ohai wrote: ↑Mon Jun 17, 2019 8:30 am

Hi. This graph is certainly interesting and shows how much expectations of "experts" have deviated from reality.

However, there is a flaw with this sort of poll. Just because the "best guess" of everyone was not correct, does not mean that their analysis was bad.

A reasonable economist would submit a possible distribution of outcomes: [10% < 2.00%, 20% between 2.00% and 2.25%, 30% between 2.25% and 2.50%... and so on]. In this case, the middle prediction could still have been the best guess. Things that have 10% likelihood happen all the time. If you asked me to guess where the stock market will end this year, I might say up 5% from now, but there is obviously a wide distribution of possible outcomes.

I don't know the confidence level of these economists on their best guess in the article, but the question definitely sets them up for apparent failure.

1) When the actual value lies far outside the entire range of

*fifty* estimates by economists, something is wrong. You can't formalize it exactly, but the fact that there are fifty economists doing fifty analyses is vaguely like looking at the distribution and a range estimate.

In this case, the spread of forecasts looks to be, ballpark, close to 1.0%, and the actual value is 0.5% below the lowest of the fifty forecasts. If you assume that the standard deviation is actually 1.0%--based on a 1% range it would be expected to be

**much** less--then the chance of a single forecast being 0.5% above the true value is about 30%. The chances of fifty of them

*all* being 0.5% or more above the true value is 0.000000000000000000000007%. That's just the roughest of rough back-of-the-envelope estimates and I might have miscounted a zero or two on my fingers either way, but you see the point. It's absurd to say "they were right, but they just had bad luck." Look, take the square root of 0.000000000000000000000007% if I somehow needed to do that to account for, two tails or something, I don't think you can come up with any plausible statistical model in which an accurate model would be that far off more than once-in-a-millennium.

2) It's a completely different set of forecasts, but GMO, to their credit, does put ranges on their own 7-year forecasts, and apparently nobody pays any attention to those ranges or GMO's track record, because when the seven years have elapsed, the actual values are

*frequently* well outside GMO's own stated full range. For example,

see this post by longinvest
3) It is much closer to the truth to say "financial forecasts are so bad it is better not to pay attention to them, and positively reckess to act on them," than to say "financial forecasts are really pretty good, it's just bad luck that we've had a thousand-year flood every year."

Annual income twenty pounds, annual expenditure nineteen nineteen and six, result happiness; Annual income twenty pounds, annual expenditure twenty pounds ought and six, result misery.