How to Combine Value and Momentum Investing Strategies

/How to Combine Value and Momentum Investing Strategies

How to Combine Value and Momentum Investing Strategies

How to Combine Value and Momentum Investing Strategies

We are probably most well known for our quantitative value investing strategies. The heart of our strategy is detailed in Wes’ book, Quantitative Value (a reader’s digest version is here). In the development of the Quantitative Value system there is no mention of the concept of “momentum investing,” which is a well-established empirical anomaly in the academic finance literature and a topic we cover extensively on the blog. And don’t get us wrong, even though we are value-investing fans, we are also huge fans of momentum investing. We have a book on that subject as well (Quantitative Momentum)

But that begs a question that we have heard many times over via our readership:

Why don’t you include momentum in your value investing process?

This is a great question and one we will answer publicly for the first time.

Many are familiar with the evidence that value and momentum investment strategies have beaten the market, historically. And the low historical correlation between value and momentum suggests there is a benefit to combining these portfolios. So why don’t we include momentum in our value investing strategy?

Well, there are a few ways to skin the value and momentum cat:

  1. One solution is to combine the exposures as seperate portfolios: part pure value; part pure momentum.
  2. Another solution is to “blend” the exposures into a single strategy: an integrated value and momentum system that weighs value and momentum factors and then holds firms with the highest combination.

As evidence-based investors, we decided early on to go with option #1. We aren’t suggesting that option #2 is a bad option, in fact, option #2 is a great solution relative to most security selection approaches peddled in the marketplace. We simply prefer option #1 because it is relatively better than option #2.

Why are pure value and momentum exposures better?

The evidence suggests that we keep highly active exposures to value and momentum in their purest forms (assuming we are doing high-conviction non-watered down versions of the anomalies). Blending the strategy dilutes the benefit of value and momentum portfolios. The summary of the benefits of a pure value and a pure momentum approach can be summarized as follows:

  • Easier ex-post assessment
    • E.g., if we mix and match value/momentum it is more difficult to identify the drivers of performance after the fact.
  • Stronger portfolio diversification benefits.
    • Pure value and pure momentum strategies have lower correlations than “blended” versions.
  • Stronger expected performance.
    • Running pure value and pure momentum in highly active forms generates higher expected performance than blended systems.

The Set-up

First, let’s set up the experiment. We will examine all firms above the NYSE 40th percentile for market-cap (currently around $1.8 billion) to avoid weird empirical effects associated with micro/small cap stocks. We will form the portfolios at a monthly frequency with the following 2 variables:

  1. Momentum = Total return over the past twelve months (ignoring the last month)
  2. Value = EBIT/(Total Enterprise Value)

We form the simple Value and Momentum portfolios as follows:

  1. EBIT VW = Highest decile of firms ranked on Value (EBIT/TEV). Portfolio is value-weighted.
  2. MOM VW = Highest decile of firms ranked on Momentum. Portfolio is value-weighted.
  3. Universe VW = Value-weight returns to the universe of firms.
  4. SP500 = S&P 500 Total return

Results are gross of management fees and transaction costs. All returns are total returns and include the reinvestment of distributions (e.g., dividends).

Here are the returns (7/1/1963-12/31/2013):

How to Combine Value and Momentum Investing Strategies_backtest 1

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. Additional information regarding the construction of these results is available upon request.

Takeaways:

  1. The universe of stocks is similar to the SP500.
  2. The top decile of Value and Momentum outperformed the index over the past 50 years.
  3. There is a rather low correlation of 0.5301 between Value and Momentum.
  4. Momentum has stronger returns than value, but much higher volatility and drawdowns. On a risk-adjusted basis they are similar.

The Test: Blended Value and Momentum vs. Pure Value and Momentum

The low correlation between value and momentum suggests there is a benefit to combining these historically high-performing portfolios. There are a few ways in which an investor can attempt to exploit these anomalies:

  1. One solution is to combine them as seperate portfolios: part pure value; part pure momentum.
  2. Another solution is to “blend” the exposures into a single strategy: an integrated value and momentum system that weighs value and momentum factors and then holds firms with the highest combination.

To help us identify the best approach we set up a small experiment.

We form the following four portfolios:

  1. EBIT VW = Highest decile of firms ranked on Value (EBIT/TEV). Portfolio is value-weighted.
  2. MOM VW = Highest decile of firms ranked on Momentum. Portfolio is value-weighted.
  3. COMBO VW = Rank firms independently on both Value and Momentum.  Add the two rankings together. Select the highest decile of firms ranked on the combined rankings. Portfolio is value-weighted.
  4. 50% EBIT/ 50% MOM VW = Each month, invest 50% in the EBIT VW portfolio, and 50% in the MOM VW portfolio. Portfolio is value-weighted.

Results are gross of management fees and transaction costs. All returns are total returns and include the reinvestment of distributions (e.g., dividends).

Here are the returns (7/1/1963-12/31/2013):

How to Combine Value and Momentum Investing Strategies_backtest 2

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. Additional information regarding the construction of these results is available upon request.

Takeaways:

  • The combination portfolio performs worse than a 50% allocation to Value and a 50% allocation to Momentum. This statement is driven by an analysis of CAGR, Sharpe and Sortino ratios.
  • The combined ranked portfolio outperforms the index over the same time period–not a “bad” portfolio construct by any means.
  • The 50% pure value and 50% pure momentum portfolio has the highest risk-adjusted returns across all portfolios.
  • In addition, we find similar results when equal weighting portfolios.

