Economic assessment of food safety standards: Costs and benefits of alternative approaches

Food Control 21 (2010) 1611–1619

Contents lists available at ScienceDirect

Food Control journal homepage: www.elsevier.com/locate/foodcont

Economic assessment of food safety standards: Costs and benefits of alternative approaches W. Bruce Traill a,*, Ariane Koenig b,c a

Professor of Food Economics, University of Reading, P.O. Box 237, Reading RG6 6AR, UK Associate Fellow, James Martin Institute for Science and Civilization, University of Oxford, UK c Conseillère du Recteur, University of Luxembourg, 162A Avenue de la Faïencerie, L-1511 Luxembourg, Luxembourg b

a r t i c l e

i n f o

Article history: Received 9 October 2008 Received in revised form 11 June 2009 Accepted 20 June 2009

Keywords: Food safety economics Impact assessment Cost–benefit QALY

a b s t r a c t This article provides an overview of economic methods to measure costs and benefits related to food safety issues. After an introduction on general economic principles, including the distinction between social and private costs and benefits, the article highlights the various methods for calculation of costs and benefits, including ‘‘willingness to pay”, amongst others. Particular attention is paid to the ‘‘quality-adjusted life years” (QALY) method for quantitatively expressing health impacts. The practice of Regulatory Impact Assessments as carried out by the UK authorities is explored in more detail as an example of cost–benefit analysis of regulatory measures. The applicability of the approaches to the various stages of the SAFE FOODS model is highlighted. Ó 2009 Elsevier Ltd. All rights reserved.

1. Introduction Formal assessment of the economic impacts of regulation is an increasing component of governments’ desire for evidence-based policy making. In particular, formal consideration of economic issues, quantification of these where possible, and publication of findings and their underlying assumptions are vital to transparent and accountable policy making. With this in mind, Guidelines have been formulated at EU and Member State levels (see for example BERR, 2008; European Commission, 2005). In some Member States, formal, quantified Regulatory Impact Assessments are becoming routine and influential. The European Commission guidelines are intended for the assessment of proposed policies and legislation, but the recommendations have not systematically been extended to the implementation of legislation and decision-making on specific risk issues. In particular in the area of food safety there is a

Abbreviations: BSE, Bovine spongiform encephalopathy; CIAA, Confederation of the food and drink industries of the EU; COI, Cost of illness; CUA, Cost utility analysis; EFSA, European Food Safety Authority; ERS, Economic Research Service of USDA; EU, European Union; FSA, Food Standards Agency (UK); HACCP, Hazard Analysis Critical Control Points; NICE, National Institute for Health and Clinical Excellence (UK); NTD, Neural tube defect; OECD, Organisation for Economic Cooperation and Development; QALY, Quality Adjusted Life Year; RIA, Regulatory Impact Assessment; UK, United Kingdom; USDA, United States Department of Agriculture; VSL, Value of Statistical Life; WTO, World Trade Organisation. * Corresponding author. E-mail address: [email protected] (W.B. Traill). 0956-7135/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.foodcont.2009.06.018

lack of detailed policy guidance for the formal and systematic consideration of the distribution of risks, costs and benefits (environmental, economic and social), and consequently there is a lack of transparency how such considerations are weighed into decisionmaking (Koenig et al., 2008). The SAFE FOODS project goes further than the various sets of guidelines in proposing that economic, social and environmental aspects of food safety should be considered throughout the risk analysis process: during problem framing, assessment, evaluation, decision-making, monitoring and review. The aim of the present paper is to review alternative approaches to the economic analysis of the distribution of costs and benefits from a conceptual and a pragmatic perspective. We indicate how and where the economic approach fits within the overall SAFE FOODS framework and how it contributes to transparent and accountable risk management. We also highlight tensions in setting common standards appropriate to all Member States and in conformity with WTO treaty obligations. This paper is not a comprehensive review of food safety economics which has many strands beyond economic evaluation of proposed regulations. These include: the use of food safety regulations as non-tariff barriers, especially to developing countries (e.g. Henson & Loader 1999); methods for the assessment of compliance costs of alternative food safety approaches such as HACCP (e.g. Boland, Hoffman, & Fox, 2007); alternative mechanisms for food safety control such as legal liability (e.g. Kolstad, Ulen, & Johnson, 1990); analysis of how firms respond to food safety regulations (e.g. Caswell & Johnson, 1991); and the political economy of food

1612

W.B. Traill, A. Koenig / Food Control 21 (2010) 1611–1619

safety regulation (e.g. Henson, Loader, & Traill, 1995). A recent book by Hoffman and Taylor (2005), also addresses many of these issues.

Implicit price per unit of safety

Marginal social cost

2. Overview of economic assessment Pm

2.1. Basic economic principles of resource allocation Markets work by coordinating the tastes and technology of consumers, producers, middlemen, and sellers given well-established property rights set by government. Markets exist for all types of foods (e.g. burgers, chips, restaurant meals, take-aways, food ingredients); and markets also exist for gyms, taxis, bicycles, health services, weight-loss clinics, and insurance. The power of the marketplace rests in its ability to help people make trade-offs such that goods and services move from low valued uses to higher value uses. The market price is the key here—the price reveals the relative scarcity of the good or service based on the classic ideas of ‘‘supply and demand”. If people demand more of a good (holding supply constant), the price goes up and, as a consequence, more resources are allocated to production of the good. Economics makes the point that when markets work, they work well—they lead people to decisions that maximize the benefits to society. If all the key conditions to support a well-functioning market exist, economics tells us markets should be left alone and the final market equilibrium would reflect the best possible outcome for society given the distribution of income.1 People achieve the best outcome through self-interest regulated by competition. Key conditions include well-defined property rights (and the means to enforce them), no externalities, no monopoly power, and nobody holds an information advantage so as to distort market prices. However, for some products and services, these market conditions are more applicable than for others. If these conditions are not applicable, markets can fail. In these cases, equilibrium prices and quantities do not capture the total social costs and benefits; some form of governmental intervention is needed to rebalance the equation between private desires and social goals. The critical question becomes understanding exactly how the market has failed, and what governments can do to realign private and social costs and benefits without generating unintended consequences of otherwise well-aimed policies, and ensuring intervention is cost effective. (There are, however, also other reasons for which policies can fail, the topic of flawed policies and government failure is, however, beyond the scope of this paper.) For market failures in relation to food safety, Information problems are usually the main cause. In a well-working market, full information is available to all market participants. In the case of food safety there are two possible information problems, the first, asymmetric information, when the seller is better informed than the buyer and could, in principle, pass off unsafe food as safe. Due diligence laws with legal liability, and the costs to firms of lost reputation make this unlikely in developed countries. The second, imperfect information, means that consumers are unable to tell by inspection whether food is safe; possible solutions include labelling (e.g. in relation to allergens) and standards (product or process). It is widely accepted that food safety is a clear example of an imperfectly functioning market and that governments have an important role to play in regulating food safety. A critical issue is: what is the appropriate level of safety governments should establish? It is widely accepted that zero risk is unachievable and 1 Economic theory suggests income redistribution is better achieved through fiscal policy instruments than through sectoral policies.

