Adolescent exposure to and attitudes toward violence: Empirical evidence from Bangladesh

Children and Youth Services Review 98 (2019) 85–95

Contents lists available at ScienceDirect

Children and Youth Services Review journal homepage: www.elsevier.com/locate/childyouth

Adolescent exposure to and attitudes toward violence: Empirical evidence from Bangladesh K.A.S. Murshida, Nadine Shaanta Murshidb, a b

T

⁎

Bangladesh Institute of Development Studies, Dhaka, Bangladesh School of Social Work, University at Buffalo, Buffalo, NY, USA

A R T I C LE I N FO

A B S T R A C T

Keywords: Exposure to violence Attitudes toward violence Adolescents Asia Bangladesh

Children and adolescents are increasingly exposed to violence amidst structural change across the world, raising the question whether development and violence are profoundly connected. Using data from a sample of 520 adolescents from rural and urban Bangladesh, this study explores the nature, structure, prevalence, correlates, and context of exposure to violence in terms of individual, household, and community characteristics. Further, it examines the association between adolescents' exposure to violence and their attitudes toward violence. Exploratory factor analysis (EFA) was used to validate scales of exposure and attitudes toward violence. EFA revealed that the items that indicate exposure to violence can be divided into high exposure items (> 10%) and low exposure items (< 10%) that can be divided into four distinct factors: witnessing garden-variety violence, witnessing severe violence, sexual harassment, and direct experience of violence. The attitude toward violence factors that were salient for this population were fear, disposition to retaliate, and willingness to use a lethal weapon in the process. Additionally, we found that fear was significantly informed by exposure to violence, disposition to retaliate was found to be widespread and significantly higher among girls, while, disposition to use weapons was linked to direct experiences of violence. Different types of violence were found to be associated with different locational and individual characteristics.

1. Introduction Violence is widespread in today's world and persists amidst high economic growth, widening inequality, rapid urbanization and migration, and technological shifts. Violence occurs in both private and public domains and its news and images are carried prominently by both old and new media. Low income countries like Bangladesh are no exception and exposure seem high although there is little systematic evidence on prevalence, nature, structure, tendencies or trends, especially in the context of adolescent experience. This is important to estimate given that exposure to violence normalizes violence, extant research shows, which is socially reproduced and intergenerationally carried over (Ehrensaft et al., 2003; Fehringer & Hindin, 2009; Murshid, 2012). Theory suggests violence is a socially learned behavior; individuals exposed to violence are more likely to have violence-prone schemas and scripts that they learn to use as a tool of conflict management (Bandura, 1978). This occurs through a mechanism that first informs individuals' attitudes about violence. In Bangladesh, research on violence has predominantly focused on violence against women and experience of sexual violence as children

⁎

(Bhattacharya, 2016; Dalal, Lee, Gifford, Institutionen för vård och, & Högskolan i, 2012; Finkelhor, Turner, Shattuck, & Hamby, 2015). The role of exposure to other forms of violence, such as neighborhood, street, or gang violence remain understudied. This is an important line of inquiry given that parents and communities are often unaware of the exposure that their children have to violence, or of the effect that exposure to violence has on their lives. It is also important to locate specific forms of violence (e.g. sexual violence against children) within the broader social context where many forms of violence coexist. The key area of inquiry in this study is to estimate the prevalence of exposure to violence and show how exposure from various sources inform adolescents' attitude toward violence in Bangladesh. This article extends the use of social learning theory in a low-middle-income country context where violence has taken on new forms with religious underpinnings. This examination adds to the knowledge base of a growing body of work that shows that those who are exposed to violence favor a culture of violence (Cristina, Yolanda, Yolanda, Esteban, & Laura, 2018; Sarah, Justin, Daniel Ewon, & Marc, 2015). We argue that exposure to violence acculturates and desensitizes individuals to violence (Tarabah, Badr, Usta, & Doyle, 2016). Having accepting attitudes

Corresponding author at: School of Social Work, 685 Baldy Hall, University at Buffalo, State University of New York, Buffalo, NY 14260, United States. E-mail addresses: [email protected] (K.A.S. Murshid), nadinemu@buffalo.edu (N.S. Murshid).

https://doi.org/10.1016/j.childyouth.2018.12.025 Received 11 September 2018; Received in revised form 24 December 2018; Accepted 24 December 2018 Available online 26 December 2018 0190-7409/ © 2018 Published by Elsevier Ltd.

Children and Youth Services Review 98 (2019) 85–95

K.A.S. Murshid, N.S. Murshid

backward – Gaibandha; 4. Migrants (international) – Comilla; and 5. Migrants (domestic) – Faridpur.

toward violence is associated with increased perpetration and victimization of violence, and other maladaptive behaviors which makes it an important area of study (Caputo, Frick, & Brodsky, 1999; Eckhardt, Samper, Suhr, & Holtzworth-Munroe, 2012). The contribution of this paper is to provide a systematic investigation of the association between exposure to violence and attitudes toward violence among adolescents in Bangladesh using cross-sectional data. It also attempts to unravel key correlates of exposure in terms of individual, household, and area characteristics.

2.2. Exposure to violence Research identifies a host of ways in which violence exposure takes place: conventional crime such as robbery, weapon-assisted attacks, threats of attack, hate crimes; child maltreatment, including physical, emotional abuse, neglect; peer and sibling victimization, including physical and emotional violence as well as threats of violence; sexual victimization, including attempted and completed rape; witnessing and indirect victimization at the family and community level including hearing about violence or hearing the sound of violent acts such as beating, gunshots, and cries; school violence and threat including acid threat and property damage; and internet violence and victimization including threats and sexual solicitation (Finkelhor et al., 2009). Other research on exposure to violence focuses on family and community (Margolin & Gordis, 2000). Examples include intimate partner violence, child maltreatment, assault, robbery, and street fights. More recently, some research has looked at military combat (MacManus et al., 2015), as well as violence on media, such as television, and social media, which may be seen as connected to exposure to military combat as images of combat are often distributed through media (Hopwood & Schutte, 2017). Others have pointed to the persistence of wars around the world over the last few decades as creating a culture of war, conflict, and combat that normalizes violence (Giroux & Evans, 2015). Similarly, scholars speak of a widespread culture of rape and violence against women that dehumanizes certain groups of individuals, namely girls, women, and other minority groups (Phipps, Ringrose, Renold, & Jackson, 2018). Indeed, a process of othering through dehumanization makes violence against certain groups acceptable, even defensible, thus contributing to an overall culture of violence (Rai, Valdesolo, & Graham, 2017). Still others show how community violence has an impact on youth (Fowler, Tompsett, Braciszewski, Jacques-Tiura, & Baltes, 2009). Exposure to violence in Bangladesh occurs at home as well as outside the home. On the one hand, children and adolescents are exposed to intimate partner violence between their parents (Murshid & Murshid, 2018). They also experience violence at home as physical and verbal forms of punishment are acceptable forms of parenting; indeed, violence against children is likely to occur in families in which intimate partner violence occurs (Guedes, Bott, Garcia-Moreno, & Colombini, 2016). On the other hand, children experience violence at school, as many schools use physical punishment as a way to control, punish, and discipline students (Malak, Sharma, & Deppeler, 2015). Meanwhile, they are exposed to community violence as political strife and street violence occur so frequently that they are quite an integral part of the social environment (Salehyan & Hendrix, 2014). This is further intensified by neighborhood and class characteristics such that individuals in low-income neighborhoods are more likely to be exposed to violence. For example, recent research shows that children are running away from home because of poverty, abuse, and family disorganization (Reza, 2016), while IPV is known to be highly prevalent across class and social barriers (Murshid, 2017). Given that all these groups of children and adolescents are exposed to violence in their different spaces, we argue for the use of social learning theory to explain why exposure to violence may inform their attitudes toward violence. Indeed, exposure to violence is associated with negative behavioral and psychological outcomes that may continue into adulthood, including trauma and mental illness (Cook et al., 2017), substance use (Fagan, Wright, & Pinchevsky, 2014), aggressive behavior (Landau et al., 2015), as well as use of violence as adults (Murshid & Murshid, 2018). However, these negative outcomes are more likely when exposure to violence leads to accepting attitudes toward violence, making it important to determine whether exposure to violence has an impact

