AGGRESSIVE BEHAVIOR Volume 41, pages 335–345 (2015)

The Interrelation Between Victimization and Bullying Inside Young Offender Institutions Jenny Häufle* and Daniel Wolter University of Cologne, Cologne, Germany

.. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. Bullying and victimization are serious problems within prisons. Young Offender Institutions (YOIs), in particular, suffer from high rates of inmate‐on‐inmate violence. More recent theories about the development of bullying in closed custody institutions imply a relationship between the experience of victimization and the usage of bullying. In our study, we test this linkage using longitudinal survey data taken at two time‐points from 473 inmates (aged 15–24) inside three YOIs in Germany. We first analyze the extent of bullying and victimization, and then used a longitudinal structural equation model to predict inmate bullying behavior at time 2 based on victimization that occurred at time 1. Age is used as a predictor variable to account for differences in the amount of victimization and bullying. Results suggest that bullying and victimization are high in the YOIs, which were subject to research. Most inmates reported being a bully and a victim at the same time. Younger inmates use more direct physical bullying but not psychological bullying. An increase in psychological bullying over time can significantly be explained by victimization at an earlier measurement time point. Our study therefore supports recent theoretical assumptions about the development of bullying behavior. Possible implications for prevention and intervention are discussed. Aggr. Behav. 41:335–345, 2015. © 2014 Wiley Periodicals, Inc.

.. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. Keywords: bullying; victimization; young offender institutions; juvenile delinquency; structural equation modeling

INTRODUCTION

Aggression and victimization inside prisons occur on a day‐to‐day basis (Edgar, O’Donnell, & Martin, 2003; Ireland, 2005a; O’Donnell & Edgar, 1999). Different studies have shown that aggression can be seen as a common behavioral strategy used by the inmates to integrate themselves into the prison subculture (Ireland & Ireland, 2008) and may even help them survive in the “total environment” of prison (Bottoms, 1999; Goffman, 1961). Moreover, the structure of prison itself facilitates routine victimization (Edgar, O’Donnell, et al., 2003; Seymour, 2003). Victimization can have serious ramifications for the well‐being of the prisoners, resulting in, for example, depression, somatic symptoms, anxiety, and even suicidal behavior (Blaauw, Winkel, & Kerkhof, 2001; Grennan & Woodhams, 2007; Ireland, 2005a; Johnson Listwan, Colvin, Hanley, & Flannery, 2010; Maitland & Sluder, 1996). Furthermore, the negative impact of aggression can extend to life beyond prison as those people engaging in aggressive behavior inside prison may have a harder time reintegrating themselves into society after their release (South & Wood, 2006). It is therefore important to understand the occurrence and the causes of aggression and victimization inside prisons, and to lay the basis for effective prevention strategies. © 2014 Wiley Periodicals, Inc.

Aggression and Bullying Inside Prison—Definition and Forms In prison, aggression between inmates is often referred to as bullying (e.g., Ireland, 2000; Ireland & Ireland, 2008), whereas a broader definition of bullying is required to account for the specifics of the prisoners’ subculture and behavioral misconduct between inmates (Ireland, 2000; South & Wood, 2006). Bullying is a subcategory of aggression with its own characteristics and, although it overlaps with violence, bullying can take place without violence (Olweus, 1996, 2011). The definition of bullying proposed by Ireland (2002a)

Contract grant sponsor: German Research Foundation (Deutsche Forschungsgemeinschaft, DFG). Daniel Wolter currently works at the Cologne University of Applied Science in the Institute of Management and Organisation in Social Work (IMOS), Cologne, Germany.  Correspondence to: Jenny Häufle, Institute of Criminology at the University of Cologne, Albertus‐Magnus‐Platz, 50923 Cologne, Germany. E‐mail: [email protected] Received 20 December 2013; Accepted 6 May 2014 DOI: 10.1002/ab.21545 Published online 26 June 2014 in Wiley Online Library (wileyonlinelibrary.com).

