Bayesian inference for time varying partial adjustment model with application to intraday price discovery

Price discovery is an important built-in function of financial markets and the central issue in the market microstructure research. Market participants need to know whether the price discovery has been achieved or how much progress has been made in order to trade at an appropriate price they consider. Since various economic events such as earnings announcement affect the price discovery, the intraday transition of price discovery varies date-by-date. In this study, we propose a statistical method to see when and how fast the intraday price discovery progresses using the high frequency price series on a daily basis. The proposed method consists of estimating three candidate models which gauge the different types of price discovery progress, i.e. no progress, smooth progress and abrupt progress, and selecting the most appropriate model based on Bayesian approach. We conduct simulation analysis to assess the performance of our proposed method and confirm that the method depicts the state of price discovery appropriately. The empirical study using the Japanese stock market index shows that the proposed method well categorizes the intraday price discovery progresses on a daily basis.