5
Analytic Methods for Assessing Effects of New Source Review Rule Changes

INTRODUCTION

In principle, numerous methods could be used to assess the effects of the U.S. Environmental Protection Agency’s (EPA’s) recent changes to the New Source Review (NSR) rules. Some methods focus on the response of individual firms or facilities; some focus on entire industrial sectors; and some attempt to assess the responses of multiple sectors or the entire economy. An assessment of all of the factors of interest requires an evaluation of how firms, industrial sectors, or the economy will alter their investments and operations (including pollution control and pollution prevention) in response to changes in the NSR rules and the resulting changes in efficiency and pollutant emissions. The assessment also involves an evaluation of how the emission changes might affect air quality and human exposures and the resulting health consequences of those exposures.

The methods used in evaluating responses to changed NSR rules begin with assumptions about how individual firms or industries respond to regulatory incentives and constraints. In some cases, these assumptions are based on empirical information involving interpretations of historical data, surveys, case studies, or anecdotal reports. In more formal analyses, the assumptions usually also incorporate theoretical constructs that have been developed in the field of economics. The usefulness of



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 134
Interim Report of the Committee on Changes in New Source Review Programs for Stationary Sources of Air Pollutants 5 Analytic Methods for Assessing Effects of New Source Review Rule Changes INTRODUCTION In principle, numerous methods could be used to assess the effects of the U.S. Environmental Protection Agency’s (EPA’s) recent changes to the New Source Review (NSR) rules. Some methods focus on the response of individual firms or facilities; some focus on entire industrial sectors; and some attempt to assess the responses of multiple sectors or the entire economy. An assessment of all of the factors of interest requires an evaluation of how firms, industrial sectors, or the economy will alter their investments and operations (including pollution control and pollution prevention) in response to changes in the NSR rules and the resulting changes in efficiency and pollutant emissions. The assessment also involves an evaluation of how the emission changes might affect air quality and human exposures and the resulting health consequences of those exposures. The methods used in evaluating responses to changed NSR rules begin with assumptions about how individual firms or industries respond to regulatory incentives and constraints. In some cases, these assumptions are based on empirical information involving interpretations of historical data, surveys, case studies, or anecdotal reports. In more formal analyses, the assumptions usually also incorporate theoretical constructs that have been developed in the field of economics. The usefulness of

OCR for page 134
Interim Report of the Committee on Changes in New Source Review Programs for Stationary Sources of Air Pollutants the alternative methods available and the selection of a particular one depends on the methods’ accuracy in representing responses to regulatory incentives and constraints, their sensitivity to the particular regulatory changes being assessed, and their ability to accurately estimate the outcomes of interest in the assessment. Different indicators can be used to assess magnitudes and trends in pollution prevention and control, energy efficiency, emissions, air quality, and health effects (e.g., NRC 1999; Esty 2001; Hayward 2004). Table 5-1 lists possible indicators for each. Many of these indicators vary over time and space or from plant to plant, and some degree of averaging or smoothing may need to be done before the data can be analyzed. In many cases, the data currently are not available from a single comprehensive source (or even distributed among many sources), and thus incomplete data would be used for drawing inferences. Furthermore, the list of measures in Table 5-1 includes factors that are quantitative and directly indicative of the targeted outcome, such as the emissions from individual plants, industries, and states, as well as other factors that are more qualitative and difficult to measure, such as the rate of innovation for pollution prevention and control technology. Because many of the outcomes and indicators in Table 5-1 are affected by a number of factors beyond the realm of the NSR rules (or even pollution control laws in general), such as economic conditions, government investment in research and development (R&D), fuel supplies and prices, and meteorological conditions, these other factors and data should also be considered in analyses that attempt to assess the likely impact of NSR rule changes on the outcomes of interest. Thus, any assessment involves (explicitly or implicitly) comparing two different estimates: an estimate of what would have happened had the rule changes not occurred and an estimate of what will happen with the rule changes. Both are subject to substantial uncertainty, and, as discussed in Chapter 6, it will be necessary to consider a range of possible scenarios for the economic and environmental assumptions that are applied to estimate and compare outcomes of the revised NSR rules with outcomes of the NSR rules before the revisions. The remainder of this chapter reviews the major approaches and methods that have been, or might be, used to assess the impacts of changes in the NSR rules on the outcomes in Table 5-1 at the level of the firm, the industrial sector, and the economy. The committee considered it important to review the full range of methods available for this purpose to determine the extent to which the different approaches could assist in responding to the committee’s charge. This survey is deliberately broad

OCR for page 134
Interim Report of the Committee on Changes in New Source Review Programs for Stationary Sources of Air Pollutants TABLE 5-1 Possible Indicators for Assessing the Outcomes of Interest Outcome Possible Indicators to Assess Outcome Pollution control • Innovation in new technologies   ○ Expenditures for research and development   ○ Inventions and patents • Implementation of new technologies   ○ Adoption by industry and utilities • Improvements in use (“learning by doing”)   ○ Performance histories for selected technologies Pollution prevention (source reduction) • Innovation, implementation, and improvements in industrial processes to be less polluting ○ Expenditures for research and development ○ Adoption by industry and utilities ○ Performance histories of selected technologies ○ Trends in emissions generated per unit of product produced • Life-cycle material-use impacts, considering economy-wide impacts through the supply chain and product delivery use, reuse, and disposal ○ Number of products introduced into commerce with reduced hazardous properties ○ Substitution of materials with less polluting substances Energy efficiency • Innovation, implementation, and improvement in use of new technologies that enable energy efficiency in electricity generation and industrial processes • Energy efficiency of operating units and plants • Industry sector-wide energy use • Life-cycle energy-use impacts, considering economy-wide impacts through the supply chain and product delivery, use, reuse, and disposal Emissions • Trends in emissions for individual units, plants, industries, states, regions, and the nation as a whole • Relationships between emissions and unit and plant operating costs and use • Life-cycle emission impacts Air quality • Ambient concentrations of relevant emitted primary pollutants and pollutants formed in the atmosphere over various spatial and temporal scales. Health effects • Human exposure and dose • Mortality and disease ○ Population incidence ○ Incidence for particular subpopulations (regional, socioeconomic) ○ Risks to highly exposed individuals

