Direct Answer
Direct answer: Time-series momentum (TSMOM) is a return strategy that goes long asset classes with positive trailing 12-month returns and short (or reduces to zero) those with negative trailing 12-month returns, scaled to a target volatility level. Across equities, bonds, commodities, and currencies, TSMOM has delivered positive returns in nearly every year over a century of data, with its strongest performance during severe equity bear markets — a property called crisis alpha. This is the key distinguishing feature: unlike most diversifiers (which tend to have low correlation with equities in normal markets but higher correlation during crises), trend-following generates its highest returns during prolonged equity drawdowns, precisely when traditional diversification tends to fail.
The evidence for cross-asset TSMOM is documented in Moskowitz, Ooi, and Pedersen (2012) across 58 markets and 25+ years, and in academic and practitioner research extending back to 1880 across available historical data. The strategy is implemented institutionally through CTA (commodity trading advisor) funds and managed futures programs. As an overlay on a 60/40 portfolio, a 10–20% allocation to a diversified trend-following program has historically reduced maximum drawdown and improved Sharpe ratio without significantly reducing long-run expected return.
Key Takeaways
- TSMOM has positive returns in most years across all major asset classes: The strategy earns by being long trending assets and short or neutral declining assets, capturing return in both up and down markets as long as trends persist.
- Crisis alpha is the defining feature: TSMOM earns its strongest returns during prolonged equity bear markets (2000–2002, 2008–2009, 2022) because equities trend down for long enough for the strategy to establish and hold a short or underweight equity position.
- The average correlation with equities is near zero, but the tail correlation is negative: This negative tail correlation is more valuable than zero average correlation because diversification is most needed in downturns.
- Momentum crashes are real and must be sized for: When a prolonged trend reverses sharply (as equities did in March 2009 and March 2020), TSMOM gives back significant gains in a short period. These crashes are correlated with liquidity crises and typically last 1–3 months before the new trend re-establishes.
- Volatility-scaling is essential: Raw TSMOM without volatility scaling concentrates risk in the most volatile asset classes. Scaling position size inversely to recent volatility equalizes risk contribution across asset classes and improves Sharpe ratio.
- Longer lookback periods reduce momentum crash risk: 12-month lookback TSMOM has better crisis alpha but higher crash risk than 3-month or 1-month lookback. Blending lookback periods (e.g., 1, 3, 6, 12 months) smooths the return profile.
- Implementation requires liquid futures markets: TSMOM is most effectively implemented using liquid exchange-traded futures (equity index futures, Treasury futures, commodity futures, currency futures) rather than ETFs, because futures allow economical short exposure and leverage without margin loans.
- Overlay size of 10–20% is typical in institutional portfolios: Larger allocations reduce equity beta meaningfully, which helps in bear markets but also drags in strong equity bull markets. The overlay is sized to improve the portfolio Sharpe ratio, not to maximize crisis alpha at the cost of normal-market return.
Core Concepts
1. Time-Series Momentum: Definition and Mechanism
Time-series momentum (TSMOM) for a single asset is defined as: if the asset's return over the past 12 months (excluding the most recent month) is positive, go long the asset in the next month; if negative, go short or be neutral. This is a trend-following rule: it bets that assets trending upward will continue upward and assets trending downward will continue downward. The rule is applied simultaneously to all available asset classes, and position sizes are scaled to equalize risk contribution across assets (each asset contributes the same target volatility to the total TSMOM portfolio, typically 40–60% annualized volatility).
The mechanism generating TSMOM returns has two components. First, initial underreaction: investors process information about fundamental changes (rising interest rates, improving corporate earnings, commodity supply disruption) slowly, causing prices to trend gradually toward fair value rather than adjusting immediately. Second, feedback momentum: price increases attract more buyers (momentum-chasing investors), further extending the trend beyond its fundamental value before eventual reversal. Both mechanisms predict return persistence over 1–12 month horizons.
The mathematical return of TSMOM is approximately: r_TSMOM ≈ |r_signal| × sign(r_signal), where r_signal is the trailing 12-month return. Assets with large positive or large negative trailing returns contribute the most to TSMOM, because they represent strong trends that are most likely to persist and because they are sized proportionally to their momentum signal strength in the typical implementation. Assets with returns near zero (no clear trend) contribute little to the TSMOM portfolio.
