The Autoregressive Conditional Duration Model And Empirical Research

Researches in the market microstructure focus on the processes of discovering the trading prices and the operating mechanisms of the trading prices. Low frequency trading data is far from enough for this king of researches. With the rapid development of computer science, it makes possible for people to research into the market microstructure with high frequency data. However, We may encounter many problems when we analyze the high frequency data with the existing econometric models. Therefore researches of the econometric methods for analyzing the high frequency data are of great importance. The empirical researches indicate that the Autoregressive Conditional Duration Model exhibit superior features in dealing with the high frequency financial data. The development of the model and its empirical research can help people understand the market operating mechanism and provide guidance in constructing market system in China. Relevant researches conducted in China have not been retrieved in literature. The present research is conducted in the following logic sequences. Firstly, this thesis extends the hypotheses of the distribution of the model; and further discusses the estimating methods for the model. Considering the microstructure of the stock market in China, this thesis selects 18 of 180 index stocks; and the pretreatment of the data indicates the model is valid. Then, using these data, the empirical research is conducted to identify the reasonable distribution of the model. The empirical result indicates that the ACD model does a good job in analyzing the high frequency data. With a new variable of the microstructure of the market and an extended ACD model, the empirical research confirms some important hypotheses of the microstructure of China's financial market. To further explore the characteristics of the microstructure of the stock market, this article also discusses the measurement of liquidity. This thesis shows: (1) the intraday duration has the time-of–day effect, and presents a trend of U-shape; (2) the ACD model can well capture the durations dynamics of our financial trading data, with one lagged model, the Weibull distribution has excellent applicability; and (3) the VNET model is very efficient in measuring the liquidity of the stock.