Monday, December 8, 2008

The Chop!

These are truly interesting times. Since October 1st the average intraday range of SPX based on percentage of the opening price has been 5.99%. For the period from 1962 until September 30th, that same value has been 1.44%. Rounding off the October 1 – present period to make a 50 session average of this intraday range finds that there has been no period since 1962 that has seen continued intraday volatility on a level remotely like this.

The last peak was in October 2002 just a bit above 3% and the period around the 1987 crash was 3.50%, though that latter reading is skewed considerably by a 3 day run of 20.47%, 12.96% and 9.5%. Excluding these three sessions and the volatility in the days before and after is essentially unremarkable. Here’s a plot of the 50 day average of SPX intraday range as percent of day open:

Some keen observers will be saying to themselves, “Hey, that looks a whole lot like the plot of the VIX and SPX.” And of course, they would be right for reasons I will get to below. But just so those clever folks can pat themselves on the back for something, here’s the chart of the VIX and SPX. (People that can recognize a VIX chart on sight need all the support they can get, after all.)

The reason that these charts are essentially identical (except for the noise in the VIX plot) lies in how the VIX is calculated. On this particular topic, it really pays to read the CBOE VIX white paper that lays out the calculation method. While the calculation itself is straight-forward, it will likely cause some head-scratching the next time you hear someone telling you that the VIX is predicting this or that for the markets. Here’s the basic overview with a few details left off for brevity:

The VIX is essentially a summation of the call/put average option prices (weighted for strike and time) on the SPX that is limited by two consecutive strike prices with zero bids. (For example: with SPX at 900, there might be no bids at 840 or 850 for puts and so 860 would mark the lowest term limit included in the VIX calculation. The call side works conversely.) The practical consequences of a calculation like this should be immediately obvious. When the SPX covers an intraday range of 6% as in the past couple of months, the number of strikes that will attract bids grows considerably and thus the terms summed in the equation grows. There have been numerous articles written and observations made about mean-reversion in the VIX. While the observation is more or less correct, in theory at least, the VIX has no upper limit.

So that’s all the VIX is – a weighted summation of a varying number terms based on the prices derivatives traders are willing to pay for contracts. This is only predictive if you believe that derivatives participants tend to be overly complacent during good times and too willing to pay too much for a hedge during bad times. But that assertion deserves a closer look so the next chart will show the VIX and SPX closes since the creation of the VIX. (It is worth noting here that the calculation of the VIX has changed in that time – the CBOE white paper details the date and nature of the changes.)

The correlation in the long term is 0.1716 – not really a strong value and more interestingly implying a direct rather than inverse relation. But that’s not being very fair to those people that believe the VIX has more to tell over a shorter term. In that respect they are correct: over any given 5 trading periods, the average VIX/SPX correlation is -0.6560. This is a pretty good result and the negative value is expected considering the typically inverse nature of the relation. However, this does not necessarily mean there is any causative or predictive nature to this relation. For that, offset data of the VIX and SPX closes can be used.

For me, a simple way to check whether or not one series is more predictive or reflective is simply to offset the data sets by varying periods and check the correlations against the original set. The inclusion of a chart for visual reference also helps (though not included here). As noted before, the average 5 period VIX/SPX correlation is -0.6560 and the entire set correlation is 0.1716. The table below shows the correlation rapidly breaks down to essentially none in the average 5-period correlation. The entire set correlation actually rises incrementally. To me, this implies little of predictive value and again shows that the VIX is simply reflecting the range of SPX.

Bottom line, using the VIX as a predictive tool on its own is of dubious merits and could lead to far worse. There are perhaps some possibilities in using some TA on the VIX to produce more reliable signals but the VIX on its own... you'd do as well flipping a coin.

- Data used is current up to December 5th.