Multiple Approaches to Environmental Decisions

Douglas M. Brown

The Logistics Management Institute Bethesda, Maryland

I. THE IMPORTANCE OF DECISION MAKING

It would be difficult to overstate the importance of the environment as a policy issue. Aside from the ecological implications of decisions in many "nonenvironmental" policy fields, environmental policies have impacts on other policy fields. The recent controversy over whether protecting the spotted owl should weigh more or less heavily than protecting the jobs of timber industry workers is not going to be solved here. The important thing is to realize that environmental policies, often considered to be based on scientific analysis, must include consideration of nonscientific issues such as fiscal realities, economic growth policies, and cultural values. Even race has surfaced as an issue in this field [1],

Because of the weakness of the current state of the art in fundamental measurability of environmental policies, an appreciation of the impact of such policies can only be hinted at by using other proxy measures. While environmental activists prefer to see environmental protection as a universally superior good not subject to such comparisons, the fact is that protective activities incur costs. Whether they are continuing expenses or just investments that will result in lower costs later on is a matter of interest, but it does not relieve society of the obligation to pay the bills until the investments mature.

In the end, all policy costs are experienced by society's consumers and taxpayers. Individual firms, of course, can be punished with criminal sanctions or forced into bankruptcy over environmental breaches; but as a rule governments and entire sectors of industry simply pass the costs along in the form of coerced tax hikes or industry-wide price hikes. Thus, neither government or industry (in a wide sense) "pays" for environmental protection except to the extent that when consumers or taxpayers find themselves with no more money to spend, the popular taste for government or the industrial product may evaporate.

Those who believed that the 1970s' Great Society programs or the 1980s' defense buildups nearly achieved this national policy bankruptcy point should look closely at the environment. The cost to the taxpayer of dealing with environmental problems is expected to exceed 6% of the Gross National Product by mid-decade. That is the size of the entire defense budget during the peak of the Reagan buildup. Simply cleaning up known hazardous waste sites at federally owned facilities is expected to cost over $125 billion [2], with the figure being adjusted upwards every year. Newly recognized threats are being discussed that will only add to the potential size of this burden [3,4]. In addition, consumers will bear an additional burden. Some of this burden is hidden in the price of consumer goods, as those private-sector firms that continue in business rather than declaring bankruptcy either pay for required cleanups, self-insure against the need to clean up in the future, or develop new processes to avoid becoming a party to a future cleanup. In some cases, businesses will choose to go out of business rather than risk personal or corporate liabilities of staggering proportions. At the least, this will reduce the number of choices available to consumers, and at the worst, employees will be thrown out of work and further impact on other taxpayers.

All in all, there are compelling fiscal and social reasons for ensuring that our very real environmental problems are identified and dealt with in prudent and responsible ways. Money spent on the environment either directly by governments on behalf of consumers or indirectly by consumers through higher prices charged by producers as a result of regulations, cannot be spent on other worthy causes such as consumer and national savings and debt reduction, urban issues, transportation networks, and national security: whatever your policy preferences are, environmental spending competes with it. And, if not properly thought out, environmental policy can compete with itself. For instance, the EPA has spent years convincing the public that toxic waste sites are a tremendous source of health risks and must be dealt with promptly whatever the cost. The publication of the Unfinished Business report [4] requires the EPA to reeducate the public that in its new view many other threats are more risky; EPA competes for funding for those higher risks against an established, costly effort that the EPA itself established and plans to continue.

A micro version of the same argument can be made at the level of the individual producer facility. Statutory responsibility or not, an organization cannot devote so many resources to environmental protection that it can no longer afford to remain in business. Environmental actions, even where deemed socially worthy, must compete for funding with other programs, and where the available funding does not cover all perceived needs, then environmental spending itself must be prioritized. In short, for regulator, policy analyst, and facility manager, a sound basis for making environmental decisions is essential to the development and effective execution of a holistic, complex, and credible program for the protection of health and resources. While some may argue for a policy based strictly on scientific evidence, others argue for environmental policies based on emotion, and yet others argue that the costs of delay on the one hand and regulation on the other are socially destructive, environmental managers are faced with a situation where something has to be done that will satisfy all sides without bankruptcy. Thus, while decision theory is not the cornerstone of environmental science, it may well be the keystone of environmental management.

II. ENVIRONMENTAL ROLES

Environmental threats are produced and dealt with by organizations whose missions are broader than simply protection of the environment. Environmental agencies, however, have the mission of ensuring that producers do not forget their environmental responsibilities. Those responsibilities are to the third player in this process: the public. Through the political process, the public caused an environmental policy to be put in place to protect health and the environment, and at the grass-roots level the public maintains oversight on the specific actions of both reg-

Environmental Management Table 1 Policy Roles

Characteristic

Enforcer

Producer

Public

Policy viewpoint Single focus on narrow

Complete operation with Does not generally under-environment as part of stand or get involved, the whole.

portion of environmental issues.

Mission Prevent all pollution and punish all polluters.

Produce products and Seeks products and pay for environmental protection, protection.

Objectives Focus resources on worst threat.

Resources Cleanup costs do not detract from primary mission.

Approach Extract payment for pollution.

Policy effect Deters, does not repair, problem.

Cleanup costs are taken from funds otherwise available for mission.

Focus on avoidance, then cleanup.

Minimize resource drain. No

Avoids, evades, or repairs problem.

Pay in either case, through taxes or prices.

No perceived threat is acceptable.

Passivity until aroused; then paranoia.

Suffers consequences.

ulators and producers. The differences in the roles of regulators, producers, and the public are summarized in Table 1.

