# Diversifying Return Stacked Allocations (Part 1) Source: https://www.yieldcurve.pro/blog/diversifying-return-stacked-allocations-001 Published: 2024-02-20 Tags: Bonds, Diversification, ETFs, Return Stacking, Trend Following _So You Have Chosen Your Return Stacked Asset Classes. Now What?_ # You Chose A Return Stacked ETF. Now What?
## Return...Stacking?
Return Stacking is a relatively hot topic as of late. The Return Stacked website states that... > At its core, Return Stacking is the idea of layering one investment return on > top of another, achieving more than \$1.00 of exposure for each \$1.00 invested. The Return Stacked folks indicate that some of the reasons one would want to own such a beast include: - Pursuing Diversification Without Sacrifice - Potential for Enhanced Returns - Potential to Improve Diversification They explain things in more detail than we provide here so be sure to peruse their website to get all the essentials. Many of us in quantitative finance first learned about this type of approach from the well known 1996 paper by Cliff Asness Why Not 100% Equities. In that paper, he discusses the virtues of taking a traditional 60/40 stock/bond portfolio and levering it up until it has the same volatility as a pure stocks. For the same units of risk, he showed that diversification improves the overall risk-adjusted returns above stocks alone. Recently, he revisited the issue with a follow-up sub-titled I Can’t Believe We Are Doing This One Again. One of the first investible implementations of this idea in ETF from that we know of was developed by WisdomTree. Here is a great write-up describing the 3 different flavors of their 90/60 ETFs. Also, in a collection of podcasts, we recall hearing Eric Crittenden (founder and Chief Investment Officer of Standpoint Funds) describe his frustration trying to convince clients to diversify their traditional stock portfolios by adding an allocation to trend following. If memory serves, the issue was client hesitance to reduce their equity exposure to make room for a trend following allocation. Simply getting them over their suspicion of trend following was a challenge as well. His solution was to offer a pre-packaged portfolio split between traditional equity beta and diversified trend following [^1]. In our opinion, the WisdomTree **HOW** combined with the Standpoint **WHY** set the stage for the Return Stacked ETFs shown in Table 1:
| ETF | Asset Class 1 | Asset Class 2 | Net Assets | | :--- | :-------------- | :-------------- | :--------- | | RSBT | US Bonds | Managed Futures | $50M | | RSSB | Global Equities | US Treasuries | $69M | | RSST | US Equities | Managed Futures | $67M |
#### **Table 1**: Return Stacked ETFs
In this vein, we would like to develop a simple framework to pose and answer the question suggested by the title of this blog post. Essentially, we would like a simple yet effective way (once we have chosen one of the Return Stacked ETFs in Table 1) to select asset classes that are additive. We will get to what we mean by **additive** in the next section. Since the ETFs in Table 1 are new with limited return history, we are going to specify proxy ETFs to model the return streams of the two sub-asset classes harvested by each Stack Returned fund. Table 2 shows how we will map them to liquid securities that are readily available to most retail investors.
| Asset Class | ETF Proxy | Description | Inception | | :-------------- | :-------- | :------------------------------------ | :-------- | | Global Equities | VT | Vanguard Total World Stock Index Fund | 2009 | | Managed Futures | WDTI | WisdomTree Managed Futures Strategy | 2011 | | US Bonds | AGG | iShares Core US Aggregate Bond | 2004 | | US Equities | SPY | SPDR S&P 500 | 1994 | | US Treasuries | IEF | iShares 10 Year Treasury | 2002 |
#### **Table 2**: Asset Class ETF Proxies
Are these perfect proxies for the underlyng asset classes in the Return Stacked ETFs? We are not certain but they are a reasonable place to start.
