Update on my predictions for 2022
At the end of 2021 I started this Substack, and I kicked it off by writing some predictions for the year. This is the first time I’ve made commitments to predictions publicly, and even though I don’t think the format worked all that well I think the accountability was beneficial.
In my view, one of the best effects of this kind of exercise is not actually the attention paid to exact numbers and probabilities, but the simple fact of paying attention to facts that would otherwise get missed over the course of the year. There are several places below where I learned about events that had taken place in the previous year only because I took the time to web search in order to grade my predictions. It’s a good reminder that the news cycle moves fast, and memories fade fast.
I’ll first talk through my predictions and how they turned out.
Covid-19
Additional domestic restrictions of some kind imposed in England
by end of January: 90%: No
In 2022: 99%: No
UK government mandates full closure of pubs and restaurants (excluding take-away service) in England for some period of time > 1 week
by end of January: 60%: No
In 2022: 75%: No
Some English citizens in lockdown (only allowed to leave their home for specified list of purposes) for some period of time
by end of January: 25%: No
In 2022: 60%: No
Wrong on all of these (although two of them were non-committal 60%, and one was a prediction against). What went wrong here? At the time of writing, Covid cases were rising and it was implausible that vaccines or treatments were going to prevent the impact on the health system being significantly better than the previous wave.
I think what happened was that the Omicron variant led to less serious symptoms, and also that a fast rising wave is also one that declines relatively quickly after its peak.
Even so, there’s no excuse for being wrong on a 99% prediction. This might be an example of my prediction being skewed by wishing to be proved wrong in a pleasant way. Restrictions that you’ve mentally prepared for are better than the alternative.
To my credit, I think at the time most people I spoke to were talking about coming lockdowns, and I was correct to doubt that that would happen. I could have been bolder, though.
Patients admitted to hospital per day (7-day moving average) doesn’t exceed:
2000
In January: 75%: No
In 2022: 50%: No
1500
In January: 60%: No
In 2022: 10%: No
1000
In January: 25%: No
In 2022: 5%: No
I didn’t clarify England vs. UK; I think at the time the UK government were still publishing UK-wide statistics so let’s assume that.
All these thresholds were in fact exceeded, although the threshold of 2000 per day was only very narrowly exceeded on one occasion in January, so if you’re prepared to squint a bit and give me the benefit of the doubt on the England vs. UK thing, my predictions were pretty reasonable here.
Bear in mind that I wrote these predictions well before 31st December, when numbers were going up sharply and a few days made a lot of difference; the numbers were also being published on several days behind the date in question, so given the information available to me my predictions were much less safe than it would first appear.
UK 1st dose vaccination rate (population 12 and over) exceeds:
90%
In January: 95%: Yes*
92%
In January: 40%: No
In 1st quarter: 50%: No
95%
In January: 10%: No
In 1st quarter: 40%: No
By the end of January, first dose vaccination in England was maybe 43,185,000. The best I could get to for population over 12 (from here) was 48,000,000. That comes to 89.9% by the end of January, and I’m going to allow myself the small amount of leeway.
I think I correctly predicted that the rate was going to plateau after that. It really did, look at this graph:
Omicron variant is widely regarded to have a lower case fatality rate than Delta:
By end January: 40%: Yes
By end Q1: 60%: Yes
It made it into The New York Times by early January. We can quibble whether this was widely accepted and when, but I think it’s reasonable to say that it was widely accepted by the end of January.
As before, at the time I made these predictions I was sticking my neck out more than it might appear, given how quickly this changed.
UK rolls out a vaccination programme involving a vaccine targeting a different strain (not original Wuhan variant)
By end Q1: 25%: No
In 2022: 75%: Yes
I missed this news, but the September booster program delivered a bivalent vaccine.
A diagnostic test or treatment for long Covid will be approved in the UK during 2022: 60%: No
At some point in 2022:
Germany sees daily case rate exceed 75k (averaged over a week): 40%: Yes
China sees daily case rate exceed 5k (averaged over a week): 50%: Yes
Greece sees daily case rate exceed 7.5k (averaged over a week): 40%: Yes
None of these were good predictions. They came out of an opinion I had that the countries with unusually good outcomes up to that point had not in fact found a permanent solution, but were likely to have problems down the line and revert to something more like the mean. I completely failed to put this in the form of a testable prediction, but I stand by the fact that this opinion was at the time slightly contrarian and in fact right.