The most familiar enforcement agency is a police department. It has a single focus on a statutory area (in this case, public order). The primary responsibilities are (preferably) to deter crime or, when that fails, to seek out and apprehend criminals, which itself deters further crimes. In addition to fines levied through the punishment process, such departments may collect fees to recoup their costs of doing business, thereby reducing the burden on their budget (and, in theory, on the taxpayer).

Frequently, however, there is no payor, either because the guilty party has not been apprehended or because the enormity of the crime makes financial restitution, even with damages, unacceptable or so high as to be unpayable. Such cases, which are the norm more than the exception, make it necessary for the department to absorb the cost of enforcement; but such expenses are budgeted for and appropriated over and above the cost of normal operations, not at the expense of those operations. We expect them to deter and apprehend, not to repair, the problem of crime. In those few jurisdictions where victim compensation is considered, it does not come out of the police operating budget. It is not generally expected that the enforcer will have to police itself (although such occasions do arise, and have arisen in most of our major cities over the past 15 years, and are generally poorly handled).

In some jurisdictions, we see a cooperative enforcement approach [5]: "community policing" or the "cop on the beat," to match the environmental metaphor with a police example. Nonetheless, those cooperative approaches are part of an overall enforcement strategy; the regulator coaches the producer toward compliance rather than taking over the operating responsibility itself. Finally, we expect the police to focus on the worst problem—catching murderers rather than staking out shoplifters.

The producers have a completely different set of responsibilities, the foremost of which is the fact that they must continue on with their production task in order to survive; environmental issues are a secondary concern. When a cleanup does become necessary, the producer must pay, and its payment comes out of its normal operating expenses. To some degree, consumers will absorb some of this cost, but in general, passing on too much of the cost will simply drive the producer out of business (unless the producer happens to be a government agency). With limited discretionary funding available for environmental restoration, then, producers need to accomplish as much as possible with the resources they have.

Normally, government regulatory agencies act as enforcers. Once enforcement has occurred, however, one is faced with the need to restore the situation. Then, and especially in the case of the Superfund, the government becomes a "producer" and needs to act and think like one.

Finally, there is the public. The public tends to be easily excited over health and safety issues, although a much smaller (but more active) group maintains vigilance over non-human health and natural resources, and a very small group is both active in and knowledgeable about global ecology issues. In addition, the public is concerned about jobs, general economic issues, property values, and the quality of life in communities. Thus, the public concerns tend to be more diffuse than the single focus enjoyed by enforcers and producers. Because of that dif-fuseness, the public seldom speaks with a coherent voice, which makes it easier for activists and extremists on all sides of an issue to misrepresent or override the public will.

While the federal government's National Environmental Policy Act provides processes for public involvement, as do a number of state statutes, there is at present no real requirement to go along with public preferences as long as the pro forma requirements are met. Thus, on any given decision, the public can be and often is ignored. The more this happens, of course, the more the public comes to see the regulator as well as the producers as its enemy, especially as these parties will be on different sides at different times.

Given this disparity in roles, it may come as no surprise that there are different perspectives on what the general objectives of the environmental effort should be. I have identified seven primary goals; others may exist. While most of them appear desirable, at least in isolation, they are not all consistent. They are presented here in no particular order; indeed, the ordering process itself is one of the most important facets of the environmental decisionmaking process.

Risk. Eliminate risks to human health and the environment.

Cost. Engage in environmental projects that achieve organizational objectives without posing an unacceptable risk to the organization's economic competitiveness or viability. Time. Accomplish objectives rapidly, if in fact there are risks to the environment and especially to our health.

Acceptability. Satisfy public and organizational expectations. All environmental decisions, whether taken by public or private organizations, occur under the observation of a political structure and still remain consistent with the organization's own value system.1 Deterrence. Ensure that environmental offenders, and particularly the worst offenders, are caught and prosecuted.

Administration. Minimize the debate over the intent and application of environmental laws and regulations.

Practicality. Develop policy alternatives that are executable, compromising the ideal to maximize what can be accomplished.

If the budget is unlimited, decision making or prioritization is unnecessary, one simply does everything that is wanted as soon as it is feasible to do so. However, increasing sophistication in regulations and increasing effectiveness of environmental compliance efforts are

'This objective was titled "Politics" in earlier work [1]. Some managers object to the idea that "politics" intrudes on the pristine pursuit of public and ecological health but feel quite comfortable with the need to tolerate public value systems. "Acceptability" seems less threatening.

Environmental Management Table 2 Decision Approaches

Knowledge condition

Approach

Certainty

Universal (one-rule) solution; Multiple-rule solutions (e.g., cost-benefit)

Statistical uncertainty Static Dynamic Outcome uncertainty

Expected value Simulation Aspiration level Subjective decisions Minimax

Political allocation

Total uncertainty combining to produce requirements estimates for environmental work that are increasing exponentially. For federal agencies alone, working off existing project requirements and dealing with a number of pending interagency agreements with the EPA or state agencies could require three- to fivefold increases in current environmental spending, even assuming that the historical inflation of environmental costs can be controlled. But as soon as we say that some projects will have to be delayed to meet more reasonable funding expectations, the question of priorities emerges: which projects should be delayed, and why?

In short, under reasonable resource conditions, not all of these objectives can be accommodated in full simultaneously. A blind emphasis on speed usually wastes money. A total focus on scientifically assessed risk may be impractical if the necessary science is incomplete. Administrative simplicity may be translated into rigidity, resulting in the carrying out of regulations blindly, resulting in large costs with no appreciable improvement in the environment. And so on. It is in just such cases—multiple, conflicting objectives—that the use of decision methodologies is needed.

