As an investor in volatility ETPs such as VXX, SVXY, XIV, and UVXY, I always keep an eye out for spot VIX. Granted, these ETFs do not track VIX, they track VIX futures. Most of their returns come from contango (backwardation) in VIX futures, and only a little bit from the changes in VIX. This is because spot VIX is far more stable than the futures during extended periods of contango (backwardation) where the front month future must come very close to spot VIX by expiration, the second month future can't be too high than the first month as otherwise it just has to decay more in the next month to match the then spot VIX, and contango (backwardation) starts off the futures far away from spot VIX.

Still, volatility ETP investors should watch the spot VIX value closely. If it is off too far from its reasonable value, there could be sudden moves in spot VIX itself that erodes the gain (loss) from contango (backwardation). One such period was in May and June when the Fed's trial balloon of an early QE tapering wrecked havoc on volatility ETPs. Order has been restored since, but it was not a pleasant time for volatility ETP investors as the two following charts show.

VXX Total Return Price data by YCharts

As the chart shows, A sharp increase in VIX turned the tables on volatility ETPs. The usually dependable guaranteed-to-lose-money VXX actually gained, and the usually dependable guaranteed-to-gain-money XIV actually lost. Of course, as VIX came back to normal levels since then, VXX has dropped sharply, and XIV has gained sharply as well.

So, VIX levels are actually important, even though random spikes are temporary and usually rectified in time. Right now, there is a lot of buzz about how VIX is too low by historical standards. True, as average VIX is around 19, and today VIX is around 13. However, in this article I will attempt to show that in fact, VIX has farther to fall than spike up.

While VIX reflects the averaged out implied volatility of a series of options on the S&P500 (with a complicated formula, which is not a strict average, but it will do for now for the purpose of this article), and as such reflects the future and not the past, in practice it is much related to historical volatility of daily changes in S&P500. Think of it this way: if S&P has been sharply going up and down, then the seller of the options would need more protection from massive swings coming up. Hence, when historical volatility is up, so is VIX, and vice versa. This doesn't make sense, as options sellers should be thinking of future events that can spike volatility (the coming debt crisis, for example), but in practice they just look backward. Past performance is no guarantee for future performance, but human beings being what they are, they extrapolate.

I went back to the origin of VIX in 1990 and calculated for each day the following historical volatilities.

- 2 weeks (HV10)
- 1 month (HV21)
- 6 weeks (HV30)
- 3 months (HV63)
- 6 months (HV127)
- 1 year (HV254)

HVx denotes the historical volatility (annualized) for daily changes in the S&P500 for the past x days.

I then calculated the correlation coefficient of all these with the spot VIX for the whole life of VIX. Here's what I got.

HV |
Correlation coefficient |

HV 10 | 85.4% |

HV 21 | 89.0% |

HV 30 | 89.4% |

HV 63 | 87.3% |

HV 127 | 81.3% |

HV 254 | 69.7% |

Clearly, all the near term HVs are very closely correlated with VIX, with the 6 week memory being the strongest. Can one predict the spot VIX using these HV values? A multiple regression is in order. I forced the regression line to go through 0, as, if the HVs are 0 for all these time periods, surely even the most nervous option seller wouldn't worry about any change. This is the result from Excel.

For those unfamiliar with multiple regression, don't worry. I will point out the salient parts.

Let's look at the R^{2} of the distribution in the first section. It is very high, at 97%. Which means this analysis is pretty accurate.

Please ignore the second section, unless you are statistically oriented.

Then let's look at the coefficients column in the last section. This means you can use the following formula to predict spot VIX using historical volatility values.

Spot VIX = 0.24*HV10 + 0.15*HV21 + 0.18*HV30 + 0.17*HV63 + 0.05*HV127 + 0.38*HV254

Note that this is not a written in stone forever formula. As more and more data points come in, it will change. But since this analysis is done over 23 years, it is almost as stable as it gets.

Of course, there is some error in statistical models, and we need ranges for predictions. Looking at the right hand side of the last section, you get the coefficients for 95% confidence intervals and 68% confidence intervals. To get to the spot VIX values at these intervals (which, incidentally, map to two- and one-sigma deviations from the average spot VIX) simply change the coefficients in the above equation to match that of the respective columns.

So this is a lot of analysis, but what does that tell us? Well, as of 7/31, these are the values for the various HVs.

HV |
Value |

HV 10 | 4.5% |

HV 21 | 7.0% |

HV 30 | 12.2% |

HV 63 | 12.2% |

HV 127 | 12.0% |

HV 254 | 11.9% |

In other words, historical volatility has been really, really low. There is no reason for option sellers to demand more premium. Plugging in these values the above formula for spot VIX yields the following:

Prediction |
Value |

95% conf. int. (lower) |
8.9% |

68% conf. int. (lower) |
10.2% |

VIX (predicted) |
11.6% |

VIX (actual) |
13.5% |

68% conf. int. (upper) |
13.0% |

95% conf. int. (upper) |
14.4% |

In other words, given the low level of historic volatility, there is a 95% chance that spot VIX will be between 8.9% to 14.4%, a 68% chance that it will be between 10.2% to 13.0%, and a "fair value" (if you will, based on the last twenty years' relationship between historical volatility and spot VIX), VIX should be at 11.6%. However, on 7/31, it was at 13.5%.

This to me indicates that if anything, spot VIX is too high and needs to fall towards the 12% range. On 8/1, spot VIX did fall, and now is at 12.9%, which puts it in the range of the 68% probability or one-sigma variance. Pretty handy model, shall we say, in predicting future values of VIX based on historical volatilities?

So, what does this mean? First, it means that average values of VIX over its 23-year history is completely bogus in deciding what it should be today. VIX depends on market conditions. If the market is stable, VIX should be low. Market stability is measured by HVs. The multiple regression analysis above can be used to predict reasonable range of values for VIX using HVs given current market conditions.

Next, and more importantly, it means that VIX has very limited upside from here. If anything, it will fall further. This, of course, is great news for volatility investors as the ETPs will continue to behave in the manner they usually do, i.e., the long volatility ETPs such as VXX and UVXY will keep dropping, and the short volatility ETPs such as SVXY and XIV will continue up.

Finally, it means that a reasonable value for VIX by the August expiration lies closer to 12 than 13. I had earlier built some models that show where the volatility ETPs will be at various values of VIX. Interested readers can look this article up and make investment decisions accordingly.

**Disclosure: **I am long XIV, SVXY. I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.

**Additional disclosure:** I hold puts on VXX and UVXY