This post is available as PDF download here.

Summary

  • As more thematic products come to the market, it can be difficult for investors to decide how to allocate to them, even if they believe in their future potential.
  • The sector disruptors are a suite of products that focus on areas of the economy that are heavily influenced by new technologies.
  • Taking a factor-based approach using, for example, low volatility and momentum has the potential to boost returns while simultaneously managing risk.
  • While the short history of these products does not lend itself to a thorough statistical analysis, relying on factors that have been shown to be robust can be a way to balance the tension between a naïve allocation and the tendency to be too active simply for the fear of missing out.

Investing in entire sectors of the economy is commonly a way investors express their own views on returns without bearing the risk the goes with picking individual or small handfuls of stocks. ETFs make this task especially simple.

But what defines a sector?

Most investors are familiar with the broad sector classifications set forth by the Global Industry Classification Standard (GICS): energy, consumer discretionary, technology, health care, etc. However, these groupings are about to change.

Next month, the changes made last year to GICS will propagate through to many popular ETFs. The changes affect 10% of the market cap of the S&P 500 and will focus on the Technology, Telecommunications, and Consumer Discretionary sectors.

The telecommunications sector will be renamed “communication services” and it will be expanded to include some big consumer discretionary names (e.g. Netflix, Walt Disney, and Comcast) and technology names (e.g. Facebook and Alphabet), affecting over 20% of the market cap in each of these sectors. Aside from splitting up the FAANG stocks among three sectors instead of two, the changes will bring the S&P 500 closer to equal weight from a sector perspective and will move the smallest and oftentimes forgotten telecommunications sector to the middle of the market cap pack under its rebranded name.

With the further adoption of ETFs, the definition of a “sector” has been greatly refined and subdivided. For example, investors can invest in semiconductors, aerospace, biotech, agribusiness, and gaming, to name a few.

These thematic sector products can be interesting for investors who desire to access a subset of companies within a targeted industry. The question is often how to allocate to them.

In this commentary, we will explore a factor-based approach to allocating across a suite of these thematic sectors.

Sector Disruptors

GlobalX has a suite of thematic sectors with a “bad boy” slant: the sector disruptors.[1] These sectors aim to profit from advances in technology and a changing consumer base that demands new approaches within traditional sectors.

There is a rough correspondence between the sector disruptors and the traditional ones they have the potential to disrupt. While there are not currently disruptors in the consumer staples, materials, or utilities sectors, most of the larger sectors are covered.

Traditional SectorSector Disruptor
Consumer DiscretionaryMillennials (MILN)
EnergyLithium (LIT)
FinancialsFinTech (FINX)
Health CareHealth & Wellness (BFIT)
IndustrialsRobotics & AI (BOTZ)
TechnologyInternet of Things (SNSR)
TelecomSocial Media (SOCL)

 Source: GlobalX

From a performance standpoint, an equal weight portfolio of these sector disruptors, starting from when there is data for 4 of them, has outperformed the S&P 500 by 140 bps annually (14.4% compared to 13.0%). It has also outperformed on a risk-adjusted basis (0.90 compared to 0.84).

Source: Solactive, Indexx, and CSI Analytics. Calculations by Newfound. Past performance is not a predictor of future results.  All information is backtested and hypothetical and does not reflect the actual strategy managed by Newfound Research.  Performance is net of all fees except for underlying ETF expense ratios.  Returns assume the reinvestment of all dividends, capital gains, and other earnings.  Data from 12/31/2011 to 7/31/2018.

Compared to the broader market, most of these sectors are heavily tilted toward mid and small cap stocks, which makes sense with how niche some of the markets are.

Source: ETFdb.com. Data as of 7/31/2018.

These ETFs also invest globally, except for the Millennials ETF (MILN), which is all U.S.-based. The Internet of Things ETF (SNSR) and the FinTech ETF (FINX) both tilt toward the U.S., but others concentrate more heavily on other countries. For example, the Robotics and AI ETF (BOTZ) has a 43% allocation to Japan and the Social Media ETF (SOCL) has 31% of its holdings in China.

Source: ETFdb.com. Data as of 7/31/2018.

One thing of note when investing in sectors like these is that the traditional sector demarcations can be very blurred. As could probably be surmised from the market trends these ETFs are trying to capitalize on, there is a large focus on the technology sector.

