|Home | About | Journals | Submit | Contact Us | Français|
Data on police stops can be examined to reflect on the relative “opacity” of these encounters and how aggregate patterns on the nature—not just the volume—of reported stops relate to public scrutiny of the police. We hypothesize that public scrutiny on police stops is positively related to the prevalence of opaque stop practices across dimensions of “intrusiveness,” “rationale,” and “setting” derived from agency records. We further argue that this relationship is influenced by neighborhood conditions in the form of concentrated disadvantage, residential instability, and heterogeneity. To develop these ideas, we draw on a publicly available NYPD dataset on police stops to specify a series of fixed and random effects models that describe variation in recorded stop practices across precincts (N=74) and overtime (T=7, 2007–2013). We relate these practices to neighborhood conditions derived from the Census and examine their association with rates of SQF complaints to the CCRB. Results show considerable variation in indicators of opacity, particularly across precincts. More importantly, we also find that rates of complaints are higher in precincts that have more vaguely defined, intrusive stops. Results also suggest that concentrated disadvantage is independently and positively related with higher rates of public scrutiny of the police.
An essential component of policing is to interact with the public. These exchanges are shaped by discretion of officers and citizens as well as neighborhood conditions and formal and informal protocols within law enforcement agencies. Stop, question, and frisk (SQF) is a conventional police-initiated tactic that encompasses most of these influences. In the last decade, however, the practice of SQF in New York City and similar tactics elsewhere in the USA has been the subject of critical research and litigation claiming that it is racially biased, unjustly targets innocent citizens, and produces mixed results in terms of crime prevention1–3; but see 4.
Less is known, however, about the relationship between SQF and the patterning of citizen grievances filed against the police in connection to police stops. With some exceptions,see e.g.,5 it is unclear whether the nature or quality of stops—not just their volume—relates to heightened formal public scrutiny of the police and whether this association varies across neighborhoods and overtime. We explore these issues by examining the ecological relationship between recorded precinct-level SQF procedures in New York City and formal grievances on SQF practices filed with the agency in charge of investigating these claims, the Civilian Complaint Review Board (CCRB). Specifically, we examine variation in SQF complaint rates per 1000 stops for 74 precincts in New York City (2007–2013) and relate these to variation in key aggregate indicators of the specific nature of SQF practices. We argue that specific properties of stops associated with the recorded rationale, setting, or degree of intrusiveness are markers of their “opacity,” or potential for more strained encounters, police abuse, or misconduct.
Assessing precinct-level trends in recorded police stops and their link to formal resident complaints offers a unique opportunity to reexamine police legitimacy from a novel perspective that supplements research on stops and police attitudes, citizen perceptions of encounters, and crime reporting with a more ecological focus on aggregate recorded stop practices and grievances.2,6 Consistent with this framework, we define opaque SQF procedures as specific attributes of stops more likely to elicit feelings of procedural injustice or unfairness among citizens exposed to these encounters. Further, by studying precincts, we focus on an important intermediate level of analysis for policy and research, halfway between specific contacts and more general discussion about “social forces” shaping police-citizen encounters. This is of interest to the public health field as grievances reflect specific instances where the public not only acknowledges a specific form of police misconduct but also mobilizes the law for assistance. Further, it matters because negative interactions between the public and the police in connection to stops may have lasting psychological and physical health effects7 and may require enhanced mechanisms of mobilization and accountability.8
Citizen complaints of police behavior have been a long-standing topic for research on public health and community social control.9 Studies have focused primarily on describing the most frequent types of complaints, with more systematic attention to how complaints about alleged police excessive use of force relate to attributes of officers and police departments.10,11 By contrast, neighborhood conditions have been relatively neglected from complaint-based assessments of policing misconduct.12 Instead, researchers have typically used measures of local conditions to examine police practices using survey-based measures of police perceptions and contacts.13,14 This leads to the question of whether examining use of force complaints as isolated incidents connected to specific officers, rather than systemic issues, or disconnected from other practices such as stops or patrol can be a detriment to addressing the institution of policing and its relationship to the community.
