APA Writing Assignment


Journal of Family Violence

(2011) 26:163–169 DOI


Predicting Officer Physical Assaults at Domestic Assault Calls Richard R. Johnson

Published online: 18 January 2011 # Springer Science + Business Media, LLC 2011

Abstract Police work is rife with the potential for physical harm, and domestic assault calls are one source of assaults on officers. Inability to accurately predict what circumstances might lead to an attack is one cause of officer cynicism and paranoia. Having the ability to identify which suspects pose the greatest threat of assault would allow officers to take appropriate protective measures. Using data collected from 1,951 domestic assault calls across three cities, the present study compared characteristics and behaviors of the batterers with whether or not the batterers physically assaulted the responding officers. Findings revealed five significant batterer characteristics (employment status, shared residence with abuse victim, alcohol consumption, property damage, and hostile demeanor toward officers) that successfully predicted officer assaults. These risk factors may be incorporated into police safety training in the response to family violence calls.

Keywords Domestic violence. Officer safety. Officer assaults

The potential for assaults on police officers handling domestic assault calls has been the subject of some debate over the last few decades. While some studies have found that domestic assault calls only contribute to a very small portion of the assaults that officers experience (Garner and Clemmer 1986; Hirschel et al. 1994; Konstantin 1984), studies in different communities have suggested that domestic assault calls are one of the three most common situations where officer assaults occur (Ellis et al. 1993;

Stanford and Mowry 1990; Uchida et al. 1987). Even if other types of calls have higher rates of officer assault, this does not necessarily mean that domestic assault calls are “safe.” On the contrary, as FBI statistics indicate that from 1980 through 2003 more than 224,000 police officers in the U.S. were assaulted while handling family disturbance calls (Johnson 2007, 2008). Furthermore, while survival rate for officers assaulted in the U.S. has been steadily increasing over the last 30 years, the survival rate for officers assaulted at domestic violence calls has failed to increase for officers assaulted at domestic assault calls. It is the unpredictability and constant potential for danger in the form of a violent physical attack that distinguishes the work environment of the police officer. Inability to accurately predict what circumstances might lead

to an attack is suggested to be a primary cause of police cynicism and officer mistrust of the public, causing officers to emphasize keeping their guard up at all times (Barker 1999; Skolnick 1966; Wilson 1968). This is impossible to achieve, however, and could lead to paranoia or citizen complaints for overly aggressive behavior. Therefore, increasing an officer’s ability to predict which domestic assault situations pose the highest potential for assault could be very helpful. Being able to identify for officers which circumstances pose threats of assault would allow officers to take protective measures. The present study sought to utilize characteristics of the domestic batterer and the situational circumstances of the incident to predict assaults on police officers attending domestic assault calls. This study extended the previous research on police officer assaults by moving beyond simply searching for correlates of assault and developing prediction odds ratios for assaults. A simple correlate usually refers to a factor that occurs at the same time as the outcome, while predictor variables predict an increased probability of the outcome (Loeber 1990). This study

Page 164

sought to predict increased probabilities of assaults on officers. Furthermore, this research targeted assaults at domestic calls, which some previous research has suggested occur somewhat differently than assaults on officers performing other duties (Johnson 2007, 2008). It was hoped that key characteristics could be identified that would be useful in predicting the likelihood of an officer assault and permit the utilization of proactive safety tactics by officers.

Literature Review

While there has been a fair amount of research on the correlates of police use of force against citizens (see for example Alpert and Dunham 1999; Geller and Toch 1995; Terrill and Mastrofski 2002), there has only been limited research on the correlates of physical assaults against police officers. Meyer et al. (1979) reviewed the characteristics of assailants in 1,143 assaults on police officers from 37 municipal law enforcement agencies in five U.S. states. The majority of the assailants in these assaults were young (between ages 18 and 25), male, members of racial minority groups, unemployed, had been consuming alcohol, and had displayed an angry or belligerent demeanor prior to committing the assault. A number of these individuals also had personal problems such as a lack of a high school education, a criminal record, and family relationship difficulties (Meyer et al. 1979). More recently, Pinizzotto et al. (1998) reviewed the assailant characteristics in all law enforcement officer murders reported to the FBI from 1985 to 1994. Again they found that the majority of the assailants were young (between the ages of 15 and 29), male, members of racial minority groups, unemployed, low

socioeconomic status, unmarried, under the influence of drugs or alcohol, and had a criminal record.

