Mapping high-income taxpayers in Berlin using kernel-smoothed proportions from aggregated georeferenced data

The rare access to exact official geocoordinates opens new methodological possibilities for analyzing highly sensitive tax data. We explore their visualization potential and systematically evaluate aggregation as an anonymization strategy, with particular attention to its methodological and analytical implications. For an analysis of high-income taxpayers in Berlin, Germany, the focus is on the presentation of regional shares. In addition to frequency maps, smoothed representations using kernel density estimation are analyzed in particular, and their cartographic characteristics are discussed. Due to the high sensitivity of individual-level data, such data are generally not published, which is why anonymization is required in official statistics. This applies in particular to the group of high-income taxpayers. Using exact data as a gold standard makes it possible to systematically analyze the distortions caused by aggregation, one of the most commonly used anonymization methods in official statistics. In order to correct these distortions, a measurement error model is employed that explicitly accounts for the aggregation process and produces smoothed kernel density estimates for interpretable cartographic representations. In addition, the measurement error model is linked with census information to demonstrate a realistic application scenario. Local and global error measures are intended to empirically substantiate the improvement achieved through the use of the measurement error model.