OCR for page 134
Interim Report of the Committee on Changes in New Source Review Programs for Stationary Sources of Air Pollutants because different approaches are likely to be required for different industries, and because a final choice of methods has not been made. Furthermore, it is important to understand the general assumptions of each approach so that their practical and theoretical limitations are clear. Once the changes in emissions are estimated, other methods are used to estimate the resulting changes in ambient concentrations, exposures, dose, and human health and environmental risks. A preliminary assessment of the potential of these tools to be used in our evaluation of NSR rule changes is then provided. The assessment approaches discussed in this report will be relevant to the committee’s final report. No assessment results are provided in this interim report. FRAMEWORKS FOR ASSESSING THE IMPACT OF REGULATION In this section, we review the various approaches that can be used to estimate economic behavior in response to regulations at the level of the firm, the industrial sector, and the economy, as well as methods for evaluating the air-quality and public-health impacts of these responses. Where formal methods have been developed and applied, we identify the candidate models available, the types of variables that they estimate, the kinds of input data that they require, and their potential relevancy for evaluating the impacts of the recent changes in the NSR regulations on efficiency and emissions. In applying and interpreting these various models, important issues arise concerning the way statistical procedures are used and model uncertainty is interpreted. As such, we also briefly review key methods and issues for statistical estimation and uncertainty analysis. Assessments of Individual Firm Behavior Decisions to undertake plant maintenance and alterations and/or decisions to implement new or different pollution control technologies are made at the level of the individual firm or facility. Their decisions reflect the constraints and incentives of environmental regulation as well as economic and financial conditions, available information, alternative investment possibilities that compete for the firm’s resources, and individual firm preferences (including tolerance for risk).

OCR for page 134
Interim Report of the Committee on Changes in New Source Review Programs for Stationary Sources of Air Pollutants Assessments of firm behavior in response to regulation can be based on anecdotal reports, directed case studies, surveys of multiple firms, and conceptual economic models. Anecdotal reports, case studies, and surveys have been done by EPA and specific state, environmental, and industry groups to address some of the issues concerning the likely impact of the NSR rules and rule changes. Although studies of this type have the potential to provide important information, they also are subject to bias depending on how issues are framed, survey participants are selected, and questions are asked (Yin 1994; Cox et al. 1995; Stake 1995). As part of our final report, we will evaluate the usefulness of a number of these studies for addressing the issues in our charge. To the extent that their information is representative and pertinent, the insights from empirical studies of the type described above help to inform economic models that characterize and predict how firms will behave in response to different incentives. Economic models estimate behavior based on principles of rational choice and profit maximization (e.g., Tietenberg 2003; O’Sullivan and Sheffrin 2005). Process engineering models that estimate the performance (for example, efficiency), emissions, and cost, given alternative capital investments and operating decisions at individual facilities, can be included as a part of, or a precursor to, these models (e.g., Allen and Rosselot 1997; Lewin 2003). Economic theory of firm behavior provides a useful window into how firms make choices and how they would likely alter their investment, input use, production, and emissions in response to changes in environmental regulations such as the NSR rule changes. Economists assume that firms exist to make profits and that their fundamental objective is to maximize profits by keeping costs low and revenues high. The effects of environmental regulation on firms’ decisions will depend on the stringency and form of the regulation and on the incentives that the regulation provides for firms to adjust their behavior (Magat 1978; Milliman and Prince 1989; Helfand 1991; Montero 2002). The economics paradigm can be at odds with how business leaders might describe what motivates their actions. Firm managers often deny that they are motivated solely by profits, arguing that firms have other goals, such as maximizing market share or even broader social goals that guide their decisions. Indeed, the long-term economic performance of a company can be affected by its commitment to environmental quality. For example, many firms now recognize that consumer confidence and allegiance can be influenced by environmental performance, and that employee health and productivity are likewise affected (Grabosky 1994;

