Market-Based Risk Dynamics in Eco-Resource Financial Sectors and Energy Finance: Evidence from Conventional and Islamic Real Estate Assets Using TVP-VAR and LSTM-NN

This study examines whether conventional and Islamic real estate indices are associated with different patterns of financial connectedness and long-memory behavior in selected eco-resource sectors. The analysis focuses on four resource-related financial markets—water, food, agriculture and livestock, and reduced-energy sector exposure—and evaluates how the inclusion of different real estate indices changes the connectedness structure of this system. Bayesian Time-Varying Parameter Vector Autoregression (TVP-VAR) is used to estimate time-varying connectedness and spillover dynamics, while Long Short-Term Memory Neural Networks (LSTM-NN) are applied as a complementary tool to assess long-memory and forecasting-related patterns in the connectedness series. Compared with using either method alone, this design captures both the evolving network structure of market-based risk transmission and the persistence of connectedness patterns over time. Using market data from 20 September 2016 to 9 January 2026, the results show that conventional real estate indices are generally associated with stronger connectedness in the eco-resource financial network, suggesting greater potential for market-based risk transmission. In contrast, Islamic real estate indices exhibit comparatively lower connectedness and more persistent long-memory behavior in the examined sample. These findings indicate that real estate asset heterogeneity matters for understanding financial connectedness among selected sustainability-related sectors. The study contributes to sustainable finance by showing how conventional and Islamic real estate assets may play different roles in the financial connectedness of resource-related markets.