Abstract
Due to their high rates of parental maltreatment and violence exposure, youth in the foster care system are considered particularly vulnerable to experiencing intimate partner violence (IPV) in adolescence and young adulthood. Those who have emancipated from foster care may be at a heightened risk, as they are significantly more likely to struggle in a variety of critical domains (i.e., mental health, substance use, and delinquency). This longitudinal study is the first to explore the impact of demographic, individual, family, and foster care system factors on IPV involvement for foster care alumni at age 23/24. Analyses were conducted on three waves of quantitative data from the Midwest Evaluation of the Adult Functioning of Former Foster Youth (the Midwest Study). We find that approximately 21% of the young adults in our sample were involved in some type of IPV at age 23/24, with bidirectional violence the most commonly reported form. Males were more likely than females to report IPV victimization, whereas females were more likely than males to report IPV perpetration and bidirectional violence. Young adults who reported parental IPV prior to foster care entry were more likely to be involved in bidirectionally violent partnerships than nonviolent partnerships in young adulthood, as were young adults who reported neglect by a foster caregiver and those who reported greater placement instability while in the foster care system. Anxiety at baseline increased the odds of IPV perpetration at age 23/24, and posttraumatic stress disorder (PTSD) at baseline decreased the odds of IPV perpetration at age 23/24. Understanding the characteristics and experiences that place these young adults at risk for IPV will allow for more effective and targeted prevention efforts.
Introduction
Intimate partner violence (IPV), defined as physical, sexual, and/or psychological abuse perpetrated by one member of a romantic partnership against the other member, is dangerously pervasive in the United States. Rates of IPV among adolescents and young adults (ages 18-25) can range from 9.2% to 90%, depending on measures and definitions of violence (Eaton, Davis, Barrios, Brener, & Noonan, 2007; Halpern, Oslak, Young, Martin, & Kupper, 2001; Halpern, Spriggs, Martin, & Kupper, 2009; Harper, Austin, Cercone, & Arias, 2005; Spencer & Bryant, 2000). Approximately 32% to 39% of adolescent males and 43% to 52% of adolescent females report having been physically aggressive with their intimate partner at least once (Cascardi, Avery-Leaf, & O’Leary, 1994; O’Keefe, 1997). Victims of IPV are more likely to report a range of serious physical and psychiatric problems, including anxiety, depression, sleep difficulties, and suicidal ideation (Afifi et al., 2009; Black et al., 2011). For some young women, IPV is fatal; one analysis of girls murdered between 1993 and 1999 found that 10% of 12- to 15-year-old girls and 22% of 16- to 19-year-old girls were killed by an intimate partner (Hickman, Jaycox, & Aronoff, 2004).
IPV among adolescents and young adults differs from adult IPV in a number of ways. For one, it is more common; most victims of IPV first experience victimization prior to age 25 (Black et al., 2011; Faulkner, Goldstein, & Wekerle, 2014). Furthermore, IPV among adolescents appears to be more reciprocal in nature than IPV among adults; numerous studies find that both sexes report perpetration and victimization (Bookwala, Frieze, Smith, & Ryan, 1992; Chiodo et al., 2012; Gray & Foshee, 1997; Whitaker, Haileyesus, Swahn, & Saltzman, 2007). In addition, IPV in adolescence and young adulthood takes place in relationships that have fewer established norms, relational skills, and power dynamics than adult relationships, so that youth may not regard abusive behavior as problematic (Faulkner et al., 2014; Wekerle et al., 2009).
Risk and Protective Factors for IPV in Adolescence and Young Adulthood
Researchers have attempted to identify factors that may place certain adolescents and young adults at greater risk for becoming involved in IPV. These risk factors include demographic variables (race, gender, etc.), individual-level variables (mental illness, substance abuse, etc.), and family system variables (child maltreatment, parental IPV, etc.).
Analyses of nationally representative datasets have shown that youth of color are at greater risk for experiencing IPV than White youth (Halpern et al., 2009; Howard & Wang, 2003a, 2003b; Howard, Wang, & Yan, 2007; Howard, Wang, & Yan, 2008; Reingle, Staras, Jennings, Branchini, & Maldonado-Molina, 2012). Foshee, Reyes, and Ennett (2010) found that Black adolescents were 4 times as likely as White adolescents to perpetrate IPV. However, other studies have found no link between race/ethnicity and IPV victimization (Vézina & Hébert, 2007). Some researchers have theorized that race/ethnicity may function as a proxy for exposure to community violence or neighborhood disadvantage (Fox, Benson, DeMaris, & Wyk, 2002; Vézina & Hébert, 2007), while others suggest that strong cultural identity can function as a protective factor against IPV (Sabina, Cuevas, & Cotignola-Pickens, 2016).
