Beyond the Dark Figure: A Causal and Probabilistic Analysis of Crimes Against Women in India

Most research on violence against women in India depends on administrative police records. However, this raw data often suffers from severe reporting biases and fails to account for massive population differences between districts. To address these flaws, this study applies causal inference, hierarchical modeling, and machine learning to better understand local crime dynamics. By combining two decades of National Crime Records Bureau (NCRB) statistics with 2011 Census demographics, we map both the spatial distribution and the true sociological drivers of this violence. We found no evidence of geographic clustering; a Moran's I analysis shows the violence is distributed randomly across the country. Furthermore, our mixed-effects hierarchical models indicate that broad state-level policies explain less than 1% of the variance in local crime rates. Furthermore, by applying Pearl's do-calculus and Directed Acyclic Graphs (DAGs), we isolate the causal effects of the male-female literacy gap and urbanization, proving that urbanization is the dominant driver of reporting rates. A Bayesian Structural Time Series (BSTS) correctly isolates the passage of the 2013 Criminal Law (Amendment) Act as an administrative reporting shock yielding an 88,879 annual case surge rather than a behavioral epidemic. Finally, Bayesian Belief Networks are utilized to model highly nonlinear sociological interactions, providing a mathematical framework for proactive risk assessment.