i realise this is a discussion about economics not finance - but i think this little story sheds some light on this. early into a job on wall street i found that the companies default risk model could be improved (against the 'training'/regression data they had already spent ages cleaning) by adding a neural network that was trained to predict the error in the model.
did they use it? no. why not? because they felt it was important that they fully understand the forces at play that cause the default rates to move in that direction. you could say that they were worried about models overfitting the data - sure, that's kinda it. but the point is the entire nature of the market can change overnight and kind of go into a different state.
so what does that have to do with economics? well i guess maybe some of the same reasoning is at stake - people who think "we're proved wrong so often we need to retreat back to just those few things we can really rely on being true" and are worried about being "too clever" with all this math stuff.
now its true that cargo cult like application of mathematical formula has been (imho) part cause for many truly dreadful economic theories. but still, they are wrong. if you don't understand the nature of far from equilibrium, complex systems, you'll never even begin to understand the forces that drive the economics of the real world.
Current mainstream economic theories underfit like crazy, that's their virtue. The problem is, they aren't based on knowledge of what actually happens, but on either a 19th century armchair philosopher's guess at what happens, or principles which have been chosen in order to create an under-fitting model.
So while they don't overfit, they also are't based on any real knowledge of the system.
It seems the other extreme is elaborate overfitting models.
Steve Keen uses fairly simple models, but tries to base them on actual reality (i.e. broad money is created by the banks, not the government, while Ben Bernanke's explanation of the Great Depression apparently focuses on M0, as he accuses the government of not minting enough coins with the gold they had in reserves, with the implicit assumption that the money would be multiplied if the government would only print it. In reality, the money was not multiplied, because the banks didn't feel like lending.)
did they use it? no. why not? because they felt it was important that they fully understand the forces at play that cause the default rates to move in that direction. you could say that they were worried about models overfitting the data - sure, that's kinda it. but the point is the entire nature of the market can change overnight and kind of go into a different state.
so what does that have to do with economics? well i guess maybe some of the same reasoning is at stake - people who think "we're proved wrong so often we need to retreat back to just those few things we can really rely on being true" and are worried about being "too clever" with all this math stuff.
now its true that cargo cult like application of mathematical formula has been (imho) part cause for many truly dreadful economic theories. but still, they are wrong. if you don't understand the nature of far from equilibrium, complex systems, you'll never even begin to understand the forces that drive the economics of the real world.