Statisticians are familiar with the “IntraOcular Trauma” test, first invented by Joe Berkson at the Mayo Clinic. Plot the data, and if the result hits you between the eyes, it’s significant. Fewer people have heard of the “Digital Nonlocability” test, because I just invented it, but it’s similar in nature — if you can’t put your finger on something in a 50 year chart, it’s probably not important.
Here’s a chart which I’ve removed the scales and titles from. I will tell you one thing about it, and that is that, according to the very well respected CEO of one of America’s largest banks, this chart has a “once in three billion years” event in it. Can you put your finger on that event?
Actually, I’ve just realised that putting your finger on it isn’t going to work here, because you’ll have to scroll down to find the answer, so the smeary fingerprint on your screen won’t be in the same place. So memorise where you thought the unprecedented event happened and check whether you got it right.
Yep, that tiny little insignificant bump is what Jamie Dimon was talking about in his letter to shareholders when he said:
It doesn’t make you lose your faith in statistics. It makes you seriously wonder whether he knows what he’s going on about.
Because, of course, as is quite clear to see, the movement in Treasury yields associated with the treasury flash-crash was not unprecedented at all, and wasn’t even all that big in the wider scheme of things, particularly when you think back to the days in which the bond market had real volatility in the 1970s and the Volcker years. The only numeraire which makes it look like a very big move is if you decide to scale everything by a rolling standard deviation with a relatively short window for calculation. But isn’t this a very good reason to not use that as your basis for measurement?
There’s a bit of institutional history to know about here. The “Value at Risk” measurement system was invented at JP Morgan. It was, specifically, invented as a way of reporting to the CEO a single number which summarised all the risks that the bank was taking across all its different portfolios. And it was based on rolling-window estimates of standard deviations. It turned out to be an absolutely unmitigated disaster as a risk management tool, because more or less immediately it was introduced, everyone involved with its production seemed to forget that it was only meant to act as a very rough summary statistic for daily reporting and to stimulate discussion, and started to act like the thing was an insight into reality itself.
And here we are, billions of dollars of lost money later, and the CEO of JP Morgan, with all of his risk team behind him, has convinced himself that 0.385 of a percentage point is a really big movement, that it demonstrates that regulatory policy is fundamentally misguided, and that even the basis of statistics themselves might be wrong. And he’s done so on the basis of a windowed standard deviation calculation. Two things here:
First, as James Saft says, when Jamie Dimon asks why his bank has to carry out stress tests on the basis of assumptions which include bad decisions made by other banks in the last crisis — well, this is why, Jamie, it’s because you say things like that. If this is the way you think about risk management, the Fed is very much within its rights to suspect that all the things you got right in 2007–9 were more by luck than judgement.
And second, if the worst thing that people can come up with to scare us about the possible unintended consequences of current regulation is the 2013 taper shock, then … well, as a Sum Of All Fears, it doesn’t look too scary, does it? The fear that Dimon is discussing here is the possibility that with the withdrawal of big banks from market making, trading capacity is not sufficient to allow all investors to adjust their positions when the monetary policy cycle turns. That might lead to a global liquidity crunch, along the lines of the 1994 bond market crisis.
Hmmm, the 1994 bond market crisis, eh?
There’s something going on here … I just can’t put my finger on it.