Bias in the Map, Bias in the Results? Validating OpenStreetMap Data for Individual-Level Analysis

The increasing availability of georeferenced data has enabled political scientists to examine how local environments shape political attitudes and behavior, contributing to a broader spatial turn in the discipline. A growing body of research links features of the built environment to individual political behavioral and attitudinal outcomes. Central to this research is the choice of geospatial data source, with an increasing reliance on crowdsourced geodata, such as OpenStreetMap (OSM). Despite its widespread use, concerns about data quality persist, and the validity of OSM-based measures of the built environment remains insufficiently tested. Focusing on the relationship between objective infrastructural deprivation, as captured by local public service provision, and citizens’ perceptions of infrastructural deprivation, this paper presents a validation study of OSM-based measures and examines whether errors in OSM data translate into biased substantive conclusions. Using Germany as a case study, we compare OSM point-of-interest (POI) data with official POI data from the German Federal Agency for Cartography and Geodesy. Specifically, we construct municipality-level indicators of the availability of hospitals, schools, and supermarkets in individuals’ residential environments to proxy objective infrastructural deprivation. We first assess spatial coverage differences between the two data sources and then evaluate how these discrepancies affect regression estimates in our substantive application. The analysis integrates georeferenced survey data from the 2021 German Longitudinal Election Study with the above contextual indicators. Our findings show that OSM data closely approximates official data in terms of overall spatial coverage, with only minor differences across service categories. Importantly, these discrepancies have minimal impact on substantive results: regression coefficients derived from OSM and official POI data are highly similar across multiple model specifications. This suggests that, despite known imperfections, OSM provides sufficiently reliable measures of infrastructural deprivation for individual-level political outcomes. These findings contribute to research on measurement error and data quality in political methodology by demonstrating that inaccuracies in crowdsourced geodata do not necessarily propagate into biased estimates. More broadly, the study provides empirical support for the use of OSM in survey-based political science research while underscoring the importance of context-specific validation.