This post is available as a PDF download here.
This post ended up being more timely than we could have ever imagined as Credit Suisse announced that it would accelerate XIV's maturity after the ETN lost more than 90% of its value from Friday's close.
We are leaving the post up for transparency and to highlight the importance of fully understanding the risk of any investment. We tried our best in the original post to clearly describe the downsides associated with using inverse volatility strategies. We have a saying at Newfound that "risk cannot be destroyed, only transformed." Part of living with that statement in mind is being aware that (a) there are always risks associated with an investment and (b) just because you may not be able to spot the risk doesn't mean they aren't there. In fact, if you can't spot the risk, you are probably better off staying away entirely.
Two other observations:
- I had CNBC on for most of the day today. Predictably, the talking heads were doing quite a bit of extrapolating. They used XIV's failure as an opportunity to criticize ETPs generally. XIV behaved exactly as we would have expected given what went on in the market. XIV is designed to deliver inverse exposure to a weighted average of the front two months of VIX futures contracts. This is exactly what it did. The events of the last two days did not unearth some unknown risks of ETNs and it certainly should not cause any alarm for holders of plain vanilla ETFs like SPY.
- CNBC also placed plenty of blame on "leverage" generally. Leverage is just a tool. It isn't inherently good or bad. Sure, when used irresponsibly, it can lead to disaster. However, when used prudently and in moderation it can be very useful, especially when used to move investors towards more risk efficient, diversified portfolios. Just two quick examples of this. First, the levered ETF portfolio we discussed, despite being leveraged, actually outperformed SPY due to the diversification provided by Treasuries. Second, the most well-known risk parity funds, most of which use leverage, also outperformed equities due to their superior diversification profile.
- We believe that capital efficiency should remain a paramount objective for investors.
- The prudent use of leverage can help investors employ more risk efficient portfolios without necessarily sacrificing potential returns.
- Many investors, however, do not have access to leverage (be it via borrowing or derivatives). They may, however, have access to leverage via the variety of ETFs and ETNs available in the market.
- We explore three ways that investors could do this using high beta ETFs, levered ETFs, and derivative ETNs.
- Each method comes with its own set of risks, but these options provide investors with practical ways to convert high risk-adjusted returns into higher absolute returns without borrowing money or using margin.
Over the last few months, we’ve been advocating strongly for capital efficiency as a way to deal with potentially muted returns in major asset classes over the next decade. In Portable Beta: Making the Most of the Returns You’re Already Getting, we laid out four ways that investors can achieve greater efficiency:
- Reduce fees to take home more of what you earn.
- Express active views more purely so that we are not caught paying active management prices for closet beta.
- Focus on risk management by “diversifying your diversifiers” with strategies like trend following that can help increase exposure to higher return asset classes without necessarily increasing the overall portfolio risk profile.
- Utilize modest leverage so that investors can create more risk-efficient portfolios without necessarily sacrificing potential return.
In Levered ETFs for the Long Run, we focused on the last approach: introducing modest leverage. Acknowledging that many investors do not have access to traditional tools for achieving leverage (e.g. margin, derivatives), we explored the common criticisms of levered ETFs and demonstrated how they may be used in the context of a portfolio to introduce diversifying exposure without necessarily sacrificing returns.
Continued innovation in exchange traded products means that in reality there are actually a number of ways that we can introduce leverage into a balanced portfolio without any direct borrowing. In this week’s commentary, we evaluate three approaches:
- High Beta ETFs
- Levered ETFs
- Derivative ETNs
We will describe each of these strategies in subsequent sections.
To evaluate each strategy, we will use a 1.7x levered 60/40 portfolio as the benchmark. The 60/40 portfolio consists of a 60% allocation to the SPDR S&P 500 ETF (ticker: SPY) and a 40% allocation to the iShares 7-10 Year Treasury ETF (ticker: IEF). We choose 1.7x leverage as, with the benefit of hindsight, this leads to equity-like volatility over the period studied. We consider daily, monthly, quarterly, and annual reset versions of this benchmark.
We can see that the leverage reset frequency does not materially impact risk or return. Therefore, we will simply use the monthly reset version going forward in the interest of simplicity.
Strategy #1: High Beta ETFs
This strategy does not use any leverage or derivatives, but rather replaces SPY and IEF with higher beta stock and bond exposures. We replace SPY with the PowerShares S&P 500 High Beta ETF (ticker: SPHB) and IEF with the iShares 20+ Year Treasury ETF (ticker: TLT). In other words, we move from market-cap weighted equities to equities with betas in the top 20% of the S&P 500 and from intermediate to high-duration bonds Treasuries.
This approach to introducing leverage is far inferior to actual leverage of the benchmark 60/40 portfolio.
