Remember the scandal over the faulty GM ignition switch? If you believe Congress and the press, GM’s Ions and Cobalts are rolling death traps. I argued in Slate that the statistics show they are not. The incremental effect of the faulty ignition switch on the risk of getting killed in an accident is close to zero. Two members of Congress asked NHTSA why it hadn’t told people to stop driving the cars until they’re repaired. In response NHTSA explains:

In other words, the cars aren’t dangerous. So to all appearances this really was a classic Kuranian-Sunsteinian availability cascade.

How is it possible that Justice Scalia and his clerks missed the error in his statement that the EPA had tried to use cost-benefit analysis to justify the regulation at issue in Whitman? How did the other justices and their clerks miss it, too? The answer is that the mistake made no difference whatsoever to the legal analysis. It signifies nothing about anything except that people sometimes make mistakes.

One of the many ironies of the argle-bargle in the press and the blogs–aside from the several listed by Jonathan Adler–is that everyone gets the political valence of the mistake wrong. The narrative is that the mistake exposes once and for all what we knew all along: that Scalia allows his political biases (or even partisan biases) to guide his legal reasoning. But Scalia was objecting to EPA’s use of cost-benefit analysis–a decision-procedure normally favored by conservatives and loathed by most (although not all) liberals–when the statute (in his view) did not allow for it. Scalia’s preferred resolution of this case would have sent the regulation back to EPA for revision, it is true. But it would also have required EPA in the revised regulation to impose stricter controls at greater cost to industry.

Indeed, the majority makes just this point. In response, Scalia denies that his interpretation would lead to overregulation, but I confess I can’t follow his convoluted reasoning. If the EPA, wielding cost-benefit analysis, chose the most efficient regulation, then any other regulation must by definition be less efficient (or I suppose equally efficient, which I think is Scalia’s limited claim, but that also seems quite unlikely). If downstream states are going to reach attainment at higher cost to upstream states, then this cost will be passed on to industry in upstream states. Because statutes are written in broad terms and directed at reducing harmful activity, CBA typically leads to less regulation than that required by a literal interpretation of a statute, not more. That’s why liberals have traditionally been hostile to it.

Many years ago, Cass Sunstein wrote a paper called Cost-Benefit Default Principles, which argued that regulatory agencies are free to use cost-benefit analysis when statutes are ambiguous. EPA v. EME Homer City Generation seems to confirm this thesis, but arguably goes beyond it, at least if you accept Scalia’s argument that the text is clear and clearly forbids EPA to take account of cost. The case may instead stand for the existence of a stronger cost-benefit presumption that entitles a regulator to use cost-benefit analysis unless Congress explicitly forbids it to. Thus, a “gap” is not needed. The justification would be that Congress ought to direct regulators to take account of costs, perhaps that cost-benefit analysis is a quasi-constitutional commitment, so statutes should be “interpreted” as much as possible to avoid conflicting with this commitment.

Cost-benefit analysis as a constitutional commitment, quasi- or other? Could there be a justification for such a claim? Possibly. Consider the longstanding worry that the New Deal regulatory state gave excessive discretion to regulators, and the defunct efforts to rein them in with the nondelegation doctrine. Scalia’s insistence on deferring to the text, however loopy the text might be, is one possible response, but a more realistic one, I think, is rather to confine regulators’ discretion by insisting that they use cost-benefit analysis. The Court did not go quite so far, but took a step in that direction.

This claim, which appears in many news articles, should sound strange to a bankruptcy lawyer. Owners of defective cars manufactured before GM’s bankruptcy in 2009 have claims again Old GM, which no longer exists, not New GM. In fact, as is standard in large corporate bankruptcies, a new entity was created, a pot of money that victims of Old GM can sue. That entity is called Motors Liquidation Co. I don’t know the details, but it appears–and this is standard–that some of Old GM’s assets were put in ML, so that victims who discover themselves as such post-2009 can collect damages against somebody. Of course, those victims (probably) cannot collect 100 cents on the dollar, but that is always what happens in bankruptcy.

Old GM’s shareholders were wiped out; New GM’s shareholders are Old GM’s creditors or people to whom those creditors sold their shares. There is no reason to hold any of these people liable for the sins of Old. An easy way to think about this is to imagine that Old GM had been liquidated, its Ion and Cobalt factories sold to investors. It would make no sense to make those investors liable for defects in cars manufactured in those factories before they bought them.

