Cross-Sectional vs Time-Series Momentum

Relative winners versus trend-following, compared.

The phrase "momentum strategy" covers two genuinely different bets that happen to share a name. Cross sectional vs time series momentum are often discussed in the same breath — both exploit the tendency of recent price performance to persist — but they differ in what they are actually long, what risks they carry, and when they tend to earn their returns. Conflating them leads to strategy construction errors, incorrect risk budgeting, and confusion when one earns while the other bleeds. The distinction is worth understanding precisely before any capital is committed to either approach.

For practitioners building systematic equity or multi-asset portfolios, the Quant Galore Momentum Index provides a live benchmark for the cross-sectional variant applied to U.S. equities — useful context for the discussion that follows.

What cross-sectional momentum actually measures

Cross-sectional momentum ranks assets relative to one another. The universe — whether S&P 500 constituents, global equities, or a mixed multi-asset set — is sorted by trailing return over a lookback window, typically 12 months with the most recent month excluded to avoid short-term reversal. Names in the top decile or quintile are bought; names in the bottom decile or quintile are sold short. The resulting portfolio is dollar-neutral: long exposure and short exposure are balanced, and the portfolio has close to zero net market beta by construction.

The bet is purely on relative performance. A cross-sectional momentum portfolio earns when recent winners continue to outperform recent losers within the universe — not when the market itself goes up or down. If every asset in the universe rallies 15%, the long and short legs both gain roughly 15% and the net return is approximately zero. The strategy does not express a view on market direction. It expresses a view on dispersion and on the persistence of relative rankings.

This is the momentum factor as it appears in most academic factor literature — Jegadeesh and Titman's original work, the Fama-French five-factor extensions, and most equity factor attribution frameworks. When a multi-factor equity manager reports a positive loading on momentum, it is almost always the cross-sectional variant being measured.

What time-series momentum actually measures

Time-series momentum, also called trend following, judges each asset against its own past rather than against the rest of the universe. For each instrument, the question is simple: has this asset's own trailing return been positive or negative? If the 12-month return is positive, go long. If it is negative, go short. Each asset is sized independently — there is no ranking against peers and no requirement for the portfolio to be dollar-neutral in aggregate.

Moskowitz, Ooi, and Pedersen's research across dozens of futures markets demonstrated that this time-series signal carries significant predictive power over a 12-month horizon. Managed futures funds and CTAs have operationalized this logic for decades, though they implement it across multiple lookback windows and asset classes simultaneously.

The critical structural consequence is that a time-series momentum portfolio can be net long across all assets simultaneously — and frequently is. When global markets trend upward together, the strategy buys everything. When markets trend down together, it sells everything short. This produces a meaningful directional exposure that varies over time. Unlike the cross-sectional version, the aggregate beta of a time-series momentum portfolio is not constrained to zero. It fluctuates based on how many instruments happen to be in positive-return regimes at any given moment.

Where they overlap and where they diverge

The two strategies share an empirical foundation: assets that have performed well recently tend to continue performing well. That commonality makes their returns correlated in some environments. In a rising market with strong dispersion, both tend to work: cross-sectional benefits because the spread between winners and losers is wide, and time-series benefits because many names are in positive-return regimes.

The divergence is most visible in two scenarios. The first is a broad, undifferentiated rally. When a market-wide event lifts most assets by a similar magnitude — a central bank pivot, a systemic risk-on shift — time-series momentum turns positive on many instruments simultaneously. Cross-sectional momentum earns little, because the relative rankings within the universe may not have changed meaningfully. Every name is trending up, but the long portfolio is not outperforming the short portfolio by any more than usual.

The second scenario is a market with high dispersion but no directional trend. A year in which some sectors surge and others collapse — with the broad index flat — is ideal for cross-sectional momentum and unhelpful for time-series momentum. Cross-sectional ranks cleanly separate winners from losers; time-series cannot confidently go long or short the flat-returning assets at the index level.

The overlap in return is real but incomplete. Empirically, the correlation between the two strategies is positive but well below one — often in the 0.3 to 0.5 range across equity implementations — which means holding both adds genuine diversification relative to holding either alone. This sits in contrast to momentum versus mean reversion, where the strategies are more structurally opposed and diversification logic is different.

Risk profiles and the crisis-alpha question

Risk structure is where the differences between the two strategies become practically important for portfolio construction.