These results should not be too surprising. After all, we are taking the wrong side of the well-documented low volatility factor. Decades of evidence suggest that lower beta stocks and lower duration bonds offer a superior risk-adjusted return profile relative to their higher volatility counterparts. The lower Sharpe ratios of higher beta securities creates a significant headwind. As we increase our beta, and with it our portfolio volatility, our expected return per unit of risk goes down. Therefore, while we are taking more risk, our actual compensation for it may not even exceed the expectation of a non-levered portfolio.
A 60/40 portfolio of a low volatility equity ETF and a lower duration bond ETF and levered up to the same volatility as the 1.7x 60/40 portfolio would likely generate higher returns.
Strategy #2: Levered ETFs
In this strategy, we replace SPY and IEF with 2x levered versions. As an example, this strategy could be implemented using the ProShares Ultra S&P 500 (ticker: SSO) and the ProShares Ultra 7-10 Year Treasury (ticker: UST). For this analysis, we replicated SSO and UST since they did not launch until 2006 and 2010, respectively.
This effectively creates 2x levered exposure to a 60/40. As a result, it has significantly higher volatility than our 1.7x levered benchmark. To make our comparisons apples to apples, we blend the Levered ETF 60/40 with the unlevered 60/40 so that the overall portfolio volatility matches that of our benchmark.
The Levered ETF portfolio actually does quite well relative to the benchmark. In fact, of the 104bps lag in annualized return, about half can be attributed to the higher management fees on SSO/UST relative to SPY/IEF.
Strategy #3: Derivative ETNs
In this strategy, we replace SPY and IEF with ETNs that offer exposure to derivatives-based indices whose returns are correlated to those of traditional stocks and bonds.
On the equity side, we focus on inverse volatility strategies. Specifically, we will consider the index tracked by the VelocityShares Daily Inverse VIX Short-Term ETN (ticker: XIV). XIV offers exposure to a rolling short position in VIX futures and is therefore “short volatility.”
The correlation between monthly excess returns for XIV’s underlying index and SPY is 0.77. The beta of the index to SPY is 3.4 and so it certainly offers a form of leveraged exposure to the equity market.
On the bond side, we use the Janus Velocity Short LIBOR Index. This index is tracked by the VelocityShares Short LIBOR ETN (ticker: DLBR), a new offering launched in August of last year. The index provides long exposure to the first 8 quarterly Eurodollar futures contracts.
Eurodollars are unsecured time deposits held outside of the U.S. (including by foreign branches of U.S. banks) and denominated in U.S. dollars. Even though these deposits are held outside of the U.S., they are still a key funding source for U.S. banks since banks can borrow offshore in the Eurodollar market and then transfer funds onshore. In fact, U.S. banks generally consider funding through Fed Funds and Eurodollars as close substitutes. LIBOR is the implied interest rate at which banks in London offer Eurodollars.
Eurodollar futures contracts are quoted as 100 minus LIBOR. As an example, if LIBOR for a particular maturity is at 1.25%, the contract will be quoted at 98.75. As such, being long Eurodollar futures is equivalent to being short LIBOR (i.e. we profit when rates go down), just like standard bond positions.
The specific number of Eurodollar futures contracts held is calculated such that the index delivers a return equal to the weighted average percentage change in LIBOR across the maturities held. Note that we can think about this as if we have exposure to the 1-Year Spot LIBOR index.
As an example, if rates fall 25bps from 5.00% to 4.75%, the index will return approximately 5.00% (0.25% / 5.00%). When rates are lower (higher), the impact of the same 25bps move will be higher (lower). A 25bs decrease with rates starting at 10.00% will produce a gain of 2.50%. A 25bps decrease with rates starting at 2.50% will produce a gain of 10.00%.
This payoff structure creates issues when rates are low. To mitigate this, the index linearly reduces its exposure to rate moves below 2.50%; the exposure is 100% when the 1-year spot LIBOR index is at 2.50% and 0% when the 1-year spot LIBOR is at 0.00%.
The following table illustrates the gain/loss realized from a +/- 25bps move at various starting rate levels.
In the backtest, the index boasts a 0.71 correlation to IEF and 2.9 beta to IEF, suggesting potential usefulness as levered bond exposure.
One important thing to keep in mind with DLBR is that because of how its payoff is structured relative to LIBOR, its volatility profile could change materially depending on LIBOR’s level. Assuming that LIBOR volatility was constant over time, we would expect DLBR’s volatility to be relatively lower when LIBOR was below 2.5% (where the participation rate is reduced), highest when LIBOR is just above 2.5%, and then declining as LIBOR rises. In reality, rate volatility is not constant, but nevertheless the pattern still roughly holds when we approximate returns by applying the payoff formula to historical 1-year LIBOR.
