Capital Income; Inequality; Top Income Shares; Measurement Error
In this paper, the author studies how the contrasting coverage of labour and capital incomes affects inequality estimates. The author uses national accounts as a benchmark to evaluate the scope of household surveys, for a number of countries, and tax data for the United States. Due to both measurement error and conceptual differences, capital income is always more underestimated. In most countries, the gap grows during the last two decades.Based on accounting identities, the author shows that inequality estimates are likely affected in level, trend and composition. Surveys thus largely exaggerate the impact of changes in the labour income distribution, while they undermine the capital share and its dynamics. As a reference, in a panel of nineteen countries, households collect half of total capital income, as opposed to corporations; but surveys only capture close to twenty percent of that half, versus seventy percent of total labour income.For any quantile group - e.g. the top 10% or bottom 50% share - a unit increase of its labour income share translates into an increase of nine tenths of a unit in the overall share, for capital income, the effect is only one tenth of a unit. Gaps are narrower but still present in tax data.