Journal of Risk Research 6 (3), 213–231 (2003)
Risk-based regulation of technical risks: Lessons learnt from case studies in Switzerland THOMAS FLÜELER* Senior Research Associate, Swiss Federal Institute of Technology (ETH), Zurich, Chair of Environmental Sciences: Natural and Social Science Interface; Umweltrecherchen & -gutachten, Hausen AG, Switzerland HANSJÖRG SEILER Extraordinary Professor of Public Law, University of Lucerne, Judge at the Administrative Court of the Canton of Bern, Extraordinary Judge at the Swiss Federal Supreme Court, Münsingen, Switzerland Abstract In administrative law, various regulation strategies are used to limit technical risks. Frequently, their approach is deterministic. Risk-based regulation seeks to make law more efficient as well as more transparent. Its aim is to replace prescriptive, deterministic regulations by goal-oriented, probabilistic regulations, based on the criteria of cost-effectiveness and limitation of individual risks. The overall goal is to achieve more safety at less cost. The project “Risk-Based Regulation”, (19961999) was intended to evaluate the feasibility of the approach from both technical and legal perspectives in the Swiss context. Nine case studies were carried out: storage and management of explosives (both military and civil), occupational safety, non-occupational accident prevention (mainly road accidents), fire protection, transportation of dangerous goods, waste disposal (conventional toxic landfills and radioactive repositories), and nuclear (reactor) safety. This paper summarises final results (Seiler, 2000) of the case studies (see also Seiler, 1998, and Flüeler and Seiler, 1999) and draws general conclusions on the possibilities and limitations of implementing a risk-based approach. Its findings should be useful in formulating standardised approaches as envisaged by the European Commission under the heading of a “Compass for Risk Analysis” (EC-JRC, 2000), and the safety guidelines of the so-called “Swiss Agency for Technical Safety” (SATS), a new institution to be set up in Switzerland to integrate all federal regulatory bodies dealing with technical risks. Keywords: risk-based regulation, technical risks, probabilistic vs. deterministic approach, cost-effectiveness, individual risk, marginal cost *
Author to whom correspondence should be addressed. E-mail:
[email protected]
1
Journal of Risk Research 6 (3), 213–231 (2003)
1. Introduction There is little consistency across existing approaches to the management of risks that are posed by technical products and systems. Risk management legislation is based on various regulating strategies to limit technical risks, using general clauses such as “hazard” or “safety” and setting limits for emissions, immissions, concentrations, etc. References are made to the state of the art, technical experience, or procedural measures like environmental impact statements and audits. Traditional regulation frequently adopts a deterministic approach by prescribing precise measures whose execution is relatively simple: Thereby it can unambiguously be determined whether the set goals are reached or not. This, however, does not necessarily assure an optimum safety level. In addition, different technical fields are not homogeneously treated. The result of these mixed and often inconsistent approaches frequently is inefficient and inflexible risk management. Invested financial and other means do not result in optimum safety and protection of environment (Pildes and Sunstein, 1995). Such discrepancies have called for a more coherent and efficient strategy, a prominent requirement made by the economic analysis of law (cf. Graham, 1994, 1996, Graham and Wiener, 1995, Tengs et al. 1995). The substantive background is twofold: On the one hand there is increased cost consciousness in economics and politics in view of globalisation and competition, on the other hand, law tends, upon factual and public pressure, to regulate risks more strictly and to increase liability respectively. This is in line with the EC New Approach, 1985, and Global Approach, 1989 (Commission of the European Communities, 1999). Several recommendations intending harmonisation have recently been made. An example is the proposal of an “International Science, Risk and Public Policy Commission” (Muntzing, 1999) or an “International Risk and Governance Council” to be linked to OECD, respectively, or of a “UN Risk Assessment Panel” (German Advisory Council on Global Change, 1999). Finally, the envisaged European “Compass for Risk Analysis” is “to develop a standardised approach (‘template’) to risk assessment and risk management at a generic technical level applicable to different industrial sectors” and “to develop and operate an open reference system to provide guidance for its practical implementation and use” (EC-JRC, 2000). Whether we like it or not, society is seen to accept certain risks and, consequently, certain harm or damage. From a strict right-oriented perspective this is not tolerable but it corresponds to a utilitarian perspective which in real life, at least to some extent, is followed. Risk-based regulation (RBR) permits to analyse this utilitarian perspective systematically.
