Accounting vs technical information: what matters more for stock return predictability?
用机器学习模型比较技术数据和会计信息对股票收益预测的效果，发现技术数据预测更准、组合收益更高，尤其在短期和波动市场，但换手率更高；会计信息在长期和低估值不确定性时表现更好。
We employ machine learning models to determine what matters more for stock return predictability: technical data or accounting information. Technical data holds an advantage—it consistently yields more accurate forecasts and higher portfolio returns. This superiority is not limited to the U.S. market but extends to major developed markets worldwide, at times showing even stronger effects. Furthermore, it remains remarkably robust across firm sizes and time periods. However, its edge is most pronounced at short horizons and comes at the cost of higher turnover. Accounting signals, while weaker overall, perform better over longer horizons and support lower-cost implementation. Finally, technical strategies excel in volatile, hard-to-value contexts, whereas accounting-based models fare better when valuation uncertainty is low.