336

Häufle and Wolter

includes (a) being a victim of direct and/or indirect aggression that (b) happens on a weekly basis by (c) the same or different perpetrator(s). Furthermore, it extends prior definitions by (d) stating that “single incidences of aggression can […] be viewed as bullying, particularly when they are severe and when the individual either believes or fears that they are at risk of future victimization by the same perpetrator or others” (p. 26), (e) by withdrawing the need for an intention of the perpetrator to harm the victim as long as the victim believes that he has been aggressed, and (f) by acknowledging that the imbalance of power between the perpetrator and the victim does not need to be manifest but implied. Ireland and Ireland (2008) further point out that bullying behavior inside prison cannot be limited to single aspects of aggression and can take different forms, for example, physical, verbal, and coercive forms of aggression (Ireland & Ireland, 2003; Kury & Smartt, 2002; Lahm, 2007; Wolff, Blitz, Shi, Siegel, & Bachman, 2007). The differentiation between direct aggression (including verbal and physical forms, such as name calling and punching) and indirect aggression (or relational aggression, such as excluding someone, gossiping, or spreading rumors) as described by Björkqvist (e.g., Björkqvist, 1994; Björkqvist, Österman, & Kaukiainen, 1992) has also proven to be important for prison research into bullying. Indirect aggression is found to be closely related to more subtle aspects of direct aggression such as verbal aggression (Ireland & Ireland, 2008). Such indirect verbal forms of bullying behavior occur most often inside prisons together with physical bullying (Connell & Farrington, 1996; Edgar & O’Donnell, 1998; Ireland, 2005a; Ireland & Power, 2004; Kury & Smartt, 2002; Maitland & Sluder, 1998; South & Wood, 2006). Theories of Bullying Development From a developmental perspective on bullying in the prison setting, it is especially important to consider what leads inmates to engage in bullying behavior and what kind of bullying they may choose. Björkqvist, Österman, et al. (1992) propose in their Developmental Model of Aggression that aggressive styles are subject to developmental change during childhood and adolescence. According to the model, young children use mostly physical forms of aggression since they lack verbal skills. In the course of their development, they then adopt verbal forms of aggression and later more subtle and sophisticated strategies, which are carried out covertly and allow the aggressor to stay unidentified. The three strategies of aggression (physical, direct verbal, and indirect aggression) are seen as developmental stages that partly follow and partly overlap each other (Björkqvist, Österman, Aggr. Behav.

et al., 1992, p. 58). Although the model is based on the developmental period of childhood and adolescence, evidence was found that the model can be applied to older imprisoned age groups (Ireland, 2002b). Furthermore, the context in which different forms of aggression occur is relevant and the chosen strategy of aggression depends on the situation (Björkqvist & Niemelä, 1992). In her review of bullying literature, Ireland (2000) concludes that bullying inside prisons is largely a product of the environment (p. 205). Theories about the development of bullying inside the prison context take into account the special characteristics of the prison environment and acknowledge the importance of victimization as a cause for bullying. The Applied Fear Response Model (Ireland, 2005b, 2011, 2012) and the Multifactor Model of Bullying in Secure Settings (Ireland, 2012) imply a link between fear of victimization and the use of bullying to protect oneself from further victimization. Bullying as a “behavioral consequence of victimization and fear” (McCorkle, 1992, p. 162), can be seen as an adaptive strategy of the inmates to deal with the situation, and as a reactive form of aggression (Dodge, 1991) toward the immediate threat of victimization. Inmates who experienced victimization may use bullying behavior to protect themselves from further victimization. Incorporating Theories About Bullying Inside Prison Brought together, the different theories imply that younger inmates may engage more in physical bullying whereas older inmates may use more subtle forms of bullying. Younger inmates are more likely than older ones to employ aggressive precautions to avoid being victimized (McCorkle, 1992). Thus, bullying can be seen as an adaptive social problem‐solving strategy for young inmates (Ireland & Murray, 2005) since other strategies can develop at an older age (Spain, 2005). Edgar, Martin, and O‘Donnell (2003) found that the direct experience of victimization is most related to fear and therefore to bullying. The link between victimization and bullying might therefore be strongest for inmates in YOIs, which house solely adolescents and young adults, who might experience more direct bullying than adults. Accordingly, higher rates of physical victimization were found for younger inmates (Ireland, 2012; Ireland & Power, 2004; Lahm, 2007; Maitland & Sluder, 1996) and age was found as a predictor of prison misconduct (Endrass, Rossegger, Noll, & Urbaniok, 2008; Gendreau, Goggin, & Law, 1997). Younger inmates also use more direct aggression than adult inmates (Ireland, 2011; McCorkle, 1992). Ireland (2002b) found evidence that juvenile inmates differ from young inmates only with regard to direct—not indirect—bullying, which partly