This chapter is not intended as a single reference for explaining the details of decision theory. Such texts are available—indeed required—in every college's management course work.2 This chapter proposes rather to explain why decision analysis is needed in environmental decisions as much as any other.

In general, decision-making methodologies are selected based on the degree of uncertainty surrounding a condition or situation; Table 2 shows that alignment.

A. Certainty

Under the condition of certainty, we know each of the outcomes; it is simply a matter of choosing the program that benefits us the most. In that case, there is a universally superior solution or decision rule. Even where such a rule is deemed to exist, it must be tested to determine its universality and the existence of underlying rule structures, which may require reversion to a more complex decision approach. For instance, the accepted rule may be that projects that reduce the most human health risks will take priority over all others. The real world presents us

2There are too many such texts to list, and each university has its own preferred texts. Each manager has, or should have, several. On my bookshelf, perhaps the most frequently used is Fleischer's Engineering Economy [6]. Others include Lapin's Quantitative Methods for Business Decisions [7], Quade's Analysis for Public Decisions [8], and Douglas's Managerial Economics [9].

III. DECISION-MAKING METHODOLOGIES

with resource constraints, not necessarily limited to constraints on ready operating funds: the total capacity of administrative, technical, industrial, capital, and labor resources forms an effective obstacle to unlimited activity levels. Given a clear goal and normal capacity constraints, the basic rule can be expected to meet a quick challenge: if two projects are equally effective in risk reduction but there are not enough resources to accomplish both, another criterion must be applied as a tie-breaker. And, as a result of limited resources, at some point the residual funding may be adequate only for a low-cost but also low-priority project while a higher priority but more expensive activity must be forgone.

Given a high degree of certainty, we know all potential cases and all potential outcomes (and the relationships between the two). Then rule-based decisions are the appropriate solution to ensure that the best possible outcome is achieved from each possible situation. The decision makers can incorporate multiple factors simply by adding more complex rules, but in a rule-based system the rules cannot be waived. Nor should there be any reason to do so, because all possible cases and outcomes can be predicted and the best course of action identified.

Generally, we do not have such perfect information. Nonetheless many environmental decisions are based entirely on methods that address certainty. This occurs for two reasons: the real information is unknown or the real information is too complex to be used in its full detail.

When the real information is unknown, decision makers should apply appropriate decision approaches under uncertainty, such as those displayed in Table 2. However, environmental decisions are frequently made by bureaucracies (governmental organizations or large industries) that do not subscribe to subjectivity or political acceptability as an explanation for how decisions were reached [10]. Thus, where the real information is too complex to be analyzed effectively (as is often the case), managers simplify it to its essentials [11]. If a mode will describe the behavior of a natural phenomenon adequately (e.g., saying that the prevailing wind is at 5 mph from the east, when in fact over the course of a year it blows from most directions at various speeds for some period of time), and if the resulting policies do not appear to be too badly flawed, the approach is validated in terms of the value of the time and effort required of the manager. Indeed, contrast this behavior with the warning to remain conscious of the value of perfect information issued later in this chapter.

Again, however, true certainty seldom prevails. And in acting "as if," we are simply making assumptions: taking a mode to be the only possible outcome rather than the most frequent one. The real situation may take one of four alternative forms, representing increasing degrees of uncertainty. Statistical uncertainty means that we have a good idea of what might happen and that we have some probabilistic statements of how, when, or how often. Technically speaking, this is called decision making under conditions of risk, but because of the extensive use of the word "risk" in the environmental sense of a threat, I have called it statistical uncertainty instead.

B. Statistical Uncertainty

When statistical uncertainty exists, we must use our statistical knowledge to make decisions "as if" the outcome were to be as predicted by the probabilities, with appropriate consideration for the fact that it might not. This is basically done using an expected value approach, which some of you may know as a weighted average approach.

Rule-based approaches become meaningless if we have uncertainty, because we do not know what situation exists or which rule should be applied. Despite the masses of data collected to support science's continuing assault on the mysteries of the ecosystem, we remain extremely uncertain in our understanding both of the system as a whole and of most of our individual "facts." There is actually nothing wrong with an uncertain situation; in fact, our economic system and our national security systems function tolerably well under admitted uncertainty. The Apollo missions reached the moon and returned safely under conditions of uncertainty. Decisions under uncertainty are quite possible: the recognition of uncertainty simply requires some adjustments to our decision pattern. Indeed, the bulk of decision theories address this uncertainty directly, either in factoring in the possibility that multiple outcomes may occur (through weighting) or through safety factors.

The more serious error is to conduct business under uncertain conditions as if there were no uncertainty. Unfortunately, in many cases, environmental policies have been devised and implemented in exactly that manner. For instance, one of the primary "rules" that dominated implementation of the Superfund program was the "worst-first" rule, under which priority of effort was given to the most hazardous known contaminated sites. This rule was implemented through a detailed process resulting in a numeric score that was deemed at the time to be somehow a measure of the site's risk. As detailed site investigations were pursued, it began to appear that in many cases the scores did not reflect risk at all [12]. Part of the reason for this is that the scoring system used is mathematically skewed so that the worst are probably not first [13]. Subsequently, the EPA issued a revision of its scoring systems, but EPA is refusing to rescore the sites now found on the National Priorities List. Although the reasons for that decision are more related to politics and face saving, it forms a good example of the fact that under uncertainty the desired or statistically most likely outcome may not in fact occur.

In addition to using expected value methodologies to overcome statistical uncertainty problems, the technique for calculating the value of perfect information is one with which all environmental decision makers should be familiar (at least in concept). The specific equation is not particularly relevant, because there is often no "variance" (one of the terms in the perfect information equation) in environmental data, which tend to be highly site-specific. But a great portion of the funds in environmental activities are expended on "just one more round of testing," a round that generally proves as inconclusive as the original round.

