclear. In this new phase, which one could call the Battle for Donbass, things are far less fluid and seem more clear.

As I noted in my recent China post, the Ukraine epistemic environment involves a shooting war yet is relatively straightforward. The disagreements feel less like disagreements over facts and more like disagreements over models of what is likely to happen going forward. There is a Russian story of what is happening, but it is not attempting to be credible. There is uncertainly over details, especially in terms of war crimes, but none of the true disputes seem important.

Early on, I felt pressure to check developments every few hours. Then that became every day, and now it feels more like every several days. This matches the pattern of Covid, both in March 2020 and then again with Omicron and to a lesser extent a few other big developments along the way, and also my experience with other major news events. Things don’t ‘go away’ so much as the situation becomes clear and new developments slow down. Then, hopefully, things eventually ‘go away’ but that usually takes longer.

This will centrally be another pass through the prediction markets, primarily Metaculus.

While I have a compiled list of new bits of information, it seems scarcely necessary – the sequence of what happened with the sinking of the Muscovy – Russia claiming it is fine, Russian Telegram groups saying it’s not fine and vowing revenge, Russia claiming it was damaged but still afloat, Wikipedia listing the ship as ‘on fire,’ the sinking, calls for revenge for the sinking, a word about Russia claiming it put part of the True Cross on the ship, an analysis of the wreck, an analysis of the attack, . One country after another realizing (correctly) they should embrace nuclear power. Maps that mostly stopped changing. Speculations about Russian losses and decline in capacity. More about Ukraine’s advantage at night. Detailed assessments of the state of play that don’t change much and all anticipate Russia having extreme difficulties advancing. Putin claiming all is going according to plan. America favoring ethanol production over food production because the food situation is fine. That sort of thing.

In terms of the short-to-medium term, the Battle for Donbass has begun. Reports are that Russia is making non-zero but not very substantial progress. I expect similar to continue until such time as Russia’s forces are exhausted or they stop trying. The next key event will be victory day on May 9, and what Russia does then after (I am expecting) having made no important progress in the two weeks leading up to it.

Metaculus

I have begun work on a draft post detailing improvements I’d like to make to Metaculus. Various things got in the way of finding the time to complete it for now. I hope to come back to it at some point.

The thing about these markets is that very, very little has changed in the past 10 days. All the lines are straight. At some point I got tired of porting over those straight lines and stopped doing it for questions that did not seem interesting right now.

The interesting questions are the new questions that were added, which were the motivation to look around again.

Will Odessa be under Russian control on June 1, 2022? No.

Will Kharkiv be under Russian control on June 1, 2022? No.

Will Cherkasy be under Russian control on June 1, 2022? No.

Will Kyiv be under Russian control on June 1, 2022? No.

Will 3 of 6 threatened cities be under Russian control on June 1, 2022? 5% → 4%.

Not sure why this one is so high.

Will Mariupol be under Russian control on June 1, 2022? 92% → 95%.

I’m staying a little lower because I think a general Russian collapse by then is not so impossible, and this would be one of Ukraine’s first major targets in a counteroffensive.

Will Belarus invade Ukraine by June 1, 2022? 9% → 5%.

If anything this seems high, reasons to do this seem much weaker now.

Will Kyiv be under Russian control before 2023? 3% → 3%.

This certainly feels low given that Russia would doubtless take Kyiv given the opportunity, and there are a number of outlier scenarios that are possible. Even though I’m generally bearish on Russia’s chances, I still kept at 4% here.

Will Russia recognize Transnistria in 2022? 26% → 25%.

Russia dropped a massive hint that this is their next target and that their war aims in Ukraine reflect this, and Metaculus did not move, nor has it moved on Russia’s failures to make progress. Those point in different directions so I’ll stay at 20%.

Will there be 10,000 or more Russo-Ukrainian war deaths in 2022? Yes.

Will there be 25,000 or more Russo-Ukrainian war deaths in 2002? Yes.

Will there be 50,000 or more Russo-Ukrainian war deaths in 2022? 85% → 93%.

This also already happened. The uncertainty is whether or not the official count will register sufficient deaths, which could fail to happen if things somehow wound down soon.

Will there be 100,000 or more Russo-Ukrainian war deaths in 2022? 32% → 40%.

If the war continues as more than a frozen conflict, this will happen before the end of the year. That doesn’t mean the war will continue, and it doesn’t mean the official counts will recognize what happened. My guess is this has not yet updated enough.

How many civilian casualties will there be before May 2022? 7,430. I find Metaculus poor at providing insight into such questions, because it’s more about how the count is being done than about the facts on the ground, but it’s much better than not having an estimate of any kind. In general, ‘anything at all’ will almost always be more useful than ‘hell if I know.’

It also provides context to the questions about how many war deaths there will be, since this is an estimate of how many civilian deaths will be considered to ‘count’ there.

How much foreign aid will USA provide Ukraine in 2022? 16.4 billion (unchanged).

This hasn’t budged in over a month which reinforces my view that these distributions are bad idea. Numbers should have adjusted in some direction at some point but they didn’t because no one ‘can even’ in terms of putting in the work here. Better to use Yes/No markets.

