1 Sven Ove Hansson Policy Forum, NewS Falkenberg 2001-04-04 Applying the Precautionary Principle to Persistent and Bioaccumulating Substances The purpose of this talk is to clarify the relationship between science and policy in applications of the precautionary principle in regulation of chemicals that are persistent and/or liable to bioaccumulate. What are the questions for scientists? What are the questions for policy-makers? Where and how do they meet? I will begin by trying to identify what the precautionary principle is, then counter some of the most common objections to the principle, discuss in general terms some ways in which it can be operationalized, and then finally propose how it can be operationalized in the regulation of substances that are a potential danger to the environment. This is a report of joint research with my graduate student Per Sandin. 1. What is the Precautionary Principle? The fact that the phrase „the precautionary principle‟ is used with the definite article indicates that there is assumed to be only one, unique precautionary principle. Unfortunately, there is no consensus as to what this principle means. However, most formulations of the principle have a basic structure in 2 common, or at least they have certain basic elements in common. Consider this version of the precautionary principle: It is mandatory to limit, regulate, or prevent potentially dangerous act- ions before scientific proof is established. We find here four elements that can also be found in most other versions of the principle, namely: (1) a threat, expressed in the phrase „potentially dangerous technologies‟; (2) a uncertainty, expressed in the phrase „even before scientific proof is established‟; (3) a action, expressed in the phrase „regulate‟; (4) a prescription dimension, expressed in the phrase „should‟. If you systematically study the available formulations of the precautionary principle - as Per Sandin has done - you will find that the differences between them can be described in terms of how these four elements are expressed. We can describe these as the four central dimensions of the concept. The term "dimension" is adequate since the phrases used for of each of the four elements can be ordered according to whether they contribute to making the overall formulation of the principle stronger or weaker. Hence, most phrases expressing the threat dimension state, indicate the severity of the threat that is required to trigger the principle. Examples of such phrases expressing different degrees of severity are „serious or irreversible damage‟ and „non-negligible harm‟. 3 Typical phrases used to express the uncertainty element are: „lack of full scientific certainty‟ and „before a causal link has been established by absolutely clear scientific evidence‟. Clearly, in order for precaution to be relevant, there has to be some sort of uncertainty. If deterministic knowledge of a threat were available, we would talk of „prevention‟ rather than „precaution‟. The uncertainty dimension states how (scientifically) plausible a threat must be in order to trigger precaution. Hence, the threat and uncertainty dimension jointly indicate when the precautionary principle should be applied. The remaining two dimensions tell us how to apply it. The action dimension concerns the intended response to the threat. Phrases used for this include „cost effective measures to prevent environmental degradation‟ and, merely, „preventive measures‟. Phrases expressing the action dimension are typically not very specific. The phrases in the prescription dimension state, finally, tell us what the normative status of the action is. The phrases used for this dimension are typically much weaker in international agreements than in statements of the principles proposed e.g. by individual researchers. Hence, the Rio Declaration only states that, if the threat is severe enough, lack of full scientific certainty shall not be used as a reason for postponing cost-effective measures to prevent environmental degradation. It prescribes no action. Unfortunately, many of the existing formulations of the precautionary principle are ridden with imprecision. Operationalisation is necessary before the principle can be used. This, however, is not unique. 4 2. Objections to the Precautionary Principle Gail Charnley, previous president of the Society for Risk Analysis, has repeatedly attacked the precautionary principle in words such as the following: ". It‟s about religion. In one corner, we have risk analysis–the practice of using science to draw conclusions about the likelihood that something will happen– and in the other corner, we have the belief that instead of science, the precautionary principle will somehow solve all our problems.” She is not the only critic. A survey of the literature shows that critics have declared that the precautionary principle... ... is ill-defined ... is absolutist ... leads to increased risk-taking ... is a value judgement or an “ideology” , and ... is unscientific or marginalises the role of science . Is it ill-defined? Is it "too vague to serve as a regulatory standard”, as