This post is available as a PDF here.

Summary­­

  • The term “multi-asset” appears in many investment strategies and applies to both balanced funds and target date retirement funds.
  • However, multi-asset strategies may be concentrated in a limited set of asset classes, and the performance of these asset classes may be driven by an even more limited set of risk factors.
  • By looking through the lenses of performance, asset classes, risk factors, and principal components, we can get an increasingly granular view of multi-asset strategies.
  • We believe that investors often overestimate how diversified multi-asset strategies are, which can expose their portfolios to unanticipated volatility and drawdowns.

The term “multi-asset” is thrown around frequently in the industry without a uniform definition applied across the board. According to Investopedia, it is “a combination of asset classes (such as cash, equity or bonds) used as an investment … The weights and types of classes vary according to the individual investor.”

So in a sense, the term is a catch-all for strategies that invest in more than one asset class, regardless of how they do so or what they actually invest in. In general, one would expect these funds to reap the benefits of diversification, with reduced volatility and drawdowns relative to many single asset class funds. Unfortunately, as we will see, that isn’t always the case.

Balanced funds and target date funds are two examples of multi-asset funds. These generally hold a mix of equities and bonds tailored to a specific risk profile the possibly varies in a predetermined way.

However, categorizing “multi-asset” funds that do not specifically fall into these categories is tough. For example, there are 54 funds in the Morningstar database that contain “multi asset” in their name.  These funds fall into 11 categories ranging from equity-based to bond to alternative.

  1. Allocation--15% to 30% Equity
  2. Allocation--30% to 50% Equity
  3. Allocation--50% to 70% Equity
  4. Allocation--70% to 85% Equity
  5. Diversified Emerging Markets
  6. Large Growth
  7. Managed Futures
  8. Multialternative
  9. Multisector Bond
  10. Tactical Allocation
  11. World Allocation

The categories, paired with the fact that many of the funds either have “income” or “growth” in their names, can provide some guidance on what to expect with these investments, but how do you know if your multi-asset strategy is actually multi-asset in the sense that you expect (i.e. diversified)?

To answer this question, we will look at multi-asset funds through a variety of lenses, in order of increasing analytical depth: performance, asset class, risk exposure, and principal components. While each has its merits and drawbacks, they can all be tools to better determine how to utilize multi-asset strategies within a portfolio.

 

Multi-Asset Through a Performance Lens

Perhaps the simplest way to start is the way most people do: let’s looks at performance. For this analysis, we will narrow down our multi-asset fund universe to those with at least five years of data (28 funds of the original 54).

“Multi-Asset” Fund 5-Year Return vs. Risk (Volatility)

Source: Morningstar, CSI Analytics. Analysis by Newfound. Data from 4/30/2012 to 4/30/2017.

We see zero correlation between risk and return for these funds, which indicates that they are all extremely different in terms of either investment universe or process. It is possible that there is some semblance of an efficient frontier emerging in the chart, but without more information on the individual funds, we can’t be sure.

What we do know is that it was possible to choose volatile funds with high or low returns or non-volatile funds with high or low returns. The desired diversification within the multi-asset universe is not clear across the board.

There is a bit more clarity if we take maximum drawdown as our risk measure but only insofar as very large drawdowns are detrimental to returns, a fact which we have shown many times before.  There is a slight inverse relationship between return and drawdown in the entire set of funds. However, funds with relatively high drawdowns still showed strong returns.

“Multi-Asset” Fund 5-Year Return vs. Risk (Drawdown)

Source: Morningstar, CSI Analytics. Analysis by Newfound. Data from 4/30/2012 to 4/30/2017.

As a point of reference, SPY, ACWI, and AGG are shown on these charts. The fact that both equity benchmarks outperformed nearly all multi-asset funds over the period is a testament to the state of diversification in the bull market in the past 5 years.

Pure performance is often not very informative without the appropriate context.  In order to better judge the diversified nature of a multi-asset strategy, we need to dig deeper.

 

Multi-Asset Through an Asset Class Lens

We can start by looking at what each fund actually invests in. For this to have meaning, we have to balance granularity with overlap, so we will concentrate on the broad asset classes of cash, U.S. and non-U.S. equity and bonds, and other, which captures everything else.

“Multi-Asset” Fund Allocation Statistics

 CashU.S. EquityNon-U.S. EquityU.S. BondNon-U.S. BondOther
Average10%22%18%29%14%6%
Median7%22%13%28%10%5%
Standard Deviation12%18%17%25%12%6%
Minimum0%-10%0%-23%-8%-2%
Maximum51%79%71%93%51%22%

Source: Morningstar. Data as of 4/30/2017.

Even though the average and median allocations are balanced, the individual funds are far from it, with 16 of the 54 funds totally excluding or even shorting certain asset classes.

Sixteen of the funds also have larger than 50% allocations to a single asset class. Some of the funds are essentially equity or bond funds with a little of the other thrown in. However, as a tactical manager, we can forgive temporary shifts among asset classes, even in a multi-asset fund. We’ll assess this in a bit.

