AI-Based Scenario Simulation for U.S. Debt Trajectories Under Fiscal Policy Changes

Abstract

Policymakers are getting more stressed about whether public debt can actually be paid back. Budgets change; economies shift; interest rate environments move up plus down. Most people still use basic math rules to guess where debt goes. These old rules explain how debt grows based on spending plus economic expansion. Theoretical clarity is nice. These simple models often miss the weird, complicated patterns seen in real history. This research project builds an AI tool to track where sovereign debt is heading. It connects machine learning models with specific scenario tests plus risk checks. The data comes from the Federal Reserve Economic Data (FRED) system. It pulls together tax receipts, federal spending, interest rates, GDP growth, plus the debt-to-GDP ratio. The study tested several ways to predict debt one year out. It looked at Ridge Regression, XGBoost, plus a standard math benchmark based on fiscal rules. The Ridge Regression model gave the most accurate short-term guesses. It worked better than the fancy XGBoost model. It even beat the standard math benchmark in this specific test. After finding the best model, a simulation engine projected debt paths 15 years into the future. It looked at different choices plus economic surprises. These included spending more money, cutting budgets, high interest rates, plus periods of stagflation. Monte Carlo simulations helped map out the uncertainty. This provided a look at the odds of various long-term debt levels. Policy changes, plus the difference between interest rates plus growth, are the main drivers of debt over time. Past debt levels keep the predictions anchored. Mixing machine learning, scenario tests, plus explainable AI techniques helps data-driven models work alongside old economics. It offers a more flexible way to see how fiscal policy might play out.