This post is available as a PDF download here.
- Factor investing (value, momentum, low volatility, carry, trend, etc.) is well-known in equities but less discussed in other asset classes.
- However, many of these factors are just as prevalent in other asset classes, such as bonds, commodities, and currencies.
- In this case study, we explore the risk-adjusted carry and trend factors that we seek to incorporate in our own Multi-Asset Income portfolio.
People who read our commentary are often somewhat confused by the fact that we do not actually offer any single-factor equity products (e.g. a value portfolio). We do write about them quite frequently, after all.
The answer is because factors are a pervasive concept. Studying factors in equities can give insight into incorporating factor concepts into multi-asset portfolios.
In this commentary, we wanted to present a case study of how we incorporate factors into our Multi-Asset Income portfolio.
Factors, Factors Everywhere
As it turns out, factors that are popular in the equity space often translate well (or sometimes better) to the multi-asset space. (To be clear, when we say the “multi-asset” space, we mean anything in the investment hierarchy above security selection. In our definition, this would include sector selection, country / region selection, and asset class selection.)
For example, in all the following definitions, we can more-or-less replace “stocks” with “assets” and use the same definition.
- Value: Buy cheap stocks; sell expensive ones.
- Momentum: Buy strong performers; sell expensive ones.
- Low-volatility: Buy low volatility stocks; sell high volatility ones.
- Carry: Buy high yielding stocks; sell low yielding ones.
- Trend: Buy positively trending stocks; sell negatively trending ones.
Of course, not everything is an exact 1-for-1 equivalent. Defining “value” in the equity space is much easier than, say, the bond or commodity space, where we might have to look at something like real yields or 3-5 year performance (often called “long-term mean-reversion” or “contrarian timing”).
When we then stretch the problem to look beyond a single asset class, we must be careful how we compare metrics. For example, the high value in one asset class may be inferior to the high value we are seeing in another asset class. Standardization is important for comparisons.
While factor investing has, historically, been an equity study, some factors are actually more prevalent in the multi-asset space. For example, commodity trading advisors (CTAs) and those offering managed futures strategies have been leveraging the trend factor in multi-asset portfolios for decades.
Similarly, “carry” is concept more frequently tied to fixed income or currency investing than it is in equities, largely because “high yield” in the equity space is a cousin of the value-factor.
What we see is that many of the same concepts we study in equities can be readily applied to multi-asset portfolios.
Case Study: Introducing Newfound’s Multi-Asset Income Portfolio
Newfound’s Multi-Asset Income portfolio seeks to provide access to both traditional (e.g. global Treasuries, dividend stocks, and corporate bonds) and non-traditional (e.g. high yield bonds, REITs, MLPs, bank loans, preferreds, and EM debt) income-based asset classes within a disciplined risk management framework that seeks to avoid significant drawdowns.
How, exactly, do we aim to do that? We look to leverage three factors:
- Carry: Overweight assets offering a higher yield; underweight assets offering a lower yield.
- Low-volatility: Overweight assets with lower volatility levels; underweight assets with higher volatility levels.
- Trend: Avoid asset classes exhibiting negative trends.
Carry is the return earned by an asset assuming its price does not change. It is most often captured by buying high-yielding assets and shorting low-yielding assets. In many ways, carry is closely tied to value – or contrarian – investing. All else held equal, if an asset’s price falls, its yield will rise assuming that the distributions do not change significantly. Therefore, as relative yields change over time, carry investing can overweight assets selling at a discount to their intrinsic value.
Low-volatility is an anomaly whereby assets exhibiting lower volatility have historically outperformed assets exhibiting higher volatility on a risk-adjusted basis. This violates neoclassical financial theory where investors should earn more reward for bearing more risk. Explanations for this anomaly include investors’ preference for “lottery tickets” and an aversion to leverage; in both cases, the biases cause investors to over-allocate to higher volatility assets and drive down expected returns.
Carry and low-volatility, on their own, have some problems.
For example, the carry factor says we should invest in things that offer more return if we assume prices do not change. Often, however, yields are structurally higher because risk is higher. High yield bonds offer a higher yield than U.S. Treasuries because the companies issuing them are less credit-worthy: they have a higher risk of defaulting on the debt. Hence, investors require a higher return.
Going long high-yielding assets and short low-yielding assets, then, is basically the same as financing the purchase of high risk assets by short-selling low-risk assets. If we assume volatility and returns are negatively correlated, this is a position that makes us “short volatility.”
