Market Timing and Risk Management, Part 2 – Momentum

I started a new series in February on Market Timing Risk Management (part 1 was on macroeconomics) but never got beyond the first part. So, finally, here’s the second installment! Part 2 is about momentum (sometimes called trend-following) and this is a topic requested by many readers in the comments section and via email. Specifically, many readers had read Meb Faber’s working paper on this topic, which by the way is the Number 1 most popular paper on SSRN with 200,000+ downloads. I always responded that read that paper and found it quite intriguing but never followed up with any detailed explanations for why I like this approach. Hence, today’s blog post!

And just for the record, I should repeat what I’ve said before in the first part: I have not suddenly become an equity day-trader. I am (mostly) a passive investor who likes to buy and hold equities. But with my early retirement around the corner and my research on Safe Withdrawal Rates and the menace of “Sequence Risk,” I have that nagging question on my mind: Are the instances where an investor would be better off throwing in the towel and selling equities to hedge against Sequence Risk? At the very least, I’d like to have some rules and necessary conditions that need to be satisfied before I would even consider reducing my equity exposure. I think of this as insurance against overreacting to short-term market volatility!

So, without further ado, here’s my take on the momentum signal…

Momentum 101

Using momentum as a way to tactically time the equity allocation (as well as many other assets classes) is not rocket science! There are many different versions of this rule and one of the simplest, yet still very effective, version is the one favored by Meb Faber: Every month-end, calculate the average of the last ten monthly S&P500 closing values (including the current one) and compare the current level to the 10-month average.

  • If the current index level is above the 10-month moving average, then you want to be invested in equities.
  • If the current index level is below the 10-month moving average, then you want to be out of the equity market. 

It’s really that simple. Why ten months? Well, the research shows that any time horizon between, say, 5 and 12 months seems to work, but the “sweet spot” is around 10 months. It’s a short enough horizon that you’ll still capture the major turning points in the stock market in a timely fashion, but long enough to not overreact to every little blip in the market. Though even with this rule there have been a lot of false alarms, more on that later!

Let’s look at a simple example, see below. For the calendar month of October 2007, the S&P500 closed at 1549.38. The average closing S&P500 levels over the nine preceding months plus October itself was 1482.77. Since the October close was above that average we’d keep our equity exposure:

S&P500 in 2007. If the index level is above the 10-month rolling average, stay invested in equities!

Forward one month and the index dropped to 1481.14, which is below the 10-month rolling average. That’s a sell signal!

S&P500 in 2007. One month later you’d sell equities!

Talking about 2007, the 10M-momentum signal would have been amazing at timing the exit from the equity market in 2007. You’d have missed the equity market peak by only one month and avoided the whole mess during the global financial crisis. The buy signal would have come in May 2009, only about three months after the market bottom. An investor following this rule would have avoided a 34% drop in the (total return) index!

10-month momentum worked beautifully during the Global Financial Crisis!

But, for full disclosure, the momentum signal also creates a lot of false signals! For short-term market fluctuations, this signal sucks! Here’s one example from 2015. You’d have sold after a temporary drop in August 2015 but then the market rallies again and you’ll buy back equities at a higher price! We’ll see below that there were quite a few such false alarms!

If the equity drawdown is too short (as it usually is outside of a recession) then momentum will do poorly!

Simulations 1871-2017

Let’s look at what happens when we apply this simple momentum rule to my historical equity returns. One thing I still have to specify is where we’d invest the money we pull out of the stock market. Apparently, Meb Faber assumes you invest in a low-risk short duration investment like cash (e.g. 3-month or 1-year T-bill). I will look at two different options:

  1. longer-duration Treasury bond, i.e., 10-year bonds.
  2. cash, e.g., short-term T-bills, similar to a money market account (and similar to Meb Faber’s calculations).

In the chart below is the cumulative growth of $1 invested in January 1871. And yes, this already inflation-adjusted! Because of the tremendous equity return potential, we have to plot this using on a log-scale on the y-axis!

  • With the S&P500 (with dividends reinvested), $1 would have grown to over $14,400 by 2017!
  • With the momentum strategy, replacing stocks with 10-year bonds while the sell signal was in effect, $1 would have grown to over $190,600 by 2017!
  • Almost equally impressive, shifting to cash (i.e., short-term T-bills) during the sell signals would have grown the $1 investment to over $116,000 by 2017!
Following the 10M-MA momentum strategy to time equities would have greatly enhanced the performance relative to a passive all-equity portfolio!

How about some more detailed stats, ERN-style? Here are the (real, inflation-adjusted) return stats for the three asset classes stocks, bonds, and cash as well as the two alternative momentum strategies. (Note: return and risk are calculated on an annualized basis, returns are compound returns). I also calculate the return/risk ratio and the maximum drawdown percentages. Some noteworthy results:

  • The momentum strategies yield higher average returns with lower risk and lower drawdowns! You almost double the return per unit of risk!
  • The Stock->Bond signal gives you slightly better outcomes, especially when looking at the latter part of the sample. Thanks in part to the strong bond bull market since 1982!
  • Both bonds and cash had some nasty drawdown percentages during the 1970s and 1980s! That’s because I calculate the returns as real returns. In nominal terms, of course, a money market account would never have any drawdown!
Return stats (all adjusted for CPI-inflation!!!): Timing the equity exposure with a 10-month momentum increases the average annual return and reduces risk and drawdowns!