Will YouTube be blocked in Russia in 2022? 74% → 70%.

I do see the back and forth demands and irritation, but if this was going to happen, why hasn’t it happened already?

Will there be a serious radiation incident at a nuclear power plant in Ukraine by 2024? 9% → 9%

This should have declined more as Russia withdraw from Chernobyl and generally made it clear it wasn’t looking to execute a scorched Earth strategy via nuclear plants. Marginally risky fights are still possible, but if you told me there was a ‘serious radiation incident’ my mind no longer jumps to a power plant.

Will Russia use Chemical Weapons in Ukraine in 2022? 50% → 53%.

The movement of this market over time does not make sense to me. If Russia was going to use chemical weapons, it already had a number of ‘opportunities’ to do so. The most likely spot was Mariupol, but after declaring victory and not using it outside the plant where remaining forces are while calling off the attack, it seems like they have decided against that. The commander who would do it has been in place for weeks. Every day they don’t do it, the chance they do it goes down. Also, the reaction to a plausible claim of use of such weapons was both measured (everyone waited to see if it was true, and it seems it wasn’t) and also made it clear such use would be viewed as a serious escalation. Instead, the probability remains elevated after the incident, which seems wrong to me unless people are baking in a chance it already happened and will be ruled that way.

Will 5mm+ Ukrainian refugees seek assistance overseas? Yes.

Will Finland join NATO by 2024? 90% → 90%.

At this point I do not know what stops them from doing so short of WW3. It is clear that every threat drives them to join faster, not to rethink their position. The combination of sabre rattling and inability to actually rattle a proper meaningful sabre went over quite poorly.

Will Sweden join NATO by 2024? 90% → 85%.

I don’t know what this decline is about but it seems wrong to me at least in magnitude. I guess I could possibly see Sweden deciding it doesn’t need to join, but it seems rather unlikely.

Will at least three European countries refuse to buy natural gas from Russia in 2022? Yes.

Will Putin be President of Russia on February 1, 2023? 82% → 85%.

Is there any way he’s not President and it’s not a ‘regime change or coup’? There must be, or the next market is going to contradict this one directly. I’m going to assume it’s not as clear cut as that in terms of descriptions.

Will a coup or regime change take place in Russia by 2024? 12% → 13%.

There are a lot of signs that Russians are very not happy with the lack of progress in the war, and that final failure would indeed get a strongly negative reaction. There are also signs of widespread support for the war. Still, this is a two year period, which has a not-that-low baseline rate of regime change or coups in such situations, and it seems highly unlikely the war ends in a way that is reassuring for Putin or Russia, especially given inflated expectations.

Will there be a bilateral cease-fire or peace agreement before 2023? 75% → 60%

This seems optimistic even with the downward shift. Talks have broken down as Russia has made it clear it has territorial ambitions beyond Donbass and Crimea that it intends to press and Ukraine becomes increasingly unwilling to compromise on territory given its successes combined with Russia’s war crimes. That doesn’t mean a cease fire is impossible, but it seems more likely that even if the war mostly ends it ends via an unofficial winding down of the fighting, as Russia can’t officially agree to a cease fire for internal reasons.

Will Ukraine formally recognize a former Ukrainian territory as independent before 2024? 19%

Question wording is weird because it implies that Ukraine recognizing Crimea as Russian wouldn’t count, only recognizing the Republic of Crimea, but I’m going to disregard that as not the intent. The real question is, how likely is it that Ukraine actually is willing to formally recognize the loss of Crimea? My instinct here is not likely. It is a big deal to them to formally do this so I do not expect it.

Will Ukraine formally pledge not to join NATO by 2024? 55%.

Such a declaration would be part of a formal peace deal, but short of that what is the interest of Ukraine in formalizing this? Why give that concession for any other reason, even if there is no practical hope of joining? I don’t see it, and I don’t see a formal peace as likely, and a lot of the times it does happen it involves a total victory by one side or the other that renders the promise less obvious or even unlikely.

Will Ukraine have a security guarantee from another country before 2024? 40%.

I continue to neither see this as useful or likely. And in the worlds where it is likely, it is even less useful.

Will Ukraine join the EU before 2024? 9% → 10%.

Ukraine is a candidate now, but it still seems like this timeline outright forces the EU to say ‘screw all the rules’ to make this happen, and that is exactly the thing the EU never says.

Will Russia have military control over any formerly Ukrainian territories other than LNR, DNR or Crimea on January 1, 2024? 70%.

This question was clarified to refer only to military control, including any group getting Russian assistance, with no minimum size requirement for territory or people involved. So there’s a lot of ‘backdoor covers’ here where the real answer is essentially no but this is a yes anyway. I do not like Russia’s chances to hold the land bridge or other major territorial concessions for this long a period, but it is definitely possible, and also possible that they hold some spec that’s technically not in Crimea or what not, so I can’t go that low.

Will Ukraine control at least 90% of the Donetsk/Luhansk Oblasts on Jan 1, 2024? 11%.

