ARTICLE

10.1177/1098611103253456 Kaminski et al. / USE OF FORCE POLICE QUARTERLY (Vol. 7, No. 3, September 2004)

THE USE OF FORCE BETWEEN THE POLICE AND PERSONS WITH IMPAIRED JUDGMENT ROBERT J. KAMINSKI

University of South Carolina

CLETE DIGIOVANNI

National Naval Medical Center

RAYMOND DOWNS

National Institute of Justice

This article examines the effects of perceived impaired judgment due to mental illness, drugs, or alcohol among a sample of arrestees on police use of force and other outcomes. Using a combined measure of impairment, bivariate analysis of more than 2,000 arrests made by officers from a large southeastern municipal police department suggests that persons with perceived impaired judgment were only mildly problematic for police in this study. Multiple regression analysis of the effects of perceived judgmental impairment on use of force indicates it significantly increased the odds that higher levels of force were used, but the influence of this factor was less than the influence of other factors commonly examined in use-of-force studies. Additional analysis using a disaggregated measure of perceived impaired judgment reveals that the significant effect of the combined measure is a function of suspected drug intoxication rather than suspected alcohol intoxication or mental illness. Keywords: police; use of force; impairment; drugs; alcohol; mental illness

Most experienced police officers can recount at least one particularly violent encounter with a suspect impaired by intoxicating substances or mental illness. These impaired individuals sometimes are reported to possess extraordinary strength and to be immune to the effects of a variety of POLICE QUARTERLY Vol. 7 No. 3, September 2004 311–338 DOI: 10.1177/1098611103253456 © 2004 Sage Publications

312

POLICE QUARTERLY (Vol. 7, No. 3, September 2004)

less-lethal weapons (e.g., pepper spray, baton strikes, beanbag rounds). Gaining control of these persons, therefore, requires substantial effort on the part of police, and, unfortunately, in some cases the use of deadly force is necessary. Although persons whose judgment is impaired from mental illness or intoxicating substances do not necessarily resist police or become violent when confronted, police are concerned about interactions with these individuals. Two situations that often give police officers cause for concern are when suspects appear to be under the influence of alcohol or drugs and when civilians appear to suffer from serious mental or emotional impairments. The concern stems from the fact that in such situations a person’s rational faculties appear impaired. In dealing with problem situations, officers most often talk their way, rather than force their way, into solutions. For this reason, when a civilian is in a highly irrational state of mind, the chances of the police officer having to use force presumably increase and the possibility of injury to both officer and civilian increases as well (Adams, 1999, p. 7). These concerns are valid as there is evidence that arrests of persons impaired by mental illness, drugs, and/or alcohol are common. Substantially more than half of male arrestees in urban areas are impaired by illicit drugs or psychiatric disorders such as dementia or schizophrenia (Taylor, Fitzgerald, Hunt, Reardon, & Brownstein, 2001; Teplin, 2000), and many studies show that police encounters with impaired persons are more likely to involve the use of force than are encounters with unimpaired persons (Crawford & Burns, 1998; Friedrich, 1980; Engel, Sobol, & Worden, 2000; Garner, Buchanan, Schade, & Hepburn, 1996; Garner & Maxwell, 2001; Garner, Maxwell, & Heraux, 2002; Terrill & Mastrofski, 2002; Worden, 1995; for exceptions, see Alpert & Dunham, 1999; Bayley & Garofalo, 1989). Studies of assaults on police indicate that the majority of police assailants are under the influence of drugs or alcohol or suffer from an emotional or psychiatric disorder (Croft, 1985; Kaminski & Sorensen, 1995; Stetser, 2001) and that attacks by mentally ill persons may more often result in The opinions expressed are those of the authors, and they do not necessarily reflect the opinions or policies of the National Institute of Justice or the Department of the Navy. Direct correspondence to Robert J. Kaminski. Robert J. Kaminski is an assistant professor in the Department of Criminology and Criminal Justice, University of South Carolina. He received his Ph.D in criminal justice from the University at Albany, State University of New York. His research interests are primarily in the areas of violence against the police, police use of force, and public perceptions of the police.

Kaminski et al. / USE OF FORCE

313

injury to police officers (Margarita, 1980). Other research shows that arrestees under the influence of drugs are less susceptible to the effects of oleoresin capsicum (OC) or “pepper spray,” whereas suspects under the influence of alcohol may be somewhat more susceptible (Kaminski, Edwards, & Johnson, 1999). Results from the same study suggest also that mentally ill persons are less susceptible to the effects of OC, but the relationship was less robust. Despite substantial evidence suggesting a positive relationship between impairment and use of force, there is substantial variability in findings when specific forms of impairment are examined. One study, for example, found mental impairment to be statistically unrelated to use of force (Terrill & Mastrofski, 2002), and another found drug and alcohol intoxication to be weak or insignificant predictors (Bayley & Garofalo, 1989, p. 9). Other research, however, found statistically significant and positive associations between alcohol impairment and use of force (Friedrich, 1980; Garner et al., 1996) but not drug impairment (Garner et al., 1996). Yet three other studies that used a combined measure of alcohol and drug intoxication found statistically significant and positive associations (Engel et al., 2000; Garner et al., 2002; Terrill & Mastrofski, 2002). Research on the relationship between impairment and use-of-force related injuries found that suspects under the influence of alcohol or drugs (combined measure) were no more likely to injure police than sober suspects (Alpert & Dunham, 1999), and another study found no relationship between assailant alcohol use and officer injury (Ellis, Choi, & Blaus, 1993). Other research, however, suggests an inverse relationship between officer injury and suspect alcohol use; for example, sober assailants were more likely to injure officers than assailants under the influence of alcohol (Kaminski & Sorensen, 1995). In summary, a clear association between impairment and use of force has not been established in the literature. This study seeks to help clarify the relationship by examining the effect of impairment on the likelihood that force is used during arrests, controlling for the effects of several other factors. Impairment is entered into a regression model using a combined measure as well as a disaggregated one to assess the independent effects of different types of impairment (drug intoxication, alcohol intoxication, mental illness).1 In addition, we examine bivariate relationships between general impairment and several other factors, such as the level of threat presented by arrestees, arrestee escape behavior, the amount of physical effort police expend to effect arrests, the types of force used during encounters, the effectiveness of that force, officer and suspect injuries, and so forth.

314

POLICE QUARTERLY (Vol. 7, No. 3, September 2004)

DATA AND METHOD This study was conducted in a large southeastern municipal police department. The agency employed approximately 500 sworn officers and served a population of about 200,000 persons at the time of data collection. New recruits receive several hours of training on how to handle mentally disturbed persons and persons under the influence of drugs or alcohol.2 From August 2000 through February 2001, officers were asked to complete a questionnaire on each person they arrested and booked; routine traffic violations and other encounters that did not result in a booking of the prisoner (e.g., emergency commitments) were excluded from the survey.3 To promote consistency in the completion of the survey forms and to promote officer cooperation, the instrument and the purpose of the research were discussed at all roll-calls at the inception of the study. We also accompanied patrol officers on ride-alongs to further explain the research and allay suspicions regarding the project. Arresting officers gave completed questionnaires to the jailer at a central booking facility at the time they turned over prisoners, and the jailer put the questionnaires into a locked box. One officer, who had a key to the box, periodically emptied it and mailed the batch of completed forms to the study investigators. The questionnaire (see Appendix A) was designed to be anonymous to increase cooperation from officers and help ensure valid responses. From the information requested, it was unlikely that arresting officers, arrestees, or incidents could be identified. Furthermore, the chief ordered that no supervisor review the completed forms, and that order was understood by the patrol officers as well as the department’s leadership. During the 7-month study period, 2,227 persons were arrested by the department and booked at the jail. The number of completed survey forms received was 2,060, providing an overall compliance of 92.5%.

