We use high-resolution spatial data to build a novel global annual gridded GDP dataset at 1°, 0.5°, and 0.25° resolutions from 2012 onward. Our random forest model trained on local and national GDP ...