Not logged in.
Quick Search - Contribution
Contribution Details
Type | Master's Thesis |
Scope | Discipline-based scholarship |
Title | The impact of weight constraints on minimum variance portfolios: An empirical out-of-sample investigation |
Organization Unit | |
Authors |
|
Supervisors |
|
Language |
|
Institution | University of Zurich |
Faculty | Faculty of Economics, Business Administration and Information Technology |
Number of Pages | 59 |
Date | 2015 |
Abstract Text | Theoretically optimal portfolios based on the mean-variance framework often show poor out-of-sample performances. This is mainly due to errors occurring in the estimation of the input parameters to the portfolio optimization, i.e. the means, variances and covariances of the asset returns. Since especially the mean estimator proofs to be particularly error prone, some of the most promising approaches in terms of out-of-sample performance improvement can be found in models including an enhanced estimation of the covariance matrix for minimum variance portfolios. This thesis focuses on the findings that weight constraints can improve the stability and performance of minimum variance portfolios. It presents and tests a model introduced by Behr et al. (2013) that combines shrinkage properties of weight constraints described by Jagannathan and Ma (2003) with the shrinkage theory of Ledoit and Wolf (2003, 2004), in order to find optimal weight constraints. The thesis investigates the impact of these optimized constraints on the out-of-sample performance of equity portfolios, and tests the model against alternative portfolio strategies. Furthermore the impact of heuristic and optimized weight constraints on the variance-covariance characteristics of asset returns is tested by assessing the implicit beliefs inherent in the constrained portfolio optimization problem. The results show that the tightness of constraints imposed by the model of Behr et al. (2013) is dependent on the number of assets included in the investment universe. While the constraints are loose for low numbers of assets, they impose a strong structure for larger portfolios, especially if the available amount of return observations is limited. The resulting portfolios yield lower out-of-sample variances and higher out-of-sample Sharpe ratios than the equally weighted portfolio and minimum variance portfolios including none or heuristic weight constraints on average, but do not consistently outperform each of the portfolio strategies for the individual data sets. |
Export | BibTeX |