We characterize the bias in cross-sectional Hill estimates caused by common underlying factors and propose two simple-to-implement remedies. To test for the presence, direction and size of the bias, we use monthly US stock returns and annual US Census county population data.
