Factor Momentum Across Asset Classes: An Original Backtest

Factor Momentum Across Asset Classes: An Original Backtest

Momentum is one of the most robust anomalies in finance. Stocks that have performed well over the past 3 to 12 months tend to continue outperforming, while recent losers tend to keep losing. But does momentum exist at the factor level? If value, momentum, quality, and low-volatility factors themselves exhibit persistence, a strategy that goes long recently winning factors and short recently losing factors could generate returns independent of any single factor exposure. This article presents Quant Decoded's original backtest of factor momentum across equity markets, currencies, commodities, and fixed income, using standard academic factor definitions and multiple lookback windows.

The results are clear: factor momentum is statistically significant, economically meaningful, and distinct from stock-level momentum. The strongest effects appear at 1-month and 12-month lookback periods, with annualized long-short returns ranging from 3.8% to 7.2% depending on the asset class and specification. These findings align with and extend the academic literature, notably Arnott et al. (2023) and Gupta and Kelly (2019), while offering new granularity on cross-asset factor persistence.

Why Factor Momentum Matters Now

Factor investing has grown from an academic curiosity to a multi-trillion-dollar industry. As of early 2026, factor-based ETFs and smart-beta strategies manage over $2.5 trillion globally. Yet most factor allocation frameworks treat each factor as a static, unconditional exposure: hold value, hold momentum, hold quality, and rebalance periodically.

This static approach ignores a critical empirical regularity. Factor returns are persistent. When value outperforms over a given quarter, it tends to continue outperforming in the next quarter. When low-volatility underperforms, it tends to keep underperforming. This persistence creates an opportunity for dynamic factor allocation, rotating into factors with recent strength and away from those with recent weakness.

The intuition behind factor momentum mirrors the intuition behind stock momentum. Behavioral explanations include investor underreaction to information (Gupta and Kelly, 2019), slow capital reallocation across strategies, and institutional herding that extends factor trends beyond what fundamentals justify. Risk-based explanations include time-varying risk premiums tied to macroeconomic regimes: when inflation rises, value tends to outperform growth persistently as discount rates shift, creating factor-level momentum that reflects rational repricing rather than mispricing.

Understanding factor momentum matters for two practical reasons. First, it can improve the timing of factor tilts in a diversified portfolio. Second, it helps explain why naive factor diversification (equal-weighting all factors at all times) leaves returns on the table relative to a momentum-informed allocation.

Data and Methodology

Factor Definitions

This backtest uses five canonical long-short factors, constructed following standard academic methodology:

Factor returns are sourced from the Ken French Data Library (HML, SMB, UMD) and the AQR Data Library (QMJ, BAB) for US equities. International equity factor data covers developed markets (ex-US) and emerging markets. The sample period runs from January 1990 through December 2025, providing 432 monthly observations.

Factor Momentum Construction

The factor momentum strategy ranks the five factors based on their trailing total return over a lookback window of L months. Each month, it goes long the top-performing factor(s) and short the bottom-performing factor(s) over the prior L months.

Four lookback windows are tested: L = 1 month, L = 3 months, L = 6 months, and L = 12 months. For the primary specification, the strategy goes long the single best-performing factor and short the single worst-performing factor (1/1 portfolio). Robustness checks include a 2/2 specification (long the top two, short the bottom two) and an equal-weighted momentum-score approach.

Portfolio returns are calculated monthly, with no leverage applied beyond what is inherent in the underlying long-short factor portfolios. Transaction costs are addressed in the robustness section.

Cross-Asset Extension

Beyond equities, factor momentum is tested across three additional asset classes:

Currencies: carry, value (PPP deviation), and momentum factors constructed from G10 currency pairs.

Commodities: carry (roll yield), momentum, and value (deviation from 5-year average) across 24 commodity futures.

Fixed Income: term (duration), carry (yield curve slope), and momentum across sovereign bond markets of 10 developed economies.

Cross-asset factor data is sourced from publicly available datasets including the AQR Data Library and central bank publications, covering the period from January 1995 through December 2025.

Results: Equity Factor Momentum

Performance by Lookback Period

The table below reports the annualized return, volatility, Sharpe ratio, and maximum drawdown of the 1/1 factor momentum strategy (long best factor, short worst factor) in US equities, across four lookback windows.

Several patterns stand out. The 1-month lookback delivers the highest annualized return (7.2%) and the best Sharpe ratio (0.61), suggesting strong short-term persistence in factor returns. This finding aligns with Gupta and Kelly (2019), who documented similar short-term factor momentum in a broader sample.

The 12-month lookback produces the second-strongest results (6.4% annualized, Sharpe 0.57), consistent with the familiar annual momentum pattern observed at the stock level. The intermediate lookback periods (3 and 6 months) produce weaker but still positive returns, with t-statistics above 2.0, indicating statistical significance at conventional levels.

The 1-month lookback and 12-month lookback display a U-shaped pattern that mirrors the stock momentum literature, where very short-term and medium-term lookbacks outperform intermediate windows.

Factor-by-Factor Autocorrelation

To understand what drives the aggregate factor momentum result, the table below reports the first-order autocorrelation of monthly returns for each individual factor.

Momentum (UMD) exhibits the highest autocorrelation across all lookback horizons, confirming that the momentum factor is itself the most momentum-like factor. Value (HML) and low volatility (BAB) show meaningful persistence at the 12-month horizon, consistent with the idea that macroeconomic regime shifts (inflation, interest rate cycles) create sustained factor trends.

Size (SMB) shows the weakest autocorrelation, suggesting that size premium fluctuations are more noise-driven and less amenable to timing via a momentum signal.

Is Factor Momentum Just Stock Momentum?

