Right to comply
Regardless of regulatory policies du jour, electric generating
companies will face continued pressures to comply with changing
markets and environmental regulations. Whether local rules
addressing water quality or international treaties addressing
climate change, generation owners will continue to face critical ?
and costly ? decisions about the choice of pollution control
technologies, sales or purchases of specific units, and even
real-time operating procedures. With potentially hundreds of
millions of dollars at stake for a single equipment investment,
developing optimal environmental compliance strategies is a crucial
component towards improving the ?bottom line.? As a result,
Navigant's clients are beginning to ask how they can develop the
best possible compliance strategies that address the inherent
uncertainty about the future.
While there is a growing body of knowledge describing new analytical techniques that directly address future uncertainties, some companies still use older, deterministic techniques that cannot account for uncertainty in a meaningful way. For example, they may fail to consider how future uncertainty will affect their investment options today or fail to incorporate the value of flexible strategies that may be available.
The uncertainties about future environmental regulations, emissions allowance prices, fuel prices, and the mix of generating plants are interdependent. All will affect the future price of electricity. Thus, developing robust and flexible environmental compliance strategies are critical inputs into overall business investment decisions. In this report, therefore, we provide an overview of how Navigant Consulting applies decision analysis techniques to develop effective and robust environmental compliance strategies.
How clean is clean?
While all forms of electric generation, even ?clean? renewables, have environmental impacts, it is fossil-fueled generating plants in the United States, especially large coal-fired units, which have encountered the largest swarm of changing environmental regulations in the past, and which face unprecedented uncertainties in the future. Under the broad guidelines of the 1970 Clean Air Act and its major Amendments in 1977 and 1990, the US Environmental Protection Agency (US EPA) implemented many ?command-and-control? regulations that created an alphabet soup of regulatory acronyms, including BACT, RACT, NSPS, and LAER. These regulations mandated specific environmental system investments, including combustion modifications, low-NOx burners, continuous emissions monitors (CEMs), and ? in the case of new units ? scrubbers and SCR controls, rather than performance standards allowing plants to reduce pollution in the most cost-effective manner.
The 1990 CAA Amendments marked the beginning of a major change in environmental protection philosophy that has taken root at the state and federal levels ? and is increasingly embraced internationally. While these Amendments mandated significant sulfur dioxide (SO2) emissions reductions for utilities as part of the Acid Rain program, they allowed a flexible, market-based approach for the first time, using tradable emissions allowances rather than prescribed control technologies. This allowed generation plant owners more latitude in determining whether to make large capital investments in their plants or, as an alternative, to purchase emissions allowances in the market. While some environmentalists still view allowance trading skeptically, the economic and environmental benefits have been clearly demonstrated through substantially reduced compliance costs and greater emissions reductions than once imagined.
Some 10 years later, there is general agreement that these market-based ?cap and trade? allowance programs work well and will continue to be a vital component for achieving future environmental protection objectives. The Acid Rain Program paved the way for the cap and trade program for oxides of nitrogen (NOx) and, in the future, potentially mercury (Hg) emissions as well. Internationally, plans to limit greenhouse gas emissions also have embraced similar trading mechanisms.
While wider use of market-based compliance programs has been welcomed in the electric generation industry, environmental compliance planning at many companies has not changed all that much from the command-and-control era. Notably, many companies have not addressed uncertainties over the nature and scope of future environmental regulations using new analytical tools, especially decision analysis techniques. Decision analysis is an ideal? and in many cases the only ? approach available that will allow companies to understand and harness for their maximum benefit the compliance flexibility built into recent and emerging regulatory programs.
What is decision analysis?
Decision analysis is a structured approach to problem solving. It is both a quantitative and qualitative approach to illuminating, structuring, and solving complex problems that encompass time and uncertainty. Although frequently equated with the use of decision trees, decision analysis is far broader. Whereas decision trees represent a mathematical technique that can be used to solve decision analysis problems, decision analysis incorporates the values, preferences, and information of decision-makers. Decision analysis builds decision trees, from their ?roots? to their outermost ?branches.?
Good decisions vs. good outcomes
A common misconception of strategic decision making under uncertainty is the difference between decisions and outcomes. A good decision is one that is approached logically, by using available information and accounting for future uncertainties. A good decision recognizes the costs associated with gathering information. A good decision is robust, in that it makes sense over a wide range of possible outcomes. In contrast, a good outcome is simply a result that the decision maker likes. For example, buying hundreds of lottery tickets each week is not a good investment decision ... unless you win a multi-million dollar jackpot.
Unfortunately, because none of us is clairvoyant, good decisions do not necessarily guarantee good outcomes. For example, when you purchase a new car, you are often given the option to purchase an extended warranty. On average, such a warranty will cost you more than the average future cost of repairing that car. Otherwise, the extended warranty provider, being astute, would expect to lose money and not offer such warranties. If you forego the extended warranty, you will have made a ?good? decision in terms of reducing your average ownership cost. But if you are unlucky enough to have your engine seize up moments after the standard warranty expires, you will almost certainly view the outcome of your decision as ?bad.? The point is that, on average, good decisions produce good outcomes and bad decisions produce bad outcomes ? but not always.
