A global prediction model for sudden stops of capital flows using decision trees

Area of Science:

• International Monetary Systems
• Macroeconomic Policy
• Financial Stability

Background:

• Capital flows offer benefits but pose risks to open economies.
• Existing models for predicting sudden stops have limitations in accuracy and scope.
• Current models are primarily developed for emerging countries, lacking global applicability.

Purpose of the Study:

• To develop a novel prediction model for sudden stop events of capital flows.
• To extend prediction capabilities to both emerging and developed countries.
• To accurately estimate future sudden stop scenarios on a global scale.

Main Methods:

• Applied a decision tree method to a sample of 103 countries (73 emerging, 30 developed).
• Utilized sample combinations to account for regional heterogeneity in warning indicators.
• Leveraged decision trees' ability to learn from sequential data and time series for long-term dependencies.

Main Results:

• Achieved excellent prediction results for sudden stop events.
• Demonstrated the model's effectiveness across diverse country groups.
• Highlighted the model's capacity for accurate global scenario estimation.

Conclusions:

• The new model significantly improves the prediction of capital flow sudden stops.
• It provides valuable tools for macroeconomic policy to mitigate financial risks.
• The model contributes to achieving global financial stability.