## Framework
The framework we will use [^2], [^3] is the multivariate regression given by the following equation: $$ r_t - r_f = \alpha + \beta_1\left(r_{1t} - r_f\right) + \beta_2\left(r_{2t} - r_f\right) + \epsilon_t $$ where $$r_t := \text{diversifying asset return}$$ $$r_{1t} := \text{asset class 1 return}$$ $$r_{2t} := \text{asset class 2 return}$$ $$r_f := \text{risk free rate}$$ $$\alpha_t := \text{intercept}$$ $$\beta_1 := \text{asset class 1 exposure}$$ $$\beta_2 := \text{asset class 2 exposure}$$
It is a simple yet useful model. For our purposes, we focus on the intercept and asset class exposures (the alphas and the betas). For now, we are going to concentrate on **RSST**. We will turn our attention to the other two ETFs in subsequent posts. Once an investor has chosen this particular Return Stacked ETF (and as a result, US Equities and Managed Futures) what other asset classes do we want to add to the mix to improve diversification and risk-adjusted returns. In order to measure how much improvement a given ETF adds to RSST we define a simple score whose purpose is to measure how much additional alpha is added per unit of beta. The naive score we have chosen takes the form
$$\text{Score} = \frac{\alpha}{|\beta_1| + |\beta_2|}$$
This ratio rewards assests with higher alpha and penalizes those with betas that deviate from zero. Tables 3 through Table 9 show an assortment of ETFs that we have cherry-picked to put through this meatgrinder framework. Each table shows the start and end dates that correspond to the cross-section of data available for each target ETF, SPY, and WDTI. We turn the computational crank and let the subsequent statistics fall out.
As shown in Table 3, GLD provides the best balance of alpha per unit of beta in this limited set of commodities ETFs.
| ETF | Description | Alpha | SPY Beta | WDTI Beta | Score | Start | End | |:------|:------------|--------:|-----------:|------------:|--------:|--------:|------:| | GLD | Gold | 1.77 | 0.00 | -0.06 | 30.71 | 2011 | 2023 | | DBC | DB Index | -1.07 | 0.37 | 2.36 | -0.39 | 2011 | 2023 | | USO | Oil | -5.01 | 0.71 | 6.65 | -0.68 | 2011 | 2023 | | GSG | GSCI | -2.95 | 0.45 | 3.15 | -0.82 | 2011 | 2023 |
#### **Table 3**: Commodity ETFs
DBC, USO, and GSG all exhibit high exposure to WDTI which should not be too surprising. They also do not provide positive alpha.
Table 4 shows that LQD (investment grade credit) as well AGG (the broader credit index) appear to be additive.
| ETF | Description | Alpha | SPY Beta | WDTI Beta | Score | Start | End | |:------|:------------------------------|--------:|-----------:|------------:|--------:|--------:|------:| | LQD | iBoxx $ Investment Grade | 2.58 | 0.04 | 0.56 | 4.28 | 2011 | 2023 | | AGG | Core US Aggregate Bond | 1.61 | -0.02 | 0.41 | 3.77 | 2011 | 2023 | | JNK | Bloomberg Barclays High Yield | 0.18 | 0.34 | 0.79 | 0.16 | 2011 | 2023 | | HYG | iBoxx $ High Yield | -0.12 | 0.36 | 0.57 | -0.13 | 2011 | 2023 |
#### **Table 4**: Credit ETFs
For this time period, high yield is not additive. The currency ETFs in Table 5 suggest that UUP (bullish dollar fund) is additive.
| ETF | Description | Alpha | SPY Beta | WDTI Beta | Score | Start | End | |:------|:---------------|--------:|-----------:|------------:|--------:|--------:|------:| | UUP | US Dollar Fund | 3.97 | -0.06 | 0.46 | 7.62 | 2011 | 2023 | | FXE | Euro Currency | -3.78 | 0.07 | 0.38 | -8.39 | 2011 | 2023 |
#### **Table 5**: FX ETFs
The alpha is modest but the exposures to SPY and WDTI are muted leading to a large score. Exposure to the Euro provides negative alpha.