The decision analysis process
The decision analysis ?process? can best be thought of as a logical, step-by-step approach to working through a complex and, perhaps, initially unstructured or even unclear problem. In thinking about alternative environmental compliance strategies, for example, a generating company will be faced with a number of important issues. These may include: (1) the stringency, scope, and timing of future environmental regulations; (2) the costs and benefits of competing technologies that promise to reduce pollutant emissions; (3) the costs and benefits of using allowances, fuel switching or other operational practices to achieve compliance; (4) the choices for allocating resources among different generating plants; (5) the purchase and sale decisions that can re-balance the firm's overall generating asset portfolio; (6) the effects of new transmission and generating facilities on existing plants' operations and dispatch; and (7) the ability to influence future regulations.
As shown in Figure 1, the decision analysis process can be separated into three distinct phases: deterministic, probabilistic, and informational.
Deterministic phase
The deterministic phase begins ? oddly enough ? with determining the problem to be solved. That is done by first defining goals and identifying a set of alternative decisions that can be made. Although defining goals may sound rather obvious, we often see how the urge to ?do something? often pre-empts the need to first determine what is to be achieved. Regardless of how sophisticated the analytical approach, neither a good decision can be made nor even the relevant decision alternatives defined, until what is to be achieved is well understood. The history of large-scale investment projects is littered with billions of dollars wasted on poorly conceived and ill-founded projects that owed more to their sponsors' ambitions than any semblance of rational decision making.
The next stage in the deterministic phase is to develop a model that incorporates the values that will measure ?success.? For example, in developing an environmental compliance strategy, one company may wish to invest in new pollution control equipment today before new regulations are issued in order to influence future regulatory direction and minimize the risk of costly future regulatory ?surprises.? Another company may wish to avoid squandering limited capital assets that can be used for other important investments.
The final stage of the deterministic phase is to weed out factors that, even though subject to future uncertainty, have little effect on the value model results. This can be tremendously useful by allowing the decision maker to focus limited resources on just those factors that are important. Sometimes, the factors that intuitively ?should? be important aren't, while those believed to be secondary can be critical to the choice of strategy today.
Probabilistic phase
The probabilistic phase takes the remaining key factors and determines their effects on the overall values of the decision alternatives. The key step in this phase is representing the uncertainties of the important decision variables. For example, the future price of electricity could be expected to affect substantially a generating plant owner's decision to invest several hundred million dollars to upgrade an older generating plant or to retire it. In the probabilistic phase, we would describe the price uncertainty by determining an overall probability distribution of future electric prices. We would also account for the relationships among the different factors, such as electric and fuels prices, and the prices of emissions allowances.
The combination of the deterministic and probabilistic phases allows us to build the decision tree. The tree will consist of ?roots? ? the initial decision options available? ? and ?branches? ? the different paths for the uncertainties, plus any future decisions that are available. At the end of this phase, we can determine the best environmental compliance strategy today, consistent with the company's environmental objectives. Of course, the actual outcome of the strategy will depend on how the future plays out. It can also depend on the decision maker's attitude toward future risk: a risk-averse decision maker, for example, might prefer an ?insurance? strategy that is more costly today, but minimizes the chance of costly future surprises.
Informational phase
The informational phase determines whether it makes economic sense to gather more information that would reduce the uncertainty facing the decision maker. The value of that new information can then be compared to its cost. In essence, therefore, the informational phase determines the value of reducing or eliminating certain uncertainties, as if one could hire a clairvoyant to reveal the future. We do this by assuming there really is a clairvoyant who can tell us what the future will look like. For example, we might estimate that, by eliminating uncertainty in emissions allowance prices over the next two years, the expected value of investing in a state-of-the-art emissions reduction technology would increase by $10 million. Knowing that, the cost of different allowance hedging strategies (e.g., a mixture of allowance puts and calls), could be determined. If those strategies had an expected cost less than $10 million, it would make sense to invest in them.
At the end of the informational phase, one of three things can occur: either a strategic decision will be made or more information will be acquired. For example, more information could be gathered that would lead to revising the entire value model or determining that another decision alternative, such as an emissions allowance hedging strategy, warrants further exploration. And so the process would continue until all cost-effective options for information gathering have been exhausted and a final compliance strategy decision made.
The process described in Figure 1 may look cumbersome and slow, but it need not be. Indeed, it should not be. As with most things in life, decision makers have to balance the costs and the benefits of analysis. Spending months of employees' time to find the cheapest source of pencils is foolish; spending an extra month or so deciding on a billion dollar R&D program is probably not. In fact, it is the deterministic phase that will (and should) take the most time. Defining your goals, and determining a balance among potentially competing goals (e.g., profit vs. environmental stewardship, risk vs. reward, short-term vs. long-term results, etc.) can be a struggle, but well worth the effort.