The role of gender as a risk factor for IPV in adolescence and young adulthood is similarly complex. Although research indicates relatively high rates of IPV perpetration by females as well as males (Foshee et al., 2010), these acts may have different meanings, effects, and consequences for males and females within the context of the relationship (Johnson, Giordano, Longmore, & Manning, 2014). Both boys and girls tend to identify anger their primary reason for engaging in violent behavior, but girls are more likely to use violence as means of self-defense or in response to emotional hurt and boys are more likely to report that they engaged in violent behavior to control their partner, or in response to feelings of jealousy (Hickman et al., 2004; O’Keefe, 1997; O’Keefe & Treister, 1998). In addition, male IPV perpetration against females has more severe physical and psychological consequences than female IPV perpetration against males (Capaldi, Kim, & Shortt, 2004). Females are also at much greater risk for sexual victimization within a dating relationship (Foshee, Benefield, Ennett, Bauman, & Suchindran, 2004).
Exposure to family violence, whether through experiencing parental maltreatment or witnessing parental IPV, is the most commonly investigated risk factor for experiencing IPV in as adolescence. Studies have repeatedly shown that youth who are victims of abuse in childhood are at increased risk for perpetrating violence in response to perceived conflict, particularly in their dating relationships in adolescence and young adulthood (Fang & Corso, 2007; Foshee, Bauman, & Linder, 1999; Herrenkohl et al., 2004; Jonson-Reid, Scott, McMillen, & Edmond, 2007; O’Donnell et al., 2006; White & Widom, 2003; Wolfe, Scott, Wekerle, & Pittman, 2001). Fang and Corso (2007) found that formerly maltreated female youth are 8.7% to 10.4% and formerly maltreated male youth are 1.3% to 17.2% more likely to perpetrate IPV in young adulthood than their nonmaltreated counterparts. A history of childhood maltreatment also places adolescents at greater risk for IPV victimization (Banyard, Arnold, & Smith, 2000; Foshee et al., 2004; Manseau, Fernet, Hébert, Collin-Vézina, & Blais, 2008; O’Donnell et al., 2006; Renner & Whitney, 2012; Sabina, Cuevas, & Cotignola-Pickens, 2016). Research has explored various factors that may mediate the relationship between childhood maltreatment and IPV, including emotion dysregulation (Gratz, Paulson, Jakupcak, & Tull, 2009) and symptoms of posttraumatic stress disorder (PTSD; Taft, Schumm, Marshall, Panuzio, & Holzworth-Munroe, 2008; Wekerle et al., 2009; Wekerle et al., 2001; Wolfe, Wekerle, Scott, Straatman, & Grasley, 2004).
Witnessing parental IPV is common for children who have been maltreated by their parents (C. E. Cox, Kotch, & Everson, 2003; Dixon, Hamilton-Giachritsis, Browne, & Ostapuik, 2007; Edleson, 1999; Hazen, Connelly, Roesch, Hough, & Landsverk, 2009; Jouriles, McDonald, Smith Slep, Heyman, & Garrido, 2008; Osofsky, 2003; Rumm, Cummings, Krauss, Bell, & Rivara, 2000; Tajima, 2000). Youth who have witnessed parental IPV are significantly more likely than their counterparts to engage in IPV perpetration (Kinsfogel & Grych, 2004; O’Donnell et al., 2006; O’Keefe, 1997; C. A. Smith, Greenman, Thornberry, Henry, & Ireland, 2015) and victimization (O’Donnell et al., 2006) in adolescence and young adulthood. Widom described the increase in violence perpetration by individuals who were maltreated or witnessed violence as children as a “cycle of violence” (Widom, 1989, 1996) drawing on social learning theory (Bandura, 1977). This theory suggests that children model their behavior after important adult figures, such as parents, and thus, children who witness parental violence are more likely to emulate their parents by employing violence in response to conflict. Children who grow up witnessing parental IPV may also learn to tolerate violence in relationships and consider it an acceptable strategy for resolving differences (Manseau et al., 2008).
Mental illness has also been identified as a risk factor for IPV in adolescence and young adulthood. Specifically, depressive symptoms and a history of suicidality have been identified as risk factors for IPV victimization among both females (Cleveland, Herrera, & Stuewig, 2003; Foshee et al., 2004; Howard & Wang, 2003b; Howard, Wang, & Yan, 2007; Renner & Whitney, 2012) and males (Howard & Wang, 2003a; Howard et al., 2008; Renner & Whitney, 2012; Yan, Howard, Beck, Shattuck, & Hallmark-Kerr, 2010). In addition, depression has been identified as a predictor of IPV perpetration, particularly for females (Foshee, Linder, MacDougall, & Bangdiwala, 2001; Foshee et al., 2010; Reingle et al., 2012; Renner & Whitney, 2012; C. A. Smith et al., 2015), as has anxiety for White adolescents (Foshee et al., 2010). Adolescents experiencing strong life dissatisfaction may seek out high-risk behaviors in an effort to cope with feelings of distress; additionally, feelings of sadness and loneliness may motivate these youth to tolerate abuse in intimate relationships out of fear of abandonment or rejection (Vézina & Hébert, 2007). Trauma symptoms have also been found to predict IPV perpetration (Hahn, Aldarondo, Silverman, McCormick, & Koenen, 2015; Riggs, Caulfield, & Street, 2000; Taft et al., 2008; Wolfe et al., 2004).
