Forget mean/variance Whereas Peter Schwartz’s approach is in the domain of business strategy, Susan Gosling’s scenarios analysis is focused on investment strategy: how institutional investors should allocate to different asset classes to get the best future return and risk outcomes. It too involves imagining narratives about potential futures. However, it is more structured in that it requires explicit forecasts of investment returns from each candidate asset grouping for each scenario. History is often used as a guide to developing these forecasts, but judgment is critical, and the approach encourages the imagination of futures that have not been encountered in the past. Before working with Susan on this concept, I would look at a historical series of multiple decades as a guide to estimating the statistical distribution of returns for different asset classes. This history would tend to drive the covariance matrix assumed in the asset allocation process. I would guesstimate the means of the return distributions using forward-looking judgment, based on the notions such as economic growth, productivity improvements, real interest rates and the equity risk premium. I then used a simple form of scenario analysis, in which I would perturb the assumed means (and sometimes the components of the covariance matrix) in a type of sensitivity analysis that recognised uncertainty inherent in the assumptions.

My purpose was to discover a set of candidate asset allocations that were robust under a range of conditions and recognised the known issues in the mean/variance model, and in the assumptions used by it. In other words, I was guarding against the “garbage in/ garbage out” phenomenon. Susan turned that approach on its head. She eschewed mean/variance analysis, which essentially assumes that investment returns conform to a simple parametric distribution, such as the normal distribution. Instead, return distributions in Susan’s approach are non-parametric. They are built from the bottom up using judgment in the form of plausible narratives about a set of different future worlds, and how asset returns may look in those different futures. The result is a non-parametric distribution of returns, often with the leptokurtic characteristics observed in historical returns. How do you come up with scenarios? Start with getting some thoughtful people in a room, with plenty of good coffee and post-it notes for a no-holds brainstorming session. Many of the techniques Peter Schwartz discusses in his book are useful. Create narratives.

Make it fun. Don’t let anyone say: “The chances of that happening are slim”. The idea should be to flush out the possible, not the probable. Do this as a group in an environment of creative dialogue. You should also inject external thinking into the process if possible, but only after the creativity of the team wanes. You should guard against external ideas stifling originality. With that caveat, here are some ideas I recently came across in a survey conducted by The Economist Intelligence Unit. That team developed 24 scenarios and asked 800 respondents to assess the likelihood of each, and its impact on their investment portfolio. The six scenarios viewed as most likely were: 1. Further political turmoil in the Middle East; 2. The Internet and social media are a catalyst behind rapid political and economic change around the world; 3. Pension funding crisis deepens in developed countries; 4. High inflation forces policy tightening in emerging markets; 5. Widespread social unrest caused by rising food and commodity prices; and 6. Oil price spikes to US$150 a barrel. Interestingly, only the second scenario was viewed as having a positive impact on investments.