Some people (including me) have argued that altruists often benefit from leveraging their investments. Recently, it has become easier to use leverage thanks to the emergence of return stacked funds.

This is not financial advice.

Contents

What is return stacking?

Return stacking is a way of getting up to leveraged exposure to multiple return streams simultaneously. For example, RSSB invests 100% into global equities and 100% into US Treasury bonds, effectively giving it 2:1 leverage on a diversified stock/bond portfolio.

A return stacked ETF is a type of leveraged ETF. But whereas traditional leveraged ETFs (such as SPXL) lever up a single index like the S&P 500, a return stacked fund holds multiple asset classes.

Return stacked ETFs have lower management fees than single-index leveraged ETFs, and (with low confidence) they appear to have lower overhead costs for reasons that are not entirely clear to me (my guess is a combination of cheaper borrowing costs + transaction costs).

An overview of return stacking funds

There are four brands of return stacking funds that I know of:

  1. the eponymous Return Stacked ETFs (RSSB, RSST, RSBT, RSSY, RSBY, RSBA, BTGD, RDMIX)
  2. WisdomTree Capital Efficient ETFs (NTSX, NTSI, NTSE, GDE, GDMN)
  3. PIMCO StocksPLUS Long Duration Fund (PSLDX)
  4. Evoke Advisors Ultra Risk Parity ETF (UPAR)

The PIMCO fund has been around since 2007, but the others only launched within the last few years.

What do each of these funds invest in?

Six of the funds stack traditional asset classes:

Fund first asset class second asset class leverage
RSSB 100% global stocks 100% US Treasury bonds 2:1
NTSX 90% US stocks 60% US Treasury bonds 1.5:1
NTSI 90% international stocks 60% US Treasury bonds 1.5:1
NTSE 90% emerging market stocks 60% US Treasury bonds 1.5:1
UPAR too many for this table*   1.68:1
PSLDX ~100% US stocks** ~100% bonds** ~2:1**

*UPAR targets 17.5% U.S. equities, 7% international equities, 10.5% emerging markets equities, 21% commodity producer equities, 14% gold, 49% TIPS, and 49% Treasuries for a total allocation of 168%.

**PSLDX percentages are only approximate because the fund is actively managed and its holdings vary over time.

These funds stack traditional asset classes with alternatives:

Fund first asset class second asset class leverage
RSST 100% US stocks 100% managed futures* 2:1*
RSBT 100% US bonds1 100% managed futures* 2:1*
RSSY 100% US stocks 100% futures yield* 2:1*
RSBY 100% US bonds 100% futures yield* 2:1*
RSBA 100% US Treasury bonds 100% merger arbitrage* 2:1*
RDMIX 50/50 US stocks/bonds 100% systematic macro* 2:1*
BTGD 100% bitcoin 100% gold 2:1
GDE 90% US stocks 90% gold 1.8:1
GDMN 90% gold miner stocks 90% gold 1.8:1

*Managed futures (a.k.a. trendfollowing), futures yield (a.k.a. carry or roll yield), merger arbitrage, and systematic macro are all long/short strategies, not simple assets that you can buy and hold. So it’s somewhat arbitrary to say that the funds invest 100% into those strategies.

The true cost of return stacked ETFs

In a previous post, I looked at how a leveraged index fund should perform and compared that against how leveraged ETFs actually did perform. I found that the ETFs consistently cost more than expected, by an average of about one percentage point.

I attempted to do the same analysis for return stacked ETFs. These ETFs are harder to replicate because they don’t track indexes, so I don’t have high confidence in the results. That said, my numbers suggest that return stacked ETFs are more cost-effective than conventional leveraged ETFs.

I was able to replicate RSSB and NTSX:

ETF Leverage Stock ETF(s) Bond Fund(s)
RSSB 100% + 100% VTI + VXUS2 bond futures ladder3
NTSX 90% + 60% SPY (S&P 500) bond futures ladder4

I also attempted to replicate NTSI, PSLDX, and GDE, but I couldn’t find benchmarks that tracked them well enough.

I calculated excess costs of each fund as the hypothetical return you’d get if you levered up the benchmark (borrowing at the 3-month T-bill rate), minus the actual historical return of the fund.

The following table shows the total excess cost and after-fee cost for the return stacked ETFs. Excess cost is shown per 100% leverage (the excess on NTSX is multiplied by two because it only has 50% leverage). After-fee cost gives the excess cost minus the difference expense ratios between the ETF and the benchmark—this represents the “unexpected” portion of the cost, since you expect to pay the expense ratio no matter what. r gives the correlation between the return stacked ETF and the benchmark. I calculated the average annual cost for each ETF starting from the earliest year for which the ETF had a full year of data.

