Application of Randomized Response Techniques Using Dichotomous Response for Mean Wage in Czechia and Slovakia

Ondřej Vozár, Luboš Marek
Statistika, 105(1): 5–17
https://doi.org/10.54694/stat.2024.42

Abstract
Research of controversial topics (drug consumption, corruption) requires reliable estimates of population means of sensitive variables (spending on drugs, illegal sources of income). To avoid non-response and fabricated responses surveys using randomized response techniques (RRT) for quantitative variables are conducted. The paper focuses on surveys conducted regularly in time to study evolution of population mean of a sensitive variable. This topic has not been explored for RRT yet. In applications of RRT is critical a choice of its parameters. The goal is to find rules of thumb if mean of a sensitive variable change in time. We focus only on methods using dichotomous variable (Antoch et al., 2022), because they were designed for variables with many values (Vozar and Marek, 2023). The different scenarios are applied on the Czech and Slovak wage data from Average Earnings Information System in years 2017–2019 using prior information. The scenarios evaluated in extensive simulation study focusing both on mean wage and year-on-year growth.

Keywords
Randomized response techniques, dichotomous response, wage distribution, survey sampling, comparability over time, population mean