Household survey data and tax data both suffer from measurement concerns at the top of the income distribution. This column analyses data from the US to investigate when and why the two data sources diverge. The authors conclude that the source of the divergence lies in the measurement of non-labour income as tax rules change over time.