"The Fear Index"

"Dear investor. Over the past 3 years, Hoffmann has generated Alpha of 112%. That''s why we twice have been voted Algorithmic Hedge Fund of the Year. It's now our opinion that our new system, VIXAL-4, is robust enough to cope with portfolio extension.... I agree, we can't analyze every aspect of human behavior in the markets and its likely trigger over the past 20 years, however much data now is available and however fast our hardware scans it. We realized from the start we would have to narrow the focus right down. The solution we came up with was to pick on one particular emotion for which we have substantive data : Fear".

Yet another brilliant book by Robert Harris. But the novel really kicks off just minutes after this investors' meeting when compliance enters the scene during the risk committee : "What we really need to focus on is the level of our risk exposure once the markets rally; it's not only the size of our exposure, tens of millions in options since Tuesday night, but the fact that they are all huge one-way bets ; - Next to our shorts on individual stocks, how many contracts has VIXAL accumulated on the VIX ? - Twenty thousand ; frankly, I am seriously worried : we are long gold, long dollar, short every equity futures you can imagine, our book has gone all out of shape"

And at that point I started wondering about a lot of things. I won't open the debate on technical analysis and self fulfilling prophecies. But this High Frequency trading has every ingredient to exponentially force markets to escalate. Recent stats on their growing importance show that they cover already at least 60% - 70% more likely - of daily volume on the US stock market : all pre-programmed trades by super computers based upon historical correlation, executed and settled in nano seconds without the intervention of human action. But there is more.

The huge one-way bets also reminded me on the remarks on CDO algorithmics (Gaussian Coppula) and its inventor David Li. After the debacle, he testified himself : "The most dangerous part is when everybody believes everything which comes out of the model". And in addition, the more people in the market rely on the same model, a new correlation risk appears with all structures designed the same way. Hence, if there is a storm, all boats capsize the same way.

Applying the previous remark on this half fictive, half true story. Algorithmic high frequency trading relies very much on correlation. And the algorithm is conceived in such a way that it can "learn". Which most likely will have tendencies to create even more stronger correlations in the market and potential strong flash-crashes. Some months ago, we blogged on RO-RO, the so-called risk-on-risk-off trades and performed a simple cross-correlation matrix on various financial assets. Already then, we had some evidence on strong significant correlations between various assets and with the sample, we also had a majority of one-way correlations amongst a set of various variables (AUD/JPY, S&P index, EUR/USD, bank spread, CRB commodity index etc). I performed the same matrix today on 2011 and we came to the same hard conclusion : highly significant cross-correlations and coefficients which are high as well (cfr infra, apologies for the quality of image)

But the thing which bothers me the most, a paradox but may be not in view of the fact that we are dealing here with human beings : we are "scientifically" programming our own "flash crash" annex "black swan" events.