Arnold Kling  

Overconfidence

PRINT
The Ingredients and the Recipe... Why I Believe in Moderate Glob...

At the conference where I gave my talk that mentioned recipes and ingredients, I met Edgar Capen, co-author of a classic paper that coined the term "winner's curse" for auctions.

His main theme is that people under-estimate uncertainty. If someone says, "the chance of X falling between A and B is 90 percent," the true probability is likely to be 30 percent. He likes to do experiments where he puts a large number of beads in a jar and asks people to guess the number of beads. One point he makes is that the "wisdom of crowds" result (that the average estimate tends to be pretty good) often depends on keeping outlier estimates in the sample. It is the extreme contrarianism of one person in the crowd that keeps the crowd's estimate from going off track. I wonder if anyone has compared median estimates with mean estimates in "wisdom of crowd" studies.

He said that some of his results on overconfidence are in a 1976 paper in the Journal of Petroleum Technology. He shares my strong skepticism about climate modeling.

UPDATE: Bryan refers to Robin Hanson's post on skeptics who won't bet on their skepticism. I have two reactions.

1. I probably would be willing to buy puts and calls on future global temperature relative to the consensus forecast. I think that the consensus is a classic case of a subjective confidence interval that is overstated relative to reality.

2. A slightly harder task is to formulate betting markets on climate models, which is where my skepticism really is strong. Off hand, I can't come up with a way to bet on the accuracy of model parameters, as opposed to just betting on future temperature.


Comments and Sharing





COMMENTS (7 to date)
Buzzcut writes:

How "Talebian". I guess I'm being forced to think more highly of "The Black Swan".

To be sure, I didn't have any quams with Taleb's conclusions, just with the fact that the guy is a major A-hole.

Phil writes:

Why not have a market where participants can define their own contracts? Then it's likely that you (or someone) will come up with something that you're willing to bet on.

I think there's a market like that already, but, alas, not for real money.

Brad Hutchings writes:

"Heh" to Buzzcut. I just read the section last night in TBS where NNT discusses overconfidence, so this post was a nice surprise this morning. It is funny that he concludes that section essentially saying that his Mom might agree with you ;-).

But to introduce another Taleb thread into this... The only safe bet would be that if he entered this debate, he would kick more butt than Fred Thompson at a butt-kicking convention. Look, I really don't expect anything profound to come out of Rush Limbaugh's or any other talk show host's piehole, which is why I switched to sports talk more than a decade ago. Rush, offering to bet that NYC won't be under water or Greenland won't melt, is betting against particular black swans, succumbing to the same fallacy that the global warming narrators have. That proponents would not take Rush up on those particular bets (absent a very wide range of particular black swans to bet in the market) shows that proponents are a bit more honest with themselves than in public.

To answer Bryan's question... An interesting bet would be that no city of some threshold size in some large geographic region (east and gulf coasts) would be underwater due to a hurricane over some time period -- regardless of "global warming". I don't know if I'd take that bet. New Orleans ended up underwater because levies broke, not because the storm dumped all that water on the city itself.

Michael Sullivan writes:

That proponents would not take Rush up on those particular bets (absent a very wide range of particular black swans to bet in the market) shows that proponents are a bit more honest with themselves than in public.

Which proponents have claimed any one of these things as a likely (or even anything but extreme way out) proposition? I'm not aware of any. I suppose Greenland glaciers melting is a feasible long term consequence of unchecked greenhouse emissions, but I don't know anyone who'd expect it within the next 100-200 years. Certainly not the standard climte models. I wouldn't bet on it happening in any time frame since ingenious human response to the problem is incredibly uncertain over such a long time scale.

Brad Hutchings writes:

Michael, you may have missed Al Gore's An Inconvenient Truth. Just sayin...

SheetWise writes:

"Off hand, I can't come up with a way to bet on the accuracy of model parameters, as opposed to just betting on future temperature."

Many climate models predicted that the last hurricane season would be the worst in history -- as a result of global warming.

Ooops.

When the people who design the models show confidence in the model by selectively predicting, I have to assume they're reporting in areas where they've noticed the highest correlation. But there is no correlation -- just past performance.

Scoop writes:

I'd guess this has been mentioned here before, but there is a very crude futures market on global warming: seaside real estate. If anyone actually believed the worst-case scenarios, it would be much cheaper than it is.

Comments for this entry have been closed
Return to top