超高频期货市场微观结构噪音实证研究

发布时间：2018-01-02 18:00

本文关键词：超高频期货市场微观结构噪音实证研究　出处：《西南财经大学》2013年硕士论文　论文类型：学位论文

【摘要】：期货市场是金融市场和国民经济的重要组成部分,具有价格发现和套期保值两大重要功能。不仅对股票市场、外汇市场等其他虚拟经济有巨大的影响,也与现货市场直接关联,调控着实体经济的运行。随着近年来国际贸易和经济一体化的迅猛发展,期货市场跨越国界配置资源的作用越发显现,成为抢夺国际资源定价权和话语权的重要阵地。期货市场的地位在不断增强。因此,理解和研究我国期货市场的微观结构及其运作方式,对于揭示我国期货产品价格形成规律,理解期货市场波动率有重要的意义。认识市场微观构件如何影响市场价格变动趋势,可以直接运用到实际期货产品投资实践,有助于我们评价期货市场质量和效率,促使我国期货市场交易制度、规则的完善。市场微观结构理论的出现和发展,为深入理解和研究期货市场微观结构提供了方法和手段。莫琳·奥哈拉将市场微观结构的研究界定为“研究确定交易规则下资产交易的过程和结果”。强调从确定的交易机制之下的规则出发,来分析价格决定过程。麦德哈文(Ananth Madhavan,2000)将市场微观结构定义为“一个研究投资者潜在的交易需求转化为最终交易价格和交易量过程的金融学领域”。关注交易机制本身对价格形成的影响,是市场微观结构研究的一大特点。微观结构噪音的相关研究正是这样的背景提出和发展的。根据市场微观结构的相关理论,现实市场不同于完美信息市场,存在着交易成本、信息不对称等市场摩擦,证券价格在交易过程会受到这些摩擦因素的影响而偏离完美市场下资产的均衡价格。在交易过程中导致证券价格偏离均衡价格的因素的总和即为微观结构噪音。微观结构噪音是产生于交易过程的一系列市场摩擦,包括买卖价差、价格的离散变化、非同步交易、信息不对称、市场对大宗交易的逐渐反应、订单流的策略成分、市场参与者的流动性需要、市场参与者存货控制效应等。高频数据和超高频数据记录了市场所有的交易信息,其运用促进了微观结构噪音的研究,也为深入研究我国期货市场提供了重要的实证数据。传统的资产价格波动研究主要以低频数据为主,采样频率较低容易造成的信息流失,难以准确的刻画市场特征。高频数据和超高频数据包含了所有的交易信息,弥补了低频研究的不足。但超高频数据具有“采样频率高、不等的时间间隔、价格的离散变化、交易的周期模式、交易的多重性”等特性,使得低频下的波动估计不再是无偏估计。事实上,当采样频率越高,基于GARCH,SV模型不能取得较好的估计效果,RV模型的估计量的渐进一致性受到的影响越严重。一个主要的原因就是存在着市场微观结构噪音的干扰,上述波动估计量将不再收敛于积分波动率。Zhang等(2005)提出了二尺度已实现波动率(TSRV),充分利用全样本观测数据,将观测价格的波动分解为来自真实价格的波动和微观结构噪音的波动,得出了市场波动的渐近一致估计量。对市场微观噪音的研究,可以充分揭示具体交易规则之下,证券市场的价格行为多大程度上偏离了完美市场假设下的均衡价格行为。可以排除观测价格存在的非信息干扰,回归资产正确的价格。其次,应用微观结构噪音的的分离框架,可以有效估计价格过程的波动率状况,为寻求市场交易机会和构建投资组合提供帮助。应用噪音的研究,也能够形成对市场流动性和市场有效性的评价,从而帮助构建一个高效率、高质量的证券期货市场。鉴于目前国内在期货市场微观结构方面的研究相对匮乏,对微观结构噪音的实证研究更是几乎处于空白状态。本文以500ms高频采样数据和处理之后的得到的超高频交易数据为样本,研究期货市场价格波动状况,成交量变动规律,以及期货日内交易的一些特征。同时将引入TSRV方法对期货市场高频采样条件下的波动率进行估计,对微观结构噪音做出分离并分析其构成因素。本文的主要结构和安排如下：第一章是绪论部分。简述了期货市场微观结构噪音研究的研究背景和意义。第二章是相关理论和文献综述。第三章简要介绍了我国期货市场及其微观结构特征。第四章是文章的实证数据描述以及数据处理方法的介绍,包括了期货交易的日历效应和日内分时特征。第五章是本文的一个重点,介绍了微观结构噪音估计的原理和本文选取的TSRV噪音估计方法。然后基于逐笔交易数据对不同期货品种价格波动率和市场噪音水平进行了分离和估计。第六章分析了微观结构噪音的影响因素,并基于超高频数据实证回归分析了相关因素对噪音的贡献。第七章是本文的结语,总结了全文的研究和得出的结论,并就研究中存在的一些问题进行了总结,指出了未来可能的研究方向。最后是本文的参考文献,致谢和附录。本文研究得出的一些图表难以在正文中全部给出,可以在附录中找到。