Predictions for 2022

for Scotland, the UK and the world

Here I will make some predictions for 2022, with probabilities for each one. At the end of the year I will score them. All the below predictions imply "during 2022", unless something else is stated. For some of the predictions I've added notes in italics giving my reasons. Each predictions has a code in [...] to use when I score them at the end of the year.

Scottish Politics

[Sec30] The Scottish government will ask the UK government for a section 30 order to facilitate an independence referendum: 30%. They might, but the way they are dragging their feet I'm not too hopeful.

[GrantSec30] Conditional on the Scottish government asking for a section 30 order, Westminster will grant one: 15%. Westminster wants to hang on to Scotland. they have to weigh the damage that a referendum might do (if the result is pro-independence) versus whether they will look intransigent to other countries by not offering one.

[IndyRef] There will be a referendum on Scottish independence: 3%. The Scottish Government are dragging their feet on this.

[Indy] Conditional on there being an independence referendum, independence will win: 65%. The main reason I place this so high is that (as in 2014) the unionist argument will boil down to "Scotland is too crap to rule itself", which many will find grating.

[UDI] Scotland will unilaterally declare independence: 1%. If this does happen, it will happen after 2022, when the Scottish government has exhausted other avenues.

[NSFM] Nicola Sturgeon will be First Minister of Scotland on 31 December 2022: 90%. I don't see her resigning this early into a Scottish parliament.

UK politics

[GE] There will be a general election in 2022: 15%. The Tories are unlikely to call one unless they think they're very likely to win. The most likely way this could happen is if Johnson is replaced and the new PM calls a snap election during their honeymoon period.

[BJPM] Boris Johnson will be prime Minister on 31 December 2022: 35%. The Tories are ruthless about getting rid of underperforming leaders.

[KSLab] Keir Starmer will be leader of the Labour Party on 31 December 2022: 90%. Labour are less ruthless than the Tories at getting rid of leaders, and if Johnson is looking bad, Starmer starts to look good by comparison.

Covid

The number of UK covid deaths in 2022, according to Coronavirus update will be:

• [Covid20k] greater than 20,000: 80%

• [Covid40k] greater than 40,000: 45%

• [Covid80k] greater than 80,000: 20%

The world

[InvUkraine] Russia invades Ukraine: 30%. Russia might well invade Ukraine. though I suspect Putin's strategy is more to wear down Ukraine and the West by making people think it is likely, then not doing it, then doing it later.

[InvTaiwan] China invades Taiwan: 3%. I don't think China will invade Taiwan in 2022, though they might if they think the USA is very weak or if Xi feels the need to divert public opinion from domestic toubles.

[EMFP] Emmanuel Macron wins the 2022 Frence presidential election: 65%. His three main challengers are all to the right of him, so in the 2nd round he'll pick up most of the left vote. Also, Macron wins pairwise opinion polls against Marine Le Pen, Valérie Pécresse, and Éric Zemmour.

[Xi] Xi Jinping is still in power in China: 96%

[Putin] Vladimir Putin is still in power in Russia: 94%

[MbS] Mohammed bin Salman is still in power in Saudi Arabia: 93%

[Biden] Joe Biden is still in power in the USA: 95%

How I will score these

Imagine if there were 100 independent events, and you predicted that each one would happen with a 20% probability. Then you would expect that 100*20% = 20 of them would happen. if a lot more than 20 happened, the 20% probability you assigned to them would be too low; conversely if a lot fewer than 20 happened, the 20% probability you assigned to them would be too high. If the number that happen is close to 20, then you can say you have good predicting power; if not you need to adjust the likelihood of the predictions you make.

A probability of y% that event X will happen is also a probability of (100-y)% that X won't happen. For events with a probability of >50%, I will invert the prediction to get a prediction in the range 1-49%. I won't predict anything at 50% as if I did I wouldn't know whether to invert it or not.

Then I will put the predictions in buckets, 10 points wide: 0-10%, 10-20%, 20-30%, 30-40%, 40-49%. Consider the 20-30% bucket. If I make 5 predictions at 20%, 2 at 25% and 4 at 30% then that's a combined 5*0.2 + 2*0.25 + 4*0.3 = 2.7 so I'd expect about 2 or 3 of those events to happen.

At the end of 2022 I can rate myself by seeing how accurate the predictions were. See Scott Alexander's graph for his 2019 predictions (note, though, that I'm using 0-49%, whereas he's using 50-100%; this results in the same graph, just rotated 180 degrees around the (50,50) point).