Forecasting the road traffic fatality rate in US based on autoregressive integrated moving average models

Road traffic fatalities (RTF) are a leading cause of death in the US. To come up with effective interventions, there is a need to recognize how RTF will change and grow over time. In this paper, a Box-Jenkins model is proposed for predicting the time trends of the RTF rate in the US. In this regard, the monthly time trends of the RTF rate from January ۱۹۹۹ to December ۲۰۱۸ are employed. The best models are selected based on the lowest AIC criterion. To confirm the un-correlation, stationarity, and zero mean of the residuals, the Ljung-Box (LB) test, Kolmogorov- Kolmogorov-Smirnov (KS) test of normality, and residual plots are applied. Besides, the out-of-sample prediction validity of the SARIMA model is checked by comparing the results to the outcomes of the classical smoothing models. The results indicate that the SARIMA (۲,۱,۰) (۱,۰,۱) and SARIMA (۰,۱,۱) (۱,۰,۱) outperform the exponential smoothing methods in out-of-sample prediction based on the lowest MAPE values. In addition, the models revealed good fits based on the confirmed white noise characteristics of the residuals. The developed instruments could be employed by policymakers to evaluate the intervention’s effectiveness, and to forecast the future time trends of RTF rate in the US.