Building Models in Economics

by Patrick L. Warren

Every Fall, I teach a PhD course that is putatively on the topic of political economy, but at least half the course is dedicated to the practice of building economic models. The philosophy of modelling I expouse is firmly practical, and it starts with some mystery about the world that we don't fully understand:

- Why do the EPA administrator and head of OIRA fight publically about policy when both are appointed by the same person? (My answer)

-Why do non-profit hospitals outsource some services at lower rates than for-profits but outsource others and approximately the same rate? (My answer)

Students are then required to investigate the institutions and exsting literature in a enough detail to come up with their own intuitive explanation for this stylized fact. Only then does the modelling begin. They must formalize their explanation in an economic model, often drawing on some of the tools/building blocks we've learned in class. The goals, are (at least) three fold:

1. Derive the conditions under which the intuititve reasoning holds up (if ever) for the case under consideration.

2. Expand the scope of the argument to cases beyond the motivating one, if feasible. 

3. Uncover implications of the theory that were not obvious from the intuitive version.

Non-intuitive implications of the intuitive explantion are particularly important, here, because they are an extremely helpful testing ground. It's silly to "test" your model on the case that inspired it. Of course it will explain the "data" in that case, you invented the model exactly to hit that target. The proof of the model is whether it predicts things it was not aimed at, but you can never know whether it will until you start. It's alway exciting to find out. (Of course, if it doesn't work out, you can always tweak the model. I wonder if we should preregister models, too?)

My thinking re: the goal of economic models has moved over time, and I really enjoyed a recent paper by Gilboa et al on models as analogies that has affected my thinking on the subject. You should check it out.