Wednesday, September 30, 2015


Okay, so people make fun of the bad forecasts of economists.  Fair enough.  But economists are trying to forecast actions of sentient, prospectively focused creatures.

Hurricanes are just big unruly piles of wind.  But we can't forecast THOSE either.  Check out this "ensemble" forecast for the next week (click for an even more incoherent image):
Of course, if we can't predict the actions of "big unruly piles of wind" I guess forecasts about Congressional votes are also off the table. 


Simon Spero said...

This is why you shouldn't look inside the black box when the motor is running.

With ensemble learning, the more diverse the component models are, the better the overall results tend to be.

(Figures 38 and 39 show the model performance when they come together to form Voltron.

M. Janousek, P. Bauer, J. Bidlot, L. Ferranti, T. Hewson, F. Prates, D.S. Richardson and F. Vitart
Evaluation of ECMWF forecasts, including 2013-2014 upgrades
December 2014


Jason Smith said...

Hurricanes are part of a system where the microfoundations are known and that is known to be nonlinear. Neither of those have been remotely established for economics.

Rick Samborski said...

I don't blame economists for not being able to predict the economy. I don't even blame them for trying to predict the economy. I blame them for pretending that they can predict the economy.

Anonymous said...

OK, meteorologists can't accurately forecast the weather for the coming weekend, but at least governments can spend billions to mitigate "global warming," based on century-ahead forecasts by prescient-omniscient environmentalists. If they too are prone to erroneous prognostication, they should switch to merely forecasting "climate change." That's not too tough; can't go wrong; adopt the epistemology of Heraclitus. Oh, wait: they DID make that switch!! Wise move. Why not also our weathermen? "My forecast for the coming weekend is that things will CHANGE."