Some more info on the drawdowns. Again, with the drawdown I measure how far the total cumulative return drops below the all-time-high prior to that point. Keeping an eye on drawdowns is particularly important for retirees because of Sequence of Return Risk (see part 14 and 15 in the SWR series!). Remember, you can have a great average return throughout your retirement but if the drawdown early on in your retirement is too deep and/or too long you’ll still run out of money. In any case, the drawdowns would have been greatly reduced with the momentum signals as we can see from the plot below. Especially during the last two bear markets 2001-2003 and 2007-2009, when the momentum strategies barely lost more than 15% from the peak!

Using a 10-month momentum strategy greatly reduces drawdowns and thus Sequence of Return Risk!

A closer look at the 64 sell and buy signals since 1926

Let’s get a better sense of when the momentum signal works and when it doesn’t. Remember, the way the signal is constructed, your returns will be identical to the passive equity strategy if the momentum signal is bullish on stocks (i.e., the S&P500 is above its 10-month rolling average). Differences in returns come about only during the times when you sell equities and temporarily hold bonds or cash. There were 64 such instances since 1926 (I look at the shorter subsample to save space) with varying durations. In the right two columns, I calculate by how much the respective momentum signals would have out/under-performed the S&P500. Out of 64 sell signals, I counted only 25 that “worked,” when buying bonds and only 17 when shifting to cash. Those are pretty measly hit ratios of 40% and 27%, respectively. The reason the momentum signal was so successful in the overall stats is that during the failures, the momentum strategy underperformed by “only” single digits (with a few exceptions, mostly during the Great Depression), but when it works it sometimes hits the ball out of the park! 170% during the Great Depression, two back-to-back 30+% in the early 2000s and 66% during the Global Financial Crisis!

All 64 buy/sell signals since 1926 from a 10-month momentum signal. Duration of being out of the market and outperformance relative to buy-and-hold. Most of the time one would have lost, but the few large gains more than made up for those losses!

Let’s sort the sell signals by how much excess return they would have added over an equity buy-and-hold signal. I also add two more columns: how many of the months between the sell and buy signal occurred during a recession (as defined by the NBER) as well as what percentage of the months out of the market that occurred during a recession. Notice a pattern? Most of the great success stories of the momentum signal occur during a recession and most of the false alarms occur during economic expansions! That makes sense, because as I have written in a post before: macroeconomics and the stock market are joint at the hip: The equity drawdowns around recessions are those that are deep enough and long enough to be correctly timed with this momentum signal! And the false positives normally occur during economic expansions when the stock market mean-reverts too rapidly for the 10-month momentum signal to properly pick up what’s going on!

Sorting by the efficacy of the momentum sell/buy signals: the signal was most effective during the large recessions and generated mostly false positives during economic expansions!

Why a momentum strategy works: the stock market is no Random Walk

None of this momentum business would ever work in real life if the stock market followed a true Random Walk (with upward drift, of course) in the strict mathematical sense. Past returns would be completely uncorrelated with and independent of future returns and thus irrelevant for predicting future returns! But as I mentioned in a post from a few months ago, the stock market isn’t exactly a random walk. And again, to calm down the critics and the trolls, I’m not saying that the stock market is perfectly predictable. I am merely saying that it tends to occasionally overreact on both the upside (dot-com bubble) and downside (Global Financial Crisis). The transition path between these cycles of greed and fear seems to be gradual and slow enough for a momentum signal to work!

How about taxes and transaction costs?

Personally, if I ever wanted to implement this strategy for timing my equity exposure, I’d probably avoid selling large quantities of stocks in taxable accounts. The tax benefit of deferring capital gains and paying the IRS only once when you actually need the money probably outweighs the timing benefits of the momentum signal. But within tax-advantaged accounts (IRAs, Roth IRAs, 401ks, etc.) where I can easily move money without tax consequences and without transaction costs (just swap one mutual fund for another) I will definitely consider this momentum signal!


The past performance of this momentum strategy has been quite amazing. You lower the drawdowns and risk while increasing average returns at the same time. But this momentum signal isn’t fool-proof either. There are many false alarms! During short-lived equity pullbacks, you are almost guaranteed to sell stocks close to the market bottom, only to buy them back at a higher level during the following recovery. But in the historical simulations, the numerous small “failures” of the momentum signal were offset by several large home runs when you could have avoided the big equity drawdowns!

So, even though I am still a passive buy-and-hold investor at heart, I think it would be a mistake to ignore this signal. At the very least, I’d add this momentum signal to the three macroeconomic signals I outlined in the first part of this series (yield curve, unemployment claims, ISM index)! Actually, monitoring both macroeconomic and equity-market-based signals might even be the solution to avoiding some of the false alarms that the momentum signal inevitably generates. Recall that most of the false alarms occur during economic expansions and most of the successful sell signals occurred around NBER recessions!

We hope you enjoyed today’s post! Please leave your feedback in the comments section below!

158 thoughts on “Market Timing and Risk Management, Part 2 – Momentum

  1. Nice piece, though I am always curious how this strategy would have held up in the real world of taxes, transaction costs and investment fees. For most of the time period, there was no tax-free trading retirement accounts, low commission or low cost index funds. LT Cap gains rates were up to 40% at times, short term cap gains even higher, commissions rates were 1% each way, and a well diversified mutual fund (no index trackers in the day) cost 1% plus loads.

    Factor in all those costs and I imagine the excess returns would evaporate! It would be interesting to see a researcher try to proxy them, haircutting the strategy’s performance with a rough estimation. Granted, in the more recent era of retirement accounts, 15% cap gains(LT, ST still at ordinary income rates), near commission free trading and 0.05% index funds, the implementation costs for the strategy have been dramatically reduced, but statistically not for long enough to draw firm conclusions.

    1. Think he said there were under 100 transactions in the huge period he looked at. Employing a 5% envelope minimizes the transactions while still offering protection.