Applying decision analysis to develop an environmental compliance strategy
The previous section described the decision analysis process in broad terms. But how is it actually done? How are goals determined? How are values developed and used to measure those goals? What do the results of the analysis mean? How do the solutions differ from standard discounted cash flow analysis? What's wrong with just using sensitivity analysis? Where is the fit between decision analysis and ?real options? analysis and other valuation techniques? While comprehensive answers to these questions would require several (dull) books' worth of space, we can provide a flavor for those answers by considering a typical environmental compliance problem.
The decision problem
The owner of an older, mid-sized (500MW) coal-fired generating plant in the Midwest faces some important decisions in order to achieve regulatory compliance for his plant because of the impending NOx regulations and more stringent Acid Rain program provisions. In addition, the owner is aware of the potential for new regulations on mercury emissions, as well as some form of greenhouse gas regulation, whether or not the Kyoto Protocol is ever implemented. Perhaps the owner is also facing EPA allegations that its New Source Review requirements have been violated at the plant, which would that require new emission controls be installed.
The owner must determine whether he should make the necessary capital improvements by choosing one of several competing emissions control technologies, or adopt a ?wait-and-see? approach by purchasing the additional NOx allowances the plant will require. The control technologies under consideration are selective catalytic reduction (SCR), a proven means of significantly reducing NOx emissions, or electro-catalytic oxidation (ECO), an experimental technology whose vendor contends that it can reduce NOx, SO2, fine particulates, and mercury. While ECO is comparatively less expensive to install and operate, it is unknown how effectively it will reduce NOx emissions. Lastly, to complicate matters further, another company has made an offer to purchase the plant for a fixed price.
Decision metrics
The owner realizes there is a risk-return tradeoff. While accepting the sale offer would eliminate all future uncertainty, the owner needs to determine the overall range of value for the other decision alternatives. Thus, the owner decides that an expected net present value (ENPV) analysis is appropriate. The results of that analysis will provide the highest expected-value decision, as well as risk-profiles for all of the alternatives available. The owner is also interested in the value of a ?wait-and-see? approach (sometimes called a ?real-option? value), although he recognizes that the offer to buy the plant might be removed or radically different than the existing offer.
With these in mind, the owner develops a decision framework that encompasses both today's and future decision alternatives. This is shown in Figure 2.
Figure 2 presents the sequence of decisions faced by the generation plant owner. (It is a decision tree without any uncertainties.) After making an initial decision from several options (including a decision to ?wait-and-see?), the owner assumes he will revisit the situation three years from now in 2004, when the new NOx allowance reductions take full effect. At that time, the owner can again make a decision to install either SCR or ECO, again wait-and-see, or possibly sell the plant.1 (Future decision opportunities are usually called ?downstream? decisions.) Another downstream decision is assumed to take place in 2007, when new mercury regulations may take effect and carbon emissions reductions could begin.
Identifying and structuring the uncertainties
In determining the most critical uncertainties, the first step is to use economic principles to identify the most likely market-based and non-market-based variables. Market-based variables will include the future price structure of electricity, and the prices of NOx and SO2 emissions allowances. The price of coal is also uncertain, but that price has generally been far less volatile than the prices of natural gas and electricity owing to the nature of coal purchase contracts. Non-market variables relate to factors such as regulatory requirements, politics, public relations, and the like. In recent times, these factors have been a source of greater concern for many generation companies than market variables.
The next step is to consider the potential relationships between market (and non-market) uncertainties. The reason is that the individual markets are influenced by many factors. The price of electricity, for example, will be influenced by the price of natural gas, to the extent that natural gas is the marginal fuel of choice. These relationships can be graphically represented by what is called an ?influence diagram,? an example of which is shown in Figure 3.
Figure 3 shows the anticipated relationship between future environmental regulations and their affect on the price of SO2 and NOx allowances, and the latter's affect on the market price of electricity. It also shows an influence between environmental regulations and the price of natural gas, and the effect of natural gas prices on the price of electricity.
While influence diagrams are not necessary to develop a decision model and determine a preferred compliance strategy, they are often an excellent way of structuring uncertainties. By using an influence diagram, the uncertainties can be better structured. For example, because of the influences affecting the future price of electricity, an overall probability distribution of that price at a given time would depend on the price of allowances and natural gas.2 This is shown on Figure 4.
In developing an environmental compliance strategy, it is the non-market uncertainties that are likely to be the most problematic. Among the many issues that a plant may contend with are further reductions in NOx and SO2, fine particulate requirements, mercury controls, regional haze provisions, and greenhouse gas limitations.