Marginal social benefit A

Qm

B

Level of safety

Fig. 1. Demand for and supply of food safety.

attempting to achieve zero risk would be a waste of valuable resources. Economists consider there is an optimal level of food safety at which the additional (marginal) costs of a higher level of safety are equal to their additional (marginal) benefits. Fig. 1 (from Henson & Traill, 1993) is the usual representation showing that the additional cost per unit to achieve ever higher levels of safety is an increasing function (the marginal social cost curve is upward sloping); in other words it is cheap and easy to improve safety from a very low level, but additional improvements are ever more costly. In contrast, the marginal social benefits which represent society’s additional willingness to pay to avoid ill-health and the costs of treating ill-health fall as the level of safety increases, represented by the downward sloping marginal social benefit curve; the more unsafe food is, the more people would be willing to pay for a small increase in safety, but once food is already relatively safe, people are less willing to pay for improvements.2 At a point such as A in Fig. 1, the amount society is willing to pay exceeds the amount it would cost to improve safety, so it is worth allocating resources to produce more safety; at B, costs exceed benefits, so too many resources are being devoted to safety. Only at Qm are costs equal to benefits per unit of safety representing efficient resource allocation (Pm is the corresponding ‘shadow’ price per unit of safety, the implicit value attached by society to safety). 2.2. Cost–benefit analysis The absence of a well functioning market for food safety necessitates the measurement of marginal social costs and benefits to determine whether a higher level of safety provided by a proposed regulation is justified. This is the basis for an economic cost–benefit assessment as part of the assessment and evaluation of regulatory options in food risk analysis. We proceed to discuss measurement approaches for cost–benefit assessment. Many of the alternative techniques for evaluating policies, such as costeffectiveness, cost-of-illness and cost-utility analysis consider, for simplicity and practicality, only a sub-set of the costs and benefits. By discussing the ‘complete’ list of costs and benefits first we are in a better position to judge how much information is lost in the simpler approaches. Economists generally consider cost–benefit analysis to be the gold standard in measuring the impact of a policy intervention, but it must be recognised that it does not consider everything that politicians, industry and other stakeholders may consider important; for example it is a static measure that does not consider the impact of an intervention on economic growth through slowing 2 Note that the curves correspond to supply and demand curves in a conventional market and have the same interpretation.

W.B. Traill, A. Koenig / Food Control 21 (2010) 1611–1619

innovation. This is because these things are largely unknowable, let alone measurable. Regulatory Impact Assessments (RIAs) frequently make reference to such dynamic aspects of regulation and the usual approach is to list them in a qualitative manner. Suggestions for approaches to identify, assess and evaluate other potential consequences of alternative regulatory options beyond the field of economics are addressed in other papers in this Special Issue (see for example Dreyer, Renn, Cope, & Frewer, 2008; Koenig & Kaiser, 2008). 2.3. Measuring costs and benefits 2.3.1. General Measuring costs and benefits of a policy intervention is easier said than done; by definition there is not a market whereby people demonstrate how much they are willing to pay by buying or not buying a product and whereby firms can respond if their costs are covered. Alternative procedures are needed. On the benefit side of the equation it is useful to distinguish between private benefits which accrue directly to the individual consuming the food (reduced risk of pain and suffering from ill-health and possibly death), and public benefits, public sector savings from reduced visits to doctors and hospitals, reduced medication and reduced costs associated with missed days of work.3 As we shall see, these are often intermingled in policy evaluation. Costs are borne by private firms through increased requirements for inspection, control, labelling, reorganisation of production, etc.4; and by governments in administration and inspection as well as, in principle, though rarely measured, the deadweight welfare loss associated with taxation.5 The literature contains a number of reviews of the theoretical strengths and weaknesses of alternative measurement instruments (e.g. Antle, 1998; OECD, 2002), but some of the approaches have a severe practical shortcoming: they are dependent on data that would be prohibitively expensive to collect. As Davies (1997) says, cost–benefit analysis is itself not costless and should only be performed if the benefits from doing so outweigh the costs. In the following sections, we discuss the data requirements of cost–benefit analysis and propose approaches which should not be too onerous in most cases. Individual cases will justify alternative degrees of completeness of the assessment depending on their complexity and their economic and social significance. Some may merit significant time and money collecting detailed data, others may only merit back-of-envelope calculations. The amount of time and money devoted to data collection and other such issues should be determined in the framing component of the SAFE FOODS approach. Cost–benefit assessment may be applied ex ante to aid the decision whether to introduce a regulation or ex post to determine if it has been successful. In the SAFE FOODS framework, these correspond to assessment and evaluation; and monitoring, respectively. Theoretically the approaches are identical though ex post data

3 It may be argued that firms also derive benefits from regulation because standards are essential to enable markets to function effectively, and without effective markets firms cannot profitably operate. However, except in dramatic cases like BSE, food safety regulation is incremental and the market will continue to function with or without an additional piece of regulation. These benefits are not therefore considered in cost-benefit analysis. 4 Some of these costs may be passed on to consumers in the form of higher prices, so the cost burden is ‘shared’ by producers and consumers. The interest in cost– benefit analysis is in measuring the magnitude of the costs, not determining who ultimately pays them. 5 A food safety regulation may impose other externalities, for example environmental. These could be included in an all-encompassing cost–benefit approach or assessed separately, for example in an environmental impact assessment as proposed in the SAFE FOODS approach.

