Tuesday, September 15, 2015

Reader Question: Leveraged ETFs

A reader, J.H. asks the following thoughtful question about leveraged ETFs (edited for length):
I am familiar with the decay factor of leveraged ETFs over the long term. However, it seems that using 50% cash, 50% 2X ETF, rebalanced say annually, mirrors the underlying 1X ETF very closely. In fact it is a bit better on a risk adjusted basis.

Blue portfolio: 100% SPY (an S&P 500 index ETF)
Red portfolio: 50% SHV (short-term U.S. bonds), 50% SSO (a 2X leveraged S&P 500 ETF)

These two portfolios gave nearly identical returns from 2008 to the present. I found this to be contradictory to everything I read about leveraged ETFs. A six year back-test should be enough to unveil the presumed decay, but I don't see it.

I am particularly interested in this way of investing as it provides a way to beat SPY without taking all the risk of being all-in. If on the cash component one can earn more than what SSO pays to borrow to buy stock (these days, one can make 2.3% or so using a GIC ladder), then I would say this approach can provide an easy 0.5 - 1% premium over a traditional all-equity portfolio. In essence, the 50% cash deleverages the leverage and can add alpha.

But I do know there is no free lunch in the investment world so I must be missing something. It cannot be that easy. A backtest using synthesized 2X returns that go back to as far as the S&P500 would be an interesting test to validate the theory.
Here is a chart of the comparison from J.H.’s backtest at Portfolio Visualizer:

As we can see, the two portfolios end up at the same place, but with somewhat different paths. Unfortunately, this is a side effect of the powerful bull market we’ve seen for 6 years and not a sign that J.H. has found a way to make an extra 1% on his money for free. I can explain this without resorting to math.

A 2X-leveraged ETF takes its cash, borrows an equal amount and uses it all to buy an index (the S&P 500 in this case). The leveraged ETF may actually use derivatives rather than borrow, but that isn’t important here.

Every day, the leveraged ETF rebalances so that it remains 200% exposed to the S&P 500. This means that if the S&P 500 goes up, the leveraged ETF borrows more and buys more stock. If the S&P 500 goes down, the leveraged ETF sells some stock and reduces its debt. On a day-to-day basis, this is essentially a momentum strategy. Unfortunately, the markets tend to revert to the mean; leveraged ETFs buy high and sell low. I gave a more detailed explanation of this effect in a past post.

If leveraged ETFs tend to leak, then how do we explain J.H.’s backtest in the chart above? The answer is the 6-year bull run in stocks. During a strong year for stocks, the leveraged ETF (SSO) maintains 200% exposure to stocks, but SSO grows to be more than 50% of the red portfolio. So, the red portfolio’s exposure to stocks grows above 100% from beginning to end of the year.

On average during a good year, the red portfolio is more than 100% exposed to S&P 500 stocks. So, the reason the red portfolio has kept up since 2008 is that this has been a great period to be more than 100% exposed to stocks. This extra exposure exactly offsets the inherent leakiness of the leveraged ETF.

To test this explanation, we can check to see what happens if we eliminate the annual portfolio rebalancing. We should expect the red portfolio to perform much better because its exposure to stocks will keep climbing above 100% instead of getting reset to 100% each year. Here is the chart without rebalancing:

As we can see, the red portfolio won easily. But this only happened because we are focused on a great period of time for stocks. Averaged over all conditions, the 50/50 portfolio with a leveraged ETF will not keep up with the plain vanilla index portfolio. So, there’s no free lunch.


  1. Michael, thanks so much for taking the time to write about this topic.

    I am not too sure about the bull case. Take Brazil in the last 5 years for instance where the country ETF EWZ lost around 50% of its value, yet look at 50% 2X Brazil backtest.

    Not too bad, eh?

    Re-balancing is a crucial part of the strategy in order to mirror the underlying 1X ETF without letting the drift buildup and hence the gap widen.

    I think a non trending sideways market is where the strategy really fails. Take the MSCI emerging market based ETF in the last five years where it practically went nowhere: 50% in 2X Emerging market backtest.

    Still, the gap is not all that big at all considering all the choppiness.

    Where I think the strategy can get interesting, and I was hoping you would touch on that with your math background, is:

    1) re-balance more effectively instead of a fixed interval like quarterly or yearly. For example, using bands, the 50/50 cash/2X ratio can sometimes drift to 45/55 (55 in a 2X ETF equals 10% leverage), or perhaps it should be allowed to drift further (40/60?, meaning 20% leverage), or 60/40 (i.e. under-invested, being 80% in).

    2) But the cash/fixed income portion is where this kind of strategy gets really sexy and interesting. One can split it so that there is just enough liquidity (perhaps 20%) to service any buys in the short term (if the 2X ETF drifts down by a lot), but can keep the rest in longer duration high grade bonds that earn more. In today's low interest rate env, 3-4% is achievable using a combination of GIC and high grader ladders.

    Obviously, this type of strategy would work in a tax sheltered account much better.

    1. @Anonymous: In the case where stocks drop sharply, the red portfolio will tend to become less than 100% exposed to stocks. This gives a boost that, in Brazil's case, is almost exactly offset by the leakiness of leveraged ETFs.

      I'm afraid I'm not optimistic about this strategy. If you examine enough cases and vary the rebalancing interval enough, you will find examples where it looks good, but this will just be data mining.

    2. I am not comfortable with it either. It is an experiment and it is all theoretical. I was just intrigued by some of the backtests and was expecting different results. I think the ultimate test would be one that goes back in time long enough, but such leveraged products only go back a few years so there are not enough historical data to test with. I suppose one could synthesize the returns from the underlying 1X, taking into account the interest at the time that such products would have to pay to pay additional stock. But even then, I am not sure how accurate of a test it would be. Probably possible and might make for a good project for me in the future.

      Thanks again,

  2. I prefer to use tqqq as my 3x etf. Pair it up with cash not bonds. 50/50. Target would be to beat the sp500 by 1.5x. Cash would be used to buy on the "dips" or even better on the crashes. tqqq underlying etf qqq has a long term upward trend. Expense of tqqq same as sso but much much more bang for the buck. Still it will take courage to buy more when your stock side has been decimated and fear is rampant.

    1. @Anonymous: My guess is that this strategy would look good until shortly before blowing up.