Open Finance, Complexity, and Institutional Survival: A Graph-Based Approach to Sustaining Financial Inclusion in Latin America

Abstract This paper develops a network-theoretic framework for analyzing institutional resilience within open finance ecosystems, focusing on the coordination structures that influence the survival of regulated entities under evolving digital architectures. Drawing on open supervisory data from Brazil and Mexico, we construct financial visibility networks and estimate institutional hazard through a hybrid model that integrates Graph Convolutional Networks (GCNs) with DeepSurv survival analysis. We find that institutions occupying peripheral network positions—characterized by low degree centrality and high modularity—face a hazard ratio of 2.31 (p < 0.01) for early exit within their first two years, compared to structurally embedded peers. The GCN–DeepSurv approach outperforms traditional Cox models, improving concordance by 12.7% and reducing integrated Brier scores by 18%. While regulatory interoperability reforms increase overall network cohesion, fragility persists among marginal actors. By embedding topological features into survival modeling, the study bridges advances in graph machine learning with supervisory economics, reconceptualizing financial inclusion in terms of structural survivability. The framework offers a replicable and policy-relevant early warning tool, contributing to the literature on economic interaction, coordination, and resilience in complex financial systems. JEL: G21; L51; O33; C63; O54