Back

Blogs

Deconstructing Active: Multi-Factor ETFs for Designing Investment Programs

Note: an edited version of this article was first published on ETF.com.

Several ETF.com articles have been published on ‘Multi-Factor Exchange-Traded Funds (ETFs)’ specifically commenting over issues of model design (see articles from Larry Swedroe and Corey Hoffstein).  This article expands upon the perspective provided by Todd Rosenbluth (CFRA) on the importance of looking underneath the hood when evaluating multi-factor ETFs.  For this article, I ran several U.S.-focused multi-factor ETFs through Bloomberg’s Risk Model as a means to objectively compare the ETFs’ risk profiles, further demonstrating the importance of ‘knowing what you’re buying.’

As a quick reminder, multi-factor ETF sponsors build baskets of equity securities using a rules-based approach that screens or scores stocks based on attractive attributes such as cheap valuation, high dividend yield, high quality, positive momentum, and/or low volatility.  The sponsors can then apply separate rules for combining these attributes to determine portfolio membership and weighting.  Multi-factor models can range from a two-factor signal (i.e. Fama/French small cap value) to ones capturing multiple signals across the style box spectrum.  Multi-factor ETFs provide investors with cost-effective quantitative-based strategies historically offerred by traditional quantitative managers employing factor-based investing.  For more background information on factor-based investing, I suggest reading MSCI’s Foundations of Factor Investing.  Although barely in existence a year ago, a handful of U.S.-focused multi-factor ETFs have already garnered $3.3 billion in assets and the popularity of multi-factor investing has spilled over into international markets. 

Multi-factor ETFs encapsulate the disruptions and innovations wrought by the advent of ETF technology and are positioned to partake in the secular shift from traditional active management to passive indexing.  Multi-factor ETFs embody the deconstruction of traditional active management, namely much of what is delivered by a traditional active management investment program can be more cost-effectively captured using a multi-factor approach.  The traditional style box is under siege as investors have come to realize that traditional value and growth managers get much of their relative performance from common risk factors rather than superior stock-picking.  Does this hasten the demise of active management?  Not necessarily, but single and multi-factor ETF programs can free up fee and risk budgets that would then enable investors to more efficiently allocate client capital to those true alpha-seeking opportunities provided by skilled managers (this assumes an investor can identify those managers ahead of time).  The increasing challenge for active managers is the ability to differentiate aspects of their process that cannot be easily captured with rules-based approaches, just as we wrote in the July ETF.com article, “Here’s How Active Can Compete With Low Cost ETFs.” 

Evaluating Multi-Factor ETFs – Understand What Is Being Delivered

As with all ETFs, investors evaluating multi-factor ETFs should consider both basket design and expense ratios.  U.S. large cap multi-factor ETFs will be priced much more cheaply than small caps and ex-U.S. markets.  It is even more critical in understanding what exposures are provided by a multi-factor ETF in the context of the role that ETF plays in the broader investment program.   Multi-factor investing can provide superior risk-adjusted returns over a market-cap portfolio, but it also introduces more moving parts that the investor must now keep track of.  The investor will want to ensure that multi-factor exposures complement other aspects of the program such that investors do not find themselves doubling up on a factor such as value or momentum. 

Multi-factor investing also introduces active risk, or tracking error, which must be monitored by the investor.  One of the challenges in multi-factor investing is to account for the cyclicality in factor performance (or as MSCI puts it, “there is no free lunch attached to factor investing…”).  This is where understanding ETF design becomes important.  For instance, a debate has been brewing among academics on what constitutes a superior approach when combining factors where such combinations can potentially mitigate factor cyclicality through diversification.

There are two primary ways for combining factors: the sleeve approach and the integrated approach as spelled out by Larry Swedroe and Corey Hoffstein in the ETF.com articles referenced above.  Kal Ghayur, whose team oversees Goldman Sachs’ ActiveBeta strategies, along with Ronan Heaney and Stephen Platt, recently published a paper (“Constructing Long-Only Multi-Factor Strategies: Portfolio Blending versus Signal Blending” – the abstract can be found at SSRN) addressing the claims put forth by other studies that an integrated signal approach produces superior risk-adjusted results over a sleeve approach.  By way of disclosure, ActiveBeta uses a sleeve approach by combining separate factor portfolios into one combined portfolio rather than combining the factor signals into one composite signal before constructing the basket. 

In their paper, Ghayur et al acknowledge the superior risk-adjusted results of the integrated approach for portfolios targeting higher factor exposures or higher tracking error, but not at lower factor exposures/lower tracking error levels (which is one reason Goldman’s ActiveBeta ETFs run lower tracking error than their peers – see below).  According to the paper, “…[the interaction effects between factors] are overwhelmed by the high concentration and stock-specific risk in the [sleeve approach] at high levels of factor exposures.”  The sleeve-approach will find itself “highly concentrated in individual factor portfolios” at higher tracking error levels while the integrated approach “retains the benefits of diversification.”  Therefore, the non-factor sources of volatility (i.e. stock-specific risk) are more prominent in a sleeve approach than an integrated approach for higher levels of tracking error.  

Multi-Factor ETFs Through the Eyes of a Risk Model

As Ghayur et al emphasize in the paper, your starting point matters when building an investment program around multi-factor ETFs.  Lower tracking error multi-factor ETFs are generally cheaper and free up more of the investor’s fee and risk budget to complement other components of the program, whereas higher tracking error multi-factor ETFs serve better as all-in-one solutions.

