Even governments with detailed administrative data have difficulties detecting tax evasion because of firms’ concealing behavior. How can a researcher with access to more commonly available data —like manufacturing surveys— estimate tax evasion through cost overreporting?