Trend Investing with Newfound’s U.S. Trend Index

If you are interested in shifting your portfolio towards a trend strategy, but maybe have some hesitation with certain basic trend models or are not sure how to move away from a buy-and-hold portfolio, Newfound Research has recently released a new trend index based on the U.S. stock market that you can track in your account for free.

As any regular reader of this blog knows, I am a fan of investing with the trend. For example, I believe Dual Momentum is a great choice for individual investors who want diversification and risk mitigation but are managing their own account with time and portfolio size constraints. I personally use both Dual Momentum and a slightly more complex trend strategy in my investment portfolio.

There are countless ways to easily measure trends in equity markets. Dual Momentum uses a 12-month lookback strategy. If the measured asset had a positive return above the risk-free rate over the past twelve months, you invest; if not, move to bonds.

If you had applied this 12-month lookback time-series momentum technique to the S&P 500 and monitored just once a month, you would have outperformed the index by about 200 basis points per year over the past five decades. These one-signal models are easy to track and follow and have shown a lot of benefits for investors. They deftly avoid extended market drawdowns, see nice gains in uptrends, and don't introduce the added risk and complexity of short selling (which is extremely difficult to do profitably).

However, one-signal systems are not without their risks and drawbacks. Over the past few months this has been widely discussed in the quantitative finance world in relation to Dual Momentum.

We don't know which signal will be the best in the future, so reliance on one signal could provide significant underperformance compared with investor expectations going forward. There's the added downside that a one-signal model has you all in or all out of the market.

For more information on these discussions and the added complexities surrounding them, look at:

A Better Look at Dual Momentum Fragility

Fragility Case Study: Dual Momentum GEM (Newfound)

Global Equity Momentum: A Craftsman's Perspective (ReSolve)

Valeriy Zakamulin's Research Papers on Trend Measurements

Whither Fragility: Dual Momentum GEM (Antonacci)

There are ways to reduce the risks associated with one-signal models, but they are not always easy to formulate or track for your average self-directed investor. It may not even be feasible or practical if your account size is limited.

However, part of the drawbacks on implementing more complex models correctly has changed thanks to Corey Hoffstein and the team at Newfound Research.

Newfound Research U.S. Trend Equity Index

Newfound Index Information with Link to Updated Excel File

The Newfound team have put together a long/flat U.S. stock trend index that is updated daily on their website. It is completely free and easy to follow. The depth of information put into this system, as well as the complexity of the numerous calculations involved, make this type of index very valuable.

Speaking as someone who is fairly proficient in Excel and has done a lot of backtesting, I can assure you it would take a lot of time executing data analysis, cleaning, and formulation on a spreadsheet to come up with even a semblance of this strategy on your own. The amount of data and calculation involved would probably have crashed your average computer fifteen years ago.

In this index, Newfound has taken three separate trend measurement styles and calculated them across 120 time periods for each style. This is a longer-term system that covers trends approximately six through twelve months in length.

  1. Time-series momentum: This is close to the trend measurement used in Dual Momentum. It simply looks back a specific number of days and determines the direction of the trend based on a positive or negative return during the period.
  2. Price minus exponential moving average: This is the simple moving average method where the trend is determined to be up if the price is above the specified moving average and down if it is below the moving average.
  3. Moving average cross-over: This method uses a shorter exponential moving average and a longer exponential moving average. If the shorter moving average is above the longer moving average the trend is up.

Put together, this complex index gives an investor an approximately equal exposure to 360 different individual trend models. In this index, Newfound has divided the 360 trend models into twenty separate tranches where each tranche is corrected to the signal weighting every twenty trading days.

To get a picture of the results, I've put together a few basic charts comparing the Newfound Research U.S. Trend Equity Index to the MSCI USA Index going back through 1970. On their site Newfound has data on the index from 1927.

Performance Chart

Credit: TheRichMoose.com, Newfound Research, MSCI Inc.

