Author: Ron Dembo
In our last blog How to look ahead in a radically uncertain world, we saw how subject experts can help us identify risk factors and their causality. And we also saw how from their knowledge and experience these experts are able to make informed estimates about future states of the world.
Where they disagree, and by how much, corresponds to the level of uncertainty surrounding a particular risk factor. But how do we read this uncertainty? How can we understand what this collective wisdom means for us as risk thinkers?
Understanding an Uncertainty Distribution
For complex issues where we are acquiring the judgement of perhaps several hundred or even thousands of experts through polls or machine learning, we need a way to organize their responses and display them visually. Only by doing this can we appreciate the majority and minority views, the groupings, and, most importantly, the level of uncertainty that exists among them.
We can do this with a distribution graph, like the one below, which was created by surveying experts on the cost of renewable energy in 2050 – a highly uncertain outcome that will be a risk factor in many companies’ strategies for the future.
From this graph, we can see the whole range of possible future values for a risk factor, as viewed by our experts. And, importantly, we can see where the consensus lies and the weighting behind it, and also where the extremes lie and their respective weightings.
Crucially, this distribution captures uncertainty over the particular issue at the moment in time when the survey was taken.
Looking at the Extremes
If we were to manage risk in this range that our experts predicted, we would be likely to perform well. This range is extreme, however, compared with any kind of risk management that is currently done with standard risk tools. And that’s one of the most important benefits of obtaining our expert judgements: they provide us with a good indication of the extreme upsides and the extreme downsides that are possible – they have a strong ability to encompass the full range of possible futures.
Although only a very small number of participants suggested extreme values, as risk thinkers we pay the most attention to these tails of the distribution. After all, when we are dealing with radical uncertainty, we know that it is the extremities of the distribution – the multiple-digit standard deviation values – that have the greatest impact despite being the most infrequent. That is where the nastiest and the best possible events occur, so that is where we must focus.
What Do We Do with these Uncertainty Distributions?
Each of the multiple factors that contribute to an issue under analysis will hold the possibility for radically uncertain events, so we must ensure we are taking account of these when we do our risk thinking. Our next step is to build this expert judgement and the uncertainty it contains for each factor into a set of scenarios that can describe the full range of potential future outcomes. In doing so, we create a formal, structured approach to scenario generation that is able, with some limitations, to capture both the black swans and the white elephants that are so highly significant in risk management.