The Interrelation Between Victimization and Bullying

supports Björkqvist’s developmental theory. She also found a higher overall extent of bullying for juvenile than for young inmates. In Germany, Wirth (2006) found that over half (55.90%) of the inmates with a record of a violent offense in their personal files in 2005 were under 25 years of age. Kury and Smartt (2002) found that 51% of the young inmates participating in their study had been victimized at least once during their incarceration. In their cross‐sectional study, Bieneck and Pfeiffer (2012) found that 57% of their juvenile respondents experienced indirect aggression (including, e.g., the spreading of lies or deliberately ignoring someone) within the last month compared to 50% in the case of adult prisoners. Special attention therefore needs to be paid to young inmates. Still, youth correctional facilities remain under‐researched, especially in Germany, partly because most studies rely on data taken from inmates’ official personal records (Kury & Smartt, 2002). However, so far no study has used a longitudinal design to test such interrelationships between victimization and bullying. In the current study, we assess the interplay between bullying and victimization inside YOIs in Germany using self‐report data provided by the young inmates. The goal of the research presented is to extend former studies in other countries by employing current theories on bullying and victimization inside prison to the data. First, we will examine the amount of bullying and victimization in the facilities under research. Second, the predicted link between victimization and the later occurrence of bullying is analyzed using longitudinal structural equation modeling of two measurement time points. It is predicted that (1) victimization predicts later direct and indirect bullying behavior and, furthermore, that (2) younger inmates present a higher amount of direct physical bullying and a higher amount of victimization than older inmates. METHODS

Procedure Data for the study are drawn from the longitudinal research project “Violence and Suicide in Youth Correctional Facilities” (Neubacher, Oelsner, Boxberg, & Schmidt, 2011). The goal of the project is to analyze the adaptation process of the respondents to the prison context. The project combines a quantitative survey of four measurement time points with qualitative interviews with young inmates. The design of the project is similar to a cohort‐sequential design: At each measurement time point all inmates willing to participate in the study took the survey (cross‐sectional data for each measurement time point). Inmates who participated at more than one measurement time point provided longitudinal data. The lag between each of the measurement time points was

