How Cost-Benefit Analysis Incorporates and Worsens Feasibility Analysis’s Flaws

David M. Driesen - Syracuse University College of Law

Posted in ,

Jonathan Masur and Eric Posner’s neglect of key normative arguments (discussed in my previous post)1 stems in part from a preoccupation with flaws in the agency practice of feasibility analysis. I agree with Masur and Posner’s characterization of that practice as less than wholly satisfactory and suggested as much in an article that they discuss extensively, Distributing the Costs of Environmental, Health, and Safety Protection: The Feasibility Principle, Cost-Benefit Analysis, and Regulatory Reform.2 But all of the significant flaws that they associate with feasibility analysis complicate cost-benefit analysis (CBA) as well. Moreover, CBA maximizes decision costs and the potential for ad hoc judgments that they attribute solely to feasibility analysis. A comparative analysis of the technical problems Masur and Posner identify with feasibility analysis follows to demonstrate this point.

I.  Clarity of Guidance

Masur and Posner assume that feasibility analysis provides “no theoretical way” to determine the correct balance between employment and health/safety and that CBA does.3 The feasibility principle demands maximization of environmental and health benefits up to the point where plant closings begin to occur. Masur and Posner may not like this criterion, but the criterion is quite clear in principle about the level of stringency required in the many cases where contemplated technologies do not lead to any shutdowns of facilities whatsoever.4 They make this clarity appear to disappear by selecting cases for study in which agencies predict some plant closures. This selection works well as a method for highlighting the feasibility principle’s weaknesses in hard cases, thereby facilitating a normative debate, but it slights the feasibility principle’s capacity to resolve many cases with relative ease.

To make the strongest possible case for CBA’s relative clarity, assume that Masur and Posner adopt the efficiency criterion: costs should equal benefits at the margin. This criterion, while not clear in practice (as we shall see), is very clear in theory. It achieves this clarity by leaving out all consideration of distributional equity—in other words, through very significant neglect of important aspects of overall well-being, Masur and Posner’s normative touchstone. To achieve a comparable degree of precision, one would have to translate the “widespread plant shutdowns” into a similar mathematical expression—for example, permitting no more than 10 percent of plants to shutdown. While Masur and Posner condemn this rule as arbitrary, it does not seem any more arbitrary than decisions establishing a speed limit at fifty-five miles per hour instead of sixty-five miles per hour. Establishing clear rules through legislative decisionmaking probably requires some fairly arbitrary judgments. Any clear rule will fit some cases poorly, as illustrated by the poor fit between the rule that costs should not exceed benefits and my previous post’s example of high aggregate costs distributed to generate minor price rises in large numbers of television sets. But if clear guidance is a paramount consideration, one can obtain that by refining, rather than abandoning, the feasibility principle.

Masur and Posner criticize the feasibility principle for failing to tell agencies how far to go. But their own examples demonstrate that CBA provides even less guidance. For example, an exposure limit of 1 µg/m3 for hexavalent chrome produces total costs of $570 million and a benefits range between $53 million and $1.382 billion. It is impossible to determine whether costs exceed benefits or not. The same is true for five of the six regulatory options that the Occupational Health and Safety Administration (OSHA) considered, because all five produce costs within the plausible range of benefits:

No normative criterion associated with CBA tells the regulatory agency whether to choose 0.5 ug/m3, 1 ug/m3, 5 ug/m3, 10 ug/m3, or 20 ug/m3 (which is 40 times as lax as the 0.5 limit).

OSHA tried to circumvent that difficulty by providing median net benefit numbers. But in three of the five cases those medians provide a range of net benefits between positive and negative, thus leaving a hapless OSHA, if its statute permitted it to follow the efficiency criterion, with a choice between a standard of 1 ug/m3 and limits ten or twenty times as lax. Even Masur and Posner concede that CBA only narrows the range to a choice between levels of 1 ug/m3 and ten times that amount of exposure.

While Masur and Posner bury this fact, OSHA’s completed feasibility analysis gave OSHA clear guidance about which level to choose under the feasibility principle. OSHA concluded that, at levels more stringent than 5 ug/m3, its regulation would destroy at least one industry, but at 5 ug/m3 few if any plant closures would occur. The feasibility principle therefore pointed rather clearly to regulation at 5 ug/m3.