Summary

Overall, the evidence suggests that a blended strategy, which combines Value and Momentum into a single unified process, is worse than allocating 50% of your capital to a pure value investing fund, and 50% to a pure momentum investing fund. This may have implications for how investors allocate to value and momentum anomalies. Of course, one must consider that we have only analyzed simple value and simple momentum strategies. Perhaps there are more sophisticated techniques to make “blended” val/mom better than allocations to pure value and momentum. Some evidence from other authors finds conflicting evidence. We’ve done our own extensive testing reconciling the various findings, and we think the analysis above highlights in a SIMPLE way that combo portfolios are relatively better than blended portfolios. That said, we are open to additional input and testing from the broader research community.

How to Combine Value and Momentum Investing Strategies (Part 2/2)

In the prior section I wrote about ways to combine value investing and momentum investing. The high level takeaway from that article was to keep value and momentum as separate exposures. This conclusion was based on ranking firms on their combined value and momentum rankings, which can be described as follows:

  1. Rank all stocks on value
  2. Rank all stocks on momentum
  3. Average the ranks

Here, I want to examine how sequential rankings affect returns. Sequential ranking can be described as follows:

  1. Rank all stocks on value
  2. Within value, rank on momentum
  3. Buy cheap stocks with the highest momentum
  4. Repeat steps 1-3 but start with momentum and then value (so you end up buying the highest momentum stocks that are the cheapest)

We find that sorting stocks on value and then momentum has been a historically successful strategy. However, sorting stocks on value and then quality, which is inline with a fundamental value philosophy, works just as well, if not better. We hypothesize that the “momentum” effect identified within the cheap stocks bucket, is really a proxy for strong fundamentals and positive operational momentum among the cheapest, highest quality value stocks. Our preference–based on empirical and philosophical grounds–is to go with a more “pure” value philosophy that focuses on buying the cheapest, highest quality value stocks, as opposed to a muddled value approach that buys the cheapest, highest momentum value stocks.

Value Investing Portfolio Set-up:

First, let’s set up the experiment. We will examine all non-financial firms above the NYSE 40th percentile for market-cap (currently around $1.8 billion) to avoid weird empirical effects associated with micro/small cap stocks. We will form the portfolios at a monthly frequency with the following 2 variables:

  1. Momentum = Total return over the past twelve months (ignoring the last month)
  2. Value = EBIT/(Total Enterprise Value)

The simple value investing portfolios are formed monthly as follows:

  1. EBIT Decile EW MR = Highest decile of firms ranked on Value (EBIT/TEV). Portfolio is equal-weighted and rebalanced monthly.
  2. EBIT (19/20) EW MR = Split the top decile ranked on Value in two. Keep the the 90%-95% cheapest firms (EBIT/TEV). Portfolio is equal-weighted and rebalanced monthly.
  3. EBIT (20/20) EW MR = Split the top decile ranked on Value in two. Keep the the 95%-100% cheapest firms (EBIT/TEV). Portfolio is equal-weighted and rebalanced monthly.
  4. SP500 EW = S&P 500 equal-weight Total return

Results are gross of management fees and transaction costs. All returns are total returns and include the reinvestment of distributions (e.g., dividends). Transaction costs at a monthly rebalance level could be substantial, but we leave it readers to make a determination on these costs because investors face vastly different trading cost regimes.

Here are the returns (1/1/1974-12/31/2014):

Value Investing Portfolio Results:

value investing portfolio results

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. Additional information regarding the construction of these results is available upon request.

Takeaways:

  1. Value investing has outperformed over the past 41 years.
  2. Breaking down the top decile into 5% buckets does not improve returns. Deep value works, but over-concentration in cheap stocks simply adds risk, but not return.

Next we turn our attention to momentum investing portfolios.

Momentum Investing Portfolios Set-up:

The simple Momentum Investing portfolios are formed monthly as follows:

  1. MOM Decile EW MR = Highest decile of firms ranked on Momentum. Portfolio is equal-weighted and rebalanced monthly.
  2. MOM (19/20) EW MR = Split the top decile ranked on Momentum in two. Keep the the 90%-95% highest Momentum firms. Portfolio is equal-weighted and rebalanced monthly.
  3. MOM (20/20) EW MR = Split the top decile ranked on Momentum in two. Keep the the 95%-100% highest Momentum firms. Portfolio is equal-weighted and rebalanced monthly.
  4. SP500 EW = S&P 500 equal-weight Total return

Results are gross of management fees and transaction costs. All returns are total returns and include the reinvestment of distributions (e.g., dividends). Transaction costs at a monthly rebalance level could be substantial, but we leave it readers to make a determination on these costs because investors face vastly different trading cost regimes.

Here are the returns (1/1/1974-12/31/2014):

Momentum Investing Portfolio Results:

momentum investing portfolio results

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. Additional information regarding the construction of these results is available upon request.

Takeaways:

  1. Momentum investing has outperformed over the past 41 years.
  2. Buying the “highest” momentum stocks (top 5%) has marginal effects (higher CAGR, lower Sharpe ratios).

Combining Value and Momentum:

Here we want to combine Value and Momentum by sequentially sorting on the two variables.

Split the Momentum Decile by Value.