There are a number of activities that are worthwhile of themselves but very much subject to the question of perfect information. Facility managers are asked to comb large tracts of land looking for endangered species that might be there but often are not. Products are banned or consigned to expensive disposal programs because they might be dangerous. Although protective measures are laudatory, all environmental professionals (whether they are regulators, facility managers, or taking the public's perspective) need to be conscious of the fact that at some point enough is enough. Part of any action discussion should be an explicit understanding of what "enough" is, what (specifically) is expected to be gained by achieving "enough" as opposed to some lower level of effort, and what the costs and other implications of getting to "enough" will be.

C. Dynamic Uncertainty

Many managerial texts address uncertainty as if it remained constant, albeit unknown. In the environmental world, things do not remain the same. Technology advances or is discovered to be ineffective, regulatory requirements change, the ecosystem changes, and specific activities evolve. Additionally, many decision theory models assume a single decision, even if a protracted and complex one. Environmental compliance activities often do not fit this mold. They occur within a continually evolving process where the outcome at any point is dictated only in part by the objective facts of the original circumstance. As a result, more complex models than the basic decision tree are needed to deal with the statistical uncertainty faced by environmental managers.

As an example, compliance requirements for wastewater treatment facilities have changed only superficially over the past two decades. Therefore, techniques of static statistical analysis are useful when looking at a large population (the view enjoyed by the EPA of a large number of wastewater plants, for instance) because the baseline for such a view has been well established and changes slowly. And EPA-level decisions, because of the mission of regulation and enforcement, address how to regulate entire populations and how to scan those populations to identify recalcitrants.

For a single facility, the domain of most environmental managers, this aggregate information is of little value. Decisions are not made on how to comply; that has already been spelled out. Generally, they are not made on whether to comply. The question is how to maintain the operation in compliance with no deficiencies, or at least as few as are practical, in the face of any number of unpredictable events.

The ability to comply with regulatory requirements is only partially a result of the original adequacy of a plant's technical design (the only static aspect of the operation). Much more important are a myriad of specific circumstances arising, or not arising, on any given day. What is needed is a model of the entire system in operation, from water flow production to effluent disposal, so that each point where a problem might occur can be identified and dealt with.

Even where a violation may occur, it does not follow that an enforcement action will result. That too is a dynamic process, dependent in part on the plant's record from the past, on teambuilding efforts by the regulated facility staff, on the sheer coincidence of inspection scheduling on days when the facility does or does not suffer a reversal of fortune, and on whether or not the facility has implemented any of the improvements recommended by the regulators on their previous visits.

This situation, in which the probabilistic variables themselves vary over time, can only be addressed by simulation tools. Such tools can represent the running of a scenario over multiple iterations to represent the effects of the passage of time or to try out the effect of assigning different probabilities to variables. Thus, in addition to the decision trees and contingency tables often found in managerial textbooks, simulation techniques must be employed.

Generally, managerial texts restrict their discussion (if any) of simulations to the Monte Carlo technique. This does provide a very powerful tool. However, even in its most basic form it requires recomputation of known equations in a number of iterations, which effectively demands automated tools. As simulations are being applied to increasingly challenging problems in industry and commerce, the computational power of simulations is being enhanced with graphics to provide comprehensibility to problems and solutions that would otherwise be nothing but piles of computer printout.

For environmental purposes, which tend to address problems an order of magnitude more complex than industrial process modeling, dynamic simulations that incorporate visual effects are needed both to complete a reasonably accurate representation and to enable the functional manager to see what the computer is trying to communicate. It is also important to understand that elaborate graphical presentations, even though they may be very data-intensive, are not useful to a rational decision process unless they can be used to develop relationships among the data. High-end geographic information systems generally display information in multiple layers, making it accessible to intuitive analysis. They may provide a database management system that permits the display of selected information, but usually they have no capacity for mathematical analysis.

Simulation tools are very rarely found in environmental use, despite their obvious utility. One reason is the complexity of environmental issues, which often forces a long tool develop ment time. Also, the policy system under which simulations are developed (as with many other areas of environmental activity) allows for endless research and refinement rather than product fielding.

Another reason is that a circular logic is in place. There is understandable reluctance at the EPA to certify any software products that have not been thoroughly validated by the EPA. However, the experience of many producers with the EPA's unwillingness to accept any way but the EPA way has made them wary of attempting to present EPA with any new products or approaches, and warier still of investing in the creation of such new ways until the EPA approves the product for general use. As a result, managerial innovation is restricted.

The final reason is that for many environmental managers on the job today, the computer remains a fearsome tool. But the very complexity of environmental issues and the increasing cost of environmental solutions will soon make an effective decision-making process dependent on the ability to exploit the power of automation. More familiarity with the use, applications, and abuse of computers in general and simulations in particular is required of environmental professionals.

D. Outcome Uncertainty

Outcome uncertainty is technical talk that means basically we have no idea of whether, how, or when things might happen, although we may be able to make some guesses as to what the things that might happen are. Under outcome uncertainty, we are forced into more subjective decision-making processes. The "aspiration level" approach [6] is seldom acknowledged but frequently seen. In such an approach, not having confidence in predictions of what may happen, and finding the risk unacceptable, we decide tb guard against the worst possible event.

This is like the U.S. defense strategy: a nuclear attack by the former Soviet Union was always considered extremely unlikely, but the potential damage that such an attack could cause was so great that we spent enormous sums protecting ourselves as best we could from such an event, even though we knew that by doing so many other needs would simply have to go unaddressed. The present worst-first policy is also an aspiration-level policy: it assumes that we know little about costs and remedies of cleanups, but we believe that we have identified the worst situations and we are committed to removing those situations, beginning with the worst.

