Be sure to check out Part 1 and Part 2 to familiarize yourself with the Five Factors
Key Takeaways:
- Our Factor Investing Model will be based on 5 common factors (value, quality, low volatility, momentum and size). During 3Q 2021, momentum did well while the other factors were flat to down. Over the past year, the size premium (small-cap stocks over large-cap stocks) led the way.
- In this report, we will be covering how to construct a baseline factor portfolio based on the five factors we have introduced in Part 1 and Part 2 of this series.
- We introduced the value and quality factors in the first part of this three-part series. In Part II we introduced low volatility, momentum, and size.
- Static factor portfolios have struggled over the past 18 months, and have not contributed consistent positive return since about 2013.
- A simple factor timing overlay, however, has performed better, particularly during the recent downturns in the value and low-volatility factors.
Multi-Factor Portfolios
Constructing a Baseline Factor Portfolio
One of the benefits of factor investing is that the factors themselves are not perfectly correlated. From Fig. 1, we see that some pairs of factors exhibit high correlations (particularly low-volatility and quality), while value and momentum tend to be negatively correlated. Size, on the other hand, shows relatively low correlations to the other factors. Also, the factors tend to outperform at different times – for example, the size factor recently saw strong returns over the past twelve months, while low-volatility struggled.
Fig. 1 – Correlation of Factor Premia
We can construct a multi-factor portfolio to take advantage of the factors not being perfectly correlated and tending to outperform at different times. We construct a simple portfolio consisting of the five factors, with the weights assigned proportional to the inverse of each factor’s trailing volatility.
This portfolio, which is rebalanced monthly, can be viewed as a proxy for static multi-factor investing, as when the portfolio does well, factors (on average) are outperforming their benchmarks. The performance of this static multi-factor portfolio is shown in Fig. 2.
What it means: When building a portfolio, we can reduce the overall risk of the portfolio through diversification – combining assets that move independently from one another. The diversification benefit is controlled by the correlation among the individual portfolio building blocks. The lower the correlation, the greater the diversification benefit provided. By investing in multiple factors, we take advantage of diversification and reduce the overall portfolio risk.
When constructing the static factor portfolio, we want to make sure that each factor is contributing an equal amount of risk. Recall from parts 1 and 2 that the factors have different levels of risk – the size factor undergoes much larger swings in performance than does the quality factor.
In building the multi-factor portfolio, we weight the factors “proportional to the inverse of the trailing volatility.” This just means that we allocate more weight to lower-risk factors, like quality, than to higher-risk factors, like size. The result is that each factor contributes the same amount of risk to the overall portfolio.
The factors we highlight as part of the multi-factor portfolio have been studied in academia for decades, and have been shown to contribute excess return over the long run. A portfolio of multiple factors aims to capture multiple sources of excess return. Investing in a portfolio of sectors, on the other hand, will not contribute excess return over the long run, as by definition, the market itself is a multi-sector portfolio. While timing sector exposures can generate alpha (as is done in the FSI Sector Allocation product) we would not expect a static multi-sector portfolio to contribute excess return over the long run.
Fig. 2 – 5-Factor Portfolio Performance
The performance of the 5-factor portfolio indicates that factor investing, in general, did well from 2001 through the middle of 2013, outperforming the benchmark S&P 500 by about 2.5% per year. Since the middle of 2013, however, the factor investing portfolio has trended lower, with significant underperformance seen since the start of 2020. For whatever reason (increased attention and/or crowding into traditional factors, continued unconventional monetary policy by central banks, etc.) the static multi-factor portfolio has generally fared poorly for the better part of the past decade.
These factors can continue to be used to gauge risks inherent in equity portfolios (i.e. as risk factors), but a static strategy that invests in these factors will likely produce little to no alpha going forward. Factor investors must adapt. We propose several approaches that factor investors can utilize:
- Redefining traditional factors, for example, by developing new ways to define value or quality.
- Using data outside of the traditional company sources (so-called alternative data) to construct new factors.
- Factor timing, which seeks to overweight factors as they are likely to outperform, and underweight (or short) factors when they are primed to underperform.
- Utilizing new modeling techniques to combine the underlying factors in new ways.
A Dynamic Factor Rotation Strategy
To illustrate a simple example, we take the same 5-factor portfolio described above, using inverse-volatility weights, but we also employ a rotation overlay. We compute the trailing 3-month (volatility-adjusted) momentum of each of the 5 factors, and overweight the factor with the best momentum score by 10%, while underweighting the factor with the worst trailing momentum score by 10%. This strategy, which we call “factor rotation” can be seen as the light blue line in Fig. 3 below. For comparison, we also show the performance of the static multi-factor portfolio in dark blue.
What it means: The goal here is to show that, by tilting weights toward factors we think will work going forward, we can outperform the static multi-factor strategy described above.
To evaluate our new factor tilting strategy, we need to compare it to a benchmark – so we use the static multi-factor portfolio described above as a benchmark.
The approach we use to tilt weights is simple. Each month, we start with the benchmark static multi-factor portfolio. We add 10% weight to the factor that did the best over the past 3 months, while taking 10% weight away from the factor that did the worst. We call the resulting portfolio (after the 10% weight adjustment) the “dynamic multi-factor portfolio” (or “factor rotation” in Fig. 3 below).
Fig. 3 – Dynamic Rotation Strategy vs. Static Strategy for Factors
The dynamic strategy that rotates across factors shows slightly better performance than its static counterpart. While the dynamic strategy introduces slightly more turnover, the amount of additional turnover is minimal, as the strategy is rebalanced monthly, we employ a relatively minor overweight (underweight) of 10% to the favored (disfavored) factor, and we are measuring performance using highly liquid ETFs. The dynamic strategy showed considerable outperformance over the static strategy since the start of 2020, when value and low-volatility saw a significant erosion in performance.
Fig. 4 shows the cumulative outperformance of the factor rotation strategy vs. its static counterpart. While the dynamic strategy also increases volatility, its risk-adjusted return (0.38 for the full period) exceeds that of the static strategy (0.22 for the full period). This chart also shows that the dynamic strategy has produced consistent outperformance starting around the middle of 2016.
Fig. 4 – Dynamic vs. Static Factor Rotation Portfolio (Relative Return)
Applying a simple factor rotation overlay can improve a static factor portfolio. More complicated approaches, when applied properly, can further boost returns.
What it means: The results show that by applying a simple tilting approach, we can improve performance over a static multi-factor portfolio. The dynamic multi-factor portfolio, which applies the 10% over/underweight tilt, has consistently outperformed the benchmark static multi-factor portfolio, averaging 70 basis points of outperformance per year.
The approach to factor tilting we use here is simple and conservative, yet still produces excess return. In the future, we will introduce new factor tilting approaches to further boost returns.
We also will continue to monitor and regularly update the performance of both the dynamic and static multi-factor portfolios, and highlight ETFs that can be used to implement these portfolios.