one critic said? Well, basically the answer is yes. On the other hand this lack of specificity is in not unique to the precautionary principle. The same objection can be raised against many other decision rules including the rule “perform only those risk reductions that are scientifically justified”, that critics seem to prefer. More importantly, this is a deficiency that can be remedied. I will come back later to how this can be done. Is the Precautionary Principle absolutist? Some critics seems to interpret it as being insensitive to scientific facts about the probabilities 5 associated with different risks. Hence, it is claimed, the precautionary principle would require us to prohibit everything that might be dangerous. Hence, the National Association of Swedish Fishermen, in their comment on the suggested new Swedish Environmental Code, held that the precautionary principle as applied to fisheries would mean that no fishing at all could be undertaken. Such arguments are clearly based on a misconstruction of the precautionary principle. The principle requires that actions be taken in certain cases in the absence of full scientific certainty. This, however, does not mean that precautionary measures are required when there is no particular evidence, scientific or other, of the presence of a possible hazard. Indeed, we have not been able to find any authoritative formulation or interpretation of the principle that supports such an extreme requirement. To the contrary, some documents e.g. from the European Union explicitly demand that the possibility of harm at least should be indentified. Does the Precautionary Principle lead to increased risk-taking? This has been proposed to be the effect of taking when precautionary measures against the use of pesticides in developing countries. Pesticides may be a threat to the environment, and, it can be argued that for reasons of precaution, they should not be used. But that would in some cases lead to an increased risk of crop failure and consequently of famine which, for reasons of precaution, should be avoided. More generally speaking, precautionary measures impose costs, and they might therefore in the end lead to worse effects than if they had not been carried out. However, although there is a problem here, does not depend on the precautionary principle itself but on the limited framing of the decision problem to which it is applied. When delineating a decision problem, one has 6 to draw the line somewhere, and determine a “horizon” for the decision . If the horizon is too narrow, then decisions will be recommended that are suboptimal in a wider perspective, and this applies irrespective of what decision rule is being used. If we apply any decision rule to an isolated issue, then the decision may very well be different from what it would have been if we had applied the same decision rule to a more widely defined decision. The precautionary principle does not differ from other decision rules in this respect. Is the Precautionary Principle a value judgement? Some critics have argued that it is a value judgement or an “ideology”, not a factual judgement. It is claimed that the precautionary principle merely expresses a subjective attitude of fear against risk taking, and therefore can neither be confirmed nor falsified by scientific studies. Since science only deals with factual truths, not subjective attitudes towards risk taking, the precautionary principle simply leaves no room for a scientific approach to risk analysis. Or so it is said. In order to appraise this argument, let us return to the a feature of the precautionary principle, namely that precautions are required in the absence of full scientific evidence. Hence, according to the precautionary principle, the level of evidence at which precautions should be taken is situated below the level of full scientific evidence. This is clearly a value judgement. However, let us consider the alternative standpoint, that the level at which precautions should be taken coincides with the level of full scientific evidence. This is to no less degree a value judgement. Hence, to the extent that the charge “It is a value judgment” holds against the precautionary principle, it also holds against its rivals. 7 Finally, is the Precautionary Principle unscientific? According to some critics, the precautionary principle fails to pay enough respect to science, since it requires that precautionary measures be taken also against threats for which full scientific evidence has not been established. For instance, the use of a chemical substance can be prohibited by the precautionary principle even if we do not know whether it (say) threatens the aquatic environment or not. In spite of its convincing first appearances, this argument breaks down as soon as sufficient attention is paid to its key term „unscientific‟. There are two meanings of this word. A statement is unscientific in what we may call the weak sense if it is not based on science. It is unscientific in what we may call the strong sense if it contradicts science. Creationism is unscientific in the strong sense. Your