MEASURES OF IMPAIRMENT Arrestee impairment is defined as officers’ perceptions of whether subjects were mentally ill, on drugs, or intoxicated on alcohol. Arrestees who were observed drinking or were thought to be only under the influence of alcohol—but apparently were not intoxicated on alcohol—are not counted as being judgmentally impaired. When officers were not sure whether an arrestee was impaired by mental illness, drugs, or alcohol, they were counted as having intact judgment (i.e., not impaired). By this definition,

Kaminski et al. / USE OF FORCE

315

373 (18.2%) of arrestees were considered judgmentally impaired. Of those impaired, officers reported that 49 (2.4%) were mentally ill, 149 (7.3%) were under the influence of drugs, and 179 (9.2%) were intoxicated on alcohol.4 Similar to several previous studies on use of force, we use a combined measure of impairment for most analyses in this article (see Engel et al., 2000; Garner et al., 2002; Terrill & Mastrofski, 2002). However, because prior research also shows substantial variability regarding the effects of different types of impairment on the use of force, we also estimate a regression model using a disaggregated measure to assess their individual influences.

ANALYSIS The analysis proceeds in three steps. First, basic descriptive statistics are provided. Second, we explore bivariate relationships between impairment and several other variables, many of which have not been examined previously. We limited these analyses to bivariate relationships because of small cell sizes for many of the variables of interest. Tests of independence are conducted using Pearson chi-square (χ2) and the strength of associations is measured using eta (η).5 Third, we examine the relationship of impairment to use of force, controlling for the effects of several other variables. Because the dependent variable consists of three categories, multinomial logistic regression analysis is used.

DESCRIPTIVE STATISTICS Officers with 5 or fewer years of total experience in law enforcement accounted for 58.6% of the arrests, and 85.1% of the arresting officers were male. Of the arrestees, 1,647 (80.1%) were male; 758 (36.9%) were younger than 24 years of age, 618 (30.1%) were 25 to 35 years of age, and 677 (33.0%) were 36 years or older. African American arrestees numbered 1,444 (70.3%) and Caucasians numbered 567 (27.6%). 6 Based on their assessment at the time of arrest, officers believed that 373 (18.2%) arrestees had impaired judgment, regardless of cause (psychiatric disorder, alcohol intoxication [not mere use], or illicit drug intoxication). An additional 436 (21.3%) arrestees had been drinking but are not counted among those perceived to be judgmentally impaired. The vast majority of arrestees, 1,550 (75.7%), exhibited no threatening behavior at the time of their arrest; 297 (14.5%) were upset, agitated, or

316

POLICE QUARTERLY (Vol. 7, No. 3, September 2004)

angry, and 72 (3.5%) were verbally abusive. Eighty-five arrestees (4.2%) engaged in mild struggling, and 43 (2.1%) were physically combative or assaultive. The arresting officers suspected or confirmed that 126 persons (6.2%) had weapons and were uncertain of the weapon status of 64 (3.1%) of those whom they arrested. One hundred sixty-eight persons (8.2%) were arrested after a car and/or foot pursuit by officers. No force beyond a firm grip was required to effect 1,814 (88.6%) of the arrests. Officers judged that moderate effort was required to arrest 194 (9.5%) persons and extreme effort for 42 (2.1%) persons. Beyond firm grip, force used by officers in making arrests consisted of open-hand techniques in 196 (9.6%), punching or kicking in 22 (1.1%), pepper spray in 13 (0.6%), baton controls in 3 (0.1%), baton strikes in 4 (0.2%), neck restraints in 3 (0.1%), the show of a police canine in 9 (0.4%), and the pointing but not firing of a firearm in 42 (2.0%). Table 1 lists the officers’subjective sense of the effectiveness of all forms of force they used in making the arrests reported in this study. The single most effective force was the presence on the scene of a police canine; officers rated it as very effective in 9 out of 9 arrests. The pointing but not discharging of a firearm, used in 42 arrests, was very effective in 39 (92.3%). Verbal tactics, use of a firm grip, and open-hand techniques generally were rated as being very effective in most arrests, about 80%. Other types of force were used relatively infrequently. Most arrested persons were transported from the scene of their arrest in handcuffs as the only restraint (1,973, or 96.8%). Nine prisoners (0.4%) were in handcuffs and leg restraints, and 2 prisoners (0.1%) were hobbled. Officers reported that they used no restraints during transportation of 55 (2.7%) of the arrestees. The majority of arrestees were transported in a sitting position (89.6%), including the 13 who were sprayed with oleoresin capsicum. Of the restrained arrestees placed in a prone position, 4 handcuffed prisoners and 1 in handcuffs and leg restraints were transported while lying on their sides. The 2 hobbled prisoners were reportedly transported in a sitting position. During arrests, 32 prisoners and 18 police officers were injured. Seventeen of the injured arrestees required no medical care, 14 required medical care that did not result in hospitalization, and 1 was admitted to a hospital for at least one overnight stay. Of the 18 police officers who sustained injuries during the arrest process, 12 required no medical care and the other 6 received medical care that did not require overnight admission to a hospital.

Kaminski et al. / USE OF FORCE

317

TABLE 1. Effectiveness of Tactics Used During Arrests 1 = Not Effective Verbal Firm grip Open-hand techniques Punching/kicking Pepper spray Baton controls Baton strikes Neck restraint Canine Point firearm

Count % Count % Count % Count % Count % Count % Count % Count % Count % Count %

233 11.7 74 7.4 11 5.6 3 13.6 2 15.4 1 33.3 0 .0 1 33.3 0 .0 2 4.8

2 = Moderately Effective 186 9.3 113 11.2 31 15.8 9 40.9 4 30.8 1 33.3 2 50.0 0 .0 0 .0 1 2.4

3 = Very Effective

Total

1,576 79.0 819 81.4 154 78.6 10 45.5 7 53.8 1 33.3 2 50.0 2 66.7 9 100.0 39 92.9

1,995 100.0 1,006 100.0 196 100.0 22 100.0 13 100.0 3 100.0 4 100.0 3 100.0 9 100.0 42 100.0

Note: Verbal refers to verbal commands/reasoning with suspect only; firm grip refers to mild/firm grip or holding; open-hand techniques refers to pushing, shoving, wrist locks, grappling, etc., excluding strikes; punching/kicking excludes weapon strikes; canine force refers to presence of dog (show) at scene and not necessarily release of dog. Types of force not used include restraining nets, electric stun devises, beanbag rounds, and discharging of firearm.

BIVARIATE ANALYSES As indicated in Table 2, officers reported that substantially more effort was required to arrest persons when they perceived them as having impaired rather than intact judgment. Specifically, officers indicated that 3.5% of impaired persons required extreme effort to effect arrests versus 1.7% for those with intact judgment. Impaired persons also more often required moderate effort than did nonimpaired persons (13.7% vs. 8.6%, respectively). Although impairment explains little of the variation in level of effort (η = .130), the association between impairment and level of effort is statistically significant, χ2 = 34.59, p ≤ .000. Table 3 shows the relationship between impairment and the level of threat presented by arrestees and escape behavior. Regardless of cause, impaired judgment was significantly associated with more threatening

318

POLICE QUARTERLY (Vol. 7, No. 3, September 2004)

TABLE 2. Level of Physical Effort by Suspect Impairment Suspect Judgmentally Impaired? 0 = Not Judgmentally Impaired

How much effort did it take to gain control of suspect? 1 = No effort 2 = Little effort 3 = Moderate effort 4 = Extreme effort

1 = Judgmentally Impaired

Count

Col. %

Count

Col. %

1,060 434 143 29

63.6 26.1 8.6 1.7

177 130 51 14

47.7 35.0 13.7 3.5

Note: χ2 = 34.59; p ≤ .000; η = .130.

