Overall Conclusions and Recommendations
The committee considered a number of analytic approaches to evaluate the effects of the New Source Review (NSR) rule changes. As described in more detail below, each of the approaches has its strengths and weaknesses. If any future assessments of the effects of the NSR rule changes are to be made, the committee recommends that both empirical analysis (that is, of permitting data or investment activities) and modeling (that is, of sectoral responses to regulatory changes or air-quality effects of emission changes) be used. The committee concludes that anecdotal information is not a sufficient basis upon which to draw conclusions about the NSR rule changes. Use of anecdotal information by itself is insufficient because there is a wide variety of ways that an operator of a specific facility might respond to changes in the NSR rules, depending on such factors as the type of industry, economic conditions, and other regulations affecting the facility. In addition, it is difficult to ensure that a representative sample across facilities has been obtained and that the information provided is valid. Although anecdotal information could be used to identify the types of investment projects that are being discouraged by NSR requirements, firms could have an incentive to exaggerate such cases and argue for less stringent NSR rules.
Some qualitative information can play a role in future analyses. For example, the committee recommends that qualitative information be collected from firms about their perceptions of the status of the NSR program for specifying future econometric models. However, that information must be collected using formal interviewing protocols and should be collected prospectively on a large and representative sample to ensure that data will be useful. More generally, information from individuals and organizations can
be more formally elicited, through Delphi approaches, other types of formal expert elicitation protocols, and other processes that could take qualitative insights from area experts and translate them into useful information for quantitative analysis. If such information is not collected in a formal way on a representative population, it cannot be used as a basis for answering the questions in the committee’s charge.
The committee finds that developing an econometric structural or behavioral model of firm-level investment decisions is not feasible for evaluating the effects of the NSR rule changes. To incorporate all effects of the NSR rule changes appropriately, data would be required on investment projects that were carried out, that were considered but never carried out, and that were modified to avoid NSR. Because these last two effects are unobservable, they could be incorporated only by collecting anecdotal information. That approach is not favored by the committee and could not be used in any case for quantitative analyses.
One analytic approach considered by the committee was to analyze NSR permit data, which could be used to determine how NSR permitting activity changed as a result of the rule changes. However, as discussed in Chapter 3, current databases are inadequate for such an analysis. In addition, these data would capture only projects that were actually done, omitting investments that may have been forgone to avoid NSR. If the databases also included minor-construction permit information at the state level for investment projects that no longer needed NSR permits, such analyses might plausibly capture the major effects of the rule changes. However, given the current state of the data, a reduced-form econometric analysis would be needed to capture effects associated with NSR on investment decisions and emissions.
Data on investment activity (such as those contained in the Longitudinal Research Database of the Census Bureau) could theoretically be used to evaluate the rule changes with econometric methods, especially given different dates of implementation of the rule changes among states. Such analyses would not be possible for several years for a number of reasons: the data become available only after a 3-year lag, many states will not be affected by the rule changes until 2006 or later, and investment decisions can take years to be carried out. In addition, although the diversity in the timing of implementation of the NSR rule changes across states facilitates this reduced-form analysis, interpretation is complicated by the geographic clustering of some industry sectors and attainment-nonattainment status. Also, the NSR permit review process could differ across states in speed and predictability, with slow or uncertain permit approval in some states serving as a major discouragement to investment activity. The importance of delaying the investment process could be especially important for facilities other than the electricity-generating sector, whose firms are attempting to
respond to rapidly changing business conditions. The committee considers this use of econometric methods to be a promising analytic approach, and the requisite data collection should commence as soon as possible, as described in the recommendations presented later in this chapter. Because of the changes in the enforcement of NSR rules over time, as discussed in Chapter 2, an analysis of historical data on actual and allowable emissions could be linked to information from the time that NSR enforcement was especially stringent. General inferences could then be made about the effects of changes in the stringency of NSR rules on emissions in the future. However, given changing economic conditions and regulatory requirements, this approach is probably not sufficiently sensitive to capture the incremental effects of NSR. Such analyses could be useful prospectively as a complement to bottom-up sectoral simulation models.
