How Mean Views Affect Volatility in Entropy Pooling

Summary

This note raises a modeling question about entropy pooling, a method that adjusts a prior distribution to satisfy views while staying close to that prior according to relative entropy. The author observes that imposing a higher expected return, without an explicit volatility view, often appears to preserve the original volatility. This can seem counterintuitive if the investor expects higher returns to be associated with higher risk.

The document asks whether this behavior is an intentional consequence of the method or a weakness, and whether a remedy exists. It cites a paper on flexible views as background but provides no answer, derivation, experiment, or proposed adjustment. The key implication is that a mean constraint alone need not encode a return-risk relationship: a desired association between expected returns and volatility may need to be stated through additional distributional or covariance views. The observation is presented as the author’s experience, not as a general result established across models or datasets.

Key ideas

• Entropy pooling finds a distribution close to a prior while satisfying specified views.
• A mean-return view alone may leave volatility near its prior level, according to the author’s observation.
• The note questions whether this behavior is a limitation or an intended consequence of the method.
• It offers no solution or empirical evidence, leaving the issue open for further analysis.

Tags

# Unintuitive results from Entropy Pooling


# Unintuitive results from Entropy Pooling












Entropy pooling can be used to find a distribution which

- is close to the original distribution that you already have, where closeness is measured by relative entropy,

- satisfies various views that you might have on the means or covariances.

For more on entropy pooling see for example the 2010 paper “Fully Flexible Views: Theory and Practice” by Meucci.

I find that entropy pooling gives unintuitive results when my view is that the mean of the returns will be a lot higher than usual.

When doing so, I would expect, intuitively, that the higher return would come with higher risk. I might not be having any view on the volatilities, but ordinarily, asset classes that deliver high returns come with higher volatilities.

But I find that entropy pooling typically preserves the volatility if your view is simply that the mean return will be higher.

Is there any way to remedy this? Is this considered a weakness of entropy pooling, or an intentional design?

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.