behavior, ranging from agitation and anger to mild struggling, than was associated with intact judgment, χ2 = 57.90, p ≤ .000; η = .133. However, impaired persons were no more likely than unimpaired persons to be combative/assaultive (2.2% vs. 2.1%, respectively). There were no significant differences between the two groups in their propensity to flee on foot or in a vehicle, χ2 = 2.36, p = .124; η = .034, or in their possession, display, or use of a weapon, χ2 = 2.83, p = .092; η = .037. Most persons, whether with intact or impaired judgment, were arrested successfully with no or minimal physical force (see Table 4). Verbal commands alone were moderately or very effective in 1,455 (89.5%) of 1,626 persons arrested with intact judgment and in 305 (83.2%) of 367 persons with impaired judgment. For firm grip/holding, the figures were 730 (93.0%) of 786 persons with intact judgment and 202 (91.4%) of 221 persons with impaired judgment. Pepper spray was moderately or very effective in 5 of 6 unimpaired arrestees (83.3%) and in 6 of 7 arrestees with impaired judgment (85.7%). The presence of a police dog was very effective in the arrest of all 6 persons with intact judgment for whom this method of force was used and in all 3 persons with impaired judgment. (Tests of significance were not conducted in Table 4 due to the sparse data structure.) More than 97% of police officers and arrestees, whether arrestees were judgmentally impaired or not, received no injuries of any kind during the arrest process. As shown in Table 5, among those arrestees, χ2 = 0.32, p = .858; η = .004, and police officers, χ2 = 0.93, p = .336; η = .021, who received injuries requiring medical care, impairment did not appear to be a factor.

Kaminski et al. / USE OF FORCE

319

TABLE 3. Level of Threat and Escape Behavior by Suspect Impairment Suspect Judgmentally Impaired? 0 = Not Judgmentally Impaired Count Most threatening behavior

1 = Not threatening 1,316 2 = Upset/agitated/angry 228 3 = Verbally abusive 43 4 = Mild struggling 54 5 = Combative/assaultive 35 Car/foot pursuit 0 = No pursuit 1,543 1 = Car/foot pursuit 130 Suspect had/used weapon? 1 = Yes 96 2 = No 1,523 3 = Not sure 48

1 = Judgmentally Impaired

Col. %

Count

Col. %

78.5 13.6 2.6 3.2 2.1 92.2 7.8 5.8 91.4 2.9

234 69 29 31 8 335 38 30 325 16

63.1 18.6 7.8 8.4 2.2 89.8 10.2 8.1 87.6 4.3

Note: Impaired × Threat: χ2 = 57.90, p ≤ .000; η = .133. Impaired × Pursuit: χ2 = 2.36, p = .124; η = .034. Impaired × Weapon: χ2 = 2.83; p = .092; η = .037 (weapon coded 1 = yes, 0 = no/not sure).

To summarize, the bivariate analyses suggest that judgmentally impaired persons appear more threatening to officers than unimpaired persons, and they require more effort by officers to arrest. The level of threat presented, however, generally was low; it was primarily in the form of agitation, verbal abuse, or mild physical resistance and not assaults on police. Impairment was not associated with arrestee flight or weapon status, and no differences were observed between the two groups regarding the effectiveness of tactics used by officers in making arrests (e.g., verbal tactics, firm grip/holding, pepper spray, show of canine). Furthermore, impairment was unrelated to the likelihood of officer and suspect injury.

MULTIPLE REGRESSION ANALYSIS Officers reported that approximately 30% of the persons arrested during the study period were intoxicated or under the influence of alcohol; 7.3% were on drugs, and 2.4% appeared to be mentally ill. Excluding those merely under the influence of alcohol and taking into account that some arrestees were classified by officers as being impaired by two or more factors simultaneously (intoxicated and/or on drugs and/or mentally ill), we found that 18.2% of the 2,060 persons arrested and surveyed in this study appeared to be impaired at the time of their arrest. Thus, officers encoun-

320 % Count % Count % Count % Count % Count % Count % Count %

Count % Count % Count

137 8.4 80 10.2 21 14.8 5 41.7 3 50.0 1 33.3 2 50.0 0 0.0 0 0.0 1 3.0

171 10.5 55 7.0 8 5.6 2 16.7 1 16.7 1 33.3 0 0.0 1 33.3 0 0.0 2 6.1

2 = Moderately Effective

113 79.6 5 41.7 2 33.3 1 33.3 2 50.0 2 66.7 6 100.0 30 90.9

1,318 81.1 650 82.85

3 = Very Effective

142 100.0 12 100.0 6 100.0 3 100.0 4 100.0 3 100.0 6 100.0 33 100.0

1,626 100.0 785 100.0

Total

3 5.6 1 10.0 1 14.3 0 0.0 0 0.0 0 0.0 0 0.0 0 0.0

62 16.9 19 8.6

1 = Not Effective

Suspect Judgmentally Impaired?

10 18.5 4 40.0 1 14.3 0 0.0 0 0.0 0 0.0 0 0.0 0 0.0

49 13.4 33 14.9

2 = Moderately Effective

41 75.9 5 50.0 5 71.4 0 0.0 0 0.0 0 0.0 3 100.0 9 100.0

256 69.8 169 76.5

3 = Very Effective

1 = Judgmentally Impaired

54 100.0 10 100.0 7 100.0 0 0.0 0 0.0 0 0.0 3 100.0 9 100.0

367 100.0 221 100.0

Total

Note: Verbal refers to verbal commands/reasoning with suspect only; firm grip refers to mild/firm grip or holding; open-hand techniques refers to pushing, shoving, wrist locks, grappling, etc., excluding strikes; punching/kicking excludes weapon strikes; canine force refers to presence of dog (show) at scene and not necessarily release of dog. Types of force not used include restraining nets, electric stun devises, beanbag rounds, and discharging of firearm.

Point firearm

Canine

Neck restraint

Baton strikes

Baton controls

Pepper spray

Punching/kicking

Open-hand techniques

Firm grip

Verbal

1 = Not Effective

0 = Not Judgmentally Impaired

TABLE 4. Effectiveness of Tactics by Arrestee Impairment

Kaminski et al. / USE OF FORCE

321

TABLE 5. Suspect and Officer Injuries Requiring Medical Treatment by Suspect Impairment Suspect Judgmentally Impaired? 0 = Not Judgmentally Impaired

Suspect injured? Officer injured?

0 = Not injured 1 = Injured 0 = Not injured 1 = Injured

1 = Judgmentally Impaired

Count

Col. %

Count

Col. %

1,649 12 1,659 4

99.3 .7 99.8 .2

367 3 368 2

99.2 .8 99.5 .5

2

Note: Impaired × Suspect Injured: χ = 0.32, p = .858; η = .004; 1 cell (25.0%) has an expected count of 2 less than 5. Impaired × Officer Injured: χ = 0.93, p = .336; η = .021; 2 cells (50.0%) have an expected count of less than 5.

tered impaired persons in approximately one of every five arrests, a substantial fraction. The earlier bivariate tests of the relationships between suspect impairment and officer and suspect injuries are limited because very few injuries occurred during the study period. The small numbers make it difficult to study systematically the relationship between impairment and injury rates, and we should not place too much confidence in those findings. An alternative approach to studying risk is to examine the relationships between the force used in arrests, suspect and officer characteristics, and situational factors. Although most officers and suspects are not injured when force is employed, it is reasonable to assume that arrests requiring the use of force increase the risk of injury to officers and suspects. Panels A and B in Table 6 show, in fact, that the risk of injury to suspects, 2 χ = 142.84, p ≤ .000; η = .27, and officers, χ2 = 123.73, p ≤ .000; η = .25 increases significantly as the level of force increases from no physical force (verbal tactics only), to firm grip/holding, to greater than firm grip/holding. Thus, it is useful to know whether suspect impairment is associated with the probability that force is used in arrests. Because use of force occurs with enough frequency, we are able to employ multiple regression to assess the relationship between impairment and use of force. Table 7 provides the variable names, description, coding, frequencies, and percentages for the variables used in the regression analysis. The dependent variable force consists of three categories. These are (a) no physical force (verbal tactics only), (b) the use of a firm grip or holding only (including routine cuffing), and (c) force greater than a firm grip or holding

