Merih Angin - IMF Decides, Machine Learns: An AI Approach to IOs
Türkiye, Istanbul
Study location | Türkiye, Istanbul |
---|---|
Type | Summer Research Program - Undergraduate, full-time |
Language requirements | English |
---|
Other requirements | At least 2 reference(s) must be provided. |
---|
Overview
This project develops a new model by utilizing a wide literature on the IMF to present a comprehensive model that takes into account all actors influencing the design of IMF programs. The following questions will be at the heart of this research: Which factors influence the terms of an IMF program (loan size, number of conditions, and granted conditionality exemptions)? And how do these factors play a role in shaping program design? In this context, the project will focus on three aspects of IMF loans: (1) the size of IMF loans; (2) the number of conditions attached to loans; and (3) the number of conditionality exemptions granted to borrowing countries during program implementation. This research, creating a comprehensive and new methodology to understand IMF program design, will shed light on the processes leading to diversity in the conditions of IMF loans. At the same time, by creating an original model, this project aims to provide a significant and scalable tool for international political economy researchers and policy makers to model the program design and implementation processes with high predictive power. The project aims to make a significant contribution to the literature by creating a comprehensive machine learning (ML) model to predict the loan size, the number of IMF conditions, and the number of conditionality exemptions during program implementation. This model will complement traditional statistical models by integrating a greater number of variables and providing high prediction accuracy. The research will also create a natural language processing (NLP) tool capable of automatically and quickly analyzing the minutes of IMF Executive Board meetings, capturing alliances between representatives of different countries and the stance of the G5 with high accuracy, and recognizing the sentiments of individual Executive Board members. Another contribution of the project will be the creation of a comprehensive dataset on the various factors affecting IMF program design, covering IMF programs between 1978-2014. Examining IMF behavior while collaborating with European Union institutions, focusing on the cases of Romania and Greece, will be the fourth contribution provided by the project to illuminate existing causal mechanisms.