Of course, I completely failed to grasp quite how much Omicron was going to be a whole new ball game in terms of numbers. I also failed to foresee the China situation in any detail, beyond “zero Covid probably won’t deliver”.
UK politics
Boris Johnson remains leader of conservative party
For 1st 6 months: 75%: Yes
For all of 2022: 60%: No
Keir Starmer remains leader of labour party
For 1st 6 months: 95%: Yes
For all of 2022: 90%: Yes
There is no general election in the UK: 90%: Yes
I was a little pessimistic about Boris Johnson surviving in post for the whole year, I could perhaps have stuck my neck out a bit more here.
Local elections / by-elections in first 6 months show a clear swing against the Conservatives: 90%: Yes
I’m going to give myself this one:
It wasn’t quite the whitewash that some would have predicted, but I think it counts as a clear shift.
No clear evidence comes to light (photographic or on-the-record named source personally witnessing) of Conservatives attending an indoor party (more than 8 people) in Winter 2020/21
Boris Johnson: 75%: No
Any member of cabinet at the time: 50%: No
Firstly, I’m not sure why I bracketed this to “Winter 2020/21”, except that all the discussion in the press had been about Christmas parties. I think I was wrong about this, but only because the report by Sue Gray made everything public. I haven’t checked, but I don’t think in the time after I wrote that anyone went on the record to the press.
In my defence, at the time I was writing the speculation was all about some whistle-blower going on the record and presumably triggering a huge shift in public opinion. I think I was right in my estimation that this wasn’t going to happen; my reasoning was that by the end of 2021 anyone who was going to come forward had already come forward. I suspected that there had been parties, but I also suspected that people who hadn’t gone on the record had some reason not to do so, and weren’t just waiting for their moment.
So perhaps I was correct on some loosely defined ideas here, but I wholly failed to turn it into a bulletproof prediction.
Supply chain disruption credibly linked to Brexit is no worse in 2022 than in 2021: 80%: Yes
It’s a bit hard to judge this one, but a web search on “brexit supply chain disruption” turned up the top 20 most relevant articles being in 2020 or 2021, none in 2022. I think if it had continued to get worse and been credibly linked, we would see more relevant articles.
At least one new trade deal between the UK and a country >1million population: 90%: Yes
There was an agreement signed with New Zealand in February 2022. There was also a “digital trade deal” with Singapore signed in February 2022.
UK has no new trade deal with any of US, China, India or Brazil: 75%: Yes
Inflation becomes a significant political issue in 2022 (explicitly framed as an economic phenomenon, not simply complaints about the cost of living): 75%: Yes
I think I’ll give myself this one. See e.g. The Guardian here with a page talking about inflation rates. I think I was expecting that there might be more of an undeniable runaway price-wage spiral, rather than just talking about real economic shocks such as the (unknown at the time of my predictions) war in Ukraine. Arguably I failed on this one, but it was sufficiently vague that I think I’m going to give myself benefit of the doubt.
UK does not enter recession during 2022: 90%: Yes, technically
Technically the last quarter of anaemic growth saved us. But also my prediction was clearly failing to properly account for the risks, so call this a bad prediction.
US politics
A bill nominally claiming to be the “build back better” bill is passed into law: 60%: No
I was expecting that if the bill didn’t pass in its existing state it would be rewritten into something that would pass, hence the “nominal” part of my prediction. I think it’s fair to say that within these bounds, this didn’t happen.
2022 mid-term elections:
Democrats lose control of the senate: 90%: No
Democrats lose control of the house: 50%: Yes
This is an embarrassing one. I think many other pundits would have said the same thing, but that doesn’t really excuse me.
Donald Trump is not allowed back onto Twitter: 75%: No
I completely failed to foresee Elon Musk here. I think had the existing management stayed in place he would have stayed banned, but a bad prediction is a bad prediction.
No new anti-censorship (e.g. proudly including Donald Trump) social network launches that maintains 1m+ regular users: 75%: Yes
The word “new” is doing some work here, because I think it rules out Parler and Gab since they existed when I wrote that.
