Producing Scenarios by the Hundred: How Statistical Approaches are Transforming Foresight Methods

Most foresight exercises requiring a quantitative assessment of the scenarios proceed in two stages. They begin with a qualitative exploration of determining factors, from which a limited range of sets of hypotheses are chosen, producing a –similarly limited– range of scenarios. A mathematical model is then applied which translates the selected hypotheses into input data, in order to quantify the various scenarios identified. Nevertheless, as this article stresses, the increasing complexity of contexts and the possible interplay between the different factors at work in the various fields of study have cast doubt on the appropriateness of this mode of operation. The representativeness of hypotheses selected is no longer any guarantee of the quality of the scenarios deriving from them. This is why, in this article, Guivarch and Rozenberg propose a different way of proceeding, retaining all the determining factors identified in the field of study and running the models hundreds or thousands of times to get as many scenarios as possible from them. Qualitative choices are then made in a second phase, in the “target-space” of these scenarios. Guivarch and Rozenberg outline this alternative method here, which consists in building databases of scenarios in order to explore the areas of uncertainty. They also indicate the advantages this represents for foresight studies and present an actual example of the use of this method in the field of climate change economics.