Abstract
Important gaps exist in our understanding of aggressive behavior and the extent to which aggression involves one or more types of victims. This information is critical for determining the utility of integrated approaches for violence prevention versus continuation of independent efforts for reducing community violence, partner violence, and child maltreatment. To better understand the overlap in aggressive behaviors within the general population, the authors examine the co-occurrence of self-reports of physically striking strangers, acquaintances, intimate partners, and children among a nationally representative sample of 3,024 U.S. adults. The findings from this cross-sectional random digit dial telephone survey show that more than a third of the population reports engaging in at least one form of aggression and that, of these, a third had perpetrated violence against more than one type of victim. The percent of respondents who reported perpetrating violence against more than one type of victim range from 13% (percent of those striking a friend or acquaintance who also struck a child) to 34% (percent of those striking a friend or acquaintance who also struck a stranger). Furthermore, engaging in one type of aggression substantially increases the odds of engaging in another from 1.5 to 4 times. The findings suggest potential value in pursuing both integrated and independent approaches in research and prevention.
Efforts to understand, prevent, and respond to violence tend to be siloed by the relationship between perpetrators and victims (Daro, Edleson, & Pinderhughes, 2004; Moffitt, Krueger, Caspi, & Fagan, 2000). For example, researchers in the field of partner violence typically focus only on partner violence and often do not measure perpetration of child maltreatment or community violence. Similarly, programs and policies addressing each type of violence also tend to be developed and implemented independently. In part, this may have resulted from the diverse fields involved in addressing each type of violence. Violence against strangers has largely been approached from the field of criminology and the criminal justice system resulting primarily in a focus on punishment. Child maltreatment has been addressed more in the fields of medicine, social work, and psychology, with an emphasis on treatment and safety. Partner violence prevention and control efforts emerged from grass-roots organizations and research centered in gender, resulting in a focus on victim safety and support and perpetrator accountability. The result of the different origins has been independent funding streams, theories, definitions, research, partnerships, prevention, and treatment efforts.
Some in the violence research and prevention field have called for more integrated approaches to reducing aggressive behavior (e.g., for all types—Krug, Dahlberg, Mercy, Zwi, & Wilson, 2002; for child maltreatment and partner violence—Knickerbocker, Heyman, Smith Slep, Jouriles, & McDonald, 2007; Smith Slep & O’Leary, 2001; Tolan, Gorman-Smith, & Henry, 2006; Whitaker, Lutzker, & Shelley, 2005; partner violence and stranger violence—Fagan & Browne, 1994; Fagan & Wexler, 1987; Moffitt et al., 2000; child maltreatment, partner violence, and nonfamily violence—Daro et al, 2004; Guterman, 2004; Hotaling, Straus, & Lincoln, 1990) based on evidence of their co-occurrence or shared risk and protective factors. Integrated approaches might make more efficient use of scarce resources.
Although there are many studies documenting the associations among violence toward different types of victims and their occurrence over the life course, the actual percentage of perpetrators who perpetrate against more than one type of victim in the general population has been reported less often. For example, the oft-cited review of co-occurrence of child maltreatment and partner violence (Appel & Holden, 1998) identified only two studies reporting population-based estimates of co-occurrence, the two National Family Violence Surveys (NFVS; Straus & Gelles, 1990; Straus, Gelles, & Steinmetz, 1980). Although both used the Conflict Tactics Scale (CTS) and random sampling techniques to obtain a large nationally representative sample, the first NFVS conducted in-person interviews of adults in intact couples and the second used random digit dial (RDD) telephone interviews and also included recently divorced or separated adults in the sample. Regardless of the differences in methods, rates of co-occurrence in the overall sample of partner violence and child abuse in the past year for the first and second NFVS among families with at least one minor in the house were almost identical: 7% and 6%, respectively (22% and 18% among families reporting any aggression; Hotaling et al., 1990).
Subsequent reviews (Herrenkohl, Sousa, Tajima, Herrenkohl, & Moylan, 2008; Knickerbocker et al., 2007) have identified two additional population-based surveys assessing the overlap between physical aggression toward a child and a partner. In the stratified multistage random sample of almost 2,000 children born mostly to unwed parents in large cities (Fragile Families and Child Well-Being Study), 39% of the sample participants interviewed reported both partner aggression (any physical, sexual, or psychological aggression ever by current partner using items adapted from the CTS) and spanking (in the past month) of their 3-year-old child (Taylor, Lee, Guterman, & Rice, 2010). Among those who reported either partner aggression or spanking, 45% reported both behaviors. The second population-based survey conducted in-person interviews using the CTS among an RDD sample of intact couples with children aged between of 3 and 7 years in a Northeastern suburban county in the United States and found a similar rate of overlap; 45% of families in the sample (50% of those reporting either form of aggression) reported both partner aggression and aggression toward a child in the past year (Smith Slep & O’Leary, 2005).