Of course the question is whether Truth Social counts here. I specifically used the word “maintains” because social networks can often have a burst of signups but it doesn’t really count unless they stick around. The numbers aren’t easy to get hold of, but Wikipedia quotes 1.7 million unique visitors (not the same thing as users with accounts) in September 2022, and also says that it was falling down the App Store rankings from June to October, ending up #75 in the social networking category.
It’s arguable, but I’m going to give myself the benefit of the doubt here.
Interest in Donald Trump (as measured by Google Trends) remains lower than at any time during his presidency: 75%: No
Hmm, I really had to write “at any time”, didn’t I? I was certainly right that interest in Donald Trump remained markedly lower. But there was a spike in November 2022 around the mid-terms that exceeded the lowest point in interest in 2019, so I got this wrong as stated.
No credible non-partisan news sources report that Donald Trump is monetarily funding particular candidates or particular issues in the mid-terms: 75%: Yes
I couldn’t find anything, not specifically about monetary funding during the midterms. There’s lots of discussion of Trump’s endorsement. There’s this article from early in the year suggesting that Trump’s PAC had supported some candidates by around $200,000 so I suppose it’s likely that at least some money was donated. But the people who were talking as if Trump would use his fortune to buy political influence were mostly wrong, I think.
US does not enter recession during 2022: 90%: Yes
Joe Biden’s approval rating:
Remains net negative throughout 2022: 80%: Yes
Remains above throughout 2022: 60%
I think this last one was supposed to have a number in it, but I don’t know what number I had in mind. I guess I’ll have to ignore it.
World politics
No military contact (people dead non-accidentally) between China and Taiwan: 90%: Yes
No new armed conflict involving a NATO power: 75%: Yes
I certainly didn’t foresee the conflict in Ukraine, even though I think there was some speculation about it by the time I was writing that. But I think I was right, given that I bounded to just NATO powers.
Direct flights between the UK and China resume: 75%: Yes
There aren’t a whole lot of flights, but according to this some direct flights restarted in August 2022.
Travel from the UK to China remains very difficult (e.g. 14 day quarantine) throughout 2022: 60%: Yes
Lucky escape on this one. The rules changed 8 days into 2023.
No new country establishes full diplomatic relations with Taiwan: 90%: Yes
Winter Olympics goes ahead on schedule: 99%: Yes
Football World Cup goes ahead on schedule: 95%: Yes
Technology
Level 3 self-driving cars are not being sold commercially and used anywhere in the world: 90%: Yes
VR headsets remain niche: 80%: Yes
This is a bit of a judgment call. There’s been some progress in 2022 I suppose, but I don’t see any qualitative difference in how niche they are.
No widespread major improvement in video call technology: 75%: Yes
No new country declares bitcoin to be legal tender: 90%: No
Well I never. The Central African Republic went and did it. It’s hard to imagine that this policy change has had any real impact, but I didn’t put any kind of “major country” or “bitcoin is actually in widespread use” qualification on that, so I lose this one.
Scores
50%: 2 right, 3 wrong = 40%
60%: 4 right, 9 wrong = 30%
75%: 15 right, 5 wrong = 80%
90%: 11 right, 3 wrong = 80%
95%: 4 right, 0 wrong = 100%
99%: 1 right, 1 wrong = 50%
Conclusion
It’s hard to be both interesting and accurate. I didn’t make many 99% predictions (and hence the numbers were useless) because I didn’t want to write a long list of uncontroversial things.
Qualitatively, it felt like a weakness was me trying to be contrarian, being somewhat right on the facts but not having written a sufficiently water-tight condition to make that come out in evaluating my predictions.
The main purpose of this was for me to learn from it, and I guess I have. Overall I was over-confident, I haven’t calculated standard deviations but I doubt any of my scores were under-confident in a statistically significant way.
I’ve moved all my prediction activity to Metaculus, where I’m currently ranked 264th on the site. Maintaining predictions that are like-for-like comparable with other people has been a good practice for me, even though the positive-sum scoring system on Metaculus makes it too easy to get ahead just by making obvious predictions on a huge number of questions.