Figure 5 illustrates a scenario regarding carbon emissions. The nature of future regulation and their monetary impact is unknown. The plant's owner may believe that, owing to something similar to the Kyoto Protocol, a carbon tax is most likely to be imposed in the year 2006. The owner assigns a probability distribution to the imposition date, as well as a probability distribution for the amount of the tax itself. The plant's owner also believes that the imposition year is likely to affect the amount of the tax, reasoning that the sooner the tax was imposed, the smaller it would be. In that case, the probability distribution for the tax amount would be conditional on the year imposed, as shown in Figure 5.
Converting uncertainties into discrete events
Except in some special circumstances, the probability distributions shown in Figures 3 and 4 need to be converted to discrete events. For example, the probability distribution for the amount of a carbon tax might have been determined to have extremes of $0 and $100/ton. Since there are an infinite number of values in between, some way of converting that continuous distribution into several discrete events must be performed.
Fortunately, there are numerical techniques to do this that are standard in statistical and decision analysis software.3 Usually, a probability distribution will be converted into three or four discrete events. For example, the carbon tax year and amount might be converted as shown in Figure 6:
Another useful tool, which can determine and rank the most critical uncertainties, is the ?tornado diagram.? This eponymous diagram ranks, in order of their effect on present value, the different uncertainties. Tornado diagrams are useful in several ways. First, they can provide insight as to those uncertainties that matter most. In Figure 7, for example, the diagram shows that natural gas price uncertainty has the greatest effect on the coal plant's ENPV, while mercury (Hg) regulations matter the least. Thus, more attention should be focused on natural gas prices in the modeling process. Secondly, tornado diagrams can provide insights into potential modeling difficulties. For example, if electric price uncertainty shows little impact, that may indicate a flaw in the underlying strategy valuation model.
Constructing the decision tree and determining an appropriate compliance strategy
Having identified a set of decision alternatives, determined a value measure, and identified and developed discrete representations of key uncertainties, we next draw and solve a decision tree. A decision tree simply provides a schematic representation of a problem. Decision trees contain ?branches,? each of which describes a different path along a realized state-of-the-world. Complex decision trees can have thousands or even millions of different paths. This is why it is important to first determine the most critical uncertainties in order to focus on the variables most likely to influence the choice of strategy.
Figure 8 illustrates a simplified version of the decision tree, first depicted in Figure 2, which uses the term ?regulatory uncertainty? to represent the overall uncertainty of future allowance prices, the price of electricity, greenhouse gas regulations, etc.)
Once the decision tree is constructed, it is solved by working from the end of the tree towards the tree's ?root.? The solution reflects the expected value of the lowest-cost path. The solution will also provide the overall probability distribution of results, which will provide the plant owner with information about the risks of the highest ENPV strategy.
The solution is illustrated in Figure 9, which shows that a ?wait-and-see? approach is preferable, given the assumptions and facts in this case study. The first downstream decision in 2004 may appear to be a counter-intuitive: sell the plant if environmental regulations are severe and investing in new emissions controls if not.
Of course, it is critical to realize that downstream decisions are not ?cast in stone.? The primary objective of the analysis is to determine what ought to be done today, given the uncertainties of the future. This is the true strength of the decision analysis framework: it can easily incorporate future flexibility and calculate the value of that flexibility.
Lastly, Figure 10 illustrates the risk profile of the preferred alternative compared to the ?no risk? strategy of selling the plant today. While selling the plant today has no risk, the sale price is $13 million less than the expected value of the ?wait-and-see? strategy. That price difference represents a risk premium: if the plant owner accepts the purchase offer, he will, in effect, be paying $13 million to avoid the risks associated with keeping the generation plant.
Summary
In writing this report, we have sought to present a brief overview of decision analysis and its application to environmental compliance strategies. Decision analysis is a common-sense, logical approach to frame and clarify decisions in the presence of uncertainty. But it is neither a computer model that makes decisions nor a substitute for careful thinking. Rather, it is a way of identifying, structuring, and solving problems. Applying these tools to determine appropriate environmental compliance strategies doesn't mean that decisions will suddenly become easy and intuitive. Like all tools, decision analysis can be used well or poorly. There will always be risks that cannot be well-defined, assumptions that are wrong, and factors that are simply overlooked.
Given the uncertain future, companies wishing to develop environmental compliance strategies face inherent risks. The most thoroughly researched strategy can fail horribly. But the alternative ? approaching investment decisions in a haphazard and illogical manner ? will almost certainly be much worse.
Navigant Consulting has extensive knowledge, experience, and resources to help companies develop their environmental compliance strategies. Please contact Dr. Jonathan Lesser, Senior Managing Economist, at (802) 865-2121, Tom Mills, Director, at (202) 973-2941, or Andrew Greene, Principal, at (781) 564-9750, for more information on Navigant Consulting's capabilities.
1 One thing to note is that the decision tree necessarily simplifies the decision process. Clearly, the owner could make a decision today, change his mind tomorrow and make another decision, and so forth. While that might be more ?realistic,? it would significantly complicate the analysis without adding much value.