| ETF | Description | Alpha | SPY Beta | WDTI Beta | Score | Start | End | |:------|:--------------------------|--------:|-----------:|------------:|--------:|--------:|------:| | VNQ | Real Estate Index | -1.13 | 0.77 | 0.08 | -1.33 | 2011 | 2023 | | IYR | US Real Estate | -1.86 | 0.77 | -0.25 | -1.82 | 2011 | 2023 | | RWX | International Real Estate | -7.05 | 0.74 | -0.31 | -6.69 | 2011 | 2023 |
#### **Table 6**: REIT ETFs
Table 7 says that rates are additive from 2011 to 2023 (note that SGOV's inception year is 2020) as the higher scoring ETFs benefit from low-ish exposures to SPY and WDTI.
| ETF | Description | Alpha | SPY Beta | WDTI Beta | Score | Start | End | |:------|:---------------------|--------:|-----------:|------------:|--------:|--------:|------:| | TLH | 10-20 Year Treasury | 3.68 | -0.17 | -0.00 | 20.63 | 2011 | 2023 | | IEI | 3-7 Year Treasury | 1.24 | -0.05 | 0.05 | 12.29 | 2011 | 2023 | | TLT | 20+ Year Treasury | 6.66 | -0.29 | -0.35 | 10.42 | 2011 | 2023 | | IEF | 7-10 Year Treasury | 2.56 | -0.11 | -0.14 | 9.96 | 2011 | 2023 | | SHY | 1-3 Year Treasury | 0.18 | -0.01 | 0.09 | 1.84 | 2011 | 2023 | | SHV | Short Treasury | 0.41 | -0.00 | 0.28 | 1.45 | 2011 | 2023 | | SGOV | 0-3 Month Treasury | 0.77 | -0.00 | 0.74 | 1.04 | 2020 | 2023 |
#### **Table 7**: Rates ETFs
Interestingly, rates exposure to WDTI decreases monotonically with tenor. ETFs tracking tenors above 10 years add value. As expected, Table 8 demonstrates that sector ETFs have near unit exposure to SPY.
| ETF | Description | Alpha | SPY Beta | WDTI Beta | Score | Start | End | |:------|:------------------------|--------:|-----------:|------------:|--------:|--------:|------:| | XLP | Consumer Staples | 2.50 | 0.62 | 0.21 | 3.02 | 2011 | 2023 | | XLK | Technology | 2.23 | 1.14 | -0.59 | 1.29 | 2011 | 2023 | | XLY | Consumer Discretionary | 2.64 | 1.05 | 1.06 | 1.24 | 2011 | 2023 | | XLV | Health Care | 1.76 | 0.85 | -0.75 | 1.09 | 2011 | 2023 | | XLU | Utilities | 1.16 | 0.51 | -1.88 | 0.49 | 2011 | 2023 | | XLI | Industrials | -0.35 | 1.03 | -0.58 | -0.22 | 2011 | 2023 | | XLF | Finance | -1.14 | 1.11 | -0.86 | -0.58 | 2011 | 2023 | | XLC | Communication | -4.78 | 1.06 | 3.04 | -1.17 | 2018 | 2023 | | XLB | Materials | -2.82 | 1.05 | -0.48 | -1.84 | 2011 | 2023 | | XLE | Energy | -2.77 | 1.07 | -0.22 | -2.14 | 2011 | 2023 | | XLRE | Real Estate | -2.20 | 0.73 | -0.19 | -2.41 | 2015 | 2023 |
#### **Table 8**: Sector ETFs
On an alpha basis some are mildly additive. The statistics computed for XLC and XLRE should be taken with a grain of salt considering the limited amount of data. Adding XLP (Consumer Staples) looks slightly interesting. Table 9 says that QQQ and IWF (the only equity ETFs that have positive Score) are similar to XLK from Table 8.
