On Balance: A New Rationale for Not Including Certain Impacts in Benefit-Cost Analysis
This post is a summary of a paper I’ve written called “What’s in, what’s out? Towards a rigorous definition of the boundaries of benefit-cost analysis,” forthcoming in Economics and Philosophy. Students are typically told that benefit-cost analysis is an application of the potential Pareto criterion, which defines net benefit as the difference between the willingness to pay of winners for their gains from a policy and the willingness to accept of losers for their losses. If the difference is positive, the policy is a potential Pareto improvement, and we say that it generates positive net benefits. Economic philosophers have presented many objections to this definition, but none of these objections refutes the basic logic.
The implication of the potential Pareto criterion is that every policy impact for which individuals can express a willingness to pay should be included in a benefit-cost calculation. Otherwise, a full tally of the compensation the losers would require and the benefits that make the winners willing to provide compensation, won’t have been made. At the same time, however, the consensus of scholars and practitioners seems to be that some things for which there is willingness to pay should not be included: distributional concerns, rights, moral sentiments, and the like. For good measure I have confirmed, using contingent valuation, that people are willing to pay for these kinds of impacts.
The executive orders that require benefit-cost analysis for federal regulations (for example, Executive Orders 12291 and 12866), guidance documents that various governments have generated for standardization of benefit-cost analysis (for example, Circular A-4), and benefit-cost analysis textbooks (for example, the magisterial Boardman et al textbook) all reflect this consensus.
One possibility for resolving the apparent conflict between the principle of including all willingness to pay and the recommendation to exclude some willingness to pay would be to embrace the potential Pareto criterion definition and agree that rights, equity, and moral sentiments should be monetized and included in benefit-cost analysis. I’m not a fan of that approach, for reasons that will become clear. Instead, I propose a definition of benefit-cost analysis, based on the concept of consumer sovereignty.
The doctrine of consumer sovereignty is that individuals should be the arbiters of what is produced and consumed, based on their preferences as expressed in their willingness to pay. Why do advocates of the doctrine feel that consumers should have the power to decide? Because it should be none of a public decision maker’s business what people value, or why. Other than the fact that individuals value them, there are no other reasons why a decision maker should be concerned with the things people value. Setting aside the inequity of some having more purchasing power than others, resources should flow to the goods people value most, and those goods should flow to those who value them most. I agree with this spirit of this doctrine, and, along with John Stuart Mill, in “On Liberty,” I think that it is fundamental to the governance of a liberal society. However, what isn’t discussed is whether we want to apply this doctrine to everything for which people have willingness to pay. Are there some things for which we think that the reasons people have for valuing them, as well as the reasons those things matter to society, should be the business of a public decision maker? That is the question here.
To see how this question arises in benefit-cost analysis, consider how the doctrine of consumer sovereignty can be extended to benefit-cost analysis. It would say that policy decisions should be based on individuals’ preferences, as expressed in their willingness to pay for the impacts of policies. A policy generates net benefits if individuals, through their willingness to pay, tell us it generates net benefits. There is nothing more to it. Benefit-cost analysis supports the doctrine of consumer sovereignty because it doesn’t tell us why individuals value what they gain or lose from a policy, and it doesn’t tell us anything about why the impacts of the policy might matter to society over and above individuals’ preferences. Willingness to pay, and hence benefit-cost analysis, does not contain any of that information.
The question is, do we want all policy impacts to be determined by consumer sovereignty? If a policy affects access to smart phones – by reducing a tariff for example – is it any of a policy maker’s business why people have willingness to pay for those smart phones? I would say no.
By contrast, if a policy affects the rights of some group, is it really none of a policy maker’s business why members of society have willingness to pay for that group’s rights? And is the fact that people have preferences over other people’s rights the only reason those rights matter? I elicited the willingness to pay of an online convenience sample for the right of transgender individuals to serve in the military. A non-trivial subset had negative willingness to pay. If their aggregate willingness to pay outweighed the aggregate willingness to pay of those who support transgender rights, would that mean that granting those rights would generate negative net benefits to society? That can’t be right.
What is the fundamental problem with valuing rights using willingness to pay or accept? It is that in the case of rights, decision makers should not defer to the preferences (expressed as willingness to pay) of individuals, because why people have those preferences matters, and because rights matter for reasons over and above the preferences of individuals. Willingness to pay doesn’t convey the kind of information that decision makers need to know about rights.
Regarding the preferences of individuals, decision makers need to know the reasons behind those preferences to know what is being valued. Benefit-cost analysis does not contain any of that information, nor any of the reasons rights matter, over and above individual preferences. That is why, in my view, rights should not be monetized and included in benefit-cost analysis.
Formally, my definition of benefit-cost analysis is that it is the aggregate of individual willingness to pay for those policy impacts for which consumer sovereignty should govern. (And, if you prefer, willingness to accept in the case of losses.) The test of whether an impact should be governed by consumer sovereignty is two-fold. 1) It should not matter to society why individuals value the impact, and 2) there are no other reasons the impact should matter to society, over and above the fact that it matters to the individuals who receive it. If one considers the list of impacts that the consensus would exclude – distributional concerns, rights, moral sentiments, and the like – they are those that do not satisfy these criteria. There will be cases in which opinions could differ as to whether an impact should be included by my definition. For example, the argument could be made that a preference for the well-being of others, altruism, should be given the deference of consumer sovereignty. I would argue that it matters why individuals have altruistic preferences, inasmuch as in-group prejudice might be considered an un-meritorious source of altruism. Also, if willingness to pay for others to be made better off is included, it might be hard to make the case that willingness to pay for some groups to be made worse off should not also be included, and I think most people would agree that this would be an undesirable inclusion. Opinions could vary, and altruism might be a boundary case. But I believe that for the most part, my definition cleanly cleaves impacts into those that belong and those that don’t, and if adopted, I believe it would keep benefit-cost analysis on a firm foundation.