Patent Innovation Factor in Stocks
Log in to collectAcademic paper
Strategy in a nutshell
This annual strategy identifies firms with high exposure to emerging technologies using U.S. patent data and a deep learning graph neural network (GNN) model. Patents are clustered based on textual and citation network information to classify inventions as new or existing technologies. Firms are sorted into portfolios based on exposure, forming a new-minus-old (NMO) factor, with size-adjusted, value-weighted portfolios rebalanced each year.
Economic rationale
Exposure to new technologies is associated with higher risk and uncertainty. Firms leading in emerging tech clusters earn higher returns, as shown by portfolio analysis. Leveraging patent text and network data via deep learning captures innovative firm characteristics predictive of stock performance.
Backtest performance
Annualised return5.54%
Volatility11.83%
Sharpe ratio0.47
Win rate71%