A critical question is whether factor momentum is distinct from stock-level momentum, or simply a repackaging of the same underlying signal. To test this, the factor momentum strategy returns are regressed on the Fama-French-Carhart four-factor model (market, size, value, and stock momentum).

The 1-month and 12-month specifications produce statistically significant alphas (t-statistics of 3.12 and 2.94 respectively) after controlling for stock momentum (UMD). The low R-squared values (4-7%) and small UMD betas confirm that factor momentum is largely orthogonal to stock momentum. This is a key finding: factor momentum captures a return pattern that stock momentum does not, making the two strategies complementary rather than redundant.

Arnott et al. (2023) reach a similar conclusion using a different methodology, demonstrating that factor momentum persists after controlling for stock-level momentum in both US and international samples.

Results: Cross-Asset Factor Momentum

Performance Across Asset Classes

The table below reports the annualized return and Sharpe ratio of the 1-month lookback factor momentum strategy across four asset classes.

Factor momentum is positive and statistically significant in every asset class tested. US equities produce the strongest absolute returns, but currencies deliver a competitive Sharpe ratio (0.52) despite lower absolute returns due to lower volatility. The cross-asset consistency of the factor momentum effect argues against a data-mining explanation; the same pattern appearing independently in equities, currencies, commodities, and bonds is unlikely to be a statistical artifact.

Diversified Cross-Asset Factor Momentum

Combining the individual asset-class factor momentum strategies into an equal-weighted cross-asset portfolio produces a diversified factor momentum strategy with improved risk-adjusted returns.

The cross-asset portfolio sacrifices some absolute return relative to US-only factor momentum but delivers a meaningfully higher Sharpe ratio (0.83 vs. 0.61) and a substantially shallower maximum drawdown (-10.1% vs. -18.3%). The diversification benefit arises because factor momentum signals across asset classes are weakly correlated; the factors that are winning in equities are not necessarily the same factors that are winning in commodities or fixed income.

Robustness Checks

Transaction Costs

Factor momentum strategies, particularly at the 1-month lookback, involve frequent rebalancing. To assess whether returns survive realistic transaction costs, the backtest is re-run with estimated round-trip costs of 10 basis points per factor rebalance (accounting for the fact that long-short factor portfolios involve trading many underlying securities).

The 12-month lookback is the most cost-efficient specification, retaining 94% of gross returns after transaction costs. The 1-month lookback remains profitable but loses roughly 25% of its gross return to trading costs. For retail investors implementing factor momentum through ETFs (where rebalancing costs are substantially lower than for direct long-short factor construction), net returns would be higher than these estimates suggest.

Sub-Period Analysis

The backtest period is split into two equal halves to check for stability across time.

Factor momentum returns decline modestly in the second half of the sample, consistent with the general pattern of declining factor premiums as factor investing has become more crowded. However, returns remain economically meaningful and statistically significant in both sub-periods.

Regime Dependence

Factor momentum performance varies with the macroeconomic environment. During expansionary regimes (as defined by NBER recession dates), factor momentum returns average 7.8% annualized. During recessions, returns drop to 3.1% but remain positive. The strategy does not crash during recessions, unlike stock-level momentum, which has historically suffered severe drawdowns during sharp market reversals (the "momentum crash" phenomenon documented by Daniel and Moskowitz, 2016).

This recession resilience is intuitive. Factor momentum trades the cross-section of factors, which are themselves long-short and approximately market-neutral. A market crash harms stock momentum (which is long high-beta winners and short low-beta losers) but does not systematically disadvantage a strategy that is simply rotating among factors based on recent relative performance.

Limitations

Several caveats apply to these results. First, the backtest uses factor return series constructed from academic datasets, not live trading returns. Implementation shortfall, liquidity constraints, and factor construction differences between academic portfolios and investable products can reduce realized returns relative to backtested returns.

Second, the five factors used in the equity analysis are the canonical academic factors. The factor momentum effect may differ when applied to proprietary factor definitions used by ETF providers, which often blend multiple signals or apply different weighting schemes.

Third, the cross-asset factor definitions are not as standardized as equity factors. Currency value (PPP deviation), commodity value (mean reversion to 5-year average), and fixed income carry (yield curve slope) are reasonable but not universally agreed-upon definitions. Results may be sensitive to alternative factor constructions in these asset classes.

Fourth, forward-looking factor momentum implementation requires choosing a lookback window ex ante. While the 1-month and 12-month lookbacks perform best in sample, there is no guarantee this U-shaped pattern will persist out of sample. A blended approach using multiple lookback windows may be more robust than committing to a single specification.

Practical Takeaways for Investors

For retail investors who allocate to factor ETFs, the findings suggest a straightforward implementation. Track the trailing 12-month returns of major factor ETFs (value, momentum, quality, low-volatility, size). Overweight the factors with the strongest recent performance and underweight or avoid the weakest. Rebalance quarterly to balance signal strength against transaction costs.

A simpler implementation involves just two factors. When value has outperformed momentum over the trailing 12 months, tilt toward value. When momentum has led, tilt toward momentum. This binary rotation captures a meaningful portion of the full factor momentum effect because the value-momentum pair exhibits the strongest negative correlation and the most pronounced persistence patterns among the five factors tested.

For institutional investors, the cross-asset results suggest that factor momentum can serve as an allocation overlay across asset-class factor strategies, improving the Sharpe ratio of a multi-asset factor portfolio from roughly 0.6 to above 0.8 through diversification and dynamic timing.

Factor momentum is not a free lunch. It requires active rebalancing, involves model risk in the choice of lookback window and factor definitions, and has shown some evidence of declining premiums as factor investing has grown. But the evidence presented here, consistent with the broader academic literature, suggests that factor returns are not random walks. They exhibit meaningful persistence that can be exploited through systematic strategies.