> merely because under those assumptions it's actually feasible to build models
I think it's a really bad excuse in age of computer modelling and when we have mathematical tools to deal with dynamical systems.
I think it's a really bad excuse in age of computer modelling and when we have mathematical tools to deal with dynamical systems.
I mean, "rational expectations" is nothing to do with the relative simplicity of the mathematics (the models it largely replaced were worked out with pen and paper) and everything to do with the argument that economic models shouldn't rely on people making a particular type of systematic error to show a desired outcome (e.g. if policy changes required economic actors to ignore the implications of the policy change to work, they probably wouldn't actually work that well). Essentially it's the absence of an assumption about human behaviour, and there's entire classes of economic models devoted to saying "even if people on average anticipate the future and economies function perfectly smoothly, simply introducing X into the model means that you still get recessions and still get a benefit from a policy response to it", which is a more powerful argument than "if I've calibrated all these parameters and specified all these functions about all these hundreds of different types of agents' planning correctly, this policy will work", especially if you're trying to disprove arguments that the economy will sort itself out eventually.
That doesn't mean there isn't a place for complex computational models and even throwing data at ML algorithms to see what sticks, or that a general equilibrium model to predict actual changes in an economy isn't fairly unlikely yield accurate results over the long term, but they're performing an entirely different function from reasoning that X will [not] affect Y even if or only if all else is held constant or assumed to respond logically.
Same goes for Keen's models, they are dynamic and pretty simple. He even writes in his books, to paraphrase, "equilibrium is feasible" was a good excuse at the beginning of 20th century, when Marshall came up with supply/demand model, but it's not today, when we can actually analyze dynamical systems mathematically.
My memory on that is little hazy (it's been maybe 15 years ago I read about it), but as I remember this was the case about Blatt's model as well. And IIRC it's based on earlier ideas by Keynes (uncertainty is a different thing than risk).
I also recall nice idea from Paul Ormerod (but it could have been somebody else or folklore) who had an interesting model of economic agents - do either one of the 3 things:
- the thing that you always did
- the thing that others are doing
- another thing that you think might work
This also leads to an interesting class of models (different from rational expectations) and it's not making too much assumptions about humans.
Again, this shows that Blatt and Keen (and other post-keynesians) are woefully underappreciated in economics.
I've got plenty of time for the post-Keynesians but most of them (particularly in the Keen Godley/Lavoie Social Accounting Matrix style) really aren't doing more complicated mathematics so much as choosing to have models far more sensitive to specified lag structures and/or using different assumptions about human behaviour (The flip side is that Keen's hypersensitive-to-how-it's-specified banking system is better in many respects than a macro model with no banking system or credit constraints) For much of the last century the Cambridge post Keynesians distinguished themselves by doing a lot less modelling than their neoclassical counterparts.
IIRC prominent economists actively ridiculed the idea of a "Manhattan Project" a few years back. The idea was to pull in experts from mathematics, physics, biology and computer science to invigorate the field via cross-pollination like occurred with mathematics and physics in the 70s and economists did not like it.
Update: apparently it did lead to a conference, at least, and Nassim Taleb was one of the participants.