2. Risk-based regulation: concept and approach The risk-based approach aims at eliminating the deficiencies mentioned above by pursuing the subsequent basic ideas and key elements of risk management:
2
Journal of Risk Research 6 (3), 213–231 (2003)
• Goal- and performance-oriented: Law should not prescribe specific and punctual safety measures but determine quantitative maximum tolerable risk levels. As in the concept of New Public Management (cf. Osborne and Gaebler, 1992) targets are defined but not the way to reach them. Those subject to law are free by which means they intend to meet the requirements. It is the regulators’ duty to verify whether the risk limit requirements are met. Considerations on limitations follow in Section 3. • Standardised: RBR is able to accommodate a diverse set of heterogeneous risks, applying comparable and consistent criteria for determining the quantitative tolerable risk level in the technical fields where identical damage indicators may be applied. The differences between densely regulated risk fields with presumably high safety levels and areas with relatively low safety standards are to be reduced. Insecurities in law application and discrepancies in resource allocation shall be eliminated to the best possible ratio. The approach allows comparison of risks for identical entities due to different sources, e.g., for human health due to work or traffic risks, natural disasters or diseases. If properly implemented, in a transparent manner, risks due to stochastic events like accidents may be compared to risks from low-level impacts like the exposure to toxic substances. • Probabilistic: The approach acknowledges–and should be able to accommodate–the fact that risks can never be completely eliminated but only be optimised. It tries to limit them according to rational criteria. Indeed, traditional regulation also followed risk considerations but mostly not in a quantitative way and often not systematically. When following these basic elements, there has to be distinguished between the direct and the indirect application of RBR: • Directly applied, particular cases are analysed with regard to specific protection goals. The risk of the respective activity has to be quantified and compared to the predefined protection goal. Generally, this is only possible for special cases characterised by substantial hazards as well as economic importance. In these cases a fullscope risk analysis is recommended. For other installations it is a prerequisite to have standardised procedures. • An indirect application is recommended where the risk is either heavily dependent on actor behaviour and on badly quantifiable singular influences or where the expenditure to carry out a risk analysis is too great. Here RBR is followed in the sense of a general directive, an evaluation criterion for a prescriptive legal decree or as an economic sector’s commitment to reduce the average individual risks within its field of influence. The requirements are framed in a way that the result is in compliance with the protection goals. A simple case of indirect application is the control of speed limits instead of a–direct–risk analysis for every driver in every circumstance. The authors are aware of the fact that RBR has been proposed and carried out in many sectors, many countries and, in some cases, for two decades or more1. To our knowl-
1
Forerunners are United Kingdom: Health and Safety at Work Act of 1974, HSE, 1989 (cf. Cassidy, 1996). – Netherlands: Risk Premises (Directorate General, 1989), Major Hazards Decree of 1992 (cf. Ale et al., 1996). – USA: Executive Order 12291 of 1981 requiring net benefits if new norms should be estab-
3
Journal of Risk Research 6 (3), 213–231 (2003)
edge, though, the approach has so far not been rigorously and quantitatively analysed in various and different technical fields under a consistent set of technical and legal issues. The authors propose to determine protection goals by way of two criteria (Seiler, 1998, 1999, Flüeler and Seiler, 1999)2: • An absolute individual fatal risk limit is a risk threshold above which no individual may be exposed; it protects individuals from risks not tolerated by society. • A (marginal) cost-effectiveness criterion serves to limit the collective risk caused by a product or a technical system and to allocate resources in an economic optimum. It specifies whether, and to what extent, a risk should be reduced below the risk threshold above. It is only applied after the individual risk limit has been complied with, so economic efficiency is clearly subordinate to minimum equity requirements. Its basic idea is an economic optimisation. Risks are only tolerable if the benefit of the riskbearing activity exceeds the respective imposed risk. Since an integral cost-benefit analysis is very complex and often not feasible, instead, the cost-effectiveness of risk-reducing measures is assessed; risks due to an activity are to be lowered to the extent that the costs of the risk-reducing measures are below the thereby reduced (monetarized) risk (see Fig. 1). If the measures from left to right–along the curve–are implemented, the maximum reduction of risk for a given level of costs is achieved. In other words, a collective risk of a system is tolerable if at least all measures (or combinations thereof) are taken whose cost-effectiveness is smaller than or equal to the marginal costs3. Each technical system may be characterised by its typical risk reduction curve in the risk/cost diagram.
lished. (cf. Sunstein, 1996), EPA, 1987, Presidential/Congressional Commission, 1997 (cf. also Morgan et al., 2000, for other agencies’ activities). 2 The reasoning is remarkably in line with Morgan’s independently suggested ”hybrid strategy” which combines an individual ”equity threshold” with subsequent efficiency concerns (Morgan, 2000). 3 The term „marginal costs“ stands for the limiting or threshold value of cost-effectiveness to be determined for a specific system. In fact, the quantitative values reproduced in the Tables are utilised in the respective technical fields in Switzerland. The marginal costs for lethal risks is the maximum society is willing to pay for saving a human life.
4
Journal of Risk Research 6 (3), 213–231 (2003)
Collective risk (reduction) R
optimum package of safety measures marginal cost of risk reduction
Cost for safety measures C dC/dR
Fig. 1. When appraising the collective risk of an activity, the investment for safety has to be related to the risk reduction achieved. The basic idea of the marginal cost criterion is to select the most cost-effective safety measures. One proceeds with safety measures until a certain cost/risk ratio is attained. Those measures are compulsory whose cost-effectiveness is better than the marginal cost value curve. Example: Assume that the marginal cost value to prevent one fatality is 1 MUSD and two safety measure options exist: Measure A costs 10 MUSD/y and reduces the collective risk by 20 fatalities (ratio 10/20=0.5), measure B costs 5 MUSD/y and prevents 3 additional lethal cases (5/3=1.7). With regard to cost-effectiveness, measure A has to be implemented whereas measure B does not.
The cost-effectiveness criterion ought not to be confounded with risk-benefits considerations, since the aim is not whether or not to build or retain an installation but to evaluate certain additional safety measures, i.e., to get the maximum safety increase out of the available resources. The marginal cost approach is utilitarian; whereas for society it results in an optimum, it cannot avoid that some individuals must bear unduly high individual risks. From a basic rights perspective, this is offensive. To take this into account and to prevent clustering of collective risks, the marginal cost criterion is, as suggested, combined with an absolute individual risk limit. This perspective demonstrates the prevalence of protection over economic analysis4. Consequently, the risk limitation follows from two rules: Rule 1: The individual fatal risk resulting from a risk source and laid upon the most exposed individual has an upper limit of 10-x per year. Rule 2: In addition, the risk is to be reduced in so far as the costs of the risk minimising measures are lower than y monetary units per reduced risk unit. 4
In agreement with the U.S. Presidential/Congressional Commission on Risk Assessment: ”The Commission supports the use of economic analysis as a consideration, but not as the overriding determinant of risk management decisions …. We call for explicit descriptions of the assumptions, data sources, sources of uncertainty, and distribution of benefits and costs across society …” (Presidential/Congressional Commission, 1997).