1613

availability is less of a problem so more sophisticated methods may be appropriate, particularly if the regulation explicitly specifies, as it ideally should, the appropriate data to be collected for subsequent monitoring.6 2.3.2. Calculation of costs There are two main approaches to measuring industry costs, direct measurement of compliance costs (sometimes called accountancy or engineering approaches) and indirect econometric approaches. Only the former is applicable ex ante; the latter requires actual data on costs incurred in implementing the regulation so is suitable for ex post monitoring. The compliance cost approach is already widely applied by governments in RIAs. The procedure identifies the individual stages of production and the changes needed to meet the requirements of the proposed new regulation, values those changes and sums them. Examples include capital costs of new equipment, operation and maintenance of new equipment, requirements for specialist inputs (e.g. laboratory technicians for monitoring and inspection) and increased wastage (see e.g. OECD, 2002). Calculation of the costs of compliance are based on a survey of, or consultation with, firms implying that the survey should be large enough to be representative of companies of different sizes, using different production techniques. This method has the scope for bias: companies have an incentive to over-state costs, regulators with an interest in implementing a proposed regulation have an incentive to understate the costs; it may be difficult to identify all of the inputs required to comply with a regulation and their levels of use; the method does not allow for the dynamics of competition that might drive some firms out of business, especially small firms if the fixed costs represent a large component of the total (e.g. HACCP), and permit other firms to expand, meaning overall industry costs would be overstated; and finally, firms adapt to regulations by innovating and modifying their production processes, whereas the compliance cost approach assumes an unchanged process plus add-ons for improved safety, which is likely to overestimate costs in the long run.7 It is important that only those costs necessary to comply with a proposed regulation are included. In practice it is highly probable that some firms already meet the requirements of the regulation, either to satisfy existing requirements of buyers (for example retailer private standards) or with a view to enhancing or maintaining their brand reputation. Only additional costs should be included in the calculation, which in some cases may only be borne by small firms or firms supplying non-traditional supply chains. All of these difficulties should be borne in mind by the analyst. Ex post, many of them cease to be problematical when real reactions of firms can be observed. The econometric approach uses detailed company-level data on inputs and outputs from industry databases such as the US Census Bureau to statistically relate firm costs to the safety procedures used. It has the advantage of allowing for the adaptations firms make to their production methods in complying with a new regulation. It allows for the fact that firms adjust their production processes to minimise costs following the regulation whereas compliance cost methods assume these to be cast in stone. Government costs of regulation should be added to industry costs to obtain an estimate of the total costs of a regulation. 6 For example, ex ante analysis requires estimates of the likely impact of regulation on the probability of ill health. Ex post analysis can use actual data on the change in incidence of ill-health provided it is collected. There remain complications in ex post analysis, notably the fact that other factors may change over the evaluation period, making it necessary to develop a statistical model to isolate the effects of the regulation (see for example Blundell & Costa Dias, 2000). 7 Though R&D and other costs associated with innovation partially off-set the overestimation of costs.

1614

W.B. Traill, A. Koenig / Food Control 21 (2010) 1611–1619

As mentioned before, there may be other costs of regulation that defy quantification: industry often argues about the overall burden of regulation (CIAA, 2006) and governments may have an objective of reducing the overall regulatory burden, but in practice trying to cost the marginal burden of one more piece of legislation is never done; it is conceivable there could be competition implications if the regulation reduced the number of firms in an already oligopolistic industry, but again the implications are virtually impossible to quantify and rarely likely to be important; small firms may be adversely affected and RIAs often require a specific mention of the impact on small firms. 2.3.3. Calculation of benefits Social benefits from a food safety regulation include (a) private benefits (reduced pain and suffering, longer life), and (b) public benefits (reduced costs of medical treatment and lost productivity). Measuring public benefits is, effectively, an accounting exercise, but it is not an easy one because of data problems; private benefits are more conceptually difficult because they involve the valuation of health and life. We begin with private benefits. 2.3.3.1. Private benefits. Economists use willingness to pay for the benefits of a regulation as their measure of private benefit; theoretically a person is willing to pay up to the point where their gain in utility from the reduced risk of ill-health or death equals their loss in utility from having less money (to spend on other things). The trick is to calculate each consumer’s maximum willingness to pay. It is not the intention in this paper to delve deeply into the alternative elicitation techniques, their strengths and weaknesses and likely tendency to lead to exaggeration or underestimation of benefits. Methods, which include contingent valuation, conjoint analysis, choice experiments and experimental auctions have been reviewed elsewhere (see e.g. Hammitt, 2005). They depend upon surveys, interviews or other interaction with the general public and are therefore relatively expensive in money and time. A criticism of willingness-to-pay approaches, and hence of cost– benefit analysis itself, is that they give the greatest weight to the richest members of society who can afford to pay the greatest amount for a reduction in their risk of ill-health/death. This means policy interventions that favour the rich are more likely to have a favourable benefit/cost ratio than ones favouring the poor. Economic purists would argue this is equally true of market transactions where the rich have much higher spending power so much greater influence on what is produced than the poor; they would further argue that the distribution of wealth is a political issue and food policy is an inappropriate and inefficient means of overcoming societal inequalities. Strictly speaking this is true, but it overstates the power of the political system to intervene to change the distribution of wealth and ignores the fact that governments, implicitly or explicitly (e.g. the UK), include reducing health inequalities as a goal of public health policy. If income distributional issues are hard to deal with within a country, they are more complicated between countries with different incomes. Refer back to Fig. 1 and imagine a rich country and poor country situation: the high income country’s marginal social benefit curve for food safety would be positioned to the right of a poor country’s, reflecting rich country consumers’ ability and willingness to pay more for any given increment in the level of safety. Even if the cost of producing safety, reflected in the marginal social cost curve, was the same in both countries, the rich country equilibrium would imply a higher level of safety and a higher per unit cost of safety than the poor country. This is a particular problem for Codex where developing countries may have higher priorities for their scarce resources, (e.g. economic growth), than to devote them to attaining western levels of food safety. Income levels vary less widely in the EU, but disparities between rich