Specifically, here are the 4 portfolios we will examine:

  1. MOM Decile, high EBIT EW MR = Highest decile of firms ranked on Momentum, then split on Value (EBIT/TEV). We keep the top half when sorted on Value. Portfolio is equal-weighted and rebalanced monthly.
  2. MOM Decile, low EBIT EW MR = Highest decile of firms ranked on Momentum, then split on Value (EBIT/TEV). We keep the bottom half when sorted on Value. Portfolio is equal-weighted and rebalanced monthly.
  3. MOM (20/20) EW MR = Split the top decile ranked on Momentum in two. Keep the the 95%-100% highest Momentum firms. Portfolio is equal-weighted and rebalanced monthly.
  4. SP500 EW = S&P 500 equal-weight Total return

Results are gross of management fees and transaction costs. All returns are total returns and include the reinvestment of distributions (e.g., dividends).

Here are the returns (1/1/1974-12/31/2014):

Momentum (and then Value) Investing Portfolio Results:

momentum and then value

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. Additional information regarding the construction of these results is available upon request.

Takeaways:

  1. Taking the top decile on Momentum and splitting on Value (EBIT/TEV) improves Sharpe and Sortino ratios (Comparing Column 1 to Columns 2 and 3).
  2. On a CAGR basis, the best bet is simply buying the top 5% of firm on Momentum.

Overall, this sequential sort (Momentum and then Value) does not drastically improve simple Momentum Investing returns.

Split the Value Decile by Momentum.

Specifically, here are the 4 portfolios we will examine:

  1. EBIT Decile, high MOM EW MR = Highest decile of firms ranked on Value (EBIT/TEV), then split on Momentum. We keep the top half when sorted on Momentum. Portfolio is equal-weighted and rebalanced monthly.
  2. EBIT Decile, low MOM EW MR = Highest decile of firms ranked on Value (EBIT/TEV), then split on Momentum. We keep the bottom half when sorted on Momentum. Portfolio is equal-weighted and rebalanced monthly.
  3. EBIT (20/20) EW MR = Split the top decile ranked on Value (EBIT/TEV) in two. Keep the the 95%-100% cheapest firms. Portfolio is equal-weighted and rebalanced monthly.
  4. SP500 EW = S&P 500 equal-weight Total return

Results are gross of management fees and transaction costs. All returns are total returns and include the reinvestment of distributions (e.g., dividends).

Here are the returns (1/1/1974-12/31/2014):

Value (and then Momentum) Investing Portfolio Results:

value and then momentum

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. Additional information regarding the construction of these results is available upon request.

Takeaways:

  1. Taking the top decile of Value firms and splitting on Momentum improves returns in general. This improves the CAGR, Sharpe and Sortino ratios (Comparing Column 1 to Columns 2 and 3).
  2. Results are similar to other studies.

Overall, it appears that splitting Value by Momentum is a good strategy to follow!

However, such a strategy does require frequent rebalancing, as momentum works best if rebalanced more frequently. This will cause higher transaction costs relative to a long-term buy-and-hold Value strategy.

So an obvious question is the following — Why don’t we integrate momentum in to our Quantitative Value strategy?

The answer is simple: momentum doesn’t increase QV performance, in expectation.

Below, we explain why we do not incorporate momentum into our value algorithm.

Quantitative Value Index Universe Results:

Here we examine how splitting a Value portfolio by Momentum compares to the Quantitative Value results.

As explained here, the Quantitative Value process involves 5 steps. Step 4 requires us to sort firms based on their quality. However, what happens if we replace this step and sort firms by their past Momentum?

We answer this below. It is important to understand that the universe for the results above is slightly different than the results below, as the Quantitative Value process requires firms to have 8 years of data (which is not required above).

Specifically, here are the 4 (annually rebalanced) portfolios we will examine:

  1. QV EW = Portfolio formed using the Quantitative Value process. Portfolio is equal-weighted and rebalanced annually.
  2. QV (High MOM) EW = Portfolio formed using the first three steps of the Quantitative Value process. However, step 4 is changed to pick the top half of firms ranked on Momentum. Portfolio is equal-weighted and rebalanced annually.
  3. QV (Low MOM) EW = Portfolio formed using the first three steps of the Quantitative Value process. However, step 4 is changed to pick the bottom half of firms ranked on Momentum. portfolio is equal-weighted and rebalanced annually.
  4. SP500 EW = S&P 500 equal-weight Total return

Results are gross of management fees and transaction costs. All returns are total returns and include the reinvestment of distributions (e.g., dividends).

Here are the returns (1/1/1974-12/31/2014):

Quantitative Value Index Investing Portfolio Results (Annual Rebalance):

Quantitative Value annual rebalance

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. Additional information regarding the construction of these results is available upon request.

Takeaways:

  1. Similar to the results above, splitting the top 10% Value firms by Momentum is a good strategy (Comparing Column 2 to Column 3).
  2. However, splitting firms by quality and keeping the top half (QV EW) is optimal when comparing CAGRs, Sharpe and Sortino ratios.
  3. These portfolios are highly correlated.

However, since we know Momentum works best at a monthly rebalance frequency, let’s examine the same portfolios which are rebalanced monthly.

Specifically, here are the 4 (monthly rebalanced) portfolios we will examine:

  1. QV EW MR = Portfolio formed using the Quantitative Value process. Portfolio is equal-weighted and rebalanced monthly.
  2. QV (High MOM) EW MR = Portfolio formed using the first three steps of the Quantitative Value process. However, step 4 is changed to pick the top half of firms ranked on Momentum. Portfolio is equal-weighted and rebalanced monthly.
  3. QV (Low MOM) EW MR = Portfolio formed using the first three steps of the Quantitative Value process. However, step 4 is changed to pick the bottom half of firms ranked on Momentum. portfolio is equal-weighted and rebalanced monthly.
  4. SP500 EW = S&P 500 equal-weight Total return

Results are gross of management fees and transaction costs. All returns are total returns and include the reinvestment of distributions (e.g., dividends).