Given current data availability, bottom-up sectoral models constitute the primary analytical approach that can be used at present. In principle, these sectoral models are most relevant to the committee’s task, because they can theoretically capture the geographic location of emissions changes, necessary for air quality and human health impact assessment. However, the uncertainties in geographically-resolved estimates are likely substantial. As noted later in this chapter, the committee can draw no conclusions from such modeling about the spatial redistribution of power generation emissions under the equipment replacement provision (ERP). In addition, these models have detailed data needs that are not being met and there are many assumptions embedded in the models that may not be realistic depictions of plant-level decision making. The committee considers this approach as an important companion to the econometric modeling described above, although sectoral models are most useful for identifying a range of possible outcomes and providing insights about general relationships rather than specific numerical forecasts. A significant strength of sectoral forecasting models is that they allow for parametric sensitivity analyses, and any future implementation of the Integrated Planning Model (IPM) or related models should carefully evaluate the sensitivity of model outputs to key inputs. Our IPM modeling efforts were only able to evaluate a small number of input parameter assumptions. Some of those assumptions, namely the assumed effect of the NSR rule changes on the rate of retrofits of selective catalytic reduction (SCR) and flue-gas desulfurization (FGD), had an important influence on the outputs of the model, and others (natural gas prices and investment costs for other generation sources) did not.
Regardless of the approach used to determine the effects of NSR rule changes on plant investment behavior and related emissions, atmospheric fate and transport modeling is required to link the emission changes with incremental changes in ambient concentrations. For health-effects assessments, a model is required that can capture detailed meteorologic factors
with appropriate geographic and temporal specificity (that is, to capture the transport and fate of emissions from a specific facility at an averaging time relevant for assessing health outcomes). Although such models are complex and contain uncertainties, the errors in estimating average exposure changes across the population associated with a given facility may be somewhat smaller than the errors in estimating the effects of a facility at a specific site, and the former calculation is more relevant to quantification of health effects. In either case, given the importance of secondary PM and O3 formation in this context, a fate and transport model must be able to model effects over substantial distances (hundreds to thousands of kilometers) and capture relevant aspects of atmospheric chemistry.
Thus, because of issues discussed in this report, the committee concludes the following:
The available information is not sufficient to quantify with a reasonable degree of certainty the potential effects of the NSR rule changes on emissions, human health, facility operating efficiency (including energy efficiency), or pollution prevention and pollution control. Modeling analysis provided insights (presented later in this chapter) into the potential effects on national emissions from the electric-power industry.
A combination of empirical analysis and modeling would be necessary to quantify the effects of the NSR rule changes and associated uncertainties. Anecdotal information by itself is insufficient to evaluate the changes in performance by a broad range of facilities after the NSR rule changes have been implemented and the effects that might occur as a result of those changes.
To carry out the recommended approaches, long-term collection of data and improved modeling techniques will be required. Our methodology and recommendations about necessary data and information and the development of better research methods are as important as the evaluations of this report regarding the 2002 and 2003 rules. Specific recommendations are presented later in this chapter.
POTENTIAL EFFECTS OF THE NEW SOURCE REVIEW RULE CHANGES
Numerous dimensions of the NSR rule changes were considered by the committee in assessing the effects of the 2002 rule changes and the ERP change promulgated in 2003: the effects on multiple air pollutants regulated under the NSR program—including sulfur dioxide (SO2), nitrogen oxides (NOx), particulate matter (PM), volatile organic compounds (VOCs), and
carbon monoxide (CO)—and exposures to the pollutants and their health effects; the effects on electricity-generating facilities and other industrial sectors; and the effects on pollution-control technology and facility efficiency.
