All analysis in this report is backtested and hypothetical and does not include management or trading costs. Please read the important disclosures at the end of this summary.
At Newfound Research, we run a number of actively risk-managed strategies. The objective of these strategies is to provide meaningful upside participation while avoiding significant drawdowns during bear markets.
To achieve this, we embrace a trend-following approach, avoiding those investments identified as having a negative trend.
One such portfolio is our Risk Managed U.S. Sectors strategy. This portfolio invests in exchange traded funds (ETFs) representing the 9 primary U.S. sectors (e.g. energy, financials, utilities, et cetera). When a sector has a positive signal, it is given an equal-weight within the portfolio. When a sector exhibits negative momentum, it is removed, and available capital is re-allocated to remaining positive sectors. Each sector is capped at a 25% allocation, however, meaning that when 3 or fewer sectors have positive trends, the portfolio will build a position in short-term U.S. Treasuries.
As a baseline comparison, we’ll use the S&P 500, which was up 1.2% in 2015.
We expect such a strategy to achieve its objective when there are a number of significant trends (either positive or negative), but struggle when the majority of sectors exhibit sideways behavior. As with most trend following strategies, sideways markets can cause whipsaw, where positions are bought at highs and sold at lows.
2015 was a difficult year for such a portfolio. After a difficult year, we believe it is important to do a deep dive to figure out exactly what went wrong and whether it raises concerns about the assumptions or processes the portfolio is built upon. (Note: While the strategy certainly underperformed the S&P 500 this year, it is important to remember a couple of things. First, the S&P 500 is not a great short-term benchmark for many active asset allocation approaches. More appropriate benchmarks (i.e. an equity position with an option overlay for risk management purposes) would show a smaller performance gap. Second, any active approach will go through periods of under- and out-performance. No strategy outperforms all of the time. Historically speaking, the underperformance this year was quite moderate relative to our expectations based on historical data).
The easiest way to begin evaluating the portfolio is to look at the trend following returns generated on a sector-by-sector basis. In other words, we will treat each sector as its own trend following system and evaluate them individually.
For this analysis we use the SPDR Select Sector ETFs.
- XLB: Materials
- XLE: Energy
- XLF: Financials
- XLI: Industrials
- XLK: Technology
- XLP: Consumer Staples
- XLU: Utilities
- XLV: Health Care
- XLY: Consumer Discretionary
What we can see is what we would expect: while several sectors eked out positive returns (e.g. XLY, XLP and XLV), other sectors were significant detractors (e.g. XLB, XLK, XLI, and XLF).
Another important way to evaluate this is how did each trend following strategy do relative to a buy and hold approach. For example, while our trend model stayed out of XLE all year, and therefore generated no return, XLE had significant losses. Therefore, compared to buy and hold, our trend following added significant value.
We can see, however, that while some of our individual sector trend following systems generated positive returns, they underperformed relative to buy and hold (e.g. XLV).
An equal-weight allocation to these 9 systems would have returned -4.7%. In other words, trend following caused a drag of -340bp on top of the strategic portfolio drag.
Acknowledging that sideways markets can be particularly difficult for a trend following model to navigate, we employ a dollar cost averaging approach. Instead of implementing each trend signal immediately, we implement it over a 4-week period (assuming the signal stays consistent). Our hypothesis is that in a sideways market, a trend signal will turn on at exactly the wrong time. Therefore, by dollar cost averaging, we can limit the cost of whipsaw.
One theme we continue to hear is that many investors believe that since markets are “moving faster,” we should act faster to avoid getting whipsawed. Our dollar cost averaging approach flies in the face of this logic. So the question is, did the approach add value in real-time in the sideways markets of 2015?
We can see that for many of the trend following systems, dollar cost averaging was very additive to performance. For example, dollar cost averaging added over 650bp of performance to the trend following systems of XLF, XLI and XLK and added 490bp to XLU. On the other hand, dollar cost averaging was a drag for XLB, causing that system to lose an extra 546bp more.