Even under such conditions, however, cost-benefit considerations are at work, although less obviously; they are squeezed in through the back door using the potential for public outrage as the vehicle. We accepted the cost of the defense program as necessary in concept and generally affordable. And our national policy is to save endangered species. However, where a human community must give up its current livelihood in order to save a species (as is threatened in the effort to preserve the spotted owl habitat), the regulation enters the political arena where the EPA may win or it may lose. The contest will be presented to the public on the one hand as preservation of the quality of human life in preference to unproven allegations of harm to what is only one of millions of species, and on the other hand as preservation of defenseless creatures against greed and callousness. The essence of this argument is the cost-effectiveness of the effort: is the public willing to pay the price for a particular environmental project?

Another method of dealing with uncertainty is subjective decision making using group processes. There are any number of approaches available, from public meetings and roundtables to expert opinions either as individual contributions or controlled through a Delphic process. An example of the weighting of subjective preferences is seen in the U.S. Department of Energy (DoE) approach to its overall capital facilities investment strategy [14]. Successful results (as proven by the subsequent accuracy of the predictions) have been experienced with using an even more structured approach known as the analytical hierarchy process (for instance, our work on community fiscal impact analysis [15]). The EPA has begun looking into that process in its negotiations with the U.S. Department of Energy over how DoE will conduct its cleanup operations [16]. The process in a nutshell is that stakeholders and/or experts come to agreement over the relative significance of factors taken two at a time; at the end of making many such comparisons, the relative significance of each of the factors can be mathematically arrayed into a numerical weight table.

Another perfectly valid approach under uncertainty is problem avoidance, an extreme form of the minimax principle in which we minimize the maximum cost without much concern for probable benefits. If the issues and the costs are imperfectly defined, or if the impact of the events (including enforcement) is seen as unlikely or highly arbitrary, the whole system may not be worth worrying about. It is probably cheaper to be dragged into court from time to time than to go around solving problems that may not exist. In our view, this is precisely the strategy in place among many producers today. We have named it "problem avoidance," although one could also characterize it as foot dragging or passive resistance.

E. Total Uncertainty

The final possible approach, under total uncertainty (if we have little confidence in the risk data or the cost data), is to simply allocate the available funds on some arbitrary or politically acceptable basis and hope for the best. That division may occur on a social basis or on a geographical basis, and there are some suggestions that this is exactly what is occurring, either deliberately or as a consequence of the problem avoidance strategy. However, using similar arguments to achieve the opposite result (as in the present enthusiasm over racial equity in environmental issues) is analytically no more pure than the original failure. The approach is not covered as fully as others in this text because such issues are not amenable to environmental management and because there is some evidence that, at least in the Superfund program, funds do flow to higher scoring sites [17]. However, the environmental manager must be aware that such a school of thought exists and that when skillfully used it can awaken powerful political pressures.

IV. HOW CERTAIN IS ENVIRONMENTAL SCIENCE?

The range of approaches noted in Section III varies according to the amount of certainty that is attached to the situation. Clearly, it varies or there would be no need for so many approaches. In this section, we briefly consider some of the issues that lend uncertainty to many cases.

Each of the major objectives to be accomplished is still subject to considerable controversy. Research needs to be applied in each of those areas if we are to move forward effectively in managing the environment and to move from the imaginary certainty of acting "as if" (when really acting under outcome uncertainty if not total uncertainty) to at least acting under statistical uncertainty. The following brief summary is provided here to allow the environmental manager a glimpse of all that we do not know.

A. Risk

The definition of chemicals as hazardous is done through a series of separate environmental regulations, and there is little comparability of threat. In many cases, two chemicals have overlapping effects levels: the question becomes whether it is worse to die immediately or to get lung cancer. In such cases, how do we establish "worse"? Even when we stick with a single chemical class, there is a great deal of doubt as to what risk, if any, a regulated material poses. While the chemical data are confusing and contradictory (see, e.g., reference 18, from which the original Hazard Ranking System (HRS) factors were derived, or reference 19), we cannot always rely on the accepted approximation systems: current measurement systems (the original and revised HRS) cannot be shown to approximate risk. In our earlier paper [13], we showed that the original HRS was mathematically inconsistent, leading not to the potential but to the actuality of score inversion. While the revised system corrected several glaring deficiencies of the original system, there are enough departures from the generally accepted threat-pathway-dose-receptor paradigm to raise doubts as to whether risk is related to cost under the revised scoring scheme.

B. Cost

Given a reasonably accurate description of the problem, a reasonably accurate cost estimate should be possible for the engineering cost of the remedy. A common criticism of current models is that they are usually off by 50% or more and usually understate the final cost. This is partly because of the lack of data points, partly a result of inaccurate input, and partly due to the continuing ballooning of legal intervention costs. However, it should be noted that a 50% error may be less than the current errors in risk estimation.

C. Time

Time has interesting impacts on environmental decision making that need to be explored. Time, for instance, is tied to risk issues. Given low risks now and high long-term risks, or moderate transient risks now, how do we choose? What about the tendency of pollution problems to expand over time?

Time has extensive impacts on the cost side of our decision model. What does the slope of the environmental learning curve look like? What is the real rate of environmental inflation, and can it be separated into its components: engineering, legal, procedural, and so on? Even more significantly, what would be the most appropriate way to integrate the concerns into a decision model, even if the answers were known?

