Child Abuse & Neglect 41 (2015) 170–181

Contents lists available at ScienceDirect

Child Abuse & Neglect

Do family order and neighbor intervention against intimate partner violence protect children from abuse? Findings from Kathmandu夽 Clifton R. Emery a,∗ , Sirjana Thapa b , Mi Hyang Do b , Ko Ling Chan c a b c

School of Social Welfare, Yonsei University, Seoul, Republic of Korea Namseoul University, Seoul, Republic of Korea Faculty of Social Sciences, University of Hong Kong, China

a r t i c l e

i n f o

Article history: Received 8 July 2014 Received in revised form 11 September 2014 Accepted 3 October 2014 Available online 22 October 2014 Keywords: Physical child abuse Intimate partner violence Informal social control Nepal ISC IPV

a b s t r a c t Drawing on previous research on intimate partner violence, child maltreatment, and informal social control, we hypothesized relationships between child abuse severity and (1) protective informal social control of intimate partner violence (ISC IPV) by neighbors, (2) intimate terrorism, (3) family order, and (4) the power of mothers in intimate relationships. In what we believe may be a first study of physical child abuse by parents in Nepal, we used a three stage cluster approach to draw a random sample of 300 families in Kathmandu. Random effects regression models were used to test the study hypotheses. The analyses found support for hypotheses one and two, but with an important caveat. Although observed (actual) protective ISC IPV had the hypothesized negative association with child abuse severity, in one of our models perceived protective ISC IPV was positively associated with child abuse severity. The models clarify that the overall direction of protective ISC IPV appears to be negative (protective), but the positive finding is important to consider for both research and practice. A significant relationship between family order and child abuse severity was found, but the direction was negative rather than positive as in hypothesis three. Implications for neighborhood research and typological research on IPV and child maltreatment are discussed. © 2014 Elsevier Ltd. All rights reserved.

Introduction The World Health Organization estimates that 25–50% of children are physically abused annually around the world (WHO, 2010). In developed countries, police or child protection authorities may intervene before most abuse becomes injurious or fatal. However, in developing countries like Nepal, such formal social control of violence in the home may be largely absent. In such places, the informal actions of bystanders like neighbors may sometimes make the difference between life and serious injury or death when severe physical abuse of a child occurs. Actions taken by ordinary citizens to protect children from abuse are a form of informal social control (Hirschi, 1969/2002; Sampson, Raudenbush, & Earls, 1997). The idea of informal social control derives from the social disorganization tradition in criminology (Sampson, Morenoff, & Gannon-Rowley, 2002) and is constituted in actions undertaken by ordinary people (not employees of the state) to achieve public order and prevent

夽 Measure development was facilitated by a grant from the H.F. Guggenheim Foundation. ∗ Corresponding author. http://dx.doi.org/10.1016/j.chiabu.2014.10.001 0145-2134/© 2014 Elsevier Ltd. All rights reserved.

C.R. Emery et al. / Child Abuse & Neglect 41 (2015) 170–181

171

crime (Sampson et al., 1997). Social disorganization theory explains the difference between criminals and non-criminals as stemming from weaker controls, rather than from differences in motivation to commit crime (Kornhauser, 1978). If social disorganization theory holds, the presence of informal social control should be associated with less child abuse. Past research on informal social control and child maltreatment has shown mixed results, with some research showing protective effects (cf. Guterman, Lee, Taylor, & Rathouz, 2009) while other research shows no relationship with physical abuse (Molnar, Buka, Brennan, Holton, & Earls, 2003), and still other research shows relationships with other forms of maltreatment, but not physical abuse (Yonas et al., 2010). However, recent research using a new measure of informal social control of child maltreatment (ISC CM) has shown negative associations between informal social control and very severe child abuse and abuse related behavior problems in Hanoi (Emery, Trung, & Wu, 2013), and child abuse injury in Seoul (Emery et al., in press-a). Yet, not all physical child abuse begins with a parent’s disciplinary action against a child. The link between physical child abuse and intimate partner violence (IPV) is well documented, varied in etiology, and strong (Appel & Holden, 1998; Emery, Kim, Song, & Song, 2013; Herrenkohl, Sousa, Tajima, Herrenkohl, & Moylan, 2008; Taylor, Guterman, Lee, & Rathouz, 2009). Hence, controlling IPV may be an important component in preventing the physical abuse of children. This paper extends the work on informal social control to examine whether intervention by neighbors against IPV protects children from abuse in Kathmandu, Nepal. Child Abuse and Intimate Partner Violence in Nepal We were somewhat stunned to be unable to discover any prevalence estimates for the rate of physical abuse of children by parents in Nepal. Because a Google Scholar search for “Nepal” and “Child Abuse” returns more than 4000 results on broader child welfare problems in Nepal, we adopted a more targeted search. A search for any publication mentioning “Nepal” and “physical child abuse” produced 34 results, none of which were quantitatively robust prevalence studies of physical abuse by parents in Nepal. Neither did specific searches for the word “Nepal” in Child Abuse & Neglect, Child Maltreatment, Child Welfare, or Children and Youth Services Review produce any relevant results. We also searched the Journal of Interpersonal Violence, Journal of Family Violence, Journal of Family Process, Pediatrics, and the Journal of Marriage and Family for “Nepal” and “child abuse. “There were no relevant results. The searches were not time-limited. In a separate search, we contacted all of the 9 larger NGOs operating in Kathmandu that work on child related issues to ask whether a study of physical violence of parents against children had been carried out previously but not published in a peer reviewed journal. None of the NGOs knew of such a study. We then queried the Nepal Health Research Council (NHRC), which registers research in the country. The head of the NHRC research team replied that no such research as yet. Ms. Ghimire (2014) added: “matters inside the household are as yet totally invisible; parents beat their children as a ‘punishment’ and it is normal to beat [children] on the face, bottom with the hands and using a stick or broom. “In geographically and culturally adjacent India, physical abuse of children by parents is high (estimated at over 40% among middle class professionals) (Segal, 1995). Intimate partner violence (IPV) is also a serious problem in Nepal. Paudel (2007) found that 35% of Nepali women experienced gender violence in their homes, while Dhakal (2008) found that 80% of violence against women is accounted for by violence perpetrated by family members. In our study of domestic violence in Kathmandu, Emery et al. (in press-b) found a somewhat higher rate; 54% of our sample reported some violence by the husband in the last year. In spite of the high rates of IPV, only 1100 cases are recorded by the police every year (Dhakal, 2008). This lack of formal social control suggests that whether and how neighbors respond to IPV may be critical in protecting women and children from severe abuse in Nepal. Moreover, in our study of domestic violence, we found that randomly selected participants were very willing to participate in the study; the response rate was 96% (Emery et al., in press-b). A high response rate is of course important for data quality, but the underlying cause of the high response rate is also important. Our qualitative experiences collecting the data, as well as those of our interviewers, suggest that there is a high level of social trust in Kathmandu, despite the 10 year civil war that ended in 2006. To us, high levels of trust suggest vitality in informal social networks, and more specifically the idea that informal social control by neighbors may be more acceptable, and hence efficacious. Nepal is 81% Hindu, 9% Buddhist, and 4.4% Muslim (Central Bureau of Statistics, 2012). The strong influence of the Hindu religion on Nepali culture means that Nepal is a caste society, like India. Although it is formally illegal, private discrimination against lower castes is still very common, and caste plays a very large role in determining education, career prospects, and other indicators of socio-economic status (Silwal, 2006). The discrimination faced by lower castes may make it difficult for them to access already scarce formal services. For this group, informal network support and assistance in controlling violence may be even more vital than for other groups in Nepal. Informal Social Control by Neighbors Research on the actions of ordinary people undertaken to achieve public order and prevent crime (informal social control) has deep roots in the social disorganization tradition of the sociology of crime (Sampson et al., 1997). Shaw and McKay (1969) were among the first researchers who tried to achieve an empirical understanding of why crime tends to cluster in some neighborhoods. Hirschi (1969/2002) went beyond population demographics to argue that crime occurs because weaker social bonds left potential criminals free to perpetrate. Although Garbarino (1976) quickly recognized and documented the important relationship between community and child maltreatment, it was not until Sampson et al. (1997) developed the concept of collective efficacy in their study of neighborhood conditions and homicide that a measure of neighborhood

