REGIME DETECTION IN THE INDIAN STOCK MARKET: AN HMM APPROACH WITH IMPLIED VOLATILITY

Abstract

This study investigates regime dynamics in the Indian equity market using a bivariate Hidden Markov Model that incorporates both NIFTY 50 returns and the India VIX as observation variables. Daily data spanning January 2010 to December 2025—a period encompassing the pre COVID expansion, the pandemic shock, and subsequent normalization—are employed to classify market conditions into three latent states. The estimated regimes are economically interpretable as Bull (positive returns, low volatility, subdued VIX), Bear (near zero returns, moderate volatility, elevated VIX), and Crisis (negative returns, high volatility, extreme VIX). The transition matrix reveals high persistence across all states and a near zero probability of direct Bull to Crisis transitions, indicating that market stress evolves gradually. Robustness checks—including winsorized returns, parametric bootstrap confidence intervals, and benchmark comparisons—confirm the stability of the classifications. Out of sample forecasts demonstrate that the model successfully anticipates crisis episodes with perfect sensitivity. A tactical allocation strategy based on the regime forecasts improves the Sharpe ratio by 48% and reduces the maximum drawdown from 40.0% to 16.8% relative to a passive buy and hold benchmark. The findings underscore the value of incorporating implied volatility into regime detection frameworks and provide actionable insights for risk management and dynamic asset allocation in emerging markets.