Comparing Factor Models with Squared Sharpe Ratios

Summary

The document raises a question about comparing nested and nonnested Fama–French factor models using squared Sharpe ratios. It refers to a proposed test that takes factor returns as inputs and does not require a separate set of test assets, then asks how the statistic is calculated in that setting.

The author wonders whether the squared Sharpe ratios of individual factors are calculated separately and added together, and points out that this could make comparisons sensitive to how many factors each model contains. The document does not provide the test’s formula, an answer, empirical evidence, or a resolution. It is therefore best read as a prompt to examine the distinction between a model-level portfolio Sharpe ratio and factor-by-factor aggregation, rather than as guidance on implementing or interpreting the test. Any conclusion about the proposed method requires consulting the cited research and its exact definition of the statistic.

Key ideas

• The document asks how a factor-only test calculates squared Sharpe ratios when comparing factor models.
• It questions whether model performance is found by summing individual factor Sharpe ratios.
• It notes that a factor-by-factor sum may complicate comparisons between models with different numbers of factors.
• The document provides no formula or answer, so the calculation remains unresolved.

Tags

# Squared Sharpe Ratio - Fama and French


# Squared Sharpe Ratio - Fama and French












I am investigating various versions of nested and nonnested Fama and French factor models. Performance of the models is compared on the basis of Squared Sharpe Ratios. Bariallas et al. (2020, JFQA) propose a test that does only require the factors as inputs and does not rely on test assets. However, I can not quite get my head around how Squared Sharpe Ratios are calculated in this case. Are they calculated for each factor individually and then summed up? This approach seems flawed, and if any, this would only make sense if the models that are compared have the same number of factors.

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.