Turning on the Light: A New Assessment of Measurement Error in International Tax Data

A central issue in international taxation is the extent to which multinational corporations shift profits to low-tax jurisdictions. Policy analysis is obscured by difficulties in quantifying foreign earnings using existing datasets, which can be affected by measurement error. This paper sheds new light on this issue by examining a key source of measurement error in administrative international tax data: aggregation error that leads to double counting of foreign income and distortions in foreign tax rate calculations. We link data from tax filings and public disclosures to construct a firm-level proxy that uses “book-tax” differences to quantify the extent to which commonly-used aggregation techniques may result in double counting. A comparison of book and tax data allows us to proxy for levels of aggregation error in tax data and reveals an increasing trend over time consistent with larger measurement error. We show that applying a simple correction significantly harmonizes measures of foreign income and tax rates across firms’ book and tax filings and resolves a systematic relationship between book-tax differences and the size of MNCs’ foreign affiliate networks. Finally, we reexamine estimates from prior literature after correcting for aggregation error and find that their conclusions are generally robust to this correction.