aesthetic judgments are unscientific in the weak but presumably not in the strong sense. The precautionary principle is certainly unscientific in the weak sense, but then so are all decision rules – including the rule that equates the evidence required for practical measures against a possible hazard with the evidence required for scientific proof that the hazard exists. On the other hand, the precautionary principle is not unscientific in the strong sense. A decision-maker who applies the precautionary principle will use the same type of scientific evidence, and assign the same relative weights to different kinds of evidence, as a decision-maker who requires full scientific evidence before actions are taken. The difference lies in the amount of such evidence that they require before they decide to act against a possible hazard. The scientific part of the process, i.e. the production and interpretation of scientific evidence, does not differ between the two decision-makers. This shows that the precautionary principle does not contradict science, and also that it does not marginalize science as a tool in decision-making. 8 3. Making the Precautionary Principle Operative throuch the Default Approach One important approach to the operationalization of the precautionary principle is to assign appropriate default values. For intrascientific purposes, the default value when we have insufficient information is “unknown”. If we are not reasonably certain of the fate of a substance in the environment or of its effects on fish, then we say that it has an unknown fate and that its effects on fish are unknown. For decision-making purposes, this is not sufficient. We have to treat the substance in one way or the other while waiting for more evidence. Hence, a substance with unknown toxicity may be treated as if it were severely toxic, as if it was moderately toxic, as if it were non-toxic, etc. This has the effect of assigning a default value of toxicity to the substance. The traditional approach has been to treat food additives with unknown properties in the same way as highly toxic such additives. (The positive list approach corresponds to the default “highly toxic”.) For general industrial substances, the opposite approach (the negative list approach) has been used. Needless to say, the latter approach is not compatible with the precautionary principle. However, the precautionary principle is compatible not only with the positive list approach, but also with other, intermediate approaches. It is essential to observe that in a science-based application of the precautionary principle, all default assumptions must be defeasible, i.e. they must be liable to adjustment when more scientific information is obtained. 9 4. Applying the Precautionary Principle to Persistent and Bioaccumulating Substances Finally, let us see what all this means in the regulation of chemicals that are persistent or liable to bioaccumulate. For that purpose, I am going to use a simple formalized framework. We assume that the objects of regulation are chemical substances or mixtures. Furthermore, we assume that the outcome of regulation consists in placing these substances in a number of regulatory categories and that the regulatory category into which a substance is placed depends on a number of objective properties that the substance has. Given these assumptions, we can treat the regulation of a substance as a decision based on a vector (n-tuple) <x1,…xn>, where each element of the vector represents an objective property of the substance, such as its LD-50 in rats, its vapour pressure, etc. We will call this a property vector. According to our assumptions, the values of the property variables should uniquely determine the regulatory category into which the substance is placed. In mathematical words, there should be a function f from the set property vectors to the set of regulatory categories, such that for each property vector <x1,…xn>, f(<x1,…xn>) is the regulatory category to which a substance belongs if it has the properties represented by this vector. Each of the variables in the property vector can take the value “not known” (denoted Ø). Clearly, the way in which f treats property vectors that contain Ø is essential for the application of the precautionary principle. In practice this means that rules for default values have to be determined, such as “if the substance has unknown toxicity, treat it in the same way as non-toxic substances”, “if the substance is persistent and bioaccumulating but has unknown toxicity, treat it as moderately toxic”, etc. 10 In order to arrive at a regulatory model of this kind, both scientific issues and policy issues must be dealt with. There are four steps in this process. These four steps are analytically distinct, but they need not be taken in successive order. They are: 1. Determining the general framework of regulation. 2. Choosing and specifying the variables of the function. 3. Choosing the function. 