Abstract
The research aims to analyse the effect of correlation method on the Minimum Spanning Tree (MST)
properties and to measure the performance of the resulting portfolios. We have employed a MST to
understand the dynamics of the South African market. As a result, we discovered relevant patterns
for risk management, portfolio design, and trading strategies. Specifically, we clustered some shares
of the Johannesburg Stock Exchange (JSE) from 2017 to 2022. We used three correlation methods
and centrality measures for portfolio selection. Sharp drops in the tree’s normalised length were observed
in 2018 and 2020. These drops may be attributed to market shocks experienced during these
years. Peripheral stocks were grouped into portfolios, and their performances were measured and compared
to a benchmark. Rank correlation portfolios outperformed the benchmark, while the benchmark
outperformed the traditional Pearson.
Keywords: Networks, Minimum spanning tree, Correlation, Centrality, Diversification, Portfolio, Backtesting