The FinTech ETF (FINX) only holds 9% financials, and the Health & Wellness ETF (BFIT) only holds 7% in healthcare as defined by traditional sector criteria.

Regardless of how we allocate, the portfolio will likely have a sizeable technology exposure.

Source: ETFdb.com. Data as of 7/31/2018.

Since we will ultimately be constructing a factor portfolio, it may be helpful to know whether any of the sectors have inherent factor exposure themselves.

We see that most of them have significant exposure to the market (i.e. beta). Some have a beta of more than 1, and the Health & Wellness ETF (BFIT) is the only one with a beta significantly less than 1.

Source: Solactive, Indexx, CSI Analytics, and AQR. Calculations by Newfound. Data from 12/31/2011 to 6/30/2018.

The only significant factor exposure in the sectors is for BFIT and SNSR to the size premium (SMB). There is no significant exposure to value (HML), momentum (UMD), quality (QMJ), or low beta (BAB). Four of the sectors (SOCL, FINX, MILN, and BOTZ) have a significant, positive alpha at the 95% confidence level, which indicates that our factor model does not account for some of the risks in these sectors. In other words, these sectors could provide diversifying sources of return.

The Factor Portfolio

As a quantitative asset manager, we prefer not to take views on the future prospects of, say, lithium technologies. However, as has been demonstrated repeatedly in academic literature and empirical studies, using established factors such as value, momentum, low volatility, and quality can lead to a systematic portfolio that has a greater potential to outperform a naïve equal weight allocation over the long run.

To test as far back as possible, we will utilize index data from Solactive (for SOCL and LIT) and Indexx (for the remaining ETF) prior to inception.

Unfortunately, with the limited access to historical valuation data (e.g. P/E, P/B, and P/S) and quality metrics (e.g. ROE, ROA, and EPS), we will only focus on factors that can be directly measured from prices.

To make our analysis more robust to model specification, we will utilize multiple measures for each factor and a span of values for the parameters.

For momentum, we will look at the following models:

  • Time Series Momentum: The return over the prior N-periods is positive.
  • Price-Minus-Moving-Average: The percentage distance of the price relative to its (N/2)-period exponentially-weighted moving average.
  • EWMA Cross-Over: The percentage distance between the (N/4)-length exponentially-weighted moving average and the (N/2)-length exponentially-weighted moving average.

For volatility, we will look at the following models:

  • Simple volatility: The annualized standard deviation of the previous N-periods.
  • EWM Volatility: The (N/2)-period exponentially weighted volatility.

These models are not independent and are not the only models we could specify. For example, in momentum, we could use a risk-adjusted, “frog in the pan”, or regression-based measure of momentum. However, by using multiple models, the goal is to reduce the risk of picking the “right” (or “wrong”) one by chance.

For time frames, we will look at the 10 periods ranging from 3 to 12 months, inclusive.

On each day, we will take the top 3 sectors according to each model and average all allocations together for each factor. We will tranche the allocations over a 21-day (essentially one-month) period to account for timing luck.

The allocations for the momentum portfolio are shown below.

All sectors have been held at some point with Millennials (MILN) and Robotics & AI (BOTZ) being favored. The annualized turnover has been 173%.

Source: Solactive, Indexx, and CSI Analytics. Calculations by Newfound. Data from 12/31/2011 to 7/31/2018.

The allocations for the low volatility portfolio are shown below.

Source: Solactive, Indexx, and CSI Analytics. Calculations by Newfound. Data from 12/31/2011 to 7/31/2018.

All sectors have again been held at some point with Millennials (MILN), Robotics & AI (BOTZ), and Health & Wellness (BFIT) being favored. The annualized turnover has been 54%.

Since we are focus on long only implementations of these factor strategies, we can construct long/short portfolios of the factor with the short leg as the equal weight sector portfolio to analyze the returns using the standard approaches employed in factor research.

Over the limited history available, we can see that both momentum and low volatility have generated returns in excess of the equal-weight portfolio, although low volatility has had a much bumpier ride.