ETF Excess Cost After Fee r Start Year
RSSB -0.55% -0.84% 0.998 2024
NTSX -0.17% -0.41% 0.997 2019

(The costs were negative, which means the real-life funds outperformed the benchmarks.)

Excess costs for each individual year for NTSX:

  2019 2020 2021 2022 2023 2024
NTSX -0.48 -3.50 2.72 1.92 -0.93 -2.15

As we can see, excess costs varied quite a bit from year to year. However, they were still generally lower than the costs of conventional leveraged ETFs.

In fact, the excess costs were negative most years. That’s surprising, since the benchmark does not account for transaction costs.

Why were return stacked funds (apparently) more cost-effective than conventional leveraged ETFs?

  1. These funds have lower expense ratios. For example, RSSB charges 0.36% and SSO (a 2x leveraged S&P 500 fund) charges 0.89%.
  2. Traditional leveraged ETFs rebalance daily. The Return Stacked and WisdomTree ETFs only rebalance if the holdings drift 5 percentage points away from the target weights. Rebalancing has transaction costs, which could be significant or could be close to zero, depending on various factors.
  3. The return stacked funds get leverage via Treasury futures, which is approximately the cheapest way to get leverage. Conventional leveraged ETFs primarily use swaps, which have an opaque pricing structure and might cost a lot more. (I have no idea how much they actually cost because the pricing is opaque.)

Those factors explain why return stacked ETFs are cheaper than 3x leveraged index ETFs. But how is it possible for a return stacked ETF to outperform a leveraged combination of index funds?

My benchmarks have some margin of error—they do not perfectly track the return stacked ETFs. Based on playing around with the implementation details of the benchmark, I believe it could be off by perhaps one percentage point.5

The most likely source of tracking error is rebalance timing. Small changes in when you rebalance can significantly change year-to-year performance, especially in years like 2024 where some asset classes perform much better than others. If stocks outpaced bonds for most of the year, and the fund was supposed to rebalance from stocks to bonds, then the real-life fund might have gained an edge over the benchmark by delaying rebalancing a little longer.

Even if these (apparently) negative costs might not persist, this still provides evidence that the return stacked ETFs have lower costs than single-asset leveraged ETFs.

Pros and cons of return stacked funds

Pros:

  • They’re a convenient way to get leverage, much more convenient than options or futures.
  • They appear to have lower all-in costs than conventional leveraged ETFs.

Cons:

  • None of them offer greater than a 100% allocation to equities.
  • Limited choices—there are only a handful of return stacked ETFs available, and they might not include the asset classes you want.
    • I personally would like to see a global stocks + managed futures ETF, but that doesn’t exist. There’s only US stocks + managed futures (RSST) and bonds + managed futures (RSBT).
  • As with other leveraged ETFs, the costs of return stacked ETFs fluctuate from year to year. Even though the costs are low on average, in any given year a return stacked ETF might perform worse than expected.
  • I could only determine the costs for two of the return stacked ETFs. The others might have higher costs.

Are bonds a good investment?

Most of the return stacked funds hold bonds. A question that some people ask:

Does it make sense to own return stacked stocks + bonds? Wouldn’t I rather have pure leveraged stocks instead?

Good question! I don’t know!

An argument against buying bonds:

Right now, the yield curve is nearly flat: yields on long-term bonds are only slightly higher than on short-term bonds. Why would you borrow at the short-term rate to earn the long-term rate if those rates are (nearly) the same?

Two counter-arguments:

  1. The efficient market hypothesis predicts that you can’t time the bond market, so you shouldn’t change how you invest based on what the yield curve looks like.
  2. A flat or inverted yield curve suggests that short-term rates will go down in the future. You might want to “lock in” the current rate by buying long-term bonds.

(Really these counter-arguments are the same—the (presumed) reason why the yield curve is flat is because the market is pricing in future changes in bond yields.)

Another argument against bonds:

In the long run, bonds have only earned a little bit of a premium over short-term T-bills. Given the overhead costs of using leverage, leveraged bonds might have near-zero or even negative expected return.

And two counter-arguments:

  1. If you can borrow at close to the risk-free rate, bonds should still have a positive long-run premium.
  2. Even if leveraged bonds have ~zero expected return, they still add value to a portfolio if they perform well during equity downturns.

Which side of the argument is correct is left as an exercise to the reader.

If you don’t want to hold bonds, there are some bondless return stacked funds available:

  1. RSST holds stocks + managed futures (a.k.a. trendfollowing).
  2. RSSY holds stocks + futures yield (a.k.a. carry or roll yield).
  3. BTGD holds bitcoin + gold.
  4. GDE holds US stocks + gold.
  5. GDMN holds gold miner stocks + gold.