        1. I guess to that point, on what data set was this model derived? Is this an in-sample v. out-sample issue? Thanks again for the in-depth analysis!

          1. Momentum has been around for as long as the stock market exists. But for full disclosure, the Faber paper has been around for only a few years. But it was published pre-Global FInancial Crisis and it would have worked spectacularly well in 2008-9. So, the simple 10-month momentum idea has been around for a long time and can be considered tested out of sample. But a lot of the “improvements” were fine-tuned to work better. So, careful, there’s in-sample bias! 🙂

    2. Note that institutional investors (pensions funds, endowments, etc.) could have used this without any tax consequences. 401k plans have been around for many years. And yet, this hasn’t been arbitraged away (yet).

  2. Can you run the SWR strategies using the Momentum stock->bond or stock->cash scenario? I am assuming that by mitigating the big drawdowns, it improves success probabilities/SWRs. However, seeing those extended periods of sub-par returns makes me wonder… What if you retire and have 10-15 yrs of sub-S&P results due to this approach (assuming a major recession does not hit or that it is a slow leak with lower than average S&P returns for the next decade)? Have you now created your own sequence of return risk by using a method that provides sub-par results in the first decade?

    1. I haven’t run that simulation yet. But most people would be willing to forego some upside in exchange for hedging the sequence risk. In other words, the SWR will be lower most of the time, but only during times when it was really high. But you will likely lift some of the extreme low outliers during the Great Depression, 1970s/80s and 2000s.

      1. I got tired of waiting for ERN to run the simulation, so I did it myself! 🙂
        See here:

        Actually, even though I have devoured every word of the SWR series, and most of the comments sections as well, I hadn’t even read this post until this weekend.

        And not being aware of Meg Faber a month ago, or even fully aware I was creating a market timing scheme, I “rolled my own” rather than using the 10 month moving average (I’m in the process of trying to use that now, and see how it compares). So the method I use there is not identical.

        But the results strongly support the point that doing such market timing is indeed a great way to forego some upside in the average case in exchange for hedging sequence risk.

  3. Hello Karsten,

    Thank you so much for doing this post!! This topic has been nagging me since it seems so logical. Do you know if we would use our individual ETF for the 200SMA or do you use the S&P 500?

    I was about to ask Meb Faber but you probably already know!

  4. That’s an interesting simple strategy if you can stick with it through the thick and thin. I’m too lazy to be active in my portfolio, and probably wouldn’t take selling out on a false alarm very well. Still, it’s nice to see what tools are out there for gauging the markets. Thanks for the analysis 🙂

    1. Same here: I would have a hard time to stick with this. We are getting dangerously close to the sell signal unless equities start moving up again!
      That’s another reason like to monitor some of the macro signals as well! This momentum business as a stand-alone just seems too risky! 🙂

      1. What would combining the macro signals with the momentum look like? I’m guessing you would monitor the 10-month moving average and if you dropped below that level, you would then go confirm with the macro signals? Would you expect those signals to coincide close enough temporally to sell at the right time? I would be interested to see an analysis of how combining the two methods would have worked historically.

  5. Seems ripe for an ETF instead of trying to follow this for an individual.

    This seems like a good possible example of something going away once people start following the strategy. If enough people start doing this, will that just make the down movements of the market that much more steep?

      1. And his fund has returned 4.76% over the past 4 years since launching. Using VTI (Total US Stock Market) returned 9.72% over that same period. Using VT (Total World Stock Market) returned 8.01%.

        That’s a bigger underperformance over a longer period than anything on ERN’s chart other than that 2 month period in 1933.

        If the guy who made the momentum signal so popular can underperform so dramatically….then what chance do any of us have?

        1. Not sure why the ETF underperformed by THAT much. It’s the expense ratio, for sure, and the absence of a bear market. But may also be due to implementing this with many different asset classes, not just VTI. So, bonds, commodities, etc. and they all had much lower returns than 9.72% over the last 4 years! 🙂

      1. I have some shares in GMOM. But it’s not been as great a performer so far as Meb’s research led me to expect…

  6. Thank you for another fantastic analysis. Since the 10 year bond purchases are temporary, I assume these are done via a bond fund. But these funds may lose money if interest rates are rising during those times. I was not sure if you took that into account or were looking at the 10 yr yield alone.

    As a “stocks for the long term investor”, I do find it harder to not flinch now that we are retired. Even though we don’t need to draw much. It’s just slightly different when you aren’t pouring money in every month. I’m curious how it will affect you. I’ll stay tuned!

    1. True: Bond funds lose value when the yield goes up. But that’s taken into account in the bond returns I use!
      Anyway, the true test comes when this momentum signal points to a “sell” next time around. Not sure if I have the stomach to sell! 🙂

    1. Yahoo finance let’s you plot funds with their moving average. 10 months is approximately 200 market days.

  7. Hi Karsten, on your point of “monitoring both macroeconomic and equity-market-based signals might even be the solution to avoiding some of the false alarms that the momentum signal inevitably generates”, I believe that this has been considered in detail over at Philosophical Economics and does indeed seem to work better than either of the signals alone – see

    Thanks for your continued great work, and here is naother vote for modelling the impact of trend following on SWRs!


    1. Philosophical Economics also has a post on the impact of bid-ask spreads when performing backtests for a trend following strategy – see below. Some of the premium may diminish in the future due to lower transaction costs (in the present compared to the past) and Rekenthaler’s Rule (“If the bozos know about it, it doesn’t work anymore”). Unfortunately, if your strategy is associated with the #1 paper on SSRN, then you can bank on lots of other folks doing the exact same thing you are doing.