member states and the transition countries are nevertheless large enough to create tensions when trying to set a common standard. Rather than try to elicit willingness to pay, the recent tendency in public health evaluations is to calculate health benefits as Quality Adjusted Life Years (QALYs) and attach an implicit or explicit financial value to the QALY. This approach, which we discuss further below, overcomes some of the income inequality problems of willingness to pay. 2.3.4. Public benefits Public benefits of regulation to improve food safety are measured as public sector cost savings. They comprise principally medical costs and lost productivity where these are paid by the public sector under a national health scheme and national employment law. Medical costs are calculated by multiplying the probability of food-related ill health by the cost of treatment; normally various health states are possible, for example Salmonella poisoning may, in its mild form, require a visit to the doctor with associated costs, or a more severe case may require hospitalisation. The most problematic issue is placing a value on death itself. Strictly speaking, death is almost costless from a medical perspective and should not therefore be assigned a value, but in practice cost of illness (COI) calculations often attach a value to life, called the value of statistical life [VSL]. This is calculated as the amount people are willing to pay for a small reduction in the probability of death; for example suppose a new standard would reduce the risk of cancer by 1 in a million and suppose further that on average people are willing to pay €1 for the greater security. A million people would in total be willing to pay €1 m and, statistically, 1 life would be saved. In this example the value of a statistical life would be €1 m. In practice when allocating scarce resources to saving lives under alternative risky situations, developed Western countries often use a figure of around €2 m. It should be noted, however, that when using such a figure alongside medical expenses of illness, we are mixing private and public benefits because medical expenses are public costs and VSL is people’s willingness to pay to lower the probability of their own deaths which reflects only private benefits. Nevertheless this is done, an example being the USDA cost of illness calculator, a demonstration of which is given in Table 1 with respect to Salmonella in 2006. Note that the VSL used here is $5.3 m, rather higher than the figure of €2 m mentioned above. The average medical cost of a case of Salmonella in the US is just over $1766 implying that the benefits of an intervention to reduce Salmonella would be ($1766 the number of cases saved). Note as well that although there were only 415 deaths from 1.4 m cases of Salmonella, the cost per case estimate is overwhelmingly dominated by the VSL component without which the cost per Salmonella case estimate would fall to $202. Thus the overall estimate depends crucially on its most problematic component. Productivity losses associated with food-related ill health are usually calculated from an estimate of the number of days work missed through illness. The numbers can easily outweigh the medical costs, for example in the UK the direct medical costs of obesity have been calculated at around £1b in 2002 whereas lost earnings attributable to obesity were put at around £3b but with a recognition this figure could be as large as £10b (Foresight, 2007). The wide range of these figures demonstrates the difficulty in measuring days of lost work attributable to a specific food-related illness. 2.4. Quality Adjusted Life Years (QALYs) We have attempted in the above sections to divide the benefits of enhanced food safety between private benefits to individuals represented by their willingness to pay for a reduced probability of food-borne ill-health and premature death, and public benefits from lower state-borne medical expenses and higher productivity.

1615

W.B. Traill, A. Koenig / Food Control 21 (2010) 1611–1619 Table 1 USDA foodborne illness cost calculator for salmonella in 2006. Foodborne illness cost calculator: Salmonella

ERS estimates that the annual economic cost of salmonellosis—the illness caused by the Salmonella bacterium—is $2467,322,866 (2006 dollars). This estimate is for all cases of salmonellosis, not just foodborne cases. The estimate includes medical costs due to illness, the cost (value) of time lost from work due to nonfatal illness, and the cost (value) of premature death. It excludes a number of other potential costs, such as those associated with chronic complications, disutility for nonfatal illness, pain and suffering, travel, childcare, etc. ERS cost estimate: Salmonella, from all sources, 2006, dollars Cost component

Did not visit physician; survived

Visited physician; survived

Hospitalized; survived

Visited physician/hospitalized; died

Total

Number of cases All illness

1224,547 Dollars

157,738

14,487

415

1397,187

Medical Medications Office visits Emergency room Outpatient visit Hospitalization Productivity, nonfatal Disutility, nonfatal Premature death ERS total cost, 2006 ERS average cost per case, 2006

0 0 0 0 0 0 59,960,648 0 0 59,960,648 49

51,516,442 0 24,744,360 6016,443 20,755,639 0 25,674,237 0 0 77,190,679 489

136,178,713 0 1136,288 1657,689 1270,829 132,113,907 6373,714 0 0 142,552,427 9840

3536,518 0 46,501 47,487 36,405 3406,126 224,978 0 2183,857,617 2187,619,112 5271,371

191,231,673 0 25,927,149 7721,619 22,062,872 135,520,033 92,233,577 0 2183,857,617 2467,322,866 1766

We saw in the previous section that there is already a tendency by public authorities to mix public and private benefits by including a value for life, VSL, in the estimate of medical expenses. This measure puts a utility value on premature death but ignores the utility of a higher probability of better health. The QALY is an attempt to measure in a single utility figure the benefits of an intervention that increases both life expectancy and the quality of life. For example, as we saw in Table 1 above, a reduction in the incidence of Salmonella cases would save a number of lives and prevent a much larger number of people becoming sick. Instead of combining them in monetary terms, as in the cost of illness approach, QALYs combine the two in utility terms by assigning to each year of ill-health a utility value that is a fraction of the value of a year of good health. For example, suppose in an average year food poisoning killed one person 10 years prematurely and made 100 more people acutely ill for all or part of that year. Suppose further the quality of their lives in the year they suffered the food poisoning was only 90% of the quality of a year spent in good health. The utility loss would be 10 + 100  0.1 = 20 QALYs. An intervention that reduced the probability of Salmonella infection by 50% would save 10 QALYs. If such an intervention cost €10 m and an alternative investment of €10 m into, say, safer rail travel saved five QALYs, then the salmonella intervention can be said to be more cost-effective. In principle all interventions should have the same cost per QALY gained, otherwise a reallocation of resources from high to low cost interventions could save more QALYs for the same total financial outlay. But how should the utility loss of ill-health be evaluated? The alternative preferred by economists is to devise experimental auctions or choice experiments to elicit people’s utility under different health states, but these approaches use willingness-to-pay and so are subject to the income distributional problems of such approaches discussed above. Alternative, procedures, now widely used, avoid willingness-to-pay by surveying people directly about their utility losses from various forms of disability. A specific valuation technique, EQ-5D, is recommended by the National Institute for Health and Clinical Excellence (NICE) in the UK and by authorities in the Netherlands, Norway, Italy, Hungary, Poland, Portugal, Canada and New Zealand (Szende, Oppe, & Devlin, 2007). In this approach, health is evaluated across five dimensions, each with three levels of health state as shown in Table 2.