The literature on substance use suggests that alcohol use is a risk factor for both IPV perpetration (Foran & O’Leary, 2008; Foshee et al., 2001; Reyes, Foshee, Bauer, & Ennett, 2014; P. H. Smith, Homish, Leonard, & Cornelius, 2012) and victimization (Cleveland et al., 2003; Eaton et al., 2007; Howard & Wang, 2003a, 2003b; Yan et al., 2010). Use of marijuana and other illicit drugs has also been identified as a risk factor for IPV perpetration (Foshee et al., 2010; Reingle et al., 2012; Reyes et al., 2014; P. H. Smith et al., 2012) and victimization (Eaton et al., 2007; Howard & Wang, 2003a, 2003b; Reingle et al., 2012; P. H. Smith et al., 2012). Regular substance use can impair judgment and use of adaptive conflict resolution skills in relationships; it can also heighten the risk for association with peers engaged in delinquent behavior and impede the development of communication and emotion regulation skills needed for healthy relationships (Reyes et al., 2014; Vézina & Hébert, 2007).
Violent behavior in adolescence, including physical fighting, carrying a weapon, and other aggressive and antisocial behaviors, is a strong risk factor for IPV perpetration and victimization (Cleveland et al., 2003; Ehrensaft et al., 2003; Fang & Corso, 2007; Foshee et al., 2010; Howard & Wang, 2003a, 2003b; Howard et al., 2007; Howard et al., 2008; Magdol, Moffitt, Caspi, & Silva, 1998; O’Donnell et al., 2006; Renner & Whitney, 2012; C. A. Smith et al., 2015; Yan et al., 2010). Engaging in delinquent behavior may put a young person at greater risk for experiencing IPV as a result of contact with peers who tolerate violence in relationships and engage in other risky behaviors such as substance abuse (East & Hokoda, 2015; Morris, Mrug, & Windle, 2015; Vézina & Hébert, 2007).
Considerably fewer researchers have explored factors that may protect against the risk of experiencing IPV in adolescence and young adulthood. Richards, Branch, and Ray (2014) found that social support from friends, but not from parents, predicted a reduced risk for physical and emotional IPV perpetration and emotional IPV victimization among adolescents in a longitudinal study. However, Staggs, Long, Mason, Krishnan, and Riger (2007) found that social support was not a protective factor against future IPV. Several studies have identified parental monitoring and closeness with parents as protective against IPV in adolescence (Cleveland et al., 2003; East & Hokoda, 2015; Maas, Fleming, Herrenkohl, & Catalano, 2010; Magdol et al., 1998; Yan et al., 2010).
IPV in Child Welfare Populations
Due to their high rates of parental maltreatment and trauma symptoms, youth in foster care are considered particularly vulnerable to experiencing IPV in young adulthood. Their interpersonal relationships may have been disrupted by frequent transitions, placement changes, and discontinuities in relationships, activities, and environments (Fong, Schwab, & Armour, 2006; Stott & Gustavsson, 2010). Despite the fact that they are at heightened risk for experiencing IPV, only a small number of empirical studies have investigated the prevalence and features of IPV in the foster care population (Jonson-Reid & Bivens, 1999; Jonson-Reid et al., 2007; Manseau et al., 2008; Wekerle et al., 2009; Wekerle et al., 2001).
In the earliest of the studies, Jonson-Reid and Bivens (1999) found that youth in foster care in California were not more likely than youth in the general population to be in violent romantic relationships, but those who were in them were significantly more likely to stay in them for longer periods of time. Higher rates of IPV victimization were associated with being female and self-reported perpetration of IPV (Jonson-Reid & Bivens, 1999). In subsequent studies, a history of maltreatment and symptoms of PTSD were associated with higher rates of IPV for female child welfare–involved adolescents (Jonson-Reid et al., 2007; Wekerle et al., 2001). Jonson-Reid et al. (2007) also found that rates of IPV in emancipating foster youth in Missouri were significantly higher than rates of IPV in the general U.S. population and that drug use was also associated with higher rates of IPV victimization and perpetration.
Manseau et al. (2008) surveyed 196 girls of ages 12 to 18 who were living in out-of-home placement under child protective services in Quebec, Canada. More than half the sample (53.1%) reported at least one experience of severe physical victimization in a dating relationship; 87.9% reported experiencing psychological aggression in a dating relationship, and 70.2% reported sexual coercion (Manseau et al., 2008). A history of parental physical abuse, having lived outside the home prior to placement, early sexual debut, and having been pregnant were identified as risk factors for severe physical victimization (Manseau et al., 2008). Wekerle et al. (2009) surveyed a random sample of youth involved in child protective services in Ontario, Canada, and found that more than half of the youth who were dating reported some form of dating violence by midadolescence (ages 14-17). Furthermore, emotional abuse was found to predict both PTSD symptomatology and dating violence (Wekerle et al., 2009).
A more recent longitudinal study found that anger mediated the relationship between childhood maltreatment and past year dating violence perpetration for a sample of child welfare–involved youth in Ontario, Canada (Faulkner et al., 2014). Posttraumatic stress, anger, depression, and witnessing IPV were correlated with perpetration of dating violence; witnessing IPV was also correlated with a measure of dating violence and marijuana use (Faulkner et al., 2014). However, maltreatment was not found to have a direct effect on either dating violence or substance use (Faulkner et al., 2014). In these studies, IPV was measured using either a modified version of Bergman’s (1992) Questionnaire on Dating Violence (Jonson-Reid & Bivens, 1999; Jonson-Reid et al., 2007) or the Conflict in Adolescent Dating Relationships Inventory (CADRI; Wolfe, Scott, Reitzel-Jaffe et al., 2001).
