The former trader turned professor, Nassim Nicholas Taleb, in his new book The Black Swan: The Impact of the Highly Improbable warns against the danger of forecasting. Forecasting is, obviously, always based on what he know and not what we don't know. A Black Swan, by the way, is an earth-shattering event we didn't predict eg, 9/11 (until Australia was discovered everyone believed that only white swans existed).
On my way to work this morning, I heard an interview with Taleb on the BBC World Service where he said that 25 year forecasts, used to justify new projects in any industry sector, were worthless.
He added that the only forecasts of any value were those for 12 months because of a reduced chance of innacuracy.
And he added: "You can guarantee that the cost of a project will be far more than you've estimated."
So sack the consultants and trust in your judgement - provided, of course, you can convince the banks.
Comments (1)
It does stand to reason that a 25 year ahead forecast would be of rather dubious value, but more due to the shape of the prediction errors which "fan out" in a somewhat quadratic manner (depending upon the underlying model). The Bank of England routinely uses fan charts to express the uncertainty in inflation forecasts, as you can see here
http://www.bankofengland.co.uk/publications/inflationreport/irfanch.htm
Note that these forecasts are typically only for a few years at most, not 25!
I am unaware of serious work on the accuracy of very long range forecasting, but if others know of it I would be interested to know via my blog.
And I have written a little on improbable events and Black Swans at
http://dsanalytics.com/dsblog/a-little-bit-about-heavy-tails-and-black-swans-and-stable-distributions_76
Posted by John Aitchison | July 18, 2007 10:40 AM
Posted on July 18, 2007 10:40