2.1. Social Multi-Criteria Evaluation
2.1.2. Multi-criteria evaluation
A multi-criteria structure consists in a set of alternatives conforming a decision space, which are evaluated under several criteria. Table 1 shows a problem structured in a a a A multi-criteria fashion: with n possible actions aj (j = 1, 2, ..., n) and m relevant evaluation criteria gi (i = 1, 2,..., m).
Criteria are at least partially contradictory, that means that one alternative can be the best under one criterion, but not necessarily under the rest of the criteria.
Table 1: Multi-criteria impact matrix
Criteria Alternatives
A1 A2 - An
g1 g1 (A1) g1 (A2) - g1 (An)
g2 g2 (A1) - -
-- - - -
-- - - -
-gm gm (A1) gm (A2) - gm (An)
There are several multi-criteria models, each one with its advantages and disadvantages: the lexicographic model, for instance, consists in applying criteria one-by-one to the set of alternatives, eliminating those options that don't fulfil the limit established by the applied criterion. The procedures continue until all selected criteria have been considered and/or until one alternative is selected.
Another sort of multi-criteria method is the ideal point approach. The idea is to find or to generate alternatives as close as possible to the ideal point: the alternative that makes discrepancies disappear. Another approach is to find the anti-ideal point, so defensible options should be as far as possible from the anti-ideal point.
Other methods are based on linear aggregation of criterion scores, which is done after a transformation of the performances by means of utility or value functions (See for instance, Keeney and Raifa, 1976). These methods are completely compensatory.
There are also the outranking methods (Roy, 1990), which are based on the concept of partial comparability. That is, the preferences between two alternatives can be modelled by means of binary relationships: indifference, strict preference, weak preference and
incomparability. These models do pair-wise comparisons of alternatives in order establish whether one alternative is at least as good as the other, according to most of the criteria. In order to do so, it is taken into account some of or all the following information:
• The number of criteria in favour of one alternative,
• The degree of importance of each criterion,
• The intensity of preference.
• The (binary) relations between the alternatives
The desirable features of a multi-criteria model in the public policy domain are discussed in Janssen and Munda (1999), Munda (2004) and Munda (2005a). In short, the model has to be as simple as possible to guarantee transparency. Non-compensation is desired to avoid that very good performances in some dimensions— for instance the economic one— overcome bad results in other dimensions— for instance ecological or social ones—, which could be
important for some social groups.
Compensation refers to the possibility that very good performances in some criteria can offset bad performances in other criteria. A Non-compensatory method doesn’t allow such a counterbalance between very good and bad performances.
In public policy problems it is useful to use indifference and preference thresholds10, which imply considering intensity of preference, and the use of weights as importance coefficients is also desired. But the mix of intensity of preference and weights leads to compensation and trade-off between criteria11. Weights as importance coefficients are used with ordinal
criterion scores to avoid compensatory aggregation procedures, a desired feature if we want to apply the Strong Sustainability principle based on weak comparability of values (for a discussion on these concepts see Martinez-Alier et al., 1998).
In methods using ordinal criterion scoring, and thus weights as important coefficients, the contribution of any criterion to the overall performance of an alternative does not depend on the intensity of preference. On the other side, in compensatory methods, the
contribution of any criterion to the overall performance of an alternative is proportional to the intensity of preference.
For instance, when comparing two alternatives with a compensatory method, the contribution of the criterion Cost of implementation will not be the same if i) cost of
alternative A=100 € and cost of alternative B=10 € or ii) cost of alternative A=100 € and cost of alternative B=50 €. The contribution of this criterion is higher in the case I). When
comparing both alternatives with a non-compensatory method, the contribution of this criterion to the overall performance of the alternatives is the same in both cases .
Moreover, within the SMCE framework the evaluation methodology should be coherent with the expected features of public involvement (improvement of democracy, appraisal of socio-environmental complexity, promotion of public acceptance of decisions and boost social learning).
Multi-criteria evaluation methods foster the practice of democracy by means of considering different criteria in the evaluation with limited or no compensation between them; it is an adequate framework to appraise complexity by means of considering several dimensions without reducing the effects of the alternatives to a single unit of measurement; and it serves as a communication tool boosting social learning.
10 The indifference threshold is the maximum difference between the criterion scores of two alternatives that makes no difference between them (under that criterion). The preference threshold is the minimum difference between the performances of two alternatives in one criterion that makes one option preferred instead of the other.
11 On the one side, weights as importance coefficients reflect the relative importance— given by the decision-maker, the analyst or the social actor— of one criterion in relation to the others. On the other hand, weights as trade-off reflect the substitution rate among criteria.
Next Chapter presents a case study in the Patagonia, Chile. It describes a SMCE process ,and compares it with the current environmental impact assessment system applied worldwide.