This is almost two years away, which is a long time. Ukraine is fully mobilized and determined to reclaim its territory if it can do so. If Russia’s military position collapses, they likely would be able to hold Crimea militarily anyway, but I would expect that they would lose Donbass entirely. I consider such scenarios highly plausible. There are certainly reasons for Putin/Russia to want to retreat back to the 2014 lines but stop there, and currently Ukraine is in no position to press an offensive, but it could quickly be a very bad idea to be a Russian force inside Ukraine.

Polymarket

The limitations of what real money markets can offer mean that a lot of these questions don’t seem as interesting as the ones asked by Metaculus.

In exchange, the answers for questions that are not mostly known already are far more reliable, since they are backed by real money, although in most cases the volumes are not that high.

The problem is that if one side is at 10% or less (and to some extent 15%-20% or less) the market gets distorted an unpredictable amount towards the unlikely outcome. Thus, it is difficult to calibrate on many of these questions once they enter that range. I suspect (but cannot prove statistically, haven’t done the research) that having previously been higher probability events makes this bias bigger.

Here are some of the more relevant numbers.

96% Zelenskyy is President of Ukraine on June 30

This gives Russia only a 4% chance of an assassination or a decapitation strike within two months. Seems overconfident.

11% Zelenskyy and Putin meet in person by June

This both seems high and seems high relative to its older values, which is the usual situation in such longshot markets. I don’t see how this meeting happens in time without precursors that are less likely than this.

9% Biden visits Ukraine by June

Not especially important, but noting that in markets trading in the single digits unreasonable amounts of hope spring eternal. This isn’t dead money but it’s too high.

85% Lukashenko remains president of Belarus through 2022

Lukashenko seen as about as vulnerable as Putin. He certainly is less popular and in some ways does look more vulnerable, but the only way I make sense of this being so low with him staying out of the war is if we think that Putin’s fall takes him out as well since the people would rise up against him. Doesn’t seem crazy, but 15% still seems like a lot here.

5% Ukraine formally cedes Crimea (or Luhansk or Donetsk) by May 31, 2022

This seems a lot like yield farming.

66% Will NATO expand in 2022?

So basically 90% at Metaculus by 2024, and 66% here by 2023. It seems odd to let things drag on that long, but the ratio doesn’t seem crazy.

26% Will Russia expand its federal subjects listed in its constitution by July 1?

This is presumably formal annexation of DNR and/or LNR. Doing so while there is still fighting over what territories they control seems like it doesn’t work, requires organizing a vote during a war fought in close proximity (even if it’s a fraud), and also makes a settlement essentially impossible so you’d want to be damn sure you wanted to go there. I don’t see real movement towards making this happen. Yet perhaps victory day is the time to do it anyway, or something like that. I’d sell.

66% Will Ukraine be an EU candidate country on July 31st?

Austria announced opposition recently and the EU generally moves slower than you can possibly imagine. Look at the dates on the candidates still on this list, wow. So I’d be inclined to always bet on things taking longer, but haven’t looked at how far in the process is required for this to count.

Noting that I also checked Insight Markets, but did not find anything enlightening.

Synthesis

Putting this together, in terms of longer term outcomes, here are the key items by 1 Jan 2024:

70% Russia controls territory beyond Donetsk/Luhansk/Crimea

55% Ukraine pledges not to join NATO

40% Ukraine gets a security guarantee

19% Ukraine recognizes loss of territory

13% Coup or regime change in Russia

11% Ukraine reclaims Donetsk/Luhansk

10% Ukraine joins EU by 1 Jan 2024

And also earlier:

65% Peace or cease fire by 2023

15% Putin NOT president of Russia 1 February 2023 (to line up with above)

53% Russia uses chemical weapons by 2023

3% Kyiv falls by 2023

Thus, we’d like to find a way to break this up into scenarios that add up to ~100% that aren’t too far out of line with any of these numbers, and see where that takes us.

If we take the 70% chance of Russia controlling other territory at face value we get this basic story:

5%: Total Russian victory

65%: Russia keeps some additional territory beyond Donbass/Crimea

19%: Ukraine does not give up additional territory but fails to reclaim Donbass

11%: Total Ukrainian victory, they reclaim 90% of Donbass

The 19% chance there seems too low to me if taken at face value, even if we accept the two extremes as reasonable. I would thus adjust for that by assuming that a large chunk of the 65% isn’t a full land bridge or other major carve out, but rather some tiny little incursion somewhere. Otherwise I don’t see this making much sense or at least being very overconfident.

Once you do that, it all seems more reasonable, so now we ask questions like when Ukraine gets its security guarantees, which seems to involve it getting such promises while substantial portions of Ukraine, that it does not recognize as lost, are under occupation. They are only 19% to recognize any loss of territory including Crimea, only 11% to get Donbass back. The other 70% (assuming zero overlap) of scenarios involve some kind of frozen conflict, at least on paper. Yet Ukraine gets assurances 40% of the time. Not sure how that is supposed to work.

It also seems like there is a clear combination of ‘if Putin falls then Russia doesn’t obviously withdraw from the war’ and also ‘if Putin withdraws from the war Putin often hangs on to power.’ I am skeptical.

6 comments

Comments sorted by top scores.