2. Cross-Sectional Momentum vs. Time-Series Momentum
Cross-sectional momentum (CSMOM) ranks asset classes by their trailing returns and goes long the top-ranked and short the bottom-ranked, regardless of the sign of returns. If all asset classes had positive returns, CSMOM would still short the worst performer. This is a relative performance strategy: it bets that recent relative leaders will continue to outperform recent relative laggards.
TSMOM takes absolute positions: an asset with a negative 12-month return is short regardless of whether other assets also had negative returns. This makes TSMOM more natural for generating crisis alpha: in a widespread equity bear market, all equity markets may have negative 12-month returns, and TSMOM goes short all of them. CSMOM would still be long some equity markets (the best-performing ones) relative to others, providing less protection.
In practice, most trend-following programs (CTAs, managed futures) implement both TSMOM and CSMOM across different lookback periods and blend them. CSMOM adds diversification because its returns depend on relative performance differences, which are present even in trending markets where all assets move in the same direction.
3. Crisis Alpha: The Evidence
Hamill, Rattray, and Van Hemert (2016) documented TSMOM returns during the five largest equity bear markets since 1880: the 1929 crash, the 1937–1938 recession, the 1973–1974 oil crisis, the 2000–2002 dot-com bust, and the 2008–2009 financial crisis. In each period, a long-only TSMOM strategy across asset classes delivered positive cumulative returns, while a 60/40 equity/bond portfolio suffered large drawdowns. The average TSMOM return during these five episodes was approximately +18% while equities averaged approximately −40%.
The mechanism of crisis alpha is as follows: equity bear markets associated with recessions or financial crises tend to be prolonged (12–36 months), giving trend-following strategies time to (a) exit long equity positions as the trailing 12-month return turns negative, (b) establish short or neutral equity exposure, and (c) simultaneously go long defensive assets (Treasuries, gold, Japanese yen, US dollar) that also trend during crisis periods. The combination — short equities, long safe havens — is the classic trend-following crisis position and is automatically assembled by any rule-based TSMOM program applied across a diversified futures universe.
The 2022 crisis was particularly interesting for trend-following: equities and bonds both declined (eliminating the traditional diversification from bonds), but commodity and currency trends were strong. A diversified TSMOM program was long energy, short equities, long US dollar — a combination that delivered strongly positive returns in 2022 while a 60/40 portfolio lost approximately 16%. This highlighted trend-following as a potential diversifier in inflationary environments where bond/equity correlation is positive.
4. Constructing a Trend-Following Overlay
A simplified trend-following overlay on a policy portfolio works as follows: maintain the full policy portfolio as the baseline; allocate a portion of portfolio value (typically 10–20%) to a trend-following futures program; fund this allocation from cash or a modest underweight in the most liquid policy assets (Treasuries or equity index futures). The overlay is delta-hedged against the policy portfolio where it has offsetting positions: if the overlay is short equity futures and the policy portfolio holds equity ETFs, the equity exposure is reduced by the amount of the futures short.
Position sizing in the overlay is typically based on volatility scaling: each futures position is sized so that its daily return contributes a target annual volatility of Y% to the overlay. With 10 asset classes each contributing 5% volatility and pairwise correlations averaging 0.2, the total overlay volatility is approximately sqrt(10 × 5² + 10×9 × 0.2 × 5²) = sqrt(250 + 450) = sqrt(700) ≈ 26% (annualized) — a high-volatility standalone strategy that is the reason it is sized as a small overlay (10–20% of portfolio) rather than a large allocation.
Implementation choices for the overlay: (1) dedicated CTA allocation — allocate 15% to a managed futures manager who runs a diversified trend-following program across 50–100 futures markets; (2) in-house replication — construct a rule-based trend-following program using 10–20 liquid futures contracts; (3) trend-following ETF — low-cost but typically with lower leverage and fewer markets than dedicated CTAs. Each approach has different cost, liquidity, and operational characteristics; institutional investors typically use dedicated CTAs, while individuals use trend-following ETFs.
5. Drawdown Profile and Momentum Crashes
The trend-following drawdown profile has a distinctive shape: relatively small drawdowns in most periods (the strategy is diversified across many uncorrelated asset class positions) with occasional sharp, short drawdowns during momentum crashes. Momentum crashes occur when a consensus trend reverses abruptly — typically when a central bank intervention, major policy announcement, or sudden regime shift forces rapid repricing. Examples: March 2009 (equity rally after stimulus announcement reversed short equity positions), May 2010 (Flash Crash), March 2020 (COVID crash and then rapid reversal within weeks).
