Momentum is the tendency for assets that have performed well (poorly) in the recent past to continue to perform well (poorly) in the future, at least for a short period of time. Initial research on momentum was published by Narasimhan Jegadeesh and Sheridan Titman, authors of the 1993 study, “Returns to Buying Winners and Selling Losers: Implications for Stock Market Efficiency.”

The type of momentum studied by Jegadeesh and Titman is called cross-sectional momentum (applied example and explanation here). It is the type of momentum used in asset pricing models. Cross-sectional momentum measures relative performance, comparing the return of an asset relative to the returns of other assets within the same asset class. Thus, in a given asset class, a cross-sectional momentum strategy might buy the 30 percent of assets with the best relative performance and short sell the 30 percent of assets with the worst relative performance. Even if all the assets had risen in value, a cross-sectional momentum strategy would still short the assets with the lowest returns.

The other type of momentum is called time-series momentum (covered recently here). Time-series momentum is also referred to as trend-following because it measures the trend of an asset with respect to its own performance. Thus, unlike cross-sectional momentum, time-series momentum is defined by absolute performance. It buys assets that have been rising in value and short sells assets that have been falling in value. In contrast to cross-sectional momentum, if all assets rise in value, then none of them would be shorted.

The research on both types of momentum has shown that their premia have been persistent across long periods of time, pervasive across geography and asset classes (stocks, bonds, commodities, and currencies), robust to various definitions (formation periods) and implementable (as it survives transaction costs).

Can Momentum Be Used to Time Anomalies? (i.e., “Factor Timing”)