Find out about the pioneering research that allows for significantly improved and simpler to manage investment portfolios than is possible with traditional portfolio optimization techniques.
Addressing The Problem Of High-Dimension, Small-Sample
A big challenge for investors and their advisers is optimizing an investment portfolio so as to generate the highest possible returns at the lowest risk. The high-dimensional, small-sample problem (HDSS) presents one such obstacle they must overcome. HDSS, in this case, is a long and wide dish with many items, but it lacks past records of all the food on its menu — thus applicable for unreliable optimization and hence most likely results to weak investments.
Chanaka Edirisinghe and Ph. D. D., Kay and Jackson Tai ’72 Senior Professor of Quantitative Finance at Rensselaer Polytechnic Institute and Jaehwan Jeong, Ph. This is where an innovative solution by Tailai Li, Ph.D., collaborator and associate professor at Radford University, comes into play. In a recent study published in The Journal of Portfolio Management, they have launched an innovative data-driven approach to augment portfolio selection under HDSS.
Their methodology remedies the issues of traditional mean-variance (MV) optimization, which often produces too much risk and portfolio silos. Edirisinghe and Jeong have made their portfolio optimization framework stronger and more efficient by using cardinality control to limit the number of assets, leverage constraints for borrowing and short-selling management, norm constraints to optimally handle asset positions.
We started by playing the Sparsity and Leverage Control card
The novelty of Edirisinghe and Jeong lies in combining sparsity and leverage controls within a data-driven paradigm. Sparsity or building on a small number of assets can help alleviate the HDSS problem as it reduces the dimensionality of the portfolio and makes more manageable.
However, what truly helps reduce risks is leverage control. The researchers effectively curtailed risk through limiting the level of borrowing or short-selling, something that could not be achieved by conventional optimization approaches leading to countless boom-bust situations.
Edirisinghe and Jeong used cross-validation to further improve the performance of their method, which would work well in new or out-of-sample data. This step helps in ensuring that the optimized portfolios are not only powerful, but also have a track record of delivering good performance consistently while working with real-life scenarios.
The conclusion is still in tact as the researchers used a case study of multiple stocks from the S&P 500 Index to test their methodology. The results were pretty remarkable, with leverage-constrained sparse portfolio selection yielding substantial performance improvements on portfolios with better controllability and reduced risk.
Conclusion
Research by Chanaka Edirisinghe and Jaehwan Jeong is an act of pioneering a new horizon for portfolio optimization. Their solution is data driven and addresses the shortcomings of traditional methods that separately manage sparsity (the number of simultaneous active bets) or leverage control; both important parts overlooked in classical markowitz type modern portfolio theory dominant among financial firm risk team methodologies. This innovation not only delivers improved portfolio performance but also improves the scalability and risk profile of investment strategies as a whole. This body of work showcases the potential provided by data-centric solutions as the landscape of finance innovations, challenging contemporary portfolio optimization becomes enriched with analytical insights.
