In January I held my annual Prediction Contest, asking people to forecast how likely they thought various events were to occur in 2025. There were 376 valid entries, making this the largest exercise of its type focused on the UK that I’m aware of1 - which in turn means the Wisdom of Crowds should have decent predictive power.
There were 45 questions, featuring UK politics and economy, global politics and world events and science, the arts and miscellany. Those looking for a fuller description of the contest can find it here.
The contest - and some headline results - has also been written up in the i newspaper in an excellent piece by Will Hazell, which also discusses forecasting more broadly, and its potential uses in government and elsewhere.
Here I’ll present the full ‘Wisdom of Crowds’ results - what people think will happen - and share my own predictions. For the results, of course, we’ll have to wait until January 2026.
There were 376 entries2, more than four times as many as last year - a big thank you to all of you who shared and promoted it. Nearly 1 in 3 (108) considered that they worked in politics, public policy journalism or similar, a group that included former MPs and advisers, think-tankers, journalists, senior civil servants and politics professors. 83% were male3 and 95% live in the UK.
What did people think would happen?
It should be emphasised that these results are not the result of a statistically weighted poll: they are simply the aggregation of a smorgasboard of people, ranging at one end from government advisers, policy wonks and journalists to, at the other, friends of mine with self-declared little interest in current affairs.
Nevertheless, there’s good evidence to show that if you average predictions in an exercise like this, the average will be better than most individual sets of predictions - and often pretty accurate. Different people have different information; positive and negative random errors cancel out, and so on
Last year, that’s exactly what we found. The Wisdom of Crowds (the average of all answers for a question) would have come 13th, outperforming 85% of individuals. It is possible to consistently beat the Wisdom of Crowds - at least three people4 have in both of the previous two contests, with the winner being the same person both times - but if you’re not sure who to trust, the Wisdom of Crowds is a good bet.
No, seriously, what did people think would happen?
The full results are here:
What does AI think will happen in 2025?
Last year
asked ChatGPT the questions on that year’s forecast. It got 0.214, just below the Wisdom of Crowds, and would have come 15th - putting it in the top 20% of contestants for that exercise.This year,
asked Perplexity+Deepseek to make a forecast based on the questions. What is perhaps most interesting is to see its chain of reasoning for each question, which you can do here - it reads to me like a smart, informed, person setting out its logic in a way that is terribly impressive5. We’ll see how it does at the end of the year.Thank you to everyone who took part - and remember to check back in next January to see the results! (This is the bit where I remind you to subscribe, if you’ve not already).
P.S. A note on the results. In previous years I just posted the top 10, with other people messaging me privately to ask for their position and score. This year, there are enough people who’ve entered that that isn’t practical - but at the same time I don’t want to rank everyone and have someone who did ‘worst of all’.
What I’ve decided to do is to list everyone who does better than chance - this is likely to be the top 65-70% of entries, using the aliases they used on registering. So that will give most people their positions while hopefully avoiding undue embarrassment to anyone in the bottom third(ish). If you were reading the form, you were asked to use an alias that ‘you don't mind me publishing on the internet in the event that you do well in the competition’ but if for some reason you’re having second thoughts, or put in your full name but now want to use an alias, feel free to message me and I will change this.
Feel free to inform me of other, bigger, ones.
There were a handful of people who wrote ‘yes’, ‘no’, and ‘maybe’ - though I have now been shown how to ensure Google Forms only accepts numerical answers so will be able to avoid this last year.
I don’t know how my overall readership breaks down, but one of the clearest immediate demonstrations to me that ‘men and women, in general, like different things’ is that around 60% of those who submit entries in the Christmas Quiz are women but that c. 80% of those who do the forecasting contest are men. We see this repeated across many areas of society, wherever people are genuinely free to make choices (and, counterintuitively, we see even greater disparities in countries such as the Nordics where there are higher levels of sex equality).
In consequence, almost any policy, judicial or HR processes that aim to get equal proportions of the sexes into any given profession or job role (as opposed to aiming at equal opportunities for those that wish to go into it) are self-evidently flawed - and their existence, if they have any teeth at all, should be considered as prima facie evidence of discrimination against the group(s) that are not being promoted in this way.
Including me!
Other than the fact that it thinks I am Scott Alexander and that there are large cash prizes available for this contest, both of which are sadly not true.
Interesting/surprising that several of your own predictions are so extreme (98, 98, 95, 5). You must feel very confident on those.
I love footnote 5 :)
I'm wondering if I've misunderstood the net migration question. Wasn't the previous year's figure close to a million? So are you saying you're 65% confident that it will have reduced by half (or more) this year? What's that based on?
Regarding footnote 3, which is more important - reducing discrimination or maximising effectiveness? If it were the case that diversity improved organisational effectiveness, or that organisation leaders thought it did, then aiming to increase diversity might well be discriminatory, but would be a rational attempt to maximise effectiveness. Would you still oppose it?