Tag: police performance

Race, Place, and Police-Caused Homicide in U.S. Municipalities

Holmes, Painter II & Smith, Justice Quarterly, 2019

The authors consider that approaches to studying police caused homicides (PCH) have focused typically on two theories, the Minority Threat hypothesis, which borrows from Conflict Theory which suggests that the amount of crime control is directly proportional to the size of the population that threatens the powerful’s interests. Framed as Minority Threat, the theory suggests the level of police caused homicide is in direct relation to the relative size of the Black population. Large populations of Black people are associated with serious criminality and urban violence and are seen as a threat. When increased crime control on the population is enacted, it will thus result in an increase in PCH. In contrast to this linear relationship model, a Power Threat hypothesis suggests a curve, where increases in crime control continues until the minority population reaches enough positions of power, to where their influence decreases the level of crime control on minority populations. The alternative theoretical perspective is the Community Violence hypothesis, which postulates that violent offending will result in more police caused homicides of suspects. Disadvantaged urban black populations have relatively high rates of violence so that Black over-representation in police caused homicides is actually a reflection of the very real threats that officers face in dealing with these greater levels of violence in these communities. Officers use deadly force  when it is necessary in the face of danger and the level of violence in these communities increases the likelihood officers will be put in those situations.

The authors suggest another theoretical approach. The Place hypothesis maintains that the residential segregation of minority populations into areas of concentrated socioeconomic disadvantage increases the likelihood of police officers employing violence against minority citizens. Police patrolling in these disadvantaged places may see minority citizens as particularly threatening, though this is a more subjective threat based on place, rather than the objective threat involved in the Community Violence hypothesis. In this theory the level of threat by minorities is based on  the segregation of the population into what are viewed as dangerous areas, and because minorities are associated with violent crime, they may be automatically viewed as a threat by being segregated in these places. However, research testing Place hypotheses about PCH has produced mixed findings but the authors suggest there may be a non-linear relationship between racial segregation into the disadvantaged areas and PCH.

The authors also considered that the relationship between Hispanics and PCH may need additional exploration. While percentage Hispanic has not typically been found to be a factor in incidence of PCH, the authors consider that group specific models (minority compared to White) may reveal disparities not evident in total incidence analysis, as well as examining the segregation aspect between White and Hispanic populations.

It should be clarified that when the authors are using structural theories like Minority Threat and Place, it is to examine whether these community structures are related to PCH but these theories operate under the unproven assumption that if there is a relationship between community structure and PCH, then that relationship exists because of  biases held by police officers against minorities. These theories, in attempting to make that connection, do not actually examine if the biases exist, nor do they take into account situational factors like suspect demeanor and behavior, the race of officers in these encounters, and attitudes in the community toward police which may either drive that statistical relationship or even negate the relationship between structural conditions and PCH.

Using data from 230 cities with over 100K population who filed Supplemental Homicide Reports with the UCR between 2008 and 2013, their outcome variable was the incidence of felon killed by police officer for the study time period (Range 0-96, Avg. 5.71, S.D. 12.92). The authors noted the small sample size but recognized that other databases include small cities and may have incomplete data,  limited methodological documentation, and a lack of verification procedures. Other variables included city population, population density and geographical region as control variables as well as percent Black and Hispanic to represent the Minority Threat hypothesis, and average violent crime rate, arrest rate per 1,000, and total number of police officers killed in the line of duty during the study period to represent the Community Violence hypothesis. To test Place hypothesis they used two variables, Black and Hispanic dissimilarity taken from the 2010 Discover America in a New Century website, which indicates the degree of separation from Whites across all neighborhoods of a city.

Using negative binomial regression because the data was a count variable, they examined total incidences, finding a larger city population was significantly related to a greater number of PCH, while the Northeast and Midwest regions were negatively associated with PCH. In total incidence, the authors did not find support for the Minority Threat hypothesis; Black percentage was significantly negatively associated with PCH (but ceased to be significant in the group specific analysis) and there was no significant association between Hispanic percentage and PCH. Finding partial support, analysis of Place showed a large significant effect in Black separation but a negligible effect with Hispanic separation. In examining the Power Threat hypothesis there was a curve-linear relationship with the most segregated cites having more incidence of PCH than less segregated cities. In support of the Community Violence hypothesis, the violent crime rate had a large statistically positive relationship with PCH (while both the overall index crime rate and property crime rate were not) as did higher arrest rates. Police officers killed in the line of duty also had a small but significant positive relationship with PCH as well. In addition the researchers also examined but failed to find a relationship between the ratio of Black and Hispanic officers to Black and Hispanic citizen population with PCH, however female officers were significantly positively associated with PCH.

