Drawing a Line: Comparing the Estimation of Top Incomes between Tax Data and Household Survey Data

Abstract

The paper uses the flexibility of household survey data to align their income categories and recipient units with the income categories and units found in data produced by tax authorities. Our analyses, based on a standardized definition of fiscal income, allow us to locate, for top-income groups, the sources of discrepancy. We find, using the cases of the United States, Germany, and France, that the results from survey-based and tax data correspond extremely well (in terms of total income, mean income, composition of income, and income shares) above the 90 th percentile and up to the top 1 percent of the distribution. Information about income composition, available in the US, allows us to investigate the determinants of this gap in the US. About three-fourths of the tax/survey gap is due to differences in nonlabor incomes, especially self-employment (business) income. The gap itself may be due to tax-induced re-classification of income from corporate to personal or/and to lower ability of surveys to capture top 1 percent incomes.

Property Tax Regressivity, the Case of Québec

“…As shown by Yonzan et al (2020), survey and tax data match very well up to the 99th percentile of the income distribution but discrepancies usually appear for the top percentile. Therefore, the main article considers the top decile as a whole, and statistics about the broken-down top decile are only presented through additional comments.…”

Is Income Inequality in Bulgaria Underestimated in Survey Data?

“…In addition, the probability of a random cluster sample of 3 060 households covering the above 1% or 0.1% should be assessed. This problem is more pronounced with a strongly distorted distribution (Yonzan et al, 2022). In particular, the Household Budget Survey in Bulgaria has been calibrated to ensure representativeness in terms of indicators such as the number of persons in the household and regional disparities, rather than in terms of income inequalities.…”

“…For this purpose, the national surveys in the USA, Germany and France included in the Luxembourg Income Study (LIS) are compared with data from the tax authorities. For the United States, in particular, 3/4 of the differences between the two types of data were found to be due to non-earnings and, in particular, selfemployment, income and incapacity of the surveys to capture the income of the richest 1% (Yonzan et al 2022).…”

“…Based on empirical analysis, they conclude that rich countries benefit from capital while poor countries benefit from labour, with income inequality higher in Latin America and significantly lower in Nordic countries. Yonzan et al (2022) summarise a number of studies in rich countries on the inability of household surveys to capture all incomes. They also examine the discrepancy between the highest incomes, where gaps in household surveys are most pronounced.…”

Missing Top Income Recipients

“…The comparisons by Yonzan et al (2020) of survey data with tax records for the U.S. suggest that the two sources only start to diverge appreciably at the very top-around the top 1%. It is the top end where tax records have a role.…”