Robust Portfolio Optimisation

Goldfarb-Iyengar (2003) · Ellipsoidal uncertainty · SOCP
Family 01 Portfolio SOCP Robust

Markowitz mean-variance takes $\mu$ and $\Sigma$ as known; in practice both are estimated with error, and Michaud (1989) showed that optimised portfolios are wildly sensitive to that error. Robust portfolio optimisation (Ben-Tal & Nemirovski 1998; Goldfarb & Iyengar 2003) replaces point estimates with uncertainty sets and maximises the worst-case expected return inside the set. The resulting programme is a second-order cone programme (SOCP), only modestly harder than a QP, and the portfolios it produces are dramatically more stable under re-estimation.

Educational Purpose
Demonstration of robust optimisation applied to portfolio choice. Not investment advice. Uncertainty-set calibration is itself a statistical problem; misspecified sets produce misleadingly precise guarantees.

The model

Uncertainty set, worst-case objective, SOCP reformulation

OR family: Portfolio optimisation Solver class: SOCP Measure: Physical $\mathbb{P}$ Realism: ★★☆ Heuristic SOCP (projected-gradient)

Notation

SymbolMeaning
$\mu_0 \in \mathbb{R}^n$Nominal (point-estimate) expected-return vector
$\Sigma \in \mathbb{R}^{n\times n}$Covariance matrix (assumed known here; Goldfarb-Iyengar also allow it uncertain)
$\mathcal{U}_\kappa$Ellipsoidal uncertainty set around $\mu_0$ with radius $\kappa$
$\kappa \ge 0$Uncertainty-set radius (0 = nominal Markowitz; large = very conservative)
$w \in \mathbb{R}^n$Portfolio weights, $\sum w_i = 1$, $w_i \ge 0$
$\lambda \ge 0$Risk-aversion parameter on portfolio variance

Ellipsoidal uncertainty set

Uncertainty set on $\mu$ $$ \mathcal{U}_\kappa \;=\; \bigl\{\, \mu \in \mathbb{R}^n \;:\; (\mu - \mu_0)^\top \Sigma^{-1} (\mu - \mu_0) \le \kappa^2 \,\bigr\} $$
Ellipsoid centred at $\mu_0$ with shape matrix $\Sigma^{-1}$; $\kappa$ is in units of standard deviations. Under a Gaussian prior on $\mu$, an $\kappa\sigma$ ellipse is a credible region at confidence $\chi^2_n(\kappa^2)$.

Robust objective

Maximise the worst-case expected return inside $\mathcal{U}_\kappa$, still subject to a variance penalty:

Worst-case problem $$ \max_{w} \; \min_{\mu \in \mathcal{U}_\kappa} \; \mu^\top w \;-\; \lambda\, w^\top \Sigma w $$ $$ \text{s.t.}\quad \mathbf{1}^\top w = 1, \; w_i \ge 0 $$

Inner worst case in closed form

The inner minimisation over the ellipsoid has a closed form:

Worst-case $\mu^\top w$ over $\mathcal{U}_\kappa$ $$ \min_{\mu \in \mathcal{U}_\kappa} \mu^\top w \;=\; \mu_0^\top w \;-\; \kappa\, \lVert \Sigma^{1/2} w \rVert_2 $$
Derivation: Lagrangian of the inner minimisation yields $\mu^* = \mu_0 - \kappa\, \Sigma w / \lVert \Sigma^{1/2} w \rVert_2$.

SOCP reformulation

Robust Markowitz (Goldfarb-Iyengar 2003) $$ \max_{w} \quad \mu_0^\top w \;-\; \kappa\, \lVert \Sigma^{1/2} w \rVert_2 \;-\; \lambda\, w^\top \Sigma w $$ $$ \text{s.t.} \quad \mathbf{1}^\top w = 1, \quad w_i \ge 0 $$
Quadratic in $w$ plus a cone term $\lVert \Sigma^{1/2} w \rVert$; representable as an SOCP and solvable by Mosek, Gurobi, CVXPY. As $\kappa \to 0$ reduces to nominal Markowitz; as $\kappa$ grows, allocations flatten toward equal-weight / minimum-variance.

