An Investigation of Models for Under-Reporting in the Analysis of Violence Against Women in Italy

Abstract

Violence against women is still one of the most widespread and persistent violations of human rights. Despite this, a significant gap of comprehensive, reliable and up-to-date figures on such a largely uncovered phenomenon remains. To develop efficient and effective policy and legal responses to gender-based violence, accurate data are necessary. Surveys specifically designed to quantify the number of victims of gender violence return prevalence estimates at a given time, and assess the under-detection of violence and its drivers. However, the last Italian Women’s Safety Survey was conducted by ISTAT in 2014. Given the substantial under-reporting affecting official counts of violence reports to the police, and the lack of recent survey data, up-to-date prevalence estimates cannot be produced. Designing ad hoc techniques suitable to pool data arising from different sources, first of all official police reports, and accounting for the under-reporting, is crucial to understand and measure violence against women to return a realistic picture of this greatly underrated phenomenon and assess its scope. We use publicly available registry data on violence reports in 2020 as a primary source to provide improved estimates of gender violence in the Italian regions, by introducing a Bayesian model that supplements the observed counts with a pool of auxiliary information, including socio-demographic indicators, data on calls from 1522 helpline number and prevalence estimates from previous surveys, while explicitly modelling the reporting process using covariates and external information. We propose using statistical models for the analysis of misreported data to improve the understanding of the problem from a methodological point of view and to get insights into the complex dynamics of the phenomenon in Italy.