5
Journal of Risk Research 6 (3), 213–231 (2003)
Several prerequisites have to be met to carry out this approach seriously: 1. Analytically, adequate methods as well as sufficient and reliable data have to be at hand (hopefully to compare various technical risks), the costs and the effectiveness of the risk-reducing measures have to be known. 2. Normatively, the legal framework has to permit the approach. The damage indicators as well as their weighting have to be defined (e.g., value of humans vs. values of natural reserves), the risk assessment methods have to be selected (risk limits: individual/collective risk and reference frame, marginal costs: monetarisation of protection goals). Often, injuries are more significant than fatalities, e.g., in occupational safety. We, therefore, propose to consider disabilities as “portions” of casualties, according to their degree of handicap: zero to 100% (100% representing death). Other lesions might be dealt with as economic damage (recovery cost and loss of wages). Since it is difficult to fulfil all the premisses, and since risk is one of many aspects in decision-making, it may be more appropriate to speak of “risk-informed” instead of “risk-based” approaches. We adhere to the term RBR, though, because its concept indeed is based on the probabilistic risk model, even if this has to be considerably enlarged compared to the traditional technical risk definition. In theory, the protection goals could be satisfactorily determined by the definition of both values x and y. This would not account for the fact that risks have to be appraised according to whether they are taken voluntarily or not, to the degree of available risk information, avoidability, influenceability or perceived benefit. This can be dealt with by differentiating values of x and y. In Switzerland, four such categories with varying values have been proposed and partly implemented5 (Merz et al., 1995): Category 1: voluntary risk exposition in order to satisfy one's own desires, e.g., dangerous sports Category 2: high degree of self-determination, direct individual benefit, e.g., car driving Category 3: low degree of self-determination, individual benefit, e.g., working conditions Category 4: involuntary, imposed risk exposition, no direct benefit, e.g., local residence of a dangerous installation Depending on the category we propose limits of 10-5 to 10-3 per year for individual fatal risks. As to risk-reducing measures, the human capital values for marginal costs are set to 1 to 20 million Swiss Francs (0.6 to 12.4 million US Dollars) per life saved accordingly. Special attention has to be given to the following aspects such as: definition of relevant valuation parameters (damage indicators, relative weights of different indicators), low
5
E.g., since the mid-1980s in the traffic sector and from the early 1980s on in the military explosives area.
6
Journal of Risk Research 6 (3), 213–231 (2003)
probability-high consequence risks, risk concentration with individuals, risk transfer from risk producers to risk bearers. Details follow in Section 3. The minimum project goal was to develop proposals if necessary for a more efficient legislation and regulation in each technical field of the respective case study. The maximum goal was to propose an overall concept for all technical risks under consideration.
3. Results and discussion 3.1. SWISS LAW NOT SYSTEMATICALLY RISK-BASED Overall, there is neither standardisation nor quantitative determination of safety criteria. However, although in some areas, the law in force has been for many years and is directly or indirectly risk-based like in the storage of military explosives field (Bienz and Niederhäuser, 2000) and in road transportation of dangerous goods (Stiefel and Vogt, 2000). A summary of the actual status of Swiss regulation in the technical areas under study is presented in Table 1.
3.2. METHODOLOGY WITH TWO CRITERIA PROVED TO BE VALUABLE If properly adapted to the respective technical risk, the RBR approach is theoretically convincing and practically feasible. Basically, analytically individual risk can be reliably quantified if large groups are involved (see Table 2). The determination of risk limits is a normative issue. Nevertheless, it may be guided by empirical arguments. The starting point is the age-dependent mortality risk every individual is exposed to. Its number is about 10-4 per year for teens of 10 to 15 years old, the age cohort with the lowest lethal risk. The philosophy requires not to unduly increase this natural lethal risk by technical risks. From this, a risk limit of 10-5 per year is derived. This value applies to risks caused by others (category 4). The more the risks may be attributed to the respective human victim, the lower the value may be. It is questionable whether a value is justified for category 1 at all. As opposed to the marginal cost approach, the individual limit deals with an equity-oriented protection of an individual from being sacrificed by society6.
6
Of course, to a certain extent law–and economy– strives for protection from self-damaging, for several reasons: Not everybody may be capable of securing an optimum personal risk prevention, damaging oneself may have negative effects on relatives, insurances or the public (public expenses for rescuing and recovery, loss of publicly created human capital), difficulties in distinguishing risk causation, self-induced or by third parties.
7
Journal of Risk Research 6 (3), 213–231 (2003)
Case study
Law Ordinance
Military explo- O sives
O
O
M
O
iR, M
Civil explosives Occupational safety
O Nonoccupational accident prevention
Fire protection Transportation of dangerous goods Landfills and contaminated sites
O, partly D O, partly D1
O, partly D1
dR
R, M
M, O, partly iR
M2, D1, dR3
Implementation
Directives, norms
dR
Directives are not universally applied.
D, M
Exceptions.
M/R
Directives probabilistic but do not replace older prescriptions. The great number of concerned and certainty in law require a measure-oriented execution but measures are risk-based. Probabilistic exemptions possible, but rare.
M
M
M
M, dR
M, dR3
M4, O/R5
Remarks
M, R
Radioactive re- O positories
R
R
Nuclear reactor O safety
M/R
M/R
Traffic safety probabilistically dealt with. Water Protection Law may be interpreted strictly deterministically. Execution in the waste field is largely deterministic, risk-based elements in the regulation of contaminated sites. Radioprotection is riskbased; targets are partly directly probabilistic. Case by case analysis, additional safety measures riskbased.
1
water protection, 2transport, 3emergency cases, 4waste, 5contaminated sites
D dR iR O M R
deterministic directly risk-based indirectly risk-based open, probabilistic approach neither required nor excluded measure-oriented risk-based
Table 1. Current regulation of various technical fields in Switzerland.
The implementation of the approach is not problematic if there is an identifiable population with an average age structure being equally exposed to a well-defined risk source. For instance, according to Merz and Thoma (2000) traffic measures for children are justified, though seniors over 80 are exposed to a higher individual death risk from traffic than 7-year-old children (5*10-4 per year vs. 10-4 per year, i.e., 5 times higher) on condition that the risk reference entity is life years and not lives. Why? Youngsters are ex-
8
Journal of Risk Research 6 (3), 213–231 (2003)
Risk category
Military explosives
Non-occupational accident prevention
1
Not determined
10-2-10-3 per year
300 %$
2
10-4 per year
10-3-10-4 per year
30 %
3
3 x 10-5 per year$$ or 5 x 10-5 per year$$$
10-4-10-5 per year
3%
4
10-5 per year
10-5-10-6 per year
0.3%
$
$
$
$
Percentage relates to the increase of the natural mortality probability dependent of the victim’s age (see $$ $$$ below), military staff, indirectly concerned persons Table 2. Individual risk limits in two technical fields as applied in Switzerland (from Bienz and Niederhäuser, 2000; Merz and Thoma, 2000 as well as bfu, 1988).