While these studies provide important evidence of IPV among youth in foster care, they also possess some limitations. Jonson-Reid and Bivens (1999) and Wekerle et al. (2009) rely on cross-sectional data, and Jonson-Reid et al. (2007) utilize longitudinal data with only 3 months in between baseline and follow-up. Jonson-Reid and Bivens surveyed a small, nonrandom sample of youth (n = 106) following a presentation on dating violence offered through an independent living program. Subsequent studies utilized random samples of adolescents with open cases in the child welfare system, but these samples were also relatively small, ranging from 158 youth to 408 youth (Faulkner et al., 2014; Jonson-Reid et al., 2007; Wekerle et al., 2009).
Furthermore, none of these studies assess youth past the age of 19. Attrition is a common problem for longitudinal studies of youth experiencing instability and frequent transitions, such as youth aging out of the foster care system (Faulkner et al., 2014); however, there is a clear need for research that follows these youth over time to explore their experiences as they move into young adulthood. There is also a need for studies with sufficient sample sizes to elucidate both risk and protective factors that affect foster youth during the transition to adulthood.
This will be the first study to specifically examine risk and protective factors for youth who are preparing to emancipate and who have experienced emancipation from the foster care system. The challenges experienced by this group of youth have been widely documented in the literature (Barth, 1990; Courtney & Heuring, 2005; Geenen & Powers, 2007; Keller, Cusick, & Courtney, 2007; McMillen & Tucker, 1999). These youth are significantly more likely to struggle in a variety of critical domains, reporting greater frequency of mental illness (Raghavan & McMillen, 2008; Zlotnick, Tam, & Soman, 2012), greater frequency of alcohol and drug use (Havlicek, Garcia, & Smith, 2013; Narendorf, Fedoravicius, McMillen, McNelly, & Robinson, 2012), and greater frequency of violent and/or delinquent behavior than their peers (Ryan, Herz, Hernandez, & Marshall, 2007; Stott, 2012). These findings may contribute to the likelihood that youth who emancipate from foster care may be more likely than their peers to become engaged in IPV post emancipation. This longitudinal study, using the largest known sample of youth emancipating from foster care, explores the impact of demographic, individual, family, and foster care factors on IPV involvement for foster care alumni at age 23/24. Understanding the characteristics and experiences that place these youth at risk for IPV will allow for more effective and targeted prevention efforts.
Research Design and Methods
Sample
Data from the Midwest Evaluation of the Adult Functioning of Former Foster Youth (the Midwest Study) will be used to explore the risks and protective factors for IPV involvement for emancipated foster youth in young adulthood. The Midwest Study is a longitudinal, biannual, five-wave study of youth making the transition from foster care to independent living in Illinois, Iowa, and Wisconsin. Youth were eligible to participate in this study if they were in the foster care system at the time of their 17th birthday, if they entered the foster care system prior to their 16th birthday and if their primary reason for placement in the foster care system was not delinquency. Interviews were conducted when youth were approximately 17, 19, 21, 23, and 25 years of age. Data from the first wave (N = 732, collected in 2002/2003 when youth were 17/18 years of age), second wave (N = 603, collected in 2004/2005 when youth were 19/20 years of age), and fourth wave (N = 602, collected in 2008/2009 when youth were 23/24 years of age) will be used for analysis in this study.
The Wave 1 interview (in 2002/2003) focused on the experiences of youth while they were still in the foster care system and included questions relating to a variety of domains: physical and mental health, social support, relationships with family, education, employment, victimization, delinquency, and receipt of services. The Wave 2 interview (in 2004/2005) included many of the same domains but focused on experiences since the first interview and, for those youth who had emancipated from care at age 18, included questions about life after foster care. The Wave 4 interview similarly covered these domains, asking specifically about experiences since the prior interview (as of the third interview, none of the youth in this sample were still in care).
Measures
Dependent variable: IPV status
An eight-question version of the Conflict Tactics Scale (CTS; Straus, Hamby, Boney-McCoy, & Sugarman, 1996) was used to establish which participants were perpetrating and/or experiencing violence in their intimate partnerships at age 23/24. These questions were only asked of those participants who identified that they were married, cohabitating with an intimate partner, or involved in a dating/romantic relationship at the time of the Wave 4 interview. “Dating” or “romantic relationship” could have included dating exclusively, dating frequently, dating infrequently, or having sexual intercourse. Four of the questions addressed whether or not the participant perpetrated or threatened to perpetrate physical or sexual violence in the year prior to the interview (see Tables 1 and 2 for interview questions). The other four questions addressed whether the participants had been threatened or experienced violence at the hands of their intimate partner in the year prior to the interview. Responses were based on recall.
CTS2 Items: Victimization.
Note. CTS = Conflict Tactics Scale.
CTS2 Items: Perpetration.
Note. CTS2 = Conflict Tactics Scale.