D. Acceptability

A great deal of work remains to be done in the integration of legitimate political and public concerns into the decision process. If political distribution of benefits is a real show-stopping issue, how can that be integrated into a decision process in an open manner so that it can be assigned a proportionate role? Is the distribution of pollution problems consistent with industrialization and hence with population density, and does this ensure an acceptably proportionate share in the program? How do we distinguish between self-interested "NIMBY-ism" and valid public safety issues, or does it not really matter?

E. Deterrence

Generally, deterrence is achieved through a combination of other objectives: effective definition of the risk factors, approaches that maximize the benefits of compliance, and a smooth administrative process that makes apprehension more likely. A great deal of research is going on today on the question of the right amount of fees to charge polluters. Less research addresses the question of how to make producers stop polluting altogether and what the costs of checking on it are.

F. Administration

Research into the other bureaucratic goal—administrative smoothness—would be well served. The primary attraction of the "as-if" approaches are that they are very easy to administer. HRS, for instance, is self-operated by polluters or agency contractors; sites simply go on a list. No effort has to be expended by the agency on the messy questions of cost, time, risk assessment, and so on; all of that is up to the polluter. From the producers' perspective, problem avoidance has advantages administratively because one has to fund only whatever the agency requires, so there is little need for decision making and a good probability that the major expense of restoration can be spread out over an almost indefinite period of time. But the primary reason for such behavior is that the process required to do anything else is so convoluted that it is not worth the cost. The simple act of declaring that a site should not be listed as dangerous becomes a very expensive process. Research is needed to address the value of each of the existing steps of the process and to determine whether it is possible to design a less cumbersome administrative process with less reliance on litigation that still protects the interests of the agencies and the producers.

G. Practicalities

Practical problems include the capacity of the environmental industry to handle the workload, the availability of technology, the capability of project managers to handle projects above a certain size and speed, and the limited experience at all levels with actually executing many types of environmental activities.

There are plenty of other practical issues within the scope of today's feasible actions. Are existing contract vehicles adequate? To what standards should contractors be held, and how effectively can they be managed? What are the lessons learned from past programs, and how are they being disseminated? Answers to any of these questions would improve the process of planning and executing environmental cleanup.

V. DECISION APPROACHES AND ENVIRONMENTAL OBJECTIVES

We noted earlier that there are several possible objectives for environmental programs and the people and organizations charged with carrying out those programs. How do the approaches we have discussed meet those objectives?

Table 3 compares the major approaches in terms of the objectives of the environmental program. The table shows a plus sign where the approach enhances that objective, a minus sign where it detracts from an objective, and a zero where it has no effect.

The single-rule approach (exemplified by the Superfund worst-first rule) addresses the risk and acceptability issues by focusing on the worst cases. In ignoring certain other objectives, the single-rule case in particular engenders a number of practical problems. Most notably, ignoring cost issues results in a number of cases of self-protective evasion by responsible parties. In addition, an oversimplified approach ignores many other practical limitations, such as the capacity of the environmental industry to undertake many projects or the availability of reliable technology with which to do the work. And it ignores the value of time in terms of inflation, opportunity costs, and continued public exposure to hazards.

More recent policies (if not many efforts) have focused on maximizing risk reduction [20], a multiple-rule policy based on both risk and cost. Cost as a consideration at least forces indirect consideration of some unspecified issues. Effectively performed cost estimates will need to consider practicality and the value of time, as a minimum. Such policies will necessarily be

Table 3 Decisions and Environmental Objectives

Static

Single Multiple (expected Aspiration Subjective

Static

Single Multiple (expected Aspiration Subjective

Table 3 Decisions and Environmental Objectives

rule

rule

value)

Simulation

level

weighting

Maximin

Politics

Risk reduction

+

+

+

+

+

-

-

-

Cost

-

+

+

+

-

0

+

0

Time

-

+

+

+

-

0

+

Acceptability

+

0

0

+

+

+

0

+

Deterrence

+

+

0

+

+

0

Administration

+

+

-

0

+

-

-

Practicalities

+

+

+

-

+

+

0

more complex to administer and may, because of their complexity, be less acceptable. Indeed, in the chart shown in Table 3, the multiple-rule-based approach appears almost ideal, with many pluses and no minuses. But this assumes that the rules can be developed in such a way as to enhance the objective, in short, that we are operating with certainty, which as we have noted above is usually not the case at all.

Operating under statistical uncertainty, we find that a more detailed understanding of the variables should result in a more targeted approach to the problem to be solved; thus risk, cost, time, and practicality are well addressed. The added complexity of the approach makes it less acceptable until it proves its worth. Likewise, deterrence may be stronger after the approach has been proved but will initially be poorer than a straightforward policy until it proves effective. And of course such a policy will always be complicated to administer.

Introducing the dynamic aspect of uncertainty would be a policy disaster without the appropriate tools. While an academic case might be made for the power of formal automated modeling and simulation in enhancing almost every objective except administration, the lack of appeal of a massive computer printout (often the only form of output available) would make it untenable to either politicians or bureau executives. However, given the right tool (a powerful simulation engine with an attractive graphic interface), this method becomes very useful. The scoring in Table 3 reflects this latter case. The effectiveness of a good simulation can help to account for practical constraints and can maximize risk reductions while minimizing time and money spent. This also maximizes the policy's deterrent effect. Effective graphics can make the conclusions more acceptable to both politicians and the public, and a well-programmed tool will ease the administrative task by making the decision rationale clearer or perhaps providing a recommendation.