OCR for page 134
Interim Report of the Committee on Changes in New Source Review Programs for Stationary Sources of Air Pollutants Hamilton 1995; Curcio and Wolf 1996; Anton et al. 2004). Nonetheless, a firm’s behavior is clearly disciplined by the realities of the market, and environmental commitments and controls can be costly. Firms that do not behave in ways that are consistent with profit maximization over the long term cannot succeed in a competitive market. The challenge for economic modeling is to characterize the firm’s resources, risks, costs, and profit opportunities that are most relevant to how a profit-maximizing firm responds to regulations and incentives. Data limitations and a lack of understanding of a firm’s constraints and opportunities can make the results of economic analyses highly uncertain, even if profit maximization is generally a good descriptor of firm behavior. The profit-maximizing paradigm described below can inform many different methods of assessing how firms would be likely to respond to regulatory changes, including case studies, surveys of firms, and more formal econometric and simulation models. The key insight from this paradigm is that in understanding how firms will respond to regulation, it is important to understand the incentives created by different forms of regulation. This is particularly true when, as is the case with the NSR rules, the firm’s actions determine whether it will ultimately be subject to the cost of complying with a regulation. Figure 5-1 is a hypothetical illustration of the trade-offs between cost and emission reductions for a firm considering different possible plant maintenance or alteration activities. The figure is simplified in several respects to highlight key implications. First, it represents a continuous and smooth range of alternatives when in fact there may be only a handful of discrete alternatives. Installation of a particular emission-control device is usually an all-or-nothing decision, and the curve is thus more properly characterized by a sequence of discontinuous steps. Second, the curve is not necessarily “drawn to scale,” exaggerating, in most cases, the costs of emission changes relative to the total cost of production. The graph identifies the emissions and costs associated with a set of possible plant maintenance and alteration decisions that a facility is assumed to be considering, while currently operating at location A, with relatively low total production costs but high emissions. The firm is considering a maintenance activity or alteration to the plant that would move it to nearby point M1, allowing it to operate with both lower cost and lower emissions. The change might result in modest improvements in operating efficiency or reliability that yield both cost and pollution benefits—a win-win outcome for the firm and the environment (this is the type of activity that proponents of the recent NSR rule changes hope to

OCR for page 134
Interim Report of the Committee on Changes in New Source Review Programs for Stationary Sources of Air Pollutants FIGURE 5-1 A hypothetical illustration of options facing a firm operating with low costs/high emissions at point A, considering a maintenance activity that would result in a shift to M1, M2, M3, or M4 but that might also trigger an NSR that would require them to shift to the high-cost/low-emissions point labeled NSR. New facilities, with new technology, might be able to operate in the low-emissions/low-cost area denoted by “New Facility?” encourage). However, if the firm fears that the proposed M1 change will trigger NSR, forcing it to move to the “NSR” location in the figure with much higher costs (and much lower emissions), it may elect to forego the M1 maintenance or alteration, thereby losing the opportunity to achieve the lower costs and the associated modest emission reductions.1 1   Although NSR rules are intended to apply only to the case in which emissions of regulated pollutants are significantly increased, an activity of the type denoted by M1 still might trigger NSR—for example, with a system of linked producers, such as utility generators. In particular, consider a single boiler that could be improved so that it generates more and increases its emissions, but with a decrease in the overall emissions from the utility system because the modified plant is more efficient than the one that it replaced. Similarly, a multiplant firm

OCR for page 134
Interim Report of the Committee on Changes in New Source Review Programs for Stationary Sources of Air Pollutants Other maintenance and alteration activities shown on the figure are also possible—opponents of the NSR rule changes fear that they would encourage more of the changes denoted as M3 and M4, resulting in higher emissions (regardless of whether they yield cost savings to the firm, although presumably, the cost-saving maintenance and alteration activities in M4 are more likely than those in M3). Even when the modest emission reductions achieved by M1 and M2 are lost in some cases, proponents of stricter criteria for triggering NSR argue that these criteria yield an overall net reduction in emissions. This happens because stricter rules encourage a number of these high-emitting plants either to make the major changes necessary to reach the low-emission levels of the NSR point on the curve (because they cannot continue to operate at the current point A without implementing the activity that now triggers NSR) or to be replaced by new facilities that do. Because of new, advanced technologies, these new facilities might even be able to achieve the lower emissions with much lower total costs, operating in the region denoted on the graph as “New Facility?”. Hart (2004) discusses how different regulatory policies can provide incentives for industry to adopt new vintages that lead to both reduced pollution and production growth. In the next section, we discuss in more detail the formulation of conceptual models of profit-maximizing behavior that underlie the trade-off between reductions in costs and emissions illustrated in Figure 5-1. Conceptual Models Conceptual models provide a formal mathematical representation of how firms make choices to maximize profits. The most common assumption in these models is that firms operate in competitive markets where they take the prices of the products that they produce and of the inputs that they use as given. The profit-maximization problem involves finding the amount of inputs to use that maximizes total profits, given a production function that relates inputs to outputs. Emissions of pollution and the capital equipment used to reduce emissions can also be repre-     might have a least-cost strategy for decreasing emissions using modifications that rebalance production so that there is an increase in emissions at some plants, despite the net reduction in total emissions.

OCR for page 134
Interim Report of the Committee on Changes in New Source Review Programs for Stationary Sources of Air Pollutants sented. A simple model of firms’ behavior typically has the following structure: Decision variablse: quantities of inputs to production that are used, including fuel, labor, capital, and pollution control equipment. Parameters: input and output prices. Constraints: what values of the decision variables are feasible or allowable. In an environmental assessment, the major constraints are as follows: A production function that defines how inputs are transformed into output. An emissions equation that calculates the amount of emissions resulting from using different combinations and amounts of inputs and relates these emission levels to emission limitations the facility must satisfy. Other environmental constraints such as restrictions on fuel input use or the use of specific pollution abatement equipment that the facility must satisfy. Objective function—a function that identifies the combination of decision variables that will maximize profits (revenues minus costs). This model typically leads firms to produce output up to the point where the marginal cost of increasing production by a single unit is equal to the price at which the firm can sell its product. When the level of the firm’s output is known, the decision becomes one of selecting the mix of inputs that minimizes the firm’s cost of production given its production function. The solution to the cost-minimization problem, for a given level of production, can be used to determine a relationship between emissions and the total cost of production, such as the one shown by the dashed line in Figure 5-1. Firms face environmental constraints that can take a variety of forms. An operating permit affecting a facility’s behavior typically imposes a cap on emissions from an operating unit within the facility. Often this cap is based on a desired maximum emission rate per unit of heat input and an assumption about maximum levels of fuel use. Environmental constraints can also take the form of requirements to install control equipment (including specific classes of control devices when technology-based rules are in place) that achieves required emission limits or requirements to use lower-polluting fuels, such as lower-sulfur coals in the generation of electricity. In some cases, firms can participate in a national or regional cap-and-trade program for emissions when the

