Volatility Comes In Clusters


Mandelbrot had it right when he said, “large changes tend to be followed by large changes, of either sign, and small changes tend to be followed by small changes.” Don’t believe him? Well its true! To illustrate this point, I’m posting three graphs of clustered volatility for General Electric (GE), Intel (INTC), and Cisco Systems (CSCO).

INTC-Clusters-062007You might wonder, why this fascination with volatility clustering? Why should you care? The answer is quite simple. If you understand that stocks and the markets go through cycles of high and low volatility, you can position yourself appropriately.

Being able to predict volatility clusters or identify a genuine buying opportunity (this is the basis for my S&P500 Volatilty Timing Model (PDF)) from just noise can make you rich. Still not convinced? Just graph in Excel the absolute (or squared) % returns for any asset of your


choosing over a long time period. You can use daily, monthly, even 5 minute time periods. You’ll see the same thing over and over again, periods of low volatility are clumped together and periods of high volatility are clumped together.

[tags]GE, INTC, CSCO, Volatility, Clusters, Investing, 401k, Timing[/tags]

  • Christian Gross

    >Being able to predict volatility clusters or identify a genuine buying opportunity

    I agree that there are clusters, but that is also apparent in randomness. If you use a good random number generator you will have periods where the random walk does nothing, and then has a burst of energy only to subside again.

    This is why I do feel trend followers do ok, because they will cash in on the cluster of randomness.

  • http://www.neuralmarkettrends.com Tom


    It’s vastly easier to identify genuine buying opportunities using volatility clusters than predicting when they happen. I’ve been timing my market purchases in my 401K for years using this method and have found that you’re lucky to have maybe one damn good buying opportunity per year (based on my proprietary model).

    Knowing where you are the cluster series allows you to take profits/close positions or enter new positions even if the series is completely at random. I tend to disagree that volatility is random, I believe it follows some sort of power law distribution but I still need to research that, its just a “gut feeling” right now. :)

  • Marius Kjeldahl

    In numerous articles on your blog you refer to your S&P500 Volatilty Timing Model, although I have not seen any details about what the model actually attempts to predict or how to profit from it. I’m not after the details (assuming you want to keep those for yourself), but I would like to know if you in your “volatility timing model” also predict market direction, or if you use it together with other models for figuring out which side of the market to play? My impression is that volatility spikes on market drops, and if you then mostly go long you are assuming mean reversion or similar?

  • admin

    My volatility timing model is nothing more that gauging “better” buy points using a combination of broad market indicators and indicies. In the old days, I used to buy funds for my 401k using the VIX as a signal.

    When the VIX spiked over 20 then I would add cash to my fund positions. Conversely, as the markets sold off in post 2000, I averaged down using the same VIX indicator.

    It became clear to me that during the recent period of low volatility we are now in, this method doesn’t work as well. So I developed another model that helps me identify better “buy” points. Sometimes spikes in the VIX proved to be less desirable buy in points then I wanted.

    How to profit from this? Well my investment horizon is several years ahead and I can deal with short downturns. You make your money when you buy an asset at a discount and when everyone is selling, I can buy at a discount.

    Hope that helps explain it better!

  • kjeldahl

    Well, it helps somehat. From what I understand spikes in volatility (as seen in the indices measured by VIX) usually happens when markets fall significantly, and simplified you usually treat these spikes as “buy points (under the assumption that the trend is up). On and individual stock level I would guess spikes in volatility could just as well happen when prices increase drastically, so general spikes (without direction) would be less useful.


  • kjeldahl

    Your next post on AMGN also provides more clues. Thanks.

  • admin

    kjeldahl: not only do I use these vol. spikes as buy points during up trends, I do it during down trends in the S&P500 or other large indices. I’m happy to ride the trends up or down if my time frame is several years to retirement.

    If I was positioning myself in a stock for a few days or weeks then I would care a lot more about the prevailing trend.