This study examined how configurations of local- and state-level policies link to fatal shooting rates across 100 US cities. Our use of CNA methods illustrates the interplay of local firearm policies with state-level regulations and demographic considerations to complement findings generated through standard quantitative approaches that identify individual risk factors for firearm-violence-related outcomes. Rather than examine the association between a single independent variable and a dependent variable, our CNA analysis identified difference-making bundles of conditions that were necessary and sufficient for the outcome of interest (low rates of fatal shootings). Starting with a large initial set of nearly 90 potential explanatory factors (50 state-level firearm policies, 34 city-level violence prevention efforts/policies, population, and region), the CNA models pinpointed five pathways that consistently and uniquely distinguished cities with comparably low fatal shooting rates.

Prior research has supported a reduction in violence associated with restrictions against those prohibited from owning a firearm [31]. Pathway A (having a population between 600,000 and 1 million AND having an emergency restraining order prohibitor in place (state) and Pathway B (having the authority to deny firearm purchase for public safety (state) AND having local efforts to improve EMS quality and response times in impacted communities (city) both highlight how state-level policies—firearm restrictions for those with a restraining order and denial of firearm purchases—can combine with population and city-level resource considerations (such as improved EMS services) to influence local firearm violence rates. Pathway A notably suggests that protective policies could operate most effectively in large cities. Pathway B findings may be interpreted such that the availability of firearm purchase denials and improved emergency services contribute to better overall public safety and more robust city-wide prevention efforts, thus shaping local contexts into places where shootings are less likely to occur.

Each pathway identified in our results is a conjunction of discrete conditions. Notably, our results do not point to a single, clear-cut pathway toward lower rates of fatal shootings. The interplay of each conjunct highlights the challenge of policy effectiveness in complex environments. Identifying and isolating the effects from one policy can be difficult, and desired outcomes are likely influenced by a combination of many different policies, programs, and related contextual factors, such as demographics and regional cultures as well as implementation effectiveness. Relatedly, the history of any given city’s policy efforts also matters. For instance, a city with a long history of local prevention efforts may be affected differently by new state regulations than a city with limited local prevention efforts (and vice versa). Yet it is notable that certain policies spanned multiple pathways in the model, including policies to prevent potentially dangerous people from accessing firearms (state) and collecting and reporting hate crime data (city).

The finding that the EMS response measure is a difference maker as part of Pathway B is particularly interesting, because it entails efforts that include the use of supplemental county EMS services to support city EMS services, utilization of city firefighters as supplemental EMS responders, incorporation of strong and clear quality standards and audits into EMS provider contracts, and the use of data and technology to improve logistics [21]. There may be variation in how services are improved across cities, but better supported EMS in general may translate into higher quality care. Ultimately, regulations are implemented and enforced in hyperlocal contexts, so uncovering an ecosystem of inter-related policies can help identify the most comprehensive and effective firearm-policy bundles available for adequate prevention [32]. The regional factors’ presence in the model also highlights how geography in a large country like the US influences conditions for firearm-policy uptake and effectiveness [33].

This study is not without limitations. First, the data included in the analysis are cross-sectional, representing a single year of information. The firearm policies included in the analysis represent provisions in force in 2023, rather than newly enacted laws from that year. As such, the policy variables capture the broader policy environment within which cities operate. Although many of the state-level firearm policies included in the analysis represent longstanding legal provisions, we do not observe policy enactment dates or duration of exposure, and the timing of local prevention efforts likely varies across cities. As a result, the identified configurations should be interpreted as links between policy environments and firearm violence rather than definitive causal effects. Some local interventions such as EMS improvements may be implemented in response to elevated violence, raising the possibility of reverse causation. Longitudinal and quasi-experimental research will be important for assessing how changes in policy environments over time influence firearm violence outcomes.

Second, while the model had high consistency and coverage scores, there were nonetheless eight cities with the outcome present that remained unaccounted for, indicating a role for additional factors beyond those included in our dataset. Firearm ownership, socioeconomic conditions, law enforcement strategies, demographic characteristics, and recent events such as the COVID-19 pandemic all likely contribute to shaping the local context of fatal shootings but were not included in the present analysis. While incorporating such environmental factors could provide additional insight into firearm violence patterns, we were cautious in bracketing our CNA analysis to focus on policy-relevant combinations of state-level policies and city-level prevention measures. We encourage future CNA researchers to consider these local contextual factors in future analyses, as well as assessing configurations of law enforcement strategies such as focused deterrence, specialized anti-gun violence task forces, citywide CCTV systems, acoustic gunshot detection technologies, and variations in departmental budgets or operational priorities alongside local prevention efforts and state firearm policies.

Third, factors not included in our model should not automatically be considered unimportant or irrelevant for the main outcome. Rather, the results indicate these factors do not meet the definition of difference makers within CNA (the minimum set of necessary and sufficient conditions for the outcome to appear) but may still play other important roles in violence reduction. Future research that includes information prior to the COVID-19 pandemic as well as more recent data after 2023 will be important for assessing whether the configurations identified here persist as policy environments and social conditions evolve.

Finally, our data are limited to measuring the presence of policies and we did not have information about the fidelity or intensity of local prevention efforts or state policy implementation. Variation in implementation efficacy likely plays a significant role in predicting policy success and helps account for some of the patterns in our data. To the extent possible, we encourage researchers to consider implementation fidelity in future CNA work. Relatedly, our analysis is limited to the use of the VPI scorecard measures, created using publicly available data and feedback from individual city representatives. As noted in the 2023 Community Justice report [21], it remains possible that scores may not fully reflect violence prevention efforts if public information was outdated and city respondents did not respond to requests for feedback.