Other Western nations have also conducted similar studies and identified similar assailant characteristics. Brown (1994), Moxey and McKenzie (1993), and Noaks and Christopher (1990) analyzed data from officer assaults in the United Kingdom. In all three of these English studies the assailants were found to be predominantly lower income males who were unemployed, had a criminal record, and were consuming alcohol at the time of the assault. Most were between the ages of 17 and 25, with an average age of 22 years old. In Australia, Mayhew (2001) analyzed official police reports and found similar characteristics were prevalent in those who assaulted officers in that nation. The Australian assailants were generally young males of lower socioeconomic status who were intoxicated, with aboriginals being over-represented among the assailants. In Iceland, Bragason (2007) conducted a survey of assaulted police officers who described the characteristics of their assailants. Again it was found that those who assaulted the officers appeared to be predominantly young indigenous males who were unemployed, had a criminal record, and had been consuming alcohol. While the previous research has done a fairly good job of describing the characteristics of the assailants who attack police officers in many nations, they all failed to make any attempt to predict officer assaults. While the person who is most likely to assault police officers is described as a lower class, adolescent male who has been consuming alcohol, it could be argued that this description fits a significant proportion of the citizens with whom the police routinely interact. Yet only small fractions of the intoxicated, lower class, adolescent males the police encounter actually assault them. So how do the assailants differ from the non-assailants that officers typically encounter? The previous literature has been silent on this question. Another weakness to the previous literature is an assumption that all assault incidents are similar. Assailant characteristics are pooled across all types of officer assault incidents. Recently there has been evidence to suggest that there may be differences in both offender characteristics, and how assaults occur, across different types of officer assault incidents (Johnson 2007). Previous studies failed to look specifically at physical assaults of officers handling domestic assault calls to determine what characteristics offenders in these specific situations have, and how these characteristics differ from batterers in domestic assault calls who did not assault the police. The present study attempted to fill this void in the literature.



Data used came from the pooling of datasets from three previous studies. A sample of domestic assault calls was needed that contained descriptive information about the batterer involved, and contained both cases that involved an assault on an officer and cases that did not.

Data from three well known domestic violence arrest experiment studies were used, the first being the Minneapolis Domestic Violence Experiment (Sherman and Rogan 1984) which involved official police reports in 1,250 domestic assault cases handled by the Minneapolis Police Department. The second was a replication of the Minneapolis experiment that was conducted in Milwaukee, Wisconsin, involving 921 domestic assault cases handled by the Milwaukee Police Department (Sherman et al. 1991). The last study was a replication study in Miami, Florida that involved 907

Page 165

domestic assault cases handled by the Metro-Dade Police Department (Pate et al. 1991). 1

Even though these data were collected in the 1980s, they remain relevant today for several reasons. First, they continue to be used in research on family violence published in prestigious social science research journals (Garner and Maxwell 2000; Hickman 2003; Maxwell et al. 2002; Piquero et al. 2006). Second, since empirical evidence suggests that assaults on officers at domestic calls, and officer survival rates from these assaults, have remained stable for more than 30 years (Johnson 2007, 2008), there is little likelihood that any possible changes in police tactics since the 1980s has occurred with respect to handling domestic calls. Finally, the emphasis of this investigation rested on offender characteristics to predict offender behavior, relating to innate human behavioral traits that are unlikely to have evolved significantly over a few decades.

Data collection method used in all three of these studies was identical. The data about the domestic assault incident came from official reports by police officers and from interviews with the domestic abuse victims conducted by the research team within a week after the incident. Information about the batterer’s behaviors before the arrival of the police was also obtained in the abuse victim interviews. Finally, data about the batterer’s prior criminal record was obtained from the state crime computer database (Pate et al. 1991; Sherman and Rogan 1984; Sherman et al. 1991).

After merging the 3,078 cases from these three studies it was discovered that 98 cases (3.2%) involved a female batterer, none of whom had assaulted an officer. Because of this lack of variation in the dependent variable for female batterers in the sample, the decision was made to limit this analysis to only male batterers. In another 33.3% of the cases the batterer had fled the scene prior to the arrival of the police, thus eliminating the possibility of an assault on the officers. These 1,026 cases were therefore eliminated from the analysis, as were three additional cases that involved an assault on officers by the abuse victim rather than the batterer.2 The final dataset for the present study consisted of data on 1,951 incidents of domestic assault that were investigated by the police.