337

approximately 3 months. Data collection took place in three YOIs in two German federal states: North Rhine‐ Westphalia and Thuringia. Only facilities for male inmates were included in the study.1 Informed consent procedures, approved by the ethics committee of the University of Cologne, were followed during data collection. Two weeks prior to each data collection, inmates interested in the study attended a presentation about the project goals, study procedure, participation options, risks and opportunities, as well as the termination of the study. After the presentations, inmates were encouraged to ask questions and were given the chance to register for study participation (survey and/ or interview). All inmates who participated in the survey signed a consent form before taking the survey. The survey was administered by two to three members of the project team in a group setting with each session involving approximately 10 inmates. Before taking the survey, inmates were shown how questionnaires were pseudonymized with a random code, and were encouraged to talk about concerns with regard to data protection and the benefit of participation. Inmates used the opportunity to ask about the study and our goals. Many expressed their willingness to support the study. The average participation rate at the different measurement time points is 68.25%. Sample Longitudinal survey data from 473 inmates participating at a minimum of two successive measurement time points are used for analyzing the development of bullying over time and the interrelation between victimization and bullying. The mean age of the participants was 20.20 years (SD ¼ 1.79). In Germany, juvenile criminal law is applied from the age of 14 until the age of 18, but it can be applied until the age of 21 if the offense is seen as typically juvenile. Juvenile criminals are imprisoned in YOIs and inmates can stay there up to the age of 24. The age range in our sample is 15 to 24 years. Most inmates in the sample have German citizenship and about half have left secondary school without a leaving certificate. Only eight inmates indicated that they have no prior criminal record. Many prisoners were convicted for more than one crime or for a combination of different crimes (e.g., assault and theft). However, many convictions were related to assault (63.75%) and property crime (19.70%). A summary of the demographic variables can be found in Table I.

1

Due to a lack of independent youth prisons for females in Germany and the small number of female youth offenders, they were excluded from this project for logistical and financial reasons. Female youth offenders are being examined in a recently started follow‐up project. Aggr. Behav.

338

Häufle and Wolter

TABLE I. Demographic Data of the Sample Mean age in years (SD) Average sentence length in months (SD) Offense Assault relateda Property crime Other Citizenship Germanb Turkish Other School leaving certificate Non Hauptschule (secondary modern school from grades 5 to 9) Realschule (secondary modern school from grades 5 to 10) Other Prior criminal record Non 1–2 More than 3

20.20 30.62 n 262 81 68 n 371 23 56 n 242 166

(1.79) (17.80) (%) (63.75) (19.70) (16.55) (%) (82.44) (5.11) (12.45) (%) (51.82) (35.55)

38 (8.14) 21 n 8 104 350

(4.49) (%) (1.73) (22.51) (75.76)

a People were categorized by their most severe offense. Thus, the category “assault” includes people that reported more than one offense, if one of them was assault‐related. b The Category “German” consists of 348 persons with German citizenship and 23 persons with dual citizenship, where one of them is German.

Measures For the analysis, measures of experiences as a perpetrator and as a victim were used. Items that represent psychological and physical bullying were used as well as items representing physical and psychological victimization. All scales were derived from the DIPC‐S by Ireland and Ireland (2008). A German translation of the DIPC‐S was used and the scale was shortened and slightly adapted for the use in the current study. The psychometric properties of the scale were tested in a pretest of the study (Neubacher, Oelsner, & Schmidt, 2013).2 Age in years was used as a predictor variable. Psychological bullying. The items of this scale include behaviors indicative of indirect and more subtle direct verbal bullying. Seven items (e.g., “I have made fun of another inmate.”, “I have intimidated someone.”, “I have deliberately ignored or excluded someone.”), referring to different incidents between inmates that occurred within the last 3 months, constitute the psychological bullying subscale. The answers were coded on a four‐point Likert scale with the values 0 2

The different scales are found to constitute distinct factors in a factor analysis with good explained variances, high factor loadings, and good internal consistencies (psychological bullying: 55% explained variance, loading all above .67, a ¼ .86; physical bullying: 67% explained variance, loading all above .75, a ¼ .91; victimization: 50% explained variance, loading all above .60, a ¼ .89). Aggr. Behav.