2. Literature review 2.1. Socio-economic make up Bangladesh has experienced surprising levels of economic and social progress in the last decade amidst persistent challenges (Asadullah, Savoia, & Mahmud, 2014). The last few decades have seen the introduction of neoliberal policies and economic liberalization by way of structural adjustment policies as advocated by the World Bank, the Asian Development Bank, and the International Monetary Fund. Between 1999 and 2009 the annual GDP growth rate averaged at about 6% while poverty rates declined by almost 17%, GDP growth rate increased from 6.2% in 2008 to 7.3% in 2017 (Ahmed and McGillivray, 2014). In the same period, female labor market participation nearly doubled, driven by increasing demand for low-skilled labor in the readymade garments sector. Meanwhile, urbanization and migration of labor – both domestic and international – have created fragmented families with large sections of the population moving from rural to urban areas or going abroad to the Middle East, South East Asia and Europe via legal or illegal means in search of employment as labor in corporations amidst restructuring of the agriculture industry during and in the aftermath of the Green Revolution (Heath, 2014; Kapiszewski, 2017; Robinson, 2015). Meanwhile, school enrollment, particularly of girls have risen considerably, surpassing that of boys, which appear to be linked to female labor participation according to a recent study by Heath and Mobarak (2015). In particular, they suggest, school enrollment of girls is linked to the availability of employment opportunities, for example in the RMG sector that rewards those who have skills and education which in turn creates demand for education for girls (Heath & Mobarak, 2015). Relatedly, fertility rates have declined, with widespread family planning initiatives across the nation that focused on using condoms and contraceptives at reasonable cost (Duvendack & Palmer-Jones, 2017). Research, however, also indicate that school enrollment for girls after primary school starts to decline as adolescent girls start experiencing sexual harassment on the way to and back from school in the absence of safe public transportation. Sexual harassment of adolescents is also associated with early marriage, as parents often feel compelled to marry off their daughters to protect against such violence (Chowdhury, 2015; Jahan, 2018; Schurmann, 2009). Bangladesh experienced increased privatization of state owned enterprises and industrialization with an emphasis on the ready-made garments sector, and rise in family capitalism (Uddin, 2005). These policies were accompanied by increases in economic growth, as mentioned above, but intensified divisions between urban and rural populations. Inequality between regions that were part of the Green Revolution and those that were not persisted, while other pockets of the country have been affected by migration of labor. This suggests that the variation in ‘structure’ and demographics in different regions could result in disproportionate social impact and violence. Therefore, in this study, we sampled adolescents from households from five areas that allow us to examine the differences between these regions, assuming that the socio-economic context of each region is different and thus differentially affect outcomes. They are identified as follows: 1. Urban (slum/non-slum) – Dhaka; 2. Green Revolution – Bogra; 3. Poor, 86

Children and Youth Services Review 98 (2019) 85–95

K.A.S. Murshid, N.S. Murshid

the experiences of youth between the ages of 12 and 19 years from a variety of contexts, we get a picture of whether the structural changes have produced violence in the last 12 to 19 years – violence of all kinds, as opposed to just violence against women and girls. Second, our intervention is theoretical. We examine whether the exposure to different kinds of violence affects various attitudes toward violence using social learning theory; but instead of examining the two broad variables, we provide a closer examination of whether the specific factors or subscales that constitute Martinez-Richters' exposure and Funk and colleagues' attitude scales are associated with each other. We do this after we test the validity of the exposure and attitude scales with a new population to create modified versions of both, which can then be used in other studies. The following research questions guide the current study: (i) What are the types and prevalence of violence that respondents are exposed to? (ii) What are the types of attitudes that respondents have regarding violence? (iii) Does exposure to violence influence respondents' attitude to violence? In other words, we test whether the type of violence plays a role in the resultant attitude, and (iv) What are the correlates of exposure to violence? We expect that exposure to violence allows for observational conditioning and experiential conditioning, when violence is experienced directly, which then inform individuals' attitudes to violence.

on adolescents' acceptance of violence in Bangladesh, where this association has not been examined. Data on violence remain scant in Bangladesh. Currently there are three sources of data on violence: the Bangladesh Bureau of Statistics (BBS), the Demographic and Health Surveys (DHS), and Ain O Shalish Kendra (ASK). BBS data only include information on violence against women that indicate that 80% of women are exposed to violence, compared to 60% as shown by DHS. None of these data sources provide estimates of adolescents' exposure to violence. 2.3. Attitude toward violence The literature on attitude toward violence has grown considerably over the last few decades. Scholars from the 1980s had consistently found gender differences in acceptance of violence, with men being more likely to favor the use of violence than women (Smith, 1984). This disparity remains, both in attitude as well as in use of violence. However, most of the literature focuses on attitude toward violence against women and on intimate relationships, with limited attention to children and young adults (Cohn, Seibert, & Zeichner, 2009;Walters, 2017; Wang, 2016). Attitude toward violence may be informed by factors such as restrictive emotionality, threats to masculinity, trait anger – emotional factors that have been found to be correlated with aggressive behavior (Cohn et al., 2009) as well as education, peer influence, and perceived parental attitudes toward violence (Walters, 2017; Wang, 2016). Others have forwarded a general aggression model that suggest that personal, biological, and situational factors may affect internal state (cognition, affect, arousal, and brain activity) that in turn may affect ability of individuals to appraise and make decisions (DeWall, Anderson, & Bushman, 2011). However, the most salient factor that informs attitude toward violence is exposure to violence, research suggests (Margolin & Gordis, 2000). This is because the social environment has a profound effect on human behavior. The types of violence that are more likely to inform adolescents' attitude toward violence include gun violence (Freeman & Roca, 2001); police brutality(Desmond, Papachristos, & Kirk, 2016); right wing authoritarianism (Benjamin, 2006); violence against women (Caron & Carter, 1997) particularly when exposure occurs within the family or community.

3. Methods 3.1. Data A household level socio-economic survey was conducted between May and December 2017 in five sites, purposively chosen to ensure sample variation that closely matches the nationally representative Household Income-Expenditure Survey: 1. Urban (slum/non-slum) – Dhaka; 2. Green Revolution – Bogra; 3. Poor, backward – Gaibandha; 4. Migrants (international) – Comilla; and 5. Migrants (domestic) – Faridpur. 3.2. Sampling 3.2.1. Sample frame We used a stratified random sampling strategy to select the respondents. From each rural district (Bogra, Comilla, Gaibandha, and Faridpur) we randomly selected five sub-districts (Upazilas), five ‘Unions’ (lowest level administrative tier), and then five villages, and conducted a village census in each to identify households with target respondents who met the inclusion criterion. See Table 1 for sample location and distribution. The inclusion criterion of the sample of the study was that respondents would have to be boys or girls between the ages of 12 and 19 years of age. This list of target respondents served as the sample frame. A total of 10 boys and 10 girls were sampled from the target list for each village yielding a total of 100 respondents per rural location (district). In households where there was more than one adolescent who met the inclusion criterion, the elder adolescent was selected. We also interviewed the heads of household from which the adolescents were selected.

