Doesn’t seem to work on the way down, either

My friend, Dennis at ISI, directed some traffic to a piece I wrote five months ago leaning into the idea of using the Citi Economic Surprise index as a market timing indicator.  Why would the market react predictably to old news? It doesn’t seem to.

At the time I wrote the post, the surprise index had just moved higher, which got people excited the market would rally.  Ok, it ended up doing so, but seemingly not because of that indicator.  At short horizons, on my simple calculations, rises of the CESI above zero from meaningfully below have been a contrary indicator, although extremely unreliably, even as that.

In this note, I would like to address the current concern, that the decline of the CESI is a bad omen.  I don’t currently have a market view; been that way for a while. [1] But the CESI does not seem to work on the way down either.

I will get to the tedious task of “testing” the CESI in a second. But before turning to that, I would like to speculate on what might actually draw attention to this silly CESI thingy.  Take a look at this chart, which shows the Citi surprise index to yesterday.  Look at that steep plunge.  That looks pretty unnerving, eh?

Screen Shot 2017-05-19 at 8.30.11 AM

The line was high and now the line is low. But let’s back that up.  More than half of that decline, i.e. to the thick horizon line, was inevitable at some point, simply on the basis that the data would eventually be inline, the definition of unremarkable.

The remaining bit reflects that the data have missed recently. Or at least that they have missed according to the CESI hodge-podge, which includes inflation as well as activity indicators, and has the bizarre feature of assigning 90 different weights over time to one economic release.

But never mind that.  This index is built not to work on the System II part of your brain, but on the System I, the plains ape bit.  You see that steep decline and you think, oh my god, things are really going to hell here.  But just ask yourself, because you probably remember, was there some huge disappointment – or set of disappointments – in the data recently, which went miraculously unnoted by others?

The obvious retort is that by “quantifying” these issues we can identify an accumulation of evidence that people have actually missed. That is the marketing pitch. But there is no compelling reason to believe it is true or ever any evidence offered in support of it.

Ok, now to the dreary bit of testing it. I did one whole test, which is one more than you will see from advocates of the CESI. I looked at whether declines of the CESI from above +50 to below zero predict changes of the S&P500 index over 3 or 6 months. No, they do not – or have not.

Screen Shot 2017-05-19 at 8.30.21 AM

I looked at declines from above 50 because I wanted to eliminate whip saws, which typically do not attract the attention even of the street. I wanted to look at declines from “meaningfully” above zero to below. But you could easily imagine another threshold or even an infinity of other ways to test this.  I chose one.  Given how strong the priors are against the idea that such a silly thing would predict, choosing more than one test might be data mining. But by all means, fill your boots.[2]

[1] Or more precisely equities seem to me to be indistinguishable from fairly priced to deliver subpar returns. They looked that way 15% ago too. My ability to distinguish is pretty limited.

[2] You might also want to check my results.  I did not actually right a bunch of if statements to find when my condition was satisfied.  Rather, I just eyeballed the chart and physically looked at the data. Reinhart-Rogoff risk would be high here.