322

POLICE QUARTERLY (Vol. 7, No. 3, September 2004)

TABLE 6. Suspect and Officer Injury by Level of Force Suspect Injured? 0 = Not Injured A. Suspectsa Force

1 = No physical force 2 = Firm grip only 3 = Greater than firm grip

Total

Count % Count % Count % Count %

1 = Injured

Total

1,010 99.7 779 99.5

3 0.3 4 0.5

1,013 100.0 783 100.0

207 89.2 1,996 98.4

25 10.8 32 1.6

232 100.0 2,028 100.0

Officer Injured? 0 = Not Injured B. Officersb Force

1 = No physical force 2 = Firm grip only 3 = Greater than firm grip

Total

Count % Count % Count % Count %

1 = Injured

Total

1,015 100.0 782 99.9

1 0.1

1,015 100.0 783 100.0

215 92.7 2,012 99.1

17 7.3 18 0.9

232 100.0 2,030 100.0

2

a. χ = 142.84; p ≤ .000; η = .27. b. χ2 = 123.73, p ≤ .000; η = .25.

(come-along holds, wrestling, punching, baton strikes, canine, etc.). We frequently refer to these categories as no force, low force, and high force, respectively. As shown in Table 7, few arrests involved force of any magnitude. Officers reported that 235 (11.5%) of arrests involved force greater than a firm grip/holding, and 1,028 (50.2%) arrests were made without resorting to physical force of any kind. Independent variables included in the regression model are those factors from the survey thought to influence the probability that physical force is used when making arrests. These variables are the suspects’ sex, age, height, weight, race/ethnicity, and level of impairment; the officers’sex and total years of experience as a police officer; whether a vehicle or foot pursuit was involved in making the arrest; whether the arrestee appeared to possess,

Kaminski et al. / USE OF FORCE

323

TABLE 7. Descriptive Statistics for Variables in the Regression Analysis Variable

Description

Forcea

Level of force used in making arrest

Suspsex Sage Sheight Sweight Nonwhiteb Impaired Copsex Yrsexp

Pursuit Weapon Threat

Coding

1 = Greater than firm grip 2 = Firm grip/holding 3 = Verbal tactics only Suspect’s sex 0 = Female 1 = Male Suspect’s age 1 = Up to 24 2 = 25 to 35 3 = 36 or older Suspect’s height 1 = Up to 5.6 2 = 5.7 to 6.0 3 = 6.1 or taller Suspect’s weight 1 = Up to 145 2 = 146 to 180 3 = 181 or heavier Suspect’s race/ethnicity 0 = White 1 = Non-White Suspect on drugs, alcohol, mentally ill 0 = Not impaired 1 = Impaired Officer’s sex 0 = Female 1 = Male Officer’s years’ experience as officer 1 = 0 to 2 2 = 3 to 5 3 = 6 to 10 4 = 11 or more Arrest involved car or foot pursuit 0 = No pursuit 1 = Pursuit Suspect possessed, displayed, or used 0 = No/not sure 1 = Yes Suspect’s most threatening behavior 1 = No threat 2 = Upset/verbally abusive 3 = Struggling/combative

Freq. 235 786 1,028 408 1,647 758 618 677 616 1,111 328 621 873 561 567 1,487 1,680 373 307 1,747 649 547 630 216 1878 168 1,913 126 1,550 369 128

% 11.5 38.4 50.2 19.9 80.1 36.9 30.1 33.0 30.0 54.1 16.0 30.2 42.5 27.3 27.6 72.4 81.8 18.2 14.9 85.1 31.8 26.8 30.9 10.6 91.8 8.2 93.8 6.2 75.7 18.0 6.3

a. Officers in a small number of arrests failed to report the use of any tactics, verbal or otherwise. We assume that these exclusions were oversights and that these arrests likely involved voice only. These cases are therefore included in category 3 of the dependent variable verbal tactics only. b. All but 43 of non-White arrestees were African American.

display, or use a weapon; and the most threatening behavior of the arrestee toward the officer before the officer handcuffed the suspect or used force. Because the dependent variable is comprised of three categories, multinomial logistic regression is used for the analysis. The logistic regression model predicts the odds of an observation being in one category of the dependent variable versus another (the reference category), given a set of explanatory variables. For this analysis, verbal tactics is the reference category for the dependent variable force.

324

POLICE QUARTERLY (Vol. 7, No. 3, September 2004)

An initial regression model indicated that officer sex and suspect age, weight, and height were unrelated to use of force. These variables were eliminated and the model reestimated (see Appendix B for the results of the full model). Excluding the nonsignificant variables had no appreciative effect on the pseudo-R2 values or on the values of the variables remaining in the model.7 Likelihood ratio tests of the variables remaining in the model, reported in Appendix C, indicate that each contributes significantly and therefore should not be excluded. The results from the trimmed model are presented in Table 8. The results indicate that confronting male arrestees (suspsex) increases significantly the odds that force greater than firm grip/holding (high force) and firm grip/holding (low force) are used over verbal tactics (no force). Specifically, the odds of high force versus no force increase by 217% (100 × [3.17 – 1]) and low force versus no force by 46% (100 × [1.46 – 1]) when arrestees are male. Additional tests (not shown) indicate male arrestees also increase significantly the odds that high force rather than low force is used (odds ratio = 2.17; p = .008). Non-White arrestees (nonwhite) increase the odds of high force versus no force by 53%, but the effect is not quite statistically significant (p = .065). However, the effect of nonwhite is statistically significant for low force versus no force, with a 49% increase in the odds that a firm grip/holding is used rather than verbal tactics when arrestees are non-White compared to when they are White (p = .003). Additional tests (not shown) indicate that being non-White does not increase significantly the odds of using high force over low force (odds ratio = 1.03; p = .892). Judgmentally impaired suspects (impaired) increase the odds of both high force and low force over no force by 57% (p = .051) and 37% (p = .023), respectively. Additional tests (not shown) indicate that impairment does not significantly increase the odds of using high levels of force over low levels of force (odds ratio = 1.15; p = .521). Generally, officers with more experience on the job (yrsexp) are less likely to use force than officers with the least amount of experience. Specifically, the odds that officers with 3 to 5 years of law enforcement experience (yrsexp2) use high force rather than no force are 62% lower than for officers with 2 years of experience or fewer (odds ratio = 0.38, p ≤ .000). Similar effects are observed for officers with 6 to 10 and 11 or more years of experience, but the effects are not statistically significant at the .05 level (p = .088 and .803, respectively). Additional analyses (not shown) indicate that officers with more experience are not significantly more likely than

Kaminski et al. / USE OF FORCE

325

TABLE 8. Multinomial Logistic Regression Model Predicting Use of Force Contrast

Variable

Greater than firm grip Intercept Suspsex Nonwhite Impaired Yrsexp(1) Yrsexp(2) Yrsexp(3) Yrsexp(4) Pursuit Weapon Threat(1) Threat(2) Threat(3) Firm grip only Intercept Suspsex Nonwhite Impaired Yrsexp(1) Yrsexp(2) Yrsexp(3) Yrsexp(4) Pursuit Weapon Threat(1) Threat(2) Threat(3)

Odds Ratio

x-Std. Odds Ratio

Std. Error

0.01 3.17 1.53 1.57 — 0.38 0.67 0.92 4.45 4.27 — 16.09 830.59 0.42 1.46 1.49 1.37 — 0.45 0.66 1.40 1.32 1.31 — 4.29 21.40

