Skip to content
All library documents

Factor-Tilting Portfolios as an Alternative for Factor Backtests

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

Summary

The document raises an alternative to the common factor backtest that buys the highest-ranked group and shorts the lowest-ranked group. It quotes a research proposal to construct a long-only portfolio with exposure to a chosen factor while targeting a specified tracking error. The question focuses on the benchmark for that tracking error and how such a portfolio would be built. It also distinguishes this proposal from factor-mimicking portfolios, which the source presents as a separate alternative.

The text does not include an answer, implementation details, empirical results, or a definition of the tracking-error benchmark. It therefore introduces a portfolio-construction idea and identifies the key ambiguity a researcher would need to resolve, but it does not provide enough information to reproduce or evaluate the method. Readers should treat the quoted proposal as a question about a backtesting design rather than as a complete factor-testing procedure.

Key ideas

  • The document contrasts factor-tilting backtests with long-short portfolios formed from extreme factor rankings.
  • The proposed portfolio is long-only and seeks exposure to a selected factor under a tracking-error target.
  • The tracking-error benchmark is left unspecified in the document.
  • Factor-mimicking portfolios are identified as a distinct alternative.
  • No construction procedure or performance evidence is supplied.

Tags

Factor investingBacktestingPortfolio construction
Full text
# How to backtest a factor by building a factor-tilting portfolio


# How to backtest a factor by building a factor-tilting portfolio












A recent research paper from Wolfe Research suggests an alternative to the traditional "long top quintile, short bottom quintile " approach to backtesting a factor. It says to

> construct a factor tilting portfolio, where a long-only portfolio with exposures to a given factor can be built with a target tracking error.

How would this even work? Tracking error to what? I know the paper isn't referring to "factor mimicking portfolios" because that's presented as a different alternative to long-short backtesting.

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.