2025-01-19-美联储-非参数时变IV-SVARs_估计与推理（英）_65页_4mb

| $\textbf{Content Summary}$ |

| $\textbf{1. Core Contribution}$ |

| | · $\textbf{Methodology}$: Nonparametric time-varying coefficient estimation in IV-SVARs using kernel-based estimators, with asymptotic distributions for identification and inference. · $\textbf{Inference}$: Handles weak identification through asymptotic distributions and confidence sets, robust to small sample sizes. · $\textbf{Application}$: Studies oil supply news shocks' effects on US industrial production, revealing time-variation and significance |

| $\textbf{2. Estimation and Inference}$ |

| | · $\textbf{Kernel Estimators}$: \

• Nonparametric approach avoiding strong parametric assumptions. \
• Reduced-form parameter estimation for consistent time-varying impulse response functions (IRFs). \
• Computationally simple, scales well with dimensionality and sample size.

| | · $\textbf{Estimators}$ \

• IV-SVAR: $\text{Consistent under weak identification}$, allows time-varying absolute IRFs. \
• Internal IV estimator: $\text{Robust to non-invertibility}$, estimates relative IRFs.

| | · $\textbf{Finite-Sample Properties}$ \

• Covers uncertainty via methods like Delta method and Anderson-Rubin test. \
• Provides weak-IV robust inference and data-driven bandwidth selection. \
• Shows reasonable empirical coverage in small samples.

| $\textbf{3. Application}$ |

| | $\textbf{Oil Market DGP}$ \

• Simulated from a TVP kernel estimator using real oil market data. · $\textbf{Results}$ \
• IRFs display substantial time-variation.
• Such patterns align with the U.S. shale oil revolution.

| | $\textbf{U.S. Industrial Production}$ \

• Time-varying IRFs reveal little significant US manufacturing response since the mid-2000s. · $\textbf{Key Patterns}$ \
• Reduced price effects and losses in the short run post-revolution. \
• Depending on industry, varied time-variation and response patterns.