2.4. Conceptual framework The use of violence, when highly prevalent in the social environment, as often is the case in rapidly changing economies, is likely to desensitize individuals, particularly children, to the shock of violence, including shock related to seeing what violence does to people: hurt, injure, maim, and kill (Jeong, 2017; Tarabah et al., 2016). Additionally, according to social learning theory, aggression or violence is a socially learned behavior (R. L. Akers & Jennings, 2015; Bandura, 1978; Oppong, 2014). The key argument scholars forward is the idea that individuals are likely to emulate scripts for which they are positively reinforced (Akers, 2017). In other words, when individuals see violence being used as a tool of control and garnering of power successfully they are more likely to mimic that behavior to achieve the same results. Similarly, if violent behavior is punished or not reinforced, it is less likely to be repeated by those who are exposed to that interaction. Moreover, violence, when normalized through high exposure, both directly and indirectly, makes the use of violence inevitable: it becomes the only salient tool of conflict management in the absence of other tools necessary to resolve conflict (Murshid & Murshid, 2015). Indeed, the repetition of violence is predicated on how exposure to violence inform attitudes toward violence, which is why it is important to assess how exposure to violence affects attitudes toward violence. Our intervention in the literature is two-fold. One, we argue that the economic and social restructuring produces tensions, but there is little knowledge on whether such tensions affect young people. By examining

3.2.2. Urban sample Sixty households were randomly chosen from slum areas in Dhaka (50% girls). A similar number were selected from adjoining non-slum areas (50% girls). The locations included Karail, Bansbari, Kalyanpur, Bhashan Tek, Bauniabad Basti. Thus, our sample consists of a total 400 rural households and 120 urban households. 3.3. Data collection, measures, scales and data Using appropriate ethics protocol that ensures human subject 87

Children and Youth Services Review 98 (2019) 85–95

K.A.S. Murshid, N.S. Murshid

protection, the data collection was conducted at the residences of the respondents by a team of female researchers trained in conducting research with this population. Specifically, the researchers, being from Bangladesh, used culturally appropriate language and social cues. The heads of household, mostly the fathers, or the mothers when the father was absent, consented to their children being interviewed. The heads of household were also interviewed after consenting to being interviewed. In the instances that the head of household was not available, their spouses were interviewed in their stead. Each of the interviews took approximately an hour to complete. The mothers of the adolescents interviewed were kept on hand in case they needed support in case of distress. None of the adolescents, however, required assistance or reported experiencing negative emotions. We used a modified Martinez and Richter Scale (Martinez & Richters, 1993; Richters & Martinez, 1993) in their translated form to measure exposure to violence and a modified Funk et al. Scale (Funk, Elliott, Urman, Flores, & Mock, 1999) to measure attitudes toward violence. The scales were first piloted to ensure that respondents would be able to understand the intent of the questions and modified as necessary. For example, questions about sexual assault were made more specific to make sure everyone understood what that meant, when it appeared to yield conflicting results. We controlled for demographic and socio-economic variables. These data were collected from the heads of household.

Table 1 Sample location and distribution. Area/Type

Location (District)

Sample (50% girls)

Urban Green Revolution area Remote/Backward/Poor Migrant area Low poverty rural area

Dhaka (Capital): slum/non-slum areas Bogra (265 km NW of Dhaka) Gaibandha (265 km N of Dhaka) Comilla (100 km E of Dhaka) Faridpur (127 km SW of Dhaka)

120 100 100 100 100

Table 2a Sample characteristics: Hijab/Nikab ratio. Area/District

Bogra Comilla Dhaka Faridpur Gaibandha

Women (%)

Ratio

hijab

niqab

46.4 58 40.4 57.0 33.3

6 58 14 55 2

12.9 100 34.7 96.5 6.1

Girls (%)

Ratio

hijab

niqab

46.6 46.4 34.2 36.1 33.3

15.5 46.4 24.7 34.4 12.3

33.3 100 72.0 95.5 36.8

seems relatively high in Bogra (19%) and Comilla (16%). The average distance to school is 2.11 km, time spent getting to school is 21 min and the cost incurred is Tk. 4.00. Around 83% of adolescents were either in secondary (47%) or primary (36%) school. Households also reported on the prevalence of hijab-niqab: Among older women, hijab use was reported by 45%, ranging from 33% in Gaibandha and Dhaka to around 58% in Comilla and Faridpur. Among school-going girls, these figures were somewhat lower at 39%, mainly due to relatively lower prevalence in Comilla and Faridpur. The hijab/ niqab ratio is also interesting varying from 34 to 100% (Table 2a). Compared to women, more girls are adopting niqab, we found, especially in Bogra, Dhaka, and Gaibandha.

3.4. Analytic strategy 3.4.1. Scale modification Exploratory factor analysis or EFA is generally used in data reduction when faced with a large number of items in a scale. EFA can also be thought of as a way to measure a latent (unobservable) variable (here susceptibility or vulnerability to violence). The purpose of EFA is to explore as clearly as possible, the underlying structure of the latent variable. The reduced set of factors require careful handling and interpretation and can be used subsequently in econometric analysis as explanatory or dependent variables in aid of further analysis. Thus first, we extract factors to understand the nature and structure of violence and attitudes to violence. Secondly, we ‘explain’ these in terms of individual, household, and area characteristics. Finally, we explore the association between exposure and attitude formation. Principal components option for exploratory factor analysis was used in Stata 12.

4.2. Household profile The five districts vary enormously in terms of basic socio-economic conditions, including poverty and other household characteristics. Thus, Gaibandha has one of the highest poverty rates in the country at 46.7% while low poverty rates are seen in Faridpur and Comilla (7.7 and 13.5%). The poverty rate in Bogra is 27.2% and in urban Dhaka, 10% (HIES, 2016). The traditional joint family structure headed by the eldest male is now changing. On average, around 18% of households are headed by a female with this figure rising to as high as 40% in some areas (e.g. Comilla which is a major migrant sending area). This is also reflected in the decline of the joint-family structure, with only 12.5% of households describing themselves as ‘joint.’ Technology and financial inclusion is expanding quickly with over 90% of households owning a mobile phone and over 33% having a ‘bKash’ account (for mobile financial services). In terms of age, education and occupation, heads of households are in the mid-forties, with 3–4 years of schooling. Interestingly, education of spouse is somewhat higher at around 6 years while average age is just below 40. Around 44% of family heads declared their main occupation either as agriculture or agricultural labor. Migrant household labor patterns were seen to vary enormously – from 3.3% of households in Faridpur to 62% in Comilla (average: 25.3%). Table 2b shows per capita land-ownership, non-land asset ownership, and availability of adult labor in the family. For obvious reasons, land ownership would be small for urban households but for rural households, we see significant variation across sample districts. This is much more subdued for the other variables, including ‘other assets.’

3.4.2. Sample characteristics and associations We computed sample and household characteristics and conducted a series of ordinary least squares (OLS) regressions to test whether factors of exposure to violence were associated with factors of attitudes toward violence. OLS regressions were also conducted to examine the predictors of exposure to violence. We had first conducted bivariate analyses to test bivariate associations and included variables in the multivariate analyses only when they were found to be significant in the bivariate analyses. In the interest of space, they have been omitted from the manuscript. 4. Results 4.1. Profile of adolescents The age-education profile of the sampled adolescents is very similar across districts, with the average age varying between 14.65 years in Faridpur to 15.59 years in Dhaka. In terms of average schooling, again the variation was small between 6.04 years in Faridpur to 7.57 years in Bogra. The gender difference is not noteworthy) In terms of the type of school attended, around 14% went to madrasas (religious schools) while 84% went to mainstream schools (the rest – a small percent – did not go to school at all). Madrasa attendance 88

Children and Youth Services Review 98 (2019) 85–95

K.A.S. Murshid, N.S. Murshid

HIES seems to be higher – however this is likely to be due to different estimation methods, especially when trying to impute “in kind” incomes.

Table 2b Household characteristics: assets and adult labor ratio. Area/district

Land-owned (decimal)/per cap

Other assets/per cap, Tk

Total fixed assets (includes land), Tk (‘000)

Adult labor ratio

Bogra Comilla Dhaka (urban) Faridpur Gaibandha

27.4 15.4 5.0 25.6 23.8

6093 5735 14,536 3252 5262

1927.7 1286.3 1828.1 725.8 1141.6

0.63 0.51 0.66 0.52 0.61

4.3. Nature and extent of exposure to violence The Richter-Martinez Scale (Martinez & Richters, 1993) was deployed with some contextualization to measure exposure to violence. The scale consists of 43 indicators measured on a 5-point Likert scale (1 = never; 5 = usually true). These items were singly examined and also processed through ‘Exploratory Factor Analysis’ (EFA) to determine if there is an underlying structure that allows the items to be collapsed into fewer, more reasonable categories as latent (unobservable) variables related to violence and vulnerability. Of the 43 items, 18 appear to be more prominent in terms of exposure incidence (‘ever happened’), as follows. These indicate high levels of exposure to violence both in terms of witnessing and direct experience, within and outside the home. The pattern of responses seems reasonable in the sense that more violent items tend to be reported at a lower rate. Table 3 shows an alarming level of violence in many forms. Another 16 items yielded low prevalence rates for violence exposure (below 10% prevalence) as seen in Table 4. These generally are more extreme forms of violence, which are expected to occur with lower frequency. These also provide an insight into what are not very important aspects of violence, which could nevertheless emerge as a bigger problem later. Another factor might be that due to the sensitive nature of some items, the lower scores could reflect significant underreporting. In general, the pattern of responses suggest that the scale used performed well with low intensity violence being much more frequently reported compared to high intensity ones in terms of witnessing or experiencing violence.

