Not if you have any training in a hard science: then you think of economics as something largely devoid of equations and made up of lots of ad-hoc regression analysis and hand waving. This might not be the fault of the economists but of the difficulty of the subject matter of course. But economists really should learn to accompany their predictions with error bars as every undergraduate physicist must.
Edit: I think Ive had my mind changed enough in this thread to edit this. It's obviously a very deep subject and I've been convinced it has predictive qualities. I definitely dont think of dry equations though, more like psychology, and politics. But I'm looking forward to learning about it and possibly having my mind changed yet further.
Or let’s say the president of Japan announces a plan to beat deflation once and for all. He says that next year, he’ll fire their central bankers and hires the ones from Venezuela, give them a 100% annual inflation target, and promise them millions of USD in bonuses if they meet or exceed that target.
Even though this plan is going into effect next year, what will you see the next day? (Assuming people believe the president is serious.) Econ predicts you will start seeing inflation immediately. And we see effects like these all the time when looking at how the actions of central bankers effects the economy – if Powell says he’s going to change interest rates, there’s a reaction in the markets when he makes the announcement and not one when he makes the actual change.
That depends on whether the stock he bought was GME.
There was even a story (no idea if true, so maybe of limited relevance...) of a reverse bank run happening in Japan, with people rushing to deposit money to save a failing bank.
Or can be demonstrably arbitraged to take advantage of the irrationality in a tangible, quantifiable and repeatable way. Put-call parity, substitution, et cetera.
Virtually all published papers in economics that use regression analysis which you mention include "error bars." The discipline calls these standard errors and confidence intervals. If a paper is non-parametrically estimating functions rather than single parameters, they might also report confidence bands.[1]
Standard errors may be calculated analytically (resting mostly on asymptotic theory) or numerically (using a bootstrap estimator for them). There is a large literature on the appropriate standard errors for various contexts, with clustered errors being mainstream in applications.[2][3] These are taken seriously in peer review, and published papers come with long lists of robustness checks which turn papers into book-length publications when you include the appendices.
[1] https://cattaneo.princeton.edu/papers/Cattaneo-Crump-Farrell... is one example.
[2] See https://economics.mit.edu/files/13927 for a recent discussion.
[3] See https://gregoryeady.com/ResearchMethodsCourse/assets/reading... for bootstrap-based methods.
Honestly the boring answer is that e.g. liquidity is the ability to change your mind instantly, which also means that any prediction can become invalid at any given moment.
If you want to increase predictability, you would have to encourage people to make decisions as soon as possible and stick with those decisions for as long as possible.