Public Budget Simulations with Machine Learning and Synthetic Data: Some Challenges and Lessons from the Mexican Case

Abstract

AbstractPublic resources are always limited and must be sufficient to meet a country’s needs in the short term. The budget process is reflected in the structure of the Revenue Law Initiative based on the question: of how much money must be raised, and in the federal spending bill: how will these resources be distributed? Regarding these seminal ideas, this paper explores the potential of machine learning (ML) techniques and synthetic data in public budget simulations. For this purpose, historical data from the Mexican government’s federal budget analytics are employed to identify which algorithms perform best in simulating the budget and the challenges in these kinds of practices. Findings discovered that Random Forest was the best-performing algorithm. The paper provides some challenges and lessons for public budgeting with ML and synthetic data. This novel study could assist public budgeting decision-making, simulating scenarios in government.KeywordsPublic budgetMachine learningArtificial intelligenceSynthetic dataSimulationsDecision-makingRandom forest