| ETF | Description | Alpha | SPY Beta | WDTI Beta | Score | Start | End | |:------|:---------------------|--------:|-----------:|------------:|--------:|--------:|------:| | QQQ | Nasdaq 100 | 2.27 | 1.14 | 0.12 | 1.80 | 2011 | 2023 | | IWF | Russell 1000 Growth | 0.67 | 1.07 | 0.21 | 0.52 | 2011 | 2023 | | IWB | Russell 1000 | -0.25 | 1.00 | -0.08 | -0.23 | 2011 | 2023 | | IWV | Russell 3000 | -0.59 | 1.02 | -0.12 | -0.52 | 2011 | 2023 | | IWD | Russell 1000 Value | -1.35 | 0.94 | -0.37 | -1.03 | 2011 | 2023 | | IWM | Russell 2000 | -3.30 | 1.14 | -0.02 | -2.84 | 2011 | 2023 | | VT | Vanguard Total World | -3.72 | 0.98 | -0.05 | -3.62 | 2011 | 2023 | | SCZ | MSCI EAFE Small Cap | -4.61 | 0.87 | -0.11 | -4.66 | 2011 | 2023 | | EEM | MSCI Emerging | -12.08 | 1.02 | -1.06 | -5.79 | 2011 | 2023 | | EFA | MSCI EAFE | -6.10 | 0.92 | -0.13 | -5.82 | 2011 | 2023 |
#### **Table 9**: Stock ETFs
## Summary
This simplistic analysis suggests that the ETFs in Table 10 are the most additive to RSST:
| ETF | Description | Alpha | SPY Beta | WDTI Beta | Score | Start | End | |:------|:-------------------------|-------:|-----------:|------------:|--------:|--------:|------:| | GLD | Gold | 1.77 | 0.00 | -0.06 | 30.71 | 2011 | 2023 | | TLH | 10-20 Year Treasury | 3.68 | -0.17 | -0.00 | 20.63 | 2011 | 2023 | | UUP | US Dollar Fund | 3.97 | -0.06 | 0.46 | 7.62 | 2011 | 2023 | | LQD | iBoxx $ Investment Grade | 2.58 | 0.04 | 0.56 | 4.28 | 2011 | 2023 | | XLP | Consumer Staples | 2.50 | 0.62 | 0.21 | 3.02 | 2011 | 2023 |
#### **Table 10**: Additive ETFs
An additive mixture of ETFs could include securities with long exposure to gold, long-dated rates, the dollar, investment grade credit, and consumer staples. Adding gold to a portfolio with rates exposure only makes sense if the rates are long-dated [^4], [^5], [^6]. By long-dated we mean anything with tenors 10+ years (TLH and TLT). As a longer term investment, gold is not additive to portolios holding tenors 1 to 10 years (SHY, IEI, and IEF). Table 10 says that adding both GLD and TLH to RSST could have positive interaction effects. Investors concerned with duration risk should think twice about adding TLH as holding long-dated rates during a period of increasing rates would be (and has been) painful. The empirical observation that holding a security that is bullish the dollar could be transitory. We need to think more deeply about why this not simply an empirical artifact. Like TLH, investors concerned with duration risk should be certain they want to own LQD. It is additive but it also has a duration approaching 10 years. One could make an argument to add XLK (Technology). However, this ETF has unit (or higher) exposure to SPY which should disqualify it. Keep in mind that we are only looking at a limited set of ETFs and for a limited amount of data (2011-2023). We will consider the other two Return Stacked ETFs (RSBT and RSSB) in subsequent posts. [^1]: If we have gotten any of these details wrong someone please send us an email. [^2]: Comovement and Predictability Relationships Between Bonds and the Cross-section of Stocks [^3]: Interest Rate Sensitivity in the Cross-Section of Stock Returns [^4]: Is Gold a Diversifier (Part 1) [^5]: Is Gold a Diversifier (Part 2) [^6]: Is Gold a Diversifier (Part 3)