Using tax records to correct for under-representation of top income sources in surveys

Top incomes tend to be under-represented in survey data, making survey income distributions unrepresentative of the true ones in terms of the amount of total income and, if some of its components (sources) are missing relatively more than others, in terms of the structure of total income by components. Since the correct structure is important for some analyses, correction is warranted. Using Croatian survey and tax records data, we show that the survey correction method of Blanchet, Flores and Morgan (2022), based on reweighting and replacement of top incomes using tax records data – which corrects total income – can fail to correct its structure by components. This happens when some income components – in our case, primarily capital income – are missing much more than others, making the structure substantially different from the correct one and difficult to correct through reweighting and replacement of total income. As a solution, we propose a pre-correction and a post-correction, each involving income replacement based on tax records data, as supplements to the Blanchet et al. method. Either supplement substantially improves the correction of both total income and its structure by source.