The electricity-generating sector was a primary focus of many of our assessments. That sector, especially coal-fired power plants, dominates national emissions of both SO2 and NOx compared with other large stationary sources. Those two pollutants are important contributors to concentrations of ambient PM with an aerodynamic diameter equal to or less than 2.5 µm (PM2.5) and NOx is an important precursor to ozone (O3), the air pollutants of principal concern from a health perspective. Power plants contribute substantially to point-source PM2.5 emissions as well, although primary PM emissions contribute less to ambient PM2.5 than do precursor emissions of SO2 and NOx. In the committee’s evaluation of modification permits from 1997 to 2002, the largest share of permits for and emissions of most pollutants was typically from the industrial class that includes electric, gas, and sanitary services.1
The ERP was the primary rule change that could have influenced the electricity-generating sector because several of the 2002 rule changes had applied to electricity-generating sources since the 1992 WEPCO rule (see Chapter 2). However, the ERP had not been implemented because of a judicial stay, and in March 2006 the rule was struck down by the D.C. Circuit Court. Therefore, there is no direct empirical evidence of the effect of the NSR rule changes on the electricity-generating sector. Thus, a bottom-up (engineering economy) simulation model was used to consider the potential effects of the ERP on emissions, facility efficiency, and other key dimensions of the committee’s charge on the basis of several key assumptions.2 The committee used the IPM to evaluate the possible effects of the ERP on emissions of SO2 and NOx, relative to previous U.S. Environmental Protection Agency (EPA) policy of prerevision NSR. Although IPM does not include the other air pollutants affected by NSR (such as VOCs, CO, and primary PM), SO2 and NOx are probably the power-plant pollutants that would contribute most to air-quality issues and related health effects.
We caution that IPM or similar models cannot be used as the sole basis for predicting the effects of the NSR rule changes on electricity-generating facility emissions. A model like IPM aggregates multiple generation facilities into agglomerated model plants. Hence, it cannot accurately simulate variations in effects at the level of the individual generating unit, so it is
insufficient for estimating local changes in emissions that could affect public health. At best, IPM is a tool for assessing effects of national, or perhaps regional, patterns of emissions, which certainly are important to public health but can hide substantial local variations in effects. Like all power-sector models, there is uncertainty in the estimates from IPM even for assessing broad patterns. Although there are many possible sources of uncertainty in the IPM inputs and structure, two key sources are briefly mentioned here: firms make decisions under conditions of uncertainty, but the model assumes perfect foresight—an unrealistic premise—and it assumes the existence of perfectly competitive markets in which owners of electricity-generating facilities act as profit maximizers that react efficiently to the economic incentives resulting from electricity markets and environmental policies.
Because IPM is not structured to simulate the effects of NSR rule changes directly, some input-parameter assumptions were required to attempt to capture these effects. In general, IPM requires assumptions regarding the representation of decision making in the industry; values of important parameters for fuel costs, plant efficiency, and performance; and relevant environmental policies and enforcement actions. Several assumptions have important implications for the results of this analysis. The key input parameters reflecting this regulatory change are highly uncertain and require subjective judgments. For example, the model must use percentage of capacity scrubbed annually as a proxy for the effect of NSR changes, but there is no quantitive basis for making that prediction. Moreover, the factual assumptions that go into the model represent projections, with varied reliability, of the future.
Unlike EPA’s modeling of the ERP in 2003 (EPA 2003c), the runs that we commissioned considered the possibility that EPA’s prior approach to routine maintenance would have compelled a significantly greater amount of coal-fired capacity than the ERP to retrofit controls, repower, or retire. That could occur if aggressive NSR enforcement strategies under the prerevision NSR approach resulted in substantial retrofit activities either through settlements or by making it clear to the regulated community that the courts would uphold the approach taken by EPA in pursuing those strategies. Because it was unknown how the prerevision law would have been interpreted in the future, given a split in court decisions, there was no firm baseline against which to evaluate the impact of the ERP. Moreover, it is not possible to judge what scenario would most likely result from NSR policy without the ERP. For instance, the model does not include decisions at the level of the generating unit about projects that might have triggered NSR under prerevision policy.
The committee also considered the impact of the ERP both under pre-2005 regulations (Title IV, enacted as part of the 1990 Clean Air Act
Amendments, and the NOx state implementation plan (SIP) call of 1998)3 and under the rules promulgated in 2005—the Clean Air Interstate Rule (CAIR), the Clean Air Mercury Rule, and best available retrofit technology under the visibility-impairment rule. CAIR, for our purposes, is the most important of the 2005 rules. A set of sensitivity analyses considered the effect of lower costs of noncoal energy sources on the conclusions, and another sensitivity run defined the lowest-cost means of achieving the national emission changes that would occur if essentially all coal capacity without FGD for control of SO2 emissions and SCR for control of NOx emissions were retrofitted with best available control technology (BACT). FGD and SCR provide the emission reduction generally required by NSR.4 Additional sensitivity analyses are recommended by the committee that it could not implement during its study, given budget and time constraints (see Chapter 6).
