A0786
Title: Cardinality constraints meet large-scale portfolio
Authors: Yuan Chen - University of Vienna (VGSF) (Austria) [presenting]
Nikolaus Hautsch - University of Vienna (Austria)
Bo Peng - University of Vienna (Austria)
Immanuel Bomze - Universitaet Wien (Austria)
Abstract: In financial econometrics, the focus is often on improving covariance matrix estimations rather than addressing optimization problems with constraints for better portfolio management. It is argued that combining these advanced estimation methods with optimization that includes specific limits, like cardinality constraints, enhances decision-making and investment strategies. Cardinality constraints limit the number of assets in a portfolio, potentially making simpler estimators like the sample covariance sufficient for investment decisions, especially when dealing with large dimensions that typically introduce significant estimation errors affecting portfolio performance. The issue of managing portfolios is also addressed when there are fewer data points than assets, leading to non-invertible, noisy covariance matrices. Cardinality constraints simplify this challenge, making it possible to aim for a global minimum variance portfolio despite these limitations. Empirically, it is found that smaller portfolios, constrained by cardinality to include only a subset of available assets, can achieve diversification similar to market portfolios while reducing transaction costs and simplifying analysis. This suggests focusing on smaller, strategically selected portfolios could offer investors efficient and cost-effective outcomes.