pected to live more years than older people do and, for them, the traffic share of their total death risk is 30 percent compared to less than one per cent with older people. (Of course, measures to protect elderly people may be suitable, too, but not strictly based on risk grounds.) Problems arise in the following circumstances: • The mechanism is more difficult the smaller and more specific the affected group is. The protection aimed at for highly exposed individuals is not achievable. • If the risk is not reliably quantifiable, it may be demonstrated that it does not exceed a certain limit. By this means, however, a strongly conservative element is built in. • If the individual risk results from multiple risk sources (e.g., road traffic) it remains unsolved who has to take risk-reducing measures. This must be determined via specific rules (cf. air protection action plans if immission limits are exceeded). In order not to violate the polluter pays principle, the polluter(s) would have to be liable for the necessary costs. • Where other than personal risks are decisive (e.g., environmental risks), absolute risk limits have to be set by other means than individual risk limits. As to the marginal cost criterion, the collective risks are for most cases well quantifiable, in some areas there exists ample experience (with explosives from the mid-1970’s, Bienz, 1994; in road traffic from the mid-1980’s, bfu, 1985). An economic perspective frequently defines marginal costs on the basis of a person’s human capital (category 1 in Table 3). From a strict economic viewpoint, this criterion would not depend on the respective category of voluntariness as introduced above because national economy suffers from the same loss regardless of suicide or homicide. But the value would surely depend on the economic position and “relevance” of the dead individual. A legal perspective, however, is the other way around, i.e., irrespective of their economic position; people are all attributed the same value following the equity principle. At the same time, voluntariness and risk attribution are to be taken into account.
9
Journal of Risk Research 6 (3), 213–231 (2003)
Hence, the determination of the marginal cost value is based on the average individual. This is deduced from equity and practicability arguments: Unlike in cases of liability, we do not deal with actual deceased but are confronted with a prevention ex ante, i.e., the victim or victims are not known yet. In some technical fields values have been defined and utilised as shown in Table 3.
Risk category
Military explosives
Non-occupational accident prevention
1
1 MCHF
2
4 MCHF
3 MCHF
3
4–10 MCHF
7 MCHF
4
20 MCHF
15 MCHF
MCHF = million Swiss Francs Table 3. Definition of marginal costs as implemented in two technical fields (from Bienz and Niederhäuser, 2000; Merz and Thoma, 2000 as well as bfu, 1988).
It is noteworthy that the gradation for the risk limit (Table 2) differs from the one for the marginal costs (Table 3): in the former case by a factor of 10 between category 2 and 4, in the latter by a factor of 5 respectively. The reason for this is the following: the individual risk limit reflects the protection of legal goods and not an economic optimisation. This protection has to be guaranteed, particularly when completely uninvolved third parties are threatened to be harmed. The above-mentioned traffic measures for children are not only justified from a risk point of view but they are also cost-effective (10 times more efficient than for seniors). This is because the “traffic risk share” of seniors over 80 years is comparatively low in relation to their total death risk, if we base it on life years instead of lives7. This example demonstrates the relevance of the selected reference. As long as the group in question corresponds to the average population, it is equivalent to use the number of lethal cases or lost life years as a reference. This premise is not valid if those concerned are significantly older or younger. This may be taken into account by determining risk limits not in absolute terms but in relation to the respective age-dependent total mortality risk (see Table 4).
7
Swiss traffic statistics show that the highest share of life years is lost by victims who are in the first half of their lives. To determine the life years lost one multiplies the number of lethal cases per age cohort with the sex-dependent average life expectancy (Merz and Thoma, 2000). This strategy follows a recommendation already given in 1985 (bfu, 1985).
10
Journal of Risk Research 6 (3), 213–231 (2003)
Risk category
Tolerable individual lethal risk as a share of the total mortality probability per year
1
300 %
2
30 %
3
3%
4
0.3%
Table 4. Variant of individual risk limit determination: percentage of total mortality probability instead of a maximum admissible fatal risk per year (from Merz and Thoma, 2000 as well as bfu, 1988).
Accordingly, the collective risk may be also based on life years if appropriate. The life values mentioned in Table 3 correspond to an average life of 40 years. The death of this person leads to a loss of, say, 40 life years if we assume an average life expectancy of 80 years. To obtain the marginal cost value for saving a human life year one has to divide the respective value of Table 3 by 40 (see Table 5).
Risk category
Marginal cost value for saving a human life year
1
25,000 CHF
2
75,000 – 100,000 CHF
3
100,000 − 250,000 CHF
4
375,000 − 500,000 CHF
CHF = Swiss Francs Table 5. Determining marginal costs per life year on an average life of 40 years and a life expectancy of 80 years (taking values from Table 3).
Major problems still exist as follows: • In some cases like nuclear reactor safety, the quantification of collective risks are bound to considerable uncertainties. Meaningful statements on absolute costeffectiveness are restricted (Schmocker et al., 2000). • Risks other than fatalities, like environmental risks, may be quantified where they are relatively small and well known (e.g., accidents concerning transportation of danger-
11
Journal of Risk Research 6 (3), 213–231 (2003)
ous goods). For conventional toxic landfills, contaminated sites (Flüeler, 2000), radioactive repositories (van Dorp, 2000) and nuclear reactor accidents (Schmocker et al., 2000) the environmental risks cannot be quantified with sufficient reliability. The characteristics of long-term and catastrophic risks make effectiveness very difficult to assess (Flüeler and van Dorp, 2000; Schmocker et al., 2000). • In some areas difficulties arise with regard to costs and the effectiveness of riskreducing measures and not so much due to the quantification of collective risks. • Generally, cost-effectiveness may only be applied in a well-defined system; if it is too small, results do not make sense, if it is too large, calculations become too complex. This has to be kept in mind when comparisons of different risks are sought. As hypothesised in Section 2, RBR may only be applied directly in special cases with considerable hazards and economic significance. As to other installations, an efficient direct application assumes standardised procedures. In all other cases the approach has to be indirect, i.e., applicable deterministic safety measures are based on risk considerations. An overview on the practicability of the approach is given in Table 6.
