Essays on Impulse Response Inference in Vector Autoregressive Models
Luo, Zizheng
0009-0000-1302-0746
:
2023-07-20
Abstract
Structural vector autoregressive models are one of the most widely applied tools in empirical macroeconomic research, and impulse responses are often employed in these models to evaluate the effect of economic shocks on variables. This dissertation provides identification, estimation, and inference results for impulse responses in several different class of structural vector autoregressive models. Chapter 1 proposes a method to construct uniformly valid confidence intervals for impulse responses in a large class of vector autoregressive models, via a uniformly valid Wald test. Our method not only allows for uniform inference on the impulse responses, but also for any function of the VAR parameters. We also provide a simulation study that not only examines the finite sample accuracy of our method, but also provides a comparison of the respective accuracy of the VAR and LP methods of constructing impulse responses. Chapter 2 demonstrates that in structural FAVAR models, the structural impulse responses can be identified by independence and non-Gaussianity assumptions on structural shocks. We propose a two-step estimator to estimate the VAR parameters and derive the relevant asymptotic properties of the estimator. We also conduct a simulation study and revisit the empirical analysis in Bernanke et al. (2005) to demonstrate the application of our method. Our method suggests a stronger and more persistent effect of monetary policy on output.