Here are the returns (1/1/1974-12/31/2014):

Quantitative Value Index Investing Portfolio Results (Monthly Rebalance):

Quantitative Value monthly rebalance

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. Additional information regarding the construction of these results is available upon request.

Takeaways:

  1. Rebalancing the Value portfolios monthly increases returns. This has been found in other studies. However, this should increase trading costs (which are not included here), and does increase drawdowns.
  2. Similar to the results above, splitting the top 10% Value firms by Momentum is a good strategy (Comparing Column 2 to Column 3).
  3. However, splitting firms by quality and keeping the top half (QV EW MR) is optimal when comparing CAGRs, Sharpe and Sortino ratios.

The last reason we prefer the QV strategy is shown in the table below:

QV and Mom correlations

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. Additional information regarding the construction of these results is available upon request.

 

As is shown in the table, the correlation between QV and a generic Momentum portfolio is 57.40%, while the cheap firms split by Momentum have a slightly higher correlation (63.44%). All else being equal, we would prefer to have a lower correlation to Momentum.

Summary

In this article we set out to decide how combining Value and Momentum screens work when using them sequentially. First off, we find that splitting a Momentum portfolio by a Value screen does not appear to add value. However, splitting a Value portfolio by Momentum does add value. So why do we not incorporate this into our Quantitative Value strategy? We show above that splitting Value firms by their Quality score (as outlined in the Quantitative Value book) achieves the highest CAGR and risk-adjusted returns (Sharpe and Sortino ratios). This strategy also has a lower correlation to a simple Momentum strategy. Why would this work? Perhaps investors overlook the quality component of the firm (the quality score may also pick up operational momentum).

One word of caution for all the above results. These are shown gross of fees and transaction costs to facilitate comparison across strategies. Clearly, transaction costs will matter on these highly active strategies (momentum in particular). One could even argue that after transaction costs and complexity these strategies no longer add value. Additionally, one must consider the consequences that taxes may have on their portfolio. Caveats aside, the results are still interesting and should be considered by professional investors focused on evidence-based investment strategies.


  • 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:

Jack Vogel
Jack Vogel, Ph.D., conducts research in empirical asset pricing and behavioral finance, and is a co-author of DIY FINANCIAL ADVISOR: A Simple Solution to Build and Protect Your Wealth. His dissertation investigates how behavioral biases affect the value anomaly. His academic background includes experience as an instructor and research assistant at Drexel University in both the Finance and Mathematics departments, as well as a Finance instructor at Villanova University. Dr. Vogel is currently a Managing Member of Alpha Architect, LLC, an SEC-Registered Investment Advisor, where he heads the research department and serves as the Chief Financial Officer. He has a PhD in Finance and a MS in Mathematics from Drexel University, and graduated summa cum laude with a BS in Mathematics and Education from The University of Scranton.

47 Comments

  1. RT1C March 26, 2015 at 4:57 pm

    Why not take apply the MOM screen only AFTER the Value screen, not in parallel? i.e., take the best value stocks, then buy those with good momentum? That is different than either combination approach you described above.
    In any case, Gary Antonacci has reported that there isn’t a benefit to combining value and momentum. http://www.dualmomentum.net/2014/09/value-and-momentum-revisited.html. His Dual Momentum strategy applies conventional momentum and then applies a trend-following overlay to transition between equities and treasuries. I’d be interested to see your replication of that work. (Incidentally, Antonacci speaks favorably of your QVAL, if one is pursuing a value strategy: http://www.dualmomentum.net/2014/10/value-investing-redux.html).

    • Steve March 26, 2015 at 10:59 pm

      Not speaking on behalf of; but the reason to test this is to see the effects of value and momentum. If you overlay value with momentum; that can work well (research shows) – but you’re still looking at value.

      The idea here is to look at value and momentum…and see what benefits (if any) can be gained by having both in a portolio; and how best to implement that. To do that, you need to treat value as value and momentum as momentum.

    • Wesley Gray, PhD
      Wesley Gray, PhD March 26, 2015 at 11:11 pm

      See Steve’s comment.

      You could try blends of value–>mom and mom–>value. Not a terrible idea. We’ve done that analysis as well and it seems to dilute the portfolio benefits of pure value vs. pure momentum.

      re Antonacci comments: We are clarifying a misunderstanding Gary and others have with respect to the difference between using cross-sectional momentum to select stocks and time-series momentum to time markets (“absolute momentum”). We have a post coming out fairly soon that highlights the clear benefits of applying cross-sectional and time-series momentum to a portfolio. A momentum-based stock portfolio with absolute momentum rules applied beats the daylight out of a S&P 500 portfolio with absolute momentum rules applied. He compares a momentum-based stock portfolio against a S&P 500 portfolio with market-timing–that is an apples to oranges comparison that is confusing the heck out of many people. Cross-section and time-series momentum are complements, not competitors!!!

      • Steve March 26, 2015 at 11:21 pm

        A momentum based stock portfolio with absolute momentum rules applied I guess would act as a bit of a natural market timer anyway…i.e. keeping you out of the market in times where the best momentum is negative anyway.