I would say the performance is very much what an investor would expect from a long-term trend style applied to a single market. We see avoidance of extended drawdowns, catches on the sharp and sudden corrections, and general outperformance over the longer term.

Drawdown Chart

Credit: TheRichMoose.com, Newfound Research, MSCI Inc.

In 1987, Newfound's index caught most of the drawdown experienced by U.S. stocks as a whole. As you might recall, the 1987 market crash was sudden and into a long-term uptrend. Every model that invests with long-term trends would have full, or nearly full exposure to the market in those conditions.

However, in every other major correction we see the Newfound index showing much smaller drawdowns. Even in 2008-2009, the worst U.S. stock bear market since the 1930s, Newfound's index lost a mere 11 percent as it methodically pulled stock exposure out of the investor portfolio.

3-Year Rolling Return Comparison

Credit: TheRichMoose.com, Newfound Research, MSCI Inc.

The lagging performance of Newfound's index in bull market phases can be seen when we look at rolling period returns. Newfound's index outperforms the U.S. market when stocks are falling and underperforms when stocks are in secular uptrends.

A drawdown has more impact on a portfolio than an upwards move (when measured in percentages). Combining the dynamic stock exposure with short-term bonds, the overall performance is positive and smoother than the S&P 500 Index.

Putting This Into Practice

If you are interested in following this model in your personal portfolio, you can click on the link I shared above and navigate to the Excel file that is available for download. The Excel file is regularly updated and contains all the information you need to follow the strategy.

Newfound has taken an easy approach with this information using two popular, low-cost ETFs: VTI (Vanguard's U.S. Total Market Index fund) and SHV (iShares' Short-term Treasury fund).

If the signal in the VTI column is "1", you fully allocate to VTI (or a similar U.S. stock ETF). If the signal in the SHV column is "1", you fully allocate to SHV (or a similar bond ETF). During transition periods there will be mixed signals across the 360 models. In this case, the column will specify the exact allocation to make to VTI and SHV.

If you don't want to make tiny trades because your portfolio is small or you are busy with other things in your life, I can't see the harm in checking the signal just once a week and making the portfolio adjustments as needed. Your returns will still be very similar to the index.

Summary

I'm excited to see good quantitative finance firms such as Newfound sharing this kind of detailed trend model with interested investors. This is quality research and quality information. My guess is only one in fifty knowledgeable investors could correctly calculate a trend index with this level of detail and complexity. At this stage in my life it is certainly well out of my league.

Not only does Newfound's U.S. Trend Equity Index show great historical results that are consistent with any long-term trend application to the U.S. stock market, the structure of the index is very diverse. This significantly reduces model specification risk.

In other words, it is highly likely that this system will provide investors with future returns that are close to what one would expect investing in any other longer-term trend model without the risk of chronic underperformance in any one trend measurement due to small changes in the market.

If the 12-month lookback period lags the 7-month lookback period over the next 20 years by 5 percent annually, no problem; each individual model only has a minute impact on the total portfolio. Same can be said for the small differences in returns between comparable time-series momentum and moving average measurements.

While this index is great in its own regard, it does not cover all the bases. It is not a replacement for Dual Momentum, shorter-term long/short strategies, or other trend strategies that provide exposure to different markets.

However, if you are invested in a passive portfolio with 60 percent U.S. stocks and 40 percent bonds, you should ask yourself why you are investing with that simple structure when a tiny bit more dynamism will drastically improve your results in absolute and risk-adjusted terms.

Also, if you are playing around with single market trend investing on the U.S. market at home, you should ask yourself if you can design a strategy that covers the range of risks even close to this one by Newfound.

Corey hinted the Newfound team will post more of these indices in the future. I'm crossing my fingers for a similar application with the MSCI EAFE and Emerging markets indices. Once those are available, it may be time to replace 12-month Dual Momentum or other more diversified strategies.