There are fewer reports of the overlap between perpetration of partner violence and community violence. A birth cohort study with 21 years of follow-up data found that 9% of the sample reported engaging in both partner violence and community crime in the past year (Moffitt et al., 2000). There are also a few studies examining the relationship between acquaintance (or peer) violence and dating violence. A survey assessing the link between peer and dating violence perpetration in the past year among students enrolled in Grades 7, 9, and 11/12 in a high-risk community in the Northeastern region of the United States found that 14% of the sample (33% of those reporting either behavior) reported both peer violence perpetration and dating violence perpetration (Swahn et al., 2008). In a Canadian cohort of boys, 13% (27% among those reporting either behavior) reported engaging in both peer violence and dating violence (Brendgen, Vitaro, Tremblay, & Lavoie, 2001) and 17% of a sample (39% of those reporting either behavior) of African American and Latino inner-city youth reported engaging in both dating violence and peer violence in the past year (Gorman-Smith, Tolan, Sheidow, & Henry, 2001).
Reports of overlaps in more than two types of violence in population-based samples are also scant. In the 1985 NFVS (Hotaling et al., 1990), 9% of the sample reported both aggression toward their partner and child and also hitting a nonfamily member in the past year (16% among those reporting any aggression). An additional population-based survey examined the co-occurrence of perpetrating violence against a partner, a child, or a nonfamily member in the past year (Orpinas, 1999). Using multistage random sampling in seven Latin American cities and Madrid, home interviews, and three items assessing each type of violence, this survey found rates of co-occurrence in the overall samples of two or more types of violence to be in the range of 0% to 11%. We were unable to calculate the percent of two or more types among those who engage in any type of aggression, from the information provided in the publication.
In sum, there are few studies reporting the percentage of aggressors who perpetrate against two or three types of victims and none assessing more than three in the general population. Although official reports (e.g., police, child protective services) are useful for understanding the characteristics of perpetrators who are identified, we know that only a small subset of all violence is reported. It can be more difficult to accurately assess levels of aggressive behavior in the general population. However, this information is critical for informing prevention strategies. Violence prevention strategies, particularly those focused on primary prevention of perpetration, need to be informed by data on those who engage in aggressive behavior.
In addition, in the studies described previously, violence toward nonfamily members has not discriminated between acquaintance and stranger violence. This distinction might be important because the conditions leading to acquaintance violence may differ from stranger violence. For example, conflict is a frequent motive for acquaintance violence (Cheng et al., 2006) whereas stranger violence may occur more often in the course of other criminal activity (Cornell, Benedek, & Benedek, 1987). Another limitation of the existing research on overlaps in the perpetration of violence across relationships is that it has focused on a limited time frame (e.g., past year or past month). Although assessing shorter periods of time should improve recall, especially of minor violent behavior, longer periods (e.g., lifetime perpetration) would increase the likelihood of capturing the behavior and can better inform primary prevention efforts.
This exploratory study extends the existing research by examining the co-occurrence of aggression against four types of victims: strangers, acquaintances, partners, and children within a nationally representative sample of U.S. adults. The findings may inform our understanding of the need for integrated approaches to studying and preventing violence.
Methods
The data analyzed in this study come from the second Injury Control and Risk Survey (ICARIS-2), Phase-2. This list-assisted random-digit-dial (RDD) telephone survey was conducted from March 12, 2007 to May 30, 2008 among English- or Spanish-speaking adults in all 50 U.S. states and the District of Columbia. The survey, sponsored by the Centers for Disease Control and Prevention’s (CDC’s) National Center for Injury Prevention and Control (NCIPC), collected data on seven injury-specific topic areas, including older (age 65+ years) adult mobility, parental supervision of children (age range: 1-10 years), injuries treated in emergency departments, traumatic brain injury, firearm ownership and access, aggressive behavior against others, and suicide attempts or self-harm. In addition, respondent- (e.g., age, education, marital status) and household-level (e.g., household income) demographic data were collected, but, because no identifying information was collected, participants remained anonymous. The survey questionnaire and protocol were approved by the CDC Institutional Review Board.