The observed dependent variable was configured to include five categories: (a) no relationship, (b) nonviolent relationship, (c) violent relationship–victim, (d) violent relationship–perpetrator, and (e) violent relationship–bidirectional violence. Those who did not identify as being in a romantic relationship at the time of the Wave 4 interview were placed in the “no relationship” category. Those who identified as being in a relationship but did not report any of the CTS items were placed in the nonviolent relationship category. Those who identified as being in a relationship and reported one or more of the CTS victim items were placed in the violent relationship–victim category. Those who identified as being in a relationship and reported one or more of the CTS perpetration items were placed in the violent relationship–perpetrator category. Those who identified as being in a relationship and reported one or more of the CTS victim items and one or more of the CTS perpetrator items were placed in the violent relationship–bidirectional violence category.
Independent variables
Demographic variables
Both race and sex were measured at the Wave 1 interview. At that time, each participant was asked to identify as White (non-Hispanic), Hispanic, Black, Asian or Pacific Islander, American Indian or Alaskan Native, or Mixed Race. Based on the infrequent reporting of categories other than White and Black, this variable was collapsed into a two-category variable: White and non-White. Gender was also measured at the Wave 1 interview when each participant was asked to identify as male or female.
Individual-level variables
Three types of youth functioning variables were included in analyses: drug and alcohol abuse, mental health status, and involvement in delinquent acts. Abuse of alcohol and drugs was measured at Wave 1 when participants were given the Composite International Diagnostic Interview (CIDI; World Health Organization, 1998) as part of the interview. This instrument asks a variety of questions in an effort to provide diagnostic data. Diagnosis of depression or dysthymia (our depression variable) and diagnosis of PTSD (our PTSD variable) also come from the CIDI. Our question about anxiety (asked at Wave 1) was less diagnostic, asking participants whether they felt “anxious, tense, or worried about everyday problems for most of a month or more.”
Our measure of delinquency was cumulative at Wave 1 (a sum of delinquent acts perpetrated in the year prior to the interview). Those participants who reported perpetrating more delinquent acts had higher delinquency scores. We included those items that were measured at each wave of the Midwest Study: damaging property, stealing, breaking/entering, use of weapon, selling drugs, and fighting (α = .7).
Family-level variables
Maltreatment prior to foster care entry was measured at Wave 1. Participants were asked whether they recalled various forms of neglect and physical abuse perpetrated by their caretakers prior to their date of entry into the foster care system. An example neglect question is “Did you ever have a serious illness or injury or physical disability, but your caretakers ignored it or failed to obtain necessary medical or remedial treatment for it?” An example physical abuse question is “Did any of your caretakers ever hit you hard with a fist or kick or slap you really hard?” Participants were also asked whether they recalled being molested or sexually assaulted at any time prior to the Wave 1 interview. If a participant reported at least one form of neglect or physical abuse, they were considered neglected or physically maltreated prior to foster care entry. If a participant reported either form of sexual abuse, they were considered sexually maltreated prior to the Wave 1 interview. The perpetrator was known for questions measuring neglect and physical abuse (the questions asked specifically about caretakers) but was unknown for questions measuring sexual abuse.
Maltreatment while in foster care was measured at Wave 2 when youth were 19 or 20. Youth were asked to recall if they had experienced various forms of neglect, physical maltreatment, and sexual maltreatment while in the foster care system (the questions were specific to foster caregivers for neglect and physical abuse; the questions were not specific to caregivers for sexual abuse). If participants reported at least one form of neglect, physical abuse, and/or sexual abuse, they were considered neglected, physically maltreated, and/or sexually maltreated while in foster care.
Parental IPV was measured at Wave 1 when youth were 17 or 18 years old. Participants were asked whether their parents or primary caregivers had engaged in “spousal abuse” prior to their date of entry into the foster care system. If participants answered “yes,” they were considered to have a history of parental IPV. Being close to an adult biological family member was also measured at Wave 1. Participants were considered to be “very close” with a biological adult family member if they stated they were very close to either of their biological parents or any of their biological grandparents.
Foster care–level variable
At Wave 1, participants were asked to recall how many “foster homes” and/or “group homes, residential treatment centers, or child caring institutions” they lived in over the course of their stay in the foster care system. The totals were summed to create a measure of out-of-home placements.
Analytic Methods
Descriptive analyses were conducted in SPSS (Version 22) to determine the rate of IPV involvement at age 23/24 in this sample of former foster youth. Bivariate analyses were conducted in an effort to determine which independent variables (at Waves 1 and 2) were correlated with IPV involvement at Wave 4 (reported in Table 3). We then estimated a multinomial logistic regression model in SPSS (Version 23) with IPV status at age 23/24 as our dependent variable. This model included Wave 1 (age 17/18) and Wave 2 (age 19/20) predictors and used “nonviolent relationship” as the comparison category.
Bivariate Statistics by IPV Status (N = 551; Reference Category = Nonviolent Relationship).
Note. Fifty-one subjects did not answer the relationship or Conflict Tactics Scale items in a way that could identify them as falling into a specific relationship category; they were classified as missing. Those included in analyses participated in Waves 1, 2, and 4; IPV = intimate partner violence; PTSD = posttraumatic stress disorder.
Variable was measured at Wave 2.
