Target alignment, dilution and forecast selection when cross-sectional forecasts share a common target

Battle Count: 0

Highly relevant for quantitative researchers dealing with cross-sectional asset ranking and forecast combination. It provides rigorous tools to diagnose why simple averaging often fails (dilution) and offers a principled selection rule to avoid admitting uninformative or negatively aligned forecasters.

The geometric decomposition is algebraically simple, but implementing the nested rolling-origin validation, HAC inference, and the three-way admission rule with simultaneous bands requires careful statistical programming. The use of LLMs adds infrastructure complexity.

The paper provides a comprehensive replication package including notebooks, code for panel construction, simulation designs, empirical panels, and configuration files. Raw market prices are excluded due to licensing but code to rebuild them is provided.