This blog post is available for download as a PDF here.

Summary

  • Trend-following strategies – such as managed futures and tactical equity – have historically provided crisis alpha against sustained drawdowns.
  • For short-horizon events (e.g. single day, week, or month events), the effectiveness of these approaches in managing risk is largely based on the luck of prior positioning.
  • For more constant protection, option-based strategies can be applied by investors. However, the cost of constant insurance can be excessive, particularly as volatility levels spike.
  • Finding that VIX-based strategies offer some of the best cost/benefit trade-off, we introduce a VIX ETN-based strategy designed to exploit the short-term momentum and long-term mean-reversion exhibited by the VIX.
  • From January 2007 to December 2016, we find that adding an 8% position to a backtested dynamically weighted VIX ETN strategy was able to reduce downside risk (CVaR) of an all-equity portfolio by 20% while simultaneously adding 27bps per year to portfolio return.

With equity valuations at elevated levels and continued concerns around rising rates, the time is right for investors to start looking beyond fixed income to manage equity tail risk.

Historically, trend-following strategies like managed futures and momentum-driven tactical asset allocation have performed well as tail risk hedges.  These strategies perform particularly well in extended sell-offs like the global financial crisis and the dot-com bust.  As a result, we believe these strategies deserve a place alongside bonds in most risk management plans.

Trend-following, however, it not a risk management holy grail.  These strategies are reactionary by definition.  They are not well-suited to provide consistent protection against rapid sell-offs that materialize out of the blue (e.g. October 1987).

For these short-lived crises, it can be tempting to look towards the constant protection of options-based strategies.

But are they worth it?

To evaluate these strategies, we will use an approach similar to the one used in this SSgA paper.  As our risk measure, we use 10% conditional value-at-risk (“CVaR”).  10% CVaR is the expected return over the worst 10% of cases (monthly returns in this case).  For the period we considered, the 10% CVaR for the S&P 500 total return index was -8.4%.

For each option-based strategy, we then find the allocation needed to reduce risk by 20% (i.e. what allocation, when combined with the S&P 500, reduces 10% CVaR from -8.4% to -6.7% = -8.4% * 80%).

For this protection, we measure the cost of each strategy as the amount of performance drag it incurs relative to an unhedged S&P 500 portfolio.  The lower the performance drag, the better.

We consider the following strategies:

  • Collar: Buy a 3-month put option (struck approximately 5% out-of-the-money), financed by the sale of a 1-month call option (struck approximately 15% out-of-the-money).
  • Zero Cost Put Spread: Buy a 1-month 2.5%-5% put spread. This put spread consists of buying a put option 2.5% out-of-the-money and selling a put option 5% out-of-the-money.  The put spread is financed by selling a call option that will offset the cost of the put spread.
  • Put: Buy 1-month put options struck 5% out-of-the-money.
  • Short-Term VIX Futures: Buy long positions in short-term (one and two month) VIX futures.
  • Mid-Term VIX Futures: Buy long positions in mid-term (four, five, six, and seven month) VIX futures.

 

Note that we are purposefully only considering very simple strategies, generally mirroring benchmark indices published by CBOE.  There are certainly more sophisticated approaches that can potentially outperform the simple strategies we consider here.

To provide context, we also compute the results for an allocation to short-term U.S. Treasuries.

For both VIX futures strategies, we consider both an overlay (where futures are purchased on top of an existing equity position) and a cash-collateralized version.  The cash-collateralized version would be akin to gaining exposure through an ETN like VXX or VXZ. In this case, notional equity exposure will be less than 100% of portfolio value.

Data Source: CBOE, S&P.  Calculations by Newfound Research.  Past performance does not guarantee future results.  Returns are gross of all fee.  Data covers the period from January 2007 to December 2016.  Negative numbers imply that the hedged strategy underperformed unhedged exposure to the S&P 500.  Analysis is based on hypothetical index returns and is backtested.

Data Source: CBOE, S&P.  Calculations by Newfound Research.  Past performance does not guarantee future results.  Returns are gross of all fee.  Data covers the period from January 2007 to December 2016.  Analysis is based on hypothetical index returns and is backtested.

Performance Data: Unhedged vs. Hedged S&P 500
January 2007 to December 2016

Data Source: CBOE, S&P.  Calculations by Newfound Research.  Past performance does not guarantee future results.  Returns are gross of all fee.  Data covers the period from January 2007 to December 2016.  Analysis is based on hypothetical index returns.  Analysis is based on hypothetical index returns and is backtested.    

