条件Beta定价模型在我国股票市场的实证研究—基于非参数方法

发布时间：2018-06-26 13:22

本文选题：条件beta定价模型 + Fama-French三因子模型　；参考：《浙江工商大学》2012年硕士论文

【摘要】：资本资产定价模型诞生半个世纪来,从非条件(静态)模型到现在的条件模型,各国学者不断通过各种检验对模型作出修正和发展。最近十多年来,我国学者对我国的股票市场,基金市场等也做了不少实证研究。总体来说,他们的研究多从修改变量和改变估计方法两个角度入手,得到的结论也多有出入。在对风险因子的选择上存在分歧,而在beta系数的研究上则以对其长期均值的显著性检验为主。本文借鉴了国外新提出的一种两步骤非参数回归分析方法对FF三因子模型进行了较细致的研究。这种方法的好处在于可以放宽对模型假设的限制,并利用迭代方法逐步逼近得到更接近真实值的估计值。本文便在此基础上对风险因子的市场溢价以及相应beta系数的变化趋势和性质做了较深入的研究,而不仅仅只是成立与否的检验。首先,本文从时间序列数据中用非参数方法估计出每个资产在每个时段的条件beta值。在这一步中,笔者参考了其他学者常用的四种状态变量,通过两两组合成了6组工具变量,并根据拟合的程度挑选出影响最大的两种变量。其次,本文从收益率和已得到的beta系数的横截面数据中估计出时变的风险因子市场溢价,并通过惩罚函数最小二乘法找出最优窗宽,对beta系数重新进行迭代估计,得到更精确的时变beta系数。然后,本文基于上述的实证结果对风险因子的市场溢价以及beta系数的性质做了详细的分析。笔者认为,FF三因子模型在我国股票市场是成立的,三个因子的存在是显著的,但并不是在所有时段均具有明显效应。另外,三种beta系数均随规模的变大而变小,但与净值市价比之间的关系则并不像先前一些学者认为的简单的线性相关；三种beta系数的波动范围随规模和净值市价比的变大而变小,但表现出一定的双峰性,在两个比较接近的数值周围波动,可能与我国股票市场数据还不够充足有关。最后,本文还对45个行业的beta系数性质做了研究。实证结果认为,市场因子的显著性很大,其效应的存在性毋庸置疑,其中新兴行业的市场因子beta系数较大,波动也较大；而规模因子和净值市价比因子在我国股票市场的确是存在的,但是受到选取的行业和时段的影响,并不是总能被观测到。
[Abstract]:Since the capital asset pricing model was born for half a century, from the non-conditional (static) model to the present conditional model, scholars from various countries have constantly revised and developed the model through various tests. In the last ten years, Chinese scholars have made many empirical studies on the stock market and fund market of our country. In general, their studies mainly focus on modifying variables and changing estimation methods, and their conclusions are different. There are differences in the choice of risk factors, while in the study of beta coefficient, the significance test of its long-term mean is the main factor. In this paper, a new two-step nonparametric regression analysis method proposed by foreign countries is used to study FF three-factor model in detail. The advantage of this method is that it can relax the restrictions on the model assumptions and use the iterative method to gradually approximate the estimated value which is closer to the real value. On this basis, this paper makes a deep study on the market premium of risk factors and the change trend and properties of the corresponding beta coefficient, not just the test of whether it is established or not. Firstly, the conditional beta value of each asset in each time period is estimated by nonparametric method from time series data. In this step the author refers to the four state variables commonly used by other scholars and combines them into six groups of tool variables and selects the two most influential variables according to the degree of fitting. Secondly, the time-varying risk factor market premium is estimated from the cross-section data of the return rate and the obtained beta coefficient, and the optimal window width is found by the penalty function least square method, and the beta coefficient is reestimated iteratively. A more accurate time-varying beta coefficient is obtained. Then, based on the above empirical results, this paper makes a detailed analysis of the market premium of risk factors and the nature of beta coefficient. The author thinks that the three-factor model is established in the stock market of our country, and the existence of the three factors is remarkable, but it does not have obvious effect in all periods. In addition, the three beta coefficients become smaller with the increase of the scale, but the relationship with the net worth market value ratio is not as simple linear correlation as some previous scholars thought; The fluctuation range of the three beta coefficients becomes smaller with the increase of the scale and the net value ratio, but it shows a certain bimodal property. The fluctuation around the two relatively close values may be related to the insufficient data of the stock market in our country. At last, this paper studies the properties of beta coefficient in 45 industries. The empirical results show that the significance of market factors is great, and the existence of market factors is beyond doubt. Among them, the beta coefficient of market factors in emerging industries is large and the volatility is large. The scale factor and the net worth market price factor do exist in China's stock market, but they are not always observed due to the influence of selected industries and time periods.
【学位授予单位】：浙江工商大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：F224;F832.51

【参考文献】