Value and Momentum Factor Portfolio Construction: Combine, Intersect, or Sequence?

//, Guest Posts, Value Investing Research, Momentum Investing Research/Value and Momentum Factor Portfolio Construction: Combine, Intersect, or Sequence?

Value and Momentum Factor Portfolio Construction: Combine, Intersect, or Sequence?

 Wes asked that I contribute to the ongoing debates regarding the construction of value and momentum portfolios.

There are three key research pieces on the topic, all with different viewpoints:

I encourage everyone to dig into the three articles above and then tackle my article below. And if you are interested in learning more about factor investing, you can sign up for our free newsletter here.

Hope you learn something…

Introduction

Being a Value investor is difficult, given that the stocks of a Value portfolios tend to have inherent issues, otherwise, these stocks would not be cheap. Being a B/M-focused value investor over the past decade has been especially hard, as the factor returns were effectively zero – plenty of pain, but no gain. Experiencing this factor cyclicality often leads investors to contemplate adding other factors in the hope of improving performance. An obvious candidate would be Momentum, as cheap and rising stocks are more appealing than cheap stocks. However, it is not quite straightforward for investors to add Momentum to a Value portfolio as there are several options available. In this short research note, we will analyze Value & Momentum portfolios created by three common multi-factor model approaches – the combination, the intersectional and the sequential models.

Methodology

We focus on the Value and Momentum factors in the US stock market. The factors are created by constructing long-short beta-neutral portfolios of the top and bottom stocks ranked by the factors. Portfolios rebalance monthly and include 10bps of transaction costs. Only companies with a market capitalization of larger than $1 billion are included.

The following three multi-factor models are utilized:

  • Combination model: Single factor portfolios are created and combined into one portfolio
  • Intersectional model: Stocks are ranked by several factors simultaneously
  • Sequential model: Stocks are sorted by factors sequentially

The portfolios are created so that they approximately contain the same number of stocks, which equates to the top and bottom 10% of the universe of 1,800 stocks. For a detailed report on these three models please see our white paper Multi-factor Models 101.

VALUE VERSUS MOMENTUM (LONG / SHORT)

The chart below shows the performance of the Value and Momentum factors in the US. We can observe that there were periods, e.g. during the Global Financial Crisis in 2007 to 2009, where the two factors exhibited significant negative correlation. The Momentum crash of 2009 is also clearly visible, which occurred when markets started recovering and the short side of Momentum outperformed the long side significantly. In theory, it should be attractive to add Momentum to a Value portfolio from a diversification perspective given the low or negative correlation.

Source: FactorResearch. The results are hypothetical results and are NOT an indicator of future results and do NOT represent returns that any investor actually attained. Indexes are unmanaged, do not reflect management or trading fees, and one cannot invest directly in an index.

VALUE & MOMENTUM FACTOR PORTFOLIOS: COMBINATION VS INTERSECTIONAL MODEL

The chart below shows the performance of two Value & Momentum factor portfolios, one created via the combination and the other via the intersectional model. We can observe that the trends are quite similar, but that the intersectional portfolio shows far higher returns. The combination model creates Value and Momentum portfolios separately and then combines these two portfolios, which likely contains conflicting positions as Value and Momentum are often negatively correlated, e.g. a stock in the long portfolio of Value might be in the short portfolio of Momentum. The intersectional model selects the stocks at the intersection of the two factors.

Source: FactorResearch. The results are hypothetical results and are NOT an indicator of future results and do NOT represent returns that any investor actually attained. Indexes are unmanaged, do not reflect management or trading fees, and one cannot invest directly in an index.

VALUE & MOMENTUM FACTOR PORTFOLIOS: SEQUENTIAL MODEL

The sequential model ranks stocks for factors sequentially, which requires investors to prioritize factors. This approach is unique given that after each ranking the universe of stocks shrinks significantly, resulting in a very concentrated portfolio. The chart below shows the performance of two portfolios, one ranked first on Value and then Momentum and the other one in the inverse order. We can observe that the profiles look almost identical, which is somewhat surprising. We do see that the portfolio ranked first on Momentum experienced a steeper decline during the Momentum crash, but the diversification benefits seem to be much lower compared to the combination or intersectional model.

Source: FactorResearch. The results are hypothetical results and are NOT an indicator of future results and do NOT represent returns that any investor actually attained. Indexes are unmanaged, do not reflect management or trading fees, and one cannot invest directly in an index.

VALUE & MOMENTUM PORTFOLIOS: MODEL COMPARISON

The chart below contrast all models and we can see that the results are overall comparable in terms of trend. Based on this analysis the intersectional model seems to be most attractive, but it somewhat depends on the observation period. If there are no strong increases or decreases in factor performance, then all models are comparable in terms of performance. However, when factor performance is particularly strong or very negative, e.g. for Momentum in 2009, then intersectional model seems to do better, which may be explained by less extreme portfolios.

