It’s been a while since the first installment of my Houston Law Review piece on Rationality or Rationalism? The Positive and Normative Flaws of Cost-Benefit Analysis. (See the printed version, 48 Hous. L. Rev. 79 (2011), or the SSRN version, for all the footnotes.) Sorry about that — I was at the Kalamazoo 46th Annual International Congress on Medieval Studies. Anyway, here’s the next installment.
* * *
First, the questions that cost–benefit analysis asks may sometimes be simply incoherent. What’s the value of not damming the Grand Canyon or not polluting the Prince William Sound? Environmental economists have a couple of theoretical answers—“[c]ompensating [v]ariation” or “[e]quivalent [v]ariation.” For any change in the level of an environmental amenity, the “compensating variation” is the change in income that, after the change, would keep me at the same utility level. (If the change is for the worse, this is the “willingness to accept compensation” to assent to the change; if the change is for the better, this is the “willingness to pay” to obtain the change.) And the “equivalent variation” is the change in income that would lead to the same utility change as the change in the environmental amenity. (If the change is for the worse, this is the willingness to pay to avoid the change; if the change is for the better, this is the willingness to accept compensation to forgo the change.)
These two amounts are generally different, at least because of income effects, so the environmental economist has to commit to whether to use compensating or equivalent variation—basically, equivalent variation implies a property right in the change while compensating variation implies a property right in the status quo, so the choice between measures will inevitably involve some noneconomic thinking. (That valuations change based on the allocation of rights is itself a challenge to the cost–benefit approach.) But once the value-laden question of who gets the implied property right is resolved, estimating willingness to pay or willingness to accept is a “merely” technical problem.
What, then, does the economist do with an environmentalist who rejects both willingness to accept and willingness to pay? Willingness to accept might be literally infinite: “There’s no amount of money I’d accept to tolerate the damming of the Grand Canyon.” One might think that willingness to pay is more serviceable because, at least, it’s bounded above by one’s total lifetime net worth; but what if one rejects the question on moral grounds?
Practitioners of contingent valuation surveys, which try to estimate natural resource values from survey data, routinely have to deal with “protest response[s],” where respondents give a willingness to pay of zero or refuse to answer the question. These protest responses are thrown out because they don’t represent true economic valuations—at least if the subject indicates that his zero bid is a protest, not a true zero valuation. But surely the problem lies more with having to shove everything into contingent valuation mode than with those pesky moral environmentalists who refuse to state a willingness to pay. These people may really not think of an amenity in monetary terms, and if you force them to reconsider their answer (at least you can warn those who state zero that their result will be either thrown out or taken to mean a valuation of literally zero), the number they come up with may be meaningless. I may value sex with a loving partner very highly, but if the question is how much I’d be willing to pay for it, we’re now measuring a different sort of experience entirely.
With these infinite valuers, or even just with people who don’t think of the environment as something monetizable, cost–benefit analysis, in demanding some (finite) number for the value of the environmental amenity, is asking the impossible. It wants to use a number that, for some of the affected population, just doesn’t exist because for these people the very question commits a category error.
Second, sometimes cost–benefit analysis may ask a question that isn’t incoherent but may not have an objective answer, even in principle. Damming the Grand Canyon isn’t a moral issue for everyone: some people have no problem treating the services of the Grand Canyon as amenities that can be bought and sold. So we’re not being incoherent in looking for these people’s valuation. The problem here is “merely” that these people may not know, and may never know, their true valuation. (In a metaphysical sense, perhaps their true valuation may not even exist.)
One’s true preferences may or may not be a brooding omnipresence. I suppose, if one could dissect my brain, one could reconstruct my complete utility function. But that doesn’t mean that, at this moment, I know what it is. Discovering my preferences is a trial-and-error learning process. I spend $5 on pizza one night and discover that, in this context (perhaps after having eaten $5 worth of pizza last night), I don’t like it as much as $5 worth of burritos. I suddenly encounter a new and unfamiliar type of cuisine and, by trying an appetizing-looking dish for the first time, discover (or create?) a new preference.
Over time, I may get a better and better sense of what fulfills my preferences, provided my preferences don’t change faster than I learn. But I only learn this because I can experience choosing options and foregoing other options in this pizza–burrito world. At any moment, given my extensive burrito experience, I might be able to tell you the maximum amount I would pay for a burrito at that moment. But because no one’s monetary contribution has ever been pivotal in preserving the Grand Canyon—not even once—and because few have even thought about the issue, there’s no way they can give a truthful answer.
Small wonder, then, that people report essentially the same willingness to pay when 2000, 20,000, or 200,000 birds are on the line: they’re reporting a number that gives them a “warm glow,” not their actual valuation of that number of birds. Or that survey respondents report an average of about $84 to prevent oil spills off the coast of Alaska—but that when they’re asked about their total willingness to pay for a range of social programs and then asked to identify how much of that amount they’d pay for environmental protection . . . and how much of that they’d pay to protect wilderness areas . . . and how much of that they’d pay to prevent human-caused problems . . . and how much of that they’d pay to prevent marine oil spills . . . and how much of that they’d pay to prevent marine oil spills in Alaska, the average response is $0.29.
It’s not that some framings give you the correct answer while others give you a biased answer. Here, everything is framing. There is no picture inside. Even if a nonuse value of the Grand Canyon exists for some people, economists haven’t figured out a theoretically defensible way to discover it.
Third, a number may exist in principle, but in practice, economics, mathematics, and the limitations of available data mean that we’re not likely to find it. In the Grand Canyon hypothetical, perhaps someone has discovered his true valuation through introspection, but why would he reveal it rather than be strategic or seek the moral approval of the interviewer? Of course, anyone interested in preserving the Grand Canyon would be well-advised to report a high willingness to pay because no one will be actually demanding his money.
Or suppose we estimate the cost of a particular environmental risk using the wages of workers exposed to different degrees of risk—we may find, for instance, that workers in a particular industry who bear extra cancer risks due to their on-the-job exposure demand higher wages. Or we may estimate the benefit of an environmental amenity by looking at the prices of houses that have that amenity to different degrees—suppose a house with a better ocean view sells for an extra $10,000. But $10,000 is only the benefit of the view to the marginal house purchaser. It’s not the benefit of the view to the inframarginal house purchasers, who may value the view at more than $10,000; nor is it the benefit of the view to those who didn’t buy the house, who may have not bought the house precisely because, while they like views, they wouldn’t pay $10,000 for one. Similarly with occupational health risks: an econometric wage study can tell us how the marginal (mobile, informed) worker values risk, but it doesn’t tell us about the inframarginal workers, who would tend to select the job because they’re less risk-averse, or those who would never even think about working in that industry precisely because they hate risk.
Cost–benefit analysts trying to back out these costs and benefits—which, being subjective, only exist in people’s heads—from other data are essentially rediscovering the socialist calculation problem. Doing good cost–benefit analysis is easier than planning an entire economy, but the same knowledge problem that dooms the latter is bad news for the former as well. While cost–benefit analysis may look like rationality, perhaps it is merely rationalism.