Reflections on Teaching Graduate Micro for the First Time

The graduate exam is done. It took them 2 hours and 45 minutes to complete it, rather than the 2 hours they are allotted here at my university for final exams. I spent the entire time rewriting the syllabus for this class, and sketching the many lessons I learned about teaching it, the pitfalls, the objectives for me and for my students, and how to improve. This class is going to be one I'm teaching from the perspective of a longrun equilibrium. I definitely will not have it up to snuff next fall, nor the fall after I'm guessing. I can already tell that there's a convergence to the longrun class I'm trying to teach that will take many periods.

For one, this semester went badly in the sense that I spent far too much time on certain subjects and never even got into competitive markets or imperfect competition, let alone partial and general equilibrium. We ended up spending a month and a half just on risk, uncertainty and game theory. Part of what slowed us down was just me now knowing, ex ante, what I should and should not cover, and so I covered everything. And in the end it felt like running waist-deep through snow. Next time, I will know exactly what to leave out. In fact, that knowledge is part of what I'm using to help me differentiate this course, which is a masters level microeconomics class, from a doctoral course. Like a PhD course, I emphasize analytical tools, like optimization, and derive economic theories from various axioms related to consumer preferences, production technologies, and market structure. But, unlike a PhD course, I will not be requiring proofs, nor will I ever much get into the n-good case anymore. For this class, I'm convinced the 2-good case is sufficient, and that can be one way that I save time without loss of generality.

I want them to learn: optimization techniques, duality, envelope theorem results, estimating elasticities and comparative statics, and the properties of various functions. But, that only takes me really through half the class I'm trying to teach. The sections I'm less comfortable with, that I covered this term, are related to information (uncertainty and risk mainly), insurance and game theory. I learned that it's entirely possible to spend 6 weeks on just those three topics. So how do I fix that? For the game theory section, I am going to fix it in simple, time-saving ways that I think will go further. Instead of writing out the definitions of the various nash equilibrium (we go into detail on dominant strategy, best response functions, nash equlibrium in pure and mixed strategies, subgame perfect nash equilibrium, backwards induction, and bayesian subgame perfect nash equilibrium), I'm going to create a single handout with these definitions and make students responsible for knowing this before coming to class. I won't write definitions on the board - or at least, I will do it much less than I did this semester, as that inevitably cost me hours, cumulatively, which itself translates into entire days of lost time. So, wherever I can, I'm going to use the board for working out problems, and handouts for listing definitions, properties of functions, and hints for solving problems.

This takes me to the second point, and that's the techniques. As my exams are mainly analytical, requiring students solve word problems mathematically, I think students need more front-end help to guide them through these kinds of problems. Students would benefit from a simple guide for how to solve a sequential game of incomplete information when trying to find the bayesian nash equilbirium. In my experience, teachers never give their students this, and as a student, it was the one thing I was always searching for. Okay, I know now the definition of Bayesian Nash Equilibrium. But how do I go from this specific game to using that definition to find the answer? Students would benefit from seeing the process in steps, wherein they know to use backwards induction to solve first the last person's best response function, and then inserting that best response function into the first person's best response function, then substituting the first person's solution into the second's to find his, etc. In other words, I'm going to aim to make this theoretical class more practical. These are masters students. They need to see where the theory comes from, become somewhat competent with optimization techniques and all related techniques, and they need to feel like they can use it themselves. So I've been reflecting on the Trivium some, and I think this class needs to incorporate some of the insights from that ancient pedagogy. Specifically, start with simple building blocks, build up to the internal logic within some given area, then extend the logic outside the system to new problems. Thus, classwork and problem sets will mainly be designed to force students to master those first two stages - the "grammar" stage and the "logic" stage if you use the Trivium language - and then tests will mainly be "rhetoric" tests. That is, application. Can they take what they've done and apply it to something unknown? Their grade in the class will depend on the degree of success at doing so.

The challenging part, though, is the development of relevant tests and exams, and then specifically timely grading with informative responses by myself. Having never taught the class, I did not know beforehand what to test them on, and so tended to just work linearly through the textbook, devoting nearly all my time just to lecture preparation, and a minority of my time to developing exams and problem sets. Not surprisingly, students felt the exams were too hard, and problem sets did not lead them to skills that made them successful on those exams. This is a function of the lack of time I poured into developing those materials. It seemed to me that the best use of my time this semester was to spend a majority of my time just re-learning this material, as it had been five years since I'd taken it myself, and I had never been all that good at it when I took it as a first year. But, now I know I have to make changes specifically in the realm of problem set and exam preparation/design, and timely, informative turnaround. The one thing that I think I could do differently is to actually put myself through the gauntlet and start working through problem sets on my own. As I tell my students, learning economics is more like learning to play the guitar than it is like reading a book. Reading a book tends to have a passive quality to it, particularly if you have a good memory and a voracious imagination and appetite for knowledge. Physics, and to a somewhat related degree economics, are unique in that they require practice, memorization, and more practice. The tools are themselves acquired through inputs of time spent (usually in agony) working on practice questions. If you cannot find some, you go to the back of the book and work those end-of-the-book chapters. When you're done with those, you tear up your answers and start over. So one thing I'll be doing differently is working through a large batch of problem sets before hand, to help me get in the right frame of mind for designing my own. This time, I depended largely on old problem sets of my own, and it helped at certain points just fine, but did not when I ended up spending forever and a day on risk and game theory.

I'm going to continue to reflect on these questions, and this semester, and hopefully turn the experience into something I can use to improve the class for the sake of the future students.