Source: FactorResearch. The results are hypothetical results and are NOT an indicator of future results and do NOT represent returns that any investor actually attained. Indexes are unmanaged, do not reflect management or trading fees, and one cannot invest directly in an index.

RISK METRICS

In addition to displaying the performance, we can also analyze the risk metrics of the different portfolios. The graphic below highlights that the intersectional portfolio generated the highest risk-return ratio for the period from 2000 to 2017 and is the only portfolio that achieves a higher ratio than the Value factor on a stand-alone basis. It is worth highlighting that Momentum crashes a very rare (1932 & 2009) and if we would have picked a different observation period, then adding Momentum might have led to more attractive multi-factor portfolios.

Source: FactorResearch. The results are hypothetical results and are NOT an indicator of future results and do NOT represent returns that any investor actually attained. Indexes are unmanaged, do not reflect management or trading fees, and one cannot invest directly in an index.

We can also analyze the maximum drawdowns, which highlight diverse results for the multi-factor portfolios. The drawdowns of the combination and intersectional portfolios are lower than that of the stand-alone Value portfolio, while this is not the case for the sequential portfolios.

Source: FactorResearch. The results are hypothetical results and are NOT an indicator of future results and do NOT represent returns that any investor actually attained. Indexes are unmanaged, do not reflect management or trading fees, and one cannot invest directly in an index.

FURTHER THOUGHTS

This short research note highlights different ways of combining Value and Momentum factors. In hindsight adding other factors, e.g. Low Volatility, should have been more profitable as Momentum in the US did not generate attractive returns during the last two decades. However, the long-term track record for these two factors is quite favorable and adding factors that have done better, i.e. performance chasing, is unlikely a recipe for success.


  • The views and opinions expressed herein are those of the author and do not necessarily reflect the views of Alpha Architect, its affiliates or its employees. Our full disclosures are available here. Definitions of common statistics used in our analysis are available here (towards the bottom).
  • Join thousands of other readers and subscribe to our blog.
  • This site provides NO information on our value ETFs or our momentum ETFs. Please refer to this site.

Print Friendly, PDF & Email

About the Author:

Nicolas Rabener
Nicolas Rabener is the Managing Director of FactorResearch, which provides quantitative solutions for factor investing. Previously he founded Jackdaw Capital, an award-winning quantitative investment manager focused on equity market neutral strategies. Before that Nicolas worked at GIC (Government of Singapore Investment Corporation) in London focused on real estate investments across the capital structure. He started his career working in investment banking at Citigroup in London and New York. Nicolas holds a Master of Finance from HHL Leipzig Graduate School of Management, is a CAIA charter holder, and enjoys endurance sports (100km Ultramarathon, Mont Blanc, Mount Kilimanjaro).

14 Comments

  1. Mike Coleman
    Mike Coleman January 19, 2018 at 5:20 pm

    If I’m reading the two articles correctly, the Combination method in this article is analogous to the “50/50” method in the AA article linked at the beginning of this article, and the Intersection method here is analogous to the “Combo” method in the other article.

    Is that correct? If so, it seems the results of this analysis may be at least somewhat contradictory to those in the referenced article.

    • Wes Gray January 19, 2018 at 5:39 pm

      Hey Mike, you got it. Here is an outline from Nicolas: https://www.factorresearch.com/research-multi-factor-models-101
      We wanted to post the article because it was somewhat contradictory with what we show in our analysis.
      Why?
      Well, this actually highlights that the issue is not an open/shut case for one methodology or the other. There is plenty of room for argument.
      In this specific case, the differences are driven by methodology and time samples. We use ebit/tev, Nicolas uses a mix of b/p and e/p and his sample period is much shorter.
      Bottom line: we like the 50/50 method because we run concentrated and feel you get the most diversification bang for your buck. But slight changes in methodology can change the conclusions…

  2. Matt Tracey
    Matt Tracey January 20, 2018 at 9:45 am

    Interesting post and indeed a topic that doesn’t seem settled. One comment and one question. The comment: seems the “combination” approach could make sense for other reasons, e.g., ability to volatility-scale the factor mix (likely results in slightly higher value weight) and the reality that the QVAL strategy’s “quality” adjustment may (correct me if I’m wrong) have some modest correlation to momentum. (I use 60/40 mix myself.)

    The question/inquiry: would be curious to see whether the analysis in this post would yield the same (directional) result if long-only factors were used instead of long-short.

  3. Edwin Madison
    Edwin Madison January 20, 2018 at 6:05 pm

    Thank you for the very interesting and timely white paper! I, like Matt, also wonder about the analysis for long only positions. I also wonder whether you have done a similar analysis comparing absolute and risk-adjusted returns for 50% small cap + 50% value (combination) versus 100% small value (intersectional). Finally, have you examined the effect of adding trend following to combination or intersectional factor portfolios?

    • Nicolas Rabener
      Nicolas Rabener January 22, 2018 at 11:17 am

      Hi Edwin, yes, we also analysed combining the Value & Size factors in the US by utilising the combination and intersectional models. The results are similar to the post above, i.e. the performance and risk-return ratios are higher and drawdowns are lower for the intersectional portfolio.