2 This results in what is called a ?conditional? probability distribution.
3 The formal name of the analytical technique most often used is ?gaussian quadrature.?
While there is a growing body of knowledge describing new analytical techniques that directly address future uncertainties, some companies still use older, deterministic techniques that cannot account for uncertainty in a meaningful way. For example, they may fail to consider how future uncertainty will affect their investment options today or fail to incorporate the value of flexible strategies that may be available.
The uncertainties about future environmental regulations, emissions allowance prices, fuel prices, and the mix of generating plants are interdependent. All will affect the future price of electricity. Thus, developing robust and flexible environmental compliance strategies are critical inputs into overall business investment decisions. In this report, therefore, we provide an overview of how Navigant Consulting applies decision analysis techniques to develop effective and robust environmental compliance strategies.
How clean is clean?
While all forms of electric generation, even ?clean? renewables, have environmental impacts, it is fossil-fueled generating plants in the United States, especially large coal-fired units, which have encountered the largest swarm of changing environmental regulations in the past, and which face unprecedented uncertainties in the future. Under the broad guidelines of the 1970 Clean Air Act and its major Amendments in 1977 and 1990, the US Environmental Protection Agency (US EPA) implemented many ?command-and-control? regulations that created an alphabet soup of regulatory acronyms, including BACT, RACT, NSPS, and LAER. These regulations mandated specific environmental system investments, including combustion modifications, low-NOx burners, continuous emissions monitors (CEMs), and ? in the case of new units ? scrubbers and SCR controls, rather than performance standards allowing plants to reduce pollution in the most cost-effective manner.
The 1990 CAA Amendments marked the beginning of a major change in environmental protection philosophy that has taken root at the state and federal levels ? and is increasingly embraced internationally. While these Amendments mandated significant sulfur dioxide (SO2) emissions reductions for utilities as part of the Acid Rain program, they allowed a flexible, market-based approach for the first time, using tradable emissions allowances rather than prescribed control technologies. This allowed generation plant owners more latitude in determining whether to make large capital investments in their plants or, as an alternative, to purchase emissions allowances in the market. While some environmentalists still view allowance trading skeptically, the economic and environmental benefits have been clearly demonstrated through substantially reduced compliance costs and greater emissions reductions than once imagined.
Some 10 years later, there is general agreement that these market-based ?cap and trade? allowance programs work well and will continue to be a vital component for achieving future environmental protection objectives. The Acid Rain Program paved the way for the cap and trade program for oxides of nitrogen (NOx) and, in the future, potentially mercury (Hg) emissions as well. Internationally, plans to limit greenhouse gas emissions also have embraced similar trading mechanisms.
While wider use of market-based compliance programs has been welcomed in the electric generation industry, environmental compliance planning at many companies has not changed all that much from the command-and-control era. Notably, many companies have not addressed uncertainties over the nature and scope of future environmental regulations using new analytical tools, especially decision analysis techniques. Decision analysis is an ideal? and in many cases the only ? approach available that will allow companies to understand and harness for their maximum benefit the compliance flexibility built into recent and emerging regulatory programs.
What is decision analysis?
Decision analysis is a structured approach to problem solving. It is both a quantitative and qualitative approach to illuminating, structuring, and solving complex problems that encompass time and uncertainty. Although frequently equated with the use of decision trees, decision analysis is far broader. Whereas decision trees represent a mathematical technique that can be used to solve decision analysis problems, decision analysis incorporates the values, preferences, and information of decision-makers. Decision analysis builds decision trees, from their ?roots? to their outermost ?branches.?
Good decisions vs. good outcomes
A common misconception of strategic decision making under uncertainty is the difference between decisions and outcomes. A good decision is one that is approached logically, by using available information and accounting for future uncertainties. A good decision recognizes the costs associated with gathering information. A good decision is robust, in that it makes sense over a wide range of possible outcomes. In contrast, a good outcome is simply a result that the decision maker likes. For example, buying hundreds of lottery tickets each week is not a good investment decision ... unless you win a multi-million dollar jackpot.
Unfortunately, because none of us is clairvoyant, good decisions do not necessarily guarantee good outcomes. For example, when you purchase a new car, you are often given the option to purchase an extended warranty. On average, such a warranty will cost you more than the average future cost of repairing that car. Otherwise, the extended warranty provider, being astute, would expect to lose money and not offer such warranties. If you forego the extended warranty, you will have made a ?good? decision in terms of reducing your average ownership cost. But if you are unlucky enough to have your engine seize up moments after the standard warranty expires, you will almost certainly view the outcome of your decision as ?bad.? The point is that, on average, good decisions produce good outcomes and bad decisions produce bad outcomes ? but not always.