The sampling frame for the survey was the Genesys Sampling System “1+ banks,” composed of blocks of 100 telephone numbers, with each block containing at least one listed residential listing. To ensure adequate racial and ethnic minority representation in the sample, telephone numbers were stratified into high (> 10% Black or Hispanic households) versus low (all others) minority strata. Seventy percent of the ICARIS-2 Phase-2 sample participants were drawn from the high-minority stratum.
One adult (aged 18 years or older) was randomly selected for interview from each household. In households with both male and female adults, a gender category was selected with a higher probability of choosing a male. The gender distribution was monitored throughout the course of the study and the probability of selecting a male adjusted as needed to obtain a final sample of approximately equal numbers of male and female respondents. In households with multiple eligible adults of the same gender, a random number generator was used to choose one adult of the selected gender.
Eighty-five percent of those eligible (cooperation rate) and 52% of those contacted (response rate) agreed to participate in the study, resulting in a total of 3,024 participants. The resulting sample was weighted to be nationally representative of noninstitutionalized adults in the United States in terms of sex, age, race and ethnicity, and region.
The analyses presented in this report are based on the four questions ICARIS-2 Phase 2 used to assess perpetration of violence. This module was the seventh in the survey. Respondents were read the following introduction prior to being asked about their own behavior:
People sometimes hit or strike other people. When people strike others it can put them and the other person at risk for being injured. It is important for us to know how often people strike other people so that we can better understand this risk. The next few questions ask about different people that you may have struck on purpose. By “struck” I mean times when you have hit, slapped, pushed, or kicked a person or times when you hit someone with an object or weapon. Sometimes people strike other people in these ways because they are angry, have some sort of problem, need something from that person, or because they are trying to protect themselves from that person, like in self-defense. We are not going to ask you why you struck someone. We are only going to ask you whether it happened. All of your answers will be kept private.
Following this introduction, respondents were asked the number of occasions they had “struck” (a stranger/a friend or acquaintance/an intimate partner/a child) since turning 18 years old. Inclusion of self-defense in this introduction was based on assumptions that this would minimize the effect of social desirability against disclosing violence and increase respondents’ comfort in disclosing perpetration (i.e., by including times when aggression is in self-defense, people who engaged in violence for other reasons would feel more comfortable disclosing). Second, if respondents were asked to only include times when they initiated the violence, the true prevalence might be underestimated because people who use violence often perceive that the other person initiated the violence or that their own violent behavior was necessary or appropriate to protect themselves even if they in fact initiated or escalated the conflict.
Respondents were interviewed using a computer-assisted telephone interviewing (CATI) system. The entire interview took approximately 17 min to complete. In recognition of their contribution to the study, participants were offered the option of receiving a US$5 phone card or approving a US$5 donation to the United Way.
All analyses used weighted data and were conducted with SUDAAN. Univariate analyses provided the percentage of individuals perpetrating each form of aggression overall and within each demographic subgroup. The percent of those striking a child was calculated only among those caring for a child. The difference between demographic subgroups was tested using the prevalence ratio test (alpha = .05) for the two subgroups being compared. Cross-tabs analyses provided the percentage of individuals aggressing toward one type of victim who also reported aggressing toward another type of victim. Chi-square tests of independence were performed to assess whether the paired observations on two variables, expressed in the contingency tables, were independent of each other. To establish the probability of striking a specific victim given that the respondent reported striking another type of victim, we divided the rate of striking Victim B for those who struck Victim A by the rate among those who did not strike Victim A (prevalence risk ratio).
Results
Among the U.S. adult population, 32% reported engaging in at least one type of aggression since turning 18 (Table 1). Thirty percent of those reporting any aggression since turning 18 (or 9% of the U.S. adult population) reported perpetrating violence against two or more types of victim. Striking a child since turning 18 was the most commonly reported type of physical aggression (14% among those who report ever caring for a child and 12% in the overall sample), followed by striking a stranger (12%), striking a partner (11%), striking a friend or acquaintance (9%). Women were significantly more likely to report striking a child or a partner (p < .001), and men were significantly more likely to report striking a friend (p < .01) or stranger (p < .001). These patterns of aggression did not differ significantly by age or income, but there were various significant differences for race and ethnicity. Non-Latino Blacks were significantly more likely than non-Latino Whites (p < .01) and Latinos to strike a child (p < .01), strike an intimate person (p < .05 and p < .001, respectively), and to perpetrate violence against two or more types of victims (p < .01 and p < .001, respectively) and more likely than non-Latino Whites to strike a stranger (p < .05). Whites were significantly more likely than Latinos to strike a friend or acquaintance (p < .05).