Equivalence to min-CVaR under Gaussian returns

When returns are jointly Gaussian, robust worst-case and mean-CVaR portfolios coincide for appropriate choices of $(\kappa, \alpha)$: $\lVert \Sigma^{1/2} w \rVert$ is proportional to the standard deviation of $\mu^\top w$, and CVaR of a Gaussian is $\sigma \phi(\Phi^{-1}(\alpha))/(1-\alpha)$. For non-Gaussian returns the two diverge — see the CVaR sub-application.

Interactive solver

Compare nominal Markowitz to robust as $\kappa$ varies

Robust vs Markowitz

Projected-gradient SOCP surrogate
8
0.80
5.0
0.40
42
Markowitz return
Robust return
Mkw std dev
Robust std dev
Turnover (L1)

Gold bars: nominal Markowitz ($\kappa = 0$). Blue bars: robust ($\kappa$ as set above). Watch the robust weights flatten out as $\kappa$ grows — shrinkage toward diversification.

Risk-return plane ($\sigma, \mathbb{E}[r]$): individual assets as dots; Markowitz and robust portfolios as starred points. The robust portfolio typically sits below-left of the nominal.

Assumptions & trade-offs

Price of robustness, calibration, variants

Modelling assumptions

  • Uncertainty only on $\mu$ in this page; Goldfarb-Iyengar also allow $\Sigma$ uncertainty via a matrix-valued ellipsoid (non-convex in general, convex under factor-model structure).
  • Ellipsoidal shape. Chosen because it yields tractable SOCP. Polyhedral sets (box or budget) yield LP formulations; discrete sets are NP-hard in general.
  • Single period, no transaction costs. As with Markowitz / CVaR.
  • $\kappa$ is a modelling choice, not estimated. Literature suggests $\kappa$ can be linked to the bootstrap standard error of $\mu_0$ or to a Bayesian credible region.

The price of robustness

Ben-Tal & Nemirovski proved that on convex problems, robustness induced by an ellipsoidal set is affordable: worst-case performance improves materially while nominal performance degrades only mildly, provided $\kappa$ is calibrated to actual parameter uncertainty rather than worst imaginable extremes. Choose $\kappa$ too aggressively and you underperform even the equal-weight portfolio; choose too conservatively and you recover Markowitz’s fragility.

Key references

Robust-optimisation theory and finance applications

Goldfarb, D. & Iyengar, G. (2003).
Robust Portfolio Selection Problems.
Mathematics of Operations Research, 28(1), 1–38. doi:10.1287/moor.28.1.1.14260
Ben-Tal, A. & Nemirovski, A. (1998).
Robust Convex Optimization.
Mathematics of Operations Research, 23(4), 769–805. doi:10.1287/moor.23.4.769
Ben-Tal, A., El Ghaoui, L. & Nemirovski, A. (2009).
Robust Optimization.
Princeton University Press. ISBN 978-0-691-14368-3.
Fabozzi, F. J., Kolm, P. N., Pachamanova, D. A. & Focardi, S. M. (2007).
Robust Portfolio Optimization and Management.
Wiley. ISBN 978-0-471-92122-6.
Michaud, R. O. (1989).
The Markowitz Optimization Enigma: Is 'Optimized' Optimal?
Financial Analysts Journal, 45(1), 31–42. doi:10.2469/faj.v45.n1.31
Black, F. & Litterman, R. (1992).
Global Portfolio Optimization.
Financial Analysts Journal, 48(5), 28–43. doi:10.2469/faj.v48.n5.28
Tütüncü, R. H. & Koenig, M. (2004).
Robust Asset Allocation.
Annals of Operations Research, 132(1), 157–187. doi:10.1023/B:ANOR.0000045281.41041.ed

Related sub-applications

Reminder
Not investment advice. Uncertainty-set calibration is itself a statistical problem; model outputs are not forecasts.