本文的研究主要得出了以下结论：第一,沪深300股指期货、天然橡胶、铜期货日历效应,各品种期货不同合约之间价格存在高度相关性,表现出同向变动趋势；股指成交量呈现当月交易活跃,其他时间交易较为清淡。临近交割日成交量先上升后下降的,类似于“M”型变化的趋势。天然橡胶期货日间成交量受现货市场的影响,存在着季节性变化的趋势。铜期货成交量与是否是主力合约有关。第二,本文分析了沪深300股指期货、天然橡胶、铜期货3个期货品种日内分时交易的特征。发现期货日内交易价格存在跳跃的行为,具有典型的离散变动特点,不能视为一个连续的价格过程。股指、铜期货的收益率在开盘和收盘附近具有类似“L”型或“倒L”型的特征,离开盘或收盘时点越近,收益率表现出较大幅度波动,随着距这些时点的间隔增加,收益率波动减小相对较为平缓。成交量有类似收益率的变动趋势。说明开盘和收盘附近存在较多的信息噪音交易。当市场处于一致的行情走势时,市场信息分歧较低,交易行为表现一致变动,日内分时特征将不明显。第三,利用已实现波动率RV、二尺度已实现波动率TSRV对市场的波动率进行了估计和比较,并对微观结构噪音进行了分离。发现以下结果：股指期货RV大致在104级别,TSRV大致在10-5级别,微观结构噪音在10-8和10-9两个级别变动。TSRV对市场噪音干扰的排除更优,是比RV更好的波动率估计量。微观结构噪音的变动趋势与成交量变动有较大关系,当成交量较低时,市场流动性价差,微观结构噪音相对较高,当交易变得活跃时,微观结构噪音下降。天然橡胶期货RV、TSRV都处于10-4水平,但显然TSRV的值更小,而其微观结构噪音水平在10-5,噪音水平与成交量有类似于股指期货的变化,成反向变动关系。铜期货RV、TSRV、微观结构噪音分别处于10-4、10-5、10-6水平。同时,微观结构噪音尖峰事件亦很好的解释了铜期货交易手续费调整和异常交易的现象。在本文的最后实证分析了微观结构噪音的影响因素,发现价差、交易规模对其有正的贡献,日内交易次数对其有负的贡献,TSRV对其无明显贡献,说明市场流动性与噪音有负相关关系。本文的创新之处在于：第一,首次运用500毫秒处理得出超高频数据分析了股指期货、天然橡胶、铜的日内分时交易特征。第二,首次引入了期货市场资产价格波动率和微观结构噪音的分离框架,并运用逐笔成交数据对股指期货、天然橡胶、铜期货三个品种的已实现波动率、二尺度已实现波动率、微观结构噪音水平分别作出了估计。第三,验证了TSRV模型在期货市场应用的稳健性,实证结果表明TSRV估计波动优于RV估计量。从而为高频条件波动率估计提出了一个参考指标,有助于更深刻地认识期货市场波动特征,更准备的构建投资组合实现风险控制。第四,实证分析了期货市场噪音的相关影响因素,验证价差、交易规模对微观结构噪音有正的影响,而日内交易频率与噪音负相关。第五,运用微观结构噪音尖峰事件,成功的解释了如市场微观结构调整(铜期货交易费率)、市场异常交易情况。本文的研究可以弥补目前在国内期货市场微观结构噪音定量研究上的空白,为以后的相关研究铺石垫瓦、抛砖引玉。
[Abstract]:The futures market is an important part of the financial market and the national economy, with the price discovery and hedging two important functions. Not only on the stock market, has great impact on the foreign exchange market and other virtual economy, and also directly related to the spot market, regulate the operation of the real economy. In recent years with the rapid development of international trade and economic integration the allocation of resources across borders, the role of the futures market is more and more, has become an important position to snatch international resource pricing and voice. The position in the futures market is growing.
