The Grattan Institute’s transport and cities program director, Marion Terrill has been urging caution in the rush to build mega-infrastructure projects for a post-Covid recovery (The Age 09/09/2020). Support for her position comes from a formidable source. Danish geographer and social policy analyst Professor Bent Flyvbjerg, in a paper to be published shortly in the journal Environmental Science and Policy argues that, in the case of certain kinds of events, we cannot rely on accumulated data from the past to predict what will happen in the future. Among other kinds of events, he lists pandemics, bushfires and floods, all too familiar to Australians.
Flyvbjerg is the Director of the Major Project Management Programme at Oxford University. He made his name through exhaustive analysis of hundreds of multibillion megaprojects worldwide such as motorways, rail lines, airports and dams. He showed that, on average, such projects failed to meet their performance goals, projects ran late and over budget. Final benefits were overestimated and costs underestimated. While a high level of performance was always argued for each individual project, the mean for actual completed projects showed otherwise.
The statistical theory predicting such underperformance is called ‘regression to the mean’. In simple terms, over-optimistic assumptions about individual cases are invalid because they ignore the impact of unpredicted random effects on the outcome. A recent example in Melbourne is the discovery of high concentrations of PFAS chemicals in contaminated soil dug out for the West Gate Tunnel. But in all such projects there are factors that cannot be predicted which can be summed up as ‘luck’. The actual performance of megaprojects, Flyvbjerg argued, reverts to the historical mean for most actual projects.
Now, in new research, he challenges the premises of his own past work which was based on seeking the mean performance of megaprojects. He now argues that regression to the mean is meaningless when considering the risks of mega-projects affected by unpredictable future variables such as climate change and pandemics.
The reason lies in the unusual extremes now known to be occurring. The distribution of the performance of any large number of actual cases normally has the shape of a bell curve with a large hump in the middle and a ‘tail’ with extreme cases showing up at each end. We don’t need to worry about the positive tail where the outcome is much better than expected. We do need to worry about the negative tail where the outcome is much worse than expected. But Flyvbjerg goes further, arguing that in the rapidly changing context of pandemic or climate change risk we have to take account not only of past events but also of the future.
When we find evidence of new and larger extremes becoming ever more frequent in any distribution of events over time (what he calls ‘fat tails’), we need to anticipate the possibility that it will be these fat tails and not the mean that will give us clues to the future. Hence his theory of ‘regression to the tail’.
Flyvbjerg argues that where regression to the tail applies, prudent decision makers and their risk managers will do two things: reduce the tail by mitigation measures, and avoid tail risk by taking a cautious approach. He particularly targets measures to rebuild the economy after the Covid 19 pandemic. These measures include giant construction projects with ‘fat-tailed’ risks such as multibillion dollar megaprojects in IT, transport, energy, water, education, housing, health and defence. The financial risks increase with the uncertainty of the future, which in a post-Covid world afflicted by global atmospheric heating is very uncertain indeed.
As far as transport is concerned, Flyvbjerg, like Terrill, points out that lockdowns and stay at home measures have reduced traffic and pollution levels. The aim should be to prevent traffic demand from returning to pre-Covid levels, to turn the focus from high risk supply-side measures (freeways, massive rail projects to meet assumed demand) to demand-side measures such as encouraging working from home for at least part of the working week. I would add, following Flyvbjerg’s risk mitigation logic that low risk supply-side measures (which may also shape demand) such as providing safe infrastructure for cyclists and pedestrians, and serious investment in fast and effective bus transport across metropolitan cities and regions must also be included to relieve traffic congestion.