Forecasting crude oil prices by a semiparametric Markov switching model: OPEC, WTI, and Brent cases

• Published in: Energy economics 2018-08, Vol.74, p.757-766
• Main Authors: Nademi, Arash, Nademi, Younes
• Format: Article
• Language: English
• Subjects: Autoregressive models; Crude oil; Crude oil price; Crude oil prices; Energy economics; Estimation; Forecasting; International economic organizations; Markov analysis; Markov chains; Markov switching AR-ARCH model; Oils; Petroleum; Prices; Pricing; Semiparametric models; Stochastic models; Studies; Switching
• ISSN: 0140-9883; 1873-6181
• DOI: 10.1016/j.eneco.2018.06.020

We use a semiparametric Markov switching AR-ARCH model to forecast the prices of OPEC, WTI, and Brent crude oils. We investigate the applicability of this model based on the proper selection of the core function in the prediction of the crude oil prices. Also, the forecasting ability of this model is compared with different ARIMA and GARCH models for both in-sample and out-of-sample horizons. The period 01-04-2010 to 31-12-2015 is used as the in-sample horizon for the estimation purposes, while the out-of-sample period for the forecasting evaluation is from 04-01-2016 to 15-12-2016. The estimation results show that the semiparametric Markov switching models forecast the crude oil prices more accurately than the ARIMA and GARCH models in both in-sample and out-of-sample horizons (1, 5, 10 and 22 steps ahead) for the studied crude oil prices. •We use a semiparametric Markov switching AR-ARCH model to forecast the prices of OPEC, WTI, and Brent crude oils•The semiparametric Markov switching ARCH model, has interesting properties for scientists to achieve a better prediction.•The results show that the semiparametric models forecast the crude oil prices more accurately than traditional models.