So, why is everybody not using such tools? As noted earlier, the EPA often refuses to endorse them, and wisely so because the state of the science is such that one can barely be said to be operating under statistical uncertainty. A case of outcome uncertainty is more often the case. Earlier, three typical approaches were noted under outcome uncertainty. The worst-first policy is an aspiration-level choice if we confess that not much is known about risks or that we care little for costs. This has the principal advantage of addressing the mission directly: every effort will be expended to meet it. This may also be well accepted until it starts to cost too much, which in the case of the environment it will. In fact, in every way, its strengths and weaknesses (as shown in Table 3) parallel those of the single-rule-based decision; indeed it differs only in that the single rule assumes that the outcome is known, whereas the aspiration level accepts the risk that the outcome may not transpire but feels it to be worth guarding against.

Environmental management today is taking a close look at subjective decision making. The principal vehicle for this in recent efforts has been the analytic hierarchy process (pairwise comparisons). In addition to being used with some success (in terms of acceptability to activists and regulators alike) at the state level, it is under development now between DoE and EPA. It maximizes participation early in the process by the groups that most delay action through legal and regulatory maneuvering. Because all parties have an opportunity to participate in the decision-making process, and a properly designed process will leave few completely disgruntled, all are co-opted into the solution. Those few who continue their challenges will have limited credibility given the ecumenical composition of those who have agreed to the solution. In addition, the process can create a rule-based framework with which to make decisions, giving the illusion of a quantitative universal solution under certainty.

This capability also presents the biggest trap presented by the method. As a rule, the process is conducted in conjunction with computer programs to present the questions, record the opinions, and perform the weighting calculations. The programs sell themselves with a pleasing and simple interface that encourages their use by novices, but the comparison structure must be set up very carefully. This has often been overlooked, with the result that the comparisons are inherently flawed.

Even where the analysis is properly performed, the illusion of quantified objectivity remains. Decisions by subjective processes necessarily mean that objective information is lacking or inadequate. The form placed on the structure simply provides a facade of quantifiability. In fact, however, the process is structuring a political decision, made in this case by consensus among experts or activists rather than duly elected or appointed authority. While there is nothing wrong with that, it is important to recognize it for the political process that it is. In short, subjective decisions maximize political acceptance and basically admit a lack of hard knowledge in other areas, especially the core objective.

Finally, within outcome uncertainty, there is the minimax approach, principally represented in environmental policy by problem avoidance. This approach is characterized by continually seeking new facts while making temporary investments to control the spread of definite risks. Since nothing is actually accomplished by this, the final cost of a solution is unaffected, and because of the tendency for environmental problems to diffuse themselves that cost may be much higher although the payment date is set further into the future. The primary positive outputs of delay are that it can be given an acceptable facade, through continuing studies (especially if nobody is concerned about costs), and it minimizes the opportunity cost to the organization, which is free to go about its business with minimal budgetary impact. In terms of our objective function, problem avoidance deals with neither risk nor effectiveness, but in a perverse sort of way it does keep an eye on costs, at least in current years, by simply refusing to do the needed work.

Finally, there is the straightforward political approach. As far as we know from the limited studies available, this is seldom seen (environmental equity authors might differ with that statement). An arbitrary decision by an administrator has the same effect. In such cases, the basis for the decision rules is very different from those we have noted so far and is unlikely to maximize any value except political acceptability.

VI. APPLICATIONS OF THE DECISION MODELS

In the remainder of this chapter we review five applications of decision theory to a single environmental problem: investment decisions. Those applications are the use of a single rule and the use of multiple rules under certainty, the use of an expected value approach in a dynamic statistical uncertainty mode, and the use of public opinion under uncertainty. However, the state of management science within the environmental field is such that only decisions under certainty can be reviewed in any detail: the other approaches are still being considered but do not as yet represent any official policy.

A. Certainty

The first cases to be analyzed use rules to solve the capital allocation question. The objective of this exercise is to conduct remediation work on a group of toxic waste sites that exceeds our ability to pay. One approach is to use a single rule: worst-first. Under that rule, all available resources will be focused on the site posing the worst risk to public health. Remaining funds will be applied to the next-worst site, and so on.

An alternative approach applies multiple rules, or at least a more complex rule. In this case, the rules are that all available funds will be devoted in sequential priority to the site where the risk reduction per dollar is the maximum. Once work on a site is started under this formula it must be finished. This rule is called the "results-first" rule [1]. The situation to be addressed is displayed in Table 4, where a third approach is also displayed for convenience. That approach, problem avoidance, is in a way a rule-based approach (do the minimum necessary for as long as possible) but, as was described earlier, is a manifestation of an approach under uncertainty.

1. Applying the Rules to a Hypothetical Problem

Table 4 shows a notional five-site universe given a budget of 100 money units (call them millions of dollars, if you will). It provides the agreed risk (in whatever units) posed by each site and the estimated cost to restore each. Additionally, the table shows the cost of "doing nothing" at each site: these are the costs of the minimum essential security measures, studies, monitoring, and so on (not to mention legal fees) that must be carried out even when no action is possible or desired.

Under the worst-first approach, only the site risk need be known. We would attempt to fund site A first; with luck, the sum of the studies over the next 12 years would result in a partial resolution of the problem, but it is more likely that at the end of the period we would have to spend another 800 units of money to execute the remedy (requiring several years to accomplish, at a rate of 100 units per year). Meantime, nothing else could have been accomplished because the balance of our funds each year (20 units) is insufficient to remediate the next site, B. Note also that remediation at either site C or E, which could have been completed with the residual funds, will not be addressed because these sites are not "worst first."