Land remains the most significant asset in rural Bangladesh with ownership being similar except in places like Comilla where the land constraint is acute (and where migration rates are high as a consequence). Non-land (‘other’) assets are few while labor assets are distributed fairly evenly. The relatively lower prevalence of adults in Comilla and Faridpur correlates with higher migration rates from these areas (See Table 2b). The status of a household is also reflected in housing, access to utilities and services and its water-sanitation facilities. 4.2.1. Housing Floors and roofs are almost uniformly made from mud/earth and corrugated iron sheets, respectively. Walls tend to be corrugated sheets or brick-cement – the latter perhaps depicting superior status (See Table 2c). 4.2.2. Utilities and sanitation The use of ‘modern’ ring-slab latrines has expanded rapidly due to government policy and GO-NGO (government organization—non-government organization) efforts. On the other hand, open defecation is almost non-existent. The main challenge is to convert the remaining ‘traditional’ latrines into slab latrines with a water seal, especially in places like Bogra and Gaibandha which are lagging behind (See Table 2d). It should be noted that ‘traditional’ latrines are often further away from the main house and could pose special difficulties, including increased risk for sexual violence, for children and women, especially when dark (Winter & Barchi, 2016). The drinking water situation in rural Bangladesh was revolutionized by the introduction of tube-wells to exploit rich groundwater sources. Thus, the technology has spread everywhere although in some areas, over-mining has led to arsenic contamination of groundwater supplies (Haque, Islam, & Zahid, 2013). In terms of rural electrification, again it has spread rapidly in recent years with minimum coverage exceeding 80%. Much of this expansion was initially solar based but appears to have become replaced by grid electricity (Table 6). How does the sample relate to the national picture? A comparison with the rural segment of Household Income-Expenditure Survey (HIES, 2016) shows that the sample characteristics are quite consistent with it, e.g. in terms of housing structure, access to electricity and drinking water and so on, indicating that the study sample is fairly representative of the population. The prevalence of traditional latrines however, is much lower for our sample (3–19%), while according to HIES 2016, this figure is 43%. In terms of per capita income levels, the

4.3.1. Exploratory factor analysis (EFA): exposure to violence The objective of EFA is to get at the underlying structure of the ‘latent’ variable that underlies the 43-item scale used. In other words, the items could load onto fewer factors, yielding some broad groupings connected by a similar thread. If these groupings make sense and are easy to interpret then we could say that the structure of the latent variable has been unearthed. Initially, several diagnostics were checked in relation to the scale used to see if these were satisfactory, including Cronbach Alpha, the Keyser-Meyer-Olkin (KMO) test and Bartlett's score. These are briefly detailed below. The Cronbach alpha measures the average inter-item covariance (in this case for 43 items), and this was 0.0356. The ‘scale reliability coefficient’ was 0.8781 while the overall KMO was 0.8563. Thus, in terms of internal consistency, scale reliability and sampling adequacy, our scale appears quite reasonable. If we were to look at each item separately, then some show evidence of inadequate sampling adequacy – generally these items were reported with much less frequency and therefore did not figure very much as factor loadings onto the main factors identified. In addition, Bartlett's test of sphericity was found to be significant suggesting that there is sufficient inter-correlation among variables to allow factor analysis.1 EFA was first conducted using the pcf (principal component factor) option with orthogonal, varimax rotation in Stata. Various other options are available for factor analysis in Stata, including pf, ipf and ml.2 Although pcf seems most widely used, we tried out the other options as well to check if we can get a cleaner, simpler structure with these. The

Table 2c Household characteristics: wall materials and house values. Area

Brick or cement wall (%)

Value of house (Tk.)

Bogra Comilla Dhaka (urban) Faridpur Gaibandha

30 14 51 09 17

120,400 79,730 254,433 59,625 156,090

1 Determinant of the correlation matrix (DET) = 0.00; Bartlett's Test: Chi-sq 8419.3, df = 903 (p value 0.00); HO: Variables are not intercorrelated. 2 pf (default): principal factors; pcf: principal-component factors; ipf: iterated principal factors; ml: maximum likelihood.

89

Children and Youth Services Review 98 (2019) 85–95

K.A.S. Murshid, N.S. Murshid

Table 2d Household characteristics: water and sanitation.

Bogra Comilla Dhaka Faridpur Gaibandha a b

Traditional (‘kutcha’) latrines %

Distance to toilet (m)

Source of drinking water: Tube well %

Electrification %

Grid %

19 03 0 09 19

10.3 03.7 10.9 04.0 08.7

99 100 1.64a 100 100

82 94 100 84 85

93.9 93.6 100 42.9b 87.1

Urban Dhaka depends primarily on tap water supply. Remaining power is from solar home systems.

unearthed). We therefore settled on the pcf option with orthogonal varimax rotation. Coefficients (factor loadings) < 0.4 were dropped. As can be seen below in Table 5, 12 factors were extracted on the basis of the default eigen value cut-off point (1) used in Stata. These together account for 62.5% of the variation in the unobserved latent variable ‘exposure’ in the sample. A scree plot (Fig. 1) is suggested in the literature to determine factors (Costello & Osborne, 2005). This suggests that the first six factors could be relevant. These together account for 43.1% of the variation in the unobserved latent variable. On further examination, e.g. of the items associated with each factor, it became clear that those loading on to factors 5 and 6 have (a) very few items, and (b) have low prevalence rates, and (c) low KMO test scores, suggesting inadequate sampling adequacy. These should be omitted as meaningful factors, thus leaving us with four remaining factors that appear valid (See Table 6). The factor loadings show how each item contributes to the factor in question, and is somewhat analogous to the weights in multiple regression (Yong & Pearce, 2013).

Table 3 High exposure items: prevalence (%). Scale item number

SL no.

%

12 2 9 40 39 23 16 41 13 30

1 2 3 4 5 6 7 8 9 10

1 5 19 43 27 11 6 4 37 3 36 33 25

11 12 13 14 15 16 17 18 19 20 21 22 23

42 17

24 25

38 8

26 27

a

Local hoodlum or gang-member.

Family member hit you Seen someone arrested Saw elders hurt each other Insecure outside neighborhood Insecure in road/school Sexual harassment Family member threatened to beat you Saw any mastanas in the locality? Beaten outside the house Ever saw someone forced to do something with sensitive/sexual organ Heard gunshots nearby Saw elders quarrel Attempt to kiss you (outside) Saw someone steal from home or shop Did you ever hit anyone badly Saw someone draw knife against someone Saw someone struck by knife Saw someone beaten up outside Situation: afraid you may die Saw drugs being sold You know anyone kidnapped? Anyone you know killed? Did you see anyone get stabbed/shot outside house Someone entered your home by force Someone threatened to beat you in public (away from home) Feel insecure at home Saw guns in someone's house

90.2 79.2 75.6 74.4 73.7 64.4 59.8 54.8 46.9 46.3 43.5 39.8 37.9 38.65 37.3 36.5 27.7 32.1 30.4 30.2 25.8 24.6 21.92

4.3.2. Exposure to violence factors 4.3.2.1. Factor 1: witnessing garden-variety violence. A total of 12 items loaded into the factor (with correlation coefficients of > 0.4). These items describe witnessing or hearing generally low level, garden-variety violence (e.g. elderly in the family quarrelling, seeing someone getting beaten, seeing mastans in the locality) – in other words, relating to low intensity, common-place violence. This explains over 13% of the variation.

