In a recent article published in the Criminologist, Richard Rosenfeld discussed forecasting national-level crime rates, addressing the complaints about delayed crime statistics released by the FBI and academics. However, Rosenfeld’s forecasts for 2021 through 2025 are unlikely to be accurate due to the inherent nature of point forecasts. Instead, it is more reasonable to provide prediction intervals, which show the potential for misleading interpretations of minor trends. The author provides their own analysis using ARIMA models and compares their forecasts to Rosenfeld’s. They acknowledge the potential for errors in long-term forecasts and emphasize the limited usefulness of macro-level crime forecasts in policymaking. The author also references other studies on crime rate trends for further perspective.
https://andrewpwheeler.com/2023/11/19/forecasts-need-to-have-error-bars/