Causal Machine Learning Applied to Macroeconomic Analysis: Applications, Challenges, and Perspectives

Abstract

Traditional econometric models such as VAR, VECM, and DSGE often struggle to identify true causal relationships in macroe-conomic analysis, particularly in data-scarce and unstable environments common in developing countries. In contrast, CML offers a promising alternative by bridging predictive modeling and causal inference. This review explores CML applications in macroeconomics, with emphasis on challenges faced in low-income regions. Findings reveal fragmented adoption of CML, especially in Africa, where informal economies, limited data, and infrastructure gaps hinder implementation. While tools like DoWhy and CausalML show potential, fewer than 1% of studies incorporate spatiotemporal analysis, a key component for policy evaluation across diverse regions. To address these gaps, the review proposes a contextualized framework built around five pillars: contextual diagnosis, adaptive tool selection, methodological flexibility, collaborative validation, and progressive implementation. This approach reframes regional constraints as opportunities for innovation, encouraging the development of CML systems that are both rigorous and locally relevant. The study highlights the importance of designing CML frameworks that are accessible, culturally adapted, and capable of supporting robust policy evaluation in developing regions. Aligning ML tools with the realities of low-resource environments can unlock new pathways for evidence-based decision-making and inclusive economic development.