Abstract

Income-based indicators are of primary importance for the research community, for policy makers, as well as for other stakeholders in monitoring social inequality, poverty and social exclusion. Across Europe, data collection for the most widely used database, the Eurostat coordinated European Statistics on Income and Living Conditions – EU-SILC, is based either on household surveys, administrative registers, or on the combination of these. In Hungary, the source of the national data collection is a stratified sample with households as sampling units. The primary raw income data is cleaned and imputed. Recently, the imputation process is based not only on information drawn from the database itself, but also on external sources, like tax records.
While trends in the national aggregates of the main income poverty indicator (at-risk-of-poverty rate – AROP) display a fairly smooth trend, simple breakdowns and other estimates (like the at-risk-of-poverty gap) are surprisingly volatile within the period of 2018-2020 (survey years). Carrying out simple simulation exercises, we can conclude that AROP estimates are sensitive to small changes in the value of the poverty threshold, especially in the period between 2018-2020. Examining the distribution of income variables (total household income and equalized household income in the 2005-2020 period), we found excessive clustering around and above the poverty line. The severity of the situation is also reflected in the fact that Hungary was not included in the UNICEF Innocenti report card 18, which was largely based on this simulation exercise.
As the detailed imputation protocol and the related codes are not accessible by external researchers, one can only speculate about the reasons behind these developments. Going more in-depth and focusing on the concentration of equivalised yearly household income around the poverty threshold, we hypothesise that the above-mentioned shift in the imputation process might play a role, but the results of the investigation indicate that other factors should also be considered. An open and trustful cooperation between the data owner (The Hungarian Central Statistical Office in this case) and the researcher community could shed light on the roots of these problems and search for remedies.