基于经验似然法的事件研究

发布时间：2018-10-05 20:35

【摘要】：经过三十多年的发展,我国的金融市场在各个方面都取得了很大的成就。金融市场的功能日趋完善,覆盖面也越来越广,在这样的金融经济背景下,怎样来衡量一个经济事件对企业产生的影响显得尤为重要。事件研究法做为一种实证分析方法已在金融市场研究中得到了广泛的应用。但对于运用事件研究法研究经济事件或金融事件来说,假设检验方法的选择是至关重要的。事件研究中经常用到假设检验方法有参数检验、符号检验和秩检验,但这些方法的使用都有各自的约束条件。比如,参数检验实际上已经明确地假设了非正常收益的分布,检验的可靠性在很大的程度上会受到分布的限制；符号检验要求收益的分布具有对称性,但日收益数据常显示出非对称性；而秩检验的缺点是要估计方差,当方差估计不准时,秩检验的检验功效就会大打折扣。相比而言,经验似然方法在不依赖于数据分布假设的条件下就可以对数据进行分析；当数据来自非正态或方差估计不稳定时,经验似然法比其他方法通常要精确；经验似然法的结果在很多情况下具有稳健性,不易受样本中异常值的影响；其次,经验似然法不需要构建轴统计量,避免了估计方差的难题。因此,我们将经验似然方法引入到事件研究中来,以弥补上述检验方法的不足。本文选取了2012年我国证券市场上两类事件进行实证研究。一类是上市公司的违规处理,另一类是上市公司实施纯派现的股利分配政策。对于这两类经济事件我们分别采用了两种模型(常量均值模型和市场模型)来计算事件窗内的非正常收益,然后分别运用参数和非参数检验方法对非正常收益的显著性进行检验。最后,从模型层面和事件层面对参数检验和非参数检验的结果进行了对比分析。通过检验结果,我们发现在上市公司的违规处理事件中,非参数与参数检验结果差异较大,很难断定公告日后公司股票价格的持续波动是否与违规处理公告相关。产生这样的结果的原因可能是由于平均累积非正常收益在公告日之前就一直处于幅度较大的波动状态,公告日后波动幅度仍然较大,但是由于公告前的大幅波动造成了显著性检验中CAR显著趋于0,也就造成了对原假设的接受。在上市公司采取纯派现股利分配政策的事件中,非参数检验结果和参数检验结果相差不大,我们有理由认为上市公司纯派现的股利政策确实对股票价格的波动产生了较大的影响。
[Abstract]:After more than 30 years of development, China's financial market has made great achievements in all aspects. The function of the financial market is becoming more and more perfect and the coverage is becoming wider and wider. Under such a financial and economic background, how to measure the impact of an economic event on an enterprise is particularly important. As an empirical analysis method, event research has been widely used in financial market research. However, the choice of hypothesis testing method is very important for the study of economic or financial events by event research. Hypothesis testing methods are often used in event research, such as parameter test, symbol test and rank test. However, the use of these methods has its own constraints. For example, the parameter test has in fact clearly assumed the distribution of abnormal income, and the reliability of the test will be limited to a large extent by the distribution; the symbolic test requires that the distribution of income be symmetrical. The disadvantage of rank test is to estimate variance. When variance estimation is not punctual, the efficiency of rank test will be greatly reduced. In contrast, the empirical likelihood method can analyze the data without dependent on the assumption of data distribution, and the empirical likelihood method is usually more accurate than other methods when the data comes from non-normal state or the variance estimation is unstable. The results of empirical likelihood method are robust in many cases and are not easily affected by the outliers in the sample. Secondly the empirical likelihood method does not need to construct axial statistics to avoid the problem of estimating variance. Therefore, we introduce the empirical likelihood method into the event study to make up for the shortcomings of the above test methods. This paper chooses two kinds of events in China's securities market in 2012 to carry on the empirical research. One is the violation of listed companies, the other is the implementation of dividend distribution policy. For these two kinds of economic events, we adopt two models (constant mean model and market model) to calculate the abnormal returns in the event window. Then the significance of abnormal income is tested by parametric and non-parametric test methods. Finally, the results of parameter test and non-parameter test are compared and analyzed from model level and event level. Through the test results, we find that the non-parametric and parametric test results are quite different in the listed companies' irregularities, and it is difficult to determine whether the continuous fluctuation of the company stock prices after the announcement is related to the illegal treatment announcement. The reason for this result may be that the average accumulated abnormal income has been in a state of large volatility before the announcement date, and the volatility after the announcement is still large. However, due to the large fluctuation before announcement, CAR tends to zero significantly in the significance test, which leads to the acceptance of the original hypothesis. In the event that the listed company adopts the policy of pure dividend distribution, the results of non-parametric test and parameter test have little difference. It is reasonable to think that the dividend policy of the listed company has a great influence on the fluctuation of stock price.
【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】：F832.51;F224

【参考文献】