基于交易费用的WCVaR投资组合模型研究

发布时间：2018-05-28 14:04

本文选题：最坏情况下的条件风险（WCVaR） + 混合分布　；参考：《广东财经大学》2014年硕士论文

【摘要】：风险度量模型CVaR(Conditional-value-at-risk,条件在险价值)有很多优点，它即弥补了VaR（Value-at-risk，在险价值）模型的不足，也满足一致性风险度量的要求，故一度受风险管理者的追捧。但在计算过程中，要求随机变量分布情况已知的前提下进行度量的，而现实中的金融市场常常受到各种复杂因素的影响，尤其我国目前证券市场发展不完善，金融市场波动较大，，随机变量分布信息无法完全知道，CVaR风险度量模型度量效率较低。随后，Zhu-FuKushima率先提出了最坏情境下的条件在险价值，简称WCVaR（Worst-case CVaR），它刻画了非完全信息下的风险，在现实中，我们无法预知某件事情的结果时，常常会考虑最坏情况发生时的情况，从而更好预知风险。本文考虑现实中资产收益率服从混合分布下的WCVaR模型，并在模型中加入比例交易费用函数，使得加入交易费用后的模型研究更贴近现实。然后利用向量自回归构建收益率未来路径，再根据上述回归后残差分布，判别残差可能服从哪几种概率分布情况，结合蒙特卡罗方法随机生成未来资产收益率情景。考虑损失函数为线性的情况下，从而将不确定的线性规划问题转化为确定的线性规划问题，利用Matlab中LP模块，即可求出模型最优解。模型结果证明，加入交易费用后，同等情况下风险相应有一定幅度增加，说明交易费用加入会相应增加风险，对现实中人们投资有一定指导性意义。
[Abstract]:The risk measurement model CVaRN Conditional-value-at-risk (conditional at risk value) has many advantages. It not only makes up for the deficiency of VaRN Value-at-riskmodel, but also meets the requirements of consistent risk measurement, so it was once sought after by risk managers. However, in the course of calculation, it is necessary to measure the distribution of random variables, but the real financial market is often affected by various complicated factors, especially the development of securities market is not perfect. Because of the volatility of the financial market, the risk measurement model of CVaR is less efficient because of the uncertainty of the random variable distribution information. Later, Zhu-Fuuseo first proposed that the worst-case condition is at risk value, or WCVaR(Worst-case Cvar Rao, which depicts the risk under incomplete information. In reality, when we cannot predict the outcome of something, we often take into account the worst-case scenario. Thus better anticipating the risks. In this paper, we consider the WCVaR model under the mixed distribution of the return rate of assets, and add the proportional transaction cost function to the model, which makes the research of the model closer to the reality after adding the transaction cost. Then the future path of return rate is constructed by using vector autoregressive method. Then according to the above regression residual distribution the probability distribution from which the residual may be obtained is determined and the future asset return scenario is randomly generated by using Monte Carlo method. When the loss function is linear, the uncertain linear programming problem is transformed into a definite linear programming problem. The optimal solution of the model can be obtained by using LP module in Matlab. The results of the model show that after the transaction cost is added, the risk will increase by a certain extent in the same situation, which means that the transaction cost will increase the risk accordingly, which is instructive to people's investment in reality.
【学位授予单位】：广东财经大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：F224;F832.51

【参考文献】