Strategies · Trend and momentum
Dual momentum: relative and absolute momentum combined, tested
Dual momentum picks the stronger of US and international stocks and moves to safety when both are falling. Exact rules, a 24-year backtest with costs, and where it lost.
On this page
Dual momentum joins two rules that this section tests separately. Relative momentum chooses between assets: hold the one that has done better. Absolute momentum decides whether to hold anything at all: stay in only if the chosen asset has itself gone up. Gary Antonacci combined them in a model he called Global Equities Momentum, which switches between US stocks, international stocks and bonds once a month.
The appeal is that each rule covers the other's weakness. A rotation is always invested and falls with the market. A time-series momentum rule can leave the market but holds only one thing. This lesson states the combined rule exactly and tests it on SPY and EFA from 2002 to 2026. It did what it promises in 2008. Over the whole period it earned 2.4 points a year less than SPY, and since 2009 less than half as much.
The idea
Take the two questions in order.
Which? Compare the 12-month return of US stocks with that of international stocks. The stronger one is the candidate. This is relative momentum, and it assumes that leadership between regions lasts for a while. It did in the 2000s, when international stocks led for years.
Whether? Check that the candidate's own 12-month return is positive. If not, both markets are down over the year, and the rule holds something safe until one of them recovers. This is absolute momentum, and it is the part that limits large losses.
Antonacci's model measures the second question against the return on Treasury bills and holds a broad bond fund when stocks fail the test. The version tested here compares with zero and holds cash. The reason is the data: the price history available for these lessons leaves out dividends and interest, and a bond fund's interest is nearly all of its return. A bond fund tested on price alone would be misrepresented, so the main test uses cash, and a bond variant is shown separately with that warning.
The rules
Two funds: SPY (the S&P 500) and EFA (developed markets outside North America). Checked once a month, on the last trading day:
- Compute each fund's 12-month return: today's close divided by the close twelve months earlier, minus one.
- Take the fund with the higher 12-month return.
- If that return is above zero, hold that fund for the next month, with the whole amount.
- If it is zero or below, hold cash for the next month.
- Trade only when the answer differs from last month's.
Because the stronger fund is tested, cash means that both funds are down over the past year.
A worked example
At the end of January 2008 the 12-month returns were −4.4% for SPY and −2.6% for EFA. EFA was the stronger of the two, but its return was below zero, so the rule moved to cash. SPY closed that month at $137.37.
It held cash for 21 months. At the end of October 2009 SPY was up 7.0% over twelve months and EFA up 19.6%. EFA was stronger and positive, so the rule bought EFA. Two months later SPY led, 23.5% to 23.2%, and the rule switched to SPY at $111.44.
The 0.3 point gap that triggered that switch is typical. When two stock markets move together, their 12-month returns are often nearly equal, and the lead changes on small differences.
What the backtest shows
The test, stated in full:
- Data. Month-end closes of SPY and EFA. EFA begins in September 2001, and twelve months supply the first comparison, so the test runs from September 2002 to October 2026: 290 months.
- Trading. The signal and the trade both use the month-end close.
- Costs. 0.10% of the amount on each sale and each purchase. A switch between the two funds is one of each.
- Left out. Dividends, interest on cash and taxes. EFA's dividend yield is higher than SPY's, so leaving dividends out understates EFA a little more. The Treasury bill hurdle of the original model is replaced by zero.
- The comparison is SPY bought once and held.
| Measure | Dual momentum | SPY, held |
|---|---|---|
| Total return | 441.9% | 824.0% |
| Return per year (CAGR) | 7.3% | 9.7% |
| Deepest fall from a peak | −25.0% | −52.2% |
| Longest time below a peak | 79 months | 64 months |
| Time invested | 80% | 100% |
| Changes of holding | 46 | 0 |
The rule held SPY in 51% of months, EFA in 29% and cash in 20%. Its 46 changes of holding cost about 0.3 points a year: with no costs it returns 7.6%.
Its longest stretch below an earlier peak, 79 months, was longer than SPY's 64. The rule fell less in 2008 and then took longer to make a new high, because it re-entered late in 2009 and lost 12.8% in 2011.
The two halves of the period
| Period | Dual momentum | SPY, held |
|---|---|---|
| September 2002 – February 2009 | 10.1% a year | −1.6% a year |
| February 2009 – October 2026 | 6.2% a year | 14.1% a year |
Through the 2009 low the rule was in its element. International stocks led from 2003 to 2007, so the relative half paid: the rule gained 17.2% in 2004, 11.3% in 2005 and 23.2% in 2006 while SPY gained 8.6%, 3.0% and 13.7%. Then the absolute half took it to cash for all of 2008 after January.
