Forecasting oil freight rates using crude oilprices : A VAR-based analysis

Sammanfattning

In this bachelor thesis, the purpose is to examine how accurately a Vector Autoregression (VAR) model can forecast the Baltic Dirty Tanker index (BDTI) using crude oil prices, compared to a Random Walk. The data used in the thesis are crude oil prices and BDTI observations from January 2015 to March 2025. Both series are modeled in first differences after stationary testing. A VAR(12) model's forecasting performance is then compared to a Random Walk benchmark over the 2024-2025 period using a one week expanding window approach. The main results show that the VAR model forecast outperforms the Random Walk in forecasting the BDTI. The VAR achieves a Mean Absolute Percentage Error of 2.89 percent compared to 3.16 percent for the Random Walk, an 8.54 percent improvement. The Diebold-Mariano test confirms that these improvements are statistically significant achieving a p-value of 0.0002. These findings confirm that crude oil prices contain useful predictive information for oil tanker freight rates, which offers practical value for shipping market participants.