My COVID-19 Forecasts for Early 2022
Predicting COVID-19 booster numbers, case numbers and deaths
I recently started reading the books Superforecasting by Phil Tetlock and Dan Gardner, but I got a bit bored with the pop science style of too many unnecessary anecdotes and started reading Expert Political Judgement—the more academic version of Superforecasting—by only Tetlock instead. It’s better! Both books are about the poor performance of experts in predicting future events and what you can do to be a good forecaster—called ‘foxes’—rather than a dogmatic and stubborn forecaster—called ‘hedgehogs’. I wanted to try my luck, so I’ve been using the website Good Judgement Open. From the FAQ:
GJ Open is a crowd forecasting site where you can hone your forecasting skills, learn about the world, and engage with other forecasters. On GJ Open, you can make forecasts about the likelihood of future events and learn how accurate you were and how your accuracy compares with the crowd. Unlike prediction markets and other forecasting sites, you can share your reasoning with other forecasters to challenge your assumptions.
There are all kinds of different questions about financial events and geopolitics that I’ve made predictions on, but I wanted to highlight my COVID-19 predictions. I am optimistic about COVID in the near future.
Zvi Mowshowitz of Don’t Worry About the Vase receives high praise for his COVID-19 predictions around the rationalist community—especially from Scott Alexander who has said “I will never bet against Zvi on anything” in an article about scoring predictions. So, it was interesting to me to hear that Zvi had made a wager about COVID mostly running it’s course by the end of February 2022. This would be great if it is true.
The bet that was made was between Zvi and Holden Karnofsky and it reads:
If at least 75% of the USA COVID-19 cases between 1/1/22 and 2/28/23 (inclusive) occur between 1/1/22 and 2/28/22 (inclusive), I pay Zvi $40.
Otherwise, Zvi pays me $60.
This bet is intended to apply to Omicron and earlier strains, and it will be a “push” if a post-Omicron strain “muddies the waters” in the following sense: counting cases from the new strain would cause me to win, and not counting them would cause Zvi to win.
We'll use Wikipedia for total COVID-19 cases and this CDC data for variant information. Each of us has the option to appeal to a third party (whom we've agreed on) to perform an adjustment for undertesting.
Others in the rationalist sphere are on Zvi’s side and Razib Khan recently remarked “I'm with Zvi.” This all provided me with at least some confidence that we might see COVID-19 coming to a close thanks to a spike in Omicron. The high numbers actually makes me happy since it isn’t paired with a lot of deaths. The data seems to be matching Zvi’s prediction. We are currently (January 10th, 2022) seeing an unprecedented spike in COVID-19 cases! So, I’m thinking I’m with Zvi too.
On GJOpen, the crowd isn’t as much with Zvi on this so I used it as a chance to prove my forecasting skills are better than the crowd and improve my Brier skill score. Here are the questions:
I think that before February 1st, 2022 is unlikely. But if Zvi is correct is is pretty likely we will see it occur before March 31st. It’s currently around 700K now and would take a sharp drop off to reach 50K. I think that is what is going to happen happen after Omicron has gone through a large enough portion of the population. What goes up has to come down and it is going up very quickly, so I think it will fall very quickly. That fall will probably happen before March. As you can see, the crowd does not agree with me. They are more pessimistic and think it’ll most likely fall between April 1st and May 31st. Not outrageous.
I put my estimate as mostly in “not before 1 July 2022” because I don’t think 165 million people will ever get the booster. I explained on GJOpen:
As of Jan 4th, 2022 there are 204M fully vaccinated individuals and 68.7M with a booster. [1] If you are unwilling to get the vaccine at this point, you are very unlikely to get the booster. A lot of people became vaccinated because of mandates and pressure. I think COVID-19 will peak in Jan-Feb and then fall off steeply. Pressure to get a booster will subside. Also, since Omicron will be more mild, people will be less inclined to get the booster. If there is a booster for Omicron, they will get that instead as we get farther along.
One very prolific forecaster Anneinak responded to my comment:
Welcome to geo-political forecasting on GJO! Thank you for following me.
You've given some good arguments to support your forecast.
One thing I'm wondering about is booster eligibility. When did we have 165m double-vaccinated people? When will they even be eligible for a booster? It seems to me that the booster curve will probably look similar to the double-shot curve, only offset by about an 6 month delay.
