Using Fama–French Factors for Return Explanation and Prediction

Summary

The document raises a central distinction in applying Fama–French factor models: explaining realized stock returns versus forecasting them in advance. It observes that the models use market portfolio returns as explanatory variables, which are only fully known after the period’s returns occur. This creates an apparent timing problem for anyone treating the model as a direct signal for trading, especially because portfolio returns aggregate securities whose outcomes are also being analyzed.

The question suggests that factor models may be more naturally used to explain returns than to predict them. However, the document contains no answer, model specification, data, or empirical evidence resolving that concern. It also does not discuss forecasting with lagged or estimated factor exposures, factor premia, or other information available before the forecast period. Its main contribution is identifying the need to distinguish explanatory regressions from genuinely out-of-sample forecasting; the practical predictive value remains undetermined here.

Key ideas

• The document questions whether realized factor returns can serve as inputs to an advance stock return forecast.
• It distinguishes explaining observed returns from predicting future returns.
• Using contemporaneous market portfolio returns raises a timing concern for trading applications.
• The document provides no evidence or model details that resolve the forecasting issue.
• Predictive use would require information and evaluation available before the target return period.

Tags

# How can one apply models such as Fama-French factor model?


# How can one apply models such as Fama-French factor model?












I'm reading into Fama-French 3- and 5-factor models. I notice that they use the returns from market portfolios to "predict" stock excess returns. But obviously we cannot know ahead of time the returns from these market portfolios, and the returns from the market portfolios already incorporate the returns from individual stocks. By the time we know the returns for those market portfolios, we'll also know the returns for all the stocks. So what's good are these models? It seems to me that the proper use of them is to explain returns, rather than predict, which means one cannot use them to trade, no?

Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.