Today Won Do Lee, Matthias Qian and myself have published our first paper on the sociospatially differentiation in person mobility levels across England at the start of the COVID-19 pandemic in Health & Place. In the paper we use data from 1.1 m mobile phones and spatial statistics to analyse how income levels and a wide range of other variables were correlated with the reduction in mobility levels in 191 clinical commissioning group (CCG) areas in March and April 2020. The CCG area classification is used because most of England’s hospital services, including care for seriously ill patients during a pandemic, are planned on this basis.
Like other studies, we find that the extent of mobility reduction is significantly higher in areas with more high income households (belonging to the top quintile of the household income distribution at the national level). The relationship between income and mobility reduction remains after controlling for spatial autocorrelation but does vary across the country: they are most pronounced in and around the post-industrial cities of northern England.
On a more technical note, geographically weighted regression models offer substantially better goodness of fit indicators than global regression models (in which one coefficient is estimated to characterize the relationship of mobility reduction with income indicators and other independent variables), with or without correction for spatial autocorrelation. Spatial heterogeneity in correlations with mobility reduction must be accounted for if the effect of, say, income is to be characterized accurately.
We are currently conducting a series of follow up analyses, looking at temporal variations in mobilty reduction over the Spring of 2020, and the complex relationships between changes in mobility levels on the one hand and (local variations in) COVID-19 infection and mortality levels.