Choosing Factors for Expected Returns and Portfolio Construction

Summary

The document asks whether a portfolio’s expected returns should use only statistically significant factors identified through Fama–MacBeth analysis or all factors in the model. The response distinguishes a statistical selection view, which would omit insignificant factors, from an economic specification view, which retains factors believed to explain asset returns.

It warns that omitting economically relevant factors may create specification bias and affect the quality of coefficient estimates. However, the reply is brief and gives no empirical comparison, portfolio construction procedure, or backtest results. It does not explain how to reconcile statistical uncertainty with economic rationale, nor address model selection, multiple testing, out-of-sample validation, or estimation risk, so the recommendation should be treated as a conceptual caution rather than a complete workflow.

Key ideas

• The question concerns which estimated factors to use when calculating expected portfolio returns.
• Statistical significance can motivate excluding factors, while economic relevance can support retaining them.
• Omitting relevant factors may lead to specification bias and unreliable estimates.
• The brief response does not provide a selection procedure or out-of-sample evidence.

Tags

# Factor model to Portfolio optimization


# Factor model to Portfolio optimization












By using the Fama and MacBeth methodology, I have identified the significant factors that explain the returns of my stocks. Now, I want to build a portfolio and backtest it.

For that, I am trying to use the Mean(factor loadings) and the respective betas. However, to calculate the expected return, should I take only the SIGNIFICANT factors or ALL THE FACTORS? P.S. : I only have one significant factor.

Thanks in advance, Vaibhav

## Answer by Alejandro Andrade (score 1)

https://quant.stackexchange.com/a/22143

Statistically speaking you should not include the factor that aren't significant. Economically speaking you should take all the factors because intuitively they explain the returns of the assets, and if you don't do it will incur in specification bias by omit the factors and will cause the the estimates aren't efficient, unbiased and consistent

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.