        I should wait to the post comes out, at least…. 🙂

      • James March 27, 2015 at 12:26 am

        A 50% mom and 50% value, such as above, with absolute momentum applied to both sounds like an interesting idea to me.

      • RT1C March 27, 2015 at 9:02 am

        Count me among the confused, then; this isn’t my understanding. Antonacci compares cross sectional momentum (absolute momentum) to time series momentum (trend following) and then combines them. He applies this to asset classes, not individual stocks, since trading costs are excessive when applying absolute momentum to individual stocks. Thus, his cross-sectional momentum is between asset classes, e.g., his GEM strategy chooses between domestic and international equities, while the trend-following component switches into treasuries if equity returns head south.

        A fair criticism, I think, of your blog above is that it doesn’t include trading costs, so it is hard to see the negative impact that has on the cross-sectional momentum component (I’m assuming the value component is less frequently traded and thus less affected). I hope in the upcoming analysis that you’ll be posting soon on cross-sectional and time-series momentum, which I am eagerly looking forward to, that you’ll take trading costs into account.

        I’d also be interested to hear your comments on why so many of the momentum funds out their (AQR etc.) have a poor performance record. Is it trading and management costs? For example, AMOMX has underperformed its benchmark Russell 1000 index.
        Additionally, I’d like very much to better understand the maximum drawdown situation. Antonacci’s GEM has very low drawdown, due to the trend-following (time-series momentum) feature. Is that fortuitous due to the time series he studied? (He has extended the time series on his blog and still shows relatively low drawdown). By comparison, the momentum component in your blog above (which is cross-sectional only, I believe?), has relatively large drawdown.
        I think there is an opportunity for a fund that effectively implements quality value and both time- and cross-sectional momentum strategies (or maybe this should be 2 funds, based on the above blog post!).

        • Jack Vogel
          Jack Vogel, PhD March 27, 2015 at 10:32 am

          Yes, Gary only looks at cross-sectional momentum within asset classes (good idea). We believe (based on the above evidence), that cross-sectional momentum works within the universe of stocks as well.

          Trading costs are not included, and these DO matter. However, even at 3-4% a year for trading costs, value and momentum would have outperformed over a 50 year period.

          As for the other fund (AMOMX), not sure why they under-performed. One explanation could be the number of holdings.

          Our upcoming post on time-series and cross-sectional momentum will address the drawdowns. Using a time-series overlay on a cross-sectional momentum strategy helps limit drawdowns.

          • Styv March 27, 2015 at 2:37 pm

            Drawdowns: Antonacci uses only monthly drawdown figures. The GEM DD in his book is -22.72.

            Rolling win %: the <50% win rate for the 50/50 strategy in 5 and 10 year periods may prove too daunting, particularly for those who need to withdraw funds on a regular basis or don't have a 50-year investment horizon. Am I misunderstanding the rolling win %?

          • Jack Vogel
            Jack Vogel, PhD March 27, 2015 at 4:29 pm

            Our drawdowns capture peak to bottom from any point in time, not just at a monthly frequency.

            Rolling win % compares column 1 against columns 2, 3 and 4. Now I see how this can be confusing! So in column 4, examining the 10-year rolling CAGR win %, this means that the pure value portfolio (column 1) beats the 50% value/50% mom portfolio (column 4) 26% of the time — or put another way, the 50% value/50% mom portfolio wins 74% of the time compared to a pure value portfolio (over a 10-year period).

            I hope that clarifies your question!

          • Styv March 28, 2015 at 1:10 am

            Thanks for the clarification!

      • dph March 27, 2015 at 3:11 pm

        Wes,

        Can you use value to select stocks and then “time-series momentum to time markets” to outperform?

        • Wesley Gray, PhD
          Wesley Gray, PhD March 27, 2015 at 3:25 pm

          you can use these rules to minimize your chance of massive drawdowns, which helps an investor preserve their capital…at least historically…

  2. Steve March 26, 2015 at 11:05 pm

    I’ve got more to ask and thoughts to share…but just one quick clarifying question:

    Should the return to the 50/50 combination be 17.07%? I don’t know why / how it could be 17.69%.

    Volatilities, drawdowns etc can change for sure. But the RETURN to a 50/50 Val/Mom stratgey should be EXACTLY the 50/50 return of the pure strategies.

    (15.89+18.25)/2 = 17.07%. The table above shows 17.69%.

    • Jack Vogel
      Jack Vogel, PhD March 26, 2015 at 11:39 pm

      They would be exactly the same if we were examining arithmetic returns. However, since CAGRs are geometric returns, there can be differences.

      • Steve March 26, 2015 at 11:43 pm

        I never thought of that…thanks for the clarification, Jack!

  3. Michael Milburn March 27, 2015 at 3:33 am

    I hope this post becomes one of the “most read” articles on the site. I greatly appreciate the case shown comparing independent value and momentum vs. blended val/mo portfolios. It’s exactly addresses a question that’s been unclear to me and is timely.

    The Cliff Asness paper “The Interaction of Value and Momentum Strategies” http://technicalanalysis.org.uk/momentum/Asne97.pdf (thanks Wes for referring this paper to me last year some time!) leaves the impression that a blended version of value and momentum can be slightly superior to value only. It suggests at minimum excluding value stocks in the bottom quintile of momentum. It also shows that high momentum stocks tend to perform well regardless of valuation, but that low valuation + momentum is somewhat better. That paper used data from nearly 20 years ago though. The analysis in this post, however, shows that blended value/momentum is really worse than pure value (EBIT/EV) in absolute terms and on risk adjusted basis.