We are also not put off by seeing large cash allocations. But given the generally positive market environment, this deserves more scrutiny.

The solution to this riddle is futures. Some “multi-asset” funds implement their strategies using derivatives and hold cash equivalents for margin requirements.

An issue with an asset class view is that many assets are driven by similar forces in the market. For example, even though the returns over the period on U.S. and non-U.S. equities were substantially different, SPY and EFA exhibited a correlation of 0.8.

Simply enforcing more stringent asset classifications is also not a remedy. Many satellite fixed income asset classes, such as emerging market bonds, bank loans, and convertible bonds can exhibit bond-like qualities on the upside and equity-like qualities on the downside.

If we want to get a better sense for whether a multi-asset fund is truly diversified, it is instructive to look at the exposure to the underlying risk factors that drive asset returns. This lens will also allow us to compare strategies that invest in futures with those that do not.

 

Multi-Asset Through a Risk Exposure Lens

There are many ways to decompose asset class returns based on risk factors, and any analysis will always be model dependent. However, using some intuition, we can get to a model that makes sense economically without overfitting the data.

We will focus on the following risk factors:

  1. Equity – both U.S. and global ex-U.S.
  2. Interest Rates – both level and slope changes in the U.S. Treasury yield curve
  3. Credit Spreads
  4. Currency – changes in the U.S. Dollar index
  5. Residual – everything not captured by these other factors

The summary statistics are shown below.

“Multi-Asset” Fund Risk Factor Exposure Statistics

 EquityRatesCreditCurrencyResidual
Average42%9%10%8%30%
Median44%0%6%4%26%
Standard Deviation32%15%14%14%23%
Minimum0%0%0%0%8%
Maximum91%60%54%55%100%
Number w/ No Exposure61611110

Source: Morningstar, MSCI, St. Louis FED, CSI Analytics. Analysis by Newfound. Data from 4/30/2012 to 4/30/2017.

What we see is a high average concentration in equity risk with a large dispersion around the mean. Six of the 28 funds have no significant exposure to equities. Far more funds have no significant exposure to interest rates, credit, and currency.[1]

If the goal is to have a balanced exposure to the risk factors, it would be nice to have a concrete way to assess the concentration for each fund.

One way to do this is to calculate the Gini score of the allocations. The Gini score is a measure of dispersion and is commonly used when analyzing the wealth distribution of a population. It ranges from 0 to 1, with 0 representing perfect equality and 1 representing perfect inequality.

In our case, the closer the Gini score of a portfolio is to 0, the closer the portfolio is to equally weighted. Some examples with five allocations (since we have five risk factors) are shown below.

Example Gini Scores

Note that many different allocations can have the same Gini score. Equally weighting 3 of the 5 exposures has a Gini score of 0.5, as does weighting them 60/10/10/10/10 and 40/30/20/10/0.

The Gini scores of the 28 funds fall in a wide range, from 0.35 to 1.0. The average coefficient is high (0.71), indicating concentration in primarily one or two risk factors.

“Multi-Asset” Fund Risk Factor Exposure Gini Score

Source: Morningstar, MSCI, St. Louis FED, CSI Analytics. Analysis by Newfound. Data from 4/30/2012 to 4/30/2017.

“Multi-Asset” Fund Risk Factor Exposure Gini Scores Summary Statistics

 Gini Score
Average0.71
Median0.73
Standard Deviation0.18
Minimum0.35
Maximum1.00

Source: Morningstar, MSCI, St. Louis FED, CSI Analytics. Analysis by Newfound. Data from 4/30/2012 to 4/30/2017.

One issue with this risk factor view is that the factors, like asset classes, are correlated.

Risk Factor Correlations


Source: MSCI, St. Louis FED, CSI Analytics. Analysis by Newfound. Data from 4/30/2012 to 4/30/2017.

The U.S. equity, credit, and currency factors had a decently large magnitude pairwise correlation[2]. A fund that bet heavily on equities may not have been much different from one with equity positions paired with some high yield.

To construct independent factors, we can use a statistical technique called Principal Component Analysis (PCA).[3]

 

Multi-Asset Through a Principal Components Lens

PCA decomposes a covariance matrix into a set of uncorrelated variables. In our case, we will use PCA on the correlation matrix (essentially a standardized covariance matrix) to account for the fact that our risk factors have very different scales and volatilities.

If we start with six risk factors, we get six principal components. Since 90% of the variance is explained by the first four components, we will focus on those.

One issue with PCA is that the components do not always have an easily ascribed meaning. They are mathematically uncorrelated, but their economic interpretation can be dubious and may not always hold over different time periods.

Nevertheless, we can take a stab at it.

PCA Risk Factor Allocations

 Component 1Component 2Component 3Component 4
U.S. Equity56%10%5%-20%
Global Equity-21%2%87%36%
Level0%84%14%-40%
Slope32%37%-29%81%
Credit-55%-2%-36%1%
Currency-48%38%-10%13%

Source: MSCI, St. Louis FED, CSI Analytics. Analysis by Newfound. Data from 4/30/2012 to 4/30/2017.

