Striking Out: Race and Support for Police Use of Force

Abstract

This research examines whites’ and blacks’ support for police use of force using a survey‐based experiment that varies the race of the offender across four different scenarios. Bivariate results show that the race of the offender influences blacks’ approval for the use of force by police, but does not affect whites’ approval. Multivariate analyses examine whether the factors influencing support for police use of force vary depending on the race of the offender. Results indicate that the predictors for approval of police use of force differ by the race of respondent, the race of offender, and the appropriateness of the use of force. The implications of the results for police–community relations are discussed.

Keywords:

• race
• police
• use of force
• attitude

Acknowledgements

We thank Stephen Mastrofski and Edward Maguire for their comments on a previous version of this manuscript.

Notes

1. In the landmark case of Graham v. Conner (1989), the U.S. Supreme Court determined that using force to facilitate an arrest must be “objectively reasonable in view of all the facts and circumstances of each particular case…” including the severity of the crime, the threat the suspect poses to officers and others, and whether the suspect is attempting to resist arrest or escape altogether. The “totality of the circumstances” calculus of reasonableness under Graham allows for the fact that police officers are often forced to make split‐second judgments—in circumstances that are tense, uncertain, and rapidly evolving—about the amount of force that is necessary in a particular situation. Further, a reasonable officer is generally defined as one who is properly trained in the legal aspects of policing and police practices regarding the use of force. Using this “reasonable man” standard, however, still leaves plenty of room for subjectivity in defining what is and is not reasonable. For example, Ross (2002) presents an assessment of over 1,200 published cases from lower federal court decisions following the Graham decision. In essence, there were substantial variations in lower court interpretations of the Graham decision, of objective reasonableness, and of the legitimacy of police use of force in a wide variety of situations. There were also substantial variations based on the type and level of force used.

2. In Tennessee v. Garner (1985) the U.S. Supreme Court provided some clarity on appropriate and inappropriate use of force (see Smith, 1998 for a summary of lower court cases in which deadly force was and was not authorized). In essence, Tennessee v. Garner prohibited police from using deadly force when a suspect was attempting to flee and he/she did not pose a significant and immediate threat of death or serious bodily harm to officers or others. However, deadly force and excessive force are not necessarily synonymous; an officer might use excessive force, but it may not be deadly force. According to Ross (2002), many lower federal courts wrestled with further defining and interpreting what excessive force means after the Graham decision, deciding where force was reasonable or excessive across four general categories of cases representing the various methods by which police typically use force in effecting arrests. Those categories included empty‐hand control techniques (which can include hitting), use of handcuffs, use of implements (including batons, flashlights, etc.), and lethal force (firearms).

3. Unlike officers, citizens are not trained in the proper and improper use of force and they generally have limited experience in such situations. Thus, it is possible that situations may occur where police officers are acting lawfully and using force appropriately, but citizens nevertheless perceive their actions as unjust.

4. Research suggests that blacks may have more favorable views of the police than whites in cities like Detroit where blacks are a significant portion of the population and hold positions of authority, such as mayor or police chief (Frank, Brandl, Cullen, & Stichman, 1996).

5. Every household that participates in the Knowledge Network panel receives free hardware, free Internet access, free e‐mail accounts, and ongoing technical support. Multi‐media surveys are delivered by e‐mail on the same standardized hardware, through a television set, about once a week. More detailed information about Knowledge Networks’ methodology and the RCPOS can be found in the appendix of Bobo and Johnson (2004).

6. Under the most stringent assumptions (taking into account the panel recruitment response rate, the household connection rate, and the within‐panel completion rate), the overall response rate was 22% for whites and 25% for blacks. A detailed comparison of the demographic characteristics of the Knowledge Networks panel and of the RCPOS sample with data from the U.S. Census was conducted and indicates that both samples are representative of the national population. Moreover, the demographic characteristics of the respondents who completed Wave 2 closely match those of the respondents who completed Wave 1, indicating that there was no systematic bias in the cases that were lost to attrition. See the appendix in Bobo and Johnson (2004) for a discussion and tables summarizing these comparative analyses.

7. Knowledge Networks computes post‐stratification weights based on Current Population Survey demographic benchmarks (gender, age, race/ethnicity, region, education) to reduce sampling error and bias due to non‐response in the fielding of a survey. All weights are scaled to the respective completed sample size.

8. It is possible that respondents may be less approving of police use of force against a teenager than against an adult. For this reason, our data are not directly comparable to the results from past research. However, the percentage of white respondents in this study who approved of police use of force across the four scenarios is quite similar to the results reported by Barkan and Cohn (1998) in their study of white respondents from the GSS, suggesting that the age of the offender (within a certain range) may not have a large influence on attitudes toward police use of force. Whether an offender’s age influences public support for police use of force is an open question for future research.