After the low, both halves hurt. US stocks led for most of 17 years, and every fall was short.
Which half did the work?
Running each half alone on the same months shows where the result comes from.
| Rule | Return per year | Deepest fall |
|---|---|---|
| SPY held throughout | 9.7% | −52.2% |
| EFA held throughout | 5.0% | −59.7% |
| Relative only: the stronger of the two, always held | 9.3% | −54.0% |
| Absolute only: SPY when its 12-month return is above zero | 7.5% | −20.8% |
| Dual momentum: both | 7.3% | −25.0% |
Choosing between the two funds, with no cash option, returned 9.3% a year, a little less than holding SPY alone, with the same deep fall. Adding the cash rule to SPY alone cut the fall to 20.8%. Putting the two together did slightly worse than the absolute rule by itself on both measures. In this period, on these two funds, the relative half was a small cost and the absolute half did all of the protecting.
Where it fails: the losing periods
Fast falls and fast recoveries. The absolute test has the same weakness as every trend rule.
| Moved to cash | Bought back | Result |
|---|---|---|
| September 2011, SPY $113.15 | October 2011, SPY $125.50 | Bought back 10.9% higher |
| December 2018, SPY $249.92 | February 2019, SPY $278.68 | Bought back 11.5% higher |
| March 2020, SPY $257.75 | May 2020, SPY $304.32 | Bought back 18.1% higher |
| April 2022, SPY $412.00 | April 2023, EFA | A year in cash; SPY was 1.0% higher |
In 2020 the rule lost 1.8% while SPY gained 16.2%: it sat through the fall of February and March, sold at the March close, and missed April and most of May.
2011. The rule lost 12.8% in a year when SPY was flat. It was in EFA from April to July while Europe's debt crisis pulled international stocks down, switched to SPY in July, moved to cash in September after the August fall, and bought back in October 10.9% higher.
Switching on noise. Between June 2017 and April 2018 the rule changed funds six times. On three of those dates the two 12-month returns were within half a point of each other. Each switch cost two commissions and decided nothing.
The recovery years. In 2009 the rule gained 3.4% and SPY 23.5%. In 2023 it gained 5.8% and SPY 24.3%. A 12-month lookback needs most of a year of recovery before it turns positive.
Other limits:
- Taxes. Forty-six changes of holding realize gains often, many within a year, which a US taxable account taxes as ordinary income (IRS Topic 409).
- The published record ends early. The model was published in a book in 2014, after circulating as a working paper. Most of the second row of the split table came after publication.
- The safe asset is not tested properly here. With IEF, a fund of 7 to 10 year Treasuries, in place of cash and counted on price alone, the rule returns 7.7% a year from August 2003 against 7.5% for the cash version over the same months. The interest IEF paid is missing from that number, so the true gap would be wider; and in 2022 the bond fund's price fell together with stocks.
Neighbouring lookbacks
| Lookback | Return per year | Deepest fall | Portfolio turned over |
|---|---|---|---|
| 3 months | 8.8% | −21.1% | 2.7 times a year |
| 6 months | 7.0% | −27.4% | 1.8 times a year |
| 9 months | 6.3% | −25.0% | 1.7 times a year |
| 12 months | 7.3% | −25.0% | 1.5 times a year |
The results range from 6.3% to 8.8% a year and do not line up: the shortest lookback is best here, which is the opposite of what the same table shows for time-series momentum on SPY alone. Unordered results from neighbouring settings are what luck looks like. None of the four matched SPY's 9.7%.
Compared with its neighbours
| Strategy | What it decides | Can hold cash |
|---|---|---|
| Relative strength rotation | Which fund | No |
| Time-series momentum | Whether to hold one fund | Yes |
| Moving average trend following | Whether to hold one fund | Yes |
| Dual momentum | Which fund, and whether | Yes |
Dual momentum's worst fall here, 25.0%, is in the same range as the single-fund trend rules on SPY. What it adds is the choice of region, and that choice is only worth something in a period when leadership between regions lasts, as it did from 2003 to 2007.
How Finabase does it
No Finabase tool runs dual momentum or tells you what to hold. These let you look at it.