Similarly, the combined costs of the Clean Air Act and Clean Water Act rules that Masur and Posner rely on to justify their technical analysis fell within the range of plausible benefits for all three options.5 No criterion associated with CBA could tell the regulator which option to choose without making a choice among plausible benefits estimates or accepting some kind of dubious averaging procedure and then ignoring the nonquantified benefits.

Arguably, the feasibility principle provided clear guidance in this case, as EPA predicted that the most stringent option would close only 9 out of 158 mills regulated—hardly a widespread shutdown. If so, then the agency shirked its duty when it chose a less stringent option, closing only two mills. A shirking of duty does not show that the duty was unclear.

Nevertheless, I agree that the concept of widespread plant closure has some ambiguity that will often matter in the minority of cases in which large numbers of plant closures are predicted. Of course, if one specified a percentage of plant closures in advance, then one would have clear guidance available for those situations.

While Masur and Posner are right that the feasibility principle provides only ambiguous guidance in some cases, they fail to recognize that, even in their chosen examples, narrow CBA provides even less guidance.  Masur and Posner miss CBA’s inability to provide clear guidance because they delve into the unattractive details of how agencies estimate the numbers for plant closures while applying no scrutiny at all to how they arrive at their estimates of costs and benefits, making the numbers in CBA appear magically from nowhere. They acknowledge CBA’s “ambiguities” in the abstract but blithely assume that agencies keeping in mind the overall goal promoting public well-being can somehow “resolve[]” these.6 Overall well-being does nothing at all to resolve the risk assessment problems generating potentially huge variability in benefits estimates and precious little to resolve other specific problems that their chosen examples illustrate. One would think that people with varying normative commitments might have different views of overall well-being, even if they all accept Matthew Adler and Eric Posner’s description of it.7

II. Generating Numbers: A Comparative Approach

Both feasibility analysis and CBA require the generation of numbers—namely, the number of predicted plant shutdowns in feasibility analysis and the dollar value of costs and benefits in CBA. Masur and Posner correctly point out that, in principle, feasibility analysis’s results may depend on industry definition, choices about technology forcing, and regulation’s path. These same variables, however, play a large role in CBA as well. In fact, CBA provides more opportunities for ad hoc judgment than feasibility analysis.

A.     Industry Definition

Masur and Posner point out that agencies must define the industry in order to carry out a feasibility analysis. And an analyst can subdivide any industry into subcategories. Because the definition of the industry can influence conclusions about whether an industry faces widespread plant shutdowns, agencies “tinker[] with industry classifications on an ad hoc basis.”8 The court reviewing the hexavalent chrome rule that they use to illustrate this problem held that OSHA’s industry classification was not arbitrary, partly because of a consistent practice of setting a uniform permitted exposure level for the entire regulated universe as a whole rather than subdividing the industry.9 Still, Masur and Posner are correct that the agency has discretion in defining an industry, such that ad hoc industry definition can occur.

But the problem of industry classification influencing results and therefore inducing tinkering exists with CBA as well. A good example of this problem comes from the Fifth Circuit’s decision overturning EPA’s phase-out of asbestos in Corrosion Proof Fittings v EPA.10 In its introduction to the case, the court explained that the rule would save either 202 or 148 lives at a cost of $450 to $800 million, about $2 to $4 million per life, putting it within the range most CBA proponents find acceptable.11 Yet, in explaining why the rule failed to satisfy the substantial evidence standard of the Toxic Substances Control Act (TSCA), the court accused EPA of spending $43 to $76 million per life saved.12 What happened? The introduction referred to the entire industry making asbestos products,13 while the passage claiming excessive costs focused on a subcategory of that industry, the manufacturers of asbestos pipe.14 In other words, the results of CBA hinge upon the definition of the industry under analysis. While Corrosion Proof Fittings involved ad hoc judicial tinkering, agencies can do the same under CBA.

B.     Existing versus Future Technology

Similarly, the problem of having to decide whether to base a rule on existing technology or on technology not yet fully developed arises for any analysis of cost, not just for feasibility analysis. The cost of meeting any level of environmental protection equals the cost of making the technological changes (broadly defined) needed to meet that level. A good example of the problem of CBA varying depending upon whether one embraces technology forcing or not comes from the CBA of climate disruption. Different analysts come up with widely varying conclusions about the costs of abating greenhouse gas emissions. Choices about how to treat the possibility of technological advancement constitute one of the most significant causes of these disparities in CBA’s results. Some analysts base their cost estimates on existing technologies or past experience, while others come to very different conclusions because they assume that abatement policies will produce technological advances that lower costs.

