An Exploration of Leveraged ETFs

Regular readers of this blog know that there’s nothing I enjoy more than finding niche, unexplored ideas. I first started exploring the world of leveraged ETFs after reading the following paper by Hendrik Bessembinder analysing the underperformance of single stock leveraged ETFs against their underlying securities. While many people in finance are aware that these ETFs are flawed products, there’s a distinct lack of research and writing on the topic. I’ve been spending the past month or so understanding these products and coming up with potential ways to exploit them profitably.

Sources of Underperformance

I want to start off by distinguishing between the two separate reasons why these products underperform their underlying. The first, more commonly known reason is “volatility decay”. To explain volatility decay I want you to imagine a 2x etf Apple stock, noting that the ETF rebalances daily to expose you to 2x the daily return. Apple drops 20% in one day, before going up 25% the next day. If you owned $100 of Apple stock, after the two days your shares would still be worth $100. Meanwhile, if you owned $100 of the 2x Apple ETF, you would only own $90 of the 2x ETF after the 2 days. The amount of volatility decay is dictated by a combination of the leverage of the ETF, the magnitude of the volatility of the underlying, and the correlation of the volatility (autocorrelation). To summarise, volatility decay kicks in with highly levered daily tracking ETFs where the underlying is experiencing large daily moves, however not consistently in the same direction. Correlated daily movements actually result in the opposite, where the leverage actually amplifies the impact (eg. a consistent run up will result in outsized gains for the levered ETF). However, due to short term stock returns being mostly random, this effect is usually muted over a reasonable time frame.

The other key way that these ETFs underperform is due to tracking error. While their target is to provide leveraged daily exposure to the underlying security, this is generally done using complex financial tools such as swaps. There are many sources of tracking error here, some that are consistent like the interest cost and expense ratio, but some that are more interesting. The following is a portfolio following the daily difference between 200% exposure to AAPL and -100% exposure to AAPU going back to the ETF’s inception. We can see that the ETF is underperforming the daily performance of its underlying by 11% annually, which is much higher than can be accounted for by market interest rates and the 1% annual expense ratio.

My understanding is that this excess tracking error comes from the excess cost of the swaps used to create the 2x exposure. For ETFs such as TQQQ and YINN implied interest rates on these swaps are usually pretty reasonable at around 50 bps above market. However, the implied interest rates for these single stock ETFs are much higher. Looking at AAPX, the T-Rex 2x Apple ETF we can observe that it is underperforming its underlying daily exposure by 11% annually. It’s really important to distinguish this sort of tracking error underperformance from the volatility decay I mentioned earlier. Volatility decay, while being an attractive source of returns, has an element of risk that is necessary to produce good returns. You need enough volatility to produce the decay, but you also need that volatility not to be autocorrelated. There are environments (we will discuss soon) where betting on this will underperform or even blow up. Meanwhile, tracking error (if structured smartly) is essentially a free lunch. While how you bet on it may carry risk, tracking error is structurally built into these products and will consistently produce underperformance relative to the underlying.

Potential Strategies

There are three different ways of taking advantage of this underperformance that I’ve identified which have varied risk and return profiles but all involve shorting the levered ETFs. The first and most obvious strategy is outright shorting them. If we know these products are going to underperform, and we want to make a directional bet on the underlying, then it makes sense to bet on it using these ETF’s. As an example, if I wanted to short $100 of MSTR I could short that outright, or I could instead short $50 of MSTU for the same amount of exposure. Below shows the daily returns of a 50% MSTR short vs a 25% MSTU short rebalanced weekly (I have factored in the 20% borrow cost here as well). We can see a clear divergence between the two, with MSTU underperforming a significant amount.

I want to point out here that these directional shorts don’t necessarily have to be betting on the underlying to drop. For retail investors with restrictions on their ability to access margin and lever up, shorting leveraged inverse products (that bet on the underlying to go down) can be an efficient way to do that. In my case as an example, Australian regulations make it very difficult to access cheap margin for long short investing, with a maximum $50k margin loan available. I have been able to increase my leverage by shorting products like SPXS (3x inverse SPY), SQQQ (3x inverse QQQ) and GLL (2x inverse gold) to gain exposure to those products. As can be seen below, a monthly rebalanced short position of these sorts of products does a reasonable job at tracking an equivalent leveraged position (especially once factoring in the fact that margin rates for retail are usually a few % above the cash rate which is being used here). I want to make a note here that using leverage is risky at the best of times, and these 3x products are incredibly complicated and risky to be short. However, I think with careful risk management and good portfolio construction this can be a useful tool in a retail investors toolbox.