Cross-sectional momentum is market-neutral and exposed to factor-specific risks. Its dominant risks are dispersion risk — the spread between winners and losers collapsing — and momentum crash risk. Momentum crashes tend to occur after sharp market drawdowns followed by rapid reversals: the strategy has accumulated a short book in beaten-down names that then snap back violently. The losses are concentrated, sharp, and often arrive when other risk assets are already stressed. This is a well-documented vulnerability. Market neutrality does not mean low risk; it means the source of risk is concentrated in relative-performance dynamics rather than market direction.

Time-series momentum carries directional exposure and, as a result, has historically demonstrated a convex payoff profile relative to equities. Because it tends to go net short during sustained equity market downturns, it has provided positive returns in several major drawdown periods. This is what practitioners call crisis alpha — the tendency of trend-following strategies to earn positive returns precisely when equity portfolios are suffering. The mechanism is not protective by design; it is a structural consequence of the strategy going short assets whose prices are falling persistently. In a sharp, rapid crash with quick reversal, the protection disappears — the strategy needs the trend to persist long enough to be captured.

Side-by-side comparison

Dimension	Cross-sectional momentum	Time-series momentum
Signal basis	Rank relative to peers in universe	Each asset vs. its own past return
Market exposure	Near-zero beta by construction	Variable net long/short; directional
Best environment	High dispersion, moderate trend	Persistent directional trends
Worst environment	Sharp reversal after drawdown (momentum crash)	Choppy, mean-reverting markets
Typical home	Equity factor portfolios, long-short funds	Managed futures, CTAs, macro
Crisis behavior	Can suffer during reversal episodes	Historically positive during prolonged drawdowns
Turnover	Moderate — driven by rank changes	Moderate to low — signals change slowly
Universe requirement	Needs a defined peer group for ranking	Each instrument evaluated independently

Implementation and data requirements

Cross-sectional momentum requires a defined universe and consistent return data for every constituent at every rebalancing date. Returns must be total returns — including dividends — and the universe must be survivorship-bias-free for backtests to be credible. Index additions and deletions must be handled correctly: the strategy must only use assets that were investable on the ranking date, not assets that were added later because they performed well. This is a data construction problem as much as a strategy design problem.

Rebalancing frequency is a significant cost driver. Monthly rebalancing is common, but high turnover in the top and bottom deciles generates transaction costs that erode live returns relative to backtested returns. Implementations that use overlapping portfolios — initiating a new cross-sectional portfolio each month and holding the prior months' portfolios simultaneously — reduce turnover while preserving signal exposure.

Time-series momentum is simpler in one sense: the signal is computed at the individual instrument level without reference to the rest of the universe. A futures trader needs only the price history of each contract. The complexity lies in position sizing. Because the strategy can be net long or net short any given instrument, and because the aggregate directional exposure fluctuates, risk scaling becomes essential. Volatility targeting — sizing each position such that its expected daily or monthly volatility contribution is constant — is the standard approach. Without it, the portfolio's risk varies enormously as more or fewer instruments enter positive-trend regimes.

Costs for time-series momentum in futures are generally lower than for equity cross-sectional momentum, because futures markets are liquid and the signals are slow-moving relative to equity factor strategies. Equity implementations of time-series momentum, applied to individual stocks rather than index futures, face higher costs and are less common in practice.

When to use each approach

The choice between strategies is not purely theoretical — it follows from the portfolio's mandate, the available instruments, and the desired risk profile.

Cross-sectional momentum belongs in equity factor portfolios where market neutrality is a constraint or a goal. It is the natural complement to value and quality factors, which it tends to have low or negative correlation with over long horizons. An equity long-short fund that wants to harvest the momentum premium while remaining insulated from broad market moves should be running cross-sectional signals. The equity factor literature's evidence base for momentum is built almost entirely on the cross-sectional variant, so that is also what is being replicated when a factor portfolio claims momentum exposure.

Time-series momentum is appropriate when the goal is convexity against equity drawdowns, when the investment universe spans multiple asset classes, or when the strategy will be implemented in futures markets where the mechanics suit the signal. CTAs and managed futures managers are the natural home. A multi-asset portfolio that wants to reduce left-tail equity exposure — without relying on options or other explicit hedges — can use trend-following as a structural diversifier, accepting the variable beta in exchange for the crisis-alpha profile.

The two are not mutually exclusive. A multi-strategy approach that allocates to both benefits from their partial correlation and meaningfully different risk exposures. Cross-sectional exposure harvests the dispersion premium inside equity markets; time-series exposure adds a directional and cross-asset layer that earns in different conditions. Together, they provide broader coverage of the momentum premium than either does alone.