But this creates an odd situation. New GM might inspect its records or receive complaints and realize that those old Ions and Cobalts are defective. Does it have an obligation to alert the owners of cars manufactured by some other company (that is, Old GM)? To fix the cars? It appears that some such obligation exists in federal regulatory law. There is an interesting question here what would happen if GM had been liquidated and these lines of cars discontinued, and its assets divided among multiple companies. I suppose in such a case, no one would bear the obligation to recall and fix. Such a rule would make liquidation more attractive to creditors than reorganization, all else equal, which is not a good thing.

So what of the claim that New GM is hiding behind bankruptcy law? Maybe the argument is that New GM delayed the recalls because any fines levied on it by NHTSA would be minimal, less than the cost of fixing the cars, while tort liability is zero thanks to the bankruptcy. This wouldn’t surprise me. But it seems like a hole in the law that Congress should patch, and the way to patch it is not self-evident–again because burdening new firms with liabilities arising from the activities of predecessors will make reorganization unattractive relative to liquidation. It is not clear that that New GM has acted in a blameworthy fashion by minimizing its responsibility for liabilities that belong to Motors Liquidation.

This very good student note examines how “dignity” has played a role in regulatory impact assessments produced by regulatory agencies. President Obama’s executive order 13,563, which renewed a longstanding requirement that regulators conduct cost-benefit analysis of major regulations, famously introduced a new provision allowing  regulators to consider the effect of a regulation on “human dignity.” Implicitly, at least, regulators were thus authorized to issue a regulation that fails a formal cost-benefit test if it advances this value. (I briefly discuss the order here.)

Many commentators worried (or hoped) that the human dignity requirement would allow regulators to issue expensive rules that failed cost-benefit analysis. Couldn’t the EPA, for example, say that pollution of all types send people to hospitals, where they undergo procedures that violate their dignity? Or could the Department of Transportation impose expensive new safety requirements on cars because people who are maimed in accidents lose their dignity? In fact, Bayefsky cites an EPA analysis that suggested that a pollution regulation would be justified in order to avoid or minimize the indignity of a protracted death “involving prolonged suffering and loss of dignity and personal control.” If such a step were to be taken, there would be little left of cost-benefit analysis.

Yet there are numerous regulations that seem to advance values relating to dignity where the benefits would be hard to quantify. Bayefsky provides a number of examples: regulations that increase access to facilities like bathrooms to people with disabilities (who might otherwise need personal assistance), reduce the incidence of prison rape, and protect patients’ health information. Can the “dignitary” benefits associated with these regulations really be monetized?

Bayefsky says no. I suspect the answer is yes, at least sometimes. The paper is well worth reading for anyone interested in this topic.

(The image is from Wikipedia.)

As it turns out, less a defense than a plea (sure to be ignored) to do a cost-benefit analysis before condemning GM.

I have dug deeper into the numbers and have found nothing to cause me to change my mind about my conclusion yesterday that GM may well have acted reasonably, and certainly that it is premature to conclude that it acted in a scandalous, criminal, or unreasonable fashion. A student of mine produced this spreadsheet, which makes clear that GM’s culpability depends on how soon it understood the problem. If within the first few years, then yes. If after 5 or 6 years, then probably not. (Her spreadsheet is based on my assumptions from yesterday, which are–caveat emptor–possibly dubious, plus she extends the time frame out another 10 years to consider cars manufactured just before the recall. She also treats this as an NPV exercise based on expected tort liability rather than a cost-benefit analysis but the conclusion would be the same assuming that the assumed tort awards reflect statistical valuation of life.) Until it is clearer who knew what when, one cannot determine whether GM acted wrongfully (in the sense of causing harm to people; it’s impossible to deny that their cars are crappy however safe they may be). I’m afraid it probably won’t be clear for years, until after lawsuits and investigations are concluded.

Cass Sunstein and Timur Kuran wrote a fine paper in 2007 that discusses regulatory panics based on what they call “availability cascades”–where some high-profile corporate misfeasance that feeds into people’s anxieties causes an enormous scandal despite no evidence that it harmed anyone. The most famous case is Love Canal, which harmed no one. Another is Alar. Also read this paper on the  Ford Pinto scandal by Gary Schwartz.The Pinto case was another regulatory panic where fears greatly outstripped the real harm and wrongdoing. The common elements in all these examples are (1) sympathetic victim groups who effectively work the media, (2) opportunistic politicians and commentators, and (3) lazy journalists.

