Skip to content
All library documents

Interpreting Factor Loadings for Custom Fama–French Benchmark Portfolios

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

Summary

The document raises a question about factor exposures when constructing equal-weighted and value-weighted benchmark portfolios from a cleaned set of U.S. listed firms. The portfolio universe contains firms across AMEX, NYSE, and NASDAQ, with membership varying through time. The author regresses standard market, size, value, profitability, investment, and momentum factors on those portfolios and asks whether a close replication of the original benchmark should produce significant loadings near one.

No answer, empirical results, or cited papers are included, so the document does not establish what the estimated betas should be. Its central research issue is how differences in firm selection, sample coverage, weighting, and portfolio construction affect measured factor exposures. The distinction between a market factor and the other style factors also matters when interpreting a regression: exposures need not all be close to one merely because a portfolio resembles a broad benchmark. The question motivates careful comparison of portfolio definitions and benchmark construction, but leaves these checks unresolved.

Key ideas

  • The document asks how factor exposures change for custom equal-weighted and value-weighted portfolios.
  • The sample includes U.S. exchange-listed firms and varies in membership over time.
  • It does not provide regression estimates or literature references that answer the question.
  • Portfolio selection and weighting can alter measured factor loadings.
  • A portfolio’s resemblance to a benchmark does not imply that every factor beta should be near one.

Tags

EquitiesFactor investingPortfolio constructionStatistics
Full text
# FF factor benchmarks


# FF factor benchmarks












I have a question to think about. I have constructed a portfolio of AMEX/NYSE/NASDAQ firms. After some data manipulation and cleaning I am left with some 4000 firms over the complete sample (max 1600 at the same time). I have excluded firms I do not have some observations for the variable I seek to examine.

Subsequently I have constructed two benchmark portfolio (EW and VW) with these respective firms. Thereafter, I regressed the fama-french-carhart factors upon these portfolios (MKT, SMB, HML, RMW, CMA, UMD).

Now my qeustion comes: How would you expect the Loadings (beta's) to behave? If I would have replicated their benchmark portfolio in the most acurate way possible. I would expect the loadings to be highly significant and close to to 1. Would that be correct? Or is that to narrow-minded.

By your knowledge are there any papers that examine the factor exposures of benchmark portfolios constructed in similar manner with similar firms as Fama and french (1992)

Hope to hear your thoughts.

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.