Prevalence of U.S. Adults Who Struck a Child, a Partner, a Friend or Acquaintance, or Stranger Since Turning 18 and Number of Types of Victims Perpetrated Against by Sex, Age, Race and Ethnicity, and Income
Analysis restricted to those who have cared for a child.
Estimate unstable.
Compared to non-Latino White.
Compared to Latinos.
p < .05. **p < .01.***p < .001.
Table 2 presents the proportion of respondents aggressing against one type of victim who also aggress against another. Ten of the 12 overlaps exceed the proportions expected by chance. The largest overlaps are found among those striking a friend/acquaintance and those striking a stranger. Specifically, whereas 34% of those striking a friend/acquaintance also struck a stranger, 29% of those striking a stranger also struck a friend. The smallest overlap was among those who struck a child who also struck a friend (13.4%).
Proportion of U.S. Adults Who Struck a Child, a Partner, a Friend or Acquaintance, or Stranger (Row) Since Turning 18, Who Also Engaged in Another Type of Aggression (Column)
Among those who have cared for a child.
p < .05. **p < .001 (based on Chi-square tests of independence).
The probability of aggressing toward a second type of victim if aggressing against a first compared to those not reporting aggressing against a first victim (prevalence risk ratio with 95% confidence intervals) is presented in Table 3. Striking one type of victim increased the probability of aggressing toward others for all types. The highest probability was found for friend/acquaintance and stranger aggression. Those who reported striking a friend or acquaintance were 4 times more likely to report also striking a stranger than those who had not struck a friend or acquaintance. Similarly, those striking a stranger were 3.7 times more likely to also strike a friend compared to those who had not struck a stranger.
Prevalence Risk Ratio (95% Confidence Interval) of Physical Aggression Toward One Type of Victim (Row) When Aggressing Toward Another Type (Column) Versus Not Aggressing Toward Victim in Row for Each Type of Victim Among U.S. Adults
Among those who have cared for a child.
Discussion
The findings of this cross-sectional survey of the perpetration of physical aggression among a large random sample of the adult population in the United States suggest that more than a third of the population has struck someone at least once since turning 18. Thirty percent of those who reported any aggression had struck more than one type of victim. Moreover, engaging in one type of aggression significantly increased the probability of engaging in another.
These findings differ somewhat from previous studies although differences in methods, measures, and populations studied limit comparisons. Prevalence rates for the perpetration of different forms of aggression vary widely depending on the time period assessed and the population sampled. Understandably, samples of populations at higher risk of aggression (e.g., youth—Gorman-Smith et al., 2005; unwed couple—Taylor et al., 2010; parents of small children—Smith Slep & O’Leary, 2005) have much higher prevalence rates and overlaps than those observed in this sample of adults. Also studies assessing multiple indicators of aggressive behavior have found higher rates of aggression toward a child or partner in nationally representative samples of adults (1975 NFVS; Straus & Gelles, 1980). However, the current results do replicate the strong associations between different forms of violence reported by others (e.g., Borowsky, Ireland, & Resnick, 2001; Flannery, Singer, & Wester, 2001; Klomek, Sourander, & Gould, 2010; Moffitt et al., 2000; Rumm, Cummings, Krauss, Bell, & Rivara, 2000; Swahn et al., 2008; Tajima, 2000).