OCR for page 134
Interim Report of the Committee on Changes in New Source Review Programs for Stationary Sources of Air Pollutants constraint takes the form of a requirement that firms hold sufficient allowances to cover their annual or seasonal emissions of a particular pollutant. All these requirements impose costs on the firms that will influence the trade-offs that they make when determining how to produce their product and how much pollution to emit. Regulated Markets When, as was traditionally the case for electricity generation, the price of the firm’s output is determined by a regulator (e.g., a state Public Utility Commission) and not by the market, the firm’s profit-maximizing problem includes an additional constraint, and the product price is no longer a parameter in the model. Typically, regulators set regulated price equal to average cost, which provides weak incentives to minimize costs. Recognizing the weak incentive properties of average cost pricing, market regulators increasingly are relying on other forms of regulation such as capping product prices to provide regulated firms with an incentive to reduce costs. In particular, as the electric-power industry in several states has been making its way through the transition from monopoly regulation to competition, prices for electric power have been capped, providing strong incentives to reduce costs. How regulators treat pollution control costs and other costs associated with environmental regulation in setting prices can have very important incentive effects on a firm’s choices over various options for complying with environmental regulation. Differences across state electric utility regulations in the treatment of emission allowances, costs of fuel switching, and costs of flue gas desulfurization (FGD) scrubbers had a definite role in shaping how electric utilities chose to comply with Title IV of the Clean Air Act (CAA) Amendments (Bohi and Burtraw 1992; Arimura 2002). Movement toward more competitive pricing of electricity generation will diminish the importance of these effects, but in certain regions of the country, such as the Southeast, deregulation of electricity-generation pricing is proceeding very slowly. Differentiated Regulation In the models discussed above, a firm has no influence on whether it is subject to a particular environmental regulation. For regulatory programs, such as new source performance standards (NSPS) and NSR, a

OCR for page 134
Interim Report of the Committee on Changes in New Source Review Programs for Stationary Sources of Air Pollutants firm does not face the regulation until it takes a particular action. If a firm builds a new facility, then it knows that facility will be subject to NSR and possibly even stricter requirements (depending on where the facility is located). If a firm makes a major modification to an existing facility that is deemed to result in a “significant” increase in pollution, then it will be subject to NSR. The revisions to the NSR program that are the subject of this report affect the conditions under which NSR applies to an alteration at a facility. To analyze such endogenously triggered regulation, a dynamic formulation of a firm’s profit maximization problem becomes more appropriate. Because major alterations to facilities are capital investments, the problem should be extended to include multiple periods and the firm’s objective should be restated as one of maximizing the present discounted value of future profits. Firms will compare discounted profits with and without the alteration and choose the course of action that appears to be the most profitable. Future costs with the alteration will include the costs of regulatory requirements triggered by NSR, and future costs without the alteration may include reduced levels of equipment reliability and other adverse outcomes. If the additional costs of complying with the NSR rule outweigh the benefits of the contemplated alteration, then the firm will not make the change. Being subject to the NSR rule may affect the payoff to the firm of different investment options and, in theory, could cause the firm to forego investments that would reduce emissions or improve energy efficiency at a facility, as illustrated in Figure 5-1. The extent to which this has happened in practice is the subject of much debate. Firms and industries indicate instances when the potential to trigger NSR requirements made or might have made plant upgrades too costly to move forward. However, there is no way to independently corroborate such reports and rigorous statistical studies of this phenomenon do not exist, party because of lack of data and the difficulty of identifying the effects of NSR given all the varied influences on investment decisions. One recent empirical study that applied statistical methods analyzing possible effects of NSR rules, as distinct from NSR rule changes (List et al. 2004), is discussed later in this chapter. Several features of the NSR rule changes, including the change in the selection of test years for emission changes and the minimum expenditure threshold for major modifications, reduce the types of investments at existing plants that will trigger NSR. By removing certain types of expenditures from the category that triggers NSR, the rule might reduce the regulatory uncertainty facing the source and lowers the cost of many types of in-