    Another data point: Patrick O’Shaugnessy has advocated a blended value and momentum. This is one reference: http://www.millennialinvest.com/blog/2014/5/12/two-ways-to-improve-the-momentum-strategy (it also discusses quality similarly) It also suggests blend for outperformance and implicates low momentum+value stocks as underperformers compared to value with better momentum.

    Again, I appreciate the viewpoint and research presented. It’s good to see such thorough results presented to compare the pure 50/50 combination with the blend. I understand you might be doing revisions of the Quantitative Value book – this topic of bivariate interaction between Value and other factors – even if just briefly showing value performance sliced by momentum, size, liquidity, quality, up mkt/down mkt (200 MA?), etc – that would make a great addition to the material and possibly make a defining book for investors like myself who are interested in quant approaches, but struggle with finding and interpreting the relevant academic papers.

  4. Sebastian Jory March 27, 2015 at 7:29 am

    Why is it you ignore the final month in your momentum approach? Because it is associated with reversals?

  5. Steve March 27, 2015 at 9:33 pm

    First – thanks again for yet another epic from Alpha Architect; just seem to be a weekly occurrence around here these days.

    My thoughts, requests, questions etc….

    – Value “persists” more than momentum. It would be great to see same research with holding periods (3,6,9,12,24 etc). I realise that your study is needed to show the effects of value/momentum in a “no cost” environment. To examine how they actually perform against each other (and in blends and combinations). For the practical application side; the holding period will bring momentum down a peg or three. Particularly if you go all the way out to 12 months. At least on US data.
    Side note: I saw a table recently on Australian data that showed momentum as far more persistent than US data. Email if you want it forwarded.

    – If you have them available and are happy to; would you post the equivalent equal weight tables?

    – The Titman (I giggle like an immature teenager every time) paper, from memory, increases the universe to include smaller stocks, uses P/B for value, and uses a, “below rank” sell signal (rather than re-balance every month).
    Do these small changes account for such a difference? I’d like to think not….

    O’Shaughnessy from Millennial Invest again increases the universe, and the hold time is 12 months. The momentum factor is 6 month momentum and his value criteria is a composite that performs similarly to EV/EBITDA alone.
    http://www.millennialinvest.com/blog/2014/12/9/value-momentum-the-tortoise-and-the-hare

    His results, (whilst not directly addressing this), seem to show that a blend is better than a 50/50 combo on a risk /return basis…and a return basis. Most interestingly from his work, I believe, is confirmation of other comments (possibly by Asness, but memory fails me) about the information effect of both factors. Looking at O’Shaughnessy’s results reveals that a val/mom blend increases returns significantly over the average of the 50/50 combo (though again I am simply averaging pure val / pure mom to come up with what the 50/50 portfolio would be).

    The information effect I mentioned is the fact that, in a “pure” val (or mom) portfolio, you will include some stocks that will be “bottom decile” on the other factor. In a long-short, they cancel each other out and you wouldn’t place the trade, but in a long only situation, your are including the “10-1” stocks.

    Although I preferred the pure approach, mixed together in a combo to suit the investor’s needs (for example, you could “hold” value for 12 months – or even longer – whilst rebalancing quarterly for momentum)……..I reluctantly came to the conclusion that a BLEND is “better” because of this information effect (and results such as O’Shaughnessy’s and Titman’s show).

    At the end of the day…for the blend to work better…one statement is fact: the stocks it “changes” from the 50/50 portfolio, must improve the result. i.e. Some stocks will overlap, whichever the model. Its the other stocks that matter. The stocks that such a blend portfolio includes (with perhaps more average but consistent scores on val and mom), must do “better” than the stocks it dumps (where a stock might have a good val score but poor mom score).

    So the real test is:

    Good Val (Mom) with below average Mom (Val) stocks…..
    versus:
    Above average stocks (but NOT best decile) on both Val AND Mom factors.

    That would be an interesting test.

    My best hypothesis (guess) at this point: it’s something to do with holding period. A one year holding period will dilute momentum so much (in a pure 50/50 approach) that a blend will come out as the better way.

    How to reconcile the Titman (chuckles) paper? By using a “below rank threshold” for sell signal, I infer that the average holding period was greater than one month. Again, strengthening my hypothesis that a blend will come out on top with a longer hold.

    Hence; I’m now desperate to see if you will test and share results with a longer hold period(s!) I’d love to see 50/50 come out on top – as that just sits right with me. But I don’t think it would. I would, in fact, be extremely surprised if it did.

    Of course; just as with your comment re: other, more sophisticated ways of measuring either value or momentum; none of this considers other ideas such as: perhaps a pure approach does better than a blend (or vice versa) when looking at super concentrated portfolios….perhaps one does better than another when overlaying other, defensive factors (‘quality’ volatility etc)

    Appreciate being able to share my thoughts on this excellent post.

    • Jack Vogel
      Jack Vogel, PhD March 31, 2015 at 11:46 am

      I agree — value persists longer than momentum. We have a post planned over the summer examining how holding periods affect momentum returns.

      EW results have same conclusions, just with higher CAGRs.

      Results should be similar using Titman measures of momentum.

      We are working on another post to show what happens when you take value and split by momentum, and take momentum and split by value — probably 2 weeks out.