In group specific analysis of Black PCH there were four predictors—black–white segregation, violent crime rate, police officers killed, and percent female officers—with statistically significant, positive relationships to PCH of Blacks. They also saw a similar non-linear effect with Black-White separation with more PCH incidence in areas of greater separation. For Hispanics, the percentage Hispanic, Hispanic-White separation, as well as the Southwest region all had statistically significant positive effects on PCH. However for Hispanics, and in accordance with the Power Threat theory there was a positive relationship with Hispanic population and incidence of PCH until Hispanics reach about 60 % of the population and the relationship reverses with PCH decreasing as Hispanic population increases and they found no non-linear relationship between Hispanic separation and PCH.

The discuss how they found support for both Community Violence and Place hypotheses and some support for all three hypothesis in group specific analyses, noting their study highlighted the importance of using both structural and event based data and variable and group specific analyses. They also note future research could examine officer race in relation to PCH as well as more detailed city and neighborhood analysis of PCH.

Holmes, M. D., Painter, M. A., & Smith, B. W. (2019). Race, place, and police-caused homicide in US municipalities. Justice Quarterly, 36(5), 751-786.

Holmes Painter, II and Smith used variables like population, and arrest rate, to examine the disparity in minority PCH but Tregle, Nix and Alpert remind us that disparity doesn’t equal bias and caution against using imperfect variables like these in examining officer involved shootings (OIS)

Disparity Does Not Mean Bias: Making Sense of Observed Racial Disparities in Fatal Officer-Involved Shootings with Multiple Benchmarks

Tregle, Nix & Alpert, Journal of Crime and Justice, 2019

Following well publicized officer involved shootings incidents starting in2014, Officer Involved Shootings (OIS) started being viewed as not isolated incidents but as a national problem involving bias on the part of the police in their interactions with minorities. However, recent agency level studies show that Blacks are not more likely to be shot by the police than Whites. Unfortunately, the government has failed as to adequately compile data related to OIS to examine this issue on a larger scale. However, in 2015, the Washington Post started compiling data related to fatal OIS, indicating that officers shoot and kill just under 1,000 people a year and 25% are black and 48% are white. While UCR data showed that Blacks made up approximately 37% of violent crime arrests, the Washington Post data revealed that in 2015 more than 80% of fatal OIS invoked a suspect with a weapon (with the UCR showing Blacks accounting for 40-44% of weapon possession arrests).

However, the authors note this data cannot show whether Blacks are more likely to be shot by the police than Whites. Simply because Blacks are over-represented in fatal shootings, relative to their population in general, does not mean there is bias toward Blacks by the police. The authors explain that using population as a measure in this way is flawed. Because, as in medical disease models, the entirety of the population do not all face the same risk of disease, nor do all members of a population face the same risk of coming into contact with the police. For example examining racial disparity in traffic stops based on racial population is inappropriate without determining what portion of the population is actually driving and thus at risk of being stopped. Another issue to contend with is that within that driving population, which groups, because of their driving behavior or vehicle condition (young people, low income citizens), might be more likely to be pulled over.

The authors examine seven variables including, population data, police-citizen interaction data  (from the Bureau of Justice Statistics’ Police Public Contact Survey (PPCS), a supplement to the National Crime Victimization Survey carried out triennial) and UCR arrest data from 2015-2017 to report the odds of Black citizens being shot, relative to White citizens. They note that many studies examining OIS showed Blacks were less likely to be shot or killed by the police compared to Whites, however some studies demonstrated opposite findings, but comparing these studies are difficult because of the use of different benchmarks. To examine whether there were any disparities between race in OIS, the authors utilized seven benchmarks to examine the issue-population, police citizen interactions (police-initiated contacts, traffic stops, and street stops), arrests (total arrests, violent crime arrests, and weapon offense arrests).