Which means we would be doing the exact opposite of what the low-volatility factor tells us to do, which is short-sell high risk assets to finance the purchase of low-risk assets, and hence be long volatility.
Instead, we believe investors should look at risk-adjusted carry, which can be measured as simply as “yield divided by volatility.”
In essence, risk-adjusted carry is similar to a “signal-blend” approach to factor investing, where we combine carry and low-volatility signals together to create a single multi-factor signal to tell us how attractive an asset is. It seeks to take advantage of the two factors simultaneously. We believe the benefit is that it allows us to focus our allocations on assets exhibiting abnormally high yields for their risk level.
To capture this factor, we rank our assets by their risk-adjusted carry scores (i.e. “yield divided by volatility”).
We will short bottom 50% and buy the top 50%. In the long leg of the trade, we weight assets in proportion to their risk-adjusted carry value, so those assets that have the highest score get the highest weight.
In the short leg of the trade, we weight assets in inverse proportion to their risk-adjusted carry value, so we are most short those assets with the worst scores.
Normally in factor research we would simply hold these two legs equally so that we are dollar-neutral. However, using volatility introduces a problem.
Namely, while low-volatility assets have historically outperformed high-volatility assets on a risk-adjusted basis, they have not necessarily done so on a total return basis. Investors cannot eat risk-adjusted returns, so without the mild application of leverage, a low-volatility strategy can still result in a lower returning strategy.
To account for this effect, we weight our exposure to these two legs based on their trailing realized variance. In doing so, we try to maintain an equal risk exposure to each leg.
We can see the hypothetical returns of this factor over time below.
Data Source: Commodity Systems Inc. Calculations by Newfound Research. All returns are hypothetical and backtested. Returns reflect the reinvestment of all distributions. Returns are gross of all fees (including any management fees and transaction costs) except for underlying ETF expense ratios. The portfolio is constructed using the following ETFs: AMJ, BKLN, BWX, CWB, EMLC, HYG, IDV, LQD, PBP, PCY, PFF, REM, TLT, VNQ, VNQI, and VYM. The hypothetical index starts on 6/30/2008. The start date is chosen based on the first month-end date that that volatility and yield information was available for at least 8 ETFs in the investable universe. The end date is 3/31/2017, the month-end date of most recent available data at the point of construction. The portfolio is reconstituted annually (using weekly overlapping portfolios). At each reconstitution, assets are ranked by their trailing 12-month volatility-adjusted distribution yield. Assets in the top 50% of ranks are held by the “long” leg; assets in the bottom 50% of ranks are held by the ”short” leg. Assets in the long leg are weighted in proportion to their volatility-adjusted distribution yield; assets in the short leg are weighted by inverse-proportion to their volatility-adjusted distribution yield. The legs are weighted in proportion to their trailing variance.
In the context of a long only portfolio, to capture this factor we would seek to overweight those assets in the long leg of the trade and underweight those assets in the short leg.
The Trend Factor
Trend – or “time-series momentum” – is the phenomenon where the price movement of assets persists in a single direction.
This anomalous behavior is most often explained through the theories of behavioral finance. Investors appear to exhibit a number of biases that lead to the emergence of trends. Anchoring bias causes them to discount new information in favor of old information, slowing the incorporation of new data into prices. The disposition effect is a tendency for investors to sell winners but hold on to losers, creating downward pressure and extending positive trends. Finally, investors tend to herd, causing prices to extend beyond fundamental value.
Systematic investors aim to exploit these biases by investing in assets with positive trends and avoiding assets with negative trends.
Specifically, in our case, we can build a factor that shorts assets exhibiting negative trends and buys assets with positive trends. The trend factor is a bit unique in that it does not necessarily always hold longs and shorts. If all assets have positive trends, it can be 100% long. Similarly, if all assets have negative trends, it can be 100% short.
We have to be somewhat careful in our construction of this factor, as not all assets are made equal. The contribution to overall portfolio risk of dividend stocks is very different than for corporate bonds. This means that when we apply our trend signals, the contribution that dividend stocks will have on overall factor risk will be much higher.
To account for this, we can employ a naïve risk-parity strategy, where assets are held in inverse proportion to their volatility. Higher volatility assets will receive lower allocations.
We can see the results of such an approach below.
