Inferring mobility reductions from COVID-19 disease spread along the urban-rural gradient

The COVID-19 pandemic reshaped human mobility through interventions and voluntary behavioral changes. Those mobility reductions helped mitigate disease spread, but factors driving variation in mobility reduction remain unclear. We introduce a Bayesian hierarchical model to quantify heterogeneity in mobility responses across time and space in Germany's 400 districts using anonymized phone data. The model successfully reproduces timing and magnitude of major reductions in mobility across districts, revealing that disease spread affected mobility reductions most strongly [effect during first wave: -23%, IQR: (-27%, -20%), second wave: -18% (-21%,-15%)], followed by temperature (median difference of 11% between summer and winter), school vacations (-4%), and public holidays (-3%). We find significant differences in mobility response along the urban-rural gradient, with large cities reducing mobility most strongly. Investigating socioeconomic influences on reaction strength reveals different patterns across waves: during the first wave, mainly population density and employment variables are significant predictors (adj. R ² = 0.44), while population density and political variables mainly explain the variance during the second wave (adj. R ² = 0.37). Our results highlight that mobility serves as a valuable behavioral proxy with near real-time availability, making it an essential information source for future outbreaks.