(Note: Skew and kurtosis both refer to the shape of the distribution of returns. A negative skew indicates that the left tail is longer (more negative returns). Kurtosis relates to the fatness of the tails. The values reported are the excess kurtosis. Values greater than 0 indicate that the distribution has fatter tails than a normal distribution.)

Why do these options-related strategies create so much performance drag?  Because being long volatility is expensive.  And unfortunately, options generally become even more expensive when they are needed the most.

Data Source: CBOE.  Calculations by Newfound Research.  Past performance does not guarantee future results. 

Having to roll over a put position in the midst of a financial crisis is akin to trying to buy fire insurance as your house burns down.

Performance drag is especially high for the collar and put spread, even relative to the put only strategy.  This is somewhat ironic since these strategies explicitly seek to control the cost of downside protection by selling a portion of equity upside via writing call options.

While the strategies using VIX futures generally outperform those that directly utilize options on the S&P 500, only mid-term VIX futures provide downside protection at a lower cost than short-term Treasuries.

Yet, VIX futures are still not perfect.  The VIX futures markets are normally in contango. Contango, characterized by an upward sloping term structure where longer-term futures contracts trade at a higher price than shorter term futures contracts, leads to negative roll yield that can cause major erosion of capital.

The S&P 500 VIX Short-Term Futures Index (tracked by VXX) and the S&P 500 VIX Mid-Term Futures Index (tracked by VXZ) have lost an annualized 42.5% and 11.9%, respectively, from January 2007 through the end of 2016.

Adding some short VIX exposure to the mix

The good news is that there is a relatively simple way to isolate many of the benefits of long VIX futures exposure in times of market crisis while also managing the cost of holding the position over the long-run.  And we can do so using liquid exchange-traded products.

All we have to do is follow a three-step process:

  1. Start with a strategic allocation between VXX (iPath® S&P 500 VIX Short-Term Futures ETN, long exposure to VIX futures) and XIV (VelocityShares Daily Inverse VIX Short-Term ETN, short exposure to VIX futures). Specifically, we will use a 50/50 blend of VXX and XIV.At first blush, this might seem counterproductive as these positions will at least partially offset each other.  Stick with us though.  We will get to the rationale in the bit.
  2. Regularly rebalance back to the strategic allocation. In this example, we will initially use a quarterly rebalance frequency.
  3. To mitigate timing risk, we will sub-divide our portfolio and use an overlapping portfolio methodology. Specifically, we use 13 overlapping portfolios, so that we rebalance one per week on a rotating 13-week schedule (note: holding for 13 weeks is approximately one quarter).

Later in this piece, we will consider other strategic allocations and rebalance frequencies as a robustness check.

We can also run this strategy with 2x levered long exposure to VIX futures by replacing VXX with UVXY (ProShares Ultra VIX Short-Term Futures ETF).  In this case, the equivalent strategic allocation to the 50/50 for VXX/XIV is 33.3/66.7 UVXY/XIV.

Data Source: CBOE.  Calculations by Newfound Research.  Past performance does not guarantee future results.  Returns are gross of all fees. Indices tracked by VXX, VXZ, and UVXY are used to compute performance.  Returns are hypothetical and backtested (in the case of the strategy indices). 

Data Source: CBOE.  Calculations by Newfound Research.  Past performance does not guarantee future results.  Returns are gross of all fees. Indices tracked by VXX, VXZ, and UVXY are used to compute performance.   Returns are hypothetical and backtested (in the case of the strategy indices). 

Both the VXX/XIV and UVXY/XIV versions outperform all of the other strategies considered.  Over the last decade, the VXX/XIV version would have achieved the 20% downside risk reduction with a 2bps increase in annualized return.  This risk reduction would have required a 12% allocation to the XIV/VXX strategy.  The UVXY/XIV version would have increased annualized return by 74bps per year and required an 8% allocation. The increased returns are akin to your insurance company paying you for the policy.

Data Source: CBOE, S&P.  Calculations by Newfound Research.  Past performance does not guarantee future results.  Returns are gross of all fee.  Data covers the period from January 2007 to December 2016.  Returns are hypothetical and backtested (in the case of the strategy indices). 

Performance Data: Unhedged vs. Hedged S&P 500
January 2007 to December 2016Data Source: CBOE, S&P.  Calculations by Newfound Research.  Past performance does not guarantee future results.  Returns are gross of all fee.  Data covers the period from January 2007 to December 2016.  Returns are hypothetical and backtested (in the case of the strategy indices). 