The committee considered both an NSR-avoidance assumption regarding prerevision NSR compliance (the basis of the previous EPA analysis), as well as assumptions that the prerevision NSR rules would lead to retrofits of pollution-control equipment that would have otherwise not taken place. For the “avoid” assumption, the committee relied on the previous EPA runs, concluding that re-running IPM would not alter the model result that the ERP would have no appreciable effect on national emissions. The IPM runs requested by the committee were used to analyze the assumption in which the prerevision NSR rules would have resulted in significantly more retrofits than otherwise. Under this assumption, the potential effects of the ERP on national emissions from electricity-generating facilities depend on whether CAIR is assumed to be in place or not. The IPM results suggest the following conclusions if indeed the prerevision NSR rules would have compelled significantly more retrofits:
For SO2, without implementation of CAIR, the ERP would be expected to result in a moderate decrease in emissions in the first 6 years or so (compared with prerevision NSR) followed by a period of relatively little change in the next 6 years or so. An initial decrease is expected under these assumptions because the ERP would substantially increase the value of the Title IV SO2 emission allowances making overcomplinace and banking in early years more attractive. However, after the first 12 years, if it is assumed that prerevision NSR rules would have caused all coal-fired electricity-generating facilities to add emission controls, the ERP would be
expected to cause a relative increase (perhaps substantial) in emissions from the electricity-generating sector compared with the prerevison NSR case (see Chapter 6).
If CAIR is assumed to be implemented, the model estimates minimal differences in total SO2 emissions between the prerevision rules and the ERP, even if all electricity-generating facilities would have added emission controls under the prerevision rules.
For NOx, without implementation of CAIR, the ERP rule changes would be expected to cause an increase in emissions after the first few years; this increase is larger when the prerevision NSR is assumed to compel greater amounts of emission controls.
If CAIR is implemented, the model estimates a minimal change in NOx emissions; although after the first 12 years or so, the ERP would be expected to cause a moderate increase in emissions if all facilities would otherwise have added emission controls under prerevision NSR rules.
Although the quantitative conclusions depend on the input parameters and constraints of the model, the qualitative conclusions are logical and probably robust. In the presence of tighter emission caps, the effect of the NSR rule changes is reduced in part because an emission increase at one facility would be offset by a decrease elsewhere. Unless enough retrofit-repower-retire activity would have occurred under the prerevision NSR rules to reduce national emissions below the emission cap, prerevision NSR policy would tend to increase the costs of control without greatly influencing national emissions. That conclusion is based in part on the structure of IPM. IPM by definition is based on economic optimization, so if a total emission cap is the only environmental constraint, the model will find the lowest-cost solution to yield that level. Imposing additional constraints can only increase the cost, given the structure of IPM. In summary, the IPM runs suggest that the promulgation of CAIR, after the establishment of the NSR rule changes, would reduce the chance that the ERP would have adverse effects on national emissions from the electricity-generating sector. However, because of the substantial uncertainties in the IPM results, no firm conclusions can be drawn about the actual effects of the revised NSR rules. (Note that CAIR is subject to judicial review.)
Changes in IPM assumptions concerning natural-gas costs and the investment cost of new renewable and integrated gasification combined-cycle generation do not change our assessment of the potential effects of the ERP on emissions from the electricity-generating sector.
