Forecasting spot rates at main routes in the dry bulk marketForecasting spot rates at main routes in the dry bulk market
Faculty of Applied Economics
Transport and Regional Economics
Maritime economics & logistics. - Basingstoke, 2003, currens
14(2012):4, p. 498-537
University of Antwerp
The dry bulk shipping market is a major component of the international shipping market and it is characterized by high risk and volatility, in view of the uncertainty caused by factors such as the global economy, the volume and pattern of seaborne trade, and government policies. In such markets, to model price behavior (of spot- or time charter rates) has always been a topic of great interest among researchers. This article makes an attempt to forecast spot rates at main routes for three types of dry bulk vessels and to find superior forecasting models that can provide better forecasts. In this article, 1-month change in the Baltic Index, representing the market sentiment, is firstly invented and incorporated into the forecasting models, and this indicator is found to be very helpful in improving prediction performance. Furthermore, some significant exogenous variables are also employed to improve forecasting performance. The results of the cointegration test reveal that there are no long-run relationships of spot prices between trading routes for all three ship sizes. Hence, except a vector error correction model, time series models, such as the ARIMA, ARIMAX, VAR and VARX, are employed in this article to make the prediction. All spot prices cover the period from January 1990 to December 2010, which is split into an estimation period and an out-of-sample forecasting period. In order to test whether the market since 2003 is significantly different from the market before, the in-sample estimation is made over two sample periods. Various models are estimated firstly over the whole period from January 1990 to June 2009, and then estimated again over the second period from January 2003 to June 2009 at all routes for three ship sizes. The period from July 2009 to December 2010 is then used to evaluate independent out-of-sample forecasts. The forecasting performance of various forecasting models is evaluated and the comparison of the forecasting capabilities between various models provides useful information in the selection of superior forecasting models, which can yield better forecasting results.