I make predictions every year to put empirical tests on my model of the world. I tend to do a lot of predictions, in order to get a larger dataset, and at the end of the year, I grade them. These were made last year in March. I’ve placed levels of confidence for each prediction with the odds I would bet on those outcomes in the vein of Bryan Caplan. I’ve created a chart at the end to show my calibration versus perfect calibration.
- Trump Approval Rating end of year <50% (Gallup): 95% ✔️
Trump Approval Rating end of year <45% (Gallup): 90%(was 45% exactly)Trump Approval Rating end of year < 40% (Gallup): 70%- US will not get involved in any new major war with death toll of > 100 US soldiers: 60% ✔️
- No single terrorist attack in the USA will kill > 100 people: 95% ✔️
- The UK will not leave the EU this year: 80% ✔️
- North Korea will still be controlled by the Kim dynasty: 95% ✔️
- North Korea will not conduct a nuclear test this year: 60% ✔️
North Korea will not conduct a missile test this year: 60%They conducted 10, with several launching many missiles- North Korea will not agree to give up nuclear weapons entirely, contingent on US troops staying in the Korean peninsula: 99% ✔️
- North Korea will not agree to give up nuclear weapons as a result of any negotiations: 90% ✔️
- Yemeni civil war will still be happening: 70% ✔️
- S&P 500 2019 >10% growth (from 2506 on Jan 1): 60% ✔️
S&P 500 will be between 2400 and 3100: 80% (80% confidence interval)was 3231- Unemployment rate December 2019 < 6%: 80% ✔️
- Unemployment rate December 2019 < 5%: 70% ✔️
- WTI Crude Oil price up by 10% (from $45.41): 70% ✔️
- Price of Bitcoin in dollars up over the year (Coinbase – 3823 Jan 1): 70% ✔️ was $7163
- Price of Bitcoin < $8,000 (does not double): 60% ✔️
- Price of Bitcoin > $1900 (does not lose half value): 70% ✔️
- Price of Bitcoin < $12,000 (does not triple): 70% ✔️
- Drivechain opcodes not soft-forked into Bitcoin: 80% ✔️
- No drivechains soft-forked into existence: 99% ✔️
- US government does not make Bitcoin ownership or exchange illegal: 95% ✔️
- Self-driving cars will not be available this year for general purchase: 95% ✔️
- Self-driving cars will not be available this year to purchase / legally operate for < $100k: 99% ✔️
- I will not be able to buy trips on self-driving cars from Uber/Lyft/Waymo in a location I am living: 95% ✔️
- I will not be able to order groceries on self-driving cars in a location I am living: 90% ✔️
- I will not be able to buy a trip on a self-driving car from Uber/Lyft/Waymo without a backup employee in the car anywhere in the US: 80% ✔️ This is tough. You can get self driving cars in Phoenix, but only if you’re part of the Waymo beta and so far they are free, so no “buying”.
- The artificial general intelligence alignment problem will not be seen as the most important problem facing humanity: 99% ✔️
- Humans will not be in lunar orbit in 2019: 99% ✔️
- SpaceX Falcon Heavy rocket will launch again this year: 90% ✔️
SpaceX will bring humans to low earth orbit: 60%- SpaceX will test the “Starship” mock up this year: 70% ✔️ (pretty sure I just meant this giant water tower thing, not a real launch)
- Mexican government does not pay for wall: 99% ✔️ (lol)
- Border wall construction not complete by end of 2019: 99% ✔️ (some construction occurred, mostly replacing existing wall)
- National Debt increases by >$1 trillion (from
$21,943,897,000,000): 90% ✔️ (was $23.201 trillion on Jan 1 2020) - There will not be a significant decrease in trade barriers between US and China from pre-2017 tariff levels: 90% ✔️
- Democratic RCP front runner will not be Bernie Sanders: 80% ✔️ (front runner on Jan 1 was Biden)
- Democratic RCP front runner will not be Kamala Harris: 80% ✔️
- Democratic RCP front runner will not be Beto O’Rourke: 80% ✔️
- Trump not removed from office or resign: 95% ✔️
Trump not impeached: 70%I was not expecting thisNo CRISPR edited babies will be born: 80%(it turns out the researcher responsible for the two 2018 CRISPR edited babies had already treated a third unborn child in 2018 when the story broke. Apparently the third baby was born in 2019 if you carefully read Xinhua, so technically this prediction is wrong, although I meant no other researcher would do anything. Remember to properly word your predictions!)- No full year US government budget will be passed (only several months spending): 90% ✔️ (they basically only do continuing resolutions now)
- Some tariffs raised: 90% ✔️ (like a bunch)
- Trump administration does not file a lawsuit against any news organization for defamation: 90% ✔️
- I got 4 of 6 predictions correct at 60% confidence
- I got 7 of 9 predictions correct at 70% confidence
- I got 7 of 9 predictions correct at 80% confidence
- I got 8 of 9 predictions correct at 90% confidence
- I got 7 of 7 predictions correct at 95% confidence
- I got 6 of 6 predictions correct at 99% confidence
Overall, not bad at all, and we should note that from last year’s grading, my 60% confidence predictions have tended to be overconfident. I only had 6 of those predictions this year, so actually 66% is the closest I could have been to perfect calibration. 70% also ended up being a bit overconfident, but a single additional missed prediction here would have dropped me down to 66% as well. Had I moved one of my correct 70% predictions to 80%, I would have been perfectly calibrated.
Combining this data and data from last year gives:
- I got 11 of 15 predictions correct (73%) at 60% confidence
- I got 14 of 20 predictions correct (70%) at 70% confidence
- I got 14 of 17 predictions correct (82%) at 80% confidence
- I got 14 of 15 predictions correct (93%) at 90% confidence
- I got 16 of 17 predictions correct (94%) at 95% confidence
- I got 10 of 10 predictions correct (100%) at 99% confidence
In 2018, as I noted in the post last year, I should have made some of my 60% predictions at a higher confidence, but other than that, these predictions are remarkably well calibrated if I do say so myself.
I hope to post my 2020 predictions soon.
You must be logged in to post a comment.