OCR for page 134
Interim Report of the Committee on Changes in New Source Review Programs for Stationary Sources of Air Pollutants lationship between nonattainment status and measures of economic activity, including employment, investment, and shipments for plants in polluting industries. He found that plants in nonattainment counties have lower employment, lower shipments, and lower total capital stock than analogous plants located in attainment areas, but the size of these effects is relatively small. Numerous other studies have used these data to study the effects of regulation on specific industries including pulp and paper (Gray and Shadebegian 2003) and oil refineries (Berman and Bui 2001). Econometric methods have also been used to study the effects of environmental regulation on R&D and innovation; many of those studies are reviewed by Jaffe et al. (2003). The empirical studies present mixed results. Lanjouw and Mody (1996) found a significant positive relationship between pollution abatement expenditures and patenting activities. Jaffe and Palmer (1997) found a positive relationship between pollution abatement expenditures and R&D expenditures but no impact of the former on patenting. Taylor et al. (2003) determined that CAA regulations have had a significant impact on air-pollution-control innovation and patenting. Other studies looked at the effects of energy price changes and explicit efficiency standards on the nature of technological change in appliances, focusing particularly on the energy-savings character of the innovations (Newell et al. 1999). Another modeling approach that can be used to explore the impact of regulation on environmental performance involves using “adaptive agents” to simulate the innovation and production activities of multiple firms competing in a product market. The decisions of the agents evolve over time in response to changing consumer preferences and demand and regulatory decisions affecting costs, prices, and profitability. These models have been used to explore the factors that affect the evolution of green products and processes (Teitelbaum 1998; Axtell et al. 2002; Bulla and Allada 2003). However, they are still in the early stages of research development, and none has yet been advanced to the point where detailed decisions on plant maintenance and replacement of the type that are important to NSR rule making can be evaluated. Estimating Effects Across Multiple Sectors of the Economy The sectoral assessments described in the section “Assessment of Sector-Wide Response” are concerned only with the direct effects of a policy on an industry and its immediate inputs and outputs. When firms

OCR for page 134
Interim Report of the Committee on Changes in New Source Review Programs for Stationary Sources of Air Pollutants modify their production levels or product designs in response to regulation or other incentives, the effects of these decisions ripple through the economy and affect other industries, including those that provide their inputs, those that use their products, quantities shipped, and associated emissions. These indirect or ripple effects can be important. One type of indirect effect results from substitution among sectors. For instance, a policy aimed at reducing emissions from a particular sector might result in increased prices for that sector’s output, shifting demands to other goods and services in the economy whose emissions then might increase. Electricity, for example, competes with natural gas, fuel oil, wood, and other fuels for home and water heating in various geographic markets in the United States, and policies that affect the price of electricity will influence the mix of fuels the residential sector consume. Another type of indirect impact is the result of upstream and downstream effects. As an example, the consideration of emissions only from the combustion of fossil fuels in power plants disregards emissions from other stages of the fuel cycle, including fuel extraction, transportation, and waste disposal. Tracing indirect effects throughout the entire economy is the focus of life-cycle analyses and macroeconomic models that compute a general equilibrium outcome for multiple sectors of the economy. Each of these methods is briefly reviewed. Studies that attempt to quantify ripple effects on the economy and the environment (as well as the direct effects from product manufacture) are referred to as life-cycle assessments (LCAs). The ripple effects occur “upstream” of the particular company, as modified orders to suppliers, their suppliers, and so forth. They also occur “downstream” of the production process because modified products and production quantity can result in changes in the emissions that occur during product use, reuse, recycling, and disposal. LCAs can be performed by one of two methods: the process modeling approach or the economic input-output approach. The process method is the underlying principle behind a variety of LCA tools, most of which have been developed in Europe (Fruhbrodt 2004): the Society of Environmental Toxicology and Chemistry (SETAC) endorses this approach (Hendrickson et al. 1997). The process modeling approach requires that each aspect of a particular product’s life cycle be analyzed and documented. The models require extensive databases on materials and manufacturing and nonmanufacturing processes and use these to estimate a wide range of economic, technical, social, and environmental

OCR for page 134
Interim Report of the Committee on Changes in New Source Review Programs for Stationary Sources of Air Pollutants impacts. Environmental impacts include factors such as resource depletion, acidification, eutrophication, global warming, human toxicity, freshwater aquatic toxicity, marine aquatic ecotoxicity, terrestrial toxicity, ozone layer depletion, tropospheric ozone creation, and radiation. Technical impacts include nonrenewable and renewable energy consumption and energy efficiency. The second approach, based on economic input-output models, considers the full set of economic transactions between different sectors in the national economy. The general framework of the economic input-output model was developed by Nobel-Prize-winning economist Wassily Leontief, and it requires that a nation’s economy be divided into sectors (typically about 500). The inputs and outputs of these sectors are then defined by the 500 × 500 matrix that quantifies the economic transactions between each. The total transactions that ripple through the economy (all upstream flows) are computed for each unit of economic activity, and these can be adjusted linearly to produce estimates for various dollar amounts of output. With this framework, an economic input-output model is capable of determining the total economic activity and associated environmental impacts from any purchase amount of a particular product or service. Because of the comprehensive data provided by the U.S. Census Bureau, the economic input-output model is able to trace even seemingly unrelated and insignificant transactions such as office computer paper used at a manufacturing plant. The Bureau of Economic Analysis also develops work files that are used to extract specific data points from the large input-output matrix produced by the U.S. Census Bureau. A model that uses this approach is the Environmental Input-Output Life Cycle Assessment program (Green Design Initiative 2004). Another method for looking at the broader economic effects of an environmental policy change is the use of macroeconomic analysis, implemented with a computable general equilibrium (CGE) model. In contrast to an input-output model, which assumes that inputs are used in fixed proportions, the CGE model allows firms to adjust their mix of inputs in response to changes in relative prices. If, for example, a change in environmental regulations increases the demand for a particular fuel, thereby increasing its price, the CGE model allows that effect to feed through to other sectors. Although this additional flexibility provides a better representation of how industries would respond to regulation-induced price changes, it comes at a cost in terms of sectoral detail. Most CGE models have only aggregate sectoral detail, dividing the entire economy into between 5 and 25 economic sectors. When augmented

