Economic research and trade policy analysis
Filter :
Language
Publication date
Content type
Series
Collection
General note and abbreviations
The statistics related to applied tariffs and imports are calculated using data which are based on the HS nomenclature adopted by the country for the reference year. For statistics on bound tariffs, the calculations are based on the approved schedule of concessions of the WTO member.
Special topic: Global trade on most-favoured-nation terms
Introduction
The World Tariff Profiles is a joint publication of the WTO, ITC and UNCTAD devoted to market access for goods.
World Tariff Profiles 2025
World Tariff Profiles 2025 provides comprehensive information on the tariffs and non-tariff measures imposed by over 170 countries and customs territories. The publication starts with a breakdown of the tariffs imposed by these economies. Tariff data are presented in comparative tables and in one-page profiles for each economy. A summary table on selected indicators on the imports and exports profile for these economies is also presented. Statistics on non tariff measures by economy and by product group complement the data on tariffs. The special topic deals with “Global trade on most-favoured-nation terms”. The publication is jointly prepared by the World Trade Organization, the UN Trade and Development (UNCTAD) and the International Trade Centre (ITC).
Global Trade Outlook and Statistics
Update: April 2025
The WTO’s “Global Trade Outlook and Statistics” presents the WTO Secretariat’s forecasts for world trade in 2025 and 2026. Breakdowns of merchandise and commercial services trade by sector and region are provided, together with details on leading traders. An analytical chapter discusses the economic effects of trade policy uncertainty. The report is timed to coincide with the release of the WTO’s latest quarterly and annual trade statistics, which can be downloaded from the WTO’s online database at stats.wto.org.
Executive Summary
The comparison of tariffs across time poses significant challenges when data are expressed in different versions of the Harmonized System (HS).
Conclusions
The study highlights that incorporating NLP techniques into HS transposition processes offers substantial potential to enhance the efficiency and accuracy of tariff analysis, making it an invaluable tool for trade statisticians.
Tariff Verification
Both the simple and complex tariff scenarios discussed above indicate that a level of human verification is still ultimately required to ensure the quality of results after the automated tariff transposition, depending on the quality of the original tariffs datasets.
Introduction
Tariff line level Harmonized System (HS) transposition beyond the harmonized 6-digit level has long been a labor-intensive process in trade statistics.
Motivation
The regular work of the Regional Trade Agreements (RTA) section in the Trade Policies Review Division (TPRD) of the World Trade Organization is one among several statistical work streams that regularly require tariff line HS transposition.
Literature Review
Natural Language Processing (NLP) is one among many sub-branches of artificial intelligence (AI) that specifically deals with text as data.
Results
After running both transposition methods, the final output is a transposed tariffs table ("TC" table, in WTO IDB parlance) providing the full set of correlated HS codes in two HS nomenclatures, with each preferential tariff line allocated a corresponding MFN duty code.
Beyond Six Digits: Automated Tariff Line HS Transposition Using Natural Language Processing
This paper explores the application of Natural Language Processing (NLP) techniques to automate Harmonized System (HS) tariff line transposition, employing a three-stage process: unique 1:1 tariff code matching (Round 1), exact description matching (Round 2), and “smart” description matching (Round 3) using Artificial Intelligence (AI) and lexical similarity methods paired with harmonized 6-digit concordance and cosine similarity. Similarity is calculated using either Term Frequency Inverse Document Frequency (TF-IDF) vectors or Sentence-BERT (SBERT) embeddings, comparing two scenarios: a straightforward case (Economy A) with standardized descriptions, and a complex case (Economy B), with more detailed technical descriptions.
Further Work
In future, further refinement and testing on diverse datasets is recommended to optimize these methods for broader application. For instance, other models such as OpenAI's Text Embedding model could be compared with SBERT to compare performance rates.