Skip to content
All library documents

OLS, Time-Series Regression, and Autocorrelation in Fama–French Models

Article Quant Q&A · Author: Priya
Read the original

Summary

The document raises two methodological questions about time-series regressions in the Fama–French asset-pricing framework. First, it asks why the method is called a time-series regression when estimation uses ordinary least squares. Second, it questions whether leaving autocorrelation unaddressed can bias results and whether adding autoregressive, moving-average, or combined ARMA terms alongside the factors could help.

No answers, data, model specification, or empirical evidence are provided, so the text does not establish that the cited regressions ignore serial dependence or that adding ARMA terms is an appropriate remedy. It usefully frames the distinction between the dimension of the data and the estimator: OLS can be applied to observations indexed over time. The autocorrelation concern also points toward checking residual dependence and inference assumptions, but the document leaves the choice of specification and correction unresolved.

Key ideas

  • OLS can estimate a regression whose observations form a time series.
  • The document asks whether autocorrelation affects Fama–French regression results.
  • It proposes adding AR, MA, or ARMA terms as a possible way to address serial dependence.
  • The text gives no empirical evidence or answer on the suitability of these adjustments.

Tags

EquitiesStatisticsFactor investing
Full text
# Why OLS in Fama French time series regression?


# Why OLS in Fama French time series regression?












I read many papers on asset pricing and have some basic doubts regarding Fama French Time series regression:

- We have time series data, but still it is a simple OLS we run in FF model. Then why it is called Fama French time series regression?

- They have done nothing for autocorrelation problem. If this is so, then the results could be bias. Can we use AR, MA or ARMA model in addition with three factors which can help for autocorrelation problem?

Kindly help me. Thank You Priya

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.