      I hope that helps!

      • Steve March 31, 2015 at 6:23 pm

        Thanks Jack – it does help, and appreciate it all!

  6. Heiko March 27, 2015 at 10:11 pm

    Valuentum (http://www.valuentum.com/) backtested a combined Value-Momentum strategy vs. pure value vs. pure momentum with a long-short portfolio in the period of 2002-2012. The results show significant outperformance of the combination strategy vs. each of the pure strategies (http://www.valuentum.com/articles/20120528). They measured value in P/E. So, the difference may be due to different valuation measures and perhaps also to different testing periods. But it looks like, depending how you tweak it, a combo strategy can outperform.

  7. Doug01 March 30, 2015 at 6:59 am

    http://www.zebracapm.com/research.php

    The above link is to Ibbotson’s paper on risk and return within the stock market. My conclusion from the paper is that value metrics separate out stocks well. Another conclusion of mine from the paper is that momentum sorts out the 25% of stocks that are going to do poorly, but not much more than that.

    I believe that DFA and Bridgeway have momentum screens, where they initially screen out negative momentum stocks. IMO, that makes sense. Don’t fight momentum. Wait until the negative momentum is gone before buying value stocks.

    • Wesley Gray, PhD
      Wesley Gray, PhD March 30, 2015 at 8:25 am

      Doug,

      Thanks for sharing.

      We agree that value can be useful. In fact, I wrote an entire book dedicated to finding out how to best accomplish that mission. Here is the reader’s digest version: http://www.alphaarchitect.com/blog/2014/10/07/our-quantitative-value-philosophy/

      Where in the paper do you see the evidence on momentum? We’ve spent a little over a year studying the subject and believe momentum is a more powerful long-term investment strategy than value…which is hard for me say as a hard-core value investor.

      I’m sure they do. You shouldn’t fight momentum–agree with that as a general statement. But it can get complicated. For example, using a momentum screen within the context of our quantitative value algorithm doesn’t work, whereas momentum does seem to “work” with generic value screens. In the QV context, it turns out the quality screens do the work of momentum for you.

      Anyway, thanks for the comments–great stuff. The post was meant to inspire discussion around the “conventional” wisdom associated with value and momentum–it worked! 🙂

      • Steve March 30, 2015 at 8:34 pm

        Wes, you said:
        ” For example, using a momentum screen within the context of our quantitative value algorithm doesn’t work, whereas momentum does seem to “work” with generic value screens. In the QV context, it turns out the quality screens do the work of momentum for you.”

        Are you saying that if, for the final cut, you did the following:

        – Chop off the bottom half (on quality) AND chop off the bottom half (on momentum) in SEPARATE sorts…(keeping what remains, which could be anywhere from a theoretical zero to forty stocks if all the quality stocks are the momentum stocks)

        ….it wouldn’t improve things? I’d be genuinely surprised (which is nothing new!)

        • Wesley Gray, PhD
          Wesley Gray, PhD March 30, 2015 at 8:41 pm

          yep, basically.

          • Steve March 30, 2015 at 8:46 pm

            Surprised! 🙂

          • Wesley Gray, PhD
            Wesley Gray, PhD March 30, 2015 at 8:49 pm

            join the club. this is what keeps research fun…you never really know the full answer until you do the work

      • Steve March 30, 2015 at 10:29 pm

        “We’ve spent a little over a year studying the subject and believe momentum is a more powerful long-term investment strategy than value…which is hard for me say as a hard-core value investor.”

        …I came to that conclusion a couple years back after reading all the papers on momentum vs the papers I’d read prior on value. Reluctantly, it seemed that “momentum” (I’m speaking of long only mom/val) is possibly a more powerful strategy.

        Interested in what you mean though, by “powerful?” I mean: greater returns, more statistical robustness / confidence that the factor is real and useful going forward. More stable results over shorter (e.g. decade) periods.

        In many countries outside of US, momentum also seems to beat value.

        However, in more recent times I’ve mellowed on momentum’s seeming superiority.

        – Momentum does beat value globally. But, taking for example, Fama and French, “Size, Value & Momentum in international stock returns” there are 2 things to note:

        On their 5×5 sorts, they actually rebalance value annually, whereas they rebalance momentum monthly. They also use P/B for value, whereas there is (perhaps) an argument to replace P/B with something like EV/EBITDA, with its international evidence.

        – Momentum beating value everywhere, is not exactly everywhere. Momentum especially does not beat value in Japan (long been the thorn in the side of momentum). You would have wanted to be a value investor (or a 50/50 val/mom – Asness’ argument) investor in Japan. (Note though, that good momentum still beats poor momentum in Japan, at least in big stocks)

        – Momentum really does seem to require attending to more frequently. O’Shaugnessy (James in “What Works on Wall Street” or Patrick from Millennial Invest) always use 12 month rebalance. Momentum is taken down about 3 pegs. Even using the table here (monthly rebalance) and using 18.25% for returns to momentum….and using 16.62% for value (taken from your post, http://www.alphaarchitect.com/blog/2015/03/19/how-rebalancing-frequency-affects-quality-and-value-investing-funds/#.VRn8GPyUfgg).

        Using real world, fairly cheap costs (Australian) in a retirement account…brings momentum (rebalanced monthly) down to about 12.5%….and value (rebalanced annually) to about 14.6% (after tax and brokerage – not including slippage).