3.3. COMPARABILITY OF TECHNICAL FIELDS: SOME RESTRICTIONS In principle, individual risks may be compared irrespective of the type of risk source. Restrictions are given where: • risks other than individual lethal risks are decisive (environmental, catastrophic) • the risk specific reference frames are not comparable since individual risks always draw upon particular reference entities (per year, per average member of a group, per km driven, at exposition only once a year, etc.). If fire risk is looked at as an isolated risk, the safety standard for individuals might be reduced since it is orders of magnitude below the margin set by the project (10-5 to 10-3 per year). But if it is analysed as part of the professional accident risk, it may not be neglected (Guggenbühler et al., 2000). These reference problems may be solved by way of the marginal cost approach since this criterion does not rely on a direct comparative risk analysis. Qualitatively comparable are the risks in the following areas: military and civil explosives (Bienz, 2000; Bienz and Niederhäuser, 2000), occupational as well as non-occupational safety (Forsblom and Wilhelm, 2000; Merz and Thoma, 2000) and fire protection (Guggenbühler et al., 2000). There is no evidence why the same standardised safety criteria should not be applied to other risk areas with similar qualities like construction safety, electrical installations, railways, air and naval traffic as well as product safety. Different risks with identical damage indicators, like e.g., for humans, are normally comparable.
12
Journal of Risk Research 6 (3), 213–231 (2003)
Case study
Practicability of Practicability of individual risk marginal cost limit criterion criterion
Direct/indirect application
Good
Good
Transportation/troops: eventually indirect, other areas: direct Direct
Good
Good
Collectively good, high expenditure with individuals
Costs and effectiveness of measures difficult to assess
Rather indirect
Generally good
Indirect
Collectively good, high expenditure for single objects Of little relevance
Collectively good, high expenditure for single objects Good
Usually indirect, for special buildings: direct Partly direct, partly indirect
Of little relevance
No quantification
Landfills: direct and indirect, contaminated sites: indirect Direct and indirect
Military explosives Civil explosives
Occupational safety
Non-occupational Collectively good, accident preven- difficult with individuals tion Fire protection Transportation of dangerous goods Landfills and contaminated sites
Problems/ remarks
Problem of implementation: knowledge base of regulatory bodies. The justification of the application of risk category 3 is questioned in view of the different working agreements in practice. Problem of reference for individual risk.
No exact quantifica- Not analysed yet tion but indication of maximum risks Of little relevance Quantification of Direct and indirect Nuclear reactor levels** 1 and 2, (details in guidelines) safety level 3 too imprecise ** Levels refer to the degree of specification in safety analyses for nuclear power plants: Level 1 deals with core damage frequencies, level 2 with source terms and releases of radioactivity, and level 3 with the overall impact on man and environment (incl. terrain contamination and economic losses).
Radioactive repositories
Table 6. Practicability of both criteria: individual risk limit and marginal cost.
In the case of qualitatively heterogeneous risks it is more difficult to formulate standardised protection goals–due to technical and political reasons. This argument is valid where: • the damage types are distinctly different, e.g., human vs. environmental (cf. landfills and contaminated sites, Flüeler, 2000), where the relative weighting of single indicators is highly sensitive (e.g., transportation of dangerous goods, Stiefel and Vogt, 2000)–a limitation to certain proxy indicators is biased
13
Journal of Risk Research 6 (3), 213–231 (2003)
• different time horizons occur, e.g., immediate vs. long-term risks (cf. waste storage, Flüeler and van Dorp, 2000) • large events are significant, e.g., low probability-high consequence risks (cf. nuclear reactor safety, Schmocker et al., 2000) • the cost-effectiveness of specific areas cannot be quantified. An overview on the comparability of the different technical fields is indicated in Table 7.
3.4. IMPROVEMENT OF EFFICIENCY The efficiency of the technical safety law could be raised in areas with quantifiable risks. Usually, measures to improve safety are also economically advantageous. As a bonus, individual risks could be lowered in most cases. Presently, we cannot make a valid statement on the efficiency on an overall economic scale. At any rate, the individual risk limit is a “target norm” and leaves it up to the responsible parties who have to choose appropriate measures to reach the goal–this by itself is a more efficient concept than the traditional ones.
3.5. SUPPORT FOR PROPORTIONALITY, EQUALITY AND PRECAUTIONARY PRINCIPLES With RBR the legal maxims of equality and proportionality may be better comprehended and tackled more systematically than with the traditional deterministic approach. Its consequent application eliminates, for instance, the enormous discrepancies in risktaking among workers of different industrial branches. The principle of proportionality is directly related to cost-effectiveness: Law should not prescribe measures if the costs associated with them are not in reasonable proportion to their benefit. RBR reduces existing different safety requirements for enterprises which, nowadays, are causing distortion in competition. RBR does not prescribe detailed safety measures but overall goals, which allow more flexibility to the implementers. A transparent risk assessment may be a useful instrument for decision-makers in priority setting and for adequately ascribing value to the options available in the fields under scrutiny. The fact that problems, restrictions and limitations are addressed shows that a thorough risk-based approach does not contradict the precautionary principle. In fact: “The precautionary principle should be considered within a structured approach to the analysis of risk which comprises of three elements: risk assessment, risk management, risk communication”, according to a recent statement of the Commission of the European Communities (Commission, 2000). The endeavour clearly is after casting light on existing evidence, eventual gaps in knowledge and scientific uncertainties. This is particularly true where subjective instead of frequentistic probability is involved, an aspect that can only be dealt with in RBR. This “may provide a basis for triggering a decision to invoke the precautionary principle” (ibid.). If risk analysis is utilised in such a manner, it may free itself from its reputation of legitimising and curative instead of preventive intent.
14
Journal of Risk Research 6 (3), 213–231 (2003)
Case study
Damages analysed Death
Military explosives Death
Remarks Dominating for externalised damages. Besides, there may exist considerable (unquantified) collateral damage which are largely internalised. Ditto.
Civil explosives Death or injury
Dominating for externalised damages. Besides, there may exist largely internalised collateral damage.
Death or injury
In traffic there is usually considerable collateral damage not considered presently (because such damage is insured to a great extent). Dominating for externalised damages (in the cases analysed). Besides, there may exist (unquantified) collateral damage (which is largely internalised). In the case studies analysed, all relevant damages are probably considered but it is important how the damages for water systems are monetarized.
