Detecting and explaining bubbles in Islamic stock markets: A dual approach with LPPLS and machine learning

This study investigates the presence and predictability of price bubbles in Islamic stock markets, challenging the proposition that their Sharia-compliant principles provide inherent resilience against such phenomena. Employing a dual methodology, we first apply the Log-Periodic Power Law Singularity (LPPLS) model to detect crash periods in the daily Dow Jones Islamic Market indices for Canada, Japan, the United Kingdom, and the United States from 1996 to 2025. Subsequently, we utilize an eXtreme Gradient Boosting (XGBoost) algorithm to identify the key macro-financial drivers of these identified bubble episodes. The results from the LPPLS analysis confirm that these indices exhibit significant bubble dynamics. The XGBoost model incorporated imbalance-aware learners and further reveals that the probability of a bubble is systematically linked to a combination of market-based and macroeconomic variables, with the stock price index, intraday volatility, long-term interest rates, and exchange rates emerging as the most significant predictors, albeit with country-specific variations.