an industry. It is not always made clear, however, from what settings the estimates were drawn. There may often be substantial variability in compliance costs depending on the characteristics of the regulated entity, including its scale and the age of its plants and equipment. For example, a plant may be so old that it would be replaced anyway, and the newer plant’s design might already incorporate a technology that is consistent with the proposed rule. In other cases, the plant might be relatively new and at a much earlier point in the company’s capital replacement cycle. In still other cases, if the cost of compliance is too great, facilities may be closed. It is also possible that, unknown to the EPA, particular facilities may have already been targeted for closure by company management even without the arrival of the new environmental requirements. A lack of such information is a common problem in the area of regulation and adds to the total uncertainty. EPA typically conducts surveys of facilities to determine the different types of technologies that are in place (EPA, 1995, 2004b). Cost estimates are usually based on estimates of the changes that would be required in the different types of facilities to comply with the standards.
EPA’s decision documents rarely present a range of costs that represents the uncertainty in estimates of engineering costs,15 and often it is not clear what assumptions underlay the computation of those cost estimates. For example, the summary of a 2000 regulatory impact analysis for arsenic in drinking water included tables listing the monetized health benefits from avoided cases of bladder and lung cancers and containing estimated compliance costs (EPA, 2000b). The table with the monetized benefits contains lower and upper estimates of benefits, which were based on the lower and the upper estimates of bladder cases avoided. No estimates that took sources of uncertainty other than human health risks into account are displayed. Estimates of the costs were provided in the summary table for two discount rates (3 percent and 7 percent), for two different plant categories, and for four different maximum contaminant levels. No analysis of other factors was displayed. As a result, the estimates did not reflect the overall variability in the cost of complying with the rule. A first step in dealing with this source of uncertainty would have been greater transparency in how the estimates were derived.
In a report that includes separate evaluations by different authors of three of EPA’s regulatory impact analyses, Harrington et al. (2009) also cited a lack of consideration of the uncertainty in many parameters that affected the regulatory impact analysis.
The use of engineering models for estimating costs raises a number of other technical issues as well. Conceptually, the relevant compliance
15 Engineering costs include the costs of purchasing, installing, and operating the technologies to meet a standard.