A growing literature has begun to broaden this perspective to include a wider range of police tactics and more explicit links to community characteristics. Police stops, in particular, have been the focus of renewed scholarship, policy deliberation, and litigation. Attention to this tactic dates back to the late 1960s with the Terry vs. Ohio (392 US 1,15) decision by the US Supreme Court ruling that officers retained the right to stop, question, and search citizens based on the presence of “reasonable suspicion” that a crime was either in progress or imminent. Legal thresholds to escalate a stop to a frisk or a search were also established but less stringently enforced.16
Incident-level data on SQF in New York City since 2003 have been made publicly available following the settlement in Daniels et al. vs City of New York.1 Analyses of these data have consistently shown that recorded stops involve a disproportionate number of minorities—particularly young African-American men—relative to their share of the population and most do no result in the finding of weapons, contraband, arrests, or summons.2–5
Research has further indicated that this patterning of stops may not only be unlawful but also counterproductive for law enforcement in terms of lower perceptions of police legitimacy and reduced likelihood of cooperation.17 However, as indicated by Carr et al.,18 resident-based initiatives of reform of police practices are elusive, as citizens may want “tougher” law enforcement despite having been themselves subject of policing practices they see as negative or disrespectful.see also,19 These findings echoed other research finding a relative gap in neighborhood-based mobilization against police stops.8 These findings encapsulate the issues with SQF, and increases in the intensity of this tactic—the dosage of law—and subsequent racial and socioeconomic disparities can be readily identified, but unpacking issues surrounding the quality of police/citizen interaction and alternatives for reform can be much more difficult.
This study examines the relationship between precinct-level SQF practices and public scrutiny in the form of recorded grievances about police stops. More specifically, we draw on publicly available data on stops and complaints aggregated to the precinct level (N=74) between 2007 and 2013 (T=7 years) to specify three domains of the relative opacity of SQF practices: (a) intrusiveness, (b) setting, and (c) rationale, to answer the following three questions: (1) How do indicators of opacity in police stops vary overtime and across precincts? (2) Are opacity measures associated to citizen complaints on stop practices? (3) Do neighborhood conditions account for the relationship between the measures of opacity and the complaint rate?
The three specified dimensions of opaque police stops capture key recorded components associated with all stops—not just those signaling more extreme encounters, like excessive force. More importantly, we seek to mitigate issues of underreporting and bias in stop recording practices by drawing on multiple items that signal different configurations of opportunities for police misconduct and citizen perceptions of unfair treatment in the context of stops.20
The data used to generate the study’s dependent variable was gathered from a complaint-level database of SQF-related grievance provided by the CCRB. We aggregated these complaints for years 2007–2013 to the precinct level and computed rates per 1000 recoded stops. Precinct-level SQF complaint rates for years 2007–2013 per 1000 population were then collated with matching aggregate data drawn from the publicly available NYPD SQF database.
Consistent with prior research, we specified a series of precinct-level indicator of socioeconomic conditions such as concentrated disadvantage, neighborhood instability, and immigration.see e.g.,3 These measures were derived from factor analyses (principal components) of key census 2010 variables including population demographics, employment, and foreign-born population. We also draw on CCRB published reports to estimate the number of full-time police officers per precinct.
To examine stop procedures signaling opacity, we aggregated discrete items in the UF-250 forms per year/precinct. Specifically, we first focused on potentially invasive procedures, meaning an interaction that escalates from a stop to a subsequent frisk, search, and arrest or summons.1,7 This measure was operationalized using three separate measures: the percentage of stops in a given precinct/year that ended in an arrest or summons, and the percentage of stops that resulted in a frisk or search. Precincts with more invasive stops may be perceived as citizens as more procedurally unsound, as they signal expanded police discretion and potential for more strained citizen encounters. We hypothesize that in indicators of opaque stops in terms of intrusiveness in police stops will be positively related to SQF complaints and neighborhood conditions (H1).
