[sumtechguy]

Most economic models are very nice in the steady state. If you get any sort of irrational actor or unaccounted variable in the model the model usually goes weird. The whole thing is at least s 20 dimensional graph just for the data. Never mind the hundreds of other things that feed it each of those systems some of which are impossible to model. It is a challenging problem but not really 'dialed in' and able to make decent predictions more than a few moments out. It is usually very good on explaining what happened but usually very ham fisted on predictions. Every once and awhile someone's model will 'get it right' at that point they go on book tours and predict the next disaster which may or may not happen. (source: degree in econ)

[huitzitziltzin]

> Most economic models are very nice in the steady state. If you get any sort of irrational actor or unaccounted variable in the model the model usually goes weird. The whole thing is at least s 20 dimensional graph just for the data.

If you're saying "it's hard," then I agree. And every academic macroeconomist would say the same thing.

> Never mind the hundreds of other things that feed it each of those systems some of which are impossible to model.

This also sounds like "it's hard," OR "you can't model everything." You can then say "and so I give up." Macroeconomists do not and for good reason: policymakers are going to use some kind of model to predict the impact of policies, or to select which policies to implement. We can either do our best to try to inform them on the basis of data, or we can throw up our hands, in which case they might well make worse choices.

> Every once and awhile someone's model will 'get it right' at that point they go on book tours and predict the next disaster which may or may not happen.

I do not see this happening personally. Nor do I think a lot of academic economists are in the business of going on book tours making confident predictions.

[sumtechguy]

The rub with >policymakers are going to use some kind of model to predict the impact of policies, or to select which policies to implement

It is the policy makers are not aware of the 3rd or even 2nd level effects on those decisions. The models do not show it to them because many cant, or they do not want to see it. There are also enough different theories that they can pick whatever sounds nice and fits what they want to say and then can lean back and say 'see the model said'. My point is they are not going to follow the 'science' they are going to have smorgasbord of whatever pet theory they want to promote. Then back the 'science' into it.

>This also sounds like "it's hard," OR "you can't model everything." You can then say "and so I give up."

What I am saying is the models are borderline not working. They 'sorta' work right up until you get an irrational actor (see recent stonks issue as an interesting case study). People are irrational but rational in a different dimension. But we have no real good way to model that. It is why almost all of these theories 'work' until you get something irrational that the model does not account for.

I am also not saying 'give up'. I am saying you need a lot more dimensions in your calculus. I am also saying many of those dimensions you will have a very hard time measuring. That is due to other external dimensions affecting those hyper dimensional curves, and even the model bending back on itself affecting things. It will also not be something you can keep in your head. Also at this point you will have to explain it in a way people can understand (any way else is the way of kafka). There are many years of theories that sound nice but do not work.

>I do not see this happening personally. Because in the majority of the cases it does not happen. Because the models usually get it wrong. But every once and awhile someone hits the lottery and leans into it. Usually around market crashes.