Source: Solactive, Indexx, and CSI Analytics. Calculations by Newfound. Past performance is not a predictor of future results.  All information is backtested and hypothetical and does not reflect the actual strategy managed by Newfound Research.  Performance is net of all fees except for underlying ETF expense ratios.  Returns assume the reinvestment of all dividends, capital gains, and other earnings.  Data from 12/31/2011 to 7/31/2018.

The returns for the momentum factor are significant at the 95% confidence level. The returns for the low volatility factor are not significant at this level, but they are on par with what a similar analysis on low volatility ETF like USMV would show.

Low VolatilityMomentum
Annualized Return1.1%5.0%
Annualized Volatility4.2%3.8%
T-statistic0.613.02

Source: Solactive, Indexx, and CSI Analytics. Calculations by Newfound. Past performance is not a predictor of future results.  All information is backtested and hypothetical and does not reflect the actual strategy managed by Newfound Research.  Performance is net of all fees except for underlying ETF expense ratios.  Returns assume the reinvestment of all dividends, capital gains, and other earnings.  Data from 12/31/2011 to 7/31/2018.

As is often the case when using factors, there are diversification benefits to using multiple factors simultaneously. The correlation between the low volatility and momentum factor is low (0.26), indicating that holding both portfolios together can potentially boost risk-adjusted returns.

While the benefits of this approach are definitely not uniformly realized over the backtest, there is still an increase in returns and a decrease in volatility.

Source: Solactive, Indexx, and CSI Analytics. Calculations by Newfound. Past performance is not a predictor of future results.  All information is backtested and hypothetical and does not reflect the actual strategy managed by Newfound Research.  Performance is net of all fees except for underlying ETF expense ratios.  Returns assume the reinvestment of all dividends, capital gains, and other earnings.  Data from 12/31/2011 to 7/31/2018.

Blended Long-Only Factor PortfolioEqual Weight Disruptor Portfolio
Annualized Return18.8%16.4%
Annualized Volatility11.7%12.3%
Sharpe Ratio1.581.30

Source: Solactive, Indexx, and CSI Analytics. Calculations by Newfound. Past performance is not a predictor of future results.  All information is backtested and hypothetical and does not reflect the actual strategy managed by Newfound Research.  Performance is net of all fees except for underlying ETF expense ratios.  Returns assume the reinvestment of all dividends, capital gains, and other earnings.  Data from 12/31/2011 to 7/31/2018.

Adding in the Value Factor

At the beginning of this analysis, we stated that the lack of fundamental metrics makes constructing quality and value factors difficult. Having these data points is important because we would expect fundamentals to mean revert more so than prices. However, if we make the assumption that the fundamentals (e.g. earnings, sales, etc.) are at least roughly constant, then we can look at mean reversion in Sharpe ratios as a proxy for value.

Even under this assumption, the difficulty in this specific case of the sector disruptors is the history is too limited to establish reasonable expectations for what the mean Sharpe ratio over the long-term even should be.

In the sectors with the longest history, such as Millennials (MILN) and the Internet of Things (SNSR), we only have enough data to calculate a five year Sharpe ratio for three years. And those three years are on a rolling basis, which means that none of the periods are independent. We would need much more data before we could see an oscillatory pattern developing and an estimate of what the long-term mean should be.

Sectors with shorter histories, like FinTech (FINX), make the prediction even more difficult.

Under this framework, sectors that are significantly above their long-term mean could be underweighted and sectors that are significantly below their long-term mean could be overweighted.

Conclusion

When new assets, such as very specific sector ETFs come to market, it can be difficult to decide how to incorporate them into a portfolio.

Factors, such as momentum and low volatility, can be used to remove investor biases that play into the hype of new products.

While the history of a new product is generally too short to do a thorough statistical analysis, factor investing has been shown to be pervasive across asset classes and geographies. Because the sector disruptors are equities, these factors should still be a way to allocate systematically in ways that can boost risk-adjusted returns over the long-run.

Ultimately, the benefits of an approach like this must be weighed against frictions like transaction costs and taxes. But a factor-based approach can be a way to avoid taking a naïve approach or an approach that leads to overly concentrated portfolio bets based on beliefs in certain disruptors that don’t pan out.

[1] https://www.globalxfunds.com/meet-the-sector-disruptors/

Just this past week, Goldman Sachs filed for a suite of very similar “motif” ETFs. Let the fee wars begin!

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.