3 ms·
> Predictions about anything are vulnerable to change from new information. The problem with predictions is that an event might be 99% likely to happen, but it
by ogogmad 4y ago
> Predictions about anything are vulnerable to change from new information.
The problem with predictions is that an event might be 99% likely to happen, but it might still be prudent to consider the 1% of cases where it doesn't happen if that's sufficiently high impact.
- spywaregorilla 4y agoConsidering a 1% event doesn't mean the best course of action still isn't to assume the 99% is going to happen. If there's a 1% chance of rain, the expected value of amount of rain is "a little bit of rain" but the forecast is still most usefully described as "no rain".
- ogogmad 4y agoThe point is that the magnitude of the outcome needs to be considered. Forecasting is about finding high probability outcomes, while not considering high-magnitude but unlikely outcomes. This makes forecasting sometimes a questionable activity.
- spywaregorilla 4y agoHow should the Fed adjust it's policies based on the San Andreas fault, solar storms flaring up electrical wires globally, China invading Taiwan, the Zaporizhzhia nuclear plant in Ukraine suffering a meltdown, a deadly covid resurgence, or zombies? Forecasting is always a questionable activity. All models are wrong, some are useful.
- ogogmad 4y agoI don't know how serious these things are, but if the San Andreas fault can tens of thousands of people (I doubt it, but I haven't looked into it), then indeed this becomes the priority. Economic policy should indeed focus on trying to prevent this outcome. And if solar flares can cause a megadeath (and again, I don't know), then this becomes a priority too. Your attempt at proof by contradiction might be a list of things to prepare for. > All models are wrong, some are useful. A forecast that causes you to act overconfidently is worse than no forecast. All models are wrong, but some are worse than nothing.
- spywaregorilla 4y ago> Your attempt at proof by contradiction might be a list of things to prepare for. It's not a proof by contradiction. It's showing that there are many long tail risks that are potentially very dangerous but also very expensive to try and prevent. It's not wrong, it's just a really difficult concept. The Fed interest's rate is a single tool that can be moved up or down. What do you do with no model? How do you make decisions? Every decision, including making no changes, is very significant.