Investigating the need for raking on already adjusted sample according to the population characteristics

Case of Croatian senior population survey

Abstract

The research paper investigates the impact of raking on survey estimates in the context of a Croatian senior population survey conducted in January 2022. Raking is a statistical technique used to improve the accuracy and precision of survey estimates by adjusting the survey sample to match known characteristics of the population. The main objective of the paper is to determine whether applying raking to a survey is beneficial when the survey sample has already been adjusted to match certain population characteristics. In this case, the survey sample was initially adjusted using quota sampling based on the 2011 census data, considering gender and NUTS2 region as variables. An important factor in this study is that a new census was conducted in 2021, which provided updated population characteristics. The research aims to compare survey point estimates using both the 2011 and 2021 population data. The research conducted raking analyses using a variety of combinations of variables, including age, gender, county, and NUTS2 region. A total of 11 different combinations of these variables were used in the raking analyses. The study also examined the impact of the different number of iterations in the raking process, specifically observing results for 5, 10, and 20 iterations. The research included statistical tests to compare point estimates derived from the survey sample to those derived from raking using different sets of population data. In almost all cases, no statistically significant differences in point estimates were found. However, statistically significant differences did emerge when comparing sample point estimates to raking point estimates based on the 2021 census data, particularly when the number of estimated raking weights was the highest.