Procedures categorizing the stop setting21 are important because environments marked by highly mobile populations (citizens moving through transit stops or housing authority residents or visitors) may be more likely to ignore or be unaware of disruptive behavior. These stops could also be conceived as an intrusion on their daily routine. Furthermore, officers not in uniform indicate a lack of officiousness and could be interpreted by other citizens as a sign of physical disorder and not a police/citizen interaction. Setting is operationalized using four precinct-level measures: percentage of stops that happen in a housing authority complex, public space, and transit station or stop, and percent of stops conducted by officers in uniform. It is hypothesized that precincts with higher levels across indicators of opaque settings will have higher rates of SQF complaints (H2).
We developed a final dimension of opacity indicator tapping into the recorded rationale for the stop.6,22 NYPD officers are required to report one or multiple reason leading to each stop from a predefined list in the UF-250 which includes items like “furtive movements” and “actions indicative of a drug transaction.” As noted by prior research, some of these items are overly vague and more closely linked to unproductive stops.see e.g.,23 We chose to measure the percentage of stops where “fits a relative description” or “suspicious bulge” because these are some of the most frequently recorded reasons for stops and they may be perceived by the community as an unclear pretext for a stop. We anticipate that precincts with higher rates of stops linked to the reasons mentioned above will have higher rates of complaints (H3).
Our dimensions of opacity do not rely on indicators of use of force in the UF-250 forms as the CCRB has a separate complaint procedure for these incidents. Furthermore, use of force is relatively rare in the context of police-recorded stops and more vulnerable to bias and underreporting.5
We specify a series of fixed and random effects models using STATA’s XT commands to assess how stop practices and community factors relate to precinct-level SQF complaint rates across 74 precincts and 7 years.2 To examine which model estimation was appropriate, a Hausman test was run finding no significant difference in core models estimated via fixed and random effect models (r=56.72, p<.001). As such, we decided to focus on the random effects models, which were not only more efficient but have the added benefit of preserving time-unvarying covariates (for ex. police officers per capita and population estimates).3 Random effects models are estimated following a stepwise approach, with models 1–3 examining how precinct-level SFQ complaint rates relate separately to the three described dimensions of opaque SQF practices: setting, rationale, and invasiveness. For each of these models, we control for rate of police stops per precinct, local population, and number of officers. Model 4 includes all dimensions of opacity at once. Model 5 adds all neighborhood conditions including concentrated disadvantage, instability, and immigration, as well as the percentage of black residents and percentage of divorced residents.
Table Table11 displays the means and standard deviations for each variable in the analysis. The average SQF complaint rate per 1000 stops is 20.38 and varies more between precincts (σ=8.50) than within precincts overtime (σ=7.97). Between 2007 and 2013, the SQF complain rate declined from 28.25 to 24.40, following a more pronounced decline in the population prevalence of stops after 2012.
With the exception of the variable measuring the percentage of recorded stops in public areas, indicators of opacity tended to vary more across precincts than within precincts overtime, particularly for measures of “setting.” This is not surprising as stops replicate the physical and social heterogeneity of precincts. In contrast, estimates of temporal and spatial variation for “rationale” and “intrusiveness” appear more closely together, confirming the importance of accounting for fixed and period effects in multivariate models. These models are presented in Table Table22.