The fact that so many cases were eliminated from the sample for this analysis could raise questions of validity, especially if the batterers excluded from the analysis differed significantly from those that remained. It order to determine the extent of this potential weakness, basic demographic characteristics available for the batterers who remained in the sample were compared with those who were excluded from the sample. These demographic characteristics are displayed in Table 1 and suggest that there were surprisingly few differences between the cases that were excluded and those that remained in the sample. In general, those batterers that were excluded from the analyses were slightly more likely to be male, were slightly younger, and less likely to be married.


The dependent variable in this analysis was whether or not the official report indicated the suspect had physically assaulted the officers handling the domestic assault call. Because the data collected was focused primarily on the batterer and the abuse victim, details were not available on whether or not the officers sustained injuries, the seriousness of the injuries, whether or not the assailant used a weapon, or when during the encounter this attack took place. While having this information would have permitted a richer analysis, the emphasis of this study, however was the prediction of assaults, no matter how minor. Just as with victims of abuse, assaults of any kind are a crime and place police officers at risk of injury and increased stress. Data were available on 12 descriptive characteristics about the batterers in the domestic assaults that served as the independent variables in this analysis. These variables and their descriptive statistics are displayed in Table 2. The first independent variable considered was the batterer’s age, and it was hypothesized that younger batterers would be more likely to assault officers than older batterers. Second, whether the batterer was married to the abuse victim was

Table 1 Comparison between sample and excluded cases Sample Cases Excluded Cases N 1,951 1,127 Percent male 100% 91% Mean age 33.4 30.4 Percent married 51% 47% Percent consuming alcohol 51% 49% Percent using drugs 22% 23% Percent damaged property 18% 16% Percent struck abuse victim 94% 91%

Page 166

Table 2 Variable descriptive statistics (N=1,951) Dependent Variable N %

N % %

Batterer physically assaulted officers

117 6%

Independent Variables

Batterer’s age Min 18 Max 79 Mean 33.4 SD 9.01

Min 18 Max 79 Mean 33.4 SD 9.01

Batterer married to abuse victim

995 51%

Batterer’s prior DV arrests

Min 0 Max 10 Mean 0.89 SD 1.08

Batterer unemployed

702 36%

Batterer has HS/GED

1,093 56% 56%

Batterer residing with abuse victim

741 38%

Batterer consuming alcohol

429 22%

Batterer using drugs

429 22%

Batterer damaged property

351 18%

Batterer struck abuse victim

1,834 94%

Batterer knew police were responding

1,034 53%

Batterer displayed hostile demeanor

1,073 55%

considered, under the assumption that married batterers may hold stronger views of privilege to control their spouses and defend against outside interference by the police. Similarly, the third independent variable was whether or not the batterer and abuse victim resided together. It was hypothesized that batterers who reside with their abuse victims would be more defensive of this territory being invaded by the arrival of an outside authority, such as the police.

The fourth and fifth independent variables dealt with substance use issues, whether or not the batterer had consumed alcohol prior to the arrival of the police, and whether the batterer had used an illegal drug before the arrival of the police. Each of these variables was expected to have a positive relationship with assaulting police officers.

The next three independent variables were believed to be measures of the batterer’s level of aggression at the time of the incident. It was reasoned that whether or not the batterer had damaged property as part of the domestic assault incident, and whether or not the abuse victim had been struck by the batterer as part of the incident, would suggest how violently aggressive the batterer was at the time of the incident. Whether or not the batterer displayed a hostile demeanor toward the officers upon their arrival at the scene was also included in the model to measure if the batterer’s level of aggression was still high when the officers finally arrived.

Whether or not the batterer had been told the police had been summoned was also considered as a possible predictor. When offenders are surprised by the appearance of the police they may be more likely to resist or be antagonistic toward the officers. It was therefore hypothesized that the batterer would be more likely to assault the officers if their arrival on the scene was unexpected, allowing him less time to compose him, flee the scene, or formulate a plan for how to respond to the officers.

The last three predictors in the model related to the batterer’s background, specifically his prior history of domestic assault arrests, his employment status, and his education level. Regarding the batterer’s prior record, all three datasets used for this analysis recorded the number of prior domestic assault arrests the batterer had received. Unfortunately, not all of the datasets included information on each batterer’s total number of prior criminal offense arrests, (which may have been a better predictor of past criminality), or all prior violent offenses (which may have been a better predictor of violent propensity). Whether or not the batterer was employed was included in the model, as was whether or not the batterer had attained a high school diploma or general education development (GED) diploma. It was hypothesized that batterers with prior domestic assault arrests, those who were unemployed, and those who had less than a high school education would be at greater risk for assaulting an officer.