The decision analysis process
The decision analysis ?process? can best be thought of as a logical, step-by-step approach to working through a complex and, perhaps, initially unstructured or even unclear problem. In thinking about alternative environmental compliance strategies, for example, a generating company will be faced with a number of important issues. These may include: (1) the stringency, scope, and timing of future environmental regulations; (2) the costs and benefits of competing technologies that promise to reduce pollutant emissions; (3) the costs and benefits of using allowances, fuel switching or other operational practices to achieve compliance; (4) the choices for allocating resources among different generating plants; (5) the purchase and sale decisions that can re-balance the firm's overall generating asset portfolio; (6) the effects of new transmission and generating facilities on existing plants' operations and dispatch; and (7) the ability to influence future regulations.
As shown in Figure 1, the decision analysis process can be separated into three distinct phases: deterministic, probabilistic, and informational.
Deterministic phase
The deterministic phase begins ? oddly enough ? with determining the problem to be solved. That is done by first defining goals and identifying a set of alternative decisions that can be made. Although defining goals may sound rather obvious, we often see how the urge to ?do something? often pre-empts the need to first determine what is to be achieved. Regardless of how sophisticated the analytical approach, neither a good decision can be made nor even the relevant decision alternatives defined, until what is to be achieved is well understood. The history of large-scale investment projects is littered with billions of dollars wasted on poorly conceived and ill-founded projects that owed more to their sponsors' ambitions than any semblance of rational decision making.
The next stage in the deterministic phase is to develop a model that incorporates the values that will measure ?success.? For example, in developing an environmental compliance strategy, one company may wish to invest in new pollution control equipment today before new regulations are issued in order to influence future regulatory direction and minimize the risk of costly future regulatory ?surprises.? Another company may wish to avoid squandering limited capital assets that can be used for other important investments.
The final stage of the deterministic phase is to weed out factors that, even though subject to future uncertainty, have little effect on the value model results. This can be tremendously useful by allowing the decision maker to focus limited resources on just those factors that are important. Sometimes, the factors that intuitively ?should? be important aren't, while those believed to be secondary can be critical to the choice of strategy today.
Probabilistic phase
The probabilistic phase takes the remaining key factors and determines their effects on the overall values of the decision alternatives. The key step in this phase is representing the uncertainties of the important decision variables. For example, the future price of electricity could be expected to affect substantially a generating plant owner's decision to invest several hundred million dollars to upgrade an older generating plant or to retire it. In the probabilistic phase, we would describe the price uncertainty by determining an overall probability distribution of future electric prices. We would also account for the relationships among the different factors, such as electric and fuels prices, and the prices of emissions allowances.
The combination of the deterministic and probabilistic phases allows us to build the decision tree. The tree will consist of ?roots? ? the initial decision options available? ? and ?branches? ? the different paths for the uncertainties, plus any future decisions that are available. At the end of this phase, we can determine the best environmental compliance strategy today, consistent with the company's environmental objectives. Of course, the actual outcome of the strategy will depend on how the future plays out. It can also depend on the decision maker's attitude toward future risk: a risk-averse decision maker, for example, might prefer an ?insurance? strategy that is more costly today, but minimizes the chance of costly future surprises.
Informational phase
The informational phase determines whether it makes economic sense to gather more information that would reduce the uncertainty facing the decision maker. The value of that new information can then be compared to its cost. In essence, therefore, the informational phase determines the value of reducing or eliminating certain uncertainties, as if one could hire a clairvoyant to reveal the future. We do this by assuming there really is a clairvoyant who can tell us what the future will look like. For example, we might estimate that, by eliminating uncertainty in emissions allowance prices over the next two years, the expected value of investing in a state-of-the-art emissions reduction technology would increase by $10 million. Knowing that, the cost of different allowance hedging strategies (e.g., a mixture of allowance puts and calls), could be determined. If those strategies had an expected cost less than $10 million, it would make sense to invest in them.
At the end of the informational phase, one of three things can occur: either a strategic decision will be made or more information will be acquired. For example, more information could be gathered that would lead to revising the entire value model or determining that another decision alternative, such as an emissions allowance hedging strategy, warrants further exploration. And so the process would continue until all cost-effective options for information gathering have been exhausted and a final compliance strategy decision made.
The process described in Figure 1 may look cumbersome and slow, but it need not be. Indeed, it should not be. As with most things in life, decision makers have to balance the costs and the benefits of analysis. Spending months of employees' time to find the cheapest source of pencils is foolish; spending an extra month or so deciding on a billion dollar R&D program is probably not. In fact, it is the deterministic phase that will (and should) take the most time. Defining your goals, and determining a balance among potentially competing goals (e.g., profit vs. environmental stewardship, risk vs. reward, short-term vs. long-term results, etc.) can be a struggle, but well worth the effort.