Analyzing the odds ratios of Blacks and Whites shot against the benchmarks, the authors first note that fatal OIS are a rare occurrence. For example, although police fatally shot 259 Black citizens in 2015, they did not use lethal force in 140,543 arrests of Black citizens for violent crimes. Similarly, while police fatally shot 497 White citizens in 2015, they did not fatally shoot suspects during 63,967 arrests of White citizens for weapons offenses. The also note that population is a flawed benchmark, that while it indicates that Blacks are over 3.5 times more likely to be shot by the police than Whites, the problem is that the majority of either population are not exposed to the risk of  being fatally shot by the police. Other benchmarks provide mixed and varying results. For clarification, note that odds ratios over 1 indicate Blacks were more likely than Whites to be shot while odds ratios less than one indicate Blacks are less likely than Whites to be shot and the horizontal line represents the confidence interval (the high likelihood that the data point lies within that range). (See Table 1)

Table 1. Black Citizen Odds Ratios of Fatal Officer Involved Shootings Benchmarks

The authors note that the popular perception that blacks are disproportionately shot by the police is based on the flawed benchmark of population, which doesn’t consider the races’ different exposure rates to the police. They suggest that arrest rates are a more appropriate measure since it represents the subset of the population that had interactions with the police that could turn deadly, working under these assumptions: (1) OIS occur in response to perceived imminently dangerous citizen behaviors, (2) Criminal behavior is a reasonable proxy for imminently dangerous behavior, and (3) Arrests are a reasonable proxy for criminal behavior. Based on total arrests, Blacks are 1.23 to 1.37 more likely to be fatally shot that Whites in that three-year period but when examining arrests that pose a greater threat to officers like those of weapons offenses or violent crimes, Blacks were slightly less likely to be fatally shot than Whites. However the authors also note that UCR data is not a  complete accounting of all police departments, with small departments being underrepresented, and that arrests are only a subset of police-citizens interactions that could escalate into lethal force incidents like traffic stops, domestics, and mentally ill and suspicious person calls. The authors state that a better benchmark might be police-citizen interactions, however the National Crime Victimization survey also has its limitations regarding who is sampled and that in regards to the risk of being shot, there are a vast number of police-citizen encounters that do not require a level of force, let alone lethal force.

An even better benchmark would be scenarios where officers drew their weapons but did not shoot, comparing shoot-no shoot would exclude interaction where it is improbable that citizens would be shot. However, this benchmark may be more appropriate at a city or agency level, as reporting standards for drawing a firearm vary widely and it may be difficult to compile national data. The authors also note that in examining OIS that the Washington Post database does not include non-fatal OIS. Data from larger cities show that non-fatal OIS range from 20-45%, and fatality may be dependent on other factors like immediacy of medical care. They also note that individual circumstances are not accounted for including suspects’ level of resistance and threatening behavior which will prompt the use of force, and level of force, which may explain some of the racial disparity. In addition, another noteworthy limitation of the study is the inability to benchmark fatal shootings of citizens who posed no imminent threat (i.e., unarmed and not aggressing).

In this case, the research question would be: In order to answer the question of whether Black citizens who pose no imminent threat are more likely to be fatally shot by police than White citizens who pose no imminent threat, given each group’s exposure to police contact, benchmarks would be needed that indicated how often officers interact with unarmed and non-aggressing citizens of each racial group. The authors conclude that the federal government should be compiling data on all OIS to better understand and analyze the conditions under which they occur and that while databases like the Washington Post’s can provide valuable information, the benchmarks used to analyze OIS have assumptions and limitations that must be acknowledged.

Tregle, B., Nix, J., & Alpert, G. P. (2019). Disparity does not mean bias: Making sense of observed racial disparities in fatal officer-involved shootings with multiple benchmarks. Journal of crime and justice, 42(1), 18-31.

While it is apparent that in order to examine any racial disparities in officer involved shooting that appropriate benchmarks be used, we also know that not all OIS are appropriate and that the police do make errors in the application of force. Taylor examined OIS and constructed a typology of police shooting errors, with suggestions on how those errors may addressed.

