Improving crime count forecasts in the city of Rio de Janeiro via reconciliation

Crime prediction based on reliable data and sound statistical methodologies can provide valuable input for tactical deployment of police resources in so-called crime hot spots and effective planning of police operations. In Rio de Janeiro, the Public Safety Secretariat uses criminal activity forecasts to detect crime patterns and evaluate police performance. This paper evaluates the impact of reconciliation on the forecasts of 271 series of registered criminal occurrences in the city of Rio de Janeiro on a monthly basis from January 2003 to December 2019. We verify that reconciliation improves crime count forecasts, especially on the most disaggregated series.