I am a big fan of managed futures trendfollowing—it’s a strategy with strong historical performance that provided protection during market downturns, and I think it’s likely to continue working in the future (for more, see Hurst et al. (2017), “A Century of Evidence on Trend-Following Investing”6). I’m ambivalent about carry (I’ve heard good arguments both for and against using it). I personally wouldn’t invest in bitcoin or gold, but if that’s your thing, return stacked ETFs give you a way to do it.

(I don’t own RSST, but I hold something similar in my own portfolio—equities (AAVM) with managed futures stacked on top.)

Source code

Source code is available on GitHub.

Acknowledgments

Thanks to Corey Hoffstein for helping me work out the implementation details of my benchmark.7

Notes

  1. “bonds” means any sort of bonds, including Treasury or corporate bonds. “Treasury bonds” means just Treasury bonds. ↩

  2. I weighted VTI at 62% and VXUS at 38% as of the beginning of 2024 because those are the weightings I get if I reverse-engineer from RSSB’s current weightings.

    It would be simpler to use VT which includes all the same stocks as VTI + VXUS. But RSSB itself holds VTI + VXUS, and I found that breaking out equities into two separate ETFs produces slightly more accurate benchmark. ↩

  3. RSSB gets exposure to bonds via an equal-weighted combination of bond futures at four maturities: 2-year, 5-year, 10-year, and long (i.e. 25- to 30-year). I replicated this using:

    • 25% S&P 2-Year U.S. Treasury Note Futures Total Return Index
    • 25% S&P 5-Year U.S. Treasury Note Futures Total Return Index
    • 25% S&P 10-Year U.S. Treasury Note Futures Total Return Index
    • 25% S&P Ultra T-Bond Futures Total Return Index

    These indexes should exactly match the bond futures that RSSB holds. ↩

  4. I wasn’t entirely sure what position to use to replicate NTSX’s bond holdings. Its materials include illustrative figures that use a 7-10 year Treasury index as a benchmark, which suggests I should use IEF or perhaps 10-year Treasury futures. But the latest holdings show that it uses a combination of bond futures of different durations.

    I found the best correlation to NTSX when using a weighted combination of four Treasury futures:

    • 12% S&P 2-Year U.S. Treasury Note Futures Total Return Index
    • 12% S&P 5-Year U.S. Treasury Note Futures Total Return Index
    • 24% S&P 10-Year U.S. Treasury Note Futures Total Return Index
    • 12% S&P Ultra T-Bond Futures Total Return Index

    As of this writing, NTSX holds 12% in 10-Year U.S. Treasury Note Futures and 12% in Ultra 10-Year U.S. Treasury Note Futures (which are like the normal 10-year futures except that they are more closely tied to a 10-year maturity). I did not use Ultra futures in my benchmark because they only launched a few years ago. ↩

  5. Some minor changes that affect the return of the benchmark:

    • The funds rebalance whenever weights drift 5% away from the target. But the prospectuses for RSSB and NTSX were not clear about what exactly that meant—I can think of at least four different interpretations. Corey Hoffstein (who co-runs RSSB) explained to me exactly how the rebalancing works, and I assumed NTSX works the same way but I don’t know for sure. Different rebalancing methods can change the average return by as much as one percentage point—a fund might get lucky and rebalance into a position right before it rockets up, or the opposite might happen.
    • Changing how the benchmark invests in bonds can change the return. GOVT outperformed the Treasury futures ladder over the sample period. Changing the NTSX benchmark to use GOVT increased its return by 24 bps (but also decreased the correlation to NTSX from 0.997 to 0.993).
    • My program might have a bug. Shortly before posting this article, I discovered that I was incorrectly calculating how much cash the benchmark needed to borrow and thus overestimating interest payments by about 40 bps per year.

    ↩

  6. Hurst, B., Ooi, Y. H., & Pedersen, L. H. (2017). A Century of Evidence on Trend-Following Investing. ↩

  7. Originally I couldn’t figure out how to get my benchmark’s correlation to RSSB higher than 0.976. Turns out I needed to compare the benchmark to RSSB’s NAV, not its daily closing price. Corey explained to me that NAV and price diverge because daily futures prices settle at 3pm but exchanges close at 4pm, so any market movements in that hour will show up in RSSB’s price but not in its NAV or in the benchmark.

    I had the same problem with my NTSX benchmark and was able to fix it the same way.

    I also originally implemented rebalancing using an incorrect method, and Corey clarified the correct method to use. (This did not improve the correlation.) ↩