— 1.58 1.21 1.19 — 0.65 0.83 0.97 1.50 1.42 — 2.91 5.08 — 1.16 1.19 1.13 — 0.70 0.83 1.11 1.08 1.07 — 1.75 2.10

.38 .30 .23 .23 — .26 .24 .34 .28 .31 — .22 .74 .16 .13 .12 .14 — .14 .13 .18 .22 .23 — .15 .74

Wald Statistic 115.97 14.84 3.42 3.80 — 13.88 2.91 0.06 27.62 22.00 — 155.70 82.43 29.70 8.61 11.84 5.15 — 35.37 10.36 3.53 1.52 1.34 — 98.27 17.05

Sig. .000 .000 .065 .051 — .000 .088 .803 .000 .000 — .000 .000 .000 .003 .001 .023 — .000 .001 .060 .217 .246 — .000 .000

–2 log likelihood = –1447.16 LR χ2(20) = 772.59, p = .0000 McFadden’s R2 = .20 Cragg and Uhler’s R2 = .37 Note: Reference category on dependent variable = verbal tactics; constants are not exponentiated.

officers with the least amount of experience (2 years or fewer) to use high force rather than low force. Examining the second contrast (firm grip only vs. verbal tactics) for yrsexp, we see compared to officers with 2 years of experience or less, officers with 3 to 5 and 6 to 10 years of experience are significantly less likely to use low force than no force (p ≤ .000 and p = .001, respectively). Officers with the most experience (11 or more years), however, are more likely to use low force than no force, but the effect is only near significant (p = .06).8

326

POLICE QUARTERLY (Vol. 7, No. 3, September 2004)

Arrests that involve pursuits (pursuit) significantly increase the odds that higher levels of force are used. Specifically, there is a 345% increase in the odds that force greater than a firm grip/holding rather than verbal tactics is used when arrests involve pursuits (p ≤ .000). Pursuits, however, do not increase the odds that firm grip/holding only is used (p = .217). Additional analyses (not shown) indicate a 239% increase in the odds that high rather than low force is used when arrests involve pursuits (odds ratio = 3.39; p ≤ .000). Suspect possession, display, or use of a weapon (weapon) significantly increases the odds that officers use high force rather than no force by 327% (p ≤ .000) but not the odds of low force versus no force (p = .246). Additional analyses (not shown) indicate weapons also increase the odds that officers use high force rather than low force by 226% (p ≤ .000). The level of threat (threat) presented by suspects significantly increased the level of force used by officers in making arrests. Specifically, the odds that officers use high and low force rather than no force increase by 1,509% and 329%, respectively, when arrestees are upset or verbally abusive (threat2) (p ≤ .000 and .000, respectively). Arrestees that struggle or are combative (threat3) increase the odds of use of high force and low force rather than no force by 82,959% (p ≤ .000) and 2,140% (p ≤ .000), respectively).9 Additional analysis (not shown) indicates that the odds of the use of high force rather than low force increase by 275% (p ≤ .000) and 3,782% (p ≤ .000) when arrestees are upset or verbally abusive (threat2) and when they struggle or are combative (threat3), respectively. In summary, the level of threat presented by arrestees; suspected possession, display, or use of weapons; foot or vehicle pursuits; and the sex of arrestees exhibit the greatest influence on the odds of higher levels of force being used. Arrestee race, impairment, and officer experience, although exhibiting statistically significant effects, appear to exert less influence. Officer sex and most suspect characteristics (age, weight, and height) are unrelated to use of force. Assessing the relative magnitudes of the effects of the independent variables on the dependent variable, however, requires a comparison of standardized estimates. Unlike continuous independent variables, interpretation of standardized coefficients from dummy-coded or categorical variables does not make substantive sense. Nevertheless, comparisons of the relative magnitudes of the effects of categorical independent variables on the odds of use of force can still be made (see Long, 1997, pp. 16-17). The column labeled “x-Std. Odds Ratio” in Table 8 displays odds ratios stan-

Kaminski et al. / USE OF FORCE

327

dardized with respect to the independent variables, or x-standardized coefficients. As suggested by the unstandardized odds ratios, the x-standardized estimates indicate that the level of threat presented by suspects exerts the greatest influence on whether force greater than a firm grip is used in making arrests (x-std. odds ratios = 5.08 and 2.91). The variable threat is followed in descending order of magnitude by suspsex (x-std. odds ratio = 1.58), pursuit (x-std. odds ratio = 1.50), weapon (x-std. odds ratio = 1.42), yrsexp2 (x-std. odds ratio = .65), nonwhite (x-std. odds ratio = 1.21), and impaired (x-std. odds ratio = 1.19). The level of threat presented by the suspect also has the greatest influence on whether a firm grip is used during arrests (x-std. odds ratios = 2.10 and 1.75), followed in descending order of magnitude by nonwhite (x-std. odds ratio = 1.19), suspsex (x-std. odds ratio = 1.16), and impaired (x-std. odds ratio = 1.13). Regarding impaired, then, the odds that higher levels of force are used increases significantly when arrestees are judgmentally impaired, but compared to the other variables in the regression model, impairment appears to have the smallest impact.10

EXAMINING PROBABILITIES An alternative approach to examining odds in the regression model is to calculate changes in the probabilities for the various levels of force given changes in values of the independent variables. Because the values of the obtained probabilities depend on the values of all the other independent variables in the model, it is common practice to set the other regressors to their mean values, which is the approach taken here. The results are presented in Table 9. The top row shows the probabilities of no force (verbal tactics only), low force (firm grip/holding only), and high force (force beyond firm grip/holding) being used during the “average arrest” in the sample, given all the independent variables in the model. Overall, we see that the probability that high force is used in any given arrest is very low (p = .06). Holding the remaining variables at their means, we observe only a small increase in the probability of the use of high force and low force when impaired changes from unimpaired to impaired (i.e., from 0 to 1). Similarly, small changes are observed when suspect sex (suspsex) changes from female to male, when race/ethnicity (nonwhite) changes from White to non-

328

POLICE QUARTERLY (Vol. 7, No. 3, September 2004)

TABLE 9. Changes in Predicted Probabilities for Changes in Independent Variables, Holding All Other Variables at Their Means

p for average person/x Impaired Suspsex Nonwhite Pursuit Weapon Threat Threat Yrsexp Yrsexp Yrsexp

∆ 0-1 ∆ 0-1 ∆ 0-1 ∆ 0-1 ∆ 0-1 ∆ 1-2 ∆ 1-3 ∆ 1-2 ∆ 1-3 ∆ 1-4

No Force

Low Force

High Force

.4815 .4152 .4581 .4538 .3676 .3680 .1865 .0115 .6294 .5522 .4267

.4579 .5094 .4694 .4819 .4496 .4510 .5847 .1940 .3313 .3952 .5346

.0606 .0755 .0725 .0644 .1828 .1810 .2288 .7944 .0393 .0526 .0487

Note: ∆ = change.

White, and when years of experience (yrsexp) as a police officer changes from 0 to 2 years to 3 to 5 years, 0 to 2 years to 6 to 10 years, and from 0 to 2 years to 11 years or more. Large changes in the probability that high force is used occurs, however, when pursuits (pursuit) were involved in making the arrest (p = .18); when arrestees were thought to possess, display, or use a weapon (weapon) (p = .18); and when threat changed from no threat to upset/verbally abusive (p = .23) and from no threat to struggling/combative (p = .79). These results indicate that the suspected presence, display, or use of a weapon; escape behavior; and the level of threat presented by arrestees most strongly affect the probability that higher levels of force are used by officers, although the level of threat presented by arrestees appears to exert the largest effect on the level of force used. Judgmental impairment increases the probability that high force is used only slightly, with the impact being somewhat larger on the probability of the use of low levels of force.

