基于不同市场状态下个股预期收益的动态投资组合策略研究

发布时间：2018-02-11 22:20

本文关键词： 马尔科夫状态转换模型 CAPM 动态投资组合策略　出处：《上海师范大学》2017年硕士论文　论文类型：学位论文

【摘要】：在研究宏观经济和金融变量的动态行为时,线性模型往往无法体现变量非线性动态过程中的非对称性、振幅的相关性和波动的集聚效应等。由于股票市场受一个潜在的市场状态(熊牛市等)影响,且投资者在不同的市况中所面临风险是不同的。所以当我们研究资本市场收益率时,将传统的Markov状态转换模型与CAPM模型相结合可以解释一些问题,但仍无法解释异方差性和极值点集聚等现象。本文在研究资本市场收益率时,考虑到市场的收益率曲线具有均值恢复,异常波动,极值点集聚等特征。所以采用了Janczura(2012)提出的校正后的Markov状态转换模型(CMRS)来预测市场收益率,并与CAPM模型相结合构建动态投资组合策略。在本文在研究过程中,首先把市场划分为牛市、熊市和稳态三种状态,并通过模型估计出每种状态下市场的预期收益率和对应概率。然后从不同行业随机选取了十只股票,利用前面得到的市场收益率预测结果来计算个股在不同市场状态下的预期收益率,进而求得个股的期望收益率,最后用线性规划的方法,根据风险最小化原则构建动态投资组合策略。本文的研究数据可以分为两个阶段,2013/1/1~2015/12/31的数据用于模型训练,参数估计,2016/1/1~2016/9/30日的数据用于对最终构建的动态投资组合策略进行模拟验证。最后实证结果表明:CMRS模型和传统的MRS模型都可以对市场的状态做出准确预测,且预测结果基本一致;由于CMRS模型可以在市场处于异常状态(|收益率|1%)时放大误差波动,在市场处于平稳状态时(|收益率|1%)缩小误差波动,所以其估计结果与传统的MRS模型相比也存在细微的差别;另外,通过实证模拟我们也发现,基于CMRS模型和MRS模型的动态投资组合策略都是有效的,但是在准确度上,特别是当市场处于平稳状态时,CMRS模型的预测结果要高于后者。本文的结论对投资者进行股票市场行情预测和构建投资组合策略具有一定的借鉴参考价值。
[Abstract]:When studying the dynamic behavior of macroeconomic and financial variables, linear models often fail to reflect the asymmetry in the nonlinear dynamic process of variables. Because the stock market is affected by a potential market state (bear bull market, etc.), Therefore, when we study the return rate of capital market, we can explain some problems by combining the traditional Markov state transition model with the CAPM model. However, we still can not explain the heteroscedasticity and extreme point agglomeration. In this paper, when we study the return rate of capital market, we consider that the return curve of the market has average recovery and abnormal fluctuation. Therefore, the corrected Markov state transition model (CMRS) proposed by Janczura (2012) is used to predict the market rate of return and combine with the CAPM model to construct a dynamic portfolio strategy. First, the market is divided into three states: bull market, bear market and steady state, and the expected rate of return and the corresponding probability of each market are estimated by the model. Then, ten stocks are randomly selected from different industries. The expected rate of return of individual stock under different market conditions is calculated by using the forecast result of market return obtained in this paper, and then the expected rate of return of individual stock is obtained. Finally, the method of linear programming is used to calculate the expected rate of return of individual stock. According to the principle of risk minimization, the dynamic portfolio strategy can be constructed. The data in this paper can be divided into two stages: 2013 / 1 / 1 / 1 / 12015 / 12 / 31 for model training. The data from 1 / 1 / 1 / 6 / 9 / 30 of 2016 is used to simulate and verify the final constructed dynamic portfolio strategy. Finally, the empirical results show that both the MRS model and the conventional MRS model can accurately predict the state of the market, and the predicted results are basically the same. Because the CMRS model can amplify the error fluctuation when the market is in the abnormal state (yield 1), and reduce the error fluctuation when the market is stationary (the yield is 1), the estimation result is also slightly different from the traditional MRS model. In addition, we also find that the dynamic portfolio strategy based on CMRS model and MRS model is effective, but accurate. Especially when the market is in a stable state, the forecast result of CMRS model is higher than that of the latter. The conclusion of this paper has a certain reference value for investors to forecast the stock market price and build the strategy of investment portfolio.
【学位授予单位】：上海师范大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：F832.51;F224

【相似文献】