The Best of Both Worlds: Forecasting US Equity Market Returns Using a Hybrid Machine Learning-Time Series Approach


Predicting long-term equity market returns is of great importance for investors to strategically allocate their assets. Harshdeep Ahluwalia, Head of Asset Allocation, Americas for Vanguard Investment Strategy Group is one of the authors invited to discuss the exploration of machine learning methods to forecast US stock returns 10-years ahead and compare the results to the traditional Shiller regression-based forecasts more commonly used in the asset-management industry.

Harshdeep discussed with Dr. Kathryn Wilkens, Curriculum Consultant for the FDP Institute, how the authors then implemented the VAR-based two-step approach of Davis et al. (2018) with machine learning techniques and allowed for unspecified nonlinear relationships (a hybrid ML-VAR approach). They found up to 56% improvement in real-time forecast accuracy for 10-year annualized US stock returns.