I will be playing around with Newfound's U.S. Trend Equity Index. I see some interesting applications when paired with a Leveraged Barbell Portfolio. It has also given me some ideas for further developing my own knowledge in more complex applications of trend investing.

Comments & Questions

All comments are moderated before being posted for public viewing. Please don't send in multiple comments if yours doesn't appear right away. It can take up to 24 hours before comments are posted.

Comments containing links or "trolling" will not be posted. Comments with profane language or those which reveal personal information will be edited by moderator.

19 Replies to “Trend Investing with Newfound’s U.S. Trend Index”

  1. Yanniel says:

    As I was reading I was wondering how to throw leverage into the mix. I’d love to read about your ideas concerning a leveraged barbell portfolio and this index from NF.

    Very interesting read. Thanks Daren.

  2. Yanniel says:

    Daren,

    I am most likely misinterpreting the charts; but anyhow here goes my dummy question:

    The returns chart in the NF link shows the NF index returns below the US equities returns. How can the performance then be better? Thanks.

  3. Daren (Editor) says:

    Are you looking at the chart on the page showing past 5 year performance? If so, that jives well with my charts. Using a 3-year smoothed return sequence we see the Newfound Index underperforming U.S. stocks since 2011. Without the smoothing, returns have been lower since 2010.

    Trend models (without leverage) always underperform the underlying index in uptrending markets since their exposure to the underlying index is reduced as that index goes down, only to be caught with lower exposure when the underlying index shoots higher continuing its uptrend.

  4. Yanniel says:

    “Are you looking at the chart on the page showing past 5 year performance?“ Yes, that’s the one I was looking at. Your “3-Year Rolling Return Comparison” falls in line with it.

    What I don’t understand is why your “Performance chart” shows MSCI US underperforming NF US Trend Equity. It seems to contradict “Trend models (without leverage) always underperform the underlying index“. I suspect I simply don’t know how to interpret your “Performance chart”.

    I appreciate greatly your feedback.

  5. I am wondering how the NF index performs against GEM Dual Momentum. Did you also have a
    look at this comparison in your backtest?

    Btw I’ve discovered your blog only a few weeks ago and it has become one of my regular go-to blogs. Thanks for providing your insights – there is plenty of intersting stuff, ideas and tips in your posts. I’ve learned a lot already and I am currently on the way of implementing my Dual Momentum strategy (I am from Germany and one hurdle is the ETF selection here…)

  6. Daren (Editor) says:

    You can take any market (ie. U.S. stocks) and divide the historical returns chart into two periods: uptrending periods and downtrending periods. When an analyst applies a trend model, particularly long-term trends, to the market there tend to be patterns that repeat themselves. The trend model will generally underperform the underlying market to a moderate extent when the underlying market is in an uptrending period. However, the trend model will significantly outperform the underlying market when the underlying marking is in a downtrending period.
    Overall, if the trend model has the investor in a safe asset when it is not invested in the market it applies the trend to, the trend model will outperform the underlying market across a full market cycle (full uptrending period AND full downtrending period). This is almost solely due to smaller drawdowns experienced by the trend model.

  7. Daren (Editor) says:

    The NF Index underperforms Dual Momentum quite substantially. But, if we understand the mechanics of each system that is expected. NFUSTE is a complex trend application on a single market. DM is a simple trend application on multiple markets.
    I suspect that Newfound will publish their more complex trend models on other markets in the coming months and we could glean some further information that may provide an alternative to DM. However, these will still underperform DM looking back. A complex system (multi-model) must always backtest worse than the best single historical system.
    The idea is that a complex system will provide more “predictable” results going forward. I don’t mean predictable in the sense that we can expect annual returns between X percent and X percent, but rather that we should expect that the complex system will provide returns near to the best single trend model going forward. For example, let’s say we live in a world with just two systems a 12-month absolute momentum or a 12-month SMA. We don’t know if a 12-month lookback or a 12-month SMA is going to be best system going forward. But we do know that a complex system that uses a blend of both models will perform better than the worst single system and worse than the best single system. It takes the guess work out of the process.