4. Choosing default values to be used when the true values of the variables cannot be determined due to lack of data. Clearly, the first of these steps is a task for policy-makers. With respect to the second step, it is necessary to distinguish between those property variables that represent inherent properties of the substance and those that relate to actual usage and to economic or social factors. Whereas the choice of the latter category of property variables is largely a policy choice, the former category should be chosen on the basis of scientific knowledge and scientific judgement. In other words, it is on the basis of scientists that we should determine whether or not, bioaccumulation, vapour pressure, toxicity in zebrafish, etc., should be used as a basis for decision-making. The third step, that of choosing the function that takes us from property vectors to regulatory categories, clearly has both scientific and policy-based components. The scientific task consists in judging the relative weights of the different property variables, and determining the “overall severity” of different combinations of properties. The policy task consists in deciding what types of measure to take against chemical hazards with these different degrees of 11 severity. I will come back later to how the scientific and the regulatory subtask can be separated in this case. The fourth step, that of determining default values (i.e. how to treat vectors containing Ø) also requires the combination of scientific and policy- related considerations. As I have already said, the second of the four tasks, namely the choice of property variables to represent the inherent properties of a substance is a scientific task. Some of these variables should represent the toxicity of the substance, whereas others should represent its fate in the environment and, in particular, the degree to which organisms will be exposed to it if it is released into the environment. The latter category are the more controversial ones. There are several problems involved in choosing and specifying the variables. In particular, we need to know to what extent the variables determine the degree of severity of the hazard, and to what degree they can be operationalised. In order to get an overview of experts' opinions, we interviewed seven Swedish scientists from different fields relevant to the risk management of chemicals, asking four questions about the choice of variables for exopsure potential and environmental fate: 1. Is the degree of persistence of a substance a good predictor of its being hazardous to the environment? 2. Is the bioaccumulative potential of a substance a good predictor of its being hazardous to the environment? 3. Are there other predictors that might be relevant? 4. Is any of the criteria more relevant than others? 12 We found a surprising lack of consensus. These experts turned out to be in almost complete disagreement about the usefulness of the property variables that are presently in use, namely persistence and bioaccumulation. Several of them mentioned other properties that could be used. We have here a case of scientific disagreement in an issue that is highly relevant to important policy decisions on environmental protection. Furthermore, this is a disagreement that seems to be possible to resolve through scientific research. It would seem reasonable, therefore, to direct significant research efforts at the issue: Which are the best predictors of environmental fate and exposure potential? Let us now turn to the choice of the function f, i.e. the function that takes us from a property vector (identifying the relevant properties of a substance) to a regulatory category (specifying the rules for production and use of the substance). The task of determining this function can be divided into two subtasks. First, given the properties of the substance, the degree to which it threatens the environment should be determined. Secondly, given that degree of threat, the appropriate regulatory action should be chosen. The first of these is a scientific task and the second a regulatory task. In mathematical terms, this amounts to treating the function f as composed of two functions: f(<x1,…xn>) = f2(f1(<x1,…xn>)) where f1 takes us from properties of the substance to its predicted degree of environmental hazardousness, and f2 from the degree of hazardousness to the chosen regulatory category. Please note that this is a mathematical model of the process, used in order to express its nature with sufficient precision. Clearly, neither 13 toxicologists nor - in particular - regulators will in practice think in terms of developing the values of a function. In discussing default values I will, for simplicity, assume that the information to be used about a substance can be reduced to three property variables p, b, and t, that represent its degrees of persistence, bioaccumulation, and toxicity. This is of course a highly simplified picture; I use it only in order to bring out some of the basic principles more clearly. We have to deal, then, with property vectors of the form <p,b,t>. Each of these three variables, we assume, can take either a nonnegative number or Ø as a value. Furthermore, we assume that p = 0 signifies that the substance is not persistent at all, that b = 0 means that it is not at all bioaccumulating and t = 0 that it is not at all toxic. The crucial issue in the application of the precautionary principle is of course how to treat property vectors that contain Ø. Consider for instance a substance which is known to be toxic and bioaccumulating, but whose degree of persistence is unknown. This corresponds to a vector <Ø,b,t> with b>0 and t>0. One possibility is of course to treat such substances – while waiting for information about their persistence – as non-persistent, i.e. to introduce the default rule f(<Ø,b,t> = f(<0,b,t>) whenever b>0 and t>0. Another possibility is to select some number p', other than 0, such that f(<Ø,b,t> = f(<p',b,t>) whenever b>0 and t>0. 