基于回归模型的指数追踪问题及实证分析

发布时间：2018-01-07 14:39

本文关键词：基于回归模型的指数追踪问题及实证分析　出处：《云南财经大学》2017年硕士论文　论文类型：学位论文

【摘要】：指数追踪是一种用少量的成分股来追踪某一市场指数走势的方法,它是消极投资组合管理策略中的一种,近年来在我国发展迅速。现在许多基金公司和投资者用指数追踪进行投资,因此,如何能够有效追踪市场指数,受到了越来越多的人关注,其中就有很多对它有兴趣的学者。学者希望从市场指数以及其成分股的历史数据中寻找到一种合适的方法,使得追踪效果更好,从而为基金公司和投资者提供更多的选择。本文建立了指数追踪的线性回归模型和分位数回归模型,然后通过三种选股法得到了三个样本股空间,最后在这三个样本股空间上对模型进行实证分析。由于线性回归模型有很多条件假设,且参数容易受到极端值的影响,为了避免这些,我们提出了指数追踪的分位数回归模型。具体来说,本文选取了2016年上半年的上证50指数为目标指数,构建两个指数追踪模型并给出约束条件,分别在三个样本股空间上选股,最后通过对真实数据的拟合来得到追踪效果较好的追踪组合,预测未来30天的收益,并与上证50的收益进行对比。本文得到的主要结论如下:(1)在三种选股方法中,最大相关系数选股法的整体追踪效果最好,且值一般是越小越好;(2)对比两个模型,在追踪效果方面,线性回归模型明显优于分位数模型,但是分位数回归模型可以获得更大的超额收益;(3)在不同的选股法下值的改变对模型的影响不同。
[Abstract]:Index tracking is a method of tracking a market index with a small number of component stocks. It is one of the negative portfolio management strategies. In recent years, there has been a rapid development in China. Now many fund companies and investors use index tracking to invest. Therefore, how to track the market index effectively has attracted more and more attention. Among them, there are many scholars who are interested in it. Scholars hope to find a suitable way from the market index and the historical data of its constituent shares to make the tracking effect better. In this paper, the linear regression model and quantile regression model of exponential tracking are established, and then three sample stock spaces are obtained by three stock selection methods. Finally, the empirical analysis of the model is carried out on the three sample stock spaces. Because the linear regression model has a lot of assumptions, and the parameters are easily affected by extreme values, in order to avoid these. In this paper, we propose a quantile regression model for exponential tracking. Specifically, we select the Shanghai Stock Exchange 50 index in the first half of 2016 as the target index, construct two index tracking models and give the constraint conditions. Stock selection in the three sample stock space, finally through the fitting of the real data to get a better tracking effect of tracking combination, forecast the next 30 days of income. The main conclusions of this paper are as follows: 1) among the three stock selection methods, the overall tracking effect of the maximum correlation coefficient stock selection method is the best, and the value is generally the smaller the better; 2) comparing the two models, the linear regression model is better than the quantile model in tracking effect, but the quantile regression model can get more excess income. 3) under different stock selection methods, the effect of value change on the model is different.
【学位授予单位】：云南财经大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：F832.51

【参考文献】