  8. Yanniel says:

    Got it. Thanks.

  9. @Daren, thanks for your reply.

    I am considering to use the NFUTSE signal in combination with DM to gradually increase/decrease my exposure when DM signal is in US stocks.

    E.g. check weekly the NFUTSE signal and adapt the exposure:
    NFUTSE for VTI above 75%: 100% invested
    NFUTSE for VTI between 50% and 75%: 75% invested
    NFUTSE for VTI between 25% and 50%: 50% invested
    NFUTSE for VTI below 25%: 25% invested

    When DM signal is not in US stocks (Non-US or cash) the NFUTSE is ignored.

    What do you think about this approach?

  10. Daren (Editor) says:

    I might not be understanding correctly, so feel free to clarify. If you decrease exposure to U.S. stocks and put that into bonds, your returns are likely to be quite a bit lower than following a basic DM model.
    With DM, there are often periods where U.S. stocks and International stocks show positive momentum. It would be preferable in that sense to shift out of U.S. stocks and into International stocks rather than bonds in most cases. But it is still not that simple.
    If your account grows sufficiently large to introduce more complexity into DM, I think it may be best to run a multi-model system such as the one I talk about in the Newfound DM system post http://therichmoose.com/post20190208/. Starting with basic 12-month momentum, it may be best to first introduce a completely different investment strategy to try diversify returns. Following this, you might slowly add other period measurements into DM. For example, you may start by adding a 6-month momentum, then add another as your account gets a little larger, and so on.

  11. My intention was to make DM a bit more reactive and I would use the NFUTSE to achieve that.
    When stocks are in a down phase with pure DM you would go from 100% in stocks to 0% in stocks when there is negative absolut momentum in the lookback period of 12-month.

    With weekly adaption using NFUSTE I would start to reduce stock exposure before that,
    as the signal there might be already e.g. 50:50 (which means I would reduce to 50% stock exposure) weeks or months before DM signals negative absolute momentum.

    Maybe this is overcomplicating things and adding another look back period (e.g. 6 month) would achive the same thing.

  12. Daren (Editor) says:

    You can make DM more reactive in several ways: increasing the frequency of calculation (ie. from monthly to daily), increasing the number of lookback periods (ie. 12-month + 6-month + 8-month), or introducing more ways of calculating trends (ie. absolute momentum + moving average + moving average crossover).
    The problem is that the number of trades made generally increases with the number of variations made on the original system (provided you don’t filter them somehow). More variations doesn’t always provide more benefits to the investor. Certainly after a certain point each additional variation on original DM has a decreasing value add to the overall system.
    In addition, each new variation needs to be properly integrated into the whole system so that it has relevant impact on all the markets tracked. This is where it gets more tricky, especially since DM is supposed to have you completely invested in 1 of 3 markets (U.S. stocks, International stocks, or Bonds). Not only do you have to apply each variation on a single market as we see with NFUSTE, but you have to compare markets to each other.
    So, lets say we have NFUSTE putting you 0.76 in VTI. Where do you put the other 0.24: in International stocks or bonds?
    If Newfound comes with a trend index of the MSCI ACWI ex-USA, that may have a score of 0.85 while NFUSTE has a score of 0.76. Do you invest in ACWI or USA? How much in each? Do you have at least 0.15 in bonds because that’s the score of the highest allocation (ACWI). If both the ACWI and USA are at 1.00, do you split your portfolio 50/50 to each? What is USA is truly in a stronger uptrend after diving into the numbers on a relative momentum basis.
    There are answers and ways of calculating the appropriate allocations, but as you can see it gets complex very quickly.

    If you are currently following DM, I believe it is best to first diversify by introducing a second trading system that tracks markets different from DM (a different return stream). After this, then consider adding more variations on DM to decrease system risk if your portfolio is big enough.
    On the other hand, if you only want to trade with the trend on U.S. stocks and don’t want foreign stocks, consider using NFUSTE to determine how much you should be allocated to U.S. stocks.

