Uncertainty in Impact and Externality Assessments - Implications for Decision-Making (13 pp)

Goal, Scope and Background Many disciplines, amongst them LCIA, environmental impact and external cost assessments, are often faced with evaluating trade-offs between two or more alternative options in terms of a range of incommensurable indicators. Using process modeling and valuation, these indicators are quantified at mid- or endpoint levels. Recent discussion amongst LCA experts showed that because of the mutually exclusive aspects of uncertainty and relevance, the midpoint/endpoint debate is controversial and difficult to reconcile. This article is aimed at a more quantitative analysis of mid- and endpoint impacts, and the implications of uncertainty for decision-making. Methods The consequences for decision-making of uncertainties of endpoints are analysed quantitatively for the example of ExternE results, by employing statistical hypothesis testing. The Analytic Hierarchy Process (AHP) is then used to demonstrate the use of multi-criteria techniques at midpoint levels. Results and Discussion Statistical hypothesis testing at the endpoint level shows that for the ExternE example, probabilities of mistakenly favouring one alternative over another when they are in reality indistinguishable can be as high as 80%. Therefore, the best estimate of external cost is inadequate for most policy making purposes. Indicators at midpoint levels are more certain, but since they are only \proxy attributes\, they carry a hidden uncertainty in their relevance. Conclusion If endpoint information is too uncertain to allow a decision to be made with reasonable confidence, then the assessment can be carried out in midpoint terms. However, midpoint indicators are generally further removed from people's experience, and less relevant to the question that people actually want to solve. Nevertheless, if this ultimate question is unanswerable (within the certainty required by the decision-maker), a decision can be made on the basis of stakeholders' subjective judgments about the more certain midpoint levels. The crucial point is that these judgments are able to intuitively incorporate many aspects that impact modeling and valuation has trouble quantifying, such as perceived risk, distribution of burdens and benefits, equity, ethical, moral, religious and political beliefs and principles, immediacy and reversibility of potential impacts, voluntariness, controllability and familiarity of exposure, or perceived incompleteness of human knowledge.