The committee also compared the cost of achieving national emission reductions through aggressive implementation of the prerevision NSR approach with the cost of achieving the same SO2 and NOx reductions by lowering the cap on total allowable emissions that is specified in CAIR
and making the cap national in scope. (NSR has local objectives as well, as mentioned in Chapter 2, so this comparison, which is limited to national emission reduction, should not be taken as attempting an overall assessment of NSR.) The IPM analysis indicates that even when the prerevision NSR approach lowers emissions below the CAIR caps, the implementation of lower caps could attain the same emission national reductions as the prerevision NSR rule at about one-third the cost or less. That is due largely to the fact that the retrofit-repower-retire scenario used in the IPM runs to capture the effects of the prerevision NSR tends to be a less-efficient means of achieving national emission reductions than market-based approaches. In the lower emission cap scenario, more low-sulfur coal and natural gas were used and fewer retrofits were made relative to the retrofit-repower-retire scenario. The estimated cost difference between a more traditional regulatory approach, such as NSR, and a market-based approach would be especially great if the application of the first type of approach were based on which plants happen to first be subject to NSR, not on first focusing on the plants that are the most cost-effective to control. Therefore, if lower national emissions of those pollutants are desired, implementation of emission caps lower than those of CAIR is a more cost-effective means of attaining national emission goals than the prerevision NSR rule. In addition, a market-based approach would give sources an incentive to reduce their emissions sooner than a more traditional regulatory approach, such as NSR, thereby reducing emissions more expeditiously. Whether to undertake such reductions is a policy matter beyond the committee’s charge or expertise and involves, among other things, a judgment as to the benefit of the reductions compared with their costs.
Because of the limitations in IPM, emissions could not be assessed at the level of the generating unit, and any effective strategy must be designed and implemented to guard against potential pitfalls, such as worsening air quality in a particular local area. It is reasonable to conclude, however, that the implementation of the ERP could lead to some local changes in SO2 and NOx emissions (as well as emissions of other power-plant pollutants, which are not covered under CAIR or other cap-and-trade programs), with increases in some locations and decreases in others. The magnitude of those changes and the number of geographic areas affected are unknown. The committee concluded that although IPM can provide some reasonable insight into national emission patterns under alternative scenarios, such insight is on too large a scale for assessment of health benefits and should not be used for such purposes. It is possible that the spatial redistribution of emissions would have some important effects, either locally or in the aggregate. As discussed in Chapter 7, NOx emission reductions in an urban area can have a very different effect from NOx emission reductions in a rural area on O3 concentrations (both in direction, at least close to the source, and
in magnitude). In addition, although SO2 emission reductions will reduce ambient sulfate (SO4) concentrations in all locations, the amount of SO4 reduction will vary geographically, and the health effects will be influenced by downwind population patterns, as well as ambient concentration patterns. Thus, even if IPM were able to provide robust national-level emission estimates, the magnitude or direction of health effects would be unclear without additional geographic specificity.
The committee notes the theoretical possibility that the ERP could reduce emissions by increasing the replacement of old higher-emitting equipment with lower-emitting equipment (although not the lowest-emitting equipment that might have been required before the ERP). Under the revised NSR rules, fewer investment projects would require NSR permits, thus reducing the costs of such projects, both in terms of avoiding NSR-permit-related emission controls and potential delays and uncertainties caused by the NSR permitting process. The newer production equipment might be cleaner than the older equipment it replaces and result in some reduction in emissions, without the addition of pollution-control equipment. Therefore, if the revised NSR rules encouraged additional investment in new equipment, the result could be a reduction in emissions at some facilities, if those facilities would have avoided triggering prerevision NSR by delaying investment in process upgrades. Key questions are whether many investment projects that were discouraged by the prerevision rules would proceed under the revised rules, and how much those projects would reduce emissions. However, available empirical data are not sufficient to formally evaluate this effect.
Other Industrial Sectors
Sectors other than electricity-generating facilities are affected not only by the ERP but also by the 2002 rule changes, which have gone into effect in some locations. However, as described in Chapter 2, the 2002 rule changes have been implemented in only some states, and sufficient data are not available to evaluate the effects of the NSR rule changes with any of the committee’s preferred analytic approaches. In addition, industry simulation models are not available for sectors other than the electricity-generating sector, and simulating plant-level decision processes and government regulatory decisions with structural and behavioral models is implausible. Therefore, the only basis today to determine the effects of the NSR rule changes on the sectors other than the electricity-generating sector would be to rely on anecdotal evidence or previous relevant case studies. As discussed in more detail below, it is the committee’s judgment that such information does not provide an adequate basis for the evaluations required in the committee’s charge.