- Strategy. The Strategy tool takes if/then allocation rules with weights, the form this rule has, and replays them over two years of real prices. Two years holds only a handful of this rule's switches, and the tool charges no trading costs, so it shows behaviour, not merit.
- Chart. The Rate of Change (ROC): percentage momentum and how to read it study with a period of 52 on a weekly chart is the 12-month return. Open it on SPY and on EFA, with the links above, to see both of the rule's inputs.
- Backtest. The Backtest tool trades indicator signals on one symbol, buys a fixed $10,000 on each and charges no costs. It cannot test a rule that chooses between two funds.
Frequently asked questions
Does dual momentum still work?
In this test it has trailed SPY since 2009, earning 6.2% a year against 14.1%. Its record before 2009 was strong. Whether it "works" depends on whether the coming years contain a long decline and a lasting change of regional leadership, which nobody can know.
What does dual momentum hold when stocks are falling?
The published model holds a broad US bond fund. The test on this page holds cash, because the data here cannot show a bond fund's interest. Either way, the safe holding is used only when both stock funds are down over twelve months, which was 20% of the months tested.
What is the difference between dual momentum and a simple rotation?
A rotation always holds the strongest fund, even when every fund is falling. Dual momentum adds a test of the strongest fund against zero and steps aside when it fails. In this test that step was the source of all the protection.
Why a 12-month lookback?
It is the lookback of the published model and of most momentum research, and it trades least. On these two funds a 3-month lookback happened to do best, which mostly shows how much one period's result depends on chance.
Summary
Dual momentum holds the stronger of two stock funds when its own 12-month return is positive and cash when it is not. On SPY and EFA from 2002 to 2026 it returned 7.3% a year against 9.7% for SPY, with a deepest fall of 25.0% against 52.2%. It was far ahead through 2009, when international stocks led and then both markets collapsed. Since then US stocks have led and falls have been short, and both parts of the rule have cost return. The protection came from the absolute test; the choice between regions added nothing in this period.
Check your understanding
4 questions on this lesson. Each answer comes with the reason, and nothing is scored.
At a month-end the 12-month returns are −4.4% for the US fund and −2.6% for the international fund. What does dual momentum hold next month?
- The international fund, because it is stronger
- Cash, because the stronger fund is still below zero
- Half of each fund
Show the answer
B. Cash, because the stronger fund is still below zero The rule first picks the stronger fund and then tests it against zero. A stronger fund that is down over the year fails the absolute test. These were the returns in January 2008.
Which part of the rule produced the smaller worst fall in this test?
- Choosing the stronger of the two funds
- Moving to cash when the chosen fund is down over twelve months
- The 0.10% trading cost
Show the answer
B. Moving to cash when the chosen fund is down over twelve months Choosing between the funds without a cash option fell 54.0%, about the same as SPY. The absolute test alone cut the fall to 20.8%. The protection came from being able to leave the market.
In 2020 the rule lost 1.8% while SPY gained 16.2%. What happened?
- It held international stocks all year
- It sold at the March close after the fall and bought back in May 18.1% higher
- It was in cash all year
Show the answer
B. It sold at the March close after the fall and bought back in May 18.1% higher A 12-month lookback reacts after a sharp fall has already happened. When the fall reverses within weeks, the rule takes the loss and misses the rebound.
Between June 2017 and April 2018 the rule switched funds six times, three of them when the two returns were within half a point. What does this illustrate?
- Relative momentum reacts to tiny differences when two markets move together, and each reaction costs money
- The rule finds many good opportunities
- International stocks were about to outperform
Show the answer
A. Relative momentum reacts to tiny differences when two markets move together, and each reaction costs money A ranking has no sense of how large a lead is. Two funds with nearly equal returns swap places on noise, and every swap is a sale and a purchase.
Sources and further reading
- Gary Antonacci, “Risk Premia Harvesting Through Dual Momentum,” first circulated as an SSRN working paper; Journal of Management & Entrepreneurship 11(1), 2017
- Gary Antonacci, Dual Momentum Investing (McGraw-Hill, 2014) — the Global Equities Momentum model
- Tobias J. Moskowitz, Yao Hua Ooi and Lasse Heje Pedersen, “Time Series Momentum,” Journal of Financial Economics 104(2), 2012, pp. 228–250
- IRS — Topic no. 409, Capital gains and losses
- Finabase chart data — monthly bars for SPY, EFA and IEF, September 2001 to October 2026; rules and results computed by Finabase
