The Inequality (or the Growth) We Measure: Data Gaps and the Distribution of Incomes

Abstract

There is a large gap between income estimates used in inequality studies and macroeconomic statistics. This makes it hard to assess how economic growth is distributed across the population, and to what extent mainstream distributional statistics are an accurate representation of income flows. We take stock of these discrepancies by confronting estimates of the income distribution from surveys, administrative records and aggregates from the system of national accounts, thoroughly documenting them over the past two decades for ten Latin American countries. We find that surveys only account for around half of the macroeconomic income in the region. Measurement gaps account for just over half of the overall gap on average, while the rest is due to conceptual differences across data sets. Measurement gaps have been growing fast for many countries, the bulk being due to non-covered capital income. We also compare the top tails in administrative data and surveys, finding diverging averages—especially for non-wage incomes—and different shapes. We discuss the degree to which inequality levels and trends could be affected. (Stone Center on Socio-Economic Inequality Working Paper)

More Unequal or Not as Rich? Revisiting the Latin American Exception

“…Thus, we reconcile all available income data to build inequality estimates that not only adjust for surveys' measurement issues, but also ensure overall macroeconomic consistency. As anticipated in Alvaredo et al (2022), the adjustments end up doubling the total income originally declared in most surveys. Hence, to ensure transparency, we show the impact that each step of our methodology has for the resulting distributions.…”

“…Figure B.5 reports the "scaling factors" we use for each component in each country, that is, we multiply each survey component by 1/scaling factor, taking comparable household incomes from the national accounts as benchmarks. As found in Alvaredo et al (2022), capital income in the survey is systematically undercovered in all countries by the largest margin, implying that we multiply the income component by a factor of 5-10 in most cases. Some income aggregates need to be deflated to arrive at the SNA benchmark, typically either imputed rents, mixed income or social benefits.…”

“…A noteworthy finding is that national income, which is measured by the SNA, is larger and has been growing faster than other income concepts traditionally used to study inequality. Whenever survey aggregates are compared to SNA aggregates, capital incomes appear to be remarkably less covered than labor incomes (Törmälehto, 2011;Bourguignon, 2015;Flores, 2021;Alvaredo et al, 2022). Such gaps make it hard to assess how macroeconomic growth is distributed among the population, and to what extent existing distributional statistics (based both on surveys and tax records) are an accurate representation of material living standards.…”

“…Some income aggregates need to be deflated to arrive at the SNA benchmark, typically either imputed rents, mixed income or social benefits. These gaps are a function of comparability issues (outlined in Table B.2) and complications with annualizing composite survey income variables from reference periods (see Alvaredo et al, 2022 for more details).…”

“…Such gaps present us with a distributional conundrum: if they are widest in capital incomes, as has been found historically and more recently, then this would entail significant repercussions for existing inequality indicators (Alvaredo et al, 2022). Moreover, if these gaps are subject to changes over time, as also appears to be the case, our assessment of inequality trends would be severely compromised.…”