David Leonhardt had an interesting column on underestimation of risk. BP’s possible underinvestment in protecting against a gross accident is exhibit one.:
The people running BP did a dreadful job of estimating the true chances of events that seemed unlikely — and may even have been unlikely — but that would bring enormous costs.
Perhaps the easiest way to see this is to consider what BP executives must be thinking today. Surely, given the expense of the clean-up and the hit to BP’s reputation, the executives wish they could go back and spend the extra money to make Deepwater Horizon safer. That they did not suggests that they figured the rig would be fine as it was.
But this does not prove the case. You may buy a stock given the odds of it going up or down. If it goes down you will regret your investment. This does not prove it was wrong to invest in the first place. It might have been right given your initial assessment. The same logic applies to BP.
This is a simple point: regret does not imply that the ex ante decision was bad. Leonhardt is a great economics commentator and journalist. The fact that he makes this elementary mistake shows how easy it is to make.
But there is another factor at work. It is impossible to determine BP’s probability assessment after the fact. They can always claim the chance of a disaster was low. There is no historical data against which to measure their assessment. All we are left with is the option to blame them even if their decision was perfect from an ex ante perspective. Blame involves saying them made a bad decision and holding them to account. This was the key element in Jeff’s earlier post on the Blame Game.
6 comments
Comments feed for this article
June 1, 2010 at 9:52 pm
Jeff Shrader
Sandeep, I think that you are not defining regret precisely enough. The way I think about it, to regret something means that one wishes one would have taken some other action. Wishing for some other action implies that the original action was sub-optimal or that one now believes the original assessment of probabilities was mistaken.
June 2, 2010 at 8:01 am
sandeep
Jeff: You are right that I did not define regret precisely. I try to keep blog posts as non-technical as possible but I am trying to get at Leonhardt’s implicit definition: Suppose you take an action A before uncertainty is realized. After uncertainty is realized, some state S is realized where you wish you had taken action B instead. This is definition of regret I was using and it proves according to Leonhardt that action A was not the optimal action.
I say that is not quite correct as B may not be better than A BEFORE uncertainty is realized. Hence, ex post regret does not prove the ex ante choice was wrong.
The article is suggesting that decision makers sometimes put high probability on some state (usually a “good” state for them), and take an ex ante decision that is optimal given that optimistic assessment. But this assessment turns out to be wrong. I do not disagree with this point and I bet people do it all the time. I just say that one of his empirical arguments in favor of it is incorrect.
Thanks for your comment.
June 1, 2010 at 10:20 pm
Zachary
This is something that has bothered me for a while about expected value. From what I understand, BP should have banked enough money to cover the expected value of a major disaster. But that means they save the cost of the disaster multiplied by the probability of it happening.
When reality rolls around, the disaster either happens or it doesn’t. If it doesn’t, saving the expected value limits your losses from saving for a nonexistent disaster. But if it does happen, the money you’ve saved is insufficient to cover it, and by a longer shot the less likely (and probably more costly) the disaster.
While expected value calculations work well when you have a large number of events to pool, it seems to be very lacking as a tool for mitigating an individual outcome.
Is there some other risk neutral strategy individuals can use the way a group uses expected value?
June 2, 2010 at 2:36 pm
Spill Logic « Eisenhood
[…] comments: But this does not prove the case. You may buy a stock given the odds of it going up or down. […]
June 3, 2010 at 12:06 pm
Dr. Harris Meyer
Jeff seems to have a point.
June 6, 2010 at 2:12 am
Regretting What I have done Correctly? | Statistically Brutal
[…] Regretting What I have done Correctly? Posted on June 6, 2010 by monoself A short article in Cheap talk […]