Data Source: Commodity Systems Inc. Calculations by Newfound Research. All returns are hypothetical and backtested. Returns reflect the reinvestment of all distributions. Returns are gross of all fees (including any management fees and transaction costs) except for underlying ETF expense ratios. The portfolio is constructed using the following ETFs: AMJ, BKLN, BWX, CWB, EMLC, HYG, IDV, LQD, PBP, PCY, PFF, REM, TLT, VNQ, VNQI, and VYM. The hypothetical index starts on 5/30/2008. The start date is chosen based on the first month-end date that that volatility and trend information was available for at least 8 ETFs in the investable universe. The end date is 3/31/2017, the date of most recent month-end available data at the point of construction. The portfolio is reconstituted monthly (using weekly overlapping portfolios). At each reconstitution, assets are given a notional weight in inverse proportion to their trailing exponentially weighted 252-day volatility. If an asset is above its 200-day moving average, the notional weight is held long; if an asset is below its 200-day moving average, it is held short.
Evidence suggests that trend-following returns are the largest when stock market returns are the most extreme. We can see this in the above graph in 2008. While the returns ultimately eroded, positive returns in this sort of environment can potentially help serve as a hedge against extreme events. This attractive feature has been given the title of “crisis risk offset” or “crisis alpha.”
In a long-only portfolio, we can implement this factor by being long those assets exhibiting positive trends, but selling out of those assets exhibiting negative trends and either reallocating to positively trending assets or holding cash.
While traditional diversification leverages asset classes, factors can also introduce diversification benefits into a portfolio. This is because the returns in factors come from the differences in returns between the long and the short legs. What causes those differences can be entirely independent from what the market is doing at large.
Not only can factor approaches potentially overlay low-correlation return streams onto a portfolio, but they can also introduce strategies that have historically performed well in environments when traditional asset class diversification has broken down. For example, it is easy to see how a trend factor, if overlaid on top of a traditional portfolio, could introduce offsetting returns in periods like 2008.
To illustrate this potential diversification benefit, we can look at the correlation between these factors and traditional asset exposures. Specifically, we will look at the MSCI ACWI (“ACWI”), the Bloomberg Barclay’s U.S. Aggregate (“AGG”), and an equal-weight portfolio of the assets in our investable universe for Multi-Asset Income (“Equal-Weight”).
We create two correlation tables: one that covers all periods and one that focuses specifically on environments where global equities have historically done poorly (i.e. monthly returns less than -2.5%).
We can see that in environments where equities do poorly – and traditional diversification has historically failed – the risk-adjusted carry factor maintained its correlation profile and the trend factor became more negatively correlated, potentially helping offset equity losses.
However, these tables include the rather anomalous period of 2008-2009. We think it is prudent to look at the post 2009 period in isolation to get a better sense of correlations in “normal” markets.
We can see that while the correlation for the trend factor goes up over the period (likely due to the fact it was net long due to all the positive trends), both factors retain a low correlation profile.
Furthermore, we can see the usual “flight to quality” emerge in the equity and bond relationship, with AGG becoming much more negatively correlated to ACWI in months where ACWI’s returns were less than -2.5%. Interestingly, we can see this also emerge in the equal-weight universe and the risk-adjusted carry factor, which both in-turn increase their correlation to fixed income.
So while we might not expect the trend factor to offer the same diversification benefits during normal bull-market pull-backs as it historically did during significant bear markets, we can potentially see some of the low-volatility benefits emerging in the risk-adjusted carry factor.
When speaking to clients about our Multi-Asset Income portfolio, we do not describe the portfolio construction process as one that is factor-based. Rather, we normally describe the methodology in two stages.
- Remove any assets exhibiting negative trends.
- Tilt the remaining assets to overweight those offering more yield per unit of volatility and underweight those offering less yield per unit of volatility.
While this description might illuminate the process, it does not explain the why. Why we perform this process is to align our portfolio as closely as possible with the factors we’ve explored herein: factors we believe have the potential to generate excess risk-adjusted returns and introduce beneficial diversification opportunities into traditionally built portfolios.
You can find more information out about our Multi-Asset Income portfolio – including a form to request more information – at https://www.thinknewfound.com/portfolio/multi-asset-income.
 For example, earnings yield (the inverse of the P/E ratio) will be proportional to dividend yield if we assume a constant payout ratio.
 This is common practice in managed futures, though they often do it in a hierarchical risk-parity fashion and then scale positions up or down to meet an overall volatility target.