Estimating Believed Knowledge of Portfolio Agents Using Inverse Optimization
Bachelor thesis in electrical engineering at the School of Electrical Engineering and Computer Science at KTH Royal Institute of Technology. Co-authored with Oscar Wink.
Abstract
In this report, we demonstrate the utility of inverse optimization in convex programming by applying it on estimating financial market beliefs and behaviors of portfolio investors. The inversion of the optimization utilized the Karush-Kuhn-Tucker optimality conditions specified for the current situation. The investor situation was simulated using the Markowitz model for optimal portfolio selection. Three model-specific implementations of inverse optimization were evaluated and the estimates were assessed by applying them in a solution to a portfolio agent sorting problem. The solution was perturbed with noise to test the robustness of the model.
The work concludes that estimation by inverse optimization of Markowitz models is possible to a satisfactory degree but requires case-specific model design.