There are so many economic indices nowadays that finding two displaying a short-term relationship is indicative of little more than human beings’ continuing wish to be able to predict the future
Société Générale’s Albert Edwards, a market commentator we respect very much here on The Value Perspective, recently wrote his weekly strategy note on the dangers of stifling dissent and the comfortable ‘groupthink’ that can come about as a result. On the first page was the following graph and – in a show of dissent that ought to please its originator – we have some issues with it.
Should we be worried?
Source: SocGen 4 May 2017, Datastream. Data Sep 2015 - May 2017.Past performance is not a guide to future performance and may not be repeated.
As you can see, one line on the graph shows the progress of the S&P500, the main US equity market, over the last 18 months or so while the other illustrates the effect of US economic data surprises over the same period (how the data is turning out versus expectations). Readings above the line point to positive surprises. Until recently, the two lines have taken a pretty similar course but the Surprise index has just fallen off a cliff.
Clearly the implication – underpinned by the graph’s title ‘Should we be worried?’ – is that the two indices have some sort of relationship and so the US stockmarket is set to follow suit. As it happens, Edwards has been predicting a market crash for some years now – a consistently downbeat outlook that has gained him the nickname ‘Dr Doom’ – but might one really now be around the corner?
Well, maybe the market will shortly follow the path of the Economic Surprises index and maybe it won’t. If it does, however, we have serious doubts here on The Value Perspective that the above graph will have any bearing on the matter. This is not because we disagree with the underlying data – the graph shows what it shows – but we do take issue with the way the data is presented.
In stockmarket terms, 18 months is not much more than a blink of an eye and so finding an apparent relationship between two indices – what is known in investment as ‘correlation’ – over such a period cannot really be seen as an argument for or against anything. To underline the point, take a look at the following graph, which shows a very similar series of data but over a much longer period of time.
Citi Economic Surprises (US) v S&P 500
Source: Bloomberg, 16 May 2017. Past performance is not a guide to future performance and may not be repeated.
And, as you can see, there is now neither rhyme nor reason to the relationship, which suggest that – while there may indeed have been some short-term correlation, there was never any causal link. What the above graph is then, really, is another example of the ‘p-hacking’ we touched on in our ever-expanding and increasingly inaccurately named Jellybean Trilogy.
The ‘p’ in that term stands for the 0.05 or 1-in-20 threshold below which any scientific or research finding becomes statistically significant and, while we are in no way suggesting any intention to mislead with the above graph, with so many thousands of economic indices alone to compare, there are always going to be some time periods where two appear convincingly correlated.
In truth, however, the future of short-term market movements – and indeed of everything else – remains stubbornly hard to predict.