Skip to content
All library documents

Using a Consistent Five-Year History for Fama-French Factors

Article Quant Q&A · Author: Tim Allen
Read the original

Summary

The document asks which stocks to include when constructing country-specific Fama-French factors for tests of portfolio returns and alphas. The researcher’s test portfolios require five years of past data, raising the question of whether factor construction should use that same history requirement or include all available stocks.

The response recommends using stocks with at least five years of data, after identifying the relevant country’s listed universe through a broad market index. Stocks lacking that history are excluded from factor construction. This offers a simple sample-alignment rule for the described research setup, but the document gives no evidence or discussion of alternative eligibility rules, survivorship bias, missing-data treatment, or whether the five-year requirement is necessary for every factor. Those choices may matter when applying the advice to other designs.

Key ideas

  • The response recommends limiting factor construction to stocks with at least five years of available data.
  • It suggests defining the country stock universe using a broad local index.
  • The recommendation aligns the history requirement for factor construction with that of the test portfolios.
  • The discussion does not address survivorship bias or compare other sample-selection rules.

Tags

EquitiesFactor investingBacktesting
Full text
# Include all stocks in Fama-French portfolio calculations?


# Include all stocks in Fama-French portfolio calculations?












I'm calculating Fama-French portfolio returns for a country, to test the sign and significance of alphas for portfolios sorted by my variable of interest. I can't use Fama-French data library factors as these are not specific enough to my country.

My question is that, if i'm assessing returns and alphas for 10 different portfolios, formed based on a variable requiring 5 years of past data - should I only include stocks with 5 years of past return data in the construction of my SMB, HML and other Fama-French factors? Or should I instead include all stocks?

Edit: Still looking for help on this if someone is able to help me out!

## Answer by Prabhnoor Duggal (score -1)

https://quant.stackexchange.com/a/55526

Include all the stocks having the data of at least last 5 years. Pick up a general index of your country(the one you want to work with) and get the data for all the stocks listed in that index. Exclude the ones that have less than 5 years of data and proceed with the rest.

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.