Significance level <.05 from chi-square test when compared with those in nonviolent relationship category. **Significance level <.01 from chi-square test when compared with those in nonviolent relationship category.
Multiple imputation (MI) was used to handle missing data in the multinomial logistic regression, as it has been shown to be superior to other techniques such as listwise deletion or other traditional methods (Allison, 2002; Cox, McIntosh, Reason, & Terenzini, 2014). According to Manly and Wells (2015), “MI uses the fundamental distribution property of each measured variable and their correlations to produce reasonable estimates of statistical inferences based on the collected information” (p. 399). Because the Midwest Study is longitudinal in nature, it is vulnerable to attrition: some data were lost at each wave of data collection. In all, 732 participants were interviewed at Wave 1, 603 participants were interviewed at Wave 2, and 602 participants were interviewed at Wave 4. There also were 51 participants who did not answer the relationship or CTS items in a way that could identify them as falling into a specific relationship category; they were classified as missing. The missing data for the variables included in this study were missing at random. All variables described above were used for imputation including the dependent variable, IPV status (Graham, 2009). Default programming was employed in the SPSS MI function, and five imputations were run. Pooled estimates were used.
Results
Descriptive and Bivariate Statistics
The most commonly reported form of IPV in the Midwest Study of those in romantic relationships at age 23/24 was in the domain of victimization: Approximately 20% of participants indicated that their partner had either threatened them with violence or pushed, shoved, or thrown something at them that could hurt in the year prior to the interview. This item was also the most frequently reported in the domain of perpetration (18%). The next most frequently reported item in both domains was regarding slapping, hitting, or kicking: Nearly 20% of participants indicated that they had been slapped, hit, or kicked by an intimate partner; nearly 16% of participants indicated that they had slapped, hit, or kicked a partner.
Youth in this sample were most frequently in nonviolent relationships at age 23/24 (44.5% of the sample). In contrast, approximately 21% of youth were in a violent relationship, with bidirectional violence being most commonly reported type (11.4% of the sample). Youth in the bidirectional category reported at least one measure of IPV victimization and at least one measure of IPV perpetration.
As reported in Table 3, we ran bivariate statistics by IPV status, using chi-square and t tests to compare youth in the (a) no relationship, (b) violent relationship–victim, (c) violent relationship–perpetrator, and (d) violent relationship–bidirectional categories with youth in the (e) nonviolent relationship category. We found significant differences by category relating to demographic, youth functioning, family system, and foster care characteristics. For demographic characteristics, gender appeared to influence category membership, as males were significantly more likely to be in the violent relationship–victim category than in the reference category (p < .01). For youth functioning characteristics, mental health status, specifically depression and PTSD, appeared to predict category membership: Depressed youth were more likely to be in the bidirectionally violent relationships than nonviolent ones (p < .05), and youth with PTSD were more likely to be in nonviolent relationships than perpetrating violence (p < .05).
Regarding family system characteristics, those who experienced parental IPV (p < .01), physical abuse prior to foster care entry (p < .05), and neglect in care (p < .01) were all more likely to fall into the bidirectionally violent category than the nonviolent relationship category. However, those who experienced sexual abuse prior to foster care entry were less likely to be victims of IPV than they were to be in nonviolent relationships (p < .01). Regarding foster care, those in bidirectionally violent relationships appeared to have experienced greater placement instability than those in any other category (reporting 7.6 [6.6] foster care placements on average compared with a sample average of 5.8 [5.9] placements). They experience significantly more placement instability than those in the nonviolent relationship category (5.0 [4.8]).
Multivariate Analyses
The logistic regression results in Table 4 show that demographic, youth functioning, family system, and foster care covariates were all predictive of IPV status at age 23/24. For family systems covariates, youth who had witnessed parental IPV were significantly more likely to be in no relationship (p = .015) or a bidirectionally violent relationship (p = .018) than in nonviolent relationships at age 23/24. Youth who identified that they were very close to at least one biological adult relative at Wave 1 were also significantly more likely to be in bidirectionally violent relationships than in nonviolent relationships (p = .018) as were those who indicated that they had experienced neglect perpetrated by a foster caregiver while in the child welfare system (p = .013). Furthermore, youth with greater placement instability were more likely to be in no relationship (p = .028) or in a bidirectionally violent relationship (p = .012) than they were to be in nonviolent relationships.
Multinomial Logistic Regression: IPV Status on Predictors (N = 579).
Note. Conducted using multiple imputation. Reference category = nonviolent relationship; PA = physical abuse; SA = sexual abuse; IPV = intimate partner violence; PTSD = posttraumatic stress disorder.
Youth who experienced sexual abuse prior to the Wave 1 interview were less likely than those who did not experience sexual abuse prior to the Wave 1 interview to report that they had experienced victimization at age 23/24. This may be due to the fact that sexual abuse prior to the Wave 1 interview was most frequently reported by females, and females were significantly less likely to report that they had experienced IPV victimization than males (p = .038). None of the other parental maltreatment variables appeared to be predictive of IPV status at age 23/24. This could be due to the fact that parental maltreatment is commonly reported by nearly all of the youth in this sample (the vast majority were in foster care for reasons having to do with parental maltreatment). There is minimal variance among study participants.