EFFECT OF TYPE OF IMPAIRMENT ON USE OF FORCE Although the focus of this study is on the relationship between general impairment and the use of force, an interesting question is whether the specific types of impairment that comprise our general measure influence the odds of use of force similarly. To examine this issue, we enter the three indi-

Kaminski et al. / USE OF FORCE

329

cators of impairment separately into the full regression model (table not shown). The results indicate that the positive relationship between general impairment and use of force observed in Table 8 is due to drug impairment. Specifically, the odds that high force versus no force is used increase by nearly 100% when arrestees are impaired by drugs (odds ratio = 1.98; p = .052), and the odds of low force versus no force increase by 66% (odds ratio = 1.66; p = .014). The effects of perceived mental status (odds ratio = 1.41; p = .520) and alcohol intoxication (odds ratio = 1.36; p = .338), although in the expected direction, are statistically unrelated to use of force. Whether we use the general measure of impairment or the specific indicators does not affect the other regressors in the model, but because of their differential impacts, it is recommended that researchers examine the effects of specific types of impairment in addition to any combined measures.

DISCUSSION AND CONCLUSIONS Bivariate analysis of the data collected in this study indicates that persons with perceived impairment of judgment generally are only mildly problematic for police. Officers did report that they appeared more threatening and required greater effort to arrest than did persons with perceived intact judgment. Tactics used to make arrests, however, appear to be equally effective on both impaired and unimpaired persons, and impaired persons were no more likely to injure police. Given the small number of cases with injury, the infrequency with which certain tactics were used, and the bivariate nature of these analyses however, the results should be considered preliminary. Larger samples or more efficient sampling methods (e.g., case-control designs) and multivariate methods would allow for a more thorough and reliable analysis of these relationships. Multiple regression analysis of the use of force indicates that impaired judgment increases the odds that force beyond verbal commands will be needed, but the influence of this factor on use of force is less than the influence of other factors (i.e., level of threat, suspected or established weapon possession, pursuit of suspect prior to arrest, and male sex of the arrestee). These findings are consistent with the examination of probabilities, which shows that impairment in judgment increases only slightly the probability that force beyond firm grip or holding is needed to effect arrests, although it increases more substantially the probability that a firm grip or holding, and

330

POLICE QUARTERLY (Vol. 7, No. 3, September 2004)

not just the issuance of verbal commands, is required. Additional analysis using a disaggregated measure of impairment revealed that the significant effect of the combined measure is a function of arrestee drug intoxication rather than arrestee alcohol intoxication or mental illness. Results using the combined impairment measure are congruent with findings from other studies using combined measures (Engel et al., 2000; Garner et al., 2002; Terrill & Mastrofski, 2002). Unfortunately, because these studies did not disaggregate their general impairment variable, we do not know the independent effects of type of impairment on use of force (although Terrill and Mastrofski examined the independent effect of mental illness). Garner et al. (1996) and Freidrich (1980), however, used disaggregated measures and found statistically significant and positive associations between alcohol impairment and use of force. Our analysis using the disaggregated measure, however, suggests that suspect alcohol intoxication is unrelated to police use of force. In addition, whereas we find that drug impairment is significantly and positively related to use of force, Garner et al. (1996) found it was statistically insignificant. Thus, our results regarding the independent effects of alcohol and drug impairment on use of force largely do not conform to the findings in previous research.11 The only study we could identify that included a measure of mental impairment in a multivariate model is that by Terrill and Mastrofski (2002), and our findings agree with theirs in that impairment by mental illness does not appear to increase the likelihood of use of force. To summarize, it appears that arrests of persons impaired by drugs increase only slightly the need for higher levels of force than arrests of unimpaired persons and those impaired by mental illness or alcohol, controlling for other factors. These other factors, such as suspected weapon possession or use, arrestee escape behavior, level of threat, and arrestee sex, are more important determinants of force than is impairment, at least in the sample of arrests analyzed here. Taken at face value, the results suggest that arrests of impaired persons generally are not very problematic for police and that field officers may not need an abundance of specialized training for handling mentally ill arrestees or those under the influence of drugs or alcohol. However, the failure to detect a large effect simply may be because arrests of impaired persons that present significant risks to police or others are rare, and a much larger sample may be required to capture enough “critical incidents” involving arrests of such individuals. Given the substantial costs and effort associated with selecting a very large sample, the case-control design seems

Kaminski et al. / USE OF FORCE

331

an especially efficient and attractive method for capturing enough critical incidents for meaningful analysis (see, e.g., Loftin & McDowall, 1998; Schlesselman, 1982). In any event, even though these critical incidents may be rare, the potentially serious consequences of officers being unprepared for them suggests that police administrators need to expend considerable resources to prepare officers to handle them in a way that minimizes the risks to arresting officers, suspects, and others at the scene.

APPENDIX A Survey Instrument Officer Information 1. Total years of experience as a police officer (regardless of department): 1[ ] 0-2 2[ ] 3-5 3[ ] 6-10 4 [ ] 11-15 5[ ] 16 or more 2. Officer Sex: 1[ ] Male 2[ ] Female Arrestee Information 3. Sex: 1[ ]Male 2[ ]Female 4. Height: 1[ ] Less than 5.0 feet 2[ ] 5.1 - 5.6 feet 3[ ] 5.7 - 6.0 feet 4[ ] 6.1 feet or taller 5. Weight: 1[ ] Less than 100 lbs. 2[ ] 100-125 3[ ]126-145 4[ ]146-165 5[ ]166-180 6[ ]181-200 7[ ] 201 or more lbs. 6. Age: 1[ ] 17 or younger 2[ ] 18-24 3[ ] 25-29 4[ ] 30-35 5[ ] 36-40 6[ ] 41-50 7 [ ] 51-60 8[ ] 61 or older 7. Race: 1[ ] Caucasian 2[ ] African American 3[ ] Hispanic 4[ ] Asian 5[ ] Other 8. Approximate build: 1[ ] Thin 2 [ ] Normal or Athletic 3 [ ] Overweight/Fat 9. Prior to arrest, did you chase the suspect on foot or in a vehicle? 1[ ] Yes—foot 2[ ] Yes—vehicle 3[ ] No—did not chase 10. Do you believe the arrestee: (check all that apply) a. Had a mental disorder? 1[ ]Yes 2[ ]No 3[ ]Not Sure b. Was under the influence of illegal drugs? 1[ ]Yes 2[ ]No 3[ ]Not Sure c. Was under the influence of alcohol? 1[ ] Yes—had been drinking 2[ ] Yes—was intoxicated 3[ ] No 4[ ] Not Sure 11. What was the most threatening behavior of the arrestee toward you BEFORE you used force/cuffed him/her? (Check only one) 1[ ] Not threatening 2[ ] Upset/Agitated/Angry 3[ ]Verbally abusive 4[ ] Mild struggling 5[ ] Combative/Assaultive

332

POLICE QUARTERLY (Vol. 7, No. 3, September 2004)

12. For each type of force listed, please check the box under the column that best describes the effectiveness of the force used by you or any other officers during the arrest. If a type of force was not used, check column 4 (Not Used). Used, Used, Used, Not Moderately Very Effective Effective Effective 1 2 3 a. Verbal commands, reasoning with suspect b. Mild or firm grip/holding c. Pushing, shoving, wrist locks, grappling, etc., excluding strikes d. Any punching or kicking, excluding weapon strikes e. Chemical irritant (pepper spray, CS/CN) f. Any form of baton restraining techniques, excluding strikes g. Strikes with impact weapon or object (baton, flashlight, vehicle) h. Carotid or other neck restraint i. Canine or horse j. Restraining net (thrown by hand) k. Restraining net (fired) l. Electric stunning device m. Beanbag round n. Pointing firearm but not discharging it o. Discharging firearm p. Other force (specify)_______________

Not Used 4

[] []

[] []

[] []

[] []

[]

[]

[]

[]

[]

[]

[]

[]

[]

[]

[]

[]