People complete a questionnaire, choosing one of the three options in each category, and an evaluation of their overall well being (between 0 and 1). With perfect health (1 1 1 1 1) anchored at 1, death at 0, a statistical analysis of a large enough survey can estimate the weights people attach to the various health states. Once the weights have been calculated (and they are publicly available in Szende et al. (2007)) a risk assessment need only specify the health benefits associated with a policy intervention in terms of the reduced probabilities of each category of ill-health in Table 2 to be able to calculate the QALY gains. Sticking to the Salmonella example, as well as the risk assessment calculating the reduced probability of each of the four levels of severity in Table 1, it would need to attach a QALY value to each by specifying their duration and impact (on the 1–3 scale of Table 2) on mobility, self care, etc. and applying the weights described by Szende et al. (2007). It is common when using QALYs to calculate the cost per QALY gained from an intervention (e.g. the UK National Institute for Health and Clinical Excellence—NICE–does this explicitly). This is

Table 2 Example of health state categories used in calculation of QALYs. Mobility 1. I have no problems in walking about 2. I have some problems in walking about 3. I am confined to bed Self care 1. I have no problems with self care 2. I have some problems washing or dressing myself 3. I am unable to wash or dress myself Usual activities (work, study, household, family or leisure) 1. I have no problems performing my usual activities 2. I have some problems performing my usual activities 3. I am unable to perform my usual activities Pain/discomfort 1. I have no pain or discomfort 2. I have moderate pain or discomfort 3. I have extreme pain or discomfort Anxiety/depression 1. I am not anxious or depressed 2. I am moderately anxious or depressed 3. I am extremely anxious or depressed

1616

W.B. Traill, A. Koenig / Food Control 21 (2010) 1611–1619

known as cost-utility analysis (CUA). An implicit value placed by government on the QALY is given by the cut-off point whereby interventions (or new medicines) are approved for use in the health service, or not approved, for example with NICE it is about £30 K (about €40,000).8 This is a reasonable starting point for the valuation of a QALY, and could in principle be used as an explicit valuation of the private benefits accruing to a health care intervention in a cost–benefit analysis. The UK FSA has begun putting a valuation on QALYs in its cost–benefit analyses which form part of its RIAs see folate (case study examples given below). However, this reintroduces the problem of income distribution in that a rich person is willing and able to pay more to avoid ill-health or death and similarly, with problems of the sort faced by Codex attempting to develop harmonised standards, a rich country is willing and able to pay much more to gain a QALY than a poor country. The discussion above might suggest that the calculation of QALYs is straightforward and uncontentious, but this is not so. First, QALYs would be improved by discounting distant over proximate years of health saved. People value a year of good health now more highly than a year of good health in 50 years time. This suggests a policy to reduce Salmonella death and illness with immediate effect should be valued more highly than an intervention that reduced the likelihood of cancer and saved the same number of lives and years of ill health in the distant future. A discount rate around 3% p.a. is typical for government projects, implying €1 in 1 year’s time is worth only 97 cents today and €1 in 20 years time is worth only 1/(1.03)20 = 55 cents now. QALYs can be discounted but often are not.9,10 We began this section with a discussion of the distinction between public and private benefits of food safety and argued that both should be considered in calculating social benefits. In this respect, we should recognise that QALYs are a measure of private benefits—and probably the most practical measure available. However, for full cost–benefit purposes, savings in medical expenses and productivity losses should be added to the value of QALY gains.

2.5. Decision criteria: data, advantages and disadvantages 2.5.1. General There are various approaches to the assessment of economic impacts of different policies that could also be used for formal economic assessments of regulatory options in the implementation of food safety law that use different amounts of the various types of information discussed above. We have already mentioned cost of illness, cost-utility and cost–benefit and now summarise and clarify further their distinctions, advantages and disadvantages. We also introduce a still simpler approach, cost-effectiveness and the extended, but less well defined, regulatory impact statement. 8 Interestingly, if the VSL is accepted to be a measure of the value an average member of society places on life, if we divide VSL by average life expectancy, L (about 45 years remaining in a population of average age around 35), it is reasonable to assume the value attached to each year is VSL/L or about €45 K if VSL is €2 m and all the years would be lived in good health. 9 DALYs (Disability Adjusted Life Years) are in principle the inverse of QALYs – whereas the QALY values a year of life in good health as having a weight of 1 with a year of death having a weight of zero, the DALY reverses these weights, so a QALY gained is the same as a DALY avoided. DALYs, first introduced by the World Bank (1993), explicitly discount the value of distant life years. Another variation is the HALY (Health Adjusted Life Year). 10 In a world of crude data, these issues may become somewhat academic. A simpler approach that avoids associating utility values to states of ill-health is to look only at the life years (LY) saved by an intervention (and these can also be discounted). This approach ignores ill-health and concentrates only on mortality, but may be acceptable if closely correlated with QALYs. Studies by Chapman et al. (2004) and Robberstad (2005) suggest in the vast majority (80%+) of cases priorities associated with alternative policy interventions are unchanged when ranking by LYs or QALYs.

2.5.1.1. Cost-effectiveness. This method of assessment does not place a value on benefits but calculates the cost per life saved or alternative outcome, e.g. cost per Salmonella case avoided. The approach is easy to use if the choice is between alternative interventions to achieve the same outcome, when the method gives a ranking of interventions. It is not helpful when a regulation has mixed benefits, saving lives and avoiding illness, nor does it give a clear rule as to whether any intervention is worthwhile. 2.5.1.2. Cost of illness. As discussed above, calculates the cost savings of regulation, generally those relating to medical treatment and productivity losses. They are therefore a direct estimate of the monetary costs to the public sector of ill-health and death and do not allow for loss in utility. As discussed, a value is sometimes placed on life, meaning the disutility of death, but not illhealth is accounted for. 2.5.1.3. Cost-utility. Similar to cost of illness, but cost-utility analysis calculates the cost per QALY gained. The approach is somewhat more demanding in terms of data than cost of illness but more appropriate when there are private as well as public benefits to regulation and it is therefore desirable to measure the utility gains from better health. Again there is need for a bench-mark figure (value of a QALY) to determine whether any intervention is worthwhile, otherwise the approach just gives a ranking of alternatives. 2.5.1.4. Cost–benefit. This criterion, discussed extensively above, computes all private and social benefits divided by all private and social costs. Intervention is worthwhile if the ratio of benefits to costs is greater than one. As discussed, from a theoretical economic perspective, private benefits should be measured in terms of willingness to pay, but this is time-consuming and may be biased in favour of the status quo income distribution. A practical alternative is to measure private benefits using QALYs and attach a monetary value to the QALY. 2.5.2. Regulatory Impact Assessments (RIAs) RIA is the term used by policy makers to assess the impacts of an intervention, but it is a fairly vague term that may encompass any of the above criteria or none of them. In some cases, RIAs may go beyond the economic criteria discussed above in order to address dynamic and/or equity issues, though usually only in qualitative terms. An example is the UK RIA in which Departments are requested to specify the geographic impact of the regulation, its impact on greenhouse gases and on competition as well as to specify the cost implications for different sized firms. It is worth highlighting some of the Guidance to civil servants preparing such documents. After identifying the need for intervention, alternative options should be developed; in the simplest case these may be ‘status quo’ and one proposed intervention, though it is often more than one. A preliminary document is prepared and there follows a consultation stage whereby the public and other stakeholders can comment and the options are refined. Following implementation there is expected to be a review to determine the ex post effectiveness of the intervention and its actual costs and benefits. This is entirely consistent with the SAFE FOODS approach. In preparing the RIA, instructions are to take account of the full range of costs and benefits, economic, social and environmental, and these should be monetised as far as possible.11 Costs and benefits should be discounted, again as discussed above, in accordance with current Treasury Guidelines (in most cases using 11 NICE in its Guide to the Methods of Technology Appraisal (NICE, 2004) does not carry out a full cost-benefit analysis but suggests that a cost per QALY saved greater than £30,000 would need good ‘other’ reasons for support such as highly innovative technology and wider social benefits (p. 33).

