Mean-CVaR Portfolio Problem Based On MIDAS Copula

In this thesis,we study the mean-CVaR portfolio optimization problem based on the mixed data sampling Copula model.When previous researchers used the traditional Copula model to study the dependencies between financial assets,they did not consider the impact of mixing data on the dependency structure.In this thesis,we apply the mixed data sampling method to Copula modeling,extend the dependence structure from the same frequency mode to the mixed frequency mode,and apply the mixed data sampling Copula model to the portfolio problem.The work of the thesis is as follows:First,we establish a mixed data sampling time-varying Copula model to measure the dependence between variables containing different frequency data.We first combine the principle of Spearman correlations to construct the realized correlations;then,we decompose the time-varying tail dependence in the traditional time-varying Copula model into long-term dependent components and short-term dependent components,and add them to obtain the specific form of the TV-Copula-RD model;Finally,the complete steps of the parameter estimation of the TV-Copula-RD model are given.Second,under the framework of mean-CVaR analysis,we re-examine the problem of portfolio decision-making in conjunction with the TV-Copula-RD model.First,combined with the nature of CVaR,the original optimization problem is transformed into a linear programming problem,and a minimum CVaR model based on TV-Copula-RD is established.Finally,Monte Carlo simulation is used to solve the mean-CVaR portfolio problem and specific steps are given.Third,we select two style indexes,growth index and value index,as the research variables;then we compare the accuracy of four different Copula models in measuring the dependence and the portfolio risk.The main empirical results show that the Copula model,which considers mixed data sampling,time-varying dependence,and asymmetric dependence,is more suitable for the description of the tail dependence,and it is also more helpful for predicting the CVaR of the portfolio.