[]

[]

[]

[]

[] [] [] [] [] [] [] [] [] []

[] [] [] [] [] [] [] [] [] []

[] [] [] [] [] [] [] [] [] []

[] [] [] [] [] [] [] [] [] []

13. How much total physical effort was needed to gain control of the arrestee? 1[ ] No effort 2[ ] Little effort 3[ ] Moderate effort 4[ ] Extreme effort 14. Did the arrestee appear to possess, or display, or use any type of weapon (gun, cutting or blunt instrument, chemical, etc.)? 1[ ] Yes 2[ ] No 3[ ] Not Sure 15. Which one of the following best describes the means of restraint used to transport the arrestee from the site of arrest? 1[ ] No restraints 2[ ] Handcuffs only 3[ ] Handcuffs and leg restraints 4[ ] Hogtie 5[ ] Restrained on a stretcher, litter, or gurney

Kaminski et al. / USE OF FORCE

333

16. What was the position of the arrestee during transport from the site of arrest? 1[ ] Sitting 2[ ] Kneeling 3[ ] Standing 4[ ] Lying face down, on stomach 5[ ] Lying face up, on back 6[ ] Lying on side 17. Where was the arrestee transported to immediately following his arrest by you? 1[ ] Precinct/district lockup facility 2[ ] Jail or central booking 3[ ] Psychiatric unit 4[ ] Medical unit 5[ ] Headquarters 6[ ] A medical or psychiatric unit first, then to any type of lockup facility/central booking/jail 7[ ] Other (specify)_______________________ 18. If the arrestee was injured as a result of the force used in making this arrest, what best describes the condition of the arrestee at your last contact with him/her? 1[ ] N/A—Suspect not injured as result of use of force 2[ ] Injured but in no need of medical care 3[ ] In need of medical care but no overnight hospital stay 4[ ] In need of medical care requiring overnight hospital stay 5[ ] Deceased 19. If any officers were injured as a result of the force used in making this arrest, what best describes the condition of the most seriously injured officer? 1[ ] N/A, no officer injured as result of use of force 2[ ] Injured but in no need of medical care 3[ ] In need of medical care but no overnight hospital stay 4[ ] In need of medical care requiring overnight hospital stay 5[ ] Deceased 20. Including you, what was the total number of officers present to assist in the arrest at the time you first attempted to arrest the suspect? ___________ 21. Including the arrestee, what was the total number of suspects present during this arrest? ____________

APPENDIX B Full Multinomial Logistic Regression Model Predicting Use of Force Force3 Greater than firm grip

β Intercept Suspsex Sage=1

–4.332 1.121 .154

Std. Error

Wald

Sig.

Exp(β)

.567 .341 .235

58.339 10.788 .430

.000 .001 .512

3.067 1.166 (continued)

334

POLICE QUARTERLY (Vol. 7, No. 3, September 2004)

APPENDIX B (continued) Force3

Firm grip only

β Sage=2 Sage=3 Sheight=1 Sheight=2 Sheight=3 Sweight=1 Sweight=2 Sweight=3 Nonwhite Impaired Copsex Yrsexp=1 Yrsexp=2 Yrsexp=3 Yrsexp=4 Pursuit Weapon Threat=1 Threat=2 Threat=3 Intercept Suspsex Sage=1 Sage=2 Sage=3 Sheight=1 Sheight=2 Sheight=3 Sweight=1 Sweight=2 Sweight=3 Nonwhite Impaired Copsex Yrsexp=1 Yrsexp=2 Yrsexp=3 Yrsexp=4 Pursuit Weapon Threat=1 Threat=2 Threat=3

–.211 0 –.143 1.931E-02 0 .153 .118 0 .406 .448 .218 0 –.971 –.440 –.108 1.444 1.423 0 2.815 6.748 –.513 .292 –7.186E-02 –6.635E-02 0 –.323 –.154 0 9.642E-02 .201 0 .409 .274 –.250 0 –.794 –.386 .364 .284 .256 0 1.469 3.091

Std. Error .250 — .358 .282 — .294 .241 — .232 .236 .296 — .260 .241 .340 .287 .311 — .225 .742 .262 .151 .125 .128 — .185 .149 — .153 .129 — .116 .139 .142 — .137 .129 .179 .225 .233 — .148 .743

Wald

Sig.

Exp(β)

.713 — .159 .005 — .272 .241 — 3.070 3.599 .541 — 13.971 3.335 .100 25.257 20.927 — 157.167 82.668 3.846 3.723 .329 .267 — 3.042 1.065 — .395 2.451 — 12.425 3.869 3.095 — 33.548 8.945 4.126 1.601 1.205 — 98.531 17.329

.399 — .691 .945 — .602 .624 — .080 .058 .462 — .000 .068 .752 .000 .000 — .000 .000 .050 .054 .566 .605 — .081 .302 — .530 .117 — .000 .049 .079 — .000 .003 .042 .206 .272 — .000 .000

.810 — .867 1.019 — 1.166 1.126 — 1.501 1.565 1.244 — .379 .644 .898 4.238 4.151 — 16.699 852.084

Note: The column labeled β are log odds; the Exp(β) are odds ratios.

1.339 .931 .936 — .724 .858 — 1.101 1.223 — 1.506 1.315 .779 — .452 .680 1.439 1.329 1.292 — 4.343 21.998

Kaminski et al. / USE OF FORCE

335

APPENDIX C Likelihood Ratio Test Statistics for Variables Retained in Regression Model Effect Intercept Suspsex Pursuit Impaired Weapon2 Nonwhite Yrsexpr Threat2r

–2 Log Likelihood of Reduced Model 674.567 695.069 703.215 680.949 695.971 687.286 734.210 1182.888

Chi-Square

df

Sig.

.000 20.503 28.648 6.382 21.404 12.719 59.644 508.321

0 2 2 2 2 2 6 4

— .000 .000 .041 .000 .002 .000 .000

NOTES 1. As pointed out by an anonymous reviewer, drugs, alcohol, and mental illness are not the only possible sources of impairment among the arrestee population. Judgmental impairment also may be associated with young or advanced age or certain medical conditions, such as epilepsy. Although not considered here, future research should consider these and other alternate sources of impairment. 2. Newly hired recruits receive 640 hours of basic training. Two hours of instruction are spent on handling mentally disturbed subjects. Another 4 hours are dedicated to recognizing, assessing, and handling irrational subjects (e.g., those under the influence of drugs and alcohol). At least 40 hours of training are devoted to use-of-force issues. Officers who complete basic training receive 4 months of on-the-job-training with a field training officer. Forty hours of annual in-service training are also required for all officers. 3. A limitation of the study, therefore, is that an unknown number of contacts with persons, impaired or not, are handled informally and do not result in arrest. We have no information regarding these contacts. Furthermore, some seriously mentally ill persons are not arrested but are transported by police to psychiatric hospitals. Although not captured in the analysis, the latter are likely few in number because bed availability and stringent criteria for admission frequently limit commitments (Teplin, 2000). 4. Only 1 arrestee was reported by police to be mentally ill, on drugs, and intoxicated on alcohol; 12 were reportedly mentally ill and on drugs; 4 mentally ill and intoxicated on alcohol; and 9 on drugs and intoxicated on alcohol. Thus, 26 (7.0%) of the impaired arrestees were perceived as being multiply impaired. In recoding these cases into specific impairment types, drugs took precedence, followed by mental illness. For example, the 1 subject reported to be mentally ill, on drugs, and intoxicated on alcohol was recoded as a drug impairment case; the 12 subjects reportedly mentally ill and intoxicated on alcohol were recoded as being impaired by mental illness.

