Research projects

In this study, as a model to explain the generating mechanism of limit orders (orders that specify the bid or ask price) in financial markets, we proposed an extension of the ACD (Autoregressive Conditional Duration) model as well as the SCD (Stochastic Conditional Duration) model in which intraday seasonality and limit order book information (the price and quantity of limit orders) are incorporated. We also developed a new efficient algorithm for Bayesian estimation of the proposed models via Markov chain Monte Carlo. We estimated the proposed models with the data of limit orders in the Tokyo Stock Exchange, and examined influences of indicators related to the market liquidity upon time intervals between limit orders.

In this study, we propose a novel estimation technique for time series models of financial high-frequency data. Specifically, we consider two types of time series models; one is a model of duration between executions of financial transactions while the other is a model of time-varying volatility (variance) in very short intervals. To make these models more realistic, we propose to incorporate intraday seasonality (a cyclical pattern of duration or volatility during trading hours) explicitly into both models and estimate it simultaneously with the model parameters. Since the proposed models are too complex to be estimated with traditional maximum likelihood estimation, we developed an efficient Bayesian Markov chain Monte Carlo (MCMC) method for these models. We applied our new method to real-world high-frequency data (commodity futures and stock prices) and demonstrated their advantage over the conventional models.