(never), 1 (seldom), 2 (sometimes), and 3 (often). Internal consistency of the scale is a ¼ .86 for the current sample. Physical bullying. Six items that asked about physical fights and incidents between inmates within the last 3 months (e.g., “I have kicked or hit another inmate.”, “I have deliberately injured another inmate.”, “I have deliberately picked a fight.”) constitute the physical bullying subscale. The answers were coded on a four‐ point Likert scale with the values 0 (never), 1 (seldom), 2 (sometimes), and 3 (often). The internal consistency of the scale is a ¼ .91 for the current study. Victimization. The victimization scale consists of ten items (e.g., “Someone has deliberately insulted me.”, “Someone has tried to turn other prisoners against me.”, “I have been deliberately pushed.”) asking the inmates about different psychological and physical victimization experiences within the last 3 months. The answers were coded on a four‐point Likert scale with the values 0 (never), 1 (seldom), 2 (sometimes), and 3 (often). For the current analysis, one victimization variable was used. Internal consistency of the whole scale is a ¼ .89; Cronbach’s alpha is .83 for both subscales. Data Analysis Data analysis consisted of three major steps. First, the frequency of bullying and victimization was analyzed for the cross‐sectional data. In a second and third step of analysis, structural equation modeling was applied to the longitudinal data to analyze the theorized interrelation between victimization and physical and psychological bullying. Measurement invariance of the used scales was needed to meaningfully interpret the change in mean scores over time (step 2). Measurement invariance was tested using the Mplus formulation (in contrast to the LISREL formulation, for an overview see Millsap & Yun‐Tein, 2010) for categorical data as items with four categories were used (Rhemtulla, Brosseau‐Liard, & Savalei, 2012). Measurement invariance in this sense requires invariance of intercepts or thresholds and invariance of factor loadings. Using the latent response variable formulation with delta parameterization, a model assuming measurement invariance was tested against a nested model assuming no such restrictions using the weighted least squares means and variance adjusted estimation (WLSMV estimator) and the difftest option in Mplus Version 6 (see Muthén & Asparouhov, 2002; Muthén & Muthén, 2010). A level of significance of a ¼ .10 was used for difference testing to reduce the beta error as measurement invariance was the desired outcome. Third, a longitudinal model was proposed that, according to the theory, predicts physical and psychological bullying at measurement time point two by victimization rates at measurement time point one. For

The Interrelation Between Victimization and Bullying

identification reasons, one factor mean needed to be fixed at zero even in the invariant case (Muthén & Asparouhov, 2002). This means that the mean of psychological bullying, the mean of physical bullying, and the mean of victimization were fixed at zero at the first measurement time point. However, this allows the mean of the latent variables at measurement time point two to be interpreted as the mean change in the latent constructs from measurement time point one to two. Model specification was conducted in accordance with the theory (see Fig. 1). Correlations between the bullying and victimization measures at each time point—not between time points— were allowed. The analysis was carried out with Mplus Version 6 (Muthén & Muthén, 2010). Again, the formulation for categorical data with delta‐parameterization and the WLSMV estimator were used. A level of significance of a ¼ .05 was used for parameter estimation. RESULTS

In every longitudinal study, attrition and missing values occur. In the current study, an additional 361 inmates, who participated at one measurement point, did not participate a second time. The main reason for dropping out of the study was the termination of the current sentence, either due to release from prison or transfer to another facility (in 73% of the cases). People who dropped out did not differ from people who stayed in the study at any of the key variables (physical bullying: U ¼ 67280.00, z ¼ 0.68, p ¼ .497; psychological bul-

339

lying: U ¼ 65608.00, z ¼ 1.22, p ¼ .221; victimization: U ¼ 65854.50, z ¼ 1.10, p ¼ .267) or any of the demographic variables. Missing values for the relevant items are under 2% for the final sample. Due to the low frequency of missing data, distortion of the results seems unlikely (Graham, Cumsille, & Elek‐Fisk, 2003). Thus, the default mode to handle missing data with a robust weighted least square estimator without covariates in the model is used for all analysis that is pairwise‐present (Asparouhov & Muthén, 2010; Muthén & Muthén, 2010). Amount of Bullying and Victimization Bullying and victimization occur rather often inside the YOIs under research. Only 125 inmates (26.48%) indicate no victimization within the last 3 months at the first measurement time point. Most often, inmates report psychological victimization (72.00%) followed by physical victimization (44.80%). The same pattern holds true for perpetuation of bullying behavior: Most inmates report psychological bullying (83.09%) followed by physical bullying (62.37%). Only 74 inmates (15.65%) have not reported perpetuating any kind of bullying of another inmate within the last 3 months of their incarceration. When looking at the co‐occurrence of bullying and victimization, 32 (6.78%) inmates neither experienced victimization nor did they bully another inmate. Three hundred five (64.62%) of the inmates reported both bullying and victimization.

