From Danny Willett to the financial crisis, why unlikely events confound forecasters so often
Life imitates art, as the sporting world has shown this week. The Grand National was won by a horse which had never previously won a steeplechase. The US golf Masters was won by Danny Willett, who nearly did not take part at all because of the birth of his son. With only nine of the 72 holes remaining, last year’s winner Jordan Spieth held a massive five shot lead, but blew it.
In soccer, Leicester City has secured a place in next season’s Champions League, and looks certain to clinch the Premier League title. The last team apart from Arsenal, Chelsea or the two Manchester teams to win was Blackburn in 1995. The most successful Rugby League side in history, Wigan, was taken to pieces 62-0 in the third heaviest defeat of its entire 121 year existence by Wakefield, a club which escaped relegation last season by the skin of its teeth.
Every single one of these events would have offered long odds against beforehand. The 33-1 of the National Winner represented by far the most likely of the four examples. Wigan has played well over 5,000 games in total, yet was slaughtered by a team it usually beats comfortably.
These things capture our interest precisely because they are unexpected. On a much more serious note, the financial crisis of the late 2000s also came as a complete surprise to most. It was a shock to financial institutions to discover that asset prices follow what is called a fat-tailed distribution. Very large changes are indeed rare events. But they are hundreds or thousands of times more likely to happen than on the assumptions which both banks and regulators made in the run up to the crisis. In their models, very large price changes could effectively never happen at all.
Humans do seem to have difficulties in anticipating the unusual. The Survey of Professional Forecasters (SPF) in America publishes information on a wide range of economic variables predicted by numerous outfits in the country. A centrepiece is the forecasts for quarterly GDP growth, where there is a track record going back almost 50 years.
One quarter ahead, the predictions are a bit like the curate’s egg, both good and bad in parts. The good bit is that, averaged out over every single quarter since 1968, the forecasts are spot on. But very large errors can be made in predicting any specific quarter. For example, when the recession had already begun, the average prediction reported by the SPF for the final quarter of 2008 was a fall in GDP of 1.1 per cent at an annualised rate. The actual number was a collapse of 8.2 per cent.
Recessions fortunately remain rare. But even just nine months into the future, the SPF has never predicted negative GDP growth at all.
I once tried to give up forecasting for Lent, but it is a difficult addiction to cure. Perhaps we should all employ an artist or science fiction writer to help us envision unlikely events.