An application of non-linear methods to the prediction of future freight rates, 2006-2008

摘要

The paper deals with the major question whether freight rate derivatives are the most accurate indicators of future freight rates. The paper proves that predicting future freight rates using non-linear models one obtains more accurate estimates. Derivatives proved to give underestimates of the actual spot freight rates within the sample. The paper goes one step further to show that the failure outlined above is due to the existence of black and pink noise in the time series. This is so as models so far assume white noise even when tests for normality, excess skewness and kurtosis are significant much above their critical values. The data covered 694 working days for 2006-2008 for derivatives and spot rates for a dry cargo vessel of 85 000 dwt. The Hurst exponent found 0.90 for spots and 0.80 for derivatives. The paper's predictions were below 10% of the actual for ten steps ahead. Derivatives were away by 28%.