Reply to: Nix and Lozada (2020), “Do police killings of unarmed persons really have spillover effects? Reanalyzing Bor et al. (2018)”

Nix and Lozada (2020) provide a critique of our 2018 paper, “Police killings and their spillover effects on the mental health of black Americans: a population-based, quasi-experimental study.” They take issue to our use of crowdsourced data from the Mapping Police Violence project database on whether or not a given victim was unarmed, our main exposure measure, and argue that 93 cases were miscoded as unarmed in these data. They then argue that recoding or dropping these 30% cases led to an attenuated and statistically non-significant estimate of the effect of these events on mental health outcomes among black American adults. In this reply, we argue that (1) our use of the Mapping Police Violence project data was carefully considered and scientifically valid both from a theoretical and analytic perspective, and (2) that the difference between our estimates and Nix and Lozada’s do not arise from “miscoding” as the authors claim, but rather a different definition of what constitutes an unarmed victim. To probe the robustness of our results, we estimate 128 regressions models representing all combinations of exclusions of the seven classes of events Nix and Lozada dispute as being unarmed. The estimates are uniformly positive (i.e., all indicative of a harmful relationship between exposure to these events at the state-level and mental health outcomes), and the range of estimates overlap substantially with the 95% confidence interval for our original point estimate.