New Directions In Econometric Modeling

Last Tuesday's Wall Street Journal featured a lengthy news story discussing the work of many new arrivals to the econometric modeling scene. Some of the recent entrants into the field use distinctly atypical approaches.

At first, some of what I read struck me as silly. Indeed, the story quotes NYU economics professor Mark Gertler as saying,

"It strikes me as not productive to say that all we have done is a complete waste. The profession is extremely competitive. If you have a better idea, it's going to win out."

Well, yes, but only over other equally-flawed and potentially limited-in-scope approaches which are all developed by economists who studied the same historic approaches.

One of the things I found distasteful about the prospect of taking a PhD in Marketing at Penn years ago was the requirement to take various econometric modeling courses. The Journal piece notes one of the new modelers, one Mr. Farmer, observing that modern "dynamic stochastic general equilibrium" models have become so complex and over-specified that convergent solutions are often impossible to satisfy all conditions and variables.

One engaging newer approach is that of using 'agents' to describe economic activity, rather than reduce all economic activity to equations of various Keynesian-era variables. Other ideas borrow from psychology and lean toward the work of Amos Tversky and Daniel Kahneman, a Nobel Laureate for his work on risk.

Another suggests approaches used for weather, traffic and epidemics, focusing on many discrete inputs, rather than a few simplifying equations and variables.

Gertler's comment brings to mind the entire question of what such models are built to do, and how they perform in terms of prediction errors. So long as one can explain a logical trail from inputs to predictions, does it really matter, other than to Gertler's sensibilities, and those of his kind, that the models aren't conventional econometrics?

I should think not. Speaking from experience, taking a fresh approach to a problem, using tools and perspectives from another field, can allow one to capture aspects of a the problem that conventional, existing approaches simply miss, out of ignorance. I've found this to be the case in my equity strategy work. Coming from a marketing and strategy background, my modeling approach to equity performance and selection is quite different than those of typical finance-trained people. And has resulted in better performance than most similar, publicly-tracked funds.

Reading the Journal story provides some shocking insights regarding what isn't included in many of the current models, e.g., central bank actions and the finance sector, generally.

After rereading the piece and reflecting on it, I think I'm more inclined to be welcoming and excited by the arrival of a group of new and varied modeling techniques onto the econometric scene.

Certainly the past several years of ineffective retreaded Keynesian-type models have produced nothing impressive.