Friday, June 04, 2010

What To Do When Regulators are Unreliable

-- Posted by Neil H. Buchanan

In my FindLaw column this week (available here), I return to the still-expanding environmental crisis in the Gulf of Mexico and discuss appropriate changes in energy policy going forward. I expand on a point from my recent Dorf on Law post on the topic regarding the unknowable nature of "rare disasters," i.e., the pure guesswork involved in assigning expected costs and probabilities to things that we have never before experienced (e.g., all-out thermonuclear war, the death of an entire section of a sea, and so on). I try to make the case that the Gulf disaster calls for a renewed assessment of all of the truly awful scenarios that we have been implicitly assuming would be somehow manageable and remediable at something approaching a reasonable cost.

The point that I want to explore further in this post is the question of how to change the way we regulate businesses in the aftermath of the Gulf disaster. In the FindLaw column, I essentially make the case that we should systematically choose higher-immediate-cost alternatives if those alternatives do not involve huge worst-case scenarios. In the context of energy, that argument implies that coal is better than nukes or oil, because the bad things that happen with coal use (worker deaths and environmental damage) do not include either "known unknowns" or "unknown unknowns" (in the inimitable phrasing of our former Secretary of Defense). Coal is awful, and it needs to be replaced with renewables as soon as possible; but at least it will not lead to the sudden wiping out of entire regions (nukes) or ecosystems (oil).

It would, of course, be possible to instruct regulatory agencies to change their environmental impact statements, etc., to take high-cost, low-probability events more seriously. In other words, rather than saying, "Don't do anything that involves a big, unknowable risk of catastrophe," we could instead say, "Regulators must do a better job of assessing all the risks of various activities, including the ones that seem remote and implausible (so that, for example, we do not issue permits to drill for oil where we lack proven methods to stop oil from gushing out of inaccessible holes)."

At its core, this difference mimics the "conduct vs. structural remedies" debate in antitrust law. There, the basic question when a company is found guilty of monopolistic behavior is either to order it to stop behaving monopolistically (conduct remedies) or to break it up (structural remedies). I argue in yesterday's FindLaw column that the one fact about regulation that we should learn from the Gulf disaster is that regulatory agencies are much worse at enforcing behavioral regulations (conduct remedies) than we previously believed. Part of the outrage of the post-April 20 world, after all, has involved the mounting number of examples of regulatory failure: waivers given to BP to skip environmental impact statements, the mess at the Minerals and Management Service (which apparently began during W's tenure but was never fixed by the Obama people), etc.

I am sure that there are plenty of Poli Sci explanations for how the capture of those agencies has intensified over the years. Whatever the explanation, however, this window into the systematic corruption of regulatory agencies suggests that categorical rules (structural remedies) will be increasingly appropriate going forward. There will still be cases where behavior must be regulated, of course, but the strong preference should now be to choose the form of legal control that does not assume that agencies are (and will remain) competent and uncorrupted.

This conclusion, of course, can (and in my opinion, definitely does) extend beyond energy and environmental policy. In fact, my view turned around pretty dramatically during the health care debate last summer, when I originally argued in favor of strong regulations rather than a public option. When it became clear that neither political party would actually put in place the necessary limitations on insurers' conduct (nor would anyone appropriately arm the agencies that would have to enforce those limitations), I argued that the public option was clearly superior to what turned out to be the ultimate bill (viewed by Obama as a "success"): more of the same, with mild changes to the rules of the road. (Don't get me wrong: the health care bill was better than nothing; but it is still fundamentally based on an untenable model of regulation.)

Similarly, the recently-passed financial regulation bills awaiting reconciliation are based on simply tightening the rules and rearranging the regulatory agencies in the financial sector. Liar's Poker author Michael Lewis recently wrote a humorous but bitterly ironic piece for The New York Times arguing that Wall Street is secretly ecstatic about the bills, despite the presence of some annoying new rules, because nothing in the bills would change the fundamental structure of the financial system. None of our supposed concerns about institutions being "too big to fail" actually infected the bill, and therefore nothing serious will change for the big players.

It is possible, of course, to take my argument to absurd extremes. I am not arguing against behavior-changing rules in every case. I am saying simply this: We now know just how likely it has become for agencies to fail at their jobs. That knowledge should cause us to make choices that reduce the need for ongoing oversight in favor of changing the size and structure of market players. The less we have to rely on watchdogs, the better.


Michael C. Dorf said...

The principle Neil proposes here is quite similar to a principle of regulation in much of Europe. The so-called "precautionary principle" has been formulated in a variety of ways. Dan Farber presented a paper recently (not yet on SSRN, apparently) in which he argued that U.S. regulators should use a medium-strength version of the precautionary principle. Whereas Europeans tend to say that where there are uncertain risks, there should be a strong presumption against the activity, American regulators tend to ignore uncertain risks. Farber proposes a middle course that takes into account small-but-uncertain-size risks of catastrophe. He's especially concerned about so-called "fat tail" distributions. In a normal distribution, the very large damage events have a vanishingly small probability as one moves out from the median-size damage events. But because of feedback effects (think about the bursting of a financial bubble or the effect of melting permafrost), some distributions have fat tails--i.e., the probability of the catastrophic events doesn't approach zero all that rapidly.

Farber's own solution didn't really attempt to quantify these issues but simply to urge American regulators to pay more attention than they do now, though less than Europeans do.

Interestingly, taking account of potential catastrophes need not mean bans. It could just mean very stringent regulations. Note that offshore drilling in the North Sea and the generation of nuclear power in Europe and Japan seem to fall into that category.

Charles T. Wolverton said...

The fundamental problem as I see it is not probability distributions with "fat tails" (notwithstanding that I have lemming-like also described LTCM-BP type problems in those terms - carelessly and incorrectly I now think), but rather that the "probability" of a rarely occurring - perhaps even never having previously occurred - real event is not meaningful. Here,"real" is meant to distinguish between possible events in the actual scenario and events in models in which simplifying assumptions allow derivation of theoretical probability distribution functions.

If in addition to the very-small-but-uncertain (and arguably meaningless) probability of the rare event there is a very-large-but-uncertain cost, you have an essentially arbitrary expected cost. In which case, the European approach (as described) seems pretty reasonable.

Neil H. Buchanan said...

I agree with Charles's comments in full. Indeed, his comments on my post from last month on this topic were very helpful in honing my thinking on the subject of unknowability and rare events.

Mike's description of Farber's paper suggests that I am more radical than Farber on the subject, which is fine with me. If the Europeans and Japanese are willing to opt for behavioral regulations over bans, that is probably because they can trust their regulators. We no longer can -- at least at the level necessary to deal with these kinds of disastrous possibilities.