Therefore, the understanding and research of China's futures market micro structure and mode of operation, to reveal China's futures price formation law, understand the futures market volatility has important significance. Know how to influence the market micro component market price movements, can be directly applied to the actual product futures investment practice, helps us to evaluate the futures market the quality and efficiency, promote China's futures market trading system, the perfection of rules.
The emergence of market microstructure theory and development, methods and means are provided for the in-depth understanding and study of futures market microstructure. Maureen O Hara will study the microscopic structure of the market is defined as "the process and results of" asset transactions to determine the transaction rules. Under the stress from the trading mechanism to determine the rules of price analysis in the decision process. Neil (Ananth Madhavan, 2000) will define the market microstructure for transforming a potential transaction demand for investors of finance "the final transaction price and transaction volume process. Pay attention to the influence of trading on the price formation mechanism itself, is a major feature of the market microstructure research.
The related research of microstructure noise is the background and development. According to the theory of market microstructure, the real market is different from the perfect information market, there is a transaction cost, information asymmetry and other market frictions, the equilibrium price of stock price will be affected by the friction factors and deviate from the perfect market assets in the transaction process. In the trading process, resulting in a total stock price deviates from the equilibrium price of the factors is the microstructure noise. The microstructure noise is a series of market frictions in the trading process, including the sale price, discrete changes in price, non synchronous trading, information asymmetry, market gradually in response to the bulk of the transaction, component order flow the liquidity needs of market participants, market participants inventory control effect.
High frequency data and ultra-high frequency data have recorded all trading information in the market, and their application has promoted the research of microstructure noise, and provided important empirical data for further study of China's futures market.
Study on the fluctuation of asset prices mainly in the traditional low frequency data, the sampling frequency is low easy to cause the loss of information, it is difficult to accurately describe the characteristics of the market. The high frequency data and high frequency data contains all of the transaction information, to make up for the lack of the low frequency. But the ultra high frequency data with high sampling frequency, unequal time interval, discrete changes in price, the transaction cycle mode, the characteristics of trading ", so that multiple low frequency fluctuation of the estimator is no longer the unbiased estimation. In fact, when the sampling frequency is high, based on GARCH, the SV model can not get good estimates of the effect, influence the progressive model by RV estimator the more serious. One of the main reasons is the existence of market microstructure noise interference, the fluctuation estimator will no longer converge to the integral fluctuation rate of.Zhang (2005) proposed two scale realized volatility (TSRV), we make full use of the whole sample observation data to decompose the fluctuation of observed price into the fluctuation of real price and microstructure noise, and obtain the asymptotic consistent estimator of market volatility.
Research on market microstructure noise, can fully reveal the specific transaction under the rules, the behavior of stock market prices of the extent to which deviates from the equilibrium price behavior under the perfect market assumption. You can exclude non interference observation price exists, the price return of assets right. Second, the separation framework of microstructure noise, can effective estimation of price process volatility, the market is looking for trading opportunities and portfolio construction help. Research and application of noise, but also to the formation of evaluation on the market liquidity and market efficiency, and help from the construction of a high efficient, high quality of the securities and futures market.
In view of the current domestic research in the futures market micro structure, the relative lack of empirical research on the microstructure noise is almost in a blank state. After taking the 500ms high frequency sampling data and processing the ultra high frequency trading data, research status of price fluctuation of futures market, change the volume, and some characteristics of futures in the deal. While the introduction of TSRV method on the futures market under the condition of high frequency sampling is used to estimate the volatility, make the separation and analysis of its constituent factors of microstructure noise.
The main structure and arrangement of this article are as follows:
The first chapter is the introduction part. The study of microstructure noise in the futures market research background and significance. The second chapter is the related theory and literature review. The third chapter briefly introduces the characteristics of China's futures market and its micro structure. The fourth chapter is the empirical data description and data processing methods in this paper are introduced, including the calendar effect of futures trading and the days when feature.
