Geographic data conversion using grid-level data

Source abstract

Researchers often need to convert variables between non-hierarchical regions. This article presents a method for calculating non-binary, population-based conversion weights for Austrian municipalities and ZIP code areas using 250-meter grid-level population data. Due to the increasing availability of small-scale grid data, this approach can be applied to other countries as well. To evaluate the accuracy of the data conversion, the article compares the converted data with the observed data for population, employment, retirees, unemployment, and supermarkets, finding median absolute percentage errors ranging from 0.25% to 2.47%. This approach substantially outperforms naive binary assignments, especially when the source regions are split relatively evenly among the target regions. Year-specific weights are provided for the years 2013 to 2020, along with implementation examples for Stata and R, to facilitate adoption.