The Volokh Conspiracy

A very interesting concurring opinion yesterday, in United States v. Craig (7th Cir. Dec. 18, 2012); here’s a slightly trimmed version:

[From the panel opinion, which Judge Posner joined.] The defendant pleaded guilty to four counts of producing child pornography. He produced them by photographing his repeated sexual assaults on a girl who was a friend of his daughters and sometimes slept over at his house. He obtained additional pornographic images of her by threatening to kill her unless she photographed herself in sexually explicit poses and emailed him the images. The abuses began when she was 11 years old and continued until she was 14....

[T]he statutory maximum sentence for each count of conviction was 30 years. (It would have been longer had the defendant had previous convictions, but he didn’t.) The judge sentenced him to the 30–year maximum on one count and to concurrent sentences of 20 years on each of the remaining three counts, but he ordered that the set of 20–year sentences be served consecutively to the 30–year sentence, making the total sentence 50 years. The judge was entitled to do this....

[Judge Posner's concurrence:] I write separately merely to remind the district judges of this circuit of the importance of careful consideration of the wisdom of imposing de facto life sentences. If the defendant in this case does not die in the next 50 years he will be 96 years old when released (though “only” 89 or 90 if he receives the maximum good-time credits that he would earn if his behavior in prison proves to be exemplary).... [I]n all likelihood the defendant will be dead before his prison term expires.

Federal imprisonment is expensive to the government; the average expense of maintaining a federal prisoner for a year is between $25,000 and $30,000, and the expense rises steeply with the prisoner’s age because the medical component of a prisoner’s expense will rise with his age, especially if he is still alive in his 70s (not to mention his 80s or 90s). It has been estimated that an elderly prisoner costs the prison system between $60,000 and $70,000 a year.

That is not a net social cost, because if free these elderly prisoners would in all likelihood receive Medicare and maybe Medicaid benefits to cover their medical expenses. But if freed before they became elderly, and employed, they would have contributed to the Medicare and Medicaid programs through payroll taxes — which is a reminder of an additional social cost of imprisonment: the loss of whatever income the prisoner might lawfully have earned had he been free, income reflecting his contribution to society through lawful employment.

The social costs of imprisonment should in principle be compared with the benefits of imprisonment to the society, consisting mainly of deterrence and incapacitation. A sentencing judge should therefore consider the incremental deterrent and incapacitative effects of a very long sentence compared to a somewhat shorter one. An impressive body of economic research ... finds for example that forgoing imprisonment as punishment of criminals whose crimes inflict little harm may save more in costs of imprisonment than the cost in increased crime that it creates. Ours is not a “little crime” case, and not even the defendant suggests that probation would be an appropriate punishment. But it is a lifetime imprisonment case, and the implications for cost, incapacitation, and deterrence create grounds for questioning that length of sentence.

For suppose the defendant had been sentenced not to 50 years in prison but to 30 years. He would then be 76 years old when released (slightly younger if he had earned the maximum good-time credits). How likely would he be to commit further crimes at that age? ... [A]lthough persons 65 and older are 13 percent of the population, they accounted for only seven-tenths of one percent of arrests in 2010. Last year 1,451 men ages 65 and older were arrested for sex offenses (excluding forcible rape and prostitution), which was less than 3 percent of the total number arrests of male sex offenders that year. Only 1.1 percent of perpetrators of all forms of crimes against children are between 70 and 75 years old and 1.3 percent between 60 and 69. How many can there be who are older than 75?

It is true that sex offenders are more likely to recidivate than other criminals, Virginia M. Kendall and T. Markus Funk, Child Exploitation and Trafficking: Examining the Global Challenges and U.S. Responses 310 (2012), because their criminal behavior is for the most part compulsive rather than opportunistic. But capacity and desire to engage in sexual activity diminish in old age. Moreover, when released, a sexual criminal is subject to registration and notification requirements that reduce access to potential victims.

As for the benefits of a lifetime sentence in deterring other sex criminals, how likely is it that if told that if apprehended and convicted he would be sentenced to 50 years in prison the defendant would not have committed the crimes for which he’s been convicted, but if told he faced a sentence of “only” 30 years he would have gone ahead and committed them? ...

Sentencing judges are not required to engage in cost-benefit analyses of optimal sentencing severity with discounting to present value. Such analyses would involve enormous guesswork because of the difficulty of assessing key variables, including one variable that I haven’t even mentioned, because I can’t imagine how it could be quantified in even the roughest way — the retributive value of criminal punishment. By that I mean the effect of punishment in assuaging the indignation that serious crime arouses and in providing a form of nonfinancial compensation to the victims.