Applying decision analysis to develop an environmental compliance strategy
The previous section described the decision analysis process in broad terms. But how is it actually done? How are goals determined? How are values developed and used to measure those goals? What do the results of the analysis mean? How do the solutions differ from standard discounted cash flow analysis? What's wrong with just using sensitivity analysis? Where is the fit between decision analysis and ?real options? analysis and other valuation techniques? While comprehensive answers to these questions would require several (dull) books' worth of space, we can provide a flavor for those answers by considering a typical environmental compliance problem.
The decision problem
The owner of an older, mid-sized (500MW) coal-fired generating plant in the Midwest faces some important decisions in order to achieve regulatory compliance for his plant because of the impending NOx regulations and more stringent Acid Rain program provisions. In addition, the owner is aware of the potential for new regulations on mercury emissions, as well as some form of greenhouse gas regulation, whether or not the Kyoto Protocol is ever implemented. Perhaps the owner is also facing EPA allegations that its New Source Review requirements have been violated at the plant, which would that require new emission controls be installed.
The owner must determine whether he should make the necessary capital improvements by choosing one of several competing emissions control technologies, or adopt a ?wait-and-see? approach by purchasing the additional NOx allowances the plant will require. The control technologies under consideration are selective catalytic reduction (SCR), a proven means of significantly reducing NOx emissions, or electro-catalytic oxidation (ECO), an experimental technology whose vendor contends that it can reduce NOx, SO2, fine particulates, and mercury. While ECO is comparatively less expensive to install and operate, it is unknown how effectively it will reduce NOx emissions. Lastly, to complicate matters further, another company has made an offer to purchase the plant for a fixed price.
Decision metrics
The owner realizes there is a risk-return tradeoff. While accepting the sale offer would eliminate all future uncertainty, the owner needs to determine the overall range of value for the other decision alternatives. Thus, the owner decides that an expected net present value (ENPV) analysis is appropriate. The results of that analysis will provide the highest expected-value decision, as well as risk-profiles for all of the alternatives available. The owner is also interested in the value of a ?wait-and-see? approach (sometimes called a ?real-option? value), although he recognizes that the offer to buy the plant might be removed or radically different than the existing offer.
With these in mind, the owner develops a decision framework that encompasses both today's and future decision alternatives. This is shown in Figure 2.
Figure 2 presents the sequence of decisions faced by the generation plant owner. (It is a decision tree without any uncertainties.) After making an initial decision from several options (including a decision to ?wait-and-see?), the owner assumes he will revisit the situation three years from now in 2004, when the new NOx allowance reductions take full effect. At that time, the owner can again make a decision to install either SCR or ECO, again wait-and-see, or possibly sell the plant.1 (Future decision opportunities are usually called ?downstream? decisions.) Another downstream decision is assumed to take place in 2007, when new mercury regulations may take effect and carbon emissions reductions could begin.
Identifying and structuring the uncertainties
In determining the most critical uncertainties, the first step is to use economic principles to identify the most likely market-based and non-market-based variables. Market-based variables will include the future price structure of electricity, and the prices of NOx and SO2 emissions allowances. The price of coal is also uncertain, but that price has generally been far less volatile than the prices of natural gas and electricity owing to the nature of coal purchase contracts. Non-market variables relate to factors such as regulatory requirements, politics, public relations, and the like. In recent times, these factors have been a source of greater concern for many generation companies than market variables.
The next step is to consider the potential relationships between market (and non-market) uncertainties. The reason is that the individual markets are influenced by many factors. The price of electricity, for example, will be influenced by the price of natural gas, to the extent that natural gas is the marginal fuel of choice. These relationships can be graphically represented by what is called an ?influence diagram,? an example of which is shown in Figure 3.
Figure 3 shows the anticipated relationship between future environmental regulations and their affect on the price of SO2 and NOx allowances, and the latter's affect on the market price of electricity. It also shows an influence between environmental regulations and the price of natural gas, and the effect of natural gas prices on the price of electricity.
While influence diagrams are not necessary to develop a decision model and determine a preferred compliance strategy, they are often an excellent way of structuring uncertainties. By using an influence diagram, the uncertainties can be better structured. For example, because of the influences affecting the future price of electricity, an overall probability distribution of that price at a given time would depend on the price of allowances and natural gas.2 This is shown on Figure 4.
In developing an environmental compliance strategy, it is the non-market uncertainties that are likely to be the most problematic. Among the many issues that a plant may contend with are further reductions in NOx and SO2, fine particulate requirements, mercury controls, regional haze provisions, and greenhouse gas limitations.
Figure 5 illustrates a scenario regarding carbon emissions. The nature of future regulation and their monetary impact is unknown. The plant's owner may believe that, owing to something similar to the Kyoto Protocol, a carbon tax is most likely to be imposed in the year 2006. The owner assigns a probability distribution to the imposition date, as well as a probability distribution for the amount of the tax itself. The plant's owner also believes that the imposition year is likely to affect the amount of the tax, reasoning that the sooner the tax was imposed, the smaller it would be. In that case, the probability distribution for the tax amount would be conditional on the year imposed, as shown in Figure 5.