W.B. Traill, A. Koenig / Food Control 21 (2010) 1611–1619

a discount rate of 3.5%12). Sensitivities and risks should be highlighted. We conclude that a RIA as applied in the UK amounts to a form of economic cost–benefit analysis with extensions. The extensions may include ethical, social and environmental impacts of the sort discussed in other papers in this issue. They may also involve dynamic economic aspects of regulations which do not form a part of traditional cost–benefit analysis. 2.5.3. Folate fortification An example of a RIA prepared by the UK Food Standards Agency illustrates a number of these points and also shows that formal quantification is possible and useful. The RIA for mandatory folate fortification of flour is interesting because it is the first to monetise QALYs13 in a cost–benefit analysis and was runner up in a UK Government Economic Service competition prize. Poor folate status in early pregnancy can cause neural tube defects (NTDs), leading to a range of possible conditions including, most commonly, Spina Bifida, and still birth. Seventy-five percent of foetuses with NTDs are terminated following pre-natal screening but there remain around 370 live births and 132 still-births of children with NTDs in the UK. The reduction in the number of still births, terminations and live births with NTDs following mandatory fortification of flour14 is calculated using US experience and UK food consumption data at between 114 and 261. For cost–benefit calculations the lower number was used, so the benefits are a conservative estimate. Using average life expectancy the total number of discounted QALY gains from reduced still births and handicapped births were 644 per year, valued, at £30,000 per QALY. Terminations were costed only at the NHS cost; it was recognised that the assumption that a still birth represents a large cost associated with a full life lost, whereas a termination has only a small medical cost of £540, is recognised to be problematic, but unimportant in this particular example given that the benefit–cost ratio is so conclusively positive (see below). The total benefit is £19.3 m/year in total.15 Costs were calculated to be only £1 m/year for folate and a monitoring cost to local authorities of £10 K, plus one-off costs to industry of re-labelling and reformulation of around £1 m (maximum) and a small administrative burden to firms for reading and understanding new regulations. Following discounting, the benefit/cost ratio is calculated at over 400 over a 5-year period indicating clear net social benefits over social costs.16,17 For other aspects of the folate RIA, it was considered that trade implications would be minimal because most UK flour is exported to Ireland which was also proposing fortification; the impact on exports to other countries was ‘‘uncertain; it would clearly depend on any reactions of our key trading partners and on underlying attribute preferences and price elasticities.” There would be no envi12

NICE also proposes using a discount rate of 3.5% (NICE, 2004). In fact they use DALYs but we need not concern ourselves with the small differences in methodology this makes. The main difference is that they use disability weights provided by medical professionals rather than from willingness to pay to avoid a condition. The data are those used by Food Standards Australia New Zealand and available on-line in Mathers, Vos, and Stevenson (1999): http://www.aihw.gov.au/publications/health/bdia/bdia.pdf. Data are available for a range of ailments. 14 Assuming fortification at a level of 300 lg/100 g (white flour only). 15 The RIA recognises there would be additional benefits of reduced health care costs of live births and home and vehicle adaptation costs to aid mobility and quality of life, but these were not calculated because the B/C ratio is favourable anyway. 16 Other options considered were markedly inferior: (1) maintain the status quo (recommendation to pregnant women to take folate supplements), (2) health education, (3) voluntary fortification. 17 Note that the alternative to valuing QALYs in a cost/benefit calculation would be to calculate the cost per QALY gained. In this case the discounted costs are £5.7 m over 5 years, the number of QALYs gained is 2,576 (644 per year for 4 years assuming no lives saved in the first year) for a cost of £2,200 per QALY. If the government was using an implicit cut-off point of £30,000 for judging an intervention worthwhile (the same as NICE), this intervention would be considered very good value for money. 13