172

C.R. Emery et al. / Child Abuse & Neglect 41 (2015) 170–181

characteristics became available to mainstream criminological survey research. For Sampson et al. (1997), collective efficacy was one part neighborhood solidarity, and one part neighborhood informal social control. They argued that strong bonds between neighbors facilitated informal social control, which then deterred crime. Collective efficacy was strongly (and negatively) correlated with Chicago homicide rates in their research, even when many other standard neighborhood demographic characteristics were held constant (ibid). Some researchers have gone so far as to claim that informal social control is more effective at controlling crime than police and other official state responses (Schwartz & Dekeseredy, 2008). Researchers studying both child maltreatment and IPV quickly embraced the promise of the collective efficacy measure. Yet in both fields, research findings were mixed. Using the collective efficacy informal social control measure, Guterman et al. (2009) found a significant relationship with physical abuse. Melton’s (2014) Strong Communities intervention appeared to increase self-reported positive interactions with children and decrease neglect, but observed changes in parents’ collective efficacy and use of physical punishment were modest. In Melton’s (2014) case, this may have been a problem of statistical power. Yonas et al. (2010) found collective efficacy to moderate the relationship between neglect and externalizing behavior problems, but did not find a moderating effect for the relationship between physical abuse and externalizing. Molnar et al. (2003) found no relationship between collective efficacy and physical abuse of children. Findings for IPV were similarly mixed. Browning (2002) found a strong negative relationship between collective efficacy and IPV. On the other hand, Dekeseredy, Schwartz, Alvi, and Tomaszewski (2003) found no relationship in their national study of Canada. Neither did Emery, Jolley, and Wu (2011) find any relationship between the collective efficacy informal social control measure and IPV desistance. On the surface, such inconsistent findings call into question the utility of the theory of informal social control with respect to IPV and child maltreatment. Emery, Trung, et al. (2013) argue that the problem is not the theory but the measure. The crux of the problem, they claim, is that the informal social control measure of collective efficacy measures informal social control of crime and disorder on the street. Person A might regard domestic violence and child abuse as crime, while person B might regard domestic violence and child abuse as a form of informal social control. For people who regard violence by fathers as necessary to deter breaking of family rules by children and wives (person B), violence is pro-social and helps to uphold the rules. In short, domestic violence and child abuse are conceived of as forms of social control. If like-minded people tend to cluster in neighborhoods, A type neighborhoods will show a negative relationship between informal social control and family violence, while B type neighborhoods will show a positive relationship between informal social control and family violence. Rather than engaging in an attempt to measure A and B type neighborhoods and then test for an interaction between informal social control and type in a predictive model for family violence, Emery, Trung, et al. (2013) simply tried to measure informal social control of child maltreatment (ISC CM) directly. They measured protective ISC CM, which tries to protect the child from harm, and punitive ISC CM, which focuses on punishing the perpetrator. They found that protective ISC CM was associated with less very severe physical abuse of children, and with fewer externalizing behavior problems when very severe abuse did occur. Such protective effects must be conservative estimates, since more severe violence must be more likely to elicit a response from neighbors than less severe violence. This trend biases associations toward positive, rather than negative correlations. No significant relationships were found for punitive ISC CM. In a parallel development and using similar logic, Emery, Wu, and Raghavan (2014) developed a measure of informal social control of intimate partner violence (ISC IPV) and tested it on a population sample of Beijing. The approach for measuring ISC IPV was somewhat different than for ISC CM. Although the questions were similar, Emery, Wu, and Raghavan (2014) coded ISC IPV as either observed (neighbors actually intervened) or perceived (respondent though neighbors would intervene). To cope with potential bias from more severe violence being more likely to elicit neighbor responses, Emery, Wu, and Raghavan (2014) restricted their analysis sample to those who reported some form of IPV in the past year, and controlled for IPV severity. They found that when neighbors did in fact intervene (observed ISC IPV), injuries from IPV were significantly lower. However, when respondents only perceived the likelihood of neighbor intervention, but it did not actually occur, IPV injuries were significantly higher. They speculated this might occur because victims who counted on a non-forthcoming neighbor response might be less likely to take action to protect themselves.

Informal Social Control of Intimate Partner Violence, and Physical Abuse of Children There is a robust correlation between physical abuse of children and IPV (cf. Appel & Holden, 1998; Emery, Kim, et al., 2013; Herrenkohl et al., 2008; Taylor et al., 2009). Although there are other potential explanations for this relationship (see Emery, Kim, et al., 2013), Knickerbocker, Heyman, Slep, Jouriles, and McDonald (2007) divide the vast majority of explanations into (1) common cause explanations, which argue that a single factor contributes to both kinds of violence, and (2) the direct contribution or spillover hypothesis. The spillover hypothesis suggests that one form of violence is the cause of the other. For example, the stress and mental health consequences of victimization may increase mothers’ risk of abusing their children, IPV may result in declines in parenting quality, or finally, children may become inadvertent victims as they seek comfort from parents during an IPV incident or attempt to stop a violent episode (Knickerbocker et al., 2007). Attempts by neighbors to regulate or reduce IPV via informal social control may have the added benefit of helping to protect children by any or all of these mechanisms. Based on the research of Emery, Trung, et al. (2013) on ISC CM, and the research of Emery, Wu, and Raghavan (2014) on ISC IPV, we hypothesized that when ISC IPV is both protective, rather than punitive, and observed, rather than just perceived likely, it would have a negative association with physical child abuse. Formally:

C.R. Emery et al. / Child Abuse & Neglect 41 (2015) 170–181

Hypothesis 1.

173

Protective, observed ISC IPV by neighbors will be negatively associated with physical child abuse severity.