To which I responded:
Thank you very much for welcoming me.
Others below have discussed the 6 month and now 5 month delay for Pfizer [1]. I am skeptical that many will rush to violate the law to prematurely get the vaccine, so I will treat the 5 month delay as a minimum. Even those who are willing will very possibly delay getting the booster because people naturally procrastinate. There is probably a population which anxiously seeks out the booster but they also anxiously sought out the vaccine. They would've already received the vaccine and the booster. Boosters will taper off as vaccines did. We can kind of see this in the data for other countries. [2] It appears to be starting for the US but the data isn't clear enough and holidays probably prevented people from getting boosters.
July 20th, 2021 was the date in which 165 million people were considered fully vaccinated. If the delay is 5 months then the earliest possible resolution if people follow the guideline would be January 20th but not everyone has Pfizer so it would be between January 20th and February 20th. That is probably the absolute lower limit. A major change in this would be a further reduction in the recommended waiting period. This is not outside the realm of possibility. "Countries such as the United Kingdom, Greece, South Korea, Germany and Israel are now allowing people to get the third dose just three months after their second shot. " [1]
The push to get everyone vaccinated came with a great deal of social pressure and incentives. The booster curve will not follow closely behind the double shot curve in my view (90%). We had to have mandates and discrimination against the unvaccinated. People feel less threatened by the omicron variant. If we look at deaths, there is not a steep rise despite the very steep rise in cases. News won't be able to note this and make people feel afraid enough to get the booster. The booster is also not specifically for Omicron. Many may consider the option of waiting for an Omicron booster.
I think that the extreme social pressure and legal mandates were necessary to get to the 204 million. I think that people are less concerned and less willing. Those that were vaccine hesitant are probably even more vaccine hesitant. Those that were hesitant but faced losing their job are still hesitant and possibly resentful. The main benefit of reduced severity came with the 2 vaccines. I think that people will see much less reason to get the booster. Even for people that believe they should get the booster, people simply put things off until a later time.
This is also partially conditional on my COVID-19 estimation. I am estimating a steep increase with a peak around mid January and harsh fall off. A reasonable person on COVID-19, from what I understand, Zvi Mowshowitz bet that: "at least 75% of the USA COVID-19 cases between 1/1/22 and 2/28/23 (inclusive) occur between 1/1/22 and 2/28/22 (inclusive) " If this is correct, then people will be much less inclined to get the booster after cases fall off so steeply. It won't feel necessary. If I am mistaken on this point, I would reduce my confidence.
[1] https://www.nbcnews.com/news/us-news/pfizer-booster-five-months-cdc-rcna10879
[2] https://ourworldindata.org/grapher/cumulative-covid-vaccine-booster-doses?country=BRA~DEU~RUS~GBR~USA
I stand by my estimate as of January 10th, 2022.
This is where I diverge the most from the consensus which is pretty dangerous. There is a consensus for 0% on under 900,000 deaths and I put it at 50%. I am much more optimistic. I’m not seeing the increase in deaths which we would expect to see with such a high caseload, but this may have been to ambitious of an estimation. I may need to update.
My comment providing my justification on January 4th:
As of January 3rd, 2022 we have a 7-day average of ~1300 deaths per day with a 7-day case average of ~481,000. I believe that we are approaching the peak soon and we will see a sharp decline in the total. The total deaths as of January 3rd, 2022 are 826K.
A consistent amount from January 3rd, 2022 to May 1st, 2022 gives us 117 days * 1300 deaths for an additional 152,100 deaths. This puts us at 826K + 152K = 978K. My COVID-19 estimation is a rapid rise and rapid fall. Omicron acting as a mandatory vaccine for a large portion of the population and preventing transfer of other variants for a while. Having such high case rates and low death rates relative, makes me very optimistic. I don't see a tremendous rise in deaths awaiting, but I do anticipate a big decline in deaths. Pfizer's treatment, milder cases, deaths not peaking, 2 vaccines + booster - should be more optimism.
I'm having a hard time seeing 1 million + deaths as likely but I could be totally wrong about the trends. If we average <632 deaths we will remain under 900,000. The other period where we kept it under this level was about May 2021 - Aug 2021. I don't see this as THAT unlikely compared to the other forecasters.
We will see what will happen! I’ll probably post to score my predictions later this year.