A1190
Title: Long-term forecasting of stock returns: Avoid overly complex machine learning and prioritize benchmarking
Authors: Parastoo Mousavi - Bayes business school (United Kingdom) [presenting]
Jens Perch Nielsen - City, University of London (United Kingdom)
Tatiana Franus - Bayes Business School, City, University of London (United Kingdom)
Abstract: The aim is to investigate long-term stock return forecasting, emphasizing the importance of systematic benchmarking of both dependent and independent variables. It is shown that adjusting independent variables for relevant benchmarks substantially improves predictive performance, a largely unexplored approach in the literature. Using a range of machine learning methods, it is further demonstrated that simple, carefully designed models provide more reliable forecasts than automated approaches in data-constrained environments.
