Application of Machine Learning to International Market and Trade Analysis

The main focus of the three studies in this dissertation is the application of advanced techniques including machine learning methods and finer data to model economic relationships: price, demand and trade patterns. Agricultural products and food are absolutely necessary for human survival, and unexpected shocks pose a threat to food availability and access in many developing countries. Along with the growth of population and income, trade in agricultural commodities has witnessed dramatic increases in recent years. In the first chapter, the role of international price volatility and inventories on domestic market price dynamics in the case of agricultural commodities is investigated using Least Absolute Shrinkage and Selection Operator (LASSO) methods. The second chapter attempts to explain global beef trade by comparing traditional econometric methods, i.e., Poisson Pseudo Maximum Likelihood (PPML), and ML techniques. The third chapter employs finer data, Harmonized System 6-digit products, to model import demand by major sources in a commodity (beef) subjected to frequent disease outbreaks.