Extended Data Fig. 1: Why the COVID-19 ‘shock’ offers new insight into questions of environmental justice.

For simplicity, imagine 5 communities across (for example) a state, represented by the five colors here. These locations each have a different racial/ethnic composition, represented here for simplicity in one dimension, as the share of the population that is Hispanic. Many observations of environmental injustice rely on cross-sectional analyses, either A without or B accounting for potential slower-moving confounds. C However, many high-frequency variables contribute to ambient pollution levels and might be correlated with geography and socioeconomic variables; panel analysis with repeat observations over time allows for inclusion of these types of covariates, and can thus account for the contributions that (for example) natural weather patterns make to exposure disparities. However, even a panel analysis is subject to potential confounding, and interpretation of residual exposure disparities as environmental injustice caused by the economy remains problematic. The COVID-19 economic shock creates a large perturbation that ‘turns off’ a portion of the economy, and thus reveals the footprint of pollution caused by that in-person economic activity. We test for whether this shock changes exposure gradients (that is, whether the shock looks like D or E, and as such whether the in-person economy is contributing to environmental racism. F The ability to account for mobility in this framework further allows the separation of very local activity from broader activity (see Supplementary Information Text). (For clarity, G shows the heterogeneous shock in time series).