21.35 19.81 18.8 11.3

4.3.2.2. Factor 2: witnessing severe violence. A total of 10 items loaded into factor 2. These include items relating to witnessing stabbing, shooting, or hearing about kidnap, or being threatened with severe violence. In other words, this factor is best described as hearing or witnessing high-intensity violence. This explains over 11% of the overall variation.

ml option resulted in a ‘Heywood case’ while pf and ipf performed less well as far as factors and associated loadings were concerned (fewer items loaded onto the factors and fewer factors could be meaningfully Table 4 Low exposure items. Item number

SL no.

7 10 14 15 18 20 21 22 24 26 28 29 31 34 33 35

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

% Seen someone get shot Seen someone point a gun at somebody Were you threatened to be killed by anyone Did someone threaten to stab or shoot you? Family member touched you despite your discomfort Seen someone throw acid Are you a victim of acid attack Were threatened with acid attack Used any weapons to threaten or attack somebody? Seen someone hurt by knife or gun attack inside home Family member forced you to do something with sensitive organs As above – by outsider Did you try to force anyone to do something with their sensitive organs? Did your family force you to stay home for any reason Did you see anyone getting killed by another person Did an outsider keep you locked up?

90

3.08 2.50 6.53 1.73 5.96 0.38 0.38 4.23 1.15 2.50 3.08 7.69 1.15 8.65 7.31 1.35

Children and Youth Services Review 98 (2019) 85–95

K.A.S. Murshid, N.S. Murshid

Table 5 Factor analysis using principal component factors.

Factor analysis/correlation Method: principal-component factors Rotation: orthogonal varimax (Kaiser off)

Number of obs = Retained factors = Number of params =

520 12 450

Factor

Variance

Difference

Proportion

Cumulative

Factor1 Factor2 Factor3 Factor4 Factor5 Factor6 Factor7 Factor8 Factor9 Factor10 Factor11 Factor12

5.69899 5.00319 2.96630 1.75987 1.57099 1.52352 1.50010 1.44941 1.44934 1.43855 1.38224 1.12813

0.69580 2.03689 1.20643 0.18888 0.04747 0.02342 0.05069 0.00007 0.01079 0.05631 0.25411 .

0.1325 0.1164 0.0690 0.0409 0.0365 0.0354 0.0349 0.0337 0.0337 0.0335 0.0321 0.0262

0.1325 0.2489 0.3179 0.3588 0.3953 0.4308 0.4657 0.4994 0.5331 0.5665 0.5987 0.6249

LR test: independent vs. saturated: chi2(903) = 8436.04 Prob>chi2 = 0.0000

4.3.2.4. Factor 4: direct experience of violence. Four items loaded into the factor with the correlation coefficients tending to be on the lower side (below 0.6). The items relate to more direct, close experience of violence at home or outside, physical or sexual, explaining 4.1% of the variation. 4.4. Nature and extent of attitudes toward violence The Funk et al. (1999) scale consisting of 17 items was fielded to obtain attitudinal data. The raw item scores are summarized below. The item scores range from 1 to 5 on an ascending scale. Any score above 1 is potentially suspect while a score above 2 can be assumed to imply some sympathy toward violence. It is interesting to see that average scores are low although there are a few that are well above 2, including items 7, 8, 11, 14, 15, 17 (see Table 7). The Cronbach alpha (scale reliability coefficient) at 0.783 and average interitem covariance of 0.1149 are satisfactory. KMO value of 0.783 is also considered reasonable. The Bartlett's test yields a Det (determinant of correlation matrix) of 0.027, a Chi-sq of 1860 (df 136) and p-value = 0.00. This is significant, suggesting sufficient inter-item variation to enable EFA.

Fig. 1. Scree plot of Eigenvalues after factor analysis.

Table 6 List of factors and associated items: exposure scale. Factors

Serial number of scale itemsa

1

2

4

5

12

13

16

27

33

34

41

2 3 4 5 6

1 19 9 20 24

6 23 27 22 28

11 38 29 24 31

16 39 38

25 40

30

36

37

42

43

a

42

4.4.1. Exploratory factor analysis (EFA): attitude toward violence In terms of eigen values six factors were initially extracted by EFA. An examination of the scree plot however suggested three factors which was also confirmed through a visual inspection of the rotated factor loadings. These three factors account for 37% of the variation, and are associated with the following items:

43

4.4.1.1. Factor 1: very fearful.

Scale details are available in annex 1.

Item 14: It is dangerous to hang out with my peers. Item 15: I am afraid I could be hurt from violent actions. Item 16: I am afraid I could even die from violence.

4.3.2.3. Factor 3: sexual harassment. Five items loaded into factor 3 relating to experiencing sexual harassment or “eve teasing” and insecurity.3 This factor explains around 7% of the variation.

3

(footnote continued) designate all manner of public sexual harassment, mostly targeting girls or young women.

This somewhat quaint phrase is widely used in the sub-continent to 91

Children and Youth Services Review 98 (2019) 85–95

K.A.S. Murshid, N.S. Murshid

Table 7 Attitudes toward violence scale (Funk et al.) applied to the Bangladeshi sample. Item

Obs

Mean

Std. Dev.

Min

Max

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17

520 520 520 520 520 520 520 520 520 520 520 520 520 520 520 520 520

1.75 1.576923 1.625 1.419231 1.894231 1.761538 2.203846 2.776923 1.755769 1.538462 2.911538 1.582692 1.696154 2.232692 2.217308 1.926923 2.138462

0.4378525 0.4984042 0.6741511 0.8959421 1.106281 0.6397717 1.240294 1.364371 0.6856395 0.5909249 1.328457 0.7684083 1.092973 1.06154 0.9253903 0.7867191 0.8165329

1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1

3 3 5 5 5 5 5 5 5 4 5 5 5 5 5 5 5

Table 8 Attitude and exposure: OLS regressions. Disaggregated by attitude factors derived from EFA Overall Attitude (aggregate Likert score)

Fearful

Self-defense

Self-defense (with weapons)

β (t ratio)

β (t ratio)

β (t ratio)

β (t ratio)

.049 (.60)

.056 (.78)

−.211 (−2.47)**

.367 (4.48)**

.076 (1.04)

.252 (2.94)**

.175 (4.33)** −.069 (−1.51) −.015 (−.12) −.444 (−1.83)* .491 (2.23)** .381 (2.58}** .027 (.26) .19 16.12 520

−.089 (−2.5**) .005 (.13) −.015 (−.13) −.851 −(3.96)** .900 (4.62)** −.356 (−2.73)** .045 (.41) .37 38.3 520

−.155 (−3.68)** .061 (1.27) .490 (3.63)** −.648 (−2.55)** .935 (4.06)** .160 (1.03) .216 (−1.72)* .11 9.19 520

Independent variables .027 Low level of (1.08) exposure to violence High level .107 exposure to (4.28)** violence Sexual harassment −.015 (−1.22) Direct experience 0.024 (1.73)* D1 (Green Rev) .087 (2.20)** D2 (Migrant) −.156 (−2.10)** D3 (Urban) .511 (7.59)** D4 (Domestic .243 Migrant) (5.39)** Constant −.09 (−.82) Adj R2 .42 F 48.5 N 520

Item 17: I have a feeling people are trying to hurt me. 4.4.1.2. Factor 2: positively disposed to violence. Item 3: Being violent is permissible to attain one's goals. Item 5: Those who are violent get respect. Item 7: If someone hurts you, you should hurt him back. Item 8: If someone speaks ill of me/my family, it is OK to hit him. Item 12: Parents should teach their children to be violent if necessary. 4.4.1.3. Factor 3: positively disposed to use of weapons.

Asterisks denote significance levels: * 10% **5% or above.

indicate ‘fear’ is positively and significantly related (β = 0.175; p < 0.01); while retaliation related variables are negative and significant. In other words, sexual harassment affects fear and tends to discourage retaliation. Direct exposure seems to have a slight impact on overall attitude (β = 0.024; p < 0.05) but this is not seen at the disaggregated level. Locational characteristics that were used as control variables are seen to be highly significant – both positive and negative. Interestingly, urban areas have a positive and significant effect (β = 0.511; p < 0.01) while migrant areas have a negative and significant effect (β = −.156; p < 0.01). It may be noted that the above regression models perform well in terms of the usual diagnostics, as indicated in Table 8. While a clear impact on attitude was uncovered, it should be pointed out that while the primary direction of causality would be from exposure to attitude, it is certainly possible for there to be a feedback loop in the reverse direction. This is difficult to capture without panel data.

