Back to search

University of Piraeus (UNIPI)

An empirical analysis of determinants for Baltic Dry Index forecasting using neural networks

Abstract

dc:description

The present thesis explores novel determinants of the Baltic Dry Index (BDI) by integrating behavioral, financial, and geopolitical dimensions into maritime economic forecasting. Motivated by the importance of the BDI not only for the dry bulk market but also for various sectors—an aspect extensively discussed throughout the study—it seeks to explain BDI movements beyond the assumptions of rational decision-making in the dry bulk market. More specifically, inspired by the growing use of Google Trends in addressing economic problems, the study examines whether information demand from individual traders—captured through maritime-related search queries—can function as a reliable proxy for dry bulk market sentiment. The findings, based on a Vector Autoregression (VAR) model, reveal a significant negative relationship between the proposed search volume indicators (SVIs), which reflect market uncertainty, and decisions regarding freight rates, eventually guiding the BDI. The effect from the latter on the former is direct, ephemeral, and negative. In contrast, the impact on dry bulk fleet development appears with a lag of approximately six months. These results support risk-aversion theory and challenge the Efficient Market Hypothesis (EMH). Furthermore, a negative bidirectional relationship between sentiment indicators and maritime stock markets is demonstrated, with the effect from the latter being similar to that on the BDI.Building on the established link between the BDI and financial markets, the study further investigates how investor sentiment in both the U.S. maritime and broader stock markets influences BDI dynamics. Utilizing a widely accepted sentiment indicator—trading volume—that has yet to be applied in maritime economics, and employing Granger causality tests and impulse response functions, the results show that investor confidence and market liquidity significantly affect corporate decisions related to freight rates and secondhand vessel investments. By comparing the effects of the general U.S. stock market sentiment index and the maritime stock sentiment index on the BDI, the study reveals the characteristics of investors in the maritime capital market. It also contributes to identification of the roles of financial market sentiment from different trader types. Additionally, it uncovers a new feedback mechanism between the dry bulk and financial markets via investor sentiment, consistent with signal theory. The thesis also addresses the underexplored effect of Chinese Trade Policy Uncertainty (CTPU) on dry bulk shipping indices. Given China’s pivotal role in global shipping as an exporter, shipbuilder, and shipowner, the study analyzes the influence of CTPU on BDI movements using VAR models. Empirical evidence shows that increases in policy uncertainty lead to temporary declines in sea-transportation demand and costs across all BDI sub-indices—effects that are consistent with option theory and persist for approximately one year. Moreover, Artificial Neural Networks (ANNs) are developed, including the examined variables, and compared with benchmark models to assess their performance in predicting BDI movements. Therefore, the study suggests various models for the accurate forecasting of the BDI on a monthly frequency. The out-of-sample predictive power of the examined variables is further evaluated by comparing the suggested models’ performance with the same models that excludes the inclusion of examined variables, reinforcing their relevance as key determinants of the BDI. Overall, the results provide practical implications for maritime stakeholders, financial analysts, and supply chain managers seeking to enhance profitability, manage risk, and promote sustainable decision-making within the shipping industry.

Degree

thesis:*
Grantor dc:publisher
University of Piraeus (UNIPI)
Year dc:date
2025

Author and committee

dc:creator, dc:contributor.*
Authors dc:creator
  • Georgoulas, Dimitrios
  • Γεωργούλας, Δημήτριος

Subjects

dc:subject × 38

Rights

Language dc:language
eng

Identifiers

dc:identifier.*
Identifier
10.12681/eadd/60506

Chain of custody

source
Harvested from
Greek National Archive of PhD Theses
Base URL
phdtheses.ekt.gr/eadd_oai/request
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
citation

Georgoulas, Dimitrios; Γεωργούλας, Δημήτριος. An empirical analysis of determinants for Baltic Dry Index forecasting using neural networks. University of Piraeus (UNIPI), 2025. http://hdl.handle.net/10442/hedi/60506