OCR for page 134
Interim Report of the Committee on Changes in New Source Review Programs for Stationary Sources of Air Pollutants with information on emission rates, CGE models can be used to look at the effect of the regulatory changes on direct emissions from regulated sectors and indirect effects on other sectors. CGE models have been implemented primarily at the scale of national economies and have been applied most frequently in the realm of environmental evaluation to address issues related to carbon control for climate change assessment, although national assessments have also been conducted for elements of the CAA (Jorgenson and Wilcoxen 1990a,b, 1993a,b; Manne et al. 1995; Fossati and Wiegard 2002). Some models go further than just looking at emissions by incorporating air transport models that estimate impacts on pollutant concentrations and dose-response functions that translate air-quality changes into effects on human health and the environment. These models, known as integrated assessment models, often also include economic estimates of the monetary value of various changes in human health and environmental endpoints. These models have been used to analyze the benefits and costs of different environmental regulations including Title IV of the 1990 CAA Amendments (EPA 1997). More information on the type of modeling used to translate emissions into environmental effects is provided in the following section of this chapter. LCAs, CGE, and integrated assessment models can provide useful information to evaluate the ripple effects of changes in production and demand that result from environmental regulation. However, given the difficulty in determining even the direct effects of NSR rules and rule changes on the production and emissions of regulated plants and industries, the use of a tool that translates these direct effects into estimates of changes in indirect economic activity and environmental emissions is premature at this time. As better estimates for direct effects are obtained and LCA and CGE tools are improved to allow more location- and plant-specific calculations, the use of these methods to estimate ripple effects should be considered. Estimating Environmental and Public Health Impacts Quantifying the influence of changes in emissions on public health and welfare is a complex, multistage process involving the integration of multiple data streams with physical, econometric, and behavioral theories using statistical models and expert opinion (NRC 2002). Models are needed to evaluate the causal pathways from the effects of NSR rule

OCR for page 134
Interim Report of the Committee on Changes in New Source Review Programs for Stationary Sources of Air Pollutants changes, beginning with the relationship between changing plant emissions and ambient air pollution, followed by personal exposure, and then human health effects. Engineering, econometric, chemical, atmospheric, biomedical, and sociological theories and data are needed to inform these relations, but available empirical information is generally insufficient to the task. In some cases, data are not available; in others, data are incomplete or observational rather than experimental so apparent relations need to be adjusted. For example, in addition to changes in NSR rules, medium- to long-term trends in ambient pollution are influenced by emissions from other sources not affected by NSR rule changes and possible trends in weather or climate. Trends in health outcomes likewise are affected by a variety of factors in addition to ambient air pollution, and even pollution effects can be modified by personal behavior (e.g., susceptible individuals may alter their behavior on high-pollution days). Additional complications arise from the need to assess relations over time and at relatively fine geographic scales. Because of the complexity of the relations and the relative lack of direct information, simulation models and complex statistical analyses are necessary to help sort out this causal network. These formal approaches are necessary to document assumptions, define and organize inputs and outputs, and, as much as possible, isolate the effects of NSR changes in the set of other candidate causes. A properly conducted and reported formal approach identifies relevant uncertainties and ensures that their influences are embedded in the outputs. Typically, the process begins with an estimate of how much emissions will change as an input, then estimates how these changes will affect exposure, and then estimates how these changes in exposure will affect human health. Ambient Concentrations and Exposure Outcomes Once changes in emissions have been estimated, atmospheric dispersion models are needed to relate emission changes to temporally and spatially indexed ambient concentrations and deposition. Pollutant fate and transport are affected by stack height and diameter, pollutant exit temperature and velocity, and other site characteristics, so differentiating among sources and source categories is important. Because relevant atmospheric conditions such as temperature, humidity, wind speed and direction, and background pollution levels vary both seasonally and spa-

OCR for page 134
Interim Report of the Committee on Changes in New Source Review Programs for Stationary Sources of Air Pollutants tially, it is important to estimate where and when emission changes occur as well as the seasonal variations in these changes. Assessments of the effects of changes in the NSR rules depend on estimating impacts of emission changes on ambient concentrations of primarily emitted pollutants such as SO2, carbon monoxide, and large or fine primary particles (PM10 and PM2.5, respectively) as well as secondarily formed fine PM (PM2.5) and ozone. In previous regulatory impact analyses of policies involving these pollutants, EPA applied multiple models with various degrees of sophistication, spatial coverage, and spatial resolution. For example, when evaluating the benefits and costs of the CAA (EPA 1999), EPA used the Urban Airshed Model (UAM) to evaluate ozone and combined the Regional Acid Deposition Model/Regional Particulate Model with the Regulatory Modeling System for Aerosols and Acid Deposition (REMSAD) to evaluate impacts on PM2.5, PM10, acid deposition, and visibility. Only a few historic episode dates were simulated and the geographic resolution was fairly coarse (56 × 56 kilometer grid spacing or greater over much of the country). More recently, EPA has used the Comprehensive Air Quality Model with Extensions (CAMx) for assessing ozone and is in the process of applying the Community Multi-scale Air Quality (CMAQ) model for its updated analyses of the Clean Air Interstate Rule. A significant contributor to uncertainty in atmospheric modeling involves the formation and subsequent fate and transport of secondary pollutants (such as sulfate particles, nitrate particles, and ozone). The UAM captures many critical factors influencing ozone formation, including the spatial distribution of emissions of NOx and VOCs (including compositional information), spatially and temporally varying wind fields, diurnal variations of solar insolation and temperature, wet and dry deposition, and the Carbon Bond IV subroutine for chemical reactions among important species (EPA 1999). But UAM has been shown to underestimate diurnal variability and has been recommended more for average patterns over longer time periods than for site-specific short-term estimates (Hogrefe et al. 2001). Similarly, REMSAD and related models contain modules for formation of secondary sulfates and nitrates, which depend on the relative ambient concentrations of sulfate, nitrate, and ammonium, solar insolation and temperature, wet and dry deposition processes, and other factors. Given the nonlinear and regionally varying relationship between changes in precursor emissions and changes in PM2.5 concentrations (West et al.

