Evaluating accession decisions in customs unions: a dynamic machine learning approach.

Area of Science:

• Economics
• Political Science
• Computer Science

Background:

• Previous research utilized machine learning for static analysis of customs unions, assessing alignment with "natural markets" versus political drivers.
• Existing methods primarily focused on evaluating the composition of customs unions at a single point in time.

Purpose of the Study:

• To develop and apply a dynamic framework for evaluating the evolution of customs unions over time.
• To analyze changes in customs union composition resulting from new member state accessions.
• To assess the alignment of global customs unions with "natural markets" using a dynamic approach.

Main Methods:

• Developed a dynamic framework building upon static machine learning approaches.
• Applied the dynamic algorithm to bilateral trade flow data for 200 countries (1958-2018).
• Evaluated the temporal evolution and accession-driven changes in customs union composition.

Main Results:

• Significant variation exists in how European Union accession rounds align with "natural markets."
• Some EU accession rounds demonstrated a stronger adherence to commercial logic than others.
• Similar patterns of variation were observed in other global customs unions.

Conclusions:

• The dynamic framework provides a more nuanced understanding of customs union evolution beyond static snapshots.
• Findings highlight the interplay between commercial logic and other factors in the formation and expansion of customs unions.
• The study offers complementary insights to static analyses, emphasizing temporal dynamics in regional economic integration.