Astronuc said:
Isn't this really about academics/theoreticians vs those (practitioners) in industry?
No it's about people that are clueless versus people who aren't.
How does one measure success? By salary? By success of one's predictions?
This is a problem in economics. If you do something wrong, and you end up blowing up the world, then obviously you were wrong, but unfortunately you would have blown up the world, and there are no do-overs.
However, if you have an economics professor that insists that the sky is pink and therefore we need to dump cyanide into the water, and when you look up and it looks blue, then it's probably a good idea to ignore what that person is saying.
One thing which is funny is when someone says that you don't exist. I've seen stupid articles from the University of Chicago saying that there was no market panic after Lehman collapsed. Except that I was there, and there was total panic in the markets until the Fed intervened and calmed everything down. One thing that is sort of interesting is how calm a market panic looks like. You actually don't have screaming people (or at least no more than usual). You just have people quietly typing orders into a computer, except that those orders are "SELL EVERYTHING SINCE THE WORLD IS FALLING TO PIECES!"
Something I do remember from 2008 was how calm everyone was, which resembles some of the history that I've read about October 1929.
Predictive analysis is only as good as the inputs (or knowledge thereof) and the integrity of the models. If the models reflect the 'actual' dynamics of the system, then one has a chance at good predictions.
All models are wrong. Some models are useful. Some models are useful because they are wrong. Except in some very, very specialized situations, numerical financial modelling is not used to make predictions, because the economy is inherently unpredictable. Even the parts of the economy that are somewhat predictable are not necessarily numerically predictable. When I buy stock in a company, I care less about the P/E ratios or the balance sheet than to read something that the CEO and the management team has written and decide for myself if I think I can *trust* them with my money. For the housing crisis, people were too busy looking at Excel spreadsheets and not looking at the "GET RICH QUICK BY REAL ESTATE" ads on late night informercials.
Typically the way that it goes is that you start out with something that says assuming A, B, and C, you will get price D. However, you never actually do get price D, but comparing price D with the actual price you observe may tell you how far off assumptions A, B, and C are. Another game is given situation X, how much is it likely to change price Y.
This is one reason banks to like physicists. Physicists tend to ask questions like, so what does this number really *MEAN*?
This is a problem in economics models, as it is in physical models, such as weather/climate models, or highly engineered systems. The more variables, the greater the challenge, especially as non-linearity increases.
One problem with economics is that I think that economics professors use too much math and too little common sense. Except in some extremely limited situations (i.e. derivatives pricing), I don't think that you can really use deterministic numerical models to model economic behavior, and even in most of these situations these models are *NOT* predictive, and will in fact totally fall apart if it turns out that things can be predicted.
One problem in trying to force the economic into deterministic numerical models is that people have free will and can make decisions whereas clouds do not.
Also there are some self-fulfiling prophecies. If people *think* that the world is going to end, and they start pull all their money out of the banks, then the world will indeed end. This can be modeled, but not deterministically.
There are a lot of mathematical or semi-mathematical techniques that you can use to model complex systems. The important question sometimes is "what is the important variable?" "are there feedback loops?" "how wrong will I be if I leave out this factor?"
Also knowing that something *is* unpredictable is sometimes a good thing. If you are selling insurance, and you've convinced yourself that something truly is unpredictable then you can do some strategies based on unpredictability.
However sometimes "evidence" of predictability doesn't come from numerical models. If I'm selling insurance, and all of a sudden, someone for no good reason seems to want lots of insurance on something crazy like insurance against the Earth being invading by green alien spacemen, then I'm going to get on the phone and talk to him and find out why.