Item 6: Feel secure if I carry gun or knife. Item: 9: It is acceptable to keep a gun/knife if you live in an unruly neighborhood. Item 10: It is good to keep a gun with you. Item 12: Parents should teach their children to be violent if necessary. 4.5. Does exposure affect attitude? The predominant attitudes toward violence revealed by the factor analysis were fear, disposition to violence/retaliation, and willingness to use a weapon in the process. We set out to examine whether the exposure factors (witnessing garden variety violence, witnessing severe violence, sexual harassment, and direct experience of violence) have a bearing on the attitudes toward violence. Simple linear OLS regressions were estimated with the aggregate score determining overall attitude toward violence as well as the three attitude factors as independent variables, along with control variables. Table 8 shows the results of the regression exercises where different types of attitude (e.g. overall, fear, being self-defensive, being prepared to use a lethal weapon) are sought to be explained by various components of exposure (low level, high level, sexual harassment, direct experience), with adjusted R-Square ranging from 0.11 to 0.42. Results indicate, generally, that low level of exposure appears not to have any effect on attitude except for ‘self-defense with weapons’, with which it is significantly but negatively associated (β = −2.11; p < 0.01). In other words, low intensity exposure actually seems to discourage retaliation. High intensity exposure generally seems to have a strong effect on attitude formation (β = 0.107, p < 0.01). The signs are all positive, and in only one case (‘self-defense’) the coefficient was not found significant. Sexual harassment, at the aggregate level appears to have no effect on attitude formation. However, at the disaggregated level we see that there are both positive and negative effects. Results

4.6. Socio-economic correlates of exposure We undertook an exploration into the socio-economic correlates of ‘exposure’ as that might be useful in yielding potential policy variables. It can be seen from Table 9 that generally, a higher ratio of adults in the family is not associated with exposure to violence. However, exposure to sexual harassment seems to be higher in families with more adults. This is difficult to explain and requires further research (β = 0.487; p < 0.01). Distance of the toilet from the main house is highly (and positively) correlated with exposure (β = 0.004; p < 0.01), except in the case of high-level or high intensity exposure. In other words, distance of the toilet is probably a reflection of unsafe, relatively isolated spaces in the home which may encourage certain types of behavior. (see See Table 10) for value inflation factors that tests multicolinearity. that 92

Children and Youth Services Review 98 (2019) 85–95

K.A.S. Murshid, N.S. Murshid

Table 9 Socio-economic correlates of exposure (OLS Regressions). Disaggregated by exposure factors derived from EFA

Independent variables Adult lab ratio Toilet distance (meter) Educ*gender (educ in years; gender dummy with female = 1) Land_hh (decimals) Madrasa dummy Female ratio Assets_hh (BDT ‘000) D1 (Green Rev) D2 (Migrant) D3 (Urban) D5 (Poor/remote) Constant Adj R2 F N

Exposure (aggregate Likert score)

Low level of exposure to violence

High level exposure to violence

Sexual harassment

Direct experience

β (t ratio)

β (t ratio)

β (t ratio)

β (t ratio)

β (t ratio)

.024 (.94) .004 (6.55)** .0006 (2.07)** −.00005 (−1.44) .005 (.31) .041 (1.46) −1.51e-08 (−.12) −.037 (−2.17)** .216 (13.25)** .307 (16.63)** −.086 (−5.16)** 1.19 (47.74)** .68 99.6 517

−.142 (1.29) .017 (5.98)** .0001 (.09) −.0001 (−.81) −.021 (−.36) −.373 (−3.19)** −1.35e-06 (−2.57)** .288 (4.07)** .317 (4.67)** 2.20 28.73)** .182 (2.64)** −.459 (−4.41)** .77 162.13

.132 (1.14) .001 (.42) .004 (2.82)** 7.43e-06 (.04) −.041 (−.65) −.297 (−2.42)** 3.08e-07 (.56) −.475 (−6.4)** 1.90 (26.7)** .066 (.82) −.607 (−8.35)** −.162 (−1.48) .76 147.9 517

.487 (2.29)** .027 (4.91)** .002 (.96) −.0005 (−1.63) 1.06 (.91) 1.81 (8.03)** 2.64e-06 (2.61)** −.544 (−3.98)** −.358 (−2.73)** −.54 (−2.66)** −.697 (−5.22)** −1.11 (−5.52)** .17 10.6 517

−.113 (−.56) .035 (6.63)** .0006 (.28) −.0006 (−1.88)* −.30 (−2.69)** −.26 (−1.20) 3.46e-06 (3.60)** 1.03 (7.94)** .618 (4.97)** −.175 (−1.25) .693 (5.45)** −.528 (−2.76)** .26 17.6 517

experience’ (β = 0.26; p > 0.05). On the other hand, there is a positive (significant) association with sexual harassment (β = 0.181; p < 0.01). Like the female ratio, asset ownership, is not associated with aggregate exposure (β = −1.51e-08; p > 0.05) although for most types of exposure (except low level), the association is positive. This is significant for sexual harassment (β = −2.64e-06; p < 0.01) and direct experience (β = 3.46e-06; p < 0.01) – suggesting, assets do not afford much protection either. The locational dummy variables were generally found to be highly significant with both positive and negative associations with exposure type. In other words, the local socio-economic environment has a profound impact on exposure to violence. The most powerful policy variable to emerge from the above is distance to the toilet. Education of girls also affords a degree of protection, especially from high-exposure levels.

Table 10 Test of multicollinearity. Variable

VIF

1/VIF

Dhaka (urban) Assets/hh (BDT Toilet distance (metres) Bogra (GR area) Comilla (migrant area) Land owned (decimal) Age Dummy (madrasa) Dummy (no school) Gender Mean VIF

1.75 1.53 1.28 1.28 1.25 1.15 1.05 1.05 1.04 1.02 1.24

.57 .65 .78 .78 .80 .87 .95 .95 .96 .98

suggests that variables are not correlated. More schooling of girls (reflected by the variable educ*gender) seems to afford partial protection from violence (β = 0.0006; p < 0.01) but not from high level exposure, to which they remain susceptible. Land owned by household is a proxy for wealth but seems not to afford a great deal of protection from violence either, except in the case of ‘direct experience’ with which the relationship is negative and significant (β = −.0006; p < 0.05). Madrasa (religious) schools are often thought to encourage violence but the evidence here does not lend this notion much support. It was not generally found to be significant and for ‘direct experience’ the association was in fact negative and significant (β = −30; p < 0.01). The ratio of females in a household does not seem to have any bearing on aggregate exposure (β = 0.041; p > 0.05). Its relationship, however, is negative and significant with both high exposure (β = −297; p < 0.01) and low exposure levels (β = −373; p < 0.01). It is also negative (but not significantly) associated with ‘direct

5. Discussion The study explored the nature and structure of adolescent exposure and attitude in a rapidly changing socio-economic and demographic context. Much has changed since the independence of Bangladesh. For example, large numbers of girls and women are now in public spaces, many of whom provide labor to local and international corporations, while many girls are in schools, indicating enormous changes in terms of demographics. Additional changes in demographics have occurred through rural-urban and international migration which has resulted in unprecedented flows of resources into rural Bangladesh via remittances sent back. Together, such economic and social shifts have led to new tensions that manifest as violence to which adolescents are frequently exposed, but there remains a lack of concerted efforts in terms of social policy that can rehabilitate and address social problems that may occur due to such exposure (Jenson & Fraser, 2006; Pan, Dexter, & Kristen, 2016; Vira, Franziska, Alistair, Boniface, & Lisa, 2017). 93

Children and Youth Services Review 98 (2019) 85–95

K.A.S. Murshid, N.S. Murshid

and high levels of violence, sexual harassment, and witnessing and/or experiencing sexual and physical violence.