  13. Yanniel says:

    Similarly to Red, an idea of combining DM with the NewFound US Trend Equity (NFUSTE) Index crossed my mind. I realized that I needed a second index for non-us stocks. For the sake of the argument let’s pretend such an index exists; let’s call it: NewFound ex US Trend Equity (NFEXUSTE).

    My idea is simple, instead of using a 12 month lookback period to determine the absolute momentum (AM) and relative momentum (RM) signals; I would use the NF indexes.

    To determine the AM signal, I would look only at NFUSTE. If the allocation to VTI is above 50% then AM favors stocks; otherwise if favors bonds. This 50% threshold is kind of arbitrary. We’d had to back test and determine what percentage threshold would be used. Maybe a 40% VTI allocation would warrant being in stocks.

    Once AM decides to be in stocks; then I would look at both NFUSTE and NFEXUSTE to determine the RM signal. The index with biggest allocation to stocks would win. If NFUSTE wins all the money goes to VTI otherwise it goes to ACWI.

    One can stop here, but we can go a step further and implement something similar to GEM70 (this is an implementation Gary described in his book where 30% of the portfolio is perpetually allocated to bonds and the remaining 70% would be as dictated by the rules of DM). Instead of allocating always 30% to bonds, one could look at the score of the NF indexes to determine a dynamic allocation to bonds.

  14. Yanniel says:

    Hi Daren,

    I am somewhat confused about something said in the ReSolve article linked above.

    “The Original GEM strategy was specified with a 12-month lookback period to measure both trend and momentum”

    Is this saying that “two” measurements are required (one for “trend” and another one for “momentum”; possibly using a MA in addition to the total return comparison) to determine the DM signal?

    Then it follows:

    “We examined GEM strategies specified with all possible combinations of absolute and relative momentum formed on 1-18 month lookbacks. With 18 possible parameterizations of trend lookbacks and 18 possible momentum lookbacks, there are 324 possible strategy combinations.”

    This again reinforces the idea that there are “two” different measurements made per lookback, one for trend and one for momentum.

    This puzzles me because I always thought DM required just one measurement: comparing the total return over the look back period.

    In the context of DM what’s the difference between “trend” and “momentum”? It is hard for me to draw the line between the two, but the ReSolve article is clearly separating those two concepts.

    Would you be so kind to drop a line or two about your take on this?

    Thanks.

  15. Daren (Editor) says:

    When I read the article I interpreted it as different ways of stating and separating relative momentum and absolute momentum. The original DM-GEM is very clear in using total return on a 12-month lookback period (in excess of the risk-free rate). Gary has talked about this single preferred measurement of momentum many times in his blog, papers, Twitter, and even on podcasts he has done.

  16. Yanniel says:

    You are right Daren. For the sake of others I’ll explain the source of my confusion.

    ReSolve refers to “trend” and “momentum”; yet Gary refers to these as “Absolute Momentum” and “Relative Momentum” respectively. Furthermore, academics often refer to “trend” and “momentum” as “Time Series Momentum” and “Cross-sectional Momentum” respectively.

    Trend following techniques can be used to determine both “Time Series Momentum” and “Cross-sectional Momentum”; so I got pretty confused when ReSolve’s article treated trend and momentum as different things.

    I went back to Gary’s book and all these concepts are explained in Chapter 7.

  17. Looks like the Excel from the link does not get updated, last entry date is April 12th.
    https://www.thinknewfound.com/newfound-research-us-trend-equity-index

    Do you know where to get the updated Excel from the Newfound site?

  18. Daren (Editor) says:

    Using the link you gave I got April 30.

  19. Hm, yeah, nevermind – looks like there is a strange browser caching happening for me. If I try with Chrome (my default browser) I get an old version. I’ve tried IE and got a recent version.

Comments are closed