The duration of momentum crash drawdowns is characteristically short (1–6 months) compared to equity bear market drawdowns (12–36 months). Recovery is typically swift because the new trend quickly establishes itself and the strategy adapts. The maximum drawdown of a diversified trend-following program over a 20-year period has historically been in the range of 15–25% — similar to a diversified bond portfolio, but with much higher annualized returns. This risk profile is the basis for including trend-following as a diversifying allocation rather than a risk-reducing one: it does not reduce total portfolio volatility much, but it substantially improves the return profile during equity bear markets.
Worked Scenario
Adding a 15% trend-following overlay to a standard 60/40 policy portfolio; funded by reducing equity to 51% and bonds to 34%.
- Base policy: 60% equity (S&P 500 proxy), 40% bonds (US Agg proxy). Historical Sharpe ratio ≈ 0.60. Maximum drawdown 2008–2009: −27%.
- Overlay: 15% CTA/TSMOM (diversified trend across equities, bonds, commodities, currencies, interest rates). Historical Sharpe ratio of CTA index ≈ 0.65. Correlation with 60/40: approximately +0.02 in normal markets, approximately −0.45 during equity bear market months.
- New portfolio: 51% equity, 34% bonds, 15% CTA. Expected return ≈ 0.51×7.0% + 0.34×4.5% + 0.15×6.0% = 3.57 + 1.53 + 0.90 = 6.0%. Slightly below the pure 60/40 expected return of 0.60×7.0 + 0.40×4.5 = 4.2 + 1.8 = 6.0%. Return is maintained (CTA expected return roughly equivalent to bonds it replaced).
- Volatility: 60/40 portfolio volatility ≈ 11%. Adding CTA overlay with near-zero average correlation: portfolio volatility ≈ sqrt(0.85²×11² + 0.15²×22²) ≈ sqrt(87.4 + 10.9) ≈ 9.9%. Slight volatility reduction from diversification.
- Crisis performance simulation (2008-style): equity falls 50%, bonds +5%, CTA +18%. Modified portfolio: 0.51×(−50%) + 0.34×5% + 0.15×18% = −25.5% + 1.7% + 2.7% = −21.1% vs. 60/40 portfolio: −27%. Maximum drawdown reduction: approximately 6 percentage points.
- 2022-style scenario (equity −20%, bonds −13%, CTA +22%): modified portfolio: 0.51×(−20%) + 0.34×(−13%) + 0.15×22% = −10.2% − 4.4% + 3.3% = −11.3% vs. 60/40 portfolio: −17.2%. Significant improvement in the scenario where bonds do not provide their traditional protection.
- Govern the overlay: specify in IPS that CTA allocation targets 15% ±5% with rebalancing band; document the crisis alpha rationale; commit to minimum 3-year evaluation horizon before adjusting.
Measurement Framework
| Measurement | What it tells you |
|---|---|
| Correlation with 60/40 portfolio (rolling 36-month) | Is the trend-following overlay providing its expected diversification? Average correlation near zero is desirable; persistently positive correlation (+0.3 or above) suggests the overlay is no longer providing the intended diversification. |
| Trend-following Sharpe ratio (rolling 3-year) | Is the overlay generating risk-adjusted returns comparable to its historical long-run average? Evaluate over 3-year minimum given the high year-to-year variability of trend-following returns. |
| Portfolio max drawdown improvement | Has adding the overlay reduced the portfolio's maximum drawdown (or projected drawdown in stress scenarios) relative to the base policy portfolio? |
| Trend-following drawdown (current) | Is the overlay currently in drawdown? How deep and how long? Drawdowns of 10–15% are normal; drawdowns exceeding 20% may warrant a management review of signal quality or implementation. |
| Implementation cost (management fee + trading costs) | What is the all-in cost of the CTA allocation? Net crisis alpha must justify the fee load. Typical CTA fees: 1–2% management, 20% performance fee for active managers; trend ETFs: 0.65–1.0% all-in. |
| Crisis performance vs. target | During months when equity returns were below −5%, did the trend-following overlay deliver positive returns? This is the primary test of whether crisis alpha is functioning as expected. |
Common Failure Modes
Exiting the Overlay During a Momentum Crash
The most common implementation failure for trend-following overlays is abandoning the strategy after a sharp short-term drawdown — exactly what the strategy is designed to deliver: short drawdowns followed by recovery. A CTA that has returned +20% during an equity bear market (2008, 2022) and then loses −12% in the first month of the equity recovery (as short equity positions are forced to close into a rising market) has behaved exactly as expected. Investors who exit at this moment lock in the loss and miss the next trend cycle.