Why does this approach work?

The reasons are threefold.

First, the systematic rebalancing takes advantage of VIX’s mean-reversionary tendencies.

Second, the short VIX futures position will at least partially offset the negative roll yield normally embedded in long futures positions.

Third, when volatility spikes do occur, the portfolio will become more and more tilted towards the long volatility side of the trade.  To the extent that the spikes continue, this can lead to sizeable gains.

Data Source: CBOE.  Calculations by Newfound Research.  Past performance does not guarantee future results.

In asset management, almost nothing comes free.  So what’s the catch with this approach?

The answer comes down to what happens between quarterly rebalances.  Between rebalances, we are essentially running a momentum strategy, albeit one that reacts much faster than your typical trend follower.  When the VIX spikes upwards, we will get longer volatility, putting ourselves in a position to profit if the spike continues.

The bad news is that we potentially give up protection to short-term, temporary bouts of volatility.  In our view, we are willing to make this trade-off as these short-lived events haven’t been the ones that typically caused the most severe and pervasive drawdowns in capital.

Checking for robustness

Checking for the robustness of any quantitative strategy is crucial to ensure that we haven’t fallen victim to data mining.  In this case, we perform the check by computing the performance drag of various alternative versions of the VXX/XIV and UVXY/XIV strategies.  Specifically, we vary both the strategic allocation between long and short futures positions and the rebalance frequency.

In the following tables, the shading of the cell is determined by the size of the allocation to the VIX strategy required to reduce CVaR by 20%.

  • Orange means an allocation greater than 50%
  • Yellow means an allocation between 25% and 50%
  • Gray means an allocation between 10% and 25%
  • Blue means an allocation less than 10%

An N/A implies that there was no allocation to that given strategy that would have achieved the desired risk reduction.  For a point of reference, our previous analysis was done using the 50/50 allocation rebalanced every 3 months (in the middle of the tables).  Positive numbers indicate that S&P 500 exposure with the hedge included outperforms the unhedged S&P 500.

Performance Drag of S&P 500 Hedged With Variations of VXX/XIV Strategy vs. Unhedged S&P 500 Exposure
January 2007 to December 2016Data Source: CBOE.  Calculations by Newfound Research.  Past performance does not guarantee future results.  Returns are gross of all fees. Indices tracked by VXX, VXZ, and UVXY are used to compute performance.  .  Returns are hypothetical and backtested. 

Performance Drag of S&P 500 Hedged With Variations of UVXY/XIV Strategy vs. Unhedged S&P 500 Exposure
January 2007 to December 2016Data Source: CBOE.  Calculations by Newfound Research.  Past performance does not guarantee future results.  Returns are gross of all fees. Indices tracked by VXX, VXZ, and UVXY are used to compute performance.  Returns are hypothetical and backtested.   

For both tables, as we move to the right we are using a strategic allocation with more and more long VIX futures exposure.  It’s not surprising given our previous results that biasing increasingly towards long VIX futures exposure will increase performance drag.

Too little VIX exposure, however, can also be damaging.  It can either entirely eliminate downside diversification or can dilute it to such an extent that unreasonably large allocations would be needed to offset equity risk.

Evidence suggests that splits between 40/60 and 70/30 for the VXX/XIV strategy and between 25/75 and 54/46 for the UVXY/XIV strategy are reasonable depending on relative preferences between risk and return.

In terms of rebalance frequency, the data is a little bit harder to interpret.  Generally speaking, we know that if we rebalance too often we won’t be able to take advantage of momentum when volatility spikes.  However, rebalancing too infrequently will lead to a portfolio that will eventually be dominated by XIV – which would be disaster for a strategy that is supposed to act as an equity hedge.

What makes interpreting the data hard in this case is the time period we studied.  As we mentioned above, the risk of less frequent rebalances is that the strategy becomes overweight XIV.  However, with the global financial crisis occurring at the beginning of the period, there wasn’t enough time for any of the iterations to get too out of whack prior to the market crash.  Intuitively, we find a rebalance frequency in the two to four month range to be reasonable.

Can we do better?

The most obvious way to potentially improve on the strategy would be to do away with the static strategic allocation.  Instead, we might think about varying the strategic allocation depending on market environment.