Beyond False Positives: A Typology of Police Shooting Errors

Taylor, Criminology and Public Policy, 2019

Taylor quotes David Kahneman saying that “There are distinctive patterns in the errors people make. Systemic errors are known as biases, and they recur predictably in particular circumstances. …The availability of diagnostic labels for [these] biases make [them] easier to anticipate, recognize, and understand”. Taylor explains that behavior tends to be systematically connected to the features of peoples’ tools, tasks, previous experiences, training, and environments and that the research findings on human error have consistently demonstrated that situations, behaviors, and decision processes that result in error tend to result in repeated errors across time and people. The examination of errors can be applied to criminal justice research, and more specifically, to police use of deadly force, and a typology of police shooting errors can be constructed.

Error should be defined as, absent any chance outside influence, when a sequence of thoughts or behaviors do not lead to their intended outcome. An officer shooting an unarmed man intentionally is not an error. It may be a violation, but it is not an error because the intent met the outcome. Systemic errors occur when people rely on pattern recognition, developed from repeated exposure to similar patterns and experiences, and automaticity, which is the development of implicit shortcuts in our cognition which speed up our decision making process with a high degree of reliability but can also lead to errors.

In the context of police shooting, errors are typically viewed as either a False Positive error, where a person is presumed to be dangerous by the officer, but is in fact not dangerous, and shot by the officer, or a False Negative, where a police officer or citizen is killed when an officer fails to shoot a dangerous individual. However the authors believe this simple typology can be expanded to cover a wider variety of scenarios, which include misses of the intended target and hits on unintended targets such as citizens and other officers

Table 1. A New Typology of Police Shooting Errors

	TARGET HIT
FIREARM DISCHARGED	Intended	Unintended
Intended	Misdiagnosis Errors	Misses
Unintended	Misapplication Errors	Unintentional Discharge

The authors explain misdiagnosis errors, similar to false-positive errors, as when the officer intended to shoot his firearm, and hit the intended target, but the outcome was unintended, i.e., a non-dangerous person was shot. In these situations, a non-dangerous person was shot in error, sometimes referred to as cell-phone shootings, mistake-of-fact shootings, and perception-only shootings. They note statistics from Los Angeles and Philadelphia that between 2013 and 2017, 14% and 10% respectively, of police shootings involved this type of error. They suggest that while more research is needed, that these errors may stem from pattern recognition. The classic and current police literature notes that through experience officers are attuned to cues of danger and impropriety and these cues prompt the reliance on pattern recognition, where these frequently experienced cues prompt the recognition of, and priming for, a dangerous situation. This leads to decision making shortcuts that prompt officers to go on alert, draw their gun and fire. However these shortcuts can lead to errors when the officer has been primed for a dangerous scenario (such as a dispatch call about a man with a gun), attends to the wrong information , or ignores or misinterprets the right information.

Misapplication errors involved the unintended firing of the firearm but a hit on the intended target. These are referred to in the literature as weapon confusion or Taser confusion shootings , where the officer intended to Taser a person but instead accidentally drew his firearm and shot. This type of error is well documented in the medical and aviation fields, where switching over to a new tool (like a Taser) or procedure has been introduced and a preoccupation or distraction is present, thus causing the misapplication and the unintended outcome. In these cases, training just to sufficiency may be insufficient as newer learned skills tended to be the first to disappear under pressure and replaced by those practiced for a longer period of time. The authors note the typical difference in training time with firearms compared to Tasers, and while it requires more research, this may be a factor in this error.

Misses are an error where the officer intends to fire his firearm but doesn’t hit the intended target, either completely missing or hitting an unintended target. Much of the research on police shooting accuracy indicates a low hit rate, typically less than 50 %, and despite changes in training methods, hasn’t improved over the past 50 years. Between 2013 and 2017, Philadelphia officer hit rates averaged 18% while in that same time period LA officer hit rates  averaged 27%, varying between 18% to 42%. This means that the error is a much more common outcome than the correct one and the authors note there is not a comparative type error in other fields and suggest much more research be conducted to determining and addressing the causes of this type of error.

