We introduce a machine studying strategy to detect value-relevant international data by modeling stock-specific, time-varying relationships between international indicators and inventory returns. An extended-short portfolio exploiting international indicators generates 12% annual irregular returns. Return predictability is extra pronounced amongst home companies, these with low international 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 priceless indicators by monitoring key worldwide buying and selling companions, monitoring shifts in financial coverage and political stability, and leveraging data significantly from under-covered rising markets.
