sandcastle

Photo: StuSeeger via Flickr.

My pal Peter Dorman is looking for answers: Does the class of economic forecasting tools known as “computable general equilibrium models” (aka CGE models) have any documented track record of success?

This may seem like an arcane point, but it’s quite relevant to climate policy. Government agencies throughout North America are using CGE models to forecast the economic impacts of various cap-and-trade proposals. But many academic economists — Dorman among them — think that the CGE models are built on sand. Says Dorman:

I think these models are so dubious theoretically and unreliable in practice that there is no case for using them.

Peter’s a smart guy, and he’s looked and looked. But he’s found no evidence that anyone’s actually tested the reliability of CGE models, to see if their predictions match up with real events as they unfold.

All it would take is a retrospective study, lining up CGE-based forecasts made in, say, the early 1990s, and comparing them with what actually happened. But search as he might, he’s found nothing of that sort in the literature.

Worse, he suspects that commodity traders and hedge fund managers — the sorts of people who get paid a lot of money to predict the future — simply ignore the results of CGE models. Sure, policy makers and academics use CGE models all the time, but there’s no real personal harm to them if the models lead them astray. For people with skin in the game, though, a bad guess is a costly one, and a good guess is worth a mint. If market traders don’t use CGE models in their work, then that’s a decent piece of evidence that those models just aren’t all that useful for predicting the future.

So for anyone out there who’s economically inclined, Peter hopes you’ll spread his challenge: If you have evidence that CGE models are worth their salt — and that their predictions are any more reliable than, say, informed guesswork — please let him, or us, know.