9. We performed a series of tests for effective ballot randomization, comparing distributions across race, gender, age, education, household income, region of residence, urbanicity, and political ideology. Of the eight separate tests, only education showed a statistically significant difference across the experimental ballots; those who received the white teenager ballot were slightly more educated than those who received the black teenager ballot.

10. Given the diversity of court decisions on the question of whether striking a fleeing offender constitutes reasonable or excessive force, it is perhaps not surprising that public perceptions on this topic vary. Although the Supreme Court argued for an “objective reasonableness” standard in the Graham decision, Ross’ (2002) analysis of court rulings post‐Graham reveals significant variations in what was deemed reasonable or excessive. According to Ross, a variety of lower federal courts held that use of force was reasonable (i.e., not excessive) when an officer/deputy punched a prisoner who was attempting to flee while being transported to jail (Wisniewski v. Kennard, 1990), used a leg sweep to take down a suspect who was resisting arrest (Wallace v. City of Shelby, 1997), used a bear hug to control a resisting arrestee (Hart v. Rogers, 1998), wrestled a fleeing suspect to the ground because the officers thought they were faced with a dangerous fleeing felon (Matasic v. City of Campbell, 1997), hit a suspect in the head with a flashlight because he was attempting to flee following a vehicle pursuit (Carver v. Bulloch, 1990), and used a TASER to restrain an arrestee who was resisting during the issuance of a citation for jaywalking. Considered collectively, it appears that lower federal court interpretations of reasonable force, following the Graham case, consistently allowed for an officer to hit a fleeing suspect depending on the totality of the circumstances of the case. These court decisions, coupled with the results of the factor analysis, provide legal and empirical support for including the “fleeing custody” scenario in the reasonable force index for white respondents.

11. The exact introduction to the stereotyping question was as follows: “Here are a few phrases that people sometimes use to describe blacks. Of course, no word or phrase fits absolutely everybody, but indicate how well you think it describes blacks as a group. On this scale, a 10 means it is a very good description of blacks and a 0 means it is a very inaccurate description of blacks. You may not agree with some of the phrases we chose, but please try to answer the question anyway. Some rating—even a neutral score of 5—is better than no rating at all. But, if you feel you absolutely cannot answer the question, go immediately to the Next Question button.” The question format mirrors that used in the GSS, except the RCPOS uses a 0–10 scale instead of a 1–7 scale. As Bobo and Zubrinsky (2000, p. 906) note: “The bipolar trait‐rating format has been shown to increase respondent comfort in expressing racial/ethnic group distinctions … and to yield reasonably reliable and valid measures of prejudice….”

12. Survey researchers have long been concerned with how to reduce item non‐response and to handle missing data in quantitative analyses (e.g., Brick & Kalton, 1996; De Leeuw, 2001). Similarly, scholars have also identified concerns with missing data in criminological research (Brame, 2000; Brame & Pasternoster, 2003). In this cross‐sectional survey, negative racial stereotyping was the only independent variable that had any meaningful missing data. We used mode substitution on this variable for three reasons: (1) the mode was the most common answer for both blacks and whites; (2) mean substitution is not truly appropriate because the variable of interest was ordinal; and (3) the mode was also the neutral midpoint of the 0–10 scale. We conducted both parametric (t‐test) and non‐parametric (Mann‐Whitney U) tests for all respondents and for blacks and whites separately on the dependent variables (reasonable and excessive force) for those who completed the stereotype item and those who did not. The results from both tests indicated that there was no significant difference on the dependent variables between those who answered the stereotype item and those who skipped the question. Similarly, chi‐square tests indicated there was no significant difference in the distribution of responses on the dependent variables for those who answered the stereotype item and those who did not. As yet another test, we examined how mode substitution would affect the mean if we considered the negative racial stereotyping variable as an interval measure rather than an ordinal measure. When the cases with missing data were excluded, the mean scores on the stereotyping variable were 4.92 for whites and 3.81 for blacks. When the missing cases were included using mode substitution, the means were 4.93 for whites and 3.89 for blacks. These changes were non‐significant, suggesting that using modal substitution did not impact the within‐group means. Finally, we reran the primary regression analyses excluding the 7% of cases with missing data on the negative racial stereotyping item. In every model, the results were almost identical to those presented in the tables below, and resulted in the same substantive conclusions. As a result of all of these tests, we concluded that mode substitution was appropriate and had no adverse effect on the findings.

13. Items included in Wave 1 of the RCPOS included fear of crime, victimization, racial stereotyping, support for three strikes laws, the belief that racial profiling is widespread, the belief that police treat whites and blacks equally, and views on the amount of police brutality against minority group members. Items included in Wave 2 included the four use of force scenarios, support for stricter parole, and support for trying juveniles in adult court. Political ideology and all demographic variables were collected when respondents initially joined the Knowledge Networks panel.