Sentiment as a shipping market predictor: Testing market-specific language models

This paper applies language models to the shipping market for the first time and studies the impact of changes in shipping market sentiment on freight rates. First, based on language models and Clarksons’ commentary reports, this paper proposes the sentiment indices for the entire shipping market and the sub-markets for bulk ships, tankers, and container ships. Second, empirical results indicate that, apart from the container shipping market sentiment index, all other shipping sentiment indices including the total shipping market sentiment index, the dry bulk shipping market sentiment index and the tanker shipping market sentiment index serve as positive predictive indicators for shipping freight rate indices. Third, this paper investigates the interaction between the shipping sentiment index and market prices through a vector autoregressive model and the Granger causality test. The authors find that the total shipping market sentiment index is the Granger cause of the Baltic Dry Index and the Baltic Dirty Tanker Index. The dry bulk shipping market sentiment index and the container shipping market sentiment index are the Granger causes of the Baltic Dry Index and the China Containerized Freight Index, respectively. Last, this paper compares the shipping sentiment index constructed by market-specific language models and lexicon-based sentiment analysis. It is evident that language models significantly outperform the lexicon-based approaches for sentiment analysis and are expected to be useful for analyzing textual sentiment in the field of asset pricing research.

• Record URL:
  • https://doi.org/10.1016/j.tre.2024.103651
• Record URL:
  • http://www.sciencedirect.com/science/article/pii/S1366554524002424
• Availability:
  • Find a library where document is available. Order URL: http://worldcat.org/issn/13665545
• Supplemental Notes:
  • © 2024 Elsevier Ltd. All rights are reserved, including those for text and data mining, AI training, and similar technologies. Abstract reprinted with permission of Elsevier.
• Authors:
  • Sui, Cong
  • 0000-0002-6023-1885
  • Wang, Shuhan
  • Zheng, Wei
• Publication Date: 2024-9

Language

• English

Media Info

• Media Type: Web
• Features: References;
• Pagination: 103651
• Serial:
  • Transportation Research Part E: Logistics and Transportation Review
  • Volume: 189
  • Issue Number: 0
  • Publisher: Elsevier
  • ISSN: 1366-5545
  • Serial URL: http://www.sciencedirect.com/science/journal/13665545

Subject/Index Terms

• TRT Terms: Freight transportation; Language; Predictive models; Rates; Shipping
• Subject Areas: Data and Information Technology; Finance; Freight Transportation;

Filing Info

• Accession Number: 01925714
• Record Type: Publication
• Files: TRIS
• Created Date: Jul 29 2024 1:41PM