Network Diversification for a Robust Portfolio Allocation
32 Pages Posted: 4 May 2022
Date Written: March 28, 2022
Abstract
Portfolio allocation strategies often seek risk budgeting and diversification by relying only on correlation matrices to model relationships between assets. Although this approach can capture, in normal times, most of the dependencies between asset prices, it faces several challenges in terms of noise resistance, capturing non-linear relations that can naturally appear in the market and extreme allocations in long-short portfolio strategies.
This paper presents novel network-based strategies that combine equal volatility allocation with network centrality measures to construct efficiently diversified portfolios and deliver stable strategies also suitable for long-short investments.
Networks can encode linear and non-linear relationships between asset prices. To encode several layers of information simultaneously multiplex networks - a particular form of a multilayer network - can be deployed. Associated centrality measures can agnostically account for each asset's (ir)relevance in diversifying the risks of the portfolios.
The results show that network-based portfolios can outperform several competing alternatives, maintaining a favourable risk characteristic.
Keywords: asset allocation, portfolio construction, graph theory, networks, multilayer networks, eigenvector centrality
JEL Classification: C15, G11, G15 ,G17, G0, G1, E44
Suggested Citation: Suggested Citation