What can economics students reasonably be expected to learn? Reading this post by Tony raised this question.
To see what I mean, consider the general skills one would need to understand panics and crashes, to take Tony's case. We'd need to know some maths and stats: what's wrong with Gaussian copulas? How do nonlinear differential equations describe bubbles? We'd need some economic history: without this, our only datapoint is 2008. We'd need detailed knowledge of regulation and market structure: what the heck is a CDO? To what extent did poor regulation cause the crisis? And we'd need some knowledge of the psychology of herding and wishful thinking, as well perhaps as some experimental economics: I'd add this paper to Tony's already-daunting reading list.
I'd be mightily impressed if a 40-year-old had all these skills. It would be utterly unrealistic to expect a 20-year-old to have them. And this is only one course.
The point here is that even the best undergraduate education can only scratch the surface of the discipline. It is therefore inevitable that students and their employers will complain about them being imperfectly prepared; as Cameron points out, such complaints are decades old. This isn't (just?) because their teachers are poor or ideologically blinkered. It's because three years is not long enough to learn very much.
So, what should economics students learn? I'd suggest that the priority should be to prepare them for lifelong learning about economics.
This should include a good grounding in conventional economics and - yes - in the mathematics required to understand it. This is necessary on practical grounds - for students going onto formal graduate study - but also intellectual ones. This is because conventional economics contains lots of truth. For example, the efficient market hypothesis gets tons of abuse, but the fact is that most stock market investors would be better off if they acted as if it were true. Also, you've got to understand conventional economics if you are to understand why and when it's wrong. I remember Manchester's Terry Peach telling me that heterodox economists must be better economists than neoclassical ones for precisely this reason.
However, learning only plain neoclassical economics isn't enough. It's like going to the gym and working only on your biceps; you'll end up a freak - albeit one capable of punching someone's lights out when they point this out. Something else is needed. That something isn't just heterodox economics, but a grounding in both statistical methods and critical thinking. By statistical methods, I don't mean deriving and proving lemmas about maximum likelihood estimators; if something's been proved once it doesn't need proving again. Instead, I mean knowing how to run statistical tests, how to interpret their results and the strengths and limitations of various statistical methods. Simon is right to say that the question "do fiscal expansions work?" should not have the answer "it all depends on whether you are a Keynesian or an Austrian." But it should lead to the question: what counts as good evidence here?
(My suspicion is that quite a bit of marginalist theory wouldn't survive a confrontation with the empirical evidence - but that's another story).
And it's here that behavioural economics enters. The point about cognitive biases is that they don't just apply to other people, but to ourselves. In this sense, learning about them should be a big part of a course on how to think properly.
The problem here is that there is a trade-off between depth and breadth; learning some things requires not learning others. But you'd expect the economics profession to be capable of dealing with trade-offs. Wouldn't you?