One of my favorite fellow research nerds is the extremely smart Charlie Bilello, Director of Research at Pension Partners. His year in review of 2016 highlighted many examples of failed forecasts, predictions, and expert opinions throughout the year. For example, Charlie shared these forecasts by RBS, Jeff Gundluch, and Dennis Gartman near the beginning of the year:
Bilello quotes Jeff Gundluch: “If you’re going to do anything in emerging market equities, my recommendation is to short them. They may fall a further 40%.” Emerging Markets, by the way, ended up over 10% on the year and bottomed out in mid-January around the time of Gundluch’s quote.
Bilello continues by referring to Dennis Gartman who “said crude oil wouldn’t ‘see $44 again” in his lifetime.” At the time, oil was trading at around $30. Bilello goes on to write, “Crude would end April over 77% from its February low, one of the largest short-term rallies in history (the spike in 1990 was larger). Yes, it was back above $44.”
Other seemingly impossible events occurred in 2016, surprising many experts and forecasts, such as the Brexit vote and the Trump election, as well as negative yields on 50-year bonds in Switzerland.
Bilello sums up the game of predictions and forecasts well at the end of his review: “Be forever humble and thankful, and leave the predictions to those whose job it is to entertain.”
“Experts aren’t that bad, right?”
They’re all bound to be wrong every now and then. Well, CXO Advisory has fortunately been tracking expert forecasts for quite some time, more than 6,500 of them from 1998 to 2012, and it doesn’t look good for the experts.
The following chart tracks the inception-to-date accuracy of all 6,582 graded forecasts in the sample. The extreme values early in the sample period relate to small cumulative samples. Terminal accuracy is 46.9%, an aggregate value very steady since the end of 2006, even as the number of forecasts continues to grow.
That’s right — only a 46.9% terminal accuracy rate. So, listen to expert forecasts or flip a coin? Take your pick.