Use your models, don't believe them!

Noah Smith complains:

"And to make it worse, most of the macro theories that economists take halfway seriously are too hard for intro kids, so they end up learning silly stuff like Mundell-Fleming and Keynesian Cross that no one even halfway believes."

But, believing is something one should never do with one's models: they are just models, and as abstractions are necessarily falsifications of the full reality being modeled. A road map is just lines on a piece of paper: it never follows each twist of a road, it doesn't show dangerous potholes, it doesn't let us know the road is now blocked by a slow-moving garbage truck. (Of course, interactive maps may show red dots when a road is backed up, but the basic point stands.)

This was a point we made when I was a partner in an asset-trading firm: our models were something we used, not believed, and as soon as they ceased to be useful, we abandoned them, and sought another useful model, without any silly concern about whether the new model was true.

And models can have many uses: they can help us make predictions, they can help us to isolate one factor operating in a complex situation, and they can help us to convey a certain view of the world.

When I teach macroeconomics, I use the Keynesian Cross for the latter purpose: I tell my students that the model is a highly simplified way of understanding what Keynesians believe goes on in a recession, and how they think we can get out of one. I stress that no one thinks that it captures even a tiny part of what goes on in a real economy. But, most of all, what I hope my students learn is what a model is, how to use it, and why it is necessarily limited.

The people in charge of Long-Term Capital Management apparently came to believe their model; why one should never do so can be seen here:

Comments

Popular posts from this blog

Libertarians, My Libertarians!

"Machine Learning"

"Pre-Galilean" Foolishness