On this page
Compare the allowed sets
Suppose an unconstrained model allows an allocation to a particular sector. A new exclusion rule requires its weight to be zero. Every portfolio satisfying the new rule was already available in the original set, but some previously allowed portfolios are no longer eligible. This is a subset relationship, not a prediction about future market returns.
Hold the inputs and objective fixed
If the objective is maximum expected return at a given risk limit, optimizing over a subset cannot produce a higher maximum than optimizing over the original set. If the objective is minimum risk for a given expected return, the subset cannot produce a lower minimum. The optimum can remain unchanged when the added rule does not exclude the previously optimal portfolio.
| Comparison | Conclusion under unchanged inputs |
|---|---|
| Maximum return at the same risk limit | Cannot increase solely because the set is restricted |
| Minimum risk at the same expected return | Cannot decrease solely because the set is restricted |
| A constraint that does not bind the optimum | The optimum may remain unchanged |
Separate a constraint from a change in estimates
An analyst may also change expected returns, risk estimates or the objective after considering new sustainability information. That is a different comparison because the optimization inputs have changed. Do not attribute the effect of new estimates to a set restriction alone.
Do not infer a performance guarantee
An exclusion followed by a mechanical reallocation is not necessarily an optimization. Its risk, tracking error and realised performance depend on the resulting exposures and market outcomes. A rule can change weights without guaranteeing a higher Sharpe ratio.