关键词: 混频数据, 自适应权重, 波动率, 渐近性质

Abstract: In this paper, a new way for extracting useful information from high-frequency data is constructed through introducing an adaptive weight function, and a new volatility GARCH model with the ability to use mixed frequency data is proposed by combining the low-frequency data with the information extracted from high-frequency data. The parameter estimation method for the suggested model is proposed and the asymptotic property of the suggested estimator is discussed in detail to obtain the corresponding central limiting theory. Simulation studies are employed to test the performance of the constructed estimator. The proposed model’ s advantages are as follows: First, the constructed model is different from the traditional ones to integrate data in which the distribution of weight depends on the order of data. Due to the fact that the independent variable of the suggested weight function can describe the character of a high-frequency transaction, the constructed model based on the new weight function will assign weight according to the character of the transaction. This assignment of weight can automatically regulate its formation according to the variation of transaction character in different trading days and the assigned weight for each high-frequency transaction is consistent with its impact on future volatility. Hence, the proposed model can exploit high-frequency data more properly. Second, more than one high-frequency data series produced in the same transaction is employed in the proposed model, which means that our model makes full use of existing data resources. These advantages of the mixed frequency GARCH model lead to better performance in predicting volatility, which has been confirmed by empirical results in the paper. Various models are used simultaneously to predict the volatility of real transaction data. The results show that the proposed model can predict volatility more accurately and robustly. The new model will enrich the technique of using mixed frequency data to analyze volatility and the related issues.

Key words: Mixed Frequency Data, Adaptive Weight, Volatility, Asymptotic Property