The governance solution is to commit in writing — in the IPS — to a minimum holding period for the trend-following allocation (typically 3 years) and to specify that short-term drawdowns (up to 20%) within the trend-following sleeve are an expected feature of the strategy, not a signal of strategy failure. Review criteria should be based on multi-year information ratios, not short-term performance.
Confusing Trend-Following with Momentum Factor in Equities
Cross-asset trend-following (TSMOM) is conceptually related to but mechanically different from equity cross-sectional momentum (the factor described in the prior guide). Cross-asset TSMOM goes long entire asset classes that are trending; equity cross-sectional momentum goes long individual stocks that have outperformed within the equity universe. The two strategies have different return profiles: TSMOM provides crisis alpha (positive during equity bear markets); equity momentum provides returns primarily in strong trending equity markets but crashes during equity market reversals. Both strategies are worth including in a diversified allocation, but they should not be confused or treated as substitutes.
Underestimating the Drag in Strong Bull Markets
In a strong, one-directional equity bull market with low volatility (2012–2019), a diversified trend-following overlay will lag significantly. The overlay is long equities (consistent with a rising 12-month return) but also holds positions in many other asset classes, and any diversification away from pure equity exposure reduces return in a year where equity returns dominate. Trend-following added approximately −1.5% per year to a 60/40 portfolio during the 2012–2019 period of strong equity performance. Investors must accept this cost as the carry of the crisis alpha option — they are paying in foregone return during good times for the protection in bad times.
Over-Leveraging the Trend-Following Position
CTA programs typically run at 3–5× gross notional leverage (each $1 of capital controls $3–5 of futures positions) to achieve their target volatility with the diversification discount of the multi-asset portfolio. This leverage is safe in normal markets because the positions are in liquid futures that can be unwound instantly. But in severe liquidity crises (March 2020), even liquid futures can gap dramatically, and highly leveraged trend-following positions can lose 20–30% in a single week. The risk is not leverage itself but the combination of leverage with inadequate position limits and insufficient stress-testing. Institutional CTA programs have sophisticated risk management; retail trend-following ETFs typically run at lower leverage (1–2×) for this reason.
Frequently Asked Questions
What is a CTA and how does it relate to trend-following?
CTA stands for commodity trading advisor — the US regulatory designation for investment managers who trade futures on behalf of clients. Most CTAs implement trend-following (momentum) strategies across diversified futures universes including equity index futures, interest rate futures, commodity futures, and currency futures. The term "CTA" is often used interchangeably with "managed futures" and "systematic trend-following" in the institutional allocation context, though some CTAs use non-trend strategies (carry, mean reversion). When investors refer to CTA allocation as a portfolio diversifier, they typically mean diversified systematic trend-following programs, not discretionary or non-trend futures strategies.
Why is trend-following called an "alternative" strategy?
Trend-following is classified as an alternative strategy because its return profile does not fit neatly into either the equity or the fixed income category. It uses leverage and short positions (unlike traditional long-only funds), trades futures rather than equities or bonds directly, has near-zero long-run correlation with traditional asset classes, and generates its best returns in market environments that are worst for equity investors. These characteristics make it genuinely "alternative" — additive to traditional portfolios rather than a substitute for any existing position. It is distinct from "alternative risk premia" or "liquid alternatives" ETFs, which are typically lower-leverage, lower-return implementations of the same concepts.
How much should a typical institution allocate to trend-following?
Most institutional allocations to dedicated trend-following programs range from 5% to 20% of total portfolio. The optimal allocation depends on the investor's primary concern: a 5% allocation provides modest crisis alpha with minimal normal-market drag; a 20% allocation provides strong crisis alpha but meaningful drag in strong equity bull markets. The typical institutional recommendation from academic research (e.g., Hurst, Ooi, Pedersen 2017) is that adding 10–15% trend-following to a 60/40 portfolio improves the Sharpe ratio and reduces maximum drawdown without materially changing the expected return. Larger allocations are justified for investors with explicit tail-risk mandates or for liability-driven investors who cannot afford large drawdowns.
Does trend-following work with ETFs or only with futures?