But virtually all sentencing, within the usually broad statutory ranges — the minimum sentence that the judge could have imposed in this case, by making the sentences on all four counts run concurrently, as he could have done, would have been 15 years, and the maximum sentence, by making them all run consecutively, as he could also have done, would have been 120 years — involves guesswork. I am merely suggesting that the cost of imprisonment of very elderly prisoners, the likelihood of recidivism by them, and the modest incremental deterrent effect of substituting a superlong sentence for a merely very long sentence, should figure in the judge’s sentencing decision.

I’m inclined to think that there’s a very strong retributive case for life imprisonment for this particular defendant, but I agree with Judge Posner that there are real costs to such sentences, and they should indeed generally be saved for situations where there is a powerful imperative for maximum retribution (short of a death sentence, which the Court has precluded for child rape and nearly all other crimes short of murder). Thanks to Jeffrey Sarles for the pointer.

In the new issue of Regulation I review Retaking Rationality: How Cost Benefit Analysis Can Better Protect the Environment and Our Health by NYU’s Richard Revesz and Michael Livermore, a progressive defense of cost-benefit analysis, albeit reformed to make it more regulation-friendly.  My review is available here (beginning on the third page of the PDF).  Here is how I conclude:

Revesz and Livermore claim their reformswould yield “an administrative state that is more efficient and fair, and deliversmore environmental, health, and safety protection for less cost.” Who could be against that? They argue that “the most appropriate and natural role for cost-benefit analysis is to help find the regulatory sweet spot, the optimal point that is between not enough and too much.” Yet at other places, it is unclear whether their aimismore “neutral” regulatory analysis or simply more regulation. While they often stress the importance of “neutral” analysis, they also presume such analysis will produce particular results, and trumpet this claim to their presumably progressive audience.

They are correct to highlight the need to consider the distributional implications of regulatory decisions. But they ignore the broader ethical questions about when government intervention in private economic decisions is appropriate. Cost-benefit analysis can inform regulatory policy, but it is
insufficient to determine when regulation is itself desirable. It is a powerful tool that can enhance the understanding of a regulation’s likely effects, but it is also prone to misuse. It can provide a veneer of technical precisionto regulatory judgments and augment the political case for action.

As the authors show, pro-regulatory interests could have much to gain by deploying a regulation-friendly cost-benefit analysis. Yet this will not make it any more “neutral,” nor will it ensure that better regulatory policies result.

In the various discussions about airport and airline security here at VC, a common response in the comments is something like this one (some version of this pops up from many commenters, and I’m just pulling up the most convenient example):

With the current procedures, flying is one of the safest things we do. Even with an occasional successful bombing, we would still be way below any level of acceptable risk. If we want to curtail civil liberties to save the lives of people flying, we should start by screening folks we allow to drive to the airport, you’re much more likely to get killed doing that.

It is a costs versus benefits argument (actually a couple of slightly different ones) pitting the costs of a successful bombing and “acceptable” risk, with a suggestion that an appropriate metric, by implication of preferring to regulate it instead, would be the drive to the airport.  It is a theme of much of the skepticism about US counterterrorism policies, a skepticism rooted in cost benefit analysis, but perhaps more accurately framed as skepticism about the proper things to be compared – what kinds of costs and what kinds of benefits?  Matthew Yglesias perhaps exemplifies the skeptical view following the Christmas attack, in a post titled “Not So Scary ‘Terror’”:

Obviously, people shouldn’t be lighting anything on fire inside airplanes. That said, all the big Christmas airline incident really shows to me is how little punch our dread terrorist adversaries really pack. Once again, this seems like a pretty unserious plot. And even if you did manage to blow up an airplane in mid-air, that would be both a very serious crime and a great tragedy, but hardly a first-order national security threat. [Edited out Peter King quote.] ...

Ultimately, it does no favors to anyone to blow this sort of thing out of proportion. The United States could not, of course, be “devastated” by anything resembling this scheme. We ought to be clear on that fact. We want to send the message around the world that this sort of vile attempt to slaughter innocent people is not, at the end of the day, anything resembling a serious challenge to American power. It’s attempted murder, it’s wrong, we should try to stop it, but it’s really not much more than that.