Unintended discharges are errors which occur when an officer did not intend to fire his weapon, had no intention of hitting a target, but the round in fact struck a target. They are typically referred to as accidental or negligent discharges. Between 2013 and 2017, 17 % of reported LA shooting incidents involved this type of error while between 2006 and 2016, the NYPD reported 19% of their shooting incidents were unintended discharges. Research indicates that unconscious touching of the trigger may be common and when combined with some exertion activity, a co-muscle activation response exerted enough pressure to discharge the weapon. A high number of accidental discharges occurred during routine weapons activity,(storing, cleaning, loading, unloading). Automaticity, where officers have done a task so many times it becomes automatic, allows them to change attentional focus and with a loss of focus on the other task, an error in unintended discharge can occur.

The authors conclude that simply trying to lump all police error shootings into a large sample and look for causal correlation is misguided as the causal mechanics vary between the types of errors but neither is it appropriate to simply look at each case as an isolated incident as causal connections to similar shooting incidents might also be missed. Utilizing this typology will more accurately discriminate between the different types of shooting errors and improve research on police shootings, and, based on the type of error, appropriate means can be employed to reduce those types of errors through policy, training or practice.

Taylor, P. L. (2019). Beyond false positives: A typology of police shooting errors. Criminology & Public Policy, 18(4), 807-822.

Eliminating errors in the use of lethal force is just one way of improving police performance, which can foster and build police legitimacy with the public. James, James, Davis, and Dotson suggest that rather than looking at outcomes to study police-citizen contacts, a more in-depth analysis of police performance that examines officer behavior while accounting for influencing factors, can not only enhance our understanding of officer decision making and behavior but also improve police performance in their contact with citizens.

Using Interval-Level Metrics to Investigate Situational-, Suspect-, and Officer-Level Predictors of Police Performance During Encounters with the Public

James, James, Davis and Dotson, Police Quarterly, 2019

The authors look at factors that may influence how police officers behave during encounters with the public, noting previous research has examined whether suspect race influences officer involved shootings or whether officers use greater force depending on suspect demeanor, or whether neighborhoods predict police-citizen outcomes. However, this research typically focuses on the outcome of the encounter, not the performance of the officer in the encounter. For example, an officer may exhibit fairness and do everything right but still generate a citizen complaint, while another officer may do everything wrong and get away with it if the citizen doesn’t bother to file a complaint. The authors examined a wide variety of 667 incident reports from a large urban department (1500 sworn officers) to examine situational, suspect, and officer level predictors on how officers perform in their interactions with the public. Utilizing a recently established and rigorously developed police encounter performance metric, the authors used interval level metrics to score officer performance across the range of these encounters which include Use of Force, Tactical Social Interaction (officer performance in routine citizen encounters), and Crisis Intervention, which involved officer performance in crisis encounter or encounter with people with mental illness.

Within these three metrics are a wide variety of performance measurements. For example, under Use of Force there are 48 performance variables within the categories of Preplan (expecting to be involved in a deadly force situation, waiting for backup) Observe/assess (correctly identifies threats, identifies pre-assault indicators, aware of what is going on in the periphery, selecting reasonable force options), Officer Behavior (paying attention to details, drawing the weapon, able to use communication skills to defuse, used appropriate level of assertiveness), Tactics (had necessary equipment, prioritizing citizen safety, prioritizing other officer safety, using cover, effectively engaging multiple opponents) and Adapt (correctly responds to a threat, recognizes need to transition to different force option, uses or compensates for environmental conditions). Tactical and Social Interaction and Crisis Intervention also utilized extensive performance variables under similar categories, including Resources, Interaction, and Closing the Encounter.

Each of these variables carried a score indicating that behavior’s impact on performance. The incident reports were than analyzed and coded if the officer took the action, or whether the officer could have taken the action but did not. Not all performance metrics were suitable for every encounter and so were not included in the scoring and analysis. The performance scores of officers are expressed as a percentage, the proportion of all behaviors that were possible in the encounters, as measured by the metrics. In addition to this, the authors also coded situational (nighttime, children present, cultural or language barrier, more than one civilian present), suspect (age, sex, race, non-compliant, armed, hostile, homeless, emotionally disturbed, substance impaired, self-harming behavior), and officer (sex) level variables and analyzed them for their effect on officer performance.