Now, some portfolio weighting schemes may de-emphasize tracking error in pursuit of maximizing multi-factor exposure but at the cost of introducing more systematic market risk.  For instance, multi-factor portfolios that do not account for tracking error will find themselves taking on more ‘size’ or small cap risk to the point that comparisons with cap-weighted benchmarks become less meaningful.  In addition, how one defines the individual factors can matter more than how one combines them.  Goldman’s ActiveBeta uses a different definition of ‘value’ by incorporating free cash flow (in addition to book/price and sales/price) which tends to emphasize less capital intensive companies that are more involved in providing services such as media.  More traditional value factors will incorporate some combination of book value, sales, earnings, and EBITDA (earnings before interest, taxes, and depreciation/amortization).  Some also incorporate dividend yield, although this factor tends to behave more differently versus other value factors. 

Hence, it is important for investors evaluating multi-factor ETFs to incorporate the findings of objective holdings-based risk models such as the ones provided by AXIOMA, MSCI BARRA, and Bloomberg.  Risk models seek to forecast the absolute risk of a portfolio based on the risks of its underlying holdings.  They then deconstruct this risk into primary risk measures such as market (beta), industry, country, and style or fundamental factors.  In other words, a risk model estimates the sources of a portfolio’s active risk, whether coming from market timing, picking/avoiding sectors and countries, and/or investing in factor portfolios.    

Here I reference Bloomberg’s Global Equity Fundamental Factor Model, profiling some major multi-factor ETFs against the Russell 1000 Index (proxied by IWB).  The table below displays the following:

1. Estimated market beta (0.90 to 1.10 is a typical range for a fully invested equity portfolio)
2. Estimated active size coefficient (negative coefficient indicates smaller cap exposure versus the benchmark).  A more negative value indicates the ETF tends to hold more small and mid-caps relative to the Russell 1000.  
3. Factor to Industry Risk Contribution Ratio.  This indicates the ratio of how much active risk is sourced from fundamental factors versus industry positioning.  Note, this ratio does not take into account interaction effects between the two, but it’s a rough approximation of how much active risk comes from factor exposures as opposed to active industry exposures. 
4. Active Exposures to Bloomberg Fundamental Factors.  These two columns display the positive and negative active exposures to common factors such as value, growth, momentum, and yield.  Those highlighted in blue indicate that the factor exposures are greater than that of the size factor. 

So, what are some of the takeaways from this table?

1. How much active risk is the multi-factor ETF taking versus a cap-weighted benchmark?  Arguably, not all the multi-factor ETFs are benchmarking themselves against the Russell 1000, but they all start with a mid-/large cap universe.  GSLC and QUS have the smallest active risk amongst the group and their lower fees reflect that; JHML has the lowest projected tracking error at 0.95% so it is difficult to justify the 0.35% expense ratio.  In contrast, DEUS has the highest projected tracking error at 3.86%, and this is largely reflected in their sizeable smaller cap tilt (see #2). 
2. How much ‘smaller cap’ risk is the multi-factor ETF taking versus a cap-weighted benchmark?  For instance, DEUS has the greatest smaller cap tilt among the profiled ETFs such that Morningstar classifies DEUS as mid-cap in its style categorization.  At the other extreme, QUS has an insignificant size tilt, so it won’t benefit as much from the small cap premium versus the other ETFs.
3. Does factor exposure comprise the majority of active risk?  Having active industry exposures is not uncommon in factor portfolios, particularly ones incorporating momentum, where as much of the historical factor premium can come from sector exposures as it does from individual stocks.  However, investors in multi-factor ETFs expect more of the active risk to come from factor exposure rather than sector or country.  TILT and ROUS have the highest factor-to-industry exposure ratios relative to their projected active risk. 
4. Where are my greatest factor exposures?  Most multi-factor ETFs will overweight value and momentum and underweight growth.  Some will emphasize quality (as proxied by profitability) and yield, as well as avoiding high volatility.  GSLC primarily emphasizes profitability above all the other factors, while ROUS has a heavy emphasis on value. 

Keep in mind that these are projections, but the academic literature is pretty consistent in its findings that risk is easier to forecast than returns.  Risk persists, so a portfolio with a small cap value tilt today will likely be influenced by how small cap value performs tomorrow, assuming little turnover.   Multi-factor ETFs serve as great starting vehicles that can serve as core holdings within a broader investment program, but investors need to be mindful of the active risk levels and factor exposures presented by multi-factor ETFs. 

At the time of this writing, 3D Asset Management did not hold any of the ETFs listed in the article. The above is the opinion of the author and should not be relied upon as investment advice or a forecast of the future. The projections or other information regarding the likelihood of various investment outcomes are hypothetical in nature, do not reflect actual investment results, and are not guarantees of future results. It is not a recommendation, offer or solicitation to buy or sell any securities or implement any investment strategy. It is for informational purposes only. The above statistics, data, anecdotes and opinions of others are assumed to be true and accurate; however, 3D Asset Management does not warrant the accuracy of any of these. There is also no assurance that any of the above is all inclusive or complete. Past performance is no guarantee of future results. None of the services offered by 3D Asset Management are insured by the FDIC, and the reader is reminded that all investments contain risk. The opinions offered above are as of January 12, 2017, and are subject to change as influencing factors change. More detail regarding 3D Asset Management, its products, services, personnel, fees and investment methodologies are available in the firm’s Form ADV Part 2, which is available upon request by calling (860) 291-1998, option 2, or emailing sales@3dadvisor.com or visiting 3D’s website atwww.3dadvisor.com.