The fifth chapter is the emphasis of this paper, this paper introduces the TSRV noise principle and microstructure noise estimation method to estimate the selection. Then the separation and estimation of different futures price volatility and market transaction data based on the level of noise.
The sixth chapter analyzes the influencing factors of microstructure noise, and the empirical regression analysis based on the ultra high frequency data with related factors on noise. The seventh chapter is the conclusion of this thesis, summarizes and draws the research conclusion, and summarizes the problems existing in the research, pointed out the direction of future studies finally is this article references acknowledgements and appendix. Some charts are drawn in this paper it is hard to give all in the text, can be found in the appendix.
The main conclusions of this paper are as follows:
First, the Shanghai and Shenzhen 300 stock index futures, natural rubber, copper futures calendar effect, the futures contract between different prices are highly correlated, showed the same change trend; the stock volume presents the active transactions, other time relatively light trading. Near the settlement volume increased after the first drop, similar to the "M" type change trend natural rubber futures day trading volume. Affected by the stock market, there is a seasonal variation trend. Copper futures trading volume and is the main contract.
Second, this paper analyzes the CSI 300 stock index futures, natural rubber, copper futures trading features 3 days of time. It is found that the trading prices of futures intraday skipping behavior, has the typical characteristics of discrete changes, can not be regarded as a continuous process. The price of stock index futures, the rate of return has characteristics similar to "L" or "inverted L" type in the vicinity of the opening and closing of the left disc or closing point closer, yield showed a relatively large fluctuation, with the distance from the point of the interval is increased, the return volatility decreases relatively smooth. Volume change trend similar yields. It shows that there are more near the opening and closing the information of noise trading. When the market is in consistent with the market trend, the differences of market information is low, trading behavior changes consistent, days when feature is not obvious.
Third, the realized volatility of RV two scale realized volatility TSRV market volatility was estimated and compared, and the microstructure noise were separated. Find the following results: RV stock index futures at approximately 104 level, TSRV is at the 10-5 level, the microscopic structure of the noise interference of noise in the market 10-8 and 10-9 two level changes of.TSRV better, is a rate estimator is better than RV fluctuations. Have a greater relationship of microstructure noise change trend and volume changes, when the volume is low, market liquidity spreads, microstructure noise is relatively high, when the transaction becomes active when the microstructure noise decreased. Natural rubber futures RV, TSRV at 10-4 level, but obviously TSRV is smaller, and the microscopic structure of the noise level in the 10-5, the noise level and volume changes similar to the stock index futures, the reverse change relationship between copper futures of RV. , TSRV, microstructure noise respectively in the 10-4,10-5,10-6 level. At the same time, the microstructure noise spike event was also a good explanation of the copper futures transaction fee adjustment and abnormal trading phenomenon.
At the end of this paper, we empirically analyze the influencing factors of microstructure noise. We find that the price difference and the scale of transaction have positive contributions to it. The number of intra day trading has a negative contribution to it, and TSRV has no significant contribution to it, which indicates that market liquidity is negatively correlated with noise.
The innovation of this article lies in:
First, for the first time by 500 milliseconds that the ultra high frequency data analysis of stock index futures, natural rubber, copper within transaction characteristics. Second, first introduced the framework of separation of asset price volatility and futures market microstructure noise, and the use of transaction transaction data of stock index futures, natural rubber, copper has achieved three varieties of two scale volatility, realized volatility, microstructure noise levels were estimated. Third, to verify the robustness of the TSRV model application in the futures market, the empirical results show that TSRV is superior to RV estimation of volatility estimator. Thus rate estimation is proposed as a reference index for high frequency fluctuations, contribute to a more profound to understand the fluctuation characteristics of the futures market, build a portfolio to achieve risk control. Fourth, empirical analysis of the factors related to the effects of noise, the futures market price verification, trading rules Die has a positive impact on the microstructure noise, and the noise trading frequency and negative correlation. Fifth days, the microstructure noise spike event, such as the successful interpretation of market microstructure adjustment (copper futures rate), market abnormal trading.
The research of this paper can make up the blank of the quantitative research on the micro structure noise in the domestic futures market.

【学位授予单位】：西南财经大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：F724.5;F224

【参考文献】