Masur and Posner point out that courts have placed a heavy burden on agencies trying to justify technology-forcing regulation, thereby making it difficult to use feasibility analysis to advance technology. There is no reason to expect CBA to help solve this problem. Indeed, by emphasizing the notion that all regulation must be cost-justified, CBA, if subject to judicial review, will likely exacerbate judicial tendencies to expect a better justification than agencies can produce for reliance on future technologies. It will no longer be enough to show that reasons exist to expect the technology to be technically feasible and not so expensive as to bankrupt anybody. Instead, the agency would have to show that it has a reasonable basis for estimating the precise cost, a difficult task with a technology not yet developed.

C.     Path Dependence and Temporal Inconsistency

Masur and Posner show that “path dependence” and “temporal inconsistency” cause feasibility analysis’s results to depend on agencies’ prior regulatory actions with respect to a particular industry.15 This means that a regulation’s acceptability might depend on when the agency chooses to promulgate it. In CBA, this problem usually becomes broader, as regulations’ acceptability can become dependent not just on the path of regulation for a particular industry, but on all regulation influencing the environmental conditions that the regulation addresses. A good example of CBA’s path dependence comes from the Clean Water Act, which aims to restore heavily damaged ecosystems through a program of regulating water intake from large industrial facilities and effluent. The water intake kills billions of fish and other aquatic organisms, thereby harming ecosystems. Suppose that EPA regulates water intake early in the statute’s life, when ecosystems are seriously degraded. The proposed regulation costs $100 million and, because a degraded ecosystem currently supports little aquatic life, saves only 5 million fish, each fish worth $10. This $50 million dollar benefit cannot justify the $100 million cost. So, CBA (or, more specifically, the criterion that costs must not exceed benefits) would prohibit regulation precisely because of ecosystem degradation, which one might otherwise treat as an indication of a need for aggressive regulation aimed at ecological recovery. Suppose now that EPA proposes the same $100 million regulation after twenty years of successful regulation of effluent. Thriving aquatic ecosystems now make regulation less important.  But the thriving ecosystem has boosted the commercial fish population so that water intake now kills 20 million fish, worth $200 million. Because the agency promulgates this regulation after other regulations, its benefits justify the cost. CBA proves not only path dependent and time inconsistent, but also, at times, utterly perverse from the standpoint of key environmental values.16

Moreover, this sort of path dependence invites ad hoc tinkering in the analysis itself. A good example comes from EPA’s recent regulation of mercury emissions from power plants. Because the technologies used to reduce mercury from power plants also reduce particulates, which are associated with tens of thousands of annual deaths, a promptly implemented mercury rule evaluated on its own would likely produce enormous benefit predictions.17 Because the Bush administration EPA chose to implement a rule aimed at particulate and other criteria pollutants before the mercury rule,18 its assessment of the mercury rule’s benefits counted only the incremental mercury benefits realized after the criteria pollutant rule was implemented.19 Hence, the agency, by manipulating the timing of the regulation, could manipulate the outcome of the CBA. The CBA of the mercury rule exhibits temporal inconsistency and path dependence,20 illustrating that yet another problem Masur and Posner imagine arising under feasibility has arisen under CBA.

Masur and Posner complain that agencies can use feasibility analysis in an ad hoc manner. The examples discussed above show that agencies can use CBA in an ad hoc manner as well. CBA multiplies the number of variables contained in the analysis, which multiplies the opportunities for ad hoc judgment. All analysis offers opportunities for ad hoc judgment, but feasibility analysis lessens the number of opportunities provided.

III. Decisionmaking Costs: CBA and Feasibility Compared

Masur and Posner blithely assure us that CBA “minimizes decision costs through the magic of quantification,” thereby suggesting that it has lower costs than feasibility analysis.21 But CBA requires analysis of technology and its costs, just as feasibility analysis does. And CBA requires very difficult quantification of environmental harms, something that feasibility analysis does not require. Because the outcome of CBA depends on the choice of which benefits to quantify and what values to attach to them, these variables regularly become matters of dispute between the Office of Management and Budget (OMB) and EPA, often leading to costly interagency debates and delays. If the cost of conducting and debating analysis is part of decisionmaking (and it is hard to see how it could not be), then CBA maximizes decision costs.

