Introduction of a panel component in the LUCAS survey and the use of a composed estimator for statistics production : 2025 edition

Each of the LUCAS samples, from the first round (2009) to the most recent in 2022, was designed as a cross-sectional sample of the entire frame (the Master Sample). Each round featured a subset of points shared with the previous round. The maximum overlap rate occurred in 2015 (87 %), while the minimum was in 2018 (34 %). Intermediate rates were 63 % in 2012 and 40 % in 2022. The incidence of overlapping components was either unplanned or, if planned, not based on a rigorous methodology to determine the optimal rate for maximising estimate accuracy. This paper presents a study aimed at determining this optimal rate. Two types of estimates will be considered: (a) level estimates, which concern the totals or means of parameters of interest at a specific date (e.g., the percentage of artificial land cover in 2022); and (b) change estimates, which describe variations in the totals or means of parameters of interest between two dates (e.g., the increase or decrease in the percentage of artificial land cover from 2018 to 2022). Furthermore, the pros and cons of adopting a composed estimator that takes full advantage of the presence of a panel component will be investigated. The first issue (optimal panel component rate) is analysed using a simulation study. In contrast, the second issue (the convenience of adopting the composed estimator) is based on comparing the accuracy of the current calibration estimator with that of the composed estimator used in the LUCAS 2022 round estimates. In addition, the application of a specific function of the R package ReGenesees, namely the svyDelta function, which facilitates the production of change estimates with accurate calculation of the related sampling variance, is illustrated. Finally, general guidance is provided regarding the potential introduction of a panel component in the LUCAS survey.