Applications of Network Theory to North America Stock Markets

Abstract

This dissertation applies network theory to investigate the structure, direction, and dynamics of systemic risk in North American equity markets, focusing on the Toronto Stock Exchange (TSX 60) and Nasdaq 100. The research combines three complementary analyses to provide a deep understanding of market connectedness across pre-COVID, pandemic, and post-pandemic periods. First, partial correlation and mutual information concepts are employed to identify edges in order to construct an undirected network for each stock market, namely the TSX 60 and the Nasdaq 100. We use several network analysis techniques, such as graphical LASSO (GLASSO), latent variable time-varying graphical lasso (LTGL), and minimum spanning tree (MST), to reveal key patterns of interconnectedness and shifts in the stock market across distinct sample periods: the “pre-COVID-19”, “COVID-19”, and “post-COVID-19” periods. Findings reveal shifts in sectoral centrality and the emergence of systemically important nodes during periods of market stress. Second, the study transitions to directed network analysis, applying a range of methodologies, including net transfer entropy, the Diebold-Yilmaz (DY) spillover index and empirical mode decomposition, to identify edges and construct a directed volatility-spillover network. The findings reveal the distinctive structures of the two stock markets in both short-term and long-term frequency data. Both the TSX 60 and Nasdaq 100 networks demonstrate a robust interconnectedness and high volatility transmission during the pandemic, reflecting heightened market uncertainty. Furthermore, the study compares the net transfer entropy and the DY spillover index and finds that the DY spillover index is also capable of identifying the volatility spillover in a smoothed pattern from a long-term frequency data. Finally, a novel integration of the directed network framework with a Susceptible - Infected - Recovered (SIR) epidemic model is used to simulate the dynamic propagation of financial contagion. By calibrating transmission (β) and recovery (γ) rates from empirical data, the model shows how network topology and sectoral positioning influence contagion speed and scale. The dissertation advances the literature on financial networks, systemic risk, and contagion modeling by uniting static structure, directional influence, and dynamic simulation within a single framework. The results provide actionable insights for policymakers, regulators, and investors aiming to identify vulnerabilities, design targeted interventions, and strengthen market resilience against future shocks.