336

POLICE QUARTERLY (Vol. 7, No. 3, September 2004)

5. The Pearson chi-square (χ2) statistic of independence is commonly used to test the significance of the relationship between categorical variables. The eta coefficient (η) is a measure of the strength of the association and is appropriate when the dependent variable is quantitative and the independent variable has ordered or unordered categories. When squared, the eta coefficient is interpreted as the proportion of the variability of the dependent variable that is explained by knowing the values of the independent variable. 6. According to the 2000 Census, 51% of the jurisdiction’s residents are African American. 7. Pseudo-R2 measures are analogous to the R2 reported in linear regression and are commonly interpreted as the amount of variation in the dependent variable explained by the independent variables. A number of pseudo-R2 measures exist, and because the reported values can vary substantially, we report two: McFadden’s R2, which is a more conservative measure, and Cragg and Uhler’s R2, which is less conservative. 8. At the request of an anonymous reviewer, we also included an interaction term between yrsexp (dichotomized) and impaired (results not shown). The interaction terms for each contrast, however, were not near statistically significant. A likelihood ratio test of the interactions overall also was statistically insignificant (p = .830). We were unable to test interactions among the different forms of impairment and officer experience due to sparse data cells. 9. The very large values for the odds ratios are not errors. Odds, like probabilities, have a lower limit of zero but unlike probabilities, odds have no upper bound and can increase greatly when probabilities change only slightly near their upper boundary of 1. As noted by Pampel (2000), Tiny changes in probabilities result in huge changes in the odds, and show that the odds increase toward infinity as the probabilities come closer and closer to one. As examples, probabilities of .99, .999, .9999 and .99999 produce odds of 99, 999, 9999, and 99999. (p. 11)

10. At the request of an anonymous reviewer, we also included an interaction term between threat (dichotomized) and impaired (results not shown). The results suggest that arrests of impaired persons who appeared threatening increased significantly the odds of the use of low force over no force (odds ratio = 2.23; p = .034) but not the odds of the use of high force over no force (odds ratio = 2.65; p = .064). A likelihood ratio test of the interaction overall, however, was statistically insignificant (p = .060). We were unable to test interactions among the different forms of impairment and level of threat due to sparse data cells. 11. There may be a number of reasons for the disparate findings, such as idiosyncrasies of research design and analytic methods. Possibilities are the use of different sampling strategies (e.g., independent observations of police-citizen encounters; arrest-based officer selfreports; official records), studies conducted in different jurisdictions in different time periods, the use of different dependent and independent variables (e.g., different thresholds for what constitutes “force”; force measured on a continuum vs. a dichotomy; “use” of alcohol vs. alcohol intoxication), and different model specifications (e.g., OLS vs. logistic regression; main effects only vs. inclusion of interactions terms). An additional factor concerns measurement error. Studies examining the effect of impairment on use of force (including the present one) rely on perceptions of whether or not a suspect is impaired and, if impaired, perceptions regarding the nature of the impairment. Acuteness of trained graduate students’ and police officers’ observational skills notwithstanding, observers will not always know when someone is mentally ill or under the

Kaminski et al. / USE OF FORCE

337

influence of drugs or alcohol; they will sometimes mistake physical illness or drug intoxication for mental illness (e.g., seizures from epilepsy, disorientation due to diabetes, bizarre behavior due to LSD), and officers sometimes fail to report information regarding impairment when filling out use-of-force or arrest reports.

REFERENCES Adams, K. (1999). What we know about police use of force. In National Institute of Justice, Use of force by police: Overview of national and local data (pp. 1-14). Washington, DC: National Institute of Justice and Bureau of Justice. Alpert, G., & Dunham, R. (1999). The force factor: Measuring and assessing police use of force and suspect resistance. In National Institute of Justice, Use of force by police: Overview of national and local data (pp. 45-60). Washington, DC: National Institute of Justice and Bureau of Justice. Bayley, D. H., & Garofalo, J. G. (1989). The management of violence by police officers. Criminology, 27, 1-25. Crawford, C., & Burns, R. (1998). Predictors of the police use of force: The application of a continuum perspective in Phoenix. Police Quarterly, 1, 41-63. Croft, E. B. (1985). Police use of force: An empirical analysis. Unpublished doctoral dissertation, University of Michigan, Ann Arbor. Ellis, D., Choi, A., & Blaus, C. (1993). Injuries to police officers attending domestic disturbances: An empirical study. Canadian Journal of Criminology, 35, 149-168. Engel, R. S., Sobol, J. J., & Worden, R. E. (2000). Further exploration of the demeanor hypothesis: The interaction effects of suspects’ characteristics and demeanor on police behavior. Justice Quarterly, 17, 236-258. Freidrich, R. J. (1980). Police use of force: Individuals, situations and organizations. Annals of the American Academy of Political and Social Science, 452, 87-97. Garner, J. H., Buchanan, J., Schade, T., & Hepburn, J. (1996). Understanding the use of force by and against the police (NCJ Report No. 158614). Washington DC: National Institute of Justice. Garner, J. H., & Maxwell, C. D. (2002). Understanding the use of force by and against the police in six jurisdictions (NCJ Report No. 196694). Washington, DC: National Institute of Justice. Garner, J. H., & Maxwell, C. D., & Heraux, C. (2002). Characteristics associated with the prevalence and severity of force used by the police. Report under review. Kaminski, R. J., Edwards, S. M., & Johnson, J. W. (1999). Assessing the incapacitative effects of pepper spray during resistive encounters with the police. Policing: An International Journal of Police Strategies and Management, 22, 7-29. Kaminski, R. J., & Sorensen, D. W. M. (1995). A multivariate analysis of individual, situational, and environmental factors associated with police assault injuries. American Journal of Police, 14, 3-48. Loftin, C., & McDowall, D. (1998). The analysis of case-control studies in criminology. The Journal of Quantitative Criminology, 4, 85-98. Long, S. J. (1997). Regression models for categorical and limited dependent variables. Thousand Oaks, CA: Sage.

338

POLICE QUARTERLY (Vol. 7, No. 3, September 2004)

Margarita, M. C. (1980). Criminal violence against police. Unpublished doctoral dissertation, University of Michigan, Ann Arbor. Pampel, F. C. (2000). Logistic regression: A primer (Sage University Papers Series on Quantitative Applications in the Social Sciences, 07-132). Thousand Oaks, CA: Sage. Schlesselman, J. J. (1982). Case-control studies: Design, conduct, analysis. New York: Oxford University Press. Stetser, M. (2001). The use of force in police control of violence: Incidents resulting in assaults on officers. New York: LFB Scholarly Publishing. Taylor, B. G., Fitzgerald, N., Hunt, D., Reardon, J. A., & Brownstein, H. H. (2001). ADAM preliminary 2000 findings on drug use and drug markets—Adult male arrestees. Washington, DC: National Institute of Justice. Teplin, L. (2000). Keeping the peace: Police discretion and mentally ill persons. National Institute of Justice Journal, 244, 8-15. Terrill, W., & Mastrofski, S. D. (2002). Situational and officer-based determinants of police coercion. Justice Quarterly, 19, 215-248. Worden, R. (1995). The causes of police brutality: Theory and evidence on police use of force. In W. Geller & H. Toch (Eds.), And justice for all: Understanding and controlling police abuse of force (pp. 31-60). Washington, DC: Police Executive Research Forum.

Robert J. Kaminski is an assistant professor in the Department of Criminology and Criminal Justice, University of South Carolina. He received his PhD in criminal justice from the University at Albany, State University of New York. His research interests are primarily in the areas of violence against the police, police use of force, and public perceptions of the police. Clete DiGiovanni was a U.S. government psychiatrist with experience in law enforcement and paramilitary operations, operational support, and research. He is now retired. Raymond Downs was a manager for the Less-Than-Lethal Technology program at the National Institute of Justice. Over a period of 10 years, he initiated and managed other technology program areas for the institute, including Safe Schools Technology and Concealed Weapons Technology. He spent most of his career as a materials scientist in nuclear energy research. He is now retired.