Results Only 117 cases involved an assault on officers, thus supporting the previous literature

that domestic assault calls are not necessarily frequent events (Hirschel et al. 1994; Konstantin 1984; Garner and Clemmer 1986). Because the dependent variable was a dichotomous nominal level measure, binary logistic regression was utilized to determine the significance of the influence the independent variables had on the probability of whether or not the batterer assaulted the police officers on the scene. The logistic regression technique also permitted the determination of the odds each predictor had of increasing the likelihood of assault, while controlling for the influence of all of the other predictors in the model. The results of this logistic regression analysis are presented in Table 3.

As can been seen in Table 3, the model chi-square was highly significant (p<.001), indicating that the combined predictors in the model did a fair job of predicting whether or not the batterer would assault the officers. The Nagelkerke pseudo R2 suggests that the independent variables in the model explained more than a quarter of the variation between cases in the odds of the batterer physically attacking the police. Within the model, however,

Page 167

Table 3 Logistic regression of batterer characteristics on officer assault (N=1,951) Variable Coefficient SE Odds Ratio Batterer age −0.016 .020 0.984 Batterer married 0.070 .327 1.073 Prior DV arrests −0.126 .219 0.882 Batterer unemployed 0.542* .308 1.582 High school/GED 0.047 .308 3.165 Resides with abuse victim

1.152** .342 3.165

Using alcohol 1.706*** .431 5.507 Using drugs −0.562 .459 0.570 Property damaged 0.652* .307 1.919 Abuse victim struck −0.656 .582 0.519 Hostile demeanor 2.530*** .480 12.555 Knew Police were coming

−0.161 .298 0.852

(Constant) −5.792*** 1.029 0.003 Model Chi-square 122.843*** Nagelkerke Pseudo R2


Significance Levels: * p<.05; ** p<.01; *** p<.001

only five of the 12 independent variables were significant in predicting the odds of an assault occurring.

The strongest predictor was whether or not the batterer displayed a hostile demeanor at the arrival of the police. When the suspect displayed a hostile demeanor toward the police, the batterer was almost 13 times more likely to assault the officers than if he did not display a hostile demeanor. The next significant predictor in strength was whether or not the batterer had been consuming alcohol prior to the arrival of the police. Batterers who had been consuming alcohol were five-and-a-half times more likely to assault officers than those who had not. If the batterer shared a residence with the victim he was more than three times more likely to physically assault the police. Batterers who had reportedly damaged property during the domestic incident were almost twice as likely to assault officers as those who had not damaged property. Those batterers who were unemployed were about one-and-a-half times more likely to assault the police as batters who were employed. Finally, all of the independent variables that revealed significant relationships to the dependent variable displayed relationships in the predicted directions.

To reveal the substantive significance these five variables had on predicting whether or not an assault on the police would occur, the cases that bore all five predictors were compared with the cases that lacked all five significant predictors. A total of 43 cases involved a batterer who displayed a hostile demeanor, had been consuming alcohol, resided with his abuse victim, had damaged property in the incident, and was unemployed. Of these cases, 25.6% (n=11) resulted in an assault on the police. On the other hand, 263 incidents in the sample failed to have any of the five significant predictor characteristics, and none of these incidents resulted in an officer assault. Therefore, it appears that simultaneous presence of all five of these predictors increases the chances of an officer being assaulted to better than one in four, while the absence of all of these predictors reduces the chances of assault to less than one in 263.

Discussion and Conclusion

The purpose of this study was to attempt to identify domestic batterer characteristics that helped predict whether or not batterers would assault police officers who responded to their acts of domestic abuse. It was hoped that this exploratory analysis would reveal characteristics that had predictive value in order to help police officers identify when they are at increased risk of being assaulted by a male domestic batterer. It appears that this study took a significant first step in identifying such risk factor characteristics.

As hostile demeanor and alcohol consumption were significant predictors of an officer assault. Prior research has suggested that domestic batterers frequently act composed when officers arrive and tend to act as if the abuse victim is the one who is acting irrationally (Buzawa and Buzawa 2003; Walker 1989). Therefore, batterers who remain hostile in their demeanor even after the police arrive appear likely to let their hostility turn to physical violence against the officers. Thus encountering a batterer who still displays a hostile demeanor when the officers arrive is an uncommon event and appears highly predictive of an impending assault.

