Market-Regime-Analysis-using-Hidden-Markov-Models
A quantitative study using Hidden Markov Models to detect market regimes in the S&P 500 and evaluate regime-based trading strategies. Combines econometric modeling with adaptive decision rules to explore how volatility states influence risk and return dynamics.
Author: Tanvi Madhaw Institution: University of Groningen – Econometrics & Operations Research
📄 Overview
This project investigates regime‐switching dynamics in financial markets using a Hidden Markov Model (HMM) applied to the S&P 500 index. The model identifies periods of relative market stability and turbulence based solely on return behavior. In addition, the project tests regime-dependent trading strategies, both rule-based (fixed probability thresholds) and adaptive (reinforcement learning). The goal is to demonstrate how probabilistic models can capture nonlinear features of asset returns and how regime awareness can improve decision-making in quantitative investment frameworks.
⚙️ Methodology
-
Data Daily S&P 500 closing prices (2010–2024) are downloaded via Yahoo Finance and converted to log returns.
-
Model Estimation Gaussian Hidden Markov Models are estimated for 2–4 regimes. Model selection is based on the Bayesian Information Criterion (BIC). The chosen model produces posterior regime probabilities P_t(Bull) P_t(Bear) .
-
Regime Interpretation
Each inferred regime is characterized by its mean return, volatility, and persistence. Transition probabilities describe the likelihood of remaining in or switching between regimes.
- Trading Strategies Threshold Rules: Go long when P_t(Bull) > θ; remain in cash otherwise.
Adaptive Rule: A tabular Q-learning algorithm learns an optimal decision boundary between long and cash positions using regime probabilities as states.
- Evaluation Performance is measured via CAGR, Sharpe ratio, and volatility, benchmarked against a buy-and-hold strategy.
🧠 Key Findings
The S&P 500 can be statistically partitioned into two dominant regimes: Bull: higher mean returns, lower volatility, persistent state. Bear: negative or low returns, high volatility, short duration. Regime transitions align closely with major macro-financial events such as the COVID-19 crash (2020) and rate-hike cycle (2022). Regime-based strategies outperform unconditional buy-and-hold on a risk-adjusted basis. The Q-learning approach yields a smoother allocation policy than static thresholds, capturing nonlinear boundaries between optimism and caution.