Family Order, Intimate Terrorism, Power, and Physical Abuse of Children Not all forms of IPV are alike, and the differences in types are likely to have important implications for physical abuse of children. Johnson (2008) argues that there are four types of intimate partner violence, the most well-known of which is intimate, or patriarchal terrorism (Johnson, 1995). Intimate terrorism is a type of domestic violence in which one partner uses violence in order to try to control the other (Johnson, 2008). For Johnson (2008), control motivated violence is likely to be more severe, injurious, and potentially lethal than violence not motivated by control. Following Holtzworth-Munroe, Meehan, Herron, Rehman, & Stuart (2000), Johnson (2008, p. 31) locates the explanation for control motivated violence in insecure childhood attachment relationships. Although there is strong theoretical reason to suppose intimate terrorism, over and above IPV, may be associated with more physical abuse of children, substantial empirical evidence is still lacking (Kelly & Johnson, 2008). IPV and intimate terrorism can be perpetrated by women as well as men. However, intimate terrorism was only measured for partner, not self in our study; as a result the small percentage of male respondents precludes an analysis of female perpetrators. Hence, we also hypothesize: Hypothesis 2. Controlling for husband’s IPV severity, intimate terrorism will be positively associated with physical child abuse severity. Emery (2011) argued that although Johnson’s typology constituted an important development in the field, it was inadequate on several counts, arguing that intimate partner violence had to be understood in terms of relationship order, power, and norms. Among Emery’s (2011) types of IPV, the most severe were thought to be those lowest and highest in order. Emery (2011) draws on Stark’s (2007) sweatshirt case to argue that IPV characterized by a plethora of norms in areas life not normally rule-governed (p. 229) is dangerous and likely to fall into Emery’s totalitarian dictatorship type. The “sweatshirt case” (Stark, 2007) discussed the ways in which an elaborate and hidden set of rules is sometimes established to control every aspect of a domestic violence victim’s life. Cheryl excelled at softball and would play well. However, after playing well for some time, her boyfriend would come onto the field saying that she was cold and offering her a sweatshirt. Cheryl’s performance would then disintegrate drastically and suddenly. Giving Cheryl the sweatshirt was a code, indicating that she had violated rules about making her boyfriend jealous and would require the sweatshirt to hide bruises from an impending punitive beating. Emery (2011) argues that the amount of rules governing behavior can be understood on a continuum of order which determines the form of domestic violence. Hence, domestic violence in the context of almost no rules governing couple behavior is conceptualized as anarchic domestic violence, but domestic violence in the context of an excessive number of rules governing behavior, concentrated power in the hands of the perpetrator, and consistent enforcement of the rules with violence is conceptualized as totalitarian domestic violence (Emery, 2011). In their study of 250 Mongolian families, Emery, Wu, and Tsolmon (2014) tested for a curvilinear relationship between family order and IPV injury. Instead they found a linear relationship. Families with high levels of order in their daily routine were at highest risk for IPV injury. This suggests that those in Emery’s (2011) totalitarian type of IPV may be at higher risk for injury than victims of other types. If this is the case, children may also be at higher risk. Hence, we also hypothesize: Hypothesis 3.

Higher levels of family order will be positively associated with child abuse severity.

In his critique of Johnson (2008), Emery (2011) argued that power, rather than control-motive, was important in distinguishing types of IPV. In our study of IPV in Kathmandu, we found that controlling for control attempts by the husband (what Johnson uses to measure intimate terrorism and we term attempted power), a wife’s self reported power in the relationship was associated with lower odds of husband violence, and was similarly associated with less husband violence severity and fewer wife IPV injuries (Emery et al., in press-b). In keeping with these findings, we predict: Hypothesis 4. severity.

Wives’ self-reported power in the relationship will be negatively associated with physical child abuse

The model also controls for the severity of husband’s IPV, the social desirability of reporting IPV, whether or not the family owns a car, household income, and the mother’s age, education, caste, and number of years in the current relationship because these variables may confound the relationships between ISC IPV, intimate terrorism, order, power, and child abuse. Necessarily, we also controlled for the original measure against which ISC IPV is compared, collective efficacy, and the age and sex of the child. Method Data The Kathmandu Families and Neighborhoods Study (KFNS) is a representative random probability cluster sample of married or partnered women in 20 Kathmandu wards (see Emery et al., in press-b). Wards were selected using probability proportional to size (PPS) sampling; households were selected using simple random sampling of ward lists. The KFNS was designed to collect data on domestic violence and community characteristics; power analysis suggested a sample size of 300

174

C.R. Emery et al. / Child Abuse & Neglect 41 (2015) 170–181

was adequate. No information was available from data on the ISC IPV or ISC CM measures when data collection commenced. For this reason we used the odds ratio from Browning’s (2002) study of collective efficacy and IPV. The odds ratio of .23 was transformed to a standardized effect size by taking the log and dividing by 1.81 (Chinn, 2000) yielding delta = .49. Using Browning’s (2002) intra-class correlation coefficient (.026) the Optimal Design program (Spybrook, Raudenbusch, Liu, Congdon, & Martinez, 2008) indicated we would have 80% power with 150 observations in 30 neighborhoods with ˛ = .05. Doubling the number of observations seemed conservative. Over a two day period, interviewers were strictly trained to protect participant privacy both during and after the interviews. Participants were told that participation was voluntary and confidential, and were provided with a small gift upon interview completion. Female partners could not be interviewed in every case, because they were not always at home even after repeated visits. When this was the case husbands, if available, were interviewed. A total of 11 people refused to be interviewed or were unavailable. The study hence had a response rate of 96%. When husband respondents and families with no minor children were removed 223 families remained. Although somewhat smaller, the analyzed sample characteristics were very similar to the sample in Emery et al. (in press-b). Among the mothers in the sample analyzed, 27% had no education, 24% had completed primary education, 35% had completed middle or high school, and 14% had any post high-school education. The average years of education for mothers in the sample was 7.4. This was quite close to the national average for females estimated by the CIA (2003 estimate). Among mothers in the analyzed sample, 32% had a monthly household income between 15,000 and 20,000 Nepalese rupees per month ($172–$229 U.S./month), which was also the median. This amount seems like a reasonable estimate, since the CIA estimates annual per capita income for Nepal at $1300 USD per annum (2012), and household incomes should be higher because data are from the capital. On average, the mothers had been married for 11 years. The average age was 33. Thirtyseven percent of the mothers were in ‘lower’ castes, 25% were ‘middle,’ and 38% were ‘high.’ Thirty-nine percent reported the family owned a car. The average household size was 4.8; the mean number of children in each family was 1.9. The children in the sample were 47% female and 53% male, and the average age was 9.2 years. Measures Physical Child Abuse Severity. The measure of child abuse severity comes from a modified version of the severe items from the Conflict Tactics Scale for parent child relationships (CTSPC) (Straus & Hamby, 1997). Straus and Hamby (1997) report good evidence of construct validity for the CTSPC, and that the scale does not appear to be confounded with social desirability response sets. The 12 month frequency of the items (1) pushed, grabbed or shoved him/her, (2) hit him/her on some other part of the body besides the bottom with something like a belt, hairbrush, a stick or some other hard object, (3) kicked, bit, or hit him/her with a fist, (4) burned or scalded him/her, and (5) threw or knocked him/her down were summed to create the scale (˛ = .70). Possible responses were none, once, twice, 3–5 times, 6–10 times, 11–20 times, and more than twenty times in the last year. The scale originally also contained the severe violence item “beat him/her up” but the fact that the item had been mistranslated was unfortunately not discovered until after the fact (back-translation did not catch this because the misunderstanding about child discipline persisted into English). Husband IPV Severity. The measure of husband’s IPV is a modified version of the Straus and Douglas (2004) Conflict Tactics Scale Short Form (CTS2S). The CTS2S has been found to have good concurrent validity with the longer CTS2 (Straus & Douglas, 2004). It includes six physical violence items: (1) slapped, (2) pushed, grabbed, or shoved, (3) hit with object, (4) punched, kicked or bit, (5) beat-up, and (6) used or threatened with a knife or gun and two injury from physical violence items (had a sprain, bruise, small cut or felt pain the next day because of a fight with the partner and had to see a doctor (MD) because of a fight with partner). Possible responses were none, once, twice, 3–5 times, 6–10 times, 11–20 times, or more than 20 times in the past year. Logistic regression coefficients from husband IPV injury regressed on number of each type of violent act were used to weight the violent acts prior to combination in a scale. Thus, a one unit increase in the husband IPV severity scale is associated with the same increase in the log-odds of injury across the entire CTS scale. We believe that this is an improvement in measurements of violence severity, because it does not equate apples to oranges (e.g. slaps to punches, or punches to being beaten up). Unfortunately, this was not possible for the child abuse severity scale because injury to the child was not measured for this data. Reliability for the husband IPV severity scale was acceptable (Cronbach’s ˛ = .81). Validity data for the CTS scale is reported in Straus and Douglas (2004). Protective ISC IPV. Protective ISC IPV was measured with responses to the question: “If my neighbors witnessed my spouse physically hurting me, my neighbors might” (1) get in between my spouse and me, (2) try to calm my spouse down by talking, and (3) criticize my spouse. Possible responses were (1) my neighbors would never do this, (2) might do this, (3) would probably do this, (4) would definitely do this, or (5) actually did this (˛ = 77). Endorsing a 5 implies violence occurred and neighbors actually intervened. To create the observed protective ISC IPV scale, the three items were coded as new variables wherein 5’s were recoded as ones and all of values as zeros; then the three items were summed. To create the perceived protective ISC IPV scale, 5’s were recoded as 4’s and the items were summed. Emery, Trung, et al. (2013) reported findings suggesting good construct validity for the ISC CM scale. Factor analysis of the ISC IPV observed scale is inappropriate