      1. That’s my concern too: The signal might get too crowded. That makes it all the more important to condition the momentum signal on something more fundamental (e.g. macro) and not just blindly follow some backward-looking moving average – something that every Bozo can do! 🙂

  8. Very interesting analysis yet again. I second the suggestion that it would be great to see you run this approach on your SWR calculations, or – dare I suggest! – add an option on your interative SWR sheet for 2 asset allocations which switch according to the momentum signal. I imagine few people would feel comfortable moving 100% in or out of equities or cash on a signal, but would be comfortable increasing or decreasing the allocations to a lesser degree. The upside would be less dramatic, but it would still be valuable to know if the return benefit yields a tangible SWR increase. If it does this could be a very significant observation for SWR planning.

  9. So happy you got back to this. Came to the same conclusion. I think using a 5% envelope minimizes the transactions in the taxable account while still offering protection.

    I wouldn’t look at this as a market beating strategy. It’s a protection strategy that is perfect for your glide path. It protects so if the period you’re looking at has a crash it will beat the market. However if you look at it since the real estate crisis it has undeformed since there hasn’t been a crash to catch.

  10. I agree that this signal shouldn’t be ignored. It seems it’s an ideal system for the Investor who actively manages their portfolio as I do. The track record is impressive and I like that it helps mitigate overall portfolio risks when the sell signals are issued.

  11. What about using an options collar (sell call buy put) in a taxable account when getting a signal? I am thinking of doing this. Main question for me is how many months out should the expiry date on the options be?

    1. Allow me to answer this. As for options collars; it is a esoteric hedging strategy but wouldn’t at all resemble the strategy mentioned in this blog posting as the SELL discipline. Options collars are complex as you have to sell calls far enough out of the money but not so far that no premium is captured to fund the purchase of puts. For a full hedge the puts need to be at the money which is very expensive (premium costs). As for duration of hedge you’d want to go pit at least 6 months and up to 12 months.
      You might want to read my own blog postings on options:

      1. Well I am thinking actually of either calls and puts 5-10% above and below the market to protect against another 2008-9 style decline. You can set up the call and puts to have a net cost of zero on set up. You can go closer to the market if you want to more closely approximate Faber’s strategy. If the model turns out to have made a mistake you can record a capital loss on the call option and put option rather than a big capital gain on selling all your stock…

        1. It seems as though you’ve done your homework on this. I fee it’s a solid way to hedge though I’m personally waiting as I feel the market is near a low and will revisit $SPX 3000 or even 3200-area before the big top ushering a 2008-type event. That’s me and only My opinion. I do like collars though. As for tax liability, I sold my equities down late last year to less than 20% allocation, I’m paying the taxes now.

        2. You’d have to periodically roll the options in your collar to a later date to maintain protection. The frequency of transaction costs (commissions, bid-ask spreads, taxes) would depend on how much time you bought each time. So that makes you want to go long-duration.

          The prices of the call and the put are almost irrelevant if they can cancel each other out. However, as the underlying changes and either the put or call becomes more expensive, you can expect changes in implird volatility to have a small effect on the overall position.

          I have switched from VTI to SPY in my portfolio to access the deeper options market for SPY. My collar features a roughly ATM long put with over 2 years duration that I plan to roll to the next year annually. I am writing covered calls every month or two to slowly pay for the put. The point is to take advantage of slow time decay for long duration options, and fast decay for short duration options. One downside is that I have tied up lots of cash in the long put, which will take a while to replace using funds from small sales of calls. Also, commissions, though small, might add up to more than my profits on time decay. Third, I have to watch the value of my long put fluctuate due to changes in IV.

          The benefit of this approach is that I can be more aggressive with near-the-money call selling (if assigned, just get back in by selling a put). I can do this without majorly limiting my upside if the market continues slowly climbing the staircase of worry. But a net is in place if it falls off the cliff of fear.

          Finally, my put is in my taxable account, but it is hedging SPY shares in tax-advantaged accounts. If I lose money on the hedge, which is probable, it’s deductable. If I need the hedge, it will be at a time I might have other tax losses to offset.

          1. Pretty cool strategy! I like the idea of “arbitraging” the different rates of decay. I noticed the same, the shorter-dated options (in my case the weekly puts) have pretty rapid time decay!

          2. This sounds good… Commissions shouldn’t amount to too much – IB charge 70 cents a contract which for SPY would cover $27k in stock. So about $6 per $100k of stock for a collar.

              1. I use 10% OTM puts on SPY two years out for a tail hedge. I can generally get them on for 1-2% annually when vol is low. I closed some in Feb for a 12% gain so they’ve more than paid for themselves. I used Meb’s 10 month moving average for the last decade, it works well, but you do get some false signals and its hard selling low and buying high, thats one reason I keep the hedges on, so I can sell them on the way down before the trend trigger, which possibly happens once a month on the last day of the month for me. One adjustment I’ve made is that I don’t completely wait for the reentry signal, but start buying in, averaging down at certain levels, down 10%, down 15%, down 20%, that way I’m locking in outperformance. It increases volatility, but also increases returns

                Early last year I started using Gary Antonocci’s Dual Momentum strategy. It uses relative and absolute performance between the S&P and the world exUS combined with a 12 month look-back. He just took his back test data back to the 50’s on his website. I also use a research service Hedgeye, to front run the signal, they map price, volume and volatility, with some GDP forecasting which is especially helpful. They called for a Q4 slowdown in August so I started migrating my work place accounts to bonds. At the end of September I also took my IRA’s to low beta, long bonds and utilities, and sold the AMZN shares I’ve held for the last year. I held onto my short SPY through last Friday when I took everything to cash. Also shorted tech through Oct, so I’m sitting on a decent YTD gain in all my accounts giving it a little break to see what the market does. We’ve been really balancing these last 3 days, I think odds are we continue down but we could bounce from here. The slowing GDP in the US and the world is what gives me that opinion, I think we’ve seen cycle highs on rates, inflation and the market. We still have a Q3 jobs report Friday which could push rates up one more time. My inclination is to short rallies in my IRA’s while keeping my work related accounts in bonds but keep dollar cost averaging, and average down my work related accounts.

