美联储-软着陆还是停滞_宏观情景概率估计框架（英）-2025_33页_1mb

Summary of "Soft Landing or Stagflation? A Framework for Estimating the Probabilities of Macro Scenarios"

Core Content

This paper presents a framework for estimating the probabilities of macroeconomic scenarios, particularly soft landing and stagflation, in the U.S. economy over a four-quarter horizon. The analysis uses a quasi-structural model that decomposes macroeconomic fluctuations into supply-driven and demand-driven components, allowing for a more nuanced understanding of inflation and growth risks.

Main Viewpoints

• The framework is built upon the BEE model (Bekaert, Engstrom, and Ermolov, 2025), which incorporates structural economic identification and non-Gaussian modeling of uncertainty.
• It leverages Survey of Professional Forecasters (SPF) data and historical macroeconomic data to generate joint predictive distributions for inflation, real GDP growth, and the unemployment rate.
• The model identifies four structural shocks: aggregate supply shocks, demand shocks, and two idiosyncratic shocks (headline inflation and unemployment).
• The model accounts for time-varying, asymmetric risks and uses the Bad Environment-Good Environment (BEGE) distribution to model the conditional distribution of shocks.

Key Information

Scenario Definitions

• Soft Landing: Four-quarter headline inflation between 1.5% and 2.5% and real GDP growth above 1%.
• Mild Stagflation: Four-quarter headline inflation above 3% and real GDP growth below 1%, excluding severe stagflation.
• Severe Stagflation: Four-quarter headline inflation above 4% and real GDP growth below 0%.

Probability Trends

• End of 2022: The probability of a stagflation scenario was about 35%, while the probability of a soft landing was below 5%.
• End of 2023: The stagflation probability dropped to 33%, and the soft landing probability increased to ~10%.
• End of 2024: The stagflation probability fell further to ~5%, and the soft landing probability rose to ~30%.
• Mid-2025: The probability of stagflation increased sharply to ~59%, while the probability of a soft landing dropped to ~10%.

Uncertainty and Risk

• The model emphasizes asymmetric risks and fat tails in the distribution of macroeconomic outcomes.
• The probability of severe stagflation remained relatively low, indicating that while stagflation risk increased, extreme outcomes were still unlikely.
• The increase in stagflation probability in mid-2025 is attributed to uncertainty surrounding the effects of tariffs on inflation and growth.

Implications

• The framework is useful for policymakers and financial market participants.
• For policymakers, it provides insight into whether inflation and growth objectives are aligned or in conflict.
• For market participants, it highlights the reemergence of stagflation risk, which may affect the pricing of inflation-sensitive assets and interest rate derivatives.

Methodology Overview

Step 1: Forecasting Means and Identifying Reduced-Form Shocks

• Uses ordinary least squares (OLS) regressions to estimate the conditional mean forecasts for four macroeconomic variables: headline inflation, core inflation, real GDP growth, and unemployment.
• The explanatory variables include the four-quarter lag of the dependent variable and the SPF forecast.
• OLS residuals represent reduced-form shocks, capturing unexpected deviations from the predicted paths.

Step 2: Identifying Structural Shocks

• Maps reduced-form shocks to four structural shocks using Keynesian sign and exclusion restrictions.
• Supply shocks are expected to increase inflation and reduce growth, while demand shocks are expected to affect both variables in the same direction.
• Loading parameters are estimated using a classical minimum distance (CMD) estimator, matching second-moment statistics (variances and correlations) of the reduced-form shocks.

Step 3: Modeling Conditional Distributions

• Uses the BEGE distribution to model the conditional distribution of each structural shock.
• The BEGE distribution allows for time-varying volatility, skewness, and kurtosis, reflecting asymmetric risk.
• The model generates multivariate predictive distributions by integrating forecast means with time-varying structural shock distributions.

Data and Estimation

• The data include realized macroeconomic outcomes and SPF forecasts for the period 1971Q3–2025Q2, excluding 2020Q1–2021Q4 due to pandemic-related anomalies.
• The SPF data is used to incorporate forward-looking expectations into the model.
• Bootstrapping techniques are used to calculate standard errors due to the overlapping nature of the data.

Limitations

• The model's ability to detect forward-looking risk is limited by its reliance on historical reduced-form shocks.
• Fixed loading parameters may not reflect evolving economic structures, potentially leading to distorted forecasts if the effects of policy shifts are misattributed.
• Statistical forecasts should be complemented with judgment and real-time analysis to account for structural changes and uncertainty.

Conclusion

The framework provides probabilistic assessments of macroeconomic outcomes, highlighting the evolving risk landscape and the potential for stagflation or soft landing. The increase in stagflation risk in mid-2025 underscores the importance of uncertainty and policy effects in shaping macroeconomic outcomes.