Jibrin, Sanusi A. Musa, Yakubu Zubair, Umar A. Saidu, Ahmed S.
Year of Publication:
[Journal:] CBN Journal of Applied Statistics [ISSN:] 2476-8472 [Publisher:] The Central Bank of Nigeria [Place:] Abuja [Volume:] 6 [Year:] 2015 [Issue:] 2 [Pages:] 59-79
The research used a long memory or Autoregressive Fractionally Integrated Moving Average model to study and forecast crude oil prices using weekly West Texas Intermediate and Brent series for the period 15/5/1987 to 20/12/2013. Fractional differencing Methods such as Local Whittle Estimator and Geweke and Porter-Hudak identified long memory characteristics in the crude oil prices. For WTI series, the Bayes Information Criteria selected 3 breaks with the first, second and last breaks captured in 1999, 2004 and 2008 respectively. Three breaks in Brent series using the Bayes Information Criteria were selected and this pointed out that Brent series has break points in 1999, 2005 and 2009. Numerous ARFIMA models were identified, selected using Akaike Information Criterion, estimated/check, in sample and out sample forecast was carried out using Box and Jenkins methodology. ARFIMA(1,0.47,2) is appropriate for West Texas Intermediate series while ARFIMA(2,0.09,0) is suitable for Brent series. One year in sample forecast indicates a small difference between the original series and the forecast results. The one year out sample forecast revealed a decline in future crude oil prices which may be good news to the consumers and bad news to the producers.
ARFIMA model Structural breaks Long memory Local Whittle Estimator and Crude oil prices