                1. Interestingly Meb’s 10 month average will trigger but not Antiocci’s Dual Momentum

                2. Thanks for your insights! Very interesting. The re-entry timing is really tough and I like the gradual buying back approach!

                  Nice tail-hedge strategy! I’ve been considering that, too, but haven’t implemented it yet. I also you considered using long VIX futures as a tail-hedge.

  12. I notice that the cash strategy outperforms the bond strategy except for the massive bull market in bonds. I imagine that the bond strategy will underperform going forward from here.

    Is it possible that the 10-month time period came from back-fitting to the data? Is there any reason that this should outperform going forward? I know you said that it comes from overreactions, but couldn’t this method result in false alarms bigger than the benefit going forward and a different length average outperforming? Similar to how if you had compared bonds to cash methods for 100 years until 1970 you would conclude that the cash method is better only to be crushed by the bond method for the next 40 years.

    1. Avoiding one single 33% drop is worth 50% in outperformance (100/67=1.5). So it would take a lot of false signals to undo that!

      Agree: I would probably implement this with Stock->Cash at least until bond yields have found a new higher plateau.

  13. Talking with my wife I just likened this strategy to walking through my backyard stepping “Left, left, right,right,right, left, right, left” to miss all the dog poop and then thinking the pattern/”strategy” will work when I continue walking through my neighbor’s yard. I feel like ultimately I’ll…step in it!

  14. Thanks for running the numbers on something that’s been talked about a lot but with a dearth of real in depth analysis! This is great work.

    One question though, say we sell out on one of the ‘false positives’ how do we afford to buy back in at a higher price? I presume if we’re using a pure buy and hold we’d have to keep a cash float.

      1. That’s not clear from the analysis to me. If you do that your return when reinvested will be less than the market return (on an absolute basis).

  15. If I am reading this correctly (perhaps not), it looks like your simulation started at one point in time, and the results flowed from there. It would be interesting, and more meaningful, to see the results starting from each year, or better yet from each month. Of course that would be a lot more work. Am I correct that the results could vary quite a bit depending on the actual starting point? Thank you, and well done!

    1. True. but since I can’t know when people like to start and I can’t list this for all starting dates we have to work with this. But you may notice that the outperformance was pretty consistent: whether the start is 1871, 1926 or 1970.

      1. I was thinking the same thing as OP… It would be interesting to compare periods of 20 years following this strategy and not following this strategy respectively. Right now, you have proven that this active strategy would have significantly outperformed a passive strategy for a period of 150 years. That’s something, but that’s not enough…

        If we could show that, starting any month during the last 150 years, for an investment period of 20 years (could be 10 or 30…) this active strategy would have outperformed a passive strategy let’s say 90% of the time, that would be amazing! And when it outperforms, by how much on average and when it underperforms, by how much on average.

        1. The share of 20-year windows that underperformed with momentum is 26%. 11.6% for 30-year windows, 8.2% for 40-year windows.
          But I think that exercise is completely nonsensical. I’m willing to underperform when the stock market rallies if this provides insurance for the big disaster scenarios when the 4% rule failed!

  16. Idea for next analysis:

    Stop-loss orders established x% below current prices only when the price falls y% below the moving average.

    The stop loss, rather than a sell order, would protect less capital but have fewer false positives. The strategy discussed in the post is dragged down by false positives.

  17. I would really like to think that this trend-following business will work. I have long thought that if one could sidestep just one severe decline in one’s investment lifetime, one could get very far ahead. If you sidestep a 50% bear and ride the ensuing bull all the way from the bottom, you could be 100% ahead of the buy-and-hold guy.

    However, I think the method presented here is just too costly. It banks on avoiding some fairly rare events and pays along the way by missing pieces of the long-run growth.

    I just proved to myself that it was too costly by doing a spreadsheet exercise on your data in 64-event table (bond column). If one had gotten spooked by the 1929 crash and implemented the 10-month moving average system on its next sell signal (11/30/1932) one would have trailed the buy-and-hold investor until 1981. Even then, the timing investor was ahead only 2%. He then fell behind again until the tech stock melt-down in 2002. At the worst, the timing investor would have been 37% behind buy-and-hold. Timing worked well post-2002, but who in his right mind is going to stick with a losing system from 1932 to 2002, seventy years?

    A huge part of the historical advantage of the moving average system is dodging the 1929 crash. The markets were a different place back then. There was no SEC. Syndicates were actively trying to boost, or having shorted, depress prices of individual issues. Speculating on margin was rampant. Margin calls on over-extended speculators exacerbated the crash. There are now regulations regarding margin limits. I would not want to bet on having to avoid a similar crash for my system to work, given how much things have changed.

    I looked at starting the moving average system after the 1937 crash. Again, it was underwater for decades, until 1981 (and only modestly ahead then), and trailed by as much as 33%. Starting after the early 1970’s crash, one did get modestly ahead in the 1980’s (18%), but trailed again, buy as much as 27% prior to the tech melt-down. A 1982 starter got as much as 38% behind. A 2010 starter is still underwater.