Several methodological factors could cause the prevalence and co-occurrence rates observed to be lower than they actually are. First, the sampling methods excluded individuals without landlines who are increasingly more likely to be young, male, mobile, and to engage in risky behaviors (Blumberg & Luke, 2009; Lee, Brick, Brown, & Grant, 2010). Second, our measures are based on retrospective self-reports and are subject to underreporting due to recall error or reluctance to disclose socially undesirable behavior. Although this is often a problem in self-reported behaviors with negative connotations, self-reports provide higher estimates of perpetration of violence in the general population compared to other available sources. For example, parents’ self-reported child disciplinary practices with the potential for injury or harm to the child was 40 times greater than official child welfare substantiated reports (Theodore et al., 2005). Moreover, the questions were framed in a manner to reduce inhibitions (e.g., the introduction referred to people sometimes striking other people for a variety of reasons, including self-defense) and respondents remained anonymous; therefore, underreporting due to social desirability should be reduced. Poor recall might have also contributed to underreporting given the longer period of assessment (i.e., lifetime perpetration), although longer periods tend to result in higher prevalence rates (Hilton, Harris, & Rice, 1998; Straus et al., 1980). We also excluded aggression perpetrated before the age of 18 when peer violence is usually higher. Third, we might have seen higher disclosure if additional questions were used to assess each aggression toward each type of victim; however, the term “struck” was defined for participants to include various forms of physical aggression. In the case of striking a child, respondents could have included physical punishment. Finally, these data are also limited by their focus on moderate physical aggression. Future research should extend these analyses to include child neglect and perpetration of psychological/verbal and sexual aggression as well as considering other targets of aggression (e.g., self or a parent). Research examining the relationship between different levels of severity and co-occurrence as well as the stability or mutability of selection of targets over time would also be useful.
Despite these and other possible limitations, the findings from this exploratory study provide an indication of the extent of the overlap in aggressive behavior across multiple types of victims within a nationally representative sample of U.S. adults. The finding that 30% of those who engaged in any aggression perpetrated violence against more than one type of victim, as well as the 18% to 50% overlap observed in other studies, suggests the potential benefits for researchers and practitioners from different fields of violence working together to better understand and address these linkages. Furthermore, the preponderance of singly aggressive individuals does not rule out the possibility of some shared causes or mechanisms. In addition, one type of violence is often a risk factor for another (Krug et al., 2002). Cross-cutting efforts would appear to be especially useful in addressing violence directed at strangers and friends or acquaintances as well as violence directed at both partners and children. If the overlap is due to common underlying causes or shared mechanisms, existing effective interventions targeting these contributing factors with the goal of reducing one form of violence could potentially have impact on another resulting in a more efficient use of scarce resources. Swahn et al. (2008) have also suggested using the detection of one type of violence as a marker for exploring and addressing other types.
There may also be value in gaining a better understanding of the potential differences between perpetrating violence in different relationships and between singly aggressive and multiply aggressive individuals. Research examining the similarities or differences in risk factors and developmental trajectories for those involved in only one versus more than one type of violence could increase our understanding of how violent behavior develops in each subgroup and help in identifying potentially different underlying processes. Different risk profiles or developmental trajectories could be indicative of differing etiologies that may require different preventive or therapeutic interventions. For example, a study examining the risk profiles of partner-only, child-only, and dually perpetrating aggressors found that dually aggressive men and women had the highest overall risk across all types of risk factors whereas singly aggressive individuals scored highest on role-specific risk factors that correspond to the type of violence they perpetrate (Smith Slep & O’Leary, 2009). As these authors suggest, this type of research can contribute to more refined theories of violence explaining different profiles and lead to the development of more tailored approaches for prevention and treatment.
Footnotes
Acknowledgements
We would like to acknowledge the contribution of members of the ICARIS-2 (Second Injury Control and Risk Survey) Phase-2 project, who designed and conducted the survey from which these data were obtained. The Project Core Group includes the following: Jieru Chen, MS; Ann Dellinger, PhD, MPH; Marcie-jo Kresnow, MS; Chester Pogostin, DVM, MPA; Richard W. Sattin, MD; and Thomas Simon, PhD, of the National Center for Injury Prevention and Control (NCIPC), Centers for Disease Control and Prevention (CDC); Diane Burkom, MA, and Jane Schulman, PhD, of Battelle Centers for Public Health Research and Evaluation (Battelle) and project associates: Victor G. Coronado, MD, MPH; Phaedra Corso, PhD; Alex E. Crosby, MD, MPH; Linda L. Dahlberg, PhD; Xiangming Fang, PhD; Arlene Greenspan, Dr. Ph, MPH; Robin Ikeda, MD; Joanne Klevens, MD, PhD; Karin Mack, PhD; and Judy Stevens, PhD, of the NCIPC-CDC, and Brenda G. Cox, PhD, and Charles Wolters, MS, of Battelle.
The findings and conclusions in this article are those of the authors and do not necessarily represent the official views of the Centers for Disease Control and Prevention.
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
The authors disclosed receipt of the financial support for the research, authorship, and/or publication of this article from the Centers for Disease Control and Prevention.