C.R. Emery et al. / Child Abuse & Neglect 41 (2015) 170–181

175

because the items are dichotomous. For the three items on the perceived protective ISC IPV scale, one factor was retained and all loadings were over .6 (the loadings were .78, .63, and .73 respectively), suggesting reasonable construct validity. Intimate Terrorism. We followed Johnson’s (2008) recommendation for measurement of IT on surveys. We used the Johnson and Leone (2005) measure of intimate terrorism implemented on National Violence Against Women Survey. These items were your husband: (1) tries to limit your contact with family and friends, (2) is jealous or possessive, (3) insists on knowing who you are with at all times, (4) puts you down in front of others, (5) makes you feel inadequate, (6) shouts or swears at you (7) prevents you from knowing about or having access to the family income. Possible responses were never, rarely, sometimes, often, and always (Cronbach’s ˛ = .77). Following Johnson (2008), we defined a respondent as reporting intimate terrorism if the sum of these items was more than two standard deviations above its mean for the sample. Johnson and Leone (2005) found good evidence of construct validity for this scale. Family Order. Family order was measured using the same items used by Emery, Wu, and Tsolmon (2014). The eight items were measured as agreement with the following statements about their families (1) we eat dinner together, (2) we eat dinner at the same time every night, (3) our house/apartment is neat and orderly, (4) our house/apartment is clean, (5) I know the daily schedules of everyone in our house, (6) I know what household chores it is my job to do, (7) I know what household chores other people in my family are supposed to do, and (8) when one person in the family is sick or really busy, other family members step in to do that person’s chores. Possible responses were never, rarely, sometimes, and often (˛ = .68). Emery, Wu, and Tsolmon (2014) report evidence of reasonably good construct validity for this scale. Power. Power in the relationship was measured by asking the women, subjectively, “on a scale of 0–100 where 0 means you have no power, 50 means you share power equally with your spouse, and 100 means you have total power: how much power do you have in your relationship with your spouse? “This approach is subject to response bias, but may be less subject to other forms of bias. For example, mothers in the sample can be presumed to know whether they really have absolute control over some decisions, or whether that decision power is simply delegated for convenience. We standardized this measure of power because even when they are significant, measures that increment on a percentile basis tend to produce microscopic regression coefficients. Social Desirability of Reporting Husband’s IPV. Social desirability bias was controlled with endorsement the following 6 items: If my spouse ever hit me, I would try to keep it secret from (1) my friends, (2) my family, (3) my neighbors, (4) my co-workers, (5) my boss, (6) everyone. Possible responses were very likely, likely, unlikely, and very unlikely. Reliability was excellent (Cronbach’s ˛ = .92). A screeplot of factors from factor analysis reveals an elbow after the first component, suggesting that a one factor solution is adequate. Loadings on the first factor were all above .7. Both of these findings suggest reasonable construct validity for the scale. Analytic Issues Analyses were conducted in Stata version 11. Item non-response caused the loss of an additional 41 cases, leaving the sample analyzed in the model at 182. Most of the missing cases occurred because of no response to the self reported power item or the respondent’s caste (30 cases). Removing these variables from the analyses did not substantively change the results. Random effects regression models were used to handle clustering within wards because a Hausman test indicated that there was no difference between ward fixed and random effects models (2 = 14, df = 17, p = .73). Random effects models are efficient and appropriate in this instance (Hsiao, 2003; Johnston & DiNardo, 1997). We elected to model child abuse as a continuous function (severity and logged severity) rather than as a dichotomous (0 or 1) function because all else held constant, statistical tests of continuous variables are more sensitive than statistical tests of dichotomous variables. Model diagnostics were run using an OLS regression model in order to make use of the full range of diagnostics available in Stata. Pregibon’s link test revealed significant non-linearity in the model (hat2 = .03, p < .01). A log transform of the child abuse severity variable eliminated this problem. Although using the log-transformed version of child abuse severity did not change the sign or significance of any of the independent variables of interest, results for several control variables changed. For this reason, we show the results for both the untransformed scale (for interpretability) and the log transformed scale (for accuracy). Remaining diagnostics were run for the log-transformed model. Because the largest Variance Inflation Factor (VIF) was 2.82 (years married) we judged multicollinearity was not a threat to our conclusions. We judged outliers as studentized residuals with an absolute value larger than 3.5 (which corresponds to p < .05 significance when the number of tests is Bonferroni corrected). None of the residuals had absolute values larger than 3, so we judged there were no significant outliers. Despite the relatively low VIFs, we were concerned that the significant negative findings for family order may have resulted from order’s strong positive correlation with husband violence severity (r = .85, p < .001). The negative significant bivariate correlation between order and child abuse severity mitigates against this. However, we also elected to test this concern with a simulation. Using the R program, we created data sets of a 1000 observations of a normally distributed family order, which was a cause of husband violence severity, while husband violence severity was, alone a cause of child abuse severity. We succeeded in inducing correlations over 85% between both our simulated husband violence severity and child

176

C.R. Emery et al. / Child Abuse & Neglect 41 (2015) 170–181

Table 1 Sample descriptive statistics. Variable Any Physical Abuse of Child Child Abuse Severity Log Child Abuse Severity Husband Violence Severity Intimate Terrorism Power (standardized) Family Order Perceived Protective ISC IPV Observed Protective ISC IPV Collective Efficacy Neighborhood Solidarity Collective Efficacy Neighborhood Informal Social Control Standardized Social Desirability of reporting IPV Does family own a car? Wife’s age Wife’s years education Household income (rupees) Wife lower caste Wife middle caste Wife high caste Years married Child’s Age Child Female Household Size Number of Children

n

Mean

223 223 223 223 223 196 223 223 223 223 223 223 220 223 223 223 205 205 205 223 223 223 223 223

27.70% 2.07 −0.32 1.59 6.70% 0.01 25.41 1.29 0.01 8.49 10.77 0.01 38.64% 32.90 7.43 17,010 37.07% 24.88% 38.05% 11.31 9.16 46.80% 4.81 1.85