From the allocations to each risk factor, we see that Component 1 is a combination of U.S. equity exposure, shrinking credit spreads, and a declining dollar. This is basically the issue we saw in the risk factor correlation matrix. Under a PCA lens, these risk factors are largely treated as one.

Component 2 is chiefly based on changes in the level of interest rates; Component 3 is global equity with some credit thrown in; and Component 4 is steepening of the yield curve.  Trust me, it is not always so straightforward…

With these components, we can do the same type of analysis as we did with the risk factors.

“Multi-Asset” Fund Risk Factor PCA Exposure Gini Scores

Source: Morningstar, MSCI, St. Louis FED, CSI Analytics. Analysis by Newfound. Data from 4/30/2012 to 4/30/2017.  

“Multi-Asset” Fund Risk Factor PCA Exposure Gini Scores Summary Statistics

 Gini Score
Average0.70
Median0.77
Standard Deviation0.23
Minimum0.20
Maximum1.00

Source: Morningstar, MSCI, St. Louis FED, CSI Analytics. Analysis by Newfound. Data from 4/30/2012 to 4/30/2017.

We now have more dispersion but still a high concentration of funds exposed to predominantly one or two risk factors.  The fund with the lowest Gini score was exposed to all of the principal component risk factors with allocations that ranged between 13% and 28%.

 

Our Take on “Multi-Asset”

We believe that multi-asset strategies should more reliably be “multi-asset”.  Our Multi-Asset Income strategy, is built with this in mind.[4]  Not only are multiple assets important, but multiple risk exposures are necessary for actual diversification, especially during a market crisis.

The strategy can be broken down into a strategic component (MAI strategic), which uses Sharpe parity to weight the 16 individual asset classes, and a tactical version (MAI tactical), which selectively removes asset classes exhibiting negative trends from the strategic portfolio.

In the previous analysis, many of the strategies showed significant concentration in a small set of risk factors. While this is not a bad thing if the best performing risk factors are the ones selected, this lack of diversification can be a shock for investors who are thinking their fund is “multi-asset”.

Using the different lenses from before, we can see where the two MAI portfolios fit in.

“Multi-Asset” Fund Risk Factor Exposure Gini Score with MAI strategy

Source: Morningstar, MSCI, St. Louis FED, CSI Analytics. Analysis by Newfound. Data from 4/30/2012 to 4/30/2017.

From a risk factor exposure perspective, the strategic portfolio in MAI has a Gini score of 0.28. The tactical portfolio has a higher Gini score of 0.47, as the portfolio shifted toward equity exposure and away from rates.

“Multi-Asset” Fund Risk Factor PCA Exposure Gini Scores with MAI strategy

Source: Morningstar, MSCI, St. Louis FED, CSI Analytics. Analysis by Newfound. Data from 4/30/2012 to 4/30/2017.

From a principal component exposure perspective, the strategic portfolio in MAI has a Gini score of 0.43. The tactical portfolio has a lower Gini score of 0.19, as the portfolio balanced the exposure from Component 1 (U.S. equity and high yield) into the other components.

 

Conclusion

Multi-asset strategies do, by definition, hold more than one asset class, but the extent to which they do this can be tenuous.  Additionally, holding multiple asset classes does not guarantee diversification among risk factors.

When selecting a multi-asset strategy, it is important to assess the asset classes that it may hold (e.g. does it only focus on core bonds and equities?) and the amount of flexibility it has when allocating (e.g. can it switch from an 80/20 portfolio to a 20/80 portfolio?). Pairing two or three strategies together can be a good way to reap the benefits of process diversification and fill in gaps when different strategies focus on different segments of the multi-asset space.

Blindly selecting a multi-asset strategy and thinking it will be diversified can be a big mistake in portfolio construction.

 

[1] Note that at least one fund had 100% exposure to the residual. There was one odd return in the data for that fund that was possibly an outlier event. From this high-level view, seeing a number like that should be a red flag that more scrutiny is required for that investment.

[2] The correlation between U.S. equities and credit is negative because we are looking at credit spreads, which move in the opposite direction of high yield bond prices.

[3] We used this before in our commentary on shocking the covariance matrix.

[4] We described this strategy in a previous commentary from a factor-based perspective.

Nathan is a Vice President at Newfound Research, a quantitative asset manager offering a suite of separately managed accounts and mutual funds. At Newfound, Nathan is responsible for investment research, strategy development, and supporting the portfolio management team.

Prior to joining Newfound, he was a chemical engineer at URS, a global engineering firm in the oil, natural gas, and biofuels industry where he was responsible for process simulation development, project economic analysis, and the creation of in-house software.

Nathan holds a Master of Science in Computational Finance from Carnegie Mellon University and graduated summa cum laude from Case Western Reserve University with a Bachelor of Science in Chemical Engineering and a minor in Mathematics.