Tuesday, 26 January 2016
Oil and stocks. Correlations that work until they don't.
There comes a time when the level of correlation hits a point where people start to notice. They will trade one asset on the back of moves of another asset because the correlation tells them that the two will move in lock step. In doing so they create a positive feedback loop of tighter and tighter correlation.
Algorithmic trading relies on correlations and there are correlations out there that only algorithms have seen and trade on. Some of them will make absolutely no sense to humankind as humankind attempts to satisfy its desire to justify correlations. The correlations algorithms trade on may be so obscure that the human brain would not even bother to look under the stones they are hidden as the number of stones to look under is almost infinite and humans like to apply filters to make the search easier. But computers? Well computers can look under stones that our logic filters would have discarded as duds. Humans have a desire to understand or justify the correlation before they trade on it. Why? Because justification is the excuse humans raise in case of failure. Without excuse there is little chance of being excused. Many years back we spotted a correlation between USD/DEM and the Icelandic fish catch. In fact it was better than USD/DEM vs Oil. Did we trade on it? No way. We would have been laughed out of the manager's office should we have had to excuse our losses.
Many years ago my friend Gerald Ashley (@Gerald_Ashley), told to me that ‘Correlations work really well - until you put the trade on’. I have never forgotten that and often quote it, normally to 12yr old quants. This statement may sound as though it flies in the face of my original comment that correlations become reinforced by those trading on them, but what it really expresses is the point where that artificial forced correlation caused by the actions of the observer (their trading) can no longer hold down the reality that the two assets are not naturally correlated but had wandered together through chance and at some point will start to diverge again causing mass liquidations of all these herded correlation trades in a classic high volatility blow up.
I raise this thought because of the current "correlation du jour'. That of oil and global stock indices. This correlation is not just real in mathematical terms but also has handy justifications that can be strapped to it to keep the human mind happy too.
Since oil started falling a year ago there have been a few justifications for why it is bad for stocks and the world in general. Without going into them all, let's just say that low oil threatens leveraged debt in that sector that if defaulted upon would have a cascade effect through other asset classes and stop the flow of money. And as we know, stopping the flow of money causes 2008 like events.
The modelling of 2008 is currently very popular and here oil prices are being used as a correlating example to support the thesis that stocks and the economy are going to collapse. Because, look see, oil fell hard in 2008 and look what happened. Well no. Of course not, oil was falling then because of actual and assumed future fall in demand due to a massive slow down of the economy led by balance sheet collapse, added to which was a huge unwinding of leveraged longs that had been chasing an explosive price spike. What happened then was completely different to today’s reason oil is falling which is that supply has exploded and exceeds demand, even though demand is actually rising. We cannot assume that there will be another 2008 because of oil being in massive supply. Quite the opposite.
But back to the Oil/SPX relationship. It has been in place for while as the leveraged debt notion persists fortified with a Chinese story that suggests recession and a collapse in oil demand. All the while the correlation is getting reinforced. We are now at a point where every asset class is looking at the oil price as an indicator.
Odd things can happen when large things are pushed around by small things. Now whilst the oil market is vast it doesn’t take that much to push the price round in $ terms relative to moving the price of, say, every stock in all the countries whose indices appear to be following oil. Which means that if you want to move a stock market at the moment all you have to do, rather than buying the huge volumes of stocks that would normally be required, is to buy a relatively small amount of oil and have the rest of the market do your job for you on assumed correlations. In simple terms . Buy liquid S+P500 in size and not move the market against you, buy oil in a market moving size (SPX follows), Sell SPX, Sell oil. Which I am wondering if some people (or trading algorithms) may now be doing. Looking at the price action over the last 24 hours and one could wonder.
But back to Gerald’s ‘Correlations work really well - until you put the trade on' comment. I think we are getting to the point where the correlation is going to break down between oil and stock in general. The secondary concerns about what a low oil price will do to leveraged oil debt has been priced. There is no new news there. If investments are destroyed to the point of restricting money in circulation then the Central Banks will act to counter and will restore it. But more importantly, oil does not normally correlate so closely to the stock market, if it correlates to it at all, so the artificial correlation imposed upon stocks and oil due to traders acting and reinforcing that correlation could well be due a blow up. This has been running a while now and when I hear of traders looking at oil for a guide as to where telecom stocks are going I think something is wrong.
It may sound sacrilegious to suggest that some correlations don’t deserve the 'r squared' they are worked out upon, but if I told you that some people think that the world’s economy is based upon iPhone sales would you still tell me I am mad?