An Enhanced Mean-Variance Framework for Robo-Advising Applications

Any robo-advisor needs to decide on a framework to model the preferences of its investors over uncertain outcomes. As of today, most robo-advisors model their investors as mean-variance optimizers. While the mean-variance framework is intuitive and optimal investment strategies have been derived in various settings, it suffers from serious drawbacks due to its time-inconsistency and non-monotonicity. We propose an enhanced mean-variance framework for robo-advising applications which is based on the equivalence between the mean-variance objective and quadratic utility functions. By introducing a flexible weight on the decreasing part of the quadratic utility function, we can alleviate the issues of time-inconsistency and non-monotonicity while keeping the features leading to the popularity of the mean-variance framework. We show how the new framework can be calibrated by means of questionnaires and discuss the advantages of the novel framework in terms of the resulting terminal wealth distributions