    I think you owe your readers a follow-up on just how ugly this can be.

    1. Trend following based on momentum DOES work. You will have a couple f false starts here and there, it happened to one of my managed accounts in 2015, the finally got to rocking in early 2016 and is still Long and strongly performing. I have a living trust account for the benefit of my children managed by a financial advisor at Merrill Lynch. He uses a sector ETF trend following system. It uses the 200-day moving ave. It is long above and in cash below, he uses the top 10 performing Sector ETF’s (based on momentum). Here’s uses same filter on this Meb Faber system, each month on the second business day of the month (don’t ask me why this particular day of mo.) his system recalculates based on the 200 day. Get this, his system recalculates the ETF’s our of the 100 he filters, some will be sold out if they cross below and some will be purchased if they cross above. In other words, the etf’s Have their own individual stop loss in addition to the large scale stop loss for the entire portfolio. I am not certain how it’s all calculated, he uses a computer model. He manages over $200million with this system and like it Mentiond, the performance has been terrific.

    2. Please read the title once more: “Risk Management”
      I can see how people want to argue that the signal may not work anymore in the future. Fine. I can’t prove the opposite. But it’s silly to argue that the signal didn’t work in the past. It did! Beautifully! Much better than, say, a 60/40 portfolio.
      Momentum underperforms in the periods when the market is rallying (60/40 would even more so). But a lot of people are willing to forego a little bit on the upside in exchange for the lower vol and lower drawdowns. That’s the whole purpose of risk management.
      Now, if you have a suggestion for an alternative method for lowering drawdowns while missing none of the upside during the major bull markets then please post your suggestion here. It takes a strategy to beat a strategy!

      1. It did not work beautifully in the past. You have cherry picked two start dates that happen to make it look good. Redraw that lead graph starting in 1932. You are not foregoing “a little on the upside”. With the 1932 start, you wind up 37.4% behind buy and hold. That is normal bear market territory. Trying to avoid a bear, this system would make one for you, albeit in slow motion.

        Look at the table of 64 events, items 11 through 19. Nine trades, eight of which are negative. The sum of these trades is MINUS 27.3%. Who is going to make the 10th trade after that run?

        Think that was a fluke? Try events 45 through 55, the 1980’s -1990’s bull market. Ten of eleven trades were negative. The sum of these trades is MINUS 46% versus simply holding the S&P500. The one positive trade in that run got back all of 1.2%.

        Anyone thinking of adopting this system owes it to themselves to really dig into this history and decide if they will have the conviction to make yet another trade when all their recent experience has been bad and their timing efforts have left them badly trailing the markets.

        1. OK, there is exactly one person that’s cherry-picking: you! I looked at long-term average returns. You conveniently pick the few starting dates of long stretches of drought for this strategy.
          Also: Do you know what all of the starting dates you mention have in common? They are all completely non-representative for today’s environment. 1932 was the bottom of the 1929-32 crash. CAPE ratio of 5.6. 1920s: CAPE in the single digits again. Today’s CAPE ratio >30. Do you agree that we are much closer to the next bear market today than in 1932 or 1982?

        1. Yeah, but the other indicators still look pretty good: Yield Curve, Unemployment, Inflation, etc.
          I’m personally also comparing the 20d SMA / 200d SMA, just to smooth things out a bit, otherwise you get too many false alarms 😉 So this time I’m not really concerned (yet).

        2. I’d use this one:
          Otherwise you got a signal in April as well 😉 Which was obviously wrong…

        3. The signal only triggers on a predetermined date, mostly the beginning or end of the month. So if you arbitrarily chose Oct 11th yea, it triggered, but most people either use month ending or beginning.

            1. Looks like the 10-month will trigger a buy this am about 2% higher than the sell. Also it’s interesting that the curve is compressing towards YTD lows at only +20bps now. False sell signal, or bear market correction? Rates and FX doesn’t seem to be buying this 90 day can kicking. Gonna be interesting. As bearish as I see everyone in my circle I’m not sure I they’re contratian to Wall Street or we’re setting up for a year end rally.

              You guys should read Dual Monentum by Antonacci, he uses a 12 month look back relative and absolute momentum, less false signals.

              1. Well, who knows, maybe the Fed is done after the December hike (2.25-2.5%) or one more in the new year. Then short-term rates are at 2.5-2.75%. The 3% on the 10Y is the new equilibrium and the curve never inverts. 🙂
                I’m telling you, bond traders are smarter than all of us!!! 🙂

                1. Yea it’s a very interesting time. If wages keep rising but inflation and growth keep slowing the Fed will be in a tough spot. I enjoyed your article about buying a house again, thanks.

  18. Great post and fascinating topic. I’m not convinced yet it’s right for me, but I’m definitely subscribed to keep reading! You probably already know this, but Portfolio Visualizer lets you backtest specific ticker symbols back to 1985 – which should account for expenses. Just playing a bit with VFINX, using this strategy in the 2000’s beat buy and hold handily using several starting points, and a lot less painful drawdowns to boot! If I decide to implement, I’d probably just do it on a portion of my 401(k) or Roth IRA, as an additional way of diversifying. I guess that’s the same as being something like 60/40 when the signal is to hold equities, or 40/60 when the signal says to sell. This strategy makes intuitive sense because it capitalizes (in a methodical way) on human psychology and investors making the same mistakes over and over. As I learned in Bernstein’s “Four Pillars of Investing.” that goes back to at least the 1600’s, and probably even earlier. Looking forward to the next one and especially the updated model you hinted at above!

      1. So when you add extra signals, when is it overfitting and when is it not?

        No easy answers, of course, but it’s not intended as a rhetorical question; genuine interested in your thoughts on this.