What emerges from this study is a high prevalence of violence that can be broadly divided in terms of four factors or types: witnessing garden variety violence, witnessing severe violence, direct sexual harassment, and direct experience of violence. Some of these are generalized while others tend to have a strong spatial dimension. Meanwhile, attitudes toward violence, when disaggregated by main factors, were found to be predominantly about fear, retaliation, and retaliation with weapons, in that order. Our study also makes clear that attitude factors are affected by different levels and types of exposure to violence, lending some support to social learning theory. For example, our findings indicate that exposure to high levels of violence is significantly associated with adolescents' wishes to retaliate with weapons, suggesting that violence is conceptualized as an antidote to violence. We did not examine the other part of social learning theory, i.e. we do not know whether adolescents used violence in retaliation, but we surmise that among those who wish to retaliate with weapons, some will indeed do so; indeed, wishing to use violence is the first step toward using violence. On the other hand, we found that those experiencing sexual harassment were more likely to feel fear, and that they were less likely to retaliate, including with violence, suggesting that sexual harassment creates a significant level of unsafety and vulnerability that precludes adolescents from feeling like they have control. Based on study results, we also show how socio-ecological factors may have an impact on exposure to violence in the first place. For example, the correlates of exposure reveal significance of individual attributes for some types of exposure and significance of individual, household, and locational variables for others. The most important, systematically significant variable to emerge was distance to toilet, suggesting that certain types of violence are carried out in accessible, presumably private spaces like toilets more readily. This supports a study from Kenya that suggests that distance to toilets and open defecation played a role in stranger rape in Kenya (Winter & Barchi, 2016). Another important finding is regarding the increased odds of sexual harassment when there are more adults in a household and may point to increases in adolescents' time spent outside the home which puts them at further risk. However, this requires further study. The study also suggests that education is a protective factor for girls; this may be because they are better able to gauge danger and are able to employ tactics and strategies, such as travelling in groups, to avoid exposure to violence – but this too requires further study. Finally, this study also serves to validate the exposure to violence and attitudes toward violence scales for Bangladesh, yielding consistent and reasonable findings including factors and correlates which can be used in future studies. The study findings have to be seen in light of several limitations. First, the study is cross-sectional in nature, which disallows causal inference. Second, there is possibility of social desirability bias which may have resulted in either over-reporting or under-reporting of violence. We had, however, made all efforts to minimize such bias by gender matching interviewers with respondents. Future studies would do well to broaden both the spatial and the temporal dimensions of studies such as these. In conclusion, the study has successfully unearthed the nature and structure of violence experienced by adolescent boys and girls in Bangladesh. This is also the first study to systematically examine the association between exposure and attitude to violence along with an examination of key individual, household and locational correlates of violence within a vulnerable sub-group of the population in a developing country setting. Overall findings suggest high prevalence of violence in the social environment, which, we posit, is at the root of creating a culture of fear, while invoking violent responses from those exposed to violence. More broadly, findings highlight how economic development comes with unintended consequences such as increased exposure to violence, which, arguably are most harmful for adolescents. A most telling sign is that almost everyone in the study was exposed to direct or indirect (seen, heard, or experienced) violence, including low

Conflict of interest None. References Ahmed, S., & McGillivray, M. (2015). Human Capital, Discrimination, and the Gender Wage Gap in Bangladesh. World Development, 67, 506–524. https://doi.org/10.1016/ j.worlddev.2014.10.017. Akers, R. (2017). Social learning and social structure: A general theory of crime and deviance. Routledge. Akers, R. L., & Jennings, W. G. (2015). Social learning theory. The handbook of criminological theory (pp. 230–240). . Asadullah, M. N., Savoia, A., & Mahmud, W. (2014). Paths to development: Is there a Bangladesh surprise? World Development, 62, 138–154. Bandura, A. (1978). Social learning theory of aggression. The Journal of Communication, 28, 12. Benjamin, A. J. (2006). The relationship between right-wing authoritarianism and attitudes toward violence: Further validation of the attitudes toward violence scale. Social Behavior and Personality: An International Journal, 34, 923–926. Bhattacharya, H. (2016). Mass media exposure and attitude towards spousal violence in India. The Social Science Journal, 53, 398. Caputo, A. A., Frick, P. J., & Brodsky, S. L. (1999). Family violence and juvenile sex offending: The potential mediating role of psychopathic traits and negative attitudes toward women. Criminal Justice and Behavior, 26, 338–356. Caron, S. L., & Carter, D. B. (1997). The relationships among sex role orientation, egalitarianism, attitudes toward sexuality, and attitudes toward violence against women. The Journal of Social Psychology, 137, 568–587. Chowdhury, E. H. (2015). Rethinking patriarchy, culture and masculinity: Transnational narratives of gender violence and human rights advocacy. Journal of International Women's Studies, 16, 98. Cohn, A. M., Seibert, L. A., & Zeichner, A. (2009). The role of restrictive emotionality, trait anger, and masculinity threat in men's perpetration of physical aggression. Psychology of Men & Masculinity, 10, 218. Cook, A., Spinazzola, J., Ford, J., Lanktree, C., Blaustein, M., Cloitre, M., et al. (2017). Complex trauma in children and adolescents. Psychiatric Annals, 35, 390–398. Costello, A. B., & Osborne, J. W. (2005). Best practices in exploratory factor analysis: Four recommendations for getting the most from your analysis. Practical Assessment, Research & Evaluation, 10, 1–9. Cristina, D., Yolanda, F., Yolanda, A., Esteban, E., & Laura, E. G. (2018). Adolescents at serious psychosocial risk: What is the role of additional exposure to violence in the home? Journal of Interpersonal Violence, 33, 865–888. Dalal, K., Lee, M. S., Gifford, M., Institutionen för vård och, n., & Högskolan i, S. (2012). Male adolescents' attitudes toward wife beating: A multi-country study in South Asia. Journal of Adolescent Health, 50, 437–442. Desmond, M., Papachristos, A. V., & Kirk, D. S. (2016). Police violence and citizen crime reporting in the black community. American Sociological Review, 81, 857–876. DeWall, C. N., Anderson, C. A., & Bushman, B. J. (2011). The general aggression model: Theoretical extensions to violence. Psychology of Violence, 1, 245. Duvendack, M., & Palmer-Jones, R. (2017). Micro-finance, women's empowerment and fertility decline in Bangladesh: How important was women's agency? Journal of Development Studies, 53, 664–683. Eckhardt, C. I., Samper, R., Suhr, L., & Holtzworth-Munroe, A. (2012). Implicit attitudes toward violence among male perpetrators of intimate partner violence: A preliminary investigation. Journal of Interpersonal Violence, 27, 471–491. Ehrensaft, M. K., Cohen, P., Brown, J., Smailes, E., Chen, H., & Johnson, J. G. (2003). Intergenerational transmission of partner violence: A 20-year prospective study. Journal of Consulting and Clinical Psychology, 71, 741. Fagan, A. A., Wright, E. M., & Pinchevsky, G. M. (2014). The protective effects of neighborhood collective efficacy on adolescent substance use and violence following exposure to violence. Journal of Youth and Adolescence, 43, 1498–1512. Fehringer, J. A., & Hindin, M. J. (2009). Like parent, like child: Intergenerational transmission of partner violence in Cebu, the Philippines. Journal of Adolescent Health, 44, 363–371. Finkelhor, D., Turner, H. A., Shattuck, A., & Hamby, S. L. (2015). Prevalence of childhood exposure to violence, crime, and abuse: Results from the National Survey of Children's Exposure to Violence. JAMA Pediatrics, 169, 746. Finkelhor, D., United States. Office of Juvenile, J, & Delinquency, P. (2009). Children's exposure to violence: A comprehensive national survey. Washington, DC: U.S. Dept. of Justice, Office of Justice Programs, Office of Juvenile Justice and Delinquency Prevention U6 – ctx_ver=Z39.88-2004&ctx_enc=info%3Aofi%2Fenc%3AUTF-8& rfr_id=info%3Asid%2Fsummon.serialssolutions.com&rft_val_fmt=info%3Aofi %2Ffmt%3Akev%3Amtx%3Abook&rft.genre=book&rft.title=Children%27s+exposure+to+violence&rft.au=Finkelhor%2C+David&rft.series=OJJDP+juvenile +justice+bulletin&rft.date=2009-01-01&rft.pub=U.S.+Dept.+of+Justice %2C+Office+of+Justice+Programs%2C+Office+of+Juvenile+Justice+and +Delinquency+Prevention&rft.externalDBID=HJZ& rft.externalDocID=003246864¶mdict=en-US U7 – eBook. Fowler, P. J., Tompsett, C. J., Braciszewski, J. M., Jacques-Tiura, A. J., & Baltes, B. B. (2009). Community violence: A meta-analysis on the effect of exposure and mental

