Chapter 1 explores how two different models of policing – reactive policing and proactive policing – understand the two most common forms of citizenry contact with the police – citizenry cooperation and citizenry deterrence. Under the reactive model of policing, responding officers help 911 callers solve personalized emergencies (Katzenbach, 1967). This differs from the proactive model of policing, where patrol officers conduct surveillance and enforcement operations independently of 911 callers (Kelling & Wilson, 1982). Both models interpret citizenry participation differently; however, none of them provide useful insight into the issue of 911 call misuse, which I define as misapplied instances of citizenry deterrence. In order for the reactive model of policing to function properly, citizens have to cooperate and report previously experienced negative outcomes (i.e., armed robberies, reported shootings, assault with a weapon, etc.). In the stylized version of this model, law enforcement does not know about the criminal unless the 911 caller submits a formal complaint. Once revealed, the police can solve crime through enforcement actions (i.e., arrests, citations, etc.). Citizenry deterrence, which involves residents calling 911 and requesting police presence in the anticipation of a negative outcome (i.e., the suspicious man in the park will rob me), can also fit into the reactive model of policing so long as the 911 calls lead to a “crime hit,” or instances where responding officers solved a crime and conducted an enforcement act (i.e., arrest, citation, etc.)^[1]. Once a crime has been solved by responding officers, citizenry participation ceases to exist in this model. This relationship between the police and citizens does not hold for the proactive model of policing. In order for this model to function properly, law enforcement aggregates information signals from previous 911 calls and uses this pooled information to develop proactive surveillance and enforcement strategies. Since law enforcement assumes a retrospective view, it can distinguish “crime hits” from “crime busts.” In using these “crime hits,” law enforcement can develop officer-initiated tactics that deter crime and wannabe criminals at their neighborhood origins (Kelling & Wilson, 1982). Neither the reactive nor the proactive models provide useful insights into this dynamic or the issue of 911 call misuse. Specifically, these models suggest that acts of citizenry deterrence and 911 call misuse will be uncorrelated with officer-initiated surveillance and enforcement strategies if they rarely result in police enforcement action, or “crime hits.”

^[1] In this paper, I ignore the issue of whether a responding officer solved the correct crime. So long as the responding officer conducted an enforcement action, then I assume law enforcement will interpret it as a successfully solved case.

To understand 911 call misuse and its downstream consequences on patrol policing, Chapter 2 introduces a coproduction model of policing that draws on previous works from political science and public administration. Brudney and England (1983) define ‘coproduction’ as “the critical mix of activities that service agents and citizens contribute to the provision of public services” (59). Several key features distinguish coproduction from the clientele participation model^[1] political scientists commonly invoke when studying local governmental service deliveries^[2]. First, coproduction “does not suggest a more responsive service bureaucracy as much as a more participative citizenry” (62). Second, coproduction “is the mix of regular producer and consumer producer inputs that contributed to the production (rather than the destruction) of services” (62). Third, coproduction “involves voluntary, cooperative action in service delivery” (62). Finally, coproduction concerns collective behaviors; “isolated instances of citizen service-directed activity will almost certainly have a negligible effect on the overall level of community services” (63). In the policing context, responding officers define “the degree to which the regular producer and [regular] consumer spheres overlap” (Brudney and England 1983, 61). Specifically, these institutional actors command the information flow between private residents and subsequent jurisdictional policing strategies. The key difference, however, involves the directional arrow between the current 911 caller and the responding officer. In this model, the current 911 caller affects the behavior of the responding officer, who then affects the behavior of the patrol officer. By weighting information signals from 911 callers and documenting their encounters, the responding officer helps to inform the patrol officer about new community safety threats, whether real or perceived. This cross-contamination best explains how even in the absence of coordinated group efforts on the part of 911 callers, demands for policing presence can spillover into jurisdictional policing strategies. Even if a responding officer knows they were summoned for inappropriate reasons, the coproduction model suggests that learning can still occur through two informational pathways. This includes the first-hand experiences responding officers have and the information they gather while traveling to 911 call locations and the second-hand accounts that parties affiliated with the 911 calls give to the responding officer.

^[1] “Citizen-initiated contacts with government agencies will be a function primarily of perceived needs, an individual’s instrumental concerns, and secondarily of the socioeconomic model, presumably the general political attitudes and information that also affect traditional forms of political participation” (Thomas, 1982, p. 518).

^[2] Where “public officials are charged with exclusive responsibility for designing and providing services to a citizenry that demands, consumes, and evaluates them” (Brudney and England 1983, 60).

In this dissertation project, I assume that cross-contamination between responding officers and their squadron members opens an information learning channel between 911 callers and officer-initiated surveillance and enforcement strategies. Unfortunately, most administrative police datasets do not provide a way to directly observe this cross-contamination and learning process^[1]. To get around this measurement problem, Chapter 3 introduces a strategy that approximates learning using call spillovers (or treatment spillovers) between the longitude/latitude coordinates of 911 calls. These call spillovers are intensity measures – calls per square kilometer, which I use as a treatment dosage for responding officer learning. As the intensity of a call increases, so too does learning, based on the two pathways, by responding officers. In addition to this core concept, this chapter provides a brief overview of point pattern data (PPD), introduces how researchers typically model PPD, describes how researchers can draw inferences using two or more PPD datasets, describes the inferential threats introduced by regionalized confounders – location-dependent confounders, and discusses how researchers can remove confounding using inverse probability of treatment weighting (IPTW). IPTW is the epidemiological analogue to instrumental variable (IV) estimation (Hogan & Lancaster, 2004), which economists and political scientists typically use. Unlike IV estimation, however, IPTW adopts a selection on observables approach towards causal inference, where the researcher selects a set of observed covariates to generate a propensity score model. Traditionally, model-misspecification and error-prone covariates bias the IPTW estimator (McCaffrey et al., 2013). In a geographic setting, however, researchers can use the longitude/latitude coordinates of spatial units and their residual errors to correct for model-misspecification and error-prone covariates.

^[1] Moving forward, I will attempt to gather this information through FOIA requests and strategic partnerships

What effect does demand for policing presence, as proxied by 911 call locations, have on the location of officer-initiated surveillance and enforcement strategies? Using publicly available data from Durham, North Carolina between 2006 and 2020, Chapter 4 tests how two neighborhood centric emergency crisis – suspicious complaints (citizenry deterrence) and weapon-related calls (citizenry cooperation) – change the intensity of directed patrols across the city, which I use as a proxy for police surveillance. Drawing upon the other chapters, I find that suspicious complaints are responsible for 10.7 percent of directed patrols across the city while weapon-related calls are responsible for 12.2 percent. This proportion is greater in majority-minority communities, however, who tend to be clustered in and around Durham’s gentrifying downtown neighborhoods. In addition, in these downtown neighborhoods, suspicious complaints explain a greater share of directed patrols than do weapon-related calls. This police responsiveness illustrates how 911 callers can indirectly affect levels of police surveillance when engaging in citizenry deterrence, such as suspicious complaints. Residents can have a bottom-up influence on officer-initiated policing strategies.