Email circulars as predictive signals in forecasting freight rates

Abstract

This thesis investigates forecasting freight rates within the Baltic Handysize Index (BHSI), concentrating on the HS7_38 Far East to Southeast Asia route. Leveraging both proprietary data – derived from pre-fixture email circulars containing available tonnage (supply) and demand in deadweight tonnage (DWT) – and publicly accessible data, including the Nominal Broad U.S. Dollar Index and Brent crude oil free on board (FOB) prices, the study use a combination of univariate and multivariate time series models. The research aims to evaluate the forecasting performance of these datasets against a naïve benchmark model to test the Efficient Markets Hypothesis (EMH). Their forecasting accuracy is measured using the mean absolute percentage error (MAPE), mean absolute scaled error (MASE), and root mean squared error (RMSE) metrics.

Using various models – the autoregressive integrated moving average (ARIMA) model, the seasonal ARIMA (SARIMA), ARIMA with eXogenous variables (ARIMAX), and seasonal ARIMAX (SARIMAX) – this thesis finds that incorporating both proprietary data and public data can lead to an improvement in forecasting accuracy. Results demonstrate that the multivariate SARIMAX model with all the variables incorporated, outperform univariate and other multivariate approaches in capturing potential underlying market dynamics, seasonality, and trends. The findings underscore the potential value of incorporating email circulars in improving forecasting accuracy, and how freight rates market can be considered inefficient.