Abstract
This study develops an integrated behavioral-theoretical and explainable machine learning framework to investigate how global information uncertainty affects real estate stock valuation in emerging markets. Grounded in Knightian uncertainty and behavioral finance theory, the analysis employs a panel dataset of 28 listed Vietnamese real estate firms from 2013 to 2024, incorporating firm fundamentals, macroeconomic controls, and four key uncertainty indices: ESG Uncertainty (ESGUI), Geopolitical Risk (GPR), World Uncertainty (WUI), and Media Sentiment (WSI). Nonlinear predictive models, particularly XGBoost optimized via Particle Swarm Optimization (PSO), are interpreted using SHAP and LIME to uncover both global and local effects. Results indicate that firm size and book value are the most influential valuation drivers, while global uncertainty indices demonstrate conditional effects. Notably, the interaction between WUI and firm size reveals an inverse U-shaped pattern, supporting the “asymmetric ambiguity response” hypothesis: under heightened uncertainty, investors gravitate towards large firms. The proposed framework offers theoretical consistency, predictive accuracy, and interpretability, contributing to the literature on asset pricing under uncertainty. Findings provide practical insights for ESG communication, risk mitigation, and regulatory design in opaque and sentiment-sensitive emerging markets.