Hidden Markov Modelling of Earthquake-Related Market Regimes and State-Dependent Forecasts: Evidence from the Turkish Stock Market

Abstract This study investigates whether major earthquakes are systematically associated with changes in the latent return and volatility regimes of the Borsa Istanbul 100 (BIST 100) index. The empirical sample contains 4,334 daily index observations from 5 January 2009 to 3 April 2026. A three-state Gaussian hidden Markov model is estimated for daily logarithmic returns, and the resulting states are interpreted ex post as crisis, uncertainty and normal regimes. The earthquake catalogue covers the common period from 19 June 2009 to 27 October 2025. Ninety-six events with magnitude M ≥ 5.5 are reduced to 74 representative events through a 48-hour main-event filter and are aligned with the contemporaneous or next trading day. The crisis regime is characterized by a mean daily return of -0.79%, volatility of 3.45% and an expected duration of 5.3 days. Earthquake-linked trading days are distributed across the crisis, uncertainty and normal regimes in counts of 7, 35 and 32, respectively. Their distribution does not differ significantly from the unconditional state composition, and the reported magnitude-group analysis does not establish an association with next-day regime changes. State-conditional forecasts nevertheless reveal substantial differences in short-horizon downside risk. The pooled results do not establish a systematic earthquake regime association, whereas the fitted model yields state-dependent predictive risk distributions. A time-varying-transition Markov-switching extension is proposed for future causal and predictive analysis.