The House Committee on Energy and Commerce is holding a hearing on whether GM delayed too long before recalling cars with defective ignition switches. It was revealed at the hearing that the cost of the replacement part is only 57 cents, and this has led to calls for the heads of those responsible. This, from Michael Moore, is typical:

I hope the criminals at General Motors will be arrested and made to pay for their premeditated decision to take human lives for a lousy ten bucks. The executives at GM knew for 13 years that their cars had a defective ignition switch that would, well, kill people. But they did a “cost-benefit analysis” and concluded that paying off the deceased’s relatives was going to be cheaper than having to install a $10 part per car.

The $10 figure is from a Washington Post article; the 57 cent figure can be found in other sources.

I was curious about whether it is true that GM’s decision to install the switch (rather than replace it with a better switch) or not to recall the car in 2007 really would have failed a cost-benefit analysis. Here are some back-of-the-envelope calculations, based on (still murky) details from a House report and newspaper stories.

Let’s suppose that back in 2005, which was the model year for the first car with the ignition switch problem, GM knew that it would install that switch in 2.6 million cars (the eventual number that would be recalled). Those cars were sold over the next 9 years. If we assume a constant number of cars sold per year, we get 290,000 cars sold per year. A little additional math shows that this amounts to 13,050,000 car-years, by which I mean the number of years in which a car with the ignition switch was owned and driven by someone.

During this period, the switch problem caused accidents that caused 13 deaths, or 0.000001 death per car-year. It is worth noting parenthetically that the probability of an average person in an average car being killed in any year is 0.0000556. So the effect of buying a GM car with an ignition switch problem increased the risk from 0.0000556 to 0.0000566.

Still, no one wants to take an unnecessary risk of death, however small. But how much would you be willing to pay to reduce the risk of dying from a car accident by 0.000001 in one year? One way to get the answer is to use the U.S. government’s valuation of a statistical life, which was around $6 million when these decisions were made. (That number is based on data that measures how much people are willing to pay to avoid very small risks of death, as in our example.) Another is to look at tort awards, which at the time averaged about $3 million for deaths. Let’s call it $4.5 million. And assume that the average buyer drives the car for 10 years. Multiplying these numbers together, a person should be willing to pay about $45 to avoid this risk of death ($4.5 million * 0.000001 per year * 10 years).

Now if GM could have fixed the problem by using a 57 cent or $10 switch at the time of manufacturing, it should have. But it appears not to have recognized the problem until most of the cars were on the road. The Washington Post says that repairing the car would take less than an hour. Let’s call it half an hour, and assume $100 per hour in labor costs, which is roughly consistent with the CEO’s testimony that it would have cost GM $100 million to  recall the 2.6 million cars on the road. If all this is true, the decision not to recall–saving $50 to avoid an expected cost of $45 per car–seems reasonable, although as we have seen the proper measurement of a loss of life is controversial. The bottom line is that GM could very well have complied with a reasonable cost-benefit analysis; if not, it was probably close. To know the truth, we need more data than are publicly available. So why the headlines? (That’s a rhetorical question.)

Are my calculations or assumptions wrong? If you see an error, please contact me here.

This book, edited by Cary Coglianese, Adam Finkel, and Christopher Carrigan, is out. My copy arrived and I have been paging through it. I became interested in this topic several years ago, and wrote a paper with Jonathan Masur arguing that regulators should monetize the expected costs from job loss caused by proposed regulations when they conduct cost-benefit analyses. Agencies have never done this, perhaps because in standard economic models, a job loss is not actually a social cost–the unemployed worker just gets another job, or capital shifts to a less regulated sector of the economy where new people are hired. But recent empirical work suggested that the social costs of a job loss could be high–in the neighborhood of $100,000–because human capital is destroyed, among other effects.

So we wrote this paper and were subsequently invited to participate in this conference, where a very strong group of people delivered papers collected in this book. Many papers dealt with the important but age-old question of whether regulation itself destroys jobs (maybe). I was more interested in whether people thought it would make sense for regulators to treat job loss as a cost in cost-benefit analysis . It looks like a maybe-to-yes with a great deal of cautious skepticism. One worry, which is a real one, is that agencies just can’t handle it. Agencies aren’t very rigorous in their existing cost-benefit methodology, and requiring them to look at job-loss effects may be too much. Still, I hope to see some experimentation in government. OIRA asked for comments on this topic a while back; I don’t know if anything came of it. There will be resistance from people who fear that our approach would reduce the amount of regulation.