Occupational safety Non-occupational accident prevention
Death
Fire protection
Transportation of dangerous goods Landfills and contaminated sites Radioactive repositories Nuclear reactor safety
Death and water systems (cost of remediation and compensating supply of drinking water) Ecological damages
Death
Death, delayed harms, terrain contamination, economic losses
Not quantified.
According to the dominant view, ecological damages are irrelevant if the protection goals for human individuals are complied with. The extensive damages to be looked at are not quantified. The proxy indicators, core damage or source term, do not permit reliable quantification of the finally resulting damage.
Table 7. For comparability of different technical fields damage indicators and their respective dominance have to be analysed.
Democratic and legal steering mechanisms may be improved with RBR. Analyses with reliably quantified risks and adequate reference frames enhance transparency because they open up the political debate and do not leave decisions completely to the administrative and expert bodies, as many traditional regulation concepts do. A direct application, though, instead of measure-oriented regulations, may enlarge the freedom of interpretation exerted by the regulatory bodies. Therefore, direct application of RBR has to be supported by measure-oriented guidelines and standardised methods for adequate risk assessments. It is not conceivable that the approach, as such, leads to a technocratic usurpation of political decisions. RBR is set up to be more than the plain public health approach, i.e., just human deaths or injuries are not enough to fully evaluate risks–the damage indica-
15
Journal of Risk Research 6 (3), 213–231 (2003)
tors have to be diversified. We concede that, in practice, risk analyses are often limited to well documented aspects. This can result in spectacular errors if well-documented risks are calculated to be low whereas other risks, with poor data quality, are high. Such sensitive situations may not be utilised as a fundamental criticism of RBR but demonstrate that the approach has to be carried out with caution. By no means, the concept should be (mis)used to diminish safety standards, e.g., by reducing (too high a) safety level in one risk area without raising it respectively in technical fields with lowlevel standards. This sometimes is the case with the public health approach where environmental impacts are perceived as minor risks compared to impacts on human beings.
3.6. ATTITUDES OF RELEVANT SWISS STAKEHOLDER GROUPS TOWARDS RISK-BASED REGULATION: OPENNESS AND SCEPTICISM The question of acceptance or non-acceptance of RBR was explicitly addressed in a subproject (Wilhelm, 2000). Various relevant stakeholder groups in Switzerland were interviewed, in an exploratory, non-representative survey to elicit paradigmatic views: judges and lawyers, politicians, scientists, representatives of environmental organisations, regulatory bodies as well as commerce, trade and industry. In a comparison with the U.S. (Tal, 1997) it is noteworthy that among Swiss stakeholders in the risk debate there has not been such a vigorous discussion on the introduction of formal procedures for risk assessment and management. Consequently, the pro and con positions have not crystallised out to now. Reservations, however, are identifiable, not only among the green movement, especially with regard to methodology and participation. Risk analyses ought not to be instrumentalized to feign accuracy in areas with no solid grounds. The risk context cannot be overcome with simple and simplified parameters. As to procedural aspects it has to be guaranteed to integrate early, depending on the field in question, the relevant societal groups and to secure on a political level the respective decisions.
3.7. SOME SPECIFICS In applicable fields, (e.g., explosives) the two-fold criterion approach has proved valuable. If a potentially affected population is very young or very old, absolute risk criteria do not make sense. More appropriately, they should be related to life years. In case of uncertainties, statistics should be carefully analysed in order not to violate the precautionary principle (appropriate use of mean, median, standard deviation, etc.). Considerable long-term risks with delayed potential damages, like from waste deposits, are not to be discounted because, doing so, we would favour the preferences of the present generations and, consequently, externalise damages to the debit of future generations8. As for
8
There exists extensive literature on discounting, a topic beyond the scope of this contribution. Discounting or not is a matter of debate. Some economists are in favour of discounting (Paté-Cornell, 1983). If the
16
Journal of Risk Research 6 (3), 213–231 (2003)
large events, damage categories and indicators other than death have to be considered (system damages, vulnerability; e.g., land contamination as a result of a large nuclear reactor accident). Here also, political and legislative decisions have to be taken.
4. Conclusions Apart from some case study recommendations for changes in specific laws–given in the case study reports referring to Swiss legislation–, we point out the following generic aspects, based on the findings in the specific technical fields, that should be pursued: • Technical risks are to be, and can be, assessed in a consistent and co-ordinated manner. • For comparable risks, homogeneous safety criteria have to be, and can be, determined. • In many technical fields, these criteria have to be specified with measure-based regulations like directives since technical complexity, the involvement of many parties, and restricted regulatory resources only allow an indirect application. • The safety criteria serve as a scale for the selection of measures and the deduction of secondary protection goals (e.g., limits). • The safety criteria are applicable in areas where the law only states undefined legal terms and where there is not a specifying regulation on the ordinance level, either. Otherwise, the respective laws and ordinances ought to be modified. The safety criteria may be laid down in a law (which would assure maximum public participation, Wilhelm, 2000). In an initial phase, however, to really get the process started, it may be preferable to set them in an ordinance or a directive to gain experience and to modify the approach before fixing it in a law (we have, hence, reached a status between our initial minimum and maximum goals). The mere definition of safety criteria evidently does by no means assure compliance with them. Their application in the specific technical fields is more challenging for all involved parties–regulators, experts, specialised institutes, professional associations, risk producers and bearers–than the simple implementation of traditional measureoriented procedures. To facilitate this task, tools and methods for risk analyses have to be developed and made available in a manageable and practical form (with appropriate indicators, measuring units, discussion of sensitivities, normative decisions, etc.). The findings should be useful in formulating the safety guidelines of the so-called “Swiss Agency for Technical Safety” (SATS), a combination of all federal regulatory bodies dealing with technical risks, to be established in Switzerland. They could also be a valuable input in the “Compass for Risk Analysis” mentioned, the project currently being set up by the European Commission.
utilisation of environmental goods is judged to be a benefit, discounting promotes the consumption of non-renewable environmental goods (EPA, 1990), thus violating the principle of sustainability.