A 60/40 blend of the indices tracked by XIV and DLBR delivers very strong risk-adjusted returns, albeit with nearly 3x more volatility than SPY (40.9% annualized volatility vs. 14.5% for SPY). As we did with the Levered ETF strategy, we blend the XIV/DLBR portfolio with the unlevered SPY/IEF 60/40 in a proportion that has the same volatility as SPY. Note that we do this with the benefit of hindsight. In reality, we would have to estimate future volatility. However, we use this approach for easier performance comparisons. The Derivative 60/40 delivers 2.6% more annualized return than the benchmark and a nearly 30% reduction in drawdown.
Of course, this type of reward doesn’t come without any additional risk. In this case, the main risk is summed up nicely from this line in XIV’s prospectus:
“The long term expected value of your ETNs is zero. If you hold your ETNs as a long-term investment, it is likely that you will lose all or a substantial portion of your investment.”
Why is this the case?
As mentioned earlier, XIV provides 1x short exposure to 1-month VIX futures, reset daily. Therefore, if 1-month VIX futures rise by 100% or more, XIV will go to $0, losing 100% of its value.
It’s important to point out here that VIX futures exposure is not the same thing as exposure to the VIX index. Historically, 1-month VIX futures have a beta to the VIX index of approximately 0.5. So, for XIV to lose all its value, we would need to see spot VIX increase by 200% or more.
Is this within the realm of possibility? Since XIV launched in 2010, the largest one-day loss (measured from the previous day’s close to the next day’s low) was 31.0%. This occurred on August 24, 2015 when VIX spiked 90.1% from 28.03 to 53.29. This VIX spike was the largest in the CBOE VIX dataset, which goes back to the early 1990s.
Fortunately, we can go back even further by looking at the VIX’s predecessor, the VXO index. VXO was launched in 1993 and has price history dating back to 1986.
There are two main differences between VIX and VXO:
- VIX is based on S&P 500 options and VXO is based on S&P 100 options.
- VIX uses a broader range of strike prices than VXO, which used only near-the-money strikes. In addition, VIX was designed so that it can be statically replicated as opposed to the VXO, which requires dynamic hedging to replicate.
Despite these differences, VIX and VXO track each other pretty closely.
Black Monday immediately jumps out from the above chart. On Friday, October 16, 1987 VXO closed at 36.37. On Monday, October 19, 1987, the index spiked to an intraday high of 152.48, an increase of 319%. A similar spike in VIX would almost certainly cause XIV to lose all of its value.
This gets to why XIV has a “long-term expected value of zero.” The longer you hold it, the more likely you are to experience a 1987 event and lose all of your money. Hence, XIV is not a great option as a standalone, buy and hold investment.
The more relevant question for us though is: “Does this invalidate the Derivatives ETN 60/40 as a viable way to introduce leverage into your portfolio over the long-run?”
We think the answer is no for two main reasons.
First, the proposed strategy does not use XIV as a standalone position, but rather as a part of a portfolio. The actual allocations for the hypothetical Derivatives 60/40 are 45.8% to SPY, 30.5% to IEF, 14.2% to XIV, and 9.5% to DLBR.
Based on the available data, the portfolio would have lost around 24% on Black Monday. While this is obviously painful, it’s not all that different than the projected 21% loss on the 1.7x levered SPY/IEF portfolio. A 1987-like event will always be extra painful for those using leverage, whether something like XIV is used or not.
Second, the proposed strategy rebalances on a monthly basis and so it doesn’t truly buy and hold XIV. Instead, any XIV gains are constantly harvested and reinvested in the other parts of the portfolio as monthly rebalances are executed. Similarly, when XIV underperforms the other positions in the portfolio, rebalances will trim other positions to invest in XIV.
But let’s say that we are really concerned about the impact of a 1987 type of outlier. What can we do? There are a few options (other than just foregoing short volatility as a means of achieving leveraged equity exposure altogether).