Our evaluations of permitting data and emission inventories can provide some insight into the sectors other than the electricity-generating sector ex-
pected to contribute most to emissions and air-quality changes as a result of the NSR rule changes. For example, for NOx emissions, the cement industry and pulp and paper mills formed a large fraction of recent NOx permitting activity for modifications and permitted emissions. The geographic clustering of NOx emissions from those sectors in Texas, Pennsylvania, Georgia, and Michigan—all of which either contain ozone or PM2.5 nonattainment areas or are upwind of states that do—indicates that further research into the influence of NSR on those sectors would be warranted. Similar conclusions can be reached for other pollutants, in that substantial permitting activity for modifications was seen in the chemical, cement, and pulp and paper industries for SO2; chemical and allied products, metal industries, and pulp and paper for PM; and pulp and paper, soybean oil mills, and lumber and wood products for VOCs. Those permitting data are insufficient for formal analysis, both because of missing information and because data are lacking on sites that did not upgrade or modify, but they provide some insight into areas on which to focus for future analyses. Reduced-form econometric analysis could also be applied to emissions data, testing another possible impact of the NSR changes.
LONG-TERM COLLECTION OF DATA AND IMPROVED MODELING TECHNIQUES NEEDED TO CARRY OUT THE RECOMMENDED APPROACHES
Overall, because of a lack of data and the limitations of current models, information is insufficient to quantify with a reasonable degree of certainty the potential effects of the NSR rule changes on emissions, on human health, on energy efficiency, or on other relevant activities at facilities subject to the revised NSR program. For any of the analytic approaches recommended by the committee, additional data collection and improved modeling methods would be warranted to improve the likelihood that the effects of the NSR rule changes could be measured. Equally or more important, the steps recommended by the committee would be valuable for measuring the effects of future regulatory changes regarding air-pollutant emissions. In the case of the NSR rule changes, not only were postimplementation data not available, but sufficient preimplementation data also were not available. Prospective data collection in areas where NSR rules are most likely to have the greatest effect could lead to more-informed policy decisions in the future.
As mentioned throughout our report, there is no central database on NSR permitting activity, and that constituted an important data gap in the committee’s deliberations. The RACT-BACT-LAER clearinghouse5 does not
readily distinguish between new sources and modifications, and the availability of the data varies by state. State permit data are generally limited and are often kept in paper form, with no attempt to be compatible with other states’ databases, and there is not much information on minor-construction permits. A standardized database program adopted by all states would make analyses of the permit data more feasible. Regardless of the database program used, the information should be collected by the states in a systematic format (same data fields, field layouts, variable definitions, and so on) and should be maintained by the EPA. We recommend that resources be made available so that EPA and other agencies can collect the information consistently in the future. The information could inform future assessments of NSR rule changes and, perhaps more importantly, could provide the foundation for prospective assessments of other future regulatory actions.
In addition, to prepare for a reduced-form econometric analysis, data should be compiled both on when the NSR rule changes became applicable for facilities in different states (these data are readily available) and on perceptions at regulated firms and among regulators regarding the rule changes, which will help identify when (and whether) investment behaviors are likely to shift. Those data should be collected in both attainment and nonattainment areas.
The Census Bureau data needed for facility-level analyses of investment behavior are already being collected. Researchers wishing to do the analyses need to have Census Bureau-approved projects in secure research data centers. Access to the data is expensive, and the analyses are time-consuming, but there is sufficient time to develop a research protocol before the adequate data would be available. These analyses could be important in evaluating NSR and other related regulations.
Bottom-up sectoral models of the electric-power industry, such as IPM, should be refined to account better for the influence of NSR and related regulations on plant-level decision making. Although that clearly is a daunting task, sequential refinements that capture the factors that influence a plant’s decision to retrofit or perform maintenance activities would be feasible. With other enhancements in plant-specific information, the model refinements could potentially improve the reliability of regional (or local) emission estimates and could allow for air-quality and health effects to be formally quantified. In addition, the committee recommends that sectoral models be refined to facilitate parametric sensitivity analyses and more formal uncertainty analysis. In particular, periodic expert review of key inputs and components of a model of regulatory importance would allow for a
more informed updating of the model. Such an investment of effort may not be warranted only to understand the effects of the NSR rule changes, but development of a better working model of the effect of regulations on plant-level decision making would help to inform numerous future analyses. In addition, the use of bottom-up sectoral models is impaired because such models do not exist for sectors other than the electricity-generating sector. Steps should be taken to compile the necessary input data to support development of models for other industry sectors, such as petroleum refining, to allow for more-informed future analyses.