14 Clearly, the degree of precaution represented by such a default rule will increase as the value of p' is increased. Much of the debate on policies referring to persistent and bioaccumulating substances has referred to the choice of a default value for substances with unknown toxicity (in our notation, t = Ø). Possibly, there may be more reluctance to assign default values above 0 to toxicity than to the other two variables discussed here. If a substance is toxic to animals, but its degrees of persistence and bioaccumulation are unknown (i.e. the situation is that represented by <Ø,Ø,t>, with t>0) , then it would probably not be treated for the time being as a substance known not to be persistent and not be be bioaccumulating. In other words, a default rule such as f(<Ø,Ø,t>) = f(<0,0,t>) would not be held to be reasonable. However, if a substance is persistent and bioaccumulating, but nothing is known about its toxicity (hence <p,b,Ø>, with p>0 and b>0), then it will in some regulatory contexts be treated as nontoxic until toxicity information arrives (i.e., the default rule f(<p,b,Ø>) = f(<p,b,0>) is applied). We believe that in a precautionary approach, default values above 0 for toxicity may have to be introduced, in other words, it may be reasonable to treat substances with unknown toxicity in the same way as substances with some known degree of toxicity, rather than in the same way as non-toxic substances. However, it is important to note that such default rules can, in a science-based approach, only be applied until the lacking scientific information arrives. Again using our mathematical framework, we propose that a default rule such as f(<p,b,Ø>) = f(<p,b,t'>), with t'> 0 15 may reasonably be applied. However, this does not mean that toxicity should be ignored when it is known. In other words, it would be difficult to defend system that had the property f(<p,b,t1>) = f(<p,b,t2>) for all p, b, t1 and t2 What approaches, then, can be taken to assigning a default value for toxicity? We have of course the traditional method of treating substances with unknown toxicity as non-toxic substances. In our present framework, this approach corresponds to setting f(<p,b,Ø>) = f(<p,b,0>). By the decision-theoretically naïve this is sometimes called "sound science". This procedure is not compatible with the precautionary principle, since it conflates “no scientific evidence of toxicity” with “scientific evidence of no toxicity”. Another approach is to treat substances with unknown toxicity in the same way as highly toxic substances. This is the method used in regulating food additives, pesticides, and medical drugs. In our present framework the use of this method would correspond to setting f(<p,b,Ø>) = f(<p,b,t'>) for some very high t', so that a substance with unknown toxicity is treated as highly toxic until empirical evidence about its toxicity becomes available. A third approach is to choose an intermediate default value based on statistical expectations. In other words, substances with unknown toxicity are assigned a default value that corresponds to the average toxicity of previously tested substances. If the substance belongs to a chemical group in which reasonably many substances have been tested, then the average in that group should be used instead of the general average. If trends of toxicity are known within the group (i.e. correlations between toxicity and chemical characteristics), then these trends can be used to set a default value (QSAR). 16 From a decision-theoretical point of view, the use of statistically determined default values represents risk-neutrality. i.e. it corresponds to expected utility maximisation. The traditional method, that of treating substances with unknown toxicity in the same way as non-toxic substances, represents a risk- seeking approach. The choice of a default value is a policy task, but one for which extensive communication with scientists is required. By developing alternative default rules with different degrees of cautiousness (including the risk-neutral that I just mentioned), scientists can contribute to making this a well-informed decision. Needless to say, default values should be replaced by empirical values as soon as they become available. The purpose of the precautionary principle is not to replace science but to help us deal with the many situations in which science cannot provide the answer.