- We could replace XIV with a lower risk short volatility alternative.For example, we could short mid-term VIX futures instead of short-term VIX futures. ZIV (VelocityShares Daily Inverse Medium-Term ETN) is an ETN that offers exposure to the S&P 500 Mid-Term Futures Inverse Daily Index. This index takes short positions in fourth, fifth, sixth, and seventh month VIX futures contracts. The benefit from a risk perspective is that this index has a beta to the VIX of around -0.2 instead of the -0.5 for short-term VIX futures. As a result, we need to see a more extreme VIX move to be wiped out (~500% spike instead of a ~200% spike, assuming long-term betas hold).The main disadvantages of moving to mid-term futures are that (i) the risk-adjusted return of shorting mid-term futures has lagged that of shorting short-term futures and (ii) a larger allocation to the derivatives-based portfolio relative to the simple 60/40 SPY/IEF is needed to hit the volatility target. In fact, when we approximate a Black Monday stress test of the portfolio with ZIV replacing XIV, we actually end up with a larger loss (-28% vs -24%) – despite the fact the ZIV position does not get wiped out – because we have to hold more ZIV to get the same amount of leverage.Another replacement for XIV would be a strategy that pairs short positions in short-term VIX futures with some type of long volatility hedge. An example would be the VelocityShares VIX Short Volatility Hedged ETN (ticker: XIVH), which tracks the S&P 500 VIX Futures Short Volatility Hedged Index. This index pairs a 90% short position in VIX futures with a 10% 2x long position in VIX futures, rebalanced quarterly on a rolling basis. The offsetting positions may be confusing, but the idea is actually interesting. We discussed it in Hedge Hunting: Wrangling the VIX.At the 10% 2x long/90% 1x short base weights and still keeping our -0.5 beta estimate to VIX, we would need to see a 285% VIX spike to get wiped out instead of the 200% spike necessary with pure XIV. In practice, the weighting is rarely 10% 2x long and 90% 2x short since the rebalancing is not daily. In fact, we found that the long volatility weighting was less than 2.5% about 7.5% of the time going back to 2005. If we were unlucky enough to have a black swan hit at a time where we were this underweight the long volatility position, a ~215% VIX move would wipe us out.We still have the same side effect, albeit to a lesser degree, that we had with ZIV in that by using a lower volatility product, we require a larger allocation to get the same amount of leverage.
- We could utilize an active strategy to trade XIV. For example, approaches using momentum, the slope of the VIX futures curve, and comparisons between implied and realized volatility all show promise. Using some combination of these approaches may indirectly reduce the impact of a large volatility spike by reducing the probability that we have a full position in XIV. Yet, events like 1987 that largely come out of nowhere are almost certainly not going to be predicted any type of active trading approach via skill.
- We could combine the Derivatives ETN 60/40 with the Levered ETF 60/40 to reduce the size of the XIV position. We present the results of this combination below. The XIV position is reduced from 14.2% to 11.0% and the simulated 1987 return improves from -24% to -23%.
Leverage is a tool. When used prudently, it can help investors potentially achieve much more risk-efficient returns. When used without care, it can lead to complete ruin.
For many investors who do not have access to traditional means of leverage, they can mimic the effect using levered ETFs and derivative ETNs (or they can do so inefficiently with high beta ETFs).
Over the period from 2005 to 2017, levered ETFs showed comparable performance to a traditionally levered 60/40 portfolio while derivative ETNs exhibited strong outperformance.
Looking back even further in history, a 1987-like event will always be extra painful for those using leverage, regardless of the method use to achieve it. However, combining the levered ETF and derivative ETN approaches may improve the outcomes in these trying market environments.
By employing leverage in creative ways, we are not destroying the risks associated with using traditional forms of leverage; we are merely transforming them. Each method comes with its own set of risks, but these options provide investors with practical ways to convert high risk-adjusted returns into higher absolute returns without using margin directly.
 For the quarterly and annual reset versions, we stagger the resets over monthly increments to reduce timing luck.
 We actually use the S&P 500 High Beta Index with a 25bps annual fee applied as a proxy for the ETF as the ETF did not launch until 2011.
 The replicated daily returns consist of three components. The first component is simply twice the daily return of the reference index (the S&P 500 in the case SSO and the Bloomberg Barclays 7-10 Year Treasury Index in the case of UST (note: UST’s reference index is actually the ICE U.S. Treasury 7-10 Year Bond Index, but the two indices are very similar, and we were able to pull a longer history of the Bloomberg Barclays index). The second component is the financing cost for borrowing. We assumed annualized financing costs equal to the 3-month Treasury bill rate plus a spread. The spread was selected such that the total return of our replication index matched the total return of the actual ETF (SSO or UST) for their shared history. The third component is fees: 90bps for SSO and 95bps for UST.
 XIV is represented by the S&P 500 VIX Short-Term Futures Inverse Daily Index with fees (135bps) netted out.
 The first contract rolls off two days prior to expiry, so for a brief period of time the ninth listed quarterly contract is also included in the index.
 This is an imperfect analysis as actual returns will not just be influenced by interest rate levels, but also the shape of the Eurodollar futures curve.
 Exposure is actually to a rolling short position in the front two VIX future contracts.
 In fact, the ETN issuer has the right to accelerate the maturity of the ETNs in the event that XIV falls by 80% or more in a single day.
 This is a gross oversimplification of relative behavior between spot VIX and VIX futures.
 Again, this is a gross oversimplification. It is certainly possible that a spot VIX move smaller than 200% could lead to a VIX futures increase of more than 100%.