Two cousins of the same family, often confused, with materially different risk characteristics. Understanding the distinction is fundamental to portfolio design.
Cross-sectional momentum ranks an asset against its peers; time-series momentum compares each asset against its own past. Both are real, both have been documented across asset classes, and both belong in a serious systematic toolkit — but the way they fail is very different.
Cross-sectional momentum
Long the top-decile performers, short the bottom-decile. It is, by construction, beta-neutral within the universe and tends to be most powerful in environments with dispersion across names. Its dominant risk is the “momentum crash” — the violent mean-reversion that follows extended trends.
Time-series momentum
Each asset is judged on its own past return; positions can be net long or net short the universe. This is the structure that drives most managed-futures programmes and is the technical basis of trend following.
When to prefer which
In sharply trending macro regimes — 2008, 2022 — time-series momentum tends to dominate, because it can carry net short exposure. In dispersion-rich, range-bound regimes, cross-sectional momentum captures relative winners better. A complete momentum sleeve usually carries both.
FAQ
Can a single signal capture both?
Approximately. Some practitioners use a single residualised return signal, but the cleanest implementations keep the two as separate sleeves with their own risk budgets.