Converting uncertainties into discrete events
Except in some special circumstances, the probability distributions shown in Figures 3 and 4 need to be converted to discrete events. For example, the probability distribution for the amount of a carbon tax might have been determined to have extremes of $0 and $100/ton. Since there are an infinite number of values in between, some way of converting that continuous distribution into several discrete events must be performed.
Fortunately, there are numerical techniques to do this that are standard in statistical and decision analysis software.3 Usually, a probability distribution will be converted into three or four discrete events. For example, the carbon tax year and amount might be converted as shown in Figure 6:
Another useful tool, which can determine and rank the most critical uncertainties, is the ?tornado diagram.? This eponymous diagram ranks, in order of their effect on present value, the different uncertainties. Tornado diagrams are useful in several ways. First, they can provide insight as to those uncertainties that matter most. In Figure 7, for example, the diagram shows that natural gas price uncertainty has the greatest effect on the coal plant's ENPV, while mercury (Hg) regulations matter the least. Thus, more attention should be focused on natural gas prices in the modeling process. Secondly, tornado diagrams can provide insights into potential modeling difficulties. For example, if electric price uncertainty shows little impact, that may indicate a flaw in the underlying strategy valuation model.
Constructing the decision tree and determining an appropriate compliance strategy
Having identified a set of decision alternatives, determined a value measure, and identified and developed discrete representations of key uncertainties, we next draw and solve a decision tree. A decision tree simply provides a schematic representation of a problem. Decision trees contain ?branches,? each of which describes a different path along a realized state-of-the-world. Complex decision trees can have thousands or even millions of different paths. This is why it is important to first determine the most critical uncertainties in order to focus on the variables most likely to influence the choice of strategy.
Figure 8 illustrates a simplified version of the decision tree, first depicted in Figure 2, which uses the term ?regulatory uncertainty? to represent the overall uncertainty of future allowance prices, the price of electricity, greenhouse gas regulations, etc.)
Once the decision tree is constructed, it is solved by working from the end of the tree towards the tree's ?root.? The solution reflects the expected value of the lowest-cost path. The solution will also provide the overall probability distribution of results, which will provide the plant owner with information about the risks of the highest ENPV strategy.
The solution is illustrated in Figure 9, which shows that a ?wait-and-see? approach is preferable, given the assumptions and facts in this case study. The first downstream decision in 2004 may appear to be a counter-intuitive: sell the plant if environmental regulations are severe and investing in new emissions controls if not.
Of course, it is critical to realize that downstream decisions are not ?cast in stone.? The primary objective of the analysis is to determine what ought to be done today, given the uncertainties of the future. This is the true strength of the decision analysis framework: it can easily incorporate future flexibility and calculate the value of that flexibility.
Lastly, Figure 10 illustrates the risk profile of the preferred alternative compared to the ?no risk? strategy of selling the plant today. While selling the plant today has no risk, the sale price is $13 million less than the expected value of the ?wait-and-see? strategy. That price difference represents a risk premium: if the plant owner accepts the purchase offer, he will, in effect, be paying $13 million to avoid the risks associated with keeping the generation plant.
Summary
In writing this report, we have sought to present a brief overview of decision analysis and its application to environmental compliance strategies. Decision analysis is a common-sense, logical approach to frame and clarify decisions in the presence of uncertainty. But it is neither a computer model that makes decisions nor a substitute for careful thinking. Rather, it is a way of identifying, structuring, and solving problems. Applying these tools to determine appropriate environmental compliance strategies doesn't mean that decisions will suddenly become easy and intuitive. Like all tools, decision analysis can be used well or poorly. There will always be risks that cannot be well-defined, assumptions that are wrong, and factors that are simply overlooked.
Given the uncertain future, companies wishing to develop environmental compliance strategies face inherent risks. The most thoroughly researched strategy can fail horribly. But the alternative ? approaching investment decisions in a haphazard and illogical manner ? will almost certainly be much worse.
Navigant Consulting has extensive knowledge, experience, and resources to help companies develop their environmental compliance strategies. Please contact Dr. Jonathan Lesser, Senior Managing Economist, at (802) 865-2121, Tom Mills, Director, at (202) 973-2941, or Andrew Greene, Principal, at (781) 564-9750, for more information on Navigant Consulting's capabilities.
1 One thing to note is that the decision tree necessarily simplifies the decision process. Clearly, the owner could make a decision today, change his mind tomorrow and make another decision, and so forth. While that might be more ?realistic,? it would significantly complicate the analysis without adding much value.
2 This results in what is called a ?conditional? probability distribution.
3 The formal name of the analytical technique most often used is ?gaussian quadrature.?
Request a Demo
Interested in IJGlobal? Request a demo to discuss a trial with a member of our team. Talk to the team to explore the value of our asset and transaction databases, our market-leading news, league tables and much more.