7.3. Luken and Fraas' Review
In their paper "The US Regulatory Analysis Framework: A Review," Luken and Fraas offer a historical perspective on the evolution of regulatory impact analysis and its current state with special emphasis on EPA. The centerpiece of the paper is a set of case studies of the analyses underlying three regulations adopted by EPA.
Through Executive Orders and guidance from the Office of Management and Budget (OMB), the requirements for regulatory impact analyses have developed beginning in 1971 only a year after creation of EPA. President Reagan’s Executive Order 12291 of 1981 and implementing OMB guidance documents receive the most attention because they were current at the time the subject document was written, i.e., President Clinton’s Executive Order did not yet exist. Executive Order 12291 set a clear economic target: “Regulatory objectives shall be chosen to maximize the net benefits to society.” Luken and Fraas embrace this target; they later use net benefits as the benchmark to measure the efficacy of the analysis process.
However, Executive Orders and OMB cannot permit EPA to base regulatory decisions on economic analysis when the underlying statutes prescribe differently. Luken and Fraas tabulate the seven statutes and their subsections to indicate where analysis is permitted and precluded. There is some ambiguity.
The paper turns to a consideration of EPA’s rulemaking during the period 1981-1992. EPA promulgated 60 major rules during that period and prepared regulatory impact statements for most of them. Luken and Fraas provide a qualitative assessment of some of them:
- "The more thorough RIAs, i.e. those that included pollutant quantities, exposures, and adverse effects, were the ones that were prepared for air regulations because of the greater availability of data and analytical techniques and, more importantly, the interest of the responsible office. In no case does it appear that EPA chose the regulatory alternative that maximized net benefits. Other factors, such as statutory requirements or concern with potential economic impacts, resulted in the selection of another option." (p. 100)
Luken and Fraas next offer examples of three “correctly prepared” RIAs and comment on their impact. First among them is the phase-down of lead in gasoline. EPA had reduced the use of lead in gasoline in the 1970s and revisited the issue in the early 1980s. An RIA compared two options: a further reduction in lead and its elimination. The former offered net benefits of $5.9 billion and the latter, $6.7 billion. Both numbers exclude benefits of reducing adult blood pressure of $27.5 billion and $30 billion, respectively. However, the investigators express some skepticism about some of the purported benefits, specifically a reduction in vehicle maintenance costs and avoided crop damage evaluated at subsidized agricultural prices rather than shadow values.
EPA chose the low lead option. It offered a net benefit relative to the status quo but less than the no lead option “because of the concern that eliminating lead altogether would have resulted in excessive valve wear, primarily in farm tractors.”
The second RIA case study concerns asbestos. EPA proposed in 1984 to phase out asbestos over a ten-year period and produced a draft RIA that found significant net benefits resulting from avoided cancers. However, that RIA’s granularity was on asbestos product groups rather than individual products. A finer grain RIA prepared in 1986 found banning some products avoided cancers at a cost exceeding $100 million each. Subsequent revisions also suggested a complete phase-out would not maximize net benefits.
In 1989, EPA adopted a rule phasing out asbestos use in most categories but not in a group of high-value uses in industrial, space and military applications where the cost per avoided cancer tended to be very high. Luken and Fraas comment on the final rule:
- "EPA did not choose the option with the greatest net social benefits. If it had, it would not have banned some general commerce product categories, where the estimated health risks appeared to be small and the cost per case avoided high. The apparent explanation for this choice was a policy decision to ban, to the extent feasible, all products in general commerce and to exempt certain critical industrial and military applications." (p. 104)
The third case study deals with radionuclides in drinking water. Luken and Fraas describe the legislative history thusly:
- "Given its frustration with the failure of EPA to regulate these contaminants, Congress passed the 1986 Amendments to the Safe Drinking Water Act. … The Amendments reduced EPA’s discretionary authority under the 1976 [Safe Drinking Water] Act by replacing the language that allowed EPA to consider benefits and costs in setting MCLs [maximum contaminant levels]. EPA is now required to set the maximum contaminant level (MCL) for each contaminant as close to the MCLG [maximum contaminant level goals] as is technically feasible." (p. 105)
EPA developed a draft RIA in 1988 that proposed MCLs for five radionuclides. In 1991, EPA completed a final RIA with different control levels. Relative to the draft standards, the final standards were relaxed but the average cost per avoided cancer varied from $2.3m to $275m over the five radionuclides. Luken and Fraas, op. cit., p. 106. Luken and Fraas state:
- "The cost of meeting the most stringent [draft] standards for these three radionuclides would have been $520m annually compared to the $90m cost of the proposed [final] standard, and the incremental reduction in excess cancer cases would have been very small (2.5 cases [annually]). For the regulation to have had net benefits, EPA would have had not to regulate … [the three radionuclides] and … [a fourth one]. … The rationale for retaining fairly stringent, but not the most stringent, standards … was to ensure that no risk … exceeded a maximum individual risk of 2 x 10-4…." (p. 106)
However, regarding the fifth radionuclide which caused nearly all the baseline cancers, the regulation was too lax:
- "Only the proposed standard for radon would result in positive net benefits, and for that contaminant, EPA could have justified, on economic grounds, … [a] more stringent standard … The EPA rationale for a less stringent radon standard was driven by monitoring considerations." (p. 106)
- Most authorizing statutes limit or preclude the use of economic analysis. The Safe Drinking Water Act case study is an example.
- “… other policy objectives prevent the choice of the most efficient alternative.” Luken and Fraas, op. cit., p. 106. To some extent, this is tautological but they suggest that political considerations were at work in the inefficient lead in gasoline choice; the politically important agricultural constituency would have born the cost of increased wear on engine valves in older tractors.
- Many RIAs were deficient. Some contained too few policy options. Many misestimated benefits. Others incorrectly monetized benefits or didn’t do so at all.
- The RIA requirement does not apply to minor programs – typically those with costs under $100 million annually – nor does it apply to some other EPA activities. However, this objection seems only tangentially related to the question of whether RIAs are assuring that regulation maximizes net benefits.
Even though RIAs are failing to push society to the margin where benefits equal costs, Luken and Fraas find virtue in them. “First, regulatory-impact analysis has resulted in numerous cases of EPA selecting more efficient regulatory alternatives than it would otherwise have.” Luken and Fraas, op. cit., p. 108. While very appealing, we can’t know how the unstated alternative would have performed. They offer several examples that appear to demonstrate that economic analysis has led regulation toward more efficient options. Further, RIAs seem to be improving over time. They offer several examples. Finally, numerous improvements in techniques have resulted from the RIA requirement. They cite the development of surrogate markets and surveys. Some methodological purists might dissent that this is progress. Refinements in hedonic techniques might have made a better example. However, this claim closely matches the second, i.e., RIAs are improving over time.