Table 5.

Characteristics of included studies

Study	Study objectives	Country	Methods of data collection	Data analysis	Correlates assessed
Abramovay et al. (1999) Chapter 3	Examine the prevalence, characteristics and behaviours of youth gangs.	Ceilandia, Planaltina and Samambaia, Brazil	A self‐report survey of 809 youth aged 15‐24 in selected residential blocks in 3 cities (data used in this review) Focus groups, interviews and case studies (qualitative data not used in this review)	For the purposes of this review we focus on the proportional differences between gang‐involved and non‐gang‐involved youth across the characteristics of interest. The authors used a mixed methods analysis, but no statistical analyses were reported. (Gang involved = present and past gang members)	Age Sex Work Education Live with both parents Violence in the family Engaged in violent situations Drug use
Celbis et al. (2012)	“To determine the prevalence of violence‐related behaviours on school property and to identify the predictors of youth violence among high school adolescents” (p. 343).	Malatya, Turkey	Self‐report questionnaire with 1175 students (747 males, 428 females). Cross‐sectional design, using stratified random sampling 6 urban high schools and 1 non‐urban high school in Malatya.	Authors reported the prevalence of violent behaviours (including gang membership) by key characteristics. A logistic regression model for violent behaviour was also reported but did not allow effect size calculation.	Sex Grade Family income School type, location Mother's education Success in school Exposure to violence (home, school, neighbourhood)
Katz and Fox (2010)	Explore prevalence of gang involvement and identify risk and protective factors associated with youth gang involvement.	Trinidad and Tobago	Data is drawn from the Trinidad and Tobago Youth Survey (TTYS), a self‐report survey completed by 2,206 students across 22 “high‐risk urban public schools” (892 males, 1,314 females).	Authors reported the prevalence of gang involvement by key characteristics, and a multinomial analysis of risk and protective factors for gang involvement (categories = never in gang, gang associate, former member, current member).	Gender, age, ethnicity Mobility, availability of handguns Commitment to school Parental attitudes Antisocial behaviour/peers Drug use/peer drug use Rewards and opportunities for prosocial behaviour Social skills, belief in moral order
Moravcová (2012)2	Assess the associations between different definitions of youth gangs and individual, social, and behavioural factors.	Czech Republic	The International Self‐Report Delinquency survey (waves 2 & 3) captured data from students in Grades 7‐9 at private and public schools (ages 12‐16; N = 6,707).	Authors performed a multinomial regression to predict an individual being categorised as a member of: a non‐gang group, a Eurogang defined gang, a self‐identified gang member, or a gang member identified on the Mokken scale. Authors reported the prevalence of 11 forms of delinquency across groups.	11 forms of delinquency Gender Age Self‐control index Personal morality index Family structure Family bonds index Truancy Risk behaviour index
Ohene et al. (2005)	Identify associations between risky behaviours and initiation of sexual activity among youth between 10‐18 years of age.	Antigua, Bahamas, Barbados, British Virgin Islands, Dominica, Grenada, Guyana, Jamaica and St Lucia (Caribbean)	Data is drawn from the Caribbean Youth Health Survey (random sampling), a self‐report survey completed by 15,695 school attending adolescents aged 10 to 18 (39% males, 61% females).	Relationships were assessed using odds ratios, stratified by gender and age group (full results were only reported stratified by gender). Statistical significance was reported. Survival analysis of factors associated with sexual initiation was also employed but did not allow effect size calculation.	Cigarette, alcohol, marijuana use Weapon carrying Runaway behaviour Skipping school Early sexual initiation
Olate et al., (2011)	A cross‐cultural comparison of high‐risk youth which examines the differences on externalizing risk behaviours in domains of school, work, sexual behaviours, substance abuse, and violence and delinquency.	Boston, USA and San Salvador, El Salvador	A cross‐sectional survey of members of two youth organisations in Boston (N= 374; 115 gang members) and one organisation in San Salvador covering four municipalities (N=208; 135 gang‐involved youth including 12 females). Mean age of El Salvadorian respondents was 20. (El Salvador data used in this review)	Authors reported the means and proportional differences between youth gang members and non‐members across five domains of externalising behaviours. T‐tests and chi‐square tests of significance were also reported.	School attendance measures Employed/employed FT Sexual behaviour Drug and alcohol use Measures of violence and delinquency (self‐report and official data)
Olate et al. (2012)	Examine the association between several risk factors and violence and delinquency in youth gang members and high‐risk youth.	San Salvador, El Salvador	Cross‐sectional survey using a non‐probability sample (N= 174) drawn from 10 urban and semi‐urban neighbourhoods within four Greater San Salvador Metropolitan municipalities. Administered by interviewers in individual or group format. Sample included 58 high‐risk non‐gang youth (13‐23 years; 36 male, 22 female) and 116 gang‐involved youth (13‐24 years; 106 male, 10 female). Appears to use the same data as Olate et al., 2011, so was treated as dependent.	Authors reported the means and proportional differences between youth gang members and non‐members across selected characteristics. T‐tests, chi‐square tests, and a correlation matrix were also reported. A logistic regression model predicting violence and delinquency was presented but did not allow effect size calculation.	Age, gender, parenthood status Measures of violence and delinquency Impulsivity, hope, empathy Unstable home, difficulty at home Expelled from school Delinquent peers Neighbourhood disorder Social support
Pyrooz and Decker (2013)	Examine the association between youth gang involvement and delinquent behaviour	Changzhi, China	Self‐report data collected from a school‐based convenience sample of 2,245 youth across six schools (mean age: 17.47; 1298 males, 865 females).	Authors reported the means and proportional differences between youth gang members and non‐members across selected characteristics. Two logistic regression models predicting offending were presented but did not allow effect size calculation.	Age, gender, minority status Broken home, parents’ education, rural Delinquency/delinquent peers Self‐control Family attachment School performance Parental monitoring Household strains
Webb et al. (2011)	A cross‐cultural comparison of the prevalence of gang involvement and the correlates of involvement for school‐age youth in China.	Hangzhou, China and a representative sample of five towns/cities in the United States.	The International Self‐Report Delinquency survey captured data from students in Grades 7‐9 at private and public schools (ages 12‐15; China N = 1,043; US N = 2,401).	Authors reported the proportional differences between youth gang members and non‐members across lifetime and last‐year prevalence of offending and victimisation. Chi‐square tests of significance were conducted.	Lifetime and last‐year prevalence of offending behaviour (drug, delinquent, criminal) Lifetime and last‐year prevalence of victimisation (robbery/extortion, assault, theft, bullying).

²Data was not available to calculate effect sizes, therefore this study was not included in the meta‐analyses.