| Summary: | 碩士 === 國立臺北商業技術學院 === 商學研究所 === 99 === The traditional portfolio is based on Mean-Variance Model (MV Model) which is built by Markowitz (1952). Moreover, Rockafellar and Uryasev (2000) think that Conditional Value at Risk (CVaR) can be used to estimate portfolio’s value at risk due to lower risk which investors concern about.
The presence of outliers will lead to estimate errors while measuring the traditional portfolio returns and risks. Based on this impact, there are many literatures which provide MV model, named robust portfolios, to modify the uncertainty of the average or variance estimation and escape the distortion of sample mean and covariance matrix.
In this study, we use some models, such as the Minimum Covariance Determinant Estimator (MCD), Orthogonalized Gnanadesikan-Kettenring Estimator (OGK), Nearest-Neighbour Variance Estimator (NNVE), variance, and CVaR, to measure the portfolios’ risk, analyze these differences between efficient frontiers, and explain the differences between these performances by using backtest. We think that these steps will help investors make decision with more choices.
We find that whether the returns are normal-distributed, or there are outliers, robust portfolio will improve the traditional MV model for investing decision-making, in which NNVE model will get the highest performance, and also MCD model will be the same with the outliers. Besides, the performances of portfolios are not the same if the outliers appear at different risk level. Finally, we suggest that robust portfolio with long-tern stocks would be a one of the investing decisions in stable economy.
|
|---|
