A random question from a Stanford University PhD micro exam looks something like this:
The "Robinson Crusoe economy" is the simplest possible general equilibrium model, and students who can solve for Crusoe's leisure and coconut consumption choices have presumably learned something about general equilibrium theory.
But they haven't learned how to theorize.
Theorizing - creating theories to understand and explain the world - starts off by identifying an interesting question. Creating exam questions is theorizing, because the question setter has to identify a puzzle. Answering exam questions is one step removed from the act of theorizing. The question has already been identified. My sense - and I could be wrong - is that a non-trivial number of post-comp PhD students feel something like this:
They can follow the most complicated plans, and build any model they are asked to build. But they struggle to come with an idea of their own; to decide what they want to build. Nothing in their training has encouraged them to fill their minds with ideas, nor taught them to distinguish the awesome from the not-so-awesome.
Identifying a problem is only the first step towards creating a theory of one's own. Theorizing is about abstraction. In the exam question above, picking coconuts serves as a metaphor for an entire economy's productive activity, and the complexities of technology are reduced to the production function: c=f(l)=l1/2. This is, as I have argued before, the power and beauty of theory: it sacrifices literal truthiness in order to capture radical (as in "at the root of") truths.
When students are taught about theory, they are presented with simple, pre-abstracted, models. It's like being asked to re-arrange the furniture in de-cluttered, minimalist home - a useful exercise that introduces the student to basic principles of design and organization. Yet when students go out to build their own theories, they are faced with explaining a world of complexity. Before they can make any progress, they have to throw a whole load of stuff out. But economic theory courses spend little time explicitly discussing methodology: what makes for a good explanation, and thus should be in the model - e.g. incentives, prices - and what can safely be bracketed out. So people end up lost.
It could also be that students struggle to theorize because they lack training in logical reasoning. Economic theory is deductive. All that math is just a way of (a) formalizing a set of assumptions and (b) deducing the implications of those assumptions. Logical reasoning is about taking a step back and asking what do these assumptions actually mean? Are they consistent with the real world? Are they capturing key aspects of the problem I'm trying to model? It makes no sense, for example, to assume constant returns to scale when modelling monopoly, because monopolies do not exist in a world with constant returns to scale.
Years ago I was chatting to a PhD student about his thesis. He said "I've got my model all worked out, I just don't know what to put in between the equations". ["How about economics?"] Training in logical reasoning - understanding that economic theorizing is a process of deduction, that the mathematical calculations are simply a way of working out the implications of the model's initial assumptions, and so each step in those calculations has an economic meaning - might help people work out what to put in between the equations. I don't know.
In theory courses - whether in economics or in other disciplines - theory becomes an object of study. Professors teach about theory - because this is how they were taught, because this is how textbooks are organized, and because it's relatively easy to teach students how to step up and solve a series of standard models.
But to teach students how to do theory? It's like riding a bicyle - ultimately, the only way to learn is just to push off and start peddling. Send people off to model the world and see what they come up with.