I don’t think this is right, for a number of reasons, starting with thinking that it is not the right way to approach cost-benefit analysis – more exactly, what one should count as categories of costs and benefits to weigh against each other.  I’ve elsewhere partly explained my views on how cost benefit analysis requires a prior view of “plausible” comparisons, arguing that skeptics like John Mueller are making “inapposite” comparisons.  But I’m interested to know what VC commenters think is the right way to approach this; I think that “plausibility” and “appositeness” of comparisons matter, and in fact form an often-covert base of assumptions in applying cost-benefit analysis, whether to skeptical or non-skeptical ends.

So let me ask.  What is right or wrong with this skeptical approach to such things as airline security on the grounds, for example, that the ride to the airport is more dangerous, or that one’s chances of getting struck by lightning are higher than getting killed by terrorism, or that even if you did manage to blow up a plane, it is not a first-order national security threat?

Let me be very clear on the question.  I am not asking your views on airport security, terrorism, Matthew Yglesias, or such things, not directly.  I am asking readers to say what is right or wrong about cost benefit analysis used in these ways – and more particularly, is it okay or not to include all of these kinds of considerations as the “frame” for costs and benefits?  The skepticism assumes that a wide range of things can be included as points of comparison – is this right, and if not, why not?  And if not, what are the limits, if any?  No rants, please, and confine responses to the methodological question about CBA and its underlying assumptions.

I am going into isolation for a bit to rewrite the last chapter in my UN book, “Returning to Earth: When and How the United States Should Engage, and Not Engage, with the United Nations” – a chapter on the end of the human rights era at the UN.  But I wanted to wish everyone a happy new year.  And thank people for the excellent comments on cost-benefit analysis and the “line in the sand” question – please keep any comments coming in, particularly as relevant to the issue of how CBA deals, or does not deal, with deliberately arbitrary lines, in a form of analysis arising out of marginal incrementalism.   (I am less interested in debates over particular issues to which CBA might be applied; I’m trying to understand it as a conceptual form and, if you like, something of its intellectual history and pedigree.)  And, finally, to commend to you this essay by Jim Manzi, “Keeping America’s Edge,” at National Affairs journal – curious to return and see what people thought of it.

Many of us who write, read, and comment on this blog work frequently with cost-benefit analysis, perhaps typically through discipline specific tools, whether in economics or finance or business or engineering or other disciplines.  In my case, in my day job I’m mostly an international business-finance professor who uses the typical, and really not very sophisticated, tools of net present value, discounted cash flow, and so on, in my day to day work.  I apply these ideas sometimes in my work in public law, but mostly these are, in my case, workaday tools in relatively narrow business contexts.

I understand CBA, that is, at the highly discipline-specific level of standard private firm decision-making.  And I think, after a fair amount of philosophical study, I understand it at the most abstract level as consequentialism, and its many philosophical arguments.  But oddly, I don’t think – despite reading a couple of textbooks and much besides on public (rather than private firm) finance theory – I really understand the “mid-tier” of cost-benefit analysis applied to public policy problems, in the way that, for example, Cass Sunstein writes about pretty much everything as a cost-benefit problem.  Yet this is where CBA seems to be most offered as a policy template – and yet which puzzles me in many ways.

The puzzles include, for example, how one compares different values that seem to me fundamentally incommensurable.  I am currently grappling with the question in precisely this “mid-tier” public policy context of proportionality in the law and ethics of war.  Among other things, it does not seem to be very much of a problem when I am working in private firm CBA, such as NPV.  Why not?  Mostly because private firm decision-making tends, as a matter of process, to force firms to compare projects that are, because of the profit nature of firms, about comparing anticipated rates of return.  Private finance might lead one to consider quite different economic activities – should we invest in Twinkle by Wenlan or steel tubing? – but ultimately we reduce to anticipated rates of return as the common question.

Public finance is different, because the goods sought are both more heterogenous – national security and health – but also often necessary, in some amounts, ratios, and costs.  That is, a private firm mostly engaged in nuclear engineering has no necessary reason, save for some calculation of efficient return, to also invest in a lingerie boutique.  But public entities do have to invest in highly heterogenous and yet essential goods.  It’s easy to say that how much of one social good versus another social good is simply a “normative” question, and once you’ve answered that, then hooray, you’re back to good old CBA – but really, that mostly simply avoids the question in order to get analysis back to a place where one is comfortable, but where the form of CBA analysis is mostly irrelevant.