94

Children and Youth Services Review 98 (2019) 85–95

K.A.S. Murshid, N.S. Murshid

Murshid, N. S., & Murshid, N. (2015). Intergenerational transmission of marital violence results from a nationally representative sample of men. Journal of Interpersonal Violence, 33(2), 211–227. https://doi.org/10.1177/0886260515604413 0886260515604413. Murshid, N. S., & Murshid, N. (2018). Intergenerational transmission of marital violence: Results from a nationally representative sample of men. Journal of Interpersonal Violence, 33, 211–227. Oppong, S. (2014). Between Bandura and Giddens: Structuration Theory in Social Psychological Research? Psychological Thought, 7, 111–123. Pan, C., Dexter, V., & Kristen, J. (2016). Community violence exposure and adolescent delinquency: Examining a spectrum of promotive factors. Youth & Society, 48, 33–57. Phipps, A., Ringrose, J., Renold, E., & Jackson, C. (2018). Rape culture, lad culture and everyday sexism: Researching, conceptualizing and politicizing new mediations of gender and sexual violence. Journal of Gender Studies, 27, 1–8. Rai, T. S., Valdesolo, P., & Graham, J. (2017). Dehumanization increases instrumental violence, but not moral violence. Proceedings of the National Academy of Sciences, 114(32), 8511–8516 201705238. Reza, M. H. (2016). Poverty, violence, and family disorganization: Three “Hydras” and their role in children's street movement in Bangladesh. Child Abuse & Neglect, 55, 62–72. Richters, J. E., & Martinez, P. (1993). The NIMH community violence project: I. Children as victims of and witnesses to violence. Psychiatry, 56, 7. Robinson, V. (2015). Bridging the gulf: The economic significance of south Asian migration to and from the Middle East. Return migration and regional economic problems (Routledge Library Editions: Economic Geography) (pp. 243). . Salehyan, I., & Hendrix, C. S. (2014). Climate shocks and political violence. Global Environmental Change, 28, 239–250. Sarah, A. S., Justin, E. H., Daniel Ewon, C., & Marc, A. Z. (2015). Predicting violent behavior: The role of violence exposure and future educational aspirations during adolescence. Journal of Adolescence, 44, 191–203. Schurmann, A. T. (2009). Review of the Bangladesh female secondary school stipend project using a social exclusion framework. Journal of Health, Population and Nutrition, 27, 505–517. Smith, T. W. (1984). The polls: Gender and attitudes toward violence. The Public Opinion Quarterly, 48, 384–396. Tarabah, A., Badr, L. K., Usta, J., & Doyle, J. (2016). Exposure to violence and childrens desensitization attitudes in Lebanon. Journal of Interpersonal Violence, 31, 3017–3038. Uddin, S. (2005). Privatization in Bangladesh: The emergence of ‘family capitalism. Development and Change, 36, 157–182. Vira, A., Franziska, M., Alistair, M., Boniface, U., & Lisa, L. (2017). Associations between adolescent experiences of violence in Malawi and gender-based attitudes, internalizing, and externalizing behaviors. Child Abuse & Neglect, 67, 305–314. Walters, G. D. (2017). Viewing the cycle of violence through a gendered pathways lens: Perceived parental tolerance of violence, peer influence, and child aggressive behavior. Journal of Interpersonal Violence0886260517702493 Available online only. Wang, L. (2016). Factors influencing attitude toward intimate partner violence. Aggression and Violent Behavior, 29, 72–78. Winter, S. C., & Barchi, F. (2016). Access to sanitation and violence against women: Evidence from Demographic Health Survey (DHS) data in Kenya. International Journal of Environmental Health Research, 26, 291–305. Yong, A. G., & Pearce, S. (2013). A beginner's guide to factor analysis: Focusing on exploratory factor analysis. Tutorials in Quantitative Methods for Psychology, 9, 79–94.

health outcomes of children and adolescents. Development and Psychopathology, 21, 227–259. Freeman, T. W., & Roca, V. (2001). Gun use, attitudes toward violence, and aggression among combat veterans with chronic posttraumatic stress disorder. The Journal of Nervous and Mental Disease, 189, 317–320. Funk, J. B., Elliott, R., Urman, M. L., Flores, G. T., & Mock, R. M. (1999). The attitudes towards violence scale: A measure for adolescents. Journal of Interpersonal Violence, 14, 1123–1136. Giroux, H. A., & Evans, B. (2015). Disposable futures: The seduction of violence in the age of spectacle. City Lights Publishers. Guedes, A., Bott, S., Garcia-Moreno, C., & Colombini, M. (2016). Bridging the gaps: A global review of intersections of violence against women and violence against children. Global Health Action, 9, 31516. Haque, M. A., Islam, M. S., & Zahid, A. (2013). Groundwater irrigation and crop economy in the lower Gangetic plain at Matbarer Char, Madaripur, South-Central Bangladesh. Journal of the Asiatic Society of Bangladesh, Science, 38, 29–39. Heath, R. (2014). Women's access to labor market opportunities, control of household resources, and domestic violence: Evidence from Bangladesh. World Development, 57, 32–46. Heath, R., & Mobarak, A. M. (2015). Manufacturing growth and the lives of Bangladeshi women. Journal of Development Economics, 115, 1–15. Hopwood, T. L., & Schutte, N. S. (2017). Psychological outcomes in reaction to media exposure to disasters and large-scale violence: A meta-analysis. Psychology of Violence, 7, 316. Household Income Expenditure Survey (2016). Bangladesh Income Expenditure Survey. (12/24/2018.) Retrieved from: https://www.ceicdata.com/en/country/bangladesh. Jahan, R. (2018). Men in seclusion, women in public: Rokeya's dream and women's struggles in Bangladesh. The challenge of local feminisms (pp. 87–109). Routledge. Jenson, J. M., & Fraser, M. W. (2006). A risk and resilience framework for child, youth, and family policy. Social policy for children & families: A risk and resilience perspective (pp. 118). . Jeong, H.-W. (2017). Peace and conflict studies: An introduction. Routledge. Kapiszewski, A. (2017). Arab versus Asian migrant workers in the GCC countries. South Asian migration to Gulf Countries (pp. 66–90). Routledge India. Landau, S. F., Dvir-Gvirsman, S., Huesmann, R., Dubow, E. F., Boxer, P., Ginges, J., et al. (2015). The effects of exposure to violence on aggressive behavior: The case of Arab and Jewish children in Israel. MacManus, D., Rona, R., Dickson, H., Somaini, G., Fear, N., & Wessely, S. (2015). Aggressive and violent behavior among military personnel deployed to Iraq and Afghanistan: Prevalence and link with deployment and combat exposure. Epidemiologic Reviews, 37, 196–212. Malak, M. S., Sharma, U., & Deppeler, J. M. (2015). Can I really teach without my magic cane?’Teachers' responses to the banning of corporal punishment. International Journal of Inclusive Education, 19, 1325–1341. Margolin, G., & Gordis, E. B. (2000). The effects of family and community violence on children. Annual review of psychology, 51, 445–479. Martinez, P., & Richters, J. E. (1993). The NIMH community violence project: II. Children's distress symptoms associated with violence exposure. Psychiatry, 56, 22. Murshid, N. S. (2012). Intergenerational transmission of domestic violence in Bangladesh. Journal of Bangladesh Studies, 14, 20–33. Murshid, N. S. (2017). Men's report of domestic violence perpetration in Bangladesh: Correlates from a nationally representative survey. Journal of Interpersonal Violence, 32, 290–307.

95