Standard deviation 6.34 .29 2.26 3.09 .25 1.02 3.78 .57 .13 2.27 2.09 1.00 .49 6.56 5.68 2220 .48 .43 .49 6.14 5.36 .50 1.50 .81

abuse severity, and between simulated family order and child abuse severity. We then regressed simulated child abuse severity on both family order and husband violence severity, and repeated the procedure 1000 times. T statistics for the family order coefficient centered around zero, and false positives were consistent with the rules of statistical chance (18 out of 1000 or 1.8% had t-statistics larger than 2, which is comparable to the expected 2.5%). Results Descriptive statistics are provided in Table 1. In addition to the sample characteristics discussed in the methods section, Table 1 suggests that nearly 28% of the sample reported some physical abuse of the child in the last year. With regard to intimate terrorism (Johnson, 2008); 6.7% of respondents were two standard deviations above the mean for this item. When bivariate statistics were used to examine basic relationships, we found a significant positive correlation between husband violence severity and child abuse severity (r = .14, p < .05). There was a significant negative correlation between wife’s power and child abuse severity (r = −.12, p < .05). There was also a significant positive correlation between intimate terrorism and child abuse severity (r = .18, p < .01). There was a significant negative correlation between family order and child abuse severity (r = −.23, p < .001). Finally, there was a strong positive correlation between perceived protective ISC IPV and child abuse severity (r = .23, p < .001), and no significant correlation between observed protective ISC IPV and child abuse severity (r = −.03, p = .60). It must be borne in mind, however, that these analyses did not have any control for the reverse causality bias that is likely for the ISC IPV – child abuse severity relationship. Table 2 presents the random effects regression results for both the child abuse severity and logged child abuse severity models. Because the child abuse severity model violates the linearity assumption in regression, we feel more comfortable using the logged child abuse severity model for inferential purposes. Hypothesis 1: Observed protective ISC IPV was negatively associated with logged child abuse severity (B = −2.76, p < .05), but perceived protective ISC IPV was not (B = .04, p = .90). The untransformed model suggests one act of protective ISC IPV by neighbors is associated with 14 fewer acts of physical child abuse each year. Hypothesis 2: When husband violence severity and other controls were introduced into the model, intimate terrorism remained a significant predictor of logged child abuse severity (B = 1.85, p < .05), but power did not (B = .002, p = .99). Regression results from the child abuse severity model suggest that being in a relationship characterized by intimate terrorism is associated with nearly 10 additional acts of physical child abuse each year. Hypothesis 3: Increases in family order were negatively associated with logged child abuse severity in the model (B = −.11, p < .01). The model for untransformed child abuse severity implies that a one unit increase in family order is associated with 0.4 fewer acts of physical child abuse each year. Power (hypothesis 4) was not associated with child abuse severity in either model. For the logged child abuse severity model, car ownership was associated with increases in physical child abuse (B = .79, p < .05), higher household incomes were associated with less (B = −1.00, p < .05), and the child’s age was associated with more child abuse (B = .13, p < .05). The mother’s age was marginally (negatively) associated with logged child abuse severity (B = −.07, p = .07). Social desirability was also associated with logged child abuse severity (B = .51, p < .01). The model explained 26% of the variance in logged child abuse severity, and 31% of untransformed child abuse severity.

C.R. Emery et al. / Child Abuse & Neglect 41 (2015) 170–181

177

Table 2 Random effects regression of child abuse severity and logged child abuse severity (N = 182). Child abuse severity Variable Husband Violence Severity Intimate Terrorism Power (standardized) Family Order Perceived Protective ISC IPV Observed Protective ISC IPV Collective Efficacy Neighborhood Solidarity Collective Efficacy Neighborhood Informal Social Control Standardized Social Desirability of reporting IPV Does family own a car? Wife’s age Wife’s years education High Household Income Middle Household Income Wife lower caste Wife middle caste Years married Child’s Age Child Female

B 0.60* 9.92*** −0.14 −0.39** 2.95** −14.09*** −0.31 −0.12 −0.36 −0.13 −0.27* −0.08 −0.42 0.10 0.41 0.73 −0.11 0.28 −0.62

SE B 0.25 2.41 0.53 0.12 1.05 3.76 0.21 0.27 0.58 1.16 0.11 0.08 1.39 1.36 1.10 1.22 0.15 0.18 0.99

Logged child abuse severity B 0.07 1.85* 0.002 −0.11** 0.04 −2.76* −0.01 0.08 0.51** 0.79* −0.07† −0.004 −1.00* −0.54 −0.22 −0.24 −0.04 0.13* −0.0005

SE B 0.08 0.81 0.18 0.04 0.35 1.27 0.07 0.09 0.20 0.39 0.04 0.03 0.47 0.46 0.37 0.41 0.05 0.06 0.33

Note: Regression standard errors corrected for clustering by ward. † p < .10. * p < .05. ** p < .01. *** p < .001.

Differences Between Models There were several differences in the findings between the untransformed and log-transformed models. The most important of these differences is the significant positive association between protective perceived ISC IPV and child abuse severity (B = 2.95, p < .01) in the untransformed model. Child’s age and high household income were not significant predictors in the untransformed model. Husband violence severity was significantly associated with child abuse severity in the untransformed model (B = .06, p < .05), but not the log-transformed model. In addition to the lack of significance in the untransformed model, the sign of the social desirability coefficient also changed (B = −.36, p = .54). Even when changes in coefficient sign do not involve coefficients that are not significant in either model, such changes in sign can be considered anomalous in a monotonically transformed model, and are hence important to note. Several other variables changed in sign (but not significance) between the models. Most noteworthy of these was power, which had a negative coefficient in the untransformed model (B = −.14, p = .80) but a positive coefficient in the log transformed model (B = .002, p = .99). Based on the size of the standard errors of these coefficients, these changes are by no means significant. However, this change in sign, accompanied as it is by other changes in coefficient sign, is worth noting. Both wife caste dummy variables, as well as the middle household income dummy variable were negative (but not significant) in the log-transformed model and positive (but not significant) in the untransformed model. Car ownership was also negative (but not significant) in the untransformed model (B = −.13, p = .91). Finally, coefficient for the collective efficacy measure of informal social control was not significant in either model, but was negative (B = −.12, p = .65) in the untransformed model and positive (B = .08, p = .38) in the log-transformed model. As noted in the analytic issues section, the untransformed model had significant non-linearity as indicated by Pregibon’s link test (hat2 = .03, p < .01). The significant link test can be caused by either non-linearity or outliers, both of which are known to bias coefficients. For this reason, findings (both inference and direction) from the logged child abuse severity model may be more reliable. Because the untransformed child abuse severity model showed signs of nonlinearity and because Emery, Wu, and Tsolmon (2014) hypothesized and tested for a curvilinear (parabolic) relationship between family order and IPV injury, but did not find a non-linear relationship, we tested for a parabolic relationship between family order and child abuse severity (results not shown). Consistent with Emery, Wu, and Tsolmon’s hypothesis that injurious family violence should be at a minimum for the middle range of order, the squared term was positive (B = .01, p = .80) but was not significant. However, introducing the order2 term into the model induces highly inflated VIFs (VIF = 71 for order2 , VIF = 70 for order), and the terms were jointly significant (2 = 10.4, df = 2, p < .01). Hence, it is not impossible that multicollinearity is responsible for the null finding for order2 , and that a larger sample size would detect a significant non-linear relationship (but power analysis indicates the sample size would need to be on the order of 4500 for 70% power).