True but trivial, in other words, as a matter of method.  CBA doesn’t really seem to answer the important question in public (as distinguished from private business) contexts, except by saying that we have to make tradeoffs.  If that is news to you – and, to be sure, it seems to be to many of my law students – then it is important, absolutely.  But beyond that, as applied to problems of incommensurables, rather than the distinctively commensurable world of private firm capital budgeting, it seems rather empty.  Both too much of a method, and too little.

A second, closely related, puzzle for me about cost-benefit analysis, particularly in the context of security debates, is that I do not understand how CBA is supposed to encompass the fundamentally strategic idea of “drawing a line in the sand.”  A line that is, by its nature, arbitrary – not to be crossed without drastic consequences that were not required just short of it, and which do not seem required just crossing it.  We understand intuitively the idea of “boundaries” as having a huge gaming function, to signal intentions and convey threats, warnings, and generally expectations about the future.  In security, diplomacy, international relations, we have no problem with borders, bright lines, fixed rules and threats, all sorts of things.

Additionally, we have a whole structure of such theory in, for example, contract and other legal areas.  Negotiation theory is full of it, for example.  And of course I can make CBA embrace such thinking by inventing a special calculus of costs and benefits associated with clear signals and expectations – but that is the virtue and vice of CBA, in that it can always invent a special calculus of costs and benefits, and finally so what?

But CBA itself does not really embrace such thinking as en essential part of its method.  It is, after all, part of the ‘marginal revolution’ – and drawing lines in the sand is not very naturally reconciled with the incremental decision-making that characterizes marginality.  This is one of the reasons why, in national security and counterterrorism, it is both so easy and, to my mind, so irrelevant to base critiques of national security measures that do involve, directly and indirectly, drawing lines in the sand, on a certain CBA-inspired skepticism about why any such line is arbitrary.  It is arbitrary, says marginalist CBA skepticism, so why get so worried about it?  This is what inspires – quite mistakenly, I think – the CBA skeptics of counterterrorism (John Mueller, as I’ve mentioned before, is perhaps the most ranting).

No airliner downed, in other words, is worth all the costs and uncertain costs and benefits of, say, war.  Sure, we can talk about a “future series of discounted-expected-downed airliners” and see if the costs of the discounted series makes war worth it.   But let’s be clear that in such case, we are forcing quite artificially a method that is rooted in incrementalism to embrace “lines” and discontinuous “cliffs.”

CBA leads to serial decision-making, precisely because it is a method based in the “marginal revolution” – that’s what’s good about it, but also what limits its methodological scope.  That’s part of what makes it so “relentlessly tactical,” as I have quoted Philip Bobbitt (one of the grand strategists, one of the few who can strategically link military history and strategy, diplomacy, and law).  As my colleague Jonathan Baker – now chief economist of the FCC – remarked to me (I don’t think he’ll mind me quoting him), cost benefit analysis is a method of analyzing and reacting to proposed courses of action, it is “not a method of generating them in the first place.”  Not merely inadequate to the analytic task, which requires forms of thinking that go beyond serial “sunk costs” and marginality – but in some way allergic to the idea of strategic forms of thinking that encompass the ideas of line-drawing, initiating action, gambit, and envelopment.  How does one draw the strategic metaphor of ‘envelopment’ into cost benefit analysis?

You can always expand the cost-benefit frame, make it sufficiently elastic to encompass all these costs and benefits across time and space, yes, sure.  But by the time you’ve done that, it seems to me you’ve invented a new method called “strategy” and called it “cost-benefit analysis.”  It is the strategic thinking that does the work, not discounted probability of gain or loss.  So here is my question:

The VC readership is pretty steeped, from various disciplinary perspectives, in forms of cost-benefit analysis.  How does, and how should, and is it really possible, for CBA to take into account the idea of “drawing a line in the sand” in the “arbitrary” sense of here and no further?  When Captain Picard (and, as Ilya points out, that’s Sir Patrick to you, now!) says, at the opening of the Borg movie, “The line must be drawn here,” what theory of decision-making justifies that?  How does it fit, if at all, into cost-benefit analysis?

(And, to be clear, no fair simply bending CBA so far out of shape that one maintains the form but the real work is done by something distinct.  Or at least, if that’s what you’re doing, say so and say why.)

Bonus question:  what is “strategy,” in this case?  I have used the term here by contrast to CBA.  I have also used it by contrast to “tactics”; note, too, that the strategy-tactics distinction does not really occur in general game theory.  What analytically does “strategy” mean, then, as against these analytic categories that have been framed as distinct from it?