A Three--Dimensional Efficient Surface for Portfolio Optimization

TL;DR

The paper addresses the limitation of the classical mean‑variance framework by incorporating network spillover risk through a symmetric spillover matrix $C$, yielding a three‑dimensional efficient surface in $(\mathbb{E}[r], \sigma, \kappa)$ with $\kappa = w^{\mathsf T} C w$. It develops a unified optimization $L_{\\lambda}(w)=\\lambda w^{\\mathsf T}\\Sigma w + (1-\\lambda) w^{\\mathsf T} C w$ and proves existence, uniqueness, and continuity of solutions, providing a closed‑form weight when $\\Sigma$ and $C$ are simultaneously diagonalizable; it also establishes a three‑fund separation into $w^{MV}$, $w^{MC}$, and $w^{\max\mu}$ under suitable conditions. The connectedness beta $\\beta_i^{(C)}=2 [C w]_i$ offers a CAPM‑like diagnostic for network risk, and a three‑fund representation clarifies how diversification changes when systemic spillovers are considered. Empirically, using S&P 500 data (2010–2024), portfolios with explicit connectedness constraints show improved downside protection during stress episodes, while high connectedness betas identify systemic hubs; a dynamic three‑fund strategy tracks the efficient frontier with favorable tracking error and cost characteristics. Overall, the framework provides a transparent theoretical basis for integrating systemic connectedness into portfolio choice and highlights how network structure can alter diversification incentives and risk management.

Abstract

The classical mean-variance framework characterizes portfolio risk solely through return variance and the covariance matrix, implicitly assuming that all relevant sources of risk are captured by second moments. In modern financial markets, however, shocks often propagate through complex networks of interconnections, giving rise to systemic and spillover risks that variance alone does not reflect. This paper develops a unified portfolio optimization framework that incorporates connectedness risk alongside expected return and variance. Using a quadratic measure of network spillovers derived from a connectedness matrix, we formulate a three-objective optimization problem and characterize the resulting three-dimensional efficient surface. We establish existence, uniqueness, and continuity of optimal portfolios under mild regularity conditions and derive closed-form solutions when short-selling is allowed. The trade-off between variance and connectedness is shown to be strictly monotone except in degenerate cases, yielding a well-defined risk-risk frontier. Under simultaneous diagonalizability of the covariance and connectedness matrices, we prove a three-fund separation theorem: all efficient portfolios can be expressed as affine combinations of a minimum-variance portfolio, a minimum-connectedness portfolio, and the tangency portfolio. The framework clarifies how network-based risk alters classical diversification results and provides a transparent theoretical foundation for incorporating systemic connectedness into portfolio choice.

Figures (3)

• Figure 1: Illustrative distribution of the top-15 connectedness betas $\beta^{(C)}_i$ on a single trading day (31 Dec 2024) using a randomly selected 100-stock subset of NYSE-listed stocks. The figure is intended solely as a didactic example to visualize the heavy-tailed nature of $\beta^{(C)}$; all formal empirical tests in Section \ref{['sec:empirical']} employ the full universe and rolling estimates.
• Figure 2: Barycentric representation of efficient portfolios spanned by the three corner funds—minimum variance (MV), minimum connectedness (MC), and tangency (Tan). Moving toward MC (resp. MV) lowers connectedness risk $\kappa$ (resp. variance $\sigma^{2}$). Dots represent the optimal portfolio for $\lambda\in\{0,0.05,\dots,1\}$ in the hybrid risk matrix $M(\lambda)=\lambda\Sigma+(1-\lambda)C$.
• Figure 3: Position of $w^{\ast}(0.4)$ relative to the simplex defined by the three corner funds. The point lies outside the shaded triangle, forcing at least one barycentric weight to be negative.

Theorems & Definitions (12)

• Proposition 1: Existence and (Conditional) Uniqueness
• proof
• Proposition 2: Closed‑Form Optimal Weights (short‑selling allowed; see Appendix A for the long‑only case)
• proof
• Proposition 3: Strict Trade–off: Negative Slope
• proof : Proof (concise convex–analytic argument)
• Proposition 4: Degenerate Case: $C=c\,\Sigma$
• proof
• Theorem 1: Conditional Three–Fund Separation
• proof