Identifying Important Factors to Prevent Juvenile Delinquency among Male and Female Adolescents: an Exploratory Analysis Using the LASSO Regression Algorithm in the Korean Children and Youth Panel Survey (KCYPS)

Juvenile delinquency is the outcome of complex interactions with multiple factors. However, few studies have explored what factors most likely contribute to delinquent behavior among female and male adolescents when all possible levels of factors are included in one model. To fill this gap, the current study investigated what factors were associated with juvenile delinquency and which factors were appeared to be significant in both female and male adolescents using machine learning algorithms. This information can be particularly informative for policymakers and researchers to capture the overall feature of delinquency. Data were derived from three-time points (8th grade, 9th grade, and 10th grade) of the nationally-representative Korean Children and Youth Panel Survey (KCYPS). The sample consisted of an almost equal number of males and females. This study employed the Least Absolute Shrinkage and Selection Operator (LASSO) regression as an exploratory analysis to identify important factors related to juvenile delinquency and compared coefficients of each factor in the model. The results showed that individual factors, including cyber delinquency, aggression, romantic relationships, following school rules, engagement in learning activities at school, academic confidence in Korean, English, and math, relationships with teachers, peer relationships, social withdrawal, and mobile phone dependency, were relatively important factors of delinquency for both females and males. These findings suggest that using LASSO regression to identify the most important factors for juvenile delinquency will provide an opportunity to understand the complex phenomenon of delinquency among female and male adolescents and can be a useful source in delinquency prevention policies in Korea.