Models 1–4 specify the relationship between the SQF complaint rate and the three dimensions of opacity considered independently and simultaneously. Model 1 (setting) shows that only the proportion of stops in public (“outside”) areas is significantly related to variation in the SQF complaint rate (b=−.073, p<.05). Model 2 (rationale) shows that both sub-dimensions of opacity are linked to higher rates of public scrutiny of SQF practices (b=.455, p<.001 for “fits description,” and b=.633, p<.001 for suspicious bulge). Similarly, Model 3 (invasiveness) indicates that two out of the three selected sub-dimensions of stops—those resulting in a subsequent frisk or search—are positively associated with variation in the SQF complain rate (b=.204, p<.001 and b=.314, p<.05, respectively). These results largely confirm hypotheses 1–3 on the positive relationship between dimensions of opacity and SQF complain rates. Across these three models, SQF complaints are also significantly higher in precincts with more police officers and those with a lower prevalence of police stops. As noted by the R-square statistics, Model 2 (rationale) is the more balanced model in terms of explained within/between variance confirming descriptive patterns in Table Table1.1. Model 4 considers all three dimensions of opacity at once. Estimates remain largely robust to this specification, although, as expected, estimates may be sensitive to the relative empirical and substantive overlap across measures and dimensions.4 Importantly, we note that estimates for indicators of rationale not only are remarkably stable but also exhibit the strongest association with variation in complaint rates.
Model 5 tests whether key neighborhood social conditions account for the observed associations between dimensions of opacity and the SQF complaint rate. Estimates of opacity are largely consistent with those presented above, with the most robust estimates linked to rationale. Contrary to research on collective efficacy,24 we find that concentrated disadvantage is associated with higher, not lower, rates of SQF complaints, controlling for other social correlates and indicators of police presence and tactics (b=2.702, p<.01) (see also percent Black). Importantly, when controlling for precinct-level socioeconomic conditions, most of the sub-dimensions of setting of the stops become negatively related to the SQF complaint rate whereas sub-dimensions of intrusiveness become less important. This suggests that neighborhood and police practices influence public scrutiny of the police independently of each other.
As noted in Table Table2,2, the explained variation of our models increases as we better-specified measures of opacity and neighborhood conditions. Model 5 accounts for roughly 53 % of the variance of the complaint rate within each precinct, 61 % of the variance between, and 57 % of the overall variance of the dependent variable. The intraclass correlation ultimately shrinks throughout the models, until roughly 41 % of the variance is explained by the standard errors of residuals across panel models. Lastly, the Wald test in each model is significant, meaning that the multiple parameters of the models are significantly different from zero and are therefore relevant in the estimation of the equations.
Our analyses show that opaque procedures associated with police stops explain a significant amount of the variation in formal SQF complaints across precincts and overtime. In particular, we note that precincts with higher average rates of stops linked to a vague rationale justifying these encounters—suspicious bulge or fits description—have higher complaint rates. The same is true of precincts recording more invasive stops (frisks), and stops carried out by non-uniformed officers, or in transit-related settings. Taken together, these results largely confirm our three working hypotheses on the positive relationship between indicators or opacity and SQF complaints.
Findings also point out to key contrasts across the three specified dimensions of opacity: Unlike setting, rationale is more closely linked to the common legal standard for stops. It is also the construct that more clearly reflects the discretion of police officers while on patrol. As such, procedures attached to justifications for stops can be managed, altered, or revamped by training, oversight, and new performance indicators. This may also apply to measures of “invasiveness.” We find that precincts that have more encounters that progress from a stop, to a frisk, and then a search, have higher rates of formal scrutiny of stops. These protocols are likely seen by citizens as progressively invasive and more likely to result in strained encounters. Further, in our more fully-specified models, we find that average rates of stops resulting in arrests/summons do not explain variation in complaint rates, confirming the importance of procedural justice concepts and constructs when assessing markers of police legitimacy and public accountability of the police. The opaque category of stop setting has a more complex relationship with complaints. In the final model, precincts that have more stops that happen in a transit hub have a predicted increase in the rate of SQF complaints filed in that precinct. Perhaps the nature of transit stops simply results in more complaints because people trying to get from one location to another are subjected to an increased inconvenience. However, precincts with more officers in uniform and more stops in housing authority complexes see a predicted decrease in their complaint rate. Officers in uniform could be perceived as having more legitimacy, and therefore, their collective discretionary use of SQF could result in fewer complaints, whereas stops in housing authority complexes may produce fewer complaints because there are fewer witnesses to corroborate the incident. Or, witnesses may be unwilling or unable to corroborate procedurally unfair incidents.
