非对称乘积误差模型及其应用

发布时间：2018-06-18 17:41

本文选题：乘积误差模型 + 非对称性　；参考：《西南财经大学》2014年硕士论文

【摘要】：自从乘积误差模型(MEM)提出后,国内外涌现了许多对MEM模型的研究,国内主要是运用MEM模型结合中国金融市场进行实证研究,关于非对称MEM模型的研究较少。事实上,在金融市场上的非负值时间序列大多都存在非对称性,我国金融市场也不例外,比如波动的不对称性、金融持续时间的非对称性等。本文基于MEM模型的基本理论,借鉴非对称ACD模型的思想和结构,探讨非对称MEM模型,将非对称结构纳入一般MEM里面,构建非对称MEM模型,并用极大似然法对经典乘积误差模型与非对称乘积误差进行估计,利用蒙特卡洛模拟比较这两种模型对具有非对称性的非负值时间序列的刻画能力。最后结合我国金融市场进行实证分析。这是对乘积误差模型理论的深入研究,弥补MEM模型对称性的缺陷,同时为非负值时间序列研究提供了一个有力的工具。本文主要研究工作： (1)根据经典的MEM模型,结合非对称GARCH模型和非对称ACD模型的构建思想,探索非对称MEM模型结构,建立非对称MEM模型,并对该模型误差项分布的选取以及模型参数估计进行介绍。 (2)为了比较经典MEM模型与非对称MEM模型对具有非对称效应的金融时间序列的刻画能力,本文运用蒙特卡洛模拟方法按照一定的数据生成过程,随机生成具有非对称性的非负值时间序列,然后通过经典MEM模型与非对称MEM模型的估计结果,根据预测能力指标比较两种模型对具有非对称性的非负值时间序列的刻画能力,结果发现非对称MEM模型能更好的刻画非负值时间序列。 (3)金融高频数据包括价格持续期、交易持续期、最高价、最低价等金融时间序列都具有典型性特征,比如高峰厚尾、自相关和长记忆性以及日内效应等基本特征,本文选取招商银行成交量、最高价、最低价等时间序列作为研究对象,验证招商银行的高频数据是否具有这些典型性特征,分析结果表明成交量、最高价、最低价都具有高峰厚尾、自相关和长记忆性以及日内效应等典型性特征。 (4)本文选取招商银行2014年年初的每隔5分钟分时的高频交易数据,选取成交量、最高价、最低价等时间序列进行研究,首先根据高频数据的典型特征,对研究对象进行数据处理,对成交量消除日内效应,通过最高最低价格的变化计算出价格指示变量等,然后建立关于交易强度的非对称MEM模型,目的是为了刻画价格变化对交易强度的非对称影响,运用极大似然估计对非对称MEM模型进行估计。实证分析结果表明：从交易强度的动态运动过程可得出,交易强度有较强的集聚特征,从模型结果发现,无论是价格正向变化还是负向变化,都会使交易强度增加,即价格波动会促进市场交易,但价格变化方向对交易强度的影响程度不同,说明交易强度具有明显的非对称性。本文创新之处：通过对非对称MEM模型的研究,一方面是对非对称GARCH模型、非对称ACD模型的拓展,为研究证券市场非负值时间序列提供一个更有力的研究工具,同时丰富了MEM模型。另一方面我们将对我国金融市场中的交易强度的非对称性进行实证研究,国内对交易强度的研究较少,主要是研究市场因素对交易强度的长短期影响,因此,本文的实证研究对于了解交易制度和市场结构对投资者和市场交易活动的影响,对于完善我国证券市场的监管具有重要的实际应用价值。本文由2011年度国家自然科学基金青年科学基金项目《新兴订单驱动市场非负值金融时间序列的乘积误差建模及应用研究》(71101118)资助完成。
[Abstract]:Since the product error model (MEM) is proposed, many researches on MEM models have emerged at home and abroad. The domestic research is mainly about the use of the MEM model and the Chinese financial market, and the research on asymmetric MEM model is less. In fact, most of the non negative time series in the financial market are asymmetric, and our financial market is also the same. No exception, such as the asymmetry of volatility and the asymmetry of financial duration. Based on the basic theory of the MEM model and using the ideas and structures of the asymmetric ACD model, the asymmetric MEM model is discussed, the asymmetric structure is incorporated into the general MEM, and the asymmetric MEM model is constructed, and the classical product error model and the nonsymmetric model are used by the maximum likelihood method. The estimation of the symmetric product error is made by using Monte Carlo simulation to compare the characterization of the two models for non negative time series with non negative values. Finally, an empirical analysis is made with the financial market in China. This is an in-depth study of the theory of the product error model, which makes up for the defects of the symmetry of the MEM model and is a non negative time series. Research provides a powerful tool.
The main research work in this paper is:
(1) according to the classical MEM model, combining asymmetric GARCH model and asymmetric ACD model, the structure of asymmetric MEM model is explored and asymmetric MEM model is established. The selection of error term distribution and parameter estimation of the model are introduced.
(2) in order to compare the characterizations of the classical MEM model and the asymmetric MEM model for the financial time series with asymmetric effects, the Monte Carlo simulation method is used to generate the non negative time series with non negative values according to a certain data generation process, and then the estimation knot of the classical MEM model and the asymmetric MEM model is obtained. According to the prediction ability index, the two models are used to characterize the non negative time series with non negative values. The results show that the non negative time series can be depicted better by the asymmetric MEM model.
(3) the financial high frequency data including the price duration, the duration of the transaction, the highest price, the lowest price and other financial time series all have the typical characteristics, such as the basic characteristics of the high peak, the autocorrelation and the long memory and the day effect. This paper selects the time series of the volume, the highest price and the lowest price of China Merchants Bank as the research object, and verifies the recruitment. Whether the high frequency data of commercial banks have these typical characteristics, the analysis results show that the volume, the highest price and the lowest price have the typical characteristics such as the peak tail, the autocorrelation and the long memory and the intraday effect.
(4) this paper selects the high frequency transaction data every 5 minutes at the beginning of 2014 of China Merchants Bank, and selects the time series of the volume, the highest price, the lowest price and so on. First, according to the typical characteristics of the high frequency data, the research object is processed, the intra day effect is eliminated and the price is calculated by the maximum minimum price. In order to describe the asymmetric effect of price change on transaction intensity, the asymmetric MEM model of transaction intensity is established. The maximum likelihood estimation is used to estimate the asymmetric MEM model. The empirical analysis shows that the dynamic process of trading intensity can be obtained from the dynamic process of trading intensity. Characteristics, from the model results, it is found that both the price positive change or the negative change will increase the transaction intensity, that is, the price fluctuation will promote the market transaction, but the price change direction has different influence on the transaction intensity, indicating that the transaction intensity has obvious asymmetry.
Innovation in this paper: through the study of asymmetric MEM model, one aspect is the expansion of asymmetric GARCH model and asymmetric ACD model, which provides a more powerful research tool for the study of non negative time series in the securities market, and enriches the MEM model. On the other hand, we will be asymmetrical to the transaction intensity in our financial market. In the empirical study, there are few studies on the intensity of transaction in China. It is mainly to study the influence of market factors on the long and short period of the transaction intensity. Therefore, the empirical study of this paper has an important practical value for understanding the influence of the trading system and market structure on the investors and the market transaction activities and improving the supervision and regulation of the securities market in China. Value.
This paper is funded by the project of product error modeling and application of the non negative financial time series of the emerging order driven market of the National Natural Science Foundation of the National Natural Science Fund of 2011 (71101118).
【学位授予单位】：西南财经大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：F224;F832.51

【参考文献】