IMF-情景综合与宏观经济风险（英）-2025.5_30页_2mb

Scenario Synthesis and Macroeconomic Risk

Abstract

This paper introduces a methodology to integrate scenario analysis and statistical forecasting by reconciling judgmental scenarios with risk forecasts. The approach uses Bayesian analysis to quantify the concordance between scenarios and a reference forecasting model. The goal is to provide a framework for systematically evaluating and combining risks from different scenarios, improving policy decision-making while supporting risk communication.

Key components include:

1. Bayesian predictive synthesis to combine baseline and alternative scenarios based on their concordance with a statistical reference distribution
2. Entropic tilting to adjust probabilistic forecasts according to partial scenario information
3. Expected misclassification rate (EMR) as a measure of probabilistic concordance
4. Synthetic or "backstop" scenarios to address potential incompleteness in the original scenario set

The methodology is applied to U.S. Federal Reserve Tealbook scenarios, showing that it can systematically evaluate scenario relevance and address risks not captured by central forecasts. It offers a transparent way to combine narrative scenarios with statistical models.

Key Recommendations

1. Methodological Innovation: The approach provides a formal way to quantitatively integrate qualitative scenario analysis with statistical risk forecasts.
2. Practical Applicability: It is demonstrated through case studies of U.S. Federal Reserve scenarios, highlighting its relevance for monetary policy.
3. Scalability: The framework scales well to large numbers of scenarios due to convex optimization properties.
4. Robustness to Uncertainty: Particularly valuable in contexts with high economic uncertainty where traditional scenarios may miss key risks.

This methodology represents a significant advance in translating qualitative economic narratives into statistically rigorous risk assessments.