Practical Constraints in Portfolio Optimization
Summary
The document surveys constraints that practitioners may impose when optimizing a portfolio. It notes that the formulation matters: a fully invested portfolio typically has weights summing to one, while an active portfolio expressed as deviations from a benchmark may require active weights to sum to zero. Other examples include long-only rules, limits on individual positions or the number of assets held or traded, bounds on asset-class and currency exposures, factor beta ranges, and restrictions tied to ratings.
It also describes implementation concerns such as liquidity requirements, limits on forced-sale haircuts, and concentration controls within an asset class. The central point is that there is no universal standard set: the relevant constraints follow from the investment mandate and practical application, and expressing them correctly for an optimizer can itself be difficult. These are illustrative examples rather than a complete catalog, and the document provides no quantitative comparison, formal optimization model, or reference text for further study.
Key ideas
- Portfolio constraints depend on the mandate and whether the optimization uses total or active weights.
- Common rules include full investment, long-only allocations, position limits, and limits on holdings or trades.
- Portfolios may be bounded by asset-class, currency, rating, or factor exposures.
- Liquidity, forced-sale risk, and concentration can also be represented as constraints.
- Translating practical requirements into optimizer inputs requires care.
Tags
Full text
# Portfolio Optimization Constraints # Portfolio Optimization Constraints Wondering which are some standard constraints in portfolio optimization in practice? For example, assuming we want to maximize expected returns subject to a risk constraint, typically we may have -constraints on gross exposure -constraints on net exposure -constraints on individual asset gross exposure etc. Is there a good document/ book that has standard constraints that are used in practice? ## Answer by vanguard2k (score 2) https://quant.stackexchange.com/a/37887 Constraints obviously also depend on the problem formulation (for example: weights sum to 1 will translate to active weights sum to 0 if we optimize relative positions) Apart from the most important ones the unit investment constraints (weights sum to 1 or money invested sums to NAV), sometimes short-selling constraints, very often you see constraints on the number of securities held or traded. Depending on the problem, also constraints on portfolio characteristics, such as FX exposure, Non-rated or asset class exposures. Unfortunately there is no standard literature on this. The reason for this being: Every optimization problem is different and the constraints come from the practical application itself. Disclosure/Addendum: One thing is considering constraints, the other one is translating them into a format the optimizer understands. More often than not, this is harder than you think. ## Answer by Johan Stax Jakobsen (score 2) https://quant.stackexchange.com/a/37897 I have worked with: Portfolio stability constraints - Long-only constraints (to avoid unrealistic short positions) - Maximum and minimum allocation to asset classes Liquidity and illiquidity constraints - Minimum posititon in liquid instruments (money market, govies, etc.) - Maximum hair-cut due to forced selling Concentration constraints - Asset within asset class concentration constraint (eaxh asset should not be 100% of the exposure in the particular asset class) Portfolio factor beta exposure - Minimum and maximum limits to the relative beta exposure of the portfolio aganst a given factor (e.g. market factor) There will certaintly be more possible constraints (see e.g. the implementation here).
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.