The next way one can take advantage of these products are through shorting both sides of a leveraged ETF. Having done a lot of backtesting, this sort of strategy doesn’t work with most of these products, however there are specific ones that it works well with. There is a sweet spot with an underlying security that is volatile enough to consistently cause underperformance through volatility decay, but not so volatile that there is a risk of blowup. For example, levered oil, gas and gold ETFs don’t work here as they are prone to large runs of momentum in either direction without seeing the reversals and choppiness in the share price we need for volatility decay to kick in. There are two ETF pairs that I have found add meaningful returns when utilising this strategy. The first is the 3x Chinese ETFs YINN and YANG (Y/Y), and the second is the 3x semiconductor ETFs SOXL and SOXS (S/S).

As we can see in the chart above, these two pairs have performed exceptionally well over the past decade, and have actually provided very effective hedges during large crashes like Covid and Liberation day. The strategies both perform reasonably well up to 2020, comfortably outperforming cash at 6% cagr each. However, the higher volatility, dip buying regime of post 2020 has seen the strategies soar, with Y/Y doing a 15% cagr at .2 beta, while S/S has done a whopping 30% cagr at a .4 beta. Personally, I believe our current volatile market structure is here to stay, but even in a return to market conditions of the 2010’s both strategies are still very attractive. These strategies can very easily be layered on top of a core equity strategy to provide a diversified source of returns. Now I want to note, these strategies do not come without risk, and I will discuss the risks and market conditions they underperform in at the end.

The final and in my opinion most interesting strategy here is going long the underlying security, while shorting the leveraged ETF. For this strategy you’re less taking advantage of the volatility decay, and more focused on the tracking error. Additionally, the risk profile of this strategy is significantly lower than the other two as the long acts as an effective hedge for the short. The index ETFs like Yinn and TQQQ don’t work well for this as their tracking error is quite low, however I’ve found that some of the individual stock ETFs provide incredible opportunities here. This trade can be broken down into two key variables, borrow cost and tracking error. As an example, while the QBTS levered ETF QBTX has a 27% annualised tracking error, its borrow cost is 25% which makes it unnattractive for this strategy. Meanwhile, for whatever reason the AVGO 2x ETF AVL has only a .6% borrow cost, so is much more attractive vs a 12.5% tracking error.

I want to note a few limitations. Firstly, transaction costs will eat away at returns. I’ve found that weekly rebalances seem to work the best with regard to limiting risk, but transaction and spread costs will cost around 1% annually, though you will also have a small amount of volatility decay working in your favour. Additionally, for people like myself with margin limitations, this strategy is less attractive as instead of going 200% AVL/-100% AVGO and pocketing the 12% spread (before transaction cost), I have to go -50% AVL/100% AVGO/50% cash, while also dealing with poor cash rates. However, that is still a 6% return before volatility decay (which will usually add a couple of percentage annually) with very low risk. Some securities that I’ve found this works well with are GMEU, MSTU, XXRP, BULU, BITX and ETHU.

Risks

The core risk that all of these strategies are exposed to is the inverse of the volatility decay that they seek to exploit. If returns are autocorrelated (repeated moves in one direction), the leveraged ETF will underperform its underlying. Imagine a stock that goes up 5% for 5 days and its 2x leveraged ETF. The underlying will be up 55%, while the 2x ETF will be up 61% as the daily leverage compounds. This works in both directions as well, underperforming with downside momentum too. Luckily, short term returns are mostly random and exhibit mean reversion, however stocks (especially some of the more volatile meme stocks) can go on aggressive short term runs that will cause these strategies to underperform. A good example of this is the recent move in IONQ, with 22% in a week resulting in a 47% gain for its 2x ETF.

Another key risk for these strategies is borrow cost. Frustratingly, it is very hard to find good data for borrow cost for these stocks, so I have run my strategies with estimates based on their recent costs. However, it should be noted that borrow costs often spike when stocks peak, or during aggressive moves. This means that borrow costs may force you to close a strategy right at the worst time.

The core finding of this exploration is that the structural flaws in leveraged ETFs—namely volatility decay and significant financing costs—create clear opportunities for investors (especially retail, who are able to apply these strategies in less liquid ETFs with minimal transaction costs). While volatility decay is dependent on market conditions (high volatility, low autocorrelation), the excess tracking error observed in many single-stock leveraged ETFs presents a compelling, more consistent source of alpha. Finally I want to reiterate to the reader to maintain caution, as these are complicated, illiquid and volatile products.