2019/05/10 by Xing Gu, Rogemar Mamon, Matt Davison +2
Decision Sciences · Economics, Econometrics and Finance · #Artificial intelligence #Complex Systems and Time Series Analysis #Computer science #Econometrics #Economics #Finance #Financial Markets and Investment Strategies #Financial economics #Financial market #Geography #Hidden Markov model #Index (typography) #Liquidity risk #Machine learning #Market liquidity #Multivariate statistics #Stock Market Forecasting Methods #Treasury
paper · doi:10.1080/00207179.2019.1616225
openalex publication_date 2019/05/10 · openalex created_date 2025/10/10 · openalex updated_date 2026/08/01
A multivariate hidden Markov model (HMM)-based approach is developed to capture simultaneously the regime-switching dynamics of four financial market indicators: Treasury-Euro Dollar rate spread, US dollar index, volatility index and S&P 500 bid-ask spread. These indicators exhibit stochasticity, mean reversion, spikes and state memory, and they are deemed to drive the main characteristics of liquidity risk and regarded to mirror financial markets' liquidity levels. In this paper, an online system is proposed in which observed indicators are processed and the results are then interfaced with an advanced alert mechanism that gives out appropriate measures. In particular, two stochastic models, with HMM-modulated parameters switching between liquidity regimes, are integrated to capture the evolutions of the four time series or their transformations. Parameter estimation is accomplished by deriving adaptive multivariate filters. Indicators' joint empirical characteristics are captured well and useful early warnings are obtained for occurrence prediction of illiquidity episodes.