How should I think about cross-sectional versus time-series factor models?

Question

AcadiFi · Accepted Answer

The easiest distinction is to ask what is being estimated first. In a time-series approach, you start with predefined factors and estimate how an asset loads on them over time. In a cross-sectional approach, you often start with many assets at one date and infer factor realizations from the cross-section using their characteristics or exposures. Both frameworks can be useful, but they answer slightly different questions.

Suppose Meridian Equity Risk wants to explain portfolio behavior. If the goal is to say how one stock historically moved with market, size, and value factors, a time-series beta framework is intuitive. If the goal is to explain how the whole universe was priced this month given a set of exposures, a cross-sectional framework may be more natural.

The tradeoff is stability versus flexibility. Time-series models can be easier to interpret but depend heavily on factor definitions and window choice. Cross-sectional models can be richer for large universes but require stronger structure and estimation discipline. Neither framework is universally superior without knowing whether forecasting, attribution, or risk control is the main goal. FRM questions usually reward candidates who understand that model choice depends on use case, data shape, and the decision the model is supposed to support. Practice more in our FRM question bank.

How should I think about cross-sectional versus time-series factor models?

Master Part II with our FRM Course

Related Questions

Practice Questions