Robo-advising: Optimal investment with mismeasured and unstable risk preferences

Abstract

We develop a robo-advising framework that incorporates interactions with a client who has time-varying risk preferences that can be mismeasured. In the presence of measurement error, a client’s welfare decreases compared to perfect measurements. The worse the quality of the measurements (higher measurement volatility), the worse this is for the client. When measurements are more volatile, interacting frequently can lead to further losses in a client’s welfare. We include a client’s willingness to interact using a budget constraint. A trade-off arises between interacting frequently to obtain up-to-date information with more volatile measurements versus less up-to-date information with more accuracy. We find that integrating client-initiated interactions and using an average of measured risk aversions both lead to more personalised investment advice. Using an alternative investment strategy rather than a time-consistent strategy can lead to improved welfare for the client.