Cost-benefit analysis and disagreements

b ] j b 0 iff W j b, a ] W j b 0 , a. where W j b, a = i N V j v i j n j x i x i b + n j y y b n + V j Z Z b 2.3 for all a and all b B. Eq. 2.3 can alternatively be written W j b, a = i N V j v i j n j x i x i b + n j y n j x i y b n + V j Z Z b 2.4 which, under certain conditions, corresponds to a weighted sum of individuals’ net willingness to pay willingness to pay minus costs, plus the welfare effect of the intrinsic value variable rights, religious considerations. This expression will be useful below. Throughout the paper, I will assume that deci- sion-makers have sufficient a priori knowledge of society to determine their subjective welfare weights V j v i j , n j x i , n j y and V j Z. 12 What they do not know, and need information about, are the impacts of alternative projects. However, we have to take into account that peo- ple have a limited ability and time to receive and understand information. Project reports from an- alysts cannot contain an unlimited amount of detailed information. Formally, assume that each decision-maker is capable of handling a maximum of K pieces of information regarding each project, where 1 B K B . 13 A piece of information is defined as a single number with an interpretation attached to it. For example, the statement ‘the average increase in the income of single mothers will be 2’ is a piece of information according to this definition. 14 If decision makers are presented with too much detailed information, they are assumed to be unable to distinguish the data they need, and are thus unable to determine W j b,a. The aim of the analysis below is to identify sets of information which enable any decision-maker to evaluate a project in accordance with her own normative views, provided that the same informa- tion must be given to all j J. Such information will be termed a sufficient welfare indicator set Brekke et al., 1996. Definition 2.1: a sufficient welfare indicator set for project b is information that enables any decision maker j J to determine W j b,a, for b B.