For example, we know that the VIX tends to mean revert (i.e. low values are followed by high values and vice versa).  So we might tilt the portfolio towards long volatility exposure (VXX or UVXY) when VIX is low and towards short volatility exposure (XIV) when VIX is high.

We will limit the strategic allocation to between 40/60 and 60/40 for the VXX/XIV strategy and the equivalent range of 25/75 to 43/57 for the UVXY/XIV strategy.

The strategic allocation is determined using the current percentile of the VIX relative to its historical values.  To avoid hindsight bias, we only use data available at each point in time.

The dynamic VXX/XIV strategy outperforms the static 50/50 by an annualized 199bps.  From a performance drag perspective, the dynamic strategy is also better, adding 27bps per year to return for a 20% risk reduction relative to a 2bps return increase for the same risk reduction with the static 50/50.

Data Source: CBOE.  Calculations by Newfound Research.  Past performance does not guarantee future results.  Returns are gross of all fees. Indices tracked by VXX, VXZ, and UVXY are used to compute performance.  Returns are hypothetical and backtested. 

The dynamic UVXY/XIV strategy outperforms the static 33/67 by an annualized 161bps.  From a performance drag perspective, the dynamic strategy is also better, adding 86bps per year to return for a 20% risk reduction relative to a 74bps return increase for the same risk reduction with the static 33/67.

Data Source: CBOE.  Calculations by Newfound Research.  Past performance does not guarantee future results.  Returns are gross of all fees. Indices tracked by VXX, VXZ, and UVXY are used to compute performance.  Returns and hypothetical and backtested. 

Implementation

VelocityShares currently offers three ETNs tracking strategies that are similar to the static UVXY/XIV strategy

  • BSWN: VelocityShares VIX Tail Risk ETN à Strategic portfolio is biased towards long VIX futures.
  • LSVX: VelocityShares VIX Variable Long/Short ETN à Strategic portfolio is neutral to VIX futures (long/short sides offset).
  • XIVH: VelocityShares VIX Short Volatility Hedged ETN à Strategic portfolio is biased towards short VIX futures.

The downside of these exchange traded wrappers is the cost.  All three carry a hefty 2.3% expense ratio, which includes both a management fee and a futures spread fee tied to cost of trading VIX futures contracts. As a result, a direct SMA implementation may be preferable for some investors.

Conclusion

While trend-following approaches have historically performed well during periods of sustained drawdowns, they often fail to manage downside risk during short-lived market crises.  The ability for a trend-following strategy to protect against events like October 1987 is largely predicated on luck, not skill.

While option-based strategies can offer a certain amount of guaranteed insurance against immediate downside losses, they are often prohibitively expensive.  In particular, if options need to be rolled during bouts of volatility, it is akin to trying to buy fire insurance while your house is burning down.  Needless to say, it is going to be expensive.

We find that of the many option-based approaches we explore, mid-term VIX futures (both cash collateralized and via a futures overlay) offer some of the best protection at the lowest cost.

For ETF investors, we introduce a strategy that leverages both long VIX and inverse VIX ETNs to harvest the short-term momentum and long-term mean-reversion exhibited by the VIX.  Over the last decade (1/31/2007 to 1/31/2017), we find that implementing such a strategy in an all-equity portfolio with an allocation sufficient to cut the downside risk (CVaR) by 20% resulted in a positive contribution to performance.

Finally, we introduce a dynamically weighted approach, seeking to more actively exploit the mean-reversionary nature of the VIX.  We find that over the backtest period, this approach further increased portfolio performance for the same level of protection.

For investors looking to move beyond traditional fixed income as a short-term crisis alpha hedge, we believe that simple strategies employing VIX ETNs may be a solution.

Justin is a Managing Director and Portfolio Manager at Newfound Research, a quantitative asset manager offering a suite of separately managed accounts and mutual funds. At Newfound, Justin is responsible for portfolio management, investment research, strategy development, and communication of the firm's views to clients.

Justin is a frequent speaker on industry panels and is a contributor to ETF Trends.

Prior to Newfound, Justin worked for J.P. Morgan and Deutsche Bank. At J.P. Morgan, he structured and syndicated ABS transactions while also managing risk on a proprietary ABS portfolio. At Deutsche Bank, Justin spent time on the event‐driven, high‐yield debt, and mortgage derivative trading desks.

Justin holds a Master of Science in Computational Finance and a Master of Business Administration from Carnegie Mellon University as a well as a BBA in Mathematics and Finance from the University of Notre Dame.