Reconstructing Volatility-Based ETPs and Some Observations

Exchanged-traded products that offer short or long exposure to volatility have exploded in popularity since their introduction in 2009 with AUM now in excess of $2 billion.  Perhaps more indicative of their popularity is their liquidity.  VXX, which offers long exposure to short-term (1 and 2 month) VIX futures, has traded more than 72 million shares a day on average over the last three months.   At this volume level, each dollar of VXX changes hands 3.5x per day on average.  To put this into context, SPY turns over about 0.2x per day.

Any analysis of trading strategies using the popular volatility ETPs is somewhat constrained by the relatively limited trading history (5 years or less in most cases).  Fortunately, only three pieces of information are needed to approximate the performance of the products prior to launch and all three are available for free.

  1. INDEX METHODOLOGY: Published by S&P (
  2. FUTURES DATA: Published by CBOE (
  3. LIST OF MARKET HOLIDAYS: Published in various locations, such as

With these three pieces of information, we are able to approximate the price movements of VXX (long short-term volatility), VXZ (long mid-term volatility), XIV (short short-term volatility), ZIV (short mid-term volatility), and any other product that tracks one of the S&P VIX Futures indices back to March 2004.

Below you can download a spreadsheet that mirrors the code we use to reconstruct the indices tracked by VXX and VXZ.  [Note: This is similar to a spreadsheet put together by Michael Stokes at of MarketSci Blog for VXX.]


As you can see from the following graphs, we are able to very closely replicate the closing prices for VXX and VXZ.  Note that we did not take into account ETP expenses in calculating our indices.  XIV and ZIV can be reconstructed by taking the inverse daily returns of VXX and VXZ, respectively.

Some quick observations (in all cases below when we say VXX we are referring to our reconstructed estimate of VXX that goes back to March 2004):

  1. $100,000 invested in our VXX in March 2004 (assuming it had existed) would be worth about $156 in January 2014 (annualized return of -48.2%)
  2. VXX has never had a positive 5-year trailing return
  3. VXX has been up in 36 of 117 months since March 2004, with an average monthly return of -3.8% and a monthly standard deviation of 19.5%
  4. VXX tends to be much more volatile on the upside than the downside (27.9% upside semi-deviation compared to 14.8% downside semi-deviation, computed assuming 0.0% expected monthly return so as to not bias the upside semi-deviation number due to the significantly negative expected return)

Some might argue that we shouldn’t be analyzing VXX as a standalone product, but rather as tool to hedge equity exposure in a broader portfolio.  After all, most insurance (auto, home, etc.) has a negative expected return in isolation, but this doesn’t stop people from holding it.  Why?  Because people look at the bigger picture.  They consider buying a home and insurance on the home as a package deal, preferring the combined return stream to taking the risk of disaster if the home is seriously damaged and no insurance is purchase (of course, this assumes that to insure or not to insure question is up the homeowner, which isn’t always the case).

Alas, even when we use VXX within a long equity portfolio the results suggest that VXX doesn’t have a permanent place within investor portfolios.  The following scatter plot presents annualized drawdown vs. Ulcer index for portfolios of SPY and VXX ranging from 0% VXX / 100% SPY to 100% VXX / 0% SPY in 5% increments.  [Note: The Ulcer index is an interesting metric that measures both severity and persistence of drawdowns over a period of time, with higher values suggesting larger and/or more persistent drawdowns.  Look for a separate blog post in the near future explaining this metric further.]

We see that:

  1. The positive return from holding SPY is completely wiped out once the VXX position exceeds 15%.
  2. Once the VXX position exceeds 18%, the Ulcer index is higher than that of a pure SPY position.  In other words, VXX actually increased downside risk in the portfolio historically if the allocation exceeded 18%.
  3. The return-to-risk ratio (annualized return to Ulcer index) is maximized between March 2004 and December 2013 when VXX receives a 0% allocation.  No amount of fixed VXX allocation over this time period would have given investors more return per unit of risk.

I could go on and on citing statistics regarding VXX, but the moral of the story is that VXX is a TRADING, not an INVESTING, instrument.  The same can be said for most other long volatility products, although to a slightly lesser degree for those that play farther out on the volatility term structure (i.e. VXZ).