17
Journal of Risk Research 6 (3), 213–231 (2003)
To make it tangible, in the Swiss context, we suggest the following procedure (which might be carried out in various countries, according to Egli Steffen, 2000 and Majer, 2000): 1. The Federal Government commissions the ministries mainly involved in safety aspects to create an inter-ministerial working group. Its task is to elaborate a guideline along the concept displayed above. The group is made up of representatives from the departments with safety-relevant duties and from involved sciences. 2. The working group proposes a draft and puts it up to a discussion with scientists, economists and the public. 3. The ministries involved file an application to the Government for passing a respective guideline. 4. Finally, an inter-ministerial office for the limitation of risks is established to pursue the topic and to accompany and evaluate the subsequent regulation by the specific departments or offices. In the case of Switzerland, such a position could be founded under the authority of the mentioned “Swiss Agency for Technical Safety”. Risk-based regulation is not a substitute for political value judgements; it is not a substitute for politics and policy but a methodical tool into which political judgements may be explicitly incorporated. Adequately implemented, RBR provides a clear separation of analytical results and political value judgements. It makes transparent what risks exist and which risks are–implicitly or explicitly–accepted. And ultimately, it forces politicians (and the public) to make–or to come up with–value judgements on the tolerability or non-tolerability of risks. Nevertheless, even if sufficient quantification is lacking, probabilistic risk analyses can provide a good risk system understanding and are useful for the identification of weak spots and (potential) problems in the respective fields. They explicitly address types of uncertainty (variability, data deviation, model and scenario issues). As such, they are a systematic decision tool for a transparent risk appraisal. These findings are as important as the precise definition of quantitative protection goals.
Acknowledgements The project “risk-based regulation” (1996 – 1999) was sponsored by: Swiss National Science Foundation under grant no. 1113-52163.97/1, Swiss Federal Department of Defense DDPS; Swiss Council for Accident Prevention (bfu); Swiss National Accident Insurance Organisation (Suva); Swiss Federal Nuclear Safety Inspectorate (HSK); (German) Gesellschaft für Anlagen- und Reaktorsicherheit GRS; Swiss National Cooperative for the Disposal of Radioactive Waste NAGRA; Bienz, Kummer & Partner AG; Gruner AG; Institute of Safety and Security and “Stiftung Risiko-Dialog”. We are indebted to our 21 colleagues for their scientific contributions in the technical fields and their interdisciplinary co-operation. We are grateful to the anonymous reviewers for their observations and suggestions. We wish to thank Ph. Tipping (HSK) for language editing.
18
Journal of Risk Research 6 (3), 213–231 (2003)
References Ale, B.J.M., Laheji, G.M.H. and Uijt de Haag, P.H.M. (1996) Zoning instruments for major accident prevention, in: C. Cacciabue and I.A. Papazoglou (eds) Probabilistic safety assessment and management ’96. Proc. ESREL’96–PSAM-III, Crete, Greece, June 24-28, 1996, London: Springer, pp. 1911-1916. bfu, Swiss Council for Accident Prevention (1985) Beurteilung des Unfallgeschehens aus der Sicht des individuellen und kollektiven Risikos. Grundlagen für ein Sicherheitskonzept im Strassenverkehr [Evaluation of accidents with regard to individual and collective risks. Fundamentals of a safety concept in road traffic], Berne: bfu. bfu, Swiss Council for Accident Prevention (1988) Praktische Grundlagen für ein Sicherheitskonzept im Strassenverkehr [Practical grounds for a safety concept in road traffic], Berne: bfu. Bienz, A.F. (1994) Swiss safety concept. Regulations and organisation in the field of military explosives safety. 5th International Symposium on Explosives Technology, NIXT ’94, Pretoria, RSA, 12-14 Oct 1994. Pretoria: National Institute for Explosives Technology (NIXT). Bienz, A.F. (2000) Risikobasierte Sicherheitsbeurteilung von zivilen Sprengstofflagern [Risk-based safety assessment of non-military storages of explosives], Berne: Stämpfli. Bienz, A.F. and Niederhäuser, F.R. (2000) Sicherheitskonzept für den Umgang mit Munition und Explosivstoffen in Armee und Militärverwaltung [Safety concept for the management of ammunition and explosives in the Armed Forces and in military administration], Berne: Stämpfli. Cassidy, K. (1996) Approaches to the risk assessment and control of major industrial chemical and other related hazards in the UK, International Journal of Environment and Pollution, 6(4-6), 361-387. Commission of the European Communities (1999) New Approach and Global Approach & Legislation. Guide to the Implementation of Directives based on New Approach and Global Approach. See http://europa.eu.int/comm/enterprise/newapproach/standardization/document/1999_ 1982_en.pdf. Commission of the European Communities (2000) Communication from the Commission on the precautionary principle. See http://europa.eu.int/comm/off/com/health_consumer/precaution_en.pdf. Directorate General for Environmental Protection at the Ministry of Housing, Physical Planning and Environment (1989) Premises for risk management. Risk limits in the context of environmental policy. Annex to the Dutch Environmental Policy Plan “Kiezen of Verliezen” (to Choose or to Lose). Second Chamber of the States General. 1988-89 session, 21137. No. 1-2 (no. 5). EC-JRC (2000) Draft consensus report: Proposal towards a project on the development of a Compass for Risk Analysis. Based on the results of the EC-JRC International Workshop on Promotion of Technical Harmonisation on Risk-Based Decision Making, Stresa/Ispra, Italy, 22-25 May, Ispra: European Communities-Joint Research Centre (workshop under http://mahbsrv.jrc.it/stresa/Risk-Workshop-2000.html). Egli Steffen, M. (2000) Risk Informed Regulation – Impressionen aus dem Ausland [Risk-informed regulation–Impressions from abroad], Berne: Stämpfli.
