DEVELOPING A PORTFOLIO BASED ON FINANCIAL NETWORK TOPOLOGY BY USING MINIMUM SPANNING TREE (MST) APPROACH

Abstract

The stock market's inherent complexity, driven by continuous growth and diverse interactions, prompted a researcher to craft a market network using a Minimum Spanning Tree (MST) network. Traditional MST methods relying on Pearson correlation may overlook nonlinear financial data, potentially providing misleading insights. Thus, this study adopts Dynamic Conditional Correlation (DCC) to construct MST networks using historical prices of Shariah-compliant stocks from Bursa Malaysia. The main aim of this study is to build networks for Malaysia Shariah-compliant stocks from 2016 to 2020 using both Pearson and DCC methods, creating investment portfolios based on network centrality for comparison. Networks are constructed annually from 2016 to 2020, as well as an overall network covering the entire period. The investment portfolios are then categorized into peripheral and central based on their centrality. The findings indicate that when using Pearson correlation to construct the network, only AwanBiru Technology Berhad (AWAN), Jaya Tiasa Holdings Berhad (JTIA), Supermax Corporation Berhad (SUPR), and WTK Holdings Berhad (WTK) appeared twice within the top 10 positions annually and in the overall network during the study period. Conversely, Sapura Energy Berhad (SAPN) and Sasbadi Holdings Berhad (SASB) appeared only once in the general network. In contrast, with DCC, only VS Industry Berhad (VS) appeared twice within the top 10 positions annually and in the overall network. Jaya Tiasa Holdings Berhad (JTIA), Malaysia Steel Works Berhad (MAST), and Melewar Industrial Group Berhad (MELE), each appeared once and also in the general network. Notably, Jaya Tiasa Holdings Berhad (JTIA) appeared in both Pearson and DCC-based networks, highlighting its significance across both methods. The uncertainty regarding the superiority of one correlation method over the other led us to explore portfolio creation based on the network. This study finds that portfolios constructed with DCC outperformed those using Pearson correlations. Furthermore, portfolios constructed with peripheral network centrality outperform central ones. This research provides valuable insights and practical benefits to various stakeholders, such as brokers, investors, researchers, and academics. It presents an alternative approach to correlation-based network formation and aids in stock selection for investment or research purposes.