Abstract
This study develops an explainable artificial intelligence framework to distinguish two questions: the strength of statistical dependence between FinTech and Shariah-compliant equity markets, and whether aligned sustainability screening adds an incremental association, termed dual-screening affinity (DSA). Using daily returns from 1 July 2016 to 1 April 2026, the matched two-by-two design compares the MSCI ACWI IMI Fintech Innovation Index (FT), the MSCI ACWI IMI Fintech Innovation Select Screened 100 Index (GFT), the S&P 500 Shariah Index (SH), and the S&P 500 ESG Shariah Index (ESGSH). Dependence is interpreted as statistical rather than causal and is examined through nonlinear association, contemporaneous and displaced-lag dependence, simultaneous downside co-exceedance, and conditional tail-state information. The framework combines a KSG mutual information lag spectrum, MINE and InfoNCE neural mutual information estimators, Bayesian downside co-exceedance, variational Bayesian deep-tail learning, and posterior Integrated Gradients. Results show strong nonlinear dependence between the selected FinTech and Shariah indices, concentrated predominantly at lag zero. Conditional Shariah tail probabilities during FinTech distress are approximately 0.70–0.71, compared with the 0.10 independence benchmark. The Bayesian deep-tail models achieve held-out ROC-AUC values of approximately 0.93–0.94 and materially outperform target-history-only Bayesian neural benchmarks. Because the baseline models include the contemporaneous FinTech return, this performance represents out-of-sample tail-state classification or nowcasting rather than one-day-ahead forecasting. Evidence for a positive Green FinTech-specific DSA is weak and model dependent: classical and neural MI estimates are negative or close to zero, Bayesian co-exceedance strengthens for both FinTech benchmarks, and the positive deep-tail affinity estimate remains highly uncertain. Explainability and ablation analyses show that the contemporaneous FinTech return accounts for approximately 42–48% of total attribution, while its removal reduces mean ROC-AUC from about 0.93 to 0.61. Overall, these selected and partly back-tested indices exhibit substantial broad statistical dependence but only weak incremental association with aligned sustainability screening. Differences in index construction and common market exposures remain unresolved; therefore, the estimates are descriptive and do not establish a causal or uniquely Green FinTech-specific mechanism.