OK, James Surowiecki sees the Wisdom of Crowds. But then Bryan Caplan noted that the average voter is irrational. And of course Philip Tetlock notes that experts are usually wrong. So you think all is left is models, a simple equation? Well, not for inflation forecasting. From Geert Bekaert's forthcoming article in the Journal of Monetary Economics:
Perhaps Nassim Taleb is onto to something: say 'something unexpected will happen', and then, when something unexpected does happen (which by definition no one predicted), point out your prescience. Brilliant!
We examine the forecasting power of four alternative methods of forecasting U.S. inflation out-of-sample: time-series ARIMA models; regressions using real activity measures motivated from the Phillips curve; term structure models that include linear, non-linear, and arbitrage-free specifications; and survey-based measures. We also investigate several methods of combining forecasts. Our results show that surveys outperform the other forecasting methods and that the term structure specifications perform relatively poorly. We find little evidence that combining forecasts produces superior forecasts to survey information alone. When combining forecasts, the data consistently places the highest weights on survey information.So, as a rule, there seem to be no rules for when picking which class of forecasters to pick from.
Perhaps Nassim Taleb is onto to something: say 'something unexpected will happen', and then, when something unexpected does happen (which by definition no one predicted), point out your prescience. Brilliant!
Eric Falkenstein - am 2007-07-03 23:37
Econocator (anonymous) meinte am 4. Jul, 18:01:
top paper
Thanks for the link Eric. This makes for fascinating reading.
F. Blair (anonymous) meinte am 5. Jul, 11:40:
surveys
I don't understand: the paper says surveys (which reflect the wisdom of crowds -- at least the wisdom of the forecaster crowd) outperform all these other strategies. How does that mean "there seem to be no rules"? There is a rule suggested by the paper: rely on the wisdom of the forecasting crowd if you want the best available answer (which is, to be sure, not going to be a perfect answer by any objective measure).
Eric Falkenstein antwortete am 5. Jul, 16:03:
But Tetlock, among others, argues simple models outperform experts. And though in this case the survey is about 'experts', in other cases the masses outperform experts.
Bernard Guerrero (anonymous) antwortete am 6. Jul, 18:12:
The obvious question then arises...
...how are the forecasters in the "expert crowd" coming up with their forecasts? :^) If they're applying similar simple algorithms to come up with their forecasts, then the paper's result would boil down to an application of the Law of Large Numbers, no? Apologies if the paper tackles this issue, I haven't popped it open yet.Tetlock's results apply, IIRC, to individual experts being treated as predictors. Aggregate a bunch of them and the errors should start to wash out. This implies, however, that a survey of algorithms should produce similar results. And you don't have to pay the algorithm.
