Asset Pricing with Attention Guided Deep Learning

Philippe Chatigny, Chengyu Zhang, Ruslan Goyenko

Abstract: Deep learning methods, which can accommodate wide ranges of various stock characteristics to identify optimal investment portfolio or stochastic discount factor (SDF), have been criticised for extracting their superior performances from difficult to arbitrage stocks, high limits-to-arbitrage market conditions or extreme turnovers. We introduce \attention-guided deep learning, which, in a data driven way, allows identifying the most influential time-varying firm characteristics contributing to SDF. Attention dramatically improves SDF performance and attention to multiple firm characteristics reduces portfolio rebalancing costs. The attention guided SDF outperforms existing models after excluding small and micro-cap stocks, avoids extreme portfolio weights, and unlike other models, exhibits the best performance during market regimes with the highest price efficiency.