      We are big believers in risk management and overlaying a trend following system would be one way of doing so. However, such an approach can be applied to almost any strategy or investible product, e.g. on the S&P 500, and the multi-factor model selection should not play a large role on this. Having said this, there is the topic of factor momentum, ie rotate amongst factors depending on their performance, where some research has been published. I recall that this research is ambiguous if factor momentum works, so we will likely publish our own analysis on this.

  4. Nicolas Rabener
    Nicolas Rabener January 21, 2018 at 12:17 am

    Hey Matt, thanks for your comment and question. I agree that the combination model has the advantage of treating factors like building blocks, which makes this model especially attractive from a tactical asset allocation perspective. However, you could also apply a volatility-weighted allocation to the intersectional model, which in the post above assumed an equal-weight allocation. At the rebalancing day, e.g. at month end, the stocks and weights for either model could be adjusted. On a side note, we did some research on equal-weight vs volatility-weighted allocation, which might be of interest:

    https://www.factorresearch.com/research-factor-allocation-101-equal-vs-volatility-weighted

    Regarding your question: We don’t have a readily available answer for this as moving from long-short to long-only has quite a few nuances, which is one of our research areas. We will publish on this in the near future, so please stay tuned.

  5. Matt Tracey
    Matt Tracey January 21, 2018 at 12:03 pm

    Thanks for the reply; all makes sense. I look forward to the long-only publication, as that will be most relevant for me (and I presume others as well). Good stuff. Appreciate it.

  6. Edwin Madison
    Edwin Madison January 22, 2018 at 7:36 pm

    Thanks very much for your response. I haven’t read anything about “factor momentum” and an associated factor rotation strategy – can you suggest a few articles? It sounds related to the Arnott/Asness debate regarding “factor timing” although I believe that debate focuses on timing factors based on their “valuation” rather than on their “momentum”. Regarding that debate, what is your view on whether buying an individual stock when its valuation is low (value investing) is different than buying a factor etf (or market or sector index) when its valuation is low (market timing)? I believe in the former (value investing) but not the later (market timing) but often wonder if that is a completely consistent stance.

    • Wes Gray January 24, 2018 at 8:26 pm

      Hey Edwin,
      The only timing mechanism we believe has legs related to trend-following mechanisms. We’ve done a lot of research over the years and there doesn’t seem to be much compelling work to suggest that market timing via other mechanisms (ie valuations) if effective. In short, we agree with Asness. If you search in our archive you will find many pieces on the topic. Here is a TF piece: https://alphaarchitect.com/2015/08/13/avoiding-the-big-drawdown-with-trend-following-investment-strategies/. Here is archive:
      https://alphaarchitect.com/research-category-list/
      Nicolas may have a different opinion.

    • Nicolas Rabener
      Nicolas Rabener January 25, 2018 at 8:26 am

      Hey Edwin, we are not aware of much research on factor momentum in the public domain. Corey from Newfound wrote a post about this a while ago, which may be of interest:

      https://blog.thinknewfound.com/2016/12/factor-rotation-possible-worth/

      We agree with Wes that factor momentum on it’s own is not very attractive, however, in combination with other variables it becomes more interesting. Naturally this depends on the investors objective, which we tend to differentiate between factor timing (difficult) and factor risk management (achievable). We’re going to publish a white paper on this topic on our website on Monday.

      Regarding your other question: I agree that consistency is important and that if you consider buying cheap stocks value investing, then buying cheap sectors, factors, indices, etc should also be called value investing. Naturally this can be differentiated by investing on a relative basis, eg a stock vs its peer group, versus investing based on its own trading history, eg S&P today versus S&P historically.

  7. Edwin Madison
    Edwin Madison January 26, 2018 at 3:01 am

    Thank you very much for the response and links, Wes and Nicolas. I also like trend following approaches, particularly your VMOT ETF. Thank you for creating that product! With respect to risk-adjusted returns, how would you expect the VMOT long only with trend following to compare with a long-short factor strategy with and without trend following or a managed futures strategy? (I believe you have products that use all of these strategies – correct?)

    I look forward to reading the new white paper!

    • Wes Gray January 29, 2018 at 7:56 am

      Hi Ed,
      We can’t discuss the VMOT ETF here, only research and index-related questions. Sorry about that — regulatory/compliance requirements. You can reach out to our firm if you have specific questions related to ETF securities.

  8. Nicolas Rabener
    Nicolas Rabener January 29, 2018 at 8:03 am

    Ed, as mentioned we published a white paper on factor allocation models today, which might be of interest:

    https://www.factorresearch.com/research-factor-allocation-models

  9. Edwin Madison
    Edwin Madison January 30, 2018 at 2:31 am

    Thanks, Wes and Nicolas! I look forward to reading the new white paper, I will reach out to your firm with questions primarily about matching specific investment strategies to specific investor situations.

Leave A Comment