Overall, across all incidents the average performance score was 80.5%. Officers scored highest in crisis encounters (83.6%), aggravated assaults (83.4%), and domestic violence incidents (82.4%) but scored lower in traffic collisions (74.8%), harassment calls (76.9%) and investigation of suspicious circumstances (76.7%). See Table 1 below with average officer performance scores and their error bars at a 95% confidence interval.

Table 1. Citizen Interaction Specific Police Officer Performance Scores

To investigate this average 20% performance deficit, the authors examined specific categories and found officers scored highly in Observe/assess (96%) and Closing (93.6%) but less proficient with Preplanning (80.5%), Adapting Tactics (83.8), and use of Tactics (84.4). They also note officers performed far better in crisis encounters (94.5%) than in routine (non-crisis) police/citizen interactions (76.9%).

When the authors examined situational factor influence on officer performance, they found similar performance irrespective of night or day, the presence of children, or the presence of cultural barriers with a slightly better performance in the presence of language barriers (84.2%) than without (81.8%) and statistically significantly better performance with more than one civilian present (81.5%) as compared to only one civilian present (78.6%). In analyzing suspect factors, performance was very similar with teens, young adults, and older adults, and slightly higher performance scores with men as opposed to women (84.7% vs 82.1 %). Officers also performed slightly better (mid 80’s percentiles) with substance impaired citizens, the homeless, self-harming individuals, hostile citizens, and armed suspects than with the opposite counterparts to these factors. Officers also had significantly better performance scores in dealing with emotionally disturbed individuals (84.8%), non-compliant citizens (86.3%) and Blacks (85.8%) compared to Whites (83.2%) or Hispanics (83.8%). While officer gender was the only officer related factor that could be analyzed in this study based on incident reports, there was no statistical difference in performance scores based on gender.

The authors suggest that the results indicate that officers perform better in crisis or “high stakes scenarios as evidenced by their higher performance in crisis incidents like domestics and aggravated assault. This may occur as officers are calling upon tasks that they excel at like vigilant situational assessment, the use of tactics, and adapting those tactics, with officers scoring high in Observe/assess. The large difference between crisis and routine encounters suggests that while measurements show that officers performed very well with performance items like clearly explaining actions, showing empathy, and demonstrating concern for the citizen but perhaps felt the need to demonstrate this more in crisis situations than in routine encounters. The finding that officers performed better with Blacks than non-blacks might be difficult to interpret. The largest differences between Blacks and non-Blacks were in the Observe/assess category, 99% compared to 95%. It could be suggested that officers have a heightened awareness because of implicit bias, unconsciously associating Blacks with weapons or danger, in line with the Minority Threat hypothesis. Alternately, officers may be paying more attention in encounters with Blacks due to a desire to perform well in these encounters and avoid being labeled as biased, with the authors noting that the department had received implicit bias training in the past year. Officers’ better performance with emotionally disturbed and non-compliant individuals suggests that while officers logically would use humanizing and de-escalation techniques in these situations, across the range of performance behaviors, indications seem to be that officers try harder during situations they perceive as more challenging.

Implications from the study suggest using performance metrics are a better way to assess officer behavior than simply analyzing outcomes, such as whether force was used or the presence of citizens complaints as they may provide a distorted picture of actual officer performance. The authors also urge the use of body worn cameras to aid in the assessment of officer performance. They also recognize that outcomes speak to fair enforcement and building public trust to enhance police legitimacy but rather than the sole measure of police encounters, both performance and outcomes can be analyzed to determine how probabilistic outcomes like use of force, or arrest, are, and how much they are dictated by good or bad officer performance. As well as being used to assess training effectiveness like Crisis Intervention Training, officers can be trained to incorporate de-escalation techniques in a broader range of scenarios where there is a likelihood of escalation, including in routine citizen encounters where techniques like empathizing, reducing the police citizen power differential, and being respectful may foster the perception of police legitimacy as well as reduce the 20% officer performance deficit.

James, L., James, S., Davis, R., & Dotson, E. (2019). Using Interval-Level Metrics to Investigate Situational-, Suspect-, and Officer-Level Predictors of Police Performance During Encounters With the Public. Police Quarterly, 22(4), 452-480.