        Obviously this is all away from the point of this post…but I’m not as sure as I (reluctantly) was that momentum is truly the “stronger” factor.

        Although this is subjective; I don’t think the intuitiveness of value can be completely dismissed. We don’t know what market states will exist…but it would seem to make sense (behaviorally) that cheap will beat expensive into the future, pretty much everywhere. Will momentum? It probably will (I have confidence it will) continue as it has done (intermediate momentum outperforms). But I can’t “feel” as intuitively “sure”.

        All this to say that for me, at least…I certainly give respect to both. I think that they are connected, in some uncorrelated way(!) My practical takeaway is simply to use both, given that my future investing is being done in a sub-period, and anything can happen in a sub-period.

  8. Ryan March 30, 2015 at 10:37 am

    I do not think it is practical to differentiate two very similar strategies using a single backtest.

    At the minimum I’d want to see a fairly comprehensive set of sensitivity analysis performed on holding period, ratio of momentum to value, and the exact measures of value/momentum that are used. I suspect the resulting variation in performance will be greater than the difference between the two strategies you have tested here.

    I understand that at some point all quantitative strategies need to take a leap of faith since true statistical correctness is impossible, but your conclusion here is rather broad.

  9. Cliff Hoover March 30, 2015 at 2:27 pm

    Hey guys,
    Why do you exclude the financials and utilities in your investable universe in QVAL? Thanks

  10. Gaurang Merani
    G-Man September 8, 2016 at 10:52 am

    Thanks a lot; I love this post!

    With regard to this:

    “the correlation between QV and a generic Momentum portfolio is 57.40%”

    Do you happen to know the correlation between QV and a generic trend following/absolute and relative momentum strategy that includes all other asset classes (rather than just equities)? I would expect it to be lower (but with possibly lower returns assuming momentum in “other” asset classes produces lower returns than in equities).

    Thanks again,
    G

    • Wesley Gray, PhD
      Wesley Gray, PhD September 12, 2016 at 9:37 am

      G,
      If you don’t have equities in there the correlation will be much lower. For example, if you run QV against a generic TF mgd futures strategy with comm/bonds the correlation is ~0

  11. Stef November 1, 2016 at 8:47 am

    Hi I have a question which is related to the theme of this post, but it involves having a value fund and then using absolute momentum to enter and exit it in order to avoid the worst drawdowns of the fund. I think Gary Antonacci suggested something like this here, but with caveats: http://www.dualmomentum.net/2014/10/value-investing-redux.html
    Have any simulations been done on this? Thanks
    Stefano

    • Wesley Gray, PhD
      Wesley Gray, PhD November 1, 2016 at 8:53 am

      Hey Stef,

      You can use absolute momentum and/or LT trend following and help manage the tail for pretty much any equity strategy, at least historically.

      With value you typically want to time off a passive index, not the value index itself. I’ve also seen arguments that applying to trend to value is a bad idea.

      • Stef November 1, 2016 at 10:25 am

        Hi Wesley thank you for your quick reply! So if I understand correctly (I’m a physicist and I’m quite new to this but I find your work fascinating) the idea would be to apply absolute momentum to an index like the S&p500 to decide when to enter/exit the value fund (even in the case of a foreign fund – the one I was thinking of is a small cap value French fund) since as explained in Antonacci’s book the US market leads the rest of the world. Is that correct?
        In case you had a reference at hand on the argument why trend applied to value funds is a bad idea, I would be most grateful!
        Thanks again,
        Kind regards
        Stef

  12. Doug01 May 15, 2017 at 9:04 am

    This is a very interesting blog post.

    “COMBO VW = Rank firms independently on both Value and Momentum. Add the two rankings together. Select the highest decile of firms ranked on the combined rankings”

    First of all, it looks there is monthly rebalancing. That will make momentum look good. With monthly rebalancing, the probability that value will add to momentum goes down. But in a taxable account, yearly rebalancing is more probable. In a tax advantaged account or ETF wrapper, rebalancing every 3-6 months is more probable. Under such circumstances, it is more likely that value will add to momentum.

    Also, you’re using deciles. If Vanguard wanted to do this, then decile analysis is relevant. But what if you repeat the analysis, and select the top 50 of firms ranked on the combined ranking?

    • Jack Vogel
      Jack Vogel, PhD May 15, 2017 at 10:15 am

      Thanks!

      Yes, we monthly rebalance as we are sticking to the academic construction for momentum. In practice, one may want to move the rebalance frequency to quarterly for momentum, as well as invest in a qualified account or use ETFs in an attempt to defer taxes.

      The top decile over this time period is around 100 stocks (varies from month to month).

  13. Gaurang Merani
    G-Man September 24, 2017 at 3:17 am

    Hi Jack,
    Not sure what happened to all the previous comments to this awesome post but I’ll kick it off again.
    Out of interest what was the correlation coefficient between “QV EW MR” and “MOM Decile EW” during the Global Financial Crisis (which I assume is considered to be from Oct 2007 to March 2009)?

  14. Boris Van Hemmen October 30, 2017 at 6:03 am

    Hi Jack,

    How did Qv compare to value/ momentum combo internationally ?

    I found that between 1992-2012 in EU using datastream data that anything with high ebit/tev was a winner. Top 100 ebitda/tev has the same returns as top 100 momentum and top 100 p/b but with less volatility and drawdowns. Top 100 ebitda/tev even beat the top 100 stocks ranked for value and then momentum.

    Thanks for the sharing.

Leave A Comment