With regard to youth functioning, those youth who reported experiencing recurrent symptoms of anxiety at Wave 1 were more likely to perpetrate IPV at age 23/24 than those who did not report experiencing anxiety (p = .012). However, those with a PTSD diagnosis at Wave 1 were significantly less likely to perpetrate IPV at age 23/24 than those without a PTSD diagnosis (p = .020).
Discussion
Four of these findings are particularly important for practitioners and administrators working with youth who are preparing to emancipate from the foster care system. First of all, we found that IPV in the emancipated foster care population is relatively common. More than 20% of the youth in our sample reported perpetrating or experiencing IPV in the year before the Wave 4 interview. Considering that youth who are involved in violent romantic relationships are more likely than their peers to experience a range of troublesome physical and psychiatric problems (including suicidal ideation and serious injury), these findings are grave. There is an immediate and imperative need to teach youth in the foster care system about healthy/unhealthy relationships and engage high-risk foster youth in programs intended to prevent IPV involvement, especially if we wish to reduce the likelihood that these foster youth will subsequently model violent behaviors in front of their children.
Similar to other studies (Foshee et al., 2010; Hickman et al., 2004; Shook, Gerrity, Jurich, & Segrist, 2000), we found young women were commonly perpetrating IPV against their male partners. It is more socially acceptable for a young woman to report slapping her male partner than it is for a young man to report slapping his female partner (social desirability bias could be impacting these findings). However, Audio-enhanced, computer-assisted self-interviewing (Audio-CASI) technology was used for the CTS questions in an effort to neutralize social desirability bias. Also similar to a number of other studies (Bookwala et al., 1992; Chiodo, et al., 2012; Gray & Foshee, 1997; Whitaker et al., 2007), we found that bidirectional IPV was frequently reported by participants: the most common type of IPV in this sample was bidirectional violence. This indicates that traditional models purporting male perpetration and female victimization may not be exclusively appropriate for young adults in the foster care population. The young women in our sample appear to be both engaging in violence and sustaining violence in their romantic relationships as opposed to being passive victims. These women could be instigating episodes of violence and/or engaging in self-defense after being struck by a partner. Despite this, the male perpetration of physical violence against female partners is often accompanied by more serious physical consequences (Capaldi, Kim, & Shortt, 2004).
Second, it comes as little surprise that IPV in the home of origin is predictive of IPV status in young adulthood. The youth in our sample who reported parental IPV in their homes of origin were significantly more likely to be in a bidirectionally violent relationship than in a nonviolent relationship. As previously mentioned, social learning theory (Bandura, 1977) dictates that youth who witness influential others engaging in certain behaviors may be more likely to emulate the behaviors they witness. In this case, we find evidence to support Widom’s “cycle of violence” hypothesis (Widom, 1989, 1996): youth who witness their parents perpetrate IPV are more likely to perpetrate and experience IPV in young adulthood than youth who do not witness their parent(s) perpetrating IPV. This may be because the youth in our sample who witnessed parental IPV have normalized the use of physical violence in moments of interpersonal conflict or tension. They may have evidence to support the idea that using violence (or threatening to use violence) may be an effective strategy when an intimate partner is engaging in activities or behaviors they find problematic.
Less clear is the finding that youth who identify as being “very close” with biological adult relatives are significantly more likely than their peers to be involved in bidirectionally violent relationships. It is theoretically possible that youth who identify as being close with adult relatives may have observed those relatives engage in IPV, making them more likely to employ these tactics themselves. However, when we used an interaction term to test this hypothesis, we found that it did not significantly predict IPV in young adulthood. Future studies, particularly those evaluating the social support networks of emancipating foster youth, may be able to investigate this relationship.
Third, the mental health predictors are particularly interesting in light of prior cross-sectional findings. Previous studies found PTSD to be predictive of IPV involvement (Hahn et al., 2015; Riggs et al., 2000; Taft et al., 2008; Wolfe et al., 2004). In contrast, we found that PTSD at 17/18 appears to be protective—those who had a PTSD diagnosis at age 17/18 were significantly less likely than their peers to report perpetrating IPV at age 23/24. Perhaps after 5 years passed, these youth chose to avoid situations that would trigger their own violent behavior and/or their partner’s violent behavior (or, by proxy, the trauma symptomatology that may be triggered by these acts of violence). Youth who are actively in the throes of trauma symptomatology (like the youth in our sample at age 17/18) may be less able to regulate emotion and engage in a thoughtful decision-making process, making them more likely to find themselves involved in violent partnerships in adolescence. They may have been more likely to be involved in violent romantic partnerships had IPV been measured at Wave 1.
We also find that youth who report experiencing anxiety at 17/18 were more likely to perpetrate violence at 23/24. Anxious youth may spend more time worrying about their relationships, particularly worrying that their partner may leave or that a more favorable partner will replace them. These youth may employ violent strategies in an effort to exert control over their partnerships. In a study on adult attachment, C. J. Allison, Bartholomew, Mayseless, and Dutton (2008) found that those adults who were anxiously attached engaged in violent behavior in an effort to gain their partner’s attention and, periodically, to solicit physical closeness.