        1. True. That’s exactly my hesitation. Looks good in past recessions. But throwing in too many additional bells and whistles would risk overfitting.
          There’s no way to exactly determine a prior what’s overfitting. We will find out if was or not the next time we have a 1929-1932-style bear market. 😉

  19. Ern
    Thx for your excellent analysis.
    Lance Roberts first clued me on this approach.
    John Hussman also made this point Back in 2005.
    I am retired and happen to believe that markets are due for a major correction in the next 12-24 months. The penalty for a 50% drawdown is extremely severe. Therefore I moved 100% into 1year treasuries yielding 2.25% and took on a consulting gig. 😀

    1. I agree with you Ven. The penalty for a huge correction is very severe. So many in the FIRE community are passive index Investors and proclaim their “long term” and are 100% in equities. This will end badly for them.
      If you’re already retired and not comfortable with potential risks in equities, the the short term treasuries you bought is a suitable investment for you. You may get ridiculed by those aforementioned articles indexers but I am also very low in equity allocations currently for the exact same reason you are in treasuries, we’re controlling risk.

          1. Is your link evaluated daily? I thought it was 200 day moving average evaluated monthly

            1. Now I look carefully, on March 30 the closing price was above 200 dma but on April 2 below it. So, if you use the closing price of the last day in the month then it didn’t cross…

  20. I wonder what the analysis would look like if you combined both the 10-month S&P average “momentum signal” and the 3 indicators from Part 1 (Yield curve, Unemployment, Mfg PMI). Would some of the false signals go away?

      1. I was looking at the yield curve and I think you are right. It’s the most reliable of the 3.

  21. I’ve been reflecting more on this over the weekend. Whilst I understand some of the criticisms above – in particular the long periods when the strategy under-performed compared to a buy-and-hold – I think there’s a missing element to those criticisms.

    Firstly, this strategy should be seen somewhat as an insurance policy against long bear runs in the market. To that end, as this fantastic analysis shows, it’s very successful at dodging those really bad periods in the economy. The last table shows that visually, with the person using this strategy, “missing out” on all of the Great Depression, Financial Crisis, late 30s recession and Dot-com Bubble successfully (amongst others). With that in mind, its not bad (an understatement?) if the strategy means you avoid the very worst and get a better return to boot? It’s kinda like having house insurance, your house burning down, but you get a much nicer house and more money at the end of it.

    Secondly, we should be mindful of the underperformance we’re talking about.We’re not talking about absolute capital losses. Rather underperformance compared to (another) very successful investing strategy. Even where this strategy struggles (the false positives between the 80s and 90s, for example), we’re not talking about large capital losses.

    Finally, the place where this strategy seems to be more hit-and-miss is where there are very sharp drops in the market: the 1933 drop, 1987 (Black Monday), 1998 (Rouble Crisis), 2010 rollercoaster. Then again, if you know a strategy to time a flash crash, I’d love to hear it!

    1. Very well said! Thanks for weighing in! That’s exactly the way see it too: it’s insurance and just like in the case.of insurance, you “underperform” slightly by a small amount but then reap the benefits every once in a while!

  22. Another question: What would the analysis show if you looked at employing this method for a 30 year or 40 year period? Similar to what the Trinity Study did. How would it affect your SWR?

  23. Bravo ERN! Another thoughtful piece with robust data analysis to articulate your point.

  24. Great job again ERN! This definitely helps the sequence of return risk. I like that you emphasize good defense too. That this happens to deliver similar/better returns to buy and hold is just a bonus.

  25. New to the Blog but loving it!

    A few ideas:

    Why not short a SPY in roughly the same volatility size as your long stock portfolio?

    Alternatively with the margin rate at IB so low and futures taxed as 1256 contracts, short the equivalent notional value in ES?
    For example $1.3m portfolio would be roughly 10 ES contracts?

    Another alternative (not the best idea if looking to protect for a massive correction) a put ratio spread? Rough example—
    Sell 2 2450 puts to finance purchase of 1 ATM a call spread?

    1. No disrespect but I think we’re over analyzing this thing and making it all too complicated. For all the hedging ideas here, how big an account t are we discussing? If it’s a million dollars or more then there’s merit to all these hedging ideas. If we’re talking $25k-$100k all this elaborate heding is laborious and costly.
      Meb Faber is a great analyst but is he a trader? It’s like most of these authors, investment advisories, newsletters, etc. They’re great analysts and present some great ideas but can they make money in the markets???
      Keeping it basic, use the moving average crossovers and move to cash during a sell signal, move back into SPY on a buy signal and leave it alone.

    2. I suggested buying puts and selling calls to establish a collar. If you have a portfolio made up of a few ETFs for this may be fairly easy and the calls will be covered calls and not require margin. The issue with shorting ES as a hedge is you will need to put up cash as margin – overnight initial margin is $8,325 per contract and if the position goes against you, you will need to add cash.

    3. I like the idea of the ES future. Easy to short without much margin and tax efficient! And it avoids realizing capital gains on the highly appreciated
      Not sure everybody wants to get into the Put spread business, but I see the potential for that one! 🙂

  26. Looks like we hit a sell signal today (depending on how you calculate the 10 month sma). Still no inverted yield curve though. I’m holding tight.

    1. I agree. I’m staying put for now. There’s no widespread talk of economic gloom and doom either. Those voices tend to get louder as we approach the cliff.

      If a person wanted to be more cautious they might start moving, say 10% of their funds. If the downward trend continues below the 10 month sma, move more funds…and then more.