Lastly, we find that neighborhood conditions matter for the specification of SQF complaints, particularly concentrated disadvantaged, with more economically deprived, minority areas associated with higher complaint rates, independent of indicators of opacity of the stops and other predictors. This suggests that perhaps in these communities, police accountability and oversight may take place through individual complaints, rather than more collective processes of mobilization.
This research furthers the scholarship on the theory of procedural justice and the policy of police procedure in three important ways. First, it provides fresh empirical evidence on how specific attributes of recorded stops seldom studied in connection to intrusiveness, rationale, and setting unfold overtime and across precincts, and in connection to formal grievances against the police. Addressing our first research question, we note that spatial variation is more important than temporal variation not only in terms of overall volume of stops but also in terms of perceived “opaque” practices. For example, Model 5 accounts for 60.7 % of the variance between precincts and 52.7 % of the variance across time. In connection to our second research question, we find that precincts with a higher fraction of recorded stops with vague justifications (the rationale dimension), as well as those with specific markers of intrusiveness and setting, exhibit relatively higher SQF complaint rates. In relation to our third research question, we find that neighborhood conditions, particularly neighborhood disadvantage, account for some of the variation in SQF complaints. Second, this study sheds light on the notion of legitimacy as a normative community concept and not solely a consequence of instrumental outcomes, such as a reduction in recorded crime or increases in sense of safety. Instead, it is based upon individual perceptions of procedural fairness: The results of this study suggest that legitimacy can be observed on a community level by examining police procedure at the aggregate precinct level and the resultant community response of such policy.
This leads to the third contribution of this study, the examination of precinct level SQF procedure as a policy concern over whether the perception of these procedures intersect with stop dosage and public scrutiny. In other words, much of the research has concentrated on sheer number of stops and their distribution throughout the city. By contrast, this study examines stop quality to begin to examine whether communities are more concerned with stop quantity or the stop quality.
This research has shown that opaque stop procedures at the precinct level are related to that precinct’s complaint rate. Though quantity of law is an important area of study, to have a better understanding of police/public interaction, the quality of police procedure and the subsequent response by the community must be explored. When citizens are presented with opaque stop procedures, they are more likely to lodge a complaint with the CCRB, and these complaints can be recognized not simply as a criticism of a singular “bad” officer but as a denunciation of the procedures prevalent in their neighborhoods. Future research and policy should address this point more decisively.
1Stops are recorded using the NYPD UF-250 form. Though these forms represent an important source of data, it is likely that due to incomplete and underreported encounters, SQF records likely undercount the number of police stops that occur each year. Despite these pitfalls, public UF-250 data is routinely used for policy and research as it is not only the “official record” but also because it allows to track police tactics over time and across areas. We mitigate issues in reporting and bias by using multiple measures of police opacity. Data retrieved from: http://www.nyc.gov/html/nypd/html/analysis_and_planning/stop_question_and_frisk_report.shtml
2The precinct covering Central Park was omitted from the analysis, because it has no residential population.
3An issue with random effects models can be whether the effect of the level of a factor is being drawn from a probability sample of that effect. Meaning, if the effect is “naturally occurring” it would be considered fixed, rather than random. However, repeated measures variables are usually treated as random, and though an argument could be made that some of the independent variables in the equation are traditionally fixed, the Hausman test confirms a random effect model to be the prudent choice.
4Bivariate correlations confirm that associations among predictors are below .65 (Pearson’s r), reducing issues of multi-collinearity. Results are available upon request.
Andres F. Rengifo, Phone: 973-353-5237, Email: ude.sregtur@ofignera.
Kurt Fowler, Email: firstname.lastname@example.org.