Recovering Implied Returns from an Unconstrained Mean-Variance Portfolio

Summary

The document asks how to infer the expected returns that would produce a given fully invested portfolio under mean-variance optimization, given its weights, risk aversion, and return covariance matrix. The accepted answer gives a simple relation: implied returns are proportional to the covariance matrix multiplied by portfolio weights, with the risk-aversion parameter as the scaling factor.

This relation offers an intuitive interpretation: assets with greater covariance exposure in the portfolio require higher expected returns under the stated optimization setup. Its applicability is narrow. The investor’s risk aversion must be known, and the answer says the formula assumes an unconstrained investor who can take both long and short positions. The question’s proposed expression includes a full-investment adjustment, but the answer does not derive or validate that constrained case; it notes that treatments including the allocation constraint exist and may introduce circularity.

Key ideas

• For the stated unconstrained mean-variance setup, implied returns equal risk aversion times covariance-weighted portfolio weights.
• The inferred returns rise with the portfolio’s covariance exposure and the assumed risk-aversion level.
• Risk aversion must be known, which is a strong requirement for recovering implied returns.
• The stated relation assumes long and short positions are allowed and does not directly cover long-only constraints.

Tags

# Markowitz Mean-Variance Implied Returns


# Markowitz Mean-Variance Implied Returns












What is the closed form solution for the following inverse Markowitz problem?

Given a mean-variance optimized fully invested portfolio $X$, a risk aversion parameter $\lambda$ and a var-covar matrix $C$. What is the formula for the (implied) returns $\mu_{impl}$ that must have been used to build the portfolio $X$?

I have a working paper (Kritzman et. al. 2008) claiming a closed form solution $\mu_{impl}$ that depends on the "expected returns" $\mu$ which seems a bit circular and is perhaps a typo.

$$ \mu_{impl} = \lambda C X^T + \frac{-\lambda + 1 C^{-1} \mu^{T}}{1 C^{-1} 1^T} 1^T $$

## Answer by Richi Wa (score 4, accepted)

https://quant.stackexchange.com/a/16732

The formula is $$ \mu = \lambda CX $$ in your notation. You find it in many places, e.g. here.

The assumption is that you know $\lambda$ which is a strong assumption. Furthermore it only holds if investors are unconstrained (long/short not long only).

It is intuitive as it says that given the weighting the return expectation increases with risk aversion and risk.

The case with the full allocation constraint (sum of weights is one) is covered by Herold but I also think that this is a bit circular ...

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.