Table 4

Example of Approaches"

Project

Risk

Cleanup cost

Risk/$

Avoidance cost

A

100

800

0.12

80

B

70

35

2

8

C

60

20

3

3

Db

40

80

0.5

40

E

20

4

5

1

"Total budget available = 100/yr. bProject D is already under way. Source: Brown et al. [1],

"Total budget available = 100/yr. bProject D is already under way. Source: Brown et al. [1],

The table shows the risk reduction per dollar ratio, which is the measure needed for a results-first decision. This is quite simply the risk divided by the cost. There is one twist to this: project D was already started, and we assume that 40 money units has already been expended on it. That is, its total cost was originally greater than 80 units and so its ratio was even less than one-half. Unfortunately, we have a unique enough situation at D that simply keeping it alive through studies is up to one-half as expensive as finishing it up (a situation that occurs more frequently than one might think). Using results first, we see that the best risk reduction per dollar gives the highest ratio, making our order of preference E, C, B, D, and then A. With this preference, we see that within our budget we can complete projects E, C, and B. We also have 41 units left over to fund some of the work at A and D. By the end of the second year we would have completed all the sites except A, and we would be left with the same problem as in the worst-First case—the inability to restore A—except that while only being 2 years further behind at A (which had not really gone anywhere anyway), we have completed all our other work and in the process we may have learned something about how to deal with site A.

Problem avoidance would require us to fund studies at all sites and to finish D, since work there was already started and it is too much trouble to stop. In either case, we do not have enough money to do the full studies required at A, so we skimp along on what we do have after funding everything else. This does happen in practice, since the duration and scope of tests and studies are fairly arbitrary and highly negotiable. The result of this strategy is that site D, which offers one of the lowest returns in risk reduction per dollar as well as being a relatively low ranking site, is the only one cleaned up, because every year the entire budget is consumed in studies. After 4 years, we have spent as much in studies at site E as we will eventually spend in cleaning it up—another event that is not infrequent in reality. Witness the fact that most of the sites removed from the National Priorities List have not been the "worst" sites that should have been first, but rather sites that proved after years of expensive studies to have posed little risk.

2. Applying Real Data

Setting up Table 4 for illustrative purposes was easy enough but may be considered to be misleading. Would the superiority of a results-first approach be demonstrated with real data? An analysis was performed using the scoring data from 1600 sites contained in the EPA's National Priorities List Technical Data Files database, and estimated remediation costs were extracted from the 523 then-existing records of decision (RODs). The result was a joint database of 412 unique sites with cost data (some RODs are for operable units within sites or are updates of earlier RODs). These data are acknowledged to be less than perfect. Aside from the fact that the HRS score may not really be a measure of risk, in some cases the factor values may have been improperly assigned during the site assessment and scoring processes. The costs are quite fluid. More important, it is not always clear in the ROD whether the data for operable units are cumulative or separate (one can sometimes get enough of an impression to know that either is used on occasion), so that the decision to add to or replace the previously recorded cost is subject to error.

The model uses an arbitrary cleanup budget ranging from 1 to 3 billion dollars per year over a 30-year period, reflecting the reasonable expected range of Superfund appropriations (presently being scaled back from $2 billion). In addition, no more than an arbitrary limit of $5 million per year may be spent on any one site; this limitation was designed to prevent the model from miraculously fixing a complex site such as Commencement Bay or Rocky Mountain Arsenal in one year with a massive infusion of cash. In fact, the environmental program

Environmental Management Table 5 Results Using NPL Data

Worst-first Results-first Results ratio

Sites cleaned HRS points reduced Total risk reduction including time effect

has shown some of the same tendency to throw away money under overstimulation as was seen in the Defense Department in the 1980s: sometimes there is a limit to how much money one can spend wisely in a given period. Finally, the model recognizes the continuation of risk incurred by failure to remediate.

Because of the funding limit by time period, and because of the life-cycle risk approach, the model requires an iterative solution based on the initial decision rules. It is important for environmental managers to note, however, that this relatively complex calculation can be (and was, for this exercise) accommodated with plain old dBase III Plus rather than using a more complex simulation tool. The working of the model is described in more detail in reference I.

The simulation determined which projects were to be funded in each year using the one-rule and multiple-rule methodologies. The results are shown in Table 5. Clearly, a more sophisticated rule system that acknowledges the effect of realistic constraints (i.e., resource limitations) outperforms one that adheres more closely to the original goal.

One might argue that for the specific sites we have selected, the HRS scores do not represent risk and do not represent toxicity or time effects. However, the purpose of this analysis is not to show how many HRS points can be cleaned up or what strategic approach is superior. Rather, it is to point out that effective management depends on effective identification of objectives, requirements, resources, and constraints. The use of tools because they are simpler (such as a one-rule system) will often be less productive than the use of tools that more closely reflect real conditions. The manager's job is to identify the relevant aspects of any decision and to ensure that all aspects are properly considered.

B. Statistical Uncertainty

The rule-based system just described was fairly complex, even when operating with the assumption of certainty. Incorporating probabilities or ranges of values into such a model as the multiple-rule model would require automated systems support. There are many models in existence that assist in the solution of particular problems. There are few, however, that provide information to a manager before an event occurs. The EPA's CAMEO model may be the best known; designed for emergency response use, it can be used for "what-if" analysis under selected circumstances. (Contact the USEPA Office of Solid Waste and Emergency Response in Washington, D.C., for additional information on CAMEO.) Even with CAMEO, however, the situation needs to be specified quite closely.

Other "screening" models have been produced under various development initiatives by the EPA and by private vendors; those models generally address a specific pollution mode (air, soil, surface water, or groundwater) and usually address fate and transport issues rather than providing impact analyses. In addition, many site-specific models have been developed for individual environmental projects, and in a number of cases those models have been placed in the

5,608 69,989 12:1

public domain in case they may be found useful. In other words, there are literally hundreds of dynamic models available; some address a known situation over time, and some are much more

Was this article helpful?

0 0

Post a comment