Objectives <p>Drawing on deterrence, situational crime prevention, and procedural justice theories, the study tests whether the association between pedestrian stops and shootings varies by stop initiation type and victim fatality.</p> Methods <p>Two-step System GMM applied to monthly New York City census-tract panel data from 2017 through 2023 (2,325 tracts × 84 months, balanced), the study models each stop type separately as endogenous while addressing simultaneity and spatial dependence, with missing data handled by multiple imputation.</p> Results <p>Witness-initiated stops exhibited the most consistent negative association with nonfatal shootings, significant in 2017, 2019, and 2020, weakening to non-significance in the peak-violence years 2021 and 2022, and re-emerging in 2023 as violence subsided. The dip-and-recovery pattern is more consistent with saturation of witness cooperation than with eroding police legitimacy alone. No stop type is significantly associated with fatal shootings, which were more prevalent in indoor settings.</p> Conclusions <p>The community-initiated source of a stop, not enforcement volume, is associated with fewer nonfatal shootings. High-volume, indiscriminate proactive stops warrant caution. The specific mechanism linking witness-initiated stops to fewer shootings remains untested.</p>