An equal-weight portfolio of the dollar cost averaged trend following models would have returned -2.7%, meaning that the dollar cost averaging helped claw back 200bp that trend following lost. So whipsaw was still a cost, but instead of -340bp, it was reduced to 140bp.
At the beginning of this article, we discussed how the portfolio rules work. Another way to interpret those rules is that the portfolio has an equal weight allocation to each of the 9 underlying trend following strategies and simply levers up and down exposure to those strategies.
For example, when a positive trend is identified in each sector, each strategy is given its base-line allocation of 1/9th of the portfolio. However, if 8 of the 9 sectors are off, those trend following strategies are entirely de-levered and the one remaining strategy is levered up 2.25x, since it is 25% of the portfolio.
While no leverage is actually applied, this model of thinking about the portfolio allows us to examine how the rules of portfolio construction affected portfolio return. We can do this simply by translating the rules at any given time into how much “leverage” we would be applying to each trading strategy.
Now, it is important first to determine that this model of portfolio construction is indeed equivalent. Below I plot the hypothetical Newfound Risk Managed U.S. Sectors model returns as well as the hypothetical results from equally-weighting exposure to the 9 trend following strategies, levered up and down based on available capital. We can see they are identical.
So then how did applying leverage to each dollar-cost averaged trend following system work?
Not well. We can see the results even more clearly on a sector-by-sector breakdown of relative performance versus buy and hold.
For nearly every single sector, relative performance was made worse by the application of “leverage” – i.e. the rules we use to build the portfolio and re-use available capital. For example, we can see that the relative loss for XLU was multiplied 3x. In aggregate, the rules caused a drag of -275bp.
We can see the 2015 total return of each step of protfolio construction below:
So we really have three significant causes of underperformance:
- An equal-weight sector approach under-performed the S&P 500 by -250bp
- Trend following created a drag of -340bp from whipsaw
- Dollar cost averaging trades over time helped reduce the whipsaw cost, adding back 200bp, putting the total cost of whipsaw from trend following at -140bp
- The rules of portfolio construction caused another -275bp drag on performance
- Should we continue to equal weight the sectors?
- Are our trend signals “broken”?
- Should we continue to use this rule set to build the portfolio?
So are our trend signals “broken”? We believe not: whipsaw is expected in a sideways market and we believe that -140bp of whipsaw cost is well within normal limits for such a period.
So what about our rules? Why bother applying the “leverage”? Why not just give each system 1/9th of the portfolio and let it sit in cash if there is not positive trend?
We apply the leverage because it helps cut down on cash drag, especially in periods where a large number of sectors turn off and only a few turn back on. Given that one half of the stated objective of the portfolio is to capture as much upside potential of the equity market as possible, we believe being invested is a critical component of achieving that goal.
Below we plot the hypothetical RMUS index versus an equal weight portfolio of the 9 dollar cost averaged trend following strategies. We can see that when the trend following strategies are in cash, and we do not re-use that cash, it can cause a significant drag.
This is made even clearer when we plot the relative performance of the two equity curves:
We can see that while the re-use of available capital has been additive to performance over the long-term, there are periods that it can cause short-term underperformance. These periods have generally been short-lived and spread out.
While trend following was not a successful approach as applied in 2015, we believe that it is critical to analyze exactly which pieces of portfolio construction were responsible for how much drag.
In our analysis, we’ve found that the equal weight sector approach was responsible for 37.8% of the under-performance, the trend following systems were responsible for 21.4% of the under-performance, and the portfolio construction rules were responsible for the remaining 40.8% of under-performance.
Based on this analysis, we do not believe that any part of our portfolio process is fundamentally broken, but are reminded of Murphy’s law of portfolio management: sometimes everything that can go wrong will.
DISCLOSURE: Important Disclosures Newfound began to actively calculate the performance of the Newfound Risk Managed U.S. Sectors index on February 26, 2015. Performance results prior to February 26, 2015 are all backtested and hypothetical. Newfound began to manag