Cross-sectional momentum: how relative performance generates systematic returns

Cross-sectional momentum is the systematic tendency for assets that have recently outperformed their peers to continue outperforming, and for recent underperformers to continue underperforming. It is one of the most extensively documented return patterns in finance, observed across equities, fixed income, currencies, and commodities in markets dating back more than a century. Cross-sectional momentum differs from time-series momentum in a specific and important way: it is a relative signal, not an absolute one—it ranks assets against each other rather than against their own history.

What cross-sectional momentum is

Time-series momentum asks: has this asset gone up or down over the past twelve months? Cross-sectional momentum asks: has this asset gone up more or less than other assets over the past twelve months? The distinction matters for portfolio construction. A time-series momentum strategy might hold cash when all assets have negative momentum—it exits the market entirely. A cross-sectional momentum strategy always holds something: the best-performing assets in the universe, whether returns are positive or negative, because the signal is relative rank rather than absolute direction.

The standard implementation, formalised by Jegadeesh and Titman (1993), uses a twelve-month formation period with a one-month skip. Assets are ranked by their total return over months two through thirteen of the lookback period (excluding the most recent month, which is skipped because very short-term returns tend to reverse due to microstructure effects). The top performers are bought; the bottom performers are sold. The portfolio is rebalanced monthly. This "12-1" formation window has proven the most robust specification across different markets and time periods.

What the evidence shows

Jegadeesh and Titman (1993) documented that buying the top decile and selling the bottom decile of US stocks, ranked by their prior twelve-month return, produced risk-adjusted returns of approximately 1% per month over 1965–1989. This finding was replicated across international equity markets by Rouwenhorst (1998) in a study of twelve European markets, and has since been confirmed in markets across Asia, Latin America, and emerging markets.

Asness, Moskowitz, and Pedersen (2013) extended cross-sectional momentum beyond equities to government bonds, currencies, and commodity futures markets, finding consistent returns across all four asset classes and across eight countries. Their study covered 1972–2011 and found that the cross-sectional momentum premium was largely uncorrelated across asset classes, suggesting a common behavioural mechanism rather than a shared risk factor. The premium persists after standard risk adjustments for market beta, value, and size.

The explanations for cross-sectional momentum cluster around two behavioural mechanisms. Investor underreaction means that information about improving fundamentals diffuses slowly through markets—prices adjust gradually over twelve months as more investors incorporate the news. Investor overreaction then means that prices overshoot as momentum attracts trend-following capital, which is why the twelve-month momentum signal is followed by a reversal over the subsequent three to five years (the value factor partially captures this long-run mean reversion).

Limitations and trade-offs

Cross-sectional momentum is subject to sharp, brief periods of underperformance known as momentum crashes. Barroso and Santa-Clara (2015) documented that momentum strategies experience their worst returns during market reversals—the months immediately following a sharp equity market decline, when recent losers (often beaten-down value stocks) rebound strongly while recent winners continue to fall. The 2009 momentum crash, following the depths of the global financial crisis, produced a drawdown of approximately 80% in a long/short momentum portfolio over a few months. Long-only momentum strategies suffer less dramatically but are still exposed to reversal risk during sharp market turns.

Capacity and transaction costs are significant practical constraints on high-frequency implementations of cross-sectional momentum. Monthly rebalancing across a diversified universe is feasible; higher-frequency implementations require very liquid assets and generate proportionately higher transaction costs. For smaller investors and multi-asset portfolios using ETFs, transaction costs are manageable; for large institutional implementations in single-stock universes, capacity constraints are real.

Cross-sectional momentum in pfolio

Cross-sectional momentum is a core selection signal in pfolio's asset allocation process. At each monthly rebalancing, assets in the investable universe are ranked by their recent relative performance, and allocation is tilted toward assets with strong relative momentum. This cross-sectional ranking complements the time-series momentum signal (described in time series momentum) to identify both absolutely and relatively strong-performing assets. The methodology is described in detail in how we build portfolios.

Related articles

• Time series momentum: how absolute price trends generate systematic returns
• Momentum investing: the evidence behind buying recent winners
• Factor investing explained: how systematic risk premia drive long-run returns
• Backtesting investment strategies: methodology, limitations, and how to avoid overfitting