We introduce a machine studying method to detect value-relevant overseas info by modeling stock-specific, time-varying relationships between overseas indicators and inventory returns. An extended-short portfolio exploiting overseas indicators generates 12% annual irregular returns. Return predictability is extra pronounced amongst home companies, these with low overseas institutional possession, and during times of low media protection and excessive mannequin settlement. Notably, efficiency concentrates on the lengthy aspect, enabling cost-effective long-only implementations. Sign significance evaluation reveals our algorithms detect precious indicators by monitoring key worldwide buying and selling companions, monitoring shifts in financial coverage and political stability, and leveraging info significantly from under-covered rising markets.
