Spatial Analysis of Child Violence Victims in West Java in 2024 Using Geographically Weighted Negative Binomial Regression

Abstract

Violence against children remains a critical issue in Indonesia, with West Java consistently reporting high numbers of reported child violence victims. This study examines socioeconomic factors associated with the count of reported child violence victims across 27 districts and cities in West Java in 2024, using secondary administrative data obtained from Open Data Jabar. The explanatory variables include poverty rate (X1), average years of schooling (X2), number of divorce cases (X3, Labor Force Participation Rate (X4), and Open Unemployment Rate (X5). Diagnostic tests indicate the presence of spatial heterogeneity and overdispersion, supporting the application of a Geographically Weighted Negative Binomial Regression (GWNBR) model with a child population offset. Model performance comparison based on in-sample fit criteria shows that the GWNBR model provides superior fit (deviance = 42.94, AICc = 99.87) compared to the global Negative Binomial Regression model (deviance = 48.27, AIC = 105.63). The GWNBR results reveal substantial spatial variation: average years of schooling (X2) is statistically significant across all 27 regions, while the number of divorce cases (X3) is significant in 23 regions. Poverty rate (X1) shows localized significance in 16 regions. Labor force participation rate (X4) and unemployment rate (X5) each exhibit significance in 6 regions, though with distinct spatial patterns. These findings highlight geographically varying risk structures that cannot be adequately captured by global models and underscore the importance of spatially adaptive modeling for informing region-specific child protection policies. Although the analysis relies on reported administrative data that may not fully represent the true underlying prevalence of child violence, the results provide valuable spatial insights relevant to policy development aligned with SDG 3, SDG 4, and SDG 16.