Perhaps Masur and Posner have in mind the costs of making decisions after the government has completed and agreed upon an analysis under the efficiency criterion, which, after all, takes the form of a mathematical equation. Even then, however, it remains hard to see how CBA “minimizes decision costs.”22 As Masur and Posner’s case studies illustrate, the agency must always decide upon the weight to be given nonquantifiable environmental benefits, as some significant benefits always defy quantification. If the CBA is scientifically honest, then the agency must also debate which points in the various overlapping quantified benefit ranges to choose. This problem only grows worse if one adopts the criterion Masur and Posner explicitly endorse—that costs should not exceed benefits—as many options have benefits exceeding costs.

By contrast, the feasibility principle makes many decisions easy once the analysis is complete, because many regulations produce no plant closures. Under those circumstances, agencies just choose the most stringent technological option. Of course, things get dicier, as Masur and Posner point out, when agencies predict some plant closures. But they have not begun to support the notion that choosing a point at which plant closures are widespread is harder than choosing which regulation maximizes net benefits when the wide range of benefit estimates and the nonquantifiables are considered.

Masur and Posner concede that it might make sense to eschew CBA if it exacerbates any agency tendencies to underregulate, but they suggest that we need a great deal of “empirical work” to overcome “one’s natural skepticism” about the idea that CBA constitutes a drag on regulation.23 In saying this, they fail to engage an enormous scholarly literature, including some by CBA proponents, showing that OMB has used CBA to slow and throttle rules in every administration and that the processes involved have killed off at least one entire regulatory program and slowed others down enormously.24 Do they have some empirical evidence to refute scholars’ assertions that, after a judicial decision demanding CBA of every option in a § 6 rulemaking under the TSCA, EPA gave up any substantial use of § 6—the principal regulatory authority EPA has for limiting the use of toxic substances?25 Do they seriously doubt the assertion that quantitative risk assessment, a procedure at the heart of CBA, doomed EPA’s pesticide program to a state of perpetual slow motion?26 Do they dispute leading scholars’ assertions that linking specific reductions of pollutants to specific results in the receiving medium, which CBA requires, has never worked well in any medium—land, air, or water?27 We do not know, because Masur and Posner have substituted their “natural skepticism” of the idea that a comprehensive quantitative analysis of all regulatory consequences might create serious burdens on regulatory programs for serious engagement with a consensus view of most of the country’s leading environmental law scholars. While my work has distinctively emphasized a normative justification for the feasibility principle, a large literature mostly preceding my work has supported feasibility analysis as necessary to avoid the well-known decisionmaking costs that CBA and risk assessment create.


CBA, not feasibility analysis, maximizes decisionmaking costs. While feasibility analysis demands some agency judgments, which can lead to ad hoc decisions, CBA requires the same sorts of judgments and provides even more opportunities for ad hoc decisions.


David M. Driesen is a University Professor at Syracuse University College of Law.

Copyright © 2011 University of Chicago Law Review

This Legal Workshop post is based on an article forthcoming in volume 35 of the Harvard Environmental Law Review.

