Using Machine Learning to Predict Stock Earnings
Log in to collectAcademic paper
You Have a Point – But a Point Is Not Enough: The Case for Distributional Forecasts of Earnings
Ilia D. Dichev; Xinyi Huang; Donald Lee; Jianxin Zhao
- Emory University
- ?Emory University - Department of Accounting
- Chinese University of Hong Kong, Shenzhen
- ?Emory University - Goizueta Business School
- ?The Chinese University of Hong Kong, Shenzhen
- ?Emory University - Dept of Biostatistics & Bioinformatics
Strategy in a nutshell
The strategy develops distributional forecasts of firm earnings using both parametric and nonparametric (BoXHED) methods. By leveraging financial predictors, it estimates the probability of firms beating or missing analyst expectations and builds long–short trading portfolios.
Economic rationale
Earnings are central to firm valuation, yet point forecasts ignore uncertainty. Distributional forecasts capture the full range of possible outcomes, improving forecast calibration and enabling stronger trading strategies around earnings surprises.
Backtest performance
Annualised return51%