OCR for page 134
Interim Report of the Committee on Changes in New Source Review Programs for Stationary Sources of Air Pollutants 1999), uncertainty beyond that of primary pollutant fate and transport should be anticipated. Many uncertainties have been addressed in recent modeling efforts. Improved emission inventories, detailed gas-phase constituents, and hundreds of reactions have been incorporated in air-pollution models. Models with explicit microphysics and chemical thermodynamics have been developed that provide mechanistic descriptions of the partitioning of gas-phase pollutants to the particle phase. These provide a more accurate description of particle evolution (Seigneur et al. 1999). Basic aspects of secondary aerosol formation prompted by ozone photochemistry vary substantially among models, with few reproducing the observed afternoon maximum in particle growth (Pun et al. 2002). This critical feature of aerosol growth is common in many regions of the country, yet many widely used models do not adequately address it. A fate-and-transport model with outputs used in a health benefits analysis does not have to accurately estimate at overly fine spatiotemporal scales. However, models are expected to perform well in estimating over long time frames and at relevant spatial scales. Local transport is insufficient, because studies have shown that a substantial portion of the health impacts of a source with an elevated stack can occur hundreds of kilometers from the stack (Levy et al. 2003). For secondary PM and ozone, such estimations are challenging, because detailed meteorologic and pollution data are required. Also, the models should be able to capture the time resolution that matches the evidence used to develop concentration-response functions. If 1-hour maximum ozone concentrations are associated with health outcomes, a model that lacks hourly concentration estimates will be deficient. Estimation of spatiotemporal exposure gradients have relied on coupling physical models with data available from ambient monitoring stations coupled with statistical interpolation and smoothing models. The best of these provide space-time point estimates and relevant uncertainties using formal Bayesian models (Christakos et al. 2001). Of course, personal exposures to pollutants can differ substantially from ambient concentrations. Efforts are being made to study the relationship between the two, but most epidemiologic, health effects studies have been based on data from monitors of ambient concentrations. Even though the locations of these monitors are not ideal for estimating population exposures (many were located to assess regulatory compliance), most health effects studies have relied on these data (NRC 2002). Many uncertainties in exposure relationships remain. For example, because

OCR for page 134
Interim Report of the Committee on Changes in New Source Review Programs for Stationary Sources of Air Pollutants people generally spend most of their time indoors, many individual factors will influence the relationship between personal exposure and ambient concentrations. However, to understand the impacts of changes in NSR, we are concerned with personal exposures to air pollutants of ambient origin. For pollutants such as PM, personal exposure to air pollutants of ambient origin is highly correlated with ambient concentrations (EPA 2003b). However, there is some evidence that ambient levels of gaseous criteria pollutants may be more strongly correlated with personal PM2.5 exposures than with personal exposures to the gases themselves (Sarnat et al. 2001). In an exposure-health assessment, these and similar uncertainties should be documented and, to the extent possible, incorporated into the analysis. Relating Ambient Concentrations and Exposure to Health Outcomes To evaluate health impacts of concentration changes, concentration-response functions are developed for key health outcomes—ranging from mild morbidity effects to premature mortality. For most health outcomes, epidemiologic studies are used to develop the concentration-response functions, with animal studies and human experimental studies providing corroborating evidence for causality (NRC 2002). Many studies are available that employ a variety of approaches. Integrating findings across multiple published studies (research synthesis; meta-analysis) is generally preferred to selecting single “representative” studies. The synthesis should be based on an underlying model or models, including multistage models that incorporate site or study characteristics if heterogeneity in effects is present (Levy et al. 2000; Dominici et al. 2003). To the extent possible, it is important to evaluate the independent effects of the pollutant in question, usually by regression adjustment for co-pollutants. Proper treatment of these issues often requires advanced statistical methods. A critical component in this stage of the analysis is the evaluation of whether thresholds for the health effect are anticipated, or, more generally, whether the concentration-response function deviates from linearity. Most key epidemiologic evidence to date has not detected thresholds or statistically significant deviations from linearity (Daniels et al. 2000; Pope et al. 2002), although these studies (and most studies) have low statistical power to address these issues. The approach by EPA generally has involved computing a baseline estimate assuming no threshold and conducting sensitivity analyses for selected plausible thresholds. This