    2. Yeah, same here. Depends on the index and how exactly you construct the signal. What if the index goes below the average intra-month? Should we wait until the last trading day?
      Better confirm this with some of the macro signals!!! 🙂

      1. I use fundamental’s and hedges to help with the false signals, but you can also quartile your portfolio and use 4 separate signals, say a 50day, 100day, 200day and 12 month or 300day traded one day a month.

  27. Good article. The MA trend cross is a good start. Reams of studies have been conducted that attempt to analyze single tactically based variables in an isolated fashion. This has been problematic, as outputs have resulted in the production of small statistical likelihoods of positive results and uneven and spurious and “false” signaling, therefore rendering them to limited efficacy. Thus, the quest for a singular, “holy grail” investment variable is futile; a single indicator cannot be expected to generate enough value as to be a viable investment strategy, yet may contribute value when used in conjunction and with others. It is the process of the combining of a few variables in a sequential and cross validation signaling format, that may tease out meaningful and useful results. In order to gain further explanatory power from signaling with less spurious results ( if one wants to hold equity assets for the longest uninterrupted and tax efficient trends), it can be important to combine objective, statistically relevant indicators / data series. Chapter 3, part 2 * describes the conundrum of using “singular” variables as tactical determinants and the use of the moving average applied to a bond fund price series in order to determine allocation into bonds vs cash.

    * ( paste link into browser address bar )

  28. I’m wondering, if companies tend to manipulate share values, typically at the end of quarters, i.e. end of months, would using a date that isn’t the end of the month (e.g. value on the 15th of each month) for calculating the averages and sell and buy signals make any difference, or would it all end up in the noise?

  29. Have you considered the possibility to switch to an inverse ETF when the sell signal is triggered? So, instead of selling our S&P500 ETF for cash or to buy bonds, we buy an inverse S&P500 ETF (such as SH). I’m wondering if the “false positive” outweigh the profits in actual bear markets…

    1. Inverse equity ETFs are expensive (high expense ratios) and due to their daily rebalancing you also have to be careful that your short exposure doesn’t wander off too much from what you actually want. Also, where does the money come from to buy the short ETF? I wouldn’t want to buy a short ETF on margin. 🙂

  30. Adding the yield curve prevents correctly timing the 1987 drop. The yield curve had last inverted in July of 1982. It helps during the recent bull market but doesn’t seem to add much during any other time frame.

    Note: I prefer using a 5% envelope around the moving average to reduce transactions and whipsawing. As a result I’m not simulating exactly the same thing ERN has.

  31. I wonder how this strategy looks like compared to the dollar cost averaging when still in accumulation phase… It looks to me you’ve made this calculation only using a single investment in the beginning?
    Also selling and writing off losses to reduce income tax is another topic that could be leveraged in a bear market during accumulation phase…

    1. Good point! I haven’t done that exercise yet. My suspicion is that market timing is a little bit less useful for folks who start out with a zero initial portfolio and keep adding. In exchange, there is an added benefit for people who are withdrawing during retirement.

  32. Hey ERN, forgive me if you addressed this in another post: Did you implement a momentum investing strategy for your portfolio? If so, how much of your portfolio did you allocate, and what are your thoughts?

    1. I have not and I don’t plan to either. I am working on a post on how to implement this in the safe withdrawal calculations, though. Mostly out of academic curiosity.
      Personally, I don’t recommend the momentum approach for the accumulation phase. But there might be some limited use while retired.

  33. Protecting against a stock market crash is great and all, but it looks like this strategy would’ve actually underperformed in the historical worst case retirement (1965-1995) so it might not actually be a good idea.

    1. Akash,

      I don’t know how you can say that – in fact this approach worked very well in raising the SWR in the 1965 time period.

      Using gold instead of bonds as the ballast for stocks works even better still.

      I implemented the 10 MMA in a hacked version of ERN’s SWR sheet; see it here:

      1. My statement was only for switching into bonds or cash, as described in the post. Obviously any strategy involving gold will vastly improve your 1965 retirement experience…

        1. If you hack the hacked sheet and overwrite the bonds+gold mashup on the additional tab with the bonds numbers, you still get a 3.88% SWR for the 10 MMA approach, an 8%+ improvement over the static 75/25 bonds portfolio.

          As ERN writes in the post above, the 3 time periods between 1969 and 1982 in the first 17 years of retirement withdrawals where the timing system has big savings for that November 1965 cohort outweigh the multiple small losses over that 30 year period.

          You are of course quite correct that it is a separate point that using gold instead of bonds as the balance/ballast for stocks – whether using a tactical asset allocation (market timing) system or a static allocation approach – has provided superior results since 1950.

          Obviously any strategy using bonds is superior during deflationary periods, especially from 1929 to 1945, and will vastly improve your retirement experience… So if you think conditions today and for the next several years are more similar to 1929, 1937 or the 19th century and run a significant risk of deflation, you should surely include a big heaping of bonds in your allocation.

          1. I’ve posted the slight modification to the hacked sheet which allows the use of bonds as the ballast for stocks for the 10 MMA.


            As I noted, using the 10 MMA with 10-year bonds as the ballast, you get a SWR of 3.88% (the low being December, 1968; November 1965 is actually slightly better)

        2. Granted. The momentum/timing worked better or worse with different alternative assets. I guess the ranking is gold>cash>bonds. But even with bonds wou’d have done better than doing nothing.

      2. Yeah, completely agree. The nice thing about those momentum strategies is that they worked very well in the past recessions and bear markets. There might be a bit of whipsaw risk, but the deep recessions, 1930, 1970, 2001-3, 2007-9 this would have worked well. Not sure why the other commenter said this.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.