Last, placement instability and neglect in care appear to predict IPV status at Wave 4. This finding reflects the importance of consistent, attentive caregiving for the prevention of IPV in young adulthood. Youth who move from placement to placement or youth who experience foster caregiver neglect may feel as though the relationship with their romantic partner is the most consistent and dependable relationship in their lives, making them more tolerant of violent behavior and more likely to remain in romantic relationships despite dangerous circumstances. Foster caregivers who are committed and attentive to the youth they house may be able to model prosocial methods of dealing with frustration, anger, and/or anxiety in romantic partnerships. They also may be able to recognize the signs of IPV (controlling behaviors, diminished self-esteem and depression, bruising and other injuries) in the youth they are housing, engaging them in discussion about their romantic partnerships in an effort to assess for safety and/or linking them with services in an effort to protect them from future involvement. Foster care administrators would be wise to train foster caregivers how to recognize signs of IPV and educate them about available intervention resources. They would also be wise to consistently assess for the quality of these relationships, both from the perspective of the foster parent and the perspective of the child.
Limitations
These findings should be interpreted in the context of study limitations. First, some of the measures used in the Midwest Study were not as specific or nuanced as would have been ideal. An abbreviated version of the CTS (CTS2; Straus et al., 1996) was chosen to accommodate time and space limitations. It was also chosen because it matched the version used in the National Longitudinal Study of Adolescent Health (Add Health; Harris et al., 2003), a nationally representative sample of adolescents with whom the youth in the Midwest Study were compared. In this version of the CTS2, the first item includes both threatening physical violence and perpetrating physical violence. These are two substantively different things. A positive answer to CTS2–1 could indicate a wide variety of activities varying in intensity and danger. It would also have been ideal to include a measure of psychological IPV (humiliation, stalking, cyber threats, and/or repeated put-downs) in the CTS2, as psychological violence is common in adolescence/young adulthood and is linked with a variety of problematic outcomes (Exner-Cortens, Eckenrode, & Rothman, 2013).
Furthermore, participants in the Midwest Study were only asked whether their primary caregiver engaged in “spouse abuse.” They were not asked whether they witnessed this abuse, what they witnessed, how often they witnessed the abuse, or who it involved. Along these same lines, the Midwest Study does not ask whether participants experienced any form of psychological abuse prior to foster care entry, a known risk factor for IPV in adolescence and young adulthood for child welfare system–involved youth (Wekerle et al., 2009). Last, we chose to categorize race as White and non-White in light of the fact that most of our participants identified as White or African American. It would have been ideal to have a more nuanced measure of race so we could examine whether various cultural backgrounds influenced the use or acceptance of violent behavior in intimate partnerships.
This study is also limited by the fact it relies on participant recall. Participants are asked when they are 17 years old (at Wave 1) about the various types of maltreatment they experienced prior to entering foster care. Many of these participants entered foster care in adolescence, making the remembering of these events more likely (participants were required to have entered foster care before their 16th birthday to participate in this study). Others, by contrast, entered foster care much earlier in life: the average age of entry for this sample was 10.7 years of age. Despite the fact that this is higher than the national average of 6.4 years of age (Fiscal Year [FY] 2012, U.S. Department of Health and Human Services, 2013), these participants may be less likely to remember the details of these episodes. Similarly, the participants who had been in foster care for extended periods of time may also be less likely to remember episodes of maltreatment that took place while they were in the foster care system.
Finally, while the Midwest Study is the largest study of emancipating foster youth, statistical power for this study was limited by the decision to divide IPV status into five categories. Studies with larger samples may have been able to identify relationships that this study failed to find. Furthermore, other studies with more variance among participants relating to parental maltreatment may have also been able to identify relationships that this study failed to find.
Despite these limitations, these findings are quite important for service providers and researchers who are interested in emancipating foster youth. Knowing that bidirectional violence is the most common form of IPV in this sample is particularly important for those looking to intervene, as is the fact that young women are more likely to find themselves in this category than young men. It comes as little surprise that these participants are more likely than their peers to report witnessing parental IPV, as exposure to IPV is one of the most consistent predictors of IPV perpetration and victimization found in the literature (Kinsfogel & Grych, 2004; O’Donnell et al., 2006; O’Keefe, 1997; C. A. Smith et al., 2015). These youth would be well served by targeted prevention programs.
Equally important is the finding that placement instability and neglect perpetrated in foster care both place youth at risk of experiencing bidirectional IPV. This is an important reminder that foster care provider relationships are critically important, both in their consistency and their quality. Youth who feel valued and connected to a foster caregiver may feel as though they have greater capacity to employ alternative mechanisms in response to frustration and relationship anxiety. They may also feel freer to leave violent partnerships, knowing that there is a supportive adult who is able to provide consistent safety and security. Future research could explore how positive, long-term relationships with foster caregivers can protect against IPV and how training foster caregivers could enable prevention and intervention for high-risk youth.
Footnotes
Acknowledgements
This research was made possible by the state child welfare agencies in Illinois, Iowa, and Wisconsin; the Bill & Melinda Gates Foundation; the William T. Grant Foundation; the Stuart Foundation; the Walter S. Johnson Foundation; and Casey Family Programs.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