OCR for page 134
Interim Report of the Committee on Changes in New Source Review Programs for Stationary Sources of Air Pollutants has been considered a reasonable approach but is only one of many candidate options (NRC 2002). Though some laboratory-controlled exposure studies of the short-term effects of air pollution are available, and there are a few “natural experiments” (Pope 1989; Friedman et al. 2001; Clancy et al. 2002; Marufu et al. 2004), most of the evidence on the health effects of ambient and indoor air pollution comes from observational studies that relate changes in health indicators to changes in exposure to air pollution. Estimates of acute exposure effects come from time-series studies. These studies relate short-term, within-location changes in air pollution to relative changes in death rates or other health outcomes. Because mortality rates is also associated with season, temperature, day of the week, and other pollutants, sophisticated statistical models using covariate adjustment and semiparametric regression are needed to adjust for long- and medium-term temporal variations and for other potential confounders. Estimates of long-term effects come from cohort studies. These studies follow individuals and use between-location variations in air pollution as the basis for estimating health effects. Both types of studies have their advantages and drawbacks, and research continues on reconciling estimates of effects (the cohort studies produce higher effects). Fitting Models and Characterizing Their Uncertainty As identified above, simulation-based and statistical models are needed to sort out key relationships in the chain from emissions to health effects. Sophisticated simulation approaches have been applied by EPA and in a wide variety of other contexts. Sophisticated statistical models are needed to integrate information from a variety of sources, gathered over different spatial and temporal scales, and with different degrees of measurement error, biasing and confounding influences (Rothman and Greenland 1998; Robins 1999; Robins et al. 1999; Mugglin et al. 2000; Pearl 2000; Zeger et al. 2000). Sensitivity analyses are especially important to quantify the robustness or fragility of conclusions to changes in model (or simulation system) form and inputs. There are a variety of methods for quantifying uncertainty in model inputs and outputs and for dealing with structural uncertainties in models and scenarios. Morgan and Henrion (1990) and Cullen and Frey (1999) provide an overview of such methods. Typically, uncertainties in the

OCR for page 134
Interim Report of the Committee on Changes in New Source Review Programs for Stationary Sources of Air Pollutants inputs to models can be quantified based upon statistical analysis of empirical data, the encoding of expert judgment in the form of probability distributions, or a combination of both. As an example, Bayesian methods provide an effective way to document assumptions; link information and expert opinion; guide analyses of complex, multilevel, nonlinear systems; ensure that all relevant uncertainties are incorporated and reported; and structure sensitivity analyses. Armitage et al. (2002) provide an excellent introduction; Carlin and Louis (2000) provide a more advanced treatment. In any modeling or simulation exercise, two kinds of uncertainty operate: inherent stochastic (also called sampling uncertainty) and modeling uncertainty (also called nonsampling uncertainty) (see Morgan and Henrion 1990). The boundary between the two is fuzzy (some nonsampling uncertainties can be embedded in an overarching model), but a distinction can be made. Modeling uncertainties tend to dominate an assessment but generally are underexplored and underreported. Although advanced statistical methods of the type described above are unlikely to be feasible for our evaluation (given both data and time limitations), we will attempt to highlight, at least qualitatively, key conceptual uncertainties in the modeling approaches that we use. Furthermore, EPA and others charged with addressing this issue over the long term should develop the capability for implementing these tools, as appropriate, in future assessments. SUMMARY EVALUATION OF APPLICABILITY OF ANALYSIS METHODS FOR ESTIMATING IMPACTS OF CHANGES IN NSR RULES Whether the formal methods described in this chapter will have sufficient sensitivity to the NSR rule changes under investigation to be able to estimate their effects accurately remains to be determined. Nonetheless, insights into the behavior of individual firms might help in estimating how individual facilities could respond to the incentives created by the rule changes. If recent historical evidence supports these behavioral models, this might then allow an assessment of at least the direction of the impacts of these changes on the outputs of concern (e.g., whether emissions are likely to increase or decrease) and possibly an estimate of the magnitude of the impact for typical facilities in different industrial sectors. Some models of the electricity-generating sector appear to be sufficiently detailed and sensitive to allow conclusions about the re-

OCR for page 134
Interim Report of the Committee on Changes in New Source Review Programs for Stationary Sources of Air Pollutants sponses of individual facilities to be aggregated to the entire industrial sector. Long-term simulations with these models could allow a first assessment of how changes in NSR rules might affect technology adoption and emission trends. However, such a model would have to be subjected to thorough sensitivity analysis to see how much the conclusions change with different input assumptions and scenarios—for example, concerning the rate of innovation, the stringency of regional or national caps on pollutant emissions, surrenders of emissions allowances under NSR settlements, and the cost of alternative electricity-generation and pollution control technologies. Furthermore, models with the capacity for representing alternative technologies in a long-term simulation are not available for other sectors, and the time and resources available to the committee are not sufficient to support the construction of sector models for this purpose. For these other sectors, therefore, any generalization from the estimates of facility-level responses to estimates of industrial-sector responses will have to be undertaken more informally. For the most part, the multisector models are even less able to represent the types of changes we are assessing than the sector models. Modifying the available models so that they can reflect these changes is substantially beyond the committee’s capacity or resources. Therefore, any intersector impacts will also have to be assessed informally, and any estimates of their direction or magnitude are likely to be highly uncertain. The most appropriate way of assessing the impacts on health and other outcomes of any emission changes estimated on the basis of the above assessments will depend substantially on the amount and quality of information resulting from these assessments. In many cases, the human health impacts, for instance, are likely to depend on which specific facilities change their emissions in response to the rule changes, who is exposed to the emissions from these facilities, and the ambient air quality in the vicinity of these facilities before the alterations occur. It is unlikely that we will be able, at least in most cases, to make estimations with such specificity. Where we cannot, attempting to undertake sophisticated modeling of human health impacts would have little validity, and we probably will be able to do little more than indicate the likely direction and possibly the rough magnitude of these impacts, if any. As discussed in Chapter 6, it will be necessary to consider a range of possible scenarios for the economic and environmental assumptions that are applied to estimate and compare outcomes from the revised NSR rules with outcomes from the NSR rules before to the revisions.