178

C.R. Emery et al. / Child Abuse & Neglect 41 (2015) 170–181

Discussion Acts of Protective ISC IPV and Intimate Terrorism The findings from the random effects regression models supported several of our hypotheses. When observed protective ISC IPV was reported (neighbors actually intervened), this was associated with significantly and substantially less child abuse severity in both models (Hypothesis 1). Intimate terrorism, on the other hand, was associated with significantly and substantially more child abuse severity in both models (hypothesis 2). Protective ISC IPV. The fact that our models consistently found support for hypothesis one has important implications for theory, research, and practice. First, at the theoretical level, they suggest that social control broadly, and protective informal social control more narrowly, may indeed be effective in helping to protect children against physical assault by their parents. This is consistent with theory and past research on informal social control and crime (cf. Hirschi, 1969/2002; Sampson et al., 1997). The findings are also consistent with the findings of Emery, Trung, et al. (2013) and Emery, Wu, and Tsolmon (2014). Based on our findings, past research, and theory, exploration of protective effects of informal intervention by other social groups (e.g. members of religious bodies, co-workers, employers) may also yield fruitful results. Perhaps more importantly, our findings suggest there may be positive externalities for informal neighbor intervention against IPV. Families in which IPV is being perpetrated are at higher risk for physical child abuse, and for more severe child abuse (Knickerbocker et al., 2007). Although there is almost certainly some omitted variable bias in estimates of the direct relationship between IPV and physical child abuse, the continued finding of a robust relationship when different models with different controls are tested suggests support for a direct pathway (the second type of explanation in Knickerbocker et al., 2007). Consistent with such a direct link, our data show that when neighbors intervene informally to protect victims of IPV, children also appear to better protected against physical abuse. It may be that IPV and physical abuse are inextricably linked, and that to solutions to the problem of the latter must invariably address the former. More research is needed to replicate these findings. Likewise, experimental methods and ISC IPV boosting community interventions need to be developed in order to ascertain the extent to which our findings are causal. At the same time, NGOs and practitioners would do well to consider these findings in the context of developing countries like Nepal, where formal social control of IPV and child abuse is often absent, and next door neighbors and community support may be the only thing standing between victims and serious injury or death. Because we did not find a relationship between the traditional collective efficacy measure of informal social control and physical child abuse severity, our findings suggest that interventions developed to improve neighborhood control of crime may not result in beneficial decreases in IPV or child abuse. Rather, community interventions may need to specifically seek to raise ISC IPV and ISC CM in order to have reductive effects on family violence. It is important to note that these findings continue to appear for protective informal social control only (Emery, Trung, et al., 2013; Emery, Wu, & Tsolmon, 2014). That is, how neighbors intervene may matter just as much as whether they intervene. Emery, Trung, and Wu (2013) suggest that informal interventions focused on punishing the perpetrator have some deterrent effect, but at the same time exacerbate immature defense mechanisms and model coercive conflict resolution, resulting in effects that are mixed in direction. If true, this research has extremely important implications for social control generally, in that overly punitive social control may exacerbate rather than reduce crime. Indeed, research on coerced mental health services and procedural justice suggest that soft control is more likely to succeed (Watson, Angell, Vidalon, & Davis, 2010) and be perceived as fair (Tyler, 2004). Intimate Terrorism. Families in the top 6% of the data for coercive control (Johnson, 2008; Stark, 2007) by husbands were at risk for significant and substantially higher frequency of child abuse. In fact, when intimate terrorism was controlled, husband violence severity was no longer even significant in the log transformed model for child abuse severity. This finding suggests the importance of continued use and development of typologies for IPV and child maltreatment. The controlling context of IPV may be as or even more important in its contribution to child abuse than IPV itself. Our findings suggest that specific efforts targeted at reducing this serious, controlling form of IPV may have disproportionately beneficial effects for protecting children. This could take the form of both increased formal monitoring of suspected intimate terrorism cases by authorities, and also efforts to educate and increase informal of this particularly serious form of IPV. This is enormously important for child well being in Nepal, since more our data suggest that more than one out of every four Kathmandu children suffered abusive physical violence at the hands of their parents in the last year. Unlike Emery et al.’s (in press-b) findings for IPV in Kathmandu, we did not find a significant relationship between the mothers’ self estimated power in the husband-wife relationship and child abuse severity when intimate terrorism is controlled. Hypothesis four was not supported. In this respect, the extension of Emery’s (2011) theoretical argument to child abuse did not result in the expected results. It is probably too early to conclude that Emery’s (2011) argument about IPV is irrelevant to child abuse, as this is a first finding and the measure of power is limited. However, we would like to stress that perhaps an inclusive, versus an either-or approach is warranted. In Emery et al. (in press-b), we argue that power is uniquely important for understanding IPV, and find evidence that is consistent with this claim. Taken together, our findings suggest that the most productive approach may be to develop a measure of power that incorporates both control attempts (Johnson, 2008) (coercive control according to Stark, 2007), and de facto power differentials in intimate relationships (Emery, 2011).

C.R. Emery et al. / Child Abuse & Neglect 41 (2015) 170–181

179

Is Perceived ISC IPV a Risk Factor? When it is measured separately from observed ISC IPV, our findings for perceived ISC IPV, as well as those of Emery, Wu, and Raghavan (2014) bear enough of a resemblance to the 2002 film Minority Report, to warrant a separate section in the discussion (Spielberg, 2002). In Spielberg’s (2002) science fiction thriller, the dissenting voice of one prognosticator from the other two becomes a matter of life and death. Although it is not significant in the log-transformed model, we found that perceived ISC IPV was positively associated with child abuse severity in a bivariate correlation and in the untransformed model. Emery, Wu, and Raghavan (2014) similarly found that perceived ISC IPV was significantly associated with IPV injury, even when they attempted to control for reverse causality bias by controlling for husband violence severity. The possibility that perceived ISC IPV may be a risk factor for IPV injury and child abuse, even as observed ISC IPV appears to be a protective factor, cannot be ignored, and is the reason the findings evoke the 2002 film. Emery, Wu, and Raghavan (2014) speculated that the positive association with IPV injury might occur because when women perceived the likelihood of intervention by neighbors, but such intervention did not occur, women might be less likely to take self-protective action in the presence of potentially life-threatening behavior. This is also a possible interpretation of our finding. Such an explanation implies that the problem lies not with the perception of likely ISC IPV, but with the failure of neighbors to act. However, another possibility is that even protective ISC IPV by neighbors is likely to be interpreted through the lens of negative affect and more particularly parental hostile attribution bias, which have been found to be positively associated with parent to child physical aggression (Mammen, Kolko, & Pilkonis, 2002). In plain speech, it is possible that perceptions of informal intervention by neighbors are translated through perpetrators’ cognitive biases that err in the direction of perceiving hostile intent. Hence, we must consider the possibility that intervention against IPV or child abuse raises the risk of more severe violence against the victim, or violence against the agent of control. At the same time, the large protective finding for observed protective ISC IPV suggests that the overall effect of ISC IPV may be protective. This is both because the finding for observed protective ISC IPV is still negatively and significantly associated with both child abuse severity and logged child abuse severity when perceived protective ISC IPV is controlled, and because the effect size for observed ISC IPV is larger than for perceived ISC IPV when regression coefficients are rendered comparable via standardization (results not shown but available on request). Given that the overall direction of the findings suggests that protective ISC IPV by neighbors is negatively related to child abuse severity and IPV injury, the question becomes how interventions might sufficiently optimize impact by damping down the positive association between perceived ISC IPV and abuse. If the reason behind the perceived finding is lack of self-protective action by victims, the appropriate approach may be to try to boost self-protection by potential victims and neighborhood solidarity at the same time that the intervention attempts to boost protective ISC IPV. Increasing neighborhood solidarity might be important both because having a stake and interest in the well-being of the family next door may make ISC IPV more likely, but also because frequent interaction with neighbors may help victims to more accurately assess the likelihood of their neighbors actually stepping in to help them. If the reason for the positive association between perceived ISC IPV and child abuse severity rests on the negative affect/hostile attribution bias explanation (Mammen et al., 2002), then efforts to develop strategic informal interventions that damp down the perpetrator’s negative affect when control occurs will be critical. The difference in protective versus punitive findings for ISC CM (Emery, Trung, et al., 2013) provide initial guidance for this. It may be the case that the softer and more protective the form the control takes, the more likely it is to be successful. Further, specific attempts to model empathy and concern for the victim may be fundamentally important to a successful ISC IPV approach, because such an approach changes the focus from the perpetrator’s action to the victim’s well being, models non-coercive conflict resolution, and de-emphasizes the potentially punitive role of the neighbor. Emery, Trung, et al. (2013) called for community interventions that have four steps: (1) community building, (2) community education, (3) informal social control training, and (4) a formal social control-informal social control coordination step. We second this, but also emphasize the importance of carrying out control in a form likely to optimize desired effects, and we call for continued research to tease out the mechanisms that explain the associations found for perceived and observed ISC IPV and, in the long run, allow for optimal impact of informal social control. Negative Association for Family Order; Counterintuitive or Not? In contradiction to hypothesis 3 and the findings of Emery, Wu, and Tsolmon (2014), we found a strong, significant negative association between family order and child abuse severity in both models. One possibility is that social disorganization (disorder) (Sampson et al., 1997; Shaw & McKay, 1969) explanations of family violence are more relevant for child abuse, but that deviant order (Emery, 2011) type explanations of family violence are more relevant for IPV. In less formal terms, highly controlled relationships may be more dangerous for women than children, while violence emerging from more general social disorganization may be more dangerous for children. We suspect, however, that this is a false dichotomy. Not least important is the connection between perpetration of intimate terrorism and child homicide (Jaffe, Campbell, Olszowy, & Hamilton, 2014), which, via its link to totalitarian control (Emery, 2011; Stark, 2007) is potentially linked to high order. The implication originally drawn by Emery, Wu, and Tsolmon (2014) from Emery’s (2011) typology of IPV was that both high and low order should be associated with more severe and injurious violence. Further, although our finding for the order squared term was not significant, it was in the expected

