Tuesday, January 10, 2012

VIX as a Predictor of Equity Returns

Marketwatch has a post Cash is still king, at least for now, which points to a model that goes to cash when the VIX is above 20:

Consider a hypothetical portfolio that switched in and out of the Wilshire 5000 index according to whether the VIX was above or below 20 — investing in the market on a given day if the VIX closed the previous session below that level, and otherwise staying in cash. This portfolio would have produced an 8.9% annualized return since 1990, when the CBOE’s data for the VIX commence, in contrast to 8.5% for buying and holding. (I chose 20 as the threshold level for illustration purposes only; it is not far from VIX’s median level over the last two decades.
I thought I'd take a look at the results using daily VIX levels and SPY data for daily equity performance. As SPY didn't launch until January 1993, the comparison isn't apples to apples in terms of equity index or timing, so I can't vouch for the accuracy of the returns in the paragraph above, but I can state the model didn't work as anticipated for the data I pulled. More important, it did show some interesting results that I will share.

The chart below shows the different "buckets" of VIX levels I utilized, which were selected based on getting a close to even distribution across these buckets (as can be seen each grouping had between 500 and 700 days of returns to analyze).

Then I simply took the daily 1-day forward returns (including dividend) associated with each bucket and calculated annualized returns and standard deviation based on a 252 day trading year.

The result?

While volatility was significantly less when VIX was lower (showing volatility is sticky), returns appear to have been better at extreme readings, perhaps when securities were selling at a huge discount. Also of note, performance appears to have been worst when VIX levels read between 17.5 - 20 and 22.5 - 25, while performance was best (in sharpe ratio terms) when the VIX read between 15 and 17.5.

A further breakout of bucket '30+' is shown below (note these buckets are not even, as most data points fall within '30-35').

My thoughts based on data from 1993 - present... when the VIX creeps higher from teens to low 20's, be cautious. When it flashes red, if you can stomach EXTREME volatility (and EXTREME drawdowns) it may be a good time to buy.


  1. I own the VIX short-term ETV, a great way to hedge the market in my view.

  2. Good spot check on that model. It's conclusion definitely is surprising given you should improve returns if you are able to buy at or close to deep market discounts.

    One thing - what's the comparison to a model that simply dollar cost averages every month over the selected time periods?

  3. Dr. Duru- average cost buys since 1993?