IMF-监测亚洲的需求和供应：一种行业层面的方法（英）-2023.10-49页_6mb

报告摘要

IMF Working Paper Summary: Monitoring Demand and Supply in Asia: An Industry Level Approach

Author: Chris Redl
Date: October 20, 2023

1. Summary

This paper introduces a method to decompose GDP growth and inflation into demand- and supply-driven components at the industry level using forecast errors from statistical models. It extends Shapiro (2022) by incorporating machine learning techniques and addressing idiosyncratic versus aggregate shocks. The analysis focuses on 12 Asian countries, the US, and Europe using industry-level data.

Key Findings:

2020 Lockdowns: Mixed demand and supply shocks drove GDP declines. Idiosyncratic demand shocks were significant in some countries (e.g., India). Supply shocks played a larger role during the post-COVID recovery (2021).
Inflation (2021-22): Supply shocks dominated in advanced economies (China, Australia, Korea), while emerging Asian economies saw significant demand-driven inflation. Commodity-related demand shocks (e.g., Indonesia) were also notable.
Keynesian Supply Shocks: Evidence of asymmetric supply shocks leading to aggregate demand-like effects exists but is country-specific and largely driven by COVID.
Growth Spillovers: Limited short-term global spillovers from China, medium-term effects from Europe and the US (around 0.3% of GDP decline). China's spillovers are stronger in Asia.
Methodological Advancements: Machine learning improves forecasting accuracy. Granular data helps distinguish between idiosyncratic and aggregate shocks.

2. Conclusion

The industry-level decomposition provides granular insights into economic shocks, aiding policymakers in designing timely responses. The analysis underscores the importance of sectoral heterogeneity in understanding demand-supply dynamics, especially during crises and recoveries.

3. Methodology

Data: Industry-level GDP, quantity, and price indices for multiple countries.
Models: Forecasting models (OLS, Ridge, Lasso, Random Forest, SVR, etc.) to predict industry-level data, followed by decomposition into demand/supply factors.
Innovation: Incorporates machine learning to handle nonlinearities and large datasets, enabling better-out-of-sample forecasting and decomposition.

展开完整摘要