Integrating Econometric and Machine Learning Approaches to Assess the Global Economic Impact of U.S. Tariff Policies

Abstract: This research paper presents an integrated framework combining econometric modeling and machine learning techniques to assess the global economic impact of U.S. tariff policies. By employing vector autoregression (VAR), global VAR (GVAR), and impulse response functions, the study quantifies the short-term and dynamic effects of tariff shocks across major economies and trade blocs. Additionally, machine learning algorithms such as Long Short-Term Memory (LSTM) networks and XGBoost are utilized to enhance long-run forecasting accuracy and capture nonlinear sectoral and regional responses. Using macroeconomic indicators including GDP, trade volumes, FDI, and commodity prices, the research identifies heterogeneous impacts and policy spillovers on emerging markets, the EU, and the Asia-Pacific region. The findings provide a replicable, data-driven simulation platform for policymakers to evaluate tariff scenarios, improve forecasting precision, and inform international trade strategy in an increasingly complex and interconnected global economy. Keywords: econometric modeling, machine learning, U.S. tariff policies, global economy, trade shocks, vector autoregression, LSTM, GVAR, XGBoost, economic forecasting, policy simulation, international trade.