1617

ronmental costs and the policy would be sustainable. There would be no competition effects (barriers to entry are high and the cost of folate fortification would not significantly change entry costs). Potential risks of the regulation would be a small potential for trade barriers against fortified products; a small and unquantifiable potential risk of vitamin B12 masking and bowel cancer; and restriction of consumer choice. 2.6. The consideration of economic impacts in the SAFE FOODS framework One main objective of the SAFE FOODS project is to enhance transparency and accountability of the current EU food safety system, whilst keeping the focus of decision-making on the protection of public health. Guidance on how to make more transparent what factors actually are considered in decision-making, apart from potential health risks, including possible economic impacts, is hence one key element of the SAFE FOODS framework. The main differences between the SAFE FOODS framework and current frameworks for risk analysis are shown in Fig. 2 and include the addition of two stages, framing and evaluation, to the routine stages of risk assessment, risk management and risk communication (see Koenig et al., 2008). The two additional stages are intended to improve the connection between risk assessment and risk management. Furthermore, they offer additional formalised processes to improve documentation on the definition of the terms of reference of the assessment and the scope, and subsequent evaluation of the assessment. The question of economic impacts is relevant at all stages of the SAFE FOODS framework, framing, assessment, evaluation, and risk management, including monitoring and review and addressed together with a short description of each stage in turn. 2.6.1. Framing The framing stage serves to gain an initial shared understanding of the issue, the objectives, and broad alternative courses of regulatory action. Areas of consensus and dissent are documented and this forms the basis for planning the decision-process. This stage can also serve for a preliminary mapping of economic aspects of the food safety issue at hand. This allows to gain an overview on existing relevant data, and to decide on what additional data may be required. The decisions on scope of the economic and social assessments in the framing stage may be amongst the most difficult and controversial, but essential to allow for a comprehensive and generally accepted economic assessment. On controversial issues, diverse views on what constitutes costs, and benefits and how they are distributed, should be documented in the ‘framing report’ and serve, among other things, as guide for further research in this area. This paper suggests to structure the overview on potentially relevant economic impacts in terms of private and public costs, and private and public benefits. This is akin to approaches commonly used in Regulatory Impact Assessments (RIAs), such as are becoming more frequent also in the UK. There should also be a preliminary analysis of already existing data and the feasibility and costs of requiring additional date for these types of assessments, versus the burden this may pose on industry and innovation (cost–benefit analysis itself is not cost-less and should only be required if benefits outweigh expected costs; Davies, 1997). In consequence, assessment of costs, benefits and their distribution may not always be warranted for each issue. Koenig and Kaiser (2008) propose the ethical matrix approach as a tool to help in deciding at the framing stage whether the implicit consideration of economics is sufficient, or whether it may be warranted to make a more formal assessment of economic impacts of alternative regulatory options. The structure and content of the ethical matrix is compatible with the

1618

W.B. Traill, A. Koenig / Food Control 21 (2010) 1611–1619

Framing • Defining objective of regulatory action • Planning the decision process

Review

Stakeholder Risk Management Implementation & Monitoring • Control • Enforcement • Monitoring of effects of decision

DecisionMaking

involvement

• Ranking and • Choice of regulatory options

Risk communication

Risk-benefit Assessment Health and Environmental Risks & Benefits Methods for risk assessment for hazard and benefit • Identification, • Characterisation • Exposure assessment

Societal Impacts Methods from: • Economics • The social sciences • Applied ethics

Evaluation Comparing perspectives on distributions of risks, costs and benefits from multiple perspectives

www.safefoods.nl Fig. 2. The SAFE FOODS risk analysis framework.

suggested approach for structuring the economic assessment. The four criteria embedded in the matrix, i.e. ‘‘absence of harm”, ‘‘do some good”, ‘‘dignity/identity” and ‘‘fairness”, that are plotted against the main affected parties with moral standing (see in particular Table 1 in Koenig and Kaiser (2008)), can easily be superimposed with the structured assessment of costs and benefits, private and public. 2.6.2. Assessment In the SAFE FOODS framework, to fit present institutional remits, EFSA is only in charge of assessing health and environmental impacts; the assessment of social or economic impacts is conducted under the auspices of DG SANCO (see Koenig et al., 2008). For this assessment, we provide recommendations of the structure and on the choice of assessment tools. 2.6.3. Evaluation This stage serves to compare the risks, costs, and benefits and their distribution. The ethical matrix could serve to structure the evaluation process. This matrix has in practice successfully been tried and proven to help stakeholders to attribute the weighting of diverse values that are at stake based on additional information from the assessment. Differences in weights between stakeholders could thus be identified (see Koenig et al., 2008). This is then documented and taken into account in decision-making. 2.6.4. Monitoring Monitoring refers to what economists call ex post analysis,18 an evaluation of policy effectiveness based on its actual impact. It would be based on the actual (rather than predicted) numbers of lives saved and ill-health avoided and actual (rather than predicted) economic costs. For example, post hoc surveys of industry may be used to determine whether anticipated impacts were realistic – if not, this may trigger not only review of the decision, but also the reframing of the question, the legislation or policy. If costs had been over-stated there may be a case for stricter standards, or, by contrast, if benefits were overstated a standard may be relaxed or abandoned. 18 The paper has, until now, been concerned with ex ante analysis—which may be viewed as a best forecast of economic impacts at the time the regulation is considered by policy makers.

2.6.5. Challenges for implementation Challenges to applying the SAFE FOODS framework and recommendations included in this paper on economic impact assessment in multi-level governance systems such as the EU and Codex, include that a certain number of the economic tools such as QALYs depend on personal preferences. For example, the weights attached to different states of ill-health in the EQ-5D approach discussed above are different in each country because the average person has different preferences in each country. Some additional research would be needed to determine a set of weights appropriate to the EU as a whole, while for Codex this would be unrealistic. However, Codex involves individual countries assessing their own interests and arriving at consensus through a process of compromise. Quantification of the benefits of a new standard as well as its costs is an important advance on current procedures in determining national interest. Income distributional inequities between countries would continue to create problems in standard and regulation setting. Within Codex the amount of money countries at different stages of development can afford to pay for each QALY of benefit will always be different, but quantification allows transparency on the compromises countries must make. Within the EU, it would be necessary to agree a common QALY value across Member States, a difficult proposition, but not more so than many of the problems faced in making trans-national laws. Economic tools often focus on aggregate impacts and neglect inequities from distributions of risks, costs, and benefits among different societal groups. In part this is because economics believes its role is to optimise decision-making given existent macro-economic conditions set by government (including income distribution). It is also a recognition of the costs of analysis—given good data, cost–benefit analysis could be carried out for each social group and a set of cost–benefit ratios computed. Someone, however, would need to attach an implicit weight to the cost– benefit ratios of each social group. However, consumer attitudes, beliefs and perceptions of food hazards and how these are managed are likely to represent important determinants of consumer confidence in food safety and the SAFE FOODS framework therefore highlights the complementarity of tools for economic, social(Dreyer et al., 2008) and ethical impact assessment (Koenig & Kaiser, 2008) and recommends their combined use to inform decisions.

