At one level, all policy analysis starts with benefit-cost analysis: on what basis could one choose among option except their advantages (benefits) and their disadvantages (costs)? That makes it puzzling, at first blush, that benefit-cost analysis should be controversial; when it comes to environmental and safety regulation, benefit-cost is beloved of industry lobbyists and loathed by activists. (Benefit-cost studies are rarely applied to questions such as going to war or extending prison sentences.)
But the puzzle has a solution: the benefit-cost analysis practiced in the regulatory process — what I teach my students to call “forensic” benefit-cost — differs in three ways from ideal, or armchair, benefit-cost, as practiced by someone trying to figure out what course of action will best serve the public interest, conceived as the sum of gains and losses to individuals. Each of those differences constitutes a frank error, and all are enforced by the courts and by the Office of Information and Regulatory Affairs (OIRA) of the Office of Management and Budget.
1. Formal benefit cost analysis counts everyone’s gains and losses equally. But common sense and the principle of diminishing marginal utility agree that a dollar’s worth of gain is more valuable to someone with few dollars than it is with someone with many. Obviously, taking $1 each from 900,000 poor people to give $1 million to a hedge-fund billionaire doesn’t reflect a social gain, but a formal benefit-cost analysis will show that it does: after all, the net benefit is $100,000. Thus gains and losses should be adjusted by (at least) dividing each gain or loss by the income or wealth of the person bearing it, so that a $20 gain to a family with an income of $20,000 weighs as a heavily as a $10,000 gain to a family with an income of $1 million.
2. Formal benefit cost analysis draws artificial lines around the impacts of a program: impacts that are very indirect, or very distant in time, or highly uncertain, which ought of course to be adjusted to reflect those facts, are instead usually excluded from the analysis as “speculative,” which amounts to treating them as being certain not to take effect. Since the costs of a regulation are mostly easy to measure, this rule tends to exclude more benefits than costs, and is thus biased against the program under consideration.
3. The same is true of gains and losses with no obvious market valuation. After great struggle, the value of preventing an early death has been set at several million dollars. But any health damage short of death, or health gain other than reduced mortality, is usually valued only in terms of lost income and medical expense, rather than at its full “willingness-to-pay” value: the amount that the person who might gain or lose would be willing to pay to enjoy the gain or avoid the loss. To value an injury that leaves someone in permanent pain only in terms of lost wages and the cost of pain medication, or an injury that leaves someone in a wheelchair only in terms of the price of the wheelchair, is obviously wrong, but it’s also the current standard.
So here’s a modest proposal: the Congress should order the administration to commission a study by the National Research Council to establish a set of standards for regulatory benefit-cost analysis, which can then be written into the statutes that require such analysis before a regulation can be issued, and adopted by OMB for benefit-cost analysis done in other administrative contexts. The legislation establishing the study should tell the NRC to develop rules that embody distributional adjustments, Bayesian weighting of uncertain gains and losses, and willingness-to-pay evaluation of gains and losses that do not come with market prices attached.