High-Frequency Volatility Modelling : A Markov-Switching Autoregressive Conditional Intensity Model

We develop a Markov-Switching Autoregressive Conditional Intensity (MS-ACI) model with time-varying transitional parameters, and show that it can be reliably estimated via the Stochastic Approximation Expectation-Maximization algorithm. Applying our model to high-frequency transaction data, we detect two distinct regimes in the intraday volatility process: a dominant volatility regime that is observable throughout the trading day representing the risk-transferring trading activity of investors, and a minor volatility regime that concentrates around market liquidity shocks which mainly capture impacts of firm-specific news arrivals. We propose a novel daily volatility decomposition based on the two detected volatility regimes