3. Cost-benefit analysis and disagreements

In a traditional cost-benefit analysis without explicit welfare weights, projects are ranked in accordance with aggregate net willingness to pay. I will disregard the substantial practical problems of eliciting individual willingness to pay for environmental goods, and simply assume that marginal willingness to pay is observable. 15 This section discusses the conditions under which aggre- gate net willingness to pay is a sufficient welfare in- dicator set — in other words: When can partici- pants in the debate form their own judgements on the basis of aggregate net willingness to pay? 13 K may be exogenously given as a natural limit to people’s cognitive capacity, or it may be the result of an optimizing process which for our purposes must be regarded as made ‘once and for all’, not being revised for every new decision problem if perceiving information is costly. 14 Here, one must exclude the possibility of using codes, like using the first digit of a number to indicate the value of one variable and the second digit to indicate another. 15 In practice, measurement problems caused by misunder- standings, strategic responses, and other factors are substan- tial. Taking these problems into account would make the analysis below even less favorable to monetary valuation. 12 Alternatively, status quo information may be provided to decision-makers before the discussion of alternative projects starts this description may have to be more thorough than that of the projects. Some social welfare functions do not require knowledge of the status quo to determine the weights V j v i j , for example, utilitarianism; however, some knowl- edge may still be required to determine n j x i andor n j y. To analyze this formally, define U i b = u i x i b,yb 3.1 as a numerical representation of individual i’s revealed or, in the case of contingent valuation, reported choices. Then, net benefits as estimated by the unweighted cost-benefit analysis can be formalized as W CB b, a = i N x i b + u i y u i x i y b 3.2 where u i y u i x i is individual i’s marginal willingness to pay for the environmental good. The cost-benefit test will generally evaluate projects in accordance with decision maker j’s individual judgements only if W j b = m j DW CB b for all b B, where m j is a positive constant. 16 From Eq. 2.4 above, we can see that this will be the case if and only if each of the three following conditions is fulfilled: V j v i j n j x i = V j v k j n j x k for all i, k N 3.3 n j x i , y; a i = C j [u i x i b, yb] 3.4 where C j is a monotonously increasing function, and V j Z Z b = 0 3.5 These conditions may need some explanation. The essence of the first is that the marginal effect on social welfare when i’s income increases is judged to be the same, regardless of who i is. This holds for a utilitarian decision-maker who be- lieves that everybody has the same marginal well- being of income. If a decision-maker believes that the marginal well-being of income varies between individuals, condition 3.3 requires that an indi- vidual well-being is weighted inversely propor- tional to that individual’s marginal well-being of income. Thus, if an extra dollar is more important to poor people than to the rich, the interests of the poor must count less in social welfare evalua- tions than those of the rich. Emphasizing income effects for low-income individuals is hence incon- sistent with 3.3, unless one believes that the marginal well-being of income is increasing in income. The second condition says that the decision- maker accepts revealed choice as an ordinal mea- sure of an individual’s well-being. Nobel laureate Amartya Sen 1977a, 1985 has repeatedly pointed out that revealed preferences may reflect other concerns than the individual’s own well-being. Sen maintains that individuals may act out of what he calls commitment: They may sometimes sacrifice personal well-being concerns to do what they believe is morally right, even if they never get any monetary or other compensation that is, even in the form of a good conscience for this. Thus, the second condition is also controversial. 17 The third condition means that non-welfaristic concerns must not be relevant for the decision- maker. Many relatively common political or ethi- cal views appear to be at odds with this criterion; for example, the belief that loss of well-being resulting from self-induced events should count less than losses that the individual had no oppor- tunity to avoid, or the view that animals have rights. It follows from the above discussion that aggre- gate net willingness to pay, W CB b, does not constitute a sufficient welfare indicator set, except under the very strong assumption that all three of the above conditions hold for every single deci- sion-maker. If a decision-maker disagrees with at least one of the conditions, the cost-benefit analy- sis will not in general evaluate alternative projects in accordance with her own views, and she cannot 17 In the present model, every decision-maker j actually has two preference orderings, represented by Wjb and v j j b. These preferences may conflict in the sense that W i b \ W i b 0 does not necessarily imply v i i b \ v i i b 0 . In a con- tingent valuation study, some respondents may respond using their social rather than personal preferences Sagoff, 1988; Vadnjal and O’Connor, 1994; Nyborg, 2000, in which case WTP will not generally be an ordinal measure of personal well-being. 16 If one only requires a ranking of projects, not a cardinal measure of DW j , then it is sufficient that DW j = f j DW CB , where f j is a monotonously increasing function. However, a cardinal measure may sometimes be needed, for example, when considering whether to strategically support another project than the most preferred. determine W j b on the basis of the net benefits estimate alone. Thus, only some decision- makers can be expected to use the net benefit estimates to rank projects. Note that this result holds even in the case of purely benevolent deci- sion-makers who act rationally to maximize social welfare, taking no other political or personal con- siderations into account in their objective functions. Using explicit distributional welfare weights does not solve the problem when there is norma- tive disagreement, unless one performs a separate analysis for every normative view — regardless of which weights one applies, some individuals will disagree with the choice of weights. It is frequently claimed that unweighted cost- benefit analysis can be defended on the basis of any welfaristic social welfare function, provided that the status quo income distribution is socially optimal see Johansson, 1993. When many indi- viduals participate in the decision-making process, however, this would require that e6ery decision- maker regards the income distribution as optimal i.e. 3.3 holds for all j J. In addition to being implausible, this may even be a logical impossibility. Further, it is often pointed out that if costless lump-sum transfers were available, undesired dis- tributional consequences could be compensated ex post. However, consider a decision-maker who supports a project with to her unacceptable dis- tributional consequences, because she believes these effects can be dealt with ex post. If other decision-makers disagree with her distributional preferences, however, they may choose not to support her ex post redistribution proposals. With normative disagreement, it will not generally be possible to redistribute income according to every decision-makers’ judgement. Knowing this, ratio- nal decision-makers will have to take distribu- tional concerns into account in the project evaluation itself, even if lump-sum transfers are technically feasible. Note also that the above results were derived for marginal projects, implying that distributional effects need not be dramatic for the above results to hold. Even if costs are marginal and equal for everyone, and with no intrinsic value variable, the ranking of projects will generally depend on the chosen welfare weights. Formally, it can be demonstrated that if one knows nothing about the social welfare weights, then two projects can be ranked, if and only if project one strictly Pareto dominates the other. This claim is stated as a proposition and proven in Appendix A. The above results are supported by the inter- view survey among Norwegian Members of Par- liament by Nyborg and Spangen 1996 and Nyborg 1998. 18 We found that cost-benefit ra- tios were rarely used to rank projects; most re- spondents said they needed more information. Moreover, leftist politicians were considerably more skeptical of cost-benefit analysis than those farther to the right one leftist respondent stated explicitly that the weighing of interests implied by cost-benefit analysis did not match hisher own views, and that this explained hisher skepticism. If left-wing politicians are less content with the current income distribution than other politicians, their skepticism may be quite reasonable. Most politicians in the survey did find cost- benefit analysis useful, but not as a device for ranking projects. A large number of road invest- ment projects were under consideration, and while politicians had to depend on the road administra- tion’s judgement to some extent, they looked for indications of projects requiring closer political attention. Cost-benefit ratios were used as such a signal; if a very unusual cost-benefit ratio high or low was reported for a project, some politicians would ask for more details. When given addi- tional information, however, they appeared to make up their minds quite independently of the cost-benefit ratio itself. Moreover, the cost-benefit ratio was clearly not the most important of such warning signals; in particular, indication of local disagreement was used almost universally as a trigger for closer political attention. Fortunately, most cost-benefit analyses report more than just aggregate net willingness to pay. 18 The survey was based on in-depth interviews with all 16 members of the Norwegian Parliament’s Standing Committee for Transport and Communication, and the interviews focused on the Committee’s treatment of a long-term national road investment plan. Nevertheless, the way such additional information is organized is extremely important. If it is too disaggregated, decision-makers may not be able to understand its implications i.e. the limit of K informational items is exceeded, and if the addi- tional data reported is too aggregated, some deci- sion-makers recieve insufficient information. The next sections are concerned with characterizing conditions under which information can be aggre- gated without losing important information, and also how information can be aggregated. In par- ticular, I will focus on whether it is desirable to value the public good and the intrinsic value variable in monetary terms.

4. Groups of equal welfare weights