Some (generally ignored) problems with Swedish register-based income data