Random Forest-Based Analysis of Risk and Importance in Selected Macroeconomic Imbalance Procedure (MIP) Scoreboard Indicators: A New Perspective on CLIFS Relevance

The relevance of Macroeconomic Imbalance Procedure (MIP) Scoreboard indicators in predicting financial crisis risk is explored in this study, with particular attention paid to how they relate to the Country-Level Index of Financial Stress (CLIFS). We apply Random Forest and Gradient Boosted Trees models in TIBCO Statistica to data from 2002 to 2021, providing a novel method of research by combining country-level financial stress indices with indicators of internal and external imbalances. The findings demonstrate that while Gradient Boosted Trees provide better predictive performance, Random Forest offers steady error rates with less overfitting. However, the latter tend to underestimate extreme values and overestimate small ones, highlighting areas for improvement. This paper's novelty lies in its interdependent analysis of these indicators, presenting a new framework for studying macroeconomic imbalances and crisis risks. The findings suggest that additional factors influencing CLIFS remain unexplored, providing avenues for further research to refine predictive models for financial crises.