W.B. Traill, A. Koenig / Food Control 21 (2010) 1611–1619

3. Conclusions Our aim in this paper has been to examine methods of economic analysis appropriate to framing, assessment, evaluation and risk management, including later monitoring of impacts of food safety regulation. We have been careful in identifying the various types of benefits and showing how these can be quantified as a basis for evidence-based policy making. Publication of the results with clear indication of the assumptions (as with the UK folate example) would ensure improved accountability and transparency. We have seen that such quantification is useful and need not be overly demanding. In particular the use of QALYs is becoming widely acceptable and either computing the cost per QALY gained from an intervention or using a monetary value for a QALY in cost– benefit analysis are realistic evaluation tools, especially since available secondary sources are now able to relate an increasing number of health states to utility levels. QALYs also benefit from not being biased towards the better off, like traditional willingness to pay approaches. We recommend this approach within the economic impact assessment proposed by the SAFE FOODS project. It should be used in ex ante evaluation and ex post evaluation (monitoring). At the EU level formal quantification is realistic, but will require further research into how to account for different preferences across nations and different ability to pay for regulation. The methods go beyond what is currently done within Codex (and some of the methods may be prone to cross-country variation, as for example in the valuation of a QALY, which is contingent on the social and economic context). But the methods are consistent with Codex’s consideration of what it calls economic interests and ‘other legitimate factors’19 We certainly do not propose that Codex should attempt to conduct a cost–benefit analysis to arrive at a single figure for all Codex countries, but suggest individual countries would benefit from the formal quantification involved in the approaches discussed. In fact Codex specifically states that ‘‘risk management should take into account the economic consequences and the feasibility of risk management options.” (p. 107 of Codex Alimentarius Commission, 2007) and ‘‘concerns related to economic interests and trade issues in general should be substantiated by quantifiable data” (p. 165 of Codex Alimentarius Commission, 2007). Acknowledgement The work was financed from the budget of the FP6 project SAFE FOODS project that is financially supported by the European Commission Directorate General for Research. References Antle, J. M. (1998). Benefits and costs of food safety regulation. Food Policy, 24(6), 605–623. BERR (2008). Impact assessment guidelines. London: Department for Business, Enterprise and Regulatory Reform. . Blundell, R., & Costa Dias, M. (2000). Evaluation methods for non-experimental data. Fiscal Studies, 21(4), 427–468. Boland, M., Hoffman, D. P., & Fox, J. A. (2007). Post-implementation costs of HACCP and SSOP’S in great plains meat plants. Journal of Food Safety, 21(3), 195–204.

19 ‘‘When elaborating and deciding upon food standards Codex Alimentarius will have regard, where appropriate, to other legitimate factors relevant for the health protection of consumers and for the promotion of fair practices in food trade.” (Codex Alimentarius Commission Procedural Manual, 16th Edition).

1619

Caswell, J. A., & Johnson, G. V. (1991). Firm strategic response to food safety and nutrition regulation. In J. A. Caswell (Ed.), Economics of food safety. New York: Elsevier Science Publishing Company. Chapman, R. H., Berger, M., Weinstein, M. C., Weeks, J. C., Goldie, S., & Neumann, P. J. (2004). When does quality-adjusting life-years matter in cost-effectiveness analysis? Health Economics, 13(5), 429–436. CIAA (2006). The Competitiveness of the EU Food and Drink Industry. Brussels: Confederation of the food and drink industries in the EU (CIAA). . Codex Alimentarius Commission (2007). Procedural manual (16th ed.). Geneva and Rome: Codex Alimentarius Commission, Joint FAO/WHO Food Standards Programme, Food and Agricultural Organisation and World Health Organisation. . Davies, R. (1997). Environmental regulation, benefit–cost analysis and the policy environment in less developed countries. Environment and Development Economics, 2(2), 206–210. Dreyer, M., Renn, O., Cope, S., & Frewer, L. (2008). Exploring the social impact of food safety issues: Including social impact assessment in food safety governance (report prepared for deliverable 6.7). Wageningen: SAFE FOODS Project, c/o RIKILT – Institute of Food Safety, Wageningen University and Research Centre. European Commission (2005). Impact assessment guidelines, SEC(2005) 791. Brussels: European Commission. . Foresight (2007). Tackling obesities: Future choices—Project report. London: Foresight Programme, Government Office for Science, Department of Innovation, University and Skills. . Hammitt, J. K. (2005). Willingness-to-pay measures of food safety regulatory benefits. In S. A. Hoffman & M. R. Taylor (Eds.), Toward safer food: Perspectives on risk and priority setting (pp. 241–260). Washington, DC: Resources for the Future Press. Henson, S. J., & Loader, R. (1999). Impact of sanitary and phytosanitary standards on developing countries and the role of the SPS Agreement. Agribusiness: An International Journal, 15(3), 355–369. Henson, S. J., Loader, R. J., & Traill, W. B. (1995). Contemporary food policy issues and the food supply chain. European Review of Agricultural Economics, 22, 271–281. Henson, S. J., & Traill, W. B. (1993). The demand for food safety: Market imperfections and the role of government. Food Policy, 18(2), 152–162. Hoffman, S. A., & Taylor, M. R. (2005). Toward safer food: Perspectives on risk and priority setting. Washington, DC: Resources for the Future Press. Koenig, A., & Kaiser, M. (2008). Considering ethics in the risk analysis of foods: A structured approach (report prepared for deliverable 6.7). Wageningen: SAFE FOODS Project, c/o RIKILT – Institute of Food Safety, Wageningen University and Research Centre. Koenig, A., Kuiper, H. A., Knudsen, I., Pascal, G., Smith, M., Kleter, G. A., et al. (2008). An introduction to risk analysis of foods. In: A. Koenig, H. A. Kuiper, G.A. Kleter (Eds.), SAFE FOODS: Overarching report on the work of the integrated platform SAFE FOODS (report prepared for deliverable 6.6). Wageningen: SAFE FOODS Project, c/ o RIKILT – Institute of Food Safety, Wageningen University and Research Centre. Kolstad, C. D., Ulen, T. S., & Johnson, G. V. (1990). Ex post liability for harm vs. ex ante safety regulation: Substitutes or complements? American Economic Review, 80(4), 888–901. Mathers, C., Vos, T., & Stevenson, C. (1999). The burden of disease and injury in Australia, PHE 17. Canberra: Australian Institute of Health and Welfare. NICE (2004). Guide to the methods of technology appraisal. London: National Institute for Clinical Excellence. . OECD (2002). Costs and benefits of new approaches to food safety regulations. Paris: Organisation for Economic Co-operation and Development. Robberstad, B. (2005). QALYs vs. DALYs vs. LYs gained: What are the differences, and what difference do they make for health care priority setting? Norsk Epidemiology, 15(2), 183–191. Szende, A., Oppe, M., & Devlin, N. (2007). EQ-5D value sets: Inventory, comparative review and user guide. New York: Springer. World Bank (1993). World development report 1993: Investing in health. New York: Oxford University Press. .