Figure A2:

Weekly Murder Rates by ZCTA and priod

The spatiotemporal visualizations above are corroborated by interrupted time series models at both the city-week and ZCTA-week levels. Similar to Table 2 in the main text, a significant increase in murders was observed, followed by modest linear decreases in the post-killing period across all specifications. However, the interaction term in Model 5, albeit similar in direction to that of Table 2 Model 5, is less strong, suggesting perhaps that the spatial heterogeneity in the post-killing effect was moderated by a factor other than racial demographic composition. However, we caution against overinterpreting this effect, as downtown ZCTAs (e.g., 55402) had a higher increase in murders, but these often involve non-residents in these ZCTAs. Such ZCTAs have a relatively advantaged residential population but experienced a larger discontinuity in murder rate, and could be suppressing the broader spatial pattern.