Forecasting crude oil price: Does exist an optimal econometric model?

The drastic reduction in oil prices after 2014 rekindled its stochastic characteristics of not settling around a mean and having unexpected high volatility. Thus, creating a branch of empirical literature devoted to the study of structural breaks in oil price longitudinal data, its treatment and forecasting. In that regard, this paper estimate and compare the accuracy measurements of different methodologies and propose the use of a Self-Exciting Threshold Auto-regressive - SETAR model. This approach automatically allows for regime switching after a threshold, hence achieving a Root Mean Square Error - RMSE of 2%, in contrast to 10% of other models commonly used. Moreover, the comparison with previous studies pointed out that the SETAR model surpasses most of the oil price prediction methods in relation to its accuracy, or because of its simplicity, since it does not require great computational effort or difficult analytical skills. •This paper compares and contrasts different approaches of oil price forecasting.•Our results provide empirical evidence of treatment and analysis of structural breaks.•Proposing the use of a SETAR model for oil price forecasting.•The SETAR model outperforms others models in forecasting accuracy.•SETAR does not require extensive knowledge in programming, Big data manipulation and Statistics.