19
Journal of Risk Research 6 (3), 213–231 (2003)
EPA, U.S. Environmental Protection Agency (1987) Unfinished business: A comparative assessment of environmental problems. Overview report, Washington, D.C.: Office of Policy Analysis. EPA Science Advisory Board (1990) Reducing Risks: Setting Priorities and Strategies for Environmental Protection, Washington D.C.: Office of Policy Analysis. Flüeler, T. (2000) Konventionell-toxische Abfälle: Deponien und Altlasten [Conventional toxic waste: landfills and contaminated sites], in Flüeler, T. and van Dorp, F., Part C. Flüeler, T. and Seiler, H. (1999) Risk-based regulation: A suitable concept to legislate and regulate technical risks? Evaluation of various case studies in Switzerland, in L.H.J. Goossens (ed) Risk Analysis: Facing the New Millennium. Proc. SRA-Europe 9th Annual Conference, Rotterdam, Oct 10-13. Delft: Delft University Press, pp. 593-597. Flüeler, T. and van Dorp, F. (2000) Radioaktive und nichtradioaktive Abfälle. Teil A: Gemeinsamkeiten und Unterschiede der Bereiche radioaktiv – nichtradioaktiv. Teil B: Endlager für radioaktive Abfälle. Teil C: Konventionell-toxische Abfälle: Deponien und Altlasten [Radioactive and non-radioactive waste. Part A: Common and different characteristics of radioactive and conventional toxic wastes. Part B: Repositories for radioactive waste. Part C: Conventional toxic waste: landfills and contaminated sites], Berne: Stämpfli. Forsblom, U. and Wilhelm, U. (2000) Sicherheit am Arbeitsplatz [Safety in working places], Berne: Stämpfli. German Advisory Council on Global Change (1999) World in transition: Strategies for managing global environmental risks. Annual Report 1998, Springer: Berlin, pp. 310311. See www.wbgu.de. Graham, J.D. (1994) The risk not reduced, Environmental Law Journal 3, 382-404. Graham, J.D. (1996) Making sense of risk. An agenda for Congress. In Hahn, R.W. (ed) Risks, costs, and lives saved. Getting better results from regulation, New York/Oxford: Oxford Univ. Press, pp. 183-207. Graham, J.D. and Wiener J.B. (1995) Risk vs. risk. Tradeoffs in protecting health and environment, Harvard: Harvard University Press. Guggenbühler, M., Respondek, J., Beck, E. and Häfelfinger, Chr. (2000) Brandschutz [Fire protection], Berne: Stämpfli. HSE, Health and Safety Executive (1987) Risk criteria for land use planning in the vicinity of major industrial hazards, London: HMSO. Majer, D. (2000) Exkurs EG-Recht [Excursus: Regulation within the European Communities], in Seiler, H. (2000), appendix 5. Merz, H. and Thoma, J. (2000): Sicherheit im nichtberuflichen Bereich des Strassenverkehrs, Sports, Haushalts und der Freizeit [Safety in the non-professional areas of road traffic, sports, household and leisure], Berne: Stämpfli. Merz, H., Schneider , T. and Bohnenblust, H. (1995): Bewertung von technischen Risiken. Beiträge zur Strukturierung und zum Stand der Kenntnisse. Modelle zur Bewertung von Todesfallrisiken [Valuation of technical risks. Contributions to structuring and to the state of the art. Models to valuate fatal risks]. Polyprojekt Risiko und Sicherheit. Dokumente. Nr. 3. Zurich: vdf-Hochschulverlag an der ETH. Morgan, M. Granger (2000) Risk management should be about efficiency and equity, Environmental Science & Technology 34(1), 32A-34A.
20
Journal of Risk Research 6 (3), 213–231 (2003)
Morgan, M. Granger et al. (2000) Categorizing risks for risk ranking, Risk Analysis 20(1), 49-58. Muntzing, L. Manning (1999) Legal framework and the use of risk in public policy decisions, Forum Engelberg 10th Conference, “Risk and Safety of Technical Systems– in View of Profound Changes”, Engelberg, 23-26 Mar, Switzerland, 15 pp. (manuscript). Osborne, D. and Gaebler T. (1992) Reinventing government. How the entrepreneurial spirit is transforming the public sector, Reading: Addison-Wesley. Paté-Cornell, M.E. (1983). Discounting in risk analysis: Capital vs. human safety. Ontario: Institute for Risk Research. Pildes, R.H. and Sunstein C.R. (1995) Reinventing the regulatory state, The University of Chicago Law Review 62, Winter 1995, 1-129, esp. 53passim. Presidential/Congressional Commission on Risk Assessment and Risk Management (1997) Risk assessment and risk management in regulatory decision-making. Final report, Vol. 2. See www.riskworld.com. Schmocker, U., Meyer, P. and Köberlein, K. (2000) Kernenergie: Reaktorsicherheit [Nuclear energy: reactor safety], Berne: Stämpfli. Seiler, H. (1998) Risk-based regulation: A convenient concept for legislation and regulation in the field of technical risks?, in P. Hubert and C. Mays (eds) Risk Analysis: Opening the Process. Proc. SRA-Europe 8th Annual Conference, Paris, Oct 11-14. Vol. 2. Fontenay-aux-Roses: Institut de Sûreté et de Protection Nucléaire (IPSN), pp. 1177-1187. Seiler, H. (1999) Risk based regulation: A legal view, International Journal of Global Energy Issues 12(1-6), 180-195. Seiler, H. (2000) Risikobasiertes Recht. Wieviel Sicherheit wollen wir? Abschlussbericht des Nationalfonds-Projekts “Risk Based Regulation - ein taugliches Konzept für das Sicherheitsrecht?” [Risk-based regulation. How much safety do we want? Final report of the Swiss National Science Foundation project “Risk-based regulation–a suitable concept for safety law?”], Berne: Stämpfli. Stiefel, U. and Vogt, S. (2000) Transport gefährlicher Güter [Transportation of dangerous goods], Berne: Stämpfli. Sunstein, C.R. (1996) Congress, constitutional moments, and the cost-benefit state, Stanford Law Review 48, 247-309. Tal, A. (1997) Assessing the environmental movement’s attitudes toward risk assessment. Environmental Science & Technology 31(10), 470A-476A. Tengs, T.O. et al. (1995) Five-hundred life-saving interventions and their costeffectiveness, Risk Analysis 15(3), 369-390. van Dorp, F. (2000) Endlager für radioaktive Abfälle [Repositories for radioactive waste], in Flüeler, T. and van Dorp, F., Part B. Wilhelm, U. (2000): Gedanken zur politischen Akzeptanz des risikobasierten Ansatzes. Resultate aus Interviews mit verschiedenen gesellschaftlichen Akteuren [Thoughts on the political acceptance of the risk-based approach. Results of interviews with different political stakeholders], Berne: Stämpfli.
21