  1. David M. Driesen, A Modest Normative Case for Feasible Regulation (Legal Workshop, Dec 13, 2010), online at (visited January 20, 2011), responding to Jonathan S. Masur and Eric A. Posner, Against Feasibility Analysis, 77 U Chi L Rev 657 (2010).
  2. David M. Driesen, Distributing the Costs of Environmental, Health, and Safety Protection: The Feasibility Principle, Cost-Benefit Analysis, and Regulatory Reform, 32 BC Envir Aff L Rev 1, 19–22 (2005) (referring to the “vagaries of implementation” and suggesting that agencies have not consistently adhered to the feasibility principle).
  3. Masur and Posner, 77 U Chi L Rev at 705–06 (cited in note 1).
  4. See Driesen, 32 BC Envir Aff L Rev at 43 (cited in note 2) (pointing out that often agencies predict no plant closures).
  5. Masur and Posner, 77 U Chi L Rev at 670–74 (cited in note 1).
  6. Id at 705.
  7. See Matthew D. Adler and Eric A. Posner, New Foundations of Cost-Benefit Analysis 185 (Harvard 2006) (defining overall welfare as the satisfaction of self-interested preferences that survive idealization).
  8. Masur and Posner, 77 U Chi L Rev at 691 (cited in note 1).
  9. See Public Citizen Health Research v Department of Labor, 557 F3d 165, 182–84 (3d Cir 2009) (rejecting an environmentalist plea to subdivide the industry in part because of consistent use of uniform standards).
  10. 947 F2d 1201 (5th Cir 1991).
  11. Id at 1208.
  12. See id at 1219 (noting parenthetically that $128 to $227 million of contemplated compliance expenditures to save three lives implies $43 to $76 million per life saved).
  13. Id at 1207–08 (associating the $2 to $4 million per life saved figure with EPA’s “rule” phasing out “most asbestos-containing products”).
  14. Corrosion Proof Fittings, 947 F2d at 1219 (associating the $43 to $76 million per life saved figure with EPA’s “ban of asbestos pipe”).
  15. Masur and Posner, 77 U Chi L Rev at 696–97 (cited in note 1).
  16. See Douglas A. Kysar, Fish Tales, in Winston Harrington, Lisa Heinzerling, and Richard D. Morgenstern, eds, Reforming Regulatory Impact Analysis 190, 209 (RFF 2009).
  17. See Catherine A. O’Neill, The Mathematics of Mercury, in Harrington, Heinzerling, and Morgenstern, eds, Reforming Regulatory Impact Analysis 108, 115–16 (cited in note 15) (describing a promptly implemented mercury rule as generating particulate “co-benefits”).
  18. Id at 111 (describing the timing of the mercury rule).
  19. Id at 113 (stating that this approach allowed EPA to avoid attributing “a sizeable category of benefits” to the mercury rule).
  20. See Alan J. Krupnick, The CAMR: An Economist’s Perspective, in Harrington, Heinzerling, and Morgenstern, eds, Reforming Regulatory Impact Analysis 142, 144–45 (cited in note 15) (agreeing that the choice of timing influenced the baseline and therefore the estimates of costs and benefits).
  21. Masur and Posner, 77 U Chi L Rev at 700 (cited in note 1).
  22. Id.
  23. Id at 711.
  24. See Richard L. Revesz and Michael A. Livermore, Retaking Rationality: How Cost-Benefit Analysis Can Better Protect the Environment and Our Health 11, 151–61 (Oxford 2008) (reviewing CBA’s role in slowing, defeating, and weakening rules and concluding that it “generally serves an antiregulatory purpose,” but supporting it with reforms designed to overcome this problem).
  25. See David M. Driesen, Is Cost-Benefit Analysis Neutral?, 77 U Colo L Rev 335, 347 (2006) (pointing out that EPA has not banned a single chemical since the Fifth Circuit subjected such actions to a cost-benefit test); Thomas O. McGarity, Professor Sunstein’s Fuzzy Math, 90 Georgetown L J 2341, 2342­–43 (2002) (describing CBA as having “stymied” regulation under the TSCA and the Federal Insecticide, Fungicide, and Rodenticide Act). Consider also Richard D. Morgenstern, ed, Economic Analyses at EPA: Assessing Regulatory Impact 199 (RFF 1997) (describing the regulation of polychlorinated biphenyls (PCBs) as the only action EPA ever took under TSCA § 6 in the wake of the Corrosion Proof Fittings decision). Note, however, that PCBs were banned long before then.
  26. See Donald T. Hornstein, Lessons from Federal Pesticide Regulation on the Paradigms and Politics of Environmental Law Reform, 10 Yale J Reg 369, 437 (1993).
  27. See Oliver A. Houck, The Clean Water Act TMDL Program: Law, Policy, and Implementation 136, 165, 194–97 (Environmental Law Institute 2d ed 2002) (making this assertion and providing examples); Adam Babich, Too Much Science in Environmental Law, 28 Colum J Envir L 119, 133–35 (2003) (finding that “{t}he most common criticism of risk-based standards is that they do not work” and providing examples of where they have failed). Consider also Amy Sinden, In Defense of Absolutes: Combating the Politics of Power in Environmental Law, 90 Iowa L Rev 1405, 1487–88 (2005) (arguing that the strict effects-based approach in the Endangered Species Act produces results “closer to where we want to be” than a balancing approach would).

Post a Comment (all fields are required)

You must be logged in to post a comment.