As for batterer alcohol consumption, the prior research on police officer assaults suggested that alcohol consumption is correlated with attacks on officers (Brown 1994; Meyer et al. 1979; Moxey and McKenzie 1993; Noaks and Christopher 1990; Pinizzotto et al. 1998), and the same was found here. Alcohol serves as a depressant, contributing to the batterer’s negative mood, and may reduce the batterer’s inhibitions about using force against the police. Alcohol consumption frequently coexists with domestic violence (Buzawa and Buzawa 2003; Saunders 1995; Straus et al. 1980), and appears to increase likelihood of the batterer assaulting the responding police officers.

Batterers who reside with their abuse victim are more likely to assault officers than those who live apart from their victim. This may be due to impressions of territorial control. Prior research has suggested that many batterers have heightened needs for dominance and control in their relationships and their home (Buzawa and Buzawa 2003; Holtzworth-Munroe and Stuart 1994; Walker 1989). They

Page 168

are threatened when they cannot successfully control the behavior of their intimate partners, and appear to be even more threatened when the police enter their home and interject themselves into the situation. In essence, the batterer has lost all influence over what he views as his human and physical property when the police intervene. While this police interference may be difficult for the batterer to accept at any location, it appears to be an even stronger insult when it occurs in his own residence.

It may also be this emphasis on dominance, power, and control that results in unemployed batterers being more likely to assault officers than batterers who are employed. Unemployed batterers may be already struggling to maintain dominance in their home. Being unable to provide for their families, they no longer have the option of using economic means to control their abuse victims. Some abuse victims are willing to tolerate their abuse because of the economic rewards afforded to her and her children if she stays in the relationship (Buzawa and Buzawa 2003; Straus et al. 1980), but if there are very few financial rewards due to the batterer’s unemployment, there are fewer incentives to stay. The batterer begins to lose some

of his power and control in the relationship. This situation can be aggravated further for the batterer if it results in his abuse victim seeking employment to support the family, thus gaining some economic independence that further threatens that batterer’s control. This situation may make the batterer more volatile than normal when the police arrive to interject themselves into that situation, and limit the batterer’s control to an even greater extent.

Destruction of property by the batterer prior to the arrival of the police was another significant predictor of officer assaults. This may be an indicator of the degree of the batterer’s rage. While whether or not the victim was struck was not a significant predictor, it is important to remember how many batterers view their use of force. Their abusive behavior is used to control their victim. When the abuse victim is completely compliant the abuse is usually very minor, often only limited to insults and innuendo. When the abuse victim’s behavior is perceived as less compliant, the severity of the abuse increases proportionally (Buzawa and Buzawa 2003; Saunders 1995; Walker 1989). When the batterer damages property, especially his own possessions, it is not a control technique as much as it is an expression of rage. This rage is then turned on the police when they arrive.

These five significant predictors could be utilized by law enforcement agencies to screen domestic assault calls for their level of danger. In many situations, police dispatchers can collect details about four of these elements from the victim or witness who reports the crime. Even if the caller is a neighbor, the dispatcher can often ascertain from the caller if the suspect resides with the victim, if he has been drinking, and if he has damaged any property. If the caller is a friend or family member, he/she should also be able to reveal if the batterer is unemployed. The dispatcher could then warn the responding officers if these four assault predictors are already present at the scene. The dispatcher can also assign additional units when these elements are present. When the officers arrive, they can determine for themselves if the fifth element, a hostile demeanor toward them, is present. If so, the officers can take defensive measures to isolate the batterer and place him in a position of disadvantage (such as on his knees with his hands interlaced on his head) before questioning him.

Finally, it is important to note for police training purposes that female victims and female batterers in domestic assault situations assault the police only very rarely. Of the 3,078 domestic assault calls in the original sampling frame, only 98 (3.2%) involved a female batterer, illustrating how rarely law enforcement officers encounter female perpetrators of domestic assault. Of these 98 incidents of female perpetrated domestic assault, none of them resulted in an assault on an officer. Three of the male perpetrated domestic assault calls did result in a female abuse victim assaulting an officer. Of the 2,980 incidents of male perpetrated domestic assault calls in the original sampling frame, less than 0.1% involved a victim assault on an officer. In this dataset, only one in 1,026 domestic assault calls handled by the police resulted in a female victim assault on an officer.

Several inferences can be drawn from the findings in this study. First, attacks on police officers while they are handling domestic assault calls do not occur very frequent- ly. In the present

study, officers were assaulted in only 6% of cases in the sample. Nevertheless, due to the sheer volume of domestic assault calls handled by the police in the U.S., there still is a significant risk of assault. For example, if the odds of assault are truly only one in 33 (3%), and a patrol officer in an urban area responds to at least two domestic assault …