The Empirical Research On Mean-CVaR And Mean-ER Portfolio Model

Foreign study about making modern portfolio investment decision has more than half century, as the basis of modern portfolio theory. But in practice mean-variance portfolio theory is not very popular. Many researchers point that this model has many deficiencyies, such as investors do not think a positive deviation from the mean of return is a risk. So they put forward some revised model, such as the risk measure: Value-at-Risk (VaR). Statistical data show that return distribution are not normal. Under this condition, VaR is not a coherent measure. So researchers put forward many other measures to revise VaR, such as Conditional Value-at-Risk (CVaR) and Expected Return (ER).CVaR is a conditional expected value of the worst losses, while ER is unconditional expected value of the loss residuals. These models are worth of studying.This thesis primary purpose is comparison the result of CVaR Constraints in portfolio and ER constraints in portfolio and choose a better risk measure, under the same confidence level, in our market. Study the Mean-CVaR and Mean-ER empirically,"extract"VaR from these two models, construct an empirical efficient frontier for VaR that acts as useful approximation to the"true"efficient frontier.The conclusion is that ER is a better measure than CVaR, especially for a lager threshold which is fixed in ER model and the position of the portfolio in two models is different.