Cross sectional versus times series momentum

Momentum has been classified as two types: cross-sectional and time series.

Most of the research and buzz has been about cross-sectional momentum whereby a large number of assets are ranked and a portfolio is created being long the top performers and short the poor performers. Buy the best and sell the worst. This can be done inside an asset class or across asset classes. The value at any time is buying the best given a ranking system of past returns. It is a tests of relative momentum preferences.

The time series approach could be thought of as a ranking the performance of asset relative to past history. This has not been as extensively tested by academic researchers because it focuses on issue of market efficiency and past performance directly; however, if you go back through past research you will find a lot of tests. It is where most of the technical traders spend their time and effort. The hurdle may be higher than cross-sectional work but there is still evidence of success using momentum. It is a more absolute test of momentum and shows that momentum is present, but there will eventually be a reversal based on different timeframes. There is under-reaction and then over-reaction which leads to  reversal and adjustment.

The current rage in modeling is to combine cross-sectional ranking with times series performance to be selective in asset choices.