The slippery slope of fat tails

A few weeks back, Gernot made reference to Martin Weitzman's work on analyzing climate change as a low probability, high-stakes event.  After a few weeks, I've had a little time to give some blog-level reflection to what Weitzman has to say and quite frankly, I'm not sure I like where the logic leads (I'm not sure I don't like where the logic doesn't lead either, but that's for another day). 

First, a little technical stuff.  Here is the abstract from Weitzman's paper:

With climate change as prototype example, this paper analyzes the implications of structural uncertainty for the economics of low-probability high-impact catastrophes. Even when updated by Bayesian learning, uncertain structural parameters induce a critical tail fattening of posterior-predictive distributions. Such fattened tails have strong implications for situations, like climate change, where a catastrophe is theoretically possible because prior knowledge cannot place sufficiently narrow bounds on overall damages. This paper shows that the economic consequences of fat-tailed structural uncertainty (along with unsureness about high-temperature damages) can readily outweigh the effects of discounting in climate-change policy analysis.

I know, blah-blah-blah.  So here's the blog version--If the likelihood of some really, really bad outcome is unknown, but unlikely and not impossible, the consequences of not avoiding the really, really bad outcome are really, really bad.  Action to avoid the really really bad (but unlikely) outcome is economically justified.

Sounds like a reasonable statement of the precautionary principle and Weitzman's paper goes a long way toward giving a mathematical formalization to the principle.  I don't want to argue with Weitzman's conclusions, but rather point out the slippery slope of using Weitzman's logic for policy analysis for other low probability high stakes events.

First let's look at an example.  Let's say two families of 5 are on vacation at a wooded resort in say, oh I don't know, Kentucky.  And let's say that upon arrival at said resort, the security personnel warn visitors to be cautious in secluded areas because the area is know for (poisonous) copperhead snakes, but it is unlikely (but not impossible) that a snake could wander--do snakes wander?--near the resort.  And for the sake of this example, let's suppose that the parties involved are city dwellers and don't know the difference between a copperhead and a harmless rat snake.  Now, suppose the six kids--all under the age of 12--are playing in the side yard when a six foot snake wanders onto the patio.  What is the correct policy decision here? 

According to Weitzmanian logic, the correct policy decision is to take action to prevent the unlikely, but not impossible, catastrophic outcome.  Why?  Because the expected losses from not acting are so high--that is, the losses if the unlikely catostrophic event happen are close to infinity--that the immediate costs of acting are warranted. 

In the snake example, the probability of a catastrophic event--death of a child--is low, but uncertain.  In such a case, it is economically justifiable to remove the threat by beheading the beast.  Wow, that actually makes me feel a little better.

Now consider another example.  Suppose that it is unlikely, but not completely impossible that a country with a totalitarian government possesses weapons capable of killing thousands, and that this country has shown a willingness in the past to use such weapons...

Like I said, a slippery slope.

Comments

[John Whitehead]

Tim,

Your examples are a bit unrealistic.

[Dano]

You are asking what is essentially a social science to have the predictive value of one of the natural sciences.

In my view it is not surprising that a person who is other-regarding is giving these weights to the chance of that outcome, as the premise and analysis is weighted with morality and ethics.

Predicting the chance of how a group will react is not the same as predicting the half-life in a collider of cluons and mohrons after bombardment by warons.

Jus' sayin'.

Best,

D

[Jim]

So pre-emptive nuclear strikes on Iran might be economically justifiable within this framework.
And?
I fail to see how this might represent a slippery slope except in the unlikely situation where decisions about interventionist foreign policy and war in the middle east are guided by measurable economic self-interest rather than international consensus or strategic national security concerns? Oh wait...

[Bob Murphy]

Yay Tim! Although maybe I shouldn't congratulate you and jinx it...

Sticking with the snake example, I think the better analogy would not be to behead the snake, but rather for your family to head back to the suburbs and camp in the back yard. Call it "alternative, renewable" camping. Sure it's not as fun as what you originally wanted to do, but safety first.

[Ironman]

I think we're all missing some obvious connections here. First, snakes die when Tim goes on vacation. Second, entire rows of beach homes vanish when Tim goes on vacation. Clearly, Tim's "vacations" represent an escalating threat to the environment.

I'm not saying that we should yet, but perhaps we should at least consider a pre-emptive nuclear strike to address the situation and prevent what would likely be even greater damage should Tim begin going to more popular travel destinations (although the "alternative, renewable" camping/sanctions approach might be effective in the short term.)

Or perhaps we could solve John's existence value problems and send Tim out on vacation wherever there have been recent Bigfoot sightings - surely that would be enough to really produce a real sasquatch cadaver!

[mark lazen]

From my comment on this post over at The Energy Collective:

I think you may be conflating two separate issues: the first is the danger of failing to address a low-likelihood, high-impact risk. The second is the possibility that the actions you take to address that risk will be totally idiotic and counterproductive.

[Tim Haab]

Mark,

You nailed it. But what seems idiotic in hindsight might seem genious or valiant or waraanted in foresight. Weitzmaniacal thinking does little to prescribe a solution. Only point put that action is warranted. Rather than eliminate the threat completely, a much more prudent strategy might be to attempt to lower the probability of the umnlikely event. Or, wait until some of the uncertainty resolves itself so we have a clear(er) picture of the likelihood of catastrophe. All fit within Weitzman's framework.

Tim

[odograph]

It strikes me reductio ad absurdum.

Sounds like a reasonable statement of the precautionary principle and Weitzman's paper goes a long way toward giving a mathematical formalization to the principle. I don't want to argue with Weitzman's conclusions, but rather point out the slippery slope of using Weitzman's logic for policy analysis for other low probability high stakes events.

As if there were a single category of "low?"

[odograph]

To expand a bit, there is tremendous data on how poorly humans understand risks in the "low" continuum. The good old Time piece, How Americans Are Living Dangerously is a good summary.

And sure, we better understand the trade-off of one risk for another ... but we don't do it by calling them all "low" and throwing up our hands.

(We probably would want to design GW responses that are either low-low risks themselves, or have other secondary benefits.)