180

C.R. Emery et al. / Child Abuse & Neglect 41 (2015) 170–181

direction. A positive coefficient for a squared term implies an upward facing parabola; that is, a curve with a minimum in the middle, rather than a maximum. Examples of extremely ordered and extremely disordered families may be comparatively rare in random samples of the population, which could compound the statistical power and multicollinearity problems. Power analysis suggests that a sample size of 4500 respondents could detect such a relationship. Smaller sample sizes that targeted advantaged and disadvantaged communities for comparison, or otherwise focused sample selection on families at risk for totalitarian dictatorship type IPV and anarchic IPV (Emery, 2011) might be able to detect a parabolic relationship with a smaller sample size. Hence, we are not persuaded that both high and low order are not risk factors for physical child abuse. Our findings suggest that further research on family order and child maltreatment is both warranted and necessary. Limitations This study cross-sectional, non-experimental, subject to reporting biases, and constitutes a first study of physical abuse of children by parents in Nepal. Findings are thus not necessarily causal relationships. Likewise, our findings can only be generalized to the city of Kathmandu. Although we have controlled for a measure of social desirability bias about IPV reporting, this bias may still be present in our data. Reliability is also limited by the fact that we only interviewed one parent from each family. All of our findings require replication. Although we attempt to control for IPV severity, reverse causality bias remains a significant threat to our analyses of ISC IPV. A second order of concern involves the anomalous sign changes for some regression coefficients between the two models. It is possible, perhaps even likely, that these occurred because of the nonlinearity detected in the untransformed model, and that the results from the log transformed model are reliable. However, we cannot rule out the possibility that in the few instances where findings were significant in one model but not in the other, that the difference was an artifact of method. The sample used for analysis was smaller than the sample size deemed reasonable by power analysis, which means that null findings in this study may have been significant with a larger sample. This does not, however, have any bearing on those variables found to be significant. The reader will also of course note that the original power analysis only required 150 observations; still fewer than those used for this analysis. It is possible that the social desirability measure captures IPV secrecy as well as social desirability. Conclusion Our results suggest that more than one in every four children in Kathmandu experiences physical abuse at the hands of their parents at least once in the past year. This is a serious risk to child well-being, especially in the context of a developing country which has relatively little formal support or control designed to deal with this problem. The main finding that acts of protective ISC IPV by neighbors is negatively associated with child abuse severity suggests that working with existing community and social network structures may be a fruitful approach to developing child protection in this context. We believe such an integrative approach much more likely to have positive effects than an approach that develops formal services and controls in isolation. At the same time, research is needed to unpack the complex pathways by which ISC IPV may affect perpetration, and community interventions need to be designed with such pathways in mind. Our findings also suggest that, controlling for all other variables in the model, disordered households and mother-father relationships characterized by coercive control constitute serious risk factors for children. Further research on theoretically informed typologies of IPV and child maltreatment is much in need. Taken together with existing research, the findings from this paper suggest the response of the person next door matters when it comes to child and family well being, and that response ought to be taken into account when policy and practice are developed. Acknowledgements We thank Hon. Eknath Dakhal, the Minister of Cooperatives and Poverty Alleviation, and Mr. Santosh Paudel who made this study possible. We also wish to thank Keshav Shrestha, the Kathmandu research team, and the participants. We would like to thank the Faculty of Social and Policy Sciences at the University of Bath for providing time and space to write up the study results. Measure development was facilitated by a grant from the H.F. Guggenheim Foundation and Namseoul University. References Appel, A., & Holden, G. (1998). The co-occurrence of spouse and physical child abuse: A review and appraisal. Journal of Family Psychology, 12(4), 578–599. Browning, C. (2002). The span of collective efficacy: Extending social disorganization theory to partner violence. Journal of Marriage and Family, 64(4), 833–850. Central Bureau of Statistics. (2012). National population and housing census 2011 (national report) (Vol. 1) Kathmandu, Nepal: Government of Nepal National Planning Commission Secretariat. Retrieved from: http://unstats.un.org/unsd/demographic/sources/census/2010 phc/ Nepal/Nepal-Census-2011-Vol1.pdf Chinn, S. (2000). A simple method for converting an odds ratio to an effect size for use in meta-analysis. Statistics in Medicine, 19, 3127–3131. Dekeseredy, W., Schwartz, M., Alvi, S., & Tomaszewski, E. (2003). Perceived collective efficacy and women’s victimization in public housing. Criminology and Criminal Justice, 3(1), 5–27. Dhakal, S. (2008). Nepalese women under the shadow of domestic violence. Lancet, 371(9612), 547–548.

C.R. Emery et al. / Child Abuse & Neglect 41 (2015) 170–181

181

Emery, C. (2011). Disorder or deviant order? Re-theorizing domestic violence in terms of order, power, and legitimacy. A typology. Aggression and Violent Behavior, 16(6), 525–540. Emery, C., Eremina, T., Yang, H., Yoo, C., Yoo, J., & Jang, J. (2014). Protective informal social control of child maltreatment, and child abuse injury in Seoul. Journal of Interpersonal Violence (in press). Emery, C., Jolley, J., & Wu, S. (2011). Desistance from intimate partner violence: The role of neighborhoods’ legal cynicism, collective efficacy, and social disorganization in Chicago neighborhoods. American Journal of Community Psychology, 48(3-4), 373L 383. Emery, C., Kim, J., Song, H., & Song, A. (2013). Child abuse as a catalyst for wife abuse? Journal of Family Violence, 28(2), 141–152. Emery, C., Trung, H., & Wu, S. (2013). Neighborhood informal social control and child maltreatment: A comparison of protective and punitive approaches. Child Abuse & Neglect, http://dx.doi.org/10.1016/j.chiabu.2013.05.002 (in press) Emery, C., Wu, S., & Raghavan, R. (2014). Why are residents of Beijing’s hutongs better protected from intimate partner violence injury? Informal social control and community life in Beijing. Injury Prevention, http://dx.doi.org/10.1136/injuryprev-2013-041117 Emery, C., Wu, S., & Tsolmon, O. (2014). The peril of order? IPV, injury, and order in Mongolian families. Journal of Interpersonal Violence, http://dx.doi.org/10.1177/0886260514532526 Emery, C., et al. (2014). Power and control in Kathmandu: A comparison of attempted power, actual power, and achieved power in understanding husband violence prevalence, severity and injury. Violence Against Women (in press-b). Garbarino, J. (1976). A preliminary study of some ecological correlates of child abuse: The impact of socioeconomic stress on mothers. Child Development, 47(1), 178–185. Ghimire, N. (2014). Personal conversation with the director of the Nepal Health Research Council Research team. 9-11-2014. Guterman, N., Lee, S., Taylor, C., & Rathouz, P. (2009). Parental perceptions of neighborhood processes, stress, personal control, and risk for physical child abuse and neglect. Child Abuse & Neglect, 33(12), 897–906. Herrenkohl, T., Sousa, C., Tajima, E., Herrenkohl, R., & Moylan, C. (2008). Intersection of child abuse and children’s exposure to domestic violence. Trauma, Violence & Abuse, 9(2), 84–99. Hirschi, T. (1969/2002). Causes of delinquency. New Brunswick, NJ: Transaction. Holtzworth-Munroe, A., Meehan, J., Herron, K., Rehman, U., & Stuart, G. (2000). Testing the Hotzworth-Munroe and Stuart (1994) batterer typology. Journal of Consulting and Clinical Psychology, 68, 1000–1019. Hsiao, C. (2003). Analysis of panel data (2nd ed.). New York, NY: Cambridge University Press. Jaffe, P., Campbell, M., Olszowy, L., & Hamilton, L. (2014). Paternal filicide in the context of domestic violence: Challenges in risk assessments and risk management for community and justice professionals. Child Abuse Review, 23(2), 142–153. Johnson, M. (1995). Patriarchal terrorism and common couple violence: Two forms of violence against women. Journal of Marriage and Family, 57, 283–294. Johnson, M. (2008). A typology of domestic violence. Intimate terrorism, violent resistance and situational couple violence. Boston, MA: Northeastern University Press. Johnson, M., & Leone, J. (2005). The differential effects of intimate terrorism and situational couple violence: Findings from the National Violence Against Women Survey. Journal of Family Issues, 26(3), 322–349. Johnston, J., & DiNardo, J. (1997). Econometric methods. New York: McGraw-Hill. Kelly, J., & Johnson, M. (2008). Differentiation among types of intimate partner violence: Research update and implications for interventions. Family Court Review, 46(3), 476–499. Kornhauser, R. (1978). Social sources of delinquency. An appraisal of analytic models. Chicago, IL: University of Chicago Press. Knickerbocker, L., Heyman, R., Slep, A., Jouriles, E., & McDonald, R. (2007). Co-occurrence of child and partner maltreatment. Definitions, prevalence, theory, and implications for assessment. European Psychologist, 12(1), 36–44. Mammen, O., Kolko, D., & Pilknois, P. (2002). Negative affect and parental aggression in child physical abuse. Child Abuse & Neglect, 26(4), 407–424. Melton, G. (2014). Strong communities for children: A community-wide approach to prevention of maltreatment. In J. Korbin, & R. Krugman (Eds.), Handbook of child maltreatment. New York: Springer. Molnar, B., Buka, S., Brennan, J., Holton, J., & Earls, F. (2003). A multilevel study of neighborhoods and parent-to-child physical aggression: Results from the project on human development in Chicago neighborhoods. Child Maltreatment, 8(2), 84–97. Paudel, G. (2007). Domestic violence against women in Nepal. Gender Technology and Development, 11(2), 199–233. Sampson, R., Morenoff, J., & Gannon-Rowley, T. (2002). Assessing neighborhood effects: Social processes and new directions in research. Annual Review of Sociology, 28, 443–478. Sampson, R., Raudenbush, S., & Earls, F. (1997). Neighborhoods and violent crime: A multilevel study of collective efficacy. Science, 277, 918L 24. Schwartz, M., & Dekeseredy, W. (2008). Interpersonal violence against women: The role of men. Journal of Contemporary Criminal Justice, 24(2), 178–185. Segal, U. (1995). Child abuse by the middle class? A study of professionals in India. Child Abuse & Neglect, 19(2), 217–231. Shaw, C., & McKay, H. (1969). Juvenile delinquency and urban areas (2nd ed.). Chicago: University of Chicago Press. Silwal, R. (2006). Caste-based discrimination in Nepal. Interview with U.N. special representative Ian Martin. Ohmy News; International Interviews,. Retrieved from: http://english.ohmynews.com/articleview/article view.asp?no=324452&rel no=1 Spielberg, S. (2002). Minority Report. Twentieth Century Fox Film, written by P.K. Dick, S. Frank & J. Cohen. Starring Tom Cruise. 145 minutes. Spybrook, J., Raudenbusch, S., Liu, X., Congdon, R., & Martinez, A. (2008). Optimal design for longitudinal and multilevel research: Documentation for the ‘Optimal Design’ software. (Unpublished manual). Retrieved from: http://www.ssicentral.com Stark, E. (2007). Coercive control. How men entrap women in personal life. New York: Oxford University Press. Straus, M., & Douglas, E. (2004). A short form of the revised conflict tactics scales, and typologies for severity and mutuality. Violence and Victims, 19(5), 507–520. Straus, M., & Hamby, S. (1997). Measuring physical and psychological maltreatment of children with the conflict tactics scales. In G. Kaufman Kantor, & J. Jasinski (Eds.), Out of the darkness: Contemporary perspectives on family violence. Thousand Oaks, CA: Sage. Taylor, C., Guterman, N., Lee, S., & Rathouz, P. (2009). Intimate partner violence, maternal stress, nativity, and risk for maternal maltreatment of young children. American Journal of Public Health, 99(1), 175–183. Tyler, T. (2004). Enhancing police legitimacy. Annals of the American Academy of Political and Social Science, 593(1), 84–99. Watson, A., Angell, B., Vidalon, T., & Davis, K. (2010). Measuring perceived procedural justice and coercion among persons with mental illness in police encounters: The police contact experience scale. Journal of Community Psychology, 38(2), 206–226. World Health Organization. (2010). Child Maltreatment. Fact Sheet No: 150. http://www.who.int/mediacentre/factsheets/fs150/en/print.html Yonas, M., Lewis, T., Hussey, J., Thompson, R., Newton, R., English, D., & Dubowitz, H. (2010). Perceptions of neighborhood collective efficacy moderate the impact of maltreatment on aggression. Child Maltreatment, 15(1), 37–47.

Do family order and neighbor intervention against intimate partner violence protect children from abuse? Findings from Kathmandu.

Drawing on previous research on intimate partner violence, child maltreatment, and informal social control, we hypothesized relationships between chil...
605KB Sizes 1 Downloads 5 Views