What Is High Implied Volatility?

Readers generally ask much better questions than my rhetorical ones. For that reason, I will see if I can set aside more of my time in this space to answer reader questions.

Recently I received several questions about how to determine whether implied volatility is high and when someone declares IV to be high, what exactly the basis for comparison is.

Ultimately, the assessment of what is high implied volatility is a subjective one, but typically the person attaching the label is making a comparison between current implied volatility levels and a historical range of either implied volatility or historical volatility levels.

To my thinking high implied volatility is best determined according to the following criteria:

1. relative to a lookback period (1 year, 6 months, 3 months or whatever – but be careful with shorter time frames) for previous implied volatility levels
   
2. relative to historical volatility (one could argue that the time frames are less important here)
   
3. relative to the current implied volatility of peers or a broader group of similar stocks (you hardly ever hear about this, but I think the comparison is relevant)

The most important piece of information to remember is that implied volatility is inherently forward looking and historical volatility is, by definition, backward looking. This is important because traders know when potential market moving events are coming and implied volatility moves accordingly. With historical volatility, on the other hand, it is much too easy to drive a car right into an unseen wall while trying to navigate by looking out the rear view mirror.

When I think about implied volatility levels, I am usually looking at ‘relative volatility’ or implied volatility as a % of the most recent 52-week range. The recent volatility trend, if any, is also worth investigating. I always check the current IV-HV spread, but to switch metaphors, it is generally better to know about the hurricane headed your way than the one that has just passed through.

Finally, it is critical to know if important events are just around the corner, such as earnings, an FDA decision, the resolution of important litigation, etc. When evaluating implied volatility for ETFs or indices (and data sensitive stocks, such as financials), upcoming events to focus on would likely be more along the lines of impending government data, proximity to Fed meetings, etc.

Comments

[Bud Haslett]

There is some interesting work about the relative volatility spread on the CBOE website at:

www.cboe.com/Instituti...

If anyone is interested...

[CharlesC]

Thanks Bud

[User 179820]

Implied vol is the market price for options. It is only tangentially and imprecisely related to views on market movements, especially if an option is long dated. Try gamma trading a three year option and you realise that market movements are almost irrelevant.

[Chungst]

Any person who trades options knows there are at least three types of volatility: historical, implied, and theoretical. Implied volatility is simply what some formula spits out as in garbage in garbage out. Given the advances in computing power of late, you would think someone would have worked out a way to solve for theoretical volatility (of an underlying security) so we don't have to work with such a horrible metric like implied volatility.

N. Taleb already shown investors many times (at least in two books no less) the folly of looking at implied volatilities. Additionally, comments by Warren Buffett over the years has demonstrated that traditional measures of risks (i.e. market volatility or stock price volatility) are grossly incorrect.

[Smart ETF]

I agree with the most recent posts that IV is problematic. The first problem is: defining what is the best time frame of data to analyze. The second problem (not necessarily pertaining to options) is determining how best to weight the data, and the third is how best to forecast the data. Because there is no right answer, we have developed a solution using GARCH models to weight the data using ‘stable’ distributions (not ‘normal’) and then run a Monte-Carlo model (simulation) for forecasting. This more accurately describes historical volatility, negates the need for IV, and replaces theoretical volatility by forecasting volatility. It works well for us, perhaps it will help one of you -