Trend-following principles can be applied using ETFs (e.g., Faber's 10-month moving average applied to asset class ETFs), but with meaningful differences from futures-based implementations. ETF-based trend-following: (1) cannot efficiently go short (requires separate inverse ETFs with negative carry); (2) incurs bid-ask costs and potential market impact on large positions; (3) may have capital gains tax consequences in taxable accounts when ETF positions are sold; (4) typically cannot use leverage efficiently. Futures-based trend-following avoids all these problems but requires a futures account, collateral management, and roll management. For most individual investors, trend-following ETFs (e.g., iMGP DBi Managed Futures Strategy ETF) are the practical implementation, accepting the higher cost and lower purity of futures replication.
Has trend-following become crowded and less effective?
This is a legitimate concern. CTA assets under management grew from approximately $200 billion in 2000 to over $400 billion by 2022. More capital chasing the same signals can erode the premium by moving prices before the strategy can benefit. However, the evidence for material crowding-related degradation is mixed: trend-following returns were weak from 2010–2019 (suggesting possible crowding or a low-trend environment) but recovered strongly in 2022 (suggesting the strategy still functions when trends are present). The consensus view: the strategy works when assets trend; crowding is a real but second-order effect. The primary determinant of performance is whether markets trend, not whether many investors are following trends.
What asset classes are included in a typical diversified trend-following program?
A diversified trend-following program typically includes 50–100 futures contracts across five broad sectors: (1) equity index futures (S&P 500, Eurostoxx, Nikkei, FTSE, Hang Seng, etc.); (2) fixed income futures (US 10-year Treasuries, German Bunds, UK Gilts, Japanese JGBs, etc.); (3) commodity futures (WTI crude, Brent crude, natural gas, gold, silver, copper, corn, wheat, soybeans); (4) currency futures (EUR/USD, GBP/USD, JPY/USD, AUD/USD, CHF/USD, etc.); (5) short-term interest rate futures (Eurodollar, SOFR). The breadth of markets is a key advantage: uncorrelated trend signals across 50+ markets provide significant diversification within the trend-following program itself.
How is trend-following different from a simple moving average strategy?
A simple moving average (SMA) strategy on a single asset (e.g., buy when price is above its 200-day SMA, sell when below) is a single-asset time-series momentum rule. Cross-asset TSMOM is the generalization: the same rule applied simultaneously to many uncorrelated asset classes, with position sizes risk-weighted to equalize volatility contributions across markets. The diversification across many uncorrelated positions is what gives TSMOM its distinctive risk profile — less dependent on any single market's trend. A single-asset SMA strategy has all its risk in one market's trending behavior; a 50-market TSMOM program diversifies across 50 simultaneous trend bets, most of which are uncorrelated with each other.
What happened to trend-following in 2022?
2022 was one of the best years for diversified trend-following strategies in the past two decades. The SG CTA Index returned approximately +26% in 2022 while the 60/40 portfolio returned approximately −16%. The key trends that drove CTA performance: rising energy prices (long crude oil, natural gas), rising interest rates (short bond futures), US dollar appreciation (long USD vs. EUR, JPY, GBP, AUD), and declining equity markets (short equity index futures). These four trends were all simultaneous and persistent — exactly the environment where cross-asset TSMOM generates its strongest returns. The 2022 episode provided the strongest out-of-sample validation of the crisis alpha thesis since 2008.
Sources and Further Verification
- Moskowitz, T.J., Ooi, Y.H., & Pedersen, L.H. (2012). "Time Series Momentum." Journal of Financial Economics, 104(2), 228–250. Foundational documentation of TSMOM across 58 liquid futures markets. doi.org/10.1016/j.jfineco.2011.11.003
- Hurst, B., Ooi, Y.H., & Pedersen, L.H. (2017). "A Century of Evidence on Trend-Following Investing." Journal of Portfolio Management, 44(1), 15–29. Historical evidence from 1880 to 2012. doi.org/10.3905/jpm.2017.44.1.015
- Hamill, C., Rattray, S., & Van Hemert, O. (2016). "Trend Following: Equity and Bond Crisis Alpha." Working paper. AQR Capital Management.
- Faber, M. (2007). "A Quantitative Approach to Tactical Asset Allocation." Journal of Wealth Management, 9(4), 69–79. Simple trend-following applied to asset class ETFs.
- SG CTA Index. SocGen maintains the Societe Generale CTA Index as the industry benchmark for managed futures performance. Available at sgcib.com.
Educational Disclaimer
This guide is for educational purposes only. Trend-following strategies involve leverage, short exposure, and futures contracts that carry specific risks including margin calls and potential losses exceeding invested capital. Past crisis alpha does not guarantee future protection. This is not investment advice. Consult a qualified financial professional before allocating to managed futures or trend-following programs.