Fig. 1. Longitudinal prediction model of the proposed link between bullying and victimization. Aggr. Behav.

340

Häufle and Wolter

Measurement Invariance for the Scales Measurement invariance was tested for the physical bullying and psychological bullying subscales in one model. The model incorporated four latent variables that represent the subscales at measurement time point one and the mean change from wave one to two. The items were allowed to load on the respective factors only. The chi‐square test for comparing a model without measurement invariance to one with measurement invariance was not significant (x2‐difference ¼ 37.89, degrees of freedom df ¼ 35, p ¼ .339). The resulting model with invariant measures over time fits the data well with a Root Mean Square Error of Approximation (RMSEA) of .06, 90% CI [.050, .060], a Comparative Fit Index (CFI) of .98, and a Tucker‐Lewis Index (TLI) of .98. All loadings on the latent variables are significant. The latent variables correlate significantly within one measurement time point and between measurement occasions. Variances of the latent variables are also significant. For psychological and physical bullying, the means of the latent variable at measurement time point two is significant and positive (apsy2‐1 ¼ 0.18, SE ¼ 0.03, p < .001; aphy2‐1 ¼ 0.11, SE ¼ 0.04, p ¼ .007) indicating an increase in psychological and physical bullying over time. For victimization, one scale was used incorporating physical and psychological victimization into one measure. The model incorporated two latent variables

representing victimization at measurement time point one and the change in victimization from measurement time point one to two. The chi‐square difference test revealed that a model assuming measurement invariance is not significantly different to a model without such assumptions (x2‐difference ¼ 32.95, df ¼ 28, p ¼ .238). The fit of the model is moderate with a RMSEA ¼ .07, 90% CI [.06, .07], CFI ¼ .95, and TLI ¼ .95. However, the modification indices indicated no misspecification of the model since they were all below ten. The loadings of the items on the victimization factor are all significant. Victimization at measurement time point one significantly correlates with the change in victimization (cv21 ¼ .51, SE ¼ .04, p < .001). The variances of both latent variables are significant. The mean change in victimization from measurement time point one to two has a negative, but non‐significant, value (av2‐1 ¼ 0.06, SE ¼ 0.05, p ¼ .190), indicating no change in victimization between the two waves of measurement. Longitudinal Prediction Model For the analysis of the longitudinal change in bullying and victimization, and the proposed link between the two variables, the invariant measurement models were incorporated into a structural equation model. The results of the analysis and the structural part of the model can be found in Figure 2 (only significant paths are shown); parameter estimates can be found in Table II. The model

Fig. 2. Results of the longitudinal prediction model. Only significant paths are shown (p < .05), variances and means are not shown. For all parameter estimates, see Table II. Aggr. Behav.

The Interrelation Between Victimization and Bullying

341

TABLE II. Parameter Estimates of the Longitudinal Prediction Model (Structural Part) and 95% CIs Measurement Time Point One (Value [CI]) Latent Variables

Intercepts

Variances

R2h

0a 0a 0a

.59 [.53, .66] .84 [.79, .89] .65 [.58, .72]

The interrelation between victimization